sched: remove sysctl control for HMP and power-aware task placement

There is no real need to control HMP and power-aware task placement at
runtime after kernel has booted. Boot-time control should be
sufficient. Not allowing for runtime (sysctl) support simplifies the
code quite a bit.

Also rename sysctl_sched_enable_hmp_task_placement to be shorter.

Change-Id: I60cae51a173c6f73b79cbf90c50ddd41a27604aa
Signed-off-by: Srivatsa Vaddagiri <vatsa@codeaurora.org>
Signed-off-by: Syed Rameez Mustafa <rameezmustafa@codeaurora.org>
[joonwoop@codeaurora.org: fixed minor conflict.  p->nr_cpus_allowed == 1
 has moved to core.c
Signed-off-by: Joonwoo Park <joonwoop@codeaurora.org>

include/linux/sched/sysctl.h

@@ -53,7 +53,6 @@ extern int sysctl_sched_freq_dec_notify_slack_pct;
 #endif
 
 #ifdef CONFIG_SCHED_HMP
-extern unsigned int sysctl_sched_enable_hmp_task_placement;
 extern unsigned int sysctl_sched_spill_nr_run;
 extern unsigned int sysctl_sched_mostly_idle_nr_run;
 extern unsigned int sysctl_sched_spill_load_pct;
@@ -62,7 +61,6 @@ extern unsigned int sysctl_sched_small_task_pct;
 extern unsigned int sysctl_sched_upmigrate_pct;
 extern unsigned int sysctl_sched_downmigrate_pct;
 extern int sysctl_sched_upmigrate_min_nice;
-extern unsigned int sysctl_sched_enable_power_aware;
 extern unsigned int sysctl_sched_powerband_limit_pct;
 extern unsigned int sysctl_sched_boost;
 

kernel/sched/core.c

@@ -1075,6 +1075,36 @@ void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
 		rq_clock_skip_update(rq, true);
 }
 
+#ifdef CONFIG_SCHED_HMP
+
+static int __init set_sched_enable_hmp(char *str)
+{
+	int enable_hmp = 0;
+
+	get_option(&str, &enable_hmp);
+
+	sched_enable_hmp = !!enable_hmp;
+
+	return 0;
+}
+
+early_param("sched_enable_hmp", set_sched_enable_hmp);
+
+static int __init set_sched_enable_power_aware(char *str)
+{
+	int enable_power_aware = 0;
+
+	get_option(&str, &enable_power_aware);
+
+	sched_enable_power_aware = !!enable_power_aware;
+
+	return 0;
+}
+
+early_param("sched_enable_power_aware", set_sched_enable_power_aware);
+
+#endif	/* CONFIG_SCHED_HMP */
+
 #if defined(CONFIG_SCHED_FREQ_INPUT) || defined(CONFIG_SCHED_HMP)
 
 /* Window size (in ns) */
@@ -1135,7 +1165,7 @@ int rq_freq_margin(struct rq *rq)
 	int margin;
 	u64 demand;
 
-	if (!sysctl_sched_enable_hmp_task_placement)
+	if (!sched_enable_hmp)
 		return INT_MAX;
 
 	demand = scale_load_to_cpu(rq->prev_runnable_sum, rq->cpu);
@@ -1393,7 +1423,7 @@ static void init_cpu_efficiency(void)
 	int i, efficiency;
 	unsigned int max = 0, min = UINT_MAX;
 
-	if (!sysctl_sched_enable_hmp_task_placement)
+	if (!sched_enable_hmp)
 		return;
 
 	for_each_possible_cpu(i) {
@@ -1436,7 +1466,7 @@ static inline void set_window_start(struct rq *rq)
 	int cpu = cpu_of(rq);
 	struct rq *sync_rq = cpu_rq(sync_cpu);
 
-	if (rq->window_start || !sysctl_sched_enable_hmp_task_placement)
+	if (rq->window_start || !sched_enable_hmp)
 		return;
 
 	if (cpu == sync_cpu) {
@@ -1720,7 +1750,7 @@ static int register_sched_callback(void)
 {
 	int ret;
 
-	if (!sysctl_sched_enable_hmp_task_placement)
+	if (!sched_enable_hmp)
 		return 0;
 
 	ret = cpufreq_register_notifier(&notifier_policy_block,
@@ -2112,8 +2142,7 @@ void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
 		p->se.nr_migrations++;
 		perf_event_task_migrate(p);
 
-		if (sysctl_sched_enable_hmp_task_placement &&
+		if (sched_enable_hmp && (p->on_rq || p->state == TASK_WAKING))
-		    (p->on_rq || p->state == TASK_WAKING))
 			fixup_busy_time(p, new_cpu);
 	}
 
@@ -3672,7 +3701,7 @@ void sched_exec(void)
 	unsigned long flags;
 	int dest_cpu;
 
-	if (sysctl_sched_enable_hmp_task_placement)
+	if (sched_enable_hmp)
 		return;
 
 	raw_spin_lock_irqsave(&p->pi_lock, flags);

kernel/sched/fair.c

@@ -2564,7 +2564,7 @@ unsigned int max_task_load(void)
 #ifdef CONFIG_SCHED_HMP
 
 /* Use this knob to turn on or off HMP-aware task placement logic */
-unsigned int __read_mostly sysctl_sched_enable_hmp_task_placement = 1;
+unsigned int __read_mostly sched_enable_hmp = 1;
 
 /* A cpu can no longer accomodate more tasks if:
  *
@@ -2585,7 +2585,7 @@ unsigned int __read_mostly sysctl_sched_mostly_idle_nr_run = 3;
  * Control whether or not individual CPU power consumption is used to
  * guide task placement.
  */
-unsigned int __read_mostly sysctl_sched_enable_power_aware = 1;
+unsigned int __read_mostly sched_enable_power_aware = 1;
 
 /*
  * This specifies the maximum percent power difference between 2
@@ -2767,7 +2767,7 @@ int sched_set_boost(int enable)
 	unsigned long flags;
 	int ret = 0;
 
-	if (!sysctl_sched_enable_hmp_task_placement)
+	if (!sched_enable_hmp)
 		return -EINVAL;
 
 	spin_lock_irqsave(&boost_lock, flags);
@@ -2887,7 +2887,7 @@ unsigned int power_cost_at_freq(int cpu, unsigned int freq)
 	struct cpu_pstate_pwr *costs;
 
 	if (!per_cpu_info || !per_cpu_info[cpu].ptable ||
-	    !sysctl_sched_enable_power_aware)
+	    !sched_enable_power_aware)
 		/* When power aware scheduling is not in use, or CPU
 		 * power data is not available, just use the CPU
 		 * capacity as a rough stand-in for real CPU power
@@ -2918,7 +2918,7 @@ static unsigned int power_cost(struct task_struct *p, int cpu)
 	unsigned int task_freq;
 	unsigned int cur_freq = cpu_rq(cpu)->cur_freq;
 
-	if (!sysctl_sched_enable_power_aware)
+	if (!sched_enable_power_aware)
 		return cpu_rq(cpu)->max_possible_capacity;
 
 	/* calculate % of max freq needed */
@@ -3075,7 +3075,7 @@ done:
 
 void inc_nr_big_small_task(struct rq *rq, struct task_struct *p)
 {
-	if (!sysctl_sched_enable_hmp_task_placement)
+	if (!sched_enable_hmp)
 		return;
 
 	if (is_big_task(p))
@@ -3086,7 +3086,7 @@ void inc_nr_big_small_task(struct rq *rq, struct task_struct *p)
 
 void dec_nr_big_small_task(struct rq *rq, struct task_struct *p)
 {
-	if (!sysctl_sched_enable_hmp_task_placement)
+	if (!sched_enable_hmp)
 		return;
 
 	if (is_big_task(p))
@@ -3154,7 +3154,7 @@ int sched_hmp_proc_update_handler(struct ctl_table *table, int write,
 	unsigned int old_val = *data;
 
 	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
-	if (ret || !write || !sysctl_sched_enable_hmp_task_placement)
+	if (ret || !write || !sched_enable_hmp)
 		return ret;
 
 	if ((sysctl_sched_downmigrate_pct > sysctl_sched_upmigrate_pct) ||
@@ -3278,7 +3278,7 @@ static inline int migration_needed(struct rq *rq, struct task_struct *p)
 	int nice = task_nice(p);
 
 	if (is_small_task(p) || p->state != TASK_RUNNING ||
-			!sysctl_sched_enable_hmp_task_placement)
+			!sched_enable_hmp)
 		return 0;
 
 	/* Todo: cgroup-based control? */
@@ -3289,7 +3289,7 @@ static inline int migration_needed(struct rq *rq, struct task_struct *p)
 	if (!task_will_fit(p, cpu_of(rq)))
 		return 1;
 
-	if (sysctl_sched_enable_power_aware &&
+	if (sched_enable_power_aware &&
 	    lower_power_cpu_available(p, cpu_of(rq)))
 		return 1;
 
@@ -3342,7 +3342,7 @@ static inline int nr_big_tasks(struct rq *rq)
 
 #else	/* CONFIG_SCHED_HMP */
 
-#define sysctl_sched_enable_power_aware 0
+#define sched_enable_power_aware 0
 
 static inline int select_best_cpu(struct task_struct *p, int target)
 {
@@ -3819,7 +3819,7 @@ add_to_scaled_stat(int cpu, struct sched_avg *sa, u64 delta)
 	u64 scaled_delta;
 	int sf;
 
-	if (!sysctl_sched_enable_hmp_task_placement)
+	if (!sched_enable_hmp)
 		return;
 
 	if (unlikely(cur_freq > max_possible_freq ||
@@ -3836,7 +3836,7 @@ add_to_scaled_stat(int cpu, struct sched_avg *sa, u64 delta)
 
 static inline void decay_scaled_stat(struct sched_avg *sa, u64 periods)
 {
-	if (!sysctl_sched_enable_hmp_task_placement)
+	if (!sched_enable_hmp)
 		return;
 
 	sa->runnable_avg_sum_scaled =
@@ -5908,7 +5908,7 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int sd_flag, int wake_f
 	int want_affine = 0;
 	int sync = wake_flags & WF_SYNC;
 
-	if (sysctl_sched_enable_hmp_task_placement)
+	if (sched_enable_hmp)
 		return select_best_cpu(p, prev_cpu);
 
 	if (sd_flag & SD_BALANCE_WAKE)
@@ -7919,7 +7919,7 @@ static struct rq *find_busiest_queue(struct lb_env *env,
 	unsigned long busiest_load = 0, busiest_capacity = 1;
 	int i;
 
-	if (sysctl_sched_enable_hmp_task_placement)
+	if (sched_enable_hmp)
 		return find_busiest_queue_hmp(env, group);
 
 	for_each_cpu_and(i, sched_group_cpus(group), env->cpus) {
@@ -8422,7 +8422,7 @@ static int idle_balance(struct rq *this_rq)
 	 */
 	rcu_read_lock();
 	sd = rcu_dereference_check_sched_domain(this_rq->sd);
-	if (sd && sysctl_sched_enable_power_aware) {
+	if (sd && sched_enable_power_aware) {
 		for_each_cpu(i, sched_domain_span(sd)) {
 			if (i == this_cpu || idle_cpu(i)) {
 				cost = power_cost_at_freq(i, 0);
@@ -8648,7 +8648,7 @@ static inline int find_new_ilb(int type)
 {
 	int ilb;
 
-	if (sysctl_sched_enable_hmp_task_placement)
+	if (sched_enable_hmp)
 		return find_new_hmp_ilb(type);
 
 	ilb = cpumask_first(nohz.idle_cpus_mask);
@@ -8902,7 +8902,7 @@ static int select_lowest_power_cpu(struct cpumask *cpus)
 	int lowest_power_cpu = -1;
 	int lowest_power = INT_MAX;
 
-	if (sysctl_sched_enable_power_aware) {
+	if (sched_enable_power_aware) {
 		for_each_cpu(i, cpus) {
 			cost = power_cost_at_freq(i, 0);
 			if (cost < lowest_power) {
@@ -9029,7 +9029,7 @@ static inline int _nohz_kick_needed(struct rq *rq, int cpu, int *type)
 {
 	unsigned long now = jiffies;
 
-	if (sysctl_sched_enable_hmp_task_placement)
+	if (sched_enable_hmp)
 		return _nohz_kick_needed_hmp(rq, cpu, type);
 
 	/*

kernel/sched/rt.c

@@ -1385,7 +1385,7 @@ select_task_rq_rt(struct task_struct *p, int cpu, int sd_flag, int flags)
 	struct task_struct *curr;
 	struct rq *rq;
 
-	if (sysctl_sched_enable_hmp_task_placement)
+	if (sched_enable_hmp)
 		return select_task_rq_rt_hmp(p, cpu, sd_flag, flags);
 
 	/* For anything but wake ups, just return the task_cpu */
@@ -1680,7 +1680,7 @@ static int find_lowest_rq(struct task_struct *task)
 	int this_cpu = smp_processor_id();
 	int cpu      = task_cpu(task);
 
-	if (sysctl_sched_enable_hmp_task_placement)
+	if (sched_enable_hmp)
 		return find_lowest_rq_hmp(task);
 
 	/* Make sure the mask is initialized first */

kernel/sched/sched.h

@@ -1016,7 +1016,8 @@ static inline unsigned long capacity_scale_cpu_freq(int cpu)
 
 #ifdef CONFIG_SCHED_HMP
 
-extern unsigned int sysctl_sched_enable_hmp_task_placement;
+extern unsigned int sched_enable_hmp;
+extern unsigned int sched_enable_power_aware;
 
 int mostly_idle_cpu(int cpu);
 extern void check_for_migration(struct rq *rq, struct task_struct *p);
@@ -1029,7 +1030,7 @@ extern unsigned int power_cost_at_freq(int cpu, unsigned int freq);
 
 #else /* CONFIG_SCHED_HMP */
 
-#define sysctl_sched_enable_hmp_task_placement 0
+#define sched_enable_hmp 0
 
 static inline void check_for_migration(struct rq *rq, struct task_struct *p) { }
 static inline void pre_big_small_task_count_change(void) { }

kernel/sysctl.c

@@ -325,13 +325,6 @@ static struct ctl_table kern_table[] = {
 	},
 #endif
 #ifdef CONFIG_SCHED_HMP
-	{
-		.procname	= "sched_enable_hmp_task_placement",
-		.data		= &sysctl_sched_enable_hmp_task_placement,
-		.maxlen		= sizeof(unsigned int),
-		.mode		= 0644,
-		.proc_handler	= proc_dointvec,
-	},
 	{
 		.procname	= "sched_small_task",
 		.data		= &sysctl_sched_small_task_pct,
@@ -395,13 +388,6 @@ static struct ctl_table kern_table[] = {
 		.mode		= 0644,
 		.proc_handler	= sched_hmp_proc_update_handler,
 	},
-	{
-		.procname	= "sched_enable_power_aware",
-		.data		= &sysctl_sched_enable_power_aware,
-		.maxlen		= sizeof(unsigned int),
-		.mode		= 0644,
-		.proc_handler	= proc_dointvec,
-	},
 	{
 		.procname	= "sched_power_band_limit",
 		.data		= &sysctl_sched_powerband_limit_pct,