On Thu, Mar 25, 2004 at 09:47:34AM +1200, Steve Wray wrote:
> We have a machine with;
> SATA, software raid, LVM2 and XFS filesystems.
> LVM2 is appropriately configured to work on software raid.
> We see corruption on XFS filesystems on real partitions on normal IDE drives
> (not SATA and no software RAID or LVM) as well as on LVM volumes on SATA
> Corruption occurs during normal use, there have been no power or kernel panic
Any chance you have a reproducible test case for hitting this?
The output from xfs_db on that inode (21214761) would be useful
information too. Also the xfs_db superblock dump (or xfs_info
output) would be handy.
> The kernel is 2.6.4
> We are seeing problems with the XFS filesystems like this for example;
> Mar 21 07:00:05 plato kernel: Filesystem "dm-0": corrupt dinode 21214761,
> extent total = 1025, nblocks = 2. Unmount and run xfs_repair.
> Mar 21 07:00:05 plato kernel: 0x0: 49 4e 81 b4 01 02 00 01 00 00 04 10 00 00
> 00 64
> Mar 21 07:00:05 plato kernel: Filesystem "dm-0": XFS internal error
> xfs_iformat(1) at line 475 of file fs/xfs/xfs_inode.c. Caller 0xc0240\