Re: [ISSUE] The time cost of IOSQE_IO_LINK

[Jens Axboe <axboe@kernel.dk>]

On 2/18/20 6:13 AM, Peter Zijlstra wrote:
> Jens, I think you want something like this on top of what you have,
> mostly it is adding sched_work_run() to exit_task_work().

It also makes it a bit cleaner, I don't like the implied task == current
we have in a bunch of spots. Folded this in (thanks!) with minor edit:

> @@ -157,10 +157,10 @@ static void __task_work_run(struct task_struct *task,
>   */
>  void task_work_run(void)
>  {
> -	__task_work_run(current, &current->task_works);
> +	__task_work_run(&current->task_works);
>  }
>  
> -void sched_work_run(struct task_struct *task)
> +void sched_work_run()
>  {
> -	__task_work_run(task, &task->sched_work);
> +	__task_work_run(&task->sched_work);
>  }

s/task/current for this last one.

-- 
Jens Axboe