On Mon, Jul 16, 2007 at 12:37:09PM +0100, Luciano Rocha wrote: > On Mon, Jul 16, 2007 at 12:18:06PM +0930, Tim wrote: > > On Sun, 2007-07-15 at 17:53 +0100, Luciano Rocha wrote: > > > The problem with hibernation would be that when resuming > > > the kernel will have an erroneous idea of the state of the filesystems > > > it was using. > > > > I would have thought that, other than for the swap drive, the system > > would flush caches and sync drives on its way down. I think it'd be > > foolish not to. > > It flushes dirty buffers, yes, but there are data not part of cache: > open inodes (even if just directories), deleted but in use inodes, > superblock data and journal state, that could have changed/will change > between the hibernation and resume. > > The suspending OS doesn't umount the filesystem (or it didn't last time > I checked). This is easy to check. Save all your data, sync, and > suspend. Then, instead of resuming, boot normally. The OS should > complain about mounting an unchecked filesystem and replaying the > journal. WARNING: re-mkswap the swap space, to prevent accidental resume in your next boot. -- lfr 0/0
Attachment:
pgp4mtKLJp57S.pgp
Description: PGP signature
