Re: [RFC PATCH v15 7/7] sched: Start blocked_on chain processing in find_proxy_task()

[Peter Zijlstra]

On Wed, Mar 12, 2025 at 03:11:37PM -0700, John Stultz wrote:
> @@ -6791,6 +6890,7 @@ static void __sched notrace __schedule(int sched_mode)
>  	struct rq_flags rf;
>  	struct rq *rq;
>  	int cpu;
> +	bool preserve_need_resched = false;
>  
>  	cpu = smp_processor_id();
>  	rq = cpu_rq(cpu);
> @@ -6858,9 +6958,12 @@ static void __sched notrace __schedule(int sched_mode)
>  		next = find_proxy_task(rq, next, &rf);
>  		if (!next)
>  			goto pick_again;
> +		if (next == rq->idle)
> +			preserve_need_resched = true;
>  	}
>  picked:
> -	clear_tsk_need_resched(prev);
> +	if (!preserve_need_resched)
> +		clear_tsk_need_resched(prev);
>  	clear_preempt_need_resched();
>  #ifdef CONFIG_SCHED_DEBUG
>  	rq->last_seen_need_resched_ns = 0;

I would really rather this was done like so:


--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -6957,12 +6944,12 @@ static void __sched notrace __schedule(i
 		if (!next)
 			goto pick_again;
 		if (next == rq->idle)
-			preserve_need_resched = true;
+			goto keep_resched;
 	}
 picked:
-	if (!preserve_need_resched)
-		clear_tsk_need_resched(prev);
+	clear_tsk_need_resched(prev);
 	clear_preempt_need_resched();
+keep_resched:
 #ifdef CONFIG_SCHED_DEBUG
 	rq->last_seen_need_resched_ns = 0;
 #endif