1 files changed, 436 insertions, 362 deletions

diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c
index 4c669121e60..c0365c14b85 100644
--- a/net/sunrpc/sched.c
+++ b/net/sunrpc/sched.c
@@ -16,17 +16,21 @@
 #include <linux/slab.h>
 #include <linux/mempool.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/spinlock.h>
 #include <linux/mutex.h>
+#include <linux/freezer.h>
 
 #include <linux/sunrpc/clnt.h>
 
+#include "sunrpc.h"
+
 #ifdef RPC_DEBUG
 #define RPCDBG_FACILITY		RPCDBG_SCHED
-#define RPC_TASK_MAGIC_ID	0xf00baa
 #endif
 
+#define CREATE_TRACE_POINTS
+#include <trace/events/sunrpc.h>
+
 /*
  * RPC slabs and memory pools
  */
@@ -38,9 +42,9 @@ static struct kmem_cache	*rpc_buffer_slabp __read_mostly;
 static mempool_t	*rpc_task_mempool __read_mostly;
 static mempool_t	*rpc_buffer_mempool __read_mostly;
 
-static void			__rpc_default_timer(struct rpc_task *task);
 static void			rpc_async_schedule(struct work_struct *);
 static void			 rpc_release_task(struct rpc_task *task);
+static void __rpc_queue_timer_fn(unsigned long ptr);
 
 /*
  * RPC tasks sit here while waiting for conditions to improve.
@@ -57,41 +61,30 @@ struct workqueue_struct *rpciod_workqueue;
  * queue->lock and bh_disabled in order to avoid races within
  * rpc_run_timer().
  */
-static inline void
-__rpc_disable_timer(struct rpc_task *task)
+static void
+__rpc_disable_timer(struct rpc_wait_queue *queue, struct rpc_task *task)
 {
+	if (task->tk_timeout == 0)
+		return;
 	dprintk("RPC: %5u disabling timer\n", task->tk_pid);
-	task->tk_timeout_fn = NULL;
 	task->tk_timeout = 0;
+	list_del(&task->u.tk_wait.timer_list);
+	if (list_empty(&queue->timer_list.list))
+		del_timer(&queue->timer_list.timer);
 }
 
-/*
- * Run a timeout function.
- * We use the callback in order to allow __rpc_wake_up_task()
- * and friends to disable the timer synchronously on SMP systems
- * without calling del_timer_sync(). The latter could cause a
- * deadlock if called while we're holding spinlocks...
- */
-static void rpc_run_timer(struct rpc_task *task)
+static void
+rpc_set_queue_timer(struct rpc_wait_queue *queue, unsigned long expires)
 {
-	void (*callback)(struct rpc_task *);
-
-	callback = task->tk_timeout_fn;
-	task->tk_timeout_fn = NULL;
-	if (callback && RPC_IS_QUEUED(task)) {
-		dprintk("RPC: %5u running timer\n", task->tk_pid);
-		callback(task);
-	}
-	smp_mb__before_clear_bit();
-	clear_bit(RPC_TASK_HAS_TIMER, &task->tk_runstate);
-	smp_mb__after_clear_bit();
+	queue->timer_list.expires = expires;
+	mod_timer(&queue->timer_list.timer, expires);
 }
 
 /*
  * Set up a timer for the current task.
  */
-static inline void
-__rpc_add_timer(struct rpc_task *task, rpc_action timer)
+static void
+__rpc_add_timer(struct rpc_wait_queue *queue, struct rpc_task *task)
 {
 	if (!task->tk_timeout)
 		return;
@@ -99,41 +92,61 @@ __rpc_add_timer(struct rpc_task *task, rpc_action timer)
 	dprintk("RPC: %5u setting alarm for %lu ms\n",
 			task->tk_pid, task->tk_timeout * 1000 / HZ);
 
-	if (timer)
-		task->tk_timeout_fn = timer;
-	else
-		task->tk_timeout_fn = __rpc_default_timer;
-	set_bit(RPC_TASK_HAS_TIMER, &task->tk_runstate);
-	mod_timer(&task->tk_timer, jiffies + task->tk_timeout);
+	task->u.tk_wait.expires = jiffies + task->tk_timeout;
+	if (list_empty(&queue->timer_list.list) || time_before(task->u.tk_wait.expires, queue->timer_list.expires))
+		rpc_set_queue_timer(queue, task->u.tk_wait.expires);
+	list_add(&task->u.tk_wait.timer_list, &queue->timer_list.list);
 }
 
-/*
- * Delete any timer for the current task. Because we use del_timer_sync(),
- * this function should never be called while holding queue->lock.
- */
-static void
-rpc_delete_timer(struct rpc_task *task)
+static void rpc_rotate_queue_owner(struct rpc_wait_queue *queue)
 {
-	if (RPC_IS_QUEUED(task))
-		return;
-	if (test_and_clear_bit(RPC_TASK_HAS_TIMER, &task->tk_runstate)) {
-		del_singleshot_timer_sync(&task->tk_timer);
-		dprintk("RPC: %5u deleting timer\n", task->tk_pid);
+	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.
  */
-static void __rpc_add_wait_queue_priority(struct rpc_wait_queue *queue, struct rpc_task *task)
+static void __rpc_add_wait_queue_priority(struct rpc_wait_queue *queue,
+		struct rpc_task *task,
+		unsigned char queue_priority)
 {
 	struct list_head *q;
 	struct rpc_task *t;
 
 	INIT_LIST_HEAD(&task->u.tk_wait.links);
-	q = &queue->tasks[task->tk_priority];
-	if (unlikely(task->tk_priority > queue->maxpriority))
-		q = &queue->tasks[queue->maxpriority];
+	if (unlikely(queue_priority > 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);
@@ -151,18 +164,24 @@ static void __rpc_add_wait_queue_priority(struct rpc_wait_queue *queue, struct r
  * improve overall performance.
  * Everyone else gets appended to the queue to ensure proper FIFO behavior.
  */
-static void __rpc_add_wait_queue(struct rpc_wait_queue *queue, struct rpc_task *task)
+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);
+		__rpc_add_wait_queue_priority(queue, task, queue_priority);
 	else if (RPC_IS_SWAPPER(task))
 		list_add(&task->u.tk_wait.list, &queue->tasks[0]);
 	else
 		list_add_tail(&task->u.tk_wait.list, &queue->tasks[0]);
-	task->u.tk_wait.rpc_waitq = queue;
+	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",
@@ -181,45 +200,23 @@ static void __rpc_remove_wait_queue_priority(struct rpc_task *task)
 		list_move(&t->u.tk_wait.list, &task->u.tk_wait.list);
 		list_splice_init(&task->u.tk_wait.links, &t->u.tk_wait.links);
 	}
-	list_del(&task->u.tk_wait.list);
 }
 
 /*
  * Remove request from queue.
  * Note: must be called with spin lock held.
  */
-static void __rpc_remove_wait_queue(struct rpc_task *task)
+static void __rpc_remove_wait_queue(struct rpc_wait_queue *queue, struct rpc_task *task)
 {
-	struct rpc_wait_queue *queue;
-	queue = task->u.tk_wait.rpc_waitq;
-
+	__rpc_disable_timer(queue, task);
 	if (RPC_IS_PRIORITY(queue))
 		__rpc_remove_wait_queue_priority(task);
-	else
-		list_del(&task->u.tk_wait.list);
+	list_del(&task->u.tk_wait.list);
 	queue->qlen--;
 	dprintk("RPC: %5u removed from queue %p \"%s\"\n",
 			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;
@@ -229,15 +226,17 @@ static void __rpc_init_priority_wait_queue(struct rpc_wait_queue *queue, const c
 		INIT_LIST_HEAD(&queue->tasks[i]);
 	queue->maxpriority = nr_queues - 1;
 	rpc_reset_waitqueue_priority(queue);
-#ifdef RPC_DEBUG
-	queue->name = qname;
-#endif
+	queue->qlen = 0;
+	setup_timer(&queue->timer_list.timer, __rpc_queue_timer_fn, (unsigned long)queue);
+	INIT_LIST_HEAD(&queue->timer_list.list);
+	rpc_assign_waitqueue_name(queue, qname);
 }
 
 void rpc_init_priority_wait_queue(struct rpc_wait_queue *queue, const char *qname)
 {
 	__rpc_init_priority_wait_queue(queue, qname, RPC_NR_PRIORITY);
 }
+EXPORT_SYMBOL_GPL(rpc_init_priority_wait_queue);
 
 void rpc_init_wait_queue(struct rpc_wait_queue *queue, const char *qname)
 {
@@ -245,20 +244,25 @@ void rpc_init_wait_queue(struct rpc_wait_queue *queue, const char *qname)
 }
 EXPORT_SYMBOL_GPL(rpc_init_wait_queue);
 
+void rpc_destroy_wait_queue(struct rpc_wait_queue *queue)
+{
+	del_timer_sync(&queue->timer_list.timer);
+}
+EXPORT_SYMBOL_GPL(rpc_destroy_wait_queue);
+
 static int rpc_wait_bit_killable(void *word)
 {
 	if (fatal_signal_pending(current))
 		return -ERESTARTSYS;
-	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;
 
-	task->tk_magic = RPC_TASK_MAGIC_ID;
 	task->tk_pid = atomic_inc_return(&rpc_pid);
 }
 #else
@@ -269,38 +273,47 @@ static inline void rpc_task_set_debuginfo(struct rpc_task *task)
 
 static void rpc_set_active(struct rpc_task *task)
 {
-	struct rpc_clnt *clnt;
-	if (test_and_set_bit(RPC_TASK_ACTIVE, &task->tk_runstate) != 0)
-		return;
+	trace_rpc_task_begin(task->tk_client, task, NULL);
+
 	rpc_task_set_debuginfo(task);
-	/* Add to global list of all tasks */
-	clnt = task->tk_client;
-	if (clnt != NULL) {
-		spin_lock(&clnt->cl_lock);
-		list_add_tail(&task->tk_task, &clnt->cl_tasks);
-		spin_unlock(&clnt->cl_lock);
-	}
+	set_bit(RPC_TASK_ACTIVE, &task->tk_runstate);
 }
 
 /*
  * Mark an RPC call as having completed by clearing the 'active' bit
+ * and then waking up all tasks that were sleeping.
  */
-static void rpc_mark_complete_task(struct rpc_task *task)
+static int rpc_complete_task(struct rpc_task *task)
 {
-	smp_mb__before_clear_bit();
+	void *m = &task->tk_runstate;
+	wait_queue_head_t *wq = bit_waitqueue(m, RPC_TASK_ACTIVE);
+	struct wait_bit_key k = __WAIT_BIT_KEY_INITIALIZER(m, RPC_TASK_ACTIVE);
+	unsigned long flags;
+	int ret;
+
+	trace_rpc_task_complete(task->tk_client, task, NULL);
+
+	spin_lock_irqsave(&wq->lock, flags);
 	clear_bit(RPC_TASK_ACTIVE, &task->tk_runstate);
-	smp_mb__after_clear_bit();
-	wake_up_bit(&task->tk_runstate, RPC_TASK_ACTIVE);
+	ret = atomic_dec_and_test(&task->tk_count);
+	if (waitqueue_active(wq))
+		__wake_up_locked_key(wq, TASK_NORMAL, &k);
+	spin_unlock_irqrestore(&wq->lock, flags);
+	return ret;
 }
 
 /*
  * Allow callers to wait for completion of an RPC call
+ *
+ * Note the use of out_of_line_wait_on_bit() rather than wait_on_bit()
+ * to enforce taking of the wq->lock and hence avoid races with
+ * rpc_complete_task().
  */
 int __rpc_wait_for_completion_task(struct rpc_task *task, int (*action)(void *))
 {
 	if (action == NULL)
 		action = rpc_wait_bit_killable;
-	return wait_on_bit(&task->tk_runstate, RPC_TASK_ACTIVE,
+	return out_of_line_wait_on_bit(&task->tk_runstate, RPC_TASK_ACTIVE,
 			action, TASK_KILLABLE);
 }
 EXPORT_SYMBOL_GPL(__rpc_wait_for_completion_task);
@@ -308,30 +321,24 @@ EXPORT_SYMBOL_GPL(__rpc_wait_for_completion_task);
 /*
  * Make an RPC task runnable.
  *
- * Note: If the task is ASYNC, this must be called with
- * the spinlock held to protect the wait queue operation.
+ * 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)
 {
-	BUG_ON(task->tk_timeout_fn);
+	bool need_wakeup = !rpc_test_and_set_running(task);
+
 	rpc_clear_queued(task);
-	if (rpc_test_and_set_running(task))
+	if (!need_wakeup)
 		return;
-	/* We might have raced */
-	if (RPC_IS_QUEUED(task)) {
-		rpc_clear_running(task);
-		return;
-	}
 	if (RPC_IS_ASYNC(task)) {
-		int status;
-
 		INIT_WORK(&task->u.tk_work, rpc_async_schedule);
-		status = queue_work(task->tk_workqueue, &task->u.tk_work);
-		if (status < 0) {
-			printk(KERN_WARNING "RPC: failed to add task to queue: error: %d!\n", status);
-			task->tk_status = status;
-			return;
-		}
+		queue_work(rpciod_workqueue, &task->u.tk_work);
 	} else
 		wake_up_bit(&task->tk_runstate, RPC_TASK_QUEUED);
 }
@@ -342,61 +349,84 @@ static void rpc_make_runnable(struct rpc_task *task)
  * NB: An RPC task will only receive interrupt-driven events as long
  * as it's on a wait queue.
  */
-static void __rpc_sleep_on(struct rpc_wait_queue *q, struct rpc_task *task,
-			rpc_action action, rpc_action timer)
+static void __rpc_sleep_on_priority(struct rpc_wait_queue *q,
+		struct rpc_task *task,
+		rpc_action action,
+		unsigned char queue_priority)
 {
 	dprintk("RPC: %5u sleep_on(queue \"%s\" time %lu)\n",
 			task->tk_pid, rpc_qname(q), jiffies);
 
-	if (!RPC_IS_ASYNC(task) && !RPC_IS_ACTIVATED(task)) {
-		printk(KERN_ERR "RPC: Inactive synchronous task put to sleep!\n");
-		return;
-	}
+	trace_rpc_task_sleep(task->tk_client, task, q);
 
-	__rpc_add_wait_queue(q, task);
+	__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(task, timer);
+	__rpc_add_timer(q, task);
 }
 
 void rpc_sleep_on(struct rpc_wait_queue *q, struct rpc_task *task,
-				rpc_action action, rpc_action timer)
+				rpc_action action)
 {
-	/* Mark the task as being activated if so needed */
-	rpc_set_active(task);
+	/* We shouldn't ever put an inactive task to sleep */
+	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.
 	 */
 	spin_lock_bh(&q->lock);
-	__rpc_sleep_on(q, task, action, timer);
+	__rpc_sleep_on_priority(q, task, action, task->tk_priority);
 	spin_unlock_bh(&q->lock);
 }
 EXPORT_SYMBOL_GPL(rpc_sleep_on);
 
+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 */
+	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.
+	 */
+	spin_lock_bh(&q->lock);
+	__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
+ * @queue: wait queue
  * @task: task to be woken up
  *
  * Caller must hold queue->lock, and have cleared the task queued flag.
  */
-static void __rpc_do_wake_up_task(struct rpc_task *task)
+static void __rpc_do_wake_up_task(struct rpc_wait_queue *queue, struct rpc_task *task)
 {
 	dprintk("RPC: %5u __rpc_wake_up_task (now %lu)\n",
 			task->tk_pid, jiffies);
 
-#ifdef RPC_DEBUG
-	BUG_ON(task->tk_magic != RPC_TASK_MAGIC_ID);
-#endif
 	/* Has the task been executed yet? If not, we cannot wake it up! */
 	if (!RPC_IS_ACTIVATED(task)) {
 		printk(KERN_ERR "RPC: Inactive task (%p) being woken up!\n", task);
 		return;
 	}
 
-	__rpc_disable_timer(task);
-	__rpc_remove_wait_queue(task);
+	trace_rpc_task_wakeup(task->tk_client, task, queue);
+
+	__rpc_remove_wait_queue(queue, task);
 
 	rpc_make_runnable(task);
 
@@ -404,53 +434,32 @@ static void __rpc_do_wake_up_task(struct rpc_task *task)
 }
 
 /*
- * Wake up the specified task
+ * Wake up a queued task while the queue lock is being held
  */
-static void __rpc_wake_up_task(struct rpc_task *task)
+static void rpc_wake_up_task_queue_locked(struct rpc_wait_queue *queue, struct rpc_task *task)
 {
-	if (rpc_start_wakeup(task)) {
-		if (RPC_IS_QUEUED(task))
-			__rpc_do_wake_up_task(task);
-		rpc_finish_wakeup(task);
+	if (RPC_IS_QUEUED(task)) {
+		smp_rmb();
+		if (task->tk_waitqueue == queue)
+			__rpc_do_wake_up_task(queue, task);
 	}
 }
 
 /*
- * Default timeout handler if none specified by user
- */
-static void
-__rpc_default_timer(struct rpc_task *task)
-{
-	dprintk("RPC: %5u timeout (default timer)\n", task->tk_pid);
-	task->tk_status = -ETIMEDOUT;
-	rpc_wake_up_task(task);
-}
-
-/*
- * Wake up the specified task
+ * Wake up a task on a specific queue
  */
-void rpc_wake_up_task(struct rpc_task *task)
+void rpc_wake_up_queued_task(struct rpc_wait_queue *queue, struct rpc_task *task)
 {
-	rcu_read_lock_bh();
-	if (rpc_start_wakeup(task)) {
-		if (RPC_IS_QUEUED(task)) {
-			struct rpc_wait_queue *queue = task->u.tk_wait.rpc_waitq;
-
-			/* Note: we're already in a bh-safe context */
-			spin_lock(&queue->lock);
-			__rpc_do_wake_up_task(task);
-			spin_unlock(&queue->lock);
-		}
-		rpc_finish_wakeup(task);
-	}
-	rcu_read_unlock_bh();
+	spin_lock_bh(&queue->lock);
+	rpc_wake_up_task_queue_locked(queue, task);
+	spin_unlock_bh(&queue->lock);
 }
-EXPORT_SYMBOL_GPL(rpc_wake_up_task);
+EXPORT_SYMBOL_GPL(rpc_wake_up_queued_task);
 
 /*
  * Wake up the next task on a priority queue.
  */
-static struct rpc_task * __rpc_wake_up_next_priority(struct rpc_wait_queue *queue)
+static struct rpc_task *__rpc_find_next_queued_priority(struct rpc_wait_queue *queue)
 {
 	struct list_head *q;
 	struct rpc_task *task;
@@ -469,8 +478,7 @@ static struct rpc_task * __rpc_wake_up_next_priority(struct rpc_wait_queue *queu
 		/*
 		 * Check if we need to switch queues.
 		 */
-		if (--queue->count)
-			goto new_owner;
+		goto new_owner;
 	}
 
 	/*
@@ -495,32 +503,54 @@ new_queue:
 new_owner:
 	rpc_set_waitqueue_owner(queue, task->tk_owner);
 out:
-	__rpc_wake_up_task(task);
 	return task;
 }
 
+static struct rpc_task *__rpc_find_next_queued(struct rpc_wait_queue *queue)
+{
+	if (RPC_IS_PRIORITY(queue))
+		return __rpc_find_next_queued_priority(queue);
+	if (!list_empty(&queue->tasks[0]))
+		return list_first_entry(&queue->tasks[0], struct rpc_task, u.tk_wait.list);
+	return NULL;
+}
+
 /*
- * Wake up the next task on the wait queue.
+ * Wake up the first task on the wait queue.
  */
-struct rpc_task * rpc_wake_up_next(struct rpc_wait_queue *queue)
+struct rpc_task *rpc_wake_up_first(struct rpc_wait_queue *queue,
+		bool (*func)(struct rpc_task *, void *), void *data)
 {
 	struct rpc_task	*task = NULL;
 
-	dprintk("RPC:       wake_up_next(%p \"%s\")\n",
+	dprintk("RPC:       wake_up_first(%p \"%s\")\n",
 			queue, rpc_qname(queue));
-	rcu_read_lock_bh();
-	spin_lock(&queue->lock);
-	if (RPC_IS_PRIORITY(queue))
-		task = __rpc_wake_up_next_priority(queue);
-	else {
-		task_for_first(task, &queue->tasks[0])
-			__rpc_wake_up_task(task);
+	spin_lock_bh(&queue->lock);
+	task = __rpc_find_next_queued(queue);
+	if (task != NULL) {
+		if (func(task, data))
+			rpc_wake_up_task_queue_locked(queue, task);
+		else
+			task = NULL;
 	}
-	spin_unlock(&queue->lock);
-	rcu_read_unlock_bh();
+	spin_unlock_bh(&queue->lock);
 
 	return task;
 }
+EXPORT_SYMBOL_GPL(rpc_wake_up_first);
+
+static bool rpc_wake_up_next_func(struct rpc_task *task, void *data)
+{
+	return true;
+}
+
+/*
+ * Wake up the next task on the wait queue.
+*/
+struct rpc_task *rpc_wake_up_next(struct rpc_wait_queue *queue)
+{
+	return rpc_wake_up_first(queue, rpc_wake_up_next_func, NULL);
+}
 EXPORT_SYMBOL_GPL(rpc_wake_up_next);
 
 /**
@@ -531,21 +561,23 @@ EXPORT_SYMBOL_GPL(rpc_wake_up_next);
  */
 void rpc_wake_up(struct rpc_wait_queue *queue)
 {
-	struct rpc_task *task, *next;
 	struct list_head *head;
 
-	rcu_read_lock_bh();
-	spin_lock(&queue->lock);
+	spin_lock_bh(&queue->lock);
 	head = &queue->tasks[queue->maxpriority];
 	for (;;) {
-		list_for_each_entry_safe(task, next, head, u.tk_wait.list)
-			__rpc_wake_up_task(task);
+		while (!list_empty(head)) {
+			struct rpc_task *task;
+			task = list_first_entry(head,
+					struct rpc_task,
+					u.tk_wait.list);
+			rpc_wake_up_task_queue_locked(queue, task);
+		}
 		if (head == &queue->tasks[0])
 			break;
 		head--;
 	}
-	spin_unlock(&queue->lock);
-	rcu_read_unlock_bh();
+	spin_unlock_bh(&queue->lock);
 }
 EXPORT_SYMBOL_GPL(rpc_wake_up);
 
@@ -558,29 +590,55 @@ EXPORT_SYMBOL_GPL(rpc_wake_up);
  */
 void rpc_wake_up_status(struct rpc_wait_queue *queue, int status)
 {
-	struct rpc_task *task, *next;
 	struct list_head *head;
 
-	rcu_read_lock_bh();
-	spin_lock(&queue->lock);
+	spin_lock_bh(&queue->lock);
 	head = &queue->tasks[queue->maxpriority];
 	for (;;) {
-		list_for_each_entry_safe(task, next, head, u.tk_wait.list) {
+		while (!list_empty(head)) {
+			struct rpc_task *task;
+			task = list_first_entry(head,
+					struct rpc_task,
+					u.tk_wait.list);
 			task->tk_status = status;
-			__rpc_wake_up_task(task);
+			rpc_wake_up_task_queue_locked(queue, task);
 		}
 		if (head == &queue->tasks[0])
 			break;
 		head--;
 	}
-	spin_unlock(&queue->lock);
-	rcu_read_unlock_bh();
+	spin_unlock_bh(&queue->lock);
 }
 EXPORT_SYMBOL_GPL(rpc_wake_up_status);
 
+static void __rpc_queue_timer_fn(unsigned long ptr)
+{
+	struct rpc_wait_queue *queue = (struct rpc_wait_queue *)ptr;
+	struct rpc_task *task, *n;
+	unsigned long expires, now, timeo;
+
+	spin_lock(&queue->lock);
+	expires = now = jiffies;
+	list_for_each_entry_safe(task, n, &queue->timer_list.list, u.tk_wait.timer_list) {
+		timeo = task->u.tk_wait.expires;
+		if (time_after_eq(now, timeo)) {
+			dprintk("RPC: %5u timeout\n", task->tk_pid);
+			task->tk_status = -ETIMEDOUT;
+			rpc_wake_up_task_queue_locked(queue, task);
+			continue;
+		}
+		if (expires == now || time_after(expires, timeo))
+			expires = timeo;
+	}
+	if (!list_empty(&queue->timer_list.list))
+		rpc_set_queue_timer(queue, expires);
+	spin_unlock(&queue->lock);
+}
+
 static void __rpc_atrun(struct rpc_task *task)
 {
-	rpc_wake_up_task(task);
+	if (task->tk_status == -ETIMEDOUT)
+		task->tk_status = 0;
 }
 
 /*
@@ -589,18 +647,37 @@ static void __rpc_atrun(struct rpc_task *task)
 void rpc_delay(struct rpc_task *task, unsigned long delay)
 {
 	task->tk_timeout = delay;
-	rpc_sleep_on(&delay_queue, task, NULL, __rpc_atrun);
+	rpc_sleep_on(&delay_queue, task, __rpc_atrun);
 }
 EXPORT_SYMBOL_GPL(rpc_delay);
 
 /*
  * Helper to call task->tk_ops->rpc_call_prepare
  */
-static void rpc_prepare_task(struct rpc_task *task)
+void rpc_prepare_task(struct rpc_task *task)
 {
-	lock_kernel();
 	task->tk_ops->rpc_call_prepare(task, task->tk_calldata);
-	unlock_kernel();
+}
+
+static void
+rpc_init_task_statistics(struct rpc_task *task)
+{
+	/* Initialize retry counters */
+	task->tk_garb_retry = 2;
+	task->tk_cred_retry = 2;
+	task->tk_rebind_retry = 2;
+
+	/* starting timestamp */
+	task->tk_start = ktime_get();
+}
+
+static void
+rpc_reset_task_statistics(struct rpc_task *task)
+{
+	task->tk_timeouts = 0;
+	task->tk_flags &= ~(RPC_CALL_MAJORSEEN|RPC_TASK_KILLED|RPC_TASK_SENT);
+
+	rpc_init_task_statistics(task);
 }
 
 /*
@@ -610,25 +687,29 @@ void rpc_exit_task(struct rpc_task *task)
 {
 	task->tk_action = NULL;
 	if (task->tk_ops->rpc_call_done != NULL) {
-		lock_kernel();
 		task->tk_ops->rpc_call_done(task, task->tk_calldata);
-		unlock_kernel();
 		if (task->tk_action != NULL) {
 			WARN_ON(RPC_ASSASSINATED(task));
 			/* Always release the RPC slot and buffer memory */
 			xprt_release(task);
+			rpc_reset_task_statistics(task);
 		}
 	}
 }
-EXPORT_SYMBOL_GPL(rpc_exit_task);
+
+void rpc_exit(struct rpc_task *task, int status)
+{
+	task->tk_status = status;
+	task->tk_action = rpc_exit_task;
+	if (RPC_IS_QUEUED(task))
+		rpc_wake_up_queued_task(task->tk_waitqueue, task);
+}
+EXPORT_SYMBOL_GPL(rpc_exit);
 
 void rpc_release_calldata(const struct rpc_call_ops *ops, void *calldata)
 {
-	if (ops->rpc_release != NULL) {
-		lock_kernel();
+	if (ops->rpc_release != NULL)
 		ops->rpc_release(calldata);
-		unlock_kernel();
-	}
 }
 
 /*
@@ -636,63 +717,63 @@ void rpc_release_calldata(const struct rpc_call_ops *ops, void *calldata)
  */
 static void __rpc_execute(struct rpc_task *task)
 {
-	int		status = 0;
+	struct rpc_wait_queue *queue;
+	int task_is_async = RPC_IS_ASYNC(task);
+	int status = 0;
 
 	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 (;;) {
-		/*
-		 * Garbage collection of pending timers...
-		 */
-		rpc_delete_timer(task);
+		void (*do_action)(struct rpc_task *);
 
 		/*
-		 * Execute any pending callback.
+		 * Execute any pending callback first.
 		 */
-		if (RPC_DO_CALLBACK(task)) {
-			/* Define a callback save pointer */
-			void (*save_callback)(struct rpc_task *);
-
+		do_action = task->tk_callback;
+		task->tk_callback = NULL;
+		if (do_action == NULL) {
 			/*
-			 * If a callback exists, save it, reset it,
-			 * call it.
-			 * The save is needed to stop from resetting
-			 * another callback set within the callback handler
-			 * - Dave
+			 * Perform the next FSM step.
+			 * tk_action may be NULL if the task has been killed.
+			 * In particular, note that rpc_killall_tasks may
+			 * do this at any time, so beware when dereferencing.
 			 */
-			save_callback=task->tk_callback;
-			task->tk_callback=NULL;
-			save_callback(task);
-		}
-
-		/*
-		 * Perform the next FSM step.
-		 * tk_action may be NULL when the task has been killed
-		 * by someone else.
-		 */
-		if (!RPC_IS_QUEUED(task)) {
-			if (task->tk_action == NULL)
+			do_action = task->tk_action;
+			if (do_action == NULL)
 				break;
-			task->tk_action(task);
 		}
+		trace_rpc_task_run_action(task->tk_client, task, task->tk_action);
+		do_action(task);
 
 		/*
 		 * Lockless check for whether task is sleeping or not.
 		 */
 		if (!RPC_IS_QUEUED(task))
 			continue;
-		rpc_clear_running(task);
-		if (RPC_IS_ASYNC(task)) {
-			/* Careful! we may have raced... */
-			if (RPC_IS_QUEUED(task))
-				return;
-			if (rpc_test_and_set_running(task))
-				return;
+		/*
+		 * The queue->lock protects against races with
+		 * rpc_make_runnable().
+		 *
+		 * Note that once we clear RPC_TASK_RUNNING on an asynchronous
+		 * rpc_task, rpc_make_runnable() can assign it to a
+		 * different workqueue. We therefore cannot assume that the
+		 * rpc_task pointer may still be dereferenced.
+		 */
+		queue = task->tk_waitqueue;
+		spin_lock_bh(&queue->lock);
+		if (!RPC_IS_QUEUED(task)) {
+			spin_unlock_bh(&queue->lock);
 			continue;
 		}
+		rpc_clear_running(task);
+		spin_unlock_bh(&queue->lock);
+		if (task_is_async)
+			return;
 
 		/* sync task: sleep here */
 		dprintk("RPC: %5u sync task going to sleep\n", task->tk_pid);
@@ -709,9 +790,7 @@ static void __rpc_execute(struct rpc_task *task)
 			dprintk("RPC: %5u got signal\n", task->tk_pid);
 			task->tk_flags |= RPC_TASK_KILLED;
 			rpc_exit(task, -ERESTARTSYS);
-			rpc_wake_up_task(task);
 		}
-		rpc_set_running(task);
 		dprintk("RPC: %5u sync task resuming\n", task->tk_pid);
 	}
 
@@ -732,28 +811,29 @@ 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_set_running(task);
-	__rpc_execute(task);
+	rpc_make_runnable(task);
+	if (!is_async)
+		__rpc_execute(task);
 }
 
 static void rpc_async_schedule(struct work_struct *work)
 {
+	current->flags |= PF_FSTRANS;
 	__rpc_execute(container_of(work, struct rpc_task, u.tk_work));
+	current->flags &= ~PF_FSTRANS;
 }
 
-struct rpc_buffer {
-	size_t	len;
-	char	data[];
-};
-
 /**
  * rpc_malloc - allocate an RPC buffer
  * @task: RPC task that will use this buffer
  * @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
@@ -766,7 +846,10 @@ struct rpc_buffer {
 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)
@@ -816,47 +899,22 @@ EXPORT_SYMBOL_GPL(rpc_free);
 static void rpc_init_task(struct rpc_task *task, const struct rpc_task_setup *task_setup_data)
 {
 	memset(task, 0, sizeof(*task));
-	setup_timer(&task->tk_timer, (void (*)(unsigned long))rpc_run_timer,
-			(unsigned long)task);
 	atomic_set(&task->tk_count, 1);
 	task->tk_flags  = task_setup_data->flags;
 	task->tk_ops = task_setup_data->callback_ops;
 	task->tk_calldata = task_setup_data->callback_data;
 	INIT_LIST_HEAD(&task->tk_task);
 
-	/* Initialize retry counters */
-	task->tk_garb_retry = 2;
-	task->tk_cred_retry = 2;
-
 	task->tk_priority = task_setup_data->priority - RPC_PRIORITY_LOW;
 	task->tk_owner = current->tgid;
 
 	/* Initialize workqueue for async tasks */
-	task->tk_workqueue = rpciod_workqueue;
-
-	task->tk_client = task_setup_data->rpc_client;
-	if (task->tk_client != NULL) {
-		kref_get(&task->tk_client->cl_kref);
-		if (task->tk_client->cl_softrtry)
-			task->tk_flags |= RPC_TASK_SOFT;
-	}
+	task->tk_workqueue = task_setup_data->workqueue;
 
 	if (task->tk_ops->rpc_call_prepare != NULL)
 		task->tk_action = rpc_prepare_task;
 
-	if (task_setup_data->rpc_message != NULL) {
-		memcpy(&task->tk_msg, task_setup_data->rpc_message, sizeof(task->tk_msg));
-		/* Bind the user cred */
-		if (task->tk_msg.rpc_cred != NULL)
-			rpcauth_holdcred(task);
-		else
-			rpcauth_bindcred(task);
-		if (task->tk_action == NULL)
-			rpc_call_start(task);
-	}
-
-	/* starting timestamp */
-	task->tk_start = jiffies;
+	rpc_init_task_statistics(task);
 
 	dprintk("RPC:       new task initialized, procpid %u\n",
 				task_pid_nr(current));
@@ -865,14 +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);
-}
-
-static void rpc_free_task(struct rcu_head *rcu)
-{
-	struct rpc_task *task = container_of(rcu, struct rpc_task, u.tk_rcu);
-	dprintk("RPC: %5u freeing task\n", task->tk_pid);
-	mempool_free(task, rpc_task_mempool);
+	return (struct rpc_task *)mempool_alloc(rpc_task_mempool, GFP_NOIO);
 }
 
 /*
@@ -885,98 +936,119 @@ struct rpc_task *rpc_new_task(const struct rpc_task_setup *setup_data)
 
 	if (task == NULL) {
 		task = rpc_alloc_task();
-		if (task == NULL)
-			goto out;
+		if (task == NULL) {
+			rpc_release_calldata(setup_data->callback_ops,
+					setup_data->callback_data);
+			return ERR_PTR(-ENOMEM);
+		}
 		flags = RPC_TASK_DYNAMIC;
 	}
 
 	rpc_init_task(task, setup_data);
-
 	task->tk_flags |= flags;
 	dprintk("RPC:       allocated task %p\n", task);
-out:
 	return task;
 }
 
-
-void rpc_put_task(struct rpc_task *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;
 
-	if (!atomic_dec_and_test(&task->tk_count))
-		return;
-	/* Release resources */
-	if (task->tk_rqstp)
-		xprt_release(task);
-	if (task->tk_msg.rpc_cred)
-		rpcauth_unbindcred(task);
-	if (task->tk_client) {
-		rpc_release_client(task->tk_client);
-		task->tk_client = NULL;
+	rpc_release_calldata(task->tk_ops, task->tk_calldata);
+
+	if (tk_flags & RPC_TASK_DYNAMIC) {
+		dprintk("RPC: %5u freeing task\n", task->tk_pid);
+		mempool_free(task, rpc_task_mempool);
 	}
-	if (task->tk_flags & RPC_TASK_DYNAMIC)
-		call_rcu_bh(&task->u.tk_rcu, rpc_free_task);
-	rpc_release_calldata(tk_ops, calldata);
 }
-EXPORT_SYMBOL_GPL(rpc_put_task);
 
-static void rpc_release_task(struct rpc_task *task)
+static void rpc_async_release(struct work_struct *work)
 {
-#ifdef RPC_DEBUG
-	BUG_ON(task->tk_magic != RPC_TASK_MAGIC_ID);
-#endif
-	dprintk("RPC: %5u release task\n", task->tk_pid);
+	rpc_free_task(container_of(work, struct rpc_task, u.tk_work));
+}
 
-	if (!list_empty(&task->tk_task)) {
-		struct rpc_clnt *clnt = task->tk_client;
-		/* Remove from client task list */
-		spin_lock(&clnt->cl_lock);
-		list_del(&task->tk_task);
-		spin_unlock(&clnt->cl_lock);
+static void rpc_release_resources_task(struct rpc_task *task)
+{
+	xprt_release(task);
+	if (task->tk_msg.rpc_cred) {
+		put_rpccred(task->tk_msg.rpc_cred);
+		task->tk_msg.rpc_cred = NULL;
 	}
-	BUG_ON (RPC_IS_QUEUED(task));
+	rpc_task_release_client(task);
+}
 
-	/* Synchronously delete any running timer */
-	rpc_delete_timer(task);
+static void rpc_final_put_task(struct rpc_task *task,
+		struct workqueue_struct *q)
+{
+	if (q != NULL) {
+		INIT_WORK(&task->u.tk_work, rpc_async_release);
+		queue_work(q, &task->u.tk_work);
+	} else
+		rpc_free_task(task);
+}
 
-#ifdef RPC_DEBUG
-	task->tk_magic = 0;
-#endif
-	/* Wake up anyone who is waiting for task completion */
-	rpc_mark_complete_task(task);
+static void rpc_do_put_task(struct rpc_task *task, struct workqueue_struct *q)
+{
+	if (atomic_dec_and_test(&task->tk_count)) {
+		rpc_release_resources_task(task);
+		rpc_final_put_task(task, q);
+	}
+}
 
-	rpc_put_task(task);
+void rpc_put_task(struct rpc_task *task)
+{
+	rpc_do_put_task(task, NULL);
 }
+EXPORT_SYMBOL_GPL(rpc_put_task);
 
-/*
- * Kill all tasks for the given client.
- * XXX: kill their descendants as well?
- */
-void rpc_killall_tasks(struct rpc_clnt *clnt)
+void rpc_put_task_async(struct rpc_task *task)
+{
+	rpc_do_put_task(task, task->tk_workqueue);
+}
+EXPORT_SYMBOL_GPL(rpc_put_task_async);
+
+static void rpc_release_task(struct rpc_task *task)
 {
-	struct rpc_task	*rovr;
+	dprintk("RPC: %5u release task\n", task->tk_pid);
 
+	WARN_ON_ONCE(RPC_IS_QUEUED(task));
+
+	rpc_release_resources_task(task);
 
-	if (list_empty(&clnt->cl_tasks))
-		return;
-	dprintk("RPC:       killing all tasks for client %p\n", clnt);
 	/*
-	 * Spin lock all_tasks to prevent changes...
+	 * Note: at this point we have been removed from rpc_clnt->cl_tasks,
+	 * so it should be safe to use task->tk_count as a test for whether
+	 * or not any other processes still hold references to our rpc_task.
 	 */
-	spin_lock(&clnt->cl_lock);
-	list_for_each_entry(rovr, &clnt->cl_tasks, tk_task) {
-		if (! RPC_IS_ACTIVATED(rovr))
-			continue;
-		if (!(rovr->tk_flags & RPC_TASK_KILLED)) {
-			rovr->tk_flags |= RPC_TASK_KILLED;
-			rpc_exit(rovr, -EIO);
-			rpc_wake_up_task(rovr);
-		}
+	if (atomic_read(&task->tk_count) != 1 + !RPC_IS_ASYNC(task)) {
+		/* Wake up anyone who may be waiting for task completion */
+		if (!rpc_complete_task(task))
+			return;
+	} else {
+		if (!atomic_dec_and_test(&task->tk_count))
+			return;
 	}
-	spin_unlock(&clnt->cl_lock);
+	rpc_final_put_task(task, task->tk_workqueue);
 }
-EXPORT_SYMBOL_GPL(rpc_killall_tasks);
 
 int rpciod_up(void)
 {
@@ -999,7 +1071,7 @@ static int rpciod_start(void)
 	 * Create the rpciod thread and wait for it to start.
 	 */
 	dprintk("RPC:       creating workqueue rpciod\n");
-	wq = create_workqueue("rpciod");
+	wq = alloc_workqueue("rpciod", WQ_MEM_RECLAIM, 1);
 	rpciod_workqueue = wq;
 	return rpciod_workqueue != NULL;
 }
@@ -1029,11 +1101,20 @@ rpc_destroy_mempool(void)
 		kmem_cache_destroy(rpc_task_slabp);
 	if (rpc_buffer_slabp)
 		kmem_cache_destroy(rpc_buffer_slabp);
+	rpc_destroy_wait_queue(&delay_queue);
 }
 
 int
 rpc_init_mempool(void)
 {
+	/*
+	 * The following is not strictly a mempool initialisation,
+	 * but there is no harm in doing it here
+	 */
+	rpc_init_wait_queue(&delay_queue, "delayq");
+	if (!rpciod_start())
+		goto err_nomem;
+
 	rpc_task_slabp = kmem_cache_create("rpc_tasks",
 					     sizeof(struct rpc_task),
 					     0, SLAB_HWCACHE_ALIGN,
@@ -1054,13 +1135,6 @@ rpc_init_mempool(void)
 						      rpc_buffer_slabp);
 	if (!rpc_buffer_mempool)
 		goto err_nomem;
-	if (!rpciod_start())
-		goto err_nomem;
-	/*
-	 * The following is not strictly a mempool initialisation,
-	 * but there is no harm in doing it here
-	 */
-	rpc_init_wait_queue(&delay_queue, "delayq");
 	return 0;
 err_nomem:
 	rpc_destroy_mempool();