On Wed, Jun 30, 2010 at 12:32:19PM +0200, John Hughes wrote:
>
> 1. fsck refuses to work if the journal device is open
We can try doing an exclusive open on the journal device, and refuse
if the journal is active. That's a good idea.
> 2. tune2fs -f -O ^has_journal modifies the filesystem but *not* the
> journal device
> 3. tune2fs -O ^has_journal refuses to work if the journal is not
> empty or the filesystem is active or the journal is open by someone.
(2) is not viable, because the journal also has a back-pointer to the
file system, so we can't just ignore the journal device. The result
in the non-snapshot case leaves dangling pointers in the journal.
(3) should be doable by trying to do an exclusive open of the journal
device and failing if the exclusive open fails.
There has been some talk about exposing the UUID via a system call,
and perhaps having the UUID explicitly change when we take a snapshot.
This would help this situation since it means that e2fsck will be able
to differentiate a snapshot copy from the original file system.
- Ted
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
