[PATCH v2] sched/fair: Sync se's load_avg with cfs_rq in reweight_task

[PATCH v2] sched/fair: Sync se's load_avg with cfs_rq in reweight_task

[Chuyi Zhou]

In reweight_task(), there are two situations:

1. The task was on_rq, then the task's load_avg is accurate because we
synchronized it with cfs_rq through update_load_avg() in dequeue_task().

2. The task is sleeping, its load_avg might not have been updated for some
time, which can result in inaccurate dequeue_load_avg() in
reweight_entity().

This patch solves this by using update_load_avg() to synchronize the
load_avg of se with cfs_rq. For tasks were on_rq, since we already update
load_avg to accurate values in dequeue_task(), this change will not have
other effects due to the short time interval between the two updates.

Signed-off-by: Chuyi Zhou <zhouchuyi@bytedance.com>
---
Changes in v2:
- change the description in commit log.
- use update_load_avg() in reweight_task() rather than in reweight_entity
suggested by chengming.
- Link to v1: https://lore.kernel.org/lkml/20240716150840.23061-1-zhouchuyi@bytedance.com/
---
 kernel/sched/fair.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 9057584ec06d..b1e07ce90284 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -3835,12 +3835,15 @@ static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
 	}
 }
 
+static inline void update_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags);
+
 void reweight_task(struct task_struct *p, const struct load_weight *lw)
 {
 	struct sched_entity *se = &p->se;
 	struct cfs_rq *cfs_rq = cfs_rq_of(se);
 	struct load_weight *load = &se->load;
 
+	update_load_avg(cfs_rq, se, 0);
 	reweight_entity(cfs_rq, se, lw->weight);
 	load->inv_weight = lw->inv_weight;
 }
-- 
2.20.1

Re: [PATCH v2] sched/fair: Sync se's load_avg with cfs_rq in reweight_task

[K Prateek Nayak]

(+ Qais)

Hello Chuyi,

On 7/20/2024 10:42 AM, Chuyi Zhou wrote:
> In reweight_task(), there are two situations:
> 
> 1. The task was on_rq, then the task's load_avg is accurate because we
> synchronized it with cfs_rq through update_load_avg() in dequeue_task().
> 
> 2. The task is sleeping, its load_avg might not have been updated for some
> time, which can result in inaccurate dequeue_load_avg() in
> reweight_entity().
> 
> This patch solves this by using update_load_avg() to synchronize the
> load_avg of se with cfs_rq. For tasks were on_rq, since we already update
> load_avg to accurate values in dequeue_task(), this change will not have
> other effects due to the short time interval between the two updates.
> 
> Signed-off-by: Chuyi Zhou <zhouchuyi@bytedance.com>
> ---
> Changes in v2:
> - change the description in commit log.
> - use update_load_avg() in reweight_task() rather than in reweight_entity
> suggested by chengming.
> - Link to v1: https://lore.kernel.org/lkml/20240716150840.23061-1-zhouchuyi@bytedance.com/
> ---
>   kernel/sched/fair.c | 3 +++
>   1 file changed, 3 insertions(+)
> 
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 9057584ec06d..b1e07ce90284 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -3835,12 +3835,15 @@ static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
>   	}
>   }
>   
> +static inline void update_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags);
> +
>   void reweight_task(struct task_struct *p, const struct load_weight *lw)
>   {
>   	struct sched_entity *se = &p->se;
>   	struct cfs_rq *cfs_rq = cfs_rq_of(se);
>   	struct load_weight *load = &se->load;
>   
> +	update_load_avg(cfs_rq, se, 0);

Seems to be necessary when we reach here from __setscheduler_params() or
set_user_nice() for a sleeping task. Please feel free to add:

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

But since we are on the subject of accurate PELT accounting, one question
I have here is whether a reweight_task() for a sleeping task race with
its wakeup? Something like the following scenario:

CPU0										CPU1
====										====
/* No rq locks held until ttwu_queue() */
try_to_wake_up(p) {
	...
	/* Migrating task */
	set_task_cpu(p, cpu) {
		/* p->sched_class->migrate_task_rq(p, new_cpu); */		/* Called with task_cpu(p)'s rq lock held */
		migrate_task_rq_fair() {					reweight_task(p) {
			/* p is still sleeping */					...
			if (!task_on_rq_migrating(p)) {					dequeue_load_avg(cfs_rq, se);
                 		remove_entity_load_avg(se);				update_load_set(&se->load, weight);
				...							enqueue_load_avg(cfs_rq, se);
			}								...
		}								}
		/* task_cpu() is updated here */
		__set_task_cpu(p, new_cpu);
	}
	ttwu_queue();
}

In theory, the remove_entity_load_avg() could record stale value of
"load_avg" that gets removed at the next dequeue if I'm not mistaken?
But I believe these small inaccuracies are tolerable since they'll decay
in a while anyways?

>   	reweight_entity(cfs_rq, se, lw->weight);
>   	load->inv_weight = lw->inv_weight;
>   }

-- 
Thanks and Regards,
Prateek

Re: [PATCH v2] sched/fair: Sync se's load_avg with cfs_rq in reweight_task

[Chengming Zhou]

On 2024/7/22 13:17, K Prateek Nayak wrote:
> (+ Qais)
> 
> Hello Chuyi,
> 
> On 7/20/2024 10:42 AM, Chuyi Zhou wrote:
>> In reweight_task(), there are two situations:
>>
>> 1. The task was on_rq, then the task's load_avg is accurate because we
>> synchronized it with cfs_rq through update_load_avg() in dequeue_task().
>>
>> 2. The task is sleeping, its load_avg might not have been updated for 
>> some
>> time, which can result in inaccurate dequeue_load_avg() in
>> reweight_entity().
>>
>> This patch solves this by using update_load_avg() to synchronize the
>> load_avg of se with cfs_rq. For tasks were on_rq, since we already update
>> load_avg to accurate values in dequeue_task(), this change will not have
>> other effects due to the short time interval between the two updates.
>>
>> Signed-off-by: Chuyi Zhou <zhouchuyi@bytedance.com>
>> ---
>> Changes in v2:
>> - change the description in commit log.
>> - use update_load_avg() in reweight_task() rather than in reweight_entity
>> suggested by chengming.
>> - Link to v1: 
>> https://lore.kernel.org/lkml/20240716150840.23061-1-zhouchuyi@bytedance.com/
>> ---
>>   kernel/sched/fair.c | 3 +++
>>   1 file changed, 3 insertions(+)
>>
>> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
>> index 9057584ec06d..b1e07ce90284 100644
>> --- a/kernel/sched/fair.c
>> +++ b/kernel/sched/fair.c
>> @@ -3835,12 +3835,15 @@ static void reweight_entity(struct cfs_rq 
>> *cfs_rq, struct sched_entity *se,
>>       }
>>   }
>> +static inline void update_load_avg(struct cfs_rq *cfs_rq, struct 
>> sched_entity *se, int flags);
>> +
>>   void reweight_task(struct task_struct *p, const struct load_weight *lw)
>>   {
>>       struct sched_entity *se = &p->se;
>>       struct cfs_rq *cfs_rq = cfs_rq_of(se);
>>       struct load_weight *load = &se->load;
>> +    update_load_avg(cfs_rq, se, 0);
> 
> Seems to be necessary when we reach here from __setscheduler_params() or
> set_user_nice() for a sleeping task. Please feel free to add:
> 
> Reviewed-by: K Prateek Nayak <kprateek.nayak@amd.com>
> 
> But since we are on the subject of accurate PELT accounting, one question
> I have here is whether a reweight_task() for a sleeping task race with
> its wakeup? Something like the following scenario:

These two paths are impossible to race, since we have the task->pi_lock.

Thanks.

> 
> CPU0                                        CPU1
> ====                                        ====
> /* No rq locks held until ttwu_queue() */
> try_to_wake_up(p) {
>      ...
>      /* Migrating task */
>      set_task_cpu(p, cpu) {
>          /* p->sched_class->migrate_task_rq(p, new_cpu); */        /* 
> Called with task_cpu(p)'s rq lock held */
>          migrate_task_rq_fair() {                    reweight_task(p) {
>              /* p is still sleeping */                    ...
>              if (!task_on_rq_migrating(p)) {                    
> dequeue_load_avg(cfs_rq, se);
>                          remove_entity_load_avg(se);                
> update_load_set(&se->load, weight);
>                  ...                            enqueue_load_avg(cfs_rq, 
> se);
>              }                                ...
>          }                                }
>          /* task_cpu() is updated here */
>          __set_task_cpu(p, new_cpu);
>      }
>      ttwu_queue();
> }
> 
> In theory, the remove_entity_load_avg() could record stale value of
> "load_avg" that gets removed at the next dequeue if I'm not mistaken?
> But I believe these small inaccuracies are tolerable since they'll decay
> in a while anyways?
> 
>>       reweight_entity(cfs_rq, se, lw->weight);
>>       load->inv_weight = lw->inv_weight;
>>   }
> 

Re: [PATCH v2] sched/fair: Sync se's load_avg with cfs_rq in reweight_task

[K Prateek Nayak]

Hello Chengming,

On 7/22/2024 12:10 PM, Chengming Zhou wrote:
> On 2024/7/22 13:17, K Prateek Nayak wrote:
>> (+ Qais)
>>
>> Hello Chuyi,
>>
>> On 7/20/2024 10:42 AM, Chuyi Zhou wrote:
>>> In reweight_task(), there are two situations:
>>>
>>> 1. The task was on_rq, then the task's load_avg is accurate because we
>>> synchronized it with cfs_rq through update_load_avg() in dequeue_task().
>>>
>>> 2. The task is sleeping, its load_avg might not have been updated for some
>>> time, which can result in inaccurate dequeue_load_avg() in
>>> reweight_entity().
>>>
>>> This patch solves this by using update_load_avg() to synchronize the
>>> load_avg of se with cfs_rq. For tasks were on_rq, since we already update
>>> load_avg to accurate values in dequeue_task(), this change will not have
>>> other effects due to the short time interval between the two updates.
>>>
>>> Signed-off-by: Chuyi Zhou <zhouchuyi@bytedance.com>
>>> ---
>>> Changes in v2:
>>> - change the description in commit log.
>>> - use update_load_avg() in reweight_task() rather than in reweight_entity
>>> suggested by chengming.
>>> - Link to v1: https://lore.kernel.org/lkml/20240716150840.23061-1-zhouchuyi@bytedance.com/
>>> ---
>>>   kernel/sched/fair.c | 3 +++
>>>   1 file changed, 3 insertions(+)
>>>
>>> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
>>> index 9057584ec06d..b1e07ce90284 100644
>>> --- a/kernel/sched/fair.c
>>> +++ b/kernel/sched/fair.c
>>> @@ -3835,12 +3835,15 @@ static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
>>>       }
>>>   }
>>> +static inline void update_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags);
>>> +
>>>   void reweight_task(struct task_struct *p, const struct load_weight *lw)
>>>   {
>>>       struct sched_entity *se = &p->se;
>>>       struct cfs_rq *cfs_rq = cfs_rq_of(se);
>>>       struct load_weight *load = &se->load;
>>> +    update_load_avg(cfs_rq, se, 0);
>>
>> Seems to be necessary when we reach here from __setscheduler_params() or
>> set_user_nice() for a sleeping task. Please feel free to add:
>>
>> Reviewed-by: K Prateek Nayak <kprateek.nayak@amd.com>
>>
>> But since we are on the subject of accurate PELT accounting, one question
>> I have here is whether a reweight_task() for a sleeping task race with
>> its wakeup? Something like the following scenario:
> 
> These two paths are impossible to race, since we have the task->pi_lock.

Ah! Yes, you are right. Thank you for the clarification :)

-- 
Thanks and Regards,
Prateek

> 
> Thanks.
> 
>>
>> [..snip..]
>>

Re: [PATCH v2] sched/fair: Sync se's load_avg with cfs_rq in reweight_task

[Qais Yousef]

On 07/22/24 10:47, K Prateek Nayak wrote:
> (+ Qais)
> 
> Hello Chuyi,
> 
> On 7/20/2024 10:42 AM, Chuyi Zhou wrote:
> > In reweight_task(), there are two situations:
> > 
> > 1. The task was on_rq, then the task's load_avg is accurate because we
> > synchronized it with cfs_rq through update_load_avg() in dequeue_task().
> > 
> > 2. The task is sleeping, its load_avg might not have been updated for some
> > time, which can result in inaccurate dequeue_load_avg() in
> > reweight_entity().
> > 
> > This patch solves this by using update_load_avg() to synchronize the
> > load_avg of se with cfs_rq. For tasks were on_rq, since we already update
> > load_avg to accurate values in dequeue_task(), this change will not have
> > other effects due to the short time interval between the two updates.
> > 
> > Signed-off-by: Chuyi Zhou <zhouchuyi@bytedance.com>
> > ---
> > Changes in v2:
> > - change the description in commit log.
> > - use update_load_avg() in reweight_task() rather than in reweight_entity
> > suggested by chengming.
> > - Link to v1: https://lore.kernel.org/lkml/20240716150840.23061-1-zhouchuyi@bytedance.com/
> > ---
> >   kernel/sched/fair.c | 3 +++
> >   1 file changed, 3 insertions(+)
> > 
> > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> > index 9057584ec06d..b1e07ce90284 100644
> > --- a/kernel/sched/fair.c
> > +++ b/kernel/sched/fair.c
> > @@ -3835,12 +3835,15 @@ static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
> >   	}
> >   }
> > +static inline void update_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags);
> > +
> >   void reweight_task(struct task_struct *p, const struct load_weight *lw)
> >   {
> >   	struct sched_entity *se = &p->se;
> >   	struct cfs_rq *cfs_rq = cfs_rq_of(se);
> >   	struct load_weight *load = &se->load;
> > +	update_load_avg(cfs_rq, se, 0);

White space and a comment perhaps?

LGTM anyway.

Re: [PATCH v2] sched/fair: Sync se's load_avg with cfs_rq in reweight_task

[Dietmar Eggemann]

On 20/07/2024 07:12, Chuyi Zhou wrote:
> In reweight_task(), there are two situations:
> 
> 1. The task was on_rq, then the task's load_avg is accurate because we
> synchronized it with cfs_rq through update_load_avg() in dequeue_task().
> 
> 2. The task is sleeping, its load_avg might not have been updated for some
> time, which can result in inaccurate dequeue_load_avg() in
> reweight_entity().
> 
> This patch solves this by using update_load_avg() to synchronize the
> load_avg of se with cfs_rq. For tasks were on_rq, since we already update
> load_avg to accurate values in dequeue_task(), this change will not have
> other effects due to the short time interval between the two updates.
> 
> Signed-off-by: Chuyi Zhou <zhouchuyi@bytedance.com>
> ---
> Changes in v2:
> - change the description in commit log.
> - use update_load_avg() in reweight_task() rather than in reweight_entity
> suggested by chengming.
> - Link to v1: https://lore.kernel.org/lkml/20240716150840.23061-1-zhouchuyi@bytedance.com/
> ---
>  kernel/sched/fair.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 9057584ec06d..b1e07ce90284 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -3835,12 +3835,15 @@ static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
>  	}
>  }
>  
> +static inline void update_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags);
> +
>  void reweight_task(struct task_struct *p, const struct load_weight *lw)
>  {
>  	struct sched_entity *se = &p->se;
>  	struct cfs_rq *cfs_rq = cfs_rq_of(se);
>  	struct load_weight *load = &se->load;
>  
> +	update_load_avg(cfs_rq, se, 0);

IIUC, you only want to sync the sleeping task with its cfs_rq. IMHO, 
sync_entity_load_avg() should be used here instead of update_load_avg(). 
The latter is doing much more than this.

>  	reweight_entity(cfs_rq, se, lw->weight);
>  	load->inv_weight = lw->inv_weight;
>  }

Maybe even do this in reweight_entity()?. You would have to do it under 
'if (!se->on_rq) in reweight_task() anyway I assume.

-->8--

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 9057584ec06d..555392be4e82 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -3669,11 +3669,31 @@ dequeue_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se)
        cfs_rq->avg.load_sum = max_t(u32, cfs_rq->avg.load_sum,
                                          cfs_rq->avg.load_avg * PELT_MIN_DIVIDER);
 }
+
+static inline u64 cfs_rq_last_update_time(struct cfs_rq *cfs_rq)
+{
+       return u64_u32_load_copy(cfs_rq->avg.last_update_time,
+                                cfs_rq->last_update_time_copy);
+}
+
+/*
+ * Synchronize entity load avg of dequeued entity without locking
+ * the previous rq.
+ */
+static void sync_entity_load_avg(struct sched_entity *se)
+{
+       struct cfs_rq *cfs_rq = cfs_rq_of(se);
+       u64 last_update_time;
+
+       last_update_time = cfs_rq_last_update_time(cfs_rq);
+       __update_load_avg_blocked_se(last_update_time, se);
+}
 #else
 static inline void
 enqueue_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se) { }
 static inline void
 dequeue_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se) { }
+static void sync_entity_load_avg(struct sched_entity *se) { }
 #endif
 
 static void reweight_eevdf(struct sched_entity *se, u64 avruntime,
@@ -3795,7 +3815,10 @@ static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
                if (!curr)
                        __dequeue_entity(cfs_rq, se);
                update_load_sub(&cfs_rq->load, se->load.weight);
+       } else if (entity_is_task(se)) {
+               sync_entity_load_avg(se);
        }
+
        dequeue_load_avg(cfs_rq, se);
 
        if (se->on_rq) {
@@ -4033,12 +4056,6 @@ static inline bool load_avg_is_decayed(struct sched_avg *sa)
 
        return true;
 }
-
-static inline u64 cfs_rq_last_update_time(struct cfs_rq *cfs_rq)
-{
-       return u64_u32_load_copy(cfs_rq->avg.last_update_time,
-                                cfs_rq->last_update_time_copy);
-}
 #ifdef CONFIG_FAIR_GROUP_SCHED
 /*
  * Because list_add_leaf_cfs_rq always places a child cfs_rq on the list
@@ -4773,19 +4790,6 @@ static inline void update_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *s
        }
 }
 
-/*
- * Synchronize entity load avg of dequeued entity without locking
- * the previous rq.
- */
-static void sync_entity_load_avg(struct sched_entity *se)
-{
-       struct cfs_rq *cfs_rq = cfs_rq_of(se);
-       u64 last_update_time;
-
-       last_update_time = cfs_rq_last_update_time(cfs_rq);
-       __update_load_avg_blocked_se(last_update_time, se);
-}
-
 /*
  * Task first catches up with cfs_rq, and then subtract
  * itself from the cfs_rq (task must be off the queue now).

Re: [PATCH v2] sched/fair: Sync se's load_avg with cfs_rq in reweight_task

[Chuyi Zhou]

Hello,

在 2024/7/22 14:34, Dietmar Eggemann 写道:
> On 20/07/2024 07:12, Chuyi Zhou wrote:
>> In reweight_task(), there are two situations:
>>
>> 1. The task was on_rq, then the task's load_avg is accurate because we
>> synchronized it with cfs_rq through update_load_avg() in dequeue_task().
>>
>> 2. The task is sleeping, its load_avg might not have been updated for some
>> time, which can result in inaccurate dequeue_load_avg() in
>> reweight_entity().
>>
>> This patch solves this by using update_load_avg() to synchronize the
>> load_avg of se with cfs_rq. For tasks were on_rq, since we already update
>> load_avg to accurate values in dequeue_task(), this change will not have
>> other effects due to the short time interval between the two updates.
>>
>> Signed-off-by: Chuyi Zhou <zhouchuyi@bytedance.com>
>> ---
>> Changes in v2:
>> - change the description in commit log.
>> - use update_load_avg() in reweight_task() rather than in reweight_entity
>> suggested by chengming.
>> - Link to v1: https://lore.kernel.org/lkml/20240716150840.23061-1-zhouchuyi@bytedance.com/
>> ---
>>   kernel/sched/fair.c | 3 +++
>>   1 file changed, 3 insertions(+)
>>
>> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
>> index 9057584ec06d..b1e07ce90284 100644
>> --- a/kernel/sched/fair.c
>> +++ b/kernel/sched/fair.c
>> @@ -3835,12 +3835,15 @@ static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
>>   	}
>>   }
>>   
>> +static inline void update_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags);
>> +
>>   void reweight_task(struct task_struct *p, const struct load_weight *lw)
>>   {
>>   	struct sched_entity *se = &p->se;
>>   	struct cfs_rq *cfs_rq = cfs_rq_of(se);
>>   	struct load_weight *load = &se->load;
>>   
>> +	update_load_avg(cfs_rq, se, 0);
> 
> IIUC, you only want to sync the sleeping task with its cfs_rq. IMHO,
> sync_entity_load_avg() should be used here instead of update_load_avg().
> The latter is doing much more than this.

Indeed, sync_entity_load_avg() is better.

> 
>>   	reweight_entity(cfs_rq, se, lw->weight);
>>   	load->inv_weight = lw->inv_weight;
>>   }
> 
> Maybe even do this in reweight_entity()?. You would have to do it under
> 'if (!se->on_rq) in reweight_task() anyway I assume.
> 

Yes, we can do it reweight_entity() and it's more clear.

Thanks for your suggestion.

Re: [PATCH v2] sched/fair: Sync se's load_avg with cfs_rq in reweight_task

[Chuyi Zhou]

-------- 转发的消息 --------
主题: Re: [PATCH v2] sched/fair: Sync se's load_avg with cfs_rq in 
reweight_task
日期: Mon, 29 Jul 2024 11:06:33 +0800
From: Chuyi Zhou <zhouchuyi@bytedance.com>
收件人: Qais Yousef <qyousef@layalina.io>, K Prateek Nayak 
<kprateek.nayak@amd.com>
抄送: mingo@redhat.com, peterz@infradead.org, juri.lelli@redhat.com, 
vincent.guittot@linaro.org, dietmar.eggemann@arm.com, 
rostedt@goodmis.org, bsegall@google.com, mgorman@suse.de, 
vschneid@redhat.com, chengming.zhou@linux.dev

Hello Qais,

在 2024/7/29 04:14, Qais Yousef 写道:
> On 07/22/24 10:47, K Prateek Nayak wrote:
>> (+ Qais)
>>
>> Hello Chuyi,
>>
>> On 7/20/2024 10:42 AM, Chuyi Zhou wrote:
>>> In reweight_task(), there are two situations:
>>>
>>> 1. The task was on_rq, then the task's load_avg is accurate because we
>>> synchronized it with cfs_rq through update_load_avg() in dequeue_task().
>>>
>>> 2. The task is sleeping, its load_avg might not have been updated for some
>>> time, which can result in inaccurate dequeue_load_avg() in
>>> reweight_entity().
>>>
>>> This patch solves this by using update_load_avg() to synchronize the
>>> load_avg of se with cfs_rq. For tasks were on_rq, since we already update
>>> load_avg to accurate values in dequeue_task(), this change will not have
>>> other effects due to the short time interval between the two updates.
>>>
>>> Signed-off-by: Chuyi Zhou <zhouchuyi@bytedance.com>
>>> ---
>>> Changes in v2:
>>> - change the description in commit log.
>>> - use update_load_avg() in reweight_task() rather than in reweight_entity
>>> suggested by chengming.
>>> - Link to v1: https://lore.kernel.org/lkml/20240716150840.23061-1-zhouchuyi@bytedance.com/
>>> ---
>>>    kernel/sched/fair.c | 3 +++
>>>    1 file changed, 3 insertions(+)
>>>
>>> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
>>> index 9057584ec06d..b1e07ce90284 100644
>>> --- a/kernel/sched/fair.c
>>> +++ b/kernel/sched/fair.c
>>> @@ -3835,12 +3835,15 @@ static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
>>>    	}
>>>    }
>>> +static inline void update_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags);
>>> +
>>>    void reweight_task(struct task_struct *p, const struct load_weight *lw)
>>>    {
>>>    	struct sched_entity *se = &p->se;
>>>    	struct cfs_rq *cfs_rq = cfs_rq_of(se);
>>>    	struct load_weight *load = &se->load;
>>> +	update_load_avg(cfs_rq, se, 0);
> 
> White space and a comment perhaps?
> 
> LGTM anyway.

Thanks for taking a look, will send another patch later.

(V3 
link:https://lore.kernel.org/lkml/20240723114247.104848-1-zhouchuyi@bytedance.com/)