On Mon, 15 May 2000, Tigran Aivazian wrote:
> On Mon, 15 May 2000, Alexander Viro wrote:
> > > However, looks like you found a VFS bug that breaks ramfs - let's fix that
> > > one
> > > first..
> >
> > I'm not that sure that VFS is at fault here - "you can't umount it, it's
> > not empty" sounds _very_ bogus.
>
> well, it sounded bogus intuitively but Linus did explain that it is a
> sensible policy for a ramfs filesystem. Are you saying that such policy
> cannot be implemented or that tha ramfs' trick of "extra" dget() is
> inadequate way of implementing it?
Such policy _can_ be implemented, but
a) extra dget() is needed for a different thing and not sufficient
(or adequate) here.
b) I think that such policy is bogus. Again, it can be
implemented, but it breaks all reflexes - I mean, no umount until
you rm -rf the thing? Closing everything is not enough?
And then there is an issue with multiple mounts - which ones should be
affected by such policy? All of them? The last one? Linus, could you
elaborate on that - what do you actually want here?