Quoting Manuel BOUYER,

> That could be the race condition I was trying to track down in the ext2fs code
> for some time (it is also derived from FFS). Quite hard to find, as this panic
> is really unfrequent on my boxes. I'll have a look deeper about that.
> If you commit this, could you call the functions ufs_* rather than ffs_*, and
> place them in a file where ext2fs could also use them ?

Most FSs need this kind of lock, so I was thinking of putting it in an
even more general place (perhaps vfs_subr.c, or a to-be-created
miscfs/genfs/genfs_subr.c, that needs to be discussed).

If I get some more positive feedback, I'll check things in.

- Frank