Skip to content

Commit 7b72d66

Browse files
committed
io_uring: gate iowait schedule on having pending requests
A previous commit made all cqring waits marked as iowait, as a way to improve performance for short schedules with pending IO. However, for use cases that have a special reaper thread that does nothing but wait on events on the ring, this causes a cosmetic issue where we know have one core marked as being "busy" with 100% iowait. While this isn't a grave issue, it is confusing to users. Rather than always mark us as being in iowait, gate setting of current->in_iowait to 1 by whether or not the waiting task has pending requests. Cc: stable@vger.kernel.org Link: https://lore.kernel.org/io-uring/CAMEGJJ2RxopfNQ7GNLhr7X9=bHXKo+G5OOe0LUq=+UgLXsv1Xg@mail.gmail.com/ Link: https://bugzilla.kernel.org/show_bug.cgi?id=217699 Link: https://bugzilla.kernel.org/show_bug.cgi?id=217700 Reported-by: Oleksandr Natalenko <oleksandr@natalenko.name> Reported-by: Phil Elwell <phil@raspberrypi.com> Tested-by: Andres Freund <andres@anarazel.de> Fixes: 8a79656 ("io_uring: Use io_schedule* in cqring wait") Signed-off-by: Jens Axboe <axboe@kernel.dk>
1 parent 07e9811 commit 7b72d66

File tree

1 file changed

+17
-6
lines changed

1 file changed

+17
-6
lines changed

io_uring/io_uring.c

Lines changed: 17 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -2493,11 +2493,20 @@ int io_run_task_work_sig(struct io_ring_ctx *ctx)
24932493
return 0;
24942494
}
24952495

2496+
static bool current_pending_io(void)
2497+
{
2498+
struct io_uring_task *tctx = current->io_uring;
2499+
2500+
if (!tctx)
2501+
return false;
2502+
return percpu_counter_read_positive(&tctx->inflight);
2503+
}
2504+
24962505
/* when returns >0, the caller should retry */
24972506
static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx,
24982507
struct io_wait_queue *iowq)
24992508
{
2500-
int token, ret;
2509+
int io_wait, ret;
25012510

25022511
if (unlikely(READ_ONCE(ctx->check_cq)))
25032512
return 1;
@@ -2511,17 +2520,19 @@ static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx,
25112520
return 0;
25122521

25132522
/*
2514-
* Use io_schedule_prepare/finish, so cpufreq can take into account
2515-
* that the task is waiting for IO - turns out to be important for low
2516-
* QD IO.
2523+
* Mark us as being in io_wait if we have pending requests, so cpufreq
2524+
* can take into account that the task is waiting for IO - turns out
2525+
* to be important for low QD IO.
25172526
*/
2518-
token = io_schedule_prepare();
2527+
io_wait = current->in_iowait;
2528+
if (current_pending_io())
2529+
current->in_iowait = 1;
25192530
ret = 0;
25202531
if (iowq->timeout == KTIME_MAX)
25212532
schedule();
25222533
else if (!schedule_hrtimeout(&iowq->timeout, HRTIMER_MODE_ABS))
25232534
ret = -ETIME;
2524-
io_schedule_finish(token);
2535+
current->in_iowait = io_wait;
25252536
return ret;
25262537
}
25272538

0 commit comments

Comments
 (0)