This has come up several times recently...it seems as though some recent
change in e2fsck must be doing things very differently to produce this
sort of problem. I think I remember reading that the only solution is to
umount the partition in question and use ext2ed on it to delete the files.
No, it's not a change in e2fsck, as far as I can tell. For some reason
we've been getting more cases where garbage is getting written into the
inode table. This is actually very troubling since you can lose a lot
of data that way. I'm not sure why it's happening, though. But for
some reason, there are cases where the Linux kernel is getting confused
enough to scribble random data on an inode table. I've gotten enough
questions about it such that I doubt it's a hardware problem. (Unless
there's some common IDE or SCSI disk which scribbles random data on the
disk during certain types of entertaining crashes or powerdown
scenarios. Unlikely, but....)
- Ted