linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH v4 0/8] sched: support schedstats for RT sched class
@ 2021-09-05 14:35 Yafang Shao
  2021-09-05 14:35 ` [PATCH v4 1/8] sched, fair: use __schedstat_set() in set_next_entity() Yafang Shao
                   ` (8 more replies)
  0 siblings, 9 replies; 29+ messages in thread
From: Yafang Shao @ 2021-09-05 14:35 UTC (permalink / raw)
  To: peterz, mingo, mgorman, juri.lelli, vincent.guittot,
	dietmar.eggemann, rostedt, bsegall, bristot
  Cc: linux-kernel, linux-rt-users, achaiken, Yafang Shao

Why do we need schedstats ?
==========================

schedstats is a useful feature to do thread-level latency analysis. Our
usecase as follows,

  Userspace Code Scope         Profiler

  {
      user_func_abc(); <----   uprobe_scope_begin() get start schedstats 
      ...
      user_func_xyz(); <----   uprobe_scope_end() get end schedstats
  }

Then with the result of (end - begin) we can get below latency details
in a specific user scope,

  scope_latency = Wait + Sleep + Blocked [1]  + Run (stime + utime)

If there's no schedstats we have to trace the heavy sched::sched_switch
and do a lot more stuff. 

[1]. With patch #4 and don't include sum_block_runtime in sum_sleep_runtime

Support schedstats for RT sched class
=====================================

If we want to use the schedstats facility to trace other sched classes, we
should make it independent of fair sched class. The struct sched_statistics
is the schedular statistics of a task_struct or a task_group. So we can
move it into struct task_struct and struct task_group to achieve the goal.

After the patch, schestats are orgnized as follows,

    struct task_struct {
       ...
       struct sched_entity se;
       struct sched_rt_entity rt;
       struct sched_dl_entity dl;
       ...
       struct sched_statistics stats;
       ...
   };

Regarding the task group, schedstats is only supported for fair group
sched, and a new struct sched_entity_stats is introduced, suggested by
Peter -

    struct sched_entity_stats {
        struct sched_entity     se;
        struct sched_statistics stats;
    } __no_randomize_layout;

Then with the se in a task_group, we can easily get the stats.

The sched_statistics members may be frequently modified when schedstats is
enabled, in order to avoid impacting on random data which may in the same
cacheline with them, the struct sched_statistics is defined as cacheline
aligned.

As this patch changes the core struct of scheduler, so I verified the
performance it may impact on the scheduler with 'perf bench sched
pipe', suggested by Mel. Below is the result, in which all the values
are in usecs/op.
                                  Before               After
      kernel.sched_schedstats=0  5.2~5.4               5.2~5.4
      kernel.sched_schedstats=1  5.3~5.5               5.3~5.5
[These data is a little difference with the earlier version, that is
 because my old test machine is destroyed so I have to use a new
 different test machine.]

Almost no impact on the sched performance.

The user can get the schedstats information in the same way in fair sched
class. For example,
       fair                            RT
       /proc/[pid]/sched               /proc/[pid]/sched

schedstats is not supported for RT group.

The sched:sched_stat_{wait, sleep, iowait, blocked} tracepoints can
be used to trace RT tasks as well.

Support schedstats for any other sched classes
==============================================

After this patchset, it is very easy to extend the schedstats to any
other sched classes. The deadline sched class is also supported in this
patchset.

Changes Since v3:
Various code improvement per Peter,
- don't support schedstats for rt group
- introduce struct sched_entity_stats for fair group 
- change the position of 'struct sched_statistics stats'
- fixes indent issue
- change the output format in /proc/[pid]/sched
- add the usecase of schedstats
- support schedstats for deadline task
- and other suggestions

Changes Since v2:
- Fixes the output format in /proc/[pid]/sched 
- Rebase it on the latest code
- Redo the performance test

Changes since v1:
- Fix the build failure reported by kernel test robot.
- Add the performance data with 'perf bench sched pipe', suggested by
  Mel.
- Make the struct sched_statistics cacheline aligned.
- Introduce task block time in schedstats

Changes since RFC:
- improvement of schedstats helpers, per Mel.
- make struct schedstats independent of fair sched class

Yafang Shao (8):
  sched, fair: use __schedstat_set() in set_next_entity()
  sched: make struct sched_statistics independent of fair sched class
  sched: make schedstats helpers independent of fair sched class
  sched: introduce task block time in schedstats
  sched, rt: support sched_stat_runtime tracepoint for RT sched class
  sched, rt: support schedstats for RT sched class
  sched, dl: support sched_stat_runtime tracepoint for deadline sched
    class
  sched, dl: support schedstats for deadline sched class

 include/linux/sched.h    |   8 +-
 kernel/sched/core.c      |  25 +++---
 kernel/sched/deadline.c  |  99 +++++++++++++++++++++-
 kernel/sched/debug.c     |  97 +++++++++++----------
 kernel/sched/fair.c      | 177 +++++++++++----------------------------
 kernel/sched/rt.c        | 130 +++++++++++++++++++++++++++-
 kernel/sched/stats.c     | 104 +++++++++++++++++++++++
 kernel/sched/stats.h     |  49 +++++++++++
 kernel/sched/stop_task.c |   4 +-
 9 files changed, 500 insertions(+), 193 deletions(-)

-- 
2.18.2


^ permalink raw reply	[flat|nested] 29+ messages in thread

* [PATCH v4 1/8] sched, fair: use __schedstat_set() in set_next_entity()
  2021-09-05 14:35 [PATCH v4 0/8] sched: support schedstats for RT sched class Yafang Shao
@ 2021-09-05 14:35 ` Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched/fair: Use " tip-bot2 for Yafang Shao
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  2021-09-05 14:35 ` [PATCH v4 2/8] sched: make struct sched_statistics independent of fair sched class Yafang Shao
                   ` (7 subsequent siblings)
  8 siblings, 2 replies; 29+ messages in thread
From: Yafang Shao @ 2021-09-05 14:35 UTC (permalink / raw)
  To: peterz, mingo, mgorman, juri.lelli, vincent.guittot,
	dietmar.eggemann, rostedt, bsegall, bristot
  Cc: linux-kernel, linux-rt-users, achaiken, Yafang Shao

schedstat_enabled() has been already checked, so we can use
__schedstat_set() directly.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Alison Chaiken <achaiken@aurora.tech>
---
 kernel/sched/fair.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 25f056d87587..1b92eec48745 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4502,7 +4502,7 @@ set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	 */
 	if (schedstat_enabled() &&
 	    rq_of(cfs_rq)->cfs.load.weight >= 2*se->load.weight) {
-		schedstat_set(se->statistics.slice_max,
+		__schedstat_set(se->statistics.slice_max,
 			max((u64)schedstat_val(se->statistics.slice_max),
 			    se->sum_exec_runtime - se->prev_sum_exec_runtime));
 	}
-- 
2.18.2


^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [PATCH v4 2/8] sched: make struct sched_statistics independent of fair sched class
  2021-09-05 14:35 [PATCH v4 0/8] sched: support schedstats for RT sched class Yafang Shao
  2021-09-05 14:35 ` [PATCH v4 1/8] sched, fair: use __schedstat_set() in set_next_entity() Yafang Shao
@ 2021-09-05 14:35 ` Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched: Make " tip-bot2 for Yafang Shao
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  2021-09-05 14:35 ` [PATCH v4 3/8] sched: make schedstats helpers " Yafang Shao
                   ` (6 subsequent siblings)
  8 siblings, 2 replies; 29+ messages in thread
From: Yafang Shao @ 2021-09-05 14:35 UTC (permalink / raw)
  To: peterz, mingo, mgorman, juri.lelli, vincent.guittot,
	dietmar.eggemann, rostedt, bsegall, bristot
  Cc: linux-kernel, linux-rt-users, achaiken, Yafang Shao, kernel test robot

If we want to use the schedstats facility to trace other sched classes, we
should make it independent of fair sched class. The struct sched_statistics
is the schedular statistics of a task_struct or a task_group. So we can
move it into struct task_struct and struct task_group to achieve the goal.

After the patch, schestats are orgnized as follows,

    struct task_struct {
       ...
       struct sched_entity se;
       struct sched_rt_entity rt;
       struct sched_dl_entity dl;
       ...
       struct sched_statistics stats;
       ...
   };

Regarding the task group, schedstats is only supported for fair group
sched, and a new struct sched_entity_stats is introduced, suggested by
Peter -

    struct sched_entity_stats {
        struct sched_entity     se;
        struct sched_statistics stats;
    } __no_randomize_layout;

Then with the se in a task_group, we can easily get the stats.

The sched_statistics members may be frequently modified when schedstats is
enabled, in order to avoid impacting on random data which may in the same
cacheline with them, the struct sched_statistics is defined as cacheline
aligned.

As this patch changes the core struct of scheduler, so I verified the
performance it may impact on the scheduler with 'perf bench sched
pipe', suggested by Mel. Below is the result, in which all the values
are in usecs/op.
                                  Before               After
      kernel.sched_schedstats=0  5.2~5.4               5.2~5.4
      kernel.sched_schedstats=1  5.3~5.5               5.3~5.5
[These data is a little difference with the earlier version, that is
 because my old test machine is destroyed so I have to use a new
 different test machine.]

Almost no impact on the sched performance.

No functional change.

[lkp@intel.com: reported build failure in earlier version]

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: kernel test robot <lkp@intel.com>
Cc: Alison Chaiken <achaiken@aurora.tech>
---
 include/linux/sched.h    |  6 +--
 kernel/sched/core.c      | 25 ++++++-----
 kernel/sched/deadline.c  |  4 +-
 kernel/sched/debug.c     | 91 +++++++++++++++++++++-------------------
 kernel/sched/fair.c      | 87 ++++++++++++++++++++++----------------
 kernel/sched/rt.c        |  4 +-
 kernel/sched/stats.h     | 19 +++++++++
 kernel/sched/stop_task.c |  4 +-
 8 files changed, 140 insertions(+), 100 deletions(-)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 1780260f237b..ed4293812ca9 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -521,7 +521,7 @@ struct sched_statistics {
 	u64				nr_wakeups_passive;
 	u64				nr_wakeups_idle;
 #endif
-};
+} ____cacheline_aligned;
 
 struct sched_entity {
 	/* For load-balancing: */
@@ -537,8 +537,6 @@ struct sched_entity {
 
 	u64				nr_migrations;
 
-	struct sched_statistics		statistics;
-
 #ifdef CONFIG_FAIR_GROUP_SCHED
 	int				depth;
 	struct sched_entity		*parent;
@@ -802,6 +800,8 @@ struct task_struct {
 	struct uclamp_se		uclamp[UCLAMP_CNT];
 #endif
 
+	struct sched_statistics         stats;
+
 #ifdef CONFIG_PREEMPT_NOTIFIERS
 	/* List of struct preempt_notifier: */
 	struct hlist_head		preempt_notifiers;
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index f963c8113375..04c04504249a 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -3489,11 +3489,11 @@ ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
 #ifdef CONFIG_SMP
 	if (cpu == rq->cpu) {
 		__schedstat_inc(rq->ttwu_local);
-		__schedstat_inc(p->se.statistics.nr_wakeups_local);
+		__schedstat_inc(p->stats.nr_wakeups_local);
 	} else {
 		struct sched_domain *sd;
 
-		__schedstat_inc(p->se.statistics.nr_wakeups_remote);
+		__schedstat_inc(p->stats.nr_wakeups_remote);
 		rcu_read_lock();
 		for_each_domain(rq->cpu, sd) {
 			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
@@ -3505,14 +3505,14 @@ ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
 	}
 
 	if (wake_flags & WF_MIGRATED)
-		__schedstat_inc(p->se.statistics.nr_wakeups_migrate);
+		__schedstat_inc(p->stats.nr_wakeups_migrate);
 #endif /* CONFIG_SMP */
 
 	__schedstat_inc(rq->ttwu_count);
-	__schedstat_inc(p->se.statistics.nr_wakeups);
+	__schedstat_inc(p->stats.nr_wakeups);
 
 	if (wake_flags & WF_SYNC)
-		__schedstat_inc(p->se.statistics.nr_wakeups_sync);
+		__schedstat_inc(p->stats.nr_wakeups_sync);
 }
 
 /*
@@ -4196,7 +4196,7 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
 
 #ifdef CONFIG_SCHEDSTATS
 	/* Even if schedstat is disabled, there should not be garbage */
-	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
+	memset(&p->stats, 0, sizeof(p->stats));
 #endif
 
 	RB_CLEAR_NODE(&p->dl.rb_node);
@@ -9553,9 +9553,9 @@ void normalize_rt_tasks(void)
 			continue;
 
 		p->se.exec_start = 0;
-		schedstat_set(p->se.statistics.wait_start,  0);
-		schedstat_set(p->se.statistics.sleep_start, 0);
-		schedstat_set(p->se.statistics.block_start, 0);
+		schedstat_set(p->stats.wait_start,  0);
+		schedstat_set(p->stats.sleep_start, 0);
+		schedstat_set(p->stats.block_start, 0);
 
 		if (!dl_task(p) && !rt_task(p)) {
 			/*
@@ -10397,11 +10397,14 @@ 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) {
+		struct sched_statistics *stats;
 		u64 ws = 0;
 		int i;
 
-		for_each_possible_cpu(i)
-			ws += schedstat_val(tg->se[i]->statistics.wait_sum);
+		for_each_possible_cpu(i) {
+			stats = __schedstats_from_se(tg->se[i]);
+			ws += schedstat_val(stats->wait_sum);
+		}
 
 		seq_printf(sf, "wait_sum %llu\n", ws);
 	}
diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index e94314633b39..51dd30990042 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -1265,8 +1265,8 @@ static void update_curr_dl(struct rq *rq)
 		return;
 	}
 
-	schedstat_set(curr->se.statistics.exec_max,
-		      max(curr->se.statistics.exec_max, delta_exec));
+	schedstat_set(curr->stats.exec_max,
+		      max(curr->stats.exec_max, delta_exec));
 
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);
diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index 49716228efb4..da923347c8f3 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -440,11 +440,14 @@ void dirty_sched_domain_sysctl(int cpu)
 static void print_cfs_group_stats(struct seq_file *m, int cpu, struct task_group *tg)
 {
 	struct sched_entity *se = tg->se[cpu];
+	struct sched_statistics *stats = __schedstats_from_se(se);
 
 #define P(F)		SEQ_printf(m, "  .%-30s: %lld\n",	#F, (long long)F)
-#define P_SCHEDSTAT(F)	SEQ_printf(m, "  .%-30s: %lld\n",	#F, (long long)schedstat_val(F))
+#define P_SCHEDSTAT(F)	SEQ_printf(m, "  .%-30s: %lld\n",	\
+		#F, (long long)schedstat_val(stats->F))
 #define PN(F)		SEQ_printf(m, "  .%-30s: %lld.%06ld\n", #F, SPLIT_NS((long long)F))
-#define PN_SCHEDSTAT(F)	SEQ_printf(m, "  .%-30s: %lld.%06ld\n", #F, SPLIT_NS((long long)schedstat_val(F)))
+#define PN_SCHEDSTAT(F)	SEQ_printf(m, "  .%-30s: %lld.%06ld\n", \
+		#F, SPLIT_NS((long long)schedstat_val(stats->F)))
 
 	if (!se)
 		return;
@@ -454,16 +457,16 @@ static void print_cfs_group_stats(struct seq_file *m, int cpu, struct task_group
 	PN(se->sum_exec_runtime);
 
 	if (schedstat_enabled()) {
-		PN_SCHEDSTAT(se->statistics.wait_start);
-		PN_SCHEDSTAT(se->statistics.sleep_start);
-		PN_SCHEDSTAT(se->statistics.block_start);
-		PN_SCHEDSTAT(se->statistics.sleep_max);
-		PN_SCHEDSTAT(se->statistics.block_max);
-		PN_SCHEDSTAT(se->statistics.exec_max);
-		PN_SCHEDSTAT(se->statistics.slice_max);
-		PN_SCHEDSTAT(se->statistics.wait_max);
-		PN_SCHEDSTAT(se->statistics.wait_sum);
-		P_SCHEDSTAT(se->statistics.wait_count);
+		PN_SCHEDSTAT(wait_start);
+		PN_SCHEDSTAT(sleep_start);
+		PN_SCHEDSTAT(block_start);
+		PN_SCHEDSTAT(sleep_max);
+		PN_SCHEDSTAT(block_max);
+		PN_SCHEDSTAT(exec_max);
+		PN_SCHEDSTAT(slice_max);
+		PN_SCHEDSTAT(wait_max);
+		PN_SCHEDSTAT(wait_sum);
+		P_SCHEDSTAT(wait_count);
 	}
 
 	P(se->load.weight);
@@ -530,9 +533,9 @@ print_task(struct seq_file *m, struct rq *rq, struct task_struct *p)
 		p->prio);
 
 	SEQ_printf(m, "%9Ld.%06ld %9Ld.%06ld %9Ld.%06ld",
-		SPLIT_NS(schedstat_val_or_zero(p->se.statistics.wait_sum)),
+		SPLIT_NS(schedstat_val_or_zero(p->stats.wait_sum)),
 		SPLIT_NS(p->se.sum_exec_runtime),
-		SPLIT_NS(schedstat_val_or_zero(p->se.statistics.sum_sleep_runtime)));
+		SPLIT_NS(schedstat_val_or_zero(p->stats.sum_sleep_runtime)));
 
 #ifdef CONFIG_NUMA_BALANCING
 	SEQ_printf(m, " %d %d", task_node(p), task_numa_group_id(p));
@@ -948,8 +951,8 @@ void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
 		"---------------------------------------------------------"
 		"----------\n");
 
-#define P_SCHEDSTAT(F)  __PS(#F, schedstat_val(p->F))
-#define PN_SCHEDSTAT(F) __PSN(#F, schedstat_val(p->F))
+#define P_SCHEDSTAT(F)  __PS(#F, schedstat_val(p->stats.F))
+#define PN_SCHEDSTAT(F) __PSN(#F, schedstat_val(p->stats.F))
 
 	PN(se.exec_start);
 	PN(se.vruntime);
@@ -962,33 +965,33 @@ void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
 	if (schedstat_enabled()) {
 		u64 avg_atom, avg_per_cpu;
 
-		PN_SCHEDSTAT(se.statistics.sum_sleep_runtime);
-		PN_SCHEDSTAT(se.statistics.wait_start);
-		PN_SCHEDSTAT(se.statistics.sleep_start);
-		PN_SCHEDSTAT(se.statistics.block_start);
-		PN_SCHEDSTAT(se.statistics.sleep_max);
-		PN_SCHEDSTAT(se.statistics.block_max);
-		PN_SCHEDSTAT(se.statistics.exec_max);
-		PN_SCHEDSTAT(se.statistics.slice_max);
-		PN_SCHEDSTAT(se.statistics.wait_max);
-		PN_SCHEDSTAT(se.statistics.wait_sum);
-		P_SCHEDSTAT(se.statistics.wait_count);
-		PN_SCHEDSTAT(se.statistics.iowait_sum);
-		P_SCHEDSTAT(se.statistics.iowait_count);
-		P_SCHEDSTAT(se.statistics.nr_migrations_cold);
-		P_SCHEDSTAT(se.statistics.nr_failed_migrations_affine);
-		P_SCHEDSTAT(se.statistics.nr_failed_migrations_running);
-		P_SCHEDSTAT(se.statistics.nr_failed_migrations_hot);
-		P_SCHEDSTAT(se.statistics.nr_forced_migrations);
-		P_SCHEDSTAT(se.statistics.nr_wakeups);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_sync);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_migrate);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_local);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_remote);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_affine);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_affine_attempts);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_passive);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_idle);
+		PN_SCHEDSTAT(sum_sleep_runtime);
+		PN_SCHEDSTAT(wait_start);
+		PN_SCHEDSTAT(sleep_start);
+		PN_SCHEDSTAT(block_start);
+		PN_SCHEDSTAT(sleep_max);
+		PN_SCHEDSTAT(block_max);
+		PN_SCHEDSTAT(exec_max);
+		PN_SCHEDSTAT(slice_max);
+		PN_SCHEDSTAT(wait_max);
+		PN_SCHEDSTAT(wait_sum);
+		P_SCHEDSTAT(wait_count);
+		PN_SCHEDSTAT(iowait_sum);
+		P_SCHEDSTAT(iowait_count);
+		P_SCHEDSTAT(nr_migrations_cold);
+		P_SCHEDSTAT(nr_failed_migrations_affine);
+		P_SCHEDSTAT(nr_failed_migrations_running);
+		P_SCHEDSTAT(nr_failed_migrations_hot);
+		P_SCHEDSTAT(nr_forced_migrations);
+		P_SCHEDSTAT(nr_wakeups);
+		P_SCHEDSTAT(nr_wakeups_sync);
+		P_SCHEDSTAT(nr_wakeups_migrate);
+		P_SCHEDSTAT(nr_wakeups_local);
+		P_SCHEDSTAT(nr_wakeups_remote);
+		P_SCHEDSTAT(nr_wakeups_affine);
+		P_SCHEDSTAT(nr_wakeups_affine_attempts);
+		P_SCHEDSTAT(nr_wakeups_passive);
+		P_SCHEDSTAT(nr_wakeups_idle);
 
 		avg_atom = p->se.sum_exec_runtime;
 		if (nr_switches)
@@ -1054,7 +1057,7 @@ void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
 void proc_sched_set_task(struct task_struct *p)
 {
 #ifdef CONFIG_SCHEDSTATS
-	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
+	memset(&p->stats, 0, sizeof(p->stats));
 #endif
 }
 
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 1b92eec48745..8b4b97453cca 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -837,8 +837,12 @@ static void update_curr(struct cfs_rq *cfs_rq)
 
 	curr->exec_start = now;
 
-	schedstat_set(curr->statistics.exec_max,
-		      max(delta_exec, curr->statistics.exec_max));
+	if (schedstat_enabled()) {
+		struct sched_statistics *stats = __schedstats_from_se(curr);
+
+		__schedstat_set(stats->exec_max,
+				max(delta_exec, stats->exec_max));
+	}
 
 	curr->sum_exec_runtime += delta_exec;
 	schedstat_add(cfs_rq->exec_clock, delta_exec);
@@ -866,39 +870,45 @@ static inline void
 update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
 	u64 wait_start, prev_wait_start;
+	struct sched_statistics *stats;
 
 	if (!schedstat_enabled())
 		return;
 
+	stats = __schedstats_from_se(se);
+
 	wait_start = rq_clock(rq_of(cfs_rq));
-	prev_wait_start = schedstat_val(se->statistics.wait_start);
+	prev_wait_start = schedstat_val(stats->wait_start);
 
 	if (entity_is_task(se) && task_on_rq_migrating(task_of(se)) &&
 	    likely(wait_start > prev_wait_start))
 		wait_start -= prev_wait_start;
 
-	__schedstat_set(se->statistics.wait_start, wait_start);
+	__schedstat_set(stats->wait_start, wait_start);
 }
 
 static inline void
 update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
-	struct task_struct *p;
+	struct sched_statistics *stats;
+	struct task_struct *p = NULL;
 	u64 delta;
 
 	if (!schedstat_enabled())
 		return;
 
+	stats = __schedstats_from_se(se);
+
 	/*
 	 * When the sched_schedstat changes from 0 to 1, some sched se
 	 * maybe already in the runqueue, the se->statistics.wait_start
 	 * will be 0.So it will let the delta wrong. We need to avoid this
 	 * scenario.
 	 */
-	if (unlikely(!schedstat_val(se->statistics.wait_start)))
+	if (unlikely(!schedstat_val(stats->wait_start)))
 		return;
 
-	delta = rq_clock(rq_of(cfs_rq)) - schedstat_val(se->statistics.wait_start);
+	delta = rq_clock(rq_of(cfs_rq)) - schedstat_val(stats->wait_start);
 
 	if (entity_is_task(se)) {
 		p = task_of(se);
@@ -908,30 +918,33 @@ update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
 			 * time stamp can be adjusted to accumulate wait time
 			 * prior to migration.
 			 */
-			__schedstat_set(se->statistics.wait_start, delta);
+			__schedstat_set(stats->wait_start, delta);
 			return;
 		}
 		trace_sched_stat_wait(p, delta);
 	}
 
-	__schedstat_set(se->statistics.wait_max,
-		      max(schedstat_val(se->statistics.wait_max), delta));
-	__schedstat_inc(se->statistics.wait_count);
-	__schedstat_add(se->statistics.wait_sum, delta);
-	__schedstat_set(se->statistics.wait_start, 0);
+	__schedstat_set(stats->wait_max,
+		      max(schedstat_val(stats->wait_max), delta));
+	__schedstat_inc(stats->wait_count);
+	__schedstat_add(stats->wait_sum, delta);
+	__schedstat_set(stats->wait_start, 0);
 }
 
 static inline void
 update_stats_enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
+	struct sched_statistics *stats;
 	struct task_struct *tsk = NULL;
 	u64 sleep_start, block_start;
 
 	if (!schedstat_enabled())
 		return;
 
-	sleep_start = schedstat_val(se->statistics.sleep_start);
-	block_start = schedstat_val(se->statistics.block_start);
+	stats = __schedstats_from_se(se);
+
+	sleep_start = schedstat_val(stats->sleep_start);
+	block_start = schedstat_val(stats->block_start);
 
 	if (entity_is_task(se))
 		tsk = task_of(se);
@@ -942,11 +955,11 @@ update_stats_enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
 		if ((s64)delta < 0)
 			delta = 0;
 
-		if (unlikely(delta > schedstat_val(se->statistics.sleep_max)))
-			__schedstat_set(se->statistics.sleep_max, delta);
+		if (unlikely(delta > schedstat_val(stats->sleep_max)))
+			__schedstat_set(stats->sleep_max, delta);
 
-		__schedstat_set(se->statistics.sleep_start, 0);
-		__schedstat_add(se->statistics.sum_sleep_runtime, delta);
+		__schedstat_set(stats->sleep_start, 0);
+		__schedstat_add(stats->sum_sleep_runtime, delta);
 
 		if (tsk) {
 			account_scheduler_latency(tsk, delta >> 10, 1);
@@ -959,16 +972,16 @@ update_stats_enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
 		if ((s64)delta < 0)
 			delta = 0;
 
-		if (unlikely(delta > schedstat_val(se->statistics.block_max)))
-			__schedstat_set(se->statistics.block_max, delta);
+		if (unlikely(delta > schedstat_val(stats->block_max)))
+			__schedstat_set(stats->block_max, delta);
 
-		__schedstat_set(se->statistics.block_start, 0);
-		__schedstat_add(se->statistics.sum_sleep_runtime, delta);
+		__schedstat_set(stats->block_start, 0);
+		__schedstat_add(stats->sum_sleep_runtime, delta);
 
 		if (tsk) {
 			if (tsk->in_iowait) {
-				__schedstat_add(se->statistics.iowait_sum, delta);
-				__schedstat_inc(se->statistics.iowait_count);
+				__schedstat_add(stats->iowait_sum, delta);
+				__schedstat_inc(stats->iowait_count);
 				trace_sched_stat_iowait(tsk, delta);
 			}
 
@@ -1030,10 +1043,10 @@ update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 		/* XXX racy against TTWU */
 		state = READ_ONCE(tsk->__state);
 		if (state & TASK_INTERRUPTIBLE)
-			__schedstat_set(se->statistics.sleep_start,
+			__schedstat_set(tsk->stats.sleep_start,
 				      rq_clock(rq_of(cfs_rq)));
 		if (state & TASK_UNINTERRUPTIBLE)
-			__schedstat_set(se->statistics.block_start,
+			__schedstat_set(tsk->stats.block_start,
 				      rq_clock(rq_of(cfs_rq)));
 	}
 }
@@ -4502,8 +4515,10 @@ set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	 */
 	if (schedstat_enabled() &&
 	    rq_of(cfs_rq)->cfs.load.weight >= 2*se->load.weight) {
-		__schedstat_set(se->statistics.slice_max,
-			max((u64)schedstat_val(se->statistics.slice_max),
+		struct sched_statistics *stats = __schedstats_from_se(se);
+
+		__schedstat_set(stats->slice_max,
+			max((u64)schedstat_val(stats->slice_max),
 			    se->sum_exec_runtime - se->prev_sum_exec_runtime));
 	}
 
@@ -5994,12 +6009,12 @@ static int wake_affine(struct sched_domain *sd, struct task_struct *p,
 	if (sched_feat(WA_WEIGHT) && target == nr_cpumask_bits)
 		target = wake_affine_weight(sd, p, this_cpu, prev_cpu, sync);
 
-	schedstat_inc(p->se.statistics.nr_wakeups_affine_attempts);
+	schedstat_inc(p->stats.nr_wakeups_affine_attempts);
 	if (target == nr_cpumask_bits)
 		return prev_cpu;
 
 	schedstat_inc(sd->ttwu_move_affine);
-	schedstat_inc(p->se.statistics.nr_wakeups_affine);
+	schedstat_inc(p->stats.nr_wakeups_affine);
 	return target;
 }
 
@@ -7803,7 +7818,7 @@ int can_migrate_task(struct task_struct *p, struct lb_env *env)
 	if (!cpumask_test_cpu(env->dst_cpu, p->cpus_ptr)) {
 		int cpu;
 
-		schedstat_inc(p->se.statistics.nr_failed_migrations_affine);
+		schedstat_inc(p->stats.nr_failed_migrations_affine);
 
 		env->flags |= LBF_SOME_PINNED;
 
@@ -7837,7 +7852,7 @@ int can_migrate_task(struct task_struct *p, struct lb_env *env)
 	env->flags &= ~LBF_ALL_PINNED;
 
 	if (task_running(env->src_rq, p)) {
-		schedstat_inc(p->se.statistics.nr_failed_migrations_running);
+		schedstat_inc(p->stats.nr_failed_migrations_running);
 		return 0;
 	}
 
@@ -7859,12 +7874,12 @@ int can_migrate_task(struct task_struct *p, struct lb_env *env)
 	    env->sd->nr_balance_failed > env->sd->cache_nice_tries) {
 		if (tsk_cache_hot == 1) {
 			schedstat_inc(env->sd->lb_hot_gained[env->idle]);
-			schedstat_inc(p->se.statistics.nr_forced_migrations);
+			schedstat_inc(p->stats.nr_forced_migrations);
 		}
 		return 1;
 	}
 
-	schedstat_inc(p->se.statistics.nr_failed_migrations_hot);
+	schedstat_inc(p->stats.nr_failed_migrations_hot);
 	return 0;
 }
 
@@ -11506,7 +11521,7 @@ int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
 		if (!cfs_rq)
 			goto err;
 
-		se = kzalloc_node(sizeof(struct sched_entity),
+		se = kzalloc_node(sizeof(struct sched_entity_stats),
 				  GFP_KERNEL, cpu_to_node(i));
 		if (!se)
 			goto err_free_rq;
diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 3daf42a0f462..95a7c3ad2dc3 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1009,8 +1009,8 @@ static void update_curr_rt(struct rq *rq)
 	if (unlikely((s64)delta_exec <= 0))
 		return;
 
-	schedstat_set(curr->se.statistics.exec_max,
-		      max(curr->se.statistics.exec_max, delta_exec));
+	schedstat_set(curr->stats.exec_max,
+		      max(curr->stats.exec_max, delta_exec));
 
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);
diff --git a/kernel/sched/stats.h b/kernel/sched/stats.h
index d8f8eb0c655b..fb6022e860af 100644
--- a/kernel/sched/stats.h
+++ b/kernel/sched/stats.h
@@ -41,6 +41,7 @@ rq_sched_info_dequeue(struct rq *rq, unsigned long long delta)
 #define   schedstat_val_or_zero(var)	((schedstat_enabled()) ? (var) : 0)
 
 #else /* !CONFIG_SCHEDSTATS: */
+
 static inline void rq_sched_info_arrive  (struct rq *rq, unsigned long long delta) { }
 static inline void rq_sched_info_dequeue(struct rq *rq, unsigned long long delta) { }
 static inline void rq_sched_info_depart  (struct rq *rq, unsigned long long delta) { }
@@ -53,8 +54,26 @@ static inline void rq_sched_info_depart  (struct rq *rq, unsigned long long delt
 # define   schedstat_set(var, val)	do { } while (0)
 # define   schedstat_val(var)		0
 # define   schedstat_val_or_zero(var)	0
+
 #endif /* CONFIG_SCHEDSTATS */
 
+#ifdef CONFIG_FAIR_GROUP_SCHED
+struct sched_entity_stats {
+	struct sched_entity     se;
+	struct sched_statistics stats;
+} __no_randomize_layout;
+#endif
+
+static inline struct sched_statistics *
+__schedstats_from_se(struct sched_entity *se)
+{
+#ifdef CONFIG_FAIR_GROUP_SCHED
+	if (!entity_is_task(se))
+		return &container_of(se, struct sched_entity_stats, se)->stats;
+#endif
+	return &task_of(se)->stats;
+}
+
 #ifdef CONFIG_PSI
 /*
  * PSI tracks state that persists across sleeps, such as iowaits and
diff --git a/kernel/sched/stop_task.c b/kernel/sched/stop_task.c
index f988ebe3febb..0b165a25f22f 100644
--- a/kernel/sched/stop_task.c
+++ b/kernel/sched/stop_task.c
@@ -78,8 +78,8 @@ static void put_prev_task_stop(struct rq *rq, struct task_struct *prev)
 	if (unlikely((s64)delta_exec < 0))
 		delta_exec = 0;
 
-	schedstat_set(curr->se.statistics.exec_max,
-			max(curr->se.statistics.exec_max, delta_exec));
+	schedstat_set(curr->stats.exec_max,
+		      max(curr->stats.exec_max, delta_exec));
 
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);
-- 
2.18.2


^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [PATCH v4 3/8] sched: make schedstats helpers independent of fair sched class
  2021-09-05 14:35 [PATCH v4 0/8] sched: support schedstats for RT sched class Yafang Shao
  2021-09-05 14:35 ` [PATCH v4 1/8] sched, fair: use __schedstat_set() in set_next_entity() Yafang Shao
  2021-09-05 14:35 ` [PATCH v4 2/8] sched: make struct sched_statistics independent of fair sched class Yafang Shao
@ 2021-09-05 14:35 ` Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched: Make " tip-bot2 for Yafang Shao
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  2021-09-05 14:35 ` [PATCH v4 4/8] sched: introduce task block time in schedstats Yafang Shao
                   ` (5 subsequent siblings)
  8 siblings, 2 replies; 29+ messages in thread
From: Yafang Shao @ 2021-09-05 14:35 UTC (permalink / raw)
  To: peterz, mingo, mgorman, juri.lelli, vincent.guittot,
	dietmar.eggemann, rostedt, bsegall, bristot
  Cc: linux-kernel, linux-rt-users, achaiken, Yafang Shao, kernel test robot

The original prototype of the schedstats helpers are

  update_stats_wait_*(struct cfs_rq *cfs_rq, struct sched_entity *se)

The cfs_rq in these helpers is used to get the rq_clock, and the se is
used to get the struct sched_statistics and the struct task_struct. In
order to make these helpers available by all sched classes, we can pass
the rq, sched_statistics and task_struct directly.

Then the new helpers are

  update_stats_wait_*(struct rq *rq, struct task_struct *p,
                      struct sched_statistics *stats)

which are independent of fair sched class.

To avoid vmlinux growing too large or introducing ovehead when
!schedstat_enabled(), some new helpers after schedstat_enabled() are also
introduced, Suggested by Mel. These helpers are in sched/stats.c,

  __update_stats_wait_*(struct rq *rq, struct task_struct *p,
                        struct sched_statistics *stats)

The size of vmlinux as follows,
                      Before          After
  Size of vmlinux     826308552       826304640
The size is a litte smaller as some functions are not inlined again after
the change.

I also compared the sched performance with 'perf bench sched pipe',
suggested by Mel. The result as followsi (in usecs/op),
                             Before                After
  kernel.sched_schedstats=0  5.2~5.4               5.2~5.4
  kernel.sched_schedstats=1  5.3~5.5               5.3~5.5

[These data is a little difference with the prev version, that is
because my old test machine is destroyed so I have to use a new
different test machine.]
Almost no difference.

No functional change.

[lkp@intel.com: reported build failure in prev version]

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: kernel test robot <lkp@intel.com>
Cc: Alison Chaiken <achaiken@aurora.tech>
---
 kernel/sched/debug.c |   4 +-
 kernel/sched/fair.c  | 140 +++++++------------------------------------
 kernel/sched/stats.c | 103 +++++++++++++++++++++++++++++++
 kernel/sched/stats.h |  30 ++++++++++
 4 files changed, 159 insertions(+), 118 deletions(-)

diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index da923347c8f3..e08eee374176 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -439,8 +439,10 @@ void dirty_sched_domain_sysctl(int cpu)
 #ifdef CONFIG_FAIR_GROUP_SCHED
 static void print_cfs_group_stats(struct seq_file *m, int cpu, struct task_group *tg)
 {
+	struct sched_statistics __maybe_unused *stats;
 	struct sched_entity *se = tg->se[cpu];
-	struct sched_statistics *stats = __schedstats_from_se(se);
+
+	stats = __schedstats_from_se(se);
 
 #define P(F)		SEQ_printf(m, "  .%-30s: %lld\n",	#F, (long long)F)
 #define P_SCHEDSTAT(F)	SEQ_printf(m, "  .%-30s: %lld\n",	\
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 8b4b97453cca..219ad90a1762 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -838,8 +838,9 @@ static void update_curr(struct cfs_rq *cfs_rq)
 	curr->exec_start = now;
 
 	if (schedstat_enabled()) {
-		struct sched_statistics *stats = __schedstats_from_se(curr);
+		struct sched_statistics __maybe_unused *stats;
 
+		stats = __schedstats_from_se(curr);
 		__schedstat_set(stats->exec_max,
 				max(delta_exec, stats->exec_max));
 	}
@@ -867,32 +868,27 @@ static void update_curr_fair(struct rq *rq)
 }
 
 static inline void
-update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
+update_stats_wait_start_fair(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
-	u64 wait_start, prev_wait_start;
 	struct sched_statistics *stats;
+	struct task_struct *p = NULL;
 
 	if (!schedstat_enabled())
 		return;
 
 	stats = __schedstats_from_se(se);
 
-	wait_start = rq_clock(rq_of(cfs_rq));
-	prev_wait_start = schedstat_val(stats->wait_start);
-
-	if (entity_is_task(se) && task_on_rq_migrating(task_of(se)) &&
-	    likely(wait_start > prev_wait_start))
-		wait_start -= prev_wait_start;
+	if (entity_is_task(se))
+		p = task_of(se);
 
-	__schedstat_set(stats->wait_start, wait_start);
+	__update_stats_wait_start(rq_of(cfs_rq), p, stats);
 }
 
 static inline void
-update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
+update_stats_wait_end_fair(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
 	struct sched_statistics *stats;
 	struct task_struct *p = NULL;
-	u64 delta;
 
 	if (!schedstat_enabled())
 		return;
@@ -908,105 +904,34 @@ update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	if (unlikely(!schedstat_val(stats->wait_start)))
 		return;
 
-	delta = rq_clock(rq_of(cfs_rq)) - schedstat_val(stats->wait_start);
-
-	if (entity_is_task(se)) {
+	if (entity_is_task(se))
 		p = task_of(se);
-		if (task_on_rq_migrating(p)) {
-			/*
-			 * Preserve migrating task's wait time so wait_start
-			 * time stamp can be adjusted to accumulate wait time
-			 * prior to migration.
-			 */
-			__schedstat_set(stats->wait_start, delta);
-			return;
-		}
-		trace_sched_stat_wait(p, delta);
-	}
 
-	__schedstat_set(stats->wait_max,
-		      max(schedstat_val(stats->wait_max), delta));
-	__schedstat_inc(stats->wait_count);
-	__schedstat_add(stats->wait_sum, delta);
-	__schedstat_set(stats->wait_start, 0);
+	__update_stats_wait_end(rq_of(cfs_rq), p, stats);
 }
 
 static inline void
-update_stats_enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
+update_stats_enqueue_sleeper_fair(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
 	struct sched_statistics *stats;
 	struct task_struct *tsk = NULL;
-	u64 sleep_start, block_start;
 
 	if (!schedstat_enabled())
 		return;
 
 	stats = __schedstats_from_se(se);
 
-	sleep_start = schedstat_val(stats->sleep_start);
-	block_start = schedstat_val(stats->block_start);
-
 	if (entity_is_task(se))
 		tsk = task_of(se);
 
-	if (sleep_start) {
-		u64 delta = rq_clock(rq_of(cfs_rq)) - sleep_start;
-
-		if ((s64)delta < 0)
-			delta = 0;
-
-		if (unlikely(delta > schedstat_val(stats->sleep_max)))
-			__schedstat_set(stats->sleep_max, delta);
-
-		__schedstat_set(stats->sleep_start, 0);
-		__schedstat_add(stats->sum_sleep_runtime, delta);
-
-		if (tsk) {
-			account_scheduler_latency(tsk, delta >> 10, 1);
-			trace_sched_stat_sleep(tsk, delta);
-		}
-	}
-	if (block_start) {
-		u64 delta = rq_clock(rq_of(cfs_rq)) - block_start;
-
-		if ((s64)delta < 0)
-			delta = 0;
-
-		if (unlikely(delta > schedstat_val(stats->block_max)))
-			__schedstat_set(stats->block_max, delta);
-
-		__schedstat_set(stats->block_start, 0);
-		__schedstat_add(stats->sum_sleep_runtime, delta);
-
-		if (tsk) {
-			if (tsk->in_iowait) {
-				__schedstat_add(stats->iowait_sum, delta);
-				__schedstat_inc(stats->iowait_count);
-				trace_sched_stat_iowait(tsk, delta);
-			}
-
-			trace_sched_stat_blocked(tsk, delta);
-
-			/*
-			 * Blocking time is in units of nanosecs, so shift by
-			 * 20 to get a milliseconds-range estimation of the
-			 * amount of time that the task spent sleeping:
-			 */
-			if (unlikely(prof_on == SLEEP_PROFILING)) {
-				profile_hits(SLEEP_PROFILING,
-						(void *)get_wchan(tsk),
-						delta >> 20);
-			}
-			account_scheduler_latency(tsk, delta >> 10, 0);
-		}
-	}
+	__update_stats_enqueue_sleeper(rq_of(cfs_rq), tsk, stats);
 }
 
 /*
  * Task is being enqueued - update stats:
  */
 static inline void
-update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
+update_stats_enqueue_fair(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 {
 	if (!schedstat_enabled())
 		return;
@@ -1016,14 +941,14 @@ update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 	 * a dequeue/enqueue event is a NOP)
 	 */
 	if (se != cfs_rq->curr)
-		update_stats_wait_start(cfs_rq, se);
+		update_stats_wait_start_fair(cfs_rq, se);
 
 	if (flags & ENQUEUE_WAKEUP)
-		update_stats_enqueue_sleeper(cfs_rq, se);
+		update_stats_enqueue_sleeper_fair(cfs_rq, se);
 }
 
 static inline void
-update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
+update_stats_dequeue_fair(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 {
 
 	if (!schedstat_enabled())
@@ -1034,7 +959,7 @@ update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 	 * waiting task:
 	 */
 	if (se != cfs_rq->curr)
-		update_stats_wait_end(cfs_rq, se);
+		update_stats_wait_end_fair(cfs_rq, se);
 
 	if ((flags & DEQUEUE_SLEEP) && entity_is_task(se)) {
 		struct task_struct *tsk = task_of(se);
@@ -4238,26 +4163,6 @@ 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 void check_schedstat_required(void)
-{
-#ifdef CONFIG_SCHEDSTATS
-	if (schedstat_enabled())
-		return;
-
-	/* Force schedstat enabled if a dependent tracepoint is active */
-	if (trace_sched_stat_wait_enabled()    ||
-			trace_sched_stat_sleep_enabled()   ||
-			trace_sched_stat_iowait_enabled()  ||
-			trace_sched_stat_blocked_enabled() ||
-			trace_sched_stat_runtime_enabled())  {
-		printk_deferred_once("Scheduler tracepoints stat_sleep, stat_iowait, "
-			     "stat_blocked and stat_runtime require the "
-			     "kernel parameter schedstats=enable or "
-			     "kernel.sched_schedstats=1\n");
-	}
-#endif
-}
-
 static inline bool cfs_bandwidth_used(void);
 
 /*
@@ -4331,7 +4236,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 		place_entity(cfs_rq, se, 0);
 
 	check_schedstat_required();
-	update_stats_enqueue(cfs_rq, se, flags);
+	update_stats_enqueue_fair(cfs_rq, se, flags);
 	check_spread(cfs_rq, se);
 	if (!curr)
 		__enqueue_entity(cfs_rq, se);
@@ -4415,7 +4320,7 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 	update_load_avg(cfs_rq, se, UPDATE_TG);
 	se_update_runnable(se);
 
-	update_stats_dequeue(cfs_rq, se, flags);
+	update_stats_dequeue_fair(cfs_rq, se, flags);
 
 	clear_buddies(cfs_rq, se);
 
@@ -4500,7 +4405,7 @@ set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 		 * a CPU. So account for the time it spent waiting on the
 		 * runqueue.
 		 */
-		update_stats_wait_end(cfs_rq, se);
+		update_stats_wait_end_fair(cfs_rq, se);
 		__dequeue_entity(cfs_rq, se);
 		update_load_avg(cfs_rq, se, UPDATE_TG);
 	}
@@ -4515,8 +4420,9 @@ set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	 */
 	if (schedstat_enabled() &&
 	    rq_of(cfs_rq)->cfs.load.weight >= 2*se->load.weight) {
-		struct sched_statistics *stats = __schedstats_from_se(se);
+		struct sched_statistics __maybe_unused *stats;
 
+		stats = __schedstats_from_se(se);
 		__schedstat_set(stats->slice_max,
 			max((u64)schedstat_val(stats->slice_max),
 			    se->sum_exec_runtime - se->prev_sum_exec_runtime));
@@ -4601,7 +4507,7 @@ static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev)
 	check_spread(cfs_rq, prev);
 
 	if (prev->on_rq) {
-		update_stats_wait_start(cfs_rq, prev);
+		update_stats_wait_start_fair(cfs_rq, prev);
 		/* Put 'current' back into the tree. */
 		__enqueue_entity(cfs_rq, prev);
 		/* in !on_rq case, update occurred at dequeue */
diff --git a/kernel/sched/stats.c b/kernel/sched/stats.c
index 3f93fc3b5648..fad781ca7791 100644
--- a/kernel/sched/stats.c
+++ b/kernel/sched/stats.c
@@ -4,6 +4,109 @@
  */
 #include "sched.h"
 
+void __update_stats_wait_start(struct rq *rq, struct task_struct *p,
+			       struct sched_statistics *stats)
+{
+	u64 wait_start, prev_wait_start;
+
+	wait_start = rq_clock(rq);
+	prev_wait_start = schedstat_val(stats->wait_start);
+
+	if (p && likely(wait_start > prev_wait_start))
+		wait_start -= prev_wait_start;
+
+	__schedstat_set(stats->wait_start, wait_start);
+}
+
+void __update_stats_wait_end(struct rq *rq, struct task_struct *p,
+			     struct sched_statistics *stats)
+{
+	u64 delta = rq_clock(rq) - schedstat_val(stats->wait_start);
+
+	if (p) {
+		if (task_on_rq_migrating(p)) {
+			/*
+			 * Preserve migrating task's wait time so wait_start
+			 * time stamp can be adjusted to accumulate wait time
+			 * prior to migration.
+			 */
+			__schedstat_set(stats->wait_start, delta);
+
+			return;
+		}
+
+		trace_sched_stat_wait(p, delta);
+	}
+
+	__schedstat_set(stats->wait_max,
+			max(schedstat_val(stats->wait_max), delta));
+	__schedstat_inc(stats->wait_count);
+	__schedstat_add(stats->wait_sum, delta);
+	__schedstat_set(stats->wait_start, 0);
+}
+
+void __update_stats_enqueue_sleeper(struct rq *rq, struct task_struct *p,
+				    struct sched_statistics *stats)
+{
+	u64 sleep_start, block_start;
+
+	sleep_start = schedstat_val(stats->sleep_start);
+	block_start = schedstat_val(stats->block_start);
+
+	if (sleep_start) {
+		u64 delta = rq_clock(rq) - sleep_start;
+
+		if ((s64)delta < 0)
+			delta = 0;
+
+		if (unlikely(delta > schedstat_val(stats->sleep_max)))
+			__schedstat_set(stats->sleep_max, delta);
+
+		__schedstat_set(stats->sleep_start, 0);
+		__schedstat_add(stats->sum_sleep_runtime, delta);
+
+		if (p) {
+			account_scheduler_latency(p, delta >> 10, 1);
+			trace_sched_stat_sleep(p, delta);
+		}
+	}
+
+	if (block_start) {
+		u64 delta = rq_clock(rq) - block_start;
+
+		if ((s64)delta < 0)
+			delta = 0;
+
+		if (unlikely(delta > schedstat_val(stats->block_max)))
+			__schedstat_set(stats->block_max, delta);
+
+		__schedstat_set(stats->block_start, 0);
+		__schedstat_add(stats->sum_sleep_runtime, delta);
+
+		if (p) {
+			if (p->in_iowait) {
+				__schedstat_add(stats->iowait_sum, delta);
+				__schedstat_inc(stats->iowait_count);
+				trace_sched_stat_iowait(p, delta);
+			}
+
+			trace_sched_stat_blocked(p, delta);
+
+			/*
+			 * Blocking time is in units of nanosecs, so shift by
+			 * 20 to get a milliseconds-range estimation of the
+			 * amount of time that the task spent sleeping:
+			 */
+			if (unlikely(prof_on == SLEEP_PROFILING)) {
+				profile_hits(SLEEP_PROFILING,
+					     (void *)get_wchan(p),
+					     delta >> 20);
+			}
+			account_scheduler_latency(p, delta >> 10, 0);
+		}
+	}
+}
+
 /*
  * Current schedstat API version.
  *
diff --git a/kernel/sched/stats.h b/kernel/sched/stats.h
index fb6022e860af..cfb0893a83d4 100644
--- a/kernel/sched/stats.h
+++ b/kernel/sched/stats.h
@@ -2,6 +2,8 @@
 
 #ifdef CONFIG_SCHEDSTATS
 
+extern struct static_key_false sched_schedstats;
+
 /*
  * Expects runqueue lock to be held for atomicity of update
  */
@@ -40,6 +42,29 @@ rq_sched_info_dequeue(struct rq *rq, unsigned long long delta)
 #define   schedstat_val(var)		(var)
 #define   schedstat_val_or_zero(var)	((schedstat_enabled()) ? (var) : 0)
 
+void __update_stats_wait_start(struct rq *rq, struct task_struct *p,
+			       struct sched_statistics *stats);
+
+void __update_stats_wait_end(struct rq *rq, struct task_struct *p,
+			     struct sched_statistics *stats);
+void __update_stats_enqueue_sleeper(struct rq *rq, struct task_struct *p,
+				    struct sched_statistics *stats);
+
+static inline void
+check_schedstat_required(void)
+{
+	if (schedstat_enabled())
+		return;
+
+	/* Force schedstat enabled if a dependent tracepoint is active */
+	if (trace_sched_stat_wait_enabled()    ||
+	    trace_sched_stat_sleep_enabled()   ||
+	    trace_sched_stat_iowait_enabled()  ||
+	    trace_sched_stat_blocked_enabled() ||
+	    trace_sched_stat_runtime_enabled())
+		printk_deferred_once("Scheduler tracepoints stat_sleep, stat_iowait, stat_blocked and stat_runtime require the kernel parameter schedstats=enable or kernel.sched_schedstats=1\n");
+}
+
 #else /* !CONFIG_SCHEDSTATS: */
 
 static inline void rq_sched_info_arrive  (struct rq *rq, unsigned long long delta) { }
@@ -55,6 +80,11 @@ static inline void rq_sched_info_depart  (struct rq *rq, unsigned long long delt
 # define   schedstat_val(var)		0
 # define   schedstat_val_or_zero(var)	0
 
+# define __update_stats_wait_start(rq, p, stats)       do { } while (0)
+# define __update_stats_wait_end(rq, p, stats)         do { } while (0)
+# define __update_stats_enqueue_sleeper(rq, p, stats)  do { } while (0)
+# define check_schedstat_required()                    do { } while (0)
+
 #endif /* CONFIG_SCHEDSTATS */
 
 #ifdef CONFIG_FAIR_GROUP_SCHED
-- 
2.18.2


^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [PATCH v4 4/8] sched: introduce task block time in schedstats
  2021-09-05 14:35 [PATCH v4 0/8] sched: support schedstats for RT sched class Yafang Shao
                   ` (2 preceding siblings ...)
  2021-09-05 14:35 ` [PATCH v4 3/8] sched: make schedstats helpers " Yafang Shao
@ 2021-09-05 14:35 ` Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched: Introduce " tip-bot2 for Yafang Shao
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  2021-09-05 14:35 ` [PATCH v4 5/8] sched, rt: support sched_stat_runtime tracepoint for RT sched class Yafang Shao
                   ` (4 subsequent siblings)
  8 siblings, 2 replies; 29+ messages in thread
From: Yafang Shao @ 2021-09-05 14:35 UTC (permalink / raw)
  To: peterz, mingo, mgorman, juri.lelli, vincent.guittot,
	dietmar.eggemann, rostedt, bsegall, bristot
  Cc: linux-kernel, linux-rt-users, achaiken, Yafang Shao

Currently in schedstats we have sum_sleep_runtime and iowait_sum, but
there's no metric to show how long the task is in D state.  Once a task in
D state, it means the task is blocked in the kernel, for example the
task may be waiting for a mutex. The D state is more frequent than
iowait, and it is more critital than S state. So it is worth to add a
metric to measure it.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Alison Chaiken <achaiken@aurora.tech>
---
 include/linux/sched.h | 2 ++
 kernel/sched/debug.c  | 6 ++++--
 kernel/sched/stats.c  | 1 +
 3 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index ed4293812ca9..4239a3fbe7f3 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -502,6 +502,8 @@ struct sched_statistics {
 
 	u64				block_start;
 	u64				block_max;
+	s64				sum_block_runtime;
+
 	u64				exec_max;
 	u64				slice_max;
 
diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index e08eee374176..32eb86932bcb 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -534,10 +534,11 @@ print_task(struct seq_file *m, struct rq *rq, struct task_struct *p)
 		(long long)(p->nvcsw + p->nivcsw),
 		p->prio);
 
-	SEQ_printf(m, "%9Ld.%06ld %9Ld.%06ld %9Ld.%06ld",
+	SEQ_printf(m, "%9lld.%06ld %9lld.%06ld %9lld.%06ld %9lld.%06ld",
 		SPLIT_NS(schedstat_val_or_zero(p->stats.wait_sum)),
 		SPLIT_NS(p->se.sum_exec_runtime),
-		SPLIT_NS(schedstat_val_or_zero(p->stats.sum_sleep_runtime)));
+		SPLIT_NS(schedstat_val_or_zero(p->stats.sum_sleep_runtime)),
+		SPLIT_NS(schedstat_val_or_zero(p->stats.sum_block_runtime)));
 
 #ifdef CONFIG_NUMA_BALANCING
 	SEQ_printf(m, " %d %d", task_node(p), task_numa_group_id(p));
@@ -968,6 +969,7 @@ void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
 		u64 avg_atom, avg_per_cpu;
 
 		PN_SCHEDSTAT(sum_sleep_runtime);
+		PN_SCHEDSTAT(sum_block_runtime);
 		PN_SCHEDSTAT(wait_start);
 		PN_SCHEDSTAT(sleep_start);
 		PN_SCHEDSTAT(block_start);
diff --git a/kernel/sched/stats.c b/kernel/sched/stats.c
index fad781ca7791..07dde2928c79 100644
--- a/kernel/sched/stats.c
+++ b/kernel/sched/stats.c
@@ -82,6 +82,7 @@ void __update_stats_enqueue_sleeper(struct rq *rq, struct task_struct *p,
 
 		__schedstat_set(stats->block_start, 0);
 		__schedstat_add(stats->sum_sleep_runtime, delta);
+		__schedstat_add(stats->sum_block_runtime, delta);
 
 		if (p) {
 			if (p->in_iowait) {
-- 
2.18.2


^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [PATCH v4 5/8] sched, rt: support sched_stat_runtime tracepoint for RT sched class
  2021-09-05 14:35 [PATCH v4 0/8] sched: support schedstats for RT sched class Yafang Shao
                   ` (3 preceding siblings ...)
  2021-09-05 14:35 ` [PATCH v4 4/8] sched: introduce task block time in schedstats Yafang Shao
@ 2021-09-05 14:35 ` Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched/rt: Support " tip-bot2 for Yafang Shao
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  2021-09-05 14:35 ` [PATCH v4 6/8] sched, rt: support schedstats " Yafang Shao
                   ` (3 subsequent siblings)
  8 siblings, 2 replies; 29+ messages in thread
From: Yafang Shao @ 2021-09-05 14:35 UTC (permalink / raw)
  To: peterz, mingo, mgorman, juri.lelli, vincent.guittot,
	dietmar.eggemann, rostedt, bsegall, bristot
  Cc: linux-kernel, linux-rt-users, achaiken, Yafang Shao

The runtime of a RT task has already been there, so we only need to
add a tracepoint.

One difference between fair task and RT task is that there is no vruntime
in RT task. To reuse the sched_stat_runtime tracepoint, '0' is passed as
vruntime for RT task.

The output of this tracepoint for RT task as follows,
          stress-9748    [039] d.h.   113.519352: sched_stat_runtime: comm=stress pid=9748 runtime=997573 [ns] vruntime=0 [ns]
          stress-9748    [039] d.h.   113.520352: sched_stat_runtime: comm=stress pid=9748 runtime=997627 [ns] vruntime=0 [ns]
          stress-9748    [039] d.h.   113.521352: sched_stat_runtime: comm=stress pid=9748 runtime=998203 [ns] vruntime=0 [ns]

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Alison Chaiken <achaiken@aurora.tech>
---
 kernel/sched/rt.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 95a7c3ad2dc3..5d251112e51c 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1012,6 +1012,8 @@ static void update_curr_rt(struct rq *rq)
 	schedstat_set(curr->stats.exec_max,
 		      max(curr->stats.exec_max, delta_exec));
 
+	trace_sched_stat_runtime(curr, delta_exec, 0);
+
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);
 
-- 
2.18.2


^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [PATCH v4 6/8] sched, rt: support schedstats for RT sched class
  2021-09-05 14:35 [PATCH v4 0/8] sched: support schedstats for RT sched class Yafang Shao
                   ` (4 preceding siblings ...)
  2021-09-05 14:35 ` [PATCH v4 5/8] sched, rt: support sched_stat_runtime tracepoint for RT sched class Yafang Shao
@ 2021-09-05 14:35 ` Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched/rt: Support " tip-bot2 for Yafang Shao
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  2021-09-05 14:35 ` [PATCH v4 7/8] sched, dl: support sched_stat_runtime tracepoint for deadline " Yafang Shao
                   ` (2 subsequent siblings)
  8 siblings, 2 replies; 29+ messages in thread
From: Yafang Shao @ 2021-09-05 14:35 UTC (permalink / raw)
  To: peterz, mingo, mgorman, juri.lelli, vincent.guittot,
	dietmar.eggemann, rostedt, bsegall, bristot
  Cc: linux-kernel, linux-rt-users, achaiken, Yafang Shao, kernel test robot

We want to measure the latency of RT tasks in our production
environment with schedstats facility, but currently schedstats is only
supported for fair sched class. This patch enable it for RT sched class
as well.

After we make the struct sched_statistics and the helpers of it
independent of fair sched class, we can easily use the schedstats
facility for RT sched class.

The schedstat usage in RT sched class is similar with fair sched class,
for example,
                fair                        RT
enqueue         update_stats_enqueue_fair   update_stats_enqueue_rt
dequeue         update_stats_dequeue_fair   update_stats_dequeue_rt
put_prev_task   update_stats_wait_start     update_stats_wait_start_rt
set_next_task   update_stats_wait_end       update_stats_wait_end_rt

The user can get the schedstats information in the same way in fair sched
class. For example,
       fair                            RT
       /proc/[pid]/sched               /proc/[pid]/sched

schedstats is not supported for RT group.

The output of a RT task's schedstats as follows,
$ cat /proc/10349/sched
...
sum_sleep_runtime                            :           972.434535
sum_block_runtime                            :           960.433522
wait_start                                   :        188510.871584
sleep_start                                  :             0.000000
block_start                                  :             0.000000
sleep_max                                    :            12.001013
block_max                                    :           952.660622
exec_max                                     :             0.049629
slice_max                                    :             0.000000
wait_max                                     :             0.018538
wait_sum                                     :             0.424340
wait_count                                   :                   49
iowait_sum                                   :           956.495640
iowait_count                                 :                   24
nr_migrations_cold                           :                    0
nr_failed_migrations_affine                  :                    0
nr_failed_migrations_running                 :                    0
nr_failed_migrations_hot                     :                    0
nr_forced_migrations                         :                    0
nr_wakeups                                   :                   49
nr_wakeups_sync                              :                    0
nr_wakeups_migrate                           :                    0
nr_wakeups_local                             :                   49
nr_wakeups_remote                            :                    0
nr_wakeups_affine                            :                    0
nr_wakeups_affine_attempts                   :                    0
nr_wakeups_passive                           :                    0
nr_wakeups_idle                              :                    0
...

The sched:sched_stat_{wait, sleep, iowait, blocked} tracepoints can
be used to trace RT tasks as well. The output of these tracepoints for a
RT tasks as follows,

- runtime
          stress-10352   [004] d.h.  1035.382286: sched_stat_runtime: comm=stress pid=10352 runtime=995769 [ns] vruntime=0 [ns]
          [vruntime=0 means it is a RT task]

- wait
          <idle>-0       [004] dN..  1227.688544: sched_stat_wait: comm=stress pid=10352 delay=46849882 [ns]

- blocked
     kworker/4:1-465     [004] dN..  1585.676371: sched_stat_blocked: comm=stress pid=17194 delay=189963 [ns]

- iowait
     kworker/4:1-465     [004] dN..  1585.675330: sched_stat_iowait: comm=stress pid=17189 delay=182848 [ns]

- sleep
           sleep-18194   [023] dN..  1780.891840: sched_stat_sleep: comm=sleep.sh pid=17767 delay=1001160770 [ns]
           sleep-18196   [023] dN..  1781.893208: sched_stat_sleep: comm=sleep.sh pid=17767 delay=1001161970 [ns]
           sleep-18197   [023] dN..  1782.894544: sched_stat_sleep: comm=sleep.sh pid=17767 delay=1001128840 [ns]
           [ In sleep.sh, it sleeps 1 sec each time. ]

[lkp@intel.com: reported build failure in earlier version]

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Cc: kernel test robot <lkp@intel.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Alison Chaiken <achaiken@aurora.tech>
---
 kernel/sched/rt.c | 124 ++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 124 insertions(+)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 5d251112e51c..bb945f8faeca 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1273,6 +1273,112 @@ static void __delist_rt_entity(struct sched_rt_entity *rt_se, struct rt_prio_arr
 	rt_se->on_list = 0;
 }
 
+static inline struct sched_statistics *
+__schedstats_from_rt_se(struct sched_rt_entity *rt_se)
+{
+#ifdef CONFIG_RT_GROUP_SCHED
+	/* schedstats is not supported for rt group. */
+	if (!rt_entity_is_task(rt_se))
+		return NULL;
+#endif
+
+	return &rt_task_of(rt_se)->stats;
+}
+
+static inline void
+update_stats_wait_start_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se)
+{
+	struct sched_statistics *stats;
+	struct task_struct *p = NULL;
+
+	if (!schedstat_enabled())
+		return;
+
+	if (rt_entity_is_task(rt_se))
+		p = rt_task_of(rt_se);
+
+	stats = __schedstats_from_rt_se(rt_se);
+	if (!stats)
+		return;
+
+	__update_stats_wait_start(rq_of_rt_rq(rt_rq), p, stats);
+}
+
+static inline void
+update_stats_enqueue_sleeper_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se)
+{
+	struct sched_statistics *stats;
+	struct task_struct *p = NULL;
+
+	if (!schedstat_enabled())
+		return;
+
+	if (rt_entity_is_task(rt_se))
+		p = rt_task_of(rt_se);
+
+	stats = __schedstats_from_rt_se(rt_se);
+	if (!stats)
+		return;
+
+	__update_stats_enqueue_sleeper(rq_of_rt_rq(rt_rq), p, stats);
+}
+
+static inline void
+update_stats_enqueue_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se,
+			int flags)
+{
+	if (!schedstat_enabled())
+		return;
+
+	if (flags & ENQUEUE_WAKEUP)
+		update_stats_enqueue_sleeper_rt(rt_rq, rt_se);
+}
+
+static inline void
+update_stats_wait_end_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se)
+{
+	struct sched_statistics *stats;
+	struct task_struct *p = NULL;
+
+	if (!schedstat_enabled())
+		return;
+
+	if (rt_entity_is_task(rt_se))
+		p = rt_task_of(rt_se);
+
+	stats = __schedstats_from_rt_se(rt_se);
+	if (!stats)
+		return;
+
+	__update_stats_wait_end(rq_of_rt_rq(rt_rq), p, stats);
+}
+
+static inline void
+update_stats_dequeue_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se,
+			int flags)
+{
+	struct task_struct *p = NULL;
+
+	if (!schedstat_enabled())
+		return;
+
+	if (rt_entity_is_task(rt_se))
+		p = rt_task_of(rt_se);
+
+	if ((flags & DEQUEUE_SLEEP) && p) {
+		unsigned int state;
+
+		state = READ_ONCE(p->__state);
+		if (state & TASK_INTERRUPTIBLE)
+			__schedstat_set(p->stats.sleep_start,
+					rq_clock(rq_of_rt_rq(rt_rq)));
+
+		if (state & TASK_UNINTERRUPTIBLE)
+			__schedstat_set(p->stats.block_start,
+					rq_clock(rq_of_rt_rq(rt_rq)));
+	}
+}
+
 static void __enqueue_rt_entity(struct sched_rt_entity *rt_se, unsigned int flags)
 {
 	struct rt_rq *rt_rq = rt_rq_of_se(rt_se);
@@ -1346,6 +1452,8 @@ static void enqueue_rt_entity(struct sched_rt_entity *rt_se, unsigned int flags)
 {
 	struct rq *rq = rq_of_rt_se(rt_se);
 
+	update_stats_enqueue_rt(rt_rq_of_se(rt_se), rt_se, flags);
+
 	dequeue_rt_stack(rt_se, flags);
 	for_each_sched_rt_entity(rt_se)
 		__enqueue_rt_entity(rt_se, flags);
@@ -1356,6 +1464,8 @@ static void dequeue_rt_entity(struct sched_rt_entity *rt_se, unsigned int flags)
 {
 	struct rq *rq = rq_of_rt_se(rt_se);
 
+	update_stats_dequeue_rt(rt_rq_of_se(rt_se), rt_se, flags);
+
 	dequeue_rt_stack(rt_se, flags);
 
 	for_each_sched_rt_entity(rt_se) {
@@ -1378,6 +1488,9 @@ enqueue_task_rt(struct rq *rq, struct task_struct *p, int flags)
 	if (flags & ENQUEUE_WAKEUP)
 		rt_se->timeout = 0;
 
+	check_schedstat_required();
+	update_stats_wait_start_rt(rt_rq_of_se(rt_se), rt_se);
+
 	enqueue_rt_entity(rt_se, flags);
 
 	if (!task_current(rq, p) && p->nr_cpus_allowed > 1)
@@ -1578,7 +1691,12 @@ static void check_preempt_curr_rt(struct rq *rq, struct task_struct *p, int flag
 
 static inline void set_next_task_rt(struct rq *rq, struct task_struct *p, bool first)
 {
+	struct sched_rt_entity *rt_se = &p->rt;
+	struct rt_rq *rt_rq = &rq->rt;
+
 	p->se.exec_start = rq_clock_task(rq);
+	if (on_rt_rq(&p->rt))
+		update_stats_wait_end_rt(rt_rq, rt_se);
 
 	/* The running task is never eligible for pushing */
 	dequeue_pushable_task(rq, p);
@@ -1652,6 +1770,12 @@ static struct task_struct *pick_next_task_rt(struct rq *rq)
 
 static void put_prev_task_rt(struct rq *rq, struct task_struct *p)
 {
+	struct sched_rt_entity *rt_se = &p->rt;
+	struct rt_rq *rt_rq = &rq->rt;
+
+	if (on_rt_rq(&p->rt))
+		update_stats_wait_start_rt(rt_rq, rt_se);
+
 	update_curr_rt(rq);
 
 	update_rt_rq_load_avg(rq_clock_pelt(rq), rq, 1);
-- 
2.18.2


^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [PATCH v4 7/8] sched, dl: support sched_stat_runtime tracepoint for deadline sched class
  2021-09-05 14:35 [PATCH v4 0/8] sched: support schedstats for RT sched class Yafang Shao
                   ` (5 preceding siblings ...)
  2021-09-05 14:35 ` [PATCH v4 6/8] sched, rt: support schedstats " Yafang Shao
@ 2021-09-05 14:35 ` Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched/dl: Support " tip-bot2 for Yafang Shao
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  2021-09-05 14:35 ` [PATCH v4 8/8] sched, dl: support schedstats " Yafang Shao
  2021-09-10 17:00 ` [PATCH v4 0/8] sched: support schedstats for RT " Peter Zijlstra
  8 siblings, 2 replies; 29+ messages in thread
From: Yafang Shao @ 2021-09-05 14:35 UTC (permalink / raw)
  To: peterz, mingo, mgorman, juri.lelli, vincent.guittot,
	dietmar.eggemann, rostedt, bsegall, bristot
  Cc: linux-kernel, linux-rt-users, achaiken, Yafang Shao

The runtime of a DL task has already been there, so we only need to
add a tracepoint.

One difference between fair task and DL task is that there is no vruntime
in dl task. To reuse the sched_stat_runtime tracepoint, '0' is passed as
vruntime for DL task.

The output of this tracepoint for DL task as follows,
             top-36462   [047] d.h.  6083.452103: sched_stat_runtime: comm=top pid=36462 runtime=409898 [ns] vruntime=0 [ns]

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Alison Chaiken <achaiken@aurora.tech>
---
 kernel/sched/deadline.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index 51dd30990042..73fb33e1868f 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -1268,6 +1268,8 @@ static void update_curr_dl(struct rq *rq)
 	schedstat_set(curr->stats.exec_max,
 		      max(curr->stats.exec_max, delta_exec));
 
+	trace_sched_stat_runtime(curr, delta_exec, 0);
+
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);
 
-- 
2.18.2


^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [PATCH v4 8/8] sched, dl: support schedstats for deadline sched class
  2021-09-05 14:35 [PATCH v4 0/8] sched: support schedstats for RT sched class Yafang Shao
                   ` (6 preceding siblings ...)
  2021-09-05 14:35 ` [PATCH v4 7/8] sched, dl: support sched_stat_runtime tracepoint for deadline " Yafang Shao
@ 2021-09-05 14:35 ` Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched/dl: Support " tip-bot2 for Yafang Shao
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  2021-09-10 17:00 ` [PATCH v4 0/8] sched: support schedstats for RT " Peter Zijlstra
  8 siblings, 2 replies; 29+ messages in thread
From: Yafang Shao @ 2021-09-05 14:35 UTC (permalink / raw)
  To: peterz, mingo, mgorman, juri.lelli, vincent.guittot,
	dietmar.eggemann, rostedt, bsegall, bristot
  Cc: linux-kernel, linux-rt-users, achaiken, Yafang Shao

After we make the struct sched_statistics and the helpers of it
independent of fair sched class, we can easily use the schedstats
facility for deadline sched class.

The schedstat usage in DL sched class is similar with fair sched class,
for example,
                    fair                        deadline
    enqueue         update_stats_enqueue_fair   update_stats_enqueue_dl
    dequeue         update_stats_dequeue_fair   update_stats_dequeue_dl
    put_prev_task   update_stats_wait_start     update_stats_wait_start_dl
    set_next_task   update_stats_wait_end       update_stats_wait_end_dl

The user can get the schedstats information in the same way in fair sched
class. For example,
           fair                            deadline
           /proc/[pid]/sched               /proc/[pid]/sched

The output of a deadline task's schedstats as follows,

$ cat /proc/69662/sched
...
se.sum_exec_runtime                          :          3067.696449
se.nr_migrations                             :                    0
sum_sleep_runtime                            :        720144.029661
sum_block_runtime                            :             0.547853
wait_start                                   :             0.000000
sleep_start                                  :      14131540.828955
block_start                                  :             0.000000
sleep_max                                    :          2999.974045
block_max                                    :             0.283637
exec_max                                     :             1.000269
slice_max                                    :             0.000000
wait_max                                     :             0.002217
wait_sum                                     :             0.762179
wait_count                                   :                  733
iowait_sum                                   :             0.547853
iowait_count                                 :                    3
nr_migrations_cold                           :                    0
nr_failed_migrations_affine                  :                    0
nr_failed_migrations_running                 :                    0
nr_failed_migrations_hot                     :                    0
nr_forced_migrations                         :                    0
nr_wakeups                                   :                  246
nr_wakeups_sync                              :                    2
nr_wakeups_migrate                           :                    0
nr_wakeups_local                             :                  244
nr_wakeups_remote                            :                    2
nr_wakeups_affine                            :                    0
nr_wakeups_affine_attempts                   :                    0
nr_wakeups_passive                           :                    0
nr_wakeups_idle                              :                    0
...

The sched:sched_stat_{wait, sleep, iowait, blocked} tracepoints can
be used to trace deadlline tasks as well.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Alison Chaiken <achaiken@aurora.tech>
---
 kernel/sched/deadline.c | 93 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 93 insertions(+)

diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index 73fb33e1868f..d2c072b0ef01 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -1474,6 +1474,82 @@ static inline bool __dl_less(struct rb_node *a, const struct rb_node *b)
 	return dl_time_before(__node_2_dle(a)->deadline, __node_2_dle(b)->deadline);
 }
 
+static inline struct sched_statistics *
+__schedstats_from_dl_se(struct sched_dl_entity *dl_se)
+{
+	return &dl_task_of(dl_se)->stats;
+}
+
+static inline void
+update_stats_wait_start_dl(struct dl_rq *dl_rq, struct sched_dl_entity *dl_se)
+{
+	struct sched_statistics *stats;
+
+	if (!schedstat_enabled())
+		return;
+
+	stats = __schedstats_from_dl_se(dl_se);
+	__update_stats_wait_start(rq_of_dl_rq(dl_rq), dl_task_of(dl_se), stats);
+}
+
+static inline void
+update_stats_wait_end_dl(struct dl_rq *dl_rq, struct sched_dl_entity *dl_se)
+{
+	struct sched_statistics *stats;
+
+	if (!schedstat_enabled())
+		return;
+
+	stats = __schedstats_from_dl_se(dl_se);
+	__update_stats_wait_end(rq_of_dl_rq(dl_rq), dl_task_of(dl_se), stats);
+}
+
+static inline void
+update_stats_enqueue_sleeper_dl(struct dl_rq *dl_rq, struct sched_dl_entity *dl_se)
+{
+	struct sched_statistics *stats;
+
+	if (!schedstat_enabled())
+		return;
+
+	stats = __schedstats_from_dl_se(dl_se);
+	__update_stats_enqueue_sleeper(rq_of_dl_rq(dl_rq), dl_task_of(dl_se), stats);
+}
+
+static inline void
+update_stats_enqueue_dl(struct dl_rq *dl_rq, struct sched_dl_entity *dl_se,
+			int flags)
+{
+	if (!schedstat_enabled())
+		return;
+
+	if (flags & ENQUEUE_WAKEUP)
+		update_stats_enqueue_sleeper_dl(dl_rq, dl_se);
+}
+
+static inline void
+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);
+
+	if (!schedstat_enabled())
+		return;
+
+	if ((flags & DEQUEUE_SLEEP)) {
+		unsigned int state;
+
+		state = READ_ONCE(p->__state);
+		if (state & TASK_INTERRUPTIBLE)
+			__schedstat_set(p->stats.sleep_start,
+					rq_clock(rq_of_dl_rq(dl_rq)));
+
+		if (state & TASK_UNINTERRUPTIBLE)
+			__schedstat_set(p->stats.block_start,
+					rq_clock(rq_of_dl_rq(dl_rq)));
+	}
+}
+
 static void __enqueue_dl_entity(struct sched_dl_entity *dl_se)
 {
 	struct dl_rq *dl_rq = dl_rq_of_se(dl_se);
@@ -1504,6 +1580,8 @@ enqueue_dl_entity(struct sched_dl_entity *dl_se, int flags)
 {
 	BUG_ON(on_dl_rq(dl_se));
 
+	update_stats_enqueue_dl(dl_rq_of_se(dl_se), dl_se, flags);
+
 	/*
 	 * If this is a wakeup or a new instance, the scheduling
 	 * parameters of the task might need updating. Otherwise,
@@ -1600,6 +1678,9 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 		return;
 	}
 
+	check_schedstat_required();
+	update_stats_wait_start_dl(dl_rq_of_se(&p->dl), &p->dl);
+
 	enqueue_dl_entity(&p->dl, flags);
 
 	if (!task_current(rq, p) && p->nr_cpus_allowed > 1)
@@ -1608,6 +1689,7 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 
 static void __dequeue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 {
+	update_stats_dequeue_dl(&rq->dl, &p->dl, flags);
 	dequeue_dl_entity(&p->dl);
 	dequeue_pushable_dl_task(rq, p);
 }
@@ -1827,7 +1909,12 @@ static void start_hrtick_dl(struct rq *rq, struct task_struct *p)
 
 static void set_next_task_dl(struct rq *rq, struct task_struct *p, bool first)
 {
+	struct sched_dl_entity *dl_se = &p->dl;
+	struct dl_rq *dl_rq = &rq->dl;
+
 	p->se.exec_start = rq_clock_task(rq);
+	if (on_dl_rq(&p->dl))
+		update_stats_wait_end_dl(dl_rq, dl_se);
 
 	/* You can't push away the running task */
 	dequeue_pushable_dl_task(rq, p);
@@ -1884,6 +1971,12 @@ static struct task_struct *pick_next_task_dl(struct rq *rq)
 
 static void put_prev_task_dl(struct rq *rq, struct task_struct *p)
 {
+	struct sched_dl_entity *dl_se = &p->dl;
+	struct dl_rq *dl_rq = &rq->dl;
+
+	if (on_dl_rq(&p->dl))
+		update_stats_wait_start_dl(dl_rq, dl_se);
+
 	update_curr_dl(rq);
 
 	update_dl_rq_load_avg(rq_clock_pelt(rq), rq, 1);
-- 
2.18.2


^ permalink raw reply related	[flat|nested] 29+ messages in thread

* Re: [PATCH v4 0/8] sched: support schedstats for RT sched class
  2021-09-05 14:35 [PATCH v4 0/8] sched: support schedstats for RT sched class Yafang Shao
                   ` (7 preceding siblings ...)
  2021-09-05 14:35 ` [PATCH v4 8/8] sched, dl: support schedstats " Yafang Shao
@ 2021-09-10 17:00 ` Peter Zijlstra
  8 siblings, 0 replies; 29+ messages in thread
From: Peter Zijlstra @ 2021-09-10 17:00 UTC (permalink / raw)
  To: Yafang Shao
  Cc: mingo, mgorman, juri.lelli, vincent.guittot, dietmar.eggemann,
	rostedt, bsegall, bristot, linux-kernel, linux-rt-users,
	achaiken

On Sun, Sep 05, 2021 at 02:35:39PM +0000, Yafang Shao wrote:
> Yafang Shao (8):
>   sched, fair: use __schedstat_set() in set_next_entity()
>   sched: make struct sched_statistics independent of fair sched class
>   sched: make schedstats helpers independent of fair sched class
>   sched: introduce task block time in schedstats
>   sched, rt: support sched_stat_runtime tracepoint for RT sched class
>   sched, rt: support schedstats for RT sched class
>   sched, dl: support sched_stat_runtime tracepoint for deadline sched
>     class
>   sched, dl: support schedstats for deadline sched class
> 
>  include/linux/sched.h    |   8 +-
>  kernel/sched/core.c      |  25 +++---
>  kernel/sched/deadline.c  |  99 +++++++++++++++++++++-
>  kernel/sched/debug.c     |  97 +++++++++++----------
>  kernel/sched/fair.c      | 177 +++++++++++----------------------------
>  kernel/sched/rt.c        | 130 +++++++++++++++++++++++++++-
>  kernel/sched/stats.c     | 104 +++++++++++++++++++++++
>  kernel/sched/stats.h     |  49 +++++++++++
>  kernel/sched/stop_task.c |   4 +-
>  9 files changed, 500 insertions(+), 193 deletions(-)

*sigh*, okay I suppose, let's try this :-)

^ permalink raw reply	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched/dl: Support sched_stat_runtime tracepoint for deadline sched class
  2021-09-05 14:35 ` [PATCH v4 7/8] sched, dl: support sched_stat_runtime tracepoint for deadline " Yafang Shao
@ 2021-09-16 11:59   ` tip-bot2 for Yafang Shao
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 0 replies; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-09-16 11:59 UTC (permalink / raw)
  To: linux-tip-commits; +Cc: Yafang Shao, Peter Zijlstra (Intel), x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     db4570a26807a7a955a5dc58e0bea3fc2b1d7c23
Gitweb:        https://git.kernel.org/tip/db4570a26807a7a955a5dc58e0bea3fc2b1d7c23
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:46 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Wed, 15 Sep 2021 17:49:00 +02:00

sched/dl: Support sched_stat_runtime tracepoint for deadline sched class

The runtime of a DL task has already been there, so we only need to
add a tracepoint.

One difference between fair task and DL task is that there is no vruntime
in dl task. To reuse the sched_stat_runtime tracepoint, '0' is passed as
vruntime for DL task.

The output of this tracepoint for DL task as follows,
             top-36462   [047] d.h.  6083.452103: sched_stat_runtime: comm=top pid=36462 runtime=409898 [ns] vruntime=0 [ns]

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210905143547.4668-8-laoar.shao@gmail.com
---
 kernel/sched/deadline.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index 51dd309..73fb33e 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -1268,6 +1268,8 @@ static void update_curr_dl(struct rq *rq)
 	schedstat_set(curr->stats.exec_max,
 		      max(curr->stats.exec_max, delta_exec));
 
+	trace_sched_stat_runtime(curr, delta_exec, 0);
+
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);
 

^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched/dl: Support schedstats for deadline sched class
  2021-09-05 14:35 ` [PATCH v4 8/8] sched, dl: support schedstats " Yafang Shao
@ 2021-09-16 11:59   ` tip-bot2 for Yafang Shao
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 0 replies; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-09-16 11:59 UTC (permalink / raw)
  To: linux-tip-commits; +Cc: Yafang Shao, Peter Zijlstra (Intel), x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     fce7dd1e9c19be245c565d1544db0dcafb230711
Gitweb:        https://git.kernel.org/tip/fce7dd1e9c19be245c565d1544db0dcafb230711
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:47 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Wed, 15 Sep 2021 17:49:00 +02:00

sched/dl: Support schedstats for deadline sched class

After we make the struct sched_statistics and the helpers of it
independent of fair sched class, we can easily use the schedstats
facility for deadline sched class.

The schedstat usage in DL sched class is similar with fair sched class,
for example,
                    fair                        deadline
    enqueue         update_stats_enqueue_fair   update_stats_enqueue_dl
    dequeue         update_stats_dequeue_fair   update_stats_dequeue_dl
    put_prev_task   update_stats_wait_start     update_stats_wait_start_dl
    set_next_task   update_stats_wait_end       update_stats_wait_end_dl

The user can get the schedstats information in the same way in fair sched
class. For example,
           fair                            deadline
           /proc/[pid]/sched               /proc/[pid]/sched

The output of a deadline task's schedstats as follows,

$ cat /proc/69662/sched
...
se.sum_exec_runtime                          :          3067.696449
se.nr_migrations                             :                    0
sum_sleep_runtime                            :        720144.029661
sum_block_runtime                            :             0.547853
wait_start                                   :             0.000000
sleep_start                                  :      14131540.828955
block_start                                  :             0.000000
sleep_max                                    :          2999.974045
block_max                                    :             0.283637
exec_max                                     :             1.000269
slice_max                                    :             0.000000
wait_max                                     :             0.002217
wait_sum                                     :             0.762179
wait_count                                   :                  733
iowait_sum                                   :             0.547853
iowait_count                                 :                    3
nr_migrations_cold                           :                    0
nr_failed_migrations_affine                  :                    0
nr_failed_migrations_running                 :                    0
nr_failed_migrations_hot                     :                    0
nr_forced_migrations                         :                    0
nr_wakeups                                   :                  246
nr_wakeups_sync                              :                    2
nr_wakeups_migrate                           :                    0
nr_wakeups_local                             :                  244
nr_wakeups_remote                            :                    2
nr_wakeups_affine                            :                    0
nr_wakeups_affine_attempts                   :                    0
nr_wakeups_passive                           :                    0
nr_wakeups_idle                              :                    0
...

The sched:sched_stat_{wait, sleep, iowait, blocked} tracepoints can
be used to trace deadlline tasks as well.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210905143547.4668-9-laoar.shao@gmail.com
---
 kernel/sched/deadline.c | 93 ++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 93 insertions(+)

diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index 73fb33e..d2c072b 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -1474,6 +1474,82 @@ static inline bool __dl_less(struct rb_node *a, const struct rb_node *b)
 	return dl_time_before(__node_2_dle(a)->deadline, __node_2_dle(b)->deadline);
 }
 
+static inline struct sched_statistics *
+__schedstats_from_dl_se(struct sched_dl_entity *dl_se)
+{
+	return &dl_task_of(dl_se)->stats;
+}
+
+static inline void
+update_stats_wait_start_dl(struct dl_rq *dl_rq, struct sched_dl_entity *dl_se)
+{
+	struct sched_statistics *stats;
+
+	if (!schedstat_enabled())
+		return;
+
+	stats = __schedstats_from_dl_se(dl_se);
+	__update_stats_wait_start(rq_of_dl_rq(dl_rq), dl_task_of(dl_se), stats);
+}
+
+static inline void
+update_stats_wait_end_dl(struct dl_rq *dl_rq, struct sched_dl_entity *dl_se)
+{
+	struct sched_statistics *stats;
+
+	if (!schedstat_enabled())
+		return;
+
+	stats = __schedstats_from_dl_se(dl_se);
+	__update_stats_wait_end(rq_of_dl_rq(dl_rq), dl_task_of(dl_se), stats);
+}
+
+static inline void
+update_stats_enqueue_sleeper_dl(struct dl_rq *dl_rq, struct sched_dl_entity *dl_se)
+{
+	struct sched_statistics *stats;
+
+	if (!schedstat_enabled())
+		return;
+
+	stats = __schedstats_from_dl_se(dl_se);
+	__update_stats_enqueue_sleeper(rq_of_dl_rq(dl_rq), dl_task_of(dl_se), stats);
+}
+
+static inline void
+update_stats_enqueue_dl(struct dl_rq *dl_rq, struct sched_dl_entity *dl_se,
+			int flags)
+{
+	if (!schedstat_enabled())
+		return;
+
+	if (flags & ENQUEUE_WAKEUP)
+		update_stats_enqueue_sleeper_dl(dl_rq, dl_se);
+}
+
+static inline void
+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);
+
+	if (!schedstat_enabled())
+		return;
+
+	if ((flags & DEQUEUE_SLEEP)) {
+		unsigned int state;
+
+		state = READ_ONCE(p->__state);
+		if (state & TASK_INTERRUPTIBLE)
+			__schedstat_set(p->stats.sleep_start,
+					rq_clock(rq_of_dl_rq(dl_rq)));
+
+		if (state & TASK_UNINTERRUPTIBLE)
+			__schedstat_set(p->stats.block_start,
+					rq_clock(rq_of_dl_rq(dl_rq)));
+	}
+}
+
 static void __enqueue_dl_entity(struct sched_dl_entity *dl_se)
 {
 	struct dl_rq *dl_rq = dl_rq_of_se(dl_se);
@@ -1504,6 +1580,8 @@ enqueue_dl_entity(struct sched_dl_entity *dl_se, int flags)
 {
 	BUG_ON(on_dl_rq(dl_se));
 
+	update_stats_enqueue_dl(dl_rq_of_se(dl_se), dl_se, flags);
+
 	/*
 	 * If this is a wakeup or a new instance, the scheduling
 	 * parameters of the task might need updating. Otherwise,
@@ -1600,6 +1678,9 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 		return;
 	}
 
+	check_schedstat_required();
+	update_stats_wait_start_dl(dl_rq_of_se(&p->dl), &p->dl);
+
 	enqueue_dl_entity(&p->dl, flags);
 
 	if (!task_current(rq, p) && p->nr_cpus_allowed > 1)
@@ -1608,6 +1689,7 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 
 static void __dequeue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 {
+	update_stats_dequeue_dl(&rq->dl, &p->dl, flags);
 	dequeue_dl_entity(&p->dl);
 	dequeue_pushable_dl_task(rq, p);
 }
@@ -1827,7 +1909,12 @@ static void start_hrtick_dl(struct rq *rq, struct task_struct *p)
 
 static void set_next_task_dl(struct rq *rq, struct task_struct *p, bool first)
 {
+	struct sched_dl_entity *dl_se = &p->dl;
+	struct dl_rq *dl_rq = &rq->dl;
+
 	p->se.exec_start = rq_clock_task(rq);
+	if (on_dl_rq(&p->dl))
+		update_stats_wait_end_dl(dl_rq, dl_se);
 
 	/* You can't push away the running task */
 	dequeue_pushable_dl_task(rq, p);
@@ -1884,6 +1971,12 @@ static struct task_struct *pick_next_task_dl(struct rq *rq)
 
 static void put_prev_task_dl(struct rq *rq, struct task_struct *p)
 {
+	struct sched_dl_entity *dl_se = &p->dl;
+	struct dl_rq *dl_rq = &rq->dl;
+
+	if (on_dl_rq(&p->dl))
+		update_stats_wait_start_dl(dl_rq, dl_se);
+
 	update_curr_dl(rq);
 
 	update_dl_rq_load_avg(rq_clock_pelt(rq), rq, 1);

^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched/rt: Support schedstats for RT sched class
  2021-09-05 14:35 ` [PATCH v4 6/8] sched, rt: support schedstats " Yafang Shao
@ 2021-09-16 11:59   ` tip-bot2 for Yafang Shao
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 0 replies; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-09-16 11:59 UTC (permalink / raw)
  To: linux-tip-commits; +Cc: Yafang Shao, Peter Zijlstra (Intel), x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     1ff69baabc2f0f3b098ad94429d3d3cb2897f097
Gitweb:        https://git.kernel.org/tip/1ff69baabc2f0f3b098ad94429d3d3cb2897f097
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:45 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Wed, 15 Sep 2021 17:49:00 +02:00

sched/rt: Support schedstats for RT sched class

We want to measure the latency of RT tasks in our production
environment with schedstats facility, but currently schedstats is only
supported for fair sched class. This patch enable it for RT sched class
as well.

After we make the struct sched_statistics and the helpers of it
independent of fair sched class, we can easily use the schedstats
facility for RT sched class.

The schedstat usage in RT sched class is similar with fair sched class,
for example,
                fair                        RT
enqueue         update_stats_enqueue_fair   update_stats_enqueue_rt
dequeue         update_stats_dequeue_fair   update_stats_dequeue_rt
put_prev_task   update_stats_wait_start     update_stats_wait_start_rt
set_next_task   update_stats_wait_end       update_stats_wait_end_rt

The user can get the schedstats information in the same way in fair sched
class. For example,
       fair                            RT
       /proc/[pid]/sched               /proc/[pid]/sched

schedstats is not supported for RT group.

The output of a RT task's schedstats as follows,
$ cat /proc/10349/sched
...
sum_sleep_runtime                            :           972.434535
sum_block_runtime                            :           960.433522
wait_start                                   :        188510.871584
sleep_start                                  :             0.000000
block_start                                  :             0.000000
sleep_max                                    :            12.001013
block_max                                    :           952.660622
exec_max                                     :             0.049629
slice_max                                    :             0.000000
wait_max                                     :             0.018538
wait_sum                                     :             0.424340
wait_count                                   :                   49
iowait_sum                                   :           956.495640
iowait_count                                 :                   24
nr_migrations_cold                           :                    0
nr_failed_migrations_affine                  :                    0
nr_failed_migrations_running                 :                    0
nr_failed_migrations_hot                     :                    0
nr_forced_migrations                         :                    0
nr_wakeups                                   :                   49
nr_wakeups_sync                              :                    0
nr_wakeups_migrate                           :                    0
nr_wakeups_local                             :                   49
nr_wakeups_remote                            :                    0
nr_wakeups_affine                            :                    0
nr_wakeups_affine_attempts                   :                    0
nr_wakeups_passive                           :                    0
nr_wakeups_idle                              :                    0
...

The sched:sched_stat_{wait, sleep, iowait, blocked} tracepoints can
be used to trace RT tasks as well. The output of these tracepoints for a
RT tasks as follows,

- runtime
          stress-10352   [004] d.h.  1035.382286: sched_stat_runtime: comm=stress pid=10352 runtime=995769 [ns] vruntime=0 [ns]
          [vruntime=0 means it is a RT task]

- wait
          <idle>-0       [004] dN..  1227.688544: sched_stat_wait: comm=stress pid=10352 delay=46849882 [ns]

- blocked
     kworker/4:1-465     [004] dN..  1585.676371: sched_stat_blocked: comm=stress pid=17194 delay=189963 [ns]

- iowait
     kworker/4:1-465     [004] dN..  1585.675330: sched_stat_iowait: comm=stress pid=17189 delay=182848 [ns]

- sleep
           sleep-18194   [023] dN..  1780.891840: sched_stat_sleep: comm=sleep.sh pid=17767 delay=1001160770 [ns]
           sleep-18196   [023] dN..  1781.893208: sched_stat_sleep: comm=sleep.sh pid=17767 delay=1001161970 [ns]
           sleep-18197   [023] dN..  1782.894544: sched_stat_sleep: comm=sleep.sh pid=17767 delay=1001128840 [ns]
           [ In sleep.sh, it sleeps 1 sec each time. ]

[lkp@intel.com: reported build failure in earlier version]

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210905143547.4668-7-laoar.shao@gmail.com
---
 kernel/sched/rt.c | 124 +++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 124 insertions(+)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 5d25111..bb945f8 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1273,6 +1273,112 @@ static void __delist_rt_entity(struct sched_rt_entity *rt_se, struct rt_prio_arr
 	rt_se->on_list = 0;
 }
 
+static inline struct sched_statistics *
+__schedstats_from_rt_se(struct sched_rt_entity *rt_se)
+{
+#ifdef CONFIG_RT_GROUP_SCHED
+	/* schedstats is not supported for rt group. */
+	if (!rt_entity_is_task(rt_se))
+		return NULL;
+#endif
+
+	return &rt_task_of(rt_se)->stats;
+}
+
+static inline void
+update_stats_wait_start_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se)
+{
+	struct sched_statistics *stats;
+	struct task_struct *p = NULL;
+
+	if (!schedstat_enabled())
+		return;
+
+	if (rt_entity_is_task(rt_se))
+		p = rt_task_of(rt_se);
+
+	stats = __schedstats_from_rt_se(rt_se);
+	if (!stats)
+		return;
+
+	__update_stats_wait_start(rq_of_rt_rq(rt_rq), p, stats);
+}
+
+static inline void
+update_stats_enqueue_sleeper_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se)
+{
+	struct sched_statistics *stats;
+	struct task_struct *p = NULL;
+
+	if (!schedstat_enabled())
+		return;
+
+	if (rt_entity_is_task(rt_se))
+		p = rt_task_of(rt_se);
+
+	stats = __schedstats_from_rt_se(rt_se);
+	if (!stats)
+		return;
+
+	__update_stats_enqueue_sleeper(rq_of_rt_rq(rt_rq), p, stats);
+}
+
+static inline void
+update_stats_enqueue_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se,
+			int flags)
+{
+	if (!schedstat_enabled())
+		return;
+
+	if (flags & ENQUEUE_WAKEUP)
+		update_stats_enqueue_sleeper_rt(rt_rq, rt_se);
+}
+
+static inline void
+update_stats_wait_end_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se)
+{
+	struct sched_statistics *stats;
+	struct task_struct *p = NULL;
+
+	if (!schedstat_enabled())
+		return;
+
+	if (rt_entity_is_task(rt_se))
+		p = rt_task_of(rt_se);
+
+	stats = __schedstats_from_rt_se(rt_se);
+	if (!stats)
+		return;
+
+	__update_stats_wait_end(rq_of_rt_rq(rt_rq), p, stats);
+}
+
+static inline void
+update_stats_dequeue_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se,
+			int flags)
+{
+	struct task_struct *p = NULL;
+
+	if (!schedstat_enabled())
+		return;
+
+	if (rt_entity_is_task(rt_se))
+		p = rt_task_of(rt_se);
+
+	if ((flags & DEQUEUE_SLEEP) && p) {
+		unsigned int state;
+
+		state = READ_ONCE(p->__state);
+		if (state & TASK_INTERRUPTIBLE)
+			__schedstat_set(p->stats.sleep_start,
+					rq_clock(rq_of_rt_rq(rt_rq)));
+
+		if (state & TASK_UNINTERRUPTIBLE)
+			__schedstat_set(p->stats.block_start,
+					rq_clock(rq_of_rt_rq(rt_rq)));
+	}
+}
+
 static void __enqueue_rt_entity(struct sched_rt_entity *rt_se, unsigned int flags)
 {
 	struct rt_rq *rt_rq = rt_rq_of_se(rt_se);
@@ -1346,6 +1452,8 @@ static void enqueue_rt_entity(struct sched_rt_entity *rt_se, unsigned int flags)
 {
 	struct rq *rq = rq_of_rt_se(rt_se);
 
+	update_stats_enqueue_rt(rt_rq_of_se(rt_se), rt_se, flags);
+
 	dequeue_rt_stack(rt_se, flags);
 	for_each_sched_rt_entity(rt_se)
 		__enqueue_rt_entity(rt_se, flags);
@@ -1356,6 +1464,8 @@ static void dequeue_rt_entity(struct sched_rt_entity *rt_se, unsigned int flags)
 {
 	struct rq *rq = rq_of_rt_se(rt_se);
 
+	update_stats_dequeue_rt(rt_rq_of_se(rt_se), rt_se, flags);
+
 	dequeue_rt_stack(rt_se, flags);
 
 	for_each_sched_rt_entity(rt_se) {
@@ -1378,6 +1488,9 @@ enqueue_task_rt(struct rq *rq, struct task_struct *p, int flags)
 	if (flags & ENQUEUE_WAKEUP)
 		rt_se->timeout = 0;
 
+	check_schedstat_required();
+	update_stats_wait_start_rt(rt_rq_of_se(rt_se), rt_se);
+
 	enqueue_rt_entity(rt_se, flags);
 
 	if (!task_current(rq, p) && p->nr_cpus_allowed > 1)
@@ -1578,7 +1691,12 @@ static void check_preempt_curr_rt(struct rq *rq, struct task_struct *p, int flag
 
 static inline void set_next_task_rt(struct rq *rq, struct task_struct *p, bool first)
 {
+	struct sched_rt_entity *rt_se = &p->rt;
+	struct rt_rq *rt_rq = &rq->rt;
+
 	p->se.exec_start = rq_clock_task(rq);
+	if (on_rt_rq(&p->rt))
+		update_stats_wait_end_rt(rt_rq, rt_se);
 
 	/* The running task is never eligible for pushing */
 	dequeue_pushable_task(rq, p);
@@ -1652,6 +1770,12 @@ static struct task_struct *pick_next_task_rt(struct rq *rq)
 
 static void put_prev_task_rt(struct rq *rq, struct task_struct *p)
 {
+	struct sched_rt_entity *rt_se = &p->rt;
+	struct rt_rq *rt_rq = &rq->rt;
+
+	if (on_rt_rq(&p->rt))
+		update_stats_wait_start_rt(rt_rq, rt_se);
+
 	update_curr_rt(rq);
 
 	update_rt_rq_load_avg(rq_clock_pelt(rq), rq, 1);

^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched/rt: Support sched_stat_runtime tracepoint for RT sched class
  2021-09-05 14:35 ` [PATCH v4 5/8] sched, rt: support sched_stat_runtime tracepoint for RT sched class Yafang Shao
@ 2021-09-16 11:59   ` tip-bot2 for Yafang Shao
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 0 replies; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-09-16 11:59 UTC (permalink / raw)
  To: linux-tip-commits; +Cc: Yafang Shao, Peter Zijlstra (Intel), x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     75ded49a19c6b655bba7ff9e5541daef4f882d78
Gitweb:        https://git.kernel.org/tip/75ded49a19c6b655bba7ff9e5541daef4f882d78
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:44 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Wed, 15 Sep 2021 17:48:59 +02:00

sched/rt: Support sched_stat_runtime tracepoint for RT sched class

The runtime of a RT task has already been there, so we only need to
add a tracepoint.

One difference between fair task and RT task is that there is no vruntime
in RT task. To reuse the sched_stat_runtime tracepoint, '0' is passed as
vruntime for RT task.

The output of this tracepoint for RT task as follows,
          stress-9748    [039] d.h.   113.519352: sched_stat_runtime: comm=stress pid=9748 runtime=997573 [ns] vruntime=0 [ns]
          stress-9748    [039] d.h.   113.520352: sched_stat_runtime: comm=stress pid=9748 runtime=997627 [ns] vruntime=0 [ns]
          stress-9748    [039] d.h.   113.521352: sched_stat_runtime: comm=stress pid=9748 runtime=998203 [ns] vruntime=0 [ns]

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210905143547.4668-6-laoar.shao@gmail.com
---
 kernel/sched/rt.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 95a7c3a..5d25111 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1012,6 +1012,8 @@ static void update_curr_rt(struct rq *rq)
 	schedstat_set(curr->stats.exec_max,
 		      max(curr->stats.exec_max, delta_exec));
 
+	trace_sched_stat_runtime(curr, delta_exec, 0);
+
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);
 

^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched: Introduce task block time in schedstats
  2021-09-05 14:35 ` [PATCH v4 4/8] sched: introduce task block time in schedstats Yafang Shao
@ 2021-09-16 11:59   ` tip-bot2 for Yafang Shao
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 0 replies; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-09-16 11:59 UTC (permalink / raw)
  To: linux-tip-commits; +Cc: Yafang Shao, Peter Zijlstra (Intel), x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     a95f0507a1a4d6de8e9c8b2b61cb733947e9422b
Gitweb:        https://git.kernel.org/tip/a95f0507a1a4d6de8e9c8b2b61cb733947e9422b
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:43 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Wed, 15 Sep 2021 17:48:59 +02:00

sched: Introduce task block time in schedstats

Currently in schedstats we have sum_sleep_runtime and iowait_sum, but
there's no metric to show how long the task is in D state.  Once a task in
D state, it means the task is blocked in the kernel, for example the
task may be waiting for a mutex. The D state is more frequent than
iowait, and it is more critital than S state. So it is worth to add a
metric to measure it.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210905143547.4668-5-laoar.shao@gmail.com
---
 include/linux/sched.h | 2 ++
 kernel/sched/debug.c  | 6 ++++--
 kernel/sched/stats.c  | 1 +
 3 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 518e19c..549018e 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -441,6 +441,8 @@ struct sched_statistics {
 
 	u64				block_start;
 	u64				block_max;
+	s64				sum_block_runtime;
+
 	u64				exec_max;
 	u64				slice_max;
 
diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index ca0dd47..935dad7 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -534,10 +534,11 @@ print_task(struct seq_file *m, struct rq *rq, struct task_struct *p)
 		(long long)(p->nvcsw + p->nivcsw),
 		p->prio);
 
-	SEQ_printf(m, "%9Ld.%06ld %9Ld.%06ld %9Ld.%06ld",
+	SEQ_printf(m, "%9lld.%06ld %9lld.%06ld %9lld.%06ld %9lld.%06ld",
 		SPLIT_NS(schedstat_val_or_zero(p->stats.wait_sum)),
 		SPLIT_NS(p->se.sum_exec_runtime),
-		SPLIT_NS(schedstat_val_or_zero(p->stats.sum_sleep_runtime)));
+		SPLIT_NS(schedstat_val_or_zero(p->stats.sum_sleep_runtime)),
+		SPLIT_NS(schedstat_val_or_zero(p->stats.sum_block_runtime)));
 
 #ifdef CONFIG_NUMA_BALANCING
 	SEQ_printf(m, " %d %d", task_node(p), task_numa_group_id(p));
@@ -971,6 +972,7 @@ void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
 		u64 avg_atom, avg_per_cpu;
 
 		PN_SCHEDSTAT(sum_sleep_runtime);
+		PN_SCHEDSTAT(sum_block_runtime);
 		PN_SCHEDSTAT(wait_start);
 		PN_SCHEDSTAT(sleep_start);
 		PN_SCHEDSTAT(block_start);
diff --git a/kernel/sched/stats.c b/kernel/sched/stats.c
index fad781c..07dde29 100644
--- a/kernel/sched/stats.c
+++ b/kernel/sched/stats.c
@@ -82,6 +82,7 @@ void __update_stats_enqueue_sleeper(struct rq *rq, struct task_struct *p,
 
 		__schedstat_set(stats->block_start, 0);
 		__schedstat_add(stats->sum_sleep_runtime, delta);
+		__schedstat_add(stats->sum_block_runtime, delta);
 
 		if (p) {
 			if (p->in_iowait) {

^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched: Make schedstats helpers independent of fair sched class
  2021-09-05 14:35 ` [PATCH v4 3/8] sched: make schedstats helpers " Yafang Shao
@ 2021-09-16 11:59   ` tip-bot2 for Yafang Shao
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 0 replies; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-09-16 11:59 UTC (permalink / raw)
  To: linux-tip-commits
  Cc: Yafang Shao, Peter Zijlstra (Intel), Mel Gorman, x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     a05296a6e2914899bc6492de92579a7ed771b7e8
Gitweb:        https://git.kernel.org/tip/a05296a6e2914899bc6492de92579a7ed771b7e8
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:42 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Wed, 15 Sep 2021 17:48:59 +02:00

sched: Make schedstats helpers independent of fair sched class

The original prototype of the schedstats helpers are

  update_stats_wait_*(struct cfs_rq *cfs_rq, struct sched_entity *se)

The cfs_rq in these helpers is used to get the rq_clock, and the se is
used to get the struct sched_statistics and the struct task_struct. In
order to make these helpers available by all sched classes, we can pass
the rq, sched_statistics and task_struct directly.

Then the new helpers are

  update_stats_wait_*(struct rq *rq, struct task_struct *p,
                      struct sched_statistics *stats)

which are independent of fair sched class.

To avoid vmlinux growing too large or introducing ovehead when
!schedstat_enabled(), some new helpers after schedstat_enabled() are also
introduced, Suggested by Mel. These helpers are in sched/stats.c,

  __update_stats_wait_*(struct rq *rq, struct task_struct *p,
                        struct sched_statistics *stats)

The size of vmlinux as follows,
                      Before          After
  Size of vmlinux     826308552       826304640
The size is a litte smaller as some functions are not inlined again after
the change.

I also compared the sched performance with 'perf bench sched pipe',
suggested by Mel. The result as followsi (in usecs/op),
                             Before                After
  kernel.sched_schedstats=0  5.2~5.4               5.2~5.4
  kernel.sched_schedstats=1  5.3~5.5               5.3~5.5

[These data is a little difference with the prev version, that is
because my old test machine is destroyed so I have to use a new
different test machine.]
Almost no difference.

No functional change.

[lkp@intel.com: reported build failure in prev version]

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lore.kernel.org/r/20210905143547.4668-4-laoar.shao@gmail.com
---
 kernel/sched/fair.c  | 134 +++++-------------------------------------
 kernel/sched/stats.c | 103 ++++++++++++++++++++++++++++++++-
 kernel/sched/stats.h |  30 +++++++++-
 3 files changed, 152 insertions(+), 115 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 500e960..e26d622 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -887,32 +887,27 @@ static void update_curr_fair(struct rq *rq)
 }
 
 static inline void
-update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
+update_stats_wait_start_fair(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
-	u64 wait_start, prev_wait_start;
 	struct sched_statistics *stats;
+	struct task_struct *p = NULL;
 
 	if (!schedstat_enabled())
 		return;
 
 	stats = __schedstats_from_se(se);
 
-	wait_start = rq_clock(rq_of(cfs_rq));
-	prev_wait_start = schedstat_val(stats->wait_start);
-
-	if (entity_is_task(se) && task_on_rq_migrating(task_of(se)) &&
-	    likely(wait_start > prev_wait_start))
-		wait_start -= prev_wait_start;
+	if (entity_is_task(se))
+		p = task_of(se);
 
-	__schedstat_set(stats->wait_start, wait_start);
+	__update_stats_wait_start(rq_of(cfs_rq), p, stats);
 }
 
 static inline void
-update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
+update_stats_wait_end_fair(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
 	struct sched_statistics *stats;
 	struct task_struct *p = NULL;
-	u64 delta;
 
 	if (!schedstat_enabled())
 		return;
@@ -928,105 +923,34 @@ update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	if (unlikely(!schedstat_val(stats->wait_start)))
 		return;
 
-	delta = rq_clock(rq_of(cfs_rq)) - schedstat_val(stats->wait_start);
-
-	if (entity_is_task(se)) {
+	if (entity_is_task(se))
 		p = task_of(se);
-		if (task_on_rq_migrating(p)) {
-			/*
-			 * Preserve migrating task's wait time so wait_start
-			 * time stamp can be adjusted to accumulate wait time
-			 * prior to migration.
-			 */
-			__schedstat_set(stats->wait_start, delta);
-			return;
-		}
-		trace_sched_stat_wait(p, delta);
-	}
 
-	__schedstat_set(stats->wait_max,
-		      max(schedstat_val(stats->wait_max), delta));
-	__schedstat_inc(stats->wait_count);
-	__schedstat_add(stats->wait_sum, delta);
-	__schedstat_set(stats->wait_start, 0);
+	__update_stats_wait_end(rq_of(cfs_rq), p, stats);
 }
 
 static inline void
-update_stats_enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
+update_stats_enqueue_sleeper_fair(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
 	struct sched_statistics *stats;
 	struct task_struct *tsk = NULL;
-	u64 sleep_start, block_start;
 
 	if (!schedstat_enabled())
 		return;
 
 	stats = __schedstats_from_se(se);
 
-	sleep_start = schedstat_val(stats->sleep_start);
-	block_start = schedstat_val(stats->block_start);
-
 	if (entity_is_task(se))
 		tsk = task_of(se);
 
-	if (sleep_start) {
-		u64 delta = rq_clock(rq_of(cfs_rq)) - sleep_start;
-
-		if ((s64)delta < 0)
-			delta = 0;
-
-		if (unlikely(delta > schedstat_val(stats->sleep_max)))
-			__schedstat_set(stats->sleep_max, delta);
-
-		__schedstat_set(stats->sleep_start, 0);
-		__schedstat_add(stats->sum_sleep_runtime, delta);
-
-		if (tsk) {
-			account_scheduler_latency(tsk, delta >> 10, 1);
-			trace_sched_stat_sleep(tsk, delta);
-		}
-	}
-	if (block_start) {
-		u64 delta = rq_clock(rq_of(cfs_rq)) - block_start;
-
-		if ((s64)delta < 0)
-			delta = 0;
-
-		if (unlikely(delta > schedstat_val(stats->block_max)))
-			__schedstat_set(stats->block_max, delta);
-
-		__schedstat_set(stats->block_start, 0);
-		__schedstat_add(stats->sum_sleep_runtime, delta);
-
-		if (tsk) {
-			if (tsk->in_iowait) {
-				__schedstat_add(stats->iowait_sum, delta);
-				__schedstat_inc(stats->iowait_count);
-				trace_sched_stat_iowait(tsk, delta);
-			}
-
-			trace_sched_stat_blocked(tsk, delta);
-
-			/*
-			 * Blocking time is in units of nanosecs, so shift by
-			 * 20 to get a milliseconds-range estimation of the
-			 * amount of time that the task spent sleeping:
-			 */
-			if (unlikely(prof_on == SLEEP_PROFILING)) {
-				profile_hits(SLEEP_PROFILING,
-						(void *)get_wchan(tsk),
-						delta >> 20);
-			}
-			account_scheduler_latency(tsk, delta >> 10, 0);
-		}
-	}
+	__update_stats_enqueue_sleeper(rq_of(cfs_rq), tsk, stats);
 }
 
 /*
  * Task is being enqueued - update stats:
  */
 static inline void
-update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
+update_stats_enqueue_fair(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 {
 	if (!schedstat_enabled())
 		return;
@@ -1036,14 +960,14 @@ update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 	 * a dequeue/enqueue event is a NOP)
 	 */
 	if (se != cfs_rq->curr)
-		update_stats_wait_start(cfs_rq, se);
+		update_stats_wait_start_fair(cfs_rq, se);
 
 	if (flags & ENQUEUE_WAKEUP)
-		update_stats_enqueue_sleeper(cfs_rq, se);
+		update_stats_enqueue_sleeper_fair(cfs_rq, se);
 }
 
 static inline void
-update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
+update_stats_dequeue_fair(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 {
 
 	if (!schedstat_enabled())
@@ -1054,7 +978,7 @@ update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 	 * waiting task:
 	 */
 	if (se != cfs_rq->curr)
-		update_stats_wait_end(cfs_rq, se);
+		update_stats_wait_end_fair(cfs_rq, se);
 
 	if ((flags & DEQUEUE_SLEEP) && entity_is_task(se)) {
 		struct task_struct *tsk = task_of(se);
@@ -4238,26 +4162,6 @@ 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 void check_schedstat_required(void)
-{
-#ifdef CONFIG_SCHEDSTATS
-	if (schedstat_enabled())
-		return;
-
-	/* Force schedstat enabled if a dependent tracepoint is active */
-	if (trace_sched_stat_wait_enabled()    ||
-			trace_sched_stat_sleep_enabled()   ||
-			trace_sched_stat_iowait_enabled()  ||
-			trace_sched_stat_blocked_enabled() ||
-			trace_sched_stat_runtime_enabled())  {
-		printk_deferred_once("Scheduler tracepoints stat_sleep, stat_iowait, "
-			     "stat_blocked and stat_runtime require the "
-			     "kernel parameter schedstats=enable or "
-			     "kernel.sched_schedstats=1\n");
-	}
-#endif
-}
-
 static inline bool cfs_bandwidth_used(void);
 
 /*
@@ -4331,7 +4235,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 		place_entity(cfs_rq, se, 0);
 
 	check_schedstat_required();
-	update_stats_enqueue(cfs_rq, se, flags);
+	update_stats_enqueue_fair(cfs_rq, se, flags);
 	check_spread(cfs_rq, se);
 	if (!curr)
 		__enqueue_entity(cfs_rq, se);
@@ -4415,7 +4319,7 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 	update_load_avg(cfs_rq, se, UPDATE_TG);
 	se_update_runnable(se);
 
-	update_stats_dequeue(cfs_rq, se, flags);
+	update_stats_dequeue_fair(cfs_rq, se, flags);
 
 	clear_buddies(cfs_rq, se);
 
@@ -4500,7 +4404,7 @@ set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 		 * a CPU. So account for the time it spent waiting on the
 		 * runqueue.
 		 */
-		update_stats_wait_end(cfs_rq, se);
+		update_stats_wait_end_fair(cfs_rq, se);
 		__dequeue_entity(cfs_rq, se);
 		update_load_avg(cfs_rq, se, UPDATE_TG);
 	}
@@ -4602,7 +4506,7 @@ static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev)
 	check_spread(cfs_rq, prev);
 
 	if (prev->on_rq) {
-		update_stats_wait_start(cfs_rq, prev);
+		update_stats_wait_start_fair(cfs_rq, prev);
 		/* Put 'current' back into the tree. */
 		__enqueue_entity(cfs_rq, prev);
 		/* in !on_rq case, update occurred at dequeue */
diff --git a/kernel/sched/stats.c b/kernel/sched/stats.c
index 3f93fc3..fad781c 100644
--- a/kernel/sched/stats.c
+++ b/kernel/sched/stats.c
@@ -4,6 +4,109 @@
  */
 #include "sched.h"
 
+void __update_stats_wait_start(struct rq *rq, struct task_struct *p,
+			       struct sched_statistics *stats)
+{
+	u64 wait_start, prev_wait_start;
+
+	wait_start = rq_clock(rq);
+	prev_wait_start = schedstat_val(stats->wait_start);
+
+	if (p && likely(wait_start > prev_wait_start))
+		wait_start -= prev_wait_start;
+
+	__schedstat_set(stats->wait_start, wait_start);
+}
+
+void __update_stats_wait_end(struct rq *rq, struct task_struct *p,
+			     struct sched_statistics *stats)
+{
+	u64 delta = rq_clock(rq) - schedstat_val(stats->wait_start);
+
+	if (p) {
+		if (task_on_rq_migrating(p)) {
+			/*
+			 * Preserve migrating task's wait time so wait_start
+			 * time stamp can be adjusted to accumulate wait time
+			 * prior to migration.
+			 */
+			__schedstat_set(stats->wait_start, delta);
+
+			return;
+		}
+
+		trace_sched_stat_wait(p, delta);
+	}
+
+	__schedstat_set(stats->wait_max,
+			max(schedstat_val(stats->wait_max), delta));
+	__schedstat_inc(stats->wait_count);
+	__schedstat_add(stats->wait_sum, delta);
+	__schedstat_set(stats->wait_start, 0);
+}
+
+void __update_stats_enqueue_sleeper(struct rq *rq, struct task_struct *p,
+				    struct sched_statistics *stats)
+{
+	u64 sleep_start, block_start;
+
+	sleep_start = schedstat_val(stats->sleep_start);
+	block_start = schedstat_val(stats->block_start);
+
+	if (sleep_start) {
+		u64 delta = rq_clock(rq) - sleep_start;
+
+		if ((s64)delta < 0)
+			delta = 0;
+
+		if (unlikely(delta > schedstat_val(stats->sleep_max)))
+			__schedstat_set(stats->sleep_max, delta);
+
+		__schedstat_set(stats->sleep_start, 0);
+		__schedstat_add(stats->sum_sleep_runtime, delta);
+
+		if (p) {
+			account_scheduler_latency(p, delta >> 10, 1);
+			trace_sched_stat_sleep(p, delta);
+		}
+	}
+
+	if (block_start) {
+		u64 delta = rq_clock(rq) - block_start;
+
+		if ((s64)delta < 0)
+			delta = 0;
+
+		if (unlikely(delta > schedstat_val(stats->block_max)))
+			__schedstat_set(stats->block_max, delta);
+
+		__schedstat_set(stats->block_start, 0);
+		__schedstat_add(stats->sum_sleep_runtime, delta);
+
+		if (p) {
+			if (p->in_iowait) {
+				__schedstat_add(stats->iowait_sum, delta);
+				__schedstat_inc(stats->iowait_count);
+				trace_sched_stat_iowait(p, delta);
+			}
+
+			trace_sched_stat_blocked(p, delta);
+
+			/*
+			 * Blocking time is in units of nanosecs, so shift by
+			 * 20 to get a milliseconds-range estimation of the
+			 * amount of time that the task spent sleeping:
+			 */
+			if (unlikely(prof_on == SLEEP_PROFILING)) {
+				profile_hits(SLEEP_PROFILING,
+					     (void *)get_wchan(p),
+					     delta >> 20);
+			}
+			account_scheduler_latency(p, delta >> 10, 0);
+		}
+	}
+}
+
 /*
  * Current schedstat API version.
  *
diff --git a/kernel/sched/stats.h b/kernel/sched/stats.h
index fb6022e..cfb0893 100644
--- a/kernel/sched/stats.h
+++ b/kernel/sched/stats.h
@@ -2,6 +2,8 @@
 
 #ifdef CONFIG_SCHEDSTATS
 
+extern struct static_key_false sched_schedstats;
+
 /*
  * Expects runqueue lock to be held for atomicity of update
  */
@@ -40,6 +42,29 @@ rq_sched_info_dequeue(struct rq *rq, unsigned long long delta)
 #define   schedstat_val(var)		(var)
 #define   schedstat_val_or_zero(var)	((schedstat_enabled()) ? (var) : 0)
 
+void __update_stats_wait_start(struct rq *rq, struct task_struct *p,
+			       struct sched_statistics *stats);
+
+void __update_stats_wait_end(struct rq *rq, struct task_struct *p,
+			     struct sched_statistics *stats);
+void __update_stats_enqueue_sleeper(struct rq *rq, struct task_struct *p,
+				    struct sched_statistics *stats);
+
+static inline void
+check_schedstat_required(void)
+{
+	if (schedstat_enabled())
+		return;
+
+	/* Force schedstat enabled if a dependent tracepoint is active */
+	if (trace_sched_stat_wait_enabled()    ||
+	    trace_sched_stat_sleep_enabled()   ||
+	    trace_sched_stat_iowait_enabled()  ||
+	    trace_sched_stat_blocked_enabled() ||
+	    trace_sched_stat_runtime_enabled())
+		printk_deferred_once("Scheduler tracepoints stat_sleep, stat_iowait, stat_blocked and stat_runtime require the kernel parameter schedstats=enable or kernel.sched_schedstats=1\n");
+}
+
 #else /* !CONFIG_SCHEDSTATS: */
 
 static inline void rq_sched_info_arrive  (struct rq *rq, unsigned long long delta) { }
@@ -55,6 +80,11 @@ static inline void rq_sched_info_depart  (struct rq *rq, unsigned long long delt
 # define   schedstat_val(var)		0
 # define   schedstat_val_or_zero(var)	0
 
+# define __update_stats_wait_start(rq, p, stats)       do { } while (0)
+# define __update_stats_wait_end(rq, p, stats)         do { } while (0)
+# define __update_stats_enqueue_sleeper(rq, p, stats)  do { } while (0)
+# define check_schedstat_required()                    do { } while (0)
+
 #endif /* CONFIG_SCHEDSTATS */
 
 #ifdef CONFIG_FAIR_GROUP_SCHED

^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched: Make struct sched_statistics independent of fair sched class
  2021-09-05 14:35 ` [PATCH v4 2/8] sched: make struct sched_statistics independent of fair sched class Yafang Shao
@ 2021-09-16 11:59   ` tip-bot2 for Yafang Shao
  2021-09-21  6:17     ` Dan Carpenter
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 1 reply; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-09-16 11:59 UTC (permalink / raw)
  To: linux-tip-commits
  Cc: Yafang Shao, Peter Zijlstra (Intel), Mel Gorman, x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     cb3e971c435d3a40af4d7ba15ffcc1e81910c8b7
Gitweb:        https://git.kernel.org/tip/cb3e971c435d3a40af4d7ba15ffcc1e81910c8b7
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:41 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Wed, 15 Sep 2021 17:48:58 +02:00

sched: Make struct sched_statistics independent of fair sched class

If we want to use the schedstats facility to trace other sched classes, we
should make it independent of fair sched class. The struct sched_statistics
is the schedular statistics of a task_struct or a task_group. So we can
move it into struct task_struct and struct task_group to achieve the goal.

After the patch, schestats are orgnized as follows,

    struct task_struct {
       ...
       struct sched_entity se;
       struct sched_rt_entity rt;
       struct sched_dl_entity dl;
       ...
       struct sched_statistics stats;
       ...
   };

Regarding the task group, schedstats is only supported for fair group
sched, and a new struct sched_entity_stats is introduced, suggested by
Peter -

    struct sched_entity_stats {
        struct sched_entity     se;
        struct sched_statistics stats;
    } __no_randomize_layout;

Then with the se in a task_group, we can easily get the stats.

The sched_statistics members may be frequently modified when schedstats is
enabled, in order to avoid impacting on random data which may in the same
cacheline with them, the struct sched_statistics is defined as cacheline
aligned.

As this patch changes the core struct of scheduler, so I verified the
performance it may impact on the scheduler with 'perf bench sched
pipe', suggested by Mel. Below is the result, in which all the values
are in usecs/op.
                                  Before               After
      kernel.sched_schedstats=0  5.2~5.4               5.2~5.4
      kernel.sched_schedstats=1  5.3~5.5               5.3~5.5
[These data is a little difference with the earlier version, that is
 because my old test machine is destroyed so I have to use a new
 different test machine.]

Almost no impact on the sched performance.

No functional change.

[lkp@intel.com: reported build failure in earlier version]

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lore.kernel.org/r/20210905143547.4668-3-laoar.shao@gmail.com
---
 include/linux/sched.h    |  6 +--
 kernel/sched/core.c      | 25 ++++++-----
 kernel/sched/deadline.c  |  4 +-
 kernel/sched/debug.c     | 92 ++++++++++++++++++++-------------------
 kernel/sched/fair.c      | 89 ++++++++++++++++++++++----------------
 kernel/sched/rt.c        |  4 +-
 kernel/sched/stats.h     | 19 ++++++++-
 kernel/sched/stop_task.c |  4 +-
 8 files changed, 143 insertions(+), 100 deletions(-)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 6421a9a..518e19c 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -460,7 +460,7 @@ struct sched_statistics {
 	u64				nr_wakeups_passive;
 	u64				nr_wakeups_idle;
 #endif
-};
+} ____cacheline_aligned;
 
 struct sched_entity {
 	/* For load-balancing: */
@@ -476,8 +476,6 @@ struct sched_entity {
 
 	u64				nr_migrations;
 
-	struct sched_statistics		statistics;
-
 #ifdef CONFIG_FAIR_GROUP_SCHED
 	int				depth;
 	struct sched_entity		*parent;
@@ -736,6 +734,8 @@ struct task_struct {
 	struct uclamp_se		uclamp[UCLAMP_CNT];
 #endif
 
+	struct sched_statistics         stats;
+
 #ifdef CONFIG_PREEMPT_NOTIFIERS
 	/* List of struct preempt_notifier: */
 	struct hlist_head		preempt_notifiers;
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 2877138..85e212d 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -3472,11 +3472,11 @@ ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
 #ifdef CONFIG_SMP
 	if (cpu == rq->cpu) {
 		__schedstat_inc(rq->ttwu_local);
-		__schedstat_inc(p->se.statistics.nr_wakeups_local);
+		__schedstat_inc(p->stats.nr_wakeups_local);
 	} else {
 		struct sched_domain *sd;
 
-		__schedstat_inc(p->se.statistics.nr_wakeups_remote);
+		__schedstat_inc(p->stats.nr_wakeups_remote);
 		rcu_read_lock();
 		for_each_domain(rq->cpu, sd) {
 			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
@@ -3488,14 +3488,14 @@ ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
 	}
 
 	if (wake_flags & WF_MIGRATED)
-		__schedstat_inc(p->se.statistics.nr_wakeups_migrate);
+		__schedstat_inc(p->stats.nr_wakeups_migrate);
 #endif /* CONFIG_SMP */
 
 	__schedstat_inc(rq->ttwu_count);
-	__schedstat_inc(p->se.statistics.nr_wakeups);
+	__schedstat_inc(p->stats.nr_wakeups);
 
 	if (wake_flags & WF_SYNC)
-		__schedstat_inc(p->se.statistics.nr_wakeups_sync);
+		__schedstat_inc(p->stats.nr_wakeups_sync);
 }
 
 /*
@@ -4134,7 +4134,7 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
 
 #ifdef CONFIG_SCHEDSTATS
 	/* Even if schedstat is disabled, there should not be garbage */
-	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
+	memset(&p->stats, 0, sizeof(p->stats));
 #endif
 
 	RB_CLEAR_NODE(&p->dl.rb_node);
@@ -9391,9 +9391,9 @@ void normalize_rt_tasks(void)
 			continue;
 
 		p->se.exec_start = 0;
-		schedstat_set(p->se.statistics.wait_start,  0);
-		schedstat_set(p->se.statistics.sleep_start, 0);
-		schedstat_set(p->se.statistics.block_start, 0);
+		schedstat_set(p->stats.wait_start,  0);
+		schedstat_set(p->stats.sleep_start, 0);
+		schedstat_set(p->stats.block_start, 0);
 
 		if (!dl_task(p) && !rt_task(p)) {
 			/*
@@ -10235,11 +10235,14 @@ 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) {
+		struct sched_statistics *stats;
 		u64 ws = 0;
 		int i;
 
-		for_each_possible_cpu(i)
-			ws += schedstat_val(tg->se[i]->statistics.wait_sum);
+		for_each_possible_cpu(i) {
+			stats = __schedstats_from_se(tg->se[i]);
+			ws += schedstat_val(stats->wait_sum);
+		}
 
 		seq_printf(sf, "wait_sum %llu\n", ws);
 	}
diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index e943146..51dd309 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -1265,8 +1265,8 @@ static void update_curr_dl(struct rq *rq)
 		return;
 	}
 
-	schedstat_set(curr->se.statistics.exec_max,
-		      max(curr->se.statistics.exec_max, delta_exec));
+	schedstat_set(curr->stats.exec_max,
+		      max(curr->stats.exec_max, delta_exec));
 
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);
diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index 317ef56..ca0dd47 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -443,9 +443,11 @@ static void print_cfs_group_stats(struct seq_file *m, int cpu, struct task_group
 	struct sched_entity *se = tg->se[cpu];
 
 #define P(F)		SEQ_printf(m, "  .%-30s: %lld\n",	#F, (long long)F)
-#define P_SCHEDSTAT(F)	SEQ_printf(m, "  .%-30s: %lld\n",	#F, (long long)schedstat_val(F))
+#define P_SCHEDSTAT(F)	SEQ_printf(m, "  .%-30s: %lld\n",	\
+		#F, (long long)schedstat_val(stats->F))
 #define PN(F)		SEQ_printf(m, "  .%-30s: %lld.%06ld\n", #F, SPLIT_NS((long long)F))
-#define PN_SCHEDSTAT(F)	SEQ_printf(m, "  .%-30s: %lld.%06ld\n", #F, SPLIT_NS((long long)schedstat_val(F)))
+#define PN_SCHEDSTAT(F)	SEQ_printf(m, "  .%-30s: %lld.%06ld\n", \
+		#F, SPLIT_NS((long long)schedstat_val(stats->F)))
 
 	if (!se)
 		return;
@@ -455,16 +457,18 @@ static void print_cfs_group_stats(struct seq_file *m, int cpu, struct task_group
 	PN(se->sum_exec_runtime);
 
 	if (schedstat_enabled()) {
-		PN_SCHEDSTAT(se->statistics.wait_start);
-		PN_SCHEDSTAT(se->statistics.sleep_start);
-		PN_SCHEDSTAT(se->statistics.block_start);
-		PN_SCHEDSTAT(se->statistics.sleep_max);
-		PN_SCHEDSTAT(se->statistics.block_max);
-		PN_SCHEDSTAT(se->statistics.exec_max);
-		PN_SCHEDSTAT(se->statistics.slice_max);
-		PN_SCHEDSTAT(se->statistics.wait_max);
-		PN_SCHEDSTAT(se->statistics.wait_sum);
-		P_SCHEDSTAT(se->statistics.wait_count);
+               struct sched_statistics *stats =  __schedstats_from_se(se);
+
+		PN_SCHEDSTAT(wait_start);
+		PN_SCHEDSTAT(sleep_start);
+		PN_SCHEDSTAT(block_start);
+		PN_SCHEDSTAT(sleep_max);
+		PN_SCHEDSTAT(block_max);
+		PN_SCHEDSTAT(exec_max);
+		PN_SCHEDSTAT(slice_max);
+		PN_SCHEDSTAT(wait_max);
+		PN_SCHEDSTAT(wait_sum);
+		P_SCHEDSTAT(wait_count);
 	}
 
 	P(se->load.weight);
@@ -531,9 +535,9 @@ print_task(struct seq_file *m, struct rq *rq, struct task_struct *p)
 		p->prio);
 
 	SEQ_printf(m, "%9Ld.%06ld %9Ld.%06ld %9Ld.%06ld",
-		SPLIT_NS(schedstat_val_or_zero(p->se.statistics.wait_sum)),
+		SPLIT_NS(schedstat_val_or_zero(p->stats.wait_sum)),
 		SPLIT_NS(p->se.sum_exec_runtime),
-		SPLIT_NS(schedstat_val_or_zero(p->se.statistics.sum_sleep_runtime)));
+		SPLIT_NS(schedstat_val_or_zero(p->stats.sum_sleep_runtime)));
 
 #ifdef CONFIG_NUMA_BALANCING
 	SEQ_printf(m, " %d %d", task_node(p), task_numa_group_id(p));
@@ -952,8 +956,8 @@ void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
 		"---------------------------------------------------------"
 		"----------\n");
 
-#define P_SCHEDSTAT(F)  __PS(#F, schedstat_val(p->F))
-#define PN_SCHEDSTAT(F) __PSN(#F, schedstat_val(p->F))
+#define P_SCHEDSTAT(F)  __PS(#F, schedstat_val(p->stats.F))
+#define PN_SCHEDSTAT(F) __PSN(#F, schedstat_val(p->stats.F))
 
 	PN(se.exec_start);
 	PN(se.vruntime);
@@ -966,33 +970,33 @@ void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
 	if (schedstat_enabled()) {
 		u64 avg_atom, avg_per_cpu;
 
-		PN_SCHEDSTAT(se.statistics.sum_sleep_runtime);
-		PN_SCHEDSTAT(se.statistics.wait_start);
-		PN_SCHEDSTAT(se.statistics.sleep_start);
-		PN_SCHEDSTAT(se.statistics.block_start);
-		PN_SCHEDSTAT(se.statistics.sleep_max);
-		PN_SCHEDSTAT(se.statistics.block_max);
-		PN_SCHEDSTAT(se.statistics.exec_max);
-		PN_SCHEDSTAT(se.statistics.slice_max);
-		PN_SCHEDSTAT(se.statistics.wait_max);
-		PN_SCHEDSTAT(se.statistics.wait_sum);
-		P_SCHEDSTAT(se.statistics.wait_count);
-		PN_SCHEDSTAT(se.statistics.iowait_sum);
-		P_SCHEDSTAT(se.statistics.iowait_count);
-		P_SCHEDSTAT(se.statistics.nr_migrations_cold);
-		P_SCHEDSTAT(se.statistics.nr_failed_migrations_affine);
-		P_SCHEDSTAT(se.statistics.nr_failed_migrations_running);
-		P_SCHEDSTAT(se.statistics.nr_failed_migrations_hot);
-		P_SCHEDSTAT(se.statistics.nr_forced_migrations);
-		P_SCHEDSTAT(se.statistics.nr_wakeups);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_sync);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_migrate);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_local);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_remote);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_affine);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_affine_attempts);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_passive);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_idle);
+		PN_SCHEDSTAT(sum_sleep_runtime);
+		PN_SCHEDSTAT(wait_start);
+		PN_SCHEDSTAT(sleep_start);
+		PN_SCHEDSTAT(block_start);
+		PN_SCHEDSTAT(sleep_max);
+		PN_SCHEDSTAT(block_max);
+		PN_SCHEDSTAT(exec_max);
+		PN_SCHEDSTAT(slice_max);
+		PN_SCHEDSTAT(wait_max);
+		PN_SCHEDSTAT(wait_sum);
+		P_SCHEDSTAT(wait_count);
+		PN_SCHEDSTAT(iowait_sum);
+		P_SCHEDSTAT(iowait_count);
+		P_SCHEDSTAT(nr_migrations_cold);
+		P_SCHEDSTAT(nr_failed_migrations_affine);
+		P_SCHEDSTAT(nr_failed_migrations_running);
+		P_SCHEDSTAT(nr_failed_migrations_hot);
+		P_SCHEDSTAT(nr_forced_migrations);
+		P_SCHEDSTAT(nr_wakeups);
+		P_SCHEDSTAT(nr_wakeups_sync);
+		P_SCHEDSTAT(nr_wakeups_migrate);
+		P_SCHEDSTAT(nr_wakeups_local);
+		P_SCHEDSTAT(nr_wakeups_remote);
+		P_SCHEDSTAT(nr_wakeups_affine);
+		P_SCHEDSTAT(nr_wakeups_affine_attempts);
+		P_SCHEDSTAT(nr_wakeups_passive);
+		P_SCHEDSTAT(nr_wakeups_idle);
 
 		avg_atom = p->se.sum_exec_runtime;
 		if (nr_switches)
@@ -1058,7 +1062,7 @@ void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
 void proc_sched_set_task(struct task_struct *p)
 {
 #ifdef CONFIG_SCHEDSTATS
-	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
+	memset(&p->stats, 0, sizeof(p->stats));
 #endif
 }
 
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 148b830..500e960 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -856,8 +856,13 @@ static void update_curr(struct cfs_rq *cfs_rq)
 
 	curr->exec_start = now;
 
-	schedstat_set(curr->statistics.exec_max,
-		      max(delta_exec, curr->statistics.exec_max));
+	if (schedstat_enabled()) {
+		struct sched_statistics *stats;
+
+		stats = __schedstats_from_se(curr);
+		__schedstat_set(stats->exec_max,
+				max(delta_exec, stats->exec_max));
+	}
 
 	curr->sum_exec_runtime += delta_exec;
 	schedstat_add(cfs_rq->exec_clock, delta_exec);
@@ -885,39 +890,45 @@ static inline void
 update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
 	u64 wait_start, prev_wait_start;
+	struct sched_statistics *stats;
 
 	if (!schedstat_enabled())
 		return;
 
+	stats = __schedstats_from_se(se);
+
 	wait_start = rq_clock(rq_of(cfs_rq));
-	prev_wait_start = schedstat_val(se->statistics.wait_start);
+	prev_wait_start = schedstat_val(stats->wait_start);
 
 	if (entity_is_task(se) && task_on_rq_migrating(task_of(se)) &&
 	    likely(wait_start > prev_wait_start))
 		wait_start -= prev_wait_start;
 
-	__schedstat_set(se->statistics.wait_start, wait_start);
+	__schedstat_set(stats->wait_start, wait_start);
 }
 
 static inline void
 update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
-	struct task_struct *p;
+	struct sched_statistics *stats;
+	struct task_struct *p = NULL;
 	u64 delta;
 
 	if (!schedstat_enabled())
 		return;
 
+	stats = __schedstats_from_se(se);
+
 	/*
 	 * When the sched_schedstat changes from 0 to 1, some sched se
 	 * maybe already in the runqueue, the se->statistics.wait_start
 	 * will be 0.So it will let the delta wrong. We need to avoid this
 	 * scenario.
 	 */
-	if (unlikely(!schedstat_val(se->statistics.wait_start)))
+	if (unlikely(!schedstat_val(stats->wait_start)))
 		return;
 
-	delta = rq_clock(rq_of(cfs_rq)) - schedstat_val(se->statistics.wait_start);
+	delta = rq_clock(rq_of(cfs_rq)) - schedstat_val(stats->wait_start);
 
 	if (entity_is_task(se)) {
 		p = task_of(se);
@@ -927,30 +938,33 @@ update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
 			 * time stamp can be adjusted to accumulate wait time
 			 * prior to migration.
 			 */
-			__schedstat_set(se->statistics.wait_start, delta);
+			__schedstat_set(stats->wait_start, delta);
 			return;
 		}
 		trace_sched_stat_wait(p, delta);
 	}
 
-	__schedstat_set(se->statistics.wait_max,
-		      max(schedstat_val(se->statistics.wait_max), delta));
-	__schedstat_inc(se->statistics.wait_count);
-	__schedstat_add(se->statistics.wait_sum, delta);
-	__schedstat_set(se->statistics.wait_start, 0);
+	__schedstat_set(stats->wait_max,
+		      max(schedstat_val(stats->wait_max), delta));
+	__schedstat_inc(stats->wait_count);
+	__schedstat_add(stats->wait_sum, delta);
+	__schedstat_set(stats->wait_start, 0);
 }
 
 static inline void
 update_stats_enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
+	struct sched_statistics *stats;
 	struct task_struct *tsk = NULL;
 	u64 sleep_start, block_start;
 
 	if (!schedstat_enabled())
 		return;
 
-	sleep_start = schedstat_val(se->statistics.sleep_start);
-	block_start = schedstat_val(se->statistics.block_start);
+	stats = __schedstats_from_se(se);
+
+	sleep_start = schedstat_val(stats->sleep_start);
+	block_start = schedstat_val(stats->block_start);
 
 	if (entity_is_task(se))
 		tsk = task_of(se);
@@ -961,11 +975,11 @@ update_stats_enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
 		if ((s64)delta < 0)
 			delta = 0;
 
-		if (unlikely(delta > schedstat_val(se->statistics.sleep_max)))
-			__schedstat_set(se->statistics.sleep_max, delta);
+		if (unlikely(delta > schedstat_val(stats->sleep_max)))
+			__schedstat_set(stats->sleep_max, delta);
 
-		__schedstat_set(se->statistics.sleep_start, 0);
-		__schedstat_add(se->statistics.sum_sleep_runtime, delta);
+		__schedstat_set(stats->sleep_start, 0);
+		__schedstat_add(stats->sum_sleep_runtime, delta);
 
 		if (tsk) {
 			account_scheduler_latency(tsk, delta >> 10, 1);
@@ -978,16 +992,16 @@ update_stats_enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
 		if ((s64)delta < 0)
 			delta = 0;
 
-		if (unlikely(delta > schedstat_val(se->statistics.block_max)))
-			__schedstat_set(se->statistics.block_max, delta);
+		if (unlikely(delta > schedstat_val(stats->block_max)))
+			__schedstat_set(stats->block_max, delta);
 
-		__schedstat_set(se->statistics.block_start, 0);
-		__schedstat_add(se->statistics.sum_sleep_runtime, delta);
+		__schedstat_set(stats->block_start, 0);
+		__schedstat_add(stats->sum_sleep_runtime, delta);
 
 		if (tsk) {
 			if (tsk->in_iowait) {
-				__schedstat_add(se->statistics.iowait_sum, delta);
-				__schedstat_inc(se->statistics.iowait_count);
+				__schedstat_add(stats->iowait_sum, delta);
+				__schedstat_inc(stats->iowait_count);
 				trace_sched_stat_iowait(tsk, delta);
 			}
 
@@ -1049,10 +1063,10 @@ update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 		/* XXX racy against TTWU */
 		state = READ_ONCE(tsk->__state);
 		if (state & TASK_INTERRUPTIBLE)
-			__schedstat_set(se->statistics.sleep_start,
+			__schedstat_set(tsk->stats.sleep_start,
 				      rq_clock(rq_of(cfs_rq)));
 		if (state & TASK_UNINTERRUPTIBLE)
-			__schedstat_set(se->statistics.block_start,
+			__schedstat_set(tsk->stats.block_start,
 				      rq_clock(rq_of(cfs_rq)));
 	}
 }
@@ -4501,8 +4515,11 @@ set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	 */
 	if (schedstat_enabled() &&
 	    rq_of(cfs_rq)->cfs.load.weight >= 2*se->load.weight) {
-		__schedstat_set(se->statistics.slice_max,
-				max((u64)se->statistics.slice_max,
+		struct sched_statistics *stats;
+
+		stats = __schedstats_from_se(se);
+		__schedstat_set(stats->slice_max,
+				max((u64)stats->slice_max,
 				    se->sum_exec_runtime - se->prev_sum_exec_runtime));
 	}
 
@@ -6013,12 +6030,12 @@ static int wake_affine(struct sched_domain *sd, struct task_struct *p,
 	if (sched_feat(WA_WEIGHT) && target == nr_cpumask_bits)
 		target = wake_affine_weight(sd, p, this_cpu, prev_cpu, sync);
 
-	schedstat_inc(p->se.statistics.nr_wakeups_affine_attempts);
+	schedstat_inc(p->stats.nr_wakeups_affine_attempts);
 	if (target == nr_cpumask_bits)
 		return prev_cpu;
 
 	schedstat_inc(sd->ttwu_move_affine);
-	schedstat_inc(p->se.statistics.nr_wakeups_affine);
+	schedstat_inc(p->stats.nr_wakeups_affine);
 	return target;
 }
 
@@ -7822,7 +7839,7 @@ int can_migrate_task(struct task_struct *p, struct lb_env *env)
 	if (!cpumask_test_cpu(env->dst_cpu, p->cpus_ptr)) {
 		int cpu;
 
-		schedstat_inc(p->se.statistics.nr_failed_migrations_affine);
+		schedstat_inc(p->stats.nr_failed_migrations_affine);
 
 		env->flags |= LBF_SOME_PINNED;
 
@@ -7856,7 +7873,7 @@ int can_migrate_task(struct task_struct *p, struct lb_env *env)
 	env->flags &= ~LBF_ALL_PINNED;
 
 	if (task_running(env->src_rq, p)) {
-		schedstat_inc(p->se.statistics.nr_failed_migrations_running);
+		schedstat_inc(p->stats.nr_failed_migrations_running);
 		return 0;
 	}
 
@@ -7878,12 +7895,12 @@ int can_migrate_task(struct task_struct *p, struct lb_env *env)
 	    env->sd->nr_balance_failed > env->sd->cache_nice_tries) {
 		if (tsk_cache_hot == 1) {
 			schedstat_inc(env->sd->lb_hot_gained[env->idle]);
-			schedstat_inc(p->se.statistics.nr_forced_migrations);
+			schedstat_inc(p->stats.nr_forced_migrations);
 		}
 		return 1;
 	}
 
-	schedstat_inc(p->se.statistics.nr_failed_migrations_hot);
+	schedstat_inc(p->stats.nr_failed_migrations_hot);
 	return 0;
 }
 
@@ -11424,7 +11441,7 @@ int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
 		if (!cfs_rq)
 			goto err;
 
-		se = kzalloc_node(sizeof(struct sched_entity),
+		se = kzalloc_node(sizeof(struct sched_entity_stats),
 				  GFP_KERNEL, cpu_to_node(i));
 		if (!se)
 			goto err_free_rq;
diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 3daf42a..95a7c3a 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1009,8 +1009,8 @@ static void update_curr_rt(struct rq *rq)
 	if (unlikely((s64)delta_exec <= 0))
 		return;
 
-	schedstat_set(curr->se.statistics.exec_max,
-		      max(curr->se.statistics.exec_max, delta_exec));
+	schedstat_set(curr->stats.exec_max,
+		      max(curr->stats.exec_max, delta_exec));
 
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);
diff --git a/kernel/sched/stats.h b/kernel/sched/stats.h
index d8f8eb0..fb6022e 100644
--- a/kernel/sched/stats.h
+++ b/kernel/sched/stats.h
@@ -41,6 +41,7 @@ rq_sched_info_dequeue(struct rq *rq, unsigned long long delta)
 #define   schedstat_val_or_zero(var)	((schedstat_enabled()) ? (var) : 0)
 
 #else /* !CONFIG_SCHEDSTATS: */
+
 static inline void rq_sched_info_arrive  (struct rq *rq, unsigned long long delta) { }
 static inline void rq_sched_info_dequeue(struct rq *rq, unsigned long long delta) { }
 static inline void rq_sched_info_depart  (struct rq *rq, unsigned long long delta) { }
@@ -53,8 +54,26 @@ static inline void rq_sched_info_depart  (struct rq *rq, unsigned long long delt
 # define   schedstat_set(var, val)	do { } while (0)
 # define   schedstat_val(var)		0
 # define   schedstat_val_or_zero(var)	0
+
 #endif /* CONFIG_SCHEDSTATS */
 
+#ifdef CONFIG_FAIR_GROUP_SCHED
+struct sched_entity_stats {
+	struct sched_entity     se;
+	struct sched_statistics stats;
+} __no_randomize_layout;
+#endif
+
+static inline struct sched_statistics *
+__schedstats_from_se(struct sched_entity *se)
+{
+#ifdef CONFIG_FAIR_GROUP_SCHED
+	if (!entity_is_task(se))
+		return &container_of(se, struct sched_entity_stats, se)->stats;
+#endif
+	return &task_of(se)->stats;
+}
+
 #ifdef CONFIG_PSI
 /*
  * PSI tracks state that persists across sleeps, such as iowaits and
diff --git a/kernel/sched/stop_task.c b/kernel/sched/stop_task.c
index f988ebe..0b165a2 100644
--- a/kernel/sched/stop_task.c
+++ b/kernel/sched/stop_task.c
@@ -78,8 +78,8 @@ static void put_prev_task_stop(struct rq *rq, struct task_struct *prev)
 	if (unlikely((s64)delta_exec < 0))
 		delta_exec = 0;
 
-	schedstat_set(curr->se.statistics.exec_max,
-			max(curr->se.statistics.exec_max, delta_exec));
+	schedstat_set(curr->stats.exec_max,
+		      max(curr->stats.exec_max, delta_exec));
 
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);

^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched/fair: Use __schedstat_set() in set_next_entity()
  2021-09-05 14:35 ` [PATCH v4 1/8] sched, fair: use __schedstat_set() in set_next_entity() Yafang Shao
@ 2021-09-16 11:59   ` tip-bot2 for Yafang Shao
  2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 0 replies; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-09-16 11:59 UTC (permalink / raw)
  To: linux-tip-commits
  Cc: Yafang Shao, Peter Zijlstra (Intel), Mel Gorman, x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     5855e81a4a3b6eb8967bff760e7d1f1b82228525
Gitweb:        https://git.kernel.org/tip/5855e81a4a3b6eb8967bff760e7d1f1b82228525
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:40 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Wed, 15 Sep 2021 17:48:58 +02:00

sched/fair: Use __schedstat_set() in set_next_entity()

schedstat_enabled() has been already checked, so we can use
__schedstat_set() directly.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lore.kernel.org/r/20210905143547.4668-2-laoar.shao@gmail.com
---
 kernel/sched/fair.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 3594884..148b830 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4501,9 +4501,9 @@ set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	 */
 	if (schedstat_enabled() &&
 	    rq_of(cfs_rq)->cfs.load.weight >= 2*se->load.weight) {
-		schedstat_set(se->statistics.slice_max,
-			max((u64)schedstat_val(se->statistics.slice_max),
-			    se->sum_exec_runtime - se->prev_sum_exec_runtime));
+		__schedstat_set(se->statistics.slice_max,
+				max((u64)se->statistics.slice_max,
+				    se->sum_exec_runtime - se->prev_sum_exec_runtime));
 	}
 
 	se->prev_sum_exec_runtime = se->sum_exec_runtime;

^ permalink raw reply related	[flat|nested] 29+ messages in thread

* Re: [tip: sched/core] sched: Make struct sched_statistics independent of fair sched class
  2021-09-16 11:59   ` [tip: sched/core] sched: Make " tip-bot2 for Yafang Shao
@ 2021-09-21  6:17     ` Dan Carpenter
  2021-09-21  7:17       ` Peter Zijlstra
  0 siblings, 1 reply; 29+ messages in thread
From: Dan Carpenter @ 2021-09-21  6:17 UTC (permalink / raw)
  To: linux-kernel
  Cc: linux-tip-commits, Yafang Shao, Peter Zijlstra (Intel), Mel Gorman, x86

On Thu, Sep 16, 2021 at 11:59:30AM -0000, tip-bot2 for Yafang Shao wrote:
> @@ -11424,7 +11441,7 @@ int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
>  		if (!cfs_rq)
>  			goto err;
>  
> -		se = kzalloc_node(sizeof(struct sched_entity),
> +		se = kzalloc_node(sizeof(struct sched_entity_stats),

This wasn't there in the original patch and it causes a Smatch warning
because "se" is declared as a "sched_entity" but it's allocating a
larger "sched_entity_stats" which contains a sched_entity.

To me, ideally, we would update the type of se.

>  				  GFP_KERNEL, cpu_to_node(i));
>  		if (!se)
>  			goto err_free_rq;

regards,
dan carpenter


^ permalink raw reply	[flat|nested] 29+ messages in thread

* Re: [tip: sched/core] sched: Make struct sched_statistics independent of fair sched class
  2021-09-21  6:17     ` Dan Carpenter
@ 2021-09-21  7:17       ` Peter Zijlstra
  2021-09-21  7:40         ` Dan Carpenter
  0 siblings, 1 reply; 29+ messages in thread
From: Peter Zijlstra @ 2021-09-21  7:17 UTC (permalink / raw)
  To: Dan Carpenter
  Cc: linux-kernel, linux-tip-commits, Yafang Shao, Mel Gorman, x86

On Tue, Sep 21, 2021 at 09:17:27AM +0300, Dan Carpenter wrote:
> On Thu, Sep 16, 2021 at 11:59:30AM -0000, tip-bot2 for Yafang Shao wrote:
> > @@ -11424,7 +11441,7 @@ int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
> >  		if (!cfs_rq)
> >  			goto err;
> >  
> > -		se = kzalloc_node(sizeof(struct sched_entity),
> > +		se = kzalloc_node(sizeof(struct sched_entity_stats),
> 
> This wasn't there in the original patch and it causes a Smatch warning

What original patch? It's part of the v4 posting.

> because "se" is declared as a "sched_entity" but it's allocating a
> larger "sched_entity_stats" which contains a sched_entity.

Yep, on purpose.

> To me, ideally, we would update the type of se.

That's a lot of churn for very little gain. I can rewrite it like:

	struct sched_entity_stats *ses = kzalloc_node(sizeof(*ses),...);
	se = &ses->se;

If that makes smatch happy. It's the exact same thing tho because we
force ses->se to be at 0 offset.

^ permalink raw reply	[flat|nested] 29+ messages in thread

* Re: [tip: sched/core] sched: Make struct sched_statistics independent of fair sched class
  2021-09-21  7:17       ` Peter Zijlstra
@ 2021-09-21  7:40         ` Dan Carpenter
  0 siblings, 0 replies; 29+ messages in thread
From: Dan Carpenter @ 2021-09-21  7:40 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: linux-kernel, linux-tip-commits, Yafang Shao, Mel Gorman, x86

On Tue, Sep 21, 2021 at 09:17:07AM +0200, Peter Zijlstra wrote:
> On Tue, Sep 21, 2021 at 09:17:27AM +0300, Dan Carpenter wrote:
> > On Thu, Sep 16, 2021 at 11:59:30AM -0000, tip-bot2 for Yafang Shao wrote:
> > > @@ -11424,7 +11441,7 @@ int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
> > >  		if (!cfs_rq)
> > >  			goto err;
> > >  
> > > -		se = kzalloc_node(sizeof(struct sched_entity),
> > > +		se = kzalloc_node(sizeof(struct sched_entity_stats),
> > 
> > This wasn't there in the original patch and it causes a Smatch warning
> 
> What original patch? It's part of the v4 posting.
> 
> > because "se" is declared as a "sched_entity" but it's allocating a
> > larger "sched_entity_stats" which contains a sched_entity.
> 
> Yep, on purpose.
> 
> > To me, ideally, we would update the type of se.
> 
> That's a lot of churn for very little gain. I can rewrite it like:
> 
> 	struct sched_entity_stats *ses = kzalloc_node(sizeof(*ses),...);
> 	se = &ses->se;
> 
> If that makes smatch happy. It's the exact same thing tho because we
> force ses->se to be at 0 offset.

I mean, I understood what the allocation was doing and that this was a
way to avoid churn.  But it's pretty confusing, because every "se" is
really an "ses" now right?

Anyway, just leave it.  It's fine.

regards,
dan carpenter

^ permalink raw reply	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched/dl: Support schedstats for deadline sched class
  2021-09-05 14:35 ` [PATCH v4 8/8] sched, dl: support schedstats " Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched/dl: Support " tip-bot2 for Yafang Shao
@ 2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 0 replies; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-10-05 14:12 UTC (permalink / raw)
  To: linux-tip-commits; +Cc: Yafang Shao, Peter Zijlstra (Intel), x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     b5eb4a5f6521d58d5564b8746701bd67a92a2b11
Gitweb:        https://git.kernel.org/tip/b5eb4a5f6521d58d5564b8746701bd67a92a2b11
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:47 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Oct 2021 15:51:53 +02:00

sched/dl: Support schedstats for deadline sched class

After we make the struct sched_statistics and the helpers of it
independent of fair sched class, we can easily use the schedstats
facility for deadline sched class.

The schedstat usage in DL sched class is similar with fair sched class,
for example,
                    fair                        deadline
    enqueue         update_stats_enqueue_fair   update_stats_enqueue_dl
    dequeue         update_stats_dequeue_fair   update_stats_dequeue_dl
    put_prev_task   update_stats_wait_start     update_stats_wait_start_dl
    set_next_task   update_stats_wait_end       update_stats_wait_end_dl

The user can get the schedstats information in the same way in fair sched
class. For example,
           fair                            deadline
           /proc/[pid]/sched               /proc/[pid]/sched

The output of a deadline task's schedstats as follows,

$ cat /proc/69662/sched
...
se.sum_exec_runtime                          :          3067.696449
se.nr_migrations                             :                    0
sum_sleep_runtime                            :        720144.029661
sum_block_runtime                            :             0.547853
wait_start                                   :             0.000000
sleep_start                                  :      14131540.828955
block_start                                  :             0.000000
sleep_max                                    :          2999.974045
block_max                                    :             0.283637
exec_max                                     :             1.000269
slice_max                                    :             0.000000
wait_max                                     :             0.002217
wait_sum                                     :             0.762179
wait_count                                   :                  733
iowait_sum                                   :             0.547853
iowait_count                                 :                    3
nr_migrations_cold                           :                    0
nr_failed_migrations_affine                  :                    0
nr_failed_migrations_running                 :                    0
nr_failed_migrations_hot                     :                    0
nr_forced_migrations                         :                    0
nr_wakeups                                   :                  246
nr_wakeups_sync                              :                    2
nr_wakeups_migrate                           :                    0
nr_wakeups_local                             :                  244
nr_wakeups_remote                            :                    2
nr_wakeups_affine                            :                    0
nr_wakeups_affine_attempts                   :                    0
nr_wakeups_passive                           :                    0
nr_wakeups_idle                              :                    0
...

The sched:sched_stat_{wait, sleep, iowait, blocked} tracepoints can
be used to trace deadlline tasks as well.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210905143547.4668-9-laoar.shao@gmail.com
---
 kernel/sched/deadline.c | 93 ++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 93 insertions(+)

diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index 73fb33e..d2c072b 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -1474,6 +1474,82 @@ static inline bool __dl_less(struct rb_node *a, const struct rb_node *b)
 	return dl_time_before(__node_2_dle(a)->deadline, __node_2_dle(b)->deadline);
 }
 
+static inline struct sched_statistics *
+__schedstats_from_dl_se(struct sched_dl_entity *dl_se)
+{
+	return &dl_task_of(dl_se)->stats;
+}
+
+static inline void
+update_stats_wait_start_dl(struct dl_rq *dl_rq, struct sched_dl_entity *dl_se)
+{
+	struct sched_statistics *stats;
+
+	if (!schedstat_enabled())
+		return;
+
+	stats = __schedstats_from_dl_se(dl_se);
+	__update_stats_wait_start(rq_of_dl_rq(dl_rq), dl_task_of(dl_se), stats);
+}
+
+static inline void
+update_stats_wait_end_dl(struct dl_rq *dl_rq, struct sched_dl_entity *dl_se)
+{
+	struct sched_statistics *stats;
+
+	if (!schedstat_enabled())
+		return;
+
+	stats = __schedstats_from_dl_se(dl_se);
+	__update_stats_wait_end(rq_of_dl_rq(dl_rq), dl_task_of(dl_se), stats);
+}
+
+static inline void
+update_stats_enqueue_sleeper_dl(struct dl_rq *dl_rq, struct sched_dl_entity *dl_se)
+{
+	struct sched_statistics *stats;
+
+	if (!schedstat_enabled())
+		return;
+
+	stats = __schedstats_from_dl_se(dl_se);
+	__update_stats_enqueue_sleeper(rq_of_dl_rq(dl_rq), dl_task_of(dl_se), stats);
+}
+
+static inline void
+update_stats_enqueue_dl(struct dl_rq *dl_rq, struct sched_dl_entity *dl_se,
+			int flags)
+{
+	if (!schedstat_enabled())
+		return;
+
+	if (flags & ENQUEUE_WAKEUP)
+		update_stats_enqueue_sleeper_dl(dl_rq, dl_se);
+}
+
+static inline void
+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);
+
+	if (!schedstat_enabled())
+		return;
+
+	if ((flags & DEQUEUE_SLEEP)) {
+		unsigned int state;
+
+		state = READ_ONCE(p->__state);
+		if (state & TASK_INTERRUPTIBLE)
+			__schedstat_set(p->stats.sleep_start,
+					rq_clock(rq_of_dl_rq(dl_rq)));
+
+		if (state & TASK_UNINTERRUPTIBLE)
+			__schedstat_set(p->stats.block_start,
+					rq_clock(rq_of_dl_rq(dl_rq)));
+	}
+}
+
 static void __enqueue_dl_entity(struct sched_dl_entity *dl_se)
 {
 	struct dl_rq *dl_rq = dl_rq_of_se(dl_se);
@@ -1504,6 +1580,8 @@ enqueue_dl_entity(struct sched_dl_entity *dl_se, int flags)
 {
 	BUG_ON(on_dl_rq(dl_se));
 
+	update_stats_enqueue_dl(dl_rq_of_se(dl_se), dl_se, flags);
+
 	/*
 	 * If this is a wakeup or a new instance, the scheduling
 	 * parameters of the task might need updating. Otherwise,
@@ -1600,6 +1678,9 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 		return;
 	}
 
+	check_schedstat_required();
+	update_stats_wait_start_dl(dl_rq_of_se(&p->dl), &p->dl);
+
 	enqueue_dl_entity(&p->dl, flags);
 
 	if (!task_current(rq, p) && p->nr_cpus_allowed > 1)
@@ -1608,6 +1689,7 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 
 static void __dequeue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 {
+	update_stats_dequeue_dl(&rq->dl, &p->dl, flags);
 	dequeue_dl_entity(&p->dl);
 	dequeue_pushable_dl_task(rq, p);
 }
@@ -1827,7 +1909,12 @@ static void start_hrtick_dl(struct rq *rq, struct task_struct *p)
 
 static void set_next_task_dl(struct rq *rq, struct task_struct *p, bool first)
 {
+	struct sched_dl_entity *dl_se = &p->dl;
+	struct dl_rq *dl_rq = &rq->dl;
+
 	p->se.exec_start = rq_clock_task(rq);
+	if (on_dl_rq(&p->dl))
+		update_stats_wait_end_dl(dl_rq, dl_se);
 
 	/* You can't push away the running task */
 	dequeue_pushable_dl_task(rq, p);
@@ -1884,6 +1971,12 @@ static struct task_struct *pick_next_task_dl(struct rq *rq)
 
 static void put_prev_task_dl(struct rq *rq, struct task_struct *p)
 {
+	struct sched_dl_entity *dl_se = &p->dl;
+	struct dl_rq *dl_rq = &rq->dl;
+
+	if (on_dl_rq(&p->dl))
+		update_stats_wait_start_dl(dl_rq, dl_se);
+
 	update_curr_dl(rq);
 
 	update_dl_rq_load_avg(rq_clock_pelt(rq), rq, 1);

^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched/dl: Support sched_stat_runtime tracepoint for deadline sched class
  2021-09-05 14:35 ` [PATCH v4 7/8] sched, dl: support sched_stat_runtime tracepoint for deadline " Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched/dl: Support " tip-bot2 for Yafang Shao
@ 2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 0 replies; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-10-05 14:12 UTC (permalink / raw)
  To: linux-tip-commits; +Cc: Yafang Shao, Peter Zijlstra (Intel), x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     95fd58e8dadb7aa707628a2187c626bb897c49ec
Gitweb:        https://git.kernel.org/tip/95fd58e8dadb7aa707628a2187c626bb897c49ec
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:46 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Oct 2021 15:51:52 +02:00

sched/dl: Support sched_stat_runtime tracepoint for deadline sched class

The runtime of a DL task has already been there, so we only need to
add a tracepoint.

One difference between fair task and DL task is that there is no vruntime
in dl task. To reuse the sched_stat_runtime tracepoint, '0' is passed as
vruntime for DL task.

The output of this tracepoint for DL task as follows,
             top-36462   [047] d.h.  6083.452103: sched_stat_runtime: comm=top pid=36462 runtime=409898 [ns] vruntime=0 [ns]

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210905143547.4668-8-laoar.shao@gmail.com
---
 kernel/sched/deadline.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index 51dd309..73fb33e 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -1268,6 +1268,8 @@ static void update_curr_dl(struct rq *rq)
 	schedstat_set(curr->stats.exec_max,
 		      max(curr->stats.exec_max, delta_exec));
 
+	trace_sched_stat_runtime(curr, delta_exec, 0);
+
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);
 

^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched/rt: Support sched_stat_runtime tracepoint for RT sched class
  2021-09-05 14:35 ` [PATCH v4 5/8] sched, rt: support sched_stat_runtime tracepoint for RT sched class Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched/rt: Support " tip-bot2 for Yafang Shao
@ 2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 0 replies; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-10-05 14:12 UTC (permalink / raw)
  To: linux-tip-commits; +Cc: Yafang Shao, Peter Zijlstra (Intel), x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     ed7b564cfdd0668efbd739d0b4e2d67797293f32
Gitweb:        https://git.kernel.org/tip/ed7b564cfdd0668efbd739d0b4e2d67797293f32
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:44 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Oct 2021 15:51:49 +02:00

sched/rt: Support sched_stat_runtime tracepoint for RT sched class

The runtime of a RT task has already been there, so we only need to
add a tracepoint.

One difference between fair task and RT task is that there is no vruntime
in RT task. To reuse the sched_stat_runtime tracepoint, '0' is passed as
vruntime for RT task.

The output of this tracepoint for RT task as follows,
          stress-9748    [039] d.h.   113.519352: sched_stat_runtime: comm=stress pid=9748 runtime=997573 [ns] vruntime=0 [ns]
          stress-9748    [039] d.h.   113.520352: sched_stat_runtime: comm=stress pid=9748 runtime=997627 [ns] vruntime=0 [ns]
          stress-9748    [039] d.h.   113.521352: sched_stat_runtime: comm=stress pid=9748 runtime=998203 [ns] vruntime=0 [ns]

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210905143547.4668-6-laoar.shao@gmail.com
---
 kernel/sched/rt.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 95a7c3a..5d25111 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1012,6 +1012,8 @@ static void update_curr_rt(struct rq *rq)
 	schedstat_set(curr->stats.exec_max,
 		      max(curr->stats.exec_max, delta_exec));
 
+	trace_sched_stat_runtime(curr, delta_exec, 0);
+
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);
 

^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched/rt: Support schedstats for RT sched class
  2021-09-05 14:35 ` [PATCH v4 6/8] sched, rt: support schedstats " Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched/rt: Support " tip-bot2 for Yafang Shao
@ 2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 0 replies; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-10-05 14:12 UTC (permalink / raw)
  To: linux-tip-commits; +Cc: Yafang Shao, Peter Zijlstra (Intel), x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     57a5c2dafca8e3ce4f70e975a9c7727b66b5071f
Gitweb:        https://git.kernel.org/tip/57a5c2dafca8e3ce4f70e975a9c7727b66b5071f
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:45 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Oct 2021 15:51:51 +02:00

sched/rt: Support schedstats for RT sched class

We want to measure the latency of RT tasks in our production
environment with schedstats facility, but currently schedstats is only
supported for fair sched class. This patch enable it for RT sched class
as well.

After we make the struct sched_statistics and the helpers of it
independent of fair sched class, we can easily use the schedstats
facility for RT sched class.

The schedstat usage in RT sched class is similar with fair sched class,
for example,
                fair                        RT
enqueue         update_stats_enqueue_fair   update_stats_enqueue_rt
dequeue         update_stats_dequeue_fair   update_stats_dequeue_rt
put_prev_task   update_stats_wait_start     update_stats_wait_start_rt
set_next_task   update_stats_wait_end       update_stats_wait_end_rt

The user can get the schedstats information in the same way in fair sched
class. For example,
       fair                            RT
       /proc/[pid]/sched               /proc/[pid]/sched

schedstats is not supported for RT group.

The output of a RT task's schedstats as follows,
$ cat /proc/10349/sched
...
sum_sleep_runtime                            :           972.434535
sum_block_runtime                            :           960.433522
wait_start                                   :        188510.871584
sleep_start                                  :             0.000000
block_start                                  :             0.000000
sleep_max                                    :            12.001013
block_max                                    :           952.660622
exec_max                                     :             0.049629
slice_max                                    :             0.000000
wait_max                                     :             0.018538
wait_sum                                     :             0.424340
wait_count                                   :                   49
iowait_sum                                   :           956.495640
iowait_count                                 :                   24
nr_migrations_cold                           :                    0
nr_failed_migrations_affine                  :                    0
nr_failed_migrations_running                 :                    0
nr_failed_migrations_hot                     :                    0
nr_forced_migrations                         :                    0
nr_wakeups                                   :                   49
nr_wakeups_sync                              :                    0
nr_wakeups_migrate                           :                    0
nr_wakeups_local                             :                   49
nr_wakeups_remote                            :                    0
nr_wakeups_affine                            :                    0
nr_wakeups_affine_attempts                   :                    0
nr_wakeups_passive                           :                    0
nr_wakeups_idle                              :                    0
...

The sched:sched_stat_{wait, sleep, iowait, blocked} tracepoints can
be used to trace RT tasks as well. The output of these tracepoints for a
RT tasks as follows,

- runtime
          stress-10352   [004] d.h.  1035.382286: sched_stat_runtime: comm=stress pid=10352 runtime=995769 [ns] vruntime=0 [ns]
          [vruntime=0 means it is a RT task]

- wait
          <idle>-0       [004] dN..  1227.688544: sched_stat_wait: comm=stress pid=10352 delay=46849882 [ns]

- blocked
     kworker/4:1-465     [004] dN..  1585.676371: sched_stat_blocked: comm=stress pid=17194 delay=189963 [ns]

- iowait
     kworker/4:1-465     [004] dN..  1585.675330: sched_stat_iowait: comm=stress pid=17189 delay=182848 [ns]

- sleep
           sleep-18194   [023] dN..  1780.891840: sched_stat_sleep: comm=sleep.sh pid=17767 delay=1001160770 [ns]
           sleep-18196   [023] dN..  1781.893208: sched_stat_sleep: comm=sleep.sh pid=17767 delay=1001161970 [ns]
           sleep-18197   [023] dN..  1782.894544: sched_stat_sleep: comm=sleep.sh pid=17767 delay=1001128840 [ns]
           [ In sleep.sh, it sleeps 1 sec each time. ]

[lkp@intel.com: reported build failure in earlier version]

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210905143547.4668-7-laoar.shao@gmail.com
---
 kernel/sched/rt.c | 124 +++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 124 insertions(+)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 5d25111..bb945f8 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1273,6 +1273,112 @@ static void __delist_rt_entity(struct sched_rt_entity *rt_se, struct rt_prio_arr
 	rt_se->on_list = 0;
 }
 
+static inline struct sched_statistics *
+__schedstats_from_rt_se(struct sched_rt_entity *rt_se)
+{
+#ifdef CONFIG_RT_GROUP_SCHED
+	/* schedstats is not supported for rt group. */
+	if (!rt_entity_is_task(rt_se))
+		return NULL;
+#endif
+
+	return &rt_task_of(rt_se)->stats;
+}
+
+static inline void
+update_stats_wait_start_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se)
+{
+	struct sched_statistics *stats;
+	struct task_struct *p = NULL;
+
+	if (!schedstat_enabled())
+		return;
+
+	if (rt_entity_is_task(rt_se))
+		p = rt_task_of(rt_se);
+
+	stats = __schedstats_from_rt_se(rt_se);
+	if (!stats)
+		return;
+
+	__update_stats_wait_start(rq_of_rt_rq(rt_rq), p, stats);
+}
+
+static inline void
+update_stats_enqueue_sleeper_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se)
+{
+	struct sched_statistics *stats;
+	struct task_struct *p = NULL;
+
+	if (!schedstat_enabled())
+		return;
+
+	if (rt_entity_is_task(rt_se))
+		p = rt_task_of(rt_se);
+
+	stats = __schedstats_from_rt_se(rt_se);
+	if (!stats)
+		return;
+
+	__update_stats_enqueue_sleeper(rq_of_rt_rq(rt_rq), p, stats);
+}
+
+static inline void
+update_stats_enqueue_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se,
+			int flags)
+{
+	if (!schedstat_enabled())
+		return;
+
+	if (flags & ENQUEUE_WAKEUP)
+		update_stats_enqueue_sleeper_rt(rt_rq, rt_se);
+}
+
+static inline void
+update_stats_wait_end_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se)
+{
+	struct sched_statistics *stats;
+	struct task_struct *p = NULL;
+
+	if (!schedstat_enabled())
+		return;
+
+	if (rt_entity_is_task(rt_se))
+		p = rt_task_of(rt_se);
+
+	stats = __schedstats_from_rt_se(rt_se);
+	if (!stats)
+		return;
+
+	__update_stats_wait_end(rq_of_rt_rq(rt_rq), p, stats);
+}
+
+static inline void
+update_stats_dequeue_rt(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se,
+			int flags)
+{
+	struct task_struct *p = NULL;
+
+	if (!schedstat_enabled())
+		return;
+
+	if (rt_entity_is_task(rt_se))
+		p = rt_task_of(rt_se);
+
+	if ((flags & DEQUEUE_SLEEP) && p) {
+		unsigned int state;
+
+		state = READ_ONCE(p->__state);
+		if (state & TASK_INTERRUPTIBLE)
+			__schedstat_set(p->stats.sleep_start,
+					rq_clock(rq_of_rt_rq(rt_rq)));
+
+		if (state & TASK_UNINTERRUPTIBLE)
+			__schedstat_set(p->stats.block_start,
+					rq_clock(rq_of_rt_rq(rt_rq)));
+	}
+}
+
 static void __enqueue_rt_entity(struct sched_rt_entity *rt_se, unsigned int flags)
 {
 	struct rt_rq *rt_rq = rt_rq_of_se(rt_se);
@@ -1346,6 +1452,8 @@ static void enqueue_rt_entity(struct sched_rt_entity *rt_se, unsigned int flags)
 {
 	struct rq *rq = rq_of_rt_se(rt_se);
 
+	update_stats_enqueue_rt(rt_rq_of_se(rt_se), rt_se, flags);
+
 	dequeue_rt_stack(rt_se, flags);
 	for_each_sched_rt_entity(rt_se)
 		__enqueue_rt_entity(rt_se, flags);
@@ -1356,6 +1464,8 @@ static void dequeue_rt_entity(struct sched_rt_entity *rt_se, unsigned int flags)
 {
 	struct rq *rq = rq_of_rt_se(rt_se);
 
+	update_stats_dequeue_rt(rt_rq_of_se(rt_se), rt_se, flags);
+
 	dequeue_rt_stack(rt_se, flags);
 
 	for_each_sched_rt_entity(rt_se) {
@@ -1378,6 +1488,9 @@ enqueue_task_rt(struct rq *rq, struct task_struct *p, int flags)
 	if (flags & ENQUEUE_WAKEUP)
 		rt_se->timeout = 0;
 
+	check_schedstat_required();
+	update_stats_wait_start_rt(rt_rq_of_se(rt_se), rt_se);
+
 	enqueue_rt_entity(rt_se, flags);
 
 	if (!task_current(rq, p) && p->nr_cpus_allowed > 1)
@@ -1578,7 +1691,12 @@ static void check_preempt_curr_rt(struct rq *rq, struct task_struct *p, int flag
 
 static inline void set_next_task_rt(struct rq *rq, struct task_struct *p, bool first)
 {
+	struct sched_rt_entity *rt_se = &p->rt;
+	struct rt_rq *rt_rq = &rq->rt;
+
 	p->se.exec_start = rq_clock_task(rq);
+	if (on_rt_rq(&p->rt))
+		update_stats_wait_end_rt(rt_rq, rt_se);
 
 	/* The running task is never eligible for pushing */
 	dequeue_pushable_task(rq, p);
@@ -1652,6 +1770,12 @@ static struct task_struct *pick_next_task_rt(struct rq *rq)
 
 static void put_prev_task_rt(struct rq *rq, struct task_struct *p)
 {
+	struct sched_rt_entity *rt_se = &p->rt;
+	struct rt_rq *rt_rq = &rq->rt;
+
+	if (on_rt_rq(&p->rt))
+		update_stats_wait_start_rt(rt_rq, rt_se);
+
 	update_curr_rt(rq);
 
 	update_rt_rq_load_avg(rq_clock_pelt(rq), rq, 1);

^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched: Introduce task block time in schedstats
  2021-09-05 14:35 ` [PATCH v4 4/8] sched: introduce task block time in schedstats Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched: Introduce " tip-bot2 for Yafang Shao
@ 2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 0 replies; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-10-05 14:12 UTC (permalink / raw)
  To: linux-tip-commits; +Cc: Yafang Shao, Peter Zijlstra (Intel), x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     847fc0cd0664fcb2a08ac66df6b85935361ec454
Gitweb:        https://git.kernel.org/tip/847fc0cd0664fcb2a08ac66df6b85935361ec454
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:43 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Oct 2021 15:51:48 +02:00

sched: Introduce task block time in schedstats

Currently in schedstats we have sum_sleep_runtime and iowait_sum, but
there's no metric to show how long the task is in D state.  Once a task in
D state, it means the task is blocked in the kernel, for example the
task may be waiting for a mutex. The D state is more frequent than
iowait, and it is more critital than S state. So it is worth to add a
metric to measure it.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210905143547.4668-5-laoar.shao@gmail.com
---
 include/linux/sched.h | 2 ++
 kernel/sched/debug.c  | 6 ++++--
 kernel/sched/stats.c  | 1 +
 3 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 2bc4c72..193e16e 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -503,6 +503,8 @@ struct sched_statistics {
 
 	u64				block_start;
 	u64				block_max;
+	s64				sum_block_runtime;
+
 	u64				exec_max;
 	u64				slice_max;
 
diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index 76fd38e..26fac5e 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -540,10 +540,11 @@ print_task(struct seq_file *m, struct rq *rq, struct task_struct *p)
 		(long long)(p->nvcsw + p->nivcsw),
 		p->prio);
 
-	SEQ_printf(m, "%9Ld.%06ld %9Ld.%06ld %9Ld.%06ld",
+	SEQ_printf(m, "%9lld.%06ld %9lld.%06ld %9lld.%06ld %9lld.%06ld",
 		SPLIT_NS(schedstat_val_or_zero(p->stats.wait_sum)),
 		SPLIT_NS(p->se.sum_exec_runtime),
-		SPLIT_NS(schedstat_val_or_zero(p->stats.sum_sleep_runtime)));
+		SPLIT_NS(schedstat_val_or_zero(p->stats.sum_sleep_runtime)),
+		SPLIT_NS(schedstat_val_or_zero(p->stats.sum_block_runtime)));
 
 #ifdef CONFIG_NUMA_BALANCING
 	SEQ_printf(m, " %d %d", task_node(p), task_numa_group_id(p));
@@ -977,6 +978,7 @@ void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
 		u64 avg_atom, avg_per_cpu;
 
 		PN_SCHEDSTAT(sum_sleep_runtime);
+		PN_SCHEDSTAT(sum_block_runtime);
 		PN_SCHEDSTAT(wait_start);
 		PN_SCHEDSTAT(sleep_start);
 		PN_SCHEDSTAT(block_start);
diff --git a/kernel/sched/stats.c b/kernel/sched/stats.c
index fad781c..07dde29 100644
--- a/kernel/sched/stats.c
+++ b/kernel/sched/stats.c
@@ -82,6 +82,7 @@ void __update_stats_enqueue_sleeper(struct rq *rq, struct task_struct *p,
 
 		__schedstat_set(stats->block_start, 0);
 		__schedstat_add(stats->sum_sleep_runtime, delta);
+		__schedstat_add(stats->sum_block_runtime, delta);
 
 		if (p) {
 			if (p->in_iowait) {

^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched: Make schedstats helpers independent of fair sched class
  2021-09-05 14:35 ` [PATCH v4 3/8] sched: make schedstats helpers " Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched: Make " tip-bot2 for Yafang Shao
@ 2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 0 replies; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-10-05 14:12 UTC (permalink / raw)
  To: linux-tip-commits
  Cc: Yafang Shao, Peter Zijlstra (Intel), Mel Gorman, x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     60f2415e19d3948641149ac6aca137a7be1d1952
Gitweb:        https://git.kernel.org/tip/60f2415e19d3948641149ac6aca137a7be1d1952
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:42 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Oct 2021 15:51:47 +02:00

sched: Make schedstats helpers independent of fair sched class

The original prototype of the schedstats helpers are

  update_stats_wait_*(struct cfs_rq *cfs_rq, struct sched_entity *se)

The cfs_rq in these helpers is used to get the rq_clock, and the se is
used to get the struct sched_statistics and the struct task_struct. In
order to make these helpers available by all sched classes, we can pass
the rq, sched_statistics and task_struct directly.

Then the new helpers are

  update_stats_wait_*(struct rq *rq, struct task_struct *p,
                      struct sched_statistics *stats)

which are independent of fair sched class.

To avoid vmlinux growing too large or introducing ovehead when
!schedstat_enabled(), some new helpers after schedstat_enabled() are also
introduced, Suggested by Mel. These helpers are in sched/stats.c,

  __update_stats_wait_*(struct rq *rq, struct task_struct *p,
                        struct sched_statistics *stats)

The size of vmlinux as follows,
                      Before          After
  Size of vmlinux     826308552       826304640
The size is a litte smaller as some functions are not inlined again after
the change.

I also compared the sched performance with 'perf bench sched pipe',
suggested by Mel. The result as followsi (in usecs/op),
                             Before                After
  kernel.sched_schedstats=0  5.2~5.4               5.2~5.4
  kernel.sched_schedstats=1  5.3~5.5               5.3~5.5

[These data is a little difference with the prev version, that is
because my old test machine is destroyed so I have to use a new
different test machine.]
Almost no difference.

No functional change.

[lkp@intel.com: reported build failure in prev version]

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lore.kernel.org/r/20210905143547.4668-4-laoar.shao@gmail.com
---
 kernel/sched/fair.c  | 134 +++++-------------------------------------
 kernel/sched/stats.c | 103 ++++++++++++++++++++++++++++++++-
 kernel/sched/stats.h |  30 +++++++++-
 3 files changed, 152 insertions(+), 115 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 97c6ecb..71b30ef 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -887,32 +887,27 @@ static void update_curr_fair(struct rq *rq)
 }
 
 static inline void
-update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
+update_stats_wait_start_fair(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
-	u64 wait_start, prev_wait_start;
 	struct sched_statistics *stats;
+	struct task_struct *p = NULL;
 
 	if (!schedstat_enabled())
 		return;
 
 	stats = __schedstats_from_se(se);
 
-	wait_start = rq_clock(rq_of(cfs_rq));
-	prev_wait_start = schedstat_val(stats->wait_start);
-
-	if (entity_is_task(se) && task_on_rq_migrating(task_of(se)) &&
-	    likely(wait_start > prev_wait_start))
-		wait_start -= prev_wait_start;
+	if (entity_is_task(se))
+		p = task_of(se);
 
-	__schedstat_set(stats->wait_start, wait_start);
+	__update_stats_wait_start(rq_of(cfs_rq), p, stats);
 }
 
 static inline void
-update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
+update_stats_wait_end_fair(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
 	struct sched_statistics *stats;
 	struct task_struct *p = NULL;
-	u64 delta;
 
 	if (!schedstat_enabled())
 		return;
@@ -928,105 +923,34 @@ update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	if (unlikely(!schedstat_val(stats->wait_start)))
 		return;
 
-	delta = rq_clock(rq_of(cfs_rq)) - schedstat_val(stats->wait_start);
-
-	if (entity_is_task(se)) {
+	if (entity_is_task(se))
 		p = task_of(se);
-		if (task_on_rq_migrating(p)) {
-			/*
-			 * Preserve migrating task's wait time so wait_start
-			 * time stamp can be adjusted to accumulate wait time
-			 * prior to migration.
-			 */
-			__schedstat_set(stats->wait_start, delta);
-			return;
-		}
-		trace_sched_stat_wait(p, delta);
-	}
 
-	__schedstat_set(stats->wait_max,
-		      max(schedstat_val(stats->wait_max), delta));
-	__schedstat_inc(stats->wait_count);
-	__schedstat_add(stats->wait_sum, delta);
-	__schedstat_set(stats->wait_start, 0);
+	__update_stats_wait_end(rq_of(cfs_rq), p, stats);
 }
 
 static inline void
-update_stats_enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
+update_stats_enqueue_sleeper_fair(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
 	struct sched_statistics *stats;
 	struct task_struct *tsk = NULL;
-	u64 sleep_start, block_start;
 
 	if (!schedstat_enabled())
 		return;
 
 	stats = __schedstats_from_se(se);
 
-	sleep_start = schedstat_val(stats->sleep_start);
-	block_start = schedstat_val(stats->block_start);
-
 	if (entity_is_task(se))
 		tsk = task_of(se);
 
-	if (sleep_start) {
-		u64 delta = rq_clock(rq_of(cfs_rq)) - sleep_start;
-
-		if ((s64)delta < 0)
-			delta = 0;
-
-		if (unlikely(delta > schedstat_val(stats->sleep_max)))
-			__schedstat_set(stats->sleep_max, delta);
-
-		__schedstat_set(stats->sleep_start, 0);
-		__schedstat_add(stats->sum_sleep_runtime, delta);
-
-		if (tsk) {
-			account_scheduler_latency(tsk, delta >> 10, 1);
-			trace_sched_stat_sleep(tsk, delta);
-		}
-	}
-	if (block_start) {
-		u64 delta = rq_clock(rq_of(cfs_rq)) - block_start;
-
-		if ((s64)delta < 0)
-			delta = 0;
-
-		if (unlikely(delta > schedstat_val(stats->block_max)))
-			__schedstat_set(stats->block_max, delta);
-
-		__schedstat_set(stats->block_start, 0);
-		__schedstat_add(stats->sum_sleep_runtime, delta);
-
-		if (tsk) {
-			if (tsk->in_iowait) {
-				__schedstat_add(stats->iowait_sum, delta);
-				__schedstat_inc(stats->iowait_count);
-				trace_sched_stat_iowait(tsk, delta);
-			}
-
-			trace_sched_stat_blocked(tsk, delta);
-
-			/*
-			 * Blocking time is in units of nanosecs, so shift by
-			 * 20 to get a milliseconds-range estimation of the
-			 * amount of time that the task spent sleeping:
-			 */
-			if (unlikely(prof_on == SLEEP_PROFILING)) {
-				profile_hits(SLEEP_PROFILING,
-						(void *)get_wchan(tsk),
-						delta >> 20);
-			}
-			account_scheduler_latency(tsk, delta >> 10, 0);
-		}
-	}
+	__update_stats_enqueue_sleeper(rq_of(cfs_rq), tsk, stats);
 }
 
 /*
  * Task is being enqueued - update stats:
  */
 static inline void
-update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
+update_stats_enqueue_fair(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 {
 	if (!schedstat_enabled())
 		return;
@@ -1036,14 +960,14 @@ update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 	 * a dequeue/enqueue event is a NOP)
 	 */
 	if (se != cfs_rq->curr)
-		update_stats_wait_start(cfs_rq, se);
+		update_stats_wait_start_fair(cfs_rq, se);
 
 	if (flags & ENQUEUE_WAKEUP)
-		update_stats_enqueue_sleeper(cfs_rq, se);
+		update_stats_enqueue_sleeper_fair(cfs_rq, se);
 }
 
 static inline void
-update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
+update_stats_dequeue_fair(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 {
 
 	if (!schedstat_enabled())
@@ -1054,7 +978,7 @@ update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 	 * waiting task:
 	 */
 	if (se != cfs_rq->curr)
-		update_stats_wait_end(cfs_rq, se);
+		update_stats_wait_end_fair(cfs_rq, se);
 
 	if ((flags & DEQUEUE_SLEEP) && entity_is_task(se)) {
 		struct task_struct *tsk = task_of(se);
@@ -4267,26 +4191,6 @@ 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 void check_schedstat_required(void)
-{
-#ifdef CONFIG_SCHEDSTATS
-	if (schedstat_enabled())
-		return;
-
-	/* Force schedstat enabled if a dependent tracepoint is active */
-	if (trace_sched_stat_wait_enabled()    ||
-			trace_sched_stat_sleep_enabled()   ||
-			trace_sched_stat_iowait_enabled()  ||
-			trace_sched_stat_blocked_enabled() ||
-			trace_sched_stat_runtime_enabled())  {
-		printk_deferred_once("Scheduler tracepoints stat_sleep, stat_iowait, "
-			     "stat_blocked and stat_runtime require the "
-			     "kernel parameter schedstats=enable or "
-			     "kernel.sched_schedstats=1\n");
-	}
-#endif
-}
-
 static inline bool cfs_bandwidth_used(void);
 
 /*
@@ -4360,7 +4264,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 		place_entity(cfs_rq, se, 0);
 
 	check_schedstat_required();
-	update_stats_enqueue(cfs_rq, se, flags);
+	update_stats_enqueue_fair(cfs_rq, se, flags);
 	check_spread(cfs_rq, se);
 	if (!curr)
 		__enqueue_entity(cfs_rq, se);
@@ -4444,7 +4348,7 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 	update_load_avg(cfs_rq, se, UPDATE_TG);
 	se_update_runnable(se);
 
-	update_stats_dequeue(cfs_rq, se, flags);
+	update_stats_dequeue_fair(cfs_rq, se, flags);
 
 	clear_buddies(cfs_rq, se);
 
@@ -4529,7 +4433,7 @@ set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 		 * a CPU. So account for the time it spent waiting on the
 		 * runqueue.
 		 */
-		update_stats_wait_end(cfs_rq, se);
+		update_stats_wait_end_fair(cfs_rq, se);
 		__dequeue_entity(cfs_rq, se);
 		update_load_avg(cfs_rq, se, UPDATE_TG);
 	}
@@ -4631,7 +4535,7 @@ static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev)
 	check_spread(cfs_rq, prev);
 
 	if (prev->on_rq) {
-		update_stats_wait_start(cfs_rq, prev);
+		update_stats_wait_start_fair(cfs_rq, prev);
 		/* Put 'current' back into the tree. */
 		__enqueue_entity(cfs_rq, prev);
 		/* in !on_rq case, update occurred at dequeue */
diff --git a/kernel/sched/stats.c b/kernel/sched/stats.c
index 3f93fc3..fad781c 100644
--- a/kernel/sched/stats.c
+++ b/kernel/sched/stats.c
@@ -4,6 +4,109 @@
  */
 #include "sched.h"
 
+void __update_stats_wait_start(struct rq *rq, struct task_struct *p,
+			       struct sched_statistics *stats)
+{
+	u64 wait_start, prev_wait_start;
+
+	wait_start = rq_clock(rq);
+	prev_wait_start = schedstat_val(stats->wait_start);
+
+	if (p && likely(wait_start > prev_wait_start))
+		wait_start -= prev_wait_start;
+
+	__schedstat_set(stats->wait_start, wait_start);
+}
+
+void __update_stats_wait_end(struct rq *rq, struct task_struct *p,
+			     struct sched_statistics *stats)
+{
+	u64 delta = rq_clock(rq) - schedstat_val(stats->wait_start);
+
+	if (p) {
+		if (task_on_rq_migrating(p)) {
+			/*
+			 * Preserve migrating task's wait time so wait_start
+			 * time stamp can be adjusted to accumulate wait time
+			 * prior to migration.
+			 */
+			__schedstat_set(stats->wait_start, delta);
+
+			return;
+		}
+
+		trace_sched_stat_wait(p, delta);
+	}
+
+	__schedstat_set(stats->wait_max,
+			max(schedstat_val(stats->wait_max), delta));
+	__schedstat_inc(stats->wait_count);
+	__schedstat_add(stats->wait_sum, delta);
+	__schedstat_set(stats->wait_start, 0);
+}
+
+void __update_stats_enqueue_sleeper(struct rq *rq, struct task_struct *p,
+				    struct sched_statistics *stats)
+{
+	u64 sleep_start, block_start;
+
+	sleep_start = schedstat_val(stats->sleep_start);
+	block_start = schedstat_val(stats->block_start);
+
+	if (sleep_start) {
+		u64 delta = rq_clock(rq) - sleep_start;
+
+		if ((s64)delta < 0)
+			delta = 0;
+
+		if (unlikely(delta > schedstat_val(stats->sleep_max)))
+			__schedstat_set(stats->sleep_max, delta);
+
+		__schedstat_set(stats->sleep_start, 0);
+		__schedstat_add(stats->sum_sleep_runtime, delta);
+
+		if (p) {
+			account_scheduler_latency(p, delta >> 10, 1);
+			trace_sched_stat_sleep(p, delta);
+		}
+	}
+
+	if (block_start) {
+		u64 delta = rq_clock(rq) - block_start;
+
+		if ((s64)delta < 0)
+			delta = 0;
+
+		if (unlikely(delta > schedstat_val(stats->block_max)))
+			__schedstat_set(stats->block_max, delta);
+
+		__schedstat_set(stats->block_start, 0);
+		__schedstat_add(stats->sum_sleep_runtime, delta);
+
+		if (p) {
+			if (p->in_iowait) {
+				__schedstat_add(stats->iowait_sum, delta);
+				__schedstat_inc(stats->iowait_count);
+				trace_sched_stat_iowait(p, delta);
+			}
+
+			trace_sched_stat_blocked(p, delta);
+
+			/*
+			 * Blocking time is in units of nanosecs, so shift by
+			 * 20 to get a milliseconds-range estimation of the
+			 * amount of time that the task spent sleeping:
+			 */
+			if (unlikely(prof_on == SLEEP_PROFILING)) {
+				profile_hits(SLEEP_PROFILING,
+					     (void *)get_wchan(p),
+					     delta >> 20);
+			}
+			account_scheduler_latency(p, delta >> 10, 0);
+		}
+	}
+}
+
 /*
  * Current schedstat API version.
  *
diff --git a/kernel/sched/stats.h b/kernel/sched/stats.h
index fb6022e..cfb0893 100644
--- a/kernel/sched/stats.h
+++ b/kernel/sched/stats.h
@@ -2,6 +2,8 @@
 
 #ifdef CONFIG_SCHEDSTATS
 
+extern struct static_key_false sched_schedstats;
+
 /*
  * Expects runqueue lock to be held for atomicity of update
  */
@@ -40,6 +42,29 @@ rq_sched_info_dequeue(struct rq *rq, unsigned long long delta)
 #define   schedstat_val(var)		(var)
 #define   schedstat_val_or_zero(var)	((schedstat_enabled()) ? (var) : 0)
 
+void __update_stats_wait_start(struct rq *rq, struct task_struct *p,
+			       struct sched_statistics *stats);
+
+void __update_stats_wait_end(struct rq *rq, struct task_struct *p,
+			     struct sched_statistics *stats);
+void __update_stats_enqueue_sleeper(struct rq *rq, struct task_struct *p,
+				    struct sched_statistics *stats);
+
+static inline void
+check_schedstat_required(void)
+{
+	if (schedstat_enabled())
+		return;
+
+	/* Force schedstat enabled if a dependent tracepoint is active */
+	if (trace_sched_stat_wait_enabled()    ||
+	    trace_sched_stat_sleep_enabled()   ||
+	    trace_sched_stat_iowait_enabled()  ||
+	    trace_sched_stat_blocked_enabled() ||
+	    trace_sched_stat_runtime_enabled())
+		printk_deferred_once("Scheduler tracepoints stat_sleep, stat_iowait, stat_blocked and stat_runtime require the kernel parameter schedstats=enable or kernel.sched_schedstats=1\n");
+}
+
 #else /* !CONFIG_SCHEDSTATS: */
 
 static inline void rq_sched_info_arrive  (struct rq *rq, unsigned long long delta) { }
@@ -55,6 +80,11 @@ static inline void rq_sched_info_depart  (struct rq *rq, unsigned long long delt
 # define   schedstat_val(var)		0
 # define   schedstat_val_or_zero(var)	0
 
+# define __update_stats_wait_start(rq, p, stats)       do { } while (0)
+# define __update_stats_wait_end(rq, p, stats)         do { } while (0)
+# define __update_stats_enqueue_sleeper(rq, p, stats)  do { } while (0)
+# define check_schedstat_required()                    do { } while (0)
+
 #endif /* CONFIG_SCHEDSTATS */
 
 #ifdef CONFIG_FAIR_GROUP_SCHED

^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched: Make struct sched_statistics independent of fair sched class
  2021-09-05 14:35 ` [PATCH v4 2/8] sched: make struct sched_statistics independent of fair sched class Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched: Make " tip-bot2 for Yafang Shao
@ 2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 0 replies; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-10-05 14:12 UTC (permalink / raw)
  To: linux-tip-commits
  Cc: Yafang Shao, Peter Zijlstra (Intel), Mel Gorman, x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     ceeadb83aea28372e54857bf88ab7e17af48ab7b
Gitweb:        https://git.kernel.org/tip/ceeadb83aea28372e54857bf88ab7e17af48ab7b
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:41 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Oct 2021 15:51:45 +02:00

sched: Make struct sched_statistics independent of fair sched class

If we want to use the schedstats facility to trace other sched classes, we
should make it independent of fair sched class. The struct sched_statistics
is the schedular statistics of a task_struct or a task_group. So we can
move it into struct task_struct and struct task_group to achieve the goal.

After the patch, schestats are orgnized as follows,

    struct task_struct {
       ...
       struct sched_entity se;
       struct sched_rt_entity rt;
       struct sched_dl_entity dl;
       ...
       struct sched_statistics stats;
       ...
   };

Regarding the task group, schedstats is only supported for fair group
sched, and a new struct sched_entity_stats is introduced, suggested by
Peter -

    struct sched_entity_stats {
        struct sched_entity     se;
        struct sched_statistics stats;
    } __no_randomize_layout;

Then with the se in a task_group, we can easily get the stats.

The sched_statistics members may be frequently modified when schedstats is
enabled, in order to avoid impacting on random data which may in the same
cacheline with them, the struct sched_statistics is defined as cacheline
aligned.

As this patch changes the core struct of scheduler, so I verified the
performance it may impact on the scheduler with 'perf bench sched
pipe', suggested by Mel. Below is the result, in which all the values
are in usecs/op.
                                  Before               After
      kernel.sched_schedstats=0  5.2~5.4               5.2~5.4
      kernel.sched_schedstats=1  5.3~5.5               5.3~5.5
[These data is a little difference with the earlier version, that is
 because my old test machine is destroyed so I have to use a new
 different test machine.]

Almost no impact on the sched performance.

No functional change.

[lkp@intel.com: reported build failure in earlier version]

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lore.kernel.org/r/20210905143547.4668-3-laoar.shao@gmail.com
---
 include/linux/sched.h    |  6 +--
 kernel/sched/core.c      | 25 ++++++-----
 kernel/sched/deadline.c  |  4 +-
 kernel/sched/debug.c     | 92 ++++++++++++++++++++-------------------
 kernel/sched/fair.c      | 89 ++++++++++++++++++++++----------------
 kernel/sched/rt.c        |  4 +-
 kernel/sched/stats.h     | 19 ++++++++-
 kernel/sched/stop_task.c |  4 +-
 8 files changed, 143 insertions(+), 100 deletions(-)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index c1a927d..2bc4c72 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -522,7 +522,7 @@ struct sched_statistics {
 	u64				nr_wakeups_passive;
 	u64				nr_wakeups_idle;
 #endif
-};
+} ____cacheline_aligned;
 
 struct sched_entity {
 	/* For load-balancing: */
@@ -538,8 +538,6 @@ struct sched_entity {
 
 	u64				nr_migrations;
 
-	struct sched_statistics		statistics;
-
 #ifdef CONFIG_FAIR_GROUP_SCHED
 	int				depth;
 	struct sched_entity		*parent;
@@ -803,6 +801,8 @@ struct task_struct {
 	struct uclamp_se		uclamp[UCLAMP_CNT];
 #endif
 
+	struct sched_statistics         stats;
+
 #ifdef CONFIG_PREEMPT_NOTIFIERS
 	/* List of struct preempt_notifier: */
 	struct hlist_head		preempt_notifiers;
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index ccb604a..a1741e0 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -3489,11 +3489,11 @@ ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
 #ifdef CONFIG_SMP
 	if (cpu == rq->cpu) {
 		__schedstat_inc(rq->ttwu_local);
-		__schedstat_inc(p->se.statistics.nr_wakeups_local);
+		__schedstat_inc(p->stats.nr_wakeups_local);
 	} else {
 		struct sched_domain *sd;
 
-		__schedstat_inc(p->se.statistics.nr_wakeups_remote);
+		__schedstat_inc(p->stats.nr_wakeups_remote);
 		rcu_read_lock();
 		for_each_domain(rq->cpu, sd) {
 			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
@@ -3505,14 +3505,14 @@ ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
 	}
 
 	if (wake_flags & WF_MIGRATED)
-		__schedstat_inc(p->se.statistics.nr_wakeups_migrate);
+		__schedstat_inc(p->stats.nr_wakeups_migrate);
 #endif /* CONFIG_SMP */
 
 	__schedstat_inc(rq->ttwu_count);
-	__schedstat_inc(p->se.statistics.nr_wakeups);
+	__schedstat_inc(p->stats.nr_wakeups);
 
 	if (wake_flags & WF_SYNC)
-		__schedstat_inc(p->se.statistics.nr_wakeups_sync);
+		__schedstat_inc(p->stats.nr_wakeups_sync);
 }
 
 /*
@@ -4196,7 +4196,7 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
 
 #ifdef CONFIG_SCHEDSTATS
 	/* Even if schedstat is disabled, there should not be garbage */
-	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
+	memset(&p->stats, 0, sizeof(p->stats));
 #endif
 
 	RB_CLEAR_NODE(&p->dl.rb_node);
@@ -9553,9 +9553,9 @@ void normalize_rt_tasks(void)
 			continue;
 
 		p->se.exec_start = 0;
-		schedstat_set(p->se.statistics.wait_start,  0);
-		schedstat_set(p->se.statistics.sleep_start, 0);
-		schedstat_set(p->se.statistics.block_start, 0);
+		schedstat_set(p->stats.wait_start,  0);
+		schedstat_set(p->stats.sleep_start, 0);
+		schedstat_set(p->stats.block_start, 0);
 
 		if (!dl_task(p) && !rt_task(p)) {
 			/*
@@ -10397,11 +10397,14 @@ 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) {
+		struct sched_statistics *stats;
 		u64 ws = 0;
 		int i;
 
-		for_each_possible_cpu(i)
-			ws += schedstat_val(tg->se[i]->statistics.wait_sum);
+		for_each_possible_cpu(i) {
+			stats = __schedstats_from_se(tg->se[i]);
+			ws += schedstat_val(stats->wait_sum);
+		}
 
 		seq_printf(sf, "wait_sum %llu\n", ws);
 	}
diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index e943146..51dd309 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -1265,8 +1265,8 @@ static void update_curr_dl(struct rq *rq)
 		return;
 	}
 
-	schedstat_set(curr->se.statistics.exec_max,
-		      max(curr->se.statistics.exec_max, delta_exec));
+	schedstat_set(curr->stats.exec_max,
+		      max(curr->stats.exec_max, delta_exec));
 
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);
diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index 34913a7..76fd38e 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -449,9 +449,11 @@ static void print_cfs_group_stats(struct seq_file *m, int cpu, struct task_group
 	struct sched_entity *se = tg->se[cpu];
 
 #define P(F)		SEQ_printf(m, "  .%-30s: %lld\n",	#F, (long long)F)
-#define P_SCHEDSTAT(F)	SEQ_printf(m, "  .%-30s: %lld\n",	#F, (long long)schedstat_val(F))
+#define P_SCHEDSTAT(F)	SEQ_printf(m, "  .%-30s: %lld\n",	\
+		#F, (long long)schedstat_val(stats->F))
 #define PN(F)		SEQ_printf(m, "  .%-30s: %lld.%06ld\n", #F, SPLIT_NS((long long)F))
-#define PN_SCHEDSTAT(F)	SEQ_printf(m, "  .%-30s: %lld.%06ld\n", #F, SPLIT_NS((long long)schedstat_val(F)))
+#define PN_SCHEDSTAT(F)	SEQ_printf(m, "  .%-30s: %lld.%06ld\n", \
+		#F, SPLIT_NS((long long)schedstat_val(stats->F)))
 
 	if (!se)
 		return;
@@ -461,16 +463,18 @@ static void print_cfs_group_stats(struct seq_file *m, int cpu, struct task_group
 	PN(se->sum_exec_runtime);
 
 	if (schedstat_enabled()) {
-		PN_SCHEDSTAT(se->statistics.wait_start);
-		PN_SCHEDSTAT(se->statistics.sleep_start);
-		PN_SCHEDSTAT(se->statistics.block_start);
-		PN_SCHEDSTAT(se->statistics.sleep_max);
-		PN_SCHEDSTAT(se->statistics.block_max);
-		PN_SCHEDSTAT(se->statistics.exec_max);
-		PN_SCHEDSTAT(se->statistics.slice_max);
-		PN_SCHEDSTAT(se->statistics.wait_max);
-		PN_SCHEDSTAT(se->statistics.wait_sum);
-		P_SCHEDSTAT(se->statistics.wait_count);
+               struct sched_statistics *stats =  __schedstats_from_se(se);
+
+		PN_SCHEDSTAT(wait_start);
+		PN_SCHEDSTAT(sleep_start);
+		PN_SCHEDSTAT(block_start);
+		PN_SCHEDSTAT(sleep_max);
+		PN_SCHEDSTAT(block_max);
+		PN_SCHEDSTAT(exec_max);
+		PN_SCHEDSTAT(slice_max);
+		PN_SCHEDSTAT(wait_max);
+		PN_SCHEDSTAT(wait_sum);
+		P_SCHEDSTAT(wait_count);
 	}
 
 	P(se->load.weight);
@@ -537,9 +541,9 @@ print_task(struct seq_file *m, struct rq *rq, struct task_struct *p)
 		p->prio);
 
 	SEQ_printf(m, "%9Ld.%06ld %9Ld.%06ld %9Ld.%06ld",
-		SPLIT_NS(schedstat_val_or_zero(p->se.statistics.wait_sum)),
+		SPLIT_NS(schedstat_val_or_zero(p->stats.wait_sum)),
 		SPLIT_NS(p->se.sum_exec_runtime),
-		SPLIT_NS(schedstat_val_or_zero(p->se.statistics.sum_sleep_runtime)));
+		SPLIT_NS(schedstat_val_or_zero(p->stats.sum_sleep_runtime)));
 
 #ifdef CONFIG_NUMA_BALANCING
 	SEQ_printf(m, " %d %d", task_node(p), task_numa_group_id(p));
@@ -958,8 +962,8 @@ void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
 		"---------------------------------------------------------"
 		"----------\n");
 
-#define P_SCHEDSTAT(F)  __PS(#F, schedstat_val(p->F))
-#define PN_SCHEDSTAT(F) __PSN(#F, schedstat_val(p->F))
+#define P_SCHEDSTAT(F)  __PS(#F, schedstat_val(p->stats.F))
+#define PN_SCHEDSTAT(F) __PSN(#F, schedstat_val(p->stats.F))
 
 	PN(se.exec_start);
 	PN(se.vruntime);
@@ -972,33 +976,33 @@ void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
 	if (schedstat_enabled()) {
 		u64 avg_atom, avg_per_cpu;
 
-		PN_SCHEDSTAT(se.statistics.sum_sleep_runtime);
-		PN_SCHEDSTAT(se.statistics.wait_start);
-		PN_SCHEDSTAT(se.statistics.sleep_start);
-		PN_SCHEDSTAT(se.statistics.block_start);
-		PN_SCHEDSTAT(se.statistics.sleep_max);
-		PN_SCHEDSTAT(se.statistics.block_max);
-		PN_SCHEDSTAT(se.statistics.exec_max);
-		PN_SCHEDSTAT(se.statistics.slice_max);
-		PN_SCHEDSTAT(se.statistics.wait_max);
-		PN_SCHEDSTAT(se.statistics.wait_sum);
-		P_SCHEDSTAT(se.statistics.wait_count);
-		PN_SCHEDSTAT(se.statistics.iowait_sum);
-		P_SCHEDSTAT(se.statistics.iowait_count);
-		P_SCHEDSTAT(se.statistics.nr_migrations_cold);
-		P_SCHEDSTAT(se.statistics.nr_failed_migrations_affine);
-		P_SCHEDSTAT(se.statistics.nr_failed_migrations_running);
-		P_SCHEDSTAT(se.statistics.nr_failed_migrations_hot);
-		P_SCHEDSTAT(se.statistics.nr_forced_migrations);
-		P_SCHEDSTAT(se.statistics.nr_wakeups);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_sync);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_migrate);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_local);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_remote);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_affine);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_affine_attempts);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_passive);
-		P_SCHEDSTAT(se.statistics.nr_wakeups_idle);
+		PN_SCHEDSTAT(sum_sleep_runtime);
+		PN_SCHEDSTAT(wait_start);
+		PN_SCHEDSTAT(sleep_start);
+		PN_SCHEDSTAT(block_start);
+		PN_SCHEDSTAT(sleep_max);
+		PN_SCHEDSTAT(block_max);
+		PN_SCHEDSTAT(exec_max);
+		PN_SCHEDSTAT(slice_max);
+		PN_SCHEDSTAT(wait_max);
+		PN_SCHEDSTAT(wait_sum);
+		P_SCHEDSTAT(wait_count);
+		PN_SCHEDSTAT(iowait_sum);
+		P_SCHEDSTAT(iowait_count);
+		P_SCHEDSTAT(nr_migrations_cold);
+		P_SCHEDSTAT(nr_failed_migrations_affine);
+		P_SCHEDSTAT(nr_failed_migrations_running);
+		P_SCHEDSTAT(nr_failed_migrations_hot);
+		P_SCHEDSTAT(nr_forced_migrations);
+		P_SCHEDSTAT(nr_wakeups);
+		P_SCHEDSTAT(nr_wakeups_sync);
+		P_SCHEDSTAT(nr_wakeups_migrate);
+		P_SCHEDSTAT(nr_wakeups_local);
+		P_SCHEDSTAT(nr_wakeups_remote);
+		P_SCHEDSTAT(nr_wakeups_affine);
+		P_SCHEDSTAT(nr_wakeups_affine_attempts);
+		P_SCHEDSTAT(nr_wakeups_passive);
+		P_SCHEDSTAT(nr_wakeups_idle);
 
 		avg_atom = p->se.sum_exec_runtime;
 		if (nr_switches)
@@ -1064,7 +1068,7 @@ void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
 void proc_sched_set_task(struct task_struct *p)
 {
 #ifdef CONFIG_SCHEDSTATS
-	memset(&p->se.statistics, 0, sizeof(p->se.statistics));
+	memset(&p->stats, 0, sizeof(p->stats));
 #endif
 }
 
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 61d3e3b..97c6ecb 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -856,8 +856,13 @@ static void update_curr(struct cfs_rq *cfs_rq)
 
 	curr->exec_start = now;
 
-	schedstat_set(curr->statistics.exec_max,
-		      max(delta_exec, curr->statistics.exec_max));
+	if (schedstat_enabled()) {
+		struct sched_statistics *stats;
+
+		stats = __schedstats_from_se(curr);
+		__schedstat_set(stats->exec_max,
+				max(delta_exec, stats->exec_max));
+	}
 
 	curr->sum_exec_runtime += delta_exec;
 	schedstat_add(cfs_rq->exec_clock, delta_exec);
@@ -885,39 +890,45 @@ static inline void
 update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
 	u64 wait_start, prev_wait_start;
+	struct sched_statistics *stats;
 
 	if (!schedstat_enabled())
 		return;
 
+	stats = __schedstats_from_se(se);
+
 	wait_start = rq_clock(rq_of(cfs_rq));
-	prev_wait_start = schedstat_val(se->statistics.wait_start);
+	prev_wait_start = schedstat_val(stats->wait_start);
 
 	if (entity_is_task(se) && task_on_rq_migrating(task_of(se)) &&
 	    likely(wait_start > prev_wait_start))
 		wait_start -= prev_wait_start;
 
-	__schedstat_set(se->statistics.wait_start, wait_start);
+	__schedstat_set(stats->wait_start, wait_start);
 }
 
 static inline void
 update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
-	struct task_struct *p;
+	struct sched_statistics *stats;
+	struct task_struct *p = NULL;
 	u64 delta;
 
 	if (!schedstat_enabled())
 		return;
 
+	stats = __schedstats_from_se(se);
+
 	/*
 	 * When the sched_schedstat changes from 0 to 1, some sched se
 	 * maybe already in the runqueue, the se->statistics.wait_start
 	 * will be 0.So it will let the delta wrong. We need to avoid this
 	 * scenario.
 	 */
-	if (unlikely(!schedstat_val(se->statistics.wait_start)))
+	if (unlikely(!schedstat_val(stats->wait_start)))
 		return;
 
-	delta = rq_clock(rq_of(cfs_rq)) - schedstat_val(se->statistics.wait_start);
+	delta = rq_clock(rq_of(cfs_rq)) - schedstat_val(stats->wait_start);
 
 	if (entity_is_task(se)) {
 		p = task_of(se);
@@ -927,30 +938,33 @@ update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
 			 * time stamp can be adjusted to accumulate wait time
 			 * prior to migration.
 			 */
-			__schedstat_set(se->statistics.wait_start, delta);
+			__schedstat_set(stats->wait_start, delta);
 			return;
 		}
 		trace_sched_stat_wait(p, delta);
 	}
 
-	__schedstat_set(se->statistics.wait_max,
-		      max(schedstat_val(se->statistics.wait_max), delta));
-	__schedstat_inc(se->statistics.wait_count);
-	__schedstat_add(se->statistics.wait_sum, delta);
-	__schedstat_set(se->statistics.wait_start, 0);
+	__schedstat_set(stats->wait_max,
+		      max(schedstat_val(stats->wait_max), delta));
+	__schedstat_inc(stats->wait_count);
+	__schedstat_add(stats->wait_sum, delta);
+	__schedstat_set(stats->wait_start, 0);
 }
 
 static inline void
 update_stats_enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
+	struct sched_statistics *stats;
 	struct task_struct *tsk = NULL;
 	u64 sleep_start, block_start;
 
 	if (!schedstat_enabled())
 		return;
 
-	sleep_start = schedstat_val(se->statistics.sleep_start);
-	block_start = schedstat_val(se->statistics.block_start);
+	stats = __schedstats_from_se(se);
+
+	sleep_start = schedstat_val(stats->sleep_start);
+	block_start = schedstat_val(stats->block_start);
 
 	if (entity_is_task(se))
 		tsk = task_of(se);
@@ -961,11 +975,11 @@ update_stats_enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
 		if ((s64)delta < 0)
 			delta = 0;
 
-		if (unlikely(delta > schedstat_val(se->statistics.sleep_max)))
-			__schedstat_set(se->statistics.sleep_max, delta);
+		if (unlikely(delta > schedstat_val(stats->sleep_max)))
+			__schedstat_set(stats->sleep_max, delta);
 
-		__schedstat_set(se->statistics.sleep_start, 0);
-		__schedstat_add(se->statistics.sum_sleep_runtime, delta);
+		__schedstat_set(stats->sleep_start, 0);
+		__schedstat_add(stats->sum_sleep_runtime, delta);
 
 		if (tsk) {
 			account_scheduler_latency(tsk, delta >> 10, 1);
@@ -978,16 +992,16 @@ update_stats_enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
 		if ((s64)delta < 0)
 			delta = 0;
 
-		if (unlikely(delta > schedstat_val(se->statistics.block_max)))
-			__schedstat_set(se->statistics.block_max, delta);
+		if (unlikely(delta > schedstat_val(stats->block_max)))
+			__schedstat_set(stats->block_max, delta);
 
-		__schedstat_set(se->statistics.block_start, 0);
-		__schedstat_add(se->statistics.sum_sleep_runtime, delta);
+		__schedstat_set(stats->block_start, 0);
+		__schedstat_add(stats->sum_sleep_runtime, delta);
 
 		if (tsk) {
 			if (tsk->in_iowait) {
-				__schedstat_add(se->statistics.iowait_sum, delta);
-				__schedstat_inc(se->statistics.iowait_count);
+				__schedstat_add(stats->iowait_sum, delta);
+				__schedstat_inc(stats->iowait_count);
 				trace_sched_stat_iowait(tsk, delta);
 			}
 
@@ -1049,10 +1063,10 @@ update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 		/* XXX racy against TTWU */
 		state = READ_ONCE(tsk->__state);
 		if (state & TASK_INTERRUPTIBLE)
-			__schedstat_set(se->statistics.sleep_start,
+			__schedstat_set(tsk->stats.sleep_start,
 				      rq_clock(rq_of(cfs_rq)));
 		if (state & TASK_UNINTERRUPTIBLE)
-			__schedstat_set(se->statistics.block_start,
+			__schedstat_set(tsk->stats.block_start,
 				      rq_clock(rq_of(cfs_rq)));
 	}
 }
@@ -4530,8 +4544,11 @@ set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	 */
 	if (schedstat_enabled() &&
 	    rq_of(cfs_rq)->cfs.load.weight >= 2*se->load.weight) {
-		__schedstat_set(se->statistics.slice_max,
-				max((u64)se->statistics.slice_max,
+		struct sched_statistics *stats;
+
+		stats = __schedstats_from_se(se);
+		__schedstat_set(stats->slice_max,
+				max((u64)stats->slice_max,
 				    se->sum_exec_runtime - se->prev_sum_exec_runtime));
 	}
 
@@ -6046,12 +6063,12 @@ static int wake_affine(struct sched_domain *sd, struct task_struct *p,
 	if (sched_feat(WA_WEIGHT) && target == nr_cpumask_bits)
 		target = wake_affine_weight(sd, p, this_cpu, prev_cpu, sync);
 
-	schedstat_inc(p->se.statistics.nr_wakeups_affine_attempts);
+	schedstat_inc(p->stats.nr_wakeups_affine_attempts);
 	if (target == nr_cpumask_bits)
 		return prev_cpu;
 
 	schedstat_inc(sd->ttwu_move_affine);
-	schedstat_inc(p->se.statistics.nr_wakeups_affine);
+	schedstat_inc(p->stats.nr_wakeups_affine);
 	return target;
 }
 
@@ -7855,7 +7872,7 @@ int can_migrate_task(struct task_struct *p, struct lb_env *env)
 	if (!cpumask_test_cpu(env->dst_cpu, p->cpus_ptr)) {
 		int cpu;
 
-		schedstat_inc(p->se.statistics.nr_failed_migrations_affine);
+		schedstat_inc(p->stats.nr_failed_migrations_affine);
 
 		env->flags |= LBF_SOME_PINNED;
 
@@ -7889,7 +7906,7 @@ int can_migrate_task(struct task_struct *p, struct lb_env *env)
 	env->flags &= ~LBF_ALL_PINNED;
 
 	if (task_running(env->src_rq, p)) {
-		schedstat_inc(p->se.statistics.nr_failed_migrations_running);
+		schedstat_inc(p->stats.nr_failed_migrations_running);
 		return 0;
 	}
 
@@ -7911,12 +7928,12 @@ int can_migrate_task(struct task_struct *p, struct lb_env *env)
 	    env->sd->nr_balance_failed > env->sd->cache_nice_tries) {
 		if (tsk_cache_hot == 1) {
 			schedstat_inc(env->sd->lb_hot_gained[env->idle]);
-			schedstat_inc(p->se.statistics.nr_forced_migrations);
+			schedstat_inc(p->stats.nr_forced_migrations);
 		}
 		return 1;
 	}
 
-	schedstat_inc(p->se.statistics.nr_failed_migrations_hot);
+	schedstat_inc(p->stats.nr_failed_migrations_hot);
 	return 0;
 }
 
@@ -11457,7 +11474,7 @@ int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
 		if (!cfs_rq)
 			goto err;
 
-		se = kzalloc_node(sizeof(struct sched_entity),
+		se = kzalloc_node(sizeof(struct sched_entity_stats),
 				  GFP_KERNEL, cpu_to_node(i));
 		if (!se)
 			goto err_free_rq;
diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 3daf42a..95a7c3a 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1009,8 +1009,8 @@ static void update_curr_rt(struct rq *rq)
 	if (unlikely((s64)delta_exec <= 0))
 		return;
 
-	schedstat_set(curr->se.statistics.exec_max,
-		      max(curr->se.statistics.exec_max, delta_exec));
+	schedstat_set(curr->stats.exec_max,
+		      max(curr->stats.exec_max, delta_exec));
 
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);
diff --git a/kernel/sched/stats.h b/kernel/sched/stats.h
index d8f8eb0..fb6022e 100644
--- a/kernel/sched/stats.h
+++ b/kernel/sched/stats.h
@@ -41,6 +41,7 @@ rq_sched_info_dequeue(struct rq *rq, unsigned long long delta)
 #define   schedstat_val_or_zero(var)	((schedstat_enabled()) ? (var) : 0)
 
 #else /* !CONFIG_SCHEDSTATS: */
+
 static inline void rq_sched_info_arrive  (struct rq *rq, unsigned long long delta) { }
 static inline void rq_sched_info_dequeue(struct rq *rq, unsigned long long delta) { }
 static inline void rq_sched_info_depart  (struct rq *rq, unsigned long long delta) { }
@@ -53,8 +54,26 @@ static inline void rq_sched_info_depart  (struct rq *rq, unsigned long long delt
 # define   schedstat_set(var, val)	do { } while (0)
 # define   schedstat_val(var)		0
 # define   schedstat_val_or_zero(var)	0
+
 #endif /* CONFIG_SCHEDSTATS */
 
+#ifdef CONFIG_FAIR_GROUP_SCHED
+struct sched_entity_stats {
+	struct sched_entity     se;
+	struct sched_statistics stats;
+} __no_randomize_layout;
+#endif
+
+static inline struct sched_statistics *
+__schedstats_from_se(struct sched_entity *se)
+{
+#ifdef CONFIG_FAIR_GROUP_SCHED
+	if (!entity_is_task(se))
+		return &container_of(se, struct sched_entity_stats, se)->stats;
+#endif
+	return &task_of(se)->stats;
+}
+
 #ifdef CONFIG_PSI
 /*
  * PSI tracks state that persists across sleeps, such as iowaits and
diff --git a/kernel/sched/stop_task.c b/kernel/sched/stop_task.c
index f988ebe..0b165a2 100644
--- a/kernel/sched/stop_task.c
+++ b/kernel/sched/stop_task.c
@@ -78,8 +78,8 @@ static void put_prev_task_stop(struct rq *rq, struct task_struct *prev)
 	if (unlikely((s64)delta_exec < 0))
 		delta_exec = 0;
 
-	schedstat_set(curr->se.statistics.exec_max,
-			max(curr->se.statistics.exec_max, delta_exec));
+	schedstat_set(curr->stats.exec_max,
+		      max(curr->stats.exec_max, delta_exec));
 
 	curr->se.sum_exec_runtime += delta_exec;
 	account_group_exec_runtime(curr, delta_exec);

^ permalink raw reply related	[flat|nested] 29+ messages in thread

* [tip: sched/core] sched/fair: Use __schedstat_set() in set_next_entity()
  2021-09-05 14:35 ` [PATCH v4 1/8] sched, fair: use __schedstat_set() in set_next_entity() Yafang Shao
  2021-09-16 11:59   ` [tip: sched/core] sched/fair: Use " tip-bot2 for Yafang Shao
@ 2021-10-05 14:12   ` tip-bot2 for Yafang Shao
  1 sibling, 0 replies; 29+ messages in thread
From: tip-bot2 for Yafang Shao @ 2021-10-05 14:12 UTC (permalink / raw)
  To: linux-tip-commits
  Cc: Yafang Shao, Peter Zijlstra (Intel), Mel Gorman, x86, linux-kernel

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     a2dcb276ff9287fcea103ca1a2436383e8583751
Gitweb:        https://git.kernel.org/tip/a2dcb276ff9287fcea103ca1a2436383e8583751
Author:        Yafang Shao <laoar.shao@gmail.com>
AuthorDate:    Sun, 05 Sep 2021 14:35:40 
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Oct 2021 15:51:44 +02:00

sched/fair: Use __schedstat_set() in set_next_entity()

schedstat_enabled() has been already checked, so we can use
__schedstat_set() directly.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lore.kernel.org/r/20210905143547.4668-2-laoar.shao@gmail.com
---
 kernel/sched/fair.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index fd41abe..61d3e3b 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4530,9 +4530,9 @@ set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 	 */
 	if (schedstat_enabled() &&
 	    rq_of(cfs_rq)->cfs.load.weight >= 2*se->load.weight) {
-		schedstat_set(se->statistics.slice_max,
-			max((u64)schedstat_val(se->statistics.slice_max),
-			    se->sum_exec_runtime - se->prev_sum_exec_runtime));
+		__schedstat_set(se->statistics.slice_max,
+				max((u64)se->statistics.slice_max,
+				    se->sum_exec_runtime - se->prev_sum_exec_runtime));
 	}
 
 	se->prev_sum_exec_runtime = se->sum_exec_runtime;

^ permalink raw reply related	[flat|nested] 29+ messages in thread

end of thread, other threads:[~2021-10-05 14:13 UTC | newest]

Thread overview: 29+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-09-05 14:35 [PATCH v4 0/8] sched: support schedstats for RT sched class Yafang Shao
2021-09-05 14:35 ` [PATCH v4 1/8] sched, fair: use __schedstat_set() in set_next_entity() Yafang Shao
2021-09-16 11:59   ` [tip: sched/core] sched/fair: Use " tip-bot2 for Yafang Shao
2021-10-05 14:12   ` tip-bot2 for Yafang Shao
2021-09-05 14:35 ` [PATCH v4 2/8] sched: make struct sched_statistics independent of fair sched class Yafang Shao
2021-09-16 11:59   ` [tip: sched/core] sched: Make " tip-bot2 for Yafang Shao
2021-09-21  6:17     ` Dan Carpenter
2021-09-21  7:17       ` Peter Zijlstra
2021-09-21  7:40         ` Dan Carpenter
2021-10-05 14:12   ` tip-bot2 for Yafang Shao
2021-09-05 14:35 ` [PATCH v4 3/8] sched: make schedstats helpers " Yafang Shao
2021-09-16 11:59   ` [tip: sched/core] sched: Make " tip-bot2 for Yafang Shao
2021-10-05 14:12   ` tip-bot2 for Yafang Shao
2021-09-05 14:35 ` [PATCH v4 4/8] sched: introduce task block time in schedstats Yafang Shao
2021-09-16 11:59   ` [tip: sched/core] sched: Introduce " tip-bot2 for Yafang Shao
2021-10-05 14:12   ` tip-bot2 for Yafang Shao
2021-09-05 14:35 ` [PATCH v4 5/8] sched, rt: support sched_stat_runtime tracepoint for RT sched class Yafang Shao
2021-09-16 11:59   ` [tip: sched/core] sched/rt: Support " tip-bot2 for Yafang Shao
2021-10-05 14:12   ` tip-bot2 for Yafang Shao
2021-09-05 14:35 ` [PATCH v4 6/8] sched, rt: support schedstats " Yafang Shao
2021-09-16 11:59   ` [tip: sched/core] sched/rt: Support " tip-bot2 for Yafang Shao
2021-10-05 14:12   ` tip-bot2 for Yafang Shao
2021-09-05 14:35 ` [PATCH v4 7/8] sched, dl: support sched_stat_runtime tracepoint for deadline " Yafang Shao
2021-09-16 11:59   ` [tip: sched/core] sched/dl: Support " tip-bot2 for Yafang Shao
2021-10-05 14:12   ` tip-bot2 for Yafang Shao
2021-09-05 14:35 ` [PATCH v4 8/8] sched, dl: support schedstats " Yafang Shao
2021-09-16 11:59   ` [tip: sched/core] sched/dl: Support " tip-bot2 for Yafang Shao
2021-10-05 14:12   ` tip-bot2 for Yafang Shao
2021-09-10 17:00 ` [PATCH v4 0/8] sched: support schedstats for RT " Peter Zijlstra

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).