3 files changed, 64 insertions, 151 deletions

diff --git a/include/linux/kthread.h b/include/linux/kthread.h
index 7dcef3317689..13d55206ccf6 100644
--- a/include/linux/kthread.h
+++ b/include/linux/kthread.h
@@ -73,7 +73,6 @@ struct kthread_worker {
 struct kthread_work {
 	struct list_head	node;
 	kthread_work_func_t	func;
-	wait_queue_head_t	done;
 	struct kthread_worker	*worker;
 };
 
@@ -85,7 +84,6 @@ struct kthread_work {
 #define KTHREAD_WORK_INIT(work, fn)	{				\
 	.node = LIST_HEAD_INIT((work).node),				\
 	.func = (fn),							\
-	.done = __WAIT_QUEUE_HEAD_INITIALIZER((work).done),		\
 	}
 
 #define DEFINE_KTHREAD_WORKER(worker)					\
@@ -95,22 +93,16 @@ struct kthread_work {
 	struct kthread_work work = KTHREAD_WORK_INIT(work, fn)
 
 /*
- * kthread_worker.lock and kthread_work.done need their own lockdep class
- * keys if they are defined on stack with lockdep enabled.  Use the
- * following macros when defining them on stack.
+ * kthread_worker.lock needs its own lockdep class key when defined on
+ * stack with lockdep enabled.  Use the following macros in such cases.
  */
 #ifdef CONFIG_LOCKDEP
 # define KTHREAD_WORKER_INIT_ONSTACK(worker)				\
 	({ init_kthread_worker(&worker); worker; })
 # define DEFINE_KTHREAD_WORKER_ONSTACK(worker)				\
 	struct kthread_worker worker = KTHREAD_WORKER_INIT_ONSTACK(worker)
-# define KTHREAD_WORK_INIT_ONSTACK(work, fn)				\
-	({ init_kthread_work((&work), fn); work; })
-# define DEFINE_KTHREAD_WORK_ONSTACK(work, fn)				\
-	struct kthread_work work = KTHREAD_WORK_INIT_ONSTACK(work, fn)
 #else
 # define DEFINE_KTHREAD_WORKER_ONSTACK(worker) DEFINE_KTHREAD_WORKER(worker)
-# define DEFINE_KTHREAD_WORK_ONSTACK(work, fn) DEFINE_KTHREAD_WORK(work, fn)
 #endif
 
 extern void __init_kthread_worker(struct kthread_worker *worker,
@@ -127,7 +119,6 @@ extern void __init_kthread_worker(struct kthread_worker *worker,
 		memset((work), 0, sizeof(struct kthread_work));		\
 		INIT_LIST_HEAD(&(work)->node);				\
 		(work)->func = (fn);					\
-		init_waitqueue_head(&(work)->done);			\
 	} while (0)
 
 int kthread_worker_fn(void *worker_ptr);
diff --git a/kernel/kthread.c b/kernel/kthread.c
index c2390f41307b..ef483220e855 100644
--- a/kernel/kthread.c
+++ b/kernel/kthread.c
@@ -591,7 +591,7 @@ static void insert_kthread_work(struct kthread_worker *worker,
 
 	list_add_tail(&work->node, pos);
 	work->worker = worker;
-	if (likely(worker->task))
+	if (!worker->current_work && likely(worker->task))
 		wake_up_process(worker->task);
 }
 
diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index 35974ac69600..7a2e449a96b1 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -265,7 +265,6 @@ struct workqueue_struct {
 
 static struct kmem_cache *pwq_cache;
 
-static int wq_numa_tbl_len;		/* highest possible NUMA node id + 1 */
 static cpumask_var_t *wq_numa_possible_cpumask;
 					/* possible CPUs of each node */
 
@@ -758,13 +757,6 @@ static bool too_many_workers(struct worker_pool *pool)
 	int nr_idle = pool->nr_idle + managing; /* manager is considered idle */
 	int nr_busy = pool->nr_workers - nr_idle;
 
-	/*
-	 * nr_idle and idle_list may disagree if idle rebinding is in
-	 * progress.  Never return %true if idle_list is empty.
-	 */
-	if (list_empty(&pool->idle_list))
-		return false;
-
 	return nr_idle > 2 && (nr_idle - 2) * MAX_IDLE_WORKERS_RATIO >= nr_busy;
 }
 
@@ -850,7 +842,7 @@ struct task_struct *wq_worker_sleeping(struct task_struct *task, int cpu)
 	pool = worker->pool;
 
 	/* this can only happen on the local cpu */
-	if (WARN_ON_ONCE(cpu != raw_smp_processor_id()))
+	if (WARN_ON_ONCE(cpu != raw_smp_processor_id() || pool->cpu != cpu))
 		return NULL;
 
 	/*
@@ -874,35 +866,22 @@ struct task_struct *wq_worker_sleeping(struct task_struct *task, int cpu)
  * worker_set_flags - set worker flags and adjust nr_running accordingly
  * @worker: self
  * @flags: flags to set
- * @wakeup: wakeup an idle worker if necessary
  *
- * Set @flags in @worker->flags and adjust nr_running accordingly.  If
- * nr_running becomes zero and @wakeup is %true, an idle worker is
- * woken up.
+ * Set @flags in @worker->flags and adjust nr_running accordingly.
  *
  * CONTEXT:
  * spin_lock_irq(pool->lock)
  */
-static inline void worker_set_flags(struct worker *worker, unsigned int flags,
-				    bool wakeup)
+static inline void worker_set_flags(struct worker *worker, unsigned int flags)
 {
 	struct worker_pool *pool = worker->pool;
 
 	WARN_ON_ONCE(worker->task != current);
 
-	/*
-	 * If transitioning into NOT_RUNNING, adjust nr_running and
-	 * wake up an idle worker as necessary if requested by
-	 * @wakeup.
-	 */
+	/* If transitioning into NOT_RUNNING, adjust nr_running. */
 	if ((flags & WORKER_NOT_RUNNING) &&
 	    !(worker->flags & WORKER_NOT_RUNNING)) {
-		if (wakeup) {
-			if (atomic_dec_and_test(&pool->nr_running) &&
-			    !list_empty(&pool->worklist))
-				wake_up_worker(pool);
-		} else
-			atomic_dec(&pool->nr_running);
+		atomic_dec(&pool->nr_running);
 	}
 
 	worker->flags |= flags;
@@ -1232,7 +1211,7 @@ static int try_to_grab_pending(struct work_struct *work, bool is_dwork,
 			pwq_activate_delayed_work(work);
 
 		list_del_init(&work->entry);
-		pwq_dec_nr_in_flight(get_work_pwq(work), get_work_color(work));
+		pwq_dec_nr_in_flight(pwq, get_work_color(work));
 
 		/* work->data points to pwq iff queued, point to pool */
 		set_work_pool_and_keep_pending(work, pool->id);
@@ -1560,7 +1539,7 @@ static void worker_enter_idle(struct worker *worker)
 			 (worker->hentry.next || worker->hentry.pprev)))
 		return;
 
-	/* can't use worker_set_flags(), also called from start_worker() */
+	/* can't use worker_set_flags(), also called from create_worker() */
 	worker->flags |= WORKER_IDLE;
 	pool->nr_idle++;
 	worker->last_active = jiffies;
@@ -1602,11 +1581,11 @@ static void worker_leave_idle(struct worker *worker)
 	list_del_init(&worker->entry);
 }
 
-static struct worker *alloc_worker(void)
+static struct worker *alloc_worker(int node)
 {
 	struct worker *worker;
 
-	worker = kzalloc(sizeof(*worker), GFP_KERNEL);
+	worker = kzalloc_node(sizeof(*worker), GFP_KERNEL, node);
 	if (worker) {
 		INIT_LIST_HEAD(&worker->entry);
 		INIT_LIST_HEAD(&worker->scheduled);
@@ -1670,6 +1649,9 @@ static void worker_detach_from_pool(struct worker *worker,
 		detach_completion = pool->detach_completion;
 	mutex_unlock(&pool->attach_mutex);
 
+	/* clear leftover flags without pool->lock after it is detached */
+	worker->flags &= ~(WORKER_UNBOUND | WORKER_REBOUND);
+
 	if (detach_completion)
 		complete(detach_completion);
 }
@@ -1678,8 +1660,7 @@ static void worker_detach_from_pool(struct worker *worker,
  * create_worker - create a new workqueue worker
  * @pool: pool the new worker will belong to
  *
- * Create a new worker which is attached to @pool.  The new worker must be
- * started by start_worker().
+ * Create and start a new worker which is attached to @pool.
  *
  * CONTEXT:
  * Might sleep.  Does GFP_KERNEL allocations.
@@ -1698,7 +1679,7 @@ static struct worker *create_worker(struct worker_pool *pool)
 	if (id < 0)
 		goto fail;
 
-	worker = alloc_worker();
+	worker = alloc_worker(pool->node);
 	if (!worker)
 		goto fail;
 
@@ -1724,6 +1705,13 @@ static struct worker *create_worker(struct worker_pool *pool)
 	/* successful, attach the worker to the pool */
 	worker_attach_to_pool(worker, pool);
 
+	/* start the newly created worker */
+	spin_lock_irq(&pool->lock);
+	worker->pool->nr_workers++;
+	worker_enter_idle(worker);
+	wake_up_process(worker->task);
+	spin_unlock_irq(&pool->lock);
+
 	return worker;
 
 fail:
@@ -1734,44 +1722,6 @@ fail:
 }
 
 /**
- * start_worker - start a newly created worker
- * @worker: worker to start
- *
- * Make the pool aware of @worker and start it.
- *
- * CONTEXT:
- * spin_lock_irq(pool->lock).
- */
-static void start_worker(struct worker *worker)
-{
-	worker->pool->nr_workers++;
-	worker_enter_idle(worker);
-	wake_up_process(worker->task);
-}
-
-/**
- * create_and_start_worker - create and start a worker for a pool
- * @pool: the target pool
- *
- * Grab the managership of @pool and create and start a new worker for it.
- *
- * Return: 0 on success. A negative error code otherwise.
- */
-static int create_and_start_worker(struct worker_pool *pool)
-{
-	struct worker *worker;
-
-	worker = create_worker(pool);
-	if (worker) {
-		spin_lock_irq(&pool->lock);
-		start_worker(worker);
-		spin_unlock_irq(&pool->lock);
-	}
-
-	return worker ? 0 : -ENOMEM;
-}
-
-/**
  * destroy_worker - destroy a workqueue worker
  * @worker: worker to be destroyed
  *
@@ -1909,23 +1859,10 @@ restart:
 	mod_timer(&pool->mayday_timer, jiffies + MAYDAY_INITIAL_TIMEOUT);
 
 	while (true) {
-		struct worker *worker;
-
-		worker = create_worker(pool);
-		if (worker) {
-			del_timer_sync(&pool->mayday_timer);
-			spin_lock_irq(&pool->lock);
-			start_worker(worker);
-			if (WARN_ON_ONCE(need_to_create_worker(pool)))
-				goto restart;
-			return true;
-		}
-
-		if (!need_to_create_worker(pool))
+		if (create_worker(pool) || !need_to_create_worker(pool))
 			break;
 
-		__set_current_state(TASK_INTERRUPTIBLE);
-		schedule_timeout(CREATE_COOLDOWN);
+		schedule_timeout_interruptible(CREATE_COOLDOWN);
 
 		if (!need_to_create_worker(pool))
 			break;
@@ -1933,6 +1870,11 @@ restart:
 
 	del_timer_sync(&pool->mayday_timer);
 	spin_lock_irq(&pool->lock);
+	/*
+	 * This is necessary even after a new worker was just successfully
+	 * created as @pool->lock was dropped and the new worker might have
+	 * already become busy.
+	 */
 	if (need_to_create_worker(pool))
 		goto restart;
 	return true;
@@ -2020,13 +1962,7 @@ __acquires(&pool->lock)
 
 	lockdep_copy_map(&lockdep_map, &work->lockdep_map);
 #endif
-	/*
-	 * Ensure we're on the correct CPU.  DISASSOCIATED test is
-	 * necessary to avoid spurious warnings from rescuers servicing the
-	 * unbound or a disassociated pool.
-	 */
-	WARN_ON_ONCE(!(worker->flags & WORKER_UNBOUND) &&
-		     !(pool->flags & POOL_DISASSOCIATED) &&
+	WARN_ON_ONCE(!(pool->flags & POOL_DISASSOCIATED) &&
 		     raw_smp_processor_id() != pool->cpu);
 
 	/*
@@ -2052,17 +1988,22 @@ __acquires(&pool->lock)
 	list_del_init(&work->entry);
 
 	/*
-	 * CPU intensive works don't participate in concurrency
-	 * management.  They're the scheduler's responsibility.
+	 * CPU intensive works don't participate in concurrency management.
+	 * They're the scheduler's responsibility.  This takes @worker out
+	 * of concurrency management and the next code block will chain
+	 * execution of the pending work items.
 	 */
 	if (unlikely(cpu_intensive))
-		worker_set_flags(worker, WORKER_CPU_INTENSIVE, true);
+		worker_set_flags(worker, WORKER_CPU_INTENSIVE);
 
 	/*
-	 * Unbound pool isn't concurrency managed and work items should be
-	 * executed ASAP.  Wake up another worker if necessary.
+	 * Wake up another worker if necessary.  The condition is always
+	 * false for normal per-cpu workers since nr_running would always
+	 * be >= 1 at this point.  This is used to chain execution of the
+	 * pending work items for WORKER_NOT_RUNNING workers such as the
+	 * UNBOUND and CPU_INTENSIVE ones.
 	 */
-	if ((worker->flags & WORKER_UNBOUND) && need_more_worker(pool))
+	if (need_more_worker(pool))
 		wake_up_worker(pool);
 
 	/*
@@ -2218,7 +2159,7 @@ recheck:
 		}
 	} while (keep_working(pool));
 
-	worker_set_flags(worker, WORKER_PREP, false);
+	worker_set_flags(worker, WORKER_PREP);
 sleep:
 	/*
 	 * pool->lock is held and there's no work to process and no need to
@@ -2311,29 +2252,27 @@ repeat:
 				move_linked_works(work, scheduled, &n);
 
 		process_scheduled_works(rescuer);
-		spin_unlock_irq(&pool->lock);
-
-		worker_detach_from_pool(rescuer, pool);
-
-		spin_lock_irq(&pool->lock);
 
 		/*
 		 * Put the reference grabbed by send_mayday().  @pool won't
-		 * go away while we're holding its lock.
+		 * go away while we're still attached to it.
 		 */
 		put_pwq(pwq);
 
 		/*
-		 * Leave this pool.  If keep_working() is %true, notify a
+		 * Leave this pool.  If need_more_worker() is %true, notify a
 		 * regular worker; otherwise, we end up with 0 concurrency
 		 * and stalling the execution.
 		 */
-		if (keep_working(pool))
+		if (need_more_worker(pool))
 			wake_up_worker(pool);
 
 		rescuer->pool = NULL;
-		spin_unlock(&pool->lock);
-		spin_lock(&wq_mayday_lock);
+		spin_unlock_irq(&pool->lock);
+
+		worker_detach_from_pool(rescuer, pool);
+
+		spin_lock_irq(&wq_mayday_lock);
 	}
 
 	spin_unlock_irq(&wq_mayday_lock);
@@ -3458,7 +3397,7 @@ static void put_unbound_pool(struct worker_pool *pool)
 		return;
 
 	/* sanity checks */
-	if (WARN_ON(!(pool->flags & POOL_DISASSOCIATED)) ||
+	if (WARN_ON(!(pool->cpu < 0)) ||
 	    WARN_ON(!list_empty(&pool->worklist)))
 		return;
 
@@ -3524,7 +3463,7 @@ static struct worker_pool *get_unbound_pool(const struct workqueue_attrs *attrs)
 	hash_for_each_possible(unbound_pool_hash, pool, hash_node, hash) {
 		if (wqattrs_equal(pool->attrs, attrs)) {
 			pool->refcnt++;
-			goto out_unlock;
+			return pool;
 		}
 	}
 
@@ -3557,12 +3496,12 @@ static struct worker_pool *get_unbound_pool(const struct workqueue_attrs *attrs)
 		goto fail;
 
 	/* create and start the initial worker */
-	if (create_and_start_worker(pool) < 0)
+	if (!create_worker(pool))
 		goto fail;
 
 	/* install */
 	hash_add(unbound_pool_hash, &pool->hash_node, hash);
-out_unlock:
+
 	return pool;
 fail:
 	if (pool)
@@ -3591,11 +3530,6 @@ static void pwq_unbound_release_workfn(struct work_struct *work)
 	if (WARN_ON_ONCE(!(wq->flags & WQ_UNBOUND)))
 		return;
 
-	/*
-	 * Unlink @pwq.  Synchronization against wq->mutex isn't strictly
-	 * necessary on release but do it anyway.  It's easier to verify
-	 * and consistent with the linking path.
-	 */
 	mutex_lock(&wq->mutex);
 	list_del_rcu(&pwq->pwqs_node);
 	is_last = list_empty(&wq->pwqs);
@@ -3692,10 +3626,7 @@ static void link_pwq(struct pool_workqueue *pwq)
 	if (!list_empty(&pwq->pwqs_node))
 		return;
 
-	/*
-	 * Set the matching work_color.  This is synchronized with
-	 * wq->mutex to avoid confusing flush_workqueue().
-	 */
+	/* set the matching work_color */
 	pwq->work_color = wq->work_color;
 
 	/* sync max_active to the current setting */
@@ -3832,7 +3763,7 @@ int apply_workqueue_attrs(struct workqueue_struct *wq,
 	if (WARN_ON((wq->flags & __WQ_ORDERED) && !list_empty(&wq->pwqs)))
 		return -EINVAL;
 
-	pwq_tbl = kzalloc(wq_numa_tbl_len * sizeof(pwq_tbl[0]), GFP_KERNEL);
+	pwq_tbl = kzalloc(nr_node_ids * sizeof(pwq_tbl[0]), GFP_KERNEL);
 	new_attrs = alloc_workqueue_attrs(GFP_KERNEL);
 	tmp_attrs = alloc_workqueue_attrs(GFP_KERNEL);
 	if (!pwq_tbl || !new_attrs || !tmp_attrs)
@@ -4080,7 +4011,7 @@ struct workqueue_struct *__alloc_workqueue_key(const char *fmt,
 
 	/* allocate wq and format name */
 	if (flags & WQ_UNBOUND)
-		tbl_size = wq_numa_tbl_len * sizeof(wq->numa_pwq_tbl[0]);
+		tbl_size = nr_node_ids * sizeof(wq->numa_pwq_tbl[0]);
 
 	wq = kzalloc(sizeof(*wq) + tbl_size, GFP_KERNEL);
 	if (!wq)
@@ -4122,7 +4053,7 @@ struct workqueue_struct *__alloc_workqueue_key(const char *fmt,
 	if (flags & WQ_MEM_RECLAIM) {
 		struct worker *rescuer;
 
-		rescuer = alloc_worker();
+		rescuer = alloc_worker(NUMA_NO_NODE);
 		if (!rescuer)
 			goto err_destroy;
 
@@ -4470,8 +4401,6 @@ static void wq_unbind_fn(struct work_struct *work)
 	struct worker *worker;
 
 	for_each_cpu_worker_pool(pool, cpu) {
-		WARN_ON_ONCE(cpu != smp_processor_id());
-
 		mutex_lock(&pool->attach_mutex);
 		spin_lock_irq(&pool->lock);
 
@@ -4543,6 +4472,7 @@ static void rebind_workers(struct worker_pool *pool)
 						  pool->attrs->cpumask) < 0);
 
 	spin_lock_irq(&pool->lock);
+	pool->flags &= ~POOL_DISASSOCIATED;
 
 	for_each_pool_worker(worker, pool) {
 		unsigned int worker_flags = worker->flags;
@@ -4632,7 +4562,7 @@ static int workqueue_cpu_up_callback(struct notifier_block *nfb,
 		for_each_cpu_worker_pool(pool, cpu) {
 			if (pool->nr_workers)
 				continue;
-			if (create_and_start_worker(pool) < 0)
+			if (!create_worker(pool))
 				return NOTIFY_BAD;
 		}
 		break;
@@ -4645,10 +4575,6 @@ static int workqueue_cpu_up_callback(struct notifier_block *nfb,
 			mutex_lock(&pool->attach_mutex);
 
 			if (pool->cpu == cpu) {
-				spin_lock_irq(&pool->lock);
-				pool->flags &= ~POOL_DISASSOCIATED;
-				spin_unlock_irq(&pool->lock);
-
 				rebind_workers(pool);
 			} else if (pool->cpu < 0) {
 				restore_unbound_workers_cpumask(pool, cpu);
@@ -4856,10 +4782,6 @@ static void __init wq_numa_init(void)
 	cpumask_var_t *tbl;
 	int node, cpu;
 
-	/* determine NUMA pwq table len - highest node id + 1 */
-	for_each_node(node)
-		wq_numa_tbl_len = max(wq_numa_tbl_len, node + 1);
-
 	if (num_possible_nodes() <= 1)
 		return;
 
@@ -4876,7 +4798,7 @@ static void __init wq_numa_init(void)
 	 * available.  Build one from cpu_to_node() which should have been
 	 * fully initialized by now.
 	 */
-	tbl = kzalloc(wq_numa_tbl_len * sizeof(tbl[0]), GFP_KERNEL);
+	tbl = kzalloc(nr_node_ids * sizeof(tbl[0]), GFP_KERNEL);
 	BUG_ON(!tbl);
 
 	for_each_node(node)
@@ -4936,7 +4858,7 @@ static int __init init_workqueues(void)
 
 		for_each_cpu_worker_pool(pool, cpu) {
 			pool->flags &= ~POOL_DISASSOCIATED;
-			BUG_ON(create_and_start_worker(pool) < 0);
+			BUG_ON(!create_worker(pool));
 		}
 	}