CVS log for src/sys/dev/pci/if_bnxreg.h

Up to [cvs.NetBSD.org] / src / sys / dev / pci

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

fix spelling mistakes, mainly in comments and log messages.

triple "r" typos, mainly s/interrrupt/interrupt/ in comments and one definition.

fix various typos, mainly in comments.

Sync with HEAD.

Curb aprint_*() abuse in bnx(4)

Sync with HEAD

 Whilespace fix. No functional change.

 Add BNX_PCICFG_DEVICE_CONTROL.

- Allow time for the management firmware to enter the running state.
- Enable/Disable management frames (NC-SI) to flow to the MCP if MFW exists.
- Print the boot code version, management firmware version and some flags.
- Gbps -> GT/s for PCIe speed.
- Use aprint_*() in bnx_print_adapter_info().
- Use braces for some macors' argument.
- Remove unused macros.
- KNF.

No functional change. This change reduces diff against DragonFly:
- Fix BNX_PCI_GRC_WINDOW_ADDR_VALUE's name. Note that this macro is not used.
- Whitespace change.
- Sort some lines.

- Use m_defrag() when bus_dmamap_load_mbuf() returned with EFBIG.
- Style change to reduce diff against {Free,Open}BSD.
- Make it compilable with BNX_DEBUG.
- Use __nothing for null macro.

No functional change:
- Fix DB_AND_RANDOMFALSE(). It has no real bug because the macro is not used
  from anywhere.
- Wse const.
- Whitespace change.

 Centralize ETHER_ALIGN into net/if_ether.h. Note that this commit also changes
if_upgt.c's ETHER_ALIGN from 0 to 2.

Sync with HEAD

Constify the microcode variables used by BNX. This moves 38 pages of kernel
memory from .data to .rodata.

Pull up the following revisions(s) (requested by msaitoh in ticket #1983):
	sys/dev/pci/pcidevs:		revisions 1.1079, 1.1134, 1.1148-1.1149, 1.1151
	sys/dev/pci/pcidevs.h:		regen
	sys/dev/pci/pcidevs_data.h:	regen
	sys/dev/pci/if_bge.c:		revisions 1.183-1.185, 1.187, 1.189-1.193, 1.195-1.199, 1.202-1.226, 1.228-1.237, 1.240-1.264, 1.267-1.276, 1.278-1.280, 1.283-1.287 via patch
	sys/dev/pci/if_bgereg.h:	revisions 1.57-1.74, 1.76-1.90 via patch
	sys/dev/pci/if_bgevar.h:	revisions 1.6, 1.10-1.13, 1.15-1.17 via patch
	sys/dev/pci/if_bnx.c:		revisions 1.32, 1.34-1.43, 1.48-1.49, 1.52
	sys/dev/pci/if_bnxreg.h:	revisions 1.8, 1.11-1.14
	sys/dev/pci/if_bnxvar.h:	revisions 1.1-1.3
	sys/dev/mii/brgphy.c:		revisions 1.53-1.63, 1.65-69, 1.72-1.74 via patch
	sys/dev/mii/brgphyreg.h:	revisions 1.5-1.8
	sys/dev/mii/miivar.h:		revisions 1.61
	sys/dev/pci/pcireg.h:		patch

Sync bge(4) up to if_bge.c rev. 1.287. Sync brgphy(4) up to 1.74.
Fix some bugs on bnx(4).

Common:
- Add device IDs for Broadcom BCM57710, BCM57711(E), BCM57712(E) and
  BCM57766 (pcidevs only).
- Fix BCM5709 PHY detection.
- Fix detection of BGEPHYF_FIBER_{MII|TBI}
- Add BCM5708S support in brgphy(4).
- Don't use the WIRESPEED function for fiber devices.
bge(4):
- Add some Fujitsu's device support from Michael Moll.
- Add BCM57762 support (PR#46961 from Ryo Onodera).
- Add Altima AC1003, APPLE BCM5701, Broadcom BCM5785F. BCM5785G,
  BCM5787F, BCM5719, BCM5720, BCM57766, BCM57782 and BCM57786.
- Fix DMA setting for read/write on conventional PCI bus devices.
  This bug was added in rev. 1.166.
- Fix printing "discarding oversize frame (len=-4)" message and
  crash by NULL pointer dereferencing.
- The BCM5785 is a PCIe chip but does not report PCIe capabilities.
  Check for this chip explicitely and enable PCIe. Fixes
  'firmware handshake timeout'.
- Allow disabling interrupt mitigation.
- Workaround for BCM5906 silicon bug. When auto-negotiation results
  in half-duplex operation, excess collision on the ethernet link may
  cause internal chip delays that may result in subsequent valid
  frames being dropped due to insufficient receive buffer resources.
  (FreeBSD: r214219, r214251, r214292).
- Allow write DMA to request larger DMA burst size to get better
  performance on BCM5785. (FreeBSD r213333: OpenBSD 1.294)
- 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. (FreeBSD r214216)
- Follow Broadcom datasheet:
  Delay 100 microseconds after enabling transmit MAC.
  Delay 10 microseconds after enabling receive MAC. (FreeBSD
  r241220)
- Insert the completion barrier between register write and the
  consecutive delay(). It will fix some device timeout problems we have
  seen before.
- Add DELAY(40) after turning on write DMA state machine.
- Add some workarounds for 5717 A0 and 5776[56] to be stable.
- Check BGE_RXBDFLAG_IPV6 flag for 5717_PLUS case. Note that
  {tcp,udp}6csum flag is currently not added in the capability.
- Add delay after clearing BGE_MACMODE_TBI_SEND_CFGS for the link
  checking.
- Do not touch the jumbo replenish threshold register on chips that do
  not have jumbo support.
- Wait for the bootcode to complete initialization for 5717 and newer
  devices.
- 5718 and 57785 document say we should wait 100us in init.
- Fix a bug that chips which have BCM5906 ASIC touch GPIO wrongly.
- Fix the setting of Tx Random Backoff Register.
- Check the hardware config words and print them.
- Set BGE_MISC_CTL's byte/word swap options before using
  bge_readmem_ind(). Fixes PR#47716.
- For BGE_IS_575X_PLUS() devices, don't set
  BGE_RXLPSTATCONTROL_DACK_FIX bits because these bits are reserved.
- Document says 5717 and newer chips have no
  BGE_PCISTATE_INTR_NOT_ACTIVE bit, so don't use the bit on those
  chips. Same as OpenBSD.
- Fix a bug that the PHY address bits in MI_MODE register is wrongly
  cleard. Set the PHY address correctly.
- Use BGE_SETBIT() instead of CSR_WRITE_4() for the BGE_MISC_LOCAL_CTL
  register to not to modify some GPIO bits.
- Set DMA watermark depend on the PCI max payload size.
- Set BGE_JUMBO_CAPABLE correctly.
- Fix a link detect bug on non-autopoll systems.
- Change the TX ring size for 5717 series and 57764 series.
- Set maximum read byte count to 2048 for PCI-X BCM5703/5704 devices.
- For PCI-X BCM5704, set maximum outstanding split transactions to 0.
- Add 40bit DMA bug workaround(BGEF_40BIT_BUG) from FreeBSD.
  This workaround is for 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. This change uses bus_dmatag_subregion(),
  so this workaround won't work on some archs which doesn't support
  bus_dmatag_subregion().
- Add 2500SX support (not tested).
- Don't use the PHY Auto Poll Mode on many chips. This fixes a bug
  that MII Fiber NIC drop packet about 50%. Tested on HP Moonshot.
- Add workaround for PR#48451. Some BCM5717-5720 based systems getNMI
  on boot. This problem doesn't occur when we don't use prefetchable
  memory in the APE area. Tested with HP MicroServer Gen8.
- 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.
- Fix a bug that BGE_PHY_TEST_CTRL_REG isn't set correctly on some
  PCIe devices.
- Use another firmware command in bge_asf_driver_up(). Same as Linux.
  This change fixes a bug that watchdog timeout occurs every 25-30
  minutes on HP ML110 G6 reported enami@ in PR#49657.
- Fix mbuf leak on failure.
- Remove PCI_PRODUCT_BROADCOM_BCM5724 and
  PCI_PRODUCT_BROADCOM_BCM5750M. These devices have not released to
  public.
- Add some workaround code for BGE_ASICREV_BCM5784 from Linux.
- Change some printf() to aprint_*().
- Fix typo in comments.
- Cleanup.
brgphy(4):
- Fix bit definition of BRGPHY_MRBE_MSG_PG5_NP_T2 from FreeBSD.
- Add BCM5481, BCM5709S, BCM5756, BCM5717C, BCM5720C, BCM5785,
  BCM57765(PR#46961), BCM57780
- In brgphyattach(), set sc_isbge, sc_isbnx and sc_phyflags before
  PHY_RESET() because brgphy_reset() refers those flags.
- Call brgpy specific autonego function in MII_TICK. Before this
  commit, only MII_MEDIACHG calls brgphy_mii_phy_auto() and MII_TICK
  calls MI mii_phy_auto(). That was not intended.
- Remove extra delay in brgphy_mii_phy_auto. Same as {Free,Open}BSD.
bnx(4):
- Add missing ifmedia_delete_instance() in bnx_detach().
- Fix a bug that BNX_NO_WOL_FLAG isn't correctly set on some chips.
  Reported by From Henning Petersen in PR#44151.
- Fix SERDES initialization.
- Get out of the interrupt handler early if !IFF_RUNNING.

Pull up the following, requested by msaitoh in #1207:

sys/dev/mii/brgphy.c				1.69, 1.71-1.76 via patch
sys/dev/mii/brgphyreg.h				1.6-1.7, 1.9
sys/dev/pci/if_bge.c				1.259,1.267-1.276 via patch
sys/dev/pci/if_bgereg.h				1.87-1.88
sys/dev/pci/if_bnx.c				1.48, 1.49, 1.52-1.57 via patch
sys/dev/pci/if_bnxreg.h				1.15-1.18
sys/dev/pci/if_bnxvar.h				1.4-1.6

	Changes for bge(4), bnx(4) and brgphy(4):
	- Add device IDs for Broadcom BCM57710, BCM57711(E), BCM57712(E) and
	  BCM5719 into pcidevs.
	- Change BCM5720's device ID from 0x1658 to 0x165f. 0x1658 is only for
	  the pre-product and not released to public.
	- Fix detection of BGEPHYF_FIBER_{MII|TBI}
	- Fix typo in comment.
	- KNF
	- brgphy(4):
	  - Add BCM5708S support.
	  - Add 2500SX support.
	  - Fix bit definition of BRGPHY_MRBE_MSG_PG5_NP_T2 from FreeBSD.
	  - Remove extra delay in brgphy_mii_phy_auto. Same as {Free,Open}BSD.
	  - Fix media detect on some Fiber chips.
	  - Add BCM5706.
	- bge(4):
	  - Print yet another hardware config word in SRAM(BGE_SRAM_DATA_CFG_5)
	    for BCM5717 and newer devices.
	  - Fix a trivial bug in #ifdef NOTYET.
	  - Add BGEPHYF_NO_WIRESPEED flag for fiber devices.
	  - Print APE firmware correctly.
	  - Don't use the PHY Auto Poll Mode on many chips. This fixes a bug
	    that MII Fiber NIC drop packets about 50%. Same as {Free,Open}BSD.
	    Tested on HP Moonshot.
	  - Add workaround for PR#48451. Some BCM5717-5720 based systems get
	    NMI on boot. This problem doesn't occur when we don't use
	    prefetchable memory in the APE area. I don't know the reason...
	    Tested with HP MicroServer Gen8.
	  - Fix a bug that bge can't be initialized correctly on some sparc64
	    systems (PR#49018).
	  - Change initializaton order as the same as the document, Linux tg3
	    and FreeBSD (part of r241436).
	  - 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. Same as FreeBSD r204978 and Linux tg3 driver.
	- bnx(4):
	  - Enclose #ifdef BNX_DEBUG correctly.
	  - Add 2500SX support.
	  - Print some information (ASCI revision, PCI status, etc).
	    From FreeBSD.

Rebase to HEAD as of a few days ago.

Rebase.

Print some information (ASIC revision, PCI status, etc). From FreeBSD.

No functional change.
- s/u_int/uint/
- Modify comment a bit (sync with OpenBSD)

No functional change:
- Tabify.
- Remove trailing white spaces.
- Fix typo in comment.

- Fix detection of BGEPHYF_FIBER_{MII|TBI}
- Add BCM5708S support in brgphy(4). The auto negotiation may have some bugs.
- Add 2500SX support (not tested).
- Fix bit definition of BRGPHY_MRBE_MSG_PG5_NP_T2 from FreeBSD.

Sync with HEAD.

sync with head

Fix BCM5709 PHY detection for ethernet PHYs (the SerDes case being already
handled):
- export bge(4) and bnx(4) CHIP ID and PHY flags to brgphy(4). Move to
"unsigned int" rather than "int", and reuse the same softc members for
chipid and phyflags (behavior controlled by the sc_isbge/isbnx boolean).
- apply bug fix for revisions A and B, so that autonegotiation can
complete (from OpenBSD).

Bug reported by Rivo Nurges via private mail, patch tested and
confirmed working by him (with thanks!)

sync with head

Sanitize names a bit, still could use more namespace cleanup

Namespace police: don't use very broad names like CTX_MASK (which accidently
collide with sparc* specific MD macros), mechanically prefix with BNX_.

Similar to bge(4), split if_bnxreg.h in two parts:
- if_bnxreg.h: general register values, flags... for bnx(4), that can
be included in other drivers
- if_bnxvar.h: PCI or device specific code (device state data, debug
macros etc.), which should remain private to bnx(4)

No comments, no objections on current-users@.

See discussion:
http://mail-index.netbsd.org/current-users/2010/12/01/msg014926.html

sync to netbsd-5

sync with head

Pull up the following revisions(s) (requested by bouyer in ticket #1268):
	sys/dev/mii/miidevs:	revision 1.90
	sys/dev/pci/pcidevs:	revision 1.1001
	sys/dev/mii/brgphy.c:	revision 1.51
	sys/dev/microcode/bnx/bnxfw.h:	revision 1.3
	sys/dev/pci/if_bnx.c:	revision 1.29
	sys/dev/pci/if_bnxreg.h:	revision 1.9
	sys/dev/pci/if_bnx.c:	revision 1.30

Add support for the Broadcom BCM5709 and BCM5716 chips. Upgrade
the B06 firmware and add the B09 firmware required by the BCM5709.
Replace TX dmamap array with a dynamically-grown list.
Support associated BCM5709CAX and BCM5709C PHYs.

Redefine bpf linkage through an always present op vector, i.e.
#if NBPFILTER is no longer required in the client.  This change
doesn't yet add support for loading bpf as a module, since drivers
can register before bpf is attached.  However, callers of bpf can
now be modularized.

Dynamically loadable bpf could probably be done fairly easily with
coordination from the stub driver and the real driver by registering
attachments in the stub before the real driver is loaded and doing
a handoff.  ... and I'm not going to ponder the depths of unload
here.

Tested with i386/MONOLITHIC, modified MONOLITHIC without bpf and rump.

Sync with OpenBSD: if_bnx.c,v 1.85. Major changes:
Add support for the Broadcom BCM5709 and BCM5716 chips. This upgrade
the B06 firmware, and add the B09 firmware required by the BCM5709.
Remplace TX dmamap array with a dynamically-grown list.

Tested with:
bnx0 at pci4 dev 0 function 0: Broadcom NetXtreme II BCM5708 1000Base-T
and
bnx0 at pci0 dev 0 function 0: Broadcom NetXtreme II BCM5709 1000Base-SX

Sync with HEAD.

Commit is split, to avoid a "too many arguments" protocol error.

sync with head.

Sync with HEAD.

Delete two unused data members, the powerhook and the shutdownhook
handles.

Sync with wrstuden-revivesa-base-2.

Sync with HEAD.

Sync with head.

Replace gcc variadic macros with c99 variadic macros.

sync with HEAD

Sync with HEAD.

sync with head.

Don't use misaligned PCI config access. Reported by Christoph Egger.

sync with head

sync with HEAD

Sync with HEAD

Don't buffer packets while the link is down, but drop them on the
floor like most other ethernet drivers do.

bnx(4) has "vanilla" MII media-handling, so use ether_mediachange
and ether_mediastatus.

For consistency with other softc members, s/ethercom/bnx_ec/.

Sync with head.

Add device suspend/resume/detach.  (Tested with drvctl -S bnx0,
drvctl -Q bnx0, drvctl -d bnx0.)

Separate device_t storage from softc storage and use CFATTACH_DECL_NEW().

sync with head.

Sync with head.

Pull up following revision(s) (requested by bouyer in ticket #1762):
	sys/dev/pci/if_bnx.c: revision 1.4
	sys/dev/pci/if_bnx.c: revision 1.5
	sys/dev/pci/if_bnx.c: revision 1.6
	sys/dev/pci/if_bnxreg.h: revision 1.2
Protect bnx_tick() with splnet.
Bring in fixes and improvements from OpenBSD:
revision 1.25
- Simplify the arguments to bnx_tx_encap.
- Don't copy the bd_chain head pointers into temporary objects, they are
available globally.
From scottl@FreeBSD
revision 1.26
Overhaul the transmit path:
- Eliminate the bnx_dmamap_arg structure.
- Refactor the loop that fills the buffer descriptor so that it can be done
with a single set of logic in a single loop instead of two sets of logic.
- Eliminate the need to cache and pass descriptor indexes between the start
loop and the encap function.
- Change the start loop to always check the ifnet sendq for more work.
From scottl@FreeBSD
revision 1.27
make the exit label naming scheme match the current function names, removes
a FreeBSD-ism from the original driver.
revision 1.28 -> 1.30
- Ensure that at least 16 TX descriptors are kept unused in the ring.
- Use more complete error handling for TX load problems.
From scottl@FreeBSD
revision 1.31
replace a few more instances of hand rolled code with the LIST_FOREACH macro.
revision 1.33
In bnx_start, check the used_tx_bd count rather than the descriptors
mbuf pointer to see if the transmit ring is full.  The mbuf pointer
is set only in the last descriptor of a multi-descriptor packet.
By relying on the mbuf pointers of the earlier descriptors, the
driver would sometimes overwrite a descriptor belonging to a
packet that wasn't completed yet.  Also, tx_chain_prod wasn't
updated inside the loop, causing the wrong descriptor to be checked
after the first iteration.  The upshot of all this was the loss of
some transmitted packets at medium to high packet rates.
In bnx_tx_encap, remove a couple of old statements that shuffled
around the tx_mbuf_map pointers.  These now correspond 1-to-1 with
the transmit descriptors, and they are not supposed to be changed.
Correct a couple of inaccurate comments.
From jdp@FreeBSD
revision 1.43
Allow the bnx(4) driver to make use of all of the available hardware
multicast hash slots. The bnx(4) hardware supports 8 slots instead of
4 like the bge(4) hardware.
From Mike Karels via FreeBSD
Tested by Brad, biorn@ and Johan M:son Lindman
RX buffers are malloced memory of 9216 bytes. This can require from 1 to
4 DMA memory segments, depending on how the buffer is in memory.
When receiving a packet, we allocate a new one to remplace the one we've
used. It can need more segments than the one it remplace, leading to
corrution of the RX descriptors, and a panic in bus_dmamap_sync() (DIAGNOSTIC
kernels) or possibly memory corruption.
Fix:
- bnx_get_buf() allocates as many buffer as possible, checking the number
  of free RX descriptors. Because one receive buffer is not guaranteed to
  be remplaced on receive, call bnx_get_buf() from bnx_tick() too.
  This also improve error handling from bnx_get_buf().
- use MCLGET() instead of MEXTMALLOC() if we're running with the standard
  ethernet MTU. This gives us more receive buffers and waste less memory.
Fix build on 32bit ports; should fix GENERIC_DIAGNOSTIC build problem reported
by Hisashi T Fujinaka on current-users.

Pull up following revision(s) (requested by bouyer in ticket #566):
	sys/dev/pci/if_bnx.c: revision 1.4
	sys/dev/pci/if_bnx.c: revision 1.5
	sys/dev/pci/if_bnx.c: revision 1.6
	sys/dev/pci/if_bnxreg.h: revision 1.2
Protect bnx_tick() with splnet.
Bring in fixes and improvements from OpenBSD:
revision 1.25
- Simplify the arguments to bnx_tx_encap.
- Don't copy the bd_chain head pointers into temporary objects, they are
available globally.
From scottl@FreeBSD
revision 1.26
Overhaul the transmit path:
- Eliminate the bnx_dmamap_arg structure.
- Refactor the loop that fills the buffer descriptor so that it can be done
with a single set of logic in a single loop instead of two sets of logic.
- Eliminate the need to cache and pass descriptor indexes between the start
loop and the encap function.
- Change the start loop to always check the ifnet sendq for more work.
From scottl@FreeBSD
revision 1.27
make the exit label naming scheme match the current function names, removes
a FreeBSD-ism from the original driver.
revision 1.28 -> 1.30
- Ensure that at least 16 TX descriptors are kept unused in the ring.
- Use more complete error handling for TX load problems.
From scottl@FreeBSD
revision 1.31
replace a few more instances of hand rolled code with the LIST_FOREACH macro.
revision 1.33
In bnx_start, check the used_tx_bd count rather than the descriptors
mbuf pointer to see if the transmit ring is full.  The mbuf pointer
is set only in the last descriptor of a multi-descriptor packet.
By relying on the mbuf pointers of the earlier descriptors, the
driver would sometimes overwrite a descriptor belonging to a
packet that wasn't completed yet.  Also, tx_chain_prod wasn't
updated inside the loop, causing the wrong descriptor to be checked
after the first iteration.  The upshot of all this was the loss of
some transmitted packets at medium to high packet rates.
In bnx_tx_encap, remove a couple of old statements that shuffled
around the tx_mbuf_map pointers.  These now correspond 1-to-1 with
the transmit descriptors, and they are not supposed to be changed.
Correct a couple of inaccurate comments.
From jdp@FreeBSD
revision 1.43
Allow the bnx(4) driver to make use of all of the available hardware
multicast hash slots. The bnx(4) hardware supports 8 slots instead of
4 like the bge(4) hardware.
From Mike Karels via FreeBSD
Tested by Brad, biorn@ and Johan M:son Lindman

RX buffers are malloced memory of 9216 bytes. This can require from 1 to
4 DMA memory segments, depending on how the buffer is in memory.
When receiving a packet, we allocate a new one to remplace the one we've
used. It can need more segments than the one it remplace, leading to
corrution of the RX descriptors, and a panic in bus_dmamap_sync() (DIAGNOSTIC
kernels) or possibly memory corruption.
Fix:
- bnx_get_buf() allocates as many buffer as possible, checking the number
  of free RX descriptors. Because one receive buffer is not guaranteed to
  be remplaced on receive, call bnx_get_buf() from bnx_tick() too.
  This also improve error handling from bnx_get_buf().
- use MCLGET() instead of MEXTMALLOC() if we're running with the standard
  ethernet MTU. This gives us more receive buffers and waste less memory.

Fix build on 32bit ports; should fix GENERIC_DIAGNOSTIC build problem reported
by Hisashi T Fujinaka on current-users.

sync with head.

Sync with head.

Protect bnx_tick() with splnet.

Bring in fixes and improvements from OpenBSD:
revision 1.25
- Simplify the arguments to bnx_tx_encap.
- Don't copy the bd_chain head pointers into temporary objects, they are
available globally.

From scottl@FreeBSD

revision 1.26
Overhaul the transmit path:
- Eliminate the bnx_dmamap_arg structure.
- Refactor the loop that fills the buffer descriptor so that it can be done
with a single set of logic in a single loop instead of two sets of logic.
- Eliminate the need to cache and pass descriptor indexes between the start
loop and the encap function.
- Change the start loop to always check the ifnet sendq for more work.

From scottl@FreeBSD

revision 1.27
make the exit label naming scheme match the current function names, removes
a FreeBSD-ism from the original driver.

revision 1.28 -> 1.30
- Ensure that at least 16 TX descriptors are kept unused in the ring.
- Use more complete error handling for TX load problems.

From scottl@FreeBSD

revision 1.31
replace a few more instances of hand rolled code with the LIST_FOREACH macro.

revision 1.33
In bnx_start, check the used_tx_bd count rather than the descriptors
mbuf pointer to see if the transmit ring is full.  The mbuf pointer
is set only in the last descriptor of a multi-descriptor packet.
By relying on the mbuf pointers of the earlier descriptors, the
driver would sometimes overwrite a descriptor belonging to a
packet that wasn't completed yet.  Also, tx_chain_prod wasn't
updated inside the loop, causing the wrong descriptor to be checked
after the first iteration.  The upshot of all this was the loss of
some transmitted packets at medium to high packet rates.

In bnx_tx_encap, remove a couple of old statements that shuffled
around the tx_mbuf_map pointers.  These now correspond 1-to-1 with
the transmit descriptors, and they are not supposed to be changed.

Correct a couple of inaccurate comments.

From jdp@FreeBSD

revision 1.43
Allow the bnx(4) driver to make use of all of the available hardware
multicast hash slots. The bnx(4) hardware supports 8 slots instead of
4 like the bge(4) hardware.

From Mike Karels via FreeBSD

Tested by Brad, biorn@ and Johan M:son Lindman

Sync with head.

sync with head.

Pull up following revision(s) (requested by bouyer in ticket #1608):
	sys/arch/i386/conf/INSTALL: revision 1.299 via patch
	sys/arch/i386/conf/XEN2_DOM0: revision 1.20 via patch
	sys/arch/i386/conf/GENERIC: revision 1.803 via patch
	sys/dev/pci/files.pci: revision 1.275 via patch
	sys/arch/i386/conf/GENERIC_LAPTOP: revision 1.218 via patch
	sys/arch/amd64/conf/GENERIC: revision 1.123 via patch
	sys/dev/pci/if_bnx.c: revision 1.1 via patch
	share/man/man4/bnx.4: revision 1.1-1.2 via patch
	distrib/sets/lists/man/mi: revision 1.959 via patch
	sys/arch/amd64/conf/INSTALL: revision 1.62 via patch
	sys/dev/microcode/bnx/bnxfw.h: revision 1.1-1.2 via patch
	sys/dev/pci/if_bnxreg.h: revision 1.1 via patch
	share/man/man4/Makefile: revision 1.417 via patch
	sys/arch/i386/conf/INSTALL_LAPTOP: revision 1.112 via patch
Add bnx(4), a driver for Broadcom NetXtreme II 10/100/1000 Ethernet device.
Ported from OpenBSD by cube@, with some bus_dma fixes by me.
Tested on i386 and amd64.
Add a NetBSD RCS tag.
Nx -> Ox; remove trailing whitespace.

Pull up following revision(s) (requested by bouyer in ticket #287):
	sys/arch/i386/conf/INSTALL: revision 1.299
	sys/arch/i386/conf/XEN2_DOM0: revision 1.20
	sys/arch/i386/conf/GENERIC: revision 1.803
	sys/dev/pci/files.pci: revision 1.275
	sys/arch/i386/conf/GENERIC_LAPTOP: revision 1.218
	sys/arch/amd64/conf/GENERIC: revision 1.123
	sys/dev/pci/if_bnx.c: revision 1.1
	share/man/man4/bnx.4: revision 1.1
	distrib/sets/lists/man/mi: revision 1.959
	sys/arch/amd64/conf/INSTALL: revision 1.62
	sys/dev/microcode/bnx/bnxfw.h: revision 1.1
	sys/arch/i386/conf/ALL: revision 1.76
	sys/dev/pci/if_bnxreg.h: revision 1.1
	share/man/man4/Makefile: revision 1.417
	sys/arch/i386/conf/INSTALL_LAPTOP: revision 1.112
Add bnx(4), a driver for Broadcom NetXtreme II 10/100/1000 Ethernet device.
Ported from OpenBSD by cube@, with some bus_dma fixes by me.
Tested on i386 and amd64.

sync with head.

file if_bnxreg.h was added on branch newlock2 on 2007-01-12 00:57:41 +0000

file if_bnxreg.h was added on branch yamt-lazymbuf on 2006-12-30 20:48:44 +0000

file if_bnxreg.h was added on branch netbsd-3 on 2006-12-23 14:43:47 +0000

file if_bnxreg.h was added on branch netbsd-4 on 2006-12-21 13:44:01 +0000

file if_bnxreg.h was added on branch yamt-splraiseipl on 2006-12-18 11:42:13 +0000

Add bnx(4), a driver for Broadcom NetXtreme II 10/100/1000 Ethernet device.
Ported from OpenBSD by cube@, with some bus_dma fixes by me.
Tested on i386 and amd64.