Merge branches 'pm-em' and 'powercap'

Merge Energy Model and power capping updates for 5.16-rc1:

 - Add support for inefficient operating performance points to the
   Energy Model and modify cpufreq to use them properly (Vincent
   Donnefort).

 - Rearrange the DTPM framework code to simplify it and make it easier
   to follow (Daniel Lezcano).

 - Fix power intialization in DTPM (Daniel Lezcano).

 - Add CPU load consideration when estimating the instaneous power
   consumption in DTPM (Daniel Lezcano).

* pm-em:
  cpufreq: mediatek-hw: Fix cpufreq_table_find_index_dl() call
  PM: EM: Mark inefficiencies in CPUFreq
  cpufreq: Use CPUFREQ_RELATION_E in DVFS governors
  cpufreq: Introducing CPUFREQ_RELATION_E
  cpufreq: Add an interface to mark inefficient frequencies
  cpufreq: Make policy min/max hard requirements
  PM: EM: Allow skipping inefficient states
  PM: EM: Extend em_perf_domain with a flag field
  PM: EM: Mark inefficient states
  PM: EM: Fix inefficient states detection

* powercap:
  powercap/drivers/dtpm: Fix power limit initialization
  powercap/drivers/dtpm: Scale the power with the load
  powercap/drivers/dtpm: Use container_of instead of a private data field
  powercap/drivers/dtpm: Simplify the dtpm table
  powercap/drivers/dtpm: Encapsulate even more the code

drivers/cpufreq/acpi-cpufreq.c

@@ -470,7 +470,8 @@ static unsigned int acpi_cpufreq_fast_switch(struct cpufreq_policy *policy,
 	if (policy->cached_target_freq == target_freq)
 		index = policy->cached_resolved_idx;
 	else
-		index = cpufreq_table_find_index_dl(policy, target_freq);
+		index = cpufreq_table_find_index_dl(policy, target_freq,
+						    false);
 
 	entry = &policy->freq_table[index];
 	next_freq = entry->frequency;

drivers/cpufreq/amd_freq_sensitivity.c

@@ -91,7 +91,8 @@ static unsigned int amd_powersave_bias_target(struct cpufreq_policy *policy,
 			unsigned int index;
 
 			index = cpufreq_table_find_index_h(policy,
-							   policy->cur - 1);
+							   policy->cur - 1,
+							   relation & CPUFREQ_RELATION_E);
 			freq_next = policy->freq_table[index].frequency;
 		}
 

drivers/cpufreq/cpufreq.c

@@ -554,7 +554,7 @@ static unsigned int __resolve_freq(struct cpufreq_policy *policy,
 unsigned int cpufreq_driver_resolve_freq(struct cpufreq_policy *policy,
 					 unsigned int target_freq)
 {
-	return __resolve_freq(policy, target_freq, CPUFREQ_RELATION_L);
+	return __resolve_freq(policy, target_freq, CPUFREQ_RELATION_LE);
 }
 EXPORT_SYMBOL_GPL(cpufreq_driver_resolve_freq);
 
@@ -2260,8 +2260,16 @@ int __cpufreq_driver_target(struct cpufreq_policy *policy,
 	    !(cpufreq_driver->flags & CPUFREQ_NEED_UPDATE_LIMITS))
 		return 0;
 
-	if (cpufreq_driver->target)
+	if (cpufreq_driver->target) {
+		/*
+		 * If the driver hasn't setup a single inefficient frequency,
+		 * it's unlikely it knows how to decode CPUFREQ_RELATION_E.
+		 */
+		if (!policy->efficiencies_available)
+			relation &= ~CPUFREQ_RELATION_E;
+
 		return cpufreq_driver->target(policy, target_freq, relation);
+	}
 
 	if (!cpufreq_driver->target_index)
 		return -EINVAL;
@@ -2523,8 +2531,15 @@ static int cpufreq_set_policy(struct cpufreq_policy *policy,
 	if (ret)
 		return ret;
 
+	/*
+	 * Resolve policy min/max to available frequencies. It ensures
+	 * no frequency resolution will neither overshoot the requested maximum
+	 * nor undershoot the requested minimum.
+	 */
 	policy->min = new_data.min;
 	policy->max = new_data.max;
+	policy->min = __resolve_freq(policy, policy->min, CPUFREQ_RELATION_L);
+	policy->max = __resolve_freq(policy, policy->max, CPUFREQ_RELATION_H);
 	trace_cpu_frequency_limits(policy);
 
 	policy->cached_target_freq = UINT_MAX;

drivers/cpufreq/cpufreq_conservative.c

@@ -111,7 +111,8 @@ static unsigned int cs_dbs_update(struct cpufreq_policy *policy)
 		if (requested_freq > policy->max)
 			requested_freq = policy->max;
 
-		__cpufreq_driver_target(policy, requested_freq, CPUFREQ_RELATION_H);
+		__cpufreq_driver_target(policy, requested_freq,
+					CPUFREQ_RELATION_HE);
 		dbs_info->requested_freq = requested_freq;
 		goto out;
 	}
@@ -134,7 +135,8 @@ static unsigned int cs_dbs_update(struct cpufreq_policy *policy)
 		else
 			requested_freq = policy->min;
 
-		__cpufreq_driver_target(policy, requested_freq, CPUFREQ_RELATION_L);
+		__cpufreq_driver_target(policy, requested_freq,
+					CPUFREQ_RELATION_LE);
 		dbs_info->requested_freq = requested_freq;
 	}
 

drivers/cpufreq/cpufreq_ondemand.c

@@ -83,9 +83,11 @@ static unsigned int generic_powersave_bias_target(struct cpufreq_policy *policy,
 	freq_avg = freq_req - freq_reduc;
 
 	/* Find freq bounds for freq_avg in freq_table */
-	index = cpufreq_table_find_index_h(policy, freq_avg);
+	index = cpufreq_table_find_index_h(policy, freq_avg,
+					   relation & CPUFREQ_RELATION_E);
 	freq_lo = freq_table[index].frequency;
-	index = cpufreq_table_find_index_l(policy, freq_avg);
+	index = cpufreq_table_find_index_l(policy, freq_avg,
+					   relation & CPUFREQ_RELATION_E);
 	freq_hi = freq_table[index].frequency;
 
 	/* Find out how long we have to be in hi and lo freqs */
@@ -118,12 +120,12 @@ static void dbs_freq_increase(struct cpufreq_policy *policy, unsigned int freq)
 
 	if (od_tuners->powersave_bias)
 		freq = od_ops.powersave_bias_target(policy, freq,
-				CPUFREQ_RELATION_H);
+						    CPUFREQ_RELATION_HE);
 	else if (policy->cur == policy->max)
 		return;
 
 	__cpufreq_driver_target(policy, freq, od_tuners->powersave_bias ?
-			CPUFREQ_RELATION_L : CPUFREQ_RELATION_H);
+			CPUFREQ_RELATION_LE : CPUFREQ_RELATION_HE);
 }
 
 /*
@@ -161,9 +163,9 @@ static void od_update(struct cpufreq_policy *policy)
 		if (od_tuners->powersave_bias)
 			freq_next = od_ops.powersave_bias_target(policy,
 								 freq_next,
-								 CPUFREQ_RELATION_L);
+								 CPUFREQ_RELATION_LE);
 
-		__cpufreq_driver_target(policy, freq_next, CPUFREQ_RELATION_C);
+		__cpufreq_driver_target(policy, freq_next, CPUFREQ_RELATION_CE);
 	}
 }
 
@@ -182,7 +184,7 @@ static unsigned int od_dbs_update(struct cpufreq_policy *policy)
 	 */
 	if (sample_type == OD_SUB_SAMPLE && policy_dbs->sample_delay_ns > 0) {
 		__cpufreq_driver_target(policy, dbs_info->freq_lo,
-					CPUFREQ_RELATION_H);
+					CPUFREQ_RELATION_HE);
 		return dbs_info->freq_lo_delay_us;
 	}
 

drivers/cpufreq/mediatek-cpufreq-hw.c

@@ -109,7 +109,7 @@ static unsigned int mtk_cpufreq_hw_fast_switch(struct cpufreq_policy *policy,
 	struct mtk_cpufreq_data *data = policy->driver_data;
 	unsigned int index;
 
-	index = cpufreq_table_find_index_dl(policy, target_freq);
+	index = cpufreq_table_find_index_dl(policy, target_freq, false);
 
 	writel_relaxed(index, data->reg_bases[REG_FREQ_PERF_STATE]);
 

drivers/cpufreq/powernv-cpufreq.c

@@ -934,7 +934,7 @@ static void powernv_cpufreq_work_fn(struct work_struct *work)
 		policy = cpufreq_cpu_get(cpu);
 		if (!policy)
 			continue;
-		index = cpufreq_table_find_index_c(policy, policy->cur);
+		index = cpufreq_table_find_index_c(policy, policy->cur, false);
 		powernv_cpufreq_target_index(policy, index);
 		cpumask_andnot(&mask, &mask, policy->cpus);
 		cpufreq_cpu_put(policy);
@@ -1022,7 +1022,7 @@ static unsigned int powernv_fast_switch(struct cpufreq_policy *policy,
 	int index;
 	struct powernv_smp_call_data freq_data;
 
-	index = cpufreq_table_find_index_dl(policy, target_freq);
+	index = cpufreq_table_find_index_dl(policy, target_freq, false);
 	freq_data.pstate_id = powernv_freqs[index].driver_data;
 	freq_data.gpstate_id = powernv_freqs[index].driver_data;
 	set_pstate(&freq_data);

drivers/cpufreq/s5pv210-cpufreq.c

@@ -243,7 +243,7 @@ static int s5pv210_target(struct cpufreq_policy *policy, unsigned int index)
 	new_freq = s5pv210_freq_table[index].frequency;
 
 	/* Finding current running level index */
-	priv_index = cpufreq_table_find_index_h(policy, old_freq);
+	priv_index = cpufreq_table_find_index_h(policy, old_freq, false);
 
 	arm_volt = dvs_conf[index].arm_volt;
 	int_volt = dvs_conf[index].int_volt;

drivers/powercap/dtpm.c

@@ -116,8 +116,6 @@ static void __dtpm_sub_power(struct dtpm *dtpm)
 		parent->power_limit -= dtpm->power_limit;
 		parent = parent->parent;
 	}
-
-	__dtpm_rebalance_weight(root);
 }
 
 static void __dtpm_add_power(struct dtpm *dtpm)
@@ -130,45 +128,45 @@ static void __dtpm_add_power(struct dtpm *dtpm)
 		parent->power_limit += dtpm->power_limit;
 		parent = parent->parent;
 	}
+}
+
+static int __dtpm_update_power(struct dtpm *dtpm)
+{
+	int ret;
+
+	__dtpm_sub_power(dtpm);
+
+	ret = dtpm->ops->update_power_uw(dtpm);
+	if (ret)
+		pr_err("Failed to update power for '%s': %d\n",
+		       dtpm->zone.name, ret);
 
-	__dtpm_rebalance_weight(root);
+	if (!test_bit(DTPM_POWER_LIMIT_FLAG, &dtpm->flags))
+		dtpm->power_limit = dtpm->power_max;
+
+	__dtpm_add_power(dtpm);
+
+	if (root)
+		__dtpm_rebalance_weight(root);
+
+	return ret;
 }
 
 /**
  * dtpm_update_power - Update the power on the dtpm
  * @dtpm: a pointer to a dtpm structure to update
- * @power_min: a u64 representing the new power_min value
- * @power_max: a u64 representing the new power_max value
  *
  * Function to update the power values of the dtpm node specified in
  * parameter. These new values will be propagated to the tree.
  *
  * Return: zero on success, -EINVAL if the values are inconsistent
  */
-int dtpm_update_power(struct dtpm *dtpm, u64 power_min, u64 power_max)
+int dtpm_update_power(struct dtpm *dtpm)
 {
-	int ret = 0;
+	int ret;
 
 	mutex_lock(&dtpm_lock);
-
-	if (power_min == dtpm->power_min && power_max == dtpm->power_max)
-		goto unlock;
-
-	if (power_max < power_min) {
-		ret = -EINVAL;
-		goto unlock;
-	}
-
-	__dtpm_sub_power(dtpm);
-
-	dtpm->power_min = power_min;
-	dtpm->power_max = power_max;
-	if (!test_bit(DTPM_POWER_LIMIT_FLAG, &dtpm->flags))
-		dtpm->power_limit = power_max;
-
-	__dtpm_add_power(dtpm);
-
-unlock:
+	ret = __dtpm_update_power(dtpm);
 	mutex_unlock(&dtpm_lock);
 
 	return ret;
@@ -359,24 +357,18 @@ static struct powercap_zone_ops zone_ops = {
 };
 
 /**
- * dtpm_alloc - Allocate and initialize a dtpm struct
- * @name: a string specifying the name of the node
- *
- * Return: a struct dtpm pointer, NULL in case of error
+ * dtpm_init - Allocate and initialize a dtpm struct
+ * @dtpm: The dtpm struct pointer to be initialized
+ * @ops: The dtpm device specific ops, NULL for a virtual node
  */
-struct dtpm *dtpm_alloc(struct dtpm_ops *ops)
+void dtpm_init(struct dtpm *dtpm, struct dtpm_ops *ops)
 {
-	struct dtpm *dtpm;
-
-	dtpm = kzalloc(sizeof(*dtpm), GFP_KERNEL);
 	if (dtpm) {
 		INIT_LIST_HEAD(&dtpm->children);
 		INIT_LIST_HEAD(&dtpm->sibling);
 		dtpm->weight = 1024;
 		dtpm->ops = ops;
 	}
-
-	return dtpm;
 }
 
 /**
@@ -436,6 +428,7 @@ int dtpm_register(const char *name, struct dtpm *dtpm, struct dtpm *parent)
 
 	if (dtpm->ops && !(dtpm->ops->set_power_uw &&
 			   dtpm->ops->get_power_uw &&
+			   dtpm->ops->update_power_uw &&
 			   dtpm->ops->release))
 		return -EINVAL;
 
@@ -455,7 +448,10 @@ int dtpm_register(const char *name, struct dtpm *dtpm, struct dtpm *parent)
 		root = dtpm;
 	}
 
-	__dtpm_add_power(dtpm);
+	if (dtpm->ops && !dtpm->ops->update_power_uw(dtpm)) {
+		__dtpm_add_power(dtpm);
+		dtpm->power_limit = dtpm->power_max;
+	}
 
 	pr_info("Registered dtpm node '%s' / %llu-%llu uW, \n",
 		dtpm->zone.name, dtpm->power_min, dtpm->power_max);
@@ -465,9 +461,9 @@ int dtpm_register(const char *name, struct dtpm *dtpm, struct dtpm *parent)
 	return 0;
 }
 
-static int __init dtpm_init(void)
+static int __init init_dtpm(void)
 {
-	struct dtpm_descr **dtpm_descr;
+	struct dtpm_descr *dtpm_descr;
 
 	pct = powercap_register_control_type(NULL, "dtpm", NULL);
 	if (IS_ERR(pct)) {
@@ -476,8 +472,8 @@ static int __init dtpm_init(void)
 	}
 
 	for_each_dtpm_table(dtpm_descr)
-		(*dtpm_descr)->init(*dtpm_descr);
+		dtpm_descr->init();
 
 	return 0;
 }
-late_initcall(dtpm_init);
+late_initcall(init_dtpm);

drivers/powercap/dtpm_cpu.c

@@ -14,6 +14,8 @@
  * The CPU hotplug is supported and the power numbers will be updated
  * if a CPU is hot plugged / unplugged.
  */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
 #include <linux/cpumask.h>
 #include <linux/cpufreq.h>
 #include <linux/cpuhotplug.h>
@@ -23,66 +25,29 @@
 #include <linux/slab.h>
 #include <linux/units.h>
 
-static struct dtpm *__parent;
-
-static DEFINE_PER_CPU(struct dtpm *, dtpm_per_cpu);
-
 struct dtpm_cpu {
+	struct dtpm dtpm;
 	struct freq_qos_request qos_req;
 	int cpu;
 };
 
-/*
- * When a new CPU is inserted at hotplug or boot time, add the power
- * contribution and update the dtpm tree.
- */
-static int power_add(struct dtpm *dtpm, struct em_perf_domain *em)
-{
-	u64 power_min, power_max;
-
-	power_min = em->table[0].power;
-	power_min *= MICROWATT_PER_MILLIWATT;
-	power_min += dtpm->power_min;
+static DEFINE_PER_CPU(struct dtpm_cpu *, dtpm_per_cpu);
 
-	power_max = em->table[em->nr_perf_states - 1].power;
-	power_max *= MICROWATT_PER_MILLIWATT;
-	power_max += dtpm->power_max;
-
-	return dtpm_update_power(dtpm, power_min, power_max);
-}
-
-/*
- * When a CPU is unplugged, remove its power contribution from the
- * dtpm tree.
- */
-static int power_sub(struct dtpm *dtpm, struct em_perf_domain *em)
+static struct dtpm_cpu *to_dtpm_cpu(struct dtpm *dtpm)
 {
-	u64 power_min, power_max;
-
-	power_min = em->table[0].power;
-	power_min *= MICROWATT_PER_MILLIWATT;
-	power_min = dtpm->power_min - power_min;
-
-	power_max = em->table[em->nr_perf_states - 1].power;
-	power_max *= MICROWATT_PER_MILLIWATT;
-	power_max = dtpm->power_max - power_max;
-
-	return dtpm_update_power(dtpm, power_min, power_max);
+	return container_of(dtpm, struct dtpm_cpu, dtpm);
 }
 
 static u64 set_pd_power_limit(struct dtpm *dtpm, u64 power_limit)
 {
-	struct dtpm_cpu *dtpm_cpu = dtpm->private;
-	struct em_perf_domain *pd;
+	struct dtpm_cpu *dtpm_cpu = to_dtpm_cpu(dtpm);
+	struct em_perf_domain *pd = em_cpu_get(dtpm_cpu->cpu);
 	struct cpumask cpus;
 	unsigned long freq;
 	u64 power;
 	int i, nr_cpus;
 
-	pd = em_cpu_get(dtpm_cpu->cpu);
-
 	cpumask_and(&cpus, cpu_online_mask, to_cpumask(pd->cpus));
-
 	nr_cpus = cpumask_weight(&cpus);
 
 	for (i = 0; i < pd->nr_perf_states; i++) {
@@ -103,34 +68,88 @@ static u64 set_pd_power_limit(struct dtpm *dtpm, u64 power_limit)
 	return power_limit;
 }
 
+static u64 scale_pd_power_uw(struct cpumask *pd_mask, u64 power)
+{
+	unsigned long max = 0, sum_util = 0;
+	int cpu;
+
+	for_each_cpu_and(cpu, pd_mask, cpu_online_mask) {
+
+		/*
+		 * The capacity is the same for all CPUs belonging to
+		 * the same perf domain, so a single call to
+		 * arch_scale_cpu_capacity() is enough. However, we
+		 * need the CPU parameter to be initialized by the
+		 * loop, so the call ends up in this block.
+		 *
+		 * We can initialize 'max' with a cpumask_first() call
+		 * before the loop but the bits computation is not
+		 * worth given the arch_scale_cpu_capacity() just
+		 * returns a value where the resulting assembly code
+		 * will be optimized by the compiler.
+		 */
+		max = arch_scale_cpu_capacity(cpu);
+		sum_util += sched_cpu_util(cpu, max);
+	}
+
+	/*
+	 * In the improbable case where all the CPUs of the perf
+	 * domain are offline, 'max' will be zero and will lead to an
+	 * illegal operation with a zero division.
+	 */
+	return max ? (power * ((sum_util << 10) / max)) >> 10 : 0;
+}
+
 static u64 get_pd_power_uw(struct dtpm *dtpm)
 {
-	struct dtpm_cpu *dtpm_cpu = dtpm->private;
+	struct dtpm_cpu *dtpm_cpu = to_dtpm_cpu(dtpm);
 	struct em_perf_domain *pd;
-	struct cpumask cpus;
+	struct cpumask *pd_mask;
 	unsigned long freq;
-	int i, nr_cpus;
+	int i;
 
 	pd = em_cpu_get(dtpm_cpu->cpu);
+
+	pd_mask = em_span_cpus(pd);
+
 	freq = cpufreq_quick_get(dtpm_cpu->cpu);
-	cpumask_and(&cpus, cpu_online_mask, to_cpumask(pd->cpus));
-	nr_cpus = cpumask_weight(&cpus);
 
 	for (i = 0; i < pd->nr_perf_states; i++) {
 
 		if (pd->table[i].frequency < freq)
 			continue;
 
-		return pd->table[i].power *
-			MICROWATT_PER_MILLIWATT * nr_cpus;
+		return scale_pd_power_uw(pd_mask, pd->table[i].power *
+					 MICROWATT_PER_MILLIWATT);
 	}
 
 	return 0;
 }
 
+static int update_pd_power_uw(struct dtpm *dtpm)
+{
+	struct dtpm_cpu *dtpm_cpu = to_dtpm_cpu(dtpm);
+	struct em_perf_domain *em = em_cpu_get(dtpm_cpu->cpu);
+	struct cpumask cpus;
+	int nr_cpus;
+
+	cpumask_and(&cpus, cpu_online_mask, to_cpumask(em->cpus));
+	nr_cpus = cpumask_weight(&cpus);
+
+	dtpm->power_min = em->table[0].power;
+	dtpm->power_min *= MICROWATT_PER_MILLIWATT;
+	dtpm->power_min *= nr_cpus;
+
+	dtpm->power_max = em->table[em->nr_perf_states - 1].power;
+	dtpm->power_max *= MICROWATT_PER_MILLIWATT;
+	dtpm->power_max *= nr_cpus;
+
+	return 0;
+}
+
 static void pd_release(struct dtpm *dtpm)
 {
-	struct dtpm_cpu *dtpm_cpu = dtpm->private;
+	struct dtpm_cpu *dtpm_cpu = to_dtpm_cpu(dtpm);
 
 	if (freq_qos_request_active(&dtpm_cpu->qos_req))
 		freq_qos_remove_request(&dtpm_cpu->qos_req);
@@ -139,44 +158,28 @@ static void pd_release(struct dtpm *dtpm)
 }
 
 static struct dtpm_ops dtpm_ops = {
-	.set_power_uw = set_pd_power_limit,
-	.get_power_uw = get_pd_power_uw,
-	.release = pd_release,
+	.set_power_uw	 = set_pd_power_limit,
+	.get_power_uw	 = get_pd_power_uw,
+	.update_power_uw = update_pd_power_uw,
+	.release	 = pd_release,
 };
 
 static int cpuhp_dtpm_cpu_offline(unsigned int cpu)
 {
-	struct cpufreq_policy *policy;
 	struct em_perf_domain *pd;
-	struct dtpm *dtpm;
-
-	policy = cpufreq_cpu_get(cpu);
-
-	if (!policy)
-		return 0;
+	struct dtpm_cpu *dtpm_cpu;
 
 	pd = em_cpu_get(cpu);
 	if (!pd)
 		return -EINVAL;
 
-	dtpm = per_cpu(dtpm_per_cpu, cpu);
+	dtpm_cpu = per_cpu(dtpm_per_cpu, cpu);
 
-	power_sub(dtpm, pd);
-
-	if (cpumask_weight(policy->cpus) != 1)
-		return 0;
-
-	for_each_cpu(cpu, policy->related_cpus)
-		per_cpu(dtpm_per_cpu, cpu) = NULL;
-
-	dtpm_unregister(dtpm);
-
-	return 0;
+	return dtpm_update_power(&dtpm_cpu->dtpm);
 }
 
 static int cpuhp_dtpm_cpu_online(unsigned int cpu)
 {
-	struct dtpm *dtpm;
 	struct dtpm_cpu *dtpm_cpu;
 	struct cpufreq_policy *policy;
 	struct em_perf_domain *pd;
@@ -184,7 +187,6 @@ static int cpuhp_dtpm_cpu_online(unsigned int cpu)
 	int ret = -ENOMEM;
 
 	policy = cpufreq_cpu_get(cpu);
-
 	if (!policy)
 		return 0;
 
@@ -192,66 +194,82 @@ static int cpuhp_dtpm_cpu_online(unsigned int cpu)
 	if (!pd)
 		return -EINVAL;
 
-	dtpm = per_cpu(dtpm_per_cpu, cpu);
-	if (dtpm)
-		return power_add(dtpm, pd);
-
-	dtpm = dtpm_alloc(&dtpm_ops);
-	if (!dtpm)
-		return -EINVAL;
+	dtpm_cpu = per_cpu(dtpm_per_cpu, cpu);
+	if (dtpm_cpu)
+		return dtpm_update_power(&dtpm_cpu->dtpm);
 
 	dtpm_cpu = kzalloc(sizeof(*dtpm_cpu), GFP_KERNEL);
 	if (!dtpm_cpu)
-		goto out_kfree_dtpm;
+		return -ENOMEM;
 
-	dtpm->private = dtpm_cpu;
+	dtpm_init(&dtpm_cpu->dtpm, &dtpm_ops);
 	dtpm_cpu->cpu = cpu;
 
 	for_each_cpu(cpu, policy->related_cpus)
-		per_cpu(dtpm_per_cpu, cpu) = dtpm;
+		per_cpu(dtpm_per_cpu, cpu) = dtpm_cpu;
 
-	sprintf(name, "cpu%d", dtpm_cpu->cpu);
+	snprintf(name, sizeof(name), "cpu%d-cpufreq", dtpm_cpu->cpu);
 
-	ret = dtpm_register(name, dtpm, __parent);
+	ret = dtpm_register(name, &dtpm_cpu->dtpm, NULL);
 	if (ret)
 		goto out_kfree_dtpm_cpu;
 
-	ret = power_add(dtpm, pd);
-	if (ret)
-		goto out_dtpm_unregister;
-
 	ret = freq_qos_add_request(&policy->constraints,
 				   &dtpm_cpu->qos_req, FREQ_QOS_MAX,
 				   pd->table[pd->nr_perf_states - 1].frequency);
 	if (ret)
-		goto out_power_sub;
+		goto out_dtpm_unregister;
 
 	return 0;
 
-out_power_sub:
-	power_sub(dtpm, pd);
-
 out_dtpm_unregister:
-	dtpm_unregister(dtpm);
+	dtpm_unregister(&dtpm_cpu->dtpm);
 	dtpm_cpu = NULL;
-	dtpm = NULL;
 
 out_kfree_dtpm_cpu:
 	for_each_cpu(cpu, policy->related_cpus)
 		per_cpu(dtpm_per_cpu, cpu) = NULL;
 	kfree(dtpm_cpu);
 
-out_kfree_dtpm:
-	kfree(dtpm);
 	return ret;
 }
 
-int dtpm_register_cpu(struct dtpm *parent)
+static int __init dtpm_cpu_init(void)
 {
-	__parent = parent;
+	int ret;
+
+	/*
+	 * The callbacks at CPU hotplug time are calling
+	 * dtpm_update_power() which in turns calls update_pd_power().
+	 *
+	 * The function update_pd_power() uses the online mask to
+	 * figure out the power consumption limits.
+	 *
+	 * At CPUHP_AP_ONLINE_DYN, the CPU is present in the CPU
+	 * online mask when the cpuhp_dtpm_cpu_online function is
+	 * called, but the CPU is still in the online mask for the
+	 * tear down callback. So the power can not be updated when
+	 * the CPU is unplugged.
+	 *
+	 * At CPUHP_AP_DTPM_CPU_DEAD, the situation is the opposite as
+	 * above. The CPU online mask is not up to date when the CPU
+	 * is plugged in.
+	 *
+	 * For this reason, we need to call the online and offline
+	 * callbacks at different moments when the CPU online mask is
+	 * consistent with the power numbers we want to update.
+	 */
+	ret = cpuhp_setup_state(CPUHP_AP_DTPM_CPU_DEAD, "dtpm_cpu:offline",
+				NULL, cpuhp_dtpm_cpu_offline);
+	if (ret < 0)
+		return ret;
+
+	ret = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "dtpm_cpu:online",
+				cpuhp_dtpm_cpu_online, NULL);
+	if (ret < 0)
+		return ret;
 
-	return cpuhp_setup_state(CPUHP_AP_DTPM_CPU_ONLINE,
-				 "dtpm_cpu:online",
-				 cpuhp_dtpm_cpu_online,
-				 cpuhp_dtpm_cpu_offline);
+	return 0;
 }
+
+DTPM_DECLARE(dtpm_cpu, dtpm_cpu_init);

include/linux/cpuhotplug.h

@@ -99,6 +99,7 @@ enum cpuhp_state {
 	CPUHP_LUSTRE_CFS_DEAD,
 	CPUHP_AP_ARM_CACHE_B15_RAC_DEAD,
 	CPUHP_PADATA_DEAD,
+	CPUHP_AP_DTPM_CPU_DEAD,
 	CPUHP_WORKQUEUE_PREP,
 	CPUHP_POWER_NUMA_PREPARE,
 	CPUHP_HRTIMERS_PREPARE,
@@ -246,7 +247,6 @@ enum cpuhp_state {
 	CPUHP_AP_MM_DEMOTION_ONLINE,
 	CPUHP_AP_X86_HPET_ONLINE,
 	CPUHP_AP_X86_KVM_CLK_ONLINE,
-	CPUHP_AP_DTPM_CPU_ONLINE,
 	CPUHP_AP_ACTIVE,
 	CPUHP_ONLINE,
 };

include/linux/dtpm.h

@@ -23,34 +23,32 @@ struct dtpm {
 	u64 power_max;
 	u64 power_min;
 	int weight;
-	void *private;
 };
 
 struct dtpm_ops {
 	u64 (*set_power_uw)(struct dtpm *, u64);
 	u64 (*get_power_uw)(struct dtpm *);
+	int (*update_power_uw)(struct dtpm *);
 	void (*release)(struct dtpm *);
 };
 
-struct dtpm_descr;
-
-typedef int (*dtpm_init_t)(struct dtpm_descr *);
+typedef int (*dtpm_init_t)(void);
 
 struct dtpm_descr {
-	struct dtpm *parent;
-	const char *name;
 	dtpm_init_t init;
 };
 
 /* Init section thermal table */
-extern struct dtpm_descr *__dtpm_table[];
-extern struct dtpm_descr *__dtpm_table_end[];
+extern struct dtpm_descr __dtpm_table[];
+extern struct dtpm_descr __dtpm_table_end[];
 
-#define DTPM_TABLE_ENTRY(name)			\
-	static typeof(name) *__dtpm_table_entry_##name	\
-	__used __section("__dtpm_table") = &name
+#define DTPM_TABLE_ENTRY(name, __init)				\
+	static struct dtpm_descr __dtpm_table_entry_##name	\
+	__used __section("__dtpm_table") = {			\
+		.init = __init,					\
+	}
 
-#define DTPM_DECLARE(name)	DTPM_TABLE_ENTRY(name)
+#define DTPM_DECLARE(name, init)	DTPM_TABLE_ENTRY(name, init)
 
 #define for_each_dtpm_table(__dtpm)	\
 	for (__dtpm = __dtpm_table;	\
@@ -62,11 +60,11 @@ static inline struct dtpm *to_dtpm(struct powercap_zone *zone)
 	return container_of(zone, struct dtpm, zone);
 }
 
-int dtpm_update_power(struct dtpm *dtpm, u64 power_min, u64 power_max);
+int dtpm_update_power(struct dtpm *dtpm);
 
 int dtpm_release_zone(struct powercap_zone *pcz);
 
-struct dtpm *dtpm_alloc(struct dtpm_ops *ops);
+void dtpm_init(struct dtpm *dtpm, struct dtpm_ops *ops);
 
 void dtpm_unregister(struct dtpm *dtpm);
 

include/linux/energy_model.h

@@ -17,19 +17,30 @@
  *		device). It can be a total power: static and dynamic.
  * @cost:	The cost coefficient associated with this level, used during
  *		energy calculation. Equal to: power * max_frequency / frequency
+ * @flags:	see "em_perf_state flags" description below.
  */
 struct em_perf_state {
 	unsigned long frequency;
 	unsigned long power;
 	unsigned long cost;
+	unsigned long flags;
 };
 
+/*
+ * em_perf_state flags:
+ *
+ * EM_PERF_STATE_INEFFICIENT: The performance state is inefficient. There is
+ * in this em_perf_domain, another performance state with a higher frequency
+ * but a lower or equal power cost. Such inefficient states are ignored when
+ * using em_pd_get_efficient_*() functions.
+ */
+#define EM_PERF_STATE_INEFFICIENT BIT(0)
+
 /**
  * struct em_perf_domain - Performance domain
  * @table:		List of performance states, in ascending order
  * @nr_perf_states:	Number of performance states
- * @milliwatts:		Flag indicating the power values are in milli-Watts
- *			or some other scale.
+ * @flags:		See "em_perf_domain flags"
  * @cpus:		Cpumask covering the CPUs of the domain. It's here
  *			for performance reasons to avoid potential cache
  *			misses during energy calculations in the scheduler
@@ -44,10 +55,22 @@ struct em_perf_state {
 struct em_perf_domain {
 	struct em_perf_state *table;
 	int nr_perf_states;
-	int milliwatts;
+	unsigned long flags;
 	unsigned long cpus[];
 };
 
+/*
+ *  em_perf_domain flags:
+ *
+ *  EM_PERF_DOMAIN_MILLIWATTS: The power values are in milli-Watts or some
+ *  other scale.
+ *
+ *  EM_PERF_DOMAIN_SKIP_INEFFICIENCIES: Skip inefficient states when estimating
+ *  energy consumption.
+ */
+#define EM_PERF_DOMAIN_MILLIWATTS BIT(0)
+#define EM_PERF_DOMAIN_SKIP_INEFFICIENCIES BIT(1)
+
 #define em_span_cpus(em) (to_cpumask((em)->cpus))
 
 #ifdef CONFIG_ENERGY_MODEL
@@ -101,6 +124,37 @@ int em_dev_register_perf_domain(struct device *dev, unsigned int nr_states,
 				bool milliwatts);
 void em_dev_unregister_perf_domain(struct device *dev);
 
+/**
+ * em_pd_get_efficient_state() - Get an efficient performance state from the EM
+ * @pd   : Performance domain for which we want an efficient frequency
+ * @freq : Frequency to map with the EM
+ *
+ * It is called from the scheduler code quite frequently and as a consequence
+ * doesn't implement any check.
+ *
+ * Return: An efficient performance state, high enough to meet @freq
+ * requirement.
+ */
+static inline
+struct em_perf_state *em_pd_get_efficient_state(struct em_perf_domain *pd,
+						unsigned long freq)
+{
+	struct em_perf_state *ps;
+	int i;
+
+	for (i = 0; i < pd->nr_perf_states; i++) {
+		ps = &pd->table[i];
+		if (ps->frequency >= freq) {
+			if (pd->flags & EM_PERF_DOMAIN_SKIP_INEFFICIENCIES &&
+			    ps->flags & EM_PERF_STATE_INEFFICIENT)
+				continue;
+			break;
+		}
+	}
+
+	return ps;
+}
+
 /**
  * em_cpu_energy() - Estimates the energy consumed by the CPUs of a
  *		performance domain
@@ -123,7 +177,7 @@ static inline unsigned long em_cpu_energy(struct em_perf_domain *pd,
 {
 	unsigned long freq, scale_cpu;
 	struct em_perf_state *ps;
-	int i, cpu;
+	int cpu;
 
 	if (!sum_util)
 		return 0;
@@ -148,11 +202,7 @@ static inline unsigned long em_cpu_energy(struct em_perf_domain *pd,
 	 * Find the lowest performance state of the Energy Model above the
 	 * requested frequency.
 	 */
-	for (i = 0; i < pd->nr_perf_states; i++) {
-		ps = &pd->table[i];
-		if (ps->frequency >= freq)
-			break;
-	}
+	ps = em_pd_get_efficient_state(pd, freq);
 
 	/*
 	 * The capacity of a CPU in the domain at the performance state (ps)

kernel/power/energy_model.c

@@ -2,7 +2,7 @@
 /*
  * Energy Model of devices
  *
- * Copyright (c) 2018-2020, Arm ltd.
+ * Copyright (c) 2018-2021, Arm ltd.
  * Written by: Quentin Perret, Arm ltd.
  * Improvements provided by: Lukasz Luba, Arm ltd.
  */
@@ -10,6 +10,7 @@
 #define pr_fmt(fmt) "energy_model: " fmt
 
 #include <linux/cpu.h>
+#include <linux/cpufreq.h>
 #include <linux/cpumask.h>
 #include <linux/debugfs.h>
 #include <linux/energy_model.h>
@@ -42,6 +43,7 @@ static void em_debug_create_ps(struct em_perf_state *ps, struct dentry *pd)
 	debugfs_create_ulong("frequency", 0444, d, &ps->frequency);
 	debugfs_create_ulong("power", 0444, d, &ps->power);
 	debugfs_create_ulong("cost", 0444, d, &ps->cost);
+	debugfs_create_ulong("inefficient", 0444, d, &ps->flags);
 }
 
 static int em_debug_cpus_show(struct seq_file *s, void *unused)
@@ -55,7 +57,8 @@ DEFINE_SHOW_ATTRIBUTE(em_debug_cpus);
 static int em_debug_units_show(struct seq_file *s, void *unused)
 {
 	struct em_perf_domain *pd = s->private;
-	char *units = pd->milliwatts ? "milliWatts" : "bogoWatts";
+	char *units = (pd->flags & EM_PERF_DOMAIN_MILLIWATTS) ?
+		"milliWatts" : "bogoWatts";
 
 	seq_printf(s, "%s\n", units);
 
@@ -63,6 +66,17 @@ static int em_debug_units_show(struct seq_file *s, void *unused)
 }
 DEFINE_SHOW_ATTRIBUTE(em_debug_units);
 
+static int em_debug_skip_inefficiencies_show(struct seq_file *s, void *unused)
+{
+	struct em_perf_domain *pd = s->private;
+	int enabled = (pd->flags & EM_PERF_DOMAIN_SKIP_INEFFICIENCIES) ? 1 : 0;
+
+	seq_printf(s, "%d\n", enabled);
+
+	return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(em_debug_skip_inefficiencies);
+
 static void em_debug_create_pd(struct device *dev)
 {
 	struct dentry *d;
@@ -76,6 +90,8 @@ static void em_debug_create_pd(struct device *dev)
 				    &em_debug_cpus_fops);
 
 	debugfs_create_file("units", 0444, d, dev->em_pd, &em_debug_units_fops);
+	debugfs_create_file("skip-inefficiencies", 0444, d, dev->em_pd,
+			    &em_debug_skip_inefficiencies_fops);
 
 	/* Create a sub-directory for each performance state */
 	for (i = 0; i < dev->em_pd->nr_perf_states; i++)
@@ -107,8 +123,7 @@ static void em_debug_remove_pd(struct device *dev) {}
 static int em_create_perf_table(struct device *dev, struct em_perf_domain *pd,
 				int nr_states, struct em_data_callback *cb)
 {
-	unsigned long opp_eff, prev_opp_eff = ULONG_MAX;
-	unsigned long power, freq, prev_freq = 0;
+	unsigned long power, freq, prev_freq = 0, prev_cost = ULONG_MAX;
 	struct em_perf_state *table;
 	int i, ret;
 	u64 fmax;
@@ -153,27 +168,22 @@ static int em_create_perf_table(struct device *dev, struct em_perf_domain *pd,
 
 		table[i].power = power;
 		table[i].frequency = prev_freq = freq;
-
-		/*
-		 * The hertz/watts efficiency ratio should decrease as the
-		 * frequency grows on sane platforms. But this isn't always
-		 * true in practice so warn the user if a higher OPP is more
-		 * power efficient than a lower one.
-		 */
-		opp_eff = freq / power;
-		if (opp_eff >= prev_opp_eff)
-			dev_dbg(dev, "EM: hertz/watts ratio non-monotonically decreasing: em_perf_state %d >= em_perf_state%d\n",
-					i, i - 1);
-		prev_opp_eff = opp_eff;
 	}
 
 	/* Compute the cost of each performance state. */
 	fmax = (u64) table[nr_states - 1].frequency;
-	for (i = 0; i < nr_states; i++) {
+	for (i = nr_states - 1; i >= 0; i--) {
 		unsigned long power_res = em_scale_power(table[i].power);
 
 		table[i].cost = div64_u64(fmax * power_res,
 					  table[i].frequency);
+		if (table[i].cost >= prev_cost) {
+			table[i].flags = EM_PERF_STATE_INEFFICIENT;
+			dev_dbg(dev, "EM: OPP:%lu is inefficient\n",
+				table[i].frequency);
+		} else {
+			prev_cost = table[i].cost;
+		}
 	}
 
 	pd->table = table;
@@ -222,6 +232,43 @@ static int em_create_pd(struct device *dev, int nr_states,
 	return 0;
 }
 
+static void em_cpufreq_update_efficiencies(struct device *dev)
+{
+	struct em_perf_domain *pd = dev->em_pd;
+	struct em_perf_state *table;
+	struct cpufreq_policy *policy;
+	int found = 0;
+	int i;
+
+	if (!_is_cpu_device(dev) || !pd)
+		return;
+
+	policy = cpufreq_cpu_get(cpumask_first(em_span_cpus(pd)));
+	if (!policy) {
+		dev_warn(dev, "EM: Access to CPUFreq policy failed");
+		return;
+	}
+
+	table = pd->table;
+
+	for (i = 0; i < pd->nr_perf_states; i++) {
+		if (!(table[i].flags & EM_PERF_STATE_INEFFICIENT))
+			continue;
+
+		if (!cpufreq_table_set_inefficient(policy, table[i].frequency))
+			found++;
+	}
+
+	if (!found)
+		return;
+
+	/*
+	 * Efficiencies have been installed in CPUFreq, inefficient frequencies
+	 * will be skipped. The EM can do the same.
+	 */
+	pd->flags |= EM_PERF_DOMAIN_SKIP_INEFFICIENCIES;
+}
+
 /**
  * em_pd_get() - Return the performance domain for a device
  * @dev : Device to find the performance domain for
@@ -335,7 +382,10 @@ int em_dev_register_perf_domain(struct device *dev, unsigned int nr_states,
 	if (ret)
 		goto unlock;
 
-	dev->em_pd->milliwatts = milliwatts;
+	if (milliwatts)
+		dev->em_pd->flags |= EM_PERF_DOMAIN_MILLIWATTS;
+
+	em_cpufreq_update_efficiencies(dev);
 
 	em_debug_create_pd(dev);
 	dev_info(dev, "EM: created perf domain\n");