On Sat, Sep 29, 2012 at 09:37:52AM +1000, Dave Chinner wrote: > On Fri, Sep 28, 2012 at 08:40:38AM -0400, Christoph Hellwig wrote: > > > > Looks good. > > > > Do you have any testcase for this? It doesn't look like an issue found > > by code inspection. > > Yeah, test 182 with my xfssyncd-die-die-die patchset triggers it > 100% reliably. FWIW, I've occasionally seen unreproducable recovery issues that could have been caused by this for many years now. This is effectively correcting a zero-day bug in the log behaviour. Here's the commit that introduced this behaviour: http://oss.sgi.com/cgi-bin/gitweb.cgi?p=archive/xfs-import.git;a=commitdiff;h=99b58644923246c710c6dc9f7b47bcae66ed3c63 Dated 4th May, 1994. It's definitely a contender for the oldest bug we've ever found in the XFS code base... Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx _______________________________________________ xfs mailing list xfs@xxxxxxxxxxx http://oss.sgi.com/mailman/listinfo/xfs
