Hi,

I recently upgraded one of our KVM servers from SL 6.0 to 6.1. Now
(yesterday) the machine was under heavy I/O load. I investigated and
saw that all virtual machines with CentOS 5.6 (we switched to SL just a
few months ago) are writing hundreds of thousands of these messages
to /var/log/messages:

Aug  8 13:15:20 infra kernel: : BUG: scheduling while atomic:
swapper/0xffff0000/0 Aug  8 13:15:20 infra kernel: : 
Aug  8 13:15:20 infra kernel: : Call Trace:
Aug  8 13:15:20 infra kernel: : [<ffffffff8006243d>]
__sched_text_start+0x7d/0xbce Aug  8 13:15:20 infra kernel: :
[<ffffffff8006bdb5>] default_idle+0x0/0x50 Aug  8 13:15:20 infra
kernel: : [<ffffffff800494dd>] cpu_idle+0xb6/0xb8 Aug  8 13:15:20 infra
kernel: : [<ffffffff80461807>] start_kernel+0x220/0x225 Aug  8 13:15:20
infra kernel: : [<ffffffff8046122f>] _sinittext+0x22f/0x236 Aug  8
13:15:20 infra kernel: : 

Now I'm not sure if this is a problem of the underlying CentOS kernel
(that one was upgraded on july, 20th) or of SL 6.1 KVM (upgraded
august, 1st). The SL 6.0/6.1 machines don't show this problem. The
5.6 machines didn't under SL 6.0.

Right now I disabled the swap which seems to help, but this can of
course only be a temporary solution.

I also saw that RH released two minor releases of qemu-kvm that aren't
yet in SL; would it make sense to try to compile the newest of them and
update?

If anyone knows something about this problem, any help is greatly
appreciated.

Regards,
Torsten