sched: Update the wakeup placement logic for fair and rt tasks

For the fair sched class, update the select_best_cpu() policy to do
power based placement. The hope is to minimize the voltage at which
the CPU runs.

While RT tasks already do power based placement, their placement
preference has to now take into account the power cost of all tasks
on a given CPU. Also remove the check for sched_boost since
sched_boost no longer intends to elevate all tasks to the highest
capacity cluster.

Change-Id: Ic6a7625c97d567254d93b94cec3174a91727cb87
Signed-off-by: Syed Rameez Mustafa <rameezmustafa@codeaurora.org>

2 files changed, 11 insertions, 24 deletions

diff --git a/include/linux/sched.h b/include/linux/sched.h
index ba49767ee09f..fa5ca5e90917 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -2339,15 +2339,6 @@ extern u32 sched_get_wake_up_idle(struct task_struct *p);
 extern int sched_set_boost(int enable);
 extern int sched_set_init_task_load(struct task_struct *p, int init_load_pct);
 extern u32 sched_get_init_task_load(struct task_struct *p);
-extern int sched_set_cpu_prefer_idle(int cpu, int prefer_idle);
-extern int sched_get_cpu_prefer_idle(int cpu);
-extern int sched_set_cpu_mostly_idle_load(int cpu, int mostly_idle_pct);
-extern int sched_get_cpu_mostly_idle_load(int cpu);
-extern int sched_set_cpu_mostly_idle_nr_run(int cpu, int nr_run);
-extern int sched_get_cpu_mostly_idle_nr_run(int cpu);
-extern int
-sched_set_cpu_mostly_idle_freq(int cpu, unsigned int mostly_idle_freq);
-extern unsigned int sched_get_cpu_mostly_idle_freq(int cpu);
 
 #else
 static inline int sched_set_boost(int enable)
diff --git a/include/trace/events/sched.h b/include/trace/events/sched.h
index 4da5f52da450..3b10396d61b6 100644
--- a/include/trace/events/sched.h
+++ b/include/trace/events/sched.h
@@ -116,9 +116,9 @@ TRACE_EVENT(sched_enq_deq_task,
 TRACE_EVENT(sched_task_load,
 
 	TP_PROTO(struct task_struct *p, int boost, int reason,
-		 int sync, int prefer_idle),
+		 int sync, int need_idle),
 
-	TP_ARGS(p, boost, reason, sync, prefer_idle),
+	TP_ARGS(p, boost, reason, sync, need_idle),
 
 	TP_STRUCT__entry(
 		__array(	char,	comm,	TASK_COMM_LEN	)
@@ -130,7 +130,7 @@ TRACE_EVENT(sched_task_load,
 		__field(	int,	boost			)
 		__field(	int,	reason			)
 		__field(	int,	sync			)
-		__field(	int,	prefer_idle		)
+		__field(	int,	need_idle		)
 	),
 
 	TP_fast_assign(
@@ -143,27 +143,24 @@ TRACE_EVENT(sched_task_load,
 		__entry->boost		= boost;
 		__entry->reason		= reason;
 		__entry->sync		= sync;
-		__entry->prefer_idle	= prefer_idle;
+		__entry->need_idle	= need_idle;
 	),
 
-	TP_printk("%d (%s): sum=%u, sum_scaled=%u, period=%u demand=%u boost=%d reason=%d sync=%d prefer_idle=%d",
+	TP_printk("%d (%s): sum=%u, sum_scaled=%u, period=%u demand=%u boost=%d reason=%d sync=%d, need_idle=%d",
 		__entry->pid, __entry->comm, __entry->sum,
 		__entry->sum_scaled, __entry->period, __entry->demand,
-		__entry->boost, __entry->reason, __entry->sync,
-		__entry->prefer_idle)
+		__entry->boost, __entry->reason, __entry->sync, __entry->need_idle)
 );
 
 TRACE_EVENT(sched_cpu_load,
 
-	TP_PROTO(struct rq *rq, int idle, int mostly_idle, u64 irqload,
-		 unsigned int power_cost, int temp),
+	TP_PROTO(struct rq *rq, int idle, u64 irqload, unsigned int power_cost, int temp),
 
-	TP_ARGS(rq, idle, mostly_idle, irqload, power_cost, temp),
+	TP_ARGS(rq, idle, irqload, power_cost, temp),
 
 	TP_STRUCT__entry(
 		__field(unsigned int, cpu			)
 		__field(unsigned int, idle			)
-		__field(unsigned int, mostly_idle		)
 		__field(unsigned int, nr_running		)
 		__field(unsigned int, nr_big_tasks		)
 		__field(unsigned int, load_scale_factor		)
@@ -180,7 +177,6 @@ TRACE_EVENT(sched_cpu_load,
 	TP_fast_assign(
 		__entry->cpu			= rq->cpu;
 		__entry->idle			= idle;
-		__entry->mostly_idle		= mostly_idle;
 		__entry->nr_running		= rq->nr_running;
 		__entry->nr_big_tasks		= rq->hmp_stats.nr_big_tasks;
 		__entry->load_scale_factor	= rq->load_scale_factor;
@@ -194,9 +190,9 @@ TRACE_EVENT(sched_cpu_load,
 		__entry->temp			= temp;
 	),
 
-	TP_printk("cpu %u idle %d mostly_idle %d nr_run %u nr_big %u lsf %u capacity %u cr_avg %llu irqload %llu fcur %u fmax %u power_cost %u cstate %d temp %d",
-	__entry->cpu, __entry->idle, __entry->mostly_idle, __entry->nr_running,
-	__entry->nr_big_tasks, __entry->load_scale_factor, __entry->capacity,
+	TP_printk("cpu %u idle %d nr_run %u nr_big %u lsf %u capacity %u cr_avg %llu irqload %llu fcur %u fmax %u power_cost %u cstate %d temp %d",
+	__entry->cpu, __entry->idle, __entry->nr_running, __entry->nr_big_tasks,
+	__entry->load_scale_factor, __entry->capacity,
 	__entry->cumulative_runnable_avg, __entry->irqload, __entry->cur_freq,
 	__entry->max_freq, __entry->power_cost, __entry->cstate, __entry->temp)
 );