Merge branch 'upstream-linus' of /linux/kernel/git/jgarzik/netdev-2.6

Merge /pub/scm/linux/kernel/git/davem/net-2.6

Merge /pub/scm/linux/kernel/git/davem/sparc-2.6

[PATCH] Fix sparse parse error in lppaca.h

sparse can't parse a struct definition in include/asm-powerpc/lppaca.h,
even though gcc can accept it.  The form looks like this:

        struct __attribute__((whatever)) foo { };

An equivalent that both gcc and sparse can handle is

        struct foo { } __attribute__((whatever));

This is the only definition of this type in the tree, and fixing it is
easier than fixing sparse.

Signed-off-by: Bryan O'Sullivan <bos@serpentine.com>
[ Side note: fixing sparse wouldn't be hard, but the "attribute at the
  end" version is the canonical one, and the one that makes sense. So
  let's just fix the kernel instead. Luc Van Oostenryck already sent
  out a sparse patch to the sparse mailing list in case anybody cares.
               -- Linus ]
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] skge: fix dma mask setup.

There are a couple of problems in the DMA setup code for skge.
* In the 64 bit case, it doesn't set the consistent mask.
* In the 32 bit case, the error check is backwards!
It likely will only be visible as a bug on 64 bit platforms.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] Fix warning with b44.c on 64bit boxes

sizeof() return is not an int, so use max_t to get the types right.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] sky2: 0.13 version

Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] sky2: more conservative transmit locking

Be more careful about transmit locking, this solves a possible race
between tx_complete and transmit, that would cause a tx timeout.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] sky2: don't inline so much

Don't need to inline quite so many routines, let the compiler
decide

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] sky2: use kzalloc

Can use kzalloc here.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] sky2: ratelimit error messages

Make sure and rate limit all the error messages that might occur. If a problem
occurs then a few messages are enough.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] sky2: optimize for 32 bit dma

Small optimization, if dma addresses are 32 bits, then high
bits are always zero.

Signed-off-by: Stephen Hemminger <shemminger@osdl.or>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] sky2: don't bother clearing status ring elements

Don't need to zero out the status ring entries after processing.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] sky2: write barrier's

Be more careful about memory barriers. The only place we really
need them is before and after updating the chip's ring interface.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] sky2: fix ram buffer for Yukon FE rev 2

Fix problems with Yukon FE rev 2 chipset. Don't cut and paste bugs in from
sk98lin driver. Change how the ram buffer is divided up, and make the math
clearer. Also, set the thresholds where rx takes precedence. The threshold
values are just guesses at this point, it might be worth tuning them later.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] sky2: version 0.12

Version update.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] sky2: call pci_set_consistent_dma_mask

Need to call pci_set_consistent_dma_mask in the case of 64 bit
DMA.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] sky2: receive buffer alignment

Need to make sure that sky2 receive buffers are 64 bit
aligned. Also, don't need to start off with GFP_ATOMIC
on initial setup.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e100: e100 whitespace fixes

e100: e100 whitespace fixes

These are whitespace only fixes.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e100: Handle the return values from pci_* functions

e100: Handle the return values from pci_* functions

This is to resolve warnings during compile time.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e100: Fix TX hang and RMCP Ping issue (due to a microcode loading issue)

e100: Fix TX hang and RMCP Ping issue (due to a microcode loading issue)

Set the end of list bit to cause the hardware's transmit state machine to
work correctly and not prevent management (BMC) traffic.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[SOUND]: sparc/cs4231: Fix some typos which wrecked the build.

Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC]: Fix sbusfb build.

sbusfb_compat_ioctl() needs to return int, not long, as that
is what the fb_ops->fb_compat_ioctl method prototype wants.

Need to git rid of the "struct file *file" first argument to
fbiogetputcmap() and fbiogscursor() to match calls done in
sbusfb_compat_ioctl().

Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC]: change if() BUG(); to BUG_ON in iommu.c

this patch changes if() BUG(); constructs
in iommu.c to BUG_ON(); so it gets save
to define BUG() and BUG_ON() to nullstatements.

Signed-off-by: Eric Sesterhenn <snakebyte@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC64]: Eliminate race condition reading Hummingbird STICK register

Ensure a consistent value is read from the STICK register by ensuring
that both high and low are read without high changing due to a roll
over of the low register.

Various Debian/SPARC users (myself include) have noticed problems with
Hummingbird based systems. The symptoms are that the system time is
seen to jump forward 3 days, 6 hours, 11 minutes give or take a few
seconds. In many cases the system then hangs some time afterwards.

I've spotted a race condition in the code to read the STICK register.
I could not work out why 3d, 6h, 11m is important but guess that it is
due to the 2^32 jump of STICK (forwards on one read and then the next
read will seem to be backwards) during a timer interrupt. I'm guessing
that a change of -2^32 will get converted to a large unsigned
increment after the arithmetic manipulation between STICK,
nanoseconds, jiffies etc.

I did a test where I modified __hbird_read_stick to artificially
inject rollover faults forcefully every few seconds. With this I saw
the clock jump over 6 times in 12 hours compared to once every month
or so.

Signed-off-by: Richard Mortimer <richm@oldelvet.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Use is_zero_ether_addr() in net/core/netpoll.c

This replaces a memcmp() with is_zero_ether_addr().

Signed-off-by: Kris Katterjohn <kjak@users.sourceforge.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[CASSINI]: Fix printk warning.

drivers/net/cassini.c:1930: warning: long unsigned int format, different type arg (arg 4)

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PKTGEN]: Replacing with (compare|is_zero)_ether_addr() and ETH_ALEN

This replaces some tests with is_zero_ether_addr(), memcmp(one, two,
6) with compare_ether_addr(one, two), and 6 with ETH_ALEN where
appropriate.

Signed-off-by: Kris Katterjohn <kjak@users.sourceforge.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: "signed long" -> "long"

Signed-off-by: Kris Katterjohn <kjak@users.sourceforge.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[EBTABLES]: Handle SCTP/DCCP in ebt_{ip,log}

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PKT_SCHED]: Handle SCTP/DCCP in sfq_hash

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

airo: Off-by-one channel fix

[PATCH] e1000: Removed unused variables and initialized variables

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Cleaned up code and removed hard coded numbers

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Added copy break code

Improves small packet performance with large amounts of reassembly being done in the stack.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Added variable to handle return values for pci_enable_* functions

This was to fix compilation warnings.  Also added log messages when pci_enable_* functions return with an error.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Added PCIe bus information

This is two patches, the first is adding additional bus information for the 8257{1|2|3} controllers.  The second patch was orginally a community patch to print bus type/speed/width, and enhanced by us.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Added firmware version reporting for 8257{1|2|3} controllers

The EEPROM image version is reported as a firmware version for these controllers.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Added hardware support for PCI express, 82546GB, and 82571 Fiber

Added 82571 fiber to WOL fix for dual port adapters.
Added support for 82546GB (Quad Copper).
Added PCIe typedef for x2, igp cable length 115, and extended TX CTRL registers.
Added parity error detection and PCIe CTRL registers.
Added EEPROM config registers.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Added cleaned_count to RX buffer allocation

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Added interrupt auto mask support

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix Netpoll issue

Fixed an issue netpoll would error out during communication, generating the following error:
--netdump[14973]: Got toomany timeouts in handshaking, ...
Even after a kernel panic, netpoll requires two way communication to successfully transfer the crash log to the remote server.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fixed frame size logic

Simplified the logic used to assign the frame_size.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix VLAN support

Fixed VLAN support by switching control over to the firmware.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix __pskb_pull_tail

Fixed by moving code to correct location (for 82572 and 82571 controllers).

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix collision distance

Fixed the collision distance for 82543 controllers and newer.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix bit 22 (TXDCTL) for 82571 & 82572 controllers

Removed duplicate code, TXDCTL and TXDCTL_COUNT_DESC are the same bit and there is no need to set it twice.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix desc. clean up

These were two separate community submitted patches.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix TX timeout logic

Fixed the TX timeout logic to use "end of packet" rather than "next to clean".  Updated message log.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix Desc. Rings and Jumbo Frames

This patch contains two fixes.  The first fix is to the tx and rx descriptor rings clean up process.  The second fix is to jumbo frames, which cleans up the code logic and removes most of the fifo related limitations on jumbo frames.  This is because the driver code now supports splitting a packet across multiple descriptors.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix TX queue length based on link speed

10/100 speeds seem to have some problems reporting false tx timeouts especially at half duplex.  Fixed by using a timeout factor to attempt to mitigate the false timeouts.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix flow control water marks

Fixed flow control water marks based on PBA size.
Store flow control state in original_fc in addition to fc.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix EEPROM read logic

Fixed read_eeprom logic to test use_eerd instead of testing for 82573 controllers.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix PHY reset when blocked

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix loopback logic

Fixed the loopback logic to work for the PCI express adapters.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix mulitple queues

Fixed stats when using multiple queues.
When multiple queues are enabled, log a message in syslog.
Fixed memory allocation for multiple queues.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix adapter structure and prepare for multique fix

Fix adapter structure to handle multiple queues and prepping the driver for full multiple queue support, some changes are ifdef'd our unless you define CONFIG_E1000_MQ.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix LED functionality for 82573

Fixed adapter identification issue.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] Fix e1000 stats

Updated the e1000_stats structure and removed mpx for rx_errors and rx_dropped.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix PHY config for 82573 controller

Added a delay to allow PHY configuration to complete before accessing NVM.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix ASF/AMT for 8257{1|2|3} controllers

The 82573 controller required different logic than 82571|2 controllers.  Corrected the reset logic for 8257{1|2|3} controllers.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix SoL/IDER link and loopback

Fix so that if a SoL/IDER session is active, do not allow operations which require a PHY reset and instead log a message.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: General Fixes

These fixes update the TX and RX ring structures.  Prepare driver for up-coming fixes.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix TSO

Fixed the TSO workaround for 82571/2 controllers.
Fixed TSO issue where a non-tso packet in a linear SKB which followed a TSO packet would get written back prematurely.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] e1000: Fix jumbo frame performance

Partition PBA for Jumbo frames based on MTU size.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: John Ronciak <john.ronciak@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] spidernet: fix missing include

This is now required to avoid
drivers/net/spider_net.c:844: error: 'IPPROTO_TCP' undeclared

Signed-off-by: Arnd Bergmann <arndb@de.ibm.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] spidernet: performance optimizations

Performance optimizations, changes in these areas:
  - RX and TX checksum offload
  - correct maximum MTU
  - don't use TX interrupts anymore, use a timer instead
  - remove some superfluous barriers
  - improve RX RAM full handling

From: Utz Bacher <utz.bacher@de.ibm.com>
Signed-off-by: Jens Osterkamp <jens.osterkamp@de.ibm.com>
Signed-off-by: Arnd Bergmann <arndb@de.ibm.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] spidernet: fix HW structures for 64 bit dma_addr_t

The driver incorrectly used dma_addr_t to describe
HW structures and consequently broke when that type
was changed in 2.6.15-rc.

This changed spidernet to use u32 for 32 bit HW defined
structure elements.

From: Jens Osterkamp <Jens.Osterkamp@de.ibm.com>
Signed-off-by: Jens Osterkamp <Jens.Osterkamp@de.ibm.com>
Signed-off-by: Arnd Bergmann <arndb@de.ibm.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] spidernet: read firmware from the OF device tree

request_firmware() is sometimes problematic, especially
in initramfs, reading the firmware from Open Firmware
is much preferrable.

We still try to get the firmware from the file system
first, in order to support old SLOF releases and to allow
updates of the spidernet firmware without reflashing
the system.

From: Jens Osterkamp <Jens.Osterkamp@de.ibm.com>
Signed-off-by: Jens Osterkamp <Jens.Osterkamp@de.ibm.com>
Signed-off-by: Arnd Bergmann <arndb@de.ibm.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] spidernet: check if firmware was loaded correctly

Uploading the device firmware may fail if wrong input data
was provided by the user. This checks for the condition.

From: Jens Osterkamp <Jens.Osterkamp@de.ibm.com>
Signed-off-by: Jens Osterkamp <Jens.Osterkamp@de.ibm.com>
Signed-off-by: Arnd Bergmann <arndb@de.ibm.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: Remove needless mask of extended intr register

All interrupts controlled by the extended mask register are also
masked by a bit in the main mask register, so there is no need to
directly manipulate the extended mask register.

Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
 mv643xx_eth.c |   81 ++++++++++++++++++----------------------------------------
 1 file changed, 26 insertions(+), 55 deletions(-)
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: Merge open and stop helper functions

Move code from helper functions mv643xx_eth_real_open and mv643xx_eth_real_stop
as they are no longer needed.

Signed-off-by Dale Farnsworth <dale@farnsworth.org>

 mv643xx_eth.c |  109 +++++++++++++++++++++++-----------------------------------
 1 file changed, 45 insertions(+), 64 deletions(-)

Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: Fix transmit skb accounting

Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
 mv643xx_eth.c |    7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: Request HW checksum generation only for IPv4

This patch removes the NETIF_F_HW_CSUM flag to be able to use other protocols
than IPv4. Hardware checksums for IPv4 should continue to work because
NETIF_F_IP_CSUM is still set.  The sanity-check has been enhanced to check
the used protocol and to not access skb->iph for non-ipv4-packets.

Signed-off-by: Wolfram Joost <pegasos@frokaschwei.de>
Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
 mv643xx_eth.c |   19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: Hold spinlocks only where needed

This driver has historically held a spin_lock during the entire open
and stop functions and while receiving multiple packets.  This is
unecessarily long and holds locks during calls that may sleep.
This patch reduces the size of windows where locks are held.

Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
 mv643xx_eth.c |  172 ++++++++++++++++++++++++++++++----------------------------
 1 file changed, 91 insertions(+), 81 deletions(-)
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: iounmap the correct SRAM buffer

Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
 mv643xx_eth.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: Fix handling of small, unaligned fragments

Fix handling of small, unaligned fragments.
It also solves a potential deadlock if skb_linearize() returns -ENOMEM.

Signed-off-by: Paul Janzen <pcj@linux.sez.to>
Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
 mv643xx_eth.c |   54 +++++++++++++++++++++++++++++++-----------------------
 1 file changed, 31 insertions(+), 23 deletions(-)
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: Receive buffers require 8 byte alignment

The Marvell mv643xx ethernet hardware requires that DMA buffers be
aligned to 8-byte boundaries.  This patch satisfies this requirement.
Buffers allocated by dev_alloc_skb() only have 4-byte alignment when
slab debugging is enabled.

Also, document that the 2-byte offset to align the IP packets on
receive is a hardware feature and is not tied to NET_IP_ALIGN.

Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
 mv643xx_eth.c |   12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: Add multicast support

This code is adapted from code in a ppc-specific version of the driver.

Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
 mv643xx_eth.c |  201 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 197 insertions(+), 4 deletions(-)
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: Fix a NULL pointer dereference

Fix a NULL pointer dereference.

Fill in the buf_ptr and byte_cnt fields of pkt_info in
eth_tx_return_desc().

Signed-off-by: Paolo Galtieri <pgaltieri@mvista.com>
Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
 mv643xx_eth.c |   51 +++++++++++++++++++++------------------------------
 1 file changed, 21 insertions(+), 30 deletions(-)
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: Fix dma_map/dma_unmap relations

If you do a dma_map_single you must do dma_unmap_single and if you do
a dma_map_page you must do a dma_unmap_page.

Signed-off-by: Paolo Galtieri <pgaltieri@mvista.com>
Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
 mv643xx_eth.c |   51 +++++++++++++++++++++------------------------------
 1 file changed, 21 insertions(+), 30 deletions(-)
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: 2.6.16 needs ip.h and in.h

Signed-off-by: Olaf Hering <olh@suse.de>
Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
 mv643xx_eth.c |    2 ++
 1 file changed, 2 insertions(+)
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

[PATCH] mv643xx_eth: Add Dale Farnsworth as a maintainer

 MAINTAINERS |    8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

Merge branch 'upstream-jgarzik' of git://git.tuxdriver.com/git/wireless-2.6

[IPV4]: rt_cache_stat can be statically defined

Using __get_cpu_var(obj) is slightly faster than per_cpu_ptr(obj,
raw_smp_processor_id()).

1) Smaller code and memory use
For static and small objects, DEFINE_PER_CPU(type, object) is preferred over a
alloc_percpu() : Better and smaller code to access them, and no extra memory
(storing the pointer, and the percpu array of pointers)

x86_64 code before patch

mov    1237577(%rip),%rax        # ffffffff803e5990 <rt_cache_stat>
not    %rax  # part of per_cpu machinery
mov    %gs:0x3c,%edx # get cpu number
movslq %edx,%rdx # extend 32 bits cpu number to 64 bits
mov    (%rax,%rdx,8),%rax # get the pointer for this cpu
incl   0x38(%rax)

x86_64 code after patch

mov    $per_cpu__rt_cache_stat,%rdx
mov    %gs:0x48,%rax # get percpu data offset
incl   0x38(%rax,%rdx,1)

2) False sharing avoidance for SMP :
For a small NR_CPUS, the array of per cpu pointers allocated in alloc_percpu()
can be <= 32 bytes. This let slab code gives a part of a cache line. If the
other part of this 64 bytes (or 128 bytes) cache line is used by a mostly
written object, we can have false sharing and expensive per_cpu_ptr() operations.

Size of rt_cache_stat is 64 bytes, so this patch is not a danger of a too big
increase of bss (in UP mode) or static per_cpu data for SMP
(PERCPU_ENOUGH_ROOM is currently 32768 bytes)

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Make second arg to skb_reserved() signed.

Some subsystems, such as PPP, can send negative values
here.  It just happened to work correctly on 32-bit with
an unsigned value, but on 64-bit this explodes.

Figured out by Paul Mackerras based upon several PPP crash
reports.

Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: ip_conntrack_proto_gre.c needs linux/interrupt.h

Signed-off-by: David S. Miller <davem@davemloft.net>

[TG3]: Refine nvram locking

Add nvram lock count so that calls to tg3_nvram_lock()/unlock() can
be nested. Add error checking to all callers of tg3_nvram_lock()
where appropriate. To prevent nvram lock failures after halting the
firmware, it is also necessary to release firmware's nvram lock in
tg3_halt_cpu().

Update version to 3.48.

Based on David Miller's initial patch.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER] ip6tables: whitespace and indent cosmetic cleanup

Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp>
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER] ip6tables: remove unused definitions

These definitions ware used for only internal use in kernel <= 2.6.13,
which had not introduced the unified parser of IPv6 extension header yet.

Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp>
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER] Makefile cleanup

These are replaced with x_tables matches and no longer exist.

Signed-off-by: Yasuyuki Kozakai <yasuyuki.kozakai@toshiba.co.jp>
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER] ip[6]t_policy: Fix compilation warnings

ip[6]t_policy argument conversion slipped when merging with x_tables

Signed-off-by: Benoit Boissinot <benoit.boissinot@ens-lyon.org>
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Fix whitespace issues in net/core/filter.c

This fixes some whitespace issues in net/core/filter.c

Signed-off-by: Kris Katterjohn <kjak@users.sourceforge.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PKT_SCHED] sch_prio: fix qdisc bands init

Currently when PRIO is configured to use N bands, it lets the packets be
directed to any of the bands 0..N-1. However, PRIO attaches a fifo qdisc
only to the bands that appear in the priomap; the rest of the N bands
remain with a noop qdisc attached. This patch changes PRIO's behavior so
that it attaches a fifo qdisc to all of the N bands.

Signed-off-by: Amnon Aaronsohn <bla@cs.huji.ac.il>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Preserve procfs IPV6 address output format

Procfs always output IPV6 addresses without the colon
characters, and we cannot change that.

Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Linux v2.6.16-rc1

Merge branch 'for-linus' of git://brick.kernel.dk/data/git/linux-2.6-block

[PATCH] Remove unused code from rioctrl.c (Last for this batch of work)

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Remove rio_table.c unused code

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Remove unused code from rio_linux.c

Signed-off-by: Linus Torvalds <torvalds@osdl.org>