[PATCH v4] workqueue: Warn when work flush own workqueue

[PATCH v4] workqueue: Warn when work flush own workqueue

[Qianli Zhao]

From: Qianli Zhao <zhaoqianli@xiaomi.com>

If a workqueue flushes itself then that will lead to
a deadlock. Print a warning and a stack trace when
this happens.

crash> ps 10856
PID    PPID  CPU       TASK       ST   COMM
10856    2   2  ffffffc873428080  UN  [kworker/u16:15]
crash> bt 10856
PID: 10856  TASK: ffffffc873428080  CPU: 2   COMMAND: "kworker/u16:15"
 #0 [ffffff80270cb9a0] __switch_to at ffffff99bba8533c
 #1 [ffffff80270cba30] __schedule at ffffff99bcda18dc
 #2 [ffffff80270cba50] schedule at ffffff99bcda1cdc
 #3 [ffffff80270cbaf0] schedule_timeout at ffffff99bcda6674
 #4 [ffffff80270cbb70] wait_for_common at ffffff99bcda2c68
 #5 [ffffff80270cbb80] wait_for_completion at ffffff99bcda2b60
 #6 [ffffff80270cbc30] flush_workqueue at ffffff99bbad7a60
 #7 [ffffff80270cbc90] drain_workqueue at ffffff99bbad80fc
 #8 [ffffff80270cbcb0] destroy_workqueue at ffffff99bbad92f8
 #9 [ffffff80270cbda0] dfc_svc_init at ffffff99bbfbfb6c
 #10 [ffffff80270cbdf0] process_one_work at ffffff99bbadc478
 #11 [ffffff80270cbe50] worker_thread at ffffff99bbadc9dc
 #12 [ffffff80270cbeb0] kthread at ffffff99bbae1f84

Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Qianli Zhao <zhaoqianli@xiaomi.com>
---
Changes in V4:
- Fix an error that when different work item used same workqueue in flush_work, patch will trigger WARNING incorrectly, reported-by "kernel test robot <lkp@intel.com>"

Changes in V3:
- update changelog
- update comment

Changes in V2:
- update changelog
- update comment

 kernel/workqueue.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index c41c3c1..6a8a4e0 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -2585,6 +2585,7 @@ static int rescuer_thread(void *__rescuer)
  * @target_work: work item being flushed (NULL for workqueue flushes)
  *
  * %current is trying to flush the whole @target_wq or @target_work on it.
+ * If a work queue flushes itself, that will lead to a deadlock.
  * If @target_wq doesn't have %WQ_MEM_RECLAIM, verify that %current is not
  * reclaiming memory or running on a workqueue which doesn't have
  * %WQ_MEM_RECLAIM as that can break forward-progress guarantee leading to
@@ -2594,13 +2595,17 @@ static void check_flush_dependency(struct workqueue_struct *target_wq,
 				   struct work_struct *target_work)
 {
 	work_func_t target_func = target_work ? target_work->func : NULL;
-	struct worker *worker;
+	struct worker *worker = current_wq_worker();
+
+	WARN_ONCE(worker && worker->current_pwq->wq == target_wq &&
+		  worker->task == current &&
+		  (target_work ? worker->current_work == target_work : true),
+		  "workqueue: current work function:%ps is flushing own workqueue:%s",
+		  worker->current_func, target_wq->name);
 
 	if (target_wq->flags & WQ_MEM_RECLAIM)
 		return;
 
-	worker = current_wq_worker();
-
 	WARN_ONCE(current->flags & PF_MEMALLOC,
 		  "workqueue: PF_MEMALLOC task %d(%s) is flushing !WQ_MEM_RECLAIM %s:%ps",
 		  current->pid, current->comm, target_wq->name, target_func);
-- 
2.7.4

Re: [PATCH v4] workqueue: Warn when work flush own workqueue

[Tejun Heo]

On Thu, Aug 27, 2020 at 02:09:11PM +0800, Qianli Zhao wrote:
> @@ -2594,13 +2595,17 @@ static void check_flush_dependency(struct workqueue_struct *target_wq,
>  				   struct work_struct *target_work)
>  {
>  	work_func_t target_func = target_work ? target_work->func : NULL;
> -	struct worker *worker;
> +	struct worker *worker = current_wq_worker();
> +
> +	WARN_ONCE(worker && worker->current_pwq->wq == target_wq &&
> +		  worker->task == current &&
> +		  (target_work ? worker->current_work == target_work : true),
> +		  "workqueue: current work function:%ps is flushing own workqueue:%s",
> +		  worker->current_func, target_wq->name);

So, the idea is that we catch these with lockdeps. Doesn't lockdep trigger
for the same condition?

Thanks.

-- 
tejun