Re: EXT-3 bug with 2.6.0-test9

[Gertjan van Wingerde <gwingerde@home.nl>]

A forced fsck of all my file-systems seems to have cured this. At least its 
not easy to trigger anymore (not even in test9).
I looked something really had gone wrong sometime. I got multiple filesystems
with errors on it (without it had been noticed; all had to have a forced fsck 
not a "normal" one).

But, so far so good.

	Gertjan.

On Tuesday 25 November 2003 21:25, Andrew Morton wrote:
> Gertjan van Wingerde <gwingerde@home.nl> wrote:
> > Hi,
> >
> > (Please CC me in any replies, as I'm not subscribed to the list)
> >
> > I've just experienced the strange behaviour that my /usr mount
> > auto-magically got mounted read-only, where it was mounted read-write
> > (obviously). Investigating the cause of this I've found the following
> > EXT-3 related BUG in my log-files:
> >
> > 	kernel BUG at fs/jbd/journal.c:1733!
>
> Yes.  This newly-added BUG check was too easy to trigger and in test10 it
> was changed to a printk-and-fix-it-up.
>
> > Nov 25 20:24:20 localhost vmunix: EXT3-fs warning (device md1):
> > ext3_unlink: Deleting nonexistent file (230991), 0
> > Nov 25 20:24:21 localhost vmunix: EXT3-fs warning (device md1):
> > ext3_unlink: Deleting nonexistent file (230990), 0
> > Nov 25 20:24:21 localhost vmunix: EXT3-fs warning (device md1):
> > ext3_unlink: Deleting nonexistent file (346096), 0
>
> And it is this stuff which triggered the bogus BUG.  I do not know why this
> happened, but it is probably some form of data loss problem at the md
> layer.
>
> It could have happened at any time after the most recent fsck, so if you
> have been running earlier kernels on that machine it could even be that the
> bug which caused this has already been fixed.
>
> You should run a fsck across all filesystems, and maybe upgrade to test10
> plus
> ftp://ftp.kernel.org/pub/linux/kernel/v2.6/snapshots/patch-2.6.0-test10-bk1
>.gz