Re: [PATCH 0/4] sched: Various reweight_entity() fixes

[K Prateek Nayak <kprateek.nayak@amd.com>]

Hello Peter,

On 2/9/2026 9:17 PM, Peter Zijlstra wrote:
> I ended up with the below; and I've already pushed out a fresh
> queue/sched/core. Could you please test again?

I haven't seen any "sum_shifts" turning non-zero on the overnight runs,
even with the larger workloads, so feel free to include:

Tested-by: K Prateek Nayak <kprateek.nayak@amd.com>

The overall patch looks good. I only have couple of cosmetic nits
inlined below:

[..snip..]

>  u64 avg_vruntime(struct cfs_rq *cfs_rq)
>  {
>  	struct sched_entity *curr = cfs_rq->curr;
> -	s64 avg = cfs_rq->sum_w_vruntime;
> -	long load = cfs_rq->sum_weight;
> +	s64 delta = 0, runtime = cfs_rq->sum_w_vruntime;

nit.

"runtime" is only ever used in "weight" != 0 case so perhaps it can be
moved there to preserve the reverse x-mas (or this whole line can be
moved up top).

> +	long weight = cfs_rq->sum_weight;
>  
> -	if (curr && curr->on_rq) {
> -		unsigned long weight = scale_load_down(curr->load.weight);
> +	if (curr && !curr->on_rq)
> +		curr = NULL;
>  
> -		avg += entity_key(cfs_rq, curr) * weight;
> -		load += weight;
> -	}
> +	if (weight) {
> +		if (curr) {
> +			unsigned long w = scale_load_down(curr->load.weight);
> +
> +			runtime += entity_key(cfs_rq, curr) * w;
> +			weight += w;
> +		}
>  
> -	if (load) {
>  		/* sign flips effective floor / ceiling */
> -		if (avg < 0)
> -			avg -= (load - 1);
> -		avg = div_s64(avg, load);
> +		if (runtime < 0)
> +			runtime -= (weight - 1);
> +
> +		delta = div_s64(runtime, weight);
> +

nit. Is this empty line necessary?

> +	} else if (curr) {
> +		/*
> +		 * When there is but one element, it is the average.
> +		 */
> +		delta = curr->vruntime - cfs_rq->zero_vruntime;
>  	}
>  
> -	return cfs_rq->zero_vruntime + avg;
> +	update_zero_vruntime(cfs_rq, delta);
> +
> +	return cfs_rq->zero_vruntime;
>  }
>  
>  /*

-- 
Thanks and Regards,
Prateek