[PATCH v3 1/2] sched: don't account throttle time for empty groups

[PATCH v3 1/2] sched: don't account throttle time for empty groups

[Josh Don]

It is easy for a cfs_rq to become throttled even when it has no enqueued
entities (for example, if we have just put_prev()'d the last runnable
task of the cfs_rq, and the cfs_rq is out of quota).

Avoid accounting this time towards total throttle time, since it
otherwise falsely inflates the stats.

Note that the dequeue path is special, since we normally disallow
migrations when a task is in a throttled hierarchy (see
throttled_lb_pair()).

Signed-off-by: Josh Don <joshdon-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>
---
v3: fix bug with nested limits
 kernel/sched/fair.c | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 6189d1a45635..72571d48559a 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4787,6 +4787,7 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
 }
 
 static void check_enqueue_throttle(struct cfs_rq *cfs_rq);
+static inline int cfs_rq_throttled(struct cfs_rq *cfs_rq);
 
 static inline bool cfs_bandwidth_used(void);
 
@@ -4873,8 +4874,14 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 
 	if (cfs_rq->nr_running == 1) {
 		check_enqueue_throttle(cfs_rq);
-		if (!throttled_hierarchy(cfs_rq))
+		if (!throttled_hierarchy(cfs_rq)) {
 			list_add_leaf_cfs_rq(cfs_rq);
+		} else {
+#ifdef CONFIG_CFS_BANDWIDTH
+			if (cfs_rq_throttled(cfs_rq) && !cfs_rq->throttled_clock)
+				cfs_rq->throttled_clock = rq_clock(rq_of(cfs_rq));
+#endif
+		}
 	}
 }
 
@@ -5480,7 +5487,9 @@ static bool throttle_cfs_rq(struct cfs_rq *cfs_rq)
 	 * throttled-list.  rq->lock protects completion.
 	 */
 	cfs_rq->throttled = 1;
-	cfs_rq->throttled_clock = rq_clock(rq);
+	SCHED_WARN_ON(cfs_rq->throttled_clock);
+	if (cfs_rq->nr_running)
+		cfs_rq->throttled_clock = rq_clock(rq);
 	return true;
 }
 
@@ -5498,7 +5507,10 @@ void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
 	update_rq_clock(rq);
 
 	raw_spin_lock(&cfs_b->lock);
-	cfs_b->throttled_time += rq_clock(rq) - cfs_rq->throttled_clock;
+	if (cfs_rq->throttled_clock) {
+		cfs_b->throttled_time += rq_clock(rq) - cfs_rq->throttled_clock;
+		cfs_rq->throttled_clock = 0;
+	}
 	list_del_rcu(&cfs_rq->throttled_list);
 	raw_spin_unlock(&cfs_b->lock);
 
-- 
2.41.0.162.gfafddb0af9-goog

[PATCH v3 2/2] sched: add throttled time stat for throttled children

[Josh Don]

We currently export the total throttled time for cgroups that are given
a bandwidth limit. This patch extends this accounting to also account
the total time that each children cgroup has been throttled.

This is useful to understand the degree to which children have been
affected by the throttling control. Children which are not runnable
during the entire throttled period, for example, will not show any
self-throttling time during this period.

Expose this in a new interface, 'cpu.stat.local', which is similar to
how non-hierarchical events are accounted in 'memory.events.local'.

Signed-off-by: Josh Don <joshdon-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>
---
v3: fix bug with nested limits
v2:
- moved export to new cpu.stat.local file, per Tejun's recommendation
 include/linux/cgroup-defs.h |  2 ++
 kernel/cgroup/cgroup.c      | 34 ++++++++++++++++++++++++++++
 kernel/sched/core.c         | 44 +++++++++++++++++++++++++++++++++++++
 kernel/sched/fair.c         | 21 +++++++++++++++++-
 kernel/sched/sched.h        |  2 ++
 5 files changed, 102 insertions(+), 1 deletion(-)

diff --git a/include/linux/cgroup-defs.h b/include/linux/cgroup-defs.h
index 8a0d5466c7be..ae20dbb885d6 100644
--- a/include/linux/cgroup-defs.h
+++ b/include/linux/cgroup-defs.h
@@ -661,6 +661,8 @@ struct cgroup_subsys {
 	void (*css_rstat_flush)(struct cgroup_subsys_state *css, int cpu);
 	int (*css_extra_stat_show)(struct seq_file *seq,
 				   struct cgroup_subsys_state *css);
+	int (*css_local_stat_show)(struct seq_file *seq,
+				   struct cgroup_subsys_state *css);
 
 	int (*can_attach)(struct cgroup_taskset *tset);
 	void (*cancel_attach)(struct cgroup_taskset *tset);
diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c
index 2d3d13e52333..bfe063225d3b 100644
--- a/kernel/cgroup/cgroup.c
+++ b/kernel/cgroup/cgroup.c
@@ -3726,6 +3726,36 @@ static int cpu_stat_show(struct seq_file *seq, void *v)
 	return ret;
 }
 
+static int __maybe_unused cgroup_local_stat_show(struct seq_file *seq,
+						 struct cgroup *cgrp, int ssid)
+{
+	struct cgroup_subsys *ss = cgroup_subsys[ssid];
+	struct cgroup_subsys_state *css;
+	int ret;
+
+	if (!ss->css_local_stat_show)
+		return 0;
+
+	css = cgroup_tryget_css(cgrp, ss);
+	if (!css)
+		return 0;
+
+	ret = ss->css_local_stat_show(seq, css);
+	css_put(css);
+	return ret;
+}
+
+static int cpu_local_stat_show(struct seq_file *seq, void *v)
+{
+	struct cgroup __maybe_unused *cgrp = seq_css(seq)->cgroup;
+	int ret = 0;
+
+#ifdef CONFIG_CGROUP_SCHED
+	ret = cgroup_local_stat_show(seq, cgrp, cpu_cgrp_id);
+#endif
+	return ret;
+}
+
 #ifdef CONFIG_PSI
 static int cgroup_io_pressure_show(struct seq_file *seq, void *v)
 {
@@ -5276,6 +5306,10 @@ static struct cftype cgroup_base_files[] = {
 		.name = "cpu.stat",
 		.seq_show = cpu_stat_show,
 	},
+	{
+		.name = "cpu.stat.local",
+		.seq_show = cpu_local_stat_show,
+	},
 	{ }	/* terminate */
 };
 
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index d9a04faa48a1..0de483190a57 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -11140,6 +11140,27 @@ static int cpu_cfs_stat_show(struct seq_file *sf, void *v)
 
 	return 0;
 }
+
+static u64 throttled_time_self(struct task_group *tg)
+{
+	int i;
+	u64 total = 0;
+
+	for_each_possible_cpu(i) {
+		total += READ_ONCE(tg->cfs_rq[i]->throttled_clock_self_time);
+	}
+
+	return total;
+}
+
+static int cpu_cfs_local_stat_show(struct seq_file *sf, void *v)
+{
+	struct task_group *tg = css_tg(seq_css(sf));
+
+	seq_printf(sf, "throttled_time %llu\n", throttled_time_self(tg));
+
+	return 0;
+}
 #endif /* CONFIG_CFS_BANDWIDTH */
 #endif /* CONFIG_FAIR_GROUP_SCHED */
 
@@ -11216,6 +11237,10 @@ static struct cftype cpu_legacy_files[] = {
 		.name = "stat",
 		.seq_show = cpu_cfs_stat_show,
 	},
+	{
+		.name = "stat.local",
+		.seq_show = cpu_cfs_local_stat_show,
+	},
 #endif
 #ifdef CONFIG_RT_GROUP_SCHED
 	{
@@ -11272,6 +11297,24 @@ static int cpu_extra_stat_show(struct seq_file *sf,
 	return 0;
 }
 
+static int cpu_local_stat_show(struct seq_file *sf,
+			       struct cgroup_subsys_state *css)
+{
+#ifdef CONFIG_CFS_BANDWIDTH
+	{
+		struct task_group *tg = css_tg(css);
+		u64 throttled_self_usec;
+
+		throttled_self_usec = throttled_time_self(tg);
+		do_div(throttled_self_usec, NSEC_PER_USEC);
+
+		seq_printf(sf, "throttled_usec %llu\n",
+			   throttled_self_usec);
+	}
+#endif
+	return 0;
+}
+
 #ifdef CONFIG_FAIR_GROUP_SCHED
 static u64 cpu_weight_read_u64(struct cgroup_subsys_state *css,
 			       struct cftype *cft)
@@ -11450,6 +11493,7 @@ struct cgroup_subsys cpu_cgrp_subsys = {
 	.css_released	= cpu_cgroup_css_released,
 	.css_free	= cpu_cgroup_css_free,
 	.css_extra_stat_show = cpu_extra_stat_show,
+	.css_local_stat_show = cpu_local_stat_show,
 #ifdef CONFIG_RT_GROUP_SCHED
 	.can_attach	= cpu_cgroup_can_attach,
 #endif
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 72571d48559a..b1bccef78b90 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4878,8 +4878,12 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 			list_add_leaf_cfs_rq(cfs_rq);
 		} else {
 #ifdef CONFIG_CFS_BANDWIDTH
+			struct rq *rq = rq_of(cfs_rq);
+
 			if (cfs_rq_throttled(cfs_rq) && !cfs_rq->throttled_clock)
-				cfs_rq->throttled_clock = rq_clock(rq_of(cfs_rq));
+				cfs_rq->throttled_clock = rq_clock(rq);
+			if (!cfs_rq->throttled_clock_self)
+				cfs_rq->throttled_clock_self = rq_clock(rq);
 #endif
 		}
 	}
@@ -5384,6 +5388,17 @@ static int tg_unthrottle_up(struct task_group *tg, void *data)
 		/* Add cfs_rq with load or one or more already running entities to the list */
 		if (!cfs_rq_is_decayed(cfs_rq))
 			list_add_leaf_cfs_rq(cfs_rq);
+
+		if (cfs_rq->throttled_clock_self) {
+			u64 delta = rq_clock(rq) - cfs_rq->throttled_clock_self;
+
+			cfs_rq->throttled_clock_self = 0;
+
+			if (SCHED_WARN_ON((s64)delta < 0))
+				delta = 0;
+
+			cfs_rq->throttled_clock_self_time += delta;
+		}
 	}
 
 	return 0;
@@ -5398,6 +5413,10 @@ static int tg_throttle_down(struct task_group *tg, void *data)
 	if (!cfs_rq->throttle_count) {
 		cfs_rq->throttled_clock_pelt = rq_clock_pelt(rq);
 		list_del_leaf_cfs_rq(cfs_rq);
+
+		SCHED_WARN_ON(cfs_rq->throttled_clock_self);
+		if (cfs_rq->nr_running)
+			cfs_rq->throttled_clock_self = rq_clock(rq);
 	}
 	cfs_rq->throttle_count++;
 
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 556496c77dc2..46f8a01ff405 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -636,6 +636,8 @@ struct cfs_rq {
 	u64			throttled_clock;
 	u64			throttled_clock_pelt;
 	u64			throttled_clock_pelt_time;
+	u64			throttled_clock_self;
+	u64			throttled_clock_self_time;
 	int			throttled;
 	int			throttle_count;
 	struct list_head	throttled_list;
-- 
2.41.0.162.gfafddb0af9-goog

Re: [PATCH v3 2/2] sched: add throttled time stat for throttled children

[Tejun Heo]

On Tue, Jun 20, 2023 at 11:32:47AM -0700, Josh Don wrote:
> We currently export the total throttled time for cgroups that are given
> a bandwidth limit. This patch extends this accounting to also account
> the total time that each children cgroup has been throttled.
> 
> This is useful to understand the degree to which children have been
> affected by the throttling control. Children which are not runnable
> during the entire throttled period, for example, will not show any
> self-throttling time during this period.
> 
> Expose this in a new interface, 'cpu.stat.local', which is similar to
> how non-hierarchical events are accounted in 'memory.events.local'.
> 
> Signed-off-by: Josh Don <joshdon-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>

Acked-by: Tejun Heo <tj-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>

Thanks.

-- 
tejun