This fixes a locking problem noted by Tim Hockin:
    * /proc/swaps uses seq_file code, calling seq_path() with swaplock held
    * seq_path() calls d_path()
    * d_path() calls mntput() which might_sleep()

We add a new semaphore protecting insertions/removals in the set of swap
components + switch of ->start()/->stop() to the same semaphore [fixes
deadlocks] + trivial cleanup of ->next().

Early POWER5 revisions (<DD2.1) have a problem requiring slbie
instructions to be repeated under some circumstances.  The patch below
adds a workaround (patch made by Anton Blanchard).

This fix is from Paul Mackerras and was applied in 2.4 sometime late
last year.

On ppc64, touching addresses between the highest other mapping and the
stack can cause the stack to be extended way, way down, rather than
causing a SEGV as you would expect.  This patch only allows the stack
mapping to be extended to cover addresses actually within the stack
(as determined by looking at the process's r1).  This fix is ported
from 2.4

This fixes failures on the LTP's shmdt01, munmap01 and munmap02 tests.

In ide_init_default_irq() patch I overlooked that ppc_ide_md.init_hwif_ports()
called from generic ide_init_hwif_ports() can set hwif->irq and it will
be overwritten by ide_init_default_irq() if CONFIG_PCI is defined.  Fix
it.

I will clean it up properly later after killing ide_init_hwif_ports() on
ARM{26}.  Doing it now is just wasted effort.

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

Patch from Dirk Behme

Make ALTERA Excalibur work again in 2.6.4. Update serial driver uart00.c and defconfig.

This is an update of patch 1833/1.

Remove #include <config/pld/hotswap.h>, it isn't necessary at all. 

atomic_dec_and_test and atomic_add_negative didn't mark their
temporary variables as early-clobber.  Fix this.

This allows us to appropriately align shared mappings on VIPT caches
with aliasing issues.

Use named initialisers for monspecs; the format of the structure
changed a while back and it broke.

Use flush_cache_vmap() after creating mappings.  Also use BUG_ON()
rather than if() BUG().

These broke when find_first_bit/find_next_bit was added.

Patch from David Vrabel

Fix a trivial typo in a dma_unregister_dev printk.

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

a ppc64 tree.

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

More functions and data that should be static.

Thanks to Dru <andru@treshna.com>, who provided an easy way to reproduce
the problem.

What we have in lib/rwsem.c:__rwsem_do_wake():
	int woken, loop;
	^^^
and several lines below:
	loop = woken;
	woken *= RWSEM_ACTIVE_BIAS-RWSEM_WAITING_BIAS;
	woken -= RWSEM_ACTIVE_BIAS;

However, rw_semaphore->count is 64-bit on Alpha, so
RWSEM_WAITING_BIAS has been defined as -0x0000000100000000L.
Obviously, this blows up in the write contention case.

Fix new ISDN CAPI's internal ncci list semaphore if
CONFIG_ISDN_CAPI_MIDDLEWARE is disabled.

Thanks to Florian Schirmer.

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

into sgi.com:/source2/xfs-linux-2.6

SGI Modid: xfs-linux:xfs-kern:170489a

SGI Modid: xfs-linux:xfs-kern:170509a

SGI Modid: xfs-linux:xfs-kern:170448a

From: Tom Rini <trini@kernel.crashing.org>

This patch updates support for the Motorola PrPMC750 platform.  Most of the
size in this patch comes from merging prpmc750_pci.c and prpmc750_setup.c into
just prpmc750.c.

From: Tom Rini <trini@kernel.crashing.org>

This patch adds openpic_hookup_cascade(offset, name, handler) which allows for
an arbitrary interrupt controller to be hooked up as a cascade to the openpic.
 This also allows for platforms to just not have a cascaded controller.

From: Tom Rini <trini@kernel.crashing.org>

This patch updates support for the SBS K2 platform.  Most of the size in this
patch comes from merging k2_pci.c and k2_setup.c into just k2.c.