[PATCH] workqueue: Skip non-freezable workqueues in freeze and thaw

[PATCH] workqueue: Skip non-freezable workqueues in freeze and thaw

[Samuel Wu]

In wq_adjust_max_active(), non-freezable workqueues always keep
saved_max_active and saved_min_active, so updating them during freeze
and thaw is a no-op. However, both freeze and thaw acquire mutexes and
adjust max_active for every wq on the wq list, which is extra work and
could increase contention.

Skip workqueues without WQ_FREEZABLE in freeze and thaw.

On a v7.3 Pixel 9 device, this consistently reduces thaw_processes
latency during system suspend by 4-5% (~0.7ms).

Signed-off-by: Samuel Wu <wusamuel@google.com>
---
 kernel/workqueue.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index edbb2aa9c950..902f84bcfd24 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -7370,6 +7370,8 @@ void freeze_workqueues_begin(void)
 	workqueue_freezing = true;
 
 	list_for_each_entry(wq, &workqueues, list) {
+		if (!(wq->flags & WQ_FREEZABLE))
+			continue;
 		mutex_lock(&wq->mutex);
 		wq_adjust_max_active(wq);
 		mutex_unlock(&wq->mutex);
@@ -7446,6 +7448,8 @@ void thaw_workqueues(void)
 
 	/* restore max_active and repopulate worklist */
 	list_for_each_entry(wq, &workqueues, list) {
+		if (!(wq->flags & WQ_FREEZABLE))
+			continue;
 		mutex_lock(&wq->mutex);
 		wq_adjust_max_active(wq);
 		mutex_unlock(&wq->mutex);
-- 
2.56.0.679.gc527c65306-goog

Re: [PATCH] workqueue: Skip non-freezable workqueues in freeze and thaw

[Tejun Heo]

Applied to wq/for-7.5.

Thanks.

--
tejun