[mirror_ubuntu-artful-kernel.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/cpu.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 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;
	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 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(int cpu)
{
	unsigned int idle_ticks, total_ticks;
	unsigned int load;
	struct cpu_dbs_info_s *this_dbs_info;
	cputime64_t cur_jiffies;

	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;
	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)
{
	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 = 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);
		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>
138a0128	24	#include <linux/cpu.h>
1da177e4 LT	25	#include <linux/sched.h>
	26	#include <linux/kmod.h>
	27	#include <linux/workqueue.h>
	28	#include <linux/jiffies.h>
	29	#include <linux/kernel_stat.h>
	30	#include <linux/percpu.h>
3fc54d37	31	#include <linux/mutex.h>
1da177e4 LT	32
	33	/*
	34	* dbs is used in this file as a shortform for demandbased switching
	35	* It helps to keep variable names smaller, simpler
	36	*/
	37
	38	#define DEF_FREQUENCY_UP_THRESHOLD (80)
c29f1403	39	#define MIN_FREQUENCY_UP_THRESHOLD (11)
1da177e4 LT	40	#define MAX_FREQUENCY_UP_THRESHOLD (100)
1da177e4 LT	41
32ee8c3e DJ	42	/*
32ee8c3e DJ	43	* The polling frequency of this governor depends on the capability of
1da177e4	44	* the processor. Default polling frequency is 1000 times the transition
32ee8c3e DJ	45	* latency of the processor. The governor will work on any processor with
32ee8c3e DJ	46	* transition latency <= 10mS, using appropriate sampling
1da177e4 LT	47	* rate.
	48	* For CPUs with transition latency > 10mS (mostly drivers with CPUFREQ_ETERNAL)
	49	* this governor will not work.
	50	* All times here are in uS.
	51	*/
32ee8c3e	52	static unsigned int def_sampling_rate;
df8b59be DJ	53	#define MIN_SAMPLING_RATE_RATIO (2)
	54	/* for correct statistics, we need at least 10 ticks between each measure */
	55	#define MIN_STAT_SAMPLING_RATE (MIN_SAMPLING_RATE_RATIO * jiffies_to_usecs(10))
	56	#define MIN_SAMPLING_RATE (def_sampling_rate / MIN_SAMPLING_RATE_RATIO)
1da177e4 LT	57	#define MAX_SAMPLING_RATE (500 * def_sampling_rate)
1da177e4 LT	58	#define DEF_SAMPLING_RATE_LATENCY_MULTIPLIER (1000)
1da177e4	59	#define TRANSITION_LATENCY_LIMIT (10 * 1000)
1da177e4 LT	60
	61	static void do_dbs_timer(void *data);
	62
	63	struct cpu_dbs_info_s {
ccb2fe20 VP	64	cputime64_t prev_cpu_idle;
ccb2fe20 VP	65	cputime64_t prev_cpu_wall;
32ee8c3e	66	struct cpufreq_policy *cur_policy;
32ee8c3e	67	unsigned int enable;
1da177e4 LT	68	};
	69	static DEFINE_PER_CPU(struct cpu_dbs_info_s, cpu_dbs_info);
	70
	71	static unsigned int dbs_enable; /* number of CPUs using this policy */
	72
4ec223d0 VP	73	/*
	74	* DEADLOCK ALERT! There is a ordering requirement between cpu_hotplug
	75	* lock and dbs_mutex. cpu_hotplug lock should always be held before
	76	* dbs_mutex. If any function that can potentially take cpu_hotplug lock
	77	* (like __cpufreq_driver_target()) is being called with dbs_mutex taken, then
	78	* cpu_hotplug lock should be taken before that. Note that cpu_hotplug lock
	79	* is recursive for the same process. -Venki
	80	*/
32ee8c3e	81	static DEFINE_MUTEX (dbs_mutex);
1da177e4 LT	82	static DECLARE_WORK (dbs_work, do_dbs_timer, NULL);
1da177e4 LT	83
6810b548 AK	84	static struct workqueue_struct *dbs_workq;
6810b548 AK	85
1da177e4	86	struct dbs_tuners {
32ee8c3e	87	unsigned int sampling_rate;
32ee8c3e DJ	88	unsigned int up_threshold;
32ee8c3e DJ	89	unsigned int ignore_nice;
1da177e4 LT	90	};
	91
	92	static struct dbs_tuners dbs_tuners_ins = {
32ee8c3e	93	.up_threshold = DEF_FREQUENCY_UP_THRESHOLD,
9cbad61b	94	.ignore_nice = 0,
1da177e4 LT	95	};
1da177e4 LT	96
ccb2fe20	97	static inline cputime64_t get_cpu_idle_time(unsigned int cpu)
dac1c1a5	98	{
ccb2fe20 VP	99	cputime64_t retval;
	100
	101	retval = cputime64_add(kstat_cpu(cpu).cpustat.idle,
	102	kstat_cpu(cpu).cpustat.iowait);
	103
	104	if (dbs_tuners_ins.ignore_nice)
	105	retval = cputime64_add(retval, kstat_cpu(cpu).cpustat.nice);
	106
	107	return retval;
dac1c1a5 DJ	108	}
dac1c1a5 DJ	109
1da177e4 LT	110	/************************ sysfs interface **********************/
	111	static ssize_t show_sampling_rate_max(struct cpufreq_policy policy, char buf)
	112	{
	113	return sprintf (buf, "%u\n", MAX_SAMPLING_RATE);
	114	}
	115
	116	static ssize_t show_sampling_rate_min(struct cpufreq_policy policy, char buf)
	117	{
	118	return sprintf (buf, "%u\n", MIN_SAMPLING_RATE);
	119	}
	120
32ee8c3e DJ	121	#define define_one_ro(_name) \
32ee8c3e DJ	122	static struct freq_attr _name = \
1da177e4 LT	123	__ATTR(_name, 0444, show_##_name, NULL)
	124
	125	define_one_ro(sampling_rate_max);
	126	define_one_ro(sampling_rate_min);
	127
	128	/* cpufreq_ondemand Governor Tunables */
	129	#define show_one(file_name, object) \
	130	static ssize_t show_##file_name \
	131	(struct cpufreq_policy unused, char buf) \
	132	{ \
	133	return sprintf(buf, "%u\n", dbs_tuners_ins.object); \
	134	}
	135	show_one(sampling_rate, sampling_rate);
1da177e4	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_rate(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
3fc54d37	146	mutex_lock(&dbs_mutex);
1da177e4	147	if (ret != 1 \|\| input > MAX_SAMPLING_RATE \|\| input < MIN_SAMPLING_RATE) {
3fc54d37	148	mutex_unlock(&dbs_mutex);
1da177e4 LT	149	return -EINVAL;
	150	}
	151
	152	dbs_tuners_ins.sampling_rate = input;
3fc54d37	153	mutex_unlock(&dbs_mutex);
1da177e4 LT	154
	155	return count;
	156	}
	157
32ee8c3e	158	static ssize_t store_up_threshold(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);
32ee8c3e	166	if (ret != 1 \|\| input > MAX_FREQUENCY_UP_THRESHOLD \|\|
c29f1403	167	input < MIN_FREQUENCY_UP_THRESHOLD) {
3fc54d37	168	mutex_unlock(&dbs_mutex);
1da177e4 LT	169	return -EINVAL;
	170	}
	171
	172	dbs_tuners_ins.up_threshold = input;
3fc54d37	173	mutex_unlock(&dbs_mutex);
1da177e4 LT	174
	175	return count;
	176	}
	177
001893cd	178	static ssize_t store_ignore_nice_load(struct cpufreq_policy *policy,
3d5ee9e5 DJ	179	const char *buf, size_t count)
	180	{
	181	unsigned int input;
	182	int ret;
	183
	184	unsigned int j;
32ee8c3e	185
3d5ee9e5 DJ	186	ret = sscanf (buf, "%u", &input);
	187	if ( ret != 1 )
	188	return -EINVAL;
	189
	190	if ( input > 1 )
	191	input = 1;
32ee8c3e	192
3fc54d37	193	mutex_lock(&dbs_mutex);
3d5ee9e5	194	if ( input == dbs_tuners_ins.ignore_nice ) { /* nothing to do */
3fc54d37	195	mutex_unlock(&dbs_mutex);
3d5ee9e5 DJ	196	return count;
	197	}
	198	dbs_tuners_ins.ignore_nice = input;
	199
ccb2fe20	200	/* we need to re-evaluate prev_cpu_idle */
dac1c1a5	201	for_each_online_cpu(j) {
ccb2fe20 VP	202	struct cpu_dbs_info_s *dbs_info;
	203	dbs_info = &per_cpu(cpu_dbs_info, j);
	204	dbs_info->prev_cpu_idle = get_cpu_idle_time(j);
	205	dbs_info->prev_cpu_wall = get_jiffies_64();
3d5ee9e5	206	}
3fc54d37	207	mutex_unlock(&dbs_mutex);
3d5ee9e5 DJ	208
	209	return count;
	210	}
	211
1da177e4 LT	212	#define define_one_rw(_name) \
	213	static struct freq_attr _name = \
	214	__ATTR(_name, 0644, show_##_name, store_##_name)
	215
	216	define_one_rw(sampling_rate);
1da177e4	217	define_one_rw(up_threshold);
001893cd	218	define_one_rw(ignore_nice_load);
1da177e4 LT	219
	220	static struct attribute * dbs_attributes[] = {
	221	&sampling_rate_max.attr,
	222	&sampling_rate_min.attr,
	223	&sampling_rate.attr,
1da177e4	224	&up_threshold.attr,
001893cd	225	&ignore_nice_load.attr,
1da177e4 LT	226	NULL
	227	};
	228
	229	static struct attribute_group dbs_attr_group = {
	230	.attrs = dbs_attributes,
	231	.name = "ondemand",
	232	};
	233
	234	/************************ sysfs end **********************/
	235
	236	static void dbs_check_cpu(int cpu)
	237	{
ccb2fe20 VP	238	unsigned int idle_ticks, total_ticks;
ccb2fe20 VP	239	unsigned int load;
1da177e4	240	struct cpu_dbs_info_s *this_dbs_info;
ccb2fe20	241	cputime64_t cur_jiffies;
1da177e4 LT	242
	243	struct cpufreq_policy *policy;
	244	unsigned int j;
	245
	246	this_dbs_info = &per_cpu(cpu_dbs_info, cpu);
	247	if (!this_dbs_info->enable)
	248	return;
	249
	250	policy = this_dbs_info->cur_policy;
ccb2fe20 VP	251	cur_jiffies = jiffies64_to_cputime64(get_jiffies_64());
	252	total_ticks = (unsigned int) cputime64_sub(cur_jiffies,
	253	this_dbs_info->prev_cpu_wall);
	254	this_dbs_info->prev_cpu_wall = cur_jiffies;
32ee8c3e	255	/*
c29f1403 DJ	256	* Every sampling_rate, we check, if current idle time is less
c29f1403 DJ	257	* than 20% (default), then we try to increase frequency
ccb2fe20	258	* Every sampling_rate, we look for a the lowest
c29f1403 DJ	259	* frequency which can sustain the load while keeping idle time over
c29f1403 DJ	260	* 30%. If such a frequency exist, we try to decrease to this frequency.
1da177e4	261	*
32ee8c3e DJ	262	* Any frequency increase takes it to the maximum frequency.
	263	* Frequency reduction happens at minimum steps of
	264	* 5% (default) of current frequency
1da177e4 LT	265	*/
1da177e4 LT	266
ccb2fe20	267	/* Get Idle Time */
9c7d269b	268	idle_ticks = UINT_MAX;
1da177e4	269	for_each_cpu_mask(j, policy->cpus) {
ccb2fe20 VP	270	cputime64_t total_idle_ticks;
ccb2fe20 VP	271	unsigned int tmp_idle_ticks;
1da177e4 LT	272	struct cpu_dbs_info_s *j_dbs_info;
1da177e4 LT	273
1da177e4	274	j_dbs_info = &per_cpu(cpu_dbs_info, j);
dac1c1a5	275	total_idle_ticks = get_cpu_idle_time(j);
ccb2fe20 VP	276	tmp_idle_ticks = (unsigned int) cputime64_sub(total_idle_ticks,
	277	j_dbs_info->prev_cpu_idle);
	278	j_dbs_info->prev_cpu_idle = total_idle_ticks;
1da177e4 LT	279
	280	if (tmp_idle_ticks < idle_ticks)
	281	idle_ticks = tmp_idle_ticks;
	282	}
ccb2fe20	283	load = (100 * (total_ticks - idle_ticks)) / total_ticks;
1da177e4	284
ccb2fe20 VP	285	/* Check for frequency increase */
ccb2fe20 VP	286	if (load > dbs_tuners_ins.up_threshold) {
c11420a6 DJ	287	/* if we are already at full speed then break out early */
	288	if (policy->cur == policy->max)
	289	return;
32ee8c3e DJ	290
32ee8c3e DJ	291	__cpufreq_driver_target(policy, policy->max,
1da177e4	292	CPUFREQ_RELATION_H);
1da177e4 LT	293	return;
	294	}
	295
	296	/* Check for frequency decrease */
c29f1403 DJ	297	/* if we cannot reduce the frequency anymore, break out early */
	298	if (policy->cur == policy->min)
	299	return;
1da177e4	300
c29f1403 DJ	301	/*
	302	* The optimal frequency is the frequency that is the lowest that
	303	* can support the current CPU usage without triggering the up
	304	* policy. To be safe, we focus 10 points under the threshold.
	305	*/
ccb2fe20 VP	306	if (load < (dbs_tuners_ins.up_threshold - 10)) {
	307	unsigned int freq_next;
	308	freq_next = (policy->cur * load) /
c29f1403	309	(dbs_tuners_ins.up_threshold - 10);
1da177e4	310
c29f1403	311	__cpufreq_driver_target(policy, freq_next, CPUFREQ_RELATION_L);
ccb2fe20	312	}
1da177e4 LT	313	}
	314
	315	static void do_dbs_timer(void *data)
32ee8c3e	316	{
1da177e4	317	int i;
4ec223d0	318	lock_cpu_hotplug();
3fc54d37	319	mutex_lock(&dbs_mutex);
6fe71165 DJ	320	for_each_online_cpu(i)
6fe71165 DJ	321	dbs_check_cpu(i);
6810b548 AK	322	queue_delayed_work(dbs_workq, &dbs_work,
6810b548 AK	323	usecs_to_jiffies(dbs_tuners_ins.sampling_rate));
3fc54d37	324	mutex_unlock(&dbs_mutex);
4ec223d0	325	unlock_cpu_hotplug();
32ee8c3e	326	}
1da177e4 LT	327
	328	static inline void dbs_timer_init(void)
	329	{
	330	INIT_WORK(&dbs_work, do_dbs_timer, NULL);
6810b548 AK	331	if (!dbs_workq)
	332	dbs_workq = create_singlethread_workqueue("ondemand");
	333	if (!dbs_workq) {
	334	printk(KERN_ERR "ondemand: Cannot initialize kernel thread\n");
	335	return;
	336	}
	337	queue_delayed_work(dbs_workq, &dbs_work,
	338	usecs_to_jiffies(dbs_tuners_ins.sampling_rate));
1da177e4 LT	339	return;
	340	}
	341
	342	static inline void dbs_timer_exit(void)
	343	{
6810b548 AK	344	if (dbs_workq)
6810b548 AK	345	cancel_rearming_delayed_workqueue(dbs_workq, &dbs_work);
1da177e4 LT	346	}
	347
	348	static int cpufreq_governor_dbs(struct cpufreq_policy *policy,
	349	unsigned int event)
	350	{
	351	unsigned int cpu = policy->cpu;
	352	struct cpu_dbs_info_s *this_dbs_info;
	353	unsigned int j;
	354
	355	this_dbs_info = &per_cpu(cpu_dbs_info, cpu);
	356
	357	switch (event) {
	358	case CPUFREQ_GOV_START:
32ee8c3e	359	if ((!cpu_online(cpu)) \|\|
1da177e4 LT	360	(!policy->cur))
	361	return -EINVAL;
	362
	363	if (policy->cpuinfo.transition_latency >
ff8c288d EP	364	(TRANSITION_LATENCY_LIMIT * 1000)) {
	365	printk(KERN_WARNING "ondemand governor failed to load "
	366	"due to too long transition latency\n");
1da177e4	367	return -EINVAL;
ff8c288d	368	}
1da177e4 LT	369	if (this_dbs_info->enable) /* Already enabled */
1da177e4 LT	370	break;
32ee8c3e	371
3fc54d37	372	mutex_lock(&dbs_mutex);
1da177e4 LT	373	for_each_cpu_mask(j, policy->cpus) {
	374	struct cpu_dbs_info_s *j_dbs_info;
	375	j_dbs_info = &per_cpu(cpu_dbs_info, j);
	376	j_dbs_info->cur_policy = policy;
32ee8c3e	377
ccb2fe20 VP	378	j_dbs_info->prev_cpu_idle = get_cpu_idle_time(j);
ccb2fe20 VP	379	j_dbs_info->prev_cpu_wall = get_jiffies_64();
1da177e4 LT	380	}
	381	this_dbs_info->enable = 1;
	382	sysfs_create_group(&policy->kobj, &dbs_attr_group);
	383	dbs_enable++;
	384	/*
	385	* Start the timerschedule work, when this governor
	386	* is used for first time
	387	*/
	388	if (dbs_enable == 1) {
	389	unsigned int latency;
	390	/* policy latency is in nS. Convert it to uS first */
df8b59be DJ	391	latency = policy->cpuinfo.transition_latency / 1000;
	392	if (latency == 0)
	393	latency = 1;
1da177e4	394
df8b59be	395	def_sampling_rate = latency *
1da177e4	396	DEF_SAMPLING_RATE_LATENCY_MULTIPLIER;
df8b59be DJ	397
	398	if (def_sampling_rate < MIN_STAT_SAMPLING_RATE)
	399	def_sampling_rate = MIN_STAT_SAMPLING_RATE;
	400
1da177e4	401	dbs_tuners_ins.sampling_rate = def_sampling_rate;
1da177e4 LT	402	dbs_timer_init();
1da177e4 LT	403	}
32ee8c3e	404
3fc54d37	405	mutex_unlock(&dbs_mutex);
1da177e4 LT	406	break;
	407
	408	case CPUFREQ_GOV_STOP:
3fc54d37	409	mutex_lock(&dbs_mutex);
1da177e4 LT	410	this_dbs_info->enable = 0;
	411	sysfs_remove_group(&policy->kobj, &dbs_attr_group);
	412	dbs_enable--;
	413	/*
	414	* Stop the timerschedule work, when this governor
	415	* is used for first time
	416	*/
32ee8c3e	417	if (dbs_enable == 0)
1da177e4	418	dbs_timer_exit();
32ee8c3e	419
3fc54d37	420	mutex_unlock(&dbs_mutex);
1da177e4 LT	421
	422	break;
	423
	424	case CPUFREQ_GOV_LIMITS:
4ec223d0	425	lock_cpu_hotplug();
3fc54d37	426	mutex_lock(&dbs_mutex);
1da177e4 LT	427	if (policy->max < this_dbs_info->cur_policy->cur)
	428	__cpufreq_driver_target(
	429	this_dbs_info->cur_policy,
32ee8c3e	430	policy->max, CPUFREQ_RELATION_H);
1da177e4 LT	431	else if (policy->min > this_dbs_info->cur_policy->cur)
	432	__cpufreq_driver_target(
	433	this_dbs_info->cur_policy,
32ee8c3e	434	policy->min, CPUFREQ_RELATION_L);
3fc54d37	435	mutex_unlock(&dbs_mutex);
4ec223d0	436	unlock_cpu_hotplug();
1da177e4 LT	437	break;
	438	}
	439	return 0;
	440	}
	441
7f335d4e	442	static struct cpufreq_governor cpufreq_gov_dbs = {
1da177e4 LT	443	.name = "ondemand",
	444	.governor = cpufreq_governor_dbs,
	445	.owner = THIS_MODULE,
	446	};
1da177e4 LT	447
	448	static int __init cpufreq_gov_dbs_init(void)
	449	{
	450	return cpufreq_register_governor(&cpufreq_gov_dbs);
	451	}
	452
	453	static void __exit cpufreq_gov_dbs_exit(void)
	454	{
6810b548 AK	455	/* Make sure that the scheduled work is indeed not running.
	456	Assumes the timer has been cancelled first. */
	457	if (dbs_workq) {
	458	flush_workqueue(dbs_workq);
	459	destroy_workqueue(dbs_workq);
	460	}
1da177e4 LT	461
	462	cpufreq_unregister_governor(&cpufreq_gov_dbs);
	463	}
	464
	465
	466	MODULE_AUTHOR ("Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>");
	467	MODULE_DESCRIPTION ("'cpufreq_ondemand' - A dynamic cpufreq governor for "
	468	"Low Latency Frequency Transition capable processors");
	469	MODULE_LICENSE ("GPL");
	470
	471	module_init(cpufreq_gov_dbs_init);
	472	module_exit(cpufreq_gov_dbs_exit);