[PATCH 3/7 V6] workqueue: add manager pointer for worker_pool

[Lai Jiangshan]

We have this plan for manage_workers(): if failed to grab
manager_mutex via mutex_trylock(), we will release gcwq->lock and then
grab manager_mutex again.

This plan will open a hole: hotplug is running after we release gcwq->lock,
and it will not handle the binding of manager. so we add ->manager
on worker_pool and let hotplug code(gcwq_unbind_fn()) handle it.

also fix too_many_workers() to use this pointer.

Signed-off-by: Lai Jiangshan <laijs@xxxxxxxxxxxxxx>
---
 kernel/workqueue.c |   12 ++++++++++--
 1 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index 3dd7ce2..b203806 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -165,6 +165,7 @@ struct worker_pool {
 	struct timer_list	idle_timer;	/* L: worker idle timeout */
 	struct timer_list	mayday_timer;	/* L: SOS timer for workers */
 
+	struct worker		*manager;	/* L: manager worker */
 	struct mutex		manager_mutex;	/* mutex manager should hold */
 	struct ida		worker_ida;	/* L: for worker IDs */
 };
@@ -680,7 +681,7 @@ 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->manager;
 	int nr_idle = pool->nr_idle + managing; /* manager is considered idle */
 	int nr_busy = pool->nr_workers - nr_idle;
 
@@ -2066,6 +2067,7 @@ static bool manage_workers(struct worker *worker)
 	if (!mutex_trylock(&pool->manager_mutex))
 		return ret;
 
+	pool->manager = worker;
 	pool->flags &= ~POOL_MANAGE_WORKERS;
 
 	/*
@@ -2076,6 +2078,8 @@ static bool manage_workers(struct worker *worker)
 	ret |= maybe_create_worker(pool);
 
 	mutex_unlock(&pool->manager_mutex);
+	pool->manager = NULL;
+
 	return ret;
 }
 
@@ -3438,9 +3442,12 @@ static void gcwq_unbind_fn(struct work_struct *work)
 	 * ones which are still executing works from before the last CPU
 	 * down must be on the cpu.  After this, they may become diasporas.
 	 */
-	for_each_worker_pool(pool, gcwq)
+	for_each_worker_pool(pool, gcwq) {
 		list_for_each_entry(worker, &pool->idle_list, entry)
 			worker->flags |= WORKER_UNBOUND;
+		if (pool->manager)
+			pool->manager->flags |= WORKER_UNBOUND;
+	}
 
 	for_each_busy_worker(worker, i, pos, gcwq)
 		worker->flags |= WORKER_UNBOUND;
@@ -3760,6 +3767,7 @@ static int __init init_workqueues(void)
 			setup_timer(&pool->mayday_timer, gcwq_mayday_timeout,
 				    (unsigned long)pool);
 
+			pool->manager = NULL;
 			mutex_init(&pool->manager_mutex);
 			ida_init(&pool->worker_ida);
 		}
-- 
1.7.4.4

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/