----- Original Message -----

On Tue, 11 Jan 2011, CAI Qian wrote:

quoted

BTW, the latest linux-next also had the similar issue.

- kswapd was running for a long time.

I'd be interested to see if this is fixed now that 2ff754fa8f41 (mm:
clear
pages_scanned only if draining a pcp adds pages to the buddy
allocator)
has been merged in Linus' tree.

I can still reproduce this similar failure on both AMD and Intel NUMA
systems using the latest linus tree with the commit you mentioned.
Unfortunately, I can't get a clear sysrq/console output of it but only
a part of it (screenshot attached).

It at least very easy to reproduce it for me by running LTP oom01 test
for both Magny-Cours and Nehalem-EX NUMA systems.

CAI Qian