> Just a moment ago I tried to remove a directory which was a file before
> the filesystem corruption (not detected by e2fsck) caused by bad RAM.
>
> The directory disappeared, but rmdir hung, I can't kill it (even -9 doesn't
> work, I guess I will have to reboot), ps shows it in the R state using over
> 90% of CPU time. The following message was found in syslog:
>
> EXT2-fs warning (device 22/1): empty_dir: bad directory (dir 46195)
>
We've had the exact same thing happen here, running a vanilla 1.3.4
kernel. What I'd take this to mean is that something between the system
call and the e2fs code is broken, though I'm not expert on such things.
I'd also take this to mean it hasn't been fixed yet. :-)
We were running rmdir on a fairly heavily used drive, mounted as
/var/spool/news. Very annoying problem. Anyone had this happen with rm, BTW?
> Just an idea: how about CRC checksums for inodes? This would allow easy
High overhead?
_________________________________
Joe Fouche | 75 B1 63 3F 0C 68 AA 77 | finger now for
Kingsport, TN, US | 3B 49 AF 89 63 BB E9 C3 | PGP public key
http://www.preferred.com/~jfouche