The patch below does not apply to the 5.10-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable@xxxxxxxxxxxxxxx>. thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >From 3b763ba1c77da5806e4fdc5684285814fe970c98 Mon Sep 17 00:00:00 2001 From: Pavel Begunkov <asml.silence@xxxxxxxxx> Date: Sun, 18 Apr 2021 14:52:08 +0100 Subject: [PATCH] io_uring: remove extra sqpoll submission halting SQPOLL task won't submit requests for a context that is currently dying, so no need to remove ctx from sqd_list prior the main loop of io_ring_exit_work(). Kill it, will be removed by io_sq_thread_finish() and only brings confusion and lockups. Cc: stable@xxxxxxxxxxxxxxx Signed-off-by: Pavel Begunkov <asml.silence@xxxxxxxxx> Link: https://lore.kernel.org/r/f220c2b786ba0f9499bebc9f3cd9714d29efb6a5.1618752958.git.asml.silence@xxxxxxxxx Signed-off-by: Jens Axboe <axboe@xxxxxxxxx> diff --git a/fs/io_uring.c b/fs/io_uring.c index e491b815df8c..4fc54cd40470 100644 --- a/fs/io_uring.c +++ b/fs/io_uring.c @@ -6750,6 +6750,10 @@ static int __io_sq_thread(struct io_ring_ctx *ctx, bool cap_entries) if (!list_empty(&ctx->iopoll_list)) io_do_iopoll(ctx, &nr_events, 0); + /* + * Don't submit if refs are dying, good for io_uring_register(), + * but also it is relied upon by io_ring_exit_work() + */ if (to_submit && likely(!percpu_ref_is_dying(&ctx->refs)) && !(ctx->flags & IORING_SETUP_R_DISABLED)) ret = io_submit_sqes(ctx, to_submit); @@ -8540,14 +8544,6 @@ static void io_ring_exit_work(struct work_struct *work) struct io_tctx_node *node; int ret; - /* prevent SQPOLL from submitting new requests */ - if (ctx->sq_data) { - io_sq_thread_park(ctx->sq_data); - list_del_init(&ctx->sqd_list); - io_sqd_update_thread_idle(ctx->sq_data); - io_sq_thread_unpark(ctx->sq_data); - } - /* * If we're doing polled IO and end up having requests being * submitted async (out-of-line), then completions can come in while
