public inbox for [email protected]
 help / color / mirror / Atom feed
From: Pavel Begunkov <[email protected]>
To: [email protected]
Cc: Jens Axboe <[email protected]>, [email protected]
Subject: [PATCH v2 13/15] io_uring: separate task_work/waiting cache line
Date: Thu, 24 Aug 2023 23:53:35 +0100	[thread overview]
Message-ID: <b7f3fcb5b6b9cca0238778262c1fdb7ada6286b7.1692916914.git.asml.silence@gmail.com> (raw)
In-Reply-To: <[email protected]>

task_work's are typically queued up from IRQ/softirq potentially by a
random CPU like in case of networking. Batch ctx fields bouncing as this
into a separate cache line.

We also move ->cq_timeouts there because waiters have to read and check
it. We can also conditionally hide ->cq_timeouts in the future from the
CQ wait path as a not really useful rudiment.

Signed-off-by: Pavel Begunkov <[email protected]>
---
 include/linux/io_uring_types.h | 19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)

diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h
index 72e609752323..5de5dffe29df 100644
--- a/include/linux/io_uring_types.h
+++ b/include/linux/io_uring_types.h
@@ -270,15 +270,25 @@ struct io_ring_ctx {
 		unsigned		cached_cq_tail;
 		unsigned		cq_entries;
 		struct io_ev_fd	__rcu	*io_ev_fd;
-		struct wait_queue_head	cq_wait;
 		unsigned		cq_extra;
 	} ____cacheline_aligned_in_smp;
 
+	/*
+	 * task_work and async notification delivery cacheline. Expected to
+	 * regularly bounce b/w CPUs.
+	 */
+	struct {
+		struct llist_head	work_llist;
+		unsigned long		check_cq;
+		atomic_t		cq_wait_nr;
+		atomic_t		cq_timeouts;
+		struct wait_queue_head	cq_wait;
+	} ____cacheline_aligned_in_smp;
+
 	struct {
 		spinlock_t		completion_lock;
 
 		bool			poll_multi_queue;
-		atomic_t		cq_wait_nr;
 
 		/*
 		 * ->iopoll_list is protected by the ctx->uring_lock for
@@ -287,14 +297,11 @@ struct io_ring_ctx {
 		 * manipulate the list, hence no extra locking is needed there.
 		 */
 		struct io_wq_work_list	iopoll_list;
-
-		struct llist_head	work_llist;
 	} ____cacheline_aligned_in_smp;
 
 	/* timeouts */
 	struct {
 		spinlock_t		timeout_lock;
-		atomic_t		cq_timeouts;
 		struct list_head	timeout_list;
 		struct list_head	ltimeout_list;
 		unsigned		cq_last_tm_flush;
@@ -314,8 +321,6 @@ struct io_ring_ctx {
 	struct wait_queue_head	sqo_sq_wait;
 	struct list_head	sqd_list;
 
-	unsigned long		check_cq;
-
 	unsigned int		file_alloc_start;
 	unsigned int		file_alloc_end;
 
-- 
2.41.0


  parent reply	other threads:[~2023-08-24 22:56 UTC|newest]

Thread overview: 17+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-08-24 22:53 [PATCH v2 00/15] caching and SQ/CQ optimisations Pavel Begunkov
2023-08-24 22:53 ` [PATCH v2 01/15] io_uring: improve cqe !tracing hot path Pavel Begunkov
2023-08-24 22:53 ` [PATCH v2 02/15] io_uring: cqe init hardening Pavel Begunkov
2023-08-24 22:53 ` [PATCH v2 03/15] io_uring: simplify big_cqe handling Pavel Begunkov
2023-08-24 22:53 ` [PATCH v2 04/15] io_uring: refactor __io_get_cqe() Pavel Begunkov
2023-08-24 22:53 ` [PATCH v2 05/15] io_uring: optimise extra io_get_cqe null check Pavel Begunkov
2023-08-24 22:53 ` [PATCH v2 06/15] io_uring: reorder cqring_flush and wakeups Pavel Begunkov
2023-08-24 22:53 ` [PATCH v2 07/15] io_uring: merge iopoll and normal completion paths Pavel Begunkov
2023-08-24 22:53 ` [PATCH v2 08/15] io_uring: force inline io_fill_cqe_req Pavel Begunkov
2023-08-24 22:53 ` [PATCH v2 09/15] io_uring: compact SQ/CQ heads/tails Pavel Begunkov
2023-08-24 22:53 ` [PATCH v2 10/15] io_uring: add option to remove SQ indirection Pavel Begunkov
2023-08-24 22:53 ` [PATCH v2 11/15] io_uring: move non aligned field to the end Pavel Begunkov
2023-08-24 22:53 ` [PATCH v2 12/15] io_uring: banish non-hot data to end of io_ring_ctx Pavel Begunkov
2023-08-24 22:53 ` Pavel Begunkov [this message]
2023-08-24 22:53 ` [PATCH v2 14/15] io_uring: move multishot cqe cache in ctx Pavel Begunkov
2023-08-24 22:53 ` [PATCH v2 15/15] io_uring: move iopoll ctx fields around Pavel Begunkov
2023-08-24 23:16 ` [PATCH v2 00/15] caching and SQ/CQ optimisations 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=b7f3fcb5b6b9cca0238778262c1fdb7ada6286b7.1692916914.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