On Wednesday, 25 October 2006 10:47, Pavel Machek wrote:
> On Wed 2006-10-25 18:38:30, David Chinner wrote:
> > On Wed, Oct 25, 2006 at 10:10:01AM +0200, Pavel Machek wrote:
> > > > Hence the only way to correctly rebuild the XFS state on resume is
> > > > to quiesce the filesystem on suspend and thaw it on resume so as to
> > > > trigger log recovery.
> > >
> > > No, during suspend/resume, memory image is saved, and no state is
> > > lost. We would not even have to do sys_sync(), and suspend/resume
> > > would still work properly.
> > It seems to me that you ensure the filesystem is synced to disk and
> > then at some point later you record the memory state of the
> > filesystem, but these happen at different times. That leaves a
> > window for things to get out of sync again, right?
> I DO NOT HAVE TO ENSURE FILESYSTEM IS SYNCED. That sys_sync() is
> Recording of memory state is atomic, and as long as noone writes to
> the disk after atomic snapshot, memory image matches what is on disk.
Well, my impression is that this is exactly what happens here: Something
in the XFS code causes metadata to be written to disk _after_ the atomic
That's why I asked if the dirty XFS metadata were flushed by a kernel thread.
You never change things by fighting the existing reality.
R. Buckminster Fuller