[RFC PATCH] sched/wait: Determine whether the wait queue is empty before waking up

From: Wang You <wangyoua@uniontech.com>
To: mingo@redhat.com, peterz@infradead.org, juri.lelli@redhat.com,
	vincent.guittot@linaro.org, dietmar.eggemann@arm.com,
	rostedt@goodmis.org, bsegall@google.com, mgorman@suse.de,
	bristot@redhat.com, vschneid@redhat.com
Cc: linux-kernel@vger.kernel.org, Wang You <wangyoua@uniontech.com>
Subject: [RFC PATCH] sched/wait: Determine whether the wait queue is empty before waking up
Date: Fri,  9 Jun 2023 13:38:12 +0800	[thread overview]
Message-ID: <20230609053812.10230-1-wangyoua@uniontech.com> (raw)

When we did some benchmark tests (such as pipe tests), we found
that the wake behavior was still triggered when the wait queue
was empty, even though it would exit later.

This processing consumes some unnecessary resources. Can we
determine at the beginning of the wake up whether there are
elements in the queue that need to be awakened? I think this
judgment is probably good for performance and should be safe.

Looking forward to your reply, thank you.

Signed-off-by: Wang You <wangyoua@uniontech.com>
---
 kernel/sched/wait.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/kernel/sched/wait.c b/kernel/sched/wait.c
index 133b74730738..17011780aa21 100644
--- a/kernel/sched/wait.c
+++ b/kernel/sched/wait.c
@@ -205,6 +205,9 @@ void __wake_up_sync_key(struct wait_queue_head *wq_head, unsigned int mode,
 	if (unlikely(!wq_head))
 		return;
 
+	if (unlikely(!wq_has_sleeper(wq_head)))
+		return;
+
 	__wake_up_common_lock(wq_head, mode, 1, WF_SYNC, key);
 }
 EXPORT_SYMBOL_GPL(__wake_up_sync_key);
-- 
2.20.1

