Re: [PATCH 5.11] io_uring: fix wqe->lock/completion_lock deadlock

[Joseph Qi <joseph.qi@linux.alibaba.com>]

On 1/27/21 7:35 AM, Pavel Begunkov wrote:
> Joseph reports following deadlock:
> 
> CPU0:
> ...
> io_kill_linked_timeout  // &ctx->completion_lock
> io_commit_cqring
> __io_queue_deferred
> __io_queue_async_work
> io_wq_enqueue
> io_wqe_enqueue  // &wqe->lock
> 
> CPU1:
> ...
> __io_uring_files_cancel
> io_wq_cancel_cb
> io_wqe_cancel_pending_work  // &wqe->lock
> io_cancel_task_cb  // &ctx->completion_lock
> 
> Only __io_queue_deferred() calls queue_async_work() while holding
> ctx->completion_lock, enqueue drained requests via io_req_task_queue()
> instead.
> 
We should follow &wqe->lock > &ctx->completion_lock from now on, right?
I was thinking getting completion_lock first before:(

Moreover, there are so many locks and no suggested locking order in
comments, so that it is hard for us to participate in the work.

> Cc: stable@vger.kernel.org # 5.9+
> Reported-by: Joseph Qi <joseph.qi@linux.alibaba.com>

Tested-by: Joseph Qi <joseph.qi@linux.alibaba.com>
> Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
> ---
>  fs/io_uring.c | 10 ++--------
>  1 file changed, 2 insertions(+), 8 deletions(-)
> 
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index bb0270eeb8cb..c218deaf73a9 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -1026,6 +1026,7 @@ static int io_setup_async_rw(struct io_kiocb *req, const struct iovec *iovec,
>  			     const struct iovec *fast_iov,
>  			     struct iov_iter *iter, bool force);
>  static void io_req_drop_files(struct io_kiocb *req);
> +static void io_req_task_queue(struct io_kiocb *req);
>  
>  static struct kmem_cache *req_cachep;
>  
> @@ -1634,18 +1635,11 @@ static void __io_queue_deferred(struct io_ring_ctx *ctx)
>  	do {
>  		struct io_defer_entry *de = list_first_entry(&ctx->defer_list,
>  						struct io_defer_entry, list);
> -		struct io_kiocb *link;
>  
>  		if (req_need_defer(de->req, de->seq))
>  			break;
>  		list_del_init(&de->list);
> -		/* punt-init is done before queueing for defer */
> -		link = __io_queue_async_work(de->req);
> -		if (link) {
> -			__io_queue_linked_timeout(link);
> -			/* drop submission reference */
> -			io_put_req_deferred(link, 1);
> -		}
> +		io_req_task_queue(de->req);
>  		kfree(de);
>  	} while (!list_empty(&ctx->defer_list));
>  }
>