Received-SPF: pass (google.com: best guess record for domain of linux-nfs-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: [PATCH RFC] sunrpc: Ensure signalled RPC tasks exit
From:   Chuck Lever <chuck.lever@oracle.com>
To:     linux-nfs@vger.kernel.org
Date:   Wed, 01 Apr 2020 15:37:50 -0400
Message-ID: <20200401193559.6487.55107.stgit@manet.1015granger.net>
User-Agent: StGit/0.17.1-dirty
MIME-Version: 1.0
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit
Sender: linux-nfs-owner@vger.kernel.org
Precedence: bulk

If an RPC task is signaled while it is running and the transport is
not connected, it will never sleep and never be terminated. This can
happen when a RPC transport is shut down: the remaining tasks are
signalled, but the transport is disconnected.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/sched.c |   14 ++++++--------
 1 file changed, 6 insertions(+), 8 deletions(-)

Interested in comments and suggestions.

Nearly every time my NFS/RDMA client unmounts when using krb5, the
umount hangs (killably). I tracked it down to an NFSv3 NULL request
that is signalled but loops and does not exit.


diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c
index 55e900255b0c..905c31f75593 100644
--- a/net/sunrpc/sched.c
+++ b/net/sunrpc/sched.c
@@ -905,6 +905,12 @@ static void __rpc_execute(struct rpc_task *task)
 		trace_rpc_task_run_action(task, do_action);
 		do_action(task);
 
+		if (RPC_SIGNALLED(task)) {
+			task->tk_rpc_status = -ERESTARTSYS;
+			rpc_exit(task, -ERESTARTSYS);
+			break;
+		}
+
 		/*
 		 * Lockless check for whether task is sleeping or not.
 		 */
@@ -912,14 +918,6 @@ static void __rpc_execute(struct rpc_task *task)
 			continue;
 
 		/*
-		 * Signalled tasks should exit rather than sleep.
-		 */
-		if (RPC_SIGNALLED(task)) {
-			task->tk_rpc_status = -ERESTARTSYS;
-			rpc_exit(task, -ERESTARTSYS);
-		}
-
-		/*
 		 * The queue->lock protects against races with
 		 * rpc_make_runnable().
 		 *