sched: Move curr check into __enqueue_entity and __dequeue_entity

[PATCH v1] sched: Move curr check into __enqueue_entity and __dequeue_entity

Posted by Jemmy Wong 3 months, 2 weeks ago

The check for se == cfs_rq->curr is common logic used before calling
__enqueue_entity and __dequeue_entity. Move this check into the respective
function bodies to simplify the code.

Signed-off-by: Jemmy Wong <jemmywong512@gmail.com>

---
 kernel/sched/fair.c | 32 +++++++++++++++-----------------
 1 file changed, 15 insertions(+), 17 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 7a14da5396fb..24cbb4b6c3fb 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -849,6 +849,9 @@ RB_DECLARE_CALLBACKS(static, min_vruntime_cb, struct sched_entity,
  */
 static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
+	if (unlikely(se == cfs_rq->curr))
+		return;
+
 	avg_vruntime_add(cfs_rq, se);
 	se->min_vruntime = se->vruntime;
 	se->min_slice = se->slice;
@@ -858,6 +861,9 @@ static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)

 static void __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
+	if (unlikely(se == cfs_rq->curr))
+		return;
+
 	rb_erase_augmented_cached(&se->run_node, &cfs_rq->tasks_timeline,
 				  &min_vruntime_cb);
 	avg_vruntime_sub(cfs_rq, se);
@@ -3797,8 +3803,6 @@ static void place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int fla
 static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
 			    unsigned long weight)
 {
-	bool curr = cfs_rq->curr == se;
-
 	if (se->on_rq) {
 		/* commit outstanding execution time */
 		update_curr(cfs_rq);
@@ -3806,8 +3810,7 @@ static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
 		se->deadline -= se->vruntime;
 		se->rel_deadline = 1;
 		cfs_rq->nr_queued--;
-		if (!curr)
-			__dequeue_entity(cfs_rq, se);
+		__dequeue_entity(cfs_rq, se);
 		update_load_sub(&cfs_rq->load, se->load.weight);
 	}
 	dequeue_load_avg(cfs_rq, se);
@@ -3834,8 +3837,7 @@ static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
 	if (se->on_rq) {
 		place_entity(cfs_rq, se, 0);
 		update_load_add(&cfs_rq->load, se->load.weight);
-		if (!curr)
-			__enqueue_entity(cfs_rq, se);
+		__enqueue_entity(cfs_rq, se);
 		cfs_rq->nr_queued++;

 		/*
@@ -5363,8 +5365,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)

 	check_schedstat_required();
 	update_stats_enqueue_fair(cfs_rq, se, flags);
-	if (!curr)
-		__enqueue_entity(cfs_rq, se);
+	__enqueue_entity(cfs_rq, se);
 	se->on_rq = 1;

 	if (cfs_rq->nr_queued == 1) {
@@ -5506,8 +5507,7 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 		se->rel_deadline = 1;
 	}

-	if (se != cfs_rq->curr)
-		__dequeue_entity(cfs_rq, se);
+	__dequeue_entity(cfs_rq, se);
 	se->on_rq = 0;
 	account_entity_dequeue(cfs_rq, se);

@@ -5624,14 +5624,14 @@ static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev)
 	/* throttle cfs_rqs exceeding runtime */
 	check_cfs_rq_runtime(cfs_rq);

+	WARN_ON_ONCE(cfs_rq->curr != prev);
 	if (prev->on_rq) {
 		update_stats_wait_start_fair(cfs_rq, prev);
-		/* Put 'current' back into the tree. */
-		__enqueue_entity(cfs_rq, prev);
 		/* in !on_rq case, update occurred at dequeue */
 		update_load_avg(cfs_rq, prev, 0);
+		cfs_rq->curr = NULL;
+		__enqueue_entity(cfs_rq, prev);
 	}
-	WARN_ON_ONCE(cfs_rq->curr != prev);
 	cfs_rq->curr = NULL;
 }

@@ -6910,12 +6910,10 @@ requeue_delayed_entity(struct sched_entity *se)
 		update_entity_lag(cfs_rq, se);
 		if (se->vlag > 0) {
 			cfs_rq->nr_queued--;
-			if (se != cfs_rq->curr)
-				__dequeue_entity(cfs_rq, se);
+			__dequeue_entity(cfs_rq, se);
 			se->vlag = 0;
 			place_entity(cfs_rq, se, 0);
-			if (se != cfs_rq->curr)
-				__enqueue_entity(cfs_rq, se);
+			__enqueue_entity(cfs_rq, se);
 			cfs_rq->nr_queued++;
 		}
 	}
--
2.43.0

Re: [PATCH v1] sched: Move curr check into __enqueue_entity and __dequeue_entity

Posted by Phil Auld 3 months, 2 weeks ago

Hi Jemmy,

On Sat, Jun 21, 2025 at 10:20:55PM +0800 Jemmy Wong wrote:
> The check for se == cfs_rq->curr is common logic used before calling
> __enqueue_entity and __dequeue_entity. Move this check into the respective
> function bodies to simplify the code.


I'm not sure this simpifies the code.  The pattern is generally that
the underscore prefixed function just does the required work and the
non-underscore-prefixed one does the checks. That's a pattern used
in lots of places.

Also, unless these end up unlined (I didn't check but they are not
explicitly called out as such) you are adding an extra function
call. 

> 
> Signed-off-by: Jemmy Wong <jemmywong512@gmail.com>
> 
> ---
>  kernel/sched/fair.c | 32 +++++++++++++++-----------------
>  1 file changed, 15 insertions(+), 17 deletions(-)
> 
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 7a14da5396fb..24cbb4b6c3fb 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -849,6 +849,9 @@ RB_DECLARE_CALLBACKS(static, min_vruntime_cb, struct sched_entity,
>   */
>  static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
>  {
> +	if (unlikely(se == cfs_rq->curr))
> +		return;
> +
>  	avg_vruntime_add(cfs_rq, se);
>  	se->min_vruntime = se->vruntime;
>  	se->min_slice = se->slice;
> @@ -858,6 +861,9 @@ static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
> 
>  static void __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
>  {
> +	if (unlikely(se == cfs_rq->curr))
> +		return;
> +
>  	rb_erase_augmented_cached(&se->run_node, &cfs_rq->tasks_timeline,
>  				  &min_vruntime_cb);
>  	avg_vruntime_sub(cfs_rq, se);
> @@ -3797,8 +3803,6 @@ static void place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int fla
>  static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
>  			    unsigned long weight)
>  {
> -	bool curr = cfs_rq->curr == se;
> -
>  	if (se->on_rq) {
>  		/* commit outstanding execution time */
>  		update_curr(cfs_rq);
> @@ -3806,8 +3810,7 @@ static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
>  		se->deadline -= se->vruntime;
>  		se->rel_deadline = 1;
>  		cfs_rq->nr_queued--;
> -		if (!curr)
> -			__dequeue_entity(cfs_rq, se);
> +		__dequeue_entity(cfs_rq, se);
>  		update_load_sub(&cfs_rq->load, se->load.weight);
>  	}
>  	dequeue_load_avg(cfs_rq, se);
> @@ -3834,8 +3837,7 @@ static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
>  	if (se->on_rq) {
>  		place_entity(cfs_rq, se, 0);
>  		update_load_add(&cfs_rq->load, se->load.weight);
> -		if (!curr)
> -			__enqueue_entity(cfs_rq, se);
> +		__enqueue_entity(cfs_rq, se);
>  		cfs_rq->nr_queued++;
> 
>  		/*
> @@ -5363,8 +5365,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
> 
>  	check_schedstat_required();
>  	update_stats_enqueue_fair(cfs_rq, se, flags);
> -	if (!curr)
> -		__enqueue_entity(cfs_rq, se);
> +	__enqueue_entity(cfs_rq, se);
>  	se->on_rq = 1;
> 
>  	if (cfs_rq->nr_queued == 1) {
> @@ -5506,8 +5507,7 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
>  		se->rel_deadline = 1;
>  	}
> 
> -	if (se != cfs_rq->curr)
> -		__dequeue_entity(cfs_rq, se);
> +	__dequeue_entity(cfs_rq, se);
>  	se->on_rq = 0;
>  	account_entity_dequeue(cfs_rq, se);
> 
> @@ -5624,14 +5624,14 @@ static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev)
>  	/* throttle cfs_rqs exceeding runtime */
>  	check_cfs_rq_runtime(cfs_rq);
> 
> +	WARN_ON_ONCE(cfs_rq->curr != prev);
>  	if (prev->on_rq) {
>  		update_stats_wait_start_fair(cfs_rq, prev);
> -		/* Put 'current' back into the tree. */
> -		__enqueue_entity(cfs_rq, prev);
>  		/* in !on_rq case, update occurred at dequeue */
>  		update_load_avg(cfs_rq, prev, 0);
> +		cfs_rq->curr = NULL;
> +		__enqueue_entity(cfs_rq, prev);
>  	}
> -	WARN_ON_ONCE(cfs_rq->curr != prev);
>  	cfs_rq->curr = NULL;
>  }

And here you are changing things around a bit more. 

Plus possibly clearing curr twice. 

Anyway, others may feel differently. But I don't think this patch is
needed.


Cheers,
Phil



> 
> @@ -6910,12 +6910,10 @@ requeue_delayed_entity(struct sched_entity *se)
>  		update_entity_lag(cfs_rq, se);
>  		if (se->vlag > 0) {
>  			cfs_rq->nr_queued--;
> -			if (se != cfs_rq->curr)
> -				__dequeue_entity(cfs_rq, se);
> +			__dequeue_entity(cfs_rq, se);
>  			se->vlag = 0;
>  			place_entity(cfs_rq, se, 0);
> -			if (se != cfs_rq->curr)
> -				__enqueue_entity(cfs_rq, se);
> +			__enqueue_entity(cfs_rq, se);
>  			cfs_rq->nr_queued++;
>  		}
>  	}
> --
> 2.43.0
> 

--