2024-01-10 13:32:10

by Junwen Wu

[permalink] [raw]
Subject: [PATCH v4] sched/stats: Fix rt/dl task's sched latency statistics error in sched_stat_wait trace_point

The sched_stat_wait tracepoint is showing unreasonably long
latencies for real-time tasks. For example:

sched_stat_wait: comm=rcu_preempt pid=14 delay=4936139545261 [ns]

This error happens when the rt task balances off the source CPU because
the dequeue operation is not updating the sched_statistics. So, follow
update_stats_wait_end_fair() and update the stats. Do the same for
SCHED_DEADLINE.

Fixes: 57a5c2dafca8 ("sched/rt: Support schedstats for RT sched class")
Fixes: b5eb4a5f6521 ("sched/dl: Support schedstats for deadline sched class")
Signed-off-by: Junwen Wu <[email protected]>
---
kernel/sched/deadline.c | 5 ++++-
kernel/sched/rt.c | 4 ++++
2 files changed, 8 insertions(+), 1 deletion(-)

diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index b28114478b82..4a9aad291fb9 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -1558,10 +1558,13 @@ update_stats_dequeue_dl(struct dl_rq *dl_rq, struct sched_dl_entity *dl_se,
int flags)
{
struct task_struct *p = dl_task_of(dl_se);
+ struct rq *rq = rq_of_dl_rq(dl_rq);

if (!schedstat_enabled())
return;
-
+ /* Mark the end of the wait period if dequeueing a waiting task.*/
+ if (p && (p != rq->curr))
+ update_stats_wait_end_dl(dl_rq, dl_se);
if ((flags & DEQUEUE_SLEEP)) {
unsigned int state;

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 6aaf0a3d6081..5cb3a54d6b13 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1360,12 +1360,16 @@ update_stats_dequeue_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se,
int flags)
{
struct task_struct *p = NULL;
+ struct rq *rq = rq_of_rt_se(rt_se);

if (!schedstat_enabled())
return;

if (rt_entity_is_task(rt_se))
p = rt_task_of(rt_se);
+ /* Mark the end of the wait period if dequeueing a waiting task. */
+ if (p && (p != rq->curr))
+ update_stats_wait_end_rt(rt_rq, rt_se);

if ((flags & DEQUEUE_SLEEP) && p) {
unsigned int state;
--
2.34.1



2024-01-11 11:53:15

by Yafang Shao

[permalink] [raw]
Subject: Re: [PATCH v4] sched/stats: Fix rt/dl task's sched latency statistics error in sched_stat_wait trace_point

On Wed, Jan 10, 2024 at 9:32 PM Junwen Wu <[email protected]> wrote:
>
> The sched_stat_wait tracepoint is showing unreasonably long
> latencies for real-time tasks. For example:
>
> sched_stat_wait: comm=rcu_preempt pid=14 delay=4936139545261 [ns]
>
> This error happens when the rt task balances off the source CPU because
> the dequeue operation is not updating the sched_statistics. So, follow
> update_stats_wait_end_fair() and update the stats. Do the same for
> SCHED_DEADLINE.
>
> Fixes: 57a5c2dafca8 ("sched/rt: Support schedstats for RT sched class")
> Fixes: b5eb4a5f6521 ("sched/dl: Support schedstats for deadline sched class")
> Signed-off-by: Junwen Wu <[email protected]>

Acked-by: Yafang Shao <[email protected]>

BTW, feel free to include the 'acked-by' in a newer version if your
modifications are only minor.

> ---
> kernel/sched/deadline.c | 5 ++++-
> kernel/sched/rt.c | 4 ++++
> 2 files changed, 8 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
> index b28114478b82..4a9aad291fb9 100644
> --- a/kernel/sched/deadline.c
> +++ b/kernel/sched/deadline.c
> @@ -1558,10 +1558,13 @@ update_stats_dequeue_dl(struct dl_rq *dl_rq, struct sched_dl_entity *dl_se,
> int flags)
> {
> struct task_struct *p = dl_task_of(dl_se);
> + struct rq *rq = rq_of_dl_rq(dl_rq);
>
> if (!schedstat_enabled())
> return;
> -
> + /* Mark the end of the wait period if dequeueing a waiting task.*/
> + if (p && (p != rq->curr))
> + update_stats_wait_end_dl(dl_rq, dl_se);
> if ((flags & DEQUEUE_SLEEP)) {
> unsigned int state;
>
> diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
> index 6aaf0a3d6081..5cb3a54d6b13 100644
> --- a/kernel/sched/rt.c
> +++ b/kernel/sched/rt.c
> @@ -1360,12 +1360,16 @@ update_stats_dequeue_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se,
> int flags)
> {
> struct task_struct *p = NULL;
> + struct rq *rq = rq_of_rt_se(rt_se);
>
> if (!schedstat_enabled())
> return;
>
> if (rt_entity_is_task(rt_se))
> p = rt_task_of(rt_se);
> + /* Mark the end of the wait period if dequeueing a waiting task. */
> + if (p && (p != rq->curr))
> + update_stats_wait_end_rt(rt_rq, rt_se);
>
> if ((flags & DEQUEUE_SLEEP) && p) {
> unsigned int state;
> --
> 2.34.1
>


--
Regards
Yafang