From: pgray <hi@pgray.dev>
To: Jens Axboe <axboe@kernel.dk>
Cc: Pavel Begunkov <asml.silence@gmail.com>,
David Wei <dw@davidwei.uk>,
io-uring@vger.kernel.org, linux-kernel@vger.kernel.org,
pgray <hi@pgray.dev>
Subject: [PATCH] io_uring/tw: avoid local task work starvation
Date: Wed, 30 Sep 2026 13:15:35 +0000 [thread overview]
Message-ID: <20260930131446.90741-1-hi@pgray.dev> (raw)
IO_LOCAL_TW_DEFAULT_MAX (20) is too small for massive backlogs
this patch fixes by:
* not counting passive completions against budget
* add a ceiling of 512: IO_LOCAL_TW_COMPL_MAX
* clearing IORING_SQ_TASKRUN when local queue empty
Link: https://github.com/axboe/liburing/issues/1645
Fixes: f46b9cdb22f7 ("io_uring: limit local tw done")
Signed-off-by: pgray <hi@pgray.dev>
---
io_uring/tw.c | 21 ++++++++++++++++-----
1 file changed, 16 insertions(+), 5 deletions(-)
diff --git a/io_uring/tw.c b/io_uring/tw.c
index f573bcc3af6a..236ed01fcfaa 100644
--- a/io_uring/tw.c
+++ b/io_uring/tw.c
@@ -16,6 +16,8 @@
#include "wait.h"
#include "mpscq.h"
+#define IO_LOCAL_TW_COMPL_MAX 512
+
static void ctx_flush_and_put(struct io_ring_ctx *ctx, io_tw_token_t tw)
{
if (!ctx)
@@ -287,17 +289,23 @@ static int __io_run_local_work_loop(struct io_ring_ctx *ctx,
int events)
{
int ret = 0;
+ int nr_events = 0;
- while (ret < events) {
+ while (nr_events < events && ret < IO_LOCAL_TW_COMPL_MAX) {
struct llist_node *node = mpscq_pop(&ctx->work_list, &ctx->work_head);
struct io_kiocb *req;
if (!node)
break;
req = container_of(node, struct io_kiocb, io_task_work.node);
- INDIRECT_CALL_2(req->io_task_work.func,
- io_poll_task_func, io_req_rw_complete,
- (struct io_tw_req){req}, tw);
+ if (req->io_task_work.func == io_req_task_complete) {
+ io_req_task_complete((struct io_tw_req){req}, tw);
+ } else {
+ INDIRECT_CALL_2(req->io_task_work.func,
+ io_poll_task_func, io_req_rw_complete,
+ (struct io_tw_req){req}, tw);
+ nr_events++;
+ }
ret++;
}
@@ -342,8 +350,11 @@ int io_run_local_work_locked(struct io_ring_ctx *ctx, int min_events)
{
struct io_tw_state ts = {};
- if (!io_local_work_pending(ctx))
+ if (!io_local_work_pending(ctx)) {
+ if (ctx->flags & IORING_SETUP_TASKRUN_FLAG)
+ atomic_andnot(IORING_SQ_TASKRUN, &ctx->rings->sq_flags);
return 0;
+ }
return __io_run_local_work(ctx, ts, min_events,
max(IO_LOCAL_TW_DEFAULT_MAX, min_events));
}
--
2.55.0
reply other threads:[~2026-09-30 13:15 UTC|newest]
Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
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=20260930131446.90741-1-hi@pgray.dev \
--to=hi@pgray.dev \
--cc=asml.silence@gmail.com \
--cc=axboe@kernel.dk \
--cc=dw@davidwei.uk \
--cc=io-uring@vger.kernel.org \
--cc=linux-kernel@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