On Thu, Aug 27, 2009 at 07:30:47PM -0700, Justin Maggard wrote:

I've been testing the latest 64-bit e2fsprogs from the git pu branch
(kernel.org 2.6.30.5/x86_64) on a 64-bit (~22TB) filesystem for a
couple days, since it seems like 32-bit e2fsprogs on a 64-bit
filesystem is going to take a while longer.  I'm able to create and
check a filesystem without any problem.  I've also run Andreas'
llverfs utility for a few hours and not had any complaints.  But, I'm
running into another strange issue.  Here's what I'm doing:
# mkfs.ext4 /dev/md0
# mount /dev/md0 /mnt
# mkdir /mnt/1 /mnt/2 /mnt/3 /mnt/4 /mnt/5
# umount /mnt
# fsck.ext4 -C0 -fy /dev/md0
** No errors at all at this point.  fsck returns 0. **
# mount /dev/md0 /mnt
The last mount command fails, and the kernel log contains:
EXT4-fs: ext4_check_descriptors: Checksum for group 0 failed (3412!=9428)
EXT4-fs: group descriptors corrupted!

Um, that's interesting.  What happens if run fsck.ext4 twice?  i.e:

# mkfs.ext4 /dev/md0
# mount /dev/md0 /mnt
# mkdir /mnt/1 /mnt/2 /mnt/3 /mnt/4 /mnt/5
# umount /mnt
# fsck.ext4 -C0 -fy /dev/md0
# fsck.ext4 -C0 -fy /dev/md0

						- Ted