sched: Introduce CONFIG_SCHED_FREQ_INPUT

Introduce a compile time flag to enable scheduler guidance of
frequency selection. This flag is also used to turn on or off
window-based load stats feature.

Having a compile time flag will let some platforms avoid any
overhead that may be present with this scheduler feature.

Change-Id: Id8dec9839f90dcac82f58ef7e2bd0ccd0b6bd16c
Signed-off-by: Srivatsa Vaddagiri <vatsa@codeaurora.org>
[rameezmustafa@codeaurora.org]: Port to msm-3.18]
Signed-off-by: Syed Rameez Mustafa <rameezmustafa@codeaurora.org>
[joonwoop@codeaurora.org: fixed minor conflict around
 sysctl_timer_migration.]
Signed-off-by: Joonwoo Park <joonwoop@codeaurora.org>

drivers/cpufreq/Kconfig

@@ -25,6 +25,16 @@ config CPU_FREQ_BOOST_SW
 	bool
 	depends on THERMAL
 
+config SCHED_FREQ_INPUT
+	bool "Scheduler inputs to cpufreq governor"
+	depends on SMP && FAIR_GROUP_SCHED
+	help
+	  This option enables support for scheduler based CPU utilization
+	  calculations which may then be used by any cpufreq governor. The
+	  scheduler keeps track of "recent" cpu demand of tasks, which can
+	  help determine need for changing frequency well in advance of what
+	  a governor would have been able to detect on its own.
+
 config CPU_FREQ_STAT
 	tristate "CPU frequency translation statistics"
 	default y

include/linux/sched.h

@@ -1424,7 +1424,9 @@ struct task_struct {
 	const struct sched_class *sched_class;
 	struct sched_entity se;
 	struct sched_rt_entity rt;
+#ifdef CONFIG_SCHED_FREQ_INPUT
 	struct ravg ravg;
+#endif
 #ifdef CONFIG_CGROUP_SCHED
 	struct task_group *sched_task_group;
 #endif

kernel/sched/core.c

@@ -1746,6 +1746,8 @@ static inline void ttwu_activate(struct rq *rq, struct task_struct *p, int en_fl
 		wq_worker_waking_up(p, cpu_of(rq));
 }
 
+#ifdef CONFIG_SCHED_FREQ_INPUT
+
 /* Window size (in ns) */
 __read_mostly unsigned int sched_ravg_window = 10000000;
 
@@ -1891,6 +1893,8 @@ void update_task_ravg(struct task_struct *p, struct rq *rq, int update_sum)
 	p->ravg.mark_start = wallclock;
 }
 
+#endif	/* CONFIG_SCHED_FREQ_INPUT */
+
 /*
  * Mark the task runnable and perform wakeup-preemption.
  */
@@ -7569,6 +7573,7 @@ void __init sched_init_smp(void)
 #endif /* CONFIG_SMP */
 
 
+#ifdef CONFIG_SCHED_FREQ_INPUT
 /*
  * Maximum possible frequency across all cpus. Task demand and cpu
  * capacity (cpu_power) metrics are scaled in reference to it.
@@ -7655,6 +7660,8 @@ static int register_sched_callback(void)
  */
 core_initcall(register_sched_callback);
 
+#endif /* CONFIG_SCHED_FREQ_INPUT */
+
 int in_sched_functions(unsigned long addr)
 {
 	return in_lock_functions(addr) ||
@@ -7792,11 +7799,13 @@ void __init sched_init(void)
 		rq->online = 0;
 		rq->idle_stamp = 0;
 		rq->avg_idle = 2*sysctl_sched_migration_cost;
+#ifdef CONFIG_SCHED_FREQ_INPUT
 		rq->cur_freq = 1;
 		rq->max_freq = 1;
 		rq->min_freq = 1;
 		rq->max_possible_freq = 1;
 		rq->cumulative_runnable_avg = 0;
+#endif
 		rq->max_idle_balance_cost = sysctl_sched_migration_cost;
 		rq->cstate = 0;
 		rq->wakeup_latency = 0;

kernel/sched/fair.c

@@ -2894,6 +2894,8 @@ static inline int idle_balance(struct rq *rq)
 
 #endif /* CONFIG_SMP */
 
+#ifdef CONFIG_SCHED_FREQ_INPUT
+
 static inline unsigned int task_load(struct task_struct *p)
 {
 	return p->ravg.demand;
@@ -2927,6 +2929,8 @@ void init_new_task_load(struct task_struct *p)
 		p->ravg.sum_history[i] = 0;
 }
 
+#endif /* CONFIG_SCHED_FREQ_INPUT */
+
 static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
 #ifdef CONFIG_SCHEDSTATS

kernel/sched/sched.h

@@ -643,12 +643,14 @@ struct rq {
 	u64 max_idle_balance_cost;
 #endif
 
+#ifdef CONFIG_SCHED_FREQ_INPUT
 	/*
 	 * max_freq = user or thermal defined maximum
 	 * max_possible_freq = maximum supported by hardware
 	 */
 	unsigned int cur_freq, max_freq, min_freq, max_possible_freq;
 	u64 cumulative_runnable_avg;
+#endif
 
 #ifdef CONFIG_IRQ_TIME_ACCOUNTING
 	u64 prev_irq_time;
@@ -915,6 +917,8 @@ static inline void sched_ttwu_pending(void) { }
 #include "stats.h"
 #include "auto_group.h"
 
+#ifdef CONFIG_SCHED_FREQ_INPUT
+
 extern unsigned int sched_ravg_window;
 extern unsigned int max_possible_freq;
 extern unsigned int min_max_freq;
@@ -934,6 +938,24 @@ dec_cumulative_runnable_avg(struct rq *rq, struct task_struct *p)
 	BUG_ON((s64)rq->cumulative_runnable_avg < 0);
 }
 
+#else	/* CONFIG_SCHED_FREQ_INPUT */
+
+static inline int pct_task_load(struct task_struct *p) { return 0; }
+
+static inline void
+inc_cumulative_runnable_avg(struct rq *rq, struct task_struct *p)
+{
+}
+
+static inline void
+dec_cumulative_runnable_avg(struct rq *rq, struct task_struct *p)
+{
+}
+
+static inline void init_new_task_load(struct task_struct *p) { }
+
+#endif	/* CONFIG_SCHED_FREQ_INPUT */
+
 #ifdef CONFIG_CGROUP_SCHED
 
 /*
@@ -1267,8 +1289,15 @@ struct sched_class {
 #endif
 };
 
+#ifdef CONFIG_SCHED_FREQ_INPUT
 extern void
 update_task_ravg(struct task_struct *p, struct rq *rq, int update_sum);
+#else	/* CONFIG_SCHED_FREQ_INPUT */
+static inline void
+update_task_ravg(struct task_struct *p, struct rq *rq, int update_sum)
+{
+}
+#endif	/* CONFIG_SCHED_FREQ_INPUT */
 
 static inline void put_prev_task(struct rq *rq, struct task_struct *prev)
 {

kernel/sysctl.c

@@ -292,6 +292,7 @@ static struct ctl_table kern_table[] = {
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec,
 	},
+#ifdef CONFIG_SCHED_FREQ_INPUT
 	{
 		.procname       = "sched_window_stats_policy",
 		.data           = &sysctl_sched_window_stats_policy,
@@ -306,6 +307,7 @@ static struct ctl_table kern_table[] = {
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec,
 	},
+#endif
 #ifdef CONFIG_SCHED_DEBUG
 	{
 		.procname	= "sched_min_granularity_ns",