[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 */

/*
 * 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 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;
	lock_cpu_hotplug();
	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);
	unlock_cpu_hotplug();
}

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:
		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)
{
	/* 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
4ec223d0 VP	74	/*
	75	* DEADLOCK ALERT! There is a ordering requirement between cpu_hotplug
	76	* lock and dbs_mutex. cpu_hotplug lock should always be held before
	77	* dbs_mutex. If any function that can potentially take cpu_hotplug lock
	78	* (like __cpufreq_driver_target()) is being called with dbs_mutex taken, then
	79	* cpu_hotplug lock should be taken before that. Note that cpu_hotplug lock
	80	* is recursive for the same process. -Venki
	81	*/
32ee8c3e	82	static DEFINE_MUTEX (dbs_mutex);
1da177e4 LT	83	static DECLARE_WORK (dbs_work, do_dbs_timer, NULL);
1da177e4 LT	84
6810b548 AK	85	static struct workqueue_struct *dbs_workq;
6810b548 AK	86
1da177e4	87	struct dbs_tuners {
32ee8c3e DJ	88	unsigned int sampling_rate;
	89	unsigned int sampling_down_factor;
	90	unsigned int up_threshold;
	91	unsigned int ignore_nice;
1da177e4 LT	92	};
	93
	94	static struct dbs_tuners dbs_tuners_ins = {
32ee8c3e DJ	95	.up_threshold = DEF_FREQUENCY_UP_THRESHOLD,
32ee8c3e DJ	96	.sampling_down_factor = DEF_SAMPLING_DOWN_FACTOR,
9cbad61b	97	.ignore_nice = 0,
1da177e4 LT	98	};
1da177e4 LT	99
dac1c1a5 DJ	100	static inline unsigned int get_cpu_idle_time(unsigned int cpu)
	101	{
	102	return kstat_cpu(cpu).cpustat.idle +
	103	kstat_cpu(cpu).cpustat.iowait +
001893cd	104	( dbs_tuners_ins.ignore_nice ?
dac1c1a5 DJ	105	kstat_cpu(cpu).cpustat.nice :
	106	0);
	107	}
	108
1da177e4 LT	109	/************************ sysfs interface **********************/
	110	static ssize_t show_sampling_rate_max(struct cpufreq_policy policy, char buf)
	111	{
	112	return sprintf (buf, "%u\n", MAX_SAMPLING_RATE);
	113	}
	114
	115	static ssize_t show_sampling_rate_min(struct cpufreq_policy policy, char buf)
	116	{
	117	return sprintf (buf, "%u\n", MIN_SAMPLING_RATE);
	118	}
	119
32ee8c3e DJ	120	#define define_one_ro(_name) \
32ee8c3e DJ	121	static struct freq_attr _name = \
1da177e4 LT	122	__ATTR(_name, 0444, show_##_name, NULL)
	123
	124	define_one_ro(sampling_rate_max);
	125	define_one_ro(sampling_rate_min);
	126
	127	/* cpufreq_ondemand Governor Tunables */
	128	#define show_one(file_name, object) \
	129	static ssize_t show_##file_name \
	130	(struct cpufreq_policy unused, char buf) \
	131	{ \
	132	return sprintf(buf, "%u\n", dbs_tuners_ins.object); \
	133	}
	134	show_one(sampling_rate, sampling_rate);
	135	show_one(sampling_down_factor, sampling_down_factor);
	136	show_one(up_threshold, up_threshold);
001893cd	137	show_one(ignore_nice_load, ignore_nice);
1da177e4	138
32ee8c3e	139	static ssize_t store_sampling_down_factor(struct cpufreq_policy *unused,
1da177e4 LT	140	const char *buf, size_t count)
	141	{
	142	unsigned int input;
	143	int ret;
	144	ret = sscanf (buf, "%u", &input);
	145	if (ret != 1 )
	146	return -EINVAL;
	147
e131832c DJ	148	if (input > MAX_SAMPLING_DOWN_FACTOR \|\| input < 1)
	149	return -EINVAL;
	150
3fc54d37	151	mutex_lock(&dbs_mutex);
1da177e4	152	dbs_tuners_ins.sampling_down_factor = input;
3fc54d37	153	mutex_unlock(&dbs_mutex);
1da177e4 LT	154
	155	return count;
	156	}
	157
32ee8c3e	158	static ssize_t store_sampling_rate(struct cpufreq_policy *unused,
1da177e4 LT	159	const char *buf, size_t count)
	160	{
	161	unsigned int input;
	162	int ret;
	163	ret = sscanf (buf, "%u", &input);
	164
3fc54d37	165	mutex_lock(&dbs_mutex);
1da177e4	166	if (ret != 1 \|\| input > MAX_SAMPLING_RATE \|\| input < MIN_SAMPLING_RATE) {
3fc54d37	167	mutex_unlock(&dbs_mutex);
1da177e4 LT	168	return -EINVAL;
	169	}
	170
	171	dbs_tuners_ins.sampling_rate = input;
3fc54d37	172	mutex_unlock(&dbs_mutex);
1da177e4 LT	173
	174	return count;
	175	}
	176
32ee8c3e	177	static ssize_t store_up_threshold(struct cpufreq_policy *unused,
1da177e4 LT	178	const char *buf, size_t count)
	179	{
	180	unsigned int input;
	181	int ret;
	182	ret = sscanf (buf, "%u", &input);
	183
3fc54d37	184	mutex_lock(&dbs_mutex);
32ee8c3e	185	if (ret != 1 \|\| input > MAX_FREQUENCY_UP_THRESHOLD \|\|
c29f1403	186	input < MIN_FREQUENCY_UP_THRESHOLD) {
3fc54d37	187	mutex_unlock(&dbs_mutex);
1da177e4 LT	188	return -EINVAL;
	189	}
	190
	191	dbs_tuners_ins.up_threshold = input;
3fc54d37	192	mutex_unlock(&dbs_mutex);
1da177e4 LT	193
	194	return count;
	195	}
	196
001893cd	197	static ssize_t store_ignore_nice_load(struct cpufreq_policy *policy,
3d5ee9e5 DJ	198	const char *buf, size_t count)
	199	{
	200	unsigned int input;
	201	int ret;
	202
	203	unsigned int j;
32ee8c3e	204
3d5ee9e5 DJ	205	ret = sscanf (buf, "%u", &input);
	206	if ( ret != 1 )
	207	return -EINVAL;
	208
	209	if ( input > 1 )
	210	input = 1;
32ee8c3e	211
3fc54d37	212	mutex_lock(&dbs_mutex);
3d5ee9e5	213	if ( input == dbs_tuners_ins.ignore_nice ) { /* nothing to do */
3fc54d37	214	mutex_unlock(&dbs_mutex);
3d5ee9e5 DJ	215	return count;
	216	}
	217	dbs_tuners_ins.ignore_nice = input;
	218
	219	/* we need to re-evaluate prev_cpu_idle_up and prev_cpu_idle_down */
dac1c1a5	220	for_each_online_cpu(j) {
3d5ee9e5 DJ	221	struct cpu_dbs_info_s *j_dbs_info;
3d5ee9e5 DJ	222	j_dbs_info = &per_cpu(cpu_dbs_info, j);
dac1c1a5	223	j_dbs_info->prev_cpu_idle_up = get_cpu_idle_time(j);
3d5ee9e5 DJ	224	j_dbs_info->prev_cpu_idle_down = j_dbs_info->prev_cpu_idle_up;
3d5ee9e5 DJ	225	}
3fc54d37	226	mutex_unlock(&dbs_mutex);
3d5ee9e5 DJ	227
	228	return count;
	229	}
	230
1da177e4 LT	231	#define define_one_rw(_name) \
	232	static struct freq_attr _name = \
	233	__ATTR(_name, 0644, show_##_name, store_##_name)
	234
	235	define_one_rw(sampling_rate);
	236	define_one_rw(sampling_down_factor);
	237	define_one_rw(up_threshold);
001893cd	238	define_one_rw(ignore_nice_load);
1da177e4 LT	239
	240	static struct attribute * dbs_attributes[] = {
	241	&sampling_rate_max.attr,
	242	&sampling_rate_min.attr,
	243	&sampling_rate.attr,
	244	&sampling_down_factor.attr,
	245	&up_threshold.attr,
001893cd	246	&ignore_nice_load.attr,
1da177e4 LT	247	NULL
	248	};
	249
	250	static struct attribute_group dbs_attr_group = {
	251	.attrs = dbs_attributes,
	252	.name = "ondemand",
	253	};
	254
	255	/************************ sysfs end **********************/
	256
	257	static void dbs_check_cpu(int cpu)
	258	{
c29f1403 DJ	259	unsigned int idle_ticks, up_idle_ticks, total_ticks;
c29f1403 DJ	260	unsigned int freq_next;
1da177e4 LT	261	unsigned int freq_down_sampling_rate;
	262	static int down_skip[NR_CPUS];
	263	struct cpu_dbs_info_s *this_dbs_info;
	264
	265	struct cpufreq_policy *policy;
	266	unsigned int j;
	267
	268	this_dbs_info = &per_cpu(cpu_dbs_info, cpu);
	269	if (!this_dbs_info->enable)
	270	return;
	271
	272	policy = this_dbs_info->cur_policy;
32ee8c3e	273	/*
c29f1403 DJ	274	* Every sampling_rate, we check, if current idle time is less
	275	* than 20% (default), then we try to increase frequency
	276	* Every sampling_rate*sampling_down_factor, we look for a the lowest
	277	* frequency which can sustain the load while keeping idle time over
	278	* 30%. If such a frequency exist, we try to decrease to this frequency.
1da177e4	279	*
32ee8c3e DJ	280	* Any frequency increase takes it to the maximum frequency.
	281	* Frequency reduction happens at minimum steps of
	282	* 5% (default) of current frequency
1da177e4 LT	283	*/
	284
	285	/* Check for frequency increase */
9c7d269b	286	idle_ticks = UINT_MAX;
1da177e4	287	for_each_cpu_mask(j, policy->cpus) {
9c7d269b	288	unsigned int tmp_idle_ticks, total_idle_ticks;
1da177e4 LT	289	struct cpu_dbs_info_s *j_dbs_info;
1da177e4 LT	290
1da177e4	291	j_dbs_info = &per_cpu(cpu_dbs_info, j);
dac1c1a5	292	total_idle_ticks = get_cpu_idle_time(j);
1da177e4 LT	293	tmp_idle_ticks = total_idle_ticks -
	294	j_dbs_info->prev_cpu_idle_up;
	295	j_dbs_info->prev_cpu_idle_up = total_idle_ticks;
	296
	297	if (tmp_idle_ticks < idle_ticks)
	298	idle_ticks = tmp_idle_ticks;
	299	}
	300
	301	/* Scale idle ticks by 100 and compare with up and down ticks */
	302	idle_ticks *= 100;
	303	up_idle_ticks = (100 - dbs_tuners_ins.up_threshold) *
6fe71165	304	usecs_to_jiffies(dbs_tuners_ins.sampling_rate);
1da177e4 LT	305
1da177e4 LT	306	if (idle_ticks < up_idle_ticks) {
dac1c1a5	307	down_skip[cpu] = 0;
790d76fa DJ	308	for_each_cpu_mask(j, policy->cpus) {
	309	struct cpu_dbs_info_s *j_dbs_info;
	310
	311	j_dbs_info = &per_cpu(cpu_dbs_info, j);
32ee8c3e	312	j_dbs_info->prev_cpu_idle_down =
790d76fa DJ	313	j_dbs_info->prev_cpu_idle_up;
790d76fa DJ	314	}
c11420a6 DJ	315	/* if we are already at full speed then break out early */
	316	if (policy->cur == policy->max)
	317	return;
32ee8c3e DJ	318
32ee8c3e DJ	319	__cpufreq_driver_target(policy, policy->max,
1da177e4	320	CPUFREQ_RELATION_H);
1da177e4 LT	321	return;
	322	}
	323
	324	/* Check for frequency decrease */
	325	down_skip[cpu]++;
	326	if (down_skip[cpu] < dbs_tuners_ins.sampling_down_factor)
	327	return;
	328
9c7d269b	329	idle_ticks = UINT_MAX;
1da177e4	330	for_each_cpu_mask(j, policy->cpus) {
9c7d269b	331	unsigned int tmp_idle_ticks, total_idle_ticks;
1da177e4 LT	332	struct cpu_dbs_info_s *j_dbs_info;
1da177e4 LT	333
1da177e4	334	j_dbs_info = &per_cpu(cpu_dbs_info, j);
dac1c1a5 DJ	335	/* Check for frequency decrease */
dac1c1a5 DJ	336	total_idle_ticks = j_dbs_info->prev_cpu_idle_up;
1da177e4 LT	337	tmp_idle_ticks = total_idle_ticks -
	338	j_dbs_info->prev_cpu_idle_down;
	339	j_dbs_info->prev_cpu_idle_down = total_idle_ticks;
	340
	341	if (tmp_idle_ticks < idle_ticks)
	342	idle_ticks = tmp_idle_ticks;
	343	}
	344
1da177e4	345	down_skip[cpu] = 0;
c29f1403 DJ	346	/* if we cannot reduce the frequency anymore, break out early */
	347	if (policy->cur == policy->min)
	348	return;
1da177e4	349
c29f1403	350	/* Compute how many ticks there are between two measurements */
1da177e4 LT	351	freq_down_sampling_rate = dbs_tuners_ins.sampling_rate *
1da177e4 LT	352	dbs_tuners_ins.sampling_down_factor;
c29f1403	353	total_ticks = usecs_to_jiffies(freq_down_sampling_rate);
1206aaac	354
c29f1403 DJ	355	/*
	356	* The optimal frequency is the frequency that is the lowest that
	357	* can support the current CPU usage without triggering the up
	358	* policy. To be safe, we focus 10 points under the threshold.
	359	*/
	360	freq_next = ((total_ticks - idle_ticks) * 100) / total_ticks;
32ee8c3e	361	freq_next = (freq_next * policy->cur) /
c29f1403	362	(dbs_tuners_ins.up_threshold - 10);
1da177e4	363
7c9d8c0e DB	364	if (freq_next < policy->min)
	365	freq_next = policy->min;
	366
c29f1403 DJ	367	if (freq_next <= ((policy->cur * 95) / 100))
c29f1403 DJ	368	__cpufreq_driver_target(policy, freq_next, CPUFREQ_RELATION_L);
1da177e4 LT	369	}
	370
	371	static void do_dbs_timer(void *data)
32ee8c3e	372	{
1da177e4	373	int i;
4ec223d0	374	lock_cpu_hotplug();
3fc54d37	375	mutex_lock(&dbs_mutex);
6fe71165 DJ	376	for_each_online_cpu(i)
6fe71165 DJ	377	dbs_check_cpu(i);
6810b548 AK	378	queue_delayed_work(dbs_workq, &dbs_work,
6810b548 AK	379	usecs_to_jiffies(dbs_tuners_ins.sampling_rate));
3fc54d37	380	mutex_unlock(&dbs_mutex);
4ec223d0	381	unlock_cpu_hotplug();
32ee8c3e	382	}
1da177e4 LT	383
	384	static inline void dbs_timer_init(void)
	385	{
	386	INIT_WORK(&dbs_work, do_dbs_timer, NULL);
6810b548 AK	387	if (!dbs_workq)
	388	dbs_workq = create_singlethread_workqueue("ondemand");
	389	if (!dbs_workq) {
	390	printk(KERN_ERR "ondemand: Cannot initialize kernel thread\n");
	391	return;
	392	}
	393	queue_delayed_work(dbs_workq, &dbs_work,
	394	usecs_to_jiffies(dbs_tuners_ins.sampling_rate));
1da177e4 LT	395	return;
	396	}
	397
	398	static inline void dbs_timer_exit(void)
	399	{
6810b548 AK	400	if (dbs_workq)
6810b548 AK	401	cancel_rearming_delayed_workqueue(dbs_workq, &dbs_work);
1da177e4 LT	402	}
	403
	404	static int cpufreq_governor_dbs(struct cpufreq_policy *policy,
	405	unsigned int event)
	406	{
	407	unsigned int cpu = policy->cpu;
	408	struct cpu_dbs_info_s *this_dbs_info;
	409	unsigned int j;
	410
	411	this_dbs_info = &per_cpu(cpu_dbs_info, cpu);
	412
	413	switch (event) {
	414	case CPUFREQ_GOV_START:
32ee8c3e	415	if ((!cpu_online(cpu)) \|\|
1da177e4 LT	416	(!policy->cur))
	417	return -EINVAL;
	418
	419	if (policy->cpuinfo.transition_latency >
ff8c288d EP	420	(TRANSITION_LATENCY_LIMIT * 1000)) {
	421	printk(KERN_WARNING "ondemand governor failed to load "
	422	"due to too long transition latency\n");
1da177e4	423	return -EINVAL;
ff8c288d	424	}
1da177e4 LT	425	if (this_dbs_info->enable) /* Already enabled */
1da177e4 LT	426	break;
32ee8c3e	427
3fc54d37	428	mutex_lock(&dbs_mutex);
1da177e4 LT	429	for_each_cpu_mask(j, policy->cpus) {
	430	struct cpu_dbs_info_s *j_dbs_info;
	431	j_dbs_info = &per_cpu(cpu_dbs_info, j);
	432	j_dbs_info->cur_policy = policy;
32ee8c3e	433
dac1c1a5	434	j_dbs_info->prev_cpu_idle_up = get_cpu_idle_time(j);
3d5ee9e5 DJ	435	j_dbs_info->prev_cpu_idle_down
3d5ee9e5 DJ	436	= j_dbs_info->prev_cpu_idle_up;
1da177e4 LT	437	}
	438	this_dbs_info->enable = 1;
	439	sysfs_create_group(&policy->kobj, &dbs_attr_group);
	440	dbs_enable++;
	441	/*
	442	* Start the timerschedule work, when this governor
	443	* is used for first time
	444	*/
	445	if (dbs_enable == 1) {
	446	unsigned int latency;
	447	/* policy latency is in nS. Convert it to uS first */
df8b59be DJ	448	latency = policy->cpuinfo.transition_latency / 1000;
	449	if (latency == 0)
	450	latency = 1;
1da177e4	451
df8b59be	452	def_sampling_rate = latency *
1da177e4	453	DEF_SAMPLING_RATE_LATENCY_MULTIPLIER;
df8b59be DJ	454
	455	if (def_sampling_rate < MIN_STAT_SAMPLING_RATE)
	456	def_sampling_rate = MIN_STAT_SAMPLING_RATE;
	457
1da177e4	458	dbs_tuners_ins.sampling_rate = def_sampling_rate;
1da177e4 LT	459	dbs_timer_init();
1da177e4 LT	460	}
32ee8c3e	461
3fc54d37	462	mutex_unlock(&dbs_mutex);
1da177e4 LT	463	break;
	464
	465	case CPUFREQ_GOV_STOP:
3fc54d37	466	mutex_lock(&dbs_mutex);
1da177e4 LT	467	this_dbs_info->enable = 0;
	468	sysfs_remove_group(&policy->kobj, &dbs_attr_group);
	469	dbs_enable--;
	470	/*
	471	* Stop the timerschedule work, when this governor
	472	* is used for first time
	473	*/
32ee8c3e	474	if (dbs_enable == 0)
1da177e4	475	dbs_timer_exit();
32ee8c3e	476
3fc54d37	477	mutex_unlock(&dbs_mutex);
1da177e4 LT	478
	479	break;
	480
	481	case CPUFREQ_GOV_LIMITS:
4ec223d0	482	lock_cpu_hotplug();
3fc54d37	483	mutex_lock(&dbs_mutex);
1da177e4 LT	484	if (policy->max < this_dbs_info->cur_policy->cur)
	485	__cpufreq_driver_target(
	486	this_dbs_info->cur_policy,
32ee8c3e	487	policy->max, CPUFREQ_RELATION_H);
1da177e4 LT	488	else if (policy->min > this_dbs_info->cur_policy->cur)
	489	__cpufreq_driver_target(
	490	this_dbs_info->cur_policy,
32ee8c3e	491	policy->min, CPUFREQ_RELATION_L);
3fc54d37	492	mutex_unlock(&dbs_mutex);
4ec223d0	493	unlock_cpu_hotplug();
1da177e4 LT	494	break;
	495	}
	496	return 0;
	497	}
	498
7f335d4e	499	static struct cpufreq_governor cpufreq_gov_dbs = {
1da177e4 LT	500	.name = "ondemand",
	501	.governor = cpufreq_governor_dbs,
	502	.owner = THIS_MODULE,
	503	};
1da177e4 LT	504
	505	static int __init cpufreq_gov_dbs_init(void)
	506	{
	507	return cpufreq_register_governor(&cpufreq_gov_dbs);
	508	}
	509
	510	static void __exit cpufreq_gov_dbs_exit(void)
	511	{
6810b548 AK	512	/* Make sure that the scheduled work is indeed not running.
	513	Assumes the timer has been cancelled first. */
	514	if (dbs_workq) {
	515	flush_workqueue(dbs_workq);
	516	destroy_workqueue(dbs_workq);
	517	}
1da177e4 LT	518
	519	cpufreq_unregister_governor(&cpufreq_gov_dbs);
	520	}
	521
	522
	523	MODULE_AUTHOR ("Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>");
	524	MODULE_DESCRIPTION ("'cpufreq_ondemand' - A dynamic cpufreq governor for "
	525	"Low Latency Frequency Transition capable processors");
	526	MODULE_LICENSE ("GPL");
	527
	528	module_init(cpufreq_gov_dbs_init);
	529	module_exit(cpufreq_gov_dbs_exit);