call_pernode_memory() is rounding memory regions to page boundaries,
but in the non-NUMA case, it's incorrectly passing the rounded start and
the original length.  Any reserved memory regions sitting beyond a
memory region that isn't nicely aligned is potentially at risk for
getting freed and trashed.  This is a fairly nasty bug that can bite in
very hard to find ways.  In my case, the EFI memory map was getting
overwritten, causing the ACPI methods to access chipset registers
cachable instead of uncachable.  It could potentially be causing some of
the random hangs and strange behavior we're seeing lately.
Signed-off-by: Alex Williamson <alex.williamson@hp.com>
Signed-off-by: David Mosberger <davidm@hpl.hp.com>

From: Anton Blanchard <anton@samba.org>

../include/linux/swap.h:extern int nr_swap_pages;       /* XXX: shouldn't this be ulong? --hch */

Sounds like it should be too me.  Some of the code checks for nr_swap_pages
< 0 so I made it a long instead.  I had to fix up the ppc64 show_mem() (Im
guessing there will be other trivial changes required in other 64bit archs,
I can find and fix those if you want).

I also noticed that the ppc64 show_mem() used ints to store page counts.
We can overflow that, so make them unsigned long.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

I found a problem with HP NUMA configuration in 2.6.7-rc3-mm1.  The
node_online_map is incorrect because of actions taken by
reassign_cpu_only_nodes in arch/ia64/mm/discontig.c.  The mm1 patch
which changes the mempolicy for the init_task uses this invalid
node_online_map and eventually uses a non-existent node's zonelist
when calling __alloc_pages.

The patch below takes care of this issue.  It puts the non-existent
nodes offline.
Signed-off-by: David Mosberger <davidm@hpl.hp.com>

This patch works around a limitation in the current NUMA code, which
doesn't support CPU-only (memory-less) nodes.  With this patch
CPU-only nodes are moved to a node with memory which is at the closest
relative distance per the SLIT information.  Any node reassignments
will result in the compression of the nodes and renumbering the nid
values where appropriate.

From: "Martin J. Bligh" <mbligh@aracnet.com>

This patch removes memblks from the kernel ...  we don't use them, and the
NUMA API that was planning to use them when they were originally designed
isn't going to use them anymore.  They're just unnecessary added complexity
now ...  time for them to go.

There's a slight complication in that ia64 uses something with a similar
name for part of its memory layout, but Jes Sorensen kindly untangled them
from each other for us.  The patch with his modifications is below.  Jes
tested it on ia64, and I testbuilt it with every config in my arsenal.

The arch-independent bootmem code now requires that arches initialize
their bootmem maps in reverse order (in particular, from high to low
addesses), otherwise alloc_bootmem_pages_low() won't work.  This change
makes the ia64 code do just that, so that machines without an IOMMU can
allocate their bounce buffers in low memory at early boot.  It also adds
a sanity check to the early init code to make sure that each node has a
local data area, because if they don't, many things will break later on
and may be hard to track down.

I stumbled on a couple trivial bugs in ia64 numa/discontig support.
The first just sets the default number of nodes to something reasonable
for a generic kernel, otherwise it's really easy to start walking over
your initdata (more error checking should probably be added).  The
second fixes a memcpy to a physical address.

This patch fixes the combination of CONFIG_DISCONTIGMEM and
CONFIG_VIRTUAL_MEM_MAP so that generic kernels will work on all ia64
platforms, including sn2, and also makes sn2 specific kernels work
(which I think is a first).

I've cleaned this patch up heavily based on feedback from yourself,
Christoph and others.  I've tested sn2, zx1, and dig (thanks Xavier!)
specific configurations, as well as a generic configuration that allows
the same binary to boot on zx1, dig, and sn2.

From: Matthew Dobson <colpatch@us.ibm.com>

Fix up the ia64 arch.

	from kernel/setup.c into two separate files: mm/{dis,contig}.c
	to handle contiguous vs. discontiguous memory layouts.

Attached is a patch for interleaved discontigmem banks.

When memory banks are interleaved between nodes,
bank ids can be overwritten by other nodes.

    doesnt -> doesn't (35 occurrences)

	Triggered by analysis done by Philipp Marek.

Here is the latest discontigmem patch for ia64 against
2.5.39 + ia64 patch + Erich's acpi_numa patch.