Merge Trond's nfs-for-next

Merging Trond's nfs-for-next branch, mainly to get
b7993cebb841b0da7a33e9d5ce301a9fd3209165 "SUNRPC: Allow rpc_create() to
request that TCP slots be unlimited", which a small piece of the
gss-proxy work depends on.

5 files changed, 106 insertions, 22 deletions

diff --git a/net/sunrpc/Kconfig b/net/sunrpc/Kconfig
index 516fe2caac2c..262caf03bd5f 100644
--- a/net/sunrpc/Kconfig
+++ b/net/sunrpc/Kconfig
@@ -24,7 +24,6 @@ config SUNRPC_XPRT_RDMA
 config SUNRPC_SWAP
 	bool
 	depends on SUNRPC
-	select NETVM
 
 config RPCSEC_GSS_KRB5
 	tristate "Secure RPC: Kerberos V mechanism"
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index dcc446e7fbf6..651245aa829a 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -414,6 +414,8 @@ struct rpc_clnt *rpc_create(struct rpc_create_args *args)
 	};
 	char servername[48];
 
+	if (args->flags & RPC_CLNT_CREATE_INFINITE_SLOTS)
+		xprtargs.flags |= XPRT_CREATE_INFINITE_SLOTS;
 	/*
 	 * If the caller chooses not to specify a hostname, whip
 	 * up a string representation of the passed-in address.
@@ -1306,6 +1308,8 @@ call_reserve(struct rpc_task *task)
 	xprt_reserve(task);
 }
 
+static void call_retry_reserve(struct rpc_task *task);
+
 /*
  * 1b.	Grok the result of xprt_reserve()
  */
@@ -1347,7 +1351,7 @@ call_reserveresult(struct rpc_task *task)
 	case -ENOMEM:
 		rpc_delay(task, HZ >> 2);
 	case -EAGAIN:	/* woken up; retry */
-		task->tk_action = call_reserve;
+		task->tk_action = call_retry_reserve;
 		return;
 	case -EIO:	/* probably a shutdown */
 		break;
@@ -1360,6 +1364,19 @@ call_reserveresult(struct rpc_task *task)
 }
 
 /*
+ * 1c.	Retry reserving an RPC call slot
+ */
+static void
+call_retry_reserve(struct rpc_task *task)
+{
+	dprint_status(task);
+
+	task->tk_status  = 0;
+	task->tk_action  = call_reserveresult;
+	xprt_retry_reserve(task);
+}
+
+/*
  * 2.	Bind and/or refresh the credentials
  */
 static void
@@ -1644,22 +1661,26 @@ call_connect_status(struct rpc_task *task)
 
 	dprint_status(task);
 
-	task->tk_status = 0;
-	if (status >= 0 || status == -EAGAIN) {
-		clnt->cl_stats->netreconn++;
-		task->tk_action = call_transmit;
-		return;
-	}
-
 	trace_rpc_connect_status(task, status);
 	switch (status) {
 		/* if soft mounted, test if we've timed out */
 	case -ETIMEDOUT:
 		task->tk_action = call_timeout;
-		break;
-	default:
-		rpc_exit(task, -EIO);
+		return;
+	case -ECONNREFUSED:
+	case -ECONNRESET:
+	case -ENETUNREACH:
+		if (RPC_IS_SOFTCONN(task))
+			break;
+		/* retry with existing socket, after a delay */
+	case 0:
+	case -EAGAIN:
+		task->tk_status = 0;
+		clnt->cl_stats->netreconn++;
+		task->tk_action = call_transmit;
+		return;
 	}
+	rpc_exit(task, status);
 }
 
 /*
diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c
index fb20f25ddec9..f8529fc8e542 100644
--- a/net/sunrpc/sched.c
+++ b/net/sunrpc/sched.c
@@ -180,6 +180,8 @@ static void __rpc_add_wait_queue(struct rpc_wait_queue *queue,
 		list_add_tail(&task->u.tk_wait.list, &queue->tasks[0]);
 	task->tk_waitqueue = queue;
 	queue->qlen++;
+	/* barrier matches the read in rpc_wake_up_task_queue_locked() */
+	smp_wmb();
 	rpc_set_queued(task);
 
 	dprintk("RPC: %5u added to queue %p \"%s\"\n",
@@ -430,8 +432,11 @@ static void __rpc_do_wake_up_task(struct rpc_wait_queue *queue, struct rpc_task
  */
 static void rpc_wake_up_task_queue_locked(struct rpc_wait_queue *queue, struct rpc_task *task)
 {
-	if (RPC_IS_QUEUED(task) && task->tk_waitqueue == queue)
-		__rpc_do_wake_up_task(queue, task);
+	if (RPC_IS_QUEUED(task)) {
+		smp_rmb();
+		if (task->tk_waitqueue == queue)
+			__rpc_do_wake_up_task(queue, task);
+	}
 }
 
 /*
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index b7478d5e7ffd..745fca3cfd36 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -948,6 +948,34 @@ void xprt_transmit(struct rpc_task *task)
 	spin_unlock_bh(&xprt->transport_lock);
 }
 
+static void xprt_add_backlog(struct rpc_xprt *xprt, struct rpc_task *task)
+{
+	set_bit(XPRT_CONGESTED, &xprt->state);
+	rpc_sleep_on(&xprt->backlog, task, NULL);
+}
+
+static void xprt_wake_up_backlog(struct rpc_xprt *xprt)
+{
+	if (rpc_wake_up_next(&xprt->backlog) == NULL)
+		clear_bit(XPRT_CONGESTED, &xprt->state);
+}
+
+static bool xprt_throttle_congested(struct rpc_xprt *xprt, struct rpc_task *task)
+{
+	bool ret = false;
+
+	if (!test_bit(XPRT_CONGESTED, &xprt->state))
+		goto out;
+	spin_lock(&xprt->reserve_lock);
+	if (test_bit(XPRT_CONGESTED, &xprt->state)) {
+		rpc_sleep_on(&xprt->backlog, task, NULL);
+		ret = true;
+	}
+	spin_unlock(&xprt->reserve_lock);
+out:
+	return ret;
+}
+
 static struct rpc_rqst *xprt_dynamic_alloc_slot(struct rpc_xprt *xprt, gfp_t gfp_flags)
 {
 	struct rpc_rqst *req = ERR_PTR(-EAGAIN);
@@ -992,7 +1020,7 @@ void xprt_alloc_slot(struct rpc_xprt *xprt, struct rpc_task *task)
 		task->tk_status = -ENOMEM;
 		break;
 	case -EAGAIN:
-		rpc_sleep_on(&xprt->backlog, task, NULL);
+		xprt_add_backlog(xprt, task);
 		dprintk("RPC:       waiting for request slot\n");
 	default:
 		task->tk_status = -EAGAIN;
@@ -1028,7 +1056,7 @@ static void xprt_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
 		memset(req, 0, sizeof(*req));	/* mark unused */
 		list_add(&req->rq_list, &xprt->free);
 	}
-	rpc_wake_up_next(&xprt->backlog);
+	xprt_wake_up_backlog(xprt);
 	spin_unlock(&xprt->reserve_lock);
 }
 
@@ -1092,7 +1120,8 @@ EXPORT_SYMBOL_GPL(xprt_free);
  * xprt_reserve - allocate an RPC request slot
  * @task: RPC task requesting a slot allocation
  *
- * If no more slots are available, place the task on the transport's
+ * If the transport is marked as being congested, or if no more
+ * slots are available, place the task on the transport's
  * backlog queue.
  */
 void xprt_reserve(struct rpc_task *task)
@@ -1107,6 +1136,32 @@ void xprt_reserve(struct rpc_task *task)
 	task->tk_status = -EAGAIN;
 	rcu_read_lock();
 	xprt = rcu_dereference(task->tk_client->cl_xprt);
+	if (!xprt_throttle_congested(xprt, task))
+		xprt->ops->alloc_slot(xprt, task);
+	rcu_read_unlock();
+}
+
+/**
+ * xprt_retry_reserve - allocate an RPC request slot
+ * @task: RPC task requesting a slot allocation
+ *
+ * If no more slots are available, place the task on the transport's
+ * backlog queue.
+ * Note that the only difference with xprt_reserve is that we now
+ * ignore the value of the XPRT_CONGESTED flag.
+ */
+void xprt_retry_reserve(struct rpc_task *task)
+{
+	struct rpc_xprt	*xprt;
+
+	task->tk_status = 0;
+	if (task->tk_rqstp != NULL)
+		return;
+
+	task->tk_timeout = 0;
+	task->tk_status = -EAGAIN;
+	rcu_read_lock();
+	xprt = rcu_dereference(task->tk_client->cl_xprt);
 	xprt->ops->alloc_slot(xprt, task);
 	rcu_read_unlock();
 }
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 3d02130828da..9c2825827dec 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2207,10 +2207,6 @@ static void xs_tcp_setup_socket(struct work_struct *work)
 		 */
 		xs_tcp_force_close(xprt);
 		break;
-	case -ECONNREFUSED:
-	case -ECONNRESET:
-	case -ENETUNREACH:
-		/* retry with existing socket, after a delay */
 	case 0:
 	case -EINPROGRESS:
 	case -EALREADY:
@@ -2221,6 +2217,10 @@ static void xs_tcp_setup_socket(struct work_struct *work)
 		/* Happens, for instance, if the user specified a link
 		 * local IPv6 address without a scope-id.
 		 */
+	case -ECONNREFUSED:
+	case -ECONNRESET:
+	case -ENETUNREACH:
+		/* retry with existing socket, after a delay */
 		goto out;
 	}
 out_eagain:
@@ -2767,9 +2767,13 @@ static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
 	struct rpc_xprt *xprt;
 	struct sock_xprt *transport;
 	struct rpc_xprt *ret;
+	unsigned int max_slot_table_size = xprt_max_tcp_slot_table_entries;
+
+	if (args->flags & XPRT_CREATE_INFINITE_SLOTS)
+		max_slot_table_size = RPC_MAX_SLOT_TABLE_LIMIT;
 
 	xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
-			xprt_max_tcp_slot_table_entries);
+			max_slot_table_size);
 	if (IS_ERR(xprt))
 		return xprt;
 	transport = container_of(xprt, struct sock_xprt, xprt);