Corrupted files in 2.1.128 (ext2)

Bjorn Nordbo (nordbo@hit.no)
Tue, 17 Nov 1998 18:00:56 +0100


This may be a bug in the kernel ext2fs driver, some other part of
the 2.1.128 kernel or a problem specific to my system.

I have been running development kernels for the last few weeks, and
installed 2.1.128 just after the release last week. On friday, I no-
ticed that one of my files had some strange characteristics: it did
not show up on the directory listings, but 'ls -l' and 'stat' found
it. When I tried to delete it, rm complained that no such file exist-
ed. The first file had inode 2130, and the second one 2054, even af-
ter I hardlinked it to the first one.

I tried to copy other files over it, but it's characteristics did
not change. When I tried to hadlink another file to it, the other
file stopped showing up in the directory listings, and got undelete-
able too. ncheck <inode> in debugfs did not find any such block, but
I have not used debugfs before, and I am not sure if I used it corr-
ect.

As I found no way of dealing with those files, I started working
with something else. A few hours later, I noticed the same pheno-
mena again with a different file on another filesystem. At this
moment, I kind of paniced, shot down to single user, synced and
hir the swedish button to force an fsck at boot time.

When it started working on /dev/sda9 (the filesystem where I first
discovered the problem) it reported unexpected inconsitency on it.
When i ran e2fsck manually on the filesystem, it reported inode
2130 and 2054 as unattached, and claimed that they had refcount 2
(should be 1). It also found that the free inode count was 1967
while it should be 1968. When the system finally came up, every-
thing was fine, the files were gone, and copies could be found in
lost+found.

I don't know what caused this. The system had been up and running
since I installed 2.1.128 and I had not noticed any problems at
all. I am still running 2.1.128, and I will try to see if I can
provoke this error to show itself again.

My system is as follows:
P100 with 96MB non-ECC RAM
NCR 53c815 SCSI-2 controller
Seagate Barracuda 15150N 4GB SCSI-2
Kernel 2.1.128
glibc 2.0.7
egcs 1.0.3a
Red Hat 5.2
ext2fs on all partitions of all the disk (except for swap)

If this is a known bug, or definately a problem which is only my
problem, please excuse me for wasting your time.

--
Bjørn Nordbø <nordbo@norskfolke.museum.no>

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/