On (22/01/08 14:57), Christoph Lameter didst pronounce:

On Tue, 22 Jan 2008, Mel Gorman wrote:

quoted

quoted

quoted

Whatever this was a problem fixed in the past or not, it's broken again now
:( . It's possible that there is a __GFP_THISNODE that can be dropped early
at boot-time that would also fix this problem in a way that doesn't
affect runtime (like altering cache_grow in my patch does).

The dropping of GFP_THISNODE has the same effect as your patch. 

The dropping of it totally? If so, this patch might fix a boot but it'll
potentially be a performance regression on NUMA machines that only have
nodes with memory, right?

No the dropping during early allocations.,

We can live with that if the machine otherwise survives during tests.
They are kicked off at the moment with CONFIG_SLAB_DEBUG set but the point
is moot if the patch doesn't work for Olaf. Am still waiting to hear if
the two patches in combination work for him.

quoted

o 0
o 2
Nodes with regular memory
o 2
Current running CPU 0 is associated with node 0
Current node is 0

So node 2 has regular memory but it's trying to use node 0 at a glance.
I've attached the patch I used against 2.6.24-rc8. It includes the revert.

We need the current processor to be attached to a node that has 
memory. We cannot fall back that early because the structures for the 
other nodes do not exist yet.

Or bodge it early in the boot process so that a node with memory is
always used.

quoted

Online nodes
o 0
o 2
Nodes with regular memory
o 2
Current running CPU 0 is associated with node 0
Current node is 0
 o kmem_list3_init

This needs to be node 2.

Rather it should be 2. I'll admit the physical setup of this machine is
.... less than ideal but clearly it's something that can happen even if
it's a bad idea.

quoted

[c0000000005c3b40] c0000000000dadec .cache_grow+0x7c/0x338
[c0000000005c3c00] c0000000000db54c .fallback_alloc+0x1c0/0x224

Fallback during bootstrap.

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab