The current code assumes a one-to-one lockowner<->lock stateid
correspondance.

Cc: stable@vger.kernel.org
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>