d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
3f929d8cdd
freebsd-dev/sys/dev/bge/if_bge.c

6601 lines
194 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
/*
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
* Broadcom BCM57xx(x)/BCM590x NetXtreme and NetLink family Ethernet driver
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
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
* Alteon Networks as part of the Tigon I and Tigon II Gigabit Ethernet
Correct spelling in comments. Submitted by: brucec
2011-05-02 21:04:23 +00:00
* MAC chips. The BCM5700, sometimes referred to as the Tigon III, has
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
* 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;
Remove duplicate const specifiers in many drivers (I hope I got all of them, please let me know if not). Most of these are of the form: static const struct bzzt_type { [...list of members...] } const bzzt_devs[] = { [...list of initializers...] }; The second const is unnecessary, as arrays cannot be modified anyway, and if the elements are const, the whole thing is const automatically (e.g. it is placed in .rodata). I have verified this does not change the binary output of a full kernel build (except for build timestamps embedded in the object files). Reviewed by: yongari, marius MFC after: 1 week
2012-11-05 19:16:27 +00:00
} 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
{ 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 },
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5717 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5718 },
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5719 },
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_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 },
Add initial BCM57765 family support. The BCM57765 family seems to have similar hardware features of BCM5718 family except the number of receive return ring is 4. The BCM57765 family is known to support IEEE 802.3az EEE(Energy Efficient Ethernet) but this change does not include EEE support code. I hope EEE is implemented in near future. This change will support BCM57761, BCM57765, BCM57781, BCM57785, BCM57791 and BCM57795. All hardware offloading features are supported and suspend/resume also should work. Many thanks to Broadcom for continuing support of FreeBSD. Tested by: Paul Thornton (prt <> prt dot org) HW donated by: Broadcom
2011-05-04 17:04:31 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM57761 },
Add preliminary support for BCM57766 ASIC. While I'm here add BCM57762 device id which is found on Apple Thunderbolt ethernet. Tested by: Richard Kuhns < rjk <> wintek dot com >
2012-11-30 01:38:00 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM57762 },
Add initial BCM57765 family support. The BCM57765 family seems to have similar hardware features of BCM5718 family except the number of receive return ring is 4. The BCM57765 family is known to support IEEE 802.3az EEE(Energy Efficient Ethernet) but this change does not include EEE support code. I hope EEE is implemented in near future. This change will support BCM57761, BCM57765, BCM57781, BCM57785, BCM57791 and BCM57795. All hardware offloading features are supported and suspend/resume also should work. Many thanks to Broadcom for continuing support of FreeBSD. Tested by: Paul Thornton (prt <> prt dot org) HW donated by: Broadcom
2011-05-04 17:04:31 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM57765 },
Add preliminary support for BCM57766 ASIC. While I'm here add BCM57762 device id which is found on Apple Thunderbolt ethernet. Tested by: Richard Kuhns < rjk <> wintek dot com >
2012-11-30 01:38:00 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM57766 },
- 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_BCM57780 },
Add initial BCM57765 family support. The BCM57765 family seems to have similar hardware features of BCM5718 family except the number of receive return ring is 4. The BCM57765 family is known to support IEEE 802.3az EEE(Energy Efficient Ethernet) but this change does not include EEE support code. I hope EEE is implemented in near future. This change will support BCM57761, BCM57765, BCM57781, BCM57785, BCM57791 and BCM57795. All hardware offloading features are supported and suspend/resume also should work. Many thanks to Broadcom for continuing support of FreeBSD. Tested by: Paul Thornton (prt <> prt dot org) HW donated by: Broadcom
2011-05-04 17:04:31 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM57781 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM57785 },
- 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_BCM57788 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM57790 },
Add initial BCM57765 family support. The BCM57765 family seems to have similar hardware features of BCM5718 family except the number of receive return ring is 4. The BCM57765 family is known to support IEEE 802.3az EEE(Energy Efficient Ethernet) but this change does not include EEE support code. I hope EEE is implemented in near future. This change will support BCM57761, BCM57765, BCM57781, BCM57785, BCM57791 and BCM57795. All hardware offloading features are supported and suspend/resume also should work. Many thanks to Broadcom for continuing support of FreeBSD. Tested by: Paul Thornton (prt <> prt dot org) HW donated by: Broadcom
2011-05-04 17:04:31 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM57791 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM57795 },
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;
Remove duplicate const specifiers in many drivers (I hope I got all of them, please let me know if not). Most of these are of the form: static const struct bzzt_type { [...list of members...] } const bzzt_devs[] = { [...list of initializers...] }; The second const is unnecessary, as arrays cannot be modified anyway, and if the elements are const, the whole thing is const automatically (e.g. it is placed in .rodata). I have verified this does not change the binary output of a full kernel build (except for build timestamps embedded in the object files). Reviewed by: yongari, marius MFC after: 1 week
2012-11-05 19:16:27 +00:00
} bge_vendors[] = {
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
{ 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;
Remove duplicate const specifiers in many drivers (I hope I got all of them, please let me know if not). Most of these are of the form: static const struct bzzt_type { [...list of members...] } const bzzt_devs[] = { [...list of initializers...] }; The second const is unnecessary, as arrays cannot be modified anyway, and if the elements are const, the whole thing is const automatically (e.g. it is placed in .rodata). I have verified this does not change the binary output of a full kernel build (except for build timestamps embedded in the object files). Reviewed by: yongari, marius MFC after: 1 week
2012-11-05 19:16:27 +00:00
} bge_revisions[] = {
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_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" },
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
{ BGE_CHIPID_BCM5717_A0, "BCM5717 A0" },
{ BGE_CHIPID_BCM5717_B0, "BCM5717 B0" },
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
{ BGE_CHIPID_BCM5719_A0, "BCM5719 A0" },
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
{ BGE_CHIPID_BCM5720_A0, "BCM5720 A0" },
Add more chipset revision IDs. Obtained from: OpenBSD
2007-03-09 01:30:23 +00:00
{ 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 initial BCM57765 family support. The BCM57765 family seems to have similar hardware features of BCM5718 family except the number of receive return ring is 4. The BCM57765 family is known to support IEEE 802.3az EEE(Energy Efficient Ethernet) but this change does not include EEE support code. I hope EEE is implemented in near future. This change will support BCM57761, BCM57765, BCM57781, BCM57785, BCM57791 and BCM57795. All hardware offloading features are supported and suspend/resume also should work. Many thanks to Broadcom for continuing support of FreeBSD. Tested by: Paul Thornton (prt <> prt dot org) HW donated by: Broadcom
2011-05-04 17:04:31 +00:00
{ BGE_CHIPID_BCM57765_A0, "BCM57765 A0" },
{ BGE_CHIPID_BCM57765_B0, "BCM57765 B0" },
- 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.
*/
Remove duplicate const specifiers in many drivers (I hope I got all of them, please let me know if not). Most of these are of the form: static const struct bzzt_type { [...list of members...] } const bzzt_devs[] = { [...list of initializers...] }; The second const is unnecessary, as arrays cannot be modified anyway, and if the elements are const, the whole thing is const automatically (e.g. it is placed in .rodata). I have verified this does not change the binary output of a full kernel build (except for build timestamps embedded in the object files). Reviewed by: yongari, marius MFC after: 1 week
2012-11-05 19:16:27 +00:00
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 initial BCM57765 family support. The BCM57765 family seems to have similar hardware features of BCM5718 family except the number of receive return ring is 4. The BCM57765 family is known to support IEEE 802.3az EEE(Energy Efficient Ethernet) but this change does not include EEE support code. I hope EEE is implemented in near future. This change will support BCM57761, BCM57765, BCM57781, BCM57785, BCM57791 and BCM57795. All hardware offloading features are supported and suspend/resume also should work. Many thanks to Broadcom for continuing support of FreeBSD. Tested by: Paul Thornton (prt <> prt dot org) HW donated by: Broadcom
2011-05-04 17:04:31 +00:00
{ BGE_ASICREV_BCM57765, "unknown BCM57765" },
Add preliminary support for BCM57766 ASIC. While I'm here add BCM57762 device id which is found on Apple Thunderbolt ethernet. Tested by: Richard Kuhns < rjk <> wintek dot com >
2012-11-30 01:38:00 +00:00
{ BGE_ASICREV_BCM57766, "unknown BCM57766" },
- 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" },
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
{ BGE_ASICREV_BCM5717, "unknown BCM5717" },
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
{ BGE_ASICREV_BCM5719, "unknown BCM5719" },
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
{ BGE_ASICREV_BCM5720, "unknown BCM5720" },
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)
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
#define BGE_IS_5717_PLUS(sc) ((sc)->bge_flags & BGE_FLAG_5717_PLUS)
Add preliminary support for BCM57766 ASIC. While I'm here add BCM57762 device id which is found on Apple Thunderbolt ethernet. Tested by: Richard Kuhns < rjk <> wintek dot com >
2012-11-30 01:38:00 +00:00
#define BGE_IS_57765_PLUS(sc) ((sc)->bge_flags & BGE_FLAG_57765_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
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
static uint32_t bge_chipid(device_t);
static const struct bge_vendor * bge_lookup_vendor(uint16_t);
static const struct bge_revision * bge_lookup_rev(uint32_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
Show PCI bus speed and width as well as running mode of PCI-X device in device attach. This would help to narrow down issue to a specific controller and operating mode of the controller. While I'm here rename BGE_MISCCFG_BOARD_ID with BGE_MISCCFG_BOARD_ID_MASK.
2012-03-12 02:42:47 +00:00
static void bge_devinfo(struct bge_softc *);
Add workaround for PCI-X BCM5704 controller that live behind AMD-8131 PCI-X bridge. The bridge seems to reorder write access to mailbox registers such that it caused watchdog timeouts by out-of-order TX completions. Tested by: Michael L. Squires <mikes <> siralan dot org > Reviewed by: jhb
2012-03-12 02:09:47 +00:00
static int bge_mbox_reorder(struct bge_softc *);
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);
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
static void bge_rxcsum(struct bge_softc *, struct bge_rx_bd *, struct mbuf *);
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 *);
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
static void bge_stats_clear_regs(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_stats_update(struct bge_softc *);
static void bge_stats_update_regs(struct bge_softc *);
Add workaround for BCM5906 controller silicon bug. If device receive two back-to-back send BDs with less than or equal to 8 total bytes then the device may hang. The two back-to-back send BDs must be in the same frame for this failure to occur. Thanks to davidch for detailed errata information. Reviewed by: davidch
2010-10-19 23:04:23 +00:00
static struct mbuf *bge_check_short_dma(struct mbuf *);
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 *,
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
uint16_t *, 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 *);
Since r117657, bge(4) does not enable buffer manager for BCM5705 or newer controllers. However, all data sheet I have access has no indication that buffer manager should not be touched on these controllers. It seems the buffer manager always runs on BCM5705 or newer controllers. Some controller(e.g. BCM5719) needs other buffer manager configuration so driver should enable buffer manager for all controllers. Both Linux and OpenBSD/NetBSD use the same approach. This change polls enable bit of block to know whether specified block was really stopped as well as enabling buffer manager for all controllers in driver initialization. Obtained from: NetBSD
2011-05-09 20:10:46 +00:00
static void bge_stop_block(struct bge_softc *, bus_size_t, 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_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 initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
static uint32_t bge_dma_swap_options(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
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
#define BGE_RESET_SHUTDOWN 0
#define BGE_RESET_START 1
#define BGE_RESET_SUSPEND 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
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
static void bge_ape_lock_init(struct bge_softc *);
static void bge_ape_read_fw_ver(struct bge_softc *);
static int bge_ape_lock(struct bge_softc *, int);
static void bge_ape_unlock(struct bge_softc *, int);
static void bge_ape_send_event(struct bge_softc *, uint32_t);
static void bge_ape_driver_state_change(struct bge_softc *, int);
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
/*
* 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);
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
static int bge_sysctl_ape_read(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
static int bge_sysctl_mem_read(SYSCTL_HANDLER_ARGS);
#endif
static void bge_add_sysctls(struct bge_softc *);
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
static void bge_add_sysctl_stats_regs(struct bge_softc *,
struct sysctl_ctx_list *, struct sysctl_oid_list *);
static void bge_add_sysctl_stats(struct bge_softc *, struct sysctl_ctx_list *,
struct sysctl_oid_list *);
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
/* MII interface */
DEVMETHOD(miibus_readreg, bge_miibus_readreg),
DEVMETHOD(miibus_writereg, bge_miibus_writereg),
DEVMETHOD(miibus_statchg, bge_miibus_statchg),
- There's no need to overwrite the default device method with the default one. Interestingly, these are actually the default for quite some time (bus_generic_driver_added(9) since r52045 and bus_generic_print_child(9) since r52045) but even recently added device drivers do this unnecessarily. Discussed with: jhb, marcel - While at it, use DEVMETHOD_END. Discussed with: jhb - Also while at it, use __FBSDID.
2011-11-22 21:28:20 +00:00
DEVMETHOD_END
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 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);
Mark all SYSCTL_NODEs static that have no corresponding SYSCTL_DECLs. The SYSCTL_NODE macro defines a list that stores all child-elements of that node. If there's no SYSCTL_DECL macro anywhere else, there's no reason why it shouldn't be static.
2011-11-07 15:43:11 +00:00
static SYSCTL_NODE(_hw, OID_AUTO, bge, CTLFLAG_RD, 0, "BGE driver parameters");
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
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
Protect bge(4) from accessing invalid NIC internal memory regions on BCM5906. Tested by: Buganini < buganini <> gmail dot com >
2010-10-12 19:31:25 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5906 &&
off >= BGE_STATS_BLOCK && off < BGE_SEND_RING_1_TO_4)
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
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;
Protect bge(4) from accessing invalid NIC internal memory regions on BCM5906. Tested by: Buganini < buganini <> gmail dot com >
2010-10-12 19:31:25 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5906 &&
off >= BGE_STATS_BLOCK && off < BGE_SEND_RING_1_TO_4)
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
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);
Add workaround for PCI-X BCM5704 controller that live behind AMD-8131 PCI-X bridge. The bridge seems to reorder write access to mailbox registers such that it caused watchdog timeouts by out-of-order TX completions. Tested by: Michael L. Squires <mikes <> siralan dot org > Reviewed by: jhb
2012-03-12 02:09:47 +00:00
if ((sc->bge_flags & BGE_FLAG_MBOX_REORDER) != 0)
CSR_READ_4(sc, off);
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 APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
/*
* Clear all stale locks and select the lock for this driver instance.
*/
static void
bge_ape_lock_init(struct bge_softc *sc)
{
uint32_t bit, regbase;
int i;
if (sc->bge_asicrev == BGE_ASICREV_BCM5761)
regbase = BGE_APE_LOCK_GRANT;
else
regbase = BGE_APE_PER_LOCK_GRANT;
/* Clear any stale locks. */
for (i = BGE_APE_LOCK_PHY0; i <= BGE_APE_LOCK_GPIO; i++) {
switch (i) {
case BGE_APE_LOCK_PHY0:
case BGE_APE_LOCK_PHY1:
case BGE_APE_LOCK_PHY2:
case BGE_APE_LOCK_PHY3:
bit = BGE_APE_LOCK_GRANT_DRIVER0;
break;
default:
if (sc->bge_func_addr != 0)
bit = BGE_APE_LOCK_GRANT_DRIVER0;
else
bit = (1 << sc->bge_func_addr);
}
APE_WRITE_4(sc, regbase + 4 * i, bit);
}
/* Select the PHY lock based on the device's function number. */
switch (sc->bge_func_addr) {
case 0:
sc->bge_phy_ape_lock = BGE_APE_LOCK_PHY0;
break;
case 1:
sc->bge_phy_ape_lock = BGE_APE_LOCK_PHY1;
break;
case 2:
sc->bge_phy_ape_lock = BGE_APE_LOCK_PHY2;
break;
case 3:
sc->bge_phy_ape_lock = BGE_APE_LOCK_PHY3;
break;
default:
device_printf(sc->bge_dev,
"PHY lock not supported on this function\n");
}
}
/*
* Check for APE firmware, set flags, and print version info.
*/
static void
bge_ape_read_fw_ver(struct bge_softc *sc)
{
const char *fwtype;
uint32_t apedata, features;
/* Check for a valid APE signature in shared memory. */
apedata = APE_READ_4(sc, BGE_APE_SEG_SIG);
if (apedata != BGE_APE_SEG_SIG_MAGIC) {
sc->bge_mfw_flags &= ~ BGE_MFW_ON_APE;
return;
}
/* Check if APE firmware is running. */
apedata = APE_READ_4(sc, BGE_APE_FW_STATUS);
if ((apedata & BGE_APE_FW_STATUS_READY) == 0) {
device_printf(sc->bge_dev, "APE signature found "
"but FW status not ready! 0x%08x\n", apedata);
return;
}
sc->bge_mfw_flags |= BGE_MFW_ON_APE;
/* Fetch the APE firwmare type and version. */
apedata = APE_READ_4(sc, BGE_APE_FW_VERSION);
features = APE_READ_4(sc, BGE_APE_FW_FEATURES);
if ((features & BGE_APE_FW_FEATURE_NCSI) != 0) {
sc->bge_mfw_flags |= BGE_MFW_TYPE_NCSI;
fwtype = "NCSI";
} else if ((features & BGE_APE_FW_FEATURE_DASH) != 0) {
sc->bge_mfw_flags |= BGE_MFW_TYPE_DASH;
fwtype = "DASH";
} else
fwtype = "UNKN";
/* Print the APE firmware version. */
device_printf(sc->bge_dev, "APE FW version: %s v%d.%d.%d.%d\n",
fwtype,
(apedata & BGE_APE_FW_VERSION_MAJMSK) >> BGE_APE_FW_VERSION_MAJSFT,
(apedata & BGE_APE_FW_VERSION_MINMSK) >> BGE_APE_FW_VERSION_MINSFT,
(apedata & BGE_APE_FW_VERSION_REVMSK) >> BGE_APE_FW_VERSION_REVSFT,
(apedata & BGE_APE_FW_VERSION_BLDMSK));
}
static int
bge_ape_lock(struct bge_softc *sc, int locknum)
{
uint32_t bit, gnt, req, status;
int i, off;
if ((sc->bge_mfw_flags & BGE_MFW_ON_APE) == 0)
return (0);
/* Lock request/grant registers have different bases. */
if (sc->bge_asicrev == BGE_ASICREV_BCM5761) {
req = BGE_APE_LOCK_REQ;
gnt = BGE_APE_LOCK_GRANT;
} else {
req = BGE_APE_PER_LOCK_REQ;
gnt = BGE_APE_PER_LOCK_GRANT;
}
off = 4 * locknum;
switch (locknum) {
case BGE_APE_LOCK_GPIO:
/* Lock required when using GPIO. */
if (sc->bge_asicrev == BGE_ASICREV_BCM5761)
return (0);
if (sc->bge_func_addr == 0)
bit = BGE_APE_LOCK_REQ_DRIVER0;
else
bit = (1 << sc->bge_func_addr);
break;
case BGE_APE_LOCK_GRC:
/* Lock required to reset the device. */
if (sc->bge_func_addr == 0)
bit = BGE_APE_LOCK_REQ_DRIVER0;
else
bit = (1 << sc->bge_func_addr);
break;
case BGE_APE_LOCK_MEM:
/* Lock required when accessing certain APE memory. */
if (sc->bge_func_addr == 0)
bit = BGE_APE_LOCK_REQ_DRIVER0;
else
bit = (1 << sc->bge_func_addr);
break;
case BGE_APE_LOCK_PHY0:
case BGE_APE_LOCK_PHY1:
case BGE_APE_LOCK_PHY2:
case BGE_APE_LOCK_PHY3:
/* Lock required when accessing PHYs. */
bit = BGE_APE_LOCK_REQ_DRIVER0;
break;
default:
return (EINVAL);
}
/* Request a lock. */
APE_WRITE_4(sc, req + off, bit);
/* Wait up to 1 second to acquire lock. */
for (i = 0; i < 20000; i++) {
status = APE_READ_4(sc, gnt + off);
if (status == bit)
break;
DELAY(50);
}
/* Handle any errors. */
if (status != bit) {
device_printf(sc->bge_dev, "APE lock %d request failed! "
"request = 0x%04x[0x%04x], status = 0x%04x[0x%04x]\n",
locknum, req + off, bit & 0xFFFF, gnt + off,
status & 0xFFFF);
/* Revoke the lock request. */
APE_WRITE_4(sc, gnt + off, bit);
return (EBUSY);
}
return (0);
}
static void
bge_ape_unlock(struct bge_softc *sc, int locknum)
{
uint32_t bit, gnt;
int off;
if ((sc->bge_mfw_flags & BGE_MFW_ON_APE) == 0)
return;
if (sc->bge_asicrev == BGE_ASICREV_BCM5761)
gnt = BGE_APE_LOCK_GRANT;
else
gnt = BGE_APE_PER_LOCK_GRANT;
off = 4 * locknum;
switch (locknum) {
case BGE_APE_LOCK_GPIO:
if (sc->bge_asicrev == BGE_ASICREV_BCM5761)
return;
if (sc->bge_func_addr == 0)
bit = BGE_APE_LOCK_GRANT_DRIVER0;
else
bit = (1 << sc->bge_func_addr);
break;
case BGE_APE_LOCK_GRC:
if (sc->bge_func_addr == 0)
bit = BGE_APE_LOCK_GRANT_DRIVER0;
else
bit = (1 << sc->bge_func_addr);
break;
case BGE_APE_LOCK_MEM:
if (sc->bge_func_addr == 0)
bit = BGE_APE_LOCK_GRANT_DRIVER0;
else
bit = (1 << sc->bge_func_addr);
break;
case BGE_APE_LOCK_PHY0:
case BGE_APE_LOCK_PHY1:
case BGE_APE_LOCK_PHY2:
case BGE_APE_LOCK_PHY3:
bit = BGE_APE_LOCK_GRANT_DRIVER0;
break;
default:
return;
}
APE_WRITE_4(sc, gnt + off, bit);
}
/*
* Send an event to the APE firmware.
*/
static void
bge_ape_send_event(struct bge_softc *sc, uint32_t event)
{
uint32_t apedata;
int i;
/* NCSI does not support APE events. */
if ((sc->bge_mfw_flags & BGE_MFW_ON_APE) == 0)
return;
/* Wait up to 1ms for APE to service previous event. */
for (i = 10; i > 0; i--) {
if (bge_ape_lock(sc, BGE_APE_LOCK_MEM) != 0)
break;
apedata = APE_READ_4(sc, BGE_APE_EVENT_STATUS);
if ((apedata & BGE_APE_EVENT_STATUS_EVENT_PENDING) == 0) {
APE_WRITE_4(sc, BGE_APE_EVENT_STATUS, event |
BGE_APE_EVENT_STATUS_EVENT_PENDING);
bge_ape_unlock(sc, BGE_APE_LOCK_MEM);
APE_WRITE_4(sc, BGE_APE_EVENT, BGE_APE_EVENT_1);
break;
}
bge_ape_unlock(sc, BGE_APE_LOCK_MEM);
DELAY(100);
}
if (i == 0)
device_printf(sc->bge_dev, "APE event 0x%08x send timed out\n",
event);
}
static void
bge_ape_driver_state_change(struct bge_softc *sc, int kind)
{
uint32_t apedata, event;
if ((sc->bge_mfw_flags & BGE_MFW_ON_APE) == 0)
return;
switch (kind) {
case BGE_RESET_START:
/* If this is the first load, clear the load counter. */
apedata = APE_READ_4(sc, BGE_APE_HOST_SEG_SIG);
if (apedata != BGE_APE_HOST_SEG_SIG_MAGIC)
APE_WRITE_4(sc, BGE_APE_HOST_INIT_COUNT, 0);
else {
apedata = APE_READ_4(sc, BGE_APE_HOST_INIT_COUNT);
APE_WRITE_4(sc, BGE_APE_HOST_INIT_COUNT, ++apedata);
}
APE_WRITE_4(sc, BGE_APE_HOST_SEG_SIG,
BGE_APE_HOST_SEG_SIG_MAGIC);
APE_WRITE_4(sc, BGE_APE_HOST_SEG_LEN,
BGE_APE_HOST_SEG_LEN_MAGIC);
/* Add some version info if bge(4) supports it. */
APE_WRITE_4(sc, BGE_APE_HOST_DRIVER_ID,
BGE_APE_HOST_DRIVER_ID_MAGIC(1, 0));
APE_WRITE_4(sc, BGE_APE_HOST_BEHAVIOR,
BGE_APE_HOST_BEHAV_NO_PHYLOCK);
APE_WRITE_4(sc, BGE_APE_HOST_HEARTBEAT_INT_MS,
BGE_APE_HOST_HEARTBEAT_INT_DISABLE);
APE_WRITE_4(sc, BGE_APE_HOST_DRVR_STATE,
BGE_APE_HOST_DRVR_STATE_START);
event = BGE_APE_EVENT_STATUS_STATE_START;
break;
case BGE_RESET_SHUTDOWN:
APE_WRITE_4(sc, BGE_APE_HOST_DRVR_STATE,
BGE_APE_HOST_DRVR_STATE_UNLOAD);
event = BGE_APE_EVENT_STATUS_STATE_UNLOAD;
break;
case BGE_RESET_SUSPEND:
event = BGE_APE_EVENT_STATUS_STATE_SUSPEND;
break;
default:
return;
}
bge_ape_send_event(sc, event | BGE_APE_EVENT_STATUS_DRIVER_EVNT |
BGE_APE_EVENT_STATUS_STATE_CHNGE);
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since 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;
Overhaul MII register access routine and remove unnecessary BGE_MI_MODE register accesses. Previously bge(4) used to read BGE_MI_MODE register to detect whether it needs to disable autopolling feature or not. Because we don't touch autopolling in other part of driver there is no reason to read BGE_MI_MODE register given that we know default value in advance. In order to achieve the goal, check whether the controller has CPMU(Central Power Mangement Unit) capability. If controller has CPMU feature, use 500KHz MII management interface(mdio/mdc) frequency regardless core clock frequency. Otherwise use default MII clock. While I'm here, add CPMU register definition. In bge_miibus_readreg(), rearrange code a bit and remove goto statement. In bge_miibus_writereg(), make sure to restore autopolling even if MII write failed. The delay time inserted after accessing BGE_MI_MODE register increased from 40us to 80us. The default PHY address is now stored in softc. All PHYs supported by bge(4) currently uses PHY address 1 but it will be changed when we add newer controllers. This change will make it easier to change default PHY address depending on PHY models. Submitted by: davidch
2010-10-06 17:35:27 +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;
sc = device_get_softc(dev);
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
if (bge_ape_lock(sc, sc->bge_phy_ape_lock) != 0)
return (0);
Overhaul MII register access routine and remove unnecessary BGE_MI_MODE register accesses. Previously bge(4) used to read BGE_MI_MODE register to detect whether it needs to disable autopolling feature or not. Because we don't touch autopolling in other part of driver there is no reason to read BGE_MI_MODE register given that we know default value in advance. In order to achieve the goal, check whether the controller has CPMU(Central Power Mangement Unit) capability. If controller has CPMU feature, use 500KHz MII management interface(mdio/mdc) frequency regardless core clock frequency. Otherwise use default MII clock. While I'm here, add CPMU register definition. In bge_miibus_readreg(), rearrange code a bit and remove goto statement. In bge_miibus_writereg(), make sure to restore autopolling even if MII write failed. The delay time inserted after accessing BGE_MI_MODE register increased from 40us to 80us. The default PHY address is now stored in softc. All PHYs supported by bge(4) currently uses PHY address 1 but it will be changed when we add newer controllers. This change will make it easier to change default PHY address depending on PHY models. Submitted by: davidch
2010-10-06 17:35:27 +00:00
/* Clear the autopoll bit if set, otherwise may trigger PCI errors. */
if ((sc->bge_mi_mode & BGE_MIMODE_AUTOPOLL) != 0) {
CSR_WRITE_4(sc, BGE_MI_MODE,
sc->bge_mi_mode & ~BGE_MIMODE_AUTOPOLL);
DELAY(80);
- 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
}
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
Overhaul MII register access routine and remove unnecessary BGE_MI_MODE register accesses. Previously bge(4) used to read BGE_MI_MODE register to detect whether it needs to disable autopolling feature or not. Because we don't touch autopolling in other part of driver there is no reason to read BGE_MI_MODE register given that we know default value in advance. In order to achieve the goal, check whether the controller has CPMU(Central Power Mangement Unit) capability. If controller has CPMU feature, use 500KHz MII management interface(mdio/mdc) frequency regardless core clock frequency. Otherwise use default MII clock. While I'm here, add CPMU register definition. In bge_miibus_readreg(), rearrange code a bit and remove goto statement. In bge_miibus_writereg(), make sure to restore autopolling even if MII write failed. The delay time inserted after accessing BGE_MI_MODE register increased from 40us to 80us. The default PHY address is now stored in softc. All PHYs supported by bge(4) currently uses PHY address 1 but it will be changed when we add newer controllers. This change will make it easier to change default PHY address depending on PHY models. Submitted by: davidch
2010-10-06 17:35:27 +00:00
/* Poll for the PHY register access to complete. */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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);
Overhaul MII register access routine and remove unnecessary BGE_MI_MODE register accesses. Previously bge(4) used to read BGE_MI_MODE register to detect whether it needs to disable autopolling feature or not. Because we don't touch autopolling in other part of driver there is no reason to read BGE_MI_MODE register given that we know default value in advance. In order to achieve the goal, check whether the controller has CPMU(Central Power Mangement Unit) capability. If controller has CPMU feature, use 500KHz MII management interface(mdio/mdc) frequency regardless core clock frequency. Otherwise use default MII clock. While I'm here, add CPMU register definition. In bge_miibus_readreg(), rearrange code a bit and remove goto statement. In bge_miibus_writereg(), make sure to restore autopolling even if MII write failed. The delay time inserted after accessing BGE_MI_MODE register increased from 40us to 80us. The default PHY address is now stored in softc. All PHYs supported by bge(4) currently uses PHY address 1 but it will be changed when we add newer controllers. This change will make it easier to change default PHY address depending on PHY models. Submitted by: davidch
2010-10-06 17:35:27 +00:00
if ((val & BGE_MICOMM_BUSY) == 0) {
DELAY(5);
val = CSR_READ_4(sc, BGE_MI_COMM);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
Overhaul MII register access routine and remove unnecessary BGE_MI_MODE register accesses. Previously bge(4) used to read BGE_MI_MODE register to detect whether it needs to disable autopolling feature or not. Because we don't touch autopolling in other part of driver there is no reason to read BGE_MI_MODE register given that we know default value in advance. In order to achieve the goal, check whether the controller has CPMU(Central Power Mangement Unit) capability. If controller has CPMU feature, use 500KHz MII management interface(mdio/mdc) frequency regardless core clock frequency. Otherwise use default MII clock. While I'm here, add CPMU register definition. In bge_miibus_readreg(), rearrange code a bit and remove goto statement. In bge_miibus_writereg(), make sure to restore autopolling even if MII write failed. The delay time inserted after accessing BGE_MI_MODE register increased from 40us to 80us. The default PHY address is now stored in softc. All PHYs supported by bge(4) currently uses PHY address 1 but it will be changed when we add newer controllers. This change will make it easier to change default PHY address depending on PHY models. Submitted by: davidch
2010-10-06 17:35:27 +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
}
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;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
Overhaul MII register access routine and remove unnecessary BGE_MI_MODE register accesses. Previously bge(4) used to read BGE_MI_MODE register to detect whether it needs to disable autopolling feature or not. Because we don't touch autopolling in other part of driver there is no reason to read BGE_MI_MODE register given that we know default value in advance. In order to achieve the goal, check whether the controller has CPMU(Central Power Mangement Unit) capability. If controller has CPMU feature, use 500KHz MII management interface(mdio/mdc) frequency regardless core clock frequency. Otherwise use default MII clock. While I'm here, add CPMU register definition. In bge_miibus_readreg(), rearrange code a bit and remove goto statement. In bge_miibus_writereg(), make sure to restore autopolling even if MII write failed. The delay time inserted after accessing BGE_MI_MODE register increased from 40us to 80us. The default PHY address is now stored in softc. All PHYs supported by bge(4) currently uses PHY address 1 but it will be changed when we add newer controllers. This change will make it easier to change default PHY address depending on PHY models. Submitted by: davidch
2010-10-06 17:35:27 +00:00
/* Restore the autopoll bit if necessary. */
if ((sc->bge_mi_mode & BGE_MIMODE_AUTOPOLL) != 0) {
CSR_WRITE_4(sc, BGE_MI_MODE, sc->bge_mi_mode);
DELAY(80);
- 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
}
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
bge_ape_unlock(sc, sc->bge_phy_ape_lock);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
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
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
if (bge_ape_lock(sc, sc->bge_phy_ape_lock) != 0)
return (0);
Overhaul MII register access routine and remove unnecessary BGE_MI_MODE register accesses. Previously bge(4) used to read BGE_MI_MODE register to detect whether it needs to disable autopolling feature or not. Because we don't touch autopolling in other part of driver there is no reason to read BGE_MI_MODE register given that we know default value in advance. In order to achieve the goal, check whether the controller has CPMU(Central Power Mangement Unit) capability. If controller has CPMU feature, use 500KHz MII management interface(mdio/mdc) frequency regardless core clock frequency. Otherwise use default MII clock. While I'm here, add CPMU register definition. In bge_miibus_readreg(), rearrange code a bit and remove goto statement. In bge_miibus_writereg(), make sure to restore autopolling even if MII write failed. The delay time inserted after accessing BGE_MI_MODE register increased from 40us to 80us. The default PHY address is now stored in softc. All PHYs supported by bge(4) currently uses PHY address 1 but it will be changed when we add newer controllers. This change will make it easier to change default PHY address depending on PHY models. Submitted by: davidch
2010-10-06 17:35:27 +00:00
/* Clear the autopoll bit if set, otherwise may trigger PCI errors. */
if ((sc->bge_mi_mode & BGE_MIMODE_AUTOPOLL) != 0) {
CSR_WRITE_4(sc, BGE_MI_MODE,
sc->bge_mi_mode & ~BGE_MIMODE_AUTOPOLL);
DELAY(80);
- 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
}
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
}
Overhaul MII register access routine and remove unnecessary BGE_MI_MODE register accesses. Previously bge(4) used to read BGE_MI_MODE register to detect whether it needs to disable autopolling feature or not. Because we don't touch autopolling in other part of driver there is no reason to read BGE_MI_MODE register given that we know default value in advance. In order to achieve the goal, check whether the controller has CPMU(Central Power Mangement Unit) capability. If controller has CPMU feature, use 500KHz MII management interface(mdio/mdc) frequency regardless core clock frequency. Otherwise use default MII clock. While I'm here, add CPMU register definition. In bge_miibus_readreg(), rearrange code a bit and remove goto statement. In bge_miibus_writereg(), make sure to restore autopolling even if MII write failed. The delay time inserted after accessing BGE_MI_MODE register increased from 40us to 80us. The default PHY address is now stored in softc. All PHYs supported by bge(4) currently uses PHY address 1 but it will be changed when we add newer controllers. This change will make it easier to change default PHY address depending on PHY models. Submitted by: davidch
2010-10-06 17:35:27 +00:00
/* Restore the autopoll bit if necessary. */
if ((sc->bge_mi_mode & BGE_MIMODE_AUTOPOLL) != 0) {
CSR_WRITE_4(sc, BGE_MI_MODE, sc->bge_mi_mode);
DELAY(80);
}
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
bge_ape_unlock(sc, sc->bge_phy_ape_lock);
Overhaul MII register access routine and remove unnecessary BGE_MI_MODE register accesses. Previously bge(4) used to read BGE_MI_MODE register to detect whether it needs to disable autopolling feature or not. Because we don't touch autopolling in other part of driver there is no reason to read BGE_MI_MODE register given that we know default value in advance. In order to achieve the goal, check whether the controller has CPMU(Central Power Mangement Unit) capability. If controller has CPMU feature, use 500KHz MII management interface(mdio/mdc) frequency regardless core clock frequency. Otherwise use default MII clock. While I'm here, add CPMU register definition. In bge_miibus_readreg(), rearrange code a bit and remove goto statement. In bge_miibus_writereg(), make sure to restore autopolling even if MII write failed. The delay time inserted after accessing BGE_MI_MODE register increased from 40us to 80us. The default PHY address is now stored in softc. All PHYs supported by bge(4) currently uses PHY address 1 but it will be changed when we add newer controllers. This change will make it easier to change default PHY address depending on PHY models. Submitted by: davidch
2010-10-06 17:35:27 +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,
On PHY write error use hex number to show the value. Add more comments.
2012-10-10 01:59:53 +00:00
"PHY write timed out (phy %d, reg %d, val 0x%04x)\n",
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, 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
- 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;
APE firmware touches EMAC Mode and TX/RX MAC Mode registers to keep the MAC connected to the outside world. So keep the accesses atomic.
2012-10-05 03:46:25 +00:00
uint32_t mac_mode, rx_mode, tx_mode;
Don't touch EMAC Mode and TX/RX MAC Mode register when driver is not running.
2012-10-05 03:35:38 +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
sc = device_get_softc(dev);
Don't touch EMAC Mode and TX/RX MAC Mode register when driver is not running.
2012-10-05 03:35:38 +00:00
if ((sc->bge_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
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
mii = device_get_softc(sc->bge_miibus);
Add more checks for resolved link speed in bge_miibus_statchg(). Link UP state could be reported first before actual completion of auto-negotiation. This change makes bge(4) reprogram BGE_MAC_MODE, BGE_TX_MODE and BGE_RX_MODE register only after controller got a valid link.
2010-10-13 21:53:37 +00:00
if ((mii->mii_media_status & (IFM_ACTIVE | IFM_AVALID)) ==
(IFM_ACTIVE | IFM_AVALID)) {
switch (IFM_SUBTYPE(mii->mii_media_active)) {
case IFM_10_T:
case IFM_100_TX:
sc->bge_link = 1;
break;
case IFM_1000_T:
case IFM_1000_SX:
case IFM_2500_SX:
if (sc->bge_asicrev != BGE_ASICREV_BCM5906)
sc->bge_link = 1;
else
sc->bge_link = 0;
break;
default:
sc->bge_link = 0;
break;
}
} else
sc->bge_link = 0;
if (sc->bge_link == 0)
return;
APE firmware touches EMAC Mode and TX/RX MAC Mode registers to keep the MAC connected to the outside world. So keep the accesses atomic.
2012-10-05 03:46:25 +00:00
/*
* APE firmware touches these registers to keep the MAC
* connected to the outside world. Try to keep the
* accesses atomic.
*/
/* Set the port mode (MII/GMII) to match the link speed. */
mac_mode = CSR_READ_4(sc, BGE_MAC_MODE) &
~(BGE_MACMODE_PORTMODE | BGE_MACMODE_HALF_DUPLEX);
tx_mode = CSR_READ_4(sc, BGE_TX_MODE);
rx_mode = CSR_READ_4(sc, BGE_RX_MODE);
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)
APE firmware touches EMAC Mode and TX/RX MAC Mode registers to keep the MAC connected to the outside world. So keep the accesses atomic.
2012-10-05 03:46:25 +00:00
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
APE firmware touches EMAC Mode and TX/RX MAC Mode registers to keep the MAC connected to the outside world. So keep the accesses atomic.
2012-10-05 03:46:25 +00:00
mac_mode |= 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
APE firmware touches EMAC Mode and TX/RX MAC Mode registers to keep the MAC connected to the outside world. So keep the accesses atomic.
2012-10-05 03:46:25 +00:00
/* Set MAC flow control behavior to match link flow control settings. */
tx_mode &= ~BGE_TXMODE_FLOWCTL_ENABLE;
rx_mode &= ~BGE_RXMODE_FLOWCTL_ENABLE;
After r207391, brgphy(4) passes resolved flow-control settings to parent driver. Use that information to configure flow-control. One drawback is there is no way to disable flow-control as we still don't have proper way to not advertise RX/TX pause capability to link partner. But I don't think it would cause severe problems and users can selectively disable flow-control in switch port.
2010-09-29 21:19:25 +00:00
if (IFM_OPTIONS(mii->mii_media_active & IFM_FDX) != 0) {
APE firmware touches EMAC Mode and TX/RX MAC Mode registers to keep the MAC connected to the outside world. So keep the accesses atomic.
2012-10-05 03:46:25 +00:00
if ((IFM_OPTIONS(mii->mii_media_active) & IFM_ETH_TXPAUSE) != 0)
tx_mode |= BGE_TXMODE_FLOWCTL_ENABLE;
if ((IFM_OPTIONS(mii->mii_media_active) & IFM_ETH_RXPAUSE) != 0)
rx_mode |= BGE_RXMODE_FLOWCTL_ENABLE;
} else
mac_mode |= BGE_MACMODE_HALF_DUPLEX;
CSR_WRITE_4(sc, BGE_MAC_MODE, mac_mode);
Add 40 microseconds delay after updating EMAC Mode register as recommended by Broadcom data sheet.
2012-10-05 06:24:22 +00:00
DELAY(40);
APE firmware touches EMAC Mode and TX/RX MAC Mode registers to keep the MAC connected to the outside world. So keep the accesses atomic.
2012-10-05 03:46:25 +00:00
CSR_WRITE_4(sc, BGE_TX_MODE, tx_mode);
CSR_WRITE_4(sc, BGE_RX_MODE, rx_mode);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/*
* 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
Add initial jumbo frame support for BCM5714/BCM5715 and BCM5780. Unlike other controllers which have more advanced jumbo support, these controllers have one send ring, one standard receive producer ring and one receive return ring. In order to receive jumbo frames on the controllers, driver now will increase Rx buffer size to 9k. Two Rx modes are supported on these controllers and I chose standard Rx BDs over extended Rx BDs. The extended Rx BD mode allows up to 4 segmentations for each Rx BDs such that kernel does not have to allocate large buffer of contiguous memory for receiving. The extended Rx BD mode is already used on controllers that have separate jumbo receive ring. However, using extended Rx BDs on BCM5714/BCM5715/BCM5780 reduces the number of Rx BDs to 256 entries which in turn may reduce the performance. Also UMA backed page allocator for jumbo frame returns contiguous memory so using extended Rx BD has no advantage on FreeBSD unless highly customized local allocator implemented in driver is used. To use jumbo buffers in standard receive ring, Rx buffer allocation handler was changed to allocate MJUM9BYTES sized mbuf. PR: kern/155192 Tested by: Vijay Singh <vijju.singh <> gmail dot com> Submitted by: mjacob (initial version)
2011-04-05 17:41:54 +00:00
if (sc->bge_flags & BGE_FLAG_JUMBO_STD &&
(sc->bge_ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN +
ETHER_VLAN_ENCAP_LEN > (MCLBYTES - ETHER_ALIGN))) {
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
m = m_getjcl(M_NOWAIT, MT_DATA, M_PKTHDR, MJUM9BYTES);
Add initial jumbo frame support for BCM5714/BCM5715 and BCM5780. Unlike other controllers which have more advanced jumbo support, these controllers have one send ring, one standard receive producer ring and one receive return ring. In order to receive jumbo frames on the controllers, driver now will increase Rx buffer size to 9k. Two Rx modes are supported on these controllers and I chose standard Rx BDs over extended Rx BDs. The extended Rx BD mode allows up to 4 segmentations for each Rx BDs such that kernel does not have to allocate large buffer of contiguous memory for receiving. The extended Rx BD mode is already used on controllers that have separate jumbo receive ring. However, using extended Rx BDs on BCM5714/BCM5715/BCM5780 reduces the number of Rx BDs to 256 entries which in turn may reduce the performance. Also UMA backed page allocator for jumbo frame returns contiguous memory so using extended Rx BD has no advantage on FreeBSD unless highly customized local allocator implemented in driver is used. To use jumbo buffers in standard receive ring, Rx buffer allocation handler was changed to allocate MJUM9BYTES sized mbuf. PR: kern/155192 Tested by: Vijay Singh <vijju.singh <> gmail dot com> Submitted by: mjacob (initial version)
2011-04-05 17:41:54 +00:00
if (m == NULL)
return (ENOBUFS);
m->m_len = m->m_pkthdr.len = MJUM9BYTES;
} else {
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
Add initial jumbo frame support for BCM5714/BCM5715 and BCM5780. Unlike other controllers which have more advanced jumbo support, these controllers have one send ring, one standard receive producer ring and one receive return ring. In order to receive jumbo frames on the controllers, driver now will increase Rx buffer size to 9k. Two Rx modes are supported on these controllers and I chose standard Rx BDs over extended Rx BDs. The extended Rx BD mode allows up to 4 segmentations for each Rx BDs such that kernel does not have to allocate large buffer of contiguous memory for receiving. The extended Rx BD mode is already used on controllers that have separate jumbo receive ring. However, using extended Rx BDs on BCM5714/BCM5715/BCM5780 reduces the number of Rx BDs to 256 entries which in turn may reduce the performance. Also UMA backed page allocator for jumbo frame returns contiguous memory so using extended Rx BD has no advantage on FreeBSD unless highly customized local allocator implemented in driver is used. To use jumbo buffers in standard receive ring, Rx buffer allocation handler was changed to allocate MJUM9BYTES sized mbuf. PR: kern/155192 Tested by: Vijay Singh <vijju.singh <> gmail dot com> Submitted by: mjacob (initial version)
2011-04-05 17:41:54 +00:00
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;
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
MGETHDR(m, M_NOWAIT, MT_DATA);
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 (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
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
m_cljget(m, M_NOWAIT, MJUM9BYTES);
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 (!(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
Fix an logic bug which caused jumbo buffers to not be synced. Reported and tested by: Michael Moll MFC after: 3 days
2011-05-02 20:45:19 +00:00
if (sc->bge_cdata.bge_rx_jumbo_chain[i] != NULL) {
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_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);
Remove extra semicolon.
2010-10-01 17:51:55 +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
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since 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);
Remove extra semicolon.
2010-10-01 17:51:55 +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
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since 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
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
/* Enable the jumbo receive producer 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
rcb = &sc->bge_ldata.bge_info.bge_jumbo_rx_rcb;
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +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)
Rename definition of BGE_SOFTWARE_GENCOMM_* to more readable ones. The origin of GENCOMM seems to come from Alteon Tigon Host/NIC interface definition where it defines general communications region which is active when firmware is loaded and running. This region was used in communication between the host and processor internal to the Tigon chip. Broadcom data sheet also defines the region as 'Software Gencomm' in NetXtreme memory map but lacks detailed description of its interface so it was hard to know which ones are used for which interface. This change shall slightly enhance readability. No functional changes.
2011-10-26 21:05:45 +00:00
bge_writemem_ind(sc, BGE_SRAM_FW_MB, BGE_SRAM_FW_MB_MAGIC);
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:
SRAM offset 0x0C04 is used by driver to inform the IPMI/ASF firmware about the various driver events like load, unload, reset, suspend, restart, and ioctl operations. Define driver's event rather than using hard-coded values. We don't still send suspend/resume event to firmware. Previously bge(4) used BGE_SDI_STATUS to send events. Because driver has to access firmware mail box to inform current state, using BGE_SDI_STATUS register was wrong. The end result was the same as BGE_SDI_STATUS is 0x0C04. No functional changes.
2011-10-26 23:52:02 +00:00
bge_writemem_ind(sc, BGE_SRAM_FW_DRV_STATE_MB,
BGE_FW_DRV_STATE_START);
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;
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
case BGE_RESET_SHUTDOWN:
SRAM offset 0x0C04 is used by driver to inform the IPMI/ASF firmware about the various driver events like load, unload, reset, suspend, restart, and ioctl operations. Define driver's event rather than using hard-coded values. We don't still send suspend/resume event to firmware. Previously bge(4) used BGE_SDI_STATUS to send events. Because driver has to access firmware mail box to inform current state, using BGE_SDI_STATUS register was wrong. The end result was the same as BGE_SDI_STATUS is 0x0C04. No functional changes.
2011-10-26 23:52:02 +00:00
bge_writemem_ind(sc, BGE_SRAM_FW_DRV_STATE_MB,
BGE_FW_DRV_STATE_UNLOAD);
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;
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
case BGE_RESET_SUSPEND:
bge_writemem_ind(sc, BGE_SRAM_FW_DRV_STATE_MB,
BGE_FW_DRV_STATE_SUSPEND);
break;
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 APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
if (type == BGE_RESET_START || type == BGE_RESET_SUSPEND)
bge_ape_driver_state_change(sc, 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
}
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:
SRAM offset 0x0C04 is used by driver to inform the IPMI/ASF firmware about the various driver events like load, unload, reset, suspend, restart, and ioctl operations. Define driver's event rather than using hard-coded values. We don't still send suspend/resume event to firmware. Previously bge(4) used BGE_SDI_STATUS to send events. Because driver has to access firmware mail box to inform current state, using BGE_SDI_STATUS register was wrong. The end result was the same as BGE_SDI_STATUS is 0x0C04. No functional changes.
2011-10-26 23:52:02 +00:00
bge_writemem_ind(sc, BGE_SRAM_FW_DRV_STATE_MB,
BGE_FW_DRV_STATE_START_DONE);
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;
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
case BGE_RESET_SHUTDOWN:
SRAM offset 0x0C04 is used by driver to inform the IPMI/ASF firmware about the various driver events like load, unload, reset, suspend, restart, and ioctl operations. Define driver's event rather than using hard-coded values. We don't still send suspend/resume event to firmware. Previously bge(4) used BGE_SDI_STATUS to send events. Because driver has to access firmware mail box to inform current state, using BGE_SDI_STATUS register was wrong. The end result was the same as BGE_SDI_STATUS is 0x0C04. No functional changes.
2011-10-26 23:52:02 +00:00
bge_writemem_ind(sc, BGE_SRAM_FW_DRV_STATE_MB,
BGE_FW_DRV_STATE_UNLOAD_DONE);
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;
}
}
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
if (type == BGE_RESET_SHUTDOWN)
bge_ape_driver_state_change(sc, 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
}
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:
SRAM offset 0x0C04 is used by driver to inform the IPMI/ASF firmware about the various driver events like load, unload, reset, suspend, restart, and ioctl operations. Define driver's event rather than using hard-coded values. We don't still send suspend/resume event to firmware. Previously bge(4) used BGE_SDI_STATUS to send events. Because driver has to access firmware mail box to inform current state, using BGE_SDI_STATUS register was wrong. The end result was the same as BGE_SDI_STATUS is 0x0C04. No functional changes.
2011-10-26 23:52:02 +00:00
bge_writemem_ind(sc, BGE_SRAM_FW_DRV_STATE_MB,
BGE_FW_DRV_STATE_START);
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;
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
case BGE_RESET_SHUTDOWN:
SRAM offset 0x0C04 is used by driver to inform the IPMI/ASF firmware about the various driver events like load, unload, reset, suspend, restart, and ioctl operations. Define driver's event rather than using hard-coded values. We don't still send suspend/resume event to firmware. Previously bge(4) used BGE_SDI_STATUS to send events. Because driver has to access firmware mail box to inform current state, using BGE_SDI_STATUS register was wrong. The end result was the same as BGE_SDI_STATUS is 0x0C04. No functional changes.
2011-10-26 23:52:02 +00:00
bge_writemem_ind(sc, BGE_SRAM_FW_DRV_STATE_MB,
BGE_FW_DRV_STATE_UNLOAD);
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;
}
}
}
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) {
Rename BGE_FW_DRV_ALIVE/BGE_FW_PAUSE to BGE_FW_CMD_DRV_ALIVE/BGE_FW_CMD_PAUSE. Also add more firmware commands(not used yet). No functional changes.
2011-10-27 20:54:53 +00:00
bge_writemem_ind(sc, BGE_SRAM_FW_CMD_MB, BGE_FW_CMD_PAUSE);
Offset 0x6810 is RX-RISC event register. Rename BGE_CPU_EVENT with BGE_RX_CPU_EVENT for readability. Additionally define BGE_TX_CPU_EVENT for TX-RSIC event register(BCM570[0-4] only).
2011-10-26 23:22:32 +00:00
CSR_WRITE_4(sc, BGE_RX_CPU_EVENT,
Rename hard-coded value 1 << 14 with BGE_RX_CPU_DRV_EVENT. This bit(SW event 7 in publicly available data sheet) is used to make RX CPU handle a firmware command and the bit is automatically cleared after RX CPU completed the command. Generally firmware command takes the following steps. 1. Write BGE_SRAM_FW_CMD_MB with a command. 2. Write BGE_SRAM_FW_CMD_LEN_MB with the length of the command in bytes. 3. Write BGE_SRAM_FW_CMD_DATA_MB with actual command data. 4. Generate BGE_RX_CPU_EVENT and let firmware handle the command. 5. Wait for the ACK of the firmware command. No functional changes.
2011-10-27 21:27:37 +00:00
CSR_READ_4(sc, BGE_RX_CPU_EVENT) | BGE_RX_CPU_DRV_EVENT);
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++ ) {
Rename hard-coded value 1 << 14 with BGE_RX_CPU_DRV_EVENT. This bit(SW event 7 in publicly available data sheet) is used to make RX CPU handle a firmware command and the bit is automatically cleared after RX CPU completed the command. Generally firmware command takes the following steps. 1. Write BGE_SRAM_FW_CMD_MB with a command. 2. Write BGE_SRAM_FW_CMD_LEN_MB with the length of the command in bytes. 3. Write BGE_SRAM_FW_CMD_DATA_MB with actual command data. 4. Generate BGE_RX_CPU_EVENT and let firmware handle the command. 5. Wait for the ACK of the firmware command. No functional changes.
2011-10-27 21:27:37 +00:00
if (!(CSR_READ_4(sc, BGE_RX_CPU_EVENT) &
BGE_RX_CPU_DRV_EVENT))
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;
DELAY(10);
}
}
}
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
static uint32_t
bge_dma_swap_options(struct bge_softc *sc)
{
uint32_t dma_options;
dma_options = BGE_MODECTL_WORDSWAP_NONFRAME |
BGE_MODECTL_BYTESWAP_DATA | BGE_MODECTL_WORDSWAP_DATA;
#if BYTE_ORDER == BIG_ENDIAN
dma_options |= BGE_MODECTL_BYTESWAP_NONFRAME;
#endif
return (dma_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
/*
- 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
{
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
uint32_t dma_rw_ctl, misc_ctl, mode_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 initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
misc_ctl = BGE_INIT;
if (sc->bge_flags & BGE_FLAG_TAGGED_STATUS)
misc_ctl |= BGE_PCIMISCCTL_TAGGED_STATUS;
pci_write_config(sc->bge_dev, BGE_PCI_MISC_CTL, misc_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
/*
* 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) {
If the maximum payload size is 256 bytes or more, set the DMA write water mark to 256 bytes. Otherwise controller will encounter DMA write under run errors and would result in RX DMA hang. If the maximum payload size is 128 bytes, the water mark is set to 128 bytes as usual. While here, set maximum read request size to 2048 for BCM5719/BCM5720. For other PCIe devices, use 4096. And reprogram the maximum read request size whenever device reset is performed.
2012-10-10 01:24:02 +00:00
if (sc->bge_mps >= 256)
dma_rw_ctl |= BGE_PCIDMARWCTL_WR_WAT_SHIFT(7);
else
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;
Add initial BCM57765 family support. The BCM57765 family seems to have similar hardware features of BCM5718 family except the number of receive return ring is 4. The BCM57765 family is known to support IEEE 802.3az EEE(Energy Efficient Ethernet) but this change does not include EEE support code. I hope EEE is implemented in near future. This change will support BCM57761, BCM57765, BCM57781, BCM57785, BCM57791 and BCM57795. All hardware offloading features are supported and suspend/resume also should work. Many thanks to Broadcom for continuing support of FreeBSD. Tested by: Paul Thornton (prt <> prt dot org) HW donated by: Broadcom
2011-05-04 17:04:31 +00:00
if (BGE_IS_5717_PLUS(sc)) {
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
dma_rw_ctl &= ~BGE_PCIDMARWCTL_DIS_CACHE_ALIGNMENT;
Add initial BCM57765 family support. The BCM57765 family seems to have similar hardware features of BCM5718 family except the number of receive return ring is 4. The BCM57765 family is known to support IEEE 802.3az EEE(Energy Efficient Ethernet) but this change does not include EEE support code. I hope EEE is implemented in near future. This change will support BCM57761, BCM57765, BCM57781, BCM57785, BCM57791 and BCM57795. All hardware offloading features are supported and suspend/resume also should work. Many thanks to Broadcom for continuing support of FreeBSD. Tested by: Paul Thornton (prt <> prt dot org) HW donated by: Broadcom
2011-05-04 17:04:31 +00:00
if (sc->bge_chipid == BGE_CHIPID_BCM57765_A0)
dma_rw_ctl &= ~BGE_PCIDMARWCTL_CRDRDR_RDMA_MRRS_MSK;
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
/*
* Enable HW workaround for controllers that misinterpret
* a status tag update and leave interrupts permanently
* disabled.
*/
if (sc->bge_asicrev != BGE_ASICREV_BCM5717 &&
sc->bge_asicrev != BGE_ASICREV_BCM57765)
dma_rw_ctl |= BGE_PCIDMARWCTL_TAGGED_STATUS_WA;
Add initial BCM57765 family support. The BCM57765 family seems to have similar hardware features of BCM5718 family except the number of receive return ring is 4. The BCM57765 family is known to support IEEE 802.3az EEE(Energy Efficient Ethernet) but this change does not include EEE support code. I hope EEE is implemented in near future. This change will support BCM57761, BCM57765, BCM57781, BCM57785, BCM57791 and BCM57795. All hardware offloading features are supported and suspend/resume also should work. Many thanks to Broadcom for continuing support of FreeBSD. Tested by: Paul Thornton (prt <> prt dot org) HW donated by: Broadcom
2011-05-04 17:04:31 +00:00
}
- 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
pci_write_config(sc->bge_dev, BGE_PCI_DMA_RW_CTL, dma_rw_ctl, 4);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Set up general mode register.
*/
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
mode_ctl = bge_dma_swap_options(sc);
if (sc->bge_asicrev == BGE_ASICREV_BCM5720) {
/* Retain Host-2-BMC settings written by APE firmware. */
mode_ctl |= CSR_READ_4(sc, BGE_MODE_CTL) &
(BGE_MODECTL_BYTESWAP_B2HRX_DATA |
BGE_MODECTL_WORDSWAP_B2HRX_DATA |
BGE_MODECTL_B2HRX_ENABLE | BGE_MODECTL_HTX2B_ENABLE);
}
mode_ctl |= BGE_MODECTL_MAC_ATTN_INTR | BGE_MODECTL_HOST_SEND_BDS |
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)
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
mode_ctl |= BGE_MODECTL_FORCE_PCI32;
- 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
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)
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
mode_ctl |= BGE_MODECTL_STACKUP;
CSR_WRITE_4(sc, BGE_MODE_CTL, mode_ctl);
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
/*
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
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
/* Set the timer prescaler (always 66 MHz). */
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;
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
uint32_t dmactl, val;
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
int i, limit;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 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 initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
if (BGE_IS_5717_PLUS(sc)) {
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_READDMA_LOWAT, 0x0);
if (sc->bge_ifp->if_mtu > ETHERMTU) {
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_MACRX_LOWAT, 0x7e);
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_HIWAT, 0xea);
} else {
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_MACRX_LOWAT, 0x2a);
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_HIWAT, 0xa0);
}
} else 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 initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
val = BGE_BMANMODE_ENABLE | BGE_BMANMODE_LOMBUF_ATTN;
/*
* Change the arbitration algorithm of TXMBUF read request to
* round-robin instead of priority based for BCM5719. When
* TXFIFO is almost empty, RDMA will hold its request until
* TXFIFO is not almost empty.
*/
if (sc->bge_asicrev == BGE_ASICREV_BCM5719)
val |= BGE_BMANMODE_NO_TX_UNDERRUN;
CSR_WRITE_4(sc, BGE_BMAN_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
Since r117657, bge(4) does not enable buffer manager for BCM5705 or newer controllers. However, all data sheet I have access has no indication that buffer manager should not be touched on these controllers. It seems the buffer manager always runs on BCM5705 or newer controllers. Some controller(e.g. BCM5719) needs other buffer manager configuration so driver should enable buffer manager for all controllers. Both Linux and OpenBSD/NetBSD use the same approach. This change polls enable bit of block to know whether specified block was really stopped as well as enabling buffer manager for all controllers in driver initialization. Obtained from: NetBSD
2011-05-09 20:10:46 +00:00
/* Poll for buffer manager start indication */
for (i = 0; i < BGE_TIMEOUT; i++) {
DELAY(10);
if (CSR_READ_4(sc, BGE_BMAN_MODE) & BGE_BMANMODE_ENABLE)
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
Since r117657, bge(4) does not enable buffer manager for BCM5705 or newer controllers. However, all data sheet I have access has no indication that buffer manager should not be touched on these controllers. It seems the buffer manager always runs on BCM5705 or newer controllers. Some controller(e.g. BCM5719) needs other buffer manager configuration so driver should enable buffer manager for all controllers. Both Linux and OpenBSD/NetBSD use the same approach. This change polls enable bit of block to know whether specified block was really stopped as well as enabling buffer manager for all controllers in driver initialization. Obtained from: NetBSD
2011-05-09 20:10:46 +00:00
if (i == BGE_TIMEOUT) {
device_printf(sc->bge_dev, "buffer manager failed to start\n");
return (ENXIO);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/* Enable flow-through queues */
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
}
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
/*
* Summary of rings supported by the controller:
*
* Standard Receive Producer Ring
* - This ring is used to feed receive buffers for "standard"
* sized frames (typically 1536 bytes) to the controller.
*
* Jumbo Receive Producer Ring
* - This ring is used to feed receive buffers for jumbo sized
* frames (i.e. anything bigger than the "standard" frames)
* to the controller.
*
* Mini Receive Producer Ring
* - This ring is used to feed receive buffers for "mini"
* sized frames to the controller.
* - This feature required external memory for the controller
* but was never used in a production system. Should always
* be disabled.
*
* Receive Return Ring
* - After the controller has placed an incoming frame into a
* receive buffer that buffer is moved into a receive return
* ring. The driver is then responsible to passing the
* buffer up to the stack. Many versions of the controller
* support multiple RR rings.
*
* Send Ring
* - This ring is used for outgoing frames. Many versions of
* the controller support multiple send rings.
*/
/* Initialize the standard receive producer 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 initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if (BGE_IS_5717_PLUS(sc)) {
/*
* Bits 31-16: Programmable ring size (2048, 1024, 512, .., 32)
* Bits 15-2 : Maximum RX frame size
* Bit 1 : 1 = Ring Disabled, 0 = Ring ENabled
* Bit 0 : Reserved
*/
rcb->bge_maxlen_flags =
BGE_RCB_MAXLEN_FLAGS(512, BGE_MAX_FRAMELEN << 2);
} else if (BGE_IS_5705_PLUS(sc)) {
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
/*
* Bits 31-16: Programmable ring size (512, 256, 128, 64, 32)
* Bits 15-2 : Reserved (should be 0)
* Bit 1 : 1 = Ring Disabled, 0 = Ring Enabled
* Bit 0 : Reserved
*/
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);
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
} else {
/*
* Ring size is always XXX entries
* Bits 31-16: Maximum RX frame size
* Bits 15-2 : Reserved (should be 0)
* Bit 1 : 1 = Ring Disabled, 0 = Ring Enabled
* Bit 0 : Reserved
*/
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
rcb->bge_maxlen_flags =
BGE_RCB_MAXLEN_FLAGS(BGE_MAX_FRAMELEN, 0);
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
}
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5717 ||
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
sc->bge_asicrev == BGE_ASICREV_BCM5719 ||
sc->bge_asicrev == BGE_ASICREV_BCM5720)
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
rcb->bge_nicaddr = BGE_STD_RX_RINGS_5717;
else
rcb->bge_nicaddr = BGE_STD_RX_RINGS;
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
/* Write the standard receive producer ring control block. */
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);
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
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
/* Reset the standard receive producer ring producer index. */
bge_writembx(sc, BGE_MBX_RX_STD_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
/*
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
* Initialize the jumbo RX producer ring control
* block. We set the 'ring disabled' bit in the
* flags field until we're actually ready to 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
* 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;
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
/* Get the jumbo receive producer ring RCB parameters. */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since 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->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);
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5717 ||
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
sc->bge_asicrev == BGE_ASICREV_BCM5719 ||
sc->bge_asicrev == BGE_ASICREV_BCM5720)
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
rcb->bge_nicaddr = BGE_JUMBO_RX_RINGS_5717;
else
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);
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
/* Program the jumbo receive producer ring RCB parameters. */
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);
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
/* Reset the jumbo receive producer ring producer index. */
bge_writembx(sc, BGE_MBX_RX_JUMBO_PROD_LO, 0);
}
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 more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
/* Disable the mini receive producer ring RCB. */
Fix a regression introduced in r213495. r213495 disabled mini receive producer ring only for BCM5700. It was believed that BCM5700 with external SSRAM is the only controller that supports mini ring but it seems all BCM570[0-4] requires to disable mini receive producer ring. Otherwise, it caused unexpected RX DMA error or watchdog timeouts. Reported by: marius, Steve Kargl <sgk <> troutmask dot apl dot washington dot edu> Tested by: marius, Steve Kargl <sgk <> troutmask dot apl dot washington dot edu>
2010-10-12 19:22:03 +00:00
if (BGE_IS_5700_FAMILY(sc)) {
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 more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
/* Reset the mini receive producer ring producer index. */
bge_writembx(sc, BGE_MBX_RX_MINI_PROD_LO, 0);
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
Apply the same workaround for SDI flow control used on BCM5906 A1 to BCM6906 A0/A2. This should fix a long standing BCM5906 A2 lockup issues. Data sheet explicitly mentions BCM5906 A0, A1 and A2 use de-pipelined mode on these revisions. Special thanks to Buganini who tried all combinations of experimental patches for more than 10 days. Tested by: Buganini <buganini <> gmail dot com >
2010-10-23 21:25:50 +00:00
/* Choose de-pipeline mode for BCM5906 A0, A1 and A2. */
if (sc->bge_asicrev == BGE_ASICREV_BCM5906) {
Use bge_chipid to compare controller ids. r214251 incorrectly used bge_chiprev. Reported by: Buganini <buganini <> gmail dot com >
2010-10-24 20:54:46 +00:00
if (sc->bge_chipid == BGE_CHIPID_BCM5906_A0 ||
sc->bge_chipid == BGE_CHIPID_BCM5906_A1 ||
sc->bge_chipid == BGE_CHIPID_BCM5906_A2)
Apply the same workaround for SDI flow control used on BCM5906 A1 to BCM6906 A0/A2. This should fix a long standing BCM5906 A2 lockup issues. Data sheet explicitly mentions BCM5906 A0, A1 and A2 use de-pipelined mode on these revisions. Special thanks to Buganini who tried all combinations of experimental patches for more than 10 days. Tested by: Buganini <buganini <> gmail dot com >
2010-10-23 21:25:50 +00:00
CSR_WRITE_4(sc, BGE_ISO_PKT_TX,
(CSR_READ_4(sc, BGE_ISO_PKT_TX) & ~3) | 2);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
* The BD ring replenish thresholds control how often the
* hardware fetches new BD's from the producer rings in host
* memory. Setting the value too low on a busy system can
* starve the hardware and recue the throughpout.
*
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if (BGE_IS_5717_PLUS(sc)) {
CSR_WRITE_4(sc, BGE_STD_REPLENISH_LWM, 32);
CSR_WRITE_4(sc, BGE_JMB_REPLENISH_LWM, 16);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
* Disable all send rings by setting the 'ring disabled' bit
* in the flags field of all the TX send ring control blocks,
* located in NIC memory.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
if (!BGE_IS_5705_PLUS(sc))
/* 5700 to 5704 had 16 send rings. */
limit = BGE_TX_RINGS_EXTSSRAM_MAX;
else
limit = 1;
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 more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
for (i = 0; i < limit; 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
}
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
/* Configure send ring 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);
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5717 ||
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
sc->bge_asicrev == BGE_ASICREV_BCM5719 ||
sc->bge_asicrev == BGE_ASICREV_BCM5720)
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
RCB_WRITE_4(sc, vrcb, bge_nicaddr, BGE_SEND_RING_5717);
else
RCB_WRITE_4(sc, vrcb, bge_nicaddr,
BGE_NIC_TXRING_ADDR(0, BGE_TX_RING_CNT));
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +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
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
/*
* Disable all receive return rings by setting the
* 'ring diabled' bit in the flags field of all the receive
* return ring control blocks, located in NIC memory.
*/
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5717 ||
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
sc->bge_asicrev == BGE_ASICREV_BCM5719 ||
sc->bge_asicrev == BGE_ASICREV_BCM5720) {
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
/* Should be 17, use 16 until we get an SRAM map. */
limit = 16;
} else if (!BGE_IS_5705_PLUS(sc))
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
limit = BGE_RX_RINGS_MAX;
Add initial BCM57765 family support. The BCM57765 family seems to have similar hardware features of BCM5718 family except the number of receive return ring is 4. The BCM57765 family is known to support IEEE 802.3az EEE(Energy Efficient Ethernet) but this change does not include EEE support code. I hope EEE is implemented in near future. This change will support BCM57761, BCM57765, BCM57781, BCM57785, BCM57791 and BCM57795. All hardware offloading features are supported and suspend/resume also should work. Many thanks to Broadcom for continuing support of FreeBSD. Tested by: Paul Thornton (prt <> prt dot org) HW donated by: Broadcom
2011-05-04 17:04:31 +00:00
else if (sc->bge_asicrev == BGE_ASICREV_BCM5755 ||
Add preliminary support for BCM57766 ASIC. While I'm here add BCM57762 device id which is found on Apple Thunderbolt ethernet. Tested by: Richard Kuhns < rjk <> wintek dot com >
2012-11-30 01:38:00 +00:00
BGE_IS_57765_PLUS(sc))
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
limit = 4;
else
limit = 1;
/* Disable all receive 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 more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
for (i = 0; i < limit; 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 more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
BGE_RCB_FLAG_RING_DISABLED);
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_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
}
/*
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
* Set up receive return ring 0. Note that the NIC address
* for RX return rings is 0x0. The return rings live entirely
* within the host, so the nicaddr field in the RCB isn't used.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
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);
Add more comments to rings supported by the controller. Different versions of controller support different number of ring control blocks such that adjust code a bit to access known number of send/receive ring control blocks. Previously bge(4) blindly accessed 16 send/receive RCBs. Also move initializing standard receive producer ring producer index, jumbo receive producer ring producer index and mini receive producer ring producer index to the end of each receive producer ring initialization. Do not assume mini receive producer ring is available only when controller has jumbo frame capability, instead explicitly check ASIC version BCM5700 to disable mini receive producer ring. Additionally always enable send ring 0 regardless of controller versions. Previously bge(4) didn't enable send ring 0 if controller is BGE_IS_5705_PLUS. Becase bge(4) need 1 send ring to send frames at least, I have no idea how it would have worked so far. Submitted by: davidch
2010-10-06 21:23:57 +00:00
RCB_WRITE_4(sc, vrcb, bge_nicaddr, 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
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 */
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
val = 0x2620;
if (sc->bge_asicrev == BGE_ASICREV_BCM5720)
val |= CSR_READ_4(sc, BGE_TX_LENGTHS) &
(BGE_TXLEN_JMB_FRM_LEN_MSK | BGE_TXLEN_CNT_DN_VAL_MSK);
CSR_WRITE_4(sc, BGE_TX_LENGTHS, 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
/*
* 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
On PHY write error use hex number to show the value. Add more comments.
2012-10-10 01:59:53 +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
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 APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
/* Allow APE to send/receive frames. */
if ((sc->bge_mfw_flags & BGE_MFW_ON_APE) != 0)
val |= BGE_MACMODE_APE_RX_EN | BGE_MACMODE_APE_TX_EN;
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 40 microseconds delay after updating EMAC Mode register as recommended by Broadcom data sheet.
2012-10-05 06:24:22 +00:00
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
/* 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
Allow write DMA to request larger DMA burst size to get better performance on BCM5785. Obtained from: OpenBSD
2010-10-01 17:46:15 +00:00
/* Request larger DMA burst size to get better performance. */
if (sc->bge_asicrev == BGE_ASICREV_BCM5785)
val |= BGE_WDMAMODE_BURST_ALL_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
/* 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 initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5717)
val |= BGE_RDMAMODE_MULT_DMA_RD_DIS;
- 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;
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if (sc->bge_flags & (BGE_FLAG_TSO | BGE_FLAG_TSO3)) {
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;
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if (sc->bge_flags & BGE_FLAG_TSO3 ||
sc->bge_asicrev == BGE_ASICREV_BCM5785 ||
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
sc->bge_asicrev == BGE_ASICREV_BCM57780)
val |= BGE_RDMAMODE_TSO6_ENABLE;
}
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
BCM5720 performance tweak from Broadcom. o Allow multiple outstanding read requests from non-LSO read DMA engine. o Allow 4KB burst length reads for non-LSO frames. o Enable 512B burst length reads for buffer descriptors. Submitted by: Geans Pin < geanspin <> broadcom dot com >
2011-12-13 20:26:46 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5720) {
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
val |= CSR_READ_4(sc, BGE_RDMA_MODE) &
BGE_RDMAMODE_H2BNC_VLAN_DET;
BCM5720 performance tweak from Broadcom. o Allow multiple outstanding read requests from non-LSO read DMA engine. o Allow 4KB burst length reads for non-LSO frames. o Enable 512B burst length reads for buffer descriptors. Submitted by: Geans Pin < geanspin <> broadcom dot com >
2011-12-13 20:26:46 +00:00
/*
* Allow multiple outstanding read requests from
* non-LSO read DMA engine.
*/
val &= ~BGE_RDMAMODE_MULT_DMA_RD_DIS;
}
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
Enable fix for read DMA FIFO overruns on controllers that have this fix. Note, we still need workaround for controllers that lacks this fix and it needs more work in RX BD updating. Submitted by: davidch
2010-10-04 18:09:01 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5761 ||
sc->bge_asicrev == BGE_ASICREV_BCM5784 ||
sc->bge_asicrev == BGE_ASICREV_BCM5785 ||
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
sc->bge_asicrev == BGE_ASICREV_BCM57780 ||
BGE_IS_5717_PLUS(sc)) {
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
dmactl = CSR_READ_4(sc, BGE_RDMA_RSRVCTRL);
/*
* Adjust tx margin to prevent TX data corruption and
* fix internal FIFO overflow.
*/
Limit applying TX data corruption and FIFO workaround to BCM5719 A0. It's believed BCM5720 does not have the issue.
2012-10-08 07:13:45 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5719 &&
sc->bge_chipid == BGE_CHIPID_BCM5719_A0) {
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
dmactl &= ~(BGE_RDMA_RSRVCTRL_FIFO_LWM_MASK |
BGE_RDMA_RSRVCTRL_FIFO_HWM_MASK |
BGE_RDMA_RSRVCTRL_TXMRGN_MASK);
dmactl |= BGE_RDMA_RSRVCTRL_FIFO_LWM_1_5K |
BGE_RDMA_RSRVCTRL_FIFO_HWM_1_5K |
BGE_RDMA_RSRVCTRL_TXMRGN_320B;
}
Enable fix for read DMA FIFO overruns on controllers that have this fix. Note, we still need workaround for controllers that lacks this fix and it needs more work in RX BD updating. Submitted by: davidch
2010-10-04 18:09:01 +00:00
/*
* Enable fix for read DMA FIFO overruns.
* The fix is to limit the number of RX BDs
* the hardware would fetch at a fime.
*/
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
CSR_WRITE_4(sc, BGE_RDMA_RSRVCTRL, dmactl |
Enable fix for read DMA FIFO overruns on controllers that have this fix. Note, we still need workaround for controllers that lacks this fix and it needs more work in RX BD updating. Submitted by: davidch
2010-10-04 18:09:01 +00:00
BGE_RDMA_RSRVCTRL_FIFO_OFLW_FIX);
}
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
BCM5720 performance tweak from Broadcom. o Allow multiple outstanding read requests from non-LSO read DMA engine. o Allow 4KB burst length reads for non-LSO frames. o Enable 512B burst length reads for buffer descriptors. Submitted by: Geans Pin < geanspin <> broadcom dot com >
2011-12-13 20:26:46 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5719) {
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
CSR_WRITE_4(sc, BGE_RDMA_LSO_CRPTEN_CTRL,
CSR_READ_4(sc, BGE_RDMA_LSO_CRPTEN_CTRL) |
BGE_RDMA_LSO_CRPTEN_CTRL_BLEN_BD_4K |
BGE_RDMA_LSO_CRPTEN_CTRL_BLEN_LSO_4K);
BCM5720 performance tweak from Broadcom. o Allow multiple outstanding read requests from non-LSO read DMA engine. o Allow 4KB burst length reads for non-LSO frames. o Enable 512B burst length reads for buffer descriptors. Submitted by: Geans Pin < geanspin <> broadcom dot com >
2011-12-13 20:26:46 +00:00
} else if (sc->bge_asicrev == BGE_ASICREV_BCM5720) {
/*
* Allow 4KB burst length reads for non-LSO frames.
* Enable 512B burst length reads for buffer descriptors.
*/
CSR_WRITE_4(sc, BGE_RDMA_LSO_CRPTEN_CTRL,
CSR_READ_4(sc, BGE_RDMA_LSO_CRPTEN_CTRL) |
BGE_RDMA_LSO_CRPTEN_CTRL_BLEN_BD_512 |
BGE_RDMA_LSO_CRPTEN_CTRL_BLEN_LSO_4K);
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
}
- 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 */
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if (sc->bge_flags & (BGE_FLAG_TSO | BGE_FLAG_TSO3))
CSR_WRITE_4(sc, BGE_SDI_MODE, BGE_SDIMODE_ENABLE |
BGE_SDIMODE_HW_LSO_PRE_DMA);
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
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
Remove one last reference of BGE_MI_MODE register for auto polling. Previously bge(4) always enabled auto polling for non-BGE_FLAG_TBI controllers. With this change, auto polling is not used anymore so polling through mii(4) was introduced. Reviewed by: davidch
2010-10-11 22:56:23 +00:00
/*
* Enable attention when the link has changed state for
* devices that use auto polling.
*/
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 a regression introduced in r213710. r213710 removed the use of auto polling such that it made all controllers obtain link status information from the state of the LNKRDY input signal. Broadcom recommends disabling auto polling such that driver should rely on PHY interrupts for link status change indications. Unfortunately it seems some controllers(BCM5703, BCM5704 and BCM5705) have PHY related issues so Linux took other approach to workaround it. bge(4) didn't follow that and it used to enable auto polling to workaround it. Restore this old behavior for BCM5700 family controllers and BCM5705 to use auto polling. For BCM5700 and BCM5701, it seems it does not need to enable auto polling but I restored it for safety. Special thanks to marius who tried lots of patches with patience. Reported by: marius Tested by: marius
2010-10-13 22:29:48 +00:00
if (sc->bge_mi_mode & BGE_MIMODE_AUTOPOLL) {
CSR_WRITE_4(sc, BGE_MI_MODE, sc->bge_mi_mode);
DELAY(80);
}
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
}
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
static const struct bge_revision *
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_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);
}
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
static const struct bge_vendor *
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_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
return (NULL);
}
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
static uint32_t
bge_chipid(device_t dev)
{
uint32_t id;
id = pci_read_config(dev, BGE_PCI_MISC_CTL, 4) >>
BGE_PCIMISCCTL_ASICREV_SHIFT;
if (BGE_ASICREV(id) == BGE_ASICREV_USE_PRODID_REG) {
/*
* Find the ASCI revision. Different chips use different
* registers.
*/
switch (pci_get_device(dev)) {
case BCOM_DEVICEID_BCM5717:
case BCOM_DEVICEID_BCM5718:
case BCOM_DEVICEID_BCM5719:
case BCOM_DEVICEID_BCM5720:
id = pci_read_config(dev,
BGE_PCI_GEN2_PRODID_ASICREV, 4);
break;
case BCOM_DEVICEID_BCM57761:
case BCOM_DEVICEID_BCM57762:
case BCOM_DEVICEID_BCM57765:
case BCOM_DEVICEID_BCM57766:
case BCOM_DEVICEID_BCM57781:
case BCOM_DEVICEID_BCM57785:
case BCOM_DEVICEID_BCM57791:
case BCOM_DEVICEID_BCM57795:
id = pci_read_config(dev,
BGE_PCI_GEN15_PRODID_ASICREV, 4);
break;
default:
id = pci_read_config(dev, BGE_PCI_PRODID_ASICREV, 4);
}
}
return (id);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* 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
{
- Remove the remaining support for older (in this case pre-7.0-RELEASE) versions of FreeBSD. In fact we are already missing a lot of conditional code necessary to support older versions of FreeBSD, including alternatives for vital functionality not yet provided by the respective subsystem back then (see for example r199663). So this change shouldn't actually break this driver on versions of FreeBSD that were supported before. Besides, this driver also isn't maintained as an multi-release version outside of the main repository, so removing the conditional code shouldn't be a problem in that regard either. - Sprinkle some more const on tables.
2010-11-30 21:08:59 +00:00
char buf[96];
char model[64];
const struct bge_revision *br;
const char *pname;
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
struct bge_softc *sc;
- Remove the remaining support for older (in this case pre-7.0-RELEASE) versions of FreeBSD. In fact we are already missing a lot of conditional code necessary to support older versions of FreeBSD, including alternatives for vital functionality not yet provided by the respective subsystem back then (see for example r199663). So this change shouldn't actually break this driver on versions of FreeBSD that were supported before. Besides, this driver also isn't maintained as an multi-release version outside of the main repository, so removing the conditional code shouldn't be a problem in that regard either. - Sprinkle some more const on tables.
2010-11-30 21:08:59 +00:00
const struct bge_type *t = bge_devs;
const struct bge_vendor *v;
uint32_t id;
uint16_t did, vid;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
sc = device_get_softc(dev);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
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)) {
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
id = bge_chipid(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
br = bge_lookup_rev(id);
- Remove the remaining support for older (in this case pre-7.0-RELEASE) versions of FreeBSD. In fact we are already missing a lot of conditional code necessary to support older versions of FreeBSD, including alternatives for vital functionality not yet provided by the respective subsystem back then (see for example r199663). So this change shouldn't actually break this driver on versions of FreeBSD that were supported before. Besides, this driver also isn't maintained as an multi-release version outside of the main repository, so removing the conditional code shouldn't be a problem in that regard either. - Sprinkle some more const on tables.
2010-11-30 21:08:59 +00:00
if (bge_has_eaddr(sc) &&
pci_get_vpd_ident(dev, &pname) == 0)
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
snprintf(model, sizeof(model), "%s", pname);
else {
v = bge_lookup_vendor(vid);
snprintf(model, sizeof(model), "%s %s",
v != NULL ? v->v_name : "Unknown",
- Remove the remaining support for older (in this case pre-7.0-RELEASE) versions of FreeBSD. In fact we are already missing a lot of conditional code necessary to support older versions of FreeBSD, including alternatives for vital functionality not yet provided by the respective subsystem back then (see for example r199663). So this change shouldn't actually break this driver on versions of FreeBSD that were supported before. Besides, this driver also isn't maintained as an multi-release version outside of the main repository, so removing the conditional code shouldn't be a problem in that regard either. - Sprinkle some more const on tables.
2010-11-30 21:08:59 +00:00
br != NULL ? br->br_name :
Mention NetLink controllers in the fallback description, too.
2013-01-26 00:41:54 +00:00
"NetXtreme/NetLink Ethernet Controller");
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
}
snprintf(buf, sizeof(buf), "%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);
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
return (BUS_PROBE_DEFAULT);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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);
Destroy DMA tag for jumbo RX buffer in device detach.
2011-12-13 20:31:57 +00:00
if (sc->bge_cdata.bge_mtag_jumbo)
bus_dma_tag_destroy(sc->bge_cdata.bge_mtag_jumbo);
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_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;
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
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
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
After r232403, DMA transactions does not cross 4GB boundary for all PCI devices. Remove driver workaround for 4GB boundary issue.
2012-03-10 06:12:14 +00:00
alignment, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
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
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;
return (0);
}
static int
bge_dma_alloc(struct bge_softc *sc)
{
bus_addr_t lowaddr;
After r232403, DMA transactions does not cross 4GB boundary for all PCI devices. Remove driver workaround for 4GB boundary issue.
2012-03-10 06:12:14 +00:00
bus_size_t rxmaxsegsz, sbsz, txsegsz, txmaxsegsz;
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
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);
Make sure to create DMA'able memory for statistics block. This was missed in r212061 and it caused crashes for 570x controllers as controller DMAed statistics to physical address 0. Reported by: kan
2010-09-07 18:29:29 +00:00
/* Create tag for statistics block. */
error = bge_dma_ring_alloc(sc, PAGE_SIZE, BGE_STATS_SZ,
&sc->bge_cdata.bge_stats_tag,
(uint8_t **)&sc->bge_ldata.bge_stats,
&sc->bge_cdata.bge_stats_map,
&sc->bge_ldata.bge_stats_paddr, "statistics block");
if (error)
return (error);
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 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. */
Apply DMA address space restriction to controllers that have 4GB DMA boundary bug and runs with PCI-X mode. watchdog timeout was observed on BCM5704 which lives behind certain PCI-X bridge(e.g. AMD 8131 PCI-X bridge). It's still not clear whether the root cause came from that PCI-X bridge or not. The watchdog timeout indicates the issue is in TX path. If the bridge reorders TX mailbox write accesses it would generate all kinds of problems but I'm not sure. This should be revisited. Tested by: Michael L. Squires (mikes <> siralan dot org)
2011-01-10 17:45:09 +00:00
if ((sc->bge_flags & BGE_FLAG_4G_BNDRY_BUG) != 0) {
/*
* XXX
* watchdog timeout issue was observed on BCM5704 which
* lives behind PCI-X bridge(e.g AMD 8131 PCI-X bridge).
Add workaround for PCI-X BCM5704 controller that live behind AMD-8131 PCI-X bridge. The bridge seems to reorder write access to mailbox registers such that it caused watchdog timeouts by out-of-order TX completions. Tested by: Michael L. Squires <mikes <> siralan dot org > Reviewed by: jhb
2012-03-12 02:09:47 +00:00
* Both limiting DMA address space to 32bits and flushing
* mailbox write seem to address the issue.
Apply DMA address space restriction to controllers that have 4GB DMA boundary bug and runs with PCI-X mode. watchdog timeout was observed on BCM5704 which lives behind certain PCI-X bridge(e.g. AMD 8131 PCI-X bridge). It's still not clear whether the root cause came from that PCI-X bridge or not. The watchdog timeout indicates the issue is in TX path. If the bridge reorders TX mailbox write accesses it would generate all kinds of problems but I'm not sure. This should be revisited. Tested by: Michael L. Squires (mikes <> siralan dot org)
2011-01-10 17:45:09 +00:00
*/
Add workaround for PCI-X BCM5704 controller that live behind AMD-8131 PCI-X bridge. The bridge seems to reorder write access to mailbox registers such that it caused watchdog timeouts by out-of-order TX completions. Tested by: Michael L. Squires <mikes <> siralan dot org > Reviewed by: jhb
2012-03-12 02:09:47 +00:00
if (sc->bge_pcixcap != 0)
Apply DMA address space restriction to controllers that have 4GB DMA boundary bug and runs with PCI-X mode. watchdog timeout was observed on BCM5704 which lives behind certain PCI-X bridge(e.g. AMD 8131 PCI-X bridge). It's still not clear whether the root cause came from that PCI-X bridge or not. The watchdog timeout indicates the issue is in TX path. If the bridge reorders TX mailbox write accesses it would generate all kinds of problems but I'm not sure. This should be revisited. Tested by: Michael L. Squires (mikes <> siralan dot org)
2011-01-10 17:45:09 +00:00
lowaddr = BUS_SPACE_MAXADDR_32BIT;
}
After r232403, DMA transactions does not cross 4GB boundary for all PCI devices. Remove driver workaround for 4GB boundary issue.
2012-03-10 06:12:14 +00:00
error = bus_dma_tag_create(bus_get_dma_tag(sc->bge_dev), 1, 0, lowaddr,
BUS_SPACE_MAXADDR, NULL, NULL, BUS_SPACE_MAXSIZE_32BIT, 0,
BUS_SPACE_MAXSIZE_32BIT, 0, NULL, NULL,
&sc->bge_cdata.bge_buffer_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 (error != 0) {
device_printf(sc->bge_dev,
"could not allocate buffer dma tag\n");
return (ENOMEM);
}
/* Create tag for Tx mbufs. */
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if (sc->bge_flags & (BGE_FLAG_TSO | BGE_FLAG_TSO3)) {
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
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. */
Add initial jumbo frame support for BCM5714/BCM5715 and BCM5780. Unlike other controllers which have more advanced jumbo support, these controllers have one send ring, one standard receive producer ring and one receive return ring. In order to receive jumbo frames on the controllers, driver now will increase Rx buffer size to 9k. Two Rx modes are supported on these controllers and I chose standard Rx BDs over extended Rx BDs. The extended Rx BD mode allows up to 4 segmentations for each Rx BDs such that kernel does not have to allocate large buffer of contiguous memory for receiving. The extended Rx BD mode is already used on controllers that have separate jumbo receive ring. However, using extended Rx BDs on BCM5714/BCM5715/BCM5780 reduces the number of Rx BDs to 256 entries which in turn may reduce the performance. Also UMA backed page allocator for jumbo frame returns contiguous memory so using extended Rx BD has no advantage on FreeBSD unless highly customized local allocator implemented in driver is used. To use jumbo buffers in standard receive ring, Rx buffer allocation handler was changed to allocate MJUM9BYTES sized mbuf. PR: kern/155192 Tested by: Vijay Singh <vijju.singh <> gmail dot com> Submitted by: mjacob (initial version)
2011-04-05 17:41:54 +00:00
if (sc->bge_flags & BGE_FLAG_JUMBO_STD)
rxmaxsegsz = MJUM9BYTES;
else
rxmaxsegsz = MCLBYTES;
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, 0,
Add initial jumbo frame support for BCM5714/BCM5715 and BCM5780. Unlike other controllers which have more advanced jumbo support, these controllers have one send ring, one standard receive producer ring and one receive return ring. In order to receive jumbo frames on the controllers, driver now will increase Rx buffer size to 9k. Two Rx modes are supported on these controllers and I chose standard Rx BDs over extended Rx BDs. The extended Rx BD mode allows up to 4 segmentations for each Rx BDs such that kernel does not have to allocate large buffer of contiguous memory for receiving. The extended Rx BD mode is already used on controllers that have separate jumbo receive ring. However, using extended Rx BDs on BCM5714/BCM5715/BCM5780 reduces the number of Rx BDs to 256 entries which in turn may reduce the performance. Also UMA backed page allocator for jumbo frame returns contiguous memory so using extended Rx BD has no advantage on FreeBSD unless highly customized local allocator implemented in driver is used. To use jumbo buffers in standard receive ring, Rx buffer allocation handler was changed to allocate MJUM9BYTES sized mbuf. PR: kern/155192 Tested by: Vijay Singh <vijju.singh <> gmail dot com> Submitted by: mjacob (initial version)
2011-04-05 17:41:54 +00:00
BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL, rxmaxsegsz, 1,
rxmaxsegsz, 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;
Oops, fix logic error introduced in r230337.
2012-01-19 20:28:58 +00:00
if (sc->bge_msi == 0)
Introduce a tunable that disables use of MSI. Non-zero value will use INTx.
2012-01-17 22:15:33 +00:00
return (0);
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
/* Disable MSI for polling(4). */
#ifdef DEVICE_POLLING
return (0);
#endif
Re-enable MSI support for those chips on which it is believed to work properly.
2006-12-22 02:59:58 +00:00
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 workaround for PCI-X BCM5704 controller that live behind AMD-8131 PCI-X bridge. The bridge seems to reorder write access to mailbox registers such that it caused watchdog timeouts by out-of-order TX completions. Tested by: Michael L. Squires <mikes <> siralan dot org > Reviewed by: jhb
2012-03-12 02:09:47 +00:00
static int
bge_mbox_reorder(struct bge_softc *sc)
{
/* Lists of PCI bridges that are known to reorder mailbox writes. */
static const struct mbox_reorder {
const uint16_t vendor;
const uint16_t device;
const char *desc;
Remove duplicate const specifiers in many drivers (I hope I got all of them, please let me know if not). Most of these are of the form: static const struct bzzt_type { [...list of members...] } const bzzt_devs[] = { [...list of initializers...] }; The second const is unnecessary, as arrays cannot be modified anyway, and if the elements are const, the whole thing is const automatically (e.g. it is placed in .rodata). I have verified this does not change the binary output of a full kernel build (except for build timestamps embedded in the object files). Reviewed by: yongari, marius MFC after: 1 week
2012-11-05 19:16:27 +00:00
} mbox_reorder_lists[] = {
Add workaround for PCI-X BCM5704 controller that live behind AMD-8131 PCI-X bridge. The bridge seems to reorder write access to mailbox registers such that it caused watchdog timeouts by out-of-order TX completions. Tested by: Michael L. Squires <mikes <> siralan dot org > Reviewed by: jhb
2012-03-12 02:09:47 +00:00
{ 0x1022, 0x7450, "AMD-8131 PCI-X Bridge" },
};
devclass_t pci, pcib;
device_t bus, dev;
Take advantage of nitems(). MFC after: 3 days
2012-06-02 19:41:28 +00:00
int i;
Add workaround for PCI-X BCM5704 controller that live behind AMD-8131 PCI-X bridge. The bridge seems to reorder write access to mailbox registers such that it caused watchdog timeouts by out-of-order TX completions. Tested by: Michael L. Squires <mikes <> siralan dot org > Reviewed by: jhb
2012-03-12 02:09:47 +00:00
pci = devclass_find("pci");
pcib = devclass_find("pcib");
dev = sc->bge_dev;
bus = device_get_parent(dev);
for (;;) {
dev = device_get_parent(bus);
bus = device_get_parent(dev);
if (device_get_devclass(dev) != pcib)
break;
Take advantage of nitems(). MFC after: 3 days
2012-06-02 19:41:28 +00:00
for (i = 0; i < nitems(mbox_reorder_lists); i++) {
Add workaround for PCI-X BCM5704 controller that live behind AMD-8131 PCI-X bridge. The bridge seems to reorder write access to mailbox registers such that it caused watchdog timeouts by out-of-order TX completions. Tested by: Michael L. Squires <mikes <> siralan dot org > Reviewed by: jhb
2012-03-12 02:09:47 +00:00
if (pci_get_vendor(dev) ==
mbox_reorder_lists[i].vendor &&
pci_get_device(dev) ==
mbox_reorder_lists[i].device) {
device_printf(sc->bge_dev,
"enabling MBOX workaround for %s\n",
mbox_reorder_lists[i].desc);
return (1);
}
}
if (device_get_devclass(bus) != pci)
break;
}
return (0);
}
Show PCI bus speed and width as well as running mode of PCI-X device in device attach. This would help to narrow down issue to a specific controller and operating mode of the controller. While I'm here rename BGE_MISCCFG_BOARD_ID with BGE_MISCCFG_BOARD_ID_MASK.
2012-03-12 02:42:47 +00:00
static void
bge_devinfo(struct bge_softc *sc)
{
uint32_t cfg, clk;
device_printf(sc->bge_dev,
"CHIP ID 0x%08x; ASIC REV 0x%02x; CHIP REV 0x%02x; ",
sc->bge_chipid, sc->bge_asicrev, sc->bge_chiprev);
if (sc->bge_flags & BGE_FLAG_PCIE)
printf("PCI-E\n");
else if (sc->bge_flags & BGE_FLAG_PCIX) {
printf("PCI-X ");
cfg = CSR_READ_4(sc, BGE_MISC_CFG) & BGE_MISCCFG_BOARD_ID_MASK;
if (cfg == BGE_MISCCFG_BOARD_ID_5704CIOBE)
clk = 133;
else {
clk = CSR_READ_4(sc, BGE_PCI_CLKCTL) & 0x1F;
switch (clk) {
case 0:
clk = 33;
break;
case 2:
clk = 50;
break;
case 4:
clk = 66;
break;
case 6:
clk = 100;
break;
case 7:
clk = 133;
break;
}
}
printf("%u MHz\n", clk);
} else {
if (sc->bge_pcixcap != 0)
printf("PCI on PCI-X ");
else
printf("PCI ");
cfg = pci_read_config(sc->bge_dev, BGE_PCI_PCISTATE, 4);
if (cfg & BGE_PCISTATE_PCI_BUSSPEED)
clk = 66;
else
clk = 33;
if (cfg & BGE_PCISTATE_32BIT_BUS)
printf("%u MHz; 32bit\n", clk);
else
printf("%u MHz; 64bit\n", clk);
}
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
uint32_t hwcfg = 0, misccfg, pcistate;
- 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];
Do not hardcode phy address. Multi-port controllers use different phy address.
2012-10-24 05:00:56 +00:00
int capmask, 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;
Rework device detach. While here, move driver lock/callout initialization to the beginning of device attach for readability.
2012-10-08 07:33:43 +00:00
BGE_LOCK_INIT(sc, device_get_nameunit(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);
Rework device detach. While here, move driver lock/callout initialization to the beginning of device attach for readability.
2012-10-08 07:33:43 +00:00
callout_init_mtx(&sc->bge_stat_ch, &sc->bge_mtx, 0);
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
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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) {
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
device_printf (sc->bge_dev, "couldn't map BAR0 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 APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
sc->bge_func_addr = pci_get_function(dev);
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
sc->bge_chipid = bge_chipid(dev);
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev = BGE_ASICREV(sc->bge_chipid);
sc->bge_chiprev = BGE_CHIPREV(sc->bge_chipid);
Overhaul MII register access routine and remove unnecessary BGE_MI_MODE register accesses. Previously bge(4) used to read BGE_MI_MODE register to detect whether it needs to disable autopolling feature or not. Because we don't touch autopolling in other part of driver there is no reason to read BGE_MI_MODE register given that we know default value in advance. In order to achieve the goal, check whether the controller has CPMU(Central Power Mangement Unit) capability. If controller has CPMU feature, use 500KHz MII management interface(mdio/mdc) frequency regardless core clock frequency. Otherwise use default MII clock. While I'm here, add CPMU register definition. In bge_miibus_readreg(), rearrange code a bit and remove goto statement. In bge_miibus_writereg(), make sure to restore autopolling even if MII write failed. The delay time inserted after accessing BGE_MI_MODE register increased from 40us to 80us. The default PHY address is now stored in softc. All PHYs supported by bge(4) currently uses PHY address 1 but it will be changed when we add newer controllers. This change will make it easier to change default PHY address depending on PHY models. Submitted by: davidch
2010-10-06 17:35:27 +00:00
/* Set default PHY address. */
Do not hardcode phy address. Multi-port controllers use different phy address.
2012-10-24 05:00:56 +00:00
sc->bge_phy_addr = 1;
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
/*
* PHY address mapping for various devices.
*
* | F0 Cu | F0 Sr | F1 Cu | F1 Sr |
* ---------+-------+-------+-------+-------+
* BCM57XX | 1 | X | X | X |
* BCM5704 | 1 | X | 1 | X |
* BCM5717 | 1 | 8 | 2 | 9 |
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
* BCM5719 | 1 | 8 | 2 | 9 |
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
* BCM5720 | 1 | 8 | 2 | 9 |
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
*
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
* | F2 Cu | F2 Sr | F3 Cu | F3 Sr |
* ---------+-------+-------+-------+-------+
* BCM57XX | X | X | X | X |
* BCM5704 | X | X | X | X |
* BCM5717 | X | X | X | X |
* BCM5719 | 3 | 10 | 4 | 11 |
* BCM5720 | X | X | X | X |
*
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
* Other addresses may respond but they are not
* IEEE compliant PHYs and should be ignored.
*/
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5717 ||
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
sc->bge_asicrev == BGE_ASICREV_BCM5719 ||
sc->bge_asicrev == BGE_ASICREV_BCM5720) {
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
if (sc->bge_chipid != BGE_CHIPID_BCM5717_A0) {
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if (CSR_READ_4(sc, BGE_SGDIG_STS) &
BGE_SGDIGSTS_IS_SERDES)
Do not hardcode phy address. Multi-port controllers use different phy address.
2012-10-24 05:00:56 +00:00
sc->bge_phy_addr = sc->bge_func_addr + 8;
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
else
Do not hardcode phy address. Multi-port controllers use different phy address.
2012-10-24 05:00:56 +00:00
sc->bge_phy_addr = sc->bge_func_addr + 1;
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
} else {
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if (CSR_READ_4(sc, BGE_CPMU_PHY_STRAP) &
BGE_CPMU_PHY_STRAP_IS_SERDES)
Do not hardcode phy address. Multi-port controllers use different phy address.
2012-10-24 05:00:56 +00:00
sc->bge_phy_addr = sc->bge_func_addr + 8;
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
else
Do not hardcode phy address. Multi-port controllers use different phy address.
2012-10-24 05:00:56 +00:00
sc->bge_phy_addr = sc->bge_func_addr + 1;
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
}
}
Overhaul MII register access routine and remove unnecessary BGE_MI_MODE register accesses. Previously bge(4) used to read BGE_MI_MODE register to detect whether it needs to disable autopolling feature or not. Because we don't touch autopolling in other part of driver there is no reason to read BGE_MI_MODE register given that we know default value in advance. In order to achieve the goal, check whether the controller has CPMU(Central Power Mangement Unit) capability. If controller has CPMU feature, use 500KHz MII management interface(mdio/mdc) frequency regardless core clock frequency. Otherwise use default MII clock. While I'm here, add CPMU register definition. In bge_miibus_readreg(), rearrange code a bit and remove goto statement. In bge_miibus_writereg(), make sure to restore autopolling even if MII write failed. The delay time inserted after accessing BGE_MI_MODE register increased from 40us to 80us. The default PHY address is now stored in softc. All PHYs supported by bge(4) currently uses PHY address 1 but it will be changed when we add newer controllers. This change will make it easier to change default PHY address depending on PHY models. Submitted by: davidch
2010-10-06 17:35:27 +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 (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 preliminary support for BCM57766 ASIC. While I'm here add BCM57762 device id which is found on Apple Thunderbolt ethernet. Tested by: Richard Kuhns < rjk <> wintek dot com >
2012-11-30 01:38:00 +00:00
case BGE_ASICREV_BCM57765:
case BGE_ASICREV_BCM57766:
sc->bge_flags |= BGE_FLAG_57765_PLUS;
/* FALLTHROUGH */
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
case BGE_ASICREV_BCM5717:
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
case BGE_ASICREV_BCM5719:
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
case BGE_ASICREV_BCM5720:
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
sc->bge_flags |= BGE_FLAG_5717_PLUS | BGE_FLAG_5755_PLUS |
BGE_FLAG_575X_PLUS | BGE_FLAG_5705_PLUS | BGE_FLAG_JUMBO |
Add initial BCM57765 family support. The BCM57765 family seems to have similar hardware features of BCM5718 family except the number of receive return ring is 4. The BCM57765 family is known to support IEEE 802.3az EEE(Energy Efficient Ethernet) but this change does not include EEE support code. I hope EEE is implemented in near future. This change will support BCM57761, BCM57765, BCM57781, BCM57785, BCM57791 and BCM57795. All hardware offloading features are supported and suspend/resume also should work. Many thanks to Broadcom for continuing support of FreeBSD. Tested by: Paul Thornton (prt <> prt dot org) HW donated by: Broadcom
2011-05-04 17:04:31 +00:00
BGE_FLAG_JUMBO_FRAME;
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5719 &&
sc->bge_chipid == BGE_CHIPID_BCM5719_A0) {
/* Jumbo frame on BCM5719 A0 does not work. */
Correctly disable jumbo frame support for BCM5719 A0.
2011-05-15 21:44:51 +00:00
sc->bge_flags &= ~BGE_FLAG_JUMBO;
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
}
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
break;
- 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 initial jumbo frame support for BCM5714/BCM5715 and BCM5780. Unlike other controllers which have more advanced jumbo support, these controllers have one send ring, one standard receive producer ring and one receive return ring. In order to receive jumbo frames on the controllers, driver now will increase Rx buffer size to 9k. Two Rx modes are supported on these controllers and I chose standard Rx BDs over extended Rx BDs. The extended Rx BD mode allows up to 4 segmentations for each Rx BDs such that kernel does not have to allocate large buffer of contiguous memory for receiving. The extended Rx BD mode is already used on controllers that have separate jumbo receive ring. However, using extended Rx BDs on BCM5714/BCM5715/BCM5780 reduces the number of Rx BDs to 256 entries which in turn may reduce the performance. Also UMA backed page allocator for jumbo frame returns contiguous memory so using extended Rx BD has no advantage on FreeBSD unless highly customized local allocator implemented in driver is used. To use jumbo buffers in standard receive ring, Rx buffer allocation handler was changed to allocate MJUM9BYTES sized mbuf. PR: kern/155192 Tested by: Vijay Singh <vijju.singh <> gmail dot com> Submitted by: mjacob (initial version)
2011-04-05 17:41:54 +00:00
sc->bge_flags |= BGE_FLAG_5714_FAMILY | BGE_FLAG_JUMBO_STD;
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;
}
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
/* Identify chips with APE processor. */
switch (sc->bge_asicrev) {
case BGE_ASICREV_BCM5717:
case BGE_ASICREV_BCM5719:
case BGE_ASICREV_BCM5720:
case BGE_ASICREV_BCM5761:
sc->bge_flags |= BGE_FLAG_APE;
break;
}
/* Chips with APE need BAR2 access for APE registers/memory. */
if ((sc->bge_flags & BGE_FLAG_APE) != 0) {
rid = PCIR_BAR(2);
sc->bge_res2 = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
RF_ACTIVE);
if (sc->bge_res2 == NULL) {
device_printf (sc->bge_dev,
"couldn't map BAR2 memory\n");
error = ENXIO;
goto fail;
}
/* Enable APE register/memory access by host driver. */
pcistate = pci_read_config(dev, BGE_PCI_PCISTATE, 4);
pcistate |= BGE_PCISTATE_ALLOW_APE_CTLSPC_WR |
BGE_PCISTATE_ALLOW_APE_SHMEM_WR |
BGE_PCISTATE_ALLOW_APE_PSPACE_WR;
pci_write_config(dev, BGE_PCI_PCISTATE, pcistate, 4);
bge_ape_lock_init(sc);
bge_ape_read_fw_ver(sc);
}
Try to finally get the point in time at which bge_add_sysctls() is called right; it needs to be called before bge_can_use_msi() but in turn requires bge_flags to be properly set. Submitted by: yongari MFC after: 3 days
2012-05-30 20:04:07 +00:00
/* Add SYSCTLs, requires the chipset family to be set. */
bge_add_sysctls(sc);
Overhaul MII register access routine and remove unnecessary BGE_MI_MODE register accesses. Previously bge(4) used to read BGE_MI_MODE register to detect whether it needs to disable autopolling feature or not. Because we don't touch autopolling in other part of driver there is no reason to read BGE_MI_MODE register given that we know default value in advance. In order to achieve the goal, check whether the controller has CPMU(Central Power Mangement Unit) capability. If controller has CPMU feature, use 500KHz MII management interface(mdio/mdc) frequency regardless core clock frequency. Otherwise use default MII clock. While I'm here, add CPMU register definition. In bge_miibus_readreg(), rearrange code a bit and remove goto statement. In bge_miibus_writereg(), make sure to restore autopolling even if MII write failed. The delay time inserted after accessing BGE_MI_MODE register increased from 40us to 80us. The default PHY address is now stored in softc. All PHYs supported by bge(4) currently uses PHY address 1 but it will be changed when we add newer controllers. This change will make it easier to change default PHY address depending on PHY models. Submitted by: davidch
2010-10-06 17:35:27 +00:00
/* Identify the chips that use an CPMU. */
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if (BGE_IS_5717_PLUS(sc) ||
sc->bge_asicrev == BGE_ASICREV_BCM5784 ||
Overhaul MII register access routine and remove unnecessary BGE_MI_MODE register accesses. Previously bge(4) used to read BGE_MI_MODE register to detect whether it needs to disable autopolling feature or not. Because we don't touch autopolling in other part of driver there is no reason to read BGE_MI_MODE register given that we know default value in advance. In order to achieve the goal, check whether the controller has CPMU(Central Power Mangement Unit) capability. If controller has CPMU feature, use 500KHz MII management interface(mdio/mdc) frequency regardless core clock frequency. Otherwise use default MII clock. While I'm here, add CPMU register definition. In bge_miibus_readreg(), rearrange code a bit and remove goto statement. In bge_miibus_writereg(), make sure to restore autopolling even if MII write failed. The delay time inserted after accessing BGE_MI_MODE register increased from 40us to 80us. The default PHY address is now stored in softc. All PHYs supported by bge(4) currently uses PHY address 1 but it will be changed when we add newer controllers. This change will make it easier to change default PHY address depending on PHY models. Submitted by: davidch
2010-10-06 17:35:27 +00:00
sc->bge_asicrev == BGE_ASICREV_BCM5761 ||
sc->bge_asicrev == BGE_ASICREV_BCM5785 ||
sc->bge_asicrev == BGE_ASICREV_BCM57780)
sc->bge_flags |= BGE_FLAG_CPMU_PRESENT;
if ((sc->bge_flags & BGE_FLAG_CPMU_PRESENT) != 0)
sc->bge_mi_mode = BGE_MIMODE_500KHZ_CONST;
else
sc->bge_mi_mode = BGE_MIMODE_BASE;
Fix a regression introduced in r213710. r213710 removed the use of auto polling such that it made all controllers obtain link status information from the state of the LNKRDY input signal. Broadcom recommends disabling auto polling such that driver should rely on PHY interrupts for link status change indications. Unfortunately it seems some controllers(BCM5703, BCM5704 and BCM5705) have PHY related issues so Linux took other approach to workaround it. bge(4) didn't follow that and it used to enable auto polling to workaround it. Restore this old behavior for BCM5700 family controllers and BCM5705 to use auto polling. For BCM5700 and BCM5701, it seems it does not need to enable auto polling but I restored it for safety. Special thanks to marius who tried lots of patches with patience. Reported by: marius Tested by: marius
2010-10-13 22:29:48 +00:00
/* Enable auto polling for BCM570[0-5]. */
if (BGE_IS_5700_FAMILY(sc) || sc->bge_asicrev == BGE_ASICREV_BCM5705)
sc->bge_mi_mode |= BGE_MIMODE_AUTOPOLL;
Overhaul MII register access routine and remove unnecessary BGE_MI_MODE register accesses. Previously bge(4) used to read BGE_MI_MODE register to detect whether it needs to disable autopolling feature or not. Because we don't touch autopolling in other part of driver there is no reason to read BGE_MI_MODE register given that we know default value in advance. In order to achieve the goal, check whether the controller has CPMU(Central Power Mangement Unit) capability. If controller has CPMU feature, use 500KHz MII management interface(mdio/mdc) frequency regardless core clock frequency. Otherwise use default MII clock. While I'm here, add CPMU register definition. In bge_miibus_readreg(), rearrange code a bit and remove goto statement. In bge_miibus_writereg(), make sure to restore autopolling even if MII write failed. The delay time inserted after accessing BGE_MI_MODE register increased from 40us to 80us. The default PHY address is now stored in softc. All PHYs supported by bge(4) currently uses PHY address 1 but it will be changed when we add newer controllers. This change will make it easier to change default PHY address depending on PHY models. Submitted by: davidch
2010-10-06 17:35:27 +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
/*
It is known that all Broadcom controllers have 4GB boundary DMA bug. Apply workaround to all controllers.
2011-10-26 18:19:50 +00:00
* All Broadcom controllers have 4GB boundary DMA 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
* 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.
*/
It is known that all Broadcom controllers have 4GB boundary DMA bug. Apply workaround to all controllers.
2011-10-26 18:19:50 +00:00
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
Broadcom says BCM5755 or higher and BCM5906 have short DMA bug. Apply workaround to these controllers.
2011-10-26 18:27:01 +00:00
/* BCM5755 or higher and BCM5906 have short DMA bug. */
if (BGE_IS_5755_PLUS(sc) || sc->bge_asicrev == BGE_ASICREV_BCM5906)
sc->bge_flags |= BGE_FLAG_SHORT_DMA_BUG;
BCM5719 cannot handle DMA requests for DMA segments that have larger than 4KB in size. However the maximum DMA segment size created in DMA tag is 4KB, so we wouldn't encounter the issue here. Just record this issue such that let developers not to create a DMA segment that is larger than 4KB for BCM5719. It's possible to split a DMA segment into multiple smaller ones in run time but I believe it's not worth to implement that.
2011-10-26 18:37:02 +00:00
/*
* BCM5719 cannot handle DMA requests for DMA segments that
* have larger than 4KB in size. However the maximum DMA
* segment size created in DMA tag is 4KB for TSO, so we
* wouldn't encounter the issue here.
*/
if (sc->bge_asicrev == BGE_ASICREV_BCM5719)
sc->bge_flags |= BGE_FLAG_4K_RDMA_BUG;
Show PCI bus speed and width as well as running mode of PCI-X device in device attach. This would help to narrow down issue to a specific controller and operating mode of the controller. While I'm here rename BGE_MISCCFG_BOARD_ID with BGE_MISCCFG_BOARD_ID_MASK.
2012-03-12 02:42:47 +00:00
misccfg = CSR_READ_4(sc, BGE_MISC_CFG) & BGE_MISCCFG_BOARD_ID_MASK;
Fix a long standing bug which regarded some revisions of controller as 5788. This caused BGE_MISC_LOCAL_CTL register is used to generate link state change interrupt for non-5788 controllers. The interrupt handler may or may not detect link state attention as status block wouldn't be updated when an interrupt was generated with BGE_MISC_LOCAL_CTL register. All controllers except 5700 and 5788 should use host coalescing mode register to trigger an interrupt.
2010-10-07 17:14:50 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5705) {
if (misccfg == BGE_MISCCFG_BOARD_ID_5788 ||
misccfg == BGE_MISCCFG_BOARD_ID_5788M)
sc->bge_flags |= BGE_FLAG_5788;
}
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
Move the limiting of the PHY to 10/100 modes of operation due to limitations of certain MAC models from brgphy(4) to bge(4) where it belongs. While at it, update the list of models having that restriction to what OpenBSD uses, which in turn seems to have obtained that information from the Linux tg3 driver.
2010-11-14 15:15:22 +00:00
capmask = BMSR_DEFCAPMASK;
if ((sc->bge_asicrev == BGE_ASICREV_BCM5703 &&
(misccfg == 0x4000 || misccfg == 0x8000)) ||
(sc->bge_asicrev == BGE_ASICREV_BCM5705 &&
pci_get_vendor(dev) == BCOM_VENDORID &&
(pci_get_device(dev) == BCOM_DEVICEID_BCM5901 ||
pci_get_device(dev) == BCOM_DEVICEID_BCM5901A2 ||
pci_get_device(dev) == BCOM_DEVICEID_BCM5705F)) ||
(pci_get_vendor(dev) == BCOM_VENDORID &&
(pci_get_device(dev) == BCOM_DEVICEID_BCM5751F ||
pci_get_device(dev) == BCOM_DEVICEID_BCM5753F ||
pci_get_device(dev) == BCOM_DEVICEID_BCM5787F)) ||
pci_get_device(dev) == BCOM_DEVICEID_BCM57790 ||
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
pci_get_device(dev) == BCOM_DEVICEID_BCM57791 ||
pci_get_device(dev) == BCOM_DEVICEID_BCM57795 ||
Move the limiting of the PHY to 10/100 modes of operation due to limitations of certain MAC models from brgphy(4) to bge(4) where it belongs. While at it, update the list of models having that restriction to what OpenBSD uses, which in turn seems to have obtained that information from the Linux tg3 driver.
2010-11-14 15:15:22 +00:00
sc->bge_asicrev == BGE_ASICREV_BCM5906) {
/* These chips are 10/100 only. */
capmask &= ~BMSR_EXTSTAT;
For fast ethernet controllers, Ethernet@WireSpeed is not defined so explicitly set BGE_PHY_NO_WIRESPEED flag.
2012-10-24 05:54:17 +00:00
sc->bge_phy_flags |= BGE_PHY_NO_WIRESPEED;
Move the limiting of the PHY to 10/100 modes of operation due to limitations of certain MAC models from brgphy(4) to bge(4) where it belongs. While at it, update the list of models having that restriction to what OpenBSD uses, which in turn seems to have obtained that information from the Linux tg3 driver.
2010-11-14 15:15:22 +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
/*
* 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
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
* known bug which can't handle TSO if Ethernet header + IP/TCP
* header is greater than 80 bytes. A workaround for the 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
* 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).
*/
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if (BGE_IS_5717_PLUS(sc)) {
/* BCM5717 requires different TSO configuration. */
sc->bge_flags |= BGE_FLAG_TSO3;
Add initial BCM5719 support. TSO and jumbo frame was intentionally disabled for BCM5719 A0 revision due to known hardware errata. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom
2011-05-12 17:15:57 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5719 &&
sc->bge_chipid == BGE_CHIPID_BCM5719_A0) {
/* TSO on BCM5719 A0 does not work. */
sc->bge_flags &= ~BGE_FLAG_TSO3;
}
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
} else if (BGE_IS_5755_PLUS(sc)) {
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
/*
* 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
Whitespace nits.
2011-10-25 20:45:14 +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.
Whitespace nits.
2011-10-25 20:45:14 +00:00
*/
Do a sweep of the tree replacing calls to pci_find_extcap() with calls to pci_find_cap() instead.
2011-03-23 13:10:15 +00:00
if (pci_find_cap(dev, PCIY_EXPRESS, &reg) == 0) {
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
/*
* 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;
If the maximum payload size is 256 bytes or more, set the DMA write water mark to 256 bytes. Otherwise controller will encounter DMA write under run errors and would result in RX DMA hang. If the maximum payload size is 128 bytes, the water mark is set to 128 bytes as usual. While here, set maximum read request size to 2048 for BCM5719/BCM5720. For other PCIe devices, use 4096. And reprogram the maximum read request size whenever device reset is performed.
2012-10-10 01:24:02 +00:00
/* Extract supported maximum payload size. */
sc->bge_mps = pci_read_config(dev, sc->bge_expcap +
PCIER_DEVICE_CAP, 2);
sc->bge_mps = 128 << (sc->bge_mps & PCIEM_CAP_MAX_PAYLOAD);
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5719 ||
sc->bge_asicrev == BGE_ASICREV_BCM5720)
If the maximum payload size is 256 bytes or more, set the DMA write water mark to 256 bytes. Otherwise controller will encounter DMA write under run errors and would result in RX DMA hang. If the maximum payload size is 128 bytes, the water mark is set to 128 bytes as usual. While here, set maximum read request size to 2048 for BCM5719/BCM5720. For other PCIe devices, use 4096. And reprogram the maximum read request size whenever device reset is performed.
2012-10-10 01:24:02 +00:00
sc->bge_expmrq = 2048;
else
sc->bge_expmrq = 4096;
pci_set_max_read_req(dev, sc->bge_expmrq);
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.)
*/
Do a sweep of the tree replacing calls to pci_find_extcap() with calls to pci_find_cap() instead.
2011-03-23 13:10:15 +00:00
if (pci_find_cap(dev, PCIY_PCIX, &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_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;
Add workaround for PCI-X BCM5704 controller that live behind AMD-8131 PCI-X bridge. The bridge seems to reorder write access to mailbox registers such that it caused watchdog timeouts by out-of-order TX completions. Tested by: Michael L. Squires <mikes <> siralan dot org > Reviewed by: jhb
2012-03-12 02:09:47 +00:00
/*
* Some PCI-X bridges are known to trigger write reordering to
* the mailbox registers. Typical phenomena is watchdog timeouts
* caused by out-of-order TX completions. Enable workaround for
* PCI-X devices that live behind these bridges.
* Note, PCI-X controllers can run in PCI mode so we can't use
* BGE_FLAG_PCIX flag to detect PCI-X controllers.
*/
if (sc->bge_pcixcap != 0 && bge_mbox_reorder(sc) != 0)
sc->bge_flags |= BGE_FLAG_MBOX_REORDER;
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;
Do a sweep of the tree replacing calls to pci_find_extcap() with calls to pci_find_cap() instead.
2011-03-23 13:10:15 +00:00
if (pci_find_cap(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
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
/*
* All controllers except BCM5700 supports tagged status but
* we use tagged status only for MSI case on BCM5717. Otherwise
* MSI on BCM5717 does not work.
*/
#ifndef DEVICE_POLLING
if (sc->bge_flags & BGE_FLAG_MSI && BGE_IS_5717_PLUS(sc))
sc->bge_flags |= BGE_FLAG_TAGGED_STATUS;
#endif
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;
}
Show PCI bus speed and width as well as running mode of PCI-X device in device attach. This would help to narrow down issue to a specific controller and operating mode of the controller. While I'm here rename BGE_MISCCFG_BOARD_ID with BGE_MISCCFG_BOARD_ID_MASK.
2012-03-12 02:42:47 +00:00
bge_devinfo(sc);
- 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
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 = 0;
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
/* No ASF if APE present. */
if ((sc->bge_flags & BGE_FLAG_APE) == 0) {
if (bge_allow_asf && (bge_readmem_ind(sc, BGE_SRAM_DATA_SIG) ==
BGE_SRAM_DATA_SIG_MAGIC)) {
if (bge_readmem_ind(sc, BGE_SRAM_DATA_CFG) &
BGE_HWCFG_ASF) {
sc->bge_asf_mode |= ASF_ENABLE;
sc->bge_asf_mode |= ASF_STACKUP;
if (BGE_IS_575X_PLUS(sc))
sc->bge_asf_mode |= ASF_NEW_HANDSHAKE;
}
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);
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
bge_sig_pre_reset(sc, BGE_RESET_START);
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;
}
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +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
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 initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if (BGE_IS_5717_PLUS(sc))
sc->bge_return_ring_cnt = BGE_RETURN_RING_CNT;
else 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;
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if ((sc->bge_flags & (BGE_FLAG_TSO | BGE_FLAG_TSO3)) != 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
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
*/
Rename definition of BGE_SOFTWARE_GENCOMM_* to more readable ones. The origin of GENCOMM seems to come from Alteon Tigon Host/NIC interface definition where it defines general communications region which is active when firmware is loaded and running. This region was used in communication between the host and processor internal to the Tigon chip. Broadcom data sheet also defines the region as 'Software Gencomm' in NetXtreme memory map but lacks detailed description of its interface so it was hard to know which ones are used for which interface. This change shall slightly enhance readability. No functional changes.
2011-10-26 21:05:45 +00:00
if (bge_readmem_ind(sc, BGE_SRAM_DATA_SIG) == BGE_SRAM_DATA_SIG_MAGIC)
hwcfg = bge_readmem_ind(sc, BGE_SRAM_DATA_CFG);
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) {
Ethernet@WireSpeed is defined for 1000baseT adapter to establish a link at a lower speed so enabling it for fiber adapters is wrong. Fix the issue by setting BGE_PHY_NO_WIRESPEED such that brgphy(4) wouldn't enable the feature. While I'm here move PHY specific feature/bug configuration to new location(just before mii attach) for readability.
2012-10-24 05:22:41 +00:00
if (BGE_IS_5705_PLUS(sc)) {
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
sc->bge_flags |= BGE_FLAG_MII_SERDES;
Ethernet@WireSpeed is defined for 1000baseT adapter to establish a link at a lower speed so enabling it for fiber adapters is wrong. Fix the issue by setting BGE_PHY_NO_WIRESPEED such that brgphy(4) wouldn't enable the feature. While I'm here move PHY specific feature/bug configuration to new location(just before mii attach) for readability.
2012-10-24 05:22:41 +00:00
sc->bge_phy_flags |= BGE_PHY_NO_WIRESPEED;
} else
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
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
Ethernet@WireSpeed is defined for 1000baseT adapter to establish a link at a lower speed so enabling it for fiber adapters is wrong. Fix the issue by setting BGE_PHY_NO_WIRESPEED such that brgphy(4) wouldn't enable the feature. While I'm here move PHY specific feature/bug configuration to new location(just before mii attach) for readability.
2012-10-24 05:22:41 +00:00
/* Set various PHY bug flags. */
if (sc->bge_chipid == BGE_CHIPID_BCM5701_A0 ||
sc->bge_chipid == BGE_CHIPID_BCM5701_B0)
sc->bge_phy_flags |= BGE_PHY_CRC_BUG;
if (sc->bge_chiprev == BGE_CHIPREV_5703_AX ||
sc->bge_chiprev == BGE_CHIPREV_5704_AX)
sc->bge_phy_flags |= BGE_PHY_ADC_BUG;
if (sc->bge_chipid == BGE_CHIPID_BCM5704_A0)
sc->bge_phy_flags |= BGE_PHY_5704_A0_BUG;
if (pci_get_subvendor(dev) == DELL_VENDORID)
sc->bge_phy_flags |= BGE_PHY_NO_3LED;
if ((BGE_IS_5705_PLUS(sc)) &&
sc->bge_asicrev != BGE_ASICREV_BCM5906 &&
sc->bge_asicrev != BGE_ASICREV_BCM5785 &&
Add preliminary support for BCM57766 ASIC. While I'm here add BCM57762 device id which is found on Apple Thunderbolt ethernet. Tested by: Richard Kuhns < rjk <> wintek dot com >
2012-11-30 01:38:00 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM57780 &&
!BGE_IS_5717_PLUS(sc)) {
Ethernet@WireSpeed is defined for 1000baseT adapter to establish a link at a lower speed so enabling it for fiber adapters is wrong. Fix the issue by setting BGE_PHY_NO_WIRESPEED such that brgphy(4) wouldn't enable the feature. While I'm here move PHY specific feature/bug configuration to new location(just before mii attach) for readability.
2012-10-24 05:22:41 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5755 ||
sc->bge_asicrev == BGE_ASICREV_BCM5761 ||
sc->bge_asicrev == BGE_ASICREV_BCM5784 ||
sc->bge_asicrev == BGE_ASICREV_BCM5787) {
if (pci_get_device(dev) != BCOM_DEVICEID_BCM5722 &&
pci_get_device(dev) != BCOM_DEVICEID_BCM5756)
sc->bge_phy_flags |= BGE_PHY_JITTER_BUG;
if (pci_get_device(dev) == BCOM_DEVICEID_BCM5755M)
sc->bge_phy_flags |= BGE_PHY_ADJUST_TRIM;
} else
sc->bge_phy_flags |= BGE_PHY_BER_BUG;
}
/*
For fast ethernet controllers, Ethernet@WireSpeed is not defined so explicitly set BGE_PHY_NO_WIRESPEED flag.
2012-10-24 05:54:17 +00:00
* Don't enable Ethernet@WireSpeed for the 5700 or the
Ethernet@WireSpeed is defined for 1000baseT adapter to establish a link at a lower speed so enabling it for fiber adapters is wrong. Fix the issue by setting BGE_PHY_NO_WIRESPEED such that brgphy(4) wouldn't enable the feature. While I'm here move PHY specific feature/bug configuration to new location(just before mii attach) for readability.
2012-10-24 05:22:41 +00:00
* 5705 A0 and A1 chips.
*/
if (sc->bge_asicrev == BGE_ASICREV_BCM5700 ||
(sc->bge_asicrev == BGE_ASICREV_BCM5705 &&
(sc->bge_chipid != BGE_CHIPID_BCM5705_A0 &&
For fast ethernet controllers, Ethernet@WireSpeed is not defined so explicitly set BGE_PHY_NO_WIRESPEED flag.
2012-10-24 05:54:17 +00:00
sc->bge_chipid != BGE_CHIPID_BCM5705_A1)))
Ethernet@WireSpeed is defined for 1000baseT adapter to establish a link at a lower speed so enabling it for fiber adapters is wrong. Fix the issue by setting BGE_PHY_NO_WIRESPEED such that brgphy(4) wouldn't enable the feature. While I'm here move PHY specific feature/bug configuration to new location(just before mii attach) for readability.
2012-10-24 05:22:41 +00:00
sc->bge_phy_flags |= BGE_PHY_NO_WIRESPEED;
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);
Move the limiting of the PHY to 10/100 modes of operation due to limitations of certain MAC models from brgphy(4) to bge(4) where it belongs. While at it, update the list of models having that restriction to what OpenBSD uses, which in turn seems to have obtained that information from the Linux tg3 driver.
2010-11-14 15:15:22 +00:00
error = mii_attach(dev, &sc->bge_miibus, ifp, bge_ifmedia_upd,
Do not hardcode phy address. Multi-port controllers use different phy address.
2012-10-24 05:00:56 +00:00
bge_ifmedia_sts, capmask, sc->bge_phy_addr, MII_OFFSET_ANY,
Move the limiting of the PHY to 10/100 modes of operation due to limitations of certain MAC models from brgphy(4) to bge(4) where it belongs. While at it, update the list of models having that restriction to what OpenBSD uses, which in turn seems to have obtained that information from the Linux tg3 driver.
2010-11-14 15:15:22 +00:00
MIIF_DOPAUSE);
Convert the PHY drivers to honor the mii_flags passed down and convert the NIC drivers as well as the PHY drivers to take advantage of the mii_attach() introduced in r213878 to get rid of certain hacks. For the most part these were: - Artificially limiting miibus_{read,write}reg methods to certain PHY addresses; we now let mii_attach() only probe the PHY at the desired address(es) instead. - PHY drivers setting MIIF_* flags based on the NIC driver they hang off from, partly even based on grabbing and using the softc of the parent; we now pass these flags down from the NIC to the PHY drivers via mii_attach(). This got us rid of all such hacks except those of brgphy() in combination with bce(4) and bge(4), which is way beyond what can be expressed with simple flags. While at it, I took the opportunity to change the NIC drivers to pass up the error returned by mii_attach() (previously by mii_phy_probe()) and unify the error message used in this case where and as appropriate as mii_attach() actually can fail for a number of reasons, not just because of no PHY(s) being present at the expected address(es). Reviewed by: jhb, yongari
2010-10-15 14:52:11 +00:00
if (error != 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
if (trys++ < 4) {
device_printf(sc->bge_dev, "Try again\n");
Do not hardcode phy address. Multi-port controllers use different phy address.
2012-10-24 05:00:56 +00:00
bge_miibus_writereg(sc->bge_dev,
sc->bge_phy_addr, 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;
}
Convert the PHY drivers to honor the mii_flags passed down and convert the NIC drivers as well as the PHY drivers to take advantage of the mii_attach() introduced in r213878 to get rid of certain hacks. For the most part these were: - Artificially limiting miibus_{read,write}reg methods to certain PHY addresses; we now let mii_attach() only probe the PHY at the desired address(es) instead. - PHY drivers setting MIIF_* flags based on the NIC driver they hang off from, partly even based on grabbing and using the softc of the parent; we now pass these flags down from the NIC to the PHY drivers via mii_attach(). This got us rid of all such hacks except those of brgphy() in combination with bce(4) and bge(4), which is way beyond what can be expressed with simple flags. While at it, I took the opportunity to change the NIC drivers to pass up the error returned by mii_attach() (previously by mii_phy_probe()) and unify the error message used in this case where and as appropriate as mii_attach() actually can fail for a number of reasons, not just because of no PHY(s) being present at the expected address(es). Reviewed by: jhb, yongari
2010-10-15 14:52:11 +00:00
device_printf(sc->bge_dev, "attaching PHYs 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
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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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.
*/
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);
Rework device detach. While here, move driver lock/callout initialization to the beginning of device attach for readability.
2012-10-08 07:33:43 +00:00
error = ENOMEM;
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
goto fail;
}
- Improve some comments. - Make bge_lookup_{rev,vendor}() static. - Factor out chip identification rather than duplicating the code. - Sanitize bge_probe() a bit (don't hardcode buffer sizes, allow bge_lookup_vendor() to return NULL so the excessive panic() three can be removed there, etc.) and return BUS_PROBE_DEFAULT rather than hardcoding 0. - According to the Linux tg3 driver, BCM57791 and BCM57795 aren't capable of Gigabit Ethernet. - Check the return value of taskqueue_start_threads().
2013-01-26 00:11:39 +00:00
error = taskqueue_start_threads(&sc->bge_tq, 1, PI_NET,
"%s taskq", device_get_nameunit(sc->bge_dev));
if (error != 0) {
device_printf(dev, "could not start threads.\n");
ether_ifdetach(ifp);
goto fail;
}
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
error = bus_setup_intr(dev, sc->bge_irq,
INTR_TYPE_NET | INTR_MPSAFE, bge_msi_intr, NULL, sc,
&sc->bge_intrhand);
} else
error = bus_setup_intr(dev, sc->bge_irq,
INTR_TYPE_NET | INTR_MPSAFE, NULL, bge_intr, sc,
&sc->bge_intrhand);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
if (error) {
Rework device detach. While here, move driver lock/callout initialization to the beginning of device attach for readability.
2012-10-08 07:33:43 +00:00
ether_ifdetach(ifp);
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "couldn't set up irq\n");
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
fail:
Rework device detach. While here, move driver lock/callout initialization to the beginning of device attach for readability.
2012-10-08 07:33:43 +00:00
if (error)
bge_detach(dev);
- 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
Rework device detach. While here, move driver lock/callout initialization to the beginning of device attach for readability.
2012-10-08 07:33:43 +00:00
if (device_is_attached(dev)) {
ether_ifdetach(ifp);
BGE_LOCK(sc);
bge_stop(sc);
BGE_UNLOCK(sc);
callout_drain(&sc->bge_stat_ch);
}
- 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
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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_attach() can fail before attaching mii(4). So make sure to check bge_miibus before detaching mii(4).
2013-01-24 04:54:53 +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);
bge_attach() can fail before attaching mii(4). So make sure to check bge_miibus before detaching mii(4).
2013-01-24 04:54:53 +00:00
else if (sc->bge_miibus != 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_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
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
if (sc->bge_res2 != NULL)
bus_release_resource(dev, SYS_RES_MEMORY,
PCIR_BAR(2), sc->bge_res2);
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;
Rework controller reset procedure. Previously driver saved BGE_PCI_PCISTATE register before issuing global reset. After issuing reset, it reads BGE_PCI_PCISTATE register again and compares the saved register value and current value. It was used to know whether the global reset operation was completed or not. Unfortunately, this logic caused several issues on recent BCM5717/ 5718/5719 and BCM5720 controllers. It seems APE firmware accesses some registers while global reset is in progress such that reading BGE_PCI_PCISTATE register after reset does not yield old pre-reset state value. This resulted in consuming too much time in global reset and sometimes it couldn't successfully complete reset. The BGE_MISCCFG_RESET_CORE_CLOCKS of BGE_MISC_CFG register is self-clearing bit so driver is able to know the reset completion. But the core-lock reset will disable indirect/flat/standard access modes such that driver cannot poll BGE_MISCCFG_RESET_CORE_CLOCKS bit of BGE_MISC_CFG register. So just wait enough time for core-clock reset to complete. Data sheet says driver should wait 100us for PCI/PCI-X devices and 100ms for PCIe devices. I chose 1ms for PCI/PCI-X since this value was used for many years in bge(4). For PCIe devices, use 100ms as recommended by data sheet. bge_chipinit() also cleared BGE_MAC_MODE register which shall clear firmware configured mode information. I think this will result in losing ASF/IPMI link in device attachment. Let bge_reset() honor firmware configured BGE_MAC_MODE register and don't announce driver is UP in bge_reset(). Firmware should have control over driver until it's fully initialized by driver. While I'm here, enable workaround for PCI-X BCM5704 A0 in bge_reset(). This will prevent internal arbitration logic from switching to the other DMA engine after a retry cycle.
2012-10-11 05:48:04 +00:00
uint32_t cachesize, command, mac_mode, mac_mode_mask, 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;
Rework controller reset procedure. Previously driver saved BGE_PCI_PCISTATE register before issuing global reset. After issuing reset, it reads BGE_PCI_PCISTATE register again and compares the saved register value and current value. It was used to know whether the global reset operation was completed or not. Unfortunately, this logic caused several issues on recent BCM5717/ 5718/5719 and BCM5720 controllers. It seems APE firmware accesses some registers while global reset is in progress such that reading BGE_PCI_PCISTATE register after reset does not yield old pre-reset state value. This resulted in consuming too much time in global reset and sometimes it couldn't successfully complete reset. The BGE_MISCCFG_RESET_CORE_CLOCKS of BGE_MISC_CFG register is self-clearing bit so driver is able to know the reset completion. But the core-lock reset will disable indirect/flat/standard access modes such that driver cannot poll BGE_MISCCFG_RESET_CORE_CLOCKS bit of BGE_MISC_CFG register. So just wait enough time for core-clock reset to complete. Data sheet says driver should wait 100us for PCI/PCI-X devices and 100ms for PCIe devices. I chose 1ms for PCI/PCI-X since this value was used for many years in bge(4). For PCIe devices, use 100ms as recommended by data sheet. bge_chipinit() also cleared BGE_MAC_MODE register which shall clear firmware configured mode information. I think this will result in losing ASF/IPMI link in device attachment. Let bge_reset() honor firmware configured BGE_MAC_MODE register and don't announce driver is UP in bge_reset(). Firmware should have control over driver until it's fully initialized by driver. While I'm here, enable workaround for PCI-X BCM5704 A0 in bge_reset(). This will prevent internal arbitration logic from switching to the other DMA engine after a retry cycle.
2012-10-11 05:48:04 +00:00
mac_mode_mask = BGE_MACMODE_HALF_DUPLEX | BGE_MACMODE_PORTMODE;
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
if ((sc->bge_mfw_flags & BGE_MFW_ON_APE) != 0)
mac_mode_mask |= BGE_MACMODE_APE_RX_EN | BGE_MACMODE_APE_TX_EN;
Rework controller reset procedure. Previously driver saved BGE_PCI_PCISTATE register before issuing global reset. After issuing reset, it reads BGE_PCI_PCISTATE register again and compares the saved register value and current value. It was used to know whether the global reset operation was completed or not. Unfortunately, this logic caused several issues on recent BCM5717/ 5718/5719 and BCM5720 controllers. It seems APE firmware accesses some registers while global reset is in progress such that reading BGE_PCI_PCISTATE register after reset does not yield old pre-reset state value. This resulted in consuming too much time in global reset and sometimes it couldn't successfully complete reset. The BGE_MISCCFG_RESET_CORE_CLOCKS of BGE_MISC_CFG register is self-clearing bit so driver is able to know the reset completion. But the core-lock reset will disable indirect/flat/standard access modes such that driver cannot poll BGE_MISCCFG_RESET_CORE_CLOCKS bit of BGE_MISC_CFG register. So just wait enough time for core-clock reset to complete. Data sheet says driver should wait 100us for PCI/PCI-X devices and 100ms for PCIe devices. I chose 1ms for PCI/PCI-X since this value was used for many years in bge(4). For PCIe devices, use 100ms as recommended by data sheet. bge_chipinit() also cleared BGE_MAC_MODE register which shall clear firmware configured mode information. I think this will result in losing ASF/IPMI link in device attachment. Let bge_reset() honor firmware configured BGE_MAC_MODE register and don't announce driver is UP in bge_reset(). Firmware should have control over driver until it's fully initialized by driver. While I'm here, enable workaround for PCI-X BCM5704 A0 in bge_reset(). This will prevent internal arbitration logic from switching to the other DMA engine after a retry cycle.
2012-10-11 05:48:04 +00:00
mac_mode = CSR_READ_4(sc, BGE_MAC_MODE) & mac_mode_mask;
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 APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
/* Take APE lock when performing reset. */
bge_ape_lock(sc, BGE_APE_LOCK_GRC);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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);
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
Rename definition of BGE_SOFTWARE_GENCOMM_* to more readable ones. The origin of GENCOMM seems to come from Alteon Tigon Host/NIC interface definition where it defines general communications region which is active when firmware is loaded and running. This region was used in communication between the host and processor internal to the Tigon chip. Broadcom data sheet also defines the region as 'Software Gencomm' in NetXtreme memory map but lacks detailed description of its interface so it was hard to know which ones are used for which interface. This change shall slightly enhance readability. No functional changes.
2011-10-26 21:05:45 +00:00
* write ~BGE_SRAM_FW_MB_MAGIC to the same location.
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
*/
Rename definition of BGE_SOFTWARE_GENCOMM_* to more readable ones. The origin of GENCOMM seems to come from Alteon Tigon Host/NIC interface definition where it defines general communications region which is active when firmware is loaded and running. This region was used in communication between the host and processor internal to the Tigon chip. Broadcom data sheet also defines the region as 'Software Gencomm' in NetXtreme memory map but lacks detailed description of its interface so it was hard to know which ones are used for which interface. This change shall slightly enhance readability. No functional changes.
2011-10-26 21:05:45 +00:00
bge_writemem_ind(sc, BGE_SRAM_FW_MB, BGE_SRAM_FW_MB_MAGIC);
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
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) {
Do not force PCIe 1.0a mode in device reset on BCM5717 and newer controllers. BCM5785 does not require PCI 1.0a mode as well during reset.
2012-10-10 02:35:10 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5785 &&
(sc->bge_flags & BGE_FLAG_5717_PLUS) == 0) {
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
Fix a long standing VCPU reset sequence bug on BCM5906. The VCPU(Virtual CPU) of BCM5906 is used to provide a mechanism to control the bootcode execution and to pick up configuration data stored inside the EEPROM. The bootcode of BCM5906 will check the BGE_VCPU_STATUS_DRV_RESET bit to decide which booting procedure to choose. Data sheet indicates the VCPU of BCM5906 should set BGE_VCPU_STATUS_DRV_RESET bit *before* VCPU reset or global reset.
2012-10-10 04:38:29 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5906) {
val = CSR_READ_4(sc, BGE_VCPU_STATUS);
CSR_WRITE_4(sc, BGE_VCPU_STATUS,
val | BGE_VCPU_STATUS_DRV_RESET);
val = CSR_READ_4(sc, BGE_VCPU_EXT_CTRL);
CSR_WRITE_4(sc, BGE_VCPU_EXT_CTRL,
val & ~BGE_VCPU_EXT_CTRL_HALT_CPU);
}
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.
*/
Make CPMU handle GPHY power down control on controllers that have CPMU capability.
2011-10-26 18:05:46 +00:00
if (BGE_IS_5705_PLUS(sc) &&
(sc->bge_flags & BGE_FLAG_CPMU_PRESENT) == 0)
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
Rework controller reset procedure. Previously driver saved BGE_PCI_PCISTATE register before issuing global reset. After issuing reset, it reads BGE_PCI_PCISTATE register again and compares the saved register value and current value. It was used to know whether the global reset operation was completed or not. Unfortunately, this logic caused several issues on recent BCM5717/ 5718/5719 and BCM5720 controllers. It seems APE firmware accesses some registers while global reset is in progress such that reading BGE_PCI_PCISTATE register after reset does not yield old pre-reset state value. This resulted in consuming too much time in global reset and sometimes it couldn't successfully complete reset. The BGE_MISCCFG_RESET_CORE_CLOCKS of BGE_MISC_CFG register is self-clearing bit so driver is able to know the reset completion. But the core-lock reset will disable indirect/flat/standard access modes such that driver cannot poll BGE_MISCCFG_RESET_CORE_CLOCKS bit of BGE_MISC_CFG register. So just wait enough time for core-clock reset to complete. Data sheet says driver should wait 100us for PCI/PCI-X devices and 100ms for PCIe devices. I chose 1ms for PCI/PCI-X since this value was used for many years in bge(4). For PCIe devices, use 100ms as recommended by data sheet. bge_chipinit() also cleared BGE_MAC_MODE register which shall clear firmware configured mode information. I think this will result in losing ASF/IPMI link in device attachment. Let bge_reset() honor firmware configured BGE_MAC_MODE register and don't announce driver is UP in bge_reset(). Firmware should have control over driver until it's fully initialized by driver. While I'm here, enable workaround for PCI-X BCM5704 A0 in bge_reset(). This will prevent internal arbitration logic from switching to the other DMA engine after a retry cycle.
2012-10-11 05:48:04 +00:00
if (sc->bge_flags & BGE_FLAG_PCIE)
DELAY(100 * 1000);
else
DELAY(1000);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
/* 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,
Align the PCI Express #defines with the style used for the PCI-X #defines. This also has the advantage that it makes the names more compact, iand also allows us to correct the non-uniform naming of the PCIM_LINK_* defines, making them all consistent amongst themselves. This is a mostly mechanical rename: s/PCIR_EXPRESS_/PCIER_/g s/PCIM_EXP_/PCIEM_/g s/PCIM_LINK_/PCIEM_LINK_/g When this is MFC'd, #defines will be added for the old names to assist out-of-tree drivers. Discussed with: jhb MFC after: 1 week
2012-09-18 22:04:59 +00:00
sc->bge_expcap + PCIER_DEVICE_CTL, 2);
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
/* Clear enable no snoop and disable relaxed ordering. */
Align the PCI Express #defines with the style used for the PCI-X #defines. This also has the advantage that it makes the names more compact, iand also allows us to correct the non-uniform naming of the PCIM_LINK_* defines, making them all consistent amongst themselves. This is a mostly mechanical rename: s/PCIR_EXPRESS_/PCIER_/g s/PCIM_EXP_/PCIEM_/g s/PCIM_LINK_/PCIEM_LINK_/g When this is MFC'd, #defines will be added for the old names to assist out-of-tree drivers. Discussed with: jhb MFC after: 1 week
2012-09-18 22:04:59 +00:00
devctl &= ~(PCIEM_CTL_RELAXED_ORD_ENABLE |
PCIEM_CTL_NOSNOOP_ENABLE);
pci_write_config(dev, sc->bge_expcap + PCIER_DEVICE_CTL,
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, 2);
If the maximum payload size is 256 bytes or more, set the DMA write water mark to 256 bytes. Otherwise controller will encounter DMA write under run errors and would result in RX DMA hang. If the maximum payload size is 128 bytes, the water mark is set to 128 bytes as usual. While here, set maximum read request size to 2048 for BCM5719/BCM5720. For other PCIe devices, use 4096. And reprogram the maximum read request size whenever device reset is performed.
2012-10-10 01:24:02 +00:00
pci_set_max_read_req(dev, sc->bge_expmrq);
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
/* Clear error status. */
Align the PCI Express #defines with the style used for the PCI-X #defines. This also has the advantage that it makes the names more compact, iand also allows us to correct the non-uniform naming of the PCIM_LINK_* defines, making them all consistent amongst themselves. This is a mostly mechanical rename: s/PCIR_EXPRESS_/PCIER_/g s/PCIM_EXP_/PCIEM_/g s/PCIM_LINK_/PCIEM_LINK_/g When this is MFC'd, #defines will be added for the old names to assist out-of-tree drivers. Discussed with: jhb MFC after: 1 week
2012-09-18 22:04:59 +00:00
pci_write_config(dev, sc->bge_expcap + PCIER_DEVICE_STA,
PCIEM_STA_CORRECTABLE_ERROR |
PCIEM_STA_NON_FATAL_ERROR | PCIEM_STA_FATAL_ERROR |
PCIEM_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);
Rework controller reset procedure. Previously driver saved BGE_PCI_PCISTATE register before issuing global reset. After issuing reset, it reads BGE_PCI_PCISTATE register again and compares the saved register value and current value. It was used to know whether the global reset operation was completed or not. Unfortunately, this logic caused several issues on recent BCM5717/ 5718/5719 and BCM5720 controllers. It seems APE firmware accesses some registers while global reset is in progress such that reading BGE_PCI_PCISTATE register after reset does not yield old pre-reset state value. This resulted in consuming too much time in global reset and sometimes it couldn't successfully complete reset. The BGE_MISCCFG_RESET_CORE_CLOCKS of BGE_MISC_CFG register is self-clearing bit so driver is able to know the reset completion. But the core-lock reset will disable indirect/flat/standard access modes such that driver cannot poll BGE_MISCCFG_RESET_CORE_CLOCKS bit of BGE_MISC_CFG register. So just wait enough time for core-clock reset to complete. Data sheet says driver should wait 100us for PCI/PCI-X devices and 100ms for PCIe devices. I chose 1ms for PCI/PCI-X since this value was used for many years in bge(4). For PCIe devices, use 100ms as recommended by data sheet. bge_chipinit() also cleared BGE_MAC_MODE register which shall clear firmware configured mode information. I think this will result in losing ASF/IPMI link in device attachment. Let bge_reset() honor firmware configured BGE_MAC_MODE register and don't announce driver is UP in bge_reset(). Firmware should have control over driver until it's fully initialized by driver. While I'm here, enable workaround for PCI-X BCM5704 A0 in bge_reset(). This will prevent internal arbitration logic from switching to the other DMA engine after a retry cycle.
2012-10-11 05:48:04 +00:00
val = BGE_PCISTATE_ROM_ENABLE | BGE_PCISTATE_ROM_RETRY_ENABLE;
if (sc->bge_chipid == BGE_CHIPID_BCM5704_A0 &&
(sc->bge_flags & BGE_FLAG_PCIX) != 0)
val |= BGE_PCISTATE_RETRY_SAME_DMA;
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
if ((sc->bge_mfw_flags & BGE_MFW_ON_APE) != 0)
val |= BGE_PCISTATE_ALLOW_APE_CTLSPC_WR |
BGE_PCISTATE_ALLOW_APE_SHMEM_WR |
BGE_PCISTATE_ALLOW_APE_PSPACE_WR;
Rework controller reset procedure. Previously driver saved BGE_PCI_PCISTATE register before issuing global reset. After issuing reset, it reads BGE_PCI_PCISTATE register again and compares the saved register value and current value. It was used to know whether the global reset operation was completed or not. Unfortunately, this logic caused several issues on recent BCM5717/ 5718/5719 and BCM5720 controllers. It seems APE firmware accesses some registers while global reset is in progress such that reading BGE_PCI_PCISTATE register after reset does not yield old pre-reset state value. This resulted in consuming too much time in global reset and sometimes it couldn't successfully complete reset. The BGE_MISCCFG_RESET_CORE_CLOCKS of BGE_MISC_CFG register is self-clearing bit so driver is able to know the reset completion. But the core-lock reset will disable indirect/flat/standard access modes such that driver cannot poll BGE_MISCCFG_RESET_CORE_CLOCKS bit of BGE_MISC_CFG register. So just wait enough time for core-clock reset to complete. Data sheet says driver should wait 100us for PCI/PCI-X devices and 100ms for PCIe devices. I chose 1ms for PCI/PCI-X since this value was used for many years in bge(4). For PCIe devices, use 100ms as recommended by data sheet. bge_chipinit() also cleared BGE_MAC_MODE register which shall clear firmware configured mode information. I think this will result in losing ASF/IPMI link in device attachment. Let bge_reset() honor firmware configured BGE_MAC_MODE register and don't announce driver is UP in bge_reset(). Firmware should have control over driver until it's fully initialized by driver. While I'm here, enable workaround for PCI-X BCM5704 A0 in bge_reset(). This will prevent internal arbitration logic from switching to the other DMA engine after a retry cycle.
2012-10-11 05:48:04 +00:00
pci_write_config(dev, BGE_PCI_PCISTATE, val, 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);
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);
}
Correct spelling in comments. Submitted by: brucec
2011-05-02 21:04:23 +00:00
/* Re-enable MSI, if necessary, 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);
Rework controller reset procedure. Previously driver saved BGE_PCI_PCISTATE register before issuing global reset. After issuing reset, it reads BGE_PCI_PCISTATE register again and compares the saved register value and current value. It was used to know whether the global reset operation was completed or not. Unfortunately, this logic caused several issues on recent BCM5717/ 5718/5719 and BCM5720 controllers. It seems APE firmware accesses some registers while global reset is in progress such that reading BGE_PCI_PCISTATE register after reset does not yield old pre-reset state value. This resulted in consuming too much time in global reset and sometimes it couldn't successfully complete reset. The BGE_MISCCFG_RESET_CORE_CLOCKS of BGE_MISC_CFG register is self-clearing bit so driver is able to know the reset completion. But the core-lock reset will disable indirect/flat/standard access modes such that driver cannot poll BGE_MISCCFG_RESET_CORE_CLOCKS bit of BGE_MISC_CFG register. So just wait enough time for core-clock reset to complete. Data sheet says driver should wait 100us for PCI/PCI-X devices and 100ms for PCIe devices. I chose 1ms for PCI/PCI-X since this value was used for many years in bge(4). For PCIe devices, use 100ms as recommended by data sheet. bge_chipinit() also cleared BGE_MAC_MODE register which shall clear firmware configured mode information. I think this will result in losing ASF/IPMI link in device attachment. Let bge_reset() honor firmware configured BGE_MAC_MODE register and don't announce driver is UP in bge_reset(). Firmware should have control over driver until it's fully initialized by driver. While I'm here, enable workaround for PCI-X BCM5704 A0 in bge_reset(). This will prevent internal arbitration logic from switching to the other DMA engine after a retry cycle.
2012-10-11 05:48:04 +00:00
/* Fix up byte swapping. */
CSR_WRITE_4(sc, BGE_MODE_CTL, bge_dma_swap_options(sc));
val = CSR_READ_4(sc, BGE_MAC_MODE);
val = (val & ~mac_mode_mask) | mac_mode;
CSR_WRITE_4(sc, BGE_MAC_MODE, val);
DELAY(40);
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
bge_ape_unlock(sc, BGE_APE_LOCK_GRC);
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);
Rename definition of BGE_SOFTWARE_GENCOMM_* to more readable ones. The origin of GENCOMM seems to come from Alteon Tigon Host/NIC interface definition where it defines general communications region which is active when firmware is loaded and running. This region was used in communication between the host and processor internal to the Tigon chip. Broadcom data sheet also defines the region as 'Software Gencomm' in NetXtreme memory map but lacks detailed description of its interface so it was hard to know which ones are used for which interface. This change shall slightly enhance readability. No functional changes.
2011-10-26 21:05:45 +00:00
val = bge_readmem_ind(sc, BGE_SRAM_FW_MB);
if (val == ~BGE_SRAM_FW_MB_MAGIC)
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
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 initial BCM57765 family support. The BCM57765 family seems to have similar hardware features of BCM5718 family except the number of receive return ring is 4. The BCM57765 family is known to support IEEE 802.3az EEE(Energy Efficient Ethernet) but this change does not include EEE support code. I hope EEE is implemented in near future. This change will support BCM57761, BCM57765, BCM57781, BCM57785, BCM57791 and BCM57795. All hardware offloading features are supported and suspend/resume also should work. Many thanks to Broadcom for continuing support of FreeBSD. Tested by: Paul Thornton (prt <> prt dot org) HW donated by: Broadcom
2011-05-04 17:04:31 +00:00
/* BCM57765 A0 needs additional time before accessing. */
if (sc->bge_chipid == BGE_CHIPID_BCM57765_A0)
DELAY(10 * 1000); /* XXX */
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
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 &&
Add initial BCM57765 family support. The BCM57765 family seems to have similar hardware features of BCM5718 family except the number of receive return ring is 4. The BCM57765 family is known to support IEEE 802.3az EEE(Energy Efficient Ethernet) but this change does not include EEE support code. I hope EEE is implemented in near future. This change will support BCM57761, BCM57765, BCM57781, BCM57785, BCM57791 and BCM57795. All hardware offloading features are supported and suspend/resume also should work. Many thanks to Broadcom for continuing support of FreeBSD. Tested by: Paul Thornton (prt <> prt dot org) HW donated by: Broadcom
2011-05-04 17:04:31 +00:00
!BGE_IS_5717_PLUS(sc) &&
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 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 initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5720)
BGE_CLRBIT(sc, BGE_CPMU_CLCK_ORIDE,
CPMU_CLCK_ORIDE_MAC_ORIDE_EN);
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);
Add initial jumbo frame support for BCM5714/BCM5715 and BCM5780. Unlike other controllers which have more advanced jumbo support, these controllers have one send ring, one standard receive producer ring and one receive return ring. In order to receive jumbo frames on the controllers, driver now will increase Rx buffer size to 9k. Two Rx modes are supported on these controllers and I chose standard Rx BDs over extended Rx BDs. The extended Rx BD mode allows up to 4 segmentations for each Rx BDs such that kernel does not have to allocate large buffer of contiguous memory for receiving. The extended Rx BD mode is already used on controllers that have separate jumbo receive ring. However, using extended Rx BDs on BCM5714/BCM5715/BCM5780 reduces the number of Rx BDs to 256 entries which in turn may reduce the performance. Also UMA backed page allocator for jumbo frame returns contiguous memory so using extended Rx BD has no advantage on FreeBSD unless highly customized local allocator implemented in driver is used. To use jumbo buffers in standard receive ring, Rx buffer allocation handler was changed to allocate MJUM9BYTES sized mbuf. PR: kern/155192 Tested by: Vijay Singh <vijju.singh <> gmail dot com> Submitted by: mjacob (initial version)
2011-04-05 17:41:54 +00:00
if (BGE_IS_JUMBO_CAPABLE(sc) &&
ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN >
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
(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;
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if (ifp->if_capenable & IFCAP_RXCSUM)
bge_rxcsum(sc, cur_rx, 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
/*
network interface driver changes: o don't strip the Ethernet header from inbound packets; pass packets up the stack intact (required significant changes to some drivers) o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN) o track ether_ifattach/ether_ifdetach API changes o track bpf changes (use BPF_TAP and BPF_MTAP) o track vlan changes (ifnet capabilities, revised processing scheme, etc.) o use if_input to pass packets "up" o call ether_ioctl for default handling of ioctls Reviewed by: many Approved by: re
2002-11-14 23:54:55 +00:00
* If we received a packet with a vlan tag,
* attach that information to the packet.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
*/
- Fix VLAN_INPUT_TAG() macro, so that it doesn't touch mtag in case if memory allocation failed. - Remove fourth argument from VLAN_INPUT_TAG(), that was used incorrectly in almost all drivers. Indicate failure with mbuf value of NULL. In collaboration with: yongari, ru, sam
2005-12-18 18:24:27 +00:00
if (have_tag) {
Move ethernet VLAN tags from mtags to its own mbuf packet header field m_pkthdr.ether_vlan. The presence of the M_VLANTAG flag on the mbuf signifies the presence and validity of its content. Drivers that support hardware VLAN tag stripping fill in the received VLAN tag (containing both vlan and priority information) into the ether_vtag mbuf packet header field: m->m_pkthdr.ether_vtag = vlan_id; /* ntohs()? */ m->m_flags |= M_VLANTAG; to mark the packet m with the specified VLAN tag. On output the driver should check the mbuf for the M_VLANTAG flag to see if a VLAN tag is present and valid: if (m->m_flags & M_VLANTAG) { ... = m->m_pkthdr.ether_vtag; /* htons()? */ ... pass tag to hardware ... } VLAN tags are stored in host byte order. Byte swapping may be necessary. (Note: This driver conversion was mechanic and did not add or remove any byte swapping in the drivers.) Remove zone_mtag_vlan UMA zone and MTAG_VLAN definition. No more tag memory allocation have to be done. Reviewed by: thompsa, yar Sponsored by: TCP/IP Optimization Fundraise 2005
2006-09-17 13:33:30 +00:00
m->m_pkthdr.ether_vtag = vlan_tag;
m->m_flags |= M_VLANTAG;
- Fix VLAN_INPUT_TAG() macro, so that it doesn't touch mtag in case if memory allocation failed. - Remove fourth argument from VLAN_INPUT_TAG(), that was used incorrectly in almost all drivers. Indicate failure with mbuf value of NULL. In collaboration with: yongari, ru, sam
2005-12-18 18:24:27 +00:00
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
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)
Fix incorrect RX BD producer updates. The producer index was already updated after allocating mbuf so driver had to use the last index instead of using next producer index. This should fix driver hang which may happen under high network load. Reported by: Igor Sysoev <is <> rambler-co dot ru>, Vlad Galu <dudu <> dudu dot ro> Tested by: Igor Sysoev <is <> rambler-co dot ru>, Vlad Galu <dudu <> dudu dot ro> MFC after: 10 days
2010-09-16 17:32:37 +00:00
bge_writembx(sc, BGE_MBX_RX_STD_PROD_LO, (sc->bge_std +
BGE_STD_RX_RING_CNT - 1) % 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
if (jumbocnt)
Fix incorrect RX BD producer updates. The producer index was already updated after allocating mbuf so driver had to use the last index instead of using next producer index. This should fix driver hang which may happen under high network load. Reported by: Igor Sysoev <is <> rambler-co dot ru>, Vlad Galu <dudu <> dudu dot ro> Tested by: Igor Sysoev <is <> rambler-co dot ru>, Vlad Galu <dudu <> dudu dot ro> MFC after: 10 days
2010-09-16 17:32:37 +00:00
bge_writembx(sc, BGE_MBX_RX_JUMBO_PROD_LO, (sc->bge_jumbo +
BGE_JUMBO_RX_RING_CNT - 1) % BGE_JUMBO_RX_RING_CNT);
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
}
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
static void
bge_rxcsum(struct bge_softc *sc, struct bge_rx_bd *cur_rx, struct mbuf *m)
{
if (BGE_IS_5717_PLUS(sc)) {
if ((cur_rx->bge_flags & BGE_RXBDFLAG_IPV6) == 0) {
if (cur_rx->bge_flags & BGE_RXBDFLAG_IP_CSUM) {
m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED;
if ((cur_rx->bge_error_flag &
BGE_RXERRFLAG_IP_CSUM_NOK) == 0)
m->m_pkthdr.csum_flags |= CSUM_IP_VALID;
}
if (cur_rx->bge_flags & BGE_RXBDFLAG_TCP_UDP_CSUM) {
m->m_pkthdr.csum_data =
cur_rx->bge_tcp_udp_csum;
m->m_pkthdr.csum_flags |= CSUM_DATA_VALID |
CSUM_PSEUDO_HDR;
}
}
} else {
if (cur_rx->bge_flags & BGE_RXBDFLAG_IP_CSUM) {
m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED;
if ((cur_rx->bge_ip_csum ^ 0xFFFF) == 0)
m->m_pkthdr.csum_flags |= CSUM_IP_VALID;
}
if (cur_rx->bge_flags & BGE_RXBDFLAG_TCP_UDP_CSUM &&
m->m_pkthdr.len >= ETHER_MIN_NOPAD) {
m->m_pkthdr.csum_data =
cur_rx->bge_tcp_udp_csum;
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
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);
On PHY write error use hex number to show the value. Add more comments.
2012-10-10 01:59:53 +00:00
/* Fetch updates from the status block. */
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
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;
On PHY write error use hex number to show the value. Add more comments.
2012-10-10 01:59:53 +00:00
/* Clear the status so the next pass only sees the changes. */
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
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;
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
uint32_t status, status_tag;
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
uint16_t rx_prod, tx_cons;
sc = (struct bge_softc *)arg;
ifp = sc->bge_ifp;
The IFF_DRV_RUNNING flag is set at the end of bge_init_locked. But before setting the flag, interrupt was already enabled such that interrupt handler could be run before setting IFF_DRV_RUNNING flag. This can lose initial link state change interrupt which in turn make bge(4) think that it still does not have valid link. Fix this race by protecting the taskqueue with a driver lock. While I'm here move reenabling interrupt code after handling of link state chage. Reviewed by: davidch
2010-10-11 23:07:12 +00:00
BGE_LOCK(sc);
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
BGE_UNLOCK(sc);
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
return;
The IFF_DRV_RUNNING flag is set at the end of bge_init_locked. But before setting the flag, interrupt was already enabled such that interrupt handler could be run before setting IFF_DRV_RUNNING flag. This can lose initial link state change interrupt which in turn make bge(4) think that it still does not have valid link. Fix this race by protecting the taskqueue with a driver lock. While I'm here move reenabling interrupt code after handling of link state chage. Reviewed by: davidch
2010-10-11 23:07:12 +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
/* 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);
On PHY write error use hex number to show the value. Add more comments.
2012-10-10 01:59:53 +00:00
/* Save producer/consumer indices. */
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_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;
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
status_tag = sc->bge_ldata.bge_status_block->bge_status_tag << 24;
On PHY write error use hex number to show the value. Add more comments.
2012-10-10 01:59:53 +00:00
/* Dirty the status flag. */
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_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);
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if ((sc->bge_flags & BGE_FLAG_TAGGED_STATUS) == 0)
status_tag = 0;
The IFF_DRV_RUNNING flag is set at the end of bge_init_locked. But before setting the flag, interrupt was already enabled such that interrupt handler could be run before setting IFF_DRV_RUNNING flag. This can lose initial link state change interrupt which in turn make bge(4) think that it still does not have valid link. Fix this race by protecting the taskqueue with a driver lock. While I'm here move reenabling interrupt code after handling of link state chage. Reviewed by: davidch
2010-10-11 23:07:12 +00:00
if ((status & BGE_STATFLAG_LINKSTATE_CHANGED) != 0)
bge_link_upd(sc);
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
/* Let controller work. */
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
bge_writembx(sc, BGE_MBX_IRQ0_LO, status_tag);
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
The IFF_DRV_RUNNING flag is set at the end of bge_init_locked. But before setting the flag, interrupt was already enabled such that interrupt handler could be run before setting IFF_DRV_RUNNING flag. This can lose initial link state change interrupt which in turn make bge(4) think that it still does not have valid link. Fix this race by protecting the taskqueue with a driver lock. While I'm here move reenabling interrupt code after handling of link state chage. Reviewed by: davidch
2010-10-11 23:07:12 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
sc->bge_rx_saved_considx != rx_prod) {
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
/* Check RX return ring producer/consumer. */
The IFF_DRV_RUNNING flag is set at the end of bge_init_locked. But before setting the flag, interrupt was already enabled such that interrupt handler could be run before setting IFF_DRV_RUNNING flag. This can lose initial link state change interrupt which in turn make bge(4) think that it still does not have valid link. Fix this race by protecting the taskqueue with a driver lock. While I'm here move reenabling interrupt code after handling of link state chage. Reviewed by: davidch
2010-10-11 23:07:12 +00:00
BGE_UNLOCK(sc);
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, 0);
The IFF_DRV_RUNNING flag is set at the end of bge_init_locked. But before setting the flag, interrupt was already enabled such that interrupt handler could be run before setting IFF_DRV_RUNNING flag. This can lose initial link state change interrupt which in turn make bge(4) think that it still does not have valid link. Fix this race by protecting the taskqueue with a driver lock. While I'm here move reenabling interrupt code after handling of link state chage. Reviewed by: davidch
2010-10-11 23:07:12 +00:00
BGE_LOCK(sc);
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 (ifp->if_drv_flags & IFF_DRV_RUNNING) {
/* Check TX ring producer/consumer. */
bge_txeof(sc, tx_cons);
Whitespace nits.
2011-10-25 20:45:14 +00:00
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
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_start_locked(ifp);
}
The IFF_DRV_RUNNING flag is set at the end of bge_init_locked. But before setting the flag, interrupt was already enabled such that interrupt handler could be run before setting IFF_DRV_RUNNING flag. This can lose initial link state change interrupt which in turn make bge(4) think that it still does not have valid link. Fix this race by protecting the taskqueue with a driver lock. While I'm here move reenabling interrupt code after handling of link state chage. Reviewed by: davidch
2010-10-11 23:07:12 +00:00
BGE_UNLOCK(sc);
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
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
Rename definition of BGE_SOFTWARE_GENCOMM_* to more readable ones. The origin of GENCOMM seems to come from Alteon Tigon Host/NIC interface definition where it defines general communications region which is active when firmware is loaded and running. This region was used in communication between the host and processor internal to the Tigon chip. Broadcom data sheet also defines the region as 'Software Gencomm' in NetXtreme memory map but lacks detailed description of its interface so it was hard to know which ones are used for which interface. This change shall slightly enhance readability. No functional changes.
2011-10-26 21:05:45 +00:00
bge_writemem_ind(sc, BGE_SRAM_FW_CMD_MB,
Rename BGE_FW_DRV_ALIVE/BGE_FW_PAUSE to BGE_FW_CMD_DRV_ALIVE/BGE_FW_CMD_PAUSE. Also add more firmware commands(not used yet). No functional changes.
2011-10-27 20:54:53 +00:00
BGE_FW_CMD_DRV_ALIVE);
Rename definition of BGE_SOFTWARE_GENCOMM_* to more readable ones. The origin of GENCOMM seems to come from Alteon Tigon Host/NIC interface definition where it defines general communications region which is active when firmware is loaded and running. This region was used in communication between the host and processor internal to the Tigon chip. Broadcom data sheet also defines the region as 'Software Gencomm' in NetXtreme memory map but lacks detailed description of its interface so it was hard to know which ones are used for which interface. This change shall slightly enhance readability. No functional changes.
2011-10-26 21:05:45 +00:00
bge_writemem_ind(sc, BGE_SRAM_FW_CMD_LEN_MB, 4);
Define BGE_FW_HB_TIMEOUT_SEC and remove one more magic value. bge(4) sends BGE_FW_CMD_DRV_ALIVE command to firmware every 2 seconds. BGE_FW_CMD_DRV_ALIVE command requires 4 bytes data. This data contains timeout value in seconds until the next BGE_FW_CMD_DRV_ALIVE command. Broadcom recommends driver set the value 3 times longer than the interval that it sends BGE_FW_CMD_DRV_ALIVE. Currently bge(4) uses 3 seconds so probably we have to increase it in future and use different ALIVE command(e.g. BGE_FW_CMD_DRV_ALIVE3). No functional changes.
2011-10-27 22:10:52 +00:00
bge_writemem_ind(sc, BGE_SRAM_FW_CMD_DATA_MB,
BGE_FW_HB_TIMEOUT_SEC);
Offset 0x6810 is RX-RISC event register. Rename BGE_CPU_EVENT with BGE_RX_CPU_EVENT for readability. Additionally define BGE_TX_CPU_EVENT for TX-RSIC event register(BCM570[0-4] only).
2011-10-26 23:22:32 +00:00
CSR_WRITE_4(sc, BGE_RX_CPU_EVENT,
Rename hard-coded value 1 << 14 with BGE_RX_CPU_DRV_EVENT. This bit(SW event 7 in publicly available data sheet) is used to make RX CPU handle a firmware command and the bit is automatically cleared after RX CPU completed the command. Generally firmware command takes the following steps. 1. Write BGE_SRAM_FW_CMD_MB with a command. 2. Write BGE_SRAM_FW_CMD_LEN_MB with the length of the command in bytes. 3. Write BGE_SRAM_FW_CMD_DATA_MB with actual command data. 4. Generate BGE_RX_CPU_EVENT and let firmware handle the command. 5. Wait for the ACK of the firmware command. No functional changes.
2011-10-27 21:27:37 +00:00
CSR_READ_4(sc, BGE_RX_CPU_EVENT) |
BGE_RX_CPU_DRV_EVENT);
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))
Whitespace nits.
2011-10-25 20:45:14 +00:00
return;
- 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
- 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
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
/* XXX Add APE heartbeat check here? */
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;
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
struct bge_mac_stats *stats;
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;
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
stats = &sc->bge_mac_stats;
stats->ifHCOutOctets +=
CSR_READ_4(sc, BGE_TX_MAC_STATS_OCTETS);
stats->etherStatsCollisions +=
CSR_READ_4(sc, BGE_TX_MAC_STATS_COLLS);
stats->outXonSent +=
CSR_READ_4(sc, BGE_TX_MAC_STATS_XON_SENT);
stats->outXoffSent +=
CSR_READ_4(sc, BGE_TX_MAC_STATS_XOFF_SENT);
stats->dot3StatsInternalMacTransmitErrors +=
CSR_READ_4(sc, BGE_TX_MAC_STATS_ERRORS);
stats->dot3StatsSingleCollisionFrames +=
CSR_READ_4(sc, BGE_TX_MAC_STATS_SINGLE_COLL);
stats->dot3StatsMultipleCollisionFrames +=
CSR_READ_4(sc, BGE_TX_MAC_STATS_MULTI_COLL);
stats->dot3StatsDeferredTransmissions +=
CSR_READ_4(sc, BGE_TX_MAC_STATS_DEFERRED);
stats->dot3StatsExcessiveCollisions +=
CSR_READ_4(sc, BGE_TX_MAC_STATS_EXCESS_COLL);
stats->dot3StatsLateCollisions +=
CSR_READ_4(sc, BGE_TX_MAC_STATS_LATE_COLL);
stats->ifHCOutUcastPkts +=
CSR_READ_4(sc, BGE_TX_MAC_STATS_UCAST);
stats->ifHCOutMulticastPkts +=
CSR_READ_4(sc, BGE_TX_MAC_STATS_MCAST);
stats->ifHCOutBroadcastPkts +=
CSR_READ_4(sc, BGE_TX_MAC_STATS_BCAST);
stats->ifHCInOctets +=
CSR_READ_4(sc, BGE_RX_MAC_STATS_OCTESTS);
stats->etherStatsFragments +=
CSR_READ_4(sc, BGE_RX_MAC_STATS_FRAGMENTS);
stats->ifHCInUcastPkts +=
CSR_READ_4(sc, BGE_RX_MAC_STATS_UCAST);
stats->ifHCInMulticastPkts +=
CSR_READ_4(sc, BGE_RX_MAC_STATS_MCAST);
stats->ifHCInBroadcastPkts +=
CSR_READ_4(sc, BGE_RX_MAC_STATS_BCAST);
stats->dot3StatsFCSErrors +=
CSR_READ_4(sc, BGE_RX_MAC_STATS_FCS_ERRORS);
stats->dot3StatsAlignmentErrors +=
CSR_READ_4(sc, BGE_RX_MAC_STATS_ALGIN_ERRORS);
stats->xonPauseFramesReceived +=
CSR_READ_4(sc, BGE_RX_MAC_STATS_XON_RCVD);
stats->xoffPauseFramesReceived +=
CSR_READ_4(sc, BGE_RX_MAC_STATS_XOFF_RCVD);
stats->macControlFramesReceived +=
CSR_READ_4(sc, BGE_RX_MAC_STATS_CTRL_RCVD);
stats->xoffStateEntered +=
CSR_READ_4(sc, BGE_RX_MAC_STATS_XOFF_ENTERED);
stats->dot3StatsFramesTooLong +=
CSR_READ_4(sc, BGE_RX_MAC_STATS_FRAME_TOO_LONG);
stats->etherStatsJabbers +=
CSR_READ_4(sc, BGE_RX_MAC_STATS_JABBERS);
stats->etherStatsUndersizePkts +=
CSR_READ_4(sc, BGE_RX_MAC_STATS_UNDERSIZE);
stats->FramesDroppedDueToFilters +=
CSR_READ_4(sc, BGE_RXLP_LOCSTAT_FILTDROP);
stats->DmaWriteQueueFull +=
CSR_READ_4(sc, BGE_RXLP_LOCSTAT_DMA_WRQ_FULL);
stats->DmaWriteHighPriQueueFull +=
CSR_READ_4(sc, BGE_RXLP_LOCSTAT_DMA_HPWRQ_FULL);
stats->NoMoreRxBDs +=
CSR_READ_4(sc, BGE_RXLP_LOCSTAT_OUT_OF_BDS);
Disable updating InputDiscards counter for BCM5717, BCM5718, BCM5719 A0 and BCM5720 A0 and add comment why driver does not try to read it.
2011-10-28 01:10:59 +00:00
/*
* XXX
* Unlike other controllers, BGE_RXLP_LOCSTAT_IFIN_DROPS
* counter of BCM5717, BCM5718, BCM5719 A0 and BCM5720 A0
* includes number of unwanted multicast frames. This comes
* from silicon bug and known workaround to get rough(not
* exact) counter is to enable interrupt on MBUF low water
* attention. This can be accomplished by setting
* BGE_HCCMODE_ATTN bit of BGE_HCC_MODE,
* BGE_BMANMODE_LOMBUF_ATTN bit of BGE_BMAN_MODE and
* BGE_MODECTL_FLOWCTL_ATTN_INTR bit of BGE_MODE_CTL.
* However that change would generate more interrupts and
* there are still possibilities of losing multiple frames
* during BGE_MODECTL_FLOWCTL_ATTN_INTR interrupt handling.
* Given that the workaround still would not get correct
* counter I don't think it's worth to implement it. So
* ignore reading the counter on controllers that have the
* silicon bug.
*/
if (sc->bge_asicrev != BGE_ASICREV_BCM5717 &&
sc->bge_chipid != BGE_CHIPID_BCM5719_A0 &&
sc->bge_chipid != BGE_CHIPID_BCM5720_A0)
stats->InputDiscards +=
CSR_READ_4(sc, BGE_RXLP_LOCSTAT_IFIN_DROPS);
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
stats->InputErrors +=
CSR_READ_4(sc, BGE_RXLP_LOCSTAT_IFIN_ERRORS);
stats->RecvThresholdHit +=
CSR_READ_4(sc, BGE_RXLP_LOCSTAT_RXTHRESH_HIT);
ifp->if_collisions = (u_long)stats->etherStatsCollisions;
ifp->if_ierrors = (u_long)(stats->NoMoreRxBDs + stats->InputDiscards +
stats->InputErrors);
}
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
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
static void
bge_stats_clear_regs(struct bge_softc *sc)
{
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
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
CSR_READ_4(sc, BGE_TX_MAC_STATS_OCTETS);
CSR_READ_4(sc, BGE_TX_MAC_STATS_COLLS);
CSR_READ_4(sc, BGE_TX_MAC_STATS_XON_SENT);
CSR_READ_4(sc, BGE_TX_MAC_STATS_XOFF_SENT);
CSR_READ_4(sc, BGE_TX_MAC_STATS_ERRORS);
CSR_READ_4(sc, BGE_TX_MAC_STATS_SINGLE_COLL);
CSR_READ_4(sc, BGE_TX_MAC_STATS_MULTI_COLL);
CSR_READ_4(sc, BGE_TX_MAC_STATS_DEFERRED);
CSR_READ_4(sc, BGE_TX_MAC_STATS_EXCESS_COLL);
CSR_READ_4(sc, BGE_TX_MAC_STATS_LATE_COLL);
CSR_READ_4(sc, BGE_TX_MAC_STATS_UCAST);
CSR_READ_4(sc, BGE_TX_MAC_STATS_MCAST);
CSR_READ_4(sc, BGE_TX_MAC_STATS_BCAST);
CSR_READ_4(sc, BGE_RX_MAC_STATS_OCTESTS);
CSR_READ_4(sc, BGE_RX_MAC_STATS_FRAGMENTS);
CSR_READ_4(sc, BGE_RX_MAC_STATS_UCAST);
CSR_READ_4(sc, BGE_RX_MAC_STATS_MCAST);
CSR_READ_4(sc, BGE_RX_MAC_STATS_BCAST);
CSR_READ_4(sc, BGE_RX_MAC_STATS_FCS_ERRORS);
CSR_READ_4(sc, BGE_RX_MAC_STATS_ALGIN_ERRORS);
CSR_READ_4(sc, BGE_RX_MAC_STATS_XON_RCVD);
CSR_READ_4(sc, BGE_RX_MAC_STATS_XOFF_RCVD);
CSR_READ_4(sc, BGE_RX_MAC_STATS_CTRL_RCVD);
CSR_READ_4(sc, BGE_RX_MAC_STATS_XOFF_ENTERED);
CSR_READ_4(sc, BGE_RX_MAC_STATS_FRAME_TOO_LONG);
CSR_READ_4(sc, BGE_RX_MAC_STATS_JABBERS);
CSR_READ_4(sc, BGE_RX_MAC_STATS_UNDERSIZE);
CSR_READ_4(sc, BGE_RXLP_LOCSTAT_FILTDROP);
CSR_READ_4(sc, BGE_RXLP_LOCSTAT_DMA_WRQ_FULL);
CSR_READ_4(sc, BGE_RXLP_LOCSTAT_DMA_HPWRQ_FULL);
CSR_READ_4(sc, BGE_RXLP_LOCSTAT_OUT_OF_BDS);
CSR_READ_4(sc, BGE_RXLP_LOCSTAT_IFIN_DROPS);
CSR_READ_4(sc, BGE_RXLP_LOCSTAT_IFIN_ERRORS);
CSR_READ_4(sc, BGE_RXLP_LOCSTAT_RXTHRESH_HIT);
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;
Make if_ierrors updated whenever any of the following counters are updated. o Number of times NIC ran out of RX buffer descriptors o Number of inbound packet errors o Number of inbound packets that were chosen to be discarded Previously only the discarded packet counter was used to update if_ierrors. This change fixes wrong if_ierrors counter on BCM570[0-4] controllers. For BCM5705 and later controllers bge(4) already correctly counted it. Reported by: Eugene Grosbein <egrosbein <> rdtc dot ru>
2012-03-12 03:47:30 +00:00
cnt = READ_STAT(sc, stats, nicNoMoreRxBDs.bge_addr_lo);
ifp->if_ierrors += (uint32_t)(cnt - sc->bge_rx_nobds);
sc->bge_rx_nobds = cnt;
cnt = READ_STAT(sc, stats, ifInErrors.bge_addr_lo);
ifp->if_ierrors += (uint32_t)(cnt - sc->bge_rx_inerrs);
sc->bge_rx_inerrs = cnt;
- Count packets discarded by RX/TX MAC (cause of FIFO overflow, etc) as input/output interface errors. - Keep values of rx/tx discards & tx collisions inside struct bge_softc. So we can keep statistic across ifconfig down/up runs (cause bringing bge up will reset chip). Approved by: glebius (mentor) MFC after: 1 week
2006-01-17 23:01:58 +00:00
cnt = READ_STAT(sc, stats, 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;
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
MGET(n, M_NOWAIT, MT_DATA);
- 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 (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);
}
Add workaround for BCM5906 controller silicon bug. If device receive two back-to-back send BDs with less than or equal to 8 total bytes then the device may hang. The two back-to-back send BDs must be in the same frame for this failure to occur. Thanks to davidch for detailed errata information. Reviewed by: davidch
2010-10-19 23:04:23 +00:00
static struct mbuf *
bge_check_short_dma(struct mbuf *m)
{
struct mbuf *n;
int found;
/*
* If device receive two back-to-back send BDs with less than
* or equal to 8 total bytes then the device may hang. The two
* back-to-back send BDs must in the same frame for this failure
* to occur. Scan mbuf chains and see whether two back-to-back
* send BDs are there. If this is the case, allocate new mbuf
* and copy the frame to workaround the silicon bug.
*/
for (n = m, found = 0; n != NULL; n = n->m_next) {
if (n->m_len < 8) {
found++;
if (found > 1)
break;
continue;
}
found = 0;
}
if (found > 1) {
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
n = m_defrag(m, M_NOWAIT);
Add workaround for BCM5906 controller silicon bug. If device receive two back-to-back send BDs with less than or equal to 8 total bytes then the device may hang. The two back-to-back send BDs must be in the same frame for this failure to occur. Thanks to davidch for detailed errata information. Reviewed by: davidch
2010-10-19 23:04:23 +00:00
if (n == NULL)
m_freem(m);
} else
n = m;
return (n);
}
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 *
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
bge_setup_tso(struct bge_softc *sc, struct mbuf *m, uint16_t *mss,
uint16_t *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
{
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. */
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
n = m_dup(m, M_NOWAIT);
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_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;
Make sure to not use stale ip/tcp header pointers. The ip/tcp header parser uses m_pullup(9) to get access to mbuf chain. m_pullup(9) can allocate new mbuf chain and free old one if the space left in the mbuf chain is not enough to hold requested contiguous bytes. Previously drivers can use stale ip/tcp header pointer if m_pullup(9) returned new mbuf chain. Reported by: Andrew Boyer (aboyer <> averesystems dot com) MFC after: 10 days
2010-10-14 18:31:40 +00:00
ip = (struct ip *)(mtod(m, char *) + sizeof(struct ether_header));
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
ip->ip_sum = 0;
ip->ip_len = htons(*mss + (ip->ip_hl << 2) + (tcp->th_off << 2));
/* Clear pseudo checksum computed by TCP stack. */
Make sure to not use stale ip/tcp header pointers. The ip/tcp header parser uses m_pullup(9) to get access to mbuf chain. m_pullup(9) can allocate new mbuf chain and free old one if the space left in the mbuf chain is not enough to hold requested contiguous bytes. Previously drivers can use stale ip/tcp header pointer if m_pullup(9) returned new mbuf chain. Reported by: Andrew Boyer (aboyer <> averesystems dot com) MFC after: 10 days
2010-10-14 18:31:40 +00:00
tcp = (struct tcphdr *)(mtod(m, char *) + 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
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
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
* we only support hardware based 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
*/
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
/* Calculate header length, incl. TCP/IP options, in 32 bit units. */
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
hlen = ((ip->ip_hl << 2) + (tcp->th_off << 2)) >> 2;
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if (sc->bge_flags & BGE_FLAG_TSO3) {
/*
* For BCM5717 and newer controllers, hardware based TSO
* uses the 14 lower bits of the bge_mss field to store the
* MSS and the upper 2 bits to store the lowest 2 bits of
* the IP/TCP header length. The upper 6 bits of the header
* length are stored in the bge_flags[14:10,4] field. Jumbo
* frames are supported.
*/
*mss |= ((hlen & 0x3) << 14);
*flags |= ((hlen & 0xF8) << 7) | ((hlen & 0x4) << 2);
} else {
/*
* For BCM5755 and newer controllers, hardware based TSO uses
* the lower 11 bits to store the MSS and the upper 5 bits to
* store the IP/TCP header length. Jumbo frames are not
* supported.
*/
*mss |= (hlen << 11);
}
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
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;
Add workaround for BCM5906 controller silicon bug. If device receive two back-to-back send BDs with less than or equal to 8 total bytes then the device may hang. The two back-to-back send BDs must be in the same frame for this failure to occur. Thanks to davidch for detailed errata information. Reviewed by: davidch
2010-10-19 23:04:23 +00:00
if ((sc->bge_flags & BGE_FLAG_SHORT_DMA_BUG) != 0 &&
m->m_next != NULL) {
*m_head = bge_check_short_dma(m);
if (*m_head == NULL)
return (ENOBUFS);
m = *m_head;
}
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->m_pkthdr.csum_flags & CSUM_TSO) != 0) {
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
*m_head = m = bge_setup_tso(sc, m, &mss, &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
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);
}
}
}
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
if ((m->m_pkthdr.csum_flags & CSUM_TSO) == 0) {
if (sc->bge_flags & BGE_FLAG_JUMBO_FRAME &&
m->m_pkthdr.len > ETHER_MAX_LEN)
csum_flags |= BGE_TXBDFLAG_JUMBO_FRAME;
if (sc->bge_forced_collapse > 0 &&
(sc->bge_flags & BGE_FLAG_PCIE) != 0 && m->m_next != NULL) {
/*
* Forcedly collapse mbuf chains to overcome hardware
* limitation which only support a single outstanding
* DMA read operation.
*/
if (sc->bge_forced_collapse == 1)
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
m = m_defrag(m, M_NOWAIT);
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
else
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
m = m_collapse(m, M_NOWAIT,
Add initial BCM5718 family support. The BCM5718 family includes the dual port BCM5717 and BCM5718 devices which are intended for mainstream workstation and entry-level server designs and represents the twelfth generation of NetXtreme Ethernet controllers. This family is the successor to the BCM5714/BCM5715 family and supports IPv4/IPv6 checksum offloading, TSO, VLAN hardware tagging, jumbo frames, MSI/MSIX, IOV, RSS and TSS. This change set supports all hardware features except IOV and RSS/TSS. Unlike its predecessors, only extended RX buffer descriptors can be posted to the jumbo producer ring. Single RX buffer descriptors for jumbo frame are not supported. RSS requires a more substantial set of changes and will apply to a larger set of NetXtreme devices so RSS/TSS multi-queue support will be implemented in a future releases. Special thanks to Broadcom who kindly sent a sample board to me and to davidch who gave provided the initial support code. Submitted by: davidch (initial version) HW donated by: Broadcom
2010-10-27 17:20:19 +00:00
sc->bge_forced_collapse);
if (m == NULL)
m = *m_head;
*m_head = m;
}
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
}
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) {
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
m = m_collapse(m, M_NOWAIT, 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 (m->m_flags & M_VLANTAG) {
csum_flags |= BGE_TXBDFLAG_VLAN_TAG;
vlan_tag = m->m_pkthdr.ether_vtag;
}
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
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;
/*
* 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;
Enable TX MAC state machine lockup fix for both BCM5755 or higher and BCM5906. Publicly available data sheet just says it may happen due to corrupted TxMbuf.
2010-10-22 18:31:44 +00:00
uint32_t mode;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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. */
Add initial jumbo frame support for BCM5714/BCM5715 and BCM5780. Unlike other controllers which have more advanced jumbo support, these controllers have one send ring, one standard receive producer ring and one receive return ring. In order to receive jumbo frames on the controllers, driver now will increase Rx buffer size to 9k. Two Rx modes are supported on these controllers and I chose standard Rx BDs over extended Rx BDs. The extended Rx BD mode allows up to 4 segmentations for each Rx BDs such that kernel does not have to allocate large buffer of contiguous memory for receiving. The extended Rx BD mode is already used on controllers that have separate jumbo receive ring. However, using extended Rx BDs on BCM5714/BCM5715/BCM5780 reduces the number of Rx BDs to 256 entries which in turn may reduce the performance. Also UMA backed page allocator for jumbo frame returns contiguous memory so using extended Rx BD has no advantage on FreeBSD unless highly customized local allocator implemented in driver is used. To use jumbo buffers in standard receive ring, Rx buffer allocation handler was changed to allocate MJUM9BYTES sized mbuf. PR: kern/155192 Tested by: Vijay Singh <vijju.singh <> gmail dot com> Submitted by: mjacob (initial version)
2011-04-05 17:41:54 +00:00
if (BGE_IS_JUMBO_CAPABLE(sc) &&
ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN >
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
(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);
Enable TX MAC state machine lockup fix for both BCM5755 or higher and BCM5906. Publicly available data sheet just says it may happen due to corrupted TxMbuf.
2010-10-22 18:31:44 +00:00
/* Enable TX MAC state machine lockup fix. */
mode = CSR_READ_4(sc, BGE_TX_MODE);
if (BGE_IS_5755_PLUS(sc) || sc->bge_asicrev == BGE_ASICREV_BCM5906)
mode |= BGE_TXMODE_MBUF_LOCKUP_FIX;
Add initial BCM5720 support. Many thanks to Broadcom for continuing support of FreeBSD. Submitted by: Geans Pin at Broadcom (initial version) H/W donated by: Broadcom
2011-10-28 01:04:40 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5720) {
mode &= ~(BGE_TXMODE_JMB_FRM_LEN | BGE_TXMODE_CNT_DN_MODE);
mode |= CSR_READ_4(sc, BGE_TX_MODE) &
(BGE_TXMODE_JMB_FRM_LEN | BGE_TXMODE_CNT_DN_MODE);
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Turn on transmitter. */
Enable TX MAC state machine lockup fix for both BCM5755 or higher and BCM5906. Publicly available data sheet just says it may happen due to corrupted TxMbuf.
2010-10-22 18:31:44 +00:00
CSR_WRITE_4(sc, BGE_TX_MODE, mode | BGE_TXMODE_ENABLE);
Follow Broadcom datasheet: Delay 100 microseconds after enabling transmit MAC. Delay 10 microseconds after enabling receive MAC.
2012-10-05 07:13:21 +00:00
DELAY(100);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
/* Turn on receiver. */
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
mode = CSR_READ_4(sc, BGE_RX_MODE);
if (BGE_IS_5755_PLUS(sc))
mode |= BGE_RXMODE_IPV6_ENABLE;
CSR_WRITE_4(sc,BGE_RX_MODE, mode | BGE_RXMODE_ENABLE);
Follow Broadcom datasheet: Delay 100 microseconds after enabling transmit MAC. Delay 10 microseconds after enabling receive MAC.
2012-10-05 07:13:21 +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
Set the number of RX frames to receive after RX MBUF low watermark has reached. This reduced number of dropped frames when flow-control is enabled. Previously it dropped incoming frames once RX MBUF low watermark has reached. The value used in MAC RX MBUF low watermark is greater than or equal to 4 so receiving two more RX frames should not be a problem. Obtained from: OpenBSD
2010-09-29 00:00:45 +00:00
/*
* Set the number of good frames to receive after RX MBUF
* Low Watermark has been reached. After the RX MAC receives
* this number of frames, it will drop subsequent incoming
* frames until the MBUF High Watermark is reached.
*/
For 57765 class controllers, set low watermark max receive frames to 1.
2013-01-24 02:25:43 +00:00
if (BGE_IS_57765_PLUS(sc))
Add initial BCM57765 family support. The BCM57765 family seems to have similar hardware features of BCM5718 family except the number of receive return ring is 4. The BCM57765 family is known to support IEEE 802.3az EEE(Energy Efficient Ethernet) but this change does not include EEE support code. I hope EEE is implemented in near future. This change will support BCM57761, BCM57765, BCM57781, BCM57785, BCM57791 and BCM57795. All hardware offloading features are supported and suspend/resume also should work. Many thanks to Broadcom for continuing support of FreeBSD. Tested by: Paul Thornton (prt <> prt dot org) HW donated by: Broadcom
2011-05-04 17:04:31 +00:00
CSR_WRITE_4(sc, BGE_MAX_RX_FRAME_LOWAT, 1);
else
CSR_WRITE_4(sc, BGE_MAX_RX_FRAME_LOWAT, 2);
Set the number of RX frames to receive after RX MBUF low watermark has reached. This reduced number of dropped frames when flow-control is enabled. Previously it dropped incoming frames once RX MBUF low watermark has reached. The value used in MAC RX MBUF low watermark is greater than or equal to 4 so receiving two more RX frames should not be a problem. Obtained from: OpenBSD
2010-09-29 00:00:45 +00:00
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
/* Clear MAC statistics. */
if (BGE_IS_5705_PLUS(sc))
bge_stats_clear_regs(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
/* 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
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
Don't touch EMAC Mode and TX/RX MAC Mode register when driver is not running.
2012-10-05 03:35:38 +00:00
bge_ifmedia_upd_locked(ifp);
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);
}
Add 40 microseconds delay after updating EMAC Mode register as recommended by Broadcom data sheet.
2012-10-05 06:24:22 +00:00
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
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);
- Remove attempts to implement setting of BMCR_LOOP/MIIF_NOLOOP (reporting IFM_LOOP based on BMCR_LOOP is left in place though as it might provide useful for debugging). For most mii(4) drivers it was unclear whether the PHYs driven by them actually support loopback or not. Moreover, typically loopback mode also needs to be activated on the MAC, which none of the Ethernet drivers using mii(4) implements. Given that loopback media has no real use (and obviously hardly had a chance to actually work) besides for driver development (which just loopback mode should be sufficient for though, i.e one doesn't necessary need support for loopback media) support for it is just dropped as both NetBSD and OpenBSD already did quite some time ago. - Let mii_phy_add_media() also announce the support of IFM_NONE. - Restructure the PHY entry points to use a structure of entry points instead of discrete function pointers, and extend this to include a "reset" entry point. Make sure any PHY-specific reset routine is always used, and provide one for lxtphy(4) which disables MII interrupts (as is done for a few other PHYs we have drivers for). This includes changing NIC drivers which previously just called the generic mii_phy_reset() to now actually call the PHY-specific reset routine, which might be crucial in some cases. While at it, the redundant checks in these NIC drivers for mii->mii_instance not being zero before calling the reset routines were removed because as soon as one PHY driver attaches mii->mii_instance is incremented and we hardly can end up in their media change callbacks etc if no PHY driver has attached as mii_attach() would have failed in that case and not attach a miibus(4) instance. Consequently, NIC drivers now no longer should call mii_phy_reset() directly, so it was removed from EXPORT_SYMS. - Add a mii_phy_dev_attach() as a companion helper to mii_phy_dev_probe(). The purpose of that function is to perform the common steps to attach a PHY driver instance and to hook it up to the miibus(4) instance and to optionally also handle the probing, addition and initialization of the supported media. So all a PHY driver without any special requirements has to do in its bus attach method is to call mii_phy_dev_attach() along with PHY-specific MIIF_* flags, a pointer to its PHY functions and the add_media set to one. All PHY drivers were updated to take advantage of mii_phy_dev_attach() as appropriate. Along with these changes the capability mask was added to the mii_softc structure so PHY drivers taking advantage of mii_phy_dev_attach() but still handling media on their own do not need to fiddle with the MII attach arguments anyway. - Keep track of the PHY offset in the mii_softc structure. This is done for compatibility with NetBSD/OpenBSD. - Keep track of the PHY's OUI, model and revision in the mii_softc structure. Several PHY drivers require this information also after attaching and previously had to wrap their own softc around mii_softc. NetBSD/OpenBSD also keep track of the model and revision on their mii_softc structure. All PHY drivers were updated to take advantage as appropriate. - Convert the mebers of the MII data structure to unsigned where appropriate. This is partly inspired by NetBSD/OpenBSD. - According to IEEE 802.3-2002 the bits actually have to be reversed when mapping an OUI to the MII ID registers. All PHY drivers and miidevs where changed as necessary. Actually this now again allows to largely share miidevs with NetBSD, which fixed this problem already 9 years ago. Consequently miidevs was synced as far as possible. - Add MIIF_NOMANPAUSE and mii_phy_flowstatus() calls to drivers that weren't explicitly converted to support flow control before. It's unclear whether flow control actually works with these but typically it should and their net behavior should be more correct with these changes in place than without if the MAC driver sets MIIF_DOPAUSE. Obtained from: NetBSD (partially) Reviewed by: yongari (earlier version), silence on arch@ and net@
2011-05-03 19:51:29 +00:00
LIST_FOREACH(miisc, &mii->mii_phys, mii_list)
PHY_RESET(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
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:
Add initial jumbo frame support for BCM5714/BCM5715 and BCM5780. Unlike other controllers which have more advanced jumbo support, these controllers have one send ring, one standard receive producer ring and one receive return ring. In order to receive jumbo frames on the controllers, driver now will increase Rx buffer size to 9k. Two Rx modes are supported on these controllers and I chose standard Rx BDs over extended Rx BDs. The extended Rx BD mode allows up to 4 segmentations for each Rx BDs such that kernel does not have to allocate large buffer of contiguous memory for receiving. The extended Rx BD mode is already used on controllers that have separate jumbo receive ring. However, using extended Rx BDs on BCM5714/BCM5715/BCM5780 reduces the number of Rx BDs to 256 entries which in turn may reduce the performance. Also UMA backed page allocator for jumbo frame returns contiguous memory so using extended Rx BD has no advantage on FreeBSD unless highly customized local allocator implemented in driver is used. To use jumbo buffers in standard receive ring, Rx buffer allocation handler was changed to allocate MJUM9BYTES sized mbuf. PR: kern/155192 Tested by: Vijay Singh <vijju.singh <> gmail dot com> Submitted by: mjacob (initial version)
2011-04-05 17:41:54 +00:00
if (BGE_IS_JUMBO_CAPABLE(sc) ||
(sc->bge_flags & BGE_FLAG_JUMBO_STD)) {
if (ifr->ifr_mtu < ETHERMIN ||
ifr->ifr_mtu > BGE_JUMBO_MTU) {
error = EINVAL;
break;
}
} else if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > ETHERMTU) {
Whitespace nits.
2011-10-25 20:45:14 +00:00
error = EINVAL;
Add initial jumbo frame support for BCM5714/BCM5715 and BCM5780. Unlike other controllers which have more advanced jumbo support, these controllers have one send ring, one standard receive producer ring and one receive return ring. In order to receive jumbo frames on the controllers, driver now will increase Rx buffer size to 9k. Two Rx modes are supported on these controllers and I chose standard Rx BDs over extended Rx BDs. The extended Rx BD mode allows up to 4 segmentations for each Rx BDs such that kernel does not have to allocate large buffer of contiguous memory for receiving. The extended Rx BD mode is already used on controllers that have separate jumbo receive ring. However, using extended Rx BDs on BCM5714/BCM5715/BCM5780 reduces the number of Rx BDs to 256 entries which in turn may reduce the performance. Also UMA backed page allocator for jumbo frame returns contiguous memory so using extended Rx BD has no advantage on FreeBSD unless highly customized local allocator implemented in driver is used. To use jumbo buffers in standard receive ring, Rx buffer allocation handler was changed to allocate MJUM9BYTES sized mbuf. PR: kern/155192 Tested by: Vijay Singh <vijju.singh <> gmail dot com> Submitted by: mjacob (initial version)
2011-04-05 17:41:54 +00:00
break;
}
Do not blindly UP the interface when interface's MTU is changed. If driver is not running there is no need to up the interface. While I'm here hold driver lock before modifying MTU as it is referenced in RX handler.
2010-10-08 17:58:07 +00:00
BGE_LOCK(sc);
Add initial jumbo frame support for BCM5714/BCM5715 and BCM5780. Unlike other controllers which have more advanced jumbo support, these controllers have one send ring, one standard receive producer ring and one receive return ring. In order to receive jumbo frames on the controllers, driver now will increase Rx buffer size to 9k. Two Rx modes are supported on these controllers and I chose standard Rx BDs over extended Rx BDs. The extended Rx BD mode allows up to 4 segmentations for each Rx BDs such that kernel does not have to allocate large buffer of contiguous memory for receiving. The extended Rx BD mode is already used on controllers that have separate jumbo receive ring. However, using extended Rx BDs on BCM5714/BCM5715/BCM5780 reduces the number of Rx BDs to 256 entries which in turn may reduce the performance. Also UMA backed page allocator for jumbo frame returns contiguous memory so using extended Rx BD has no advantage on FreeBSD unless highly customized local allocator implemented in driver is used. To use jumbo buffers in standard receive ring, Rx buffer allocation handler was changed to allocate MJUM9BYTES sized mbuf. PR: kern/155192 Tested by: Vijay Singh <vijju.singh <> gmail dot com> Submitted by: mjacob (initial version)
2011-04-05 17:41:54 +00:00
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;
Do not blindly UP the interface when interface's MTU is changed. If driver is not running there is no need to up the interface. While I'm here hold driver lock before modifying MTU as it is referenced in RX handler.
2010-10-08 17:58:07 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
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
}
Do not blindly UP the interface when interface's MTU is changed. If driver is not running there is no need to up the interface. While I'm here hold driver lock before modifying MTU as it is referenced in RX handler.
2010-10-08 17:58:07 +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
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
Fix IFCAP_TXCSUM/IFCAP_RXCSUM handling. Previously bge(4) used IFCAP_HWCSUM to know which capability should be changed such that disabling RX checksun offloading resulted in disabling TX checksum offloading.
2010-09-30 22:34:15 +00:00
if ((mask & IFCAP_TXCSUM) != 0 &&
(ifp->if_capabilities & IFCAP_TXCSUM) != 0) {
ifp->if_capenable ^= IFCAP_TXCSUM;
if ((ifp->if_capenable & IFCAP_TXCSUM) != 0)
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
Fix IFCAP_TXCSUM/IFCAP_RXCSUM handling. Previously bge(4) used IFCAP_HWCSUM to know which capability should be changed such that disabling RX checksun offloading resulted in disabling TX checksum offloading.
2010-09-30 22:34:15 +00:00
if ((mask & IFCAP_RXCSUM) != 0 &&
(ifp->if_capabilities & IFCAP_RXCSUM) != 0)
ifp->if_capenable ^= IFCAP_RXCSUM;
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++;
}
Since r117657, bge(4) does not enable buffer manager for BCM5705 or newer controllers. However, all data sheet I have access has no indication that buffer manager should not be touched on these controllers. It seems the buffer manager always runs on BCM5705 or newer controllers. Some controller(e.g. BCM5719) needs other buffer manager configuration so driver should enable buffer manager for all controllers. Both Linux and OpenBSD/NetBSD use the same approach. This change polls enable bit of block to know whether specified block was really stopped as well as enabling buffer manager for all controllers in driver initialization. Obtained from: NetBSD
2011-05-09 20:10:46 +00:00
static void
bge_stop_block(struct bge_softc *sc, bus_size_t reg, uint32_t bit)
{
int i;
BGE_CLRBIT(sc, reg, bit);
for (i = 0; i < BGE_TIMEOUT; i++) {
if ((CSR_READ_4(sc, reg) & bit) == 0)
return;
DELAY(100);
}
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 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);
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
bge_sig_pre_reset(sc, BGE_RESET_SHUTDOWN);
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
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
*/
Since r117657, bge(4) does not enable buffer manager for BCM5705 or newer controllers. However, all data sheet I have access has no indication that buffer manager should not be touched on these controllers. It seems the buffer manager always runs on BCM5705 or newer controllers. Some controller(e.g. BCM5719) needs other buffer manager configuration so driver should enable buffer manager for all controllers. Both Linux and OpenBSD/NetBSD use the same approach. This change polls enable bit of block to know whether specified block was really stopped as well as enabling buffer manager for all controllers in driver initialization. Obtained from: NetBSD
2011-05-09 20:10:46 +00:00
bge_stop_block(sc, BGE_RX_MODE, BGE_RXMODE_ENABLE);
bge_stop_block(sc, BGE_RBDI_MODE, BGE_RBDIMODE_ENABLE);
bge_stop_block(sc, BGE_RXLP_MODE, BGE_RXLPMODE_ENABLE);
if (BGE_IS_5700_FAMILY(sc))
bge_stop_block(sc, BGE_RXLS_MODE, BGE_RXLSMODE_ENABLE);
bge_stop_block(sc, BGE_RDBDI_MODE, BGE_RBDIMODE_ENABLE);
bge_stop_block(sc, BGE_RDC_MODE, BGE_RDCMODE_ENABLE);
bge_stop_block(sc, BGE_RBDC_MODE, BGE_RBDCMODE_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
/*
- 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
*/
Since r117657, bge(4) does not enable buffer manager for BCM5705 or newer controllers. However, all data sheet I have access has no indication that buffer manager should not be touched on these controllers. It seems the buffer manager always runs on BCM5705 or newer controllers. Some controller(e.g. BCM5719) needs other buffer manager configuration so driver should enable buffer manager for all controllers. Both Linux and OpenBSD/NetBSD use the same approach. This change polls enable bit of block to know whether specified block was really stopped as well as enabling buffer manager for all controllers in driver initialization. Obtained from: NetBSD
2011-05-09 20:10:46 +00:00
bge_stop_block(sc, BGE_SRS_MODE, BGE_SRSMODE_ENABLE);
bge_stop_block(sc, BGE_SBDI_MODE, BGE_SBDIMODE_ENABLE);
bge_stop_block(sc, BGE_SDI_MODE, BGE_SDIMODE_ENABLE);
bge_stop_block(sc, BGE_RDMA_MODE, BGE_RDMAMODE_ENABLE);
bge_stop_block(sc, BGE_SDC_MODE, BGE_SDCMODE_ENABLE);
if (BGE_IS_5700_FAMILY(sc))
bge_stop_block(sc, BGE_DMAC_MODE, BGE_DMACMODE_ENABLE);
bge_stop_block(sc, BGE_SBDC_MODE, BGE_SBDCMODE_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
/*
* Shut down all of the memory managers and related
* state machines.
*/
Since r117657, bge(4) does not enable buffer manager for BCM5705 or newer controllers. However, all data sheet I have access has no indication that buffer manager should not be touched on these controllers. It seems the buffer manager always runs on BCM5705 or newer controllers. Some controller(e.g. BCM5719) needs other buffer manager configuration so driver should enable buffer manager for all controllers. Both Linux and OpenBSD/NetBSD use the same approach. This change polls enable bit of block to know whether specified block was really stopped as well as enabling buffer manager for all controllers in driver initialization. Obtained from: NetBSD
2011-05-09 20:10:46 +00:00
bge_stop_block(sc, BGE_HCC_MODE, BGE_HCCMODE_ENABLE);
bge_stop_block(sc, BGE_WDMA_MODE, BGE_WDMAMODE_ENABLE);
if (BGE_IS_5700_FAMILY(sc))
bge_stop_block(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);
}
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
/* Update MAC statistics. */
if (BGE_IS_5705_PLUS(sc))
bge_stats_update_regs(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
bge_reset(sc);
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
bge_sig_legacy(sc, BGE_RESET_SHUTDOWN);
bge_sig_post_reset(sc, BGE_RESET_SHUTDOWN);
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
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);
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);
Do not hardcode phy address. Multi-port controllers use different phy address.
2012-10-24 05:00:56 +00:00
bge_miibus_readreg(sc->bge_dev, sc->bge_phy_addr,
BRGPHY_MII_ISR);
bge_miibus_writereg(sc->bge_dev, sc->bge_phy_addr,
BRGPHY_MII_IMR, BRGPHY_INTRS);
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
}
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++;
Add 40 microseconds delay after updating EMAC Mode register as recommended by Broadcom data sheet.
2012-10-05 06:24:22 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5704) {
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_CLRBIT(sc, BGE_MAC_MODE,
BGE_MACMODE_TBI_SEND_CFGS);
Add 40 microseconds delay after updating EMAC Mode register as recommended by Broadcom data sheet.
2012-10-05 06:24:22 +00:00
DELAY(40);
}
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
}
Remove one last reference of BGE_MI_MODE register for auto polling. Previously bge(4) always enabled auto polling for non-BGE_FLAG_TBI controllers. With this change, auto polling is not used anymore so polling through mii(4) was introduced. Reviewed by: davidch
2010-10-11 22:56:23 +00:00
} else if ((sc->bge_mi_mode & BGE_MIMODE_AUTOPOLL) != 0) {
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 {
/*
Remove one last reference of BGE_MI_MODE register for auto polling. Previously bge(4) always enabled auto polling for non-BGE_FLAG_TBI controllers. With this change, auto polling is not used anymore so polling through mii(4) was introduced. Reviewed by: davidch
2010-10-11 22:56:23 +00:00
* For controllers that call mii_tick, we have to poll
* link status.
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
*/
Remove one last reference of BGE_MI_MODE register for auto polling. Previously bge(4) always enabled auto polling for non-BGE_FLAG_TBI controllers. With this change, auto polling is not used anymore so polling through mii(4) was introduced. Reviewed by: davidch
2010-10-11 22:56:23 +00:00
mii = device_get_softc(sc->bge_miibus);
mii_pollstat(mii);
Add more checks for resolved link speed in bge_miibus_statchg(). Link UP state could be reported first before actual completion of auto-negotiation. This change makes bge(4) reprogram BGE_MAC_MODE, BGE_TX_MODE and BGE_RX_MODE register only after controller got a valid link.
2010-10-13 21:53:37 +00:00
bge_miibus_statchg(sc->bge_dev);
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
}
On PHY write error use hex number to show the value. Add more comments.
2012-10-10 01:59:53 +00:00
/* Disable MAC attention when link is up. */
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
static void
bge_add_sysctls(struct bge_softc *sc)
{
struct sysctl_ctx_list *ctx;
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
struct sysctl_oid_list *children;
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",
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
"MAC Register Read");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "ape_read",
CTLTYPE_INT | CTLFLAG_RW, sc, 0, bge_sysctl_ape_read, "I",
"APE Register Read");
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
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");
Rename dev.bge.%d.msi_disable to dev.bge.%d.msi which matches enable/disable and default it to on. Suggested by: jhb
2012-01-19 20:21:59 +00:00
sc->bge_msi = 1;
snprintf(tn, sizeof(tn), "dev.bge.%d.msi", unit);
TUNABLE_INT_FETCH(tn, &sc->bge_msi);
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "msi",
CTLFLAG_RD, &sc->bge_msi, 0, "Enable MSI");
Introduce a tunable that disables use of MSI. Non-zero value will use INTx.
2012-01-17 22:15:33 +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
/*
* 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))
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
bge_add_sysctl_stats_regs(sc, ctx, children);
else
bge_add_sysctl_stats(sc, ctx, children);
}
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
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
#define BGE_SYSCTL_STAT(sc, ctx, desc, parent, node, oid) \
SYSCTL_ADD_PROC(ctx, parent, OID_AUTO, oid, CTLTYPE_UINT|CTLFLAG_RD, \
sc, offsetof(struct bge_stats, node), bge_sysctl_stats, "IU", \
desc)
static void
bge_add_sysctl_stats(struct bge_softc *sc, struct sysctl_ctx_list *ctx,
struct sysctl_oid_list *parent)
{
struct sysctl_oid *tree;
struct sysctl_oid_list *children, *schildren;
tree = SYSCTL_ADD_NODE(ctx, parent, OID_AUTO, "stats", CTLFLAG_RD,
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
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",
Consistently use ifHCOutOctets/ifHCInOctets instead of Octets as these names are used in data sheet. Also use UnicastPkts, MulticastPkts and BroadcastPkts instead of UcastPkts, McastPkts and BcastPkts to clarify its meaning. Suggested by: bde
2010-10-04 18:01:23 +00:00
children, rxstats.ifHCInOctets, "ifHCInOctets");
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
BGE_SYSCTL_STAT(sc, ctx, "Fragments",
children, rxstats.etherStatsFragments, "Fragments");
BGE_SYSCTL_STAT(sc, ctx, "Inbound Unicast Packets",
Consistently use ifHCOutOctets/ifHCInOctets instead of Octets as these names are used in data sheet. Also use UnicastPkts, MulticastPkts and BroadcastPkts instead of UcastPkts, McastPkts and BcastPkts to clarify its meaning. Suggested by: bde
2010-10-04 18:01:23 +00:00
children, rxstats.ifHCInUcastPkts, "UnicastPkts");
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
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",
Consistently use ifHCOutOctets/ifHCInOctets instead of Octets as these names are used in data sheet. Also use UnicastPkts, MulticastPkts and BroadcastPkts instead of UcastPkts, McastPkts and BcastPkts to clarify its meaning. Suggested by: bde
2010-10-04 18:01:23 +00:00
children, txstats.ifHCOutOctets, "ifHCOutOctets");
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
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",
Consistently use ifHCOutOctets/ifHCInOctets instead of Octets as these names are used in data sheet. Also use UnicastPkts, MulticastPkts and BroadcastPkts instead of UcastPkts, McastPkts and BcastPkts to clarify its meaning. Suggested by: bde
2010-10-04 18:01:23 +00:00
children, txstats.ifHCOutUcastPkts, "UnicastPkts");
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
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");
}
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
#undef BGE_SYSCTL_STAT
#define BGE_SYSCTL_STAT_ADD64(c, h, n, p, d) \
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the rest of the devices.
2011-01-12 19:53:56 +00:00
SYSCTL_ADD_UQUAD(c, h, OID_AUTO, n, CTLFLAG_RD, p, d)
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
static void
bge_add_sysctl_stats_regs(struct bge_softc *sc, struct sysctl_ctx_list *ctx,
struct sysctl_oid_list *parent)
{
struct sysctl_oid *tree;
struct sysctl_oid_list *child, *schild;
struct bge_mac_stats *stats;
stats = &sc->bge_mac_stats;
tree = SYSCTL_ADD_NODE(ctx, parent, OID_AUTO, "stats", CTLFLAG_RD,
NULL, "BGE Statistics");
schild = child = SYSCTL_CHILDREN(tree);
BGE_SYSCTL_STAT_ADD64(ctx, child, "FramesDroppedDueToFilters",
&stats->FramesDroppedDueToFilters, "Frames Dropped Due to Filters");
BGE_SYSCTL_STAT_ADD64(ctx, child, "DmaWriteQueueFull",
&stats->DmaWriteQueueFull, "NIC DMA Write Queue Full");
BGE_SYSCTL_STAT_ADD64(ctx, child, "DmaWriteHighPriQueueFull",
&stats->DmaWriteHighPriQueueFull,
"NIC DMA Write High Priority Queue Full");
BGE_SYSCTL_STAT_ADD64(ctx, child, "NoMoreRxBDs",
&stats->NoMoreRxBDs, "NIC No More RX Buffer Descriptors");
BGE_SYSCTL_STAT_ADD64(ctx, child, "InputDiscards",
&stats->InputDiscards, "Discarded Input Frames");
BGE_SYSCTL_STAT_ADD64(ctx, child, "InputErrors",
&stats->InputErrors, "Input Errors");
BGE_SYSCTL_STAT_ADD64(ctx, child, "RecvThresholdHit",
&stats->RecvThresholdHit, "NIC Recv Threshold Hit");
tree = SYSCTL_ADD_NODE(ctx, schild, OID_AUTO, "rx", CTLFLAG_RD,
NULL, "BGE RX Statistics");
child = SYSCTL_CHILDREN(tree);
BGE_SYSCTL_STAT_ADD64(ctx, child, "ifHCInOctets",
&stats->ifHCInOctets, "Inbound Octets");
BGE_SYSCTL_STAT_ADD64(ctx, child, "Fragments",
&stats->etherStatsFragments, "Fragments");
Consistently use ifHCOutOctets/ifHCInOctets instead of Octets as these names are used in data sheet. Also use UnicastPkts, MulticastPkts and BroadcastPkts instead of UcastPkts, McastPkts and BcastPkts to clarify its meaning. Suggested by: bde
2010-10-04 18:01:23 +00:00
BGE_SYSCTL_STAT_ADD64(ctx, child, "UnicastPkts",
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
&stats->ifHCInUcastPkts, "Inbound Unicast Packets");
BGE_SYSCTL_STAT_ADD64(ctx, child, "MulticastPkts",
&stats->ifHCInMulticastPkts, "Inbound Multicast Packets");
BGE_SYSCTL_STAT_ADD64(ctx, child, "BroadcastPkts",
&stats->ifHCInBroadcastPkts, "Inbound Broadcast Packets");
BGE_SYSCTL_STAT_ADD64(ctx, child, "FCSErrors",
&stats->dot3StatsFCSErrors, "FCS Errors");
BGE_SYSCTL_STAT_ADD64(ctx, child, "AlignmentErrors",
&stats->dot3StatsAlignmentErrors, "Alignment Errors");
BGE_SYSCTL_STAT_ADD64(ctx, child, "xonPauseFramesReceived",
&stats->xonPauseFramesReceived, "XON Pause Frames Received");
BGE_SYSCTL_STAT_ADD64(ctx, child, "xoffPauseFramesReceived",
&stats->xoffPauseFramesReceived, "XOFF Pause Frames Received");
BGE_SYSCTL_STAT_ADD64(ctx, child, "ControlFramesReceived",
&stats->macControlFramesReceived, "MAC Control Frames Received");
BGE_SYSCTL_STAT_ADD64(ctx, child, "xoffStateEntered",
&stats->xoffStateEntered, "XOFF State Entered");
BGE_SYSCTL_STAT_ADD64(ctx, child, "FramesTooLong",
&stats->dot3StatsFramesTooLong, "Frames Too Long");
BGE_SYSCTL_STAT_ADD64(ctx, child, "Jabbers",
&stats->etherStatsJabbers, "Jabbers");
BGE_SYSCTL_STAT_ADD64(ctx, child, "UndersizePkts",
&stats->etherStatsUndersizePkts, "Undersized Packets");
tree = SYSCTL_ADD_NODE(ctx, schild, OID_AUTO, "tx", CTLFLAG_RD,
NULL, "BGE TX Statistics");
child = SYSCTL_CHILDREN(tree);
Consistently use ifHCOutOctets/ifHCInOctets instead of Octets as these names are used in data sheet. Also use UnicastPkts, MulticastPkts and BroadcastPkts instead of UcastPkts, McastPkts and BcastPkts to clarify its meaning. Suggested by: bde
2010-10-04 18:01:23 +00:00
BGE_SYSCTL_STAT_ADD64(ctx, child, "ifHCOutOctets",
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
&stats->ifHCOutOctets, "Outbound Octets");
BGE_SYSCTL_STAT_ADD64(ctx, child, "Collisions",
&stats->etherStatsCollisions, "TX Collisions");
BGE_SYSCTL_STAT_ADD64(ctx, child, "XonSent",
&stats->outXonSent, "XON Sent");
BGE_SYSCTL_STAT_ADD64(ctx, child, "XoffSent",
&stats->outXoffSent, "XOFF Sent");
BGE_SYSCTL_STAT_ADD64(ctx, child, "InternalMacTransmitErrors",
&stats->dot3StatsInternalMacTransmitErrors,
"Internal MAC TX Errors");
BGE_SYSCTL_STAT_ADD64(ctx, child, "SingleCollisionFrames",
&stats->dot3StatsSingleCollisionFrames, "Single Collision Frames");
BGE_SYSCTL_STAT_ADD64(ctx, child, "MultipleCollisionFrames",
&stats->dot3StatsMultipleCollisionFrames,
"Multiple Collision Frames");
BGE_SYSCTL_STAT_ADD64(ctx, child, "DeferredTransmissions",
&stats->dot3StatsDeferredTransmissions, "Deferred Transmissions");
BGE_SYSCTL_STAT_ADD64(ctx, child, "ExcessiveCollisions",
&stats->dot3StatsExcessiveCollisions, "Excessive Collisions");
BGE_SYSCTL_STAT_ADD64(ctx, child, "LateCollisions",
&stats->dot3StatsLateCollisions, "Late Collisions");
Consistently use ifHCOutOctets/ifHCInOctets instead of Octets as these names are used in data sheet. Also use UnicastPkts, MulticastPkts and BroadcastPkts instead of UcastPkts, McastPkts and BcastPkts to clarify its meaning. Suggested by: bde
2010-10-04 18:01:23 +00:00
BGE_SYSCTL_STAT_ADD64(ctx, child, "UnicastPkts",
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
&stats->ifHCOutUcastPkts, "Outbound Unicast Packets");
Consistently use ifHCOutOctets/ifHCInOctets instead of Octets as these names are used in data sheet. Also use UnicastPkts, MulticastPkts and BroadcastPkts instead of UcastPkts, McastPkts and BcastPkts to clarify its meaning. Suggested by: bde
2010-10-04 18:01:23 +00:00
BGE_SYSCTL_STAT_ADD64(ctx, child, "MulticastPkts",
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
&stats->ifHCOutMulticastPkts, "Outbound Multicast Packets");
Consistently use ifHCOutOctets/ifHCInOctets instead of Octets as these names are used in data sheet. Also use UnicastPkts, MulticastPkts and BroadcastPkts instead of UcastPkts, McastPkts and BcastPkts to clarify its meaning. Suggested by: bde
2010-10-04 18:01:23 +00:00
BGE_SYSCTL_STAT_ADD64(ctx, child, "BroadcastPkts",
Implement hardware MAC statistics for BCM5705 or newer Broadcom controllers. bge(4) exported MAC statistics on controllers that maintain the statistics in the NIC's internal memory. Newer controllers require register access to fetch these values. These counters provide useful information to diagnose driver issues.
2010-09-29 21:56:31 +00:00
&stats->ifHCOutBroadcastPkts, "Outbound Broadcast Packets");
}
#undef BGE_SYSCTL_STAT_ADD64
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)
{
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;
Fix long standing bge_sysctl_debug_info() issues. o Protect bge(4) status block access and register dump with driver lock. o Add missing bus_dmamap_sync() before dumping status block. o Use minimum status block size, 32 bytes, for status block dump on most controllers except BCM5700 AX/BX. While I'm here, make the handler show 5717 Plus in hardware flags.
2011-10-26 01:03:53 +00:00
int error, result, sbsz;
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
int 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;
Fix long standing bge_sysctl_debug_info() issues. o Protect bge(4) status block access and register dump with driver lock. o Add missing bus_dmamap_sync() before dumping status block. o Use minimum status block size, 32 bytes, for status block dump on most controllers except BCM5700 AX/BX. While I'm here, make the handler show 5717 Plus in hardware flags.
2011-10-26 01:03:53 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
sc->bge_chipid != BGE_CHIPID_BCM5700_C0)
sbsz = BGE_STATUS_BLK_SZ;
else
sbsz = 32;
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
sbdata = (uint16_t *)sc->bge_ldata.bge_status_block;
printf("Status Block:\n");
Fix long standing bge_sysctl_debug_info() issues. o Protect bge(4) status block access and register dump with driver lock. o Add missing bus_dmamap_sync() before dumping status block. o Use minimum status block size, 32 bytes, for status block dump on most controllers except BCM5700 AX/BX. While I'm here, make the handler show 5717 Plus in hardware flags.
2011-10-26 01:03:53 +00:00
BGE_LOCK(sc);
bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map,
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
for (i = 0x0; i < sbsz / sizeof(uint16_t); ) {
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);
Fix long standing bge_sysctl_debug_info() issues. o Protect bge(4) status block access and register dump with driver lock. o Add missing bus_dmamap_sync() before dumping status block. o Use minimum status block size, 32 bytes, for status block dump on most controllers except BCM5700 AX/BX. While I'm here, make the handler show 5717 Plus in hardware flags.
2011-10-26 01:03:53 +00:00
for (j = 0; j < 8; j++)
printf(" %04x", sbdata[i++]);
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("\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");
}
Fix long standing bge_sysctl_debug_info() issues. o Protect bge(4) status block access and register dump with driver lock. o Add missing bus_dmamap_sync() before dumping status block. o Use minimum status block size, 32 bytes, for status block dump on most controllers except BCM5700 AX/BX. While I'm here, make the handler show 5717 Plus in hardware flags.
2011-10-26 01:03:53 +00:00
BGE_UNLOCK(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("Hardware Flags:\n");
Fix long standing bge_sysctl_debug_info() issues. o Protect bge(4) status block access and register dump with driver lock. o Add missing bus_dmamap_sync() before dumping status block. o Use minimum status block size, 32 bytes, for status block dump on most controllers except BCM5700 AX/BX. While I'm here, make the handler show 5717 Plus in hardware flags.
2011-10-26 01:03:53 +00:00
if (BGE_IS_5717_PLUS(sc))
printf(" - 5717 Plus\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 bge(4) build breakage when BGE_REGISTER_DEBUG is defined.
2010-10-06 01:23:40 +00:00
if (sc->bge_phy_flags & BGE_PHY_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);
}
Add APE firmware support and improve firmware handshake procedure. This change will enable IPMI access on 5717/5718/5719/5720 and 5761 controllers. Because ASF is not available when APE firmware is present, bge_allow_asf tunable is ignored when driver detects APE firmware. Also bge(4) no longer performs two resets(one blind reset and the other reset with firmware in mind) in device attach. Now bge(4) performs a reset with enough information in bge_reset(). The APE firmware also needs special handling to make suspend/resume work but it was not implemented yet. With this change, bge(4) should work on any 5717/5718/5719/5720 controllers. Special thanks to Mike Hibler at Emulab who setup remote debugging on Dell R820. Without his help I couldn't be able to address several issues happened on Dell Rx20 systems. And many thanks to Broadcom for continuing to support FreeBSD! Submitted by: davidch (initial version) H/W donated by: Broadcom Tested by: many Tested on: Del R820/R720/R620/R420/R320 and HP Proliant DL 360 G8
2012-10-11 06:43:43 +00:00
static int
bge_sysctl_ape_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 = APE_READ_4(sc, result);
printf("reg 0x%06X = 0x%08X\n", result, val);
}
return (error);
}
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 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;
Define MAC address mail box and use it instead of using hard-coded value.
2011-10-26 21:11:40 +00:00
mac_addr = bge_readmem_ind(sc, BGE_SRAM_MAC_ADDR_HIGH_MB);
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 ((mac_addr >> 16) == 0x484b) {
ether_addr[0] = (uint8_t)(mac_addr >> 8);
ether_addr[1] = (uint8_t)mac_addr;
Define MAC address mail box and use it instead of using hard-coded value.
2011-10-26 21:11:40 +00:00
mac_addr = bge_readmem_ind(sc, BGE_SRAM_MAC_ADDR_LOW_MB);
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
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