Re: [RFC PATCH] sched: Fix performance regression introduced by mm_cid

[Peter Zijlstra]

On Mon, Apr 03, 2023 at 02:13:42PM -0400, Mathieu Desnoyers wrote:
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index 0d18c3969f90..775c7da6ec7f 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -2326,16 +2326,20 @@ static inline bool is_cpu_allowed(struct task_struct *p, int cpu)
>  static struct rq *move_queued_task(struct rq *rq, struct rq_flags *rf,
>  				   struct task_struct *p, int new_cpu)
>  {
> +	int cid;
> +
>  	lockdep_assert_rq_held(rq);
>  
>  	deactivate_task(rq, p, DEQUEUE_NOCLOCK);
>  	set_task_cpu(p, new_cpu);
> +	cid = sched_mm_cid_migrate_from(rq, p);
>  	rq_unlock(rq, rf);
>  
>  	rq = cpu_rq(new_cpu);
>  
>  	rq_lock(rq, rf);
>  	WARN_ON_ONCE(task_cpu(p) != new_cpu);
> +	sched_mm_cid_migrate_to(rq, p, cid);
>  	activate_task(rq, p, 0);
>  	check_preempt_curr(rq, p, 0);
>  

I can't find more uses? Yet there are a ton more ways to actually
migrate tasks.