Ext3 users - Jan 2001 - 0.0.3d-e: JFS: Unrecognised features on journal

I updated my kernel from 0.0.3d to 0.0.3e using the "d-e" patch in the
0.0.3e tarball.  When I try to boot the new kernel on a machine with
an ext3 root filesystem I am getting the message:

    JFS: Unrecognised features on journal

Does this sound feasible or even somewhat expected?  Is there any
interest in me doing some debugging of this?

Is there any way to deal with this without having to boot the new
kernel with ext2 and recreating the journal?

b.

-- 
Brian J. Murrell

Brian Murrell writes:
> I updated my kernel from 0.0.3d to 0.0.3e using the "d-e" patch
in the
> 0.0.3e tarball.
I presume you mean ext3-0.0.5e?
> When I try to boot the new kernel on a machine with
> an ext3 root filesystem I am getting the message:
> 
>     JFS: Unrecognised features on journal
> 
> Does this sound feasible or even somewhat expected?  Is there any
> interest in me doing some debugging of this?
This is expected, given the problems that you previously had.  It was
a bug in the older ext3 code that is now detected by the new code.
> Is there any way to deal with this without having to boot the new
> kernel with ext2 and recreating the journal?
I think you can fix it by using these options when you first mount the fs:

     noload,journal=<inum>

where <inum> is the inode number of the existing journal, as told by
dumpe2fs.
What this does is tell the kernel not to use the existing journal, and then
by supplying a "new" journal inode number it creates a new journal
there.
For root, you need to specify "rootflags=noload,journal=<inum>"
at the LILO
prompt.

I just checked, and journal_create() zeroes all of the journal blocks, so
this will fix the "Unrecognised features on journal" error.  The
problem
was that the old "update journal to v2" code didn't clear the
feature
flags, so random junk was there for "features".  This has been fixed.

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

here's a quote from Stephen:
> ext3-0.0.5d did not always initialise the features flags correctly
> when creating a new journal.  Fixed in 0.0.5e.  As long as you can
> still mount the filesystem, there's an easy upgrade path: you can
> do a clean shutdown, and on the next mount, use the lilo option "rw
> rootflags=noload,journal=<nnn>" where nnn is the inode number of
the
> journal file.
> 
> In other words, it's just like creating the journal in the first
> place, except that you also have the "noload" option to prevent
it
> from loading the existing journal first.
give that a shot, given a clean shutdown previous to using those options.
it worked perfectly for me.

-tcl.


On Wed, 24 Jan 2001, Brian J. Murrell wrote:
> I updated my kernel from 0.0.3d to 0.0.3e using the "d-e" patch
in the
> 0.0.3e tarball.  When I try to boot the new kernel on a machine with
> an ext3 root filesystem I am getting the message:
> 
>     JFS: Unrecognised features on journal
> 
> Does this sound feasible or even somewhat expected?  Is there any
> interest in me doing some debugging of this?
> 
> Is there any way to deal with this without having to boot the new
> kernel with ext2 and recreating the journal?
> 
> b.
> 
> -- 
> Brian J. Murrell
> 
> 
> 
> _______________________________________________
> Ext3-users mailing list
> Ext3-users@redhat.com
> https://listman.redhat.com/mailman/listinfo/ext3-users
>

Hi,

On Wed, Jan 24, 2001 at 09:15:15AM -0800, Brian J. Murrell
wrote:
> I updated my kernel from 0.0.3d to 0.0.3e using the "d-e" patch
in the
> 0.0.3e tarball. 
I assume you mean 0.0.5d to 0.0.5e.
> When I try to boot the new kernel on a machine with
> an ext3 root filesystem I am getting the message:
> 
>     JFS: Unrecognised features on journal
> 
> Does this sound feasible or even somewhat expected?  Is there any
> interest in me doing some debugging of this?
The ChangeLog mentions that older versions of the journal
initialisation code could incorrectly initialise some unused fields,
including the features flags.  5e gets the setting and the checking
right.
 
> Is there any way to deal with this without having to boot the new
> kernel with ext2 and recreating the journal?
Yep --- as long as you know the inode number for the journal, just
boot with the kernel options

	rw rootflags=noload,journal=<inum>

using the 0.0.5e kernel and it will overwrite the old journal with a
new, valid one.  Make sure you have done a clean reboot first, of
course!

Cheers,
 Stephen