Re: [PATCH] sched/fair: Avoid transient double accounting for delayed tasks
From: Lad, Prabhakar
Date: Fri Nov 20 2026 - 08:03:53 EST
Hi all,
Sorry for top posting.
The original commit [0] in dequeue_hierarchy() set `h_nr_runnable =
1;` where as this seems to be missed in the merge [1] which is now
commit 486c3d06b90240860e99686d29ed9f15637dc314 in next. Due to this,
I'm seeing performance drops while running tests on Renesas SoCs.
[0] https://lore.kernel.org/all/20261112083447.25603-1-kprateek.nayak@xxxxxxx/
[1] https://lore.kernel.org/all/179500916013.3191731.3149125419718630663.tip-bot2@tip-bot2/
diff from original commit:
---------------------------------
@@ -8317,16 +8307,15 @@ enqueue_task_fair(struct rq *rq, struct
task_struct *p, int flags)
static void dequeue_hierarchy(struct task_struct *p, int flags)
{
struct sched_entity *se = &p->se;
- bool task_sleep = flags & DEQUEUE_SLEEP;
- bool task_delayed = flags & DEQUEUE_DELAYED;
bool task_throttled = flags & DEQUEUE_THROTTLE;
int h_nr_idle = task_has_idle_policy(p);
- int h_nr_runnable = 0;
+ int h_nr_runnable = 1;
struct cfs_rq *cfs_rq;
bool dequeue = true;
Cheers,
Prabhakar