Merge branch 'for-3.6-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq

Pull workqueue fixes from Tejun Heo:
 "It's later than I'd like but well the timing just didn't work out this
  time.

  There are three bug fixes.  One from before 3.6-rc1 and two from the
  new CPU hotplug code.  Kudos to Lai for discovering all of them and
  providing fixes.

   * Atomicity bug when clearing a flag and setting another.  The two
     operation should have been atomic but wasn't.  This bug has existed
     for a long time but is unlikely to have actually happened.  Fix is
     safe.  Marked for -stable.

   * If CPU hotplug cycles happen back-to-back before workers finish the
     previous cycle, the states could get out of sync and it could get
     stuck.  Fixed by waiting for workers to complete before finishing
     hotplug cycle.

   * While CPU hotplug is in progress, idle workers could be depleted
     which can then lead to deadlock.  I think both happening together
     is highly unlikely but still better to fix it and the fix isn't too
     scary.

  There's another workqueue related regression which reported a few days
  ago:

    https://bugzilla.kernel.org/show_bug.cgi?id=47301

  It's a bit of head scratcher but there is a semi-reliable reproduce
  case, so I'm hoping to resolve it soonish."

* 'for-3.6-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
  workqueue: fix possible idle worker depletion across CPU hotplug
  workqueue: restore POOL_MANAGING_WORKERS
  workqueue: fix possible deadlock in idle worker rebinding
  workqueue: move WORKER_REBIND clearing in rebind_workers() to the end of the function
  workqueue: UNBOUND -> REBIND morphing in rebind_workers() should be atomic

1 files changed, 89 insertions, 21 deletions

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index 692d976..1e1373b 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -66,6 +66,7 @@ enum {
 
 	/* pool flags */
 	POOL_MANAGE_WORKERS	= 1 << 0,	/* need to manage workers */
+	POOL_MANAGING_WORKERS   = 1 << 1,       /* managing workers */
 
 	/* worker flags */
 	WORKER_STARTED		= 1 << 0,	/* started */
@@ -652,7 +653,7 @@ static bool need_to_manage_workers(struct worker_pool *pool)
 /* Do we have too many workers and should some go away? */
 static bool too_many_workers(struct worker_pool *pool)
 {
-	bool managing = mutex_is_locked(&pool->manager_mutex);
+	bool managing = pool->flags & POOL_MANAGING_WORKERS;
 	int nr_idle = pool->nr_idle + managing; /* manager is considered idle */
 	int nr_busy = pool->nr_workers - nr_idle;
 
@@ -1326,6 +1327,15 @@ static void idle_worker_rebind(struct worker *worker)
 
 	/* we did our part, wait for rebind_workers() to finish up */
 	wait_event(gcwq->rebind_hold, !(worker->flags & WORKER_REBIND));
+
+	/*
+	 * rebind_workers() shouldn't finish until all workers passed the
+	 * above WORKER_REBIND wait.  Tell it when done.
+	 */
+	spin_lock_irq(&worker->pool->gcwq->lock);
+	if (!--worker->idle_rebind->cnt)
+		complete(&worker->idle_rebind->done);
+	spin_unlock_irq(&worker->pool->gcwq->lock);
 }
 
 /*
@@ -1396,12 +1406,15 @@ retry:
 	/* set REBIND and kick idle ones, we'll wait for these later */
 	for_each_worker_pool(pool, gcwq) {
 		list_for_each_entry(worker, &pool->idle_list, entry) {
+			unsigned long worker_flags = worker->flags;
+
 			if (worker->flags & WORKER_REBIND)
 				continue;
 
-			/* morph UNBOUND to REBIND */
-			worker->flags &= ~WORKER_UNBOUND;
-			worker->flags |= WORKER_REBIND;
+			/* morph UNBOUND to REBIND atomically */
+			worker_flags &= ~WORKER_UNBOUND;
+			worker_flags |= WORKER_REBIND;
+			ACCESS_ONCE(worker->flags) = worker_flags;
 
 			idle_rebind.cnt++;
 			worker->idle_rebind = &idle_rebind;
@@ -1419,25 +1432,15 @@ retry:
 		goto retry;
 	}
 
-	/*
-	 * All idle workers are rebound and waiting for %WORKER_REBIND to
-	 * be cleared inside idle_worker_rebind().  Clear and release.
-	 * Clearing %WORKER_REBIND from this foreign context is safe
-	 * because these workers are still guaranteed to be idle.
-	 */
-	for_each_worker_pool(pool, gcwq)
-		list_for_each_entry(worker, &pool->idle_list, entry)
-			worker->flags &= ~WORKER_REBIND;
-
-	wake_up_all(&gcwq->rebind_hold);
-
-	/* rebind busy workers */
+	/* all idle workers are rebound, rebind busy workers */
 	for_each_busy_worker(worker, i, pos, gcwq) {
 		struct work_struct *rebind_work = &worker->rebind_work;
+		unsigned long worker_flags = worker->flags;
 
-		/* morph UNBOUND to REBIND */
-		worker->flags &= ~WORKER_UNBOUND;
-		worker->flags |= WORKER_REBIND;
+		/* morph UNBOUND to REBIND atomically */
+		worker_flags &= ~WORKER_UNBOUND;
+		worker_flags |= WORKER_REBIND;
+		ACCESS_ONCE(worker->flags) = worker_flags;
 
 		if (test_and_set_bit(WORK_STRUCT_PENDING_BIT,
 				     work_data_bits(rebind_work)))
@@ -1449,6 +1452,34 @@ retry:
 			    worker->scheduled.next,
 			    work_color_to_flags(WORK_NO_COLOR));
 	}
+
+	/*
+	 * All idle workers are rebound and waiting for %WORKER_REBIND to
+	 * be cleared inside idle_worker_rebind().  Clear and release.
+	 * Clearing %WORKER_REBIND from this foreign context is safe
+	 * because these workers are still guaranteed to be idle.
+	 *
+	 * We need to make sure all idle workers passed WORKER_REBIND wait
+	 * in idle_worker_rebind() before returning; otherwise, workers can
+	 * get stuck at the wait if hotplug cycle repeats.
+	 */
+	idle_rebind.cnt = 1;
+	INIT_COMPLETION(idle_rebind.done);
+
+	for_each_worker_pool(pool, gcwq) {
+		list_for_each_entry(worker, &pool->idle_list, entry) {
+			worker->flags &= ~WORKER_REBIND;
+			idle_rebind.cnt++;
+		}
+	}
+
+	wake_up_all(&gcwq->rebind_hold);
+
+	if (--idle_rebind.cnt) {
+		spin_unlock_irq(&gcwq->lock);
+		wait_for_completion(&idle_rebind.done);
+		spin_lock_irq(&gcwq->lock);
+	}
 }
 
 static struct worker *alloc_worker(void)
@@ -1794,9 +1825,45 @@ static bool manage_workers(struct worker *worker)
 	struct worker_pool *pool = worker->pool;
 	bool ret = false;
 
-	if (!mutex_trylock(&pool->manager_mutex))
+	if (pool->flags & POOL_MANAGING_WORKERS)
 		return ret;
 
+	pool->flags |= POOL_MANAGING_WORKERS;
+
+	/*
+	 * To simplify both worker management and CPU hotplug, hold off
+	 * management while hotplug is in progress.  CPU hotplug path can't
+	 * grab %POOL_MANAGING_WORKERS to achieve this because that can
+	 * lead to idle worker depletion (all become busy thinking someone
+	 * else is managing) which in turn can result in deadlock under
+	 * extreme circumstances.  Use @pool->manager_mutex to synchronize
+	 * manager against CPU hotplug.
+	 *
+	 * manager_mutex would always be free unless CPU hotplug is in
+	 * progress.  trylock first without dropping @gcwq->lock.
+	 */
+	if (unlikely(!mutex_trylock(&pool->manager_mutex))) {
+		spin_unlock_irq(&pool->gcwq->lock);
+		mutex_lock(&pool->manager_mutex);
+		/*
+		 * CPU hotplug could have happened while we were waiting
+		 * for manager_mutex.  Hotplug itself can't handle us
+		 * because manager isn't either on idle or busy list, and
+		 * @gcwq's state and ours could have deviated.
+		 *
+		 * As hotplug is now excluded via manager_mutex, we can
+		 * simply try to bind.  It will succeed or fail depending
+		 * on @gcwq's current state.  Try it and adjust
+		 * %WORKER_UNBOUND accordingly.
+		 */
+		if (worker_maybe_bind_and_lock(worker))
+			worker->flags &= ~WORKER_UNBOUND;
+		else
+			worker->flags |= WORKER_UNBOUND;
+
+		ret = true;
+	}
+
 	pool->flags &= ~POOL_MANAGE_WORKERS;
 
 	/*
@@ -1806,6 +1873,7 @@ static bool manage_workers(struct worker *worker)
 	ret |= maybe_destroy_workers(pool);
 	ret |= maybe_create_worker(pool);
 
+	pool->flags &= ~POOL_MANAGING_WORKERS;
 	mutex_unlock(&pool->manager_mutex);
 	return ret;
 }