As reported by David Jeffery: "a signal was sent to lockd while lockd was shutting down from a request to stop nfs. The signal causes lockd to call restart_grace() which puts the lockd_net structure on the grace list. If this signal is received at the wrong time, it will occur after lockd_down_net() has called locks_end_grace() but before lockd_down_net() stops the lockd thread. This leads to lockd putting the lockd_net structure back on the grace list, then exiting without anything removing it from the list."

So, perform the final locks_end_grace() from the the lockd thread; this ensures it's serialized with respect to restart_grace().

Reported-by David Jeffery <[email protected]>

Signed-off-by J. Bruce Fields <[email protected]>

Signed-off-by Sasha Levin <[email protected]>

Signed-off-by Greg Kroah-Hartman <[email protected]>

---
 fs/lockd/svc.c |    6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

fs/lockd/svc.c +4 -2

@@ -129,6 +129,8 @@ lockd(void *vrqstp)
 {
 	int		err = 0;
 	struct svc_rqst *rqstp = vrqstp;
+	struct net *net = &init_net;
+	struct lockd_net *ln = net_generic(net, lockd_net_id);
 
 	/* try_to_freeze() is called from svc_recv() */
 	set_freezable();

@@ -173,6 +175,8 @@ lockd(void *vrqstp)
 	if (nlmsvc_ops)
 		nlmsvc_invalidate_all();
 	nlm_shutdown_hosts();
+	cancel_delayed_work_sync(&ln->grace_period_end);
+	locks_end_grace(&ln->lockd_manager);
 	return 0;
 }

@@ -267,8 +271,6 @@ static void lockd_down_net(struct svc_se
 	if (ln->nlmsvc_users) {
 		if (--ln->nlmsvc_users == 0) {
 			nlm_shutdown_hosts_net(net);
-			cancel_delayed_work_sync(&ln->grace_period_end);
-			locks_end_grace(&ln->lockd_manager);
 			svc_shutdown_net(serv, net);
 			dprintk("lockd_down_net: per-net data destroyed; net=%p\n", net);
 		}

Patches currently in stable-queue which might be from [email protected] are

queue-4.4/lockd-fix-lockd-shutdown-race.patch