sched: Add tunables for static cpu and cluster cost

Add per-cpu tunable to set the extra cost to use a CPU that is idle.
Add the same for a cluster.

Change-Id: I4aa53f3c42c963df7abc7480980f747f0413d389
Signed-off-by: Olav Haugan <ohaugan@codeaurora.org>
[joonwoop@codeaurora.org: omitted changes for qhmp*.[c,h]  stripped out
 CONFIG_SCHED_QHMP in drivers/base/cpu.c and include/linux/sched.h]
Signed-off-by: Joonwoo Park <joonwoop@codeaurora.org>

7 files changed, 149 insertions, 6 deletions

diff --git a/drivers/base/cpu.c b/drivers/base/cpu.c
index 91bbb1959d8d..dee022638fe6 100644
--- a/drivers/base/cpu.c
+++ b/drivers/base/cpu.c
@@ -180,10 +180,106 @@ static struct attribute_group crash_note_cpu_attr_group = {
 };
 #endif
 
+#ifdef CONFIG_SCHED_HMP
+
+static ssize_t show_sched_static_cpu_pwr_cost(struct device *dev,
+				struct device_attribute *attr, char *buf)
+{
+	struct cpu *cpu = container_of(dev, struct cpu, dev);
+	ssize_t rc;
+	int cpuid = cpu->dev.id;
+	unsigned int pwr_cost;
+
+	pwr_cost = sched_get_static_cpu_pwr_cost(cpuid);
+
+	rc = snprintf(buf, PAGE_SIZE-2, "%d\n", pwr_cost);
+
+	return rc;
+}
+
+static ssize_t __ref store_sched_static_cpu_pwr_cost(struct device *dev,
+				struct device_attribute *attr,
+				const char *buf, size_t count)
+{
+	struct cpu *cpu = container_of(dev, struct cpu, dev);
+	int err;
+	int cpuid = cpu->dev.id;
+	unsigned int pwr_cost;
+
+	err = kstrtouint(strstrip((char *)buf), 0, &pwr_cost);
+	if (err)
+		return err;
+
+	err = sched_set_static_cpu_pwr_cost(cpuid, pwr_cost);
+
+	if (err >= 0)
+		err = count;
+
+	return err;
+}
+
+static ssize_t show_sched_static_cluster_pwr_cost(struct device *dev,
+				struct device_attribute *attr, char *buf)
+{
+	struct cpu *cpu = container_of(dev, struct cpu, dev);
+	ssize_t rc;
+	int cpuid = cpu->dev.id;
+	unsigned int pwr_cost;
+
+	pwr_cost = sched_get_static_cluster_pwr_cost(cpuid);
+
+	rc = snprintf(buf, PAGE_SIZE-2, "%d\n", pwr_cost);
+
+	return rc;
+}
+
+static ssize_t __ref store_sched_static_cluster_pwr_cost(struct device *dev,
+				struct device_attribute *attr,
+				const char *buf, size_t count)
+{
+	struct cpu *cpu = container_of(dev, struct cpu, dev);
+	int err;
+	int cpuid = cpu->dev.id;
+	unsigned int pwr_cost;
+
+	err = kstrtouint(strstrip((char *)buf), 0, &pwr_cost);
+	if (err)
+		return err;
+
+	err = sched_set_static_cluster_pwr_cost(cpuid, pwr_cost);
+
+	if (err >= 0)
+		err = count;
+
+	return err;
+}
+
+static DEVICE_ATTR(sched_static_cpu_pwr_cost, 0644,
+					show_sched_static_cpu_pwr_cost,
+					store_sched_static_cpu_pwr_cost);
+static DEVICE_ATTR(sched_static_cluster_pwr_cost, 0644,
+					show_sched_static_cluster_pwr_cost,
+					store_sched_static_cluster_pwr_cost);
+
+static struct attribute *hmp_sched_cpu_attrs[] = {
+	&dev_attr_sched_static_cpu_pwr_cost.attr,
+	&dev_attr_sched_static_cluster_pwr_cost.attr,
+	NULL
+};
+
+static struct attribute_group sched_hmp_cpu_attr_group = {
+	.attrs = hmp_sched_cpu_attrs,
+};
+
+#endif /* CONFIG_SCHED_HMP */
+
 static const struct attribute_group *common_cpu_attr_groups[] = {
 #ifdef CONFIG_KEXEC
 	&crash_note_cpu_attr_group,
 #endif
+#ifdef CONFIG_SCHED_HMP
+	&sched_hmp_cpu_attr_group,
+#endif
 	NULL
 };
 
@@ -191,6 +287,9 @@ static const struct attribute_group *hotplugable_cpu_attr_groups[] = {
 #ifdef CONFIG_KEXEC
 	&crash_note_cpu_attr_group,
 #endif
+#ifdef CONFIG_SCHED_HMP
+	&sched_hmp_cpu_attr_group,
+#endif
 	NULL
 };
 
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 9e4171146f39..77d3b4c106cd 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -2352,6 +2352,10 @@ 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_static_cpu_pwr_cost(int cpu, unsigned int cost);
+extern unsigned int sched_get_static_cpu_pwr_cost(int cpu);
+extern int sched_set_static_cluster_pwr_cost(int cpu, unsigned int cost);
+extern unsigned int sched_get_static_cluster_pwr_cost(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 3b10396d61b6..69aede209948 100644
--- a/include/trace/events/sched.h
+++ b/include/trace/events/sched.h
@@ -171,6 +171,7 @@ TRACE_EVENT(sched_cpu_load,
 		__field(unsigned int, max_freq			)
 		__field(unsigned int, power_cost		)
 		__field(	 int, cstate			)
+		__field(	 int, dstate			)
 		__field(	 int, temp			)
 	),
 
@@ -187,14 +188,16 @@ TRACE_EVENT(sched_cpu_load,
 		__entry->max_freq		= rq->max_freq;
 		__entry->power_cost		= power_cost;
 		__entry->cstate			= rq->cstate;
+		__entry->dstate			= rq->dstate;
 		__entry->temp			= temp;
 	),
 
-	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",
+	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 dstate %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)
+	__entry->max_freq, __entry->power_cost, __entry->cstate,
+	__entry->dstate, __entry->temp)
 );
 
 TRACE_EVENT(sched_set_boost,
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 1150d8665ac8..7b874fb5ebcd 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -1168,6 +1168,32 @@ static inline void clear_hmp_request(int cpu)
 	}
 }
 
+int sched_set_static_cpu_pwr_cost(int cpu, unsigned int cost)
+{
+	struct rq *rq = cpu_rq(cpu);
+
+	rq->static_cpu_pwr_cost = cost;
+	return 0;
+}
+
+unsigned int sched_get_static_cpu_pwr_cost(int cpu)
+{
+	return cpu_rq(cpu)->static_cpu_pwr_cost;
+}
+
+int sched_set_static_cluster_pwr_cost(int cpu, unsigned int cost)
+{
+	struct rq *rq = cpu_rq(cpu);
+
+	rq->static_cluster_pwr_cost = cost;
+	return 0;
+}
+
+unsigned int sched_get_static_cluster_pwr_cost(int cpu)
+{
+	return cpu_rq(cpu)->static_cluster_pwr_cost;
+}
+
 #else
 
 static inline int got_boost_kick(void)
@@ -9398,6 +9424,9 @@ void __init sched_init(void)
 		rq->cur_irqload = 0;
 		rq->avg_irqload = 0;
 		rq->irqload_ts = 0;
+		rq->static_cpu_pwr_cost = 0;
+		rq->static_cluster_pwr_cost = 0;
+
 #ifdef CONFIG_SCHED_FREQ_INPUT
 		rq->curr_runnable_sum = rq->prev_runnable_sum = 0;
 		rq->old_busy_time = 0;
diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index 5261ee0c35b9..b2bb6caa6a5b 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -324,6 +324,8 @@ do {									\
 	P(efficiency);
 	P(cur_freq);
 	P(max_freq);
+	P(static_cpu_pwr_cost);
+	P(static_cluster_pwr_cost);
 #endif
 #ifdef CONFIG_SCHED_HMP
 	P(hmp_stats.nr_big_tasks);
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 27e1a3d7bb05..f009c718fd82 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -3091,14 +3091,13 @@ static unsigned int power_cost_at_freq(int cpu, unsigned int freq)
 	BUG();
 }
 
-/* Return the cost of running task p on CPU cpu. This function
- * currently assumes that task p is the only task which will run on
- * the CPU. */
+/* Return the cost of running the total task load total_load on CPU cpu. */
 unsigned int power_cost(u64 total_load, int cpu)
 {
 	unsigned int task_freq;
 	struct rq *rq = cpu_rq(cpu);
 	u64 demand;
+	int total_static_pwr_cost = 0;
 
 	if (!sysctl_sched_enable_power_aware)
 		return rq->max_possible_capacity;
@@ -3110,7 +3109,12 @@ unsigned int power_cost(u64 total_load, int cpu)
 	task_freq = demand * rq->max_possible_freq;
 	task_freq /= 100; /* khz needed */
 
-	return power_cost_at_freq(cpu, task_freq);
+	if (idle_cpu(cpu) && rq->cstate) {
+		total_static_pwr_cost += rq->static_cpu_pwr_cost;
+		if (rq->dstate)
+			total_static_pwr_cost += rq->static_cluster_pwr_cost;
+	}
+	return power_cost_at_freq(cpu, task_freq) + total_static_pwr_cost;
 }
 
 #define UP_MIGRATION		1
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 4380cfacf1da..b38041e3df9b 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -690,6 +690,8 @@ struct rq {
 	u64 cur_irqload;
 	u64 avg_irqload;
 	u64 irqload_ts;
+	unsigned int static_cpu_pwr_cost;
+	unsigned int static_cluster_pwr_cost;
 
 #ifdef CONFIG_SCHED_FREQ_INPUT
 	unsigned int old_busy_time;