From: Martin Schwidefsky <schwidefsky@de.ibm.com>

This fixes a problem in sys_swapon that can cause the creation of invalid
swap ptes.  This has its cause in the arch-independent swap entries vs. 
the pte coded swap entries.  The swp_entry_t uses 27 bits for the offset
and 5 bits for the type.  In sys_swapon this definition is used to find how
many swap devices and how many pages on each device there can be.  But the
swap entries encoded in a pte can be subject to additional restrictions due
to the hardware besides the 27/5 division of the bits in the swp_entry_t
type.  This is solved by adding pte_to_swp_entry and swp_entry_to_pte calls
to the calculations for maximum type and offset.

In addition the s390 swap pte division for offset/type is changed from 19/6
bits to 20/5 bits.

Two callsites, 48 bytes saved

Two callsites, 456 bytes saved

Four callsites, 104 bytes saved

Three callsites, 1104 bytes saved.

Seven callsites and an out-of-line copy is a bit excessive.  562 bytes saved.

Display number of slab, mapped and pagetable pages in the sysrq-M output.

If someone runs page_address() before page_address_init(), the kernel locks
up over uninitialised spinlocks.

This only happens with the 4:4 patch, but it is more robust to run
page_address_init() before setup_arch().  page_address_init() simply
initialises statically allocated storage.

From: Chris Mason <mason@suse.com>

I think Andrew and I managed to mismerge the loop setup race fix. 
loop_set_fd is using get_capacity() to read the size of the disk and
sending that to bd_set_size.

But, it is doing this before calling set_capacity, so the size being used
is wrong.  This should clean things up.

From: David Mosberger <davidm@napali.hpl.hp.com>

Below is a warmed up version of a patch originally done by Werner Almesberger
(see http://tinyurl.com/25zra) to replace the MAX_MAP_COUNT limit with a
sysctl variable.  I thought this had gone into the tree a long time ago but
alas it has not and as luck would have it, the hard limit bit someone today
once again with a large app on a large machine.

Here is a small test app:

Spotted by Andrea: we need the barriers in there to prevent reads passing
ahead of the setting of current->state.

All architectures now make this `long', so we can remove the arch override.

From: Andy Whitcroft <apw@shadowen.org>

Two problems:

a) The memory overcommit code fails oto take into account all the pages
   which are pinned by being reserved for the hugetlbpage pool

b) We're performing overcommit accounting and checking on behalf of
   hugetlbpage vmas.

The main thrust is to ensure that VM_ACCOUNT actually only gets set on
vma's which are indeed accountable.  With that ensured much of the rest
comes out in the wash.  It also removes the hugetlb memory for the
overcommit_memory=2 case.

From: Marcus Meissner <meissner@suse.de>

After discussion on the glibc list the result was that=20 si_band is "long
int" according to POSIX:

	http://www.opengroup.org/onlinepubs/007904975/basedefs/signal.h.html

Ulrich Drepper refused a patch to fix glibc due to this reason:
	http://sources.redhat.com/ml/libc-alpha/2004-03/msg00254.html

so here is the patch to fix it in the kernel.

ppc64 and s390x were broken before and are fixed by this patch too.

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

This patch adds a debugging message to the ppc64 hugepage code when we
attempt to open the "low" (32-bit) hugepage window on PPC64, but can't
because a (non-hugepage) mapping already exists in the region.

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

On PowerPC64 the "low" hugepage range (at 2-3G for use by 32-bit processes)
needs to be activated before it can be used.  hugetlb_get_unmapped_area()
automatically activates the range for hugepage mappings in 32-bit processes
which are not MAP_FIXED.  However for MAP_FIXED mmap()s, even at a suitable
address will fail if the region is not already activated, because there is
no suitable callback from the generic MAP_FIXED code path into the arch
code.

This patch corrects this problem and allows PPC64 to do MAP_FIXED hugepage
mappings in the low hugepage range.

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

Due to a misunderstanding of pmd_offset() the PPC64 hugepage code could end
up looking at bogus pages as if they were PMD pages.

From: Anton Blanchard <anton@samba.org>

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

This creates DMA_ERROR_CODE and uses it everywhere instead of
PCI_DMA_ERROR_CODE as we really want the three DMA mapping API's to return
a single error code.  Also we now have dma_mapping_error and
vio_dma_mapping_error - and this latter and pci_dma_mapping_error both just
call the former.

Also a small fix in the vscsi - dma_map_sg returns 0 to indicate an error.

From: Matt Porter <mporter@kernel.crashing.org>

This fixes the build on non cache coherent PPC32 platforms.

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

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

into austin.ibm.com:/shaggy/bk/jfs-2.5

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

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

estimate.  We must add in for the worst case of a log stripe taking us the
full distance for a log stripe boundary.

SGI Modid: xfs-linux:xfs-kern:169304a

SGI Modid: xfs-linux:xfs-kern:169300a

SGI Modid: xfs-linux:xfs-kern:169208a

else we panic.

SGI Modid: xfs-linux:xfs-kern:169200a

SGI Modid: xfs-linux:xfs-kern:169199a

SGI Modid: xfs-linux:xfs-kern:169135a

SGI Modid: xfs-linux:xfs-kern:169048a

SGI Modid: xfs-linux:xfs-kern:169038a

SGI Modid: xfs-linux:xfs-kern:168809a

SGI Modid: xfs-linux:xfs-kern:168693a

feature bit in sb_versionnum to use to indicate that the new feature bit
field is to be used.

SGI Modid: xfs-linux:xfs-kern:168665a

Add XFS_ALLOCFREE_LOG_RES to IFREE log reservation.

SGI Modid: xfs-linux:xfs-kern:168597a

SGI Modid: xfs-linux:xfs-kern:168167a

previously saved FSTRANS state.  Otherwise we can lose process flags.

SGI Modid: xfs-linux:xfs-kern:168082a