2000-11-03 16:55:20

by jsack

[permalink] [raw]
Subject: raid5 2.4-test10: EXT2-fs corruption during sync

Running mke2fs on a raid5 which is still syncing causes a *variety* of
filesystem corruption errors:

(sample errs)
- ...kernel: EXT2-fs error (device md(9,0)): ext2_check_descriptors:
Block bitmap for group 4 not in group (block 0)!

- e2fsck 1.18, 11-Nov-1999 for EXT2 FS 0.5b, 95/08/09 Group descriptors
look bad... trying backup blocks... e2fsck: Bad magic number in
super-block while trying to open /dev/md0

The superblock could not be read

- e2fsck 1.18, 11-Nov-1999 for EXT2 FS 0.5b, 95/08/09
...Pass 5: Checking group summary information
Padding at end of block bitmap is not set. Fix<y>? yes
(/sample errs)

Raid operations seem stable after fixing errors or if mke2fs is run after
sync is complete. Delaying mke2fs is not a real workaround, because the
fs corruption also occurs when resyncing (a spare) after a disk failure.

This behaviour has also been seen with test9, but does not occur with 2.2
kernels. Non-exhaustive testing indicates it does not occur with raid1.

Where do I go from here?