Ext3 users - Nov 2001 - Often seeing i_blocks too large after recovery(ies)

I have many systems running current ext3 code.  Some of them have
hardware problems that lead to frequent reboots, and the journal
recovery has worked excellently.  However, on the periodic forced fsck
on supposedly recovered and clean filesystems, I've many times seen
fsck report that i_blocks is too large in one or more inodes.

I suspect that i_blocks is somehow escaping proper restoration during
the log replay.

This is obviously not a critical problem, but it seems to have
persisted through several minor ext3 revisions, so I thought it
would be a good idea to mention it here.

PS: The hardware problems in question are video driver lockups.  One
    machine is a desktop with nVidia's closed driver and the other
    is a ThinkPad 570 using the open neomagic driver; I don't think
    there's a specific connection between this bad hardware and the
    i_blocks problem.
-- 
Chip Salzenberg       - a.k.a. -       <chip@pobox.com>
       "Men of lofty genius are most active
        when they are doing the least work."
                           -- Leonardo da Vinci

Chip Salzenberg wrote:
> 
> I have many systems running current ext3 code.  Some of them have
> hardware problems that lead to frequent reboots, and the journal
> recovery has worked excellently.  However, on the periodic forced fsck
> on supposedly recovered and clean filesystems, I've many times seen
> fsck report that i_blocks is too large in one or more inodes.
> 
> I suspect that i_blocks is somehow escaping proper restoration during
> the log replay.
> 
Is this 2.4.16 or thereabouts?  Are you using quotas?

-

On Nov 30, 2001  11:43 -0800, Chip Salzenberg wrote:
> I have many systems running current ext3 code.  Some of them have
> hardware problems that lead to frequent reboots, and the journal
> recovery has worked excellently.  However, on the periodic forced fsck
> on supposedly recovered and clean filesystems, I've many times seen
> fsck report that i_blocks is too large in one or more inodes.
> 
> This is obviously not a critical problem, but it seems to have
> persisted through several minor ext3 revisions, so I thought it
> would be a good idea to mention it here.
Are you using the EA patches?  There are problems with using the
older EA patches with a newer kernel.

Cheers, Andreas
--
Andreas Dilger
http://sourceforge.net/projects/ext2resize/
http://www-mddsp.enel.ucalgary.ca/People/adilger/