1 files changed, 112 insertions, 60 deletions

diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c
index 1f19aa15f89..c0365c14b85 100644
--- a/net/sunrpc/sched.c
+++ b/net/sunrpc/sched.c
@@ -98,6 +98,39 @@ __rpc_add_timer(struct rpc_wait_queue *queue, struct rpc_task *task)
 	list_add(&task->u.tk_wait.timer_list, &queue->timer_list.list);
 }
 
+static void rpc_rotate_queue_owner(struct rpc_wait_queue *queue)
+{
+	struct list_head *q = &queue->tasks[queue->priority];
+	struct rpc_task *task;
+
+	if (!list_empty(q)) {
+		task = list_first_entry(q, struct rpc_task, u.tk_wait.list);
+		if (task->tk_owner == queue->owner)
+			list_move_tail(&task->u.tk_wait.list, q);
+	}
+}
+
+static void rpc_set_waitqueue_priority(struct rpc_wait_queue *queue, int priority)
+{
+	if (queue->priority != priority) {
+		/* Fairness: rotate the list when changing priority */
+		rpc_rotate_queue_owner(queue);
+		queue->priority = priority;
+	}
+}
+
+static void rpc_set_waitqueue_owner(struct rpc_wait_queue *queue, pid_t pid)
+{
+	queue->owner = pid;
+	queue->nr = RPC_BATCH_COUNT;
+}
+
+static void rpc_reset_waitqueue_priority(struct rpc_wait_queue *queue)
+{
+	rpc_set_waitqueue_priority(queue, queue->maxpriority);
+	rpc_set_waitqueue_owner(queue, 0);
+}
+
 /*
  * Add new request to a priority queue.
  */
@@ -109,9 +142,11 @@ static void __rpc_add_wait_queue_priority(struct rpc_wait_queue *queue,
 	struct rpc_task *t;
 
 	INIT_LIST_HEAD(&task->u.tk_wait.links);
-	q = &queue->tasks[queue_priority];
 	if (unlikely(queue_priority > queue->maxpriority))
-		q = &queue->tasks[queue->maxpriority];
+		queue_priority = queue->maxpriority;
+	if (queue_priority > queue->priority)
+		rpc_set_waitqueue_priority(queue, queue_priority);
+	q = &queue->tasks[queue_priority];
 	list_for_each_entry(t, q, u.tk_wait.list) {
 		if (t->tk_owner == task->tk_owner) {
 			list_add_tail(&task->u.tk_wait.list, &t->u.tk_wait.links);
@@ -133,7 +168,9 @@ static void __rpc_add_wait_queue(struct rpc_wait_queue *queue,
 		struct rpc_task *task,
 		unsigned char queue_priority)
 {
-	BUG_ON (RPC_IS_QUEUED(task));
+	WARN_ON_ONCE(RPC_IS_QUEUED(task));
+	if (RPC_IS_QUEUED(task))
+		return;
 
 	if (RPC_IS_PRIORITY(queue))
 		__rpc_add_wait_queue_priority(queue, task, queue_priority);
@@ -143,6 +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",
@@ -178,24 +217,6 @@ static void __rpc_remove_wait_queue(struct rpc_wait_queue *queue, struct rpc_tas
 			task->tk_pid, queue, rpc_qname(queue));
 }
 
-static inline void rpc_set_waitqueue_priority(struct rpc_wait_queue *queue, int priority)
-{
-	queue->priority = priority;
-	queue->count = 1 << (priority * 2);
-}
-
-static inline void rpc_set_waitqueue_owner(struct rpc_wait_queue *queue, pid_t pid)
-{
-	queue->owner = pid;
-	queue->nr = RPC_BATCH_COUNT;
-}
-
-static inline void rpc_reset_waitqueue_priority(struct rpc_wait_queue *queue)
-{
-	rpc_set_waitqueue_priority(queue, queue->maxpriority);
-	rpc_set_waitqueue_owner(queue, 0);
-}
-
 static void __rpc_init_priority_wait_queue(struct rpc_wait_queue *queue, const char *qname, unsigned char nr_queues)
 {
 	int i;
@@ -233,11 +254,11 @@ static int rpc_wait_bit_killable(void *word)
 {
 	if (fatal_signal_pending(current))
 		return -ERESTARTSYS;
-	freezable_schedule();
+	freezable_schedule_unsafe();
 	return 0;
 }
 
-#ifdef RPC_DEBUG
+#if defined(RPC_DEBUG) || defined(RPC_TRACEPOINTS)
 static void rpc_task_set_debuginfo(struct rpc_task *task)
 {
 	static atomic_t rpc_pid;
@@ -303,11 +324,17 @@ EXPORT_SYMBOL_GPL(__rpc_wait_for_completion_task);
  * Note: If the task is ASYNC, and is being made runnable after sitting on an
  * rpc_wait_queue, this must be called with the queue spinlock held to protect
  * the wait queue operation.
+ * Note the ordering of rpc_test_and_set_running() and rpc_clear_queued(),
+ * which is needed to ensure that __rpc_execute() doesn't loop (due to the
+ * lockless RPC_IS_QUEUED() test) before we've had a chance to test
+ * the RPC_TASK_RUNNING flag.
  */
 static void rpc_make_runnable(struct rpc_task *task)
 {
+	bool need_wakeup = !rpc_test_and_set_running(task);
+
 	rpc_clear_queued(task);
-	if (rpc_test_and_set_running(task))
+	if (!need_wakeup)
 		return;
 	if (RPC_IS_ASYNC(task)) {
 		INIT_WORK(&task->u.tk_work, rpc_async_schedule);
@@ -334,7 +361,7 @@ static void __rpc_sleep_on_priority(struct rpc_wait_queue *q,
 
 	__rpc_add_wait_queue(q, task, queue_priority);
 
-	BUG_ON(task->tk_callback != NULL);
+	WARN_ON_ONCE(task->tk_callback != NULL);
 	task->tk_callback = action;
 	__rpc_add_timer(q, task);
 }
@@ -343,7 +370,12 @@ void rpc_sleep_on(struct rpc_wait_queue *q, struct rpc_task *task,
 				rpc_action action)
 {
 	/* We shouldn't ever put an inactive task to sleep */
-	BUG_ON(!RPC_IS_ACTIVATED(task));
+	WARN_ON_ONCE(!RPC_IS_ACTIVATED(task));
+	if (!RPC_IS_ACTIVATED(task)) {
+		task->tk_status = -EIO;
+		rpc_put_task_async(task);
+		return;
+	}
 
 	/*
 	 * Protect the queue operations.
@@ -358,7 +390,12 @@ void rpc_sleep_on_priority(struct rpc_wait_queue *q, struct rpc_task *task,
 		rpc_action action, int priority)
 {
 	/* We shouldn't ever put an inactive task to sleep */
-	BUG_ON(!RPC_IS_ACTIVATED(task));
+	WARN_ON_ONCE(!RPC_IS_ACTIVATED(task));
+	if (!RPC_IS_ACTIVATED(task)) {
+		task->tk_status = -EIO;
+		rpc_put_task_async(task);
+		return;
+	}
 
 	/*
 	 * Protect the queue operations.
@@ -367,6 +404,7 @@ void rpc_sleep_on_priority(struct rpc_wait_queue *q, struct rpc_task *task,
 	__rpc_sleep_on_priority(q, task, action, priority - RPC_PRIORITY_LOW);
 	spin_unlock_bh(&q->lock);
 }
+EXPORT_SYMBOL_GPL(rpc_sleep_on_priority);
 
 /**
  * __rpc_do_wake_up_task - wake up a single rpc_task
@@ -400,23 +438,12 @@ 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);
-}
-
-/*
- * Tests whether rpc queue is empty
- */
-int rpc_queue_empty(struct rpc_wait_queue *queue)
-{
-	int res;
-
-	spin_lock_bh(&queue->lock);
-	res = queue->qlen;
-	spin_unlock_bh(&queue->lock);
-	return res == 0;
+	if (RPC_IS_QUEUED(task)) {
+		smp_rmb();
+		if (task->tk_waitqueue == queue)
+			__rpc_do_wake_up_task(queue, task);
+	}
 }
-EXPORT_SYMBOL_GPL(rpc_queue_empty);
 
 /*
  * Wake up a task on a specific queue
@@ -451,8 +478,7 @@ static struct rpc_task *__rpc_find_next_queued_priority(struct rpc_wait_queue *q
 		/*
 		 * Check if we need to switch queues.
 		 */
-		if (--queue->count)
-			goto new_owner;
+		goto new_owner;
 	}
 
 	/*
@@ -611,7 +637,8 @@ static void __rpc_queue_timer_fn(unsigned long ptr)
 
 static void __rpc_atrun(struct rpc_task *task)
 {
-	task->tk_status = 0;
+	if (task->tk_status == -ETIMEDOUT)
+		task->tk_status = 0;
 }
 
 /*
@@ -697,7 +724,9 @@ static void __rpc_execute(struct rpc_task *task)
 	dprintk("RPC: %5u __rpc_execute flags=0x%x\n",
 			task->tk_pid, task->tk_flags);
 
-	BUG_ON(RPC_IS_QUEUED(task));
+	WARN_ON_ONCE(RPC_IS_QUEUED(task));
+	if (RPC_IS_QUEUED(task))
+		return;
 
 	for (;;) {
 		void (*do_action)(struct rpc_task *);
@@ -762,7 +791,6 @@ static void __rpc_execute(struct rpc_task *task)
 			task->tk_flags |= RPC_TASK_KILLED;
 			rpc_exit(task, -ERESTARTSYS);
 		}
-		rpc_set_running(task);
 		dprintk("RPC: %5u sync task resuming\n", task->tk_pid);
 	}
 
@@ -783,9 +811,11 @@ static void __rpc_execute(struct rpc_task *task)
  */
 void rpc_execute(struct rpc_task *task)
 {
+	bool is_async = RPC_IS_ASYNC(task);
+
 	rpc_set_active(task);
 	rpc_make_runnable(task);
-	if (!RPC_IS_ASYNC(task))
+	if (!is_async)
 		__rpc_execute(task);
 }
 
@@ -802,7 +832,8 @@ static void rpc_async_schedule(struct work_struct *work)
  * @size: requested byte size
  *
  * To prevent rpciod from hanging, this allocator never sleeps,
- * returning NULL if the request cannot be serviced immediately.
+ * returning NULL and suppressing warning if the request cannot be serviced
+ * immediately.
  * The caller can arrange to sleep in a way that is safe for rpciod.
  *
  * Most requests are 'small' (under 2KiB) and can be serviced from a
@@ -815,7 +846,10 @@ static void rpc_async_schedule(struct work_struct *work)
 void *rpc_malloc(struct rpc_task *task, size_t size)
 {
 	struct rpc_buffer *buf;
-	gfp_t gfp = RPC_IS_SWAPPER(task) ? GFP_ATOMIC : GFP_NOWAIT;
+	gfp_t gfp = GFP_NOWAIT | __GFP_NOWARN;
+
+	if (RPC_IS_SWAPPER(task))
+		gfp |= __GFP_MEMALLOC;
 
 	size += sizeof(struct rpc_buffer);
 	if (size <= RPC_BUFFER_MAXSIZE)
@@ -889,7 +923,7 @@ static void rpc_init_task(struct rpc_task *task, const struct rpc_task_setup *ta
 static struct rpc_task *
 rpc_alloc_task(void)
 {
-	return (struct rpc_task *)mempool_alloc(rpc_task_mempool, GFP_NOFS);
+	return (struct rpc_task *)mempool_alloc(rpc_task_mempool, GFP_NOIO);
 }
 
 /*
@@ -916,16 +950,35 @@ struct rpc_task *rpc_new_task(const struct rpc_task_setup *setup_data)
 	return task;
 }
 
+/*
+ * rpc_free_task - release rpc task and perform cleanups
+ *
+ * Note that we free up the rpc_task _after_ rpc_release_calldata()
+ * in order to work around a workqueue dependency issue.
+ *
+ * Tejun Heo states:
+ * "Workqueue currently considers two work items to be the same if they're
+ * on the same address and won't execute them concurrently - ie. it
+ * makes a work item which is queued again while being executed wait
+ * for the previous execution to complete.
+ *
+ * If a work function frees the work item, and then waits for an event
+ * which should be performed by another work item and *that* work item
+ * recycles the freed work item, it can create a false dependency loop.
+ * There really is no reliable way to detect this short of verifying
+ * every memory free."
+ *
+ */
 static void rpc_free_task(struct rpc_task *task)
 {
-	const struct rpc_call_ops *tk_ops = task->tk_ops;
-	void *calldata = task->tk_calldata;
+	unsigned short tk_flags = task->tk_flags;
+
+	rpc_release_calldata(task->tk_ops, task->tk_calldata);
 
-	if (task->tk_flags & RPC_TASK_DYNAMIC) {
+	if (tk_flags & RPC_TASK_DYNAMIC) {
 		dprintk("RPC: %5u freeing task\n", task->tk_pid);
 		mempool_free(task, rpc_task_mempool);
 	}
-	rpc_release_calldata(tk_ops, calldata);
 }
 
 static void rpc_async_release(struct work_struct *work)
@@ -935,8 +988,7 @@ static void rpc_async_release(struct work_struct *work)
 
 static void rpc_release_resources_task(struct rpc_task *task)
 {
-	if (task->tk_rqstp)
-		xprt_release(task);
+	xprt_release(task);
 	if (task->tk_msg.rpc_cred) {
 		put_rpccred(task->tk_msg.rpc_cred);
 		task->tk_msg.rpc_cred = NULL;
@@ -978,7 +1030,7 @@ static void rpc_release_task(struct rpc_task *task)
 {
 	dprintk("RPC: %5u release task\n", task->tk_pid);
 
-	BUG_ON (RPC_IS_QUEUED(task));
+	WARN_ON_ONCE(RPC_IS_QUEUED(task));
 
 	rpc_release_resources_task(task);
 
@@ -1019,7 +1071,7 @@ static int rpciod_start(void)
 	 * Create the rpciod thread and wait for it to start.
 	 */
 	dprintk("RPC:       creating workqueue rpciod\n");
-	wq = alloc_workqueue("rpciod", WQ_MEM_RECLAIM, 0);
+	wq = alloc_workqueue("rpciod", WQ_MEM_RECLAIM, 1);
 	rpciod_workqueue = wq;
 	return rpciod_workqueue != NULL;
 }