SUNRPC: Return EAGAIN instead of ENOTCONN when waking up xprt->pending

While we should definitely return socket errors to the task that is
currently trying to send data, there is no need to propagate the same error
to all the other tasks on xprt->pending. Doing so actually slows down
recovery, since it causes more than one tasks to attempt socket recovery.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>

net/sunrpc/clnt.c

@@ -1032,27 +1032,20 @@ call_connect_status(struct rpc_task *task)
 	dprint_status(task);
 
 	task->tk_status = 0;
-	if (status >= 0) {
+	if (status >= 0 || status == -EAGAIN) {
 		clnt->cl_stats->netreconn++;
 		task->tk_action = call_transmit;
 		return;
 	}
 
-	/* Something failed: remote service port may have changed */
-	rpc_force_rebind(clnt);
-
 	switch (status) {
-	case -ENOTCONN:
-	case -EAGAIN:
-		task->tk_action = call_bind;
-		if (!RPC_IS_SOFT(task))
-			return;
 		/* if soft mounted, test if we've timed out */
 	case -ETIMEDOUT:
 		task->tk_action = call_timeout;
-		return;
+		break;
+	default:
+		rpc_exit(task, -EIO);
 	}
-	rpc_exit(task, -EIO);
 }
 
 /*

net/sunrpc/xprt.c

@@ -611,7 +611,7 @@ void xprt_disconnect_done(struct rpc_xprt *xprt)
 	dprintk("RPC:       disconnected transport %p\n", xprt);
 	spin_lock_bh(&xprt->transport_lock);
 	xprt_clear_connected(xprt);
-	xprt_wake_pending_tasks(xprt, -ENOTCONN);
+	xprt_wake_pending_tasks(xprt, -EAGAIN);
 	spin_unlock_bh(&xprt->transport_lock);
 }
 EXPORT_SYMBOL_GPL(xprt_disconnect_done);
@@ -629,7 +629,7 @@ void xprt_force_disconnect(struct rpc_xprt *xprt)
 	/* Try to schedule an autoclose RPC call */
 	if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
 		queue_work(rpciod_workqueue, &xprt->task_cleanup);
-	xprt_wake_pending_tasks(xprt, -ENOTCONN);
+	xprt_wake_pending_tasks(xprt, -EAGAIN);
 	spin_unlock_bh(&xprt->transport_lock);
 }
 
@@ -656,7 +656,7 @@ void xprt_conditional_disconnect(struct rpc_xprt *xprt, unsigned int cookie)
 	/* Try to schedule an autoclose RPC call */
 	if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
 		queue_work(rpciod_workqueue, &xprt->task_cleanup);
-	xprt_wake_pending_tasks(xprt, -ENOTCONN);
+	xprt_wake_pending_tasks(xprt, -EAGAIN);
 out:
 	spin_unlock_bh(&xprt->transport_lock);
 }
@@ -726,9 +726,8 @@ static void xprt_connect_status(struct rpc_task *task)
 	}
 
 	switch (task->tk_status) {
-	case -ENOTCONN:
+	case -EAGAIN:
-		dprintk("RPC: %5u xprt_connect_status: connection broken\n",
+		dprintk("RPC: %5u xprt_connect_status: retrying\n", task->tk_pid);
-				task->tk_pid);
 		break;
 	case -ETIMEDOUT:
 		dprintk("RPC: %5u xprt_connect_status: connect attempt timed "
@@ -849,15 +848,8 @@ int xprt_prepare_transmit(struct rpc_task *task)
 		err = req->rq_received;
 		goto out_unlock;
 	}
-	if (!xprt->ops->reserve_xprt(task)) {
+	if (!xprt->ops->reserve_xprt(task))
 		err = -EAGAIN;
-		goto out_unlock;
-	}
-
-	if (!xprt_connected(xprt)) {
-		err = -ENOTCONN;
-		goto out_unlock;
-	}
 out_unlock:
 	spin_unlock_bh(&xprt->transport_lock);
 	return err;

net/sunrpc/xprtsock.c

@@ -1162,7 +1162,7 @@ static void xs_tcp_state_change(struct sock *sk)
 			transport->tcp_flags =
 				TCP_RCV_COPY_FRAGHDR | TCP_RCV_COPY_XID;
 
-			xprt_wake_pending_tasks(xprt, 0);
+			xprt_wake_pending_tasks(xprt, -EAGAIN);
 		}
 		spin_unlock_bh(&xprt->transport_lock);
 		break;
@@ -1721,20 +1721,22 @@ static void xs_tcp_connect_worker4(struct work_struct *work)
 	dprintk("RPC:       %p connect status %d connected %d sock state %d\n",
 			xprt, -status, xprt_connected(xprt),
 			sock->sk->sk_state);
-	if (status < 0) {
+	switch (status) {
-		switch (status) {
+	case 0:
-			case -EINPROGRESS:
+	case -EINPROGRESS:
-			case -EALREADY:
+	case -EALREADY:
-				goto out_clear;
+		goto out_clear;
-			case -ECONNREFUSED:
+	case -ECONNREFUSED:
-			case -ECONNRESET:
+	case -ECONNRESET:
-				/* retry with existing socket, after a delay */
+		/* retry with existing socket, after a delay */
-				break;
+		break;
-			default:
+	default:
-				/* get rid of existing socket, and retry */
+		/* get rid of existing socket, and retry */
-				xs_tcp_shutdown(xprt);
+		xs_tcp_shutdown(xprt);
-		}
+		printk("%s: connect returned unhandled error %d\n",
+				__func__, status);
 	}
+	status = -EAGAIN;
 out:
 	xprt_wake_pending_tasks(xprt, status);
 out_clear:
@@ -1780,20 +1782,22 @@ static void xs_tcp_connect_worker6(struct work_struct *work)
 	status = xs_tcp_finish_connecting(xprt, sock);
 	dprintk("RPC:       %p connect status %d connected %d sock state %d\n",
 			xprt, -status, xprt_connected(xprt), sock->sk->sk_state);
-	if (status < 0) {
+	switch (status) {
-		switch (status) {
+	case 0:
-			case -EINPROGRESS:
+	case -EINPROGRESS:
-			case -EALREADY:
+	case -EALREADY:
-				goto out_clear;
+		goto out_clear;
-			case -ECONNREFUSED:
+	case -ECONNREFUSED:
-			case -ECONNRESET:
+	case -ECONNRESET:
-				/* retry with existing socket, after a delay */
+		/* retry with existing socket, after a delay */
-				break;
+		break;
-			default:
+	default:
-				/* get rid of existing socket, and retry */
+		/* get rid of existing socket, and retry */
-				xs_tcp_shutdown(xprt);
+		xs_tcp_shutdown(xprt);
-		}
+		printk("%s: connect returned unhandled error %d\n",
+				__func__, status);
 	}
+	status = -EAGAIN;
 out:
 	xprt_wake_pending_tasks(xprt, status);
 out_clear: