Bug#533616: linux-image-2.6.29-2-amd64: occasional ext3 filesystem corruption
severity 533616 important
tags 533616 moreinfo
thanks
On Fri, Jun 19, 2009 at 12:38:12PM +0200, Matijs van Zuijlen wrote:
> Yesterday, I found my root filesystem mounted read-only. Dmesg gave the
> following messages (retyped by hand, which is why the timestamps are
> missing):
>
> EXT3-fs error (device sda3): htree_dirblock_to_tree: bad entry in directory
> #91119000: directory entry accross blocks - offset=0, inode=2364050278,
> rec_len=36552, name-len=216
> Aborting journal on device sda3.
> Remounting filesystem read-only
> __journal_remove_journal_head: freeing b_comitted_data
>
> Note the ridiculously large inode number mentioned in the first line.
>
> After reboot, I needed to run fsck from the root prompt. An orphaned inode
> list was fixed, a zero dtime was fixed, block bitmap differences were
> fixed, free block counts were fixed, free inode counts were fixed. I seem
> to recall inode bitmap differences were also fixed.
Log from the fsck run? Some of the differences are normal if the journal
got aborted. But overall this looks like bad hardware, most likely
memory.
Bastian
Reply to: