Ext3 users - Oct 2001 - e2fsprogs 1.23 problem handling 2.2 version 1 format journals

Just fired up a test machine on its first 2.4 kernel - specifically
2.4.9-ac16 (includes ext3 0.9.6).  Had also upgraded e2fs tool set to
1.23.

This box has previously had 2.2 kernels with Stephen's ext3 patches, and
looks like it was last rebuilt from scratch in early March.   I suspect
it may have an old version 1 format journal on the filesystems.
[Unfortunately I managed to destroy the journals in the process - since
I had to remove and recreate them to get the machine usable]

On attempted fsck of the filesystems, I get the error
  External journal has multiple filesystem users (unsupported)

This error was thrown for all filesystems (which would all have had the
journals created at the same time).  All filesystems had previously been
cleanly unmounted.

This happens when I downgrade the kernel (but not e2fsprogs) so I guess
its the userspace journal handling of e2fsprogs

I see that Ted has produced more versions... I will update.  I'll also
add to my todo list (unless someone gives me an excuse) a task to
rebuild the kit off my Q1 build CDs and then check what happens with
different e2fsprogs versions.

BTW what does the 2.4 code do with V1 journals?

	Nigel.

On Oct 01, 2001  11:44 +0100, Nigel Metheringham wrote:
> This box has previously had 2.2 kernels with Stephen's ext3 patches,
and
> looks like it was last rebuilt from scratch in early March.   I suspect
> it may have an old version 1 format journal on the filesystems.
> [Unfortunately I managed to destroy the journals in the process - since
> I had to remove and recreate them to get the machine usable]
> 
> On attempted fsck of the filesystems, I get the error
>   External journal has multiple filesystem users (unsupported)
> 
> This error was thrown for all filesystems (which would all have had the
> journals created at the same time).  All filesystems had previously been
> cleanly unmounted.
This is a known problem of older 2.2 journal code.  It forgot to zero out
the rest of the superblock when updating the journal format, so there is
garbage in now-used fields.  This problem was fixed in later ext3-2.2 code.

Since there is no useful data in an empty journal, it is perfectly safe to
delete the journal and re-create a valid one.
> BTW what does the 2.4 code do with V1 journals?
I think it is still supported, but you can only use them in data=journal
mode.  You need to mount with "journal=update" or something like that
to get it to a V2 journal.

Cheers, Andreas
--
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert

Hi,

On Mon, Oct 01, 2001 at 11:16:28AM -0600, Andreas Dilger wrote:
> > BTW what does the 2.4 code do with V1 journals?
> 
> I think it is still supported, but you can only use them in data=journal
> mode.  You need to mount with "journal=update" or something like
that
> to get it to a V2 journal.
Yep --- the v2 journal is needed for block revoke support, and
writeback and ordered mode need that.

Cheers,
 Stephen