On Tue, 2003-04-15 at 17:57, Dave Kleikamp wrote:
> On Tuesday 15 April 2003 16:19, Gary Gaydos wrote:
> > I received an oops in jfs_tree.c:3297 while
> > doing an rm -rf * on a 2.0G filesystem, about 1.8G used.
> Had this volume ever completely filled up (such that creating something
> failed)? If so, that introduces the possibility that we aren't
> handling an out-of-space condition correctly, leading to fs corruption.
Nope, it never filled up. But I do have a corrupt /home now that you
pointed me in the right direction.
When I copy /home to that 2G scratch fs and then rm -rf the scratch
I receive the oops. The /home corruption isn't due to jfs. The
oops happened as I was finishing migrating /home from ext2 to jfs
by copying it to scratch, reformatting /home, and then deleting scratch.
> I'm not sure if I'll be able to find the root cause, but I've been
> meaning to "harden" JFS, so that instead of trapping in situations like
> these, we'd mark the volume dirty and return an error.
I don't think I'll find out what happened to corrupt the original /home
either. Your explanation makes sense. My oops isn't very interesting.
Thanks for looking at it Shaggy.