All of lore.kernel.org
 help / color / mirror / Atom feed
From: Pavel Begunkov <asml.silence@gmail.com>
To: Jens Axboe <axboe@kernel.dk>, io-uring@vger.kernel.org
Subject: [PATCH v4 11/26] io_uring: combine lock/unlock sections on exit
Date: Thu,  1 Apr 2021 15:43:50 +0100	[thread overview]
Message-ID: <a8ae0589b0ea64ad4791e2c282e4e9b713dd7024.1617287883.git.asml.silence@gmail.com> (raw)
In-Reply-To: <cover.1617287883.git.asml.silence@gmail.com>

io_ring_exit_work() already does uring_lock lock/unlock, no need to
repeat it for lock waiting trick in io_ring_ctx_free(). Move the waiting
with comments and spinlocking into io_ring_exit_work.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 fs/io_uring.c | 18 ++++++++----------
 1 file changed, 8 insertions(+), 10 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index e9bfe137270c..9ebdd288653f 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -8455,16 +8455,6 @@ static void io_req_caches_free(struct io_ring_ctx *ctx)
 
 static void io_ring_ctx_free(struct io_ring_ctx *ctx)
 {
-	/*
-	 * Some may use context even when all refs and requests have been put,
-	 * and they are free to do so while still holding uring_lock or
-	 * completion_lock, see __io_req_task_submit(). Wait for them to finish.
-	 */
-	mutex_lock(&ctx->uring_lock);
-	mutex_unlock(&ctx->uring_lock);
-	spin_lock_irq(&ctx->completion_lock);
-	spin_unlock_irq(&ctx->completion_lock);
-
 	io_sq_thread_finish(ctx);
 	io_sqe_buffers_unregister(ctx);
 
@@ -8615,6 +8605,12 @@ static void io_ring_exit_work(struct work_struct *work)
 		WARN_ON_ONCE(time_after(jiffies, timeout));
 	} while (!wait_for_completion_timeout(&ctx->ref_comp, HZ/20));
 
+	/*
+	 * Some may use context even when all refs and requests have been put,
+	 * and they are free to do so while still holding uring_lock or
+	 * completion_lock, see __io_req_task_submit(). Apart from other work,
+	 * this lock/unlock section also waits them to finish.
+	 */
 	mutex_lock(&ctx->uring_lock);
 	while (!list_empty(&ctx->tctx_list)) {
 		WARN_ON_ONCE(time_after(jiffies, timeout));
@@ -8635,6 +8631,8 @@ static void io_ring_exit_work(struct work_struct *work)
 		mutex_lock(&ctx->uring_lock);
 	}
 	mutex_unlock(&ctx->uring_lock);
+	spin_lock_irq(&ctx->completion_lock);
+	spin_unlock_irq(&ctx->completion_lock);
 
 	io_ring_ctx_free(ctx);
 }
-- 
2.24.0


  parent reply	other threads:[~2021-04-01 18:11 UTC|newest]

Thread overview: 29+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-04-01 14:43 [PATCH v4 00/26] ctx wide rsrc nodes + Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 01/26] io_uring: name rsrc bits consistently Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 02/26] io_uring: simplify io_rsrc_node_ref_zero Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 03/26] io_uring: use rsrc prealloc infra for files reg Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 04/26] io_uring: encapsulate rsrc node manipulations Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 05/26] io_uring: move rsrc_put callback into io_rsrc_data Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 06/26] io_uring: refactor io_queue_rsrc_removal() Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 07/26] io_uring: ctx-wide rsrc nodes Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 08/26] io_uring: reuse io_rsrc_node_destroy() Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 09/26] io_uring: remove useless is_dying check on quiesce Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 10/26] io_uring: refactor rw reissue Pavel Begunkov
2021-04-01 14:43 ` Pavel Begunkov [this message]
2021-04-01 14:43 ` [PATCH v4 12/26] io_uring: better ref handling in poll_remove_one Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 13/26] io_uring: remove unused hash_wait Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 14/26] io_uring: refactor io_async_cancel() Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 15/26] io_uring: improve import_fixed overflow checks Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 16/26] io_uring: store reg buffer end instead of length Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 17/26] io_uring: kill unused forward decls Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 18/26] io_uring: lock annotate timeouts and poll Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 19/26] io_uring: simplify overflow handling Pavel Begunkov
2021-04-01 14:43 ` [PATCH v4 20/26] io_uring: put link timeout req consistently Pavel Begunkov
2021-04-01 14:44 ` [PATCH v4 21/26] io_uring: deduplicate NOSIGNAL setting Pavel Begunkov
2021-04-01 14:44 ` [PATCH v4 22/26] io_uring: set proper FFS* flags on reg file update Pavel Begunkov
2021-04-01 14:44 ` [PATCH v4 23/26] io_uring: don't quiesce intial files register Pavel Begunkov
2021-04-01 14:44 ` [PATCH v4 24/26] io_uring: refactor file tables alloc/free Pavel Begunkov
2021-04-01 14:44 ` [PATCH v4 25/26] io_uring: encapsulate fixed files into struct Pavel Begunkov
2021-04-01 14:44 ` [PATCH v4 26/26] io_uring: kill outdated comment about splice punt Pavel Begunkov
2021-04-04 19:16 ` [PATCH v4 00/26] ctx wide rsrc nodes + Jens Axboe
2021-04-04 19:22   ` Pavel Begunkov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=a8ae0589b0ea64ad4791e2c282e4e9b713dd7024.1617287883.git.asml.silence@gmail.com \
    --to=asml.silence@gmail.com \
    --cc=axboe@kernel.dk \
    --cc=io-uring@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.