Re: [PATCH] sched/fair: Add null pointer check to pick_next_entity()

[Steven Rostedt]

On Thu, 20 Mar 2025 10:34:39 -0700
Pat Cody <patcody@xxxxxxxx> wrote:

> pick_eevdf() can return null, resulting in a null pointer dereference
> crash in pick_next_entity()
> 
> The other call site of pick_eevdf() can already handle a null pointer,
> and pick_next_entity() can already return null as well. Add an extra
> check to handle the null return here.

Have you actually triggered this?

> 
> Cc: stable@xxxxxxxxxx
> Fixes: f12e148892ed ("sched/fair: Prepare pick_next_task() for delayed dequeue")


No signed-off-by.

> ---
>  kernel/sched/fair.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index a553181dc764..f2157298cbce 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -5560,6 +5560,8 @@ pick_next_entity(struct rq *rq, struct cfs_rq *cfs_rq)
>  	}
>  
>  	struct sched_entity *se = pick_eevdf(cfs_rq);
> +	if (!se)
> +		return NULL;

So the place that this is called is in pick_task_fair() which has:

	cfs_rq = &rq->cfs;
	if (!cfs_rq->nr_queued)
		return NULL;

Where I believe it's a bug if cfs_rq->nr_queued > 0 and pick_eevdf() returns NULL.

Which means in should never happen. Now, to be more robust, we could add a
check for NULL, but this is not a fix, nor should it go to stable.

-- Steve



>  	if (se->sched_delayed) {
>  		dequeue_entities(rq, se, DEQUEUE_SLEEP | DEQUEUE_DELAYED);
>  		/*