[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/smp.h>
#include <linux/init.h>
#include <linux/interrupt.h>
#include <linux/ctype.h>
#include <linux/cpufreq.h>
#include <linux/sysctl.h>
#include <linux/types.h>
#include <linux/fs.h>
#include <linux/sysfs.h>
#include <linux/sched.h>
#include <linux/kmod.h>
#include <linux/workqueue.h>
#include <linux/jiffies.h>
#include <linux/kernel_stat.h>
#include <linux/percpu.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 DEF_SAMPLING_DOWN_FACTOR		(1)
#define MAX_SAMPLING_DOWN_FACTOR		(10)
#define TRANSITION_LATENCY_LIMIT		(10 * 1000)

static void do_dbs_timer(void *data);

struct cpu_dbs_info_s {
	struct cpufreq_policy *cur_policy;
	unsigned int prev_cpu_idle_up;
	unsigned int prev_cpu_idle_down;
	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 */

static DEFINE_MUTEX (dbs_mutex);
static DECLARE_WORK	(dbs_work, do_dbs_timer, NULL);

static struct workqueue_struct *dbs_workq;

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

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

static inline unsigned int get_cpu_idle_time(unsigned int cpu)
{
	return	kstat_cpu(cpu).cpustat.idle +
		kstat_cpu(cpu).cpustat.iowait +
		( dbs_tuners_ins.ignore_nice ?
		  kstat_cpu(cpu).cpustat.nice :
		  0);
}

/************************** 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(sampling_down_factor, sampling_down_factor);
show_one(up_threshold, up_threshold);
show_one(ignore_nice_load, ignore_nice);

static ssize_t store_sampling_down_factor(struct cpufreq_policy *unused,
		const char *buf, size_t count)
{
	unsigned int input;
	int ret;
	ret = sscanf (buf, "%u", &input);
	if (ret != 1 )
		return -EINVAL;

	if (input > MAX_SAMPLING_DOWN_FACTOR || input < 1)
		return -EINVAL;

	mutex_lock(&dbs_mutex);
	dbs_tuners_ins.sampling_down_factor = input;
	mutex_unlock(&dbs_mutex);

	return count;
}

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_up and prev_cpu_idle_down */
	for_each_online_cpu(j) {
		struct cpu_dbs_info_s *j_dbs_info;
		j_dbs_info = &per_cpu(cpu_dbs_info, j);
		j_dbs_info->prev_cpu_idle_up = get_cpu_idle_time(j);
		j_dbs_info->prev_cpu_idle_down = j_dbs_info->prev_cpu_idle_up;
	}
	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(sampling_down_factor);
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,
	&sampling_down_factor.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(int cpu)
{
	unsigned int idle_ticks, up_idle_ticks, total_ticks;
	unsigned int freq_next;
	unsigned int freq_down_sampling_rate;
	static int down_skip[NR_CPUS];
	struct cpu_dbs_info_s *this_dbs_info;

	struct cpufreq_policy *policy;
	unsigned int j;

	this_dbs_info = &per_cpu(cpu_dbs_info, cpu);
	if (!this_dbs_info->enable)
		return;

	policy = this_dbs_info->cur_policy;
	/*
	 * Every sampling_rate, we check, if current idle time is less
	 * than 20% (default), then we try to increase frequency
	 * Every sampling_rate*sampling_down_factor, 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
	 */

	/* Check for frequency increase */
	idle_ticks = UINT_MAX;
	for_each_cpu_mask(j, policy->cpus) {
		unsigned int tmp_idle_ticks, total_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 = total_idle_ticks -
			j_dbs_info->prev_cpu_idle_up;
		j_dbs_info->prev_cpu_idle_up = total_idle_ticks;

		if (tmp_idle_ticks < idle_ticks)
			idle_ticks = tmp_idle_ticks;
	}

	/* Scale idle ticks by 100 and compare with up and down ticks */
	idle_ticks *= 100;
	up_idle_ticks = (100 - dbs_tuners_ins.up_threshold) *
			usecs_to_jiffies(dbs_tuners_ins.sampling_rate);

	if (idle_ticks < up_idle_ticks) {
		down_skip[cpu] = 0;
		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->prev_cpu_idle_down =
					j_dbs_info->prev_cpu_idle_up;
		}
		/* 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 */
	down_skip[cpu]++;
	if (down_skip[cpu] < dbs_tuners_ins.sampling_down_factor)
		return;

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

		j_dbs_info = &per_cpu(cpu_dbs_info, j);
		/* Check for frequency decrease */
		total_idle_ticks = j_dbs_info->prev_cpu_idle_up;
		tmp_idle_ticks = total_idle_ticks -
			j_dbs_info->prev_cpu_idle_down;
		j_dbs_info->prev_cpu_idle_down = total_idle_ticks;

		if (tmp_idle_ticks < idle_ticks)
			idle_ticks = tmp_idle_ticks;
	}

	down_skip[cpu] = 0;
	/* if we cannot reduce the frequency anymore, break out early */
	if (policy->cur == policy->min)
		return;

	/* Compute how many ticks there are between two measurements */
	freq_down_sampling_rate = dbs_tuners_ins.sampling_rate *
		dbs_tuners_ins.sampling_down_factor;
	total_ticks = usecs_to_jiffies(freq_down_sampling_rate);

	/*
	 * 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.
	 */
	freq_next = ((total_ticks - idle_ticks) * 100) / total_ticks;
	freq_next = (freq_next * policy->cur) /
			(dbs_tuners_ins.up_threshold - 10);

	if (freq_next < policy->min)
		freq_next = policy->min;

	if (freq_next <= ((policy->cur * 95) / 100))
		__cpufreq_driver_target(policy, freq_next, CPUFREQ_RELATION_L);
}

static void do_dbs_timer(void *data)
{
	int i;
	mutex_lock(&dbs_mutex);
	for_each_online_cpu(i)
		dbs_check_cpu(i);
	queue_delayed_work(dbs_workq, &dbs_work,
			   usecs_to_jiffies(dbs_tuners_ins.sampling_rate));
	mutex_unlock(&dbs_mutex);
}

static inline void dbs_timer_init(void)
{
	INIT_WORK(&dbs_work, do_dbs_timer, NULL);
	if (!dbs_workq)
		dbs_workq = create_singlethread_workqueue("ondemand");
	if (!dbs_workq) {
		printk(KERN_ERR "ondemand: Cannot initialize kernel thread\n");
		return;
	}
	queue_delayed_work(dbs_workq, &dbs_work,
			   usecs_to_jiffies(dbs_tuners_ins.sampling_rate));
	return;
}

static inline void dbs_timer_exit(void)
{
	if (dbs_workq)
		cancel_rearming_delayed_workqueue(dbs_workq, &dbs_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);
		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_up = get_cpu_idle_time(j);
			j_dbs_info->prev_cpu_idle_down
				= j_dbs_info->prev_cpu_idle_up;
		}
		this_dbs_info->enable = 1;
		sysfs_create_group(&policy->kobj, &dbs_attr_group);
		dbs_enable++;
		/*
		 * 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();
		}

		mutex_unlock(&dbs_mutex);
		break;

	case CPUFREQ_GOV_STOP:
		mutex_lock(&dbs_mutex);
		this_dbs_info->enable = 0;
		sysfs_remove_group(&policy->kobj, &dbs_attr_group);
		dbs_enable--;
		/*
		 * Stop the timerschedule work, when this governor
		 * is used for first time
		 */
		if (dbs_enable == 0)
			dbs_timer_exit();

		mutex_unlock(&dbs_mutex);

		break;

	case CPUFREQ_GOV_LIMITS:
		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);
		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)
{
	/* Make sure that the scheduled work is indeed not running.
	   Assumes the timer has been cancelled first. */
	if (dbs_workq) {
		flush_workqueue(dbs_workq);
		destroy_workqueue(dbs_workq);
	}

	cpufreq_unregister_governor(&cpufreq_gov_dbs);
}


MODULE_AUTHOR ("Venkatesh Pallipadi <venkatesh.pallipadi@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>
	15	#include <linux/smp.h>
	16	#include <linux/init.h>
	17	#include <linux/interrupt.h>
	18	#include <linux/ctype.h>
	19	#include <linux/cpufreq.h>
	20	#include <linux/sysctl.h>
	21	#include <linux/types.h>
	22	#include <linux/fs.h>
	23	#include <linux/sysfs.h>
	24	#include <linux/sched.h>
	25	#include <linux/kmod.h>
	26	#include <linux/workqueue.h>
	27	#include <linux/jiffies.h>
	28	#include <linux/kernel_stat.h>
	29	#include <linux/percpu.h>
3fc54d37	30	#include <linux/mutex.h>
1da177e4 LT	31
	32	/*
	33	* dbs is used in this file as a shortform for demandbased switching
	34	* It helps to keep variable names smaller, simpler
	35	*/
	36
	37	#define DEF_FREQUENCY_UP_THRESHOLD (80)
c29f1403	38	#define MIN_FREQUENCY_UP_THRESHOLD (11)
1da177e4 LT	39	#define MAX_FREQUENCY_UP_THRESHOLD (100)
1da177e4 LT	40
32ee8c3e DJ	41	/*
32ee8c3e DJ	42	* The polling frequency of this governor depends on the capability of
1da177e4	43	* the processor. Default polling frequency is 1000 times the transition
32ee8c3e DJ	44	* latency of the processor. The governor will work on any processor with
32ee8c3e DJ	45	* transition latency <= 10mS, using appropriate sampling
1da177e4 LT	46	* rate.
	47	* For CPUs with transition latency > 10mS (mostly drivers with CPUFREQ_ETERNAL)
	48	* this governor will not work.
	49	* All times here are in uS.
	50	*/
32ee8c3e	51	static unsigned int def_sampling_rate;
df8b59be DJ	52	#define MIN_SAMPLING_RATE_RATIO (2)
	53	/* for correct statistics, we need at least 10 ticks between each measure */
	54	#define MIN_STAT_SAMPLING_RATE (MIN_SAMPLING_RATE_RATIO * jiffies_to_usecs(10))
	55	#define MIN_SAMPLING_RATE (def_sampling_rate / MIN_SAMPLING_RATE_RATIO)
1da177e4 LT	56	#define MAX_SAMPLING_RATE (500 * def_sampling_rate)
1da177e4 LT	57	#define DEF_SAMPLING_RATE_LATENCY_MULTIPLIER (1000)
e131832c DJ	58	#define DEF_SAMPLING_DOWN_FACTOR (1)
e131832c DJ	59	#define MAX_SAMPLING_DOWN_FACTOR (10)
1da177e4	60	#define TRANSITION_LATENCY_LIMIT (10 * 1000)
1da177e4 LT	61
	62	static void do_dbs_timer(void *data);
	63
	64	struct cpu_dbs_info_s {
32ee8c3e DJ	65	struct cpufreq_policy *cur_policy;
	66	unsigned int prev_cpu_idle_up;
	67	unsigned int prev_cpu_idle_down;
	68	unsigned int enable;
1da177e4 LT	69	};
	70	static DEFINE_PER_CPU(struct cpu_dbs_info_s, cpu_dbs_info);
	71
	72	static unsigned int dbs_enable; /* number of CPUs using this policy */
	73
32ee8c3e	74	static DEFINE_MUTEX (dbs_mutex);
1da177e4 LT	75	static DECLARE_WORK (dbs_work, do_dbs_timer, NULL);
1da177e4 LT	76
6810b548 AK	77	static struct workqueue_struct *dbs_workq;
6810b548 AK	78
1da177e4	79	struct dbs_tuners {
32ee8c3e DJ	80	unsigned int sampling_rate;
	81	unsigned int sampling_down_factor;
	82	unsigned int up_threshold;
	83	unsigned int ignore_nice;
1da177e4 LT	84	};
	85
	86	static struct dbs_tuners dbs_tuners_ins = {
32ee8c3e DJ	87	.up_threshold = DEF_FREQUENCY_UP_THRESHOLD,
32ee8c3e DJ	88	.sampling_down_factor = DEF_SAMPLING_DOWN_FACTOR,
9cbad61b	89	.ignore_nice = 0,
1da177e4 LT	90	};
1da177e4 LT	91
dac1c1a5 DJ	92	static inline unsigned int get_cpu_idle_time(unsigned int cpu)
	93	{
	94	return kstat_cpu(cpu).cpustat.idle +
	95	kstat_cpu(cpu).cpustat.iowait +
001893cd	96	( dbs_tuners_ins.ignore_nice ?
dac1c1a5 DJ	97	kstat_cpu(cpu).cpustat.nice :
	98	0);
	99	}
	100
1da177e4 LT	101	/************************ sysfs interface **********************/
	102	static ssize_t show_sampling_rate_max(struct cpufreq_policy policy, char buf)
	103	{
	104	return sprintf (buf, "%u\n", MAX_SAMPLING_RATE);
	105	}
	106
	107	static ssize_t show_sampling_rate_min(struct cpufreq_policy policy, char buf)
	108	{
	109	return sprintf (buf, "%u\n", MIN_SAMPLING_RATE);
	110	}
	111
32ee8c3e DJ	112	#define define_one_ro(_name) \
32ee8c3e DJ	113	static struct freq_attr _name = \
1da177e4 LT	114	__ATTR(_name, 0444, show_##_name, NULL)
	115
	116	define_one_ro(sampling_rate_max);
	117	define_one_ro(sampling_rate_min);
	118
	119	/* cpufreq_ondemand Governor Tunables */
	120	#define show_one(file_name, object) \
	121	static ssize_t show_##file_name \
	122	(struct cpufreq_policy unused, char buf) \
	123	{ \
	124	return sprintf(buf, "%u\n", dbs_tuners_ins.object); \
	125	}
	126	show_one(sampling_rate, sampling_rate);
	127	show_one(sampling_down_factor, sampling_down_factor);
	128	show_one(up_threshold, up_threshold);
001893cd	129	show_one(ignore_nice_load, ignore_nice);
1da177e4	130
32ee8c3e	131	static ssize_t store_sampling_down_factor(struct cpufreq_policy *unused,
1da177e4 LT	132	const char *buf, size_t count)
	133	{
	134	unsigned int input;
	135	int ret;
	136	ret = sscanf (buf, "%u", &input);
	137	if (ret != 1 )
	138	return -EINVAL;
	139
e131832c DJ	140	if (input > MAX_SAMPLING_DOWN_FACTOR \|\| input < 1)
	141	return -EINVAL;
	142
3fc54d37	143	mutex_lock(&dbs_mutex);
1da177e4	144	dbs_tuners_ins.sampling_down_factor = input;
3fc54d37	145	mutex_unlock(&dbs_mutex);
1da177e4 LT	146
	147	return count;
	148	}
	149
32ee8c3e	150	static ssize_t store_sampling_rate(struct cpufreq_policy *unused,
1da177e4 LT	151	const char *buf, size_t count)
	152	{
	153	unsigned int input;
	154	int ret;
	155	ret = sscanf (buf, "%u", &input);
	156
3fc54d37	157	mutex_lock(&dbs_mutex);
1da177e4	158	if (ret != 1 \|\| input > MAX_SAMPLING_RATE \|\| input < MIN_SAMPLING_RATE) {
3fc54d37	159	mutex_unlock(&dbs_mutex);
1da177e4 LT	160	return -EINVAL;
	161	}
	162
	163	dbs_tuners_ins.sampling_rate = input;
3fc54d37	164	mutex_unlock(&dbs_mutex);
1da177e4 LT	165
	166	return count;
	167	}
	168
32ee8c3e	169	static ssize_t store_up_threshold(struct cpufreq_policy *unused,
1da177e4 LT	170	const char *buf, size_t count)
	171	{
	172	unsigned int input;
	173	int ret;
	174	ret = sscanf (buf, "%u", &input);
	175
3fc54d37	176	mutex_lock(&dbs_mutex);
32ee8c3e	177	if (ret != 1 \|\| input > MAX_FREQUENCY_UP_THRESHOLD \|\|
c29f1403	178	input < MIN_FREQUENCY_UP_THRESHOLD) {
3fc54d37	179	mutex_unlock(&dbs_mutex);
1da177e4 LT	180	return -EINVAL;
	181	}
	182
	183	dbs_tuners_ins.up_threshold = input;
3fc54d37	184	mutex_unlock(&dbs_mutex);
1da177e4 LT	185
	186	return count;
	187	}
	188
001893cd	189	static ssize_t store_ignore_nice_load(struct cpufreq_policy *policy,
3d5ee9e5 DJ	190	const char *buf, size_t count)
	191	{
	192	unsigned int input;
	193	int ret;
	194
	195	unsigned int j;
32ee8c3e	196
3d5ee9e5 DJ	197	ret = sscanf (buf, "%u", &input);
	198	if ( ret != 1 )
	199	return -EINVAL;
	200
	201	if ( input > 1 )
	202	input = 1;
32ee8c3e	203
3fc54d37	204	mutex_lock(&dbs_mutex);
3d5ee9e5	205	if ( input == dbs_tuners_ins.ignore_nice ) { /* nothing to do */
3fc54d37	206	mutex_unlock(&dbs_mutex);
3d5ee9e5 DJ	207	return count;
	208	}
	209	dbs_tuners_ins.ignore_nice = input;
	210
	211	/* we need to re-evaluate prev_cpu_idle_up and prev_cpu_idle_down */
dac1c1a5	212	for_each_online_cpu(j) {
3d5ee9e5 DJ	213	struct cpu_dbs_info_s *j_dbs_info;
3d5ee9e5 DJ	214	j_dbs_info = &per_cpu(cpu_dbs_info, j);
dac1c1a5	215	j_dbs_info->prev_cpu_idle_up = get_cpu_idle_time(j);
3d5ee9e5 DJ	216	j_dbs_info->prev_cpu_idle_down = j_dbs_info->prev_cpu_idle_up;
3d5ee9e5 DJ	217	}
3fc54d37	218	mutex_unlock(&dbs_mutex);
3d5ee9e5 DJ	219
	220	return count;
	221	}
	222
1da177e4 LT	223	#define define_one_rw(_name) \
	224	static struct freq_attr _name = \
	225	__ATTR(_name, 0644, show_##_name, store_##_name)
	226
	227	define_one_rw(sampling_rate);
	228	define_one_rw(sampling_down_factor);
	229	define_one_rw(up_threshold);
001893cd	230	define_one_rw(ignore_nice_load);
1da177e4 LT	231
	232	static struct attribute * dbs_attributes[] = {
	233	&sampling_rate_max.attr,
	234	&sampling_rate_min.attr,
	235	&sampling_rate.attr,
	236	&sampling_down_factor.attr,
	237	&up_threshold.attr,
001893cd	238	&ignore_nice_load.attr,
1da177e4 LT	239	NULL
	240	};
	241
	242	static struct attribute_group dbs_attr_group = {
	243	.attrs = dbs_attributes,
	244	.name = "ondemand",
	245	};
	246
	247	/************************ sysfs end **********************/
	248
	249	static void dbs_check_cpu(int cpu)
	250	{
c29f1403 DJ	251	unsigned int idle_ticks, up_idle_ticks, total_ticks;
c29f1403 DJ	252	unsigned int freq_next;
1da177e4 LT	253	unsigned int freq_down_sampling_rate;
	254	static int down_skip[NR_CPUS];
	255	struct cpu_dbs_info_s *this_dbs_info;
	256
	257	struct cpufreq_policy *policy;
	258	unsigned int j;
	259
	260	this_dbs_info = &per_cpu(cpu_dbs_info, cpu);
	261	if (!this_dbs_info->enable)
	262	return;
	263
	264	policy = this_dbs_info->cur_policy;
32ee8c3e	265	/*
c29f1403 DJ	266	* Every sampling_rate, we check, if current idle time is less
	267	* than 20% (default), then we try to increase frequency
	268	* Every sampling_rate*sampling_down_factor, we look for a the lowest
	269	* frequency which can sustain the load while keeping idle time over
	270	* 30%. If such a frequency exist, we try to decrease to this frequency.
1da177e4	271	*
32ee8c3e DJ	272	* Any frequency increase takes it to the maximum frequency.
	273	* Frequency reduction happens at minimum steps of
	274	* 5% (default) of current frequency
1da177e4 LT	275	*/
	276
	277	/* Check for frequency increase */
9c7d269b	278	idle_ticks = UINT_MAX;
1da177e4	279	for_each_cpu_mask(j, policy->cpus) {
9c7d269b	280	unsigned int tmp_idle_ticks, total_idle_ticks;
1da177e4 LT	281	struct cpu_dbs_info_s *j_dbs_info;
1da177e4 LT	282
1da177e4	283	j_dbs_info = &per_cpu(cpu_dbs_info, j);
dac1c1a5	284	total_idle_ticks = get_cpu_idle_time(j);
1da177e4 LT	285	tmp_idle_ticks = total_idle_ticks -
	286	j_dbs_info->prev_cpu_idle_up;
	287	j_dbs_info->prev_cpu_idle_up = total_idle_ticks;
	288
	289	if (tmp_idle_ticks < idle_ticks)
	290	idle_ticks = tmp_idle_ticks;
	291	}
	292
	293	/* Scale idle ticks by 100 and compare with up and down ticks */
	294	idle_ticks *= 100;
	295	up_idle_ticks = (100 - dbs_tuners_ins.up_threshold) *
6fe71165	296	usecs_to_jiffies(dbs_tuners_ins.sampling_rate);
1da177e4 LT	297
1da177e4 LT	298	if (idle_ticks < up_idle_ticks) {
dac1c1a5	299	down_skip[cpu] = 0;
790d76fa DJ	300	for_each_cpu_mask(j, policy->cpus) {
	301	struct cpu_dbs_info_s *j_dbs_info;
	302
	303	j_dbs_info = &per_cpu(cpu_dbs_info, j);
32ee8c3e	304	j_dbs_info->prev_cpu_idle_down =
790d76fa DJ	305	j_dbs_info->prev_cpu_idle_up;
790d76fa DJ	306	}
c11420a6 DJ	307	/* if we are already at full speed then break out early */
	308	if (policy->cur == policy->max)
	309	return;
32ee8c3e DJ	310
32ee8c3e DJ	311	__cpufreq_driver_target(policy, policy->max,
1da177e4	312	CPUFREQ_RELATION_H);
1da177e4 LT	313	return;
	314	}
	315
	316	/* Check for frequency decrease */
	317	down_skip[cpu]++;
	318	if (down_skip[cpu] < dbs_tuners_ins.sampling_down_factor)
	319	return;
	320
9c7d269b	321	idle_ticks = UINT_MAX;
1da177e4	322	for_each_cpu_mask(j, policy->cpus) {
9c7d269b	323	unsigned int tmp_idle_ticks, total_idle_ticks;
1da177e4 LT	324	struct cpu_dbs_info_s *j_dbs_info;
1da177e4 LT	325
1da177e4	326	j_dbs_info = &per_cpu(cpu_dbs_info, j);
dac1c1a5 DJ	327	/* Check for frequency decrease */
dac1c1a5 DJ	328	total_idle_ticks = j_dbs_info->prev_cpu_idle_up;
1da177e4 LT	329	tmp_idle_ticks = total_idle_ticks -
	330	j_dbs_info->prev_cpu_idle_down;
	331	j_dbs_info->prev_cpu_idle_down = total_idle_ticks;
	332
	333	if (tmp_idle_ticks < idle_ticks)
	334	idle_ticks = tmp_idle_ticks;
	335	}
	336
1da177e4	337	down_skip[cpu] = 0;
c29f1403 DJ	338	/* if we cannot reduce the frequency anymore, break out early */
	339	if (policy->cur == policy->min)
	340	return;
1da177e4	341
c29f1403	342	/* Compute how many ticks there are between two measurements */
1da177e4 LT	343	freq_down_sampling_rate = dbs_tuners_ins.sampling_rate *
1da177e4 LT	344	dbs_tuners_ins.sampling_down_factor;
c29f1403	345	total_ticks = usecs_to_jiffies(freq_down_sampling_rate);
1206aaac	346
c29f1403 DJ	347	/*
	348	* The optimal frequency is the frequency that is the lowest that
	349	* can support the current CPU usage without triggering the up
	350	* policy. To be safe, we focus 10 points under the threshold.
	351	*/
	352	freq_next = ((total_ticks - idle_ticks) * 100) / total_ticks;
32ee8c3e	353	freq_next = (freq_next * policy->cur) /
c29f1403	354	(dbs_tuners_ins.up_threshold - 10);
1da177e4	355
7c9d8c0e DB	356	if (freq_next < policy->min)
	357	freq_next = policy->min;
	358
c29f1403 DJ	359	if (freq_next <= ((policy->cur * 95) / 100))
c29f1403 DJ	360	__cpufreq_driver_target(policy, freq_next, CPUFREQ_RELATION_L);
1da177e4 LT	361	}
	362
	363	static void do_dbs_timer(void *data)
32ee8c3e	364	{
1da177e4	365	int i;
3fc54d37	366	mutex_lock(&dbs_mutex);
6fe71165 DJ	367	for_each_online_cpu(i)
6fe71165 DJ	368	dbs_check_cpu(i);
6810b548 AK	369	queue_delayed_work(dbs_workq, &dbs_work,
6810b548 AK	370	usecs_to_jiffies(dbs_tuners_ins.sampling_rate));
3fc54d37	371	mutex_unlock(&dbs_mutex);
32ee8c3e	372	}
1da177e4 LT	373
	374	static inline void dbs_timer_init(void)
	375	{
	376	INIT_WORK(&dbs_work, do_dbs_timer, NULL);
6810b548 AK	377	if (!dbs_workq)
	378	dbs_workq = create_singlethread_workqueue("ondemand");
	379	if (!dbs_workq) {
	380	printk(KERN_ERR "ondemand: Cannot initialize kernel thread\n");
	381	return;
	382	}
	383	queue_delayed_work(dbs_workq, &dbs_work,
	384	usecs_to_jiffies(dbs_tuners_ins.sampling_rate));
1da177e4 LT	385	return;
	386	}
	387
	388	static inline void dbs_timer_exit(void)
	389	{
6810b548 AK	390	if (dbs_workq)
6810b548 AK	391	cancel_rearming_delayed_workqueue(dbs_workq, &dbs_work);
1da177e4 LT	392	}
	393
	394	static int cpufreq_governor_dbs(struct cpufreq_policy *policy,
	395	unsigned int event)
	396	{
	397	unsigned int cpu = policy->cpu;
	398	struct cpu_dbs_info_s *this_dbs_info;
	399	unsigned int j;
	400
	401	this_dbs_info = &per_cpu(cpu_dbs_info, cpu);
	402
	403	switch (event) {
	404	case CPUFREQ_GOV_START:
32ee8c3e	405	if ((!cpu_online(cpu)) \|\|
1da177e4 LT	406	(!policy->cur))
	407	return -EINVAL;
	408
	409	if (policy->cpuinfo.transition_latency >
ff8c288d EP	410	(TRANSITION_LATENCY_LIMIT * 1000)) {
	411	printk(KERN_WARNING "ondemand governor failed to load "
	412	"due to too long transition latency\n");
1da177e4	413	return -EINVAL;
ff8c288d	414	}
1da177e4 LT	415	if (this_dbs_info->enable) /* Already enabled */
1da177e4 LT	416	break;
32ee8c3e	417
3fc54d37	418	mutex_lock(&dbs_mutex);
1da177e4 LT	419	for_each_cpu_mask(j, policy->cpus) {
	420	struct cpu_dbs_info_s *j_dbs_info;
	421	j_dbs_info = &per_cpu(cpu_dbs_info, j);
	422	j_dbs_info->cur_policy = policy;
32ee8c3e	423
dac1c1a5	424	j_dbs_info->prev_cpu_idle_up = get_cpu_idle_time(j);
3d5ee9e5 DJ	425	j_dbs_info->prev_cpu_idle_down
3d5ee9e5 DJ	426	= j_dbs_info->prev_cpu_idle_up;
1da177e4 LT	427	}
	428	this_dbs_info->enable = 1;
	429	sysfs_create_group(&policy->kobj, &dbs_attr_group);
	430	dbs_enable++;
	431	/*
	432	* Start the timerschedule work, when this governor
	433	* is used for first time
	434	*/
	435	if (dbs_enable == 1) {
	436	unsigned int latency;
	437	/* policy latency is in nS. Convert it to uS first */
df8b59be DJ	438	latency = policy->cpuinfo.transition_latency / 1000;
	439	if (latency == 0)
	440	latency = 1;
1da177e4	441
df8b59be	442	def_sampling_rate = latency *
1da177e4	443	DEF_SAMPLING_RATE_LATENCY_MULTIPLIER;
df8b59be DJ	444
	445	if (def_sampling_rate < MIN_STAT_SAMPLING_RATE)
	446	def_sampling_rate = MIN_STAT_SAMPLING_RATE;
	447
1da177e4	448	dbs_tuners_ins.sampling_rate = def_sampling_rate;
1da177e4 LT	449	dbs_timer_init();
1da177e4 LT	450	}
32ee8c3e	451
3fc54d37	452	mutex_unlock(&dbs_mutex);
1da177e4 LT	453	break;
	454
	455	case CPUFREQ_GOV_STOP:
3fc54d37	456	mutex_lock(&dbs_mutex);
1da177e4 LT	457	this_dbs_info->enable = 0;
	458	sysfs_remove_group(&policy->kobj, &dbs_attr_group);
	459	dbs_enable--;
	460	/*
	461	* Stop the timerschedule work, when this governor
	462	* is used for first time
	463	*/
32ee8c3e	464	if (dbs_enable == 0)
1da177e4	465	dbs_timer_exit();
32ee8c3e	466
3fc54d37	467	mutex_unlock(&dbs_mutex);
1da177e4 LT	468
	469	break;
	470
	471	case CPUFREQ_GOV_LIMITS:
3fc54d37	472	mutex_lock(&dbs_mutex);
1da177e4 LT	473	if (policy->max < this_dbs_info->cur_policy->cur)
	474	__cpufreq_driver_target(
	475	this_dbs_info->cur_policy,
32ee8c3e	476	policy->max, CPUFREQ_RELATION_H);
1da177e4 LT	477	else if (policy->min > this_dbs_info->cur_policy->cur)
	478	__cpufreq_driver_target(
	479	this_dbs_info->cur_policy,
32ee8c3e	480	policy->min, CPUFREQ_RELATION_L);
3fc54d37	481	mutex_unlock(&dbs_mutex);
1da177e4 LT	482	break;
	483	}
	484	return 0;
	485	}
	486
7f335d4e	487	static struct cpufreq_governor cpufreq_gov_dbs = {
1da177e4 LT	488	.name = "ondemand",
	489	.governor = cpufreq_governor_dbs,
	490	.owner = THIS_MODULE,
	491	};
1da177e4 LT	492
	493	static int __init cpufreq_gov_dbs_init(void)
	494	{
	495	return cpufreq_register_governor(&cpufreq_gov_dbs);
	496	}
	497
	498	static void __exit cpufreq_gov_dbs_exit(void)
	499	{
6810b548 AK	500	/* Make sure that the scheduled work is indeed not running.
	501	Assumes the timer has been cancelled first. */
	502	if (dbs_workq) {
	503	flush_workqueue(dbs_workq);
	504	destroy_workqueue(dbs_workq);
	505	}
1da177e4 LT	506
	507	cpufreq_unregister_governor(&cpufreq_gov_dbs);
	508	}
	509
	510
	511	MODULE_AUTHOR ("Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>");
	512	MODULE_DESCRIPTION ("'cpufreq_ondemand' - A dynamic cpufreq governor for "
	513	"Low Latency Frequency Transition capable processors");
	514	MODULE_LICENSE ("GPL");
	515
	516	module_init(cpufreq_gov_dbs_init);
	517	module_exit(cpufreq_gov_dbs_exit);