[linux-2.6-block.git] / kernel / sched / cpufreq_schedutil.c

/*
 * CPUFreq governor based on scheduler-provided CPU utilization data.
 *
 * Copyright (C) 2016, Intel Corporation
 * Author: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 */

#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

#include <linux/cpufreq.h>
#include <linux/module.h>
#include <linux/slab.h>
#include <trace/events/power.h>

#include "sched.h"

struct sugov_tunables {
	struct gov_attr_set attr_set;
	unsigned int rate_limit_us;
};

struct sugov_policy {
	struct cpufreq_policy *policy;

	struct sugov_tunables *tunables;
	struct list_head tunables_hook;

	raw_spinlock_t update_lock;  /* For shared policies */
	u64 last_freq_update_time;
	s64 freq_update_delay_ns;
	unsigned int next_freq;

	/* The next fields are only needed if fast switch cannot be used. */
	struct irq_work irq_work;
	struct work_struct work;
	struct mutex work_lock;
	bool work_in_progress;

	bool need_freq_update;
};

struct sugov_cpu {
	struct update_util_data update_util;
	struct sugov_policy *sg_policy;

	unsigned int cached_raw_freq;

	/* The fields below are only needed when sharing a policy. */
	unsigned long util;
	unsigned long max;
	u64 last_update;
};

static DEFINE_PER_CPU(struct sugov_cpu, sugov_cpu);

/************************ Governor internals ***********************/

static bool sugov_should_update_freq(struct sugov_policy *sg_policy, u64 time)
{
	s64 delta_ns;

	if (sg_policy->work_in_progress)
		return false;

	if (unlikely(sg_policy->need_freq_update)) {
		sg_policy->need_freq_update = false;
		/*
		 * This happens when limits change, so forget the previous
		 * next_freq value and force an update.
		 */
		sg_policy->next_freq = UINT_MAX;
		return true;
	}

	delta_ns = time - sg_policy->last_freq_update_time;
	return delta_ns >= sg_policy->freq_update_delay_ns;
}

static void sugov_update_commit(struct sugov_policy *sg_policy, u64 time,
				unsigned int next_freq)
{
	struct cpufreq_policy *policy = sg_policy->policy;

	sg_policy->last_freq_update_time = time;

	if (policy->fast_switch_enabled) {
		if (sg_policy->next_freq == next_freq) {
			trace_cpu_frequency(policy->cur, smp_processor_id());
			return;
		}
		sg_policy->next_freq = next_freq;
		next_freq = cpufreq_driver_fast_switch(policy, next_freq);
		if (next_freq == CPUFREQ_ENTRY_INVALID)
			return;

		policy->cur = next_freq;
		trace_cpu_frequency(next_freq, smp_processor_id());
	} else if (sg_policy->next_freq != next_freq) {
		sg_policy->next_freq = next_freq;
		sg_policy->work_in_progress = true;
		irq_work_queue(&sg_policy->irq_work);
	}
}

/**
 * get_next_freq - Compute a new frequency for a given cpufreq policy.
 * @sg_cpu: schedutil cpu object to compute the new frequency for.
 * @util: Current CPU utilization.
 * @max: CPU capacity.
 *
 * If the utilization is frequency-invariant, choose the new frequency to be
 * proportional to it, that is
 *
 * next_freq = C * max_freq * util / max
 *
 * Otherwise, approximate the would-be frequency-invariant utilization by
 * util_raw * (curr_freq / max_freq) which leads to
 *
 * next_freq = C * curr_freq * util_raw / max
 *
 * Take C = 1.25 for the frequency tipping point at (util / max) = 0.8.
 *
 * The lowest driver-supported frequency which is equal or greater than the raw
 * next_freq (as calculated above) is returned, subject to policy min/max and
 * cpufreq driver limitations.
 */
static unsigned int get_next_freq(struct sugov_cpu *sg_cpu, unsigned long util,
				  unsigned long max)
{
	struct sugov_policy *sg_policy = sg_cpu->sg_policy;
	struct cpufreq_policy *policy = sg_policy->policy;
	unsigned int freq = arch_scale_freq_invariant() ?
				policy->cpuinfo.max_freq : policy->cur;

	freq = (freq + (freq >> 2)) * util / max;

	if (freq == sg_cpu->cached_raw_freq && sg_policy->next_freq != UINT_MAX)
		return sg_policy->next_freq;
	sg_cpu->cached_raw_freq = freq;
	return cpufreq_driver_resolve_freq(policy, freq);
}

static void sugov_update_single(struct update_util_data *hook, u64 time,
				unsigned long util, unsigned long max)
{
	struct sugov_cpu *sg_cpu = container_of(hook, struct sugov_cpu, update_util);
	struct sugov_policy *sg_policy = sg_cpu->sg_policy;
	struct cpufreq_policy *policy = sg_policy->policy;
	unsigned int next_f;

	if (!sugov_should_update_freq(sg_policy, time))
		return;

	next_f = util == ULONG_MAX ? policy->cpuinfo.max_freq :
			get_next_freq(sg_cpu, util, max);
	sugov_update_commit(sg_policy, time, next_f);
}

static unsigned int sugov_next_freq_shared(struct sugov_cpu *sg_cpu,
					   unsigned long util, unsigned long max)
{
	struct sugov_policy *sg_policy = sg_cpu->sg_policy;
	struct cpufreq_policy *policy = sg_policy->policy;
	unsigned int max_f = policy->cpuinfo.max_freq;
	u64 last_freq_update_time = sg_policy->last_freq_update_time;
	unsigned int j;

	if (util == ULONG_MAX)
		return max_f;

	for_each_cpu(j, policy->cpus) {
		struct sugov_cpu *j_sg_cpu;
		unsigned long j_util, j_max;
		s64 delta_ns;

		if (j == smp_processor_id())
			continue;

		j_sg_cpu = &per_cpu(sugov_cpu, j);
		/*
		 * If the CPU utilization was last updated before the previous
		 * frequency update and the time elapsed between the last update
		 * of the CPU utilization and the last frequency update is long
		 * enough, don't take the CPU into account as it probably is
		 * idle now.
		 */
		delta_ns = last_freq_update_time - j_sg_cpu->last_update;
		if (delta_ns > TICK_NSEC)
			continue;

		j_util = j_sg_cpu->util;
		if (j_util == ULONG_MAX)
			return max_f;

		j_max = j_sg_cpu->max;
		if (j_util * max > j_max * util) {
			util = j_util;
			max = j_max;
		}
	}

	return get_next_freq(sg_cpu, util, max);
}

static void sugov_update_shared(struct update_util_data *hook, u64 time,
				unsigned long util, unsigned long max)
{
	struct sugov_cpu *sg_cpu = container_of(hook, struct sugov_cpu, update_util);
	struct sugov_policy *sg_policy = sg_cpu->sg_policy;
	unsigned int next_f;

	raw_spin_lock(&sg_policy->update_lock);

	sg_cpu->util = util;
	sg_cpu->max = max;
	sg_cpu->last_update = time;

	if (sugov_should_update_freq(sg_policy, time)) {
		next_f = sugov_next_freq_shared(sg_cpu, util, max);
		sugov_update_commit(sg_policy, time, next_f);
	}

	raw_spin_unlock(&sg_policy->update_lock);
}

static void sugov_work(struct work_struct *work)
{
	struct sugov_policy *sg_policy = container_of(work, struct sugov_policy, work);

	mutex_lock(&sg_policy->work_lock);
	__cpufreq_driver_target(sg_policy->policy, sg_policy->next_freq,
				CPUFREQ_RELATION_L);
	mutex_unlock(&sg_policy->work_lock);

	sg_policy->work_in_progress = false;
}

static void sugov_irq_work(struct irq_work *irq_work)
{
	struct sugov_policy *sg_policy;

	sg_policy = container_of(irq_work, struct sugov_policy, irq_work);
	schedule_work_on(smp_processor_id(), &sg_policy->work);
}

/************************** sysfs interface ************************/

static struct sugov_tunables *global_tunables;
static DEFINE_MUTEX(global_tunables_lock);

static inline struct sugov_tunables *to_sugov_tunables(struct gov_attr_set *attr_set)
{
	return container_of(attr_set, struct sugov_tunables, attr_set);
}

static ssize_t rate_limit_us_show(struct gov_attr_set *attr_set, char *buf)
{
	struct sugov_tunables *tunables = to_sugov_tunables(attr_set);

	return sprintf(buf, "%u\n", tunables->rate_limit_us);
}

static ssize_t rate_limit_us_store(struct gov_attr_set *attr_set, const char *buf,
				   size_t count)
{
	struct sugov_tunables *tunables = to_sugov_tunables(attr_set);
	struct sugov_policy *sg_policy;
	unsigned int rate_limit_us;

	if (kstrtouint(buf, 10, &rate_limit_us))
		return -EINVAL;

	tunables->rate_limit_us = rate_limit_us;

	list_for_each_entry(sg_policy, &attr_set->policy_list, tunables_hook)
		sg_policy->freq_update_delay_ns = rate_limit_us * NSEC_PER_USEC;

	return count;
}

static struct governor_attr rate_limit_us = __ATTR_RW(rate_limit_us);

static struct attribute *sugov_attributes[] = {
	&rate_limit_us.attr,
	NULL
};

static struct kobj_type sugov_tunables_ktype = {
	.default_attrs = sugov_attributes,
	.sysfs_ops = &governor_sysfs_ops,
};

/********************** cpufreq governor interface *********************/

static struct cpufreq_governor schedutil_gov;

static struct sugov_policy *sugov_policy_alloc(struct cpufreq_policy *policy)
{
	struct sugov_policy *sg_policy;

	sg_policy = kzalloc(sizeof(*sg_policy), GFP_KERNEL);
	if (!sg_policy)
		return NULL;

	sg_policy->policy = policy;
	init_irq_work(&sg_policy->irq_work, sugov_irq_work);
	INIT_WORK(&sg_policy->work, sugov_work);
	mutex_init(&sg_policy->work_lock);
	raw_spin_lock_init(&sg_policy->update_lock);
	return sg_policy;
}

static void sugov_policy_free(struct sugov_policy *sg_policy)
{
	mutex_destroy(&sg_policy->work_lock);
	kfree(sg_policy);
}

static struct sugov_tunables *sugov_tunables_alloc(struct sugov_policy *sg_policy)
{
	struct sugov_tunables *tunables;

	tunables = kzalloc(sizeof(*tunables), GFP_KERNEL);
	if (tunables) {
		gov_attr_set_init(&tunables->attr_set, &sg_policy->tunables_hook);
		if (!have_governor_per_policy())
			global_tunables = tunables;
	}
	return tunables;
}

static void sugov_tunables_free(struct sugov_tunables *tunables)
{
	if (!have_governor_per_policy())
		global_tunables = NULL;

	kfree(tunables);
}

static int sugov_init(struct cpufreq_policy *policy)
{
	struct sugov_policy *sg_policy;
	struct sugov_tunables *tunables;
	unsigned int lat;
	int ret = 0;

	/* State should be equivalent to EXIT */
	if (policy->governor_data)
		return -EBUSY;

	sg_policy = sugov_policy_alloc(policy);
	if (!sg_policy)
		return -ENOMEM;

	mutex_lock(&global_tunables_lock);

	if (global_tunables) {
		if (WARN_ON(have_governor_per_policy())) {
			ret = -EINVAL;
			goto free_sg_policy;
		}
		policy->governor_data = sg_policy;
		sg_policy->tunables = global_tunables;

		gov_attr_set_get(&global_tunables->attr_set, &sg_policy->tunables_hook);
		goto out;
	}

	tunables = sugov_tunables_alloc(sg_policy);
	if (!tunables) {
		ret = -ENOMEM;
		goto free_sg_policy;
	}

	tunables->rate_limit_us = LATENCY_MULTIPLIER;
	lat = policy->cpuinfo.transition_latency / NSEC_PER_USEC;
	if (lat)
		tunables->rate_limit_us *= lat;

	policy->governor_data = sg_policy;
	sg_policy->tunables = tunables;

	ret = kobject_init_and_add(&tunables->attr_set.kobj, &sugov_tunables_ktype,
				   get_governor_parent_kobj(policy), "%s",
				   schedutil_gov.name);
	if (ret)
		goto fail;

 out:
	mutex_unlock(&global_tunables_lock);

	cpufreq_enable_fast_switch(policy);
	return 0;

 fail:
	policy->governor_data = NULL;
	sugov_tunables_free(tunables);

 free_sg_policy:
	mutex_unlock(&global_tunables_lock);

	sugov_policy_free(sg_policy);
	pr_err("initialization failed (error %d)\n", ret);
	return ret;
}

static void sugov_exit(struct cpufreq_policy *policy)
{
	struct sugov_policy *sg_policy = policy->governor_data;
	struct sugov_tunables *tunables = sg_policy->tunables;
	unsigned int count;

	cpufreq_disable_fast_switch(policy);

	mutex_lock(&global_tunables_lock);

	count = gov_attr_set_put(&tunables->attr_set, &sg_policy->tunables_hook);
	policy->governor_data = NULL;
	if (!count)
		sugov_tunables_free(tunables);

	mutex_unlock(&global_tunables_lock);

	sugov_policy_free(sg_policy);
}

static int sugov_start(struct cpufreq_policy *policy)
{
	struct sugov_policy *sg_policy = policy->governor_data;
	unsigned int cpu;

	sg_policy->freq_update_delay_ns = sg_policy->tunables->rate_limit_us * NSEC_PER_USEC;
	sg_policy->last_freq_update_time = 0;
	sg_policy->next_freq = UINT_MAX;
	sg_policy->work_in_progress = false;
	sg_policy->need_freq_update = false;

	for_each_cpu(cpu, policy->cpus) {
		struct sugov_cpu *sg_cpu = &per_cpu(sugov_cpu, cpu);

		sg_cpu->sg_policy = sg_policy;
		if (policy_is_shared(policy)) {
			sg_cpu->util = ULONG_MAX;
			sg_cpu->max = 0;
			sg_cpu->last_update = 0;
			sg_cpu->cached_raw_freq = 0;
			cpufreq_add_update_util_hook(cpu, &sg_cpu->update_util,
						     sugov_update_shared);
		} else {
			cpufreq_add_update_util_hook(cpu, &sg_cpu->update_util,
						     sugov_update_single);
		}
	}
	return 0;
}

static void sugov_stop(struct cpufreq_policy *policy)
{
	struct sugov_policy *sg_policy = policy->governor_data;
	unsigned int cpu;

	for_each_cpu(cpu, policy->cpus)
		cpufreq_remove_update_util_hook(cpu);

	synchronize_sched();

	irq_work_sync(&sg_policy->irq_work);
	cancel_work_sync(&sg_policy->work);
}

static void sugov_limits(struct cpufreq_policy *policy)
{
	struct sugov_policy *sg_policy = policy->governor_data;

	if (!policy->fast_switch_enabled) {
		mutex_lock(&sg_policy->work_lock);
		cpufreq_policy_apply_limits(policy);
		mutex_unlock(&sg_policy->work_lock);
	}

	sg_policy->need_freq_update = true;
}

static struct cpufreq_governor schedutil_gov = {
	.name = "schedutil",
	.owner = THIS_MODULE,
	.init = sugov_init,
	.exit = sugov_exit,
	.start = sugov_start,
	.stop = sugov_stop,
	.limits = sugov_limits,
};

static int __init sugov_module_init(void)
{
	return cpufreq_register_governor(&schedutil_gov);
}

static void __exit sugov_module_exit(void)
{
	cpufreq_unregister_governor(&schedutil_gov);
}

MODULE_AUTHOR("Rafael J. Wysocki <rafael.j.wysocki@intel.com>");
MODULE_DESCRIPTION("Utilization-based CPU frequency selection");
MODULE_LICENSE("GPL");

#ifdef CONFIG_CPU_FREQ_DEFAULT_GOV_SCHEDUTIL
struct cpufreq_governor *cpufreq_default_governor(void)
{
	return &schedutil_gov;
}

fs_initcall(sugov_module_init);
#else
module_init(sugov_module_init);
#endif
module_exit(sugov_module_exit);
Commit	Line	Data
9bdcb44e RW	1	/*
	2	* CPUFreq governor based on scheduler-provided CPU utilization data.
	3	*
	4	* Copyright (C) 2016, Intel Corporation
	5	* Author: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
	6	*
	7	* This program is free software; you can redistribute it and/or modify
	8	* it under the terms of the GNU General Public License version 2 as
	9	* published by the Free Software Foundation.
	10	*/
	11
60f05e86 VK	12	#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
60f05e86 VK	13
9bdcb44e RW	14	#include <linux/cpufreq.h>
	15	#include <linux/module.h>
	16	#include <linux/slab.h>
	17	#include <trace/events/power.h>
	18
	19	#include "sched.h"
	20
	21	struct sugov_tunables {
	22	struct gov_attr_set attr_set;
	23	unsigned int rate_limit_us;
	24	};
	25
	26	struct sugov_policy {
	27	struct cpufreq_policy *policy;
	28
	29	struct sugov_tunables *tunables;
	30	struct list_head tunables_hook;
	31
	32	raw_spinlock_t update_lock; /* For shared policies */
	33	u64 last_freq_update_time;
	34	s64 freq_update_delay_ns;
	35	unsigned int next_freq;
	36
	37	/* The next fields are only needed if fast switch cannot be used. */
	38	struct irq_work irq_work;
	39	struct work_struct work;
	40	struct mutex work_lock;
	41	bool work_in_progress;
	42
	43	bool need_freq_update;
	44	};
	45
	46	struct sugov_cpu {
	47	struct update_util_data update_util;
	48	struct sugov_policy *sg_policy;
	49
5cbea469 SM	50	unsigned int cached_raw_freq;
5cbea469 SM	51
9bdcb44e RW	52	/* The fields below are only needed when sharing a policy. */
	53	unsigned long util;
	54	unsigned long max;
	55	u64 last_update;
	56	};
	57
	58	static DEFINE_PER_CPU(struct sugov_cpu, sugov_cpu);
	59
	60	/********************** Governor internals *********************/
	61
	62	static bool sugov_should_update_freq(struct sugov_policy *sg_policy, u64 time)
	63	{
	64	s64 delta_ns;
	65
	66	if (sg_policy->work_in_progress)
	67	return false;
	68
	69	if (unlikely(sg_policy->need_freq_update)) {
	70	sg_policy->need_freq_update = false;
	71	/*
	72	* This happens when limits change, so forget the previous
	73	* next_freq value and force an update.
	74	*/
	75	sg_policy->next_freq = UINT_MAX;
	76	return true;
	77	}
	78
	79	delta_ns = time - sg_policy->last_freq_update_time;
	80	return delta_ns >= sg_policy->freq_update_delay_ns;
	81	}
	82
	83	static void sugov_update_commit(struct sugov_policy *sg_policy, u64 time,
	84	unsigned int next_freq)
	85	{
	86	struct cpufreq_policy *policy = sg_policy->policy;
	87
	88	sg_policy->last_freq_update_time = time;
	89
	90	if (policy->fast_switch_enabled) {
	91	if (sg_policy->next_freq == next_freq) {
	92	trace_cpu_frequency(policy->cur, smp_processor_id());
	93	return;
	94	}
	95	sg_policy->next_freq = next_freq;
	96	next_freq = cpufreq_driver_fast_switch(policy, next_freq);
	97	if (next_freq == CPUFREQ_ENTRY_INVALID)
	98	return;
	99
	100	policy->cur = next_freq;
	101	trace_cpu_frequency(next_freq, smp_processor_id());
	102	} else if (sg_policy->next_freq != next_freq) {
	103	sg_policy->next_freq = next_freq;
	104	sg_policy->work_in_progress = true;
	105	irq_work_queue(&sg_policy->irq_work);
	106	}
	107	}
	108
	109	/**
	110	* get_next_freq - Compute a new frequency for a given cpufreq policy.
5cbea469	111	* @sg_cpu: schedutil cpu object to compute the new frequency for.
9bdcb44e RW	112	* @util: Current CPU utilization.
	113	* @max: CPU capacity.
	114	*
	115	* If the utilization is frequency-invariant, choose the new frequency to be
	116	* proportional to it, that is
	117	*
	118	* next_freq = C * max_freq * util / max
	119	*
	120	* Otherwise, approximate the would-be frequency-invariant utilization by
	121	* util_raw * (curr_freq / max_freq) which leads to
	122	*
	123	* next_freq = C * curr_freq * util_raw / max
	124	*
	125	* Take C = 1.25 for the frequency tipping point at (util / max) = 0.8.
5cbea469 SM	126	*
	127	* The lowest driver-supported frequency which is equal or greater than the raw
	128	* next_freq (as calculated above) is returned, subject to policy min/max and
	129	* cpufreq driver limitations.
9bdcb44e	130	*/
5cbea469 SM	131	static unsigned int get_next_freq(struct sugov_cpu *sg_cpu, unsigned long util,
5cbea469 SM	132	unsigned long max)
9bdcb44e	133	{
5cbea469 SM	134	struct sugov_policy *sg_policy = sg_cpu->sg_policy;
5cbea469 SM	135	struct cpufreq_policy *policy = sg_policy->policy;
9bdcb44e RW	136	unsigned int freq = arch_scale_freq_invariant() ?
	137	policy->cpuinfo.max_freq : policy->cur;
	138
5cbea469 SM	139	freq = (freq + (freq >> 2)) * util / max;
	140
	141	if (freq == sg_cpu->cached_raw_freq && sg_policy->next_freq != UINT_MAX)
	142	return sg_policy->next_freq;
	143	sg_cpu->cached_raw_freq = freq;
	144	return cpufreq_driver_resolve_freq(policy, freq);
9bdcb44e RW	145	}
	146
	147	static void sugov_update_single(struct update_util_data *hook, u64 time,
	148	unsigned long util, unsigned long max)
	149	{
	150	struct sugov_cpu *sg_cpu = container_of(hook, struct sugov_cpu, update_util);
	151	struct sugov_policy *sg_policy = sg_cpu->sg_policy;
	152	struct cpufreq_policy *policy = sg_policy->policy;
	153	unsigned int next_f;
	154
	155	if (!sugov_should_update_freq(sg_policy, time))
	156	return;
	157
	158	next_f = util == ULONG_MAX ? policy->cpuinfo.max_freq :
5cbea469	159	get_next_freq(sg_cpu, util, max);
9bdcb44e RW	160	sugov_update_commit(sg_policy, time, next_f);
	161	}
	162
5cbea469	163	static unsigned int sugov_next_freq_shared(struct sugov_cpu *sg_cpu,
9bdcb44e RW	164	unsigned long util, unsigned long max)
9bdcb44e RW	165	{
5cbea469	166	struct sugov_policy *sg_policy = sg_cpu->sg_policy;
9bdcb44e RW	167	struct cpufreq_policy *policy = sg_policy->policy;
	168	unsigned int max_f = policy->cpuinfo.max_freq;
	169	u64 last_freq_update_time = sg_policy->last_freq_update_time;
	170	unsigned int j;
	171
	172	if (util == ULONG_MAX)
	173	return max_f;
	174
	175	for_each_cpu(j, policy->cpus) {
	176	struct sugov_cpu *j_sg_cpu;
	177	unsigned long j_util, j_max;
	178	s64 delta_ns;
	179
	180	if (j == smp_processor_id())
	181	continue;
	182
	183	j_sg_cpu = &per_cpu(sugov_cpu, j);
	184	/*
	185	* If the CPU utilization was last updated before the previous
	186	* frequency update and the time elapsed between the last update
	187	* of the CPU utilization and the last frequency update is long
	188	* enough, don't take the CPU into account as it probably is
	189	* idle now.
	190	*/
	191	delta_ns = last_freq_update_time - j_sg_cpu->last_update;
	192	if (delta_ns > TICK_NSEC)
	193	continue;
	194
	195	j_util = j_sg_cpu->util;
	196	if (j_util == ULONG_MAX)
	197	return max_f;
	198
	199	j_max = j_sg_cpu->max;
	200	if (j_util * max > j_max * util) {
	201	util = j_util;
	202	max = j_max;
	203	}
	204	}
	205
5cbea469	206	return get_next_freq(sg_cpu, util, max);
9bdcb44e RW	207	}
	208
	209	static void sugov_update_shared(struct update_util_data *hook, u64 time,
	210	unsigned long util, unsigned long max)
	211	{
	212	struct sugov_cpu *sg_cpu = container_of(hook, struct sugov_cpu, update_util);
	213	struct sugov_policy *sg_policy = sg_cpu->sg_policy;
	214	unsigned int next_f;
	215
	216	raw_spin_lock(&sg_policy->update_lock);
	217
	218	sg_cpu->util = util;
	219	sg_cpu->max = max;
	220	sg_cpu->last_update = time;
	221
	222	if (sugov_should_update_freq(sg_policy, time)) {
5cbea469	223	next_f = sugov_next_freq_shared(sg_cpu, util, max);
9bdcb44e RW	224	sugov_update_commit(sg_policy, time, next_f);
	225	}
	226
	227	raw_spin_unlock(&sg_policy->update_lock);
	228	}
	229
	230	static void sugov_work(struct work_struct *work)
	231	{
	232	struct sugov_policy *sg_policy = container_of(work, struct sugov_policy, work);
	233
	234	mutex_lock(&sg_policy->work_lock);
	235	__cpufreq_driver_target(sg_policy->policy, sg_policy->next_freq,
	236	CPUFREQ_RELATION_L);
	237	mutex_unlock(&sg_policy->work_lock);
	238
	239	sg_policy->work_in_progress = false;
	240	}
	241
	242	static void sugov_irq_work(struct irq_work *irq_work)
	243	{
	244	struct sugov_policy *sg_policy;
	245
	246	sg_policy = container_of(irq_work, struct sugov_policy, irq_work);
	247	schedule_work_on(smp_processor_id(), &sg_policy->work);
	248	}
	249
	250	/************************ sysfs interface **********************/
	251
	252	static struct sugov_tunables *global_tunables;
	253	static DEFINE_MUTEX(global_tunables_lock);
	254
	255	static inline struct sugov_tunables to_sugov_tunables(struct gov_attr_set attr_set)
	256	{
	257	return container_of(attr_set, struct sugov_tunables, attr_set);
	258	}
	259
	260	static ssize_t rate_limit_us_show(struct gov_attr_set attr_set, char buf)
	261	{
	262	struct sugov_tunables *tunables = to_sugov_tunables(attr_set);
	263
	264	return sprintf(buf, "%u\n", tunables->rate_limit_us);
	265	}
	266
	267	static ssize_t rate_limit_us_store(struct gov_attr_set attr_set, const char buf,
	268	size_t count)
	269	{
	270	struct sugov_tunables *tunables = to_sugov_tunables(attr_set);
	271	struct sugov_policy *sg_policy;
	272	unsigned int rate_limit_us;
	273
	274	if (kstrtouint(buf, 10, &rate_limit_us))
	275	return -EINVAL;
	276
	277	tunables->rate_limit_us = rate_limit_us;
	278
	279	list_for_each_entry(sg_policy, &attr_set->policy_list, tunables_hook)
	280	sg_policy->freq_update_delay_ns = rate_limit_us * NSEC_PER_USEC;
	281
	282	return count;
	283	}
	284
	285	static struct governor_attr rate_limit_us = __ATTR_RW(rate_limit_us);
	286
	287	static struct attribute *sugov_attributes[] = {
288	&rate_limit_us.attr,
289	NULL
290	};
291
292	static struct kobj_type sugov_tunables_ktype = {
293	.default_attrs = sugov_attributes,
294	.sysfs_ops = &governor_sysfs_ops,
295	};
296
297	/******************** cpufreq governor interface *******************/
298
299	static struct cpufreq_governor schedutil_gov;
300
301	static struct sugov_policy sugov_policy_alloc(struct cpufreq_policy policy)
302	{
303	struct sugov_policy *sg_policy;
304
305	sg_policy = kzalloc(sizeof(*sg_policy), GFP_KERNEL);
306	if (!sg_policy)
307	return NULL;
308
309	sg_policy->policy = policy;
310	init_irq_work(&sg_policy->irq_work, sugov_irq_work);
311	INIT_WORK(&sg_policy->work, sugov_work);
312	mutex_init(&sg_policy->work_lock);
313	raw_spin_lock_init(&sg_policy->update_lock);
314	return sg_policy;
315	}
316
317	static void sugov_policy_free(struct sugov_policy *sg_policy)
318	{
319	mutex_destroy(&sg_policy->work_lock);
320	kfree(sg_policy);
321	}
322
323	static struct sugov_tunables sugov_tunables_alloc(struct sugov_policy sg_policy)
324	{
325	struct sugov_tunables *tunables;
326
327	tunables = kzalloc(sizeof(*tunables), GFP_KERNEL);
328	if (tunables) {
329	gov_attr_set_init(&tunables->attr_set, &sg_policy->tunables_hook);
330	if (!have_governor_per_policy())
331	global_tunables = tunables;
332	}
333	return tunables;
334	}
335
336	static void sugov_tunables_free(struct sugov_tunables *tunables)
337	{
338	if (!have_governor_per_policy())
339	global_tunables = NULL;
340
341	kfree(tunables);
342	}
343
344	static int sugov_init(struct cpufreq_policy *policy)
345	{
346	struct sugov_policy *sg_policy;
347	struct sugov_tunables *tunables;
348	unsigned int lat;
349	int ret = 0;
350
351	/* State should be equivalent to EXIT */
352	if (policy->governor_data)
353	return -EBUSY;
354
355	sg_policy = sugov_policy_alloc(policy);
356	if (!sg_policy)
357	return -ENOMEM;
358
359	mutex_lock(&global_tunables_lock);
360
361	if (global_tunables) {
362	if (WARN_ON(have_governor_per_policy())) {
363	ret = -EINVAL;
364	goto free_sg_policy;
365	}
366	policy->governor_data = sg_policy;
367	sg_policy->tunables = global_tunables;
368
369	gov_attr_set_get(&global_tunables->attr_set, &sg_policy->tunables_hook);
370	goto out;
371	}
372
373	tunables = sugov_tunables_alloc(sg_policy);
374	if (!tunables) {
375	ret = -ENOMEM;
376	goto free_sg_policy;
377	}
378
379	tunables->rate_limit_us = LATENCY_MULTIPLIER;
380	lat = policy->cpuinfo.transition_latency / NSEC_PER_USEC;
381	if (lat)
382	tunables->rate_limit_us *= lat;
383
384	policy->governor_data = sg_policy;
385	sg_policy->tunables = tunables;
386
387	ret = kobject_init_and_add(&tunables->attr_set.kobj, &sugov_tunables_ktype,
388	get_governor_parent_kobj(policy), "%s",
389	schedutil_gov.name);
390	if (ret)
391	goto fail;
392
393	out:
394	mutex_unlock(&global_tunables_lock);
395
396	cpufreq_enable_fast_switch(policy);
397	return 0;
398
399	fail:
400	policy->governor_data = NULL;
401	sugov_tunables_free(tunables);
402
403	free_sg_policy:
404	mutex_unlock(&global_tunables_lock);
405
406	sugov_policy_free(sg_policy);
60f05e86	407	pr_err("initialization failed (error %d)\n", ret);
9bdcb44e RW	408	return ret;
	409	}
	410
e788892b	411	static void sugov_exit(struct cpufreq_policy *policy)
9bdcb44e RW	412	{
	413	struct sugov_policy *sg_policy = policy->governor_data;
	414	struct sugov_tunables *tunables = sg_policy->tunables;
	415	unsigned int count;
	416
6c9d9c81 RW	417	cpufreq_disable_fast_switch(policy);
6c9d9c81 RW	418
9bdcb44e RW	419	mutex_lock(&global_tunables_lock);
	420
	421	count = gov_attr_set_put(&tunables->attr_set, &sg_policy->tunables_hook);
	422	policy->governor_data = NULL;
	423	if (!count)
	424	sugov_tunables_free(tunables);
	425
	426	mutex_unlock(&global_tunables_lock);
	427
	428	sugov_policy_free(sg_policy);
9bdcb44e RW	429	}
	430
	431	static int sugov_start(struct cpufreq_policy *policy)
	432	{
	433	struct sugov_policy *sg_policy = policy->governor_data;
	434	unsigned int cpu;
	435
	436	sg_policy->freq_update_delay_ns = sg_policy->tunables->rate_limit_us * NSEC_PER_USEC;
	437	sg_policy->last_freq_update_time = 0;
	438	sg_policy->next_freq = UINT_MAX;
	439	sg_policy->work_in_progress = false;
	440	sg_policy->need_freq_update = false;
	441
	442	for_each_cpu(cpu, policy->cpus) {
	443	struct sugov_cpu *sg_cpu = &per_cpu(sugov_cpu, cpu);
	444
	445	sg_cpu->sg_policy = sg_policy;
	446	if (policy_is_shared(policy)) {
	447	sg_cpu->util = ULONG_MAX;
	448	sg_cpu->max = 0;
	449	sg_cpu->last_update = 0;
5cbea469	450	sg_cpu->cached_raw_freq = 0;
9bdcb44e RW	451	cpufreq_add_update_util_hook(cpu, &sg_cpu->update_util,
	452	sugov_update_shared);
	453	} else {
	454	cpufreq_add_update_util_hook(cpu, &sg_cpu->update_util,
	455	sugov_update_single);
	456	}
	457	}
	458	return 0;
	459	}
	460
e788892b	461	static void sugov_stop(struct cpufreq_policy *policy)
9bdcb44e RW	462	{
	463	struct sugov_policy *sg_policy = policy->governor_data;
	464	unsigned int cpu;
	465
	466	for_each_cpu(cpu, policy->cpus)
	467	cpufreq_remove_update_util_hook(cpu);
	468
	469	synchronize_sched();
	470
	471	irq_work_sync(&sg_policy->irq_work);
	472	cancel_work_sync(&sg_policy->work);
9bdcb44e RW	473	}
9bdcb44e RW	474
e788892b	475	static void sugov_limits(struct cpufreq_policy *policy)
9bdcb44e RW	476	{
	477	struct sugov_policy *sg_policy = policy->governor_data;
	478
	479	if (!policy->fast_switch_enabled) {
	480	mutex_lock(&sg_policy->work_lock);
bf2be2de	481	cpufreq_policy_apply_limits(policy);
9bdcb44e RW	482	mutex_unlock(&sg_policy->work_lock);
	483	}
	484
	485	sg_policy->need_freq_update = true;
9bdcb44e RW	486	}
	487
	488	static struct cpufreq_governor schedutil_gov = {
	489	.name = "schedutil",
9bdcb44e	490	.owner = THIS_MODULE,
e788892b RW	491	.init = sugov_init,
	492	.exit = sugov_exit,
	493	.start = sugov_start,
	494	.stop = sugov_stop,
	495	.limits = sugov_limits,
9bdcb44e RW	496	};
	497
	498	static int __init sugov_module_init(void)
	499	{
	500	return cpufreq_register_governor(&schedutil_gov);
	501	}
	502
	503	static void __exit sugov_module_exit(void)
	504	{
	505	cpufreq_unregister_governor(&schedutil_gov);
	506	}
	507
	508	MODULE_AUTHOR("Rafael J. Wysocki <rafael.j.wysocki@intel.com>");
	509	MODULE_DESCRIPTION("Utilization-based CPU frequency selection");
	510	MODULE_LICENSE("GPL");
	511
	512	#ifdef CONFIG_CPU_FREQ_DEFAULT_GOV_SCHEDUTIL
	513	struct cpufreq_governor *cpufreq_default_governor(void)
	514	{
	515	return &schedutil_gov;
	516	}
	517
	518	fs_initcall(sugov_module_init);
	519	#else
	520	module_init(sugov_module_init);
	521	#endif
	522	module_exit(sugov_module_exit);