Re: [PATCH 3/5] io_uring: implement our own schedule timeout handling

[Pavel Begunkov <asml.silence@gmail.com>]

On 8/21/24 15:16, Jens Axboe wrote:
> In preparation for having two distinct timeouts and avoid waking the
> task if we don't need to.
> 
> Signed-off-by: Jens Axboe <axboe@kernel.dk>
> ---
>   io_uring/io_uring.c | 37 ++++++++++++++++++++++++++++++++-----
>   io_uring/io_uring.h |  2 ++
>   2 files changed, 34 insertions(+), 5 deletions(-)
> 
> diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
> index 9e2b8d4c05db..4ba5292137c3 100644
> --- a/io_uring/io_uring.c
> +++ b/io_uring/io_uring.c
> @@ -2322,7 +2322,7 @@ static int io_wake_function(struct wait_queue_entry *curr, unsigned int mode,
>   	 * Cannot safely flush overflowed CQEs from here, ensure we wake up
>   	 * the task, and the next invocation will do it.
>   	 */
> -	if (io_should_wake(iowq) || io_has_work(iowq->ctx))
> +	if (io_should_wake(iowq) || io_has_work(iowq->ctx) || iowq->hit_timeout)

Shouldn't be needed. If the timer fires, it should wake the task,
and the task will check ->hit_timeout there and later remove the
itself from the waitqueue.

>   		return autoremove_wake_function(curr, mode, wake_flags, key);
>   	return -1;
>   }
> @@ -2350,6 +2350,34 @@ static bool current_pending_io(void)
>   	return percpu_counter_read_positive(&tctx->inflight);
>   }
...

-- 
Pavel Begunkov