From: Brian Gerst <bgerst@didntduck.org>

We don't need to keep the pointer array around after the caches are
initialized.  This doesn't affect the actual strings.
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

From: Jakub Jelinek <jakub@redhat.com>

FUTEX_REQUEUE operation has been added to the kernel mainly to improve
pthread_cond_broadcast which previously used FUTEX_WAKE INT_MAX op.
pthread_cond_broadcast releases internal condvar mutex before FUTEX_REQUEUE
operation, as otherwise the woken up thread most likely immediately sleeps
again on the internal condvar mutex until the broadcasting thread releases it.

Unfortunately this is racy and causes e.g.
http://sources.redhat.com/cgi-bin/cvsweb.cgi/libc/nptl/tst-cond16.c?rev=1.1&content-type=text/x-cvsweb-markup&cvsroot=glibc
to hang on SMP.

http://listman.redhat.com/archives/phil-list/2004-May/msg00023.html contains
analysis how the hang happens, the problem is if any thread does
pthread_cond_*wait in between releasing of the internal condvar mutex and
FUTEX_REQUEUE operation, a wrong thread might be awaken (and immediately go to
sleep again because it doesn't satisfy conditions for returning from
pthread_cond_*wait) while the right thread requeued on the associated mutex
and there would be nobody to wake that thread up.

The patch below extends FUTEX_REQUEUE operation with something FUTEX_WAIT
already uses:

FUTEX_CMP_REQUEUE is passed an additional argument which is the expected value
of *futex.  Kernel then while holding the futex locks checks if *futex !=
expected and returns -EAGAIN in that case, while if it is equal, continues
with a normal FUTEX_REQUEUE operation.  If the syscall returns -EAGAIN, NPTL
can fall back to FUTEX_WAKE INT_MAX operation which doesn't have this problem,
but is less efficient, while in the likely case that nobody hit the (small)
window the efficient FUTEX_REQUEUE operation is used.
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

From: David Gibson <david@gibson.dropbear.id.au>

Currently the iSeries virtual ethernet driver has no Tx watchdog timer. 
This makes it vulnerable to clagging up if the other end of connection is
misbehaving - in particular if it is not giving timely hypervisor level
acks to our data frams.

This patch adds a watchdog timer which resets the connection to any lpar we
seem to be having trouble sending to.  With any luck the other end might
behave better after the reset.  If not, this will at least unclag the queue
for a while so we can keep talking to the lpars which are behaving
correctly.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

From: Anton Blanchard <anton@samba.org>

Legacy iseries has problems with the bolted vmalloc patch.  This patch
disables the optimisation on iseries and creates a slb_add_bolted helper
function.

Also, while we require all SLB entries to be context switched, we werent
informing the iseries hypervisor.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

From: Anton Blanchard <anton@samba.org>

Even though we have a variable containing the number of entries in the SLB,
we hardwire the value at boot.  We also dont use the variable in the head.S
fastpath handler.

This patch removes it.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

From: Anton Blanchard <anton@samba.org>

Based on some profiles we noticed the first vmalloc region was being
continually cast out and replaced.  All modules end up there so it is one
of our hottest segments.

This patch bolts the vmalloc region into the second segment.  SLB misses on
an NFS benchmark were reduced by about 10% with this patch.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

From: Anton Blanchard <anton@samba.org>

Reset the progress indicator on iseries after boot.
Signed-off-by: John Engel <engel@us.ibm.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

From: Anton Blanchard <anton@samba.org>

Implement eeh_add_device_early and eeh_add_device_late, required for
the ppc64 PCI hotplug code.
Signed-off-by: Linas Vepstas <linas@us.ibm.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

From: Stephen Rothwell <sfr@canb.auug.org.au>

This patch fixes the virtual ethernet driver so that it will not block the
transmit queue indefinitely.  This patch appplies on top of the previous
patch from Rusty that removed skb_clone.  There is one white space fix in
hte middle of this - I hope that doesn't offend :-).
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

From: Stephen Rothwell <sfr@canb.auug.org.au>

This has only been adjusted by me to apply after David Gibson's patch to
add a watchdog timer.

The iseries_veth driver does skb_clone, it should only need to skb_get,
which is cheaper.  Should help performance a little.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

From: Stephen Rothwell <sfr@canb.auug.org.au>

This patch brings the iSeries default config up to date and changes some of
the options to what I use.  These are more sensible options (at least in my
opinion :-)).  Please apply.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

From: Jurriaan <thunder7@xs4all.nl>

I've used this patch since 2.6.5, and other users have confirmed it
solves their problems.

Basically, when not in 8bpp mode, radeonfb should use the palette when
clearing a region on screen.  This is how it's done in 2.4 and xfree, also.
Signed-off-by: Jurriaan Kalkman <thunder7@xs4all.nl>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

This fixes a couple of bugs in arch/ppc64/kernel/prom.c.  We were
missing a couple of places where we needed to use RELOC().  I added the
RELOC in one case, and in the other, moved the variable that we were
accessing onto the stack (and reduced its size).  (We use the variable
to get a property value, but we aren't interested in the value, just in
whether the property exists or not.  Since we pass the size of the
variable to the OF getprop call, it won't overflow.) The effect of
missing the RELOCs would be that random memory locations get used on IBM
pSeries systems (possibly causing random boot failures). 

The other thing that this does is add a linux,phandle property to each
node, containing the phandle for the node, which is the token that OF
uses to identify the node.  Some nodes reference other nodes by means
of their phandle.  Without the linux,phandle property, userspace code
looking at the OF device-tree image in /proc/device-tree has no way of
knowing which other node is being referenced.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

kthreads are not just for breakfast anymore.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (creator)
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

This fixes compilation of x86-64 without CONFIG_NUMA again (got broken
by the previous patchkit)

Older versions of gcc were unhappy with our previous trick, and
just separating out the __CHECKER__ case made it much simpler.

On i386, stack usually grows with "sub $0x8,%esp" and shrinks with
"add $0x8,%esp" respectively.  In some cases, though, stack grows with
"add $0xffffff80,%esp" and shrinks with "sub $0xffffff80,%esp".
Obviously, we don't want to miss those cases.

Since in either case add and sub seem to be balanced and contain the
same parameter, we don't need a second regex.  We simply accept hex
numbers of up to 8 digits and treat them as negative numbers when the
sub appears to be a little too high.

...or so I thought.  But another day of testing proved me wrong again.

Some functions do stuff like "sub $0x10,%esp", ..., "add $0x20,%esp".
In other words, add and sub are *NOT* balanced.  Manual inspection
showed that 0x20 is a more realistic number, so I accept either
variant, just in case.  We pay for this with a bunch of duplicates in
our output, but that beats missing some stack hogs.

In the long run, this script has to be replaced by gcc options,
really.  Looking at the result and guessing back is such a stupid
idea.
Signed-off-by: Jörn Engel <joern@wohnheim.fh-wedel.de>

into ppc970.osdl.org:/home/torvalds/v2.6/linux

This patch improves some of the handling of PA-RISC tulip cards.

 - Introduce HAS_SWAPPED_SEEPROM and NEEDS_FAKE_MEDIA_TABLE
 - Only trigger this code on GSC machines.  The pure PCI machines don't
   have these cards.
 - Allow the chip_name to be overridden in tulip_init_one().
 - Fix some indentation.
 - Handle the output from tulip_read_eeprom() better.

> From: Ralf Baechle <ralf@linux-mips.org>
> Subject: [PATCH] Stop queue on close in hdlcdrv
> To: Jeff Garzik <garzik@gtf.org>
> Date: Tue, 4 May 2004 13:59:15 +0200
>
> The stop method of a driver should ensure queueing is stopped ...
>
> diff -u -r1.19 hdlcdrv.c
> --- suckage/drivers/net/hamradio/hdlcdrv.c 12 Apr 2004 20:23:32 -0000

the above means the following is needed for my drivers - please apply to 2.6:

i386 gets it implicitly from somewhere but some architectures don't.
Also move <asm/uaccess.h> down so we have asm after linux headers.

s2io_rem_nic() is marked __exit and is being referred to from .data.
Signed-off-by: Andrew Morton <akpm@osdl.org>

From: "Luiz Fernando N. Capitulino" <lcapitulino@prefeitura.sp.gov.br>

drivers/net/ixgb/ixgb_main.c: In function `ixgb_intr':
drivers/net/ixgb/ixgb_main.c:1593: warning: unused variable `hw'
(catch by J. Cherry).

This happens because `hw' is only used when CONFIG_IXGB_NAPI is not set.
As `hw' is used only to have the code readable, we can use it for
!CONFIG_IXGB_NAPI too.
Signed-off-by: Andrew Morton <akpm@osdl.org>

into redhat.com:/spare/repo/net-drivers-2.6

into ppc970.osdl.org:/home/torvalds/v2.6/linux

into ppc970.osdl.org:/home/torvalds/v2.6/linux

into nuts.davemloft.net:/disk1/BK/sparc-2.6

into kernel.bkbits.net:/home/davem/bluetooth-2.6

Some RTX Telecom based USB dongles offer SCO support, but their
implementation is broken. This patch disables the use of the ISOC
interface for these devices.

If no channel is specified for a RFCOMM server socket, it will search
for the next free one and automaticly bind to it.

Proposed by Stephen Crane <steve.crane@rococosoft.com>

If no PSM is specified for a L2CAP server socket, it will search
for the next free one and automaticly bind to it.

Proposed by Stephen Crane <steve.crane@rococosoft.com>

This patch allocates the next free protocol number for the upcoming
HIDP support.

This patch updates the address of the BlueZ project and changes
some wording of the help entries.

Since we are using Bitkeeper the CVS tags are useless. This patch
removes them and makes some whitespace cleanups.

With Linux 2.6 there is no need to keep the function exports in a
central file and thus the use of syms.c makes no real sense. Remove
it and add the EXPORT_SYMBOL calls to the functions they belong to.

into ppc970.osdl.org:/home/torvalds/v2.6/linux

The timing calculations used by the PXA platforms did not always take
account of the timings handed to us by PCMCIA, and where it did, it
assumed IO timings for memory windows.

We fix this, and provide a generic function which calculates the
required timings (in nanoseconds) for IO, memory and attribute
windows.  The SOC drivers only have to convert this information to
whatever format the hardware requires to achieve at least these
timing parameters.

into ppc970.osdl.org:/home/torvalds/v2.6/linux