lkml.org 
[lkml]   [2019]   [Jan]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH] sched/debug: Show intergroup and hierarchy sum wait time of a task group
From
Date


On 2019/1/23 下午5:46, ufo19890607@gmail.com wrote:
> From: yuzhoujian <yuzhoujian@didichuxing.com>
>
> We can monitor the sum wait time of a task group since 'commit 3d6c50c27bd6
> ("sched/debug: Show the sum wait time of a task group")'. However this
> wait_sum just represents the confilct between different task groups, since
> it is simply sum the wait time of task_group's cfs_rq. And we still cannot
> evaluate the conflict between all the tasks within hierarchy of this group,
> so the hierarchy wait time is still needed.

Could you please give us a scene that we do need this hierarchy wait_sum,
despite the extra overhead?

Regards,
Michael Wang

>
> Thus we introduce hierarchy wait_sum which summarizes the total wait sum of
> all the tasks in the hierarchy of a group.
>
> The 'cpu.stat' is modified to show the statistic, like:
>
> nr_periods 0
> nr_throttled 0
> throttled_time 0
> intergroup wait_sum 2842251984
> hierarchy wait_sum 6389509389332798
>
> From now on we can monitor both the wait_sum of intergroup and hierarchy,
> which will inevitably help a system administrator know how intense the CPU
> competition is within a task group and between different task groups. We
> can calculate the wait rate of a task group based on hierarchy wait_sum and
> cpuacct.usage.
>
> For example:
> X% = (current_wait_sum - last_wait_sum) / ((current_usage -
> last_usage) + (current_wait_sum - last_wait_sum))
>
> That means the task group paid X percentage of time on runqueue waiting
> for the CPU.
>
> Signed-off-by: yuzhoujian <yuzhoujian@didichuxing.com>
> ---
> kernel/sched/core.c | 11 +++++++----
> kernel/sched/fair.c | 17 +++++++++++++++++
> kernel/sched/sched.h | 3 +++
> 3 files changed, 27 insertions(+), 4 deletions(-)
>
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index ee77636..172e6fb 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -6760,13 +6760,16 @@ static int cpu_cfs_stat_show(struct seq_file *sf, void *v)
> seq_printf(sf, "throttled_time %llu\n", cfs_b->throttled_time);
>
> if (schedstat_enabled() && tg != &root_task_group) {
> - u64 ws = 0;
> + u64 inter_ws = 0, hierarchy_ws = 0;
> int i;
>
> - for_each_possible_cpu(i)
> - ws += schedstat_val(tg->se[i]->statistics.wait_sum);
> + for_each_possible_cpu(i) {
> + inter_ws += schedstat_val(tg->se[i]->statistics.wait_sum);
> + hierarchy_ws += tg->cfs_rq[i]->hierarchy_wait_sum;
> + }
>
> - seq_printf(sf, "wait_sum %llu\n", ws);
> + seq_printf(sf, "intergroup wait_sum %llu\n", inter_ws);
> + seq_printf(sf, "hierarchy wait_sum %llu\n", hierarchy_ws);
> }
>
> return 0;
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index e2ff4b6..35e89ca 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -858,6 +858,19 @@ static void update_curr_fair(struct rq *rq)
> }
>
> static inline void
> +update_hierarchy_wait_sum(struct sched_entity *se,
> + u64 delta_wait)
> +{
> + for_each_sched_entity(se) {
> + struct cfs_rq *cfs_rq = cfs_rq_of(se);
> +
> + if (cfs_rq->tg != &root_task_group)
> + __schedstat_add(cfs_rq->hierarchy_wait_sum,
> + delta_wait);
> + }
> +}
> +
> +static inline void
> update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
> {
> struct task_struct *p;
> @@ -880,6 +893,7 @@ static void update_curr_fair(struct rq *rq)
> return;
> }
> trace_sched_stat_wait(p, delta);
> + update_hierarchy_wait_sum(se, delta);
> }
>
> __schedstat_set(se->statistics.wait_max,
> @@ -10273,6 +10287,9 @@ void init_cfs_rq(struct cfs_rq *cfs_rq)
> #ifndef CONFIG_64BIT
> cfs_rq->min_vruntime_copy = cfs_rq->min_vruntime;
> #endif
> +#ifdef CONFIG_SCHEDSTATS
> + cfs_rq->hierarchy_wait_sum = 0;
> +#endif
> #ifdef CONFIG_SMP
> raw_spin_lock_init(&cfs_rq->removed.lock);
> #endif
> diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
> index d27c1a5..c01ab99 100644
> --- a/kernel/sched/sched.h
> +++ b/kernel/sched/sched.h
> @@ -496,6 +496,9 @@ struct cfs_rq {
> #ifndef CONFIG_64BIT
> u64 min_vruntime_copy;
> #endif
> +#ifdef CONFIG_SCHEDSTATS
> + u64 hierarchy_wait_sum;
> +#endif
>
> struct rb_root_cached tasks_timeline;
>
>

\
 
 \ /
  Last update: 2019-01-25 04:13    [W:0.109 / U:2.600 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site