Ext3 users - Jun 2001 - EXT3-fs error: ext3_readdir: bad entry in directory

Hi. IIRC, I had the same problem with 2.4.3 and ext2, I'm not sure if with
the same directory entry, but at the time fsck fixed without any problems. I
only know that the error again is with my Mozilla sources.

The Mozilla sources are in another partition. I used lndir from my root fs to
symlink everything and build it. At some point c++ segfaulted (happens a lot
with my GCC 3.0), and I had to restart the build process. But it failed before
the segfault.

Then the following appeared in /var/log/syslog:

Jun 22 02:35:51 pervalidus kernel: EXT3-fs error (device ide0(3,3)):
ext3_readdir: bad entry in directory #290811: directory entry across blocks -
offset=12, inode=354971, rec_len=16396, name_len=2

All symlinks pointing to the affected directory were broken, so I noticed it was
empty.

I unmounted /usr/local/src, ran fsck, and this time it reported no errors, what
wasn't the case with ext2 (although I didn't see any changes by fsck):

# fsck -v /dev/ide/host0/bus0/target0/lun0/part3
Parallelizing fsck version 1.21 (15-Jun-2001)
e2fsck 1.21, 15-Jun-2001 for EXT2 FS 0.5b, 95/08/09
/dev/ide/host0/bus0/target0/lun0/part3: clean, 180665/512512 files,
476613/512071 blocks

I mounted the partition and everything was there.

Jun 22 02:41:32 pervalidus kernel: Journal thread starting. Commit interval 5
seconds
Jun 22 02:41:32 pervalidus kernel: EXT3-fs: mounted filesystem with ordered data
mode.

-- 
0@pervalidus.{net, {dyndns.}org} Tel: 55-21-717-2399 (Niterói-RJ BR)

OK, I found the archives with my earlier problem. It was ext2 with 2.4.4. 
http://marc.theaimsgroup.com/?l=linux-kernel&m=98929561832651&w=2

Not the same inode.

Andreas Dilger suggested hardware problem. I really don't know how to verify
this, but it's strange to see it happened with the same partition and
Mozilla.

-- 
0@pervalidus.{net, {dyndns.}org} Tel: 55-21-717-2399 (Niterói-RJ BR)

"Frédéric L. W. Meunier" wrote:
> 
> ...
> Jun 22 02:35:51 pervalidus kernel: EXT3-fs error (device ide0(3,3)):
> ext3_readdir: bad entry in directory #290811: directory entry across
> blocks - offset=12, inode=354971, rec_len=16396, name_len=2
16396 == 0x400c
> I unmounted /usr/local/src, ran fsck, and this time it reported no
> errors, what wasn't the case with ext2 (although I didn't see any
> changes by fsck):
This would indicate that the data on-disk was correct, and
the data in memory had a single-bit-error.

I'd say you have bad RAM.  Probably the bad RAM contents
happened to get written back to disk on ext2, but not with
ext3 - just a fluke.