From: Pavel Begunkov <[email protected]>
To: Jens Axboe <[email protected]>,
[email protected], [email protected]
Subject: [PATCH 2/4] io-wq: optimise locking in io_worker_handle_work()
Date: Wed, 4 Mar 2020 16:14:10 +0300 [thread overview]
Message-ID: <ca39ae1c7b5cb8c86666e3aa9b80bf191949d4c6.1583314087.git.asml.silence@gmail.com> (raw)
In-Reply-To: <[email protected]>
There are 2 optimisations:
- Now, io_worker_handler_work() do io_assign_current_work() twice per
request, and each one adds lock/unlock(worker->lock) pair. The first is
to reset worker->cur_work to NULL, and the second to set a real work
shortly after. If there is a dependant work, set it immediately, that
effectively removes the extra NULL'ing.
- And there is no use in taking wqe->lock for linked works, as they are
not hashed now. Optimise it out.
Signed-off-by: Pavel Begunkov <[email protected]>
---
fs/io-wq.c | 15 +++++++--------
1 file changed, 7 insertions(+), 8 deletions(-)
diff --git a/fs/io-wq.c b/fs/io-wq.c
index e438dc4d7cb3..473af080470a 100644
--- a/fs/io-wq.c
+++ b/fs/io-wq.c
@@ -476,7 +476,7 @@ static void io_worker_handle_work(struct io_worker *worker)
struct io_wq *wq = wqe->wq;
do {
- struct io_wq_work *work, *old_work;
+ struct io_wq_work *work;
unsigned hash = -1U;
/*
@@ -495,12 +495,13 @@ static void io_worker_handle_work(struct io_worker *worker)
spin_unlock_irq(&wqe->lock);
if (!work)
break;
+ io_assign_current_work(worker, work);
/* handle a whole dependent link */
do {
- io_assign_current_work(worker, work);
- io_impersonate_work(worker, work);
+ struct io_wq_work *old_work;
+ io_impersonate_work(worker, work);
/*
* OK to set IO_WQ_WORK_CANCEL even for uncancellable
* work, the worker function will do the right thing.
@@ -513,10 +514,8 @@ static void io_worker_handle_work(struct io_worker *worker)
old_work = work;
work->func(&work);
-
- spin_lock_irq(&worker->lock);
- worker->cur_work = NULL;
- spin_unlock_irq(&worker->lock);
+ work = (old_work == work) ? NULL : work;
+ io_assign_current_work(worker, work);
if (wq->put_work)
wq->put_work(old_work);
@@ -529,7 +528,7 @@ static void io_worker_handle_work(struct io_worker *worker)
/* dependent work is not hashed */
hash = -1U;
}
- } while (work && work != old_work);
+ } while (work);
spin_lock_irq(&wqe->lock);
} while (1);
--
2.24.0
next prev parent reply other threads:[~2020-03-04 13:15 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-03-04 13:14 [PATCH 0/4] io-wq/io_uring locking optimisation Pavel Begunkov
2020-03-04 13:14 ` [PATCH 1/4] io-wq: shuffle io_worker_handle_work() code Pavel Begunkov
2020-03-04 13:14 ` Pavel Begunkov [this message]
2020-03-04 13:14 ` [PATCH 3/4] io-wq: optimise out *next_work() double lock Pavel Begunkov
2020-03-04 13:14 ` [PATCH 4/4] io_uring/io-wq: forward submission ref to async Pavel Begunkov
2020-03-04 18:48 ` [PATCH 0/4] io-wq/io_uring locking optimisation 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=ca39ae1c7b5cb8c86666e3aa9b80bf191949d4c6.1583314087.git.asml.silence@gmail.com \
[email protected] \
[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