[deliverable/linux.git] / drivers / cpufreq / cpufreq_ondemand.c

/*
 *  drivers/cpufreq/cpufreq_ondemand.c
 *
 *  Copyright (C)  2001 Russell King
 *            (C)  2003 Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>.
 *                      Jun Nakajima <jun.nakajima@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.
 */

#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/cpufreq.h>
#include <linux/cpu.h>
#include <linux/jiffies.h>
#include <linux/kernel_stat.h>
#include <linux/mutex.h>

/*
 * dbs is used in this file as a shortform for demandbased switching
 * It helps to keep variable names smaller, simpler
 */

#define DEF_FREQUENCY_UP_THRESHOLD		(80)
#define MIN_FREQUENCY_UP_THRESHOLD		(11)
#define MAX_FREQUENCY_UP_THRESHOLD		(100)

/*
 * The polling frequency of this governor depends on the capability of
 * the processor. Default polling frequency is 1000 times the transition
 * latency of the processor. The governor will work on any processor with
 * transition latency <= 10mS, using appropriate sampling
 * rate.
 * For CPUs with transition latency > 10mS (mostly drivers with CPUFREQ_ETERNAL)
 * this governor will not work.
 * All times here are in uS.
 */
static unsigned int def_sampling_rate;
#define MIN_SAMPLING_RATE_RATIO			(2)
/* for correct statistics, we need at least 10 ticks between each measure */
#define MIN_STAT_SAMPLING_RATE			(MIN_SAMPLING_RATE_RATIO * jiffies_to_usecs(10))
#define MIN_SAMPLING_RATE			(def_sampling_rate / MIN_SAMPLING_RATE_RATIO)
#define MAX_SAMPLING_RATE			(500 * def_sampling_rate)
#define DEF_SAMPLING_RATE_LATENCY_MULTIPLIER	(1000)
#define TRANSITION_LATENCY_LIMIT		(10 * 1000)

static void do_dbs_timer(void *data);

struct cpu_dbs_info_s {
	cputime64_t prev_cpu_idle;
	cputime64_t prev_cpu_wall;
	struct cpufreq_policy *cur_policy;
 	struct work_struct work;
	unsigned int enable;
};
static DEFINE_PER_CPU(struct cpu_dbs_info_s, cpu_dbs_info);

static unsigned int dbs_enable;	/* number of CPUs using this policy */

/*
 * DEADLOCK ALERT! There is a ordering requirement between cpu_hotplug
 * lock and dbs_mutex. cpu_hotplug lock should always be held before
 * dbs_mutex. If any function that can potentially take cpu_hotplug lock
 * (like __cpufreq_driver_target()) is being called with dbs_mutex taken, then
 * cpu_hotplug lock should be taken before that. Note that cpu_hotplug lock
 * is recursive for the same process. -Venki
 */
static DEFINE_MUTEX(dbs_mutex);

static struct workqueue_struct	*kondemand_wq;

struct dbs_tuners {
	unsigned int sampling_rate;
	unsigned int up_threshold;
	unsigned int ignore_nice;
};

static struct dbs_tuners dbs_tuners_ins = {
	.up_threshold = DEF_FREQUENCY_UP_THRESHOLD,
	.ignore_nice = 0,
};

static inline cputime64_t get_cpu_idle_time(unsigned int cpu)
{
	cputime64_t retval;

	retval = cputime64_add(kstat_cpu(cpu).cpustat.idle,
			kstat_cpu(cpu).cpustat.iowait);

	if (dbs_tuners_ins.ignore_nice)
		retval = cputime64_add(retval, kstat_cpu(cpu).cpustat.nice);

	return retval;
}

/************************** sysfs interface ************************/
static ssize_t show_sampling_rate_max(struct cpufreq_policy *policy, char *buf)
{
	return sprintf (buf, "%u\n", MAX_SAMPLING_RATE);
}

static ssize_t show_sampling_rate_min(struct cpufreq_policy *policy, char *buf)
{
	return sprintf (buf, "%u\n", MIN_SAMPLING_RATE);
}

#define define_one_ro(_name)		\
static struct freq_attr _name =		\
__ATTR(_name, 0444, show_##_name, NULL)

define_one_ro(sampling_rate_max);
define_one_ro(sampling_rate_min);

/* cpufreq_ondemand Governor Tunables */
#define show_one(file_name, object)					\
static ssize_t show_##file_name						\
(struct cpufreq_policy *unused, char *buf)				\
{									\
	return sprintf(buf, "%u\n", dbs_tuners_ins.object);		\
}
show_one(sampling_rate, sampling_rate);
show_one(up_threshold, up_threshold);
show_one(ignore_nice_load, ignore_nice);

static ssize_t store_sampling_rate(struct cpufreq_policy *unused,
		const char *buf, size_t count)
{
	unsigned int input;
	int ret;
	ret = sscanf(buf, "%u", &input);

	mutex_lock(&dbs_mutex);
	if (ret != 1 || input > MAX_SAMPLING_RATE || input < MIN_SAMPLING_RATE) {
		mutex_unlock(&dbs_mutex);
		return -EINVAL;
	}

	dbs_tuners_ins.sampling_rate = input;
	mutex_unlock(&dbs_mutex);

	return count;
}

static ssize_t store_up_threshold(struct cpufreq_policy *unused,
		const char *buf, size_t count)
{
	unsigned int input;
	int ret;
	ret = sscanf(buf, "%u", &input);

	mutex_lock(&dbs_mutex);
	if (ret != 1 || input > MAX_FREQUENCY_UP_THRESHOLD ||
			input < MIN_FREQUENCY_UP_THRESHOLD) {
		mutex_unlock(&dbs_mutex);
		return -EINVAL;
	}

	dbs_tuners_ins.up_threshold = input;
	mutex_unlock(&dbs_mutex);

	return count;
}

static ssize_t store_ignore_nice_load(struct cpufreq_policy *policy,
		const char *buf, size_t count)
{
	unsigned int input;
	int ret;

	unsigned int j;

	ret = sscanf(buf, "%u", &input);
	if ( ret != 1 )
		return -EINVAL;

	if ( input > 1 )
		input = 1;

	mutex_lock(&dbs_mutex);
	if ( input == dbs_tuners_ins.ignore_nice ) { /* nothing to do */
		mutex_unlock(&dbs_mutex);
		return count;
	}
	dbs_tuners_ins.ignore_nice = input;

	/* we need to re-evaluate prev_cpu_idle */
	for_each_online_cpu(j) {
		struct cpu_dbs_info_s *dbs_info;
		dbs_info = &per_cpu(cpu_dbs_info, j);
		dbs_info->prev_cpu_idle = get_cpu_idle_time(j);
		dbs_info->prev_cpu_wall = get_jiffies_64();
	}
	mutex_unlock(&dbs_mutex);

	return count;
}

#define define_one_rw(_name) \
static struct freq_attr _name = \
__ATTR(_name, 0644, show_##_name, store_##_name)

define_one_rw(sampling_rate);
define_one_rw(up_threshold);
define_one_rw(ignore_nice_load);

static struct attribute * dbs_attributes[] = {
	&sampling_rate_max.attr,
	&sampling_rate_min.attr,
	&sampling_rate.attr,
	&up_threshold.attr,
	&ignore_nice_load.attr,
	NULL
};

static struct attribute_group dbs_attr_group = {
	.attrs = dbs_attributes,
	.name = "ondemand",
};

/************************** sysfs end ************************/

static void dbs_check_cpu(struct cpu_dbs_info_s *this_dbs_info)
{
	unsigned int idle_ticks, total_ticks;
	unsigned int load;
	cputime64_t cur_jiffies;

	struct cpufreq_policy *policy;
	unsigned int j;

	if (!this_dbs_info->enable)
		return;

	policy = this_dbs_info->cur_policy;
	cur_jiffies = jiffies64_to_cputime64(get_jiffies_64());
	total_ticks = (unsigned int) cputime64_sub(cur_jiffies,
			this_dbs_info->prev_cpu_wall);
	this_dbs_info->prev_cpu_wall = cur_jiffies;
	/*
	 * Every sampling_rate, we check, if current idle time is less
	 * than 20% (default), then we try to increase frequency
	 * Every sampling_rate, we look for a the lowest
	 * frequency which can sustain the load while keeping idle time over
	 * 30%. If such a frequency exist, we try to decrease to this frequency.
	 *
	 * Any frequency increase takes it to the maximum frequency.
	 * Frequency reduction happens at minimum steps of
	 * 5% (default) of current frequency
	 */

	/* Get Idle Time */
	idle_ticks = UINT_MAX;
	for_each_cpu_mask(j, policy->cpus) {
		cputime64_t total_idle_ticks;
		unsigned int tmp_idle_ticks;
		struct cpu_dbs_info_s *j_dbs_info;

		j_dbs_info = &per_cpu(cpu_dbs_info, j);
		total_idle_ticks = get_cpu_idle_time(j);
		tmp_idle_ticks = (unsigned int) cputime64_sub(total_idle_ticks,
				j_dbs_info->prev_cpu_idle);
		j_dbs_info->prev_cpu_idle = total_idle_ticks;

		if (tmp_idle_ticks < idle_ticks)
			idle_ticks = tmp_idle_ticks;
	}
	load = (100 * (total_ticks - idle_ticks)) / total_ticks;

	/* Check for frequency increase */
	if (load > dbs_tuners_ins.up_threshold) {
		/* if we are already at full speed then break out early */
		if (policy->cur == policy->max)
			return;

		__cpufreq_driver_target(policy, policy->max,
			CPUFREQ_RELATION_H);
		return;
	}

	/* Check for frequency decrease */
	/* if we cannot reduce the frequency anymore, break out early */
	if (policy->cur == policy->min)
		return;

	/*
	 * The optimal frequency is the frequency that is the lowest that
	 * can support the current CPU usage without triggering the up
	 * policy. To be safe, we focus 10 points under the threshold.
	 */
	if (load < (dbs_tuners_ins.up_threshold - 10)) {
		unsigned int freq_next;
		freq_next = (policy->cur * load) /
			(dbs_tuners_ins.up_threshold - 10);

		__cpufreq_driver_target(policy, freq_next, CPUFREQ_RELATION_L);
	}
}

static void do_dbs_timer(void *data)
{
	unsigned int cpu = smp_processor_id();
	struct cpu_dbs_info_s *dbs_info = &per_cpu(cpu_dbs_info, cpu);

	dbs_check_cpu(dbs_info);
	queue_delayed_work_on(cpu, kondemand_wq, &dbs_info->work,
			usecs_to_jiffies(dbs_tuners_ins.sampling_rate));
}

static inline void dbs_timer_init(unsigned int cpu)
{
	struct cpu_dbs_info_s *dbs_info = &per_cpu(cpu_dbs_info, cpu);

	INIT_WORK(&dbs_info->work, do_dbs_timer, 0);
	queue_delayed_work_on(cpu, kondemand_wq, &dbs_info->work,
			usecs_to_jiffies(dbs_tuners_ins.sampling_rate));
	return;
}

static inline void dbs_timer_exit(unsigned int cpu)
{
	struct cpu_dbs_info_s *dbs_info = &per_cpu(cpu_dbs_info, cpu);

	cancel_rearming_delayed_workqueue(kondemand_wq, &dbs_info->work);
}

static int cpufreq_governor_dbs(struct cpufreq_policy *policy,
				   unsigned int event)
{
	unsigned int cpu = policy->cpu;
	struct cpu_dbs_info_s *this_dbs_info;
	unsigned int j;

	this_dbs_info = &per_cpu(cpu_dbs_info, cpu);

	switch (event) {
	case CPUFREQ_GOV_START:
		if ((!cpu_online(cpu)) || (!policy->cur))
			return -EINVAL;

		if (policy->cpuinfo.transition_latency >
				(TRANSITION_LATENCY_LIMIT * 1000)) {
			printk(KERN_WARNING "ondemand governor failed to load "
			       "due to too long transition latency\n");
			return -EINVAL;
		}
		if (this_dbs_info->enable) /* Already enabled */
			break;

		mutex_lock(&dbs_mutex);
		dbs_enable++;
		if (dbs_enable == 1) {
			kondemand_wq = create_workqueue("kondemand");
			if (!kondemand_wq) {
				printk(KERN_ERR "Creation of kondemand failed\n");
				dbs_enable--;
				mutex_unlock(&dbs_mutex);
				return -ENOSPC;
			}
		}
		for_each_cpu_mask(j, policy->cpus) {
			struct cpu_dbs_info_s *j_dbs_info;
			j_dbs_info = &per_cpu(cpu_dbs_info, j);
			j_dbs_info->cur_policy = policy;

			j_dbs_info->prev_cpu_idle = get_cpu_idle_time(j);
			j_dbs_info->prev_cpu_wall = get_jiffies_64();
		}
		this_dbs_info->enable = 1;
		sysfs_create_group(&policy->kobj, &dbs_attr_group);
		/*
		 * Start the timerschedule work, when this governor
		 * is used for first time
		 */
		if (dbs_enable == 1) {
			unsigned int latency;
			/* policy latency is in nS. Convert it to uS first */
			latency = policy->cpuinfo.transition_latency / 1000;
			if (latency == 0)
				latency = 1;

			def_sampling_rate = latency *
					DEF_SAMPLING_RATE_LATENCY_MULTIPLIER;

			if (def_sampling_rate < MIN_STAT_SAMPLING_RATE)
				def_sampling_rate = MIN_STAT_SAMPLING_RATE;

			dbs_tuners_ins.sampling_rate = def_sampling_rate;
		}
		dbs_timer_init(policy->cpu);

		mutex_unlock(&dbs_mutex);
		break;

	case CPUFREQ_GOV_STOP:
		mutex_lock(&dbs_mutex);
		dbs_timer_exit(policy->cpu);
		this_dbs_info->enable = 0;
		sysfs_remove_group(&policy->kobj, &dbs_attr_group);
		dbs_enable--;
		if (dbs_enable == 0)
			destroy_workqueue(kondemand_wq);

		mutex_unlock(&dbs_mutex);

		break;

	case CPUFREQ_GOV_LIMITS:
		lock_cpu_hotplug();
		mutex_lock(&dbs_mutex);
		if (policy->max < this_dbs_info->cur_policy->cur)
			__cpufreq_driver_target(this_dbs_info->cur_policy,
			                        policy->max,
			                        CPUFREQ_RELATION_H);
		else if (policy->min > this_dbs_info->cur_policy->cur)
			__cpufreq_driver_target(this_dbs_info->cur_policy,
			                        policy->min,
			                        CPUFREQ_RELATION_L);
		mutex_unlock(&dbs_mutex);
		unlock_cpu_hotplug();
		break;
	}
	return 0;
}

static struct cpufreq_governor cpufreq_gov_dbs = {
	.name = "ondemand",
	.governor = cpufreq_governor_dbs,
	.owner = THIS_MODULE,
};

static int __init cpufreq_gov_dbs_init(void)
{
	return cpufreq_register_governor(&cpufreq_gov_dbs);
}

static void __exit cpufreq_gov_dbs_exit(void)
{
	cpufreq_unregister_governor(&cpufreq_gov_dbs);
}


MODULE_AUTHOR("Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>");
MODULE_AUTHOR("Alexey Starikovskiy <alexey.y.starikovskiy@intel.com>");
MODULE_DESCRIPTION("'cpufreq_ondemand' - A dynamic cpufreq governor for "
                   "Low Latency Frequency Transition capable processors");
MODULE_LICENSE("GPL");

module_init(cpufreq_gov_dbs_init);
module_exit(cpufreq_gov_dbs_exit);
Commit	Line	Data
1da177e4 LT	1	/*
	2	* drivers/cpufreq/cpufreq_ondemand.c
	3	*
	4	* Copyright (C) 2001 Russell King
	5	* (C) 2003 Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>.
	6	* Jun Nakajima <jun.nakajima@intel.com>
	7	*
	8	* This program is free software; you can redistribute it and/or modify
	9	* it under the terms of the GNU General Public License version 2 as
	10	* published by the Free Software Foundation.
	11	*/
	12
	13	#include <linux/kernel.h>
	14	#include <linux/module.h>
1da177e4	15	#include <linux/init.h>
1da177e4	16	#include <linux/cpufreq.h>
138a0128	17	#include <linux/cpu.h>
1da177e4 LT	18	#include <linux/jiffies.h>
1da177e4 LT	19	#include <linux/kernel_stat.h>
3fc54d37	20	#include <linux/mutex.h>
1da177e4 LT	21
	22	/*
	23	* dbs is used in this file as a shortform for demandbased switching
	24	* It helps to keep variable names smaller, simpler
	25	*/
	26
	27	#define DEF_FREQUENCY_UP_THRESHOLD (80)
c29f1403	28	#define MIN_FREQUENCY_UP_THRESHOLD (11)
1da177e4 LT	29	#define MAX_FREQUENCY_UP_THRESHOLD (100)
1da177e4 LT	30
32ee8c3e DJ	31	/*
32ee8c3e DJ	32	* The polling frequency of this governor depends on the capability of
1da177e4	33	* the processor. Default polling frequency is 1000 times the transition
32ee8c3e DJ	34	* latency of the processor. The governor will work on any processor with
32ee8c3e DJ	35	* transition latency <= 10mS, using appropriate sampling
1da177e4 LT	36	* rate.
	37	* For CPUs with transition latency > 10mS (mostly drivers with CPUFREQ_ETERNAL)
	38	* this governor will not work.
	39	* All times here are in uS.
	40	*/
32ee8c3e	41	static unsigned int def_sampling_rate;
df8b59be DJ	42	#define MIN_SAMPLING_RATE_RATIO (2)
	43	/* for correct statistics, we need at least 10 ticks between each measure */
	44	#define MIN_STAT_SAMPLING_RATE (MIN_SAMPLING_RATE_RATIO * jiffies_to_usecs(10))
	45	#define MIN_SAMPLING_RATE (def_sampling_rate / MIN_SAMPLING_RATE_RATIO)
1da177e4 LT	46	#define MAX_SAMPLING_RATE (500 * def_sampling_rate)
1da177e4 LT	47	#define DEF_SAMPLING_RATE_LATENCY_MULTIPLIER (1000)
1da177e4	48	#define TRANSITION_LATENCY_LIMIT (10 * 1000)
1da177e4 LT	49
	50	static void do_dbs_timer(void *data);
	51
	52	struct cpu_dbs_info_s {
ccb2fe20 VP	53	cputime64_t prev_cpu_idle;
ccb2fe20 VP	54	cputime64_t prev_cpu_wall;
32ee8c3e	55	struct cpufreq_policy *cur_policy;
2f8a835c	56	struct work_struct work;
32ee8c3e	57	unsigned int enable;
1da177e4 LT	58	};
	59	static DEFINE_PER_CPU(struct cpu_dbs_info_s, cpu_dbs_info);
	60
	61	static unsigned int dbs_enable; /* number of CPUs using this policy */
	62
4ec223d0 VP	63	/*
	64	* DEADLOCK ALERT! There is a ordering requirement between cpu_hotplug
	65	* lock and dbs_mutex. cpu_hotplug lock should always be held before
	66	* dbs_mutex. If any function that can potentially take cpu_hotplug lock
	67	* (like __cpufreq_driver_target()) is being called with dbs_mutex taken, then
	68	* cpu_hotplug lock should be taken before that. Note that cpu_hotplug lock
	69	* is recursive for the same process. -Venki
	70	*/
ffac80e9	71	static DEFINE_MUTEX(dbs_mutex);
1da177e4	72
2f8a835c	73	static struct workqueue_struct *kondemand_wq;
6810b548	74
1da177e4	75	struct dbs_tuners {
32ee8c3e	76	unsigned int sampling_rate;
32ee8c3e DJ	77	unsigned int up_threshold;
32ee8c3e DJ	78	unsigned int ignore_nice;
1da177e4 LT	79	};
	80
	81	static struct dbs_tuners dbs_tuners_ins = {
32ee8c3e	82	.up_threshold = DEF_FREQUENCY_UP_THRESHOLD,
9cbad61b	83	.ignore_nice = 0,
1da177e4 LT	84	};
1da177e4 LT	85
ccb2fe20	86	static inline cputime64_t get_cpu_idle_time(unsigned int cpu)
dac1c1a5	87	{
ccb2fe20 VP	88	cputime64_t retval;
	89
	90	retval = cputime64_add(kstat_cpu(cpu).cpustat.idle,
	91	kstat_cpu(cpu).cpustat.iowait);
	92
	93	if (dbs_tuners_ins.ignore_nice)
	94	retval = cputime64_add(retval, kstat_cpu(cpu).cpustat.nice);
	95
	96	return retval;
dac1c1a5 DJ	97	}
dac1c1a5 DJ	98
1da177e4 LT	99	/************************ sysfs interface **********************/
	100	static ssize_t show_sampling_rate_max(struct cpufreq_policy policy, char buf)
	101	{
	102	return sprintf (buf, "%u\n", MAX_SAMPLING_RATE);
	103	}
	104
	105	static ssize_t show_sampling_rate_min(struct cpufreq_policy policy, char buf)
	106	{
	107	return sprintf (buf, "%u\n", MIN_SAMPLING_RATE);
	108	}
	109
32ee8c3e DJ	110	#define define_one_ro(_name) \
32ee8c3e DJ	111	static struct freq_attr _name = \
1da177e4 LT	112	__ATTR(_name, 0444, show_##_name, NULL)
	113
	114	define_one_ro(sampling_rate_max);
	115	define_one_ro(sampling_rate_min);
	116
	117	/* cpufreq_ondemand Governor Tunables */
	118	#define show_one(file_name, object) \
	119	static ssize_t show_##file_name \
	120	(struct cpufreq_policy unused, char buf) \
	121	{ \
	122	return sprintf(buf, "%u\n", dbs_tuners_ins.object); \
	123	}
	124	show_one(sampling_rate, sampling_rate);
1da177e4	125	show_one(up_threshold, up_threshold);
001893cd	126	show_one(ignore_nice_load, ignore_nice);
1da177e4	127
32ee8c3e	128	static ssize_t store_sampling_rate(struct cpufreq_policy *unused,
1da177e4 LT	129	const char *buf, size_t count)
	130	{
	131	unsigned int input;
	132	int ret;
ffac80e9	133	ret = sscanf(buf, "%u", &input);
1da177e4	134
3fc54d37	135	mutex_lock(&dbs_mutex);
1da177e4	136	if (ret != 1 \|\| input > MAX_SAMPLING_RATE \|\| input < MIN_SAMPLING_RATE) {
3fc54d37	137	mutex_unlock(&dbs_mutex);
1da177e4 LT	138	return -EINVAL;
	139	}
	140
	141	dbs_tuners_ins.sampling_rate = input;
3fc54d37	142	mutex_unlock(&dbs_mutex);
1da177e4 LT	143
	144	return count;
	145	}
	146
32ee8c3e	147	static ssize_t store_up_threshold(struct cpufreq_policy *unused,
1da177e4 LT	148	const char *buf, size_t count)
	149	{
	150	unsigned int input;
	151	int ret;
ffac80e9	152	ret = sscanf(buf, "%u", &input);
1da177e4	153
3fc54d37	154	mutex_lock(&dbs_mutex);
32ee8c3e	155	if (ret != 1 \|\| input > MAX_FREQUENCY_UP_THRESHOLD \|\|
c29f1403	156	input < MIN_FREQUENCY_UP_THRESHOLD) {
3fc54d37	157	mutex_unlock(&dbs_mutex);
1da177e4 LT	158	return -EINVAL;
	159	}
	160
	161	dbs_tuners_ins.up_threshold = input;
3fc54d37	162	mutex_unlock(&dbs_mutex);
1da177e4 LT	163
	164	return count;
	165	}
	166
001893cd	167	static ssize_t store_ignore_nice_load(struct cpufreq_policy *policy,
3d5ee9e5 DJ	168	const char *buf, size_t count)
	169	{
	170	unsigned int input;
	171	int ret;
	172
	173	unsigned int j;
32ee8c3e	174
ffac80e9	175	ret = sscanf(buf, "%u", &input);
3d5ee9e5 DJ	176	if ( ret != 1 )
	177	return -EINVAL;
	178
	179	if ( input > 1 )
	180	input = 1;
32ee8c3e	181
3fc54d37	182	mutex_lock(&dbs_mutex);
3d5ee9e5	183	if ( input == dbs_tuners_ins.ignore_nice ) { /* nothing to do */
3fc54d37	184	mutex_unlock(&dbs_mutex);
3d5ee9e5 DJ	185	return count;
	186	}
	187	dbs_tuners_ins.ignore_nice = input;
	188
ccb2fe20	189	/* we need to re-evaluate prev_cpu_idle */
dac1c1a5	190	for_each_online_cpu(j) {
ccb2fe20 VP	191	struct cpu_dbs_info_s *dbs_info;
	192	dbs_info = &per_cpu(cpu_dbs_info, j);
	193	dbs_info->prev_cpu_idle = get_cpu_idle_time(j);
	194	dbs_info->prev_cpu_wall = get_jiffies_64();
3d5ee9e5	195	}
3fc54d37	196	mutex_unlock(&dbs_mutex);
3d5ee9e5 DJ	197
	198	return count;
	199	}
	200
1da177e4 LT	201	#define define_one_rw(_name) \
	202	static struct freq_attr _name = \
	203	__ATTR(_name, 0644, show_##_name, store_##_name)
	204
	205	define_one_rw(sampling_rate);
1da177e4	206	define_one_rw(up_threshold);
001893cd	207	define_one_rw(ignore_nice_load);
1da177e4 LT	208
	209	static struct attribute * dbs_attributes[] = {
	210	&sampling_rate_max.attr,
	211	&sampling_rate_min.attr,
	212	&sampling_rate.attr,
1da177e4	213	&up_threshold.attr,
001893cd	214	&ignore_nice_load.attr,
1da177e4 LT	215	NULL
	216	};
	217
	218	static struct attribute_group dbs_attr_group = {
	219	.attrs = dbs_attributes,
	220	.name = "ondemand",
	221	};
	222
	223	/************************ sysfs end **********************/
	224
2f8a835c	225	static void dbs_check_cpu(struct cpu_dbs_info_s *this_dbs_info)
1da177e4	226	{
ccb2fe20 VP	227	unsigned int idle_ticks, total_ticks;
ccb2fe20 VP	228	unsigned int load;
ccb2fe20	229	cputime64_t cur_jiffies;
1da177e4 LT	230
	231	struct cpufreq_policy *policy;
	232	unsigned int j;
	233
1da177e4 LT	234	if (!this_dbs_info->enable)
	235	return;
	236
	237	policy = this_dbs_info->cur_policy;
ccb2fe20 VP	238	cur_jiffies = jiffies64_to_cputime64(get_jiffies_64());
	239	total_ticks = (unsigned int) cputime64_sub(cur_jiffies,
	240	this_dbs_info->prev_cpu_wall);
	241	this_dbs_info->prev_cpu_wall = cur_jiffies;
32ee8c3e	242	/*
c29f1403 DJ	243	* Every sampling_rate, we check, if current idle time is less
c29f1403 DJ	244	* than 20% (default), then we try to increase frequency
ccb2fe20	245	* Every sampling_rate, we look for a the lowest
c29f1403 DJ	246	* frequency which can sustain the load while keeping idle time over
c29f1403 DJ	247	* 30%. If such a frequency exist, we try to decrease to this frequency.
1da177e4	248	*
32ee8c3e DJ	249	* Any frequency increase takes it to the maximum frequency.
	250	* Frequency reduction happens at minimum steps of
	251	* 5% (default) of current frequency
1da177e4 LT	252	*/
1da177e4 LT	253
ccb2fe20	254	/* Get Idle Time */
9c7d269b	255	idle_ticks = UINT_MAX;
1da177e4	256	for_each_cpu_mask(j, policy->cpus) {
ccb2fe20 VP	257	cputime64_t total_idle_ticks;
ccb2fe20 VP	258	unsigned int tmp_idle_ticks;
1da177e4 LT	259	struct cpu_dbs_info_s *j_dbs_info;
1da177e4 LT	260
1da177e4	261	j_dbs_info = &per_cpu(cpu_dbs_info, j);
dac1c1a5	262	total_idle_ticks = get_cpu_idle_time(j);
ccb2fe20 VP	263	tmp_idle_ticks = (unsigned int) cputime64_sub(total_idle_ticks,
	264	j_dbs_info->prev_cpu_idle);
	265	j_dbs_info->prev_cpu_idle = total_idle_ticks;
1da177e4 LT	266
	267	if (tmp_idle_ticks < idle_ticks)
	268	idle_ticks = tmp_idle_ticks;
	269	}
ccb2fe20	270	load = (100 * (total_ticks - idle_ticks)) / total_ticks;
1da177e4	271
ccb2fe20 VP	272	/* Check for frequency increase */
ccb2fe20 VP	273	if (load > dbs_tuners_ins.up_threshold) {
c11420a6 DJ	274	/* if we are already at full speed then break out early */
	275	if (policy->cur == policy->max)
	276	return;
32ee8c3e DJ	277
32ee8c3e DJ	278	__cpufreq_driver_target(policy, policy->max,
1da177e4	279	CPUFREQ_RELATION_H);
1da177e4 LT	280	return;
	281	}
	282
	283	/* Check for frequency decrease */
c29f1403 DJ	284	/* if we cannot reduce the frequency anymore, break out early */
	285	if (policy->cur == policy->min)
	286	return;
1da177e4	287
c29f1403 DJ	288	/*
	289	* The optimal frequency is the frequency that is the lowest that
	290	* can support the current CPU usage without triggering the up
	291	* policy. To be safe, we focus 10 points under the threshold.
	292	*/
ccb2fe20 VP	293	if (load < (dbs_tuners_ins.up_threshold - 10)) {
	294	unsigned int freq_next;
	295	freq_next = (policy->cur * load) /
c29f1403	296	(dbs_tuners_ins.up_threshold - 10);
1da177e4	297
c29f1403	298	__cpufreq_driver_target(policy, freq_next, CPUFREQ_RELATION_L);
ccb2fe20	299	}
1da177e4 LT	300	}
	301
	302	static void do_dbs_timer(void *data)
32ee8c3e	303	{
2f8a835c VP	304	unsigned int cpu = smp_processor_id();
	305	struct cpu_dbs_info_s *dbs_info = &per_cpu(cpu_dbs_info, cpu);
	306
	307	dbs_check_cpu(dbs_info);
	308	queue_delayed_work_on(cpu, kondemand_wq, &dbs_info->work,
	309	usecs_to_jiffies(dbs_tuners_ins.sampling_rate));
32ee8c3e	310	}
1da177e4	311
2f8a835c	312	static inline void dbs_timer_init(unsigned int cpu)
1da177e4	313	{
2f8a835c VP	314	struct cpu_dbs_info_s *dbs_info = &per_cpu(cpu_dbs_info, cpu);
	315
	316	INIT_WORK(&dbs_info->work, do_dbs_timer, 0);
	317	queue_delayed_work_on(cpu, kondemand_wq, &dbs_info->work,
	318	usecs_to_jiffies(dbs_tuners_ins.sampling_rate));
1da177e4 LT	319	return;
	320	}
	321
2f8a835c	322	static inline void dbs_timer_exit(unsigned int cpu)
1da177e4	323	{
2f8a835c VP	324	struct cpu_dbs_info_s *dbs_info = &per_cpu(cpu_dbs_info, cpu);
	325
	326	cancel_rearming_delayed_workqueue(kondemand_wq, &dbs_info->work);
1da177e4 LT	327	}
	328
	329	static int cpufreq_governor_dbs(struct cpufreq_policy *policy,
	330	unsigned int event)
	331	{
	332	unsigned int cpu = policy->cpu;
	333	struct cpu_dbs_info_s *this_dbs_info;
	334	unsigned int j;
	335
	336	this_dbs_info = &per_cpu(cpu_dbs_info, cpu);
	337
	338	switch (event) {
	339	case CPUFREQ_GOV_START:
ffac80e9	340	if ((!cpu_online(cpu)) \|\| (!policy->cur))
1da177e4 LT	341	return -EINVAL;
	342
	343	if (policy->cpuinfo.transition_latency >
ff8c288d EP	344	(TRANSITION_LATENCY_LIMIT * 1000)) {
	345	printk(KERN_WARNING "ondemand governor failed to load "
	346	"due to too long transition latency\n");
1da177e4	347	return -EINVAL;
ff8c288d	348	}
1da177e4 LT	349	if (this_dbs_info->enable) /* Already enabled */
1da177e4 LT	350	break;
32ee8c3e	351
3fc54d37	352	mutex_lock(&dbs_mutex);
2f8a835c VP	353	dbs_enable++;
	354	if (dbs_enable == 1) {
	355	kondemand_wq = create_workqueue("kondemand");
	356	if (!kondemand_wq) {
	357	printk(KERN_ERR "Creation of kondemand failed\n");
	358	dbs_enable--;
	359	mutex_unlock(&dbs_mutex);
	360	return -ENOSPC;
	361	}
	362	}
1da177e4 LT	363	for_each_cpu_mask(j, policy->cpus) {
	364	struct cpu_dbs_info_s *j_dbs_info;
	365	j_dbs_info = &per_cpu(cpu_dbs_info, j);
	366	j_dbs_info->cur_policy = policy;
32ee8c3e	367
ccb2fe20 VP	368	j_dbs_info->prev_cpu_idle = get_cpu_idle_time(j);
ccb2fe20 VP	369	j_dbs_info->prev_cpu_wall = get_jiffies_64();
1da177e4 LT	370	}
	371	this_dbs_info->enable = 1;
	372	sysfs_create_group(&policy->kobj, &dbs_attr_group);
1da177e4 LT	373	/*
	374	* Start the timerschedule work, when this governor
	375	* is used for first time
	376	*/
	377	if (dbs_enable == 1) {
	378	unsigned int latency;
	379	/* policy latency is in nS. Convert it to uS first */
df8b59be DJ	380	latency = policy->cpuinfo.transition_latency / 1000;
	381	if (latency == 0)
	382	latency = 1;
1da177e4	383
df8b59be	384	def_sampling_rate = latency *
1da177e4	385	DEF_SAMPLING_RATE_LATENCY_MULTIPLIER;
df8b59be DJ	386
	387	if (def_sampling_rate < MIN_STAT_SAMPLING_RATE)
	388	def_sampling_rate = MIN_STAT_SAMPLING_RATE;
	389
1da177e4	390	dbs_tuners_ins.sampling_rate = def_sampling_rate;
1da177e4	391	}
2f8a835c	392	dbs_timer_init(policy->cpu);
32ee8c3e	393
3fc54d37	394	mutex_unlock(&dbs_mutex);
1da177e4 LT	395	break;
	396
	397	case CPUFREQ_GOV_STOP:
3fc54d37	398	mutex_lock(&dbs_mutex);
2f8a835c	399	dbs_timer_exit(policy->cpu);
1da177e4 LT	400	this_dbs_info->enable = 0;
	401	sysfs_remove_group(&policy->kobj, &dbs_attr_group);
	402	dbs_enable--;
32ee8c3e	403	if (dbs_enable == 0)
2f8a835c	404	destroy_workqueue(kondemand_wq);
32ee8c3e	405
3fc54d37	406	mutex_unlock(&dbs_mutex);
1da177e4 LT	407
	408	break;
	409
	410	case CPUFREQ_GOV_LIMITS:
4ec223d0	411	lock_cpu_hotplug();
3fc54d37	412	mutex_lock(&dbs_mutex);
1da177e4	413	if (policy->max < this_dbs_info->cur_policy->cur)
ffac80e9 VP	414	__cpufreq_driver_target(this_dbs_info->cur_policy,
	415	policy->max,
	416	CPUFREQ_RELATION_H);
1da177e4	417	else if (policy->min > this_dbs_info->cur_policy->cur)
ffac80e9 VP	418	__cpufreq_driver_target(this_dbs_info->cur_policy,
	419	policy->min,
	420	CPUFREQ_RELATION_L);
3fc54d37	421	mutex_unlock(&dbs_mutex);
4ec223d0	422	unlock_cpu_hotplug();
1da177e4 LT	423	break;
	424	}
	425	return 0;
	426	}
	427
7f335d4e	428	static struct cpufreq_governor cpufreq_gov_dbs = {
ffac80e9 VP	429	.name = "ondemand",
	430	.governor = cpufreq_governor_dbs,
	431	.owner = THIS_MODULE,
1da177e4	432	};
1da177e4 LT	433
	434	static int __init cpufreq_gov_dbs_init(void)
	435	{
	436	return cpufreq_register_governor(&cpufreq_gov_dbs);
	437	}
	438
	439	static void __exit cpufreq_gov_dbs_exit(void)
	440	{
1da177e4 LT	441	cpufreq_unregister_governor(&cpufreq_gov_dbs);
	442	}
	443
	444
ffac80e9 VP	445	MODULE_AUTHOR("Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>");
	446	MODULE_AUTHOR("Alexey Starikovskiy <alexey.y.starikovskiy@intel.com>");
	447	MODULE_DESCRIPTION("'cpufreq_ondemand' - A dynamic cpufreq governor for "
	448	"Low Latency Frequency Transition capable processors");
	449	MODULE_LICENSE("GPL");
1da177e4 LT	450
	451	module_init(cpufreq_gov_dbs_init);
	452	module_exit(cpufreq_gov_dbs_exit);