sched: add affinity, task load information to sched tracepoints

Knowing the affinity mask and CPU usage of a task is helpful
in understanding the behavior of the system. Affinity information
has been added to the enq_deq trace event, and the migration
tracepoint now reports the load of the task migrated.

Change-Id: I29d8a610292b4dfeeb8fe16174e9d4dc196649b7
Signed-off-by: Steve Muckle <smuckle@codeaurora.org>
Signed-off-by: Syed Rameez Mustafa <rameezmustafa@codeaurora.org>

3 files changed, 21 insertions, 13 deletions

diff --git a/include/trace/events/sched.h b/include/trace/events/sched.h
index e9569c65330d..99f93e6a9235 100644
--- a/include/trace/events/sched.h
+++ b/include/trace/events/sched.h
@@ -57,9 +57,9 @@ TRACE_EVENT(sched_kthread_stop_ret,
  */
 TRACE_EVENT(sched_enq_deq_task,
 
-	TP_PROTO(struct task_struct *p, int enqueue),
+	TP_PROTO(struct task_struct *p, int enqueue, unsigned int cpus_allowed),
 
-	TP_ARGS(p, enqueue),
+	TP_ARGS(p, enqueue, cpus_allowed),
 
 	TP_STRUCT__entry(
 		__array(	char,	comm,	TASK_COMM_LEN	)
@@ -70,6 +70,7 @@ TRACE_EVENT(sched_enq_deq_task,
 		__field(unsigned int,	nr_running		)
 		__field(unsigned long,	cpu_load		)
 		__field(unsigned int,	rt_nr_running		)
+		__field(unsigned int,	cpus_allowed		)
 #ifdef CONFIG_SCHED_FREQ_INPUT
 		__field(unsigned int,	sum_scaled		)
 		__field(unsigned int,	period			)
@@ -86,6 +87,7 @@ TRACE_EVENT(sched_enq_deq_task,
 		__entry->nr_running	= task_rq(p)->nr_running;
 		__entry->cpu_load	= task_rq(p)->cpu_load[0];
 		__entry->rt_nr_running	= task_rq(p)->rt.rt_nr_running;
+		__entry->cpus_allowed	= cpus_allowed;
 #ifdef CONFIG_SCHED_FREQ_INPUT
 		__entry->sum_scaled	= p->se.avg.runnable_avg_sum_scaled;
 		__entry->period		= p->se.avg.runnable_avg_period;
@@ -93,15 +95,15 @@ TRACE_EVENT(sched_enq_deq_task,
 #endif
 	),
 
-	TP_printk("cpu=%d %s comm=%s pid=%d prio=%d nr_running=%u cpu_load=%lu rt_nr_running=%u"
+	TP_printk("cpu=%d %s comm=%s pid=%d prio=%d nr_running=%u cpu_load=%lu rt_nr_running=%u affine=%x"
 #ifdef CONFIG_SCHED_FREQ_INPUT
-		 "sum_scaled=%u period=%u demand=%u"
+		 " sum_scaled=%u period=%u demand=%u"
 #endif
 			, __entry->cpu,
 			__entry->enqueue ? "enqueue" : "dequeue",
 			__entry->comm, __entry->pid,
 			__entry->prio, __entry->nr_running,
-			__entry->cpu_load, __entry->rt_nr_running
+			__entry->cpu_load, __entry->rt_nr_running, __entry->cpus_allowed
 #ifdef CONFIG_SCHED_FREQ_INPUT
 			, __entry->sum_scaled, __entry->period, __entry->demand
 #endif
@@ -303,14 +305,16 @@ TRACE_EVENT(sched_switch,
  */
 TRACE_EVENT(sched_migrate_task,
 
-	TP_PROTO(struct task_struct *p, int dest_cpu),
+	TP_PROTO(struct task_struct *p, int dest_cpu,
+		 unsigned int load),
 
-	TP_ARGS(p, dest_cpu),
+	TP_ARGS(p, dest_cpu, load),
 
 	TP_STRUCT__entry(
 		__array(	char,	comm,	TASK_COMM_LEN	)
 		__field(	pid_t,	pid			)
 		__field(	int,	prio			)
+		__field(unsigned int,	load			)
 		__field(	int,	orig_cpu		)
 		__field(	int,	dest_cpu		)
 	),
@@ -319,12 +323,13 @@ TRACE_EVENT(sched_migrate_task,
 		memcpy(__entry->comm, p->comm, TASK_COMM_LEN);
 		__entry->pid		= p->pid;
 		__entry->prio		= p->prio;
+		__entry->load		= load;
 		__entry->orig_cpu	= task_cpu(p);
 		__entry->dest_cpu	= dest_cpu;
 	),
 
-	TP_printk("comm=%s pid=%d prio=%d orig_cpu=%d dest_cpu=%d",
-		  __entry->comm, __entry->pid, __entry->prio,
+	TP_printk("comm=%s pid=%d prio=%d load=%d orig_cpu=%d dest_cpu=%d",
+		  __entry->comm, __entry->pid, __entry->prio,  __entry->load,
 		  __entry->orig_cpu, __entry->dest_cpu)
 );
 
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index a96e2225755a..1352245a0fc4 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -854,7 +854,7 @@ static inline void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
 	if (!(flags & ENQUEUE_RESTORE))
 		sched_info_queued(rq, p);
 	p->sched_class->enqueue_task(rq, p, flags);
-	trace_sched_enq_deq_task(p, 1);
+	trace_sched_enq_deq_task(p, 1, cpumask_bits(&p->cpus_allowed)[0]);
 	inc_cumulative_runnable_avg(rq, p);
 }
 
@@ -864,7 +864,7 @@ static inline void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
 	if (!(flags & DEQUEUE_SAVE))
 		sched_info_dequeued(rq, p);
 	p->sched_class->dequeue_task(rq, p, flags);
-	trace_sched_enq_deq_task(p, 0);
+	trace_sched_enq_deq_task(p, 0, cpumask_bits(&p->cpus_allowed)[0]);
 	dec_cumulative_runnable_avg(rq, p);
 }
 
@@ -1702,7 +1702,7 @@ void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
 #endif
 #endif
 
-	trace_sched_migrate_task(p, new_cpu);
+	trace_sched_migrate_task(p, new_cpu, pct_task_load(p));
 
 	if (task_cpu(p) != new_cpu) {
 		if (p->sched_class->migrate_task_rq)
@@ -8070,6 +8070,8 @@ void __init sched_init(void)
 	int i, j;
 	unsigned long alloc_size = 0, ptr;
 
+	BUG_ON(num_possible_cpus() > BITS_PER_LONG);
+
 #ifdef CONFIG_FAIR_GROUP_SCHED
 	alloc_size += 2 * nr_cpu_ids * sizeof(void **);
 #endif
diff --git a/kernel/trace/trace_sched_wakeup.c b/kernel/trace/trace_sched_wakeup.c
index 9d4399b553a3..78f04e4ad829 100644
--- a/kernel/trace/trace_sched_wakeup.c
+++ b/kernel/trace/trace_sched_wakeup.c
@@ -359,7 +359,8 @@ static bool report_latency(struct trace_array *tr, cycle_t delta)
 }
 
 static void
-probe_wakeup_migrate_task(void *ignore, struct task_struct *task, int cpu)
+probe_wakeup_migrate_task(void *ignore, struct task_struct *task, int cpu,
+							unsigned int load)
 {
 	if (task != wakeup_task)
 		return;