Re: [PATCH 1/2] sched/fair: Fix panic if NEXT_BUDDY enabled

[Adam Li <adamli@os.amperecomputing.com>]

On 11/25/2024 6:16 PM, Madadi Vineeth Reddy wrote:
> On 25/11/24 07:42, Adam Li wrote:
>> Enabling NEXT_BUDDY may trigger warning and kernel panic:
> 
> IIUC, this happens due to a race condition between choosing the next_buddy and
> dequeuing the task from the runqueue as soon as its lag becomes zero, thereby causing
> task_of(se) to result in a null dereference in pick_task_fair?
> 
> Correct me if I am wrong.
> 
pick_eevdf() may return NULL, which triggers this NULL pointer dereference
at pick_next_entity():
	struct sched_entity *se = pick_eevdf(cfs_rq);
	if (se->sched_delayed)

Enabling NEXT_BUDDY is more likely to trigger this panic.

I added another patch to check return value of pick_eevdf() in v2:
https://lore.kernel.org/all/20241127055610.7076-1-adamli@os.amperecomputing.com/

Thanks for review.
-adam