Commit 906a3c6f authored by Xiaoguang Wang's avatar Xiaoguang Wang Committed by Jens Axboe
Browse files

io_uring: don't acquire uring_lock twice 



Both IOPOLL and sqes handling need to acquire uring_lock, combine
them together, then we just need to acquire uring_lock once.

Signed-off-by: default avatarXiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
Signed-off-by: default avatarJens Axboe <axboe@kernel.dk>
parent a0d9205f

fs/io_uring.c

+7 −11
Original line number Diff line number Diff line
@@ -6847,23 +6847,19 @@ static int __io_sq_thread(struct io_ring_ctx *ctx, bool cap_entries) 
	unsigned int to_submit;

	int ret = 0;



	if (!list_empty(&ctx->iopoll_list)) {

		unsigned nr_events = 0;



		mutex_lock(&ctx->uring_lock);

		if (!list_empty(&ctx->iopoll_list))

			io_do_iopoll(ctx, &nr_events, 0);

		mutex_unlock(&ctx->uring_lock);

	}



	to_submit = io_sqring_entries(ctx);

	/* if we're handling multiple rings, cap submit size for fairness */

	if (cap_entries && to_submit > 8)

		to_submit = 8;



	if (to_submit) {

	if (!list_empty(&ctx->iopoll_list) || to_submit) {

		unsigned nr_events = 0;



		mutex_lock(&ctx->uring_lock);

		if (likely(!percpu_ref_is_dying(&ctx->refs)))

		if (!list_empty(&ctx->iopoll_list))

			io_do_iopoll(ctx, &nr_events, 0);



		if (to_submit && likely(!percpu_ref_is_dying(&ctx->refs)))

			ret = io_submit_sqes(ctx, to_submit);

		mutex_unlock(&ctx->uring_lock);

	}

CRA Git | Maintained and supported by SUSTech CRA and CCSE