Re: [RFC PATCH] sched, wait: wake_up_preempt() for first-order low-latency wakeups (to use in input subsystem)

Re: [RFC PATCH] sched, wait: wake_up_preempt() for first-order low-latency wakeups (to use in input subsystem)

[Peter Zijlstra]

On Fri, Jul 05, 2013 at 02:06:32PM +0400, Kirill Tkhai wrote:
> The patch introduces new primitives for waking up tasks, latencies of which
> are appreciable by user directly, that need the first-order interactivity.
> 
> The appliable area is not wide. It's generally input subsystem and wake_ups
> of tasks waiting for user actions: keyboard press, mouse movement etc. 

Ha! before you know it tons of drivers are using this because they all
think their interrupts it the most important thing on earth.. seriously!

> diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
> index 01970c8..8b8b328 100644
> --- a/kernel/sched/rt.c
> +++ b/kernel/sched/rt.c
> @@ -1226,8 +1226,21 @@ out:
>  	return cpu;
>  }
>  
> -static void check_preempt_equal_prio(struct rq *rq, struct task_struct *p)
> +static void check_preempt_equal_prio(struct rq *rq, struct task_struct *p, int flags)
>  {
> +	/*
> +	 * If:
> +	 *
> +	 * - the newly woken task is of equal priority to the current task
> +	 * - the newly woken task is non-migratable while current is migratable
> +	 * - current will be preempted on the next reschedule
> +	 *
> +	 * we should check to see if current can readily move to a different
> +	 * cpu.  If so, we will reschedule to allow the push logic to try
> +	 * to move current somewhere else, making room for our non-migratable
> +	 * task.
> +	 */
> +
>  	if (rq->curr->nr_cpus_allowed == 1)
>  		return;
>  
> @@ -1235,7 +1248,8 @@ static void check_preempt_equal_prio(struct rq *rq, struct task_struct *p)
>  	    && cpupri_find(&rq->rd->cpupri, p, NULL))
>  		return;
>  
> -	if (!cpupri_find(&rq->rd->cpupri, rq->curr, NULL))
> +	if (!cpupri_find(&rq->rd->cpupri, rq->curr, NULL) &&
> +	    !(flags & WF_PREEMPT))
>  		return;
>  
>  	/*
> @@ -1246,7 +1260,15 @@ static void check_preempt_equal_prio(struct rq *rq, struct task_struct *p)
>  	requeue_task_rt(rq, p, 1);
>  	resched_task(rq->curr);
>  }
> +#else /* !CONFIG_SMP */
> +static void check_preempt_equal_prio(struct rq *rq, struct task_struct *p, int flags)
> +{
> +	if (!(flags & WF_PREEMPT))
> +		return;
>  
> +	requeue_task_rt(rq, p, 1);
> +	resched_task(rq->curr);
> +}

You can't do this:

  http://pubs.opengroup.org/onlinepubs/009696899/functions/xsh_chap02_08.html

SCHED_FIFO;

2) When a blocked thread becomes a runnable thread, it becomes the tail
   of the thread list for its priority.