[PATCH v4 2/3] workqueue: defer the worker wakeup outside pool->lock in __queue_work()

[Breno Leitao <leitao@debian.org>]

__queue_work() is the enqueue hot path: it inserts the work item and
calls kick_pool() while holding pool->lock. kick_pool() ends in a
wakeup, which takes the target task's rq->lock, so rq->lock nests under
pool->lock on every enqueue that wakes a worker on a contended unbound
pool.

Use kick_pool_pick() to select and claim the worker under pool->lock,
queue it on an on-stack wake_q, and issue the wakeup with wake_up_q()
right after dropping the lock. Worker selection, wake_cpu setup and
claiming the worker off pool->idle_list still happen under the lock;
only the rq->lock acquisition moves out.

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 kernel/workqueue.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index fd3b5bc78df9e..972f783f98281 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -2299,6 +2299,7 @@ static void __queue_work(int cpu, struct workqueue_struct *wq,
 {
 	struct pool_workqueue *pwq;
 	struct worker_pool *last_pool, *pool;
+	DEFINE_WAKE_Q(wakeq);
 	unsigned int work_flags;
 	unsigned int req_cpu = cpu;
 
@@ -2421,7 +2422,7 @@ static void __queue_work(int cpu, struct workqueue_struct *wq,
 
 		trace_workqueue_activate_work(work);
 		insert_work(pwq, work, &pool->worklist, work_flags);
-		kick_pool(pool);
+		kick_pool_pick(pool, &wakeq);
 	} else {
 		work_flags |= WORK_STRUCT_INACTIVE;
 		insert_work(pwq, work, &pwq->inactive_works, work_flags);
@@ -2429,6 +2430,7 @@ static void __queue_work(int cpu, struct workqueue_struct *wq,
 
 out:
 	raw_spin_unlock(&pool->lock);
+	wake_up_q(&wakeq);
 	rcu_read_unlock();
 }
 

-- 
2.53.0-Meta