[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: kernel oops


On Fri, Nov 03, 2000 at 12:44:19PM -0500, mgalgoci redhat com wrote:
> > 
> > > Nope. Not if you want to use a kernel that lacks ext3 support.
> > 
> > Wil got it right --- ext3 is just ext2 with a journal file.  If you
> > crash, then ext3 keeps a feature flag set in the superblock which
> > prevents you from mounting the filesystem as ext2 without first doing
> > a filesystem recovery, but if you unmount cleanly then you can remount
> > as ext2 without modifying the filesystem at all.
> My experience has been the former. Is it possible for the state machine that 
> manages this to get out of sync and always have the have_journal flag 
> set?

has_journal *should* always be set, because even if you mount it as
ext2, it still has a journal and that journal can still be used in the
future if you mount it as ext3 next time.

The temporary flag which indicates that recovery is necessary is the
"needs_recovery" feature, not "has_journal".

"needs_recovery" is an incompatible feature flag, meaning that if the
kernel doesn't recognise it, it will refuse to mount the filesystem.
This prevents ext3 unrecovered partitions from being mounted as ext2.

"has_journal", on the other hand, is a compatible feature.  ext2 will
not complain if asked to mount a filesystem with that bit set.  Only
fsck will complain --- you just need to use a fsck which is recent
enough to include the ext3 support code.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]