Re: sched_ext: Fix SCX_KICK_WAIT to work reliably

[Peter Zijlstra <peterz@infradead.org>]

On Tue, Oct 21, 2025 at 11:03:54AM -1000, Tejun Heo wrote:

> @@ -5208,12 +5214,11 @@ static void kick_cpus_irq_workfn(struct
>  
>  		if (cpu != cpu_of(this_rq)) {
>  			/*
> -			 * Pairs with smp_store_release() issued by this CPU in
> -			 * switch_class() on the resched path.
> +			 * Pairs with store_release in put_prev_task_scx().
>  			 *
> -			 * We busy-wait here to guarantee that no other task can
> -			 * be scheduled on our core before the target CPU has
> -			 * entered the resched path.
> +			 * We busy-wait here to guarantee that the task running
> +			 * at the time of kicking is no longer running. This can
> +			 * be used to implement e.g. core scheduling.
>  			 */
>  			while (smp_load_acquire(wait_pnt_seq) == pseqs[cpu])
>  				cpu_relax();

You could consider using:

			smp_cond_load_acquire(wait_pnt_seq, VAL !+ pseqs[cpu]);

that's the fancy way of doing a spin wait and allows architectures to
optimize (mostly arm64 at this point).