android_kernel_samsung_universal7885/drivers/cpufreq/cpufreq_ondemand_x.c

/*
 *  drivers/cpufreq/cpufreq_ondemand_x.c
 *
 *  Copyright (C)  2001 Russell King
 *            (C)  2003 Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>.
 *                      Jun Nakajima <jun.nakajima@intel.com>
 *            (c)  2013 The Linux Foundation. All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 */

#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/cpufreq.h>
#include <linux/cpu.h>
#include <linux/jiffies.h>
#include <linux/kernel_stat.h>
#include <linux/mutex.h>
#include <linux/hrtimer.h>
#include <linux/tick.h>
#include <linux/ktime.h>
#include <linux/sched.h>
#include <linux/workqueue.h>
#include <linux/slab.h>

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

/* User tunabble controls */
#define DEF_FREQUENCY_UP_THRESHOLD		(70)
#define ANY_CPU_DEF_FREQUENCY_UP_THRESHOLD	(70)
#define MULTI_CORE_DEF_FREQUENCY_UP_THRESHOLD	(70)
#define MICRO_FREQUENCY_UP_THRESHOLD		(95)

#define DEF_MIDDLE_GRID_STEP			(14)
#define DEF_HIGH_GRID_STEP			(20)
#define DEF_MIDDLE_GRID_LOAD			(55)
#define DEF_HIGH_GRID_LOAD			(79)

#define DEF_SAMPLING_DOWN_FACTOR		(1)
#define DEF_SAMPLING_RATE			(20000)

#define DEF_SYNC_FREQUENCY			(1267200)
#define DEF_OPTIMAL_FREQUENCY			(1574400)
#define DEF_OPTIMAL_MAX_FREQ			(1728000)

/* Kernel tunabble controls */
#define MICRO_FREQUENCY_MIN_SAMPLE_RATE		(10000)
#define MAX_SAMPLING_DOWN_FACTOR		(3)
#define MIN_FREQUENCY_UP_THRESHOLD		(11)
#define MAX_FREQUENCY_UP_THRESHOLD		(100)
#define MIN_FREQUENCY_DOWN_DIFFERENTIAL		(1)

/*
 * 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.
 */
#define MIN_SAMPLING_RATE_RATIO			(2)

static unsigned int min_sampling_rate;

#define LATENCY_MULTIPLIER			(1000)
#define MIN_LATENCY_MULTIPLIER			(100)
#define TRANSITION_LATENCY_LIMIT		(10 * 1000 * 1000)

static void do_dbs_timer(struct work_struct *work);

/* Sampling types */
enum {DBS_NORMAL_SAMPLE, DBS_SUB_SAMPLE};

struct cpu_dbs_info_s {
	u64 prev_cpu_idle;
	u64 prev_cpu_iowait;
	u64 prev_cpu_wall;
	u64 prev_cpu_nice;
	/*
	 * Used to keep track of load in the previous interval. However, when
	 * explicitly set to zero, it is used as a flag to ensure that we copy
	 * the previous load to the current interval only once, upon the first
	 * wake-up from idle.
	 */
	unsigned int prev_load;
	struct cpufreq_policy *cur_policy;
	struct delayed_work work;
	struct cpufreq_frequency_table *freq_table;
	unsigned int freq_lo;
	unsigned int freq_lo_jiffies;
	unsigned int freq_hi_jiffies;
	unsigned int rate_mult;
	unsigned int max_load;
	unsigned int cpu;
	unsigned int sample_type:1;
	/*
	 * percpu mutex that serializes governor limit change with
	 * do_dbs_timer invocation. We do not want do_dbs_timer to run
	 * when user is changing the governor or limits.
	 */
	struct mutex timer_mutex;

	wait_queue_head_t sync_wq;
};
static DEFINE_PER_CPU(struct cpu_dbs_info_s, od_cpu_dbs_info);

static inline void dbs_timer_init(struct cpu_dbs_info_s *dbs_info);
static inline void dbs_timer_exit(struct cpu_dbs_info_s *dbs_info);

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

/*
 * dbs_mutex protects dbs_enable and dbs_info during start/stop.
 */
static DEFINE_MUTEX(dbs_mutex);

static struct workqueue_struct *dbs_wq;

static struct dbs_tuners {
	unsigned int sampling_rate;
	unsigned int up_threshold;
	unsigned int up_threshold_multi_core;
	unsigned int optimal_freq;
	unsigned int up_threshold_any_cpu_load;
	unsigned int micro_freq_up_threshold;
	unsigned int sync_freq;
	unsigned int sampling_down_factor;
	unsigned int optimal_max_freq;
	unsigned int middle_grid_step;
	unsigned int high_grid_step;
	unsigned int middle_grid_load;
	unsigned int high_grid_load;
} dbs_tuners_ins = {
	.up_threshold_multi_core = MULTI_CORE_DEF_FREQUENCY_UP_THRESHOLD,
	.up_threshold = DEF_FREQUENCY_UP_THRESHOLD,
	.sampling_down_factor = DEF_SAMPLING_DOWN_FACTOR,
	.up_threshold_any_cpu_load = ANY_CPU_DEF_FREQUENCY_UP_THRESHOLD,
	.micro_freq_up_threshold = MICRO_FREQUENCY_UP_THRESHOLD,
	.middle_grid_step = DEF_MIDDLE_GRID_STEP,
	.high_grid_step = DEF_HIGH_GRID_STEP,
	.middle_grid_load = DEF_MIDDLE_GRID_LOAD,
	.high_grid_load = DEF_HIGH_GRID_LOAD,
	.sync_freq = DEF_SYNC_FREQUENCY,
	.optimal_freq = DEF_OPTIMAL_FREQUENCY,
	.optimal_max_freq = DEF_OPTIMAL_MAX_FREQ,
	.sampling_rate = DEF_SAMPLING_RATE,
};


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

static ssize_t show_sampling_rate_min(struct kobject *kobj,
				      struct attribute *attr, char *buf)
{
	return sprintf(buf, "%u\n", min_sampling_rate);
}

define_one_global_ro(sampling_rate_min);

/* cpufreq_ondemand_x Governor Tunables */
#define show_one(file_name, object)					\
static ssize_t show_##file_name						\
(struct kobject *kobj, struct attribute *attr, 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(up_threshold_multi_core, up_threshold_multi_core);
show_one(sampling_down_factor, sampling_down_factor);
show_one(optimal_freq, optimal_freq);
show_one(up_threshold_any_cpu_load, up_threshold_any_cpu_load);
show_one(micro_freq_up_threshold, micro_freq_up_threshold);
show_one(middle_grid_step, middle_grid_step);
show_one(high_grid_step, high_grid_step);
show_one(middle_grid_load, middle_grid_load);
show_one(high_grid_load, high_grid_load);
show_one(sync_freq, sync_freq);
show_one(optimal_max_freq, optimal_max_freq);

static ssize_t store_sampling_rate(struct kobject *a, struct attribute *b,
				   const char *buf, size_t count)
{
	unsigned int input;
	int ret = 0;
	int mpd = strcmp(current->comm, "mpdecision");

	if (mpd == 0)
		return ret;

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

	dbs_tuners_ins.sampling_rate = max(input, min_sampling_rate);
	return count;
}

static ssize_t store_sync_freq(struct kobject *a, struct attribute *b,
				   const char *buf, size_t count)
{
	unsigned int input;
	int ret = 0;
	int mpd = strcmp(current->comm, "mpdecision");

	if (mpd == 0)
		return ret;

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

	dbs_tuners_ins.sync_freq = input;
	return count;
}

static ssize_t store_optimal_freq(struct kobject *a, struct attribute *b,
				   const char *buf, size_t count)
{
	unsigned int input;
	int ret = 0;
	int mpd = strcmp(current->comm, "mpdecision");

	if (mpd == 0)
		return ret;

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

	dbs_tuners_ins.optimal_freq = input;
	return count;
}

static ssize_t store_up_threshold(struct kobject *a, struct attribute *b,
				  const char *buf, size_t count)
{
	unsigned int input;
	int ret;

	ret = sscanf(buf, "%u", &input);
	if (ret != 1 || input > MAX_FREQUENCY_UP_THRESHOLD ||
			input < MIN_FREQUENCY_UP_THRESHOLD) {
		return -EINVAL;
	}
	dbs_tuners_ins.up_threshold = input;
	return count;
}

static ssize_t store_up_threshold_multi_core(struct kobject *a,
			struct attribute *b, const char *buf, size_t count)
{
	unsigned int input;
	int ret;
	ret = sscanf(buf, "%u", &input);

	if (ret != 1 || input > MAX_FREQUENCY_UP_THRESHOLD ||
			input < MIN_FREQUENCY_UP_THRESHOLD) {
		return -EINVAL;
	}
	dbs_tuners_ins.up_threshold_multi_core = input;
	return count;
}

static ssize_t store_up_threshold_any_cpu_load(struct kobject *a,
			struct attribute *b, const char *buf, size_t count)
{
	unsigned int input;
	int ret;
	ret = sscanf(buf, "%u", &input);

	if (ret != 1 || input > MAX_FREQUENCY_UP_THRESHOLD ||
			input < MIN_FREQUENCY_UP_THRESHOLD) {
		return -EINVAL;
	}
	dbs_tuners_ins.up_threshold_any_cpu_load = input;
	return count;
}

static ssize_t store_middle_grid_step(struct kobject *a, struct attribute *b,
				   const char *buf, size_t count)
{
	unsigned int input;
	int ret;

	ret = sscanf(buf, "%u", &input);
	if (ret != 1)
		return -EINVAL;
	dbs_tuners_ins.middle_grid_step = input;
	return count;
}

static ssize_t store_high_grid_step(struct kobject *a, struct attribute *b,
				   const char *buf, size_t count)
{
	unsigned int input;
	int ret;

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

	return count;
}

static ssize_t store_optimal_max_freq(struct kobject *a, struct attribute *b,
				   const char *buf, size_t count)
{
	unsigned int input;
	int ret = 0;
	int mpd = strcmp(current->comm, "mpdecision");

	if (mpd == 0)
		return ret;

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

	dbs_tuners_ins.optimal_max_freq = input;

	return count;
}

static ssize_t store_middle_grid_load(struct kobject *a,
			struct attribute *b, const char *buf, size_t count)
{
	unsigned int input;
	int ret;

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

	return count;
}

static ssize_t store_high_grid_load(struct kobject *a,
			struct attribute *b, const char *buf, size_t count)
{
	unsigned int input;
	int ret;

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

	return count;
}

static ssize_t store_micro_freq_up_threshold(struct kobject *a,
			struct attribute *b, const char *buf, size_t count)
{
	unsigned int input;
	int ret;

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

	return count;
}

static ssize_t store_sampling_down_factor(struct kobject *a,
			struct attribute *b, const char *buf, size_t count)
{
	unsigned int input, j;
	int ret;

	ret = sscanf(buf, "%u", &input);
	if (ret != 1 || input > MAX_SAMPLING_DOWN_FACTOR || input < 1)
		return -EINVAL;
	dbs_tuners_ins.sampling_down_factor = input;

	/* Reset down sampling multiplier in case it was active */
	for_each_online_cpu(j) {
		struct cpu_dbs_info_s *dbs_info;
		dbs_info = &per_cpu(od_cpu_dbs_info, j);
		dbs_info->rate_mult = 1;
	}
	return count;
}

define_one_global_rw(sampling_rate);
define_one_global_rw(up_threshold);
define_one_global_rw(sampling_down_factor);
define_one_global_rw(up_threshold_multi_core);
define_one_global_rw(optimal_freq);
define_one_global_rw(up_threshold_any_cpu_load);
define_one_global_rw(micro_freq_up_threshold);
define_one_global_rw(sync_freq);
define_one_global_rw(optimal_max_freq);
define_one_global_rw(middle_grid_step);
define_one_global_rw(high_grid_step);
define_one_global_rw(middle_grid_load);
define_one_global_rw(high_grid_load);

static struct attribute *dbs_attributes[] = {
	&sampling_rate_min.attr,
	&sampling_rate.attr,
	&up_threshold.attr,
	&sampling_down_factor.attr,
	&up_threshold_multi_core.attr,
	&optimal_freq.attr,
	&optimal_max_freq.attr,
	&up_threshold_any_cpu_load.attr,
	&micro_freq_up_threshold.attr,
	&sync_freq.attr,
	&middle_grid_step.attr,
	&high_grid_step.attr,
	&middle_grid_load.attr,
	&high_grid_load.attr,
	NULL
};

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

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

static void dbs_freq_increase(struct cpufreq_policy *p, unsigned int freq)
{
	if (p->cur == p->max)
		return;

	__cpufreq_driver_target(p, freq, CPUFREQ_RELATION_L);
}

static void dbs_check_cpu(struct cpu_dbs_info_s *this_dbs_info)
{
	/* Extrapolated load of this CPU */
	unsigned int load_at_max_freq = 0;
	unsigned int max_load_freq;
	/* Current load across this CPU */
	unsigned int cur_load = 0;
	unsigned int max_load_other_cpu = 0;
	unsigned int sampling_rate;
	struct cpufreq_policy *policy;
	unsigned int j;

	sampling_rate = dbs_tuners_ins.sampling_rate * this_dbs_info->rate_mult;
	this_dbs_info->freq_lo = 0;

	policy = this_dbs_info->cur_policy;
	if (policy == NULL)
		return;

	/*
	 * 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 Absolute Load - in terms of freq */
	max_load_freq = 0;

	for_each_cpu(j, policy->cpus) {
		struct cpu_dbs_info_s *j_dbs_info;
		u64 cur_wall_time, cur_idle_time;
		unsigned int idle_time, wall_time;
		unsigned int load_freq;
		int freq_avg;

		j_dbs_info = &per_cpu(od_cpu_dbs_info, j);

		/*
		 * For the purpose of ondemand_x, waiting for disk IO is an
		 * indication that you're performance critical, and not that
		 * the system is actually idle. So subtract the iowait time
		 * from the cpu idle time.
		 */
		cur_idle_time = get_cpu_idle_time(j, &cur_wall_time, 0);

		wall_time = (unsigned int)
			(cur_wall_time - j_dbs_info->prev_cpu_wall);
		j_dbs_info->prev_cpu_wall = cur_wall_time;

		idle_time = (unsigned int)
			(cur_idle_time - j_dbs_info->prev_cpu_idle);
		j_dbs_info->prev_cpu_idle = cur_idle_time;

		if (unlikely(!wall_time || wall_time < idle_time))
			continue;

		/*
		 * If the CPU had gone completely idle, and a task just woke up
		 * on this CPU now, it would be unfair to calculate 'load' the
		 * usual way for this elapsed time-window, because it will show
		 * near-zero load, irrespective of how CPU intensive the new
		 * task is. This is undesirable for latency-sensitive bursty
		 * workloads.
		 *
		 * To avoid this, we reuse the 'load' from the previous
		 * time-window and give this task a chance to start with a
		 * reasonably high CPU frequency. (However, we shouldn't over-do
		 * this copy, lest we get stuck at a high load (high frequency)
		 * for too long, even when the current system load has actually
		 * dropped down. So we perform the copy only once, upon the
		 * first wake-up from idle.)
		 *
		 *
		 * Detecting this situation is easy: the governor's deferrable
		 * timer would not have fired during CPU-idle periods. Hence
		 * an unusually large 'wall_time' (as compared to the sampling
		 * rate) indicates this scenario.
		 *
		 * prev_load can be zero in two cases and we must recalculate it
		 * for both cases:
		 * - during long idle intervals
		 * - explicitly set to zero
		 */
		if (unlikely(wall_time > (2 * sampling_rate) &&
					j_dbs_info->prev_load)) {
			cur_load = j_dbs_info->prev_load;
			j_dbs_info->max_load = cur_load;
			/*
			 * Perform a destructive copy, to ensure that we copy
			 * the previous load only once, upon the first wake-up
			 * from idle.
			 */
			j_dbs_info->prev_load = 0;
		} else {
			cur_load = 100 * (wall_time - idle_time) / wall_time;
			j_dbs_info->max_load = max(cur_load, j_dbs_info->prev_load);
			j_dbs_info->prev_load = cur_load;
		}

//		freq_avg = __cpufreq_driver_getavg(policy, j);
		if (policy == NULL)
			return;
//		if (freq_avg <= 0)
//			freq_avg = policy->cur;

		load_freq = cur_load * freq_avg;
		if (load_freq > max_load_freq)
			max_load_freq = load_freq;
	}

	for_each_online_cpu(j) {
		struct cpu_dbs_info_s *j_dbs_info;
		j_dbs_info = &per_cpu(od_cpu_dbs_info, j);

		if (j == policy->cpu)
			continue;

		if (max_load_other_cpu < j_dbs_info->max_load)
			max_load_other_cpu = j_dbs_info->max_load;
	}

	/* calculate the scaled load across CPU */
	load_at_max_freq = (cur_load * policy->cur)/policy->max;

	cpufreq_notify_utilization(policy, load_at_max_freq);

	/* Check for frequency increase */
	if (max_load_freq > (dbs_tuners_ins.up_threshold * policy->cur)) {
		int freq_target, freq_div;
		freq_target = 0; freq_div = 0;

		if (load_at_max_freq > dbs_tuners_ins.high_grid_load) {
			freq_div = (policy->max * dbs_tuners_ins.high_grid_step) / 100;
			freq_target = min(policy->max, policy->cur + freq_div);
		} else if (load_at_max_freq > dbs_tuners_ins.middle_grid_load) {
			freq_div = (policy->max * dbs_tuners_ins.middle_grid_step) / 100;
			freq_target = min(policy->max, policy->cur + freq_div);
		} else {
			if (policy->max < dbs_tuners_ins.optimal_max_freq)
				freq_target = policy->max;
			else
				freq_target = dbs_tuners_ins.optimal_max_freq;
		}

		/* If switching to max speed, apply sampling_down_factor */
		if (policy->cur < policy->max)
			this_dbs_info->rate_mult =
				dbs_tuners_ins.sampling_down_factor;

		dbs_freq_increase(policy, freq_target);
		return;
	}

	if (num_online_cpus() > 1) {
		if (max_load_other_cpu >
				dbs_tuners_ins.up_threshold_any_cpu_load) {
			if (policy->cur < dbs_tuners_ins.sync_freq)
				dbs_freq_increase(policy,
						dbs_tuners_ins.sync_freq);
			return;
		}

		if (max_load_freq > (dbs_tuners_ins.up_threshold_multi_core *
								policy->cur)) {
			if (policy->cur < dbs_tuners_ins.optimal_freq)
				dbs_freq_increase(policy,
						dbs_tuners_ins.optimal_freq);
			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.
	 */
	if (max_load_freq <
	    (dbs_tuners_ins.up_threshold * policy->cur)) {
		unsigned int freq_next;
		freq_next = max_load_freq / dbs_tuners_ins.up_threshold;

		/* No longer fully busy, reset rate_mult */
		this_dbs_info->rate_mult = 1;

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

		if (num_online_cpus() > 1) {
			if (max_load_other_cpu >
					dbs_tuners_ins.up_threshold_multi_core
					&& freq_next <
					dbs_tuners_ins.sync_freq)
				freq_next = dbs_tuners_ins.sync_freq;

			if (max_load_freq >
					(dbs_tuners_ins.up_threshold_multi_core *
					policy->cur) &&
					freq_next < dbs_tuners_ins.optimal_freq)
				freq_next = dbs_tuners_ins.optimal_freq;

		}
		__cpufreq_driver_target(policy, freq_next,
				CPUFREQ_RELATION_L);
	}
}

static void do_dbs_timer(struct work_struct *work)
{
	struct cpu_dbs_info_s *dbs_info =
		container_of(work, struct cpu_dbs_info_s, work.work);
	int sample_type = dbs_info->sample_type;
	int delay;

	if (unlikely(!cpu_online(dbs_info->cpu) || !dbs_info->cur_policy))
		return;

	mutex_lock(&dbs_info->timer_mutex);

	/* Common NORMAL_SAMPLE setup */
	dbs_info->sample_type = DBS_NORMAL_SAMPLE;
	if (sample_type == DBS_NORMAL_SAMPLE) {
		dbs_check_cpu(dbs_info);
		if (dbs_info->freq_lo) {
			/* Setup timer for SUB_SAMPLE */
			dbs_info->sample_type = DBS_SUB_SAMPLE;
			delay = dbs_info->freq_hi_jiffies;
		} else {
			/* We want all CPUs to do sampling nearly on
			 * same jiffy
			 */
			delay = usecs_to_jiffies(dbs_tuners_ins.sampling_rate
				* dbs_info->rate_mult);

			if (num_online_cpus() > 1)
				delay -= jiffies % delay;
		}
	} else {
		__cpufreq_driver_target(dbs_info->cur_policy,
			dbs_info->freq_lo, CPUFREQ_RELATION_H);
		delay = dbs_info->freq_lo_jiffies;
	}
	queue_delayed_work_on(dbs_info->cpu, dbs_wq, &dbs_info->work, delay);
	mutex_unlock(&dbs_info->timer_mutex);
}

static inline void dbs_timer_init(struct cpu_dbs_info_s *dbs_info)
{
	/* We want all CPUs to do sampling nearly on same jiffy */
	int delay = usecs_to_jiffies(dbs_tuners_ins.sampling_rate);

	if (num_online_cpus() > 1)
		delay -= jiffies % delay;

	dbs_info->sample_type = DBS_NORMAL_SAMPLE;
	INIT_DEFERRABLE_WORK(&dbs_info->work, do_dbs_timer);
	queue_delayed_work_on(dbs_info->cpu, dbs_wq, &dbs_info->work, delay);
}

static inline void dbs_timer_exit(struct cpu_dbs_info_s *dbs_info)
{
	cancel_delayed_work_sync(&dbs_info->work);
}

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

	this_dbs_info = &per_cpu(od_cpu_dbs_info, cpu);

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

		mutex_lock(&dbs_mutex);

		dbs_enable++;
		for_each_cpu(j, policy->cpus) {
			struct cpu_dbs_info_s *j_dbs_info;
			unsigned int prev_load;

			j_dbs_info = &per_cpu(od_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, 0);

			prev_load = (unsigned int)
				(j_dbs_info->prev_cpu_wall - j_dbs_info->prev_cpu_idle);
			j_dbs_info->prev_load = 100 * prev_load /
				(unsigned int) j_dbs_info->prev_cpu_wall;
		}
		cpu = policy->cpu;
		this_dbs_info->cpu = cpu;
		this_dbs_info->rate_mult = 1;
		/*
		 * Start the timerschedule work, when this governor
		 * is used for first time
		 */
		if (dbs_enable == 1) {
			unsigned int latency;

			rc = sysfs_create_group(cpufreq_global_kobject,
						&dbs_attr_group);
			if (rc) {
				dbs_enable--;
				mutex_unlock(&dbs_mutex);
				return rc;
			}

			/* policy latency is in nS. Convert it to uS first */
			latency = policy->cpuinfo.transition_latency / 1000;
			if (latency == 0)
				latency = 1;
			/* Bring kernel and HW constraints together */
			min_sampling_rate = max(min_sampling_rate,
					MIN_LATENCY_MULTIPLIER * latency);
			if (latency != 1)
				dbs_tuners_ins.sampling_rate =
					max(dbs_tuners_ins.sampling_rate,
						latency * LATENCY_MULTIPLIER);

			if (dbs_tuners_ins.optimal_freq == 0)
				dbs_tuners_ins.optimal_freq = policy->cpuinfo.min_freq;

			if (dbs_tuners_ins.sync_freq == 0)
				dbs_tuners_ins.sync_freq = policy->cpuinfo.min_freq;
		}
		mutex_unlock(&dbs_mutex);

		dbs_timer_init(this_dbs_info);
		break;

	case CPUFREQ_GOV_STOP:
		dbs_timer_exit(this_dbs_info);

		mutex_lock(&dbs_mutex);

		dbs_enable--;

		/* If device is being removed, policy is no longer
		 * valid. */
		this_dbs_info->cur_policy = NULL;
		if (!dbs_enable)
			sysfs_remove_group(cpufreq_global_kobject,
					   &dbs_attr_group);

		mutex_unlock(&dbs_mutex);

		break;

	case CPUFREQ_GOV_LIMITS:
		/* If device is being removed, skip set limits */
		if (!this_dbs_info->cur_policy || !policy)
			break;
		mutex_lock(&this_dbs_info->timer_mutex);
		__cpufreq_driver_target(this_dbs_info->cur_policy,
				policy->cur, CPUFREQ_RELATION_L);
		dbs_check_cpu(this_dbs_info);
		mutex_unlock(&this_dbs_info->timer_mutex);
		break;
	}
	return 0;
}

#ifndef CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND_X
static
#endif
struct cpufreq_governor cpufreq_gov_ondemand_x = {
	.name					= "ondemand_x",
	.governor				= cpufreq_governor_dbs,
	.max_transition_latency	= TRANSITION_LATENCY_LIMIT,
	.owner					= THIS_MODULE,
};

static int __init cpufreq_gov_dbs_init(void)
{
	u64 idle_time;
	unsigned int i;
	int cpu = get_cpu();

	idle_time = get_cpu_idle_time_us(cpu, NULL);
	put_cpu();
	if (idle_time != -1ULL) {
		/* Idle micro accounting is supported. Use finer thresholds */
		dbs_tuners_ins.up_threshold = dbs_tuners_ins.micro_freq_up_threshold;
		/*
		 * In nohz/micro accounting case we set the minimum frequency
		 * not depending on HZ, but fixed (very low). The deferred
		 * timer might skip some samples if idle/sleeping as needed.
		*/
		min_sampling_rate = MICRO_FREQUENCY_MIN_SAMPLE_RATE;
	} else {
		/* For correct statistics, we need 10 ticks for each measure */
		min_sampling_rate =
			MIN_SAMPLING_RATE_RATIO * jiffies_to_usecs(10);
	}

	dbs_wq = alloc_workqueue("ondemand_x_dbs_wq", WQ_HIGHPRI, 0);
	if (!dbs_wq) {
		printk(KERN_ERR "Failed to create ondemand_x_dbs_wq workqueue\n");
		return -EFAULT;
	}
	for_each_possible_cpu(i) {
		struct cpu_dbs_info_s *this_dbs_info =
			&per_cpu(od_cpu_dbs_info, i);
		mutex_init(&this_dbs_info->timer_mutex);
		init_waitqueue_head(&this_dbs_info->sync_wq);
	}

	return cpufreq_register_governor(&cpufreq_gov_ondemand_x);
}

static void __exit cpufreq_gov_dbs_exit(void)
{
	unsigned int i;

	cpufreq_unregister_governor(&cpufreq_gov_ondemand_x);
	for_each_possible_cpu(i) {
		struct cpu_dbs_info_s *this_dbs_info =
			&per_cpu(od_cpu_dbs_info, i);
		mutex_destroy(&this_dbs_info->timer_mutex);
	}
	destroy_workqueue(dbs_wq);
}

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

#ifdef CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND_X
fs_initcall(cpufreq_gov_dbs_init);
#else
module_init(cpufreq_gov_dbs_init);
#endif
module_exit(cpufreq_gov_dbs_exit);