Re: quota problem with 2.6.15.7

[Bernd Schubert <bernd-schubert@gmx.de>]

Hello Jan,

On Tuesday 05 September 2006 11:19, Jan Kara wrote:
>   Hello,
>
> > I just wanted to enable quotas on one of our server systems and got an
> > oops. This is an opteron system with a kernel in 64bit mode.
> > As you can see, the filesystem is reiserfs.
>
>   Hmm, is this reproducible? Any chances of trying out some newer
> kernel?

[trace deleted]

>   Hmm, the trace looks strange... It is definitely mixed with some old
> data. We definitely reached reiserfs_quota_on() but didn't reach
> vfs_quota_on() so it seems we crashed somewhere in path_lookup() (also
> link_path_walk() in the beginning of the trace suggests that). That's
> generic VFS code so maybe this is nothing quota specific. So this looks
> quite hard to debug if there's no reasonable way of reproducing it.

Its reproducible, I hoped it wouldn't happen again after a reboot, but 
unfortunately it did. Trying a newer kernel, hmm, usually a problem on server 
systems. I will try to test quotas on another amd64 system first(*).

Thanks for your help,
Bernd

(*) Which would be easy, if there wouldn't be a 64bit kernel, 32bit glibc, 
nfsclient inode truncation bug, which forces me to run all of our amd64 
nfsroot-clients with a 32bit kernel, http://lkml.org/lkml/2005/2/28/172. 
Well, my fault that I still didn't submit a glibc bug report.

-- 
Bernd Schubert
PCI / Theoretische Chemie
Universität Heidelberg
INF 229
69120 Heidelberg