I stumbled across this the other day. The block layer only uses a single
memory pool for request allocation, so it's very possible for eg writes
to have allocated them all at any point in time. If that is the case and
the machine is low on memory, a reader attempting to allocate a request
and failing in blk_alloc_request() can get stuck for a long time since
no one is there to wake it up.

The solution is either to add the extra mempool so both reads and writes
have one, or attempt to handle the situation. I chose the latter, to
save the extra memory required for the additional mempool with
BLKDEV_MIN_RQ statically allocated requests per-queue.

If a read allocation fails and we have no readers in flight for this
queue, mark us rq-starved so that the next write being freed will wake
up the sleeping reader(s). Same situation would happen for writes as
well of course, it's just a lot more unlikely.
Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

"ATA over Ethernet support" should not default to 'm', it doesn't make
any sense for a special case driver to do so.
Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

A recent trivial fixup in sys_getdents64 gives gcc-2.96 indigestion in
the form of an ICE.

While upgrading to a sane gcc would be the preferred solution, rewriting
the change as follows eliminates the error for those who cannot do so.
Signed-off-by: Adam Kropelin <akropel1@rochester.rr.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

This broke since the recent MODULE_PARAM conversion on
architectures that don't have CONFIG_MTRR
Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

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

into flint.arm.linux.org.uk:/usr/src/bk/linux-2.6-rmk

Patch from Nicolas Pitre

This patch adds support for all alignments to both endianness.

Signed-off-by: Nicolas Pitre 
Signed-off-by: Russell King

Patch from Nicolas Pitre

Now that MTD XIP support is merged this part is
not relevant anymore.

Signed-off-by: Nicolas Pitre
Signed-off-by: Russell King

Signed-off-by: Russell King

Patch from Ben Dooks

MAINTAINERS entries for currently supported Simtec Electronics
development boards.

Signed-off-by: Ben Dooks
Signed-off-by: Russell King

Patch from Catalin Marinas

The patch adds the "ax" attributes to the .sched.text section to
avoid a compiler warning.

Signed-off-by: Catalin Marinas
Signed-off-by: Russell King

Patch from Richard Purdie

The Sharp SL-C7xx Series (Corgi) has 3 devices connected to the SSP
interface each needing different configurations of the port.
This code provides the necessary access and locking so drivers can
access these components. It uses the functions provided by the PXA
SSP driver to access the port.
It also adds some machine specific GPIO definitions used by this
code and adds some comments to existing definitions.

Signed-off-by: Richard Purdie
Signed-off-by: Russell King

asm/processor.h doesn't use atomic operations nor types, so
there's no need to include asm/atomic.h.
Signed-off-by: Russell King <rmk@arm.linux.org.uk>

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

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

From: 	James.Smart@Emulex.Com

This patch adds 5 more FC transport host attributes in support of HBAAPI. 
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

From: 	James.Smart@Emulex.Com

This patch deprecates the use of device_for_each_child() with stargets.
The reasoning behind this is due to issues regarding:
      Semaphores that device_for_each_child() takes
      Implicit assumptions that each child is an sdev device.

The patch adds a new helper function, starget_for_each_device(), and
replaces all previous uses of device_for_each_child().
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

From: Jesse Barnes <jbarnes@engr.sgi.com>

There are a few spots in qla1280.c that don't need a full PCI write flush
to the device, but rather a simple write ordering guarantee.  This patch
changes some of the PIO reads that cause write flushes into mmiowb calls
instead, which is a lighter weight way of ensuring ordering.
Signed-off-by: Jeremy Higdon <jeremy@sgi.com>
Signed-off-by: Jesse Barnes <jbarnes@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

From: Adrian Bunk <bunk@stusta.de>

The patch below kills kernel 2.2 #ifdef's from the SCSI aic7xxx driver.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

This is really part of the earlier changeset from David to add the
new machine vector to support certain limited range DMA cards on
zx1.  I just forgot to run "bk new" before the commit, so the newly
added files weren't checked into BK.
Signed-off-by: David Mosberger-Tang <davidm@hpl.hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>

In ChangeSet 1.2370, 2005/01/11 17:41:32-08:00, tglx@linutronix.de wrote:
>
>         [PATCH] ppc: remove duplicate define
>        
>         The MMCR0_PMXE is already defined in reg.h...

Er, no it's not. But perhaps it should be...

Ingo presumably didn't notice that ppc64 already had a functional
_raw_read_trylock when he added the #define to use the generic
version.  This just removes the #define so we use the ppc64-specific
version again.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Currently we only print the default io scheduler if the kernel chooses,
not if the user/bootloader has specified one. This patch saves the extra
line in dmesg but always notified of the default choice by appending
(default) to that line:

	..
	io scheduler noop registered
	io scheduler anticipatory registered
	io scheduler deadline registered
	io scheduler cfq registered (default)
	..

Patch originally from Srihari Vijayaraghavan, modified by me.
Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

into flint.arm.linux.org.uk:/usr/src/bk/linux-2.6-rmk

Signed-off-by: Tony Luck <tony.luck@intel.com>

The accounting can go bad in the requeue hook, it must check the
accounted flag to make sure it was previously considered in the driver.
Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Signed-off-by: Keith Owens <kaos@sgi.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

This apparently explains some weird IO failures reported in last two months.
Only non-bwx (including generic) kernels were affected.
Acked-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

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

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

Patch from Richard Purdie

Tidy up a couple of coding style issues in the Sharp SCOOP Driver

Signed-off-by: Richard Purdie
Signed-off-by: Russell King

This enables it only for debug kernels, and also makes sure
that if some external module is still broken, we don't leave
the mmap-sem locked after warning about it.

Patch from Deepak Saxena

We should be including <linux/kernel.h> to pick up various important
constants that might be needed by other include files.

Signed-off-by: Deepak Saxena
Signed-off-by: Russell King

Patch from Deepak Saxena

Signed-off-by: Deepak Saxena
Signed-off-by: Russell King

Patch from Ben Dooks

This cleans up a few items in arch/arm/mach-s3c2410
with naming of functions and a build problem. Items
which are general to s3c2410 and s3c2440 are now
named s3c24xx_ instead of s3c2410_, as well as
moving them to the correct headers.
The patch also fixes a problem where at least one
s3c2410 target had to be selected to allow an s3c2440
target to build without error.
The following have been renamed:
	s3c2410_init_irq -> s3c24xx_init_irq
	s3c2410_timer -> s3c24xx_timer

Signed-off-by: Ben Dooks
Signed-off-by: Russell King

It seems to be general consensus that its safer to require all do_brk() callers
to grab mmap_sem, and have do_brk to warn otherwise. This is what the following
patch does.

Similar version has been changed to in v2.4.
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

We had all the locking right, but we didn't check whether one of the
threads now no longer needed to expand, so we could incorrectly _shrink_
the stack in the other thread instead (not only causing segfaults, but
since we didn't do a proper unmap, we'd possibly leak pages too).

So re-check the need for expand after getting the lock.

Noticed by Paul Starzetz.

Remove asm/smp.h include, use linux/smp.h instead.
Add linux/cpu.h include.

This removes ipi_count from the globally visible per-cpu cpu data
structure.