[PATCH] workqueue: check for NULL wq in queue_delayed_work_on()

[PATCH] workqueue: check for NULL wq in queue_delayed_work_on()

[baichen]

If queue_delayed_work() is called with a NULL wq, the existing
WARN_ON_ONCE(!wq) in __queue_delayed_work() only prints a warning
but doesn't prevent the subsequent NULL pointer dereference when
accessing wq->flags, causing a panic.
Add a NULL check at the beginning of queue_delayed_work_on() to
catch this error early, print a warning with stack dump, and return
false to the caller instead of crashing later.
Additionally, if the delay time is relatively long or there are too many
kernel log prints in a short time, the previous warning message may be
overwritten, making such issues difficult to debug. Relying on kernel log
alone makes it impossible to diagnose this issue, and one must resort to
fulldump to solve it.

Cc: Tejun Heo <tj@kernel.org>,
	Lai Jiangshan <jiangshanlai@gmail.com>,
	linux-kernel@vger.kernel.org
Signed-off-by: baichen.zhang <ttdxba@gmail.com>
---
 kernel/workqueue.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index cd5171208964..6cf0c5eefa07 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -2510,7 +2510,6 @@ static void __queue_delayed_work(int cpu, struct workqueue_struct *wq,
 	struct timer_list *timer = &dwork->timer;
 	struct work_struct *work = &dwork->work;
 
-	WARN_ON_ONCE(!wq);
 	WARN_ON_ONCE(timer->function != delayed_work_timer_fn);
 	WARN_ON_ONCE(timer_pending(timer));
 	WARN_ON_ONCE(!list_empty(&work->entry));
@@ -2569,6 +2568,14 @@ bool queue_delayed_work_on(int cpu, struct workqueue_struct *wq,
 	bool ret = false;
 	unsigned long irq_flags;
 
+	if (unlikely(!wq)) {
+		WARN_ON_ONCE(1);
+		pr_warn("workqueue: %s() called with NULL wq, dumping stack:\n",
+			__func__);
+		dump_stack();
+		return false;
+	}
+
 	/* read the comment in __queue_work() */
 	local_irq_save(irq_flags);
 
-- 
2.52.0

Re: [PATCH] workqueue: check for NULL wq in queue_delayed_work_on()

[Tejun Heo]

Hello,

Calling workqueue APIs with a NULL wq is always a caller bug and adding
NULL checks to all APIs isn't the right approach. The WARN was there as a
debug aid but it doesn't actually prevent the crash, so it's misleading.
I'm removing it instead.

Thanks.

--
tejun

[PATCH wq/for-7.1] workqueue: Remove NULL wq WARN in __queue_delayed_work()

[Tejun Heo]

Remove the WARN_ON_ONCE(!wq) which doesn't serve any useful purpose.

Signed-off-by: Tejun Heo <tj@kernel.org>
---
Applied to wq/for-7.1.

 kernel/workqueue.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index c1743b20a524..63acaa3e1d6a 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -2510,7 +2510,6 @@ static void __queue_delayed_work(int cpu, struct workqueue_struct *wq,
 	struct timer_list *timer = &dwork->timer;
 	struct work_struct *work = &dwork->work;

-	WARN_ON_ONCE(!wq);
 	WARN_ON_ONCE(timer->function != delayed_work_timer_fn);
 	WARN_ON_ONCE(timer_pending(timer));
 	WARN_ON_ONCE(!list_empty(&work->entry));
--
tejun