[PATCH 3/3] sched/fair: add tracepoints for cfs throttle

[Cong Wang]

This patch introduces two sched tracepoints to track
CFS throttle and unthrottle. The use case is to measure
how long each throttle is and to track when a CPU cgroup
is throttled and unthrottled.

Sample output:

     cfs-722   [000] dN.3    51.477702: sched_cfs_throttle: path=/test cpu=0 
runtime_remaining=0
  <idle>-0     [000] d.h4    51.536659: sched_cfs_unthrottle: path=/test cpu=0 
runtime_remaining=1

Cc: Paul Turner <p...@google.com>
Cc: Peter Zijlstra <pet...@infradead.org>
Cc: Mike Galbraith <efa...@gmx.de>
Cc: Thomas Gleixner <t...@linutronix.de>
Cc: Ingo Molnar <mi...@kernel.org>
Signed-off-by: Cong Wang <xiyou.wangc...@gmail.com>
---
 include/trace/events/sched.h | 42 ++++++++++++++++++++++++++++++++++++++++++
 kernel/sched/fair.c          |  2 ++
 2 files changed, 44 insertions(+)

diff --git a/include/trace/events/sched.h b/include/trace/events/sched.h
index bc01e06bc716..3d9e00972e92 100644
--- a/include/trace/events/sched.h
+++ b/include/trace/events/sched.h
@@ -570,6 +570,48 @@ TRACE_EVENT(sched_wake_idle_without_ipi,
 
        TP_printk("cpu=%d", __entry->cpu)
 );
+
+#ifdef CONFIG_CFS_BANDWIDTH
+DECLARE_EVENT_CLASS(sched_fair,
+
+       TP_PROTO(struct cfs_rq *cfs_rq),
+
+       TP_ARGS(cfs_rq),
+
+       TP_STRUCT__entry(
+               __field(        s64,            runtime_remaining       )
+               __field(        int,            cpu                     )
+               __dynamic_array(char,           cfs_path,
+                               cgroup_path(cfs_rq->tg->css.cgroup, NULL, 0) + 
1)
+       ),
+
+       TP_fast_assign(
+               __entry->runtime_remaining = cfs_rq->runtime_remaining;
+               __entry->cpu = cpu_of(cfs_rq->rq);
+               cgroup_path(cfs_rq->tg->css.cgroup,
+                           __get_dynamic_array(cfs_path),
+                           __get_dynamic_array_len(cfs_path));
+       ),
+
+       TP_printk("path=%s cpu=%d runtime_remaining=%lld", __get_str(cfs_path),
+                 __entry->cpu, __entry->runtime_remaining)
+);
+
+DEFINE_EVENT(sched_fair, sched_cfs_throttle,
+
+       TP_PROTO(struct cfs_rq *cfs_rq),
+
+       TP_ARGS(cfs_rq)
+);
+
+DEFINE_EVENT(sched_fair, sched_cfs_unthrottle,
+
+       TP_PROTO(struct cfs_rq *cfs_rq),
+
+       TP_ARGS(cfs_rq)
+);
+#endif /* CONFIG_CFS_BANDWIDTH */
+
 #endif /* _TRACE_SCHED_H */
 
 /* This part must be outside protection */
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 8a13ee006f39..3bcc40f2f272 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4736,6 +4736,7 @@ static void throttle_cfs_rq(struct cfs_rq *cfs_rq)
 
        cfs_rq->throttled = 1;
        cfs_rq->throttled_clock = rq_clock(rq);
+       trace_sched_cfs_throttle(cfs_rq);
        raw_spin_lock(&cfs_b->lock);
        empty = list_empty(&cfs_b->throttled_cfs_rq);
 
@@ -4766,6 +4767,7 @@ void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
        se = cfs_rq->tg->se[cpu_of(rq)];
 
        cfs_rq->throttled = 0;
+       trace_sched_cfs_unthrottle(cfs_rq);
 
        update_rq_clock(rq);
 
-- 
2.13.0