On Thu, 10 Oct 2002, David Woodhouse wrote:

JFFS2 gets somewhat unhappy when the VFS calls read_inode() and
clear_inode() for the same inode simultaneously. This changeset (which can
be fed through 'grep ^# | cut -c2-' to get a unified diff if anyone cares)
fixes that particular case.

I do care. Next time please mind to send me unified diffs?

That makes me life easier :)

I still haven't decided what to do in the stable branch about the other less
likely case where the VFS calls read_inode() read_inode() clear_inode()
clear_inode() in that order, all for the same inode number. By the time the
last clear_inode() is called, some of the data structures which belonged to
the physical inode in question may have been freed, although it's unlikely
as it actually requires a flash block erase to happen in the meantime too.

Patch applied. Thanks