On 18/02/2021 22:32, Pavel Begunkov wrote:
> In case of failure io_wq_submit_work() needs to post an CQE and so
> potentially take uring_lock. The safest way to deal with it is to do
> that from under task_work where we can safely take the lock.
>
> Also, as io_iopoll_check() holds the lock tight and releases it
> reluctantly, it will play nicer in the furuter with notifying an
> iopolling task about new such pending failed requests.
>
> Signed-off-by: Pavel Begunkov <asml.silence@xxxxxxxxx>
> ---
> @@ -2371,11 +2371,22 @@ static void io_req_task_queue(struct io_kiocb *req)
> req->task_work.func = io_req_task_submit;
> ret = io_req_task_work_add(req);
> if (unlikely(ret)) {
> + ret = -ECANCELED;
That's a stupid mistake. Jens, any chance you can fold in a diff below?
> percpu_ref_get(&req->ctx->refs);
> io_req_task_work_add_fallback(req, io_req_task_cancel);
> }
> }
diff --git a/fs/io_uring.c b/fs/io_uring.c
index 1cb5e40d9822..582306b1dfd1 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -2371,7 +2371,7 @@ static void io_req_task_queue(struct io_kiocb *req)
req->task_work.func = io_req_task_submit;
ret = io_req_task_work_add(req);
if (unlikely(ret)) {
- ret = -ECANCELED;
+ req->result = -ECANCELED;
percpu_ref_get(&req->ctx->refs);
io_req_task_work_add_fallback(req, io_req_task_cancel);
}
