Re: [PATCH 4/4] fuse: drop dentry on failed revalidate

[Anand Avati]

On 8/7/13 8:44 AM, Miklos Szeredi wrote:
> On Tue, Aug 6, 2013 at 10:06 PM, Anand Avati <avati@xxxxxxxxxx> wrote:
> > On 8/6/13 7:30 AM, Miklos Szeredi wrote:
> > > From: Anand Avati <avati@xxxxxxxxxx>
> > >
> > > Drop a subtree when we find that it has moved or been delated.  This can
> > > be
> > > done as long as there are no submounts under this location.
> > >
> > > If the directory was moved and we come across the same directory in a
> > > future lookup it will be reconnected by d_materialise_unique().
> > >
> > > Signed-off-by: Anand Avati <avati@xxxxxxxxxx>
> > > Signed-off-by: Miklos Szeredi <mszeredi@xxxxxxx>
> > > ---
> > >    fs/fuse/dir.c | 7 ++++++-
> > >    1 file changed, 6 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/fs/fuse/dir.c b/fs/fuse/dir.c
> > > index 131d14b..4ba5893 100644
> > > --- a/fs/fuse/dir.c
> > > +++ b/fs/fuse/dir.c
> > > @@ -226,8 +226,13 @@ static int fuse_dentry_revalidate(struct dentry
> > > *entry, unsigned int flags)
> > >                  if (!err) {
> > >                          struct fuse_inode *fi = get_fuse_inode(inode);
> > >                          if (outarg.nodeid != get_node_id(inode)) {
> > > +                               int ret = 0;
> > > +
> > > +                               if (check_submounts_and_drop(entry) != 0)
> > > +                                       ret = 1;
> > > +
> > >                                  fuse_queue_forget(fc, forget,
> > > outarg.nodeid, 1);
> > > -                               return 0;
> > > +                               return ret;
> >
> >
> > If outarg.nodeid != get_node_id(inode), then we have to return 0 no matter
> > what (whether we successfully dropped the entry or not), no?
>
> If we return 0 in that case (we failed to invalidate the dentry), then
> the VFS will call d_invalidate() which will fail.  The result is the
> same...
>
> > Or are you
> > trying to forcefully keep the path to reach the submount alive? If so, we
> > still fail in inode_permission() .. -> getattr() of the dir inode, no?
>
> Yes.  But the path to the mountpoint should still be reachable (for
> the purpose of unmounting for example).  I'm including an interesting
> discussion between Al and Linus about this (mailing lists weren't
> CC-d, but I don't think they'd mind).


Thanks for attaching the thread. Was very educative! I still do not 
quite understand - will umount() still work when 
inode_permission()[->getattr()] on the ancestors fail (with ESTALE 
etc.)? Wouldn't path resolution itself abort and fail and therefore 
do_umount() never called? I understand that the path to the mountpoint 
being reachable through the dentry chain is a necessity for umounting, 
but is just that really sufficient?

Avati
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/