sched: Add new trace events

Add trace events for update_task_ravg(), update_history(), and
set_task_cpu(). These tracepoints are useful for monitoring the
per-task and per-runqueue demand statistics.

Change-Id: Ibec9f945074ff31d1fc1a76ae37c40c8fea8cda9
Signed-off-by: Srivatsa Vaddagiri <vatsa@codeaurora.org>

3 files changed, 152 insertions, 6 deletions

diff --git a/include/linux/sched.h b/include/linux/sched.h
index f22f67de7320..266dca375d59 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -314,6 +314,14 @@ extern char ___assert_task_state[1 - 2*!!(
 /* Task command name length */
 #define TASK_COMM_LEN 16
 
+enum task_event {
+	PUT_PREV_TASK   = 0,
+	PICK_NEXT_TASK  = 1,
+	TASK_WAKE       = 2,
+	TASK_MIGRATE    = 3,
+	TASK_UPDATE     = 4,
+};
+
 #include <linux/spinlock.h>
 
 /*
diff --git a/include/trace/events/sched.h b/include/trace/events/sched.h
index 99f93e6a9235..cfe4bc63db7c 100644
--- a/include/trace/events/sched.h
+++ b/include/trace/events/sched.h
@@ -9,6 +9,7 @@
 #include <linux/binfmts.h>
 
 struct rq;
+extern const char *task_event_names[];
 
 /*
  * Tracepoint for calling kthread_stop, performed to end a kthread:
@@ -188,6 +189,139 @@ TRACE_EVENT(sched_cpu_load,
 
 #endif	/* CONFIG_SCHED_HMP */
 
+#if defined(CONFIG_SCHED_FREQ_INPUT) || defined(CONFIG_SCHED_HMP)
+
+TRACE_EVENT(sched_update_task_ravg,
+
+	TP_PROTO(struct task_struct *p, struct rq *rq, enum task_event evt,
+								u64 wallclock),
+
+	TP_ARGS(p, rq, evt, wallclock),
+
+	TP_STRUCT__entry(
+		__array(	char,	comm,   TASK_COMM_LEN	)
+		__field(	pid_t,	pid			)
+		__field(	pid_t,	cur_pid			)
+		__field(unsigned int,	cur_freq		)
+		__field(unsigned int,	cs			)
+		__field(unsigned int,	ps			)
+		__field(	u64,	wallclock		)
+		__field(	u64,	mark_start		)
+		__field(	u64,	delta_m			)
+		__field(	u64,	win_start		)
+		__field(	u64,	delta			)
+		__field(enum task_event,	evt		)
+		__field(unsigned int,	demand			)
+		__field(unsigned int,	partial_demand		)
+		__field(unsigned int,	sum			)
+		__field(unsigned int,	prev_window		)
+		__field(	 int,	cpu			)
+	),
+
+	TP_fast_assign(
+		__entry->wallclock      = wallclock;
+		__entry->win_start      = rq->window_start;
+		__entry->delta          = (wallclock - rq->window_start);
+		__entry->evt            = evt;
+		__entry->cpu            = rq->cpu;
+		__entry->cur_pid        = rq->curr->pid;
+		__entry->cur_freq       = rq->cur_freq;
+		__entry->cs             = rq->curr_runnable_sum;
+		__entry->ps             = rq->prev_runnable_sum;
+		memcpy(__entry->comm, p->comm, TASK_COMM_LEN);
+		__entry->pid            = p->pid;
+		__entry->mark_start     = p->ravg.mark_start;
+		__entry->delta_m        = (wallclock - p->ravg.mark_start);
+		__entry->demand         = p->ravg.demand;
+		__entry->partial_demand = p->ravg.partial_demand;
+		__entry->sum            = p->ravg.sum;
+		__entry->prev_window    = p->ravg.prev_window;
+	),
+
+	TP_printk("wc %llu ws %llu delta %llu event %s cpu %d cur_freq %u cs %u ps %u cur_pid %d task %d (%s) ms %llu delta %llu demand %u partial_demand %u sum %u prev_window %u",
+		__entry->wallclock, __entry->win_start, __entry->delta,
+		task_event_names[__entry->evt], __entry->cpu,
+		__entry->cur_freq, __entry->cs, __entry->ps, __entry->cur_pid,
+		__entry->pid, __entry->comm, __entry->mark_start,
+		__entry->delta_m, __entry->demand, __entry->partial_demand,
+		__entry->sum, __entry->prev_window)
+);
+
+TRACE_EVENT(sched_update_history,
+
+	TP_PROTO(struct rq *rq, struct task_struct *p, u32 runtime, int samples,
+			int update_sum, int new_window, enum task_event evt),
+
+	TP_ARGS(rq, p, runtime, samples, update_sum, new_window, evt),
+
+	TP_STRUCT__entry(
+		__array(	char,	comm,   TASK_COMM_LEN	)
+		__field(	pid_t,	pid			)
+		__field(unsigned int,	runtime			)
+		__field(	 int,	samples			)
+		__field(	 int,	update_sum		)
+		__field(	 int,	new_window		)
+		__field(enum task_event,	evt		)
+		__field(unsigned int,	partial_demand		)
+		__field(unsigned int,	demand			)
+		__array(         u32,	hist,   RAVG_HIST_SIZE	)
+		__field(unsigned int,	nr_big_tasks		)
+		__field(unsigned int,	nr_small_tasks		)
+		__field(	 int,	cpu			)
+	),
+
+	TP_fast_assign(
+		memcpy(__entry->comm, p->comm, TASK_COMM_LEN);
+		__entry->pid            = p->pid;
+		__entry->runtime        = runtime;
+		__entry->samples        = samples;
+		__entry->update_sum     = update_sum;
+		__entry->new_window     = new_window;
+		__entry->evt            = evt;
+		__entry->partial_demand = p->ravg.partial_demand;
+		__entry->demand         = p->ravg.demand;
+		memcpy(__entry->hist, p->ravg.sum_history,
+					RAVG_HIST_SIZE*sizeof(u32));
+#ifdef CONFIG_SCHED_HMP
+		__entry->nr_big_tasks   = rq->nr_big_tasks;
+		__entry->nr_small_tasks = rq->nr_small_tasks;
+#endif
+		__entry->cpu            = rq->cpu;
+	),
+
+	TP_printk("%d (%s): runtime %u samples %d us %d nw %d event %s partial_demand %u demand %u (hist: %u %u %u %u %u) cpu %d nr_big %u nr_small %u",
+		__entry->pid, __entry->comm,
+		__entry->runtime, __entry->samples, __entry->update_sum,
+		__entry->new_window, task_event_names[__entry->evt],
+		__entry->partial_demand, __entry->demand, __entry->hist[0],
+		__entry->hist[1], __entry->hist[2], __entry->hist[3],
+		__entry->hist[4], __entry->cpu, __entry->nr_big_tasks,
+		__entry->nr_small_tasks)
+);
+
+TRACE_EVENT(sched_migration_update_sum,
+
+	TP_PROTO(struct rq *rq),
+
+	TP_ARGS(rq),
+
+	TP_STRUCT__entry(
+		__field(int,		cpu			)
+		__field(int,		cs			)
+		__field(int,		ps			)
+	),
+
+	TP_fast_assign(
+		__entry->cpu		= cpu_of(rq);
+		__entry->cs		= rq->curr_runnable_sum;
+		__entry->ps		= rq->prev_runnable_sum;
+	),
+
+	TP_printk("cpu %d: cs %u ps %u\n", __entry->cpu,
+		      __entry->cs, __entry->ps)
+);
+
+#endif /* CONFIG_SCHED_FREQ_INPUT || CONFIG_SCHED_HMP */
 
 /*
  * Tracepoint for waking up a task:
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 8888c1081ef3..69f8aef5a849 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -91,6 +91,9 @@
 #define CREATE_TRACE_POINTS
 #include <trace/events/sched.h>
 
+const char *task_event_names[] = {"PUT_PREV_TASK", "PICK_NEXT_TASK",
+				  "TASK_WAKE", "TASK_MIGRATE", "TASK_UPDATE"};
+
 ATOMIC_NOTIFIER_HEAD(migration_notifier_head);
 ATOMIC_NOTIFIER_HEAD(load_alert_notifier_head);
 
@@ -1072,12 +1075,6 @@ void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
 		rq_clock_skip_update(rq, true);
 }
 
-#define PUT_PREV_TASK  0
-#define PICK_NEXT_TASK 1
-#define TASK_WAKE      2
-#define TASK_MIGRATE   3
-#define TASK_UPDATE    4
-
 #if defined(CONFIG_SCHED_FREQ_INPUT) || defined(CONFIG_SCHED_HMP)
 
 /* Window size (in ns) */
@@ -1212,6 +1209,8 @@ compute_demand:
 		if (p->sched_class == &fair_sched_class)
 			inc_nr_big_small_task(rq, p);
 	}
+	trace_sched_update_history(rq, p, runtime, samples, update_sum,
+				   new_window, event);
 }
 
 static int __init set_sched_ravg_window(char *str)
@@ -1352,6 +1351,8 @@ static void update_task_ravg(struct task_struct *p, struct rq *rq,
 	if (event == PICK_NEXT_TASK && !p->ravg.sum)
 		rq->curr_runnable_sum += p->ravg.partial_demand;
 
+	trace_sched_update_task_ravg(p, rq, event, wallclock);
+
 	p->ravg.mark_start = wallclock;
 }
 
@@ -1839,6 +1840,9 @@ void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
 			BUG_ON((int)src_rq->prev_runnable_sum < 0);
 			BUG_ON((int)src_rq->curr_runnable_sum < 0);
 
+			trace_sched_migration_update_sum(src_rq);
+			trace_sched_migration_update_sum(dest_rq);
+
 			if (p->state == TASK_WAKING)
 				double_rq_unlock(src_rq, dest_rq);