NLM: Don't hang forever on NLM unlock requests

commit 0b760113 upstream.

If the NLM daemon is killed on the NFS server, we can currently end up
hanging forever on an 'unlock' request, instead of aborting. Basically,
if the rpcbind request fails, or the server keeps returning garbage, we
really want to quit instead of retrying.
Tested-by: Vasily Averin <vvs@sw.ru>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

fs/lockd/clntproc.c

@@ -709,7 +709,13 @@ static void nlmclnt_unlock_callback(struct rpc_task *task, void *data)
 
 	if (task->tk_status < 0) {
 		dprintk("lockd: unlock failed (err = %d)\n", -task->tk_status);
-		goto retry_rebind;
+		switch (task->tk_status) {
+		case -EACCES:
+		case -EIO:
+			goto die;
+		default:
+			goto retry_rebind;
+		}
 	}
 	if (status == NLM_LCK_DENIED_GRACE_PERIOD) {
 		rpc_delay(task, NLMCLNT_GRACE_WAIT);

include/linux/sunrpc/sched.h

@@ -84,8 +84,8 @@ struct rpc_task {
 	long			tk_rtt;		/* round-trip time (jiffies) */
 
 	pid_t			tk_owner;	/* Process id for batching tasks */
-	unsigned char		tk_priority : 2;/* Task priority */
-
+	unsigned char		tk_priority : 2,/* Task priority */
+				tk_rebind_retry : 2;
 #ifdef RPC_DEBUG
 	unsigned short		tk_pid;		/* debugging aid */
 #endif

net/sunrpc/clnt.c

@@ -1052,6 +1052,9 @@ call_bind_status(struct rpc_task *task)
 			status = -EOPNOTSUPP;
 			break;
 		}
+		if (task->tk_rebind_retry == 0)
+			break;
+		task->tk_rebind_retry--;
 		rpc_delay(task, 3*HZ);
 		goto retry_timeout;
 	case -ETIMEDOUT:

net/sunrpc/sched.c

@@ -784,6 +784,7 @@ static void rpc_init_task(struct rpc_task *task, const struct rpc_task_setup *ta
 	/* Initialize retry counters */
 	task->tk_garb_retry = 2;
 	task->tk_cred_retry = 2;
+	task->tk_rebind_retry = 2;
 
 	task->tk_priority = task_setup_data->priority - RPC_PRIORITY_LOW;
 	task->tk_owner = current->tgid;