Thread (3 messages) 3 messages, 3 authors, 2021-09-15

Re: [PATCH -next] io-wq: Remove duplicate code in io_workqueue_create()

From: Jens Axboe <axboe@kernel.dk>
Date: 2021-09-15 18:12:21
Also in: lkml

On 9/12/21 9:36 AM, Hao Xu wrote:
在 2021/9/11 下午4:58, Bixuan Cui 写道:
quoted
While task_work_add() in io_workqueue_create() is true,
then duplicate code is executed:

   -> clear_bit_unlock(0, &worker->create_state);
   -> io_worker_release(worker);
   -> atomic_dec(&acct->nr_running);
   -> io_worker_ref_put(wq);
   -> return false;

   -> clear_bit_unlock(0, &worker->create_state); // back to io_workqueue_create()
   -> io_worker_release(worker);
   -> kfree(worker);

The io_worker_release() and clear_bit_unlock() are executed twice.

Fixes: 3146cba99aa2 ("io-wq: make worker creation resilient against signals")
Signed-off-by: Bixuan Cui <redacted>
---
  fs/io-wq.c | 9 ++++-----
  1 file changed, 4 insertions(+), 5 deletions(-)
diff --git a/fs/io-wq.c b/fs/io-wq.c
index 6c55362c1f99..95d0eaed7c00 100644
--- a/fs/io-wq.c
+++ b/fs/io-wq.c
@@ -329,8 +329,10 @@ static bool io_queue_worker_create(struct io_worker *worker,
  
  	init_task_work(&worker->create_work, func);
  	worker->create_index = acct->index;
-	if (!task_work_add(wq->task, &worker->create_work, TWA_SIGNAL))
+	if (!task_work_add(wq->task, &worker->create_work, TWA_SIGNAL)) {
+		clear_bit_unlock(0, &worker->create_state);
  		return true;
+	}
  	clear_bit_unlock(0, &worker->create_state);
  fail_release:
  	io_worker_release(worker);
@@ -723,11 +725,8 @@ static void io_workqueue_create(struct work_struct *work)
  	struct io_worker *worker = container_of(work, struct io_worker, work);
  	struct io_wqe_acct *acct = io_wqe_get_acct(worker);
  
-	if (!io_queue_worker_create(worker, acct, create_worker_cont)) {
-		clear_bit_unlock(0, &worker->create_state);
-		io_worker_release(worker);
+	if (!io_queue_worker_create(worker, acct, create_worker_cont))
  		kfree(worker);
-	}
  }
  
  static bool create_io_worker(struct io_wq *wq, struct io_wqe *wqe, int index)
AFAIK, this looks reasonable for me.
I took that as a reviewed-by, let me know if that isn't correct.

-- 
Jens Axboe
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help