public inbox for [email protected]
 help / color / mirror / Atom feed
From: Pavel Begunkov <[email protected]>
To: Jens Axboe <[email protected]>, [email protected]
Subject: [PATCH 16/16] io_uring: optimise io_req_task_work_add()
Date: Fri, 19 Mar 2021 17:22:44 +0000	[thread overview]
Message-ID: <23fda3980a56cec26385a81030a6cc9e01c04bf6.1616167719.git.asml.silence@gmail.com> (raw)
In-Reply-To: <[email protected]>

Inline io_task_work_add() into io_req_task_work_add(). They both work
with a request, so keeping them separate doesn't make things much more
clear, but merging allows optimise it. Apart from small wins like not
reading req->ctx or not calculating @notify in the hot path, i.e. with
tctx->task_state set, it avoids doing wake_up_process() for every single
add, but only after actually done task_work_add().

Signed-off-by: Pavel Begunkov <[email protected]>
---
 fs/io_uring.c | 50 ++++++++++++++++++--------------------------------
 1 file changed, 18 insertions(+), 32 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index d7b4cbe2ac3a..3548b2e60ba5 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -1931,13 +1931,17 @@ static void tctx_task_work(struct callback_head *cb)
 		cond_resched();
 }
 
-static int io_task_work_add(struct task_struct *tsk, struct io_kiocb *req,
-			    enum task_work_notify_mode notify)
+static int io_req_task_work_add(struct io_kiocb *req)
 {
+	struct task_struct *tsk = req->task;
 	struct io_uring_task *tctx = tsk->io_uring;
+	enum task_work_notify_mode notify;
 	struct io_wq_work_node *node, *prev;
 	unsigned long flags;
-	int ret;
+	int ret = 0;
+
+	if (unlikely(tsk->flags & PF_EXITING))
+		return -ESRCH;
 
 	WARN_ON_ONCE(!tctx);
 
@@ -1950,14 +1954,23 @@ static int io_task_work_add(struct task_struct *tsk, struct io_kiocb *req,
 	    test_and_set_bit(0, &tctx->task_state))
 		return 0;
 
-	if (!task_work_add(tsk, &tctx->task_work, notify))
+	/*
+	 * SQPOLL kernel thread doesn't need notification, just a wakeup. For
+	 * all other cases, use TWA_SIGNAL unconditionally to ensure we're
+	 * processing task_work. There's no reliable way to tell if TWA_RESUME
+	 * will do the job.
+	 */
+	notify = (req->ctx->flags & IORING_SETUP_SQPOLL) ? TWA_NONE : TWA_SIGNAL;
+
+	if (!task_work_add(tsk, &tctx->task_work, notify)) {
+		wake_up_process(tsk);
 		return 0;
+	}
 
 	/*
 	 * Slow path - we failed, find and delete work. if the work is not
 	 * in the list, it got run and we're fine.
 	 */
-	ret = 0;
 	spin_lock_irqsave(&tctx->task_lock, flags);
 	wq_list_for_each(node, prev, &tctx->task_list) {
 		if (&req->io_task_work.node == node) {
@@ -1971,33 +1984,6 @@ static int io_task_work_add(struct task_struct *tsk, struct io_kiocb *req,
 	return ret;
 }
 
-static int io_req_task_work_add(struct io_kiocb *req)
-{
-	struct task_struct *tsk = req->task;
-	struct io_ring_ctx *ctx = req->ctx;
-	enum task_work_notify_mode notify;
-	int ret;
-
-	if (tsk->flags & PF_EXITING)
-		return -ESRCH;
-
-	/*
-	 * SQPOLL kernel thread doesn't need notification, just a wakeup. For
-	 * all other cases, use TWA_SIGNAL unconditionally to ensure we're
-	 * processing task_work. There's no reliable way to tell if TWA_RESUME
-	 * will do the job.
-	 */
-	notify = TWA_NONE;
-	if (!(ctx->flags & IORING_SETUP_SQPOLL))
-		notify = TWA_SIGNAL;
-
-	ret = io_task_work_add(tsk, req, notify);
-	if (!ret)
-		wake_up_process(tsk);
-
-	return ret;
-}
-
 static bool io_run_task_work_head(struct callback_head **work_head)
 {
 	struct callback_head *work, *next;
-- 
2.24.0


  parent reply	other threads:[~2021-03-19 17:28 UTC|newest]

Thread overview: 18+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-03-19 17:22 [PATCH 00/16] random 5.13 bits Pavel Begunkov
2021-03-19 17:22 ` [PATCH 01/16] io_uring: don't take ctx refs in task_work handler Pavel Begunkov
2021-03-19 17:22 ` [PATCH 02/16] io_uring: optimise io_uring_enter() Pavel Begunkov
2021-03-19 17:22 ` [PATCH 03/16] io_uring: optimise tctx node checks/alloc Pavel Begunkov
2021-03-19 17:22 ` [PATCH 04/16] io_uring: keep io_req_free_batch() call locality Pavel Begunkov
2021-03-19 17:22 ` [PATCH 05/16] io_uring: inline __io_queue_linked_timeout() Pavel Begunkov
2021-03-19 17:22 ` [PATCH 06/16] io_uring: optimise success case of __io_queue_sqe Pavel Begunkov
2021-03-19 17:22 ` [PATCH 07/16] io_uring: refactor io_flush_cached_reqs() Pavel Begunkov
2021-03-19 17:22 ` [PATCH 08/16] io_uring: refactor rsrc refnode allocation Pavel Begunkov
2021-03-19 17:22 ` [PATCH 09/16] io_uring: inline io_put_req and friends Pavel Begunkov
2021-03-19 17:22 ` [PATCH 10/16] io_uring: refactor io_free_req_deferred() Pavel Begunkov
2021-03-19 17:22 ` [PATCH 11/16] io_uring: add helper flushing locked_free_list Pavel Begunkov
2021-03-19 17:22 ` [PATCH 12/16] io_uring: remove __io_req_task_cancel() Pavel Begunkov
2021-03-19 17:22 ` [PATCH 13/16] io_uring: inline io_clean_op()'s fast path Pavel Begunkov
2021-03-19 17:22 ` [PATCH 14/16] io_uring: optimise io_dismantle_req() " Pavel Begunkov
2021-03-19 17:22 ` [PATCH 15/16] io_uring: abolish old io_put_file() Pavel Begunkov
2021-03-19 17:22 ` Pavel Begunkov [this message]
2021-03-19 18:59 ` [PATCH 00/16] random 5.13 bits Jens Axboe

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=23fda3980a56cec26385a81030a6cc9e01c04bf6.1616167719.git.asml.silence@gmail.com \
    [email protected] \
    [email protected] \
    [email protected] \
    /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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox