Re: [PATCH] sched/fair: fix task_numa_migrate to consider both task and group benefits

[Ethan Zhao]

On 8/29/2025 4:55 PM, Jianyong Wu wrote:
> The comment indicates that when searching for a suitable NUMA node, we
> should ensure that the selected node benefits both the task and its NUMA
> group. However, the current implementation can only guarantee that either
> the task or the group benefits, but not necessarily both.
>
> Signed-off-by: Jianyong Wu <wujianyong@xxxxxxxx>
> ---
>   kernel/sched/fair.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index b173a059315c..58c899738399 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -2568,7 +2568,7 @@ static int task_numa_migrate(struct task_struct *p)
>   			/* Only consider nodes where both task and groups benefit */
>   			taskimp = task_weight(p, nid, dist) - taskweight;
>   			groupimp = group_weight(p, nid, dist) - groupweight;
> -			if (taskimp < 0 && groupimp < 0)
> +			if (taskimp < 0 || groupimp < 0) 
Perhaps you misunderstand the comment, && means either the task or the
group has NO benefit from this migration, it wouldn't be done.
But if you replace it with ||, you will ignore the target node that
could benefit either the task or the group.

There is more logic to consider the benefit for both task & group
in the later function part.

One question, why not
if (taskimp <= 0 && groupimp <= 0) ?

Thanks,
Ethan
>   				continue;
>   
>   			env.dist = dist;