From: Jens Axboe <axboe@kernel.dk>
To: io-uring@vger.kernel.org
Cc: Jens Axboe <axboe@kernel.dk>
Subject: [PATCH 17/18] io_uring: assign new io_identity for task if members have changed
Date: Fri, 16 Oct 2020 10:02:23 -0600 [thread overview]
Message-ID: <20201016160224.1575329-18-axboe@kernel.dk> (raw)
In-Reply-To: <20201016160224.1575329-1-axboe@kernel.dk>
This avoids doing a copy for each new async IO, if some parts of the
io_identity has changed. We avoid reference counting for the normal
fast path of nothing ever changing.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
fs/io_uring.c | 19 +++++++++++++++----
include/linux/io_uring.h | 3 ++-
2 files changed, 17 insertions(+), 5 deletions(-)
diff --git a/fs/io_uring.c b/fs/io_uring.c
index bc16327a6481..5af6dafcc669 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -1066,12 +1066,18 @@ static void io_init_identity(struct io_identity *id)
*/
static inline void io_req_init_async(struct io_kiocb *req)
{
+ struct io_uring_task *tctx = current->io_uring;
+
if (req->flags & REQ_F_WORK_INITIALIZED)
return;
memset(&req->work, 0, sizeof(req->work));
req->flags |= REQ_F_WORK_INITIALIZED;
- req->work.identity = ¤t->io_uring->identity;
+
+ /* Grab a ref if this isn't our static identity */
+ req->work.identity = tctx->identity;
+ if (tctx->identity != &tctx->__identity)
+ refcount_inc(&req->work.identity->count);
}
static inline bool io_async_submit(struct io_ring_ctx *ctx)
@@ -1179,7 +1185,7 @@ static void __io_commit_cqring(struct io_ring_ctx *ctx)
static void io_put_identity(struct io_uring_task *tctx, struct io_kiocb *req)
{
- if (req->work.identity == &tctx->identity)
+ if (req->work.identity == &tctx->__identity)
return;
if (refcount_dec_and_test(&req->work.identity->count))
kfree(req->work.identity);
@@ -1254,11 +1260,12 @@ static bool io_identity_cow(struct io_kiocb *req)
refcount_inc(&id->count);
/* drop old identity, assign new one. one ref for req, one for tctx */
- if (req->work.identity != &tctx->identity &&
+ if (req->work.identity != tctx->identity &&
refcount_sub_and_test(2, &req->work.identity->count))
kfree(req->work.identity);
req->work.identity = id;
+ tctx->identity = id;
return true;
}
@@ -7684,7 +7691,8 @@ static int io_uring_alloc_task_context(struct task_struct *task)
tctx->in_idle = 0;
atomic_long_set(&tctx->req_issue, 0);
atomic_long_set(&tctx->req_complete, 0);
- io_init_identity(&tctx->identity);
+ io_init_identity(&tctx->__identity);
+ tctx->identity = &tctx->__identity;
task->io_uring = tctx;
return 0;
}
@@ -7694,6 +7702,9 @@ void __io_uring_free(struct task_struct *tsk)
struct io_uring_task *tctx = tsk->io_uring;
WARN_ON_ONCE(!xa_empty(&tctx->xa));
+ WARN_ON_ONCE(refcount_read(&tctx->identity->count) != 1);
+ if (tctx->identity != &tctx->__identity)
+ kfree(tctx->identity);
kfree(tctx);
tsk->io_uring = NULL;
}
diff --git a/include/linux/io_uring.h b/include/linux/io_uring.h
index bd3346194bca..607d14f61132 100644
--- a/include/linux/io_uring.h
+++ b/include/linux/io_uring.h
@@ -24,7 +24,8 @@ struct io_uring_task {
struct wait_queue_head wait;
struct file *last;
atomic_long_t req_issue;
- struct io_identity identity;
+ struct io_identity __identity;
+ struct io_identity *identity;
/* completion side */
bool in_idle ____cacheline_aligned_in_smp;
--
2.28.0
next prev parent reply other threads:[~2020-10-16 16:02 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-10-16 16:02 [PATCHES 0/10] Fixes queued up for 5.10 Jens Axboe
2020-10-16 16:02 ` [PATCH 01/18] io_uring: Fix sizeof() mismatch Jens Axboe
2020-10-16 16:02 ` [PATCH 02/18] readahead: use limited read-ahead to satisfy read Jens Axboe
2020-10-16 16:02 ` [PATCH 03/18] io_uring: don't clear IOCB_NOWAIT for async buffered retry Jens Axboe
2020-10-16 16:02 ` [PATCH 04/18] io_uring: don't set COMP_LOCKED if won't put Jens Axboe
2020-10-16 16:02 ` [PATCH 05/18] io_uring: don't unnecessarily clear F_LINK_TIMEOUT Jens Axboe
2020-10-16 16:02 ` [PATCH 06/18] io_uring: don't put a poll req under spinlock Jens Axboe
2020-10-16 16:02 ` [PATCH 07/18] io_uring: dig out COMP_LOCK from deep call chain Jens Axboe
2020-10-16 16:02 ` [PATCH 08/18] io_uring: fix REQ_F_COMP_LOCKED by killing it Jens Axboe
2020-10-16 16:02 ` [PATCH 09/18] Revert "io_uring: mark io_uring_fops/io_op_defs as __read_mostly" Jens Axboe
2020-10-16 16:02 ` [PATCH 10/18] io_uring: fix error path cleanup in io_sqe_files_register() Jens Axboe
2020-10-16 16:02 ` [PATCH 11/18] io-wq: assign NUMA node locality if appropriate Jens Axboe
2020-10-16 16:02 ` [PATCH 12/18] io_uring: pass required context in as flags Jens Axboe
2020-10-16 16:02 ` [PATCH 13/18] io_uring: rely solely on work flags to determine personality Jens Axboe
2020-10-16 16:02 ` [PATCH 14/18] io_uring: move io identity items into separate struct Jens Axboe
2020-10-16 16:02 ` [PATCH 15/18] io_uring: COW io_identity on mismatch Jens Axboe
2020-10-16 16:02 ` [PATCH 16/18] io_uring: store io_identity in io_uring_task Jens Axboe
2020-10-16 16:02 ` Jens Axboe [this message]
2020-10-16 16:02 ` [PATCH 18/18] io_uring: use percpu counters to track inflight requests 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=20201016160224.1575329-18-axboe@kernel.dk \
--to=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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).