On Fri, Dec 14, 2018 at 08:03:36AM +0100, Sinisa wrote: > On 12/13/18 11:05 PM, Dave Chinner wrote: > >On Wed, Dec 12, 2018 at 01:29:49PM +0100, Sinisa wrote: > >Basically, this looks and smells like a MD sync barrier race > >condition, not an XFs problem. > > But why don't we see the same issue with other filesystems? XFS has a lot more parallelism at the storage layer than other filesystems and has a different integrity synchronisation model (via IO completion processing rather than submission serialisation), so it stresses the underlying storage very differently to other filesystems. Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx
