When mixing -olock and -onolock mounts on the same client, we have to check that fl->fl_u.nfs_fl.owner is set before dereferencing it. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>