[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: That darned orphaned socket hang



Hi,

On Wed, Feb 14, 2001 at 09:06:26PM -0800, Peter J. Braam wrote:
> It also happens on local disk file systems..

Nailed it.

There were two separate bugs involved.  One was Andreas's original
orphan code: it was reading information about the orphan list outside
of the superblock lock and retaining that info once the lock was held.
The other was mine: the recovery code was reading the revoke blocks
wrongly, and if it read beyond the end of the valid revoke entries
then it was likely to encounter a zero and end up revoking the
superblock itself (which is where the orphan list head is held).

Ted, the superblock revoke also explains that ext2fs summary weirdness
we were seeing.  I'll also send you the one-liner for recovery.c in
e2fsprogs where the same bug is present at the moment.

ext3-0.0.6b will be out today to fix this.  I've been doing a heap of
recovery testing over the past 2-3 weeks getting this sorted, and the
two bugs above explain every single oddity I've seen in that time, so
there are no more outstanding known bugs in the way of a codefreeze.  

Cheers,
 Stephen





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]