xfs_trans_log_inode: OOPS in 2.6.28.7

Mon Apr 6 12:40:31 CDT 2009

On Thu, Apr 02, 2009 at 02:47:44AM +0000, Christopher Layne wrote:
> While I think I might be able to reproduce this it's not something I
> particularly want to reproduce. The precursor was cancelling an
> already running xfs_fsr, removing the .fsr_last, and immediately
> restarting it (forcing a new cycle). Right after I received an OOPS and
> all file system access was blocked necessitating a hard reboot and
> xfs_repair.

The a5a5a5 in there looks like slab poisoning.  I'll try to come up with
a testcase based on your above report to try to reproduce it.