On Sun, Feb 16, 2003 at 07:12:01PM +0100, Christian Guggenberger wrote:
> Hi all,
>
> with recent cvs Kernels (namely with
> SGI XFS snapshot 2.4.20-2003-01-14_00:43_UTC with ACLs, quota, no debug
> enabled )
> I see, after a week of heavy load, one or often more processes stuck in D
> state.
> All that processes are acessing nfs-mounts, such as : (Output of ps -efl)
>
> 000 D abc12345 921 1 0 69 0 - 404 lock_p Feb14 ? 00:00:00
> md5sum -v -c sums
>
> I have also machines accessing the same nfs-share (running 2.4.19-xfs-cvs
> of October) with no problems at all.
> I don't want to blame xfs here directly, but I like to know if someone of
> you guys is seeing a similiar behavour...
>
I have been seeing a similar issue in a couple of the QA tests.
I've traced my issue back to a recent regression on the XFS IO
path - I expect to have a fix for that checked in later today;
hopefully this will be the fix for the problem you're hitting
as well (I'm not using NFS at all here though).
cheers.
--
Nathan
|