Re: [PATCH v2 1/2] nfsd: wake waiters blocked on file_lock before deleting it

[Jeff Layton <jlayton@kernel.org>]

On Mon, Apr 22, 2019 at 7:47 PM NeilBrown <neilb@suse.com> wrote:
>
> On Mon, Apr 22 2019, Jeff Layton wrote:
>
> > After a blocked nfsd file_lock request is deleted, knfsd will send a
> > callback to the client and then free the request. Commit 16306a61d3b7
> > ("fs/locks: always delete_block after waiting.") changed it such that
> > locks_delete_block is always called on a request after it is awoken,
> > but that patch missed fixing up blocked nfsd request handling.
> >
> > Call locks_delete_block on the block to wake up any locks still blocked
> > on the nfsd lock request before freeing it. Some of its callers already
> > do this however, so just remove those calls.
> >
> > URL: https://bugzilla.kernel.org/show_bug.cgi?id=203363
> > Fixes: 16306a61d3b7 ("fs/locks: always delete_block after waiting.")
> > Reported-by: Slawomir Pryczek <slawek1211@gmail.com>
> > Cc: Neil Brown <neilb@suse.com>
> > Cc: stable@vger.kernel.org
> > Signed-off-by: Jeff Layton <jlayton@kernel.org>
> > ---
> >  fs/nfsd/nfs4state.c | 3 +--
> >  1 file changed, 1 insertion(+), 2 deletions(-)
> >
> > diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
> > index 6a45fb00c5fc..e87e15df2044 100644
> > --- a/fs/nfsd/nfs4state.c
> > +++ b/fs/nfsd/nfs4state.c
> > @@ -265,6 +265,7 @@ find_or_allocate_block(struct nfs4_lockowner *lo, struct knfsd_fh *fh,
> >  static void
> >  free_blocked_lock(struct nfsd4_blocked_lock *nbl)
> >  {
> > +     locks_delete_block(&nbl->nbl_lock);
> >       locks_release_private(&nbl->nbl_lock);
>
> Thanks for tracking this down.
>
> An implication of this bug and fix is that we need to be particularly
> careful to make sure locks_delete_block() is called on all relevant
> paths.
> Can we make that easier?  My first thought was to include the call in
> locks_release_private, but lockd calls the two quite separately and it
> certainly seems appropriate that locks_delete_block should be called
> asap, but locks_release_private() can be delayed.
>
> Also cifs calls locks_delete_block, but never calls
> locks_release_private, so it wouldn't help there.
>
> Looking at cifs, I think there is a call missing there too.
> cifs_posix_lock_set() *doesn't* always call locks_delete_block() after
> waiting.  In particular, if ->can_cache_brlcks becomes true while
> waiting then I don't think the behaviour is right.... though I'm not
> sure it is right for other reasons.  It looks like the return value
> should be 1 in that case, but it'll be zero.
>
> But back to my question about making it easier, move the BUG_ON()
> calls from locks_free_lock() into locks_release_private().
>
> ??
>

That sounds like a fine idea. I was thinking about putting all of the
BUG_ONs there in a separate function that both spots could call, but
moving them into locks_release_private should work just as well. Care
to propose a patch?
--
Jeff Layton <jlayton@kernel.org>