On Mon, Mar 28, 2022 at 05:36:50PM -0700, Darrick J. Wong wrote:
> On Thu, Mar 24, 2022 at 11:21:03AM +1100, Dave Chinner wrote:
> > From: Dave Chinner <dchinner@xxxxxxxxxx>
> >
> > If a shut races with xfs_trans_commit() and we have shut down the
> > filesystem but not the log, we will still cancel the transaction.
> > This can result in aborting dirty log items instead of committing and
> > pinning them whilst the log is still running. Hence we can end up
> > with dirty, unlogged metadata that isn't in the AIL in memory that
> > can be flushed to disk via writeback clustering.
>
> ...because we cancelled the transaction, leaving (say) an inode with
> dirty uncommited changes? And now iflush for an adjacent inode comes
> along and writes it to disk, because we haven't yet told the log to
> stop? And blammo?
.....
> If the answers are {yes, yes, yes} then yikes and:
Yes: yes, yes, yes and yes, yikes!
> Reviewed-by: Darrick J. Wong <djwong@xxxxxxxxxx>
Thanks!
-Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx
