On 20/10/2011 05:49, Ross wrote: > I don't actually have this problem, I'm just curious. What will happen > if read fails on sector in UFS2 journal area? Sector won't get > remapped until write, so reads will be failing until then. > > I've had experience with bad sectors with non-journaled filesystems — > the system was online and working. The only problem was that > particular file and messages in /var/log/messages until the Current > Pending Sector counter was cleared and sector was remapped. > > Will it be the same (online system) with bad sector in journal area? > And will it damage the filesystem?
Well, SUJ is new enough that there's no general disseminated knowledge about it, but from a quick look at the code: 1) the journal recovery is only ever done in the fsck_ffs program, so it's unlikely that the kernel will panic from a bad sector 2) if the journal cannot be read at any point, the recovery code *should* offer the user to nuke the journal and perform a normal, full fsck, which presumably should leave the file system in a usable state 3) journal writes are done via the normal mechanisms, so it also shouldn't lead to kernel panics on bad sectors, but the "normal" cannot-write-to-device kernel error message and continued operation.
Description: OpenPGP digital signature