path: root/interactive_governor-linux4.4-20160501.patch

diff -Naur a/arch/x86/include/asm/idle.h b/arch/x86/include/asm/idle.h
--- a/arch/x86/include/asm/idle.h	2016-05-01 23:53:53.784424167 +1000
+++ b/arch/x86/include/asm/idle.h	2016-05-01 23:54:08.214652936 +1000
@@ -1,13 +1,6 @@
 #ifndef _ASM_X86_IDLE_H
 #define _ASM_X86_IDLE_H
 
-#define IDLE_START 1
-#define IDLE_END 2
-
-struct notifier_block;
-void idle_notifier_register(struct notifier_block *n);
-void idle_notifier_unregister(struct notifier_block *n);
-
 #ifdef CONFIG_X86_64
 void enter_idle(void);
 void exit_idle(void);
diff -Naur a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
--- a/arch/x86/kernel/process.c	2016-05-01 23:53:57.111143565 +1000
+++ b/arch/x86/kernel/process.c	2016-05-01 23:54:08.216319629 +1000
@@ -62,19 +62,6 @@
 
 #ifdef CONFIG_X86_64
 static DEFINE_PER_CPU(unsigned char, is_idle);
-static ATOMIC_NOTIFIER_HEAD(idle_notifier);
-
-void idle_notifier_register(struct notifier_block *n)
-{
-	atomic_notifier_chain_register(&idle_notifier, n);
-}
-EXPORT_SYMBOL_GPL(idle_notifier_register);
-
-void idle_notifier_unregister(struct notifier_block *n)
-{
-	atomic_notifier_chain_unregister(&idle_notifier, n);
-}
-EXPORT_SYMBOL_GPL(idle_notifier_unregister);
 #endif
 
 /*
@@ -251,14 +238,14 @@
 void enter_idle(void)
 {
 	this_cpu_write(is_idle, 1);
-	atomic_notifier_call_chain(&idle_notifier, IDLE_START, NULL);
+	idle_notifier_call_chain(IDLE_START);
 }
 
 static void __exit_idle(void)
 {
 	if (x86_test_and_clear_bit_percpu(0, is_idle) == 0)
 		return;
-	atomic_notifier_call_chain(&idle_notifier, IDLE_END, NULL);
+	idle_notifier_call_chain(IDLE_END);
 }
 
 /* Called from interrupts to signify idle end */
diff -Naur a/Documentation/cpu-freq/governors.txt b/Documentation/cpu-freq/governors.txt
--- a/Documentation/cpu-freq/governors.txt	2016-05-01 23:54:17.309797175 +1000
+++ b/Documentation/cpu-freq/governors.txt	2016-05-01 23:54:20.669850471 +1000
@@ -28,6 +28,7 @@
 2.3  Userspace
 2.4  Ondemand
 2.5  Conservative
+2.6  Interactive
 
 3.   The Governor Interface in the CPUfreq Core
 
@@ -218,6 +219,90 @@
 speed. Load for frequency increase is still evaluated every
 sampling rate.
 
+2.6 Interactive
+---------------
+
+The CPUfreq governor "interactive" is designed for latency-sensitive,
+interactive workloads. This governor sets the CPU speed depending on
+usage, similar to "ondemand" and "conservative" governors, but with a
+different set of configurable behaviors.
+
+The tuneable values for this governor are:
+
+target_loads: CPU load values used to adjust speed to influence the
+current CPU load toward that value.  In general, the lower the target
+load, the more often the governor will raise CPU speeds to bring load
+below the target.  The format is a single target load, optionally
+followed by pairs of CPU speeds and CPU loads to target at or above
+those speeds.  Colons can be used between the speeds and associated
+target loads for readability.  For example:
+
+   85 1000000:90 1700000:99
+
+targets CPU load 85% below speed 1GHz, 90% at or above 1GHz, until
+1.7GHz and above, at which load 99% is targeted.  If speeds are
+specified these must appear in ascending order.  Higher target load
+values are typically specified for higher speeds, that is, target load
+values also usually appear in an ascending order. The default is
+target load 90% for all speeds.
+
+min_sample_time: The minimum amount of time to spend at the current
+frequency before ramping down. Default is 80000 uS.
+
+hispeed_freq: An intermediate "hi speed" at which to initially ramp
+when CPU load hits the value specified in go_hispeed_load.  If load
+stays high for the amount of time specified in above_hispeed_delay,
+then speed may be bumped higher.  Default is the maximum speed
+allowed by the policy at governor initialization time.
+
+go_hispeed_load: The CPU load at which to ramp to hispeed_freq.
+Default is 99%.
+
+above_hispeed_delay: When speed is at or above hispeed_freq, wait for
+this long before raising speed in response to continued high load.
+The format is a single delay value, optionally followed by pairs of
+CPU speeds and the delay to use at or above those speeds.  Colons can
+be used between the speeds and associated delays for readability.  For
+example:
+
+   80000 1300000:200000 1500000:40000
+
+uses delay 80000 uS until CPU speed 1.3 GHz, at which speed delay
+200000 uS is used until speed 1.5 GHz, at which speed (and above)
+delay 40000 uS is used.  If speeds are specified these must appear in
+ascending order.  Default is 20000 uS.
+
+timer_rate: Sample rate for reevaluating CPU load when the CPU is not
+idle.  A deferrable timer is used, such that the CPU will not be woken
+from idle to service this timer until something else needs to run.
+(The maximum time to allow deferring this timer when not running at
+minimum speed is configurable via timer_slack.)  Default is 20000 uS.
+
+timer_slack: Maximum additional time to defer handling the governor
+sampling timer beyond timer_rate when running at speeds above the
+minimum.  For platforms that consume additional power at idle when
+CPUs are running at speeds greater than minimum, this places an upper
+bound on how long the timer will be deferred prior to re-evaluating
+load and dropping speed.  For example, if timer_rate is 20000uS and
+timer_slack is 10000uS then timers will be deferred for up to 30msec
+when not at lowest speed.  A value of -1 means defer timers
+indefinitely at all speeds.  Default is 80000 uS.
+
+boost: If non-zero, immediately boost speed of all CPUs to at least
+hispeed_freq until zero is written to this attribute.  If zero, allow
+CPU speeds to drop below hispeed_freq according to load as usual.
+Default is zero.
+
+boostpulse: On each write, immediately boost speed of all CPUs to
+hispeed_freq for at least the period of time specified by
+boostpulse_duration, after which speeds are allowed to drop below
+hispeed_freq according to load as usual.
+
+boostpulse_duration: Length of time to hold CPU speed at hispeed_freq
+on a write to boostpulse, before allowing speed to drop according to
+load as usual.  Default is 80000 uS.
+
+
 3. The Governor Interface in the CPUfreq Core
 =============================================
 
diff -Naur a/drivers/cpufreq/cpufreq_interactive.c b/drivers/cpufreq/cpufreq_interactive.c
--- a/drivers/cpufreq/cpufreq_interactive.c	1970-01-01 10:00:00.000000000 +1000
+++ b/drivers/cpufreq/cpufreq_interactive.c	2016-05-01 23:54:11.309702016 +1000
@@ -0,0 +1,1360 @@
+/*
+ * drivers/cpufreq/cpufreq_interactive.c
+ *
+ * Copyright (C) 2010 Google, Inc.
+ *
+ * This software is licensed under the terms of the GNU General Public
+ * License version 2, as published by the Free Software Foundation, and
+ * may be copied, distributed, and modified under those terms.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * Author: Mike Chan (mike@android.com)
+ *
+ */
+
+#include <linux/cpu.h>
+#include <linux/cpumask.h>
+#include <linux/cpufreq.h>
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/rwsem.h>
+#include <linux/sched.h>
+#include <linux/sched/rt.h>
+#include <linux/tick.h>
+#include <linux/time.h>
+#include <linux/timer.h>
+#include <linux/workqueue.h>
+#include <linux/kthread.h>
+#include <linux/slab.h>
+
+#define CREATE_TRACE_POINTS
+#include <trace/events/cpufreq_interactive.h>
+
+struct cpufreq_interactive_cpuinfo {
+	struct timer_list cpu_timer;
+	struct timer_list cpu_slack_timer;
+	spinlock_t load_lock; /* protects the next 4 fields */
+	u64 time_in_idle;
+	u64 time_in_idle_timestamp;
+	u64 cputime_speedadj;
+	u64 cputime_speedadj_timestamp;
+	struct cpufreq_policy *policy;
+	struct cpufreq_frequency_table *freq_table;
+	spinlock_t target_freq_lock; /*protects target freq */
+	unsigned int target_freq;
+	unsigned int floor_freq;
+	u64 pol_floor_val_time; /* policy floor_validate_time */
+	u64 loc_floor_val_time; /* per-cpu floor_validate_time */
+	u64 pol_hispeed_val_time; /* policy hispeed_validate_time */
+	u64 loc_hispeed_val_time; /* per-cpu hispeed_validate_time */
+	struct rw_semaphore enable_sem;
+	int governor_enabled;
+};
+
+static DEFINE_PER_CPU(struct cpufreq_interactive_cpuinfo, cpuinfo);
+
+/* realtime thread handles frequency scaling */
+static struct task_struct *speedchange_task;
+static cpumask_t speedchange_cpumask;
+static spinlock_t speedchange_cpumask_lock;
+static struct mutex gov_lock;
+
+/* Target load.  Lower values result in higher CPU speeds. */
+#define DEFAULT_TARGET_LOAD 90
+static unsigned int default_target_loads[] = {DEFAULT_TARGET_LOAD};
+
+#define DEFAULT_TIMER_RATE (20 * USEC_PER_MSEC)
+#define DEFAULT_ABOVE_HISPEED_DELAY DEFAULT_TIMER_RATE
+static unsigned int default_above_hispeed_delay[] = {
+	DEFAULT_ABOVE_HISPEED_DELAY };
+
+struct cpufreq_interactive_tunables {
+	int usage_count;
+	/* Hi speed to bump to from lo speed when load burst (default max) */
+	unsigned int hispeed_freq;
+	/* Go to hi speed when CPU load at or above this value. */
+#define DEFAULT_GO_HISPEED_LOAD 99
+	unsigned long go_hispeed_load;
+	/* Target load. Lower values result in higher CPU speeds. */
+	spinlock_t target_loads_lock;
+	unsigned int *target_loads;
+	int ntarget_loads;
+	/*
+	 * The minimum amount of time to spend at a frequency before we can ramp
+	 * down.
+	 */
+#define DEFAULT_MIN_SAMPLE_TIME (80 * USEC_PER_MSEC)
+	unsigned long min_sample_time;
+	/*
+	 * The sample rate of the timer used to increase frequency
+	 */
+	unsigned long timer_rate;
+	/*
+	 * Wait this long before raising speed above hispeed, by default a
+	 * single timer interval.
+	 */
+	spinlock_t above_hispeed_delay_lock;
+	unsigned int *above_hispeed_delay;
+	int nabove_hispeed_delay;
+	/* Non-zero means indefinite speed boost active */
+	int boost_val;
+	/* Duration of a boot pulse in usecs */
+	int boostpulse_duration_val;
+	/* End time of boost pulse in ktime converted to usecs */
+	u64 boostpulse_endtime;
+	bool boosted;
+	/*
+	 * Max additional time to wait in idle, beyond timer_rate, at speeds
+	 * above minimum before wakeup to reduce speed, or -1 if unnecessary.
+	 */
+#define DEFAULT_TIMER_SLACK (4 * DEFAULT_TIMER_RATE)
+	int timer_slack_val;
+	bool io_is_busy;
+};
+
+/*
+ * HACK: FIXME: Bring back cpufreq_{get,put}_global_kobject()
+ * definition removed by upstream commit 8eec1020f0c0 "cpufreq:
+ * create cpu/cpufreq at boot time" to fix build failures.
+ */
+static int cpufreq_global_kobject_usage;
+
+int cpufreq_get_global_kobject(void)
+{
+	if (!cpufreq_global_kobject_usage++)
+		return kobject_add(cpufreq_global_kobject,
+				&cpu_subsys.dev_root->kobj, "%s", "cpufreq");
+
+	return 0;
+}
+
+void cpufreq_put_global_kobject(void)
+{
+	if (!--cpufreq_global_kobject_usage)
+		kobject_del(cpufreq_global_kobject);
+}
+
+/* For cases where we have single governor instance for system */
+static struct cpufreq_interactive_tunables *common_tunables;
+
+static struct attribute_group *get_sysfs_attr(void);
+
+static void cpufreq_interactive_timer_resched(
+	struct cpufreq_interactive_cpuinfo *pcpu)
+{
+	struct cpufreq_interactive_tunables *tunables =
+		pcpu->policy->governor_data;
+	unsigned long expires;
+	unsigned long flags;
+
+	spin_lock_irqsave(&pcpu->load_lock, flags);
+	pcpu->time_in_idle =
+		get_cpu_idle_time(smp_processor_id(),
+				  &pcpu->time_in_idle_timestamp,
+				  tunables->io_is_busy);
+	pcpu->cputime_speedadj = 0;
+	pcpu->cputime_speedadj_timestamp = pcpu->time_in_idle_timestamp;
+	expires = jiffies + usecs_to_jiffies(tunables->timer_rate);
+	mod_timer_pinned(&pcpu->cpu_timer, expires);
+
+	if (tunables->timer_slack_val >= 0 &&
+	    pcpu->target_freq > pcpu->policy->min) {
+		expires += usecs_to_jiffies(tunables->timer_slack_val);
+		mod_timer_pinned(&pcpu->cpu_slack_timer, expires);
+	}
+
+	spin_unlock_irqrestore(&pcpu->load_lock, flags);
+}
+
+/* The caller shall take enable_sem write semaphore to avoid any timer race.
+ * The cpu_timer and cpu_slack_timer must be deactivated when calling this
+ * function.
+ */
+static void cpufreq_interactive_timer_start(
+	struct cpufreq_interactive_tunables *tunables, int cpu)
+{
+	struct cpufreq_interactive_cpuinfo *pcpu = &per_cpu(cpuinfo, cpu);
+	unsigned long expires = jiffies +
+		usecs_to_jiffies(tunables->timer_rate);
+	unsigned long flags;
+
+	pcpu->cpu_timer.expires = expires;
+	add_timer_on(&pcpu->cpu_timer, cpu);
+	if (tunables->timer_slack_val >= 0 &&
+	    pcpu->target_freq > pcpu->policy->min) {
+		expires += usecs_to_jiffies(tunables->timer_slack_val);
+		pcpu->cpu_slack_timer.expires = expires;
+		add_timer_on(&pcpu->cpu_slack_timer, cpu);
+	}
+
+	spin_lock_irqsave(&pcpu->load_lock, flags);
+	pcpu->time_in_idle =
+		get_cpu_idle_time(cpu, &pcpu->time_in_idle_timestamp,
+				  tunables->io_is_busy);
+	pcpu->cputime_speedadj = 0;
+	pcpu->cputime_speedadj_timestamp = pcpu->time_in_idle_timestamp;
+	spin_unlock_irqrestore(&pcpu->load_lock, flags);
+}
+
+static unsigned int freq_to_above_hispeed_delay(
+	struct cpufreq_interactive_tunables *tunables,
+	unsigned int freq)
+{
+	int i;
+	unsigned int ret;
+	unsigned long flags;
+
+	spin_lock_irqsave(&tunables->above_hispeed_delay_lock, flags);
+
+	for (i = 0; i < tunables->nabove_hispeed_delay - 1 &&
+			freq >= tunables->above_hispeed_delay[i+1]; i += 2)
+		;
+
+	ret = tunables->above_hispeed_delay[i];
+	spin_unlock_irqrestore(&tunables->above_hispeed_delay_lock, flags);
+	return ret;
+}
+
+static unsigned int freq_to_targetload(
+	struct cpufreq_interactive_tunables *tunables, unsigned int freq)
+{
+	int i;
+	unsigned int ret;
+	unsigned long flags;
+
+	spin_lock_irqsave(&tunables->target_loads_lock, flags);
+
+	for (i = 0; i < tunables->ntarget_loads - 1 &&
+		    freq >= tunables->target_loads[i+1]; i += 2)
+		;
+
+	ret = tunables->target_loads[i];
+	spin_unlock_irqrestore(&tunables->target_loads_lock, flags);
+	return ret;
+}
+
+/*
+ * If increasing frequencies never map to a lower target load then
+ * choose_freq() will find the minimum frequency that does not exceed its
+ * target load given the current load.
+ */
+static unsigned int choose_freq(struct cpufreq_interactive_cpuinfo *pcpu,
+		unsigned int loadadjfreq)
+{
+	unsigned int freq = pcpu->policy->cur;
+	unsigned int prevfreq, freqmin, freqmax;
+	unsigned int tl;
+	int index;
+
+	freqmin = 0;
+	freqmax = UINT_MAX;
+
+	do {
+		prevfreq = freq;
+		tl = freq_to_targetload(pcpu->policy->governor_data, freq);
+
+		/*
+		 * Find the lowest frequency where the computed load is less
+		 * than or equal to the target load.
+		 */
+
+		if (cpufreq_frequency_table_target(
+			    pcpu->policy, pcpu->freq_table, loadadjfreq / tl,
+			    CPUFREQ_RELATION_L, &index))
+			break;
+		freq = pcpu->freq_table[index].frequency;
+
+		if (freq > prevfreq) {
+			/* The previous frequency is too low. */
+			freqmin = prevfreq;
+
+			if (freq >= freqmax) {
+				/*
+				 * Find the highest frequency that is less
+				 * than freqmax.
+				 */
+				if (cpufreq_frequency_table_target(
+					    pcpu->policy, pcpu->freq_table,
+					    freqmax - 1, CPUFREQ_RELATION_H,
+					    &index))
+					break;
+				freq = pcpu->freq_table[index].frequency;
+
+				if (freq == freqmin) {
+					/*
+					 * The first frequency below freqmax
+					 * has already been found to be too
+					 * low.  freqmax is the lowest speed
+					 * we found that is fast enough.
+					 */
+					freq = freqmax;
+					break;
+				}
+			}
+		} else if (freq < prevfreq) {
+			/* The previous frequency is high enough. */
+			freqmax = prevfreq;
+
+			if (freq <= freqmin) {
+				/*
+				 * Find the lowest frequency that is higher
+				 * than freqmin.
+				 */
+				if (cpufreq_frequency_table_target(
+					    pcpu->policy, pcpu->freq_table,
+					    freqmin + 1, CPUFREQ_RELATION_L,
+					    &index))
+					break;
+				freq = pcpu->freq_table[index].frequency;
+
+				/*
+				 * If freqmax is the first frequency above
+				 * freqmin then we have already found that
+				 * this speed is fast enough.
+				 */
+				if (freq == freqmax)
+					break;
+			}
+		}
+
+		/* If same frequency chosen as previous then done. */
+	} while (freq != prevfreq);
+
+	return freq;
+}
+
+static u64 update_load(int cpu)
+{
+	struct cpufreq_interactive_cpuinfo *pcpu = &per_cpu(cpuinfo, cpu);
+	struct cpufreq_interactive_tunables *tunables =
+		pcpu->policy->governor_data;
+	u64 now;
+	u64 now_idle;
+	unsigned int delta_idle;
+	unsigned int delta_time;
+	u64 active_time;
+
+	now_idle = get_cpu_idle_time(cpu, &now, tunables->io_is_busy);
+	delta_idle = (unsigned int)(now_idle - pcpu->time_in_idle);
+	delta_time = (unsigned int)(now - pcpu->time_in_idle_timestamp);
+
+	if (delta_time <= delta_idle)
+		active_time = 0;
+	else
+		active_time = delta_time - delta_idle;
+
+	pcpu->cputime_speedadj += active_time * pcpu->policy->cur;
+
+	pcpu->time_in_idle = now_idle;
+	pcpu->time_in_idle_timestamp = now;
+	return now;
+}
+
+static void cpufreq_interactive_timer(unsigned long data)
+{
+	u64 now;
+	unsigned int delta_time;
+	u64 cputime_speedadj;
+	int cpu_load;
+	struct cpufreq_interactive_cpuinfo *pcpu =
+		&per_cpu(cpuinfo, data);
+	struct cpufreq_interactive_tunables *tunables =
+		pcpu->policy->governor_data;
+	unsigned int new_freq;
+	unsigned int loadadjfreq;
+	unsigned int index;
+	unsigned long flags;
+	u64 max_fvtime;
+
+	if (!down_read_trylock(&pcpu->enable_sem))
+		return;
+	if (!pcpu->governor_enabled)
+		goto exit;
+
+	spin_lock_irqsave(&pcpu->load_lock, flags);
+	now = update_load(data);
+	delta_time = (unsigned int)(now - pcpu->cputime_speedadj_timestamp);
+	cputime_speedadj = pcpu->cputime_speedadj;
+	spin_unlock_irqrestore(&pcpu->load_lock, flags);
+
+	if (WARN_ON_ONCE(!delta_time))
+		goto rearm;
+
+	spin_lock_irqsave(&pcpu->target_freq_lock, flags);
+	do_div(cputime_speedadj, delta_time);
+	loadadjfreq = (unsigned int)cputime_speedadj * 100;
+	cpu_load = loadadjfreq / pcpu->policy->cur;
+	tunables->boosted = tunables->boost_val || now < tunables->boostpulse_endtime;
+
+	if (cpu_load >= tunables->go_hispeed_load || tunables->boosted) {
+		if (pcpu->policy->cur < tunables->hispeed_freq) {
+			new_freq = tunables->hispeed_freq;
+		} else {
+			new_freq = choose_freq(pcpu, loadadjfreq);
+
+			if (new_freq < tunables->hispeed_freq)
+				new_freq = tunables->hispeed_freq;
+		}
+	} else {
+		new_freq = choose_freq(pcpu, loadadjfreq);
+		if (new_freq > tunables->hispeed_freq &&
+				pcpu->policy->cur < tunables->hispeed_freq)
+			new_freq = tunables->hispeed_freq;
+	}
+
+	if (pcpu->policy->cur >= tunables->hispeed_freq &&
+	    new_freq > pcpu->policy->cur &&
+	    now - pcpu->pol_hispeed_val_time <
+	    freq_to_above_hispeed_delay(tunables, pcpu->policy->cur)) {
+		trace_cpufreq_interactive_notyet(
+			data, cpu_load, pcpu->target_freq,
+			pcpu->policy->cur, new_freq);
+		spin_unlock_irqrestore(&pcpu->target_freq_lock, flags);
+		goto rearm;
+	}
+
+	pcpu->loc_hispeed_val_time = now;
+
+	if (cpufreq_frequency_table_target(pcpu->policy, pcpu->freq_table,
+					   new_freq, CPUFREQ_RELATION_L,
+					   &index)) {
+		spin_unlock_irqrestore(&pcpu->target_freq_lock, flags);
+		goto rearm;
+	}
+
+	new_freq = pcpu->freq_table[index].frequency;
+
+	/*
+	 * Do not scale below floor_freq unless we have been at or above the
+	 * floor frequency for the minimum sample time since last validated.
+	 */
+	max_fvtime = max(pcpu->pol_floor_val_time, pcpu->loc_floor_val_time);
+	if (new_freq < pcpu->floor_freq &&
+	    pcpu->target_freq >= pcpu->policy->cur) {
+		if (now - max_fvtime < tunables->min_sample_time) {
+			trace_cpufreq_interactive_notyet(
+				data, cpu_load, pcpu->target_freq,
+				pcpu->policy->cur, new_freq);
+			spin_unlock_irqrestore(&pcpu->target_freq_lock, flags);
+			goto rearm;
+		}
+	}
+
+	/*
+	 * Update the timestamp for checking whether speed has been held at
+	 * or above the selected frequency for a minimum of min_sample_time,
+	 * if not boosted to hispeed_freq.  If boosted to hispeed_freq then we
+	 * allow the speed to drop as soon as the boostpulse duration expires
+	 * (or the indefinite boost is turned off).
+	 */
+
+	if (!tunables->boosted || new_freq > tunables->hispeed_freq) {
+		pcpu->floor_freq = new_freq;
+		if (pcpu->target_freq >= pcpu->policy->cur ||
+		    new_freq >= pcpu->policy->cur)
+			pcpu->loc_floor_val_time = now;
+	}
+
+	if (pcpu->target_freq == new_freq &&
+			pcpu->target_freq <= pcpu->policy->cur) {
+		trace_cpufreq_interactive_already(
+			data, cpu_load, pcpu->target_freq,
+			pcpu->policy->cur, new_freq);
+		spin_unlock_irqrestore(&pcpu->target_freq_lock, flags);
+		goto rearm;
+	}
+
+	trace_cpufreq_interactive_target(data, cpu_load, pcpu->target_freq,
+					 pcpu->policy->cur, new_freq);
+
+	pcpu->target_freq = new_freq;
+	spin_unlock_irqrestore(&pcpu->target_freq_lock, flags);
+	spin_lock_irqsave(&speedchange_cpumask_lock, flags);
+	cpumask_set_cpu(data, &speedchange_cpumask);
+	spin_unlock_irqrestore(&speedchange_cpumask_lock, flags);
+	wake_up_process(speedchange_task);
+
+rearm:
+	if (!timer_pending(&pcpu->cpu_timer))
+		cpufreq_interactive_timer_resched(pcpu);
+
+exit:
+	up_read(&pcpu->enable_sem);
+	return;
+}
+
+static void cpufreq_interactive_idle_end(void)
+{
+	struct cpufreq_interactive_cpuinfo *pcpu =
+		&per_cpu(cpuinfo, smp_processor_id());
+
+	if (!down_read_trylock(&pcpu->enable_sem))
+		return;
+	if (!pcpu->governor_enabled) {
+		up_read(&pcpu->enable_sem);
+		return;
+	}
+
+	/* Arm the timer for 1-2 ticks later if not already. */
+	if (!timer_pending(&pcpu->cpu_timer)) {
+		cpufreq_interactive_timer_resched(pcpu);
+	} else if (time_after_eq(jiffies, pcpu->cpu_timer.expires)) {
+		del_timer(&pcpu->cpu_timer);
+		del_timer(&pcpu->cpu_slack_timer);
+		cpufreq_interactive_timer(smp_processor_id());
+	}
+
+	up_read(&pcpu->enable_sem);
+}
+
+static int cpufreq_interactive_speedchange_task(void *data)
+{
+	unsigned int cpu;
+	cpumask_t tmp_mask;
+	unsigned long flags;
+	struct cpufreq_interactive_cpuinfo *pcpu;
+
+	while (1) {
+		set_current_state(TASK_INTERRUPTIBLE);
+		spin_lock_irqsave(&speedchange_cpumask_lock, flags);
+
+		if (cpumask_empty(&speedchange_cpumask)) {
+			spin_unlock_irqrestore(&speedchange_cpumask_lock,
+					       flags);
+			schedule();
+
+			if (kthread_should_stop())
+				break;
+
+			spin_lock_irqsave(&speedchange_cpumask_lock, flags);
+		}
+
+		set_current_state(TASK_RUNNING);
+		tmp_mask = speedchange_cpumask;
+		cpumask_clear(&speedchange_cpumask);
+		spin_unlock_irqrestore(&speedchange_cpumask_lock, flags);
+
+		for_each_cpu(cpu, &tmp_mask) {
+			unsigned int j;
+			unsigned int max_freq = 0;
+			struct cpufreq_interactive_cpuinfo *pjcpu;
+			u64 hvt = ~0ULL, fvt = 0;
+
+			pcpu = &per_cpu(cpuinfo, cpu);
+			if (!down_read_trylock(&pcpu->enable_sem))
+				continue;
+			if (!pcpu->governor_enabled) {
+				up_read(&pcpu->enable_sem);
+				continue;
+			}
+
+			for_each_cpu(j, pcpu->policy->cpus) {
+				pjcpu = &per_cpu(cpuinfo, j);
+
+				fvt = max(fvt, pjcpu->loc_floor_val_time);
+				if (pjcpu->target_freq > max_freq) {
+					max_freq = pjcpu->target_freq;
+					hvt = pjcpu->loc_hispeed_val_time;
+				} else if (pjcpu->target_freq == max_freq) {
+					hvt = min(hvt, pjcpu->loc_hispeed_val_time);
+				}
+			}
+			for_each_cpu(j, pcpu->policy->cpus) {
+				pjcpu = &per_cpu(cpuinfo, j);
+				pjcpu->pol_floor_val_time = fvt;
+			}
+
+			if (max_freq != pcpu->policy->cur) {
+				__cpufreq_driver_target(pcpu->policy,
+							max_freq,
+							CPUFREQ_RELATION_H);
+				for_each_cpu(j, pcpu->policy->cpus) {
+					pjcpu = &per_cpu(cpuinfo, j);
+					pjcpu->pol_hispeed_val_time = hvt;
+				}
+			}
+			trace_cpufreq_interactive_setspeed(cpu,
+						     pcpu->target_freq,
+						     pcpu->policy->cur);
+
+			up_read(&pcpu->enable_sem);
+		}
+	}
+
+	return 0;
+}
+
+static void cpufreq_interactive_boost(struct cpufreq_interactive_tunables *tunables)
+{
+	int i;
+	int anyboost = 0;
+	unsigned long flags[2];
+	struct cpufreq_interactive_cpuinfo *pcpu;
+
+	tunables->boosted = true;
+
+	spin_lock_irqsave(&speedchange_cpumask_lock, flags[0]);
+
+	for_each_online_cpu(i) {
+		pcpu = &per_cpu(cpuinfo, i);
+		if (tunables != pcpu->policy->governor_data)
+			continue;
+
+		spin_lock_irqsave(&pcpu->target_freq_lock, flags[1]);
+		if (pcpu->target_freq < tunables->hispeed_freq) {
+			pcpu->target_freq = tunables->hispeed_freq;
+			cpumask_set_cpu(i, &speedchange_cpumask);
+			pcpu->pol_hispeed_val_time =
+				ktime_to_us(ktime_get());
+			anyboost = 1;
+		}
+		spin_unlock_irqrestore(&pcpu->target_freq_lock, flags[1]);
+	}
+
+	spin_unlock_irqrestore(&speedchange_cpumask_lock, flags[0]);
+
+	if (anyboost)
+		wake_up_process(speedchange_task);
+}
+
+static int cpufreq_interactive_notifier(
+	struct notifier_block *nb, unsigned long val, void *data)
+{
+	struct cpufreq_freqs *freq = data;
+	struct cpufreq_interactive_cpuinfo *pcpu;
+	int cpu;
+	unsigned long flags;
+
+	if (val == CPUFREQ_POSTCHANGE) {
+		pcpu = &per_cpu(cpuinfo, freq->cpu);
+		if (!down_read_trylock(&pcpu->enable_sem))
+			return 0;
+		if (!pcpu->governor_enabled) {
+			up_read(&pcpu->enable_sem);
+			return 0;
+		}
+
+		for_each_cpu(cpu, pcpu->policy->cpus) {
+			struct cpufreq_interactive_cpuinfo *pjcpu =
+				&per_cpu(cpuinfo, cpu);
+			if (cpu != freq->cpu) {
+				if (!down_read_trylock(&pjcpu->enable_sem))
+					continue;
+				if (!pjcpu->governor_enabled) {
+					up_read(&pjcpu->enable_sem);
+					continue;
+				}
+			}
+			spin_lock_irqsave(&pjcpu->load_lock, flags);
+			update_load(cpu);
+			spin_unlock_irqrestore(&pjcpu->load_lock, flags);
+			if (cpu != freq->cpu)
+				up_read(&pjcpu->enable_sem);
+		}
+
+		up_read(&pcpu->enable_sem);
+	}
+	return 0;
+}
+
+static struct notifier_block cpufreq_notifier_block = {
+	.notifier_call = cpufreq_interactive_notifier,
+};
+
+static unsigned int *get_tokenized_data(const char *buf, int *num_tokens)
+{
+	const char *cp;
+	int i;
+	int ntokens = 1;
+	unsigned int *tokenized_data;
+	int err = -EINVAL;
+
+	cp = buf;
+	while ((cp = strpbrk(cp + 1, " :")))
+		ntokens++;
+
+	if (!(ntokens & 0x1))
+		goto err;
+
+	tokenized_data = kmalloc(ntokens * sizeof(unsigned int), GFP_KERNEL);
+	if (!tokenized_data) {
+		err = -ENOMEM;
+		goto err;
+	}
+
+	cp = buf;
+	i = 0;
+	while (i < ntokens) {
+		if (sscanf(cp, "%u", &tokenized_data[i++]) != 1)
+			goto err_kfree;
+
+		cp = strpbrk(cp, " :");
+		if (!cp)
+			break;
+		cp++;
+	}
+
+	if (i != ntokens)
+		goto err_kfree;
+
+	*num_tokens = ntokens;
+	return tokenized_data;
+
+err_kfree:
+	kfree(tokenized_data);
+err:
+	return ERR_PTR(err);
+}
+
+static ssize_t show_target_loads(
+	struct cpufreq_interactive_tunables *tunables,
+	char *buf)
+{
+	int i;
+	ssize_t ret = 0;
+	unsigned long flags;
+
+	spin_lock_irqsave(&tunables->target_loads_lock, flags);
+
+	for (i = 0; i < tunables->ntarget_loads; i++)
+		ret += sprintf(buf + ret, "%u%s", tunables->target_loads[i],
+			       i & 0x1 ? ":" : " ");
+
+	sprintf(buf + ret - 1, "\n");
+	spin_unlock_irqrestore(&tunables->target_loads_lock, flags);
+	return ret;
+}
+
+static ssize_t store_target_loads(
+	struct cpufreq_interactive_tunables *tunables,
+	const char *buf, size_t count)
+{
+	int ntokens;
+	unsigned int *new_target_loads = NULL;
+	unsigned long flags;
+
+	new_target_loads = get_tokenized_data(buf, &ntokens);
+	if (IS_ERR(new_target_loads))
+		return PTR_RET(new_target_loads);
+
+	spin_lock_irqsave(&tunables->target_loads_lock, flags);
+	if (tunables->target_loads != default_target_loads)
+		kfree(tunables->target_loads);
+	tunables->target_loads = new_target_loads;
+	tunables->ntarget_loads = ntokens;
+	spin_unlock_irqrestore(&tunables->target_loads_lock, flags);
+	return count;
+}
+
+static ssize_t show_above_hispeed_delay(
+	struct cpufreq_interactive_tunables *tunables, char *buf)
+{
+	int i;
+	ssize_t ret = 0;
+	unsigned long flags;
+
+	spin_lock_irqsave(&tunables->above_hispeed_delay_lock, flags);
+
+	for (i = 0; i < tunables->nabove_hispeed_delay; i++)
+		ret += sprintf(buf + ret, "%u%s",
+			       tunables->above_hispeed_delay[i],
+			       i & 0x1 ? ":" : " ");
+
+	sprintf(buf + ret - 1, "\n");
+	spin_unlock_irqrestore(&tunables->above_hispeed_delay_lock, flags);
+	return ret;
+}
+
+static ssize_t store_above_hispeed_delay(
+	struct cpufreq_interactive_tunables *tunables,
+	const char *buf, size_t count)
+{
+	int ntokens;
+	unsigned int *new_above_hispeed_delay = NULL;
+	unsigned long flags;
+
+	new_above_hispeed_delay = get_tokenized_data(buf, &ntokens);
+	if (IS_ERR(new_above_hispeed_delay))
+		return PTR_RET(new_above_hispeed_delay);
+
+	spin_lock_irqsave(&tunables->above_hispeed_delay_lock, flags);
+	if (tunables->above_hispeed_delay != default_above_hispeed_delay)
+		kfree(tunables->above_hispeed_delay);
+	tunables->above_hispeed_delay = new_above_hispeed_delay;
+	tunables->nabove_hispeed_delay = ntokens;
+	spin_unlock_irqrestore(&tunables->above_hispeed_delay_lock, flags);
+	return count;
+
+}
+
+static ssize_t show_hispeed_freq(struct cpufreq_interactive_tunables *tunables,
+		char *buf)
+{
+	return sprintf(buf, "%u\n", tunables->hispeed_freq);
+}
+
+static ssize_t store_hispeed_freq(struct cpufreq_interactive_tunables *tunables,
+		const char *buf, size_t count)
+{
+	int ret;
+	long unsigned int val;
+
+	ret = kstrtoul(buf, 0, &val);
+	if (ret < 0)
+		return ret;
+	tunables->hispeed_freq = val;
+	return count;
+}
+
+static ssize_t show_go_hispeed_load(struct cpufreq_interactive_tunables
+		*tunables, char *buf)
+{
+	return sprintf(buf, "%lu\n", tunables->go_hispeed_load);
+}
+
+static ssize_t store_go_hispeed_load(struct cpufreq_interactive_tunables
+		*tunables, const char *buf, size_t count)
+{
+	int ret;
+	unsigned long val;
+
+	ret = kstrtoul(buf, 0, &val);
+	if (ret < 0)
+		return ret;
+	tunables->go_hispeed_load = val;
+	return count;
+}
+
+static ssize_t show_min_sample_time(struct cpufreq_interactive_tunables
+		*tunables, char *buf)
+{
+	return sprintf(buf, "%lu\n", tunables->min_sample_time);
+}
+
+static ssize_t store_min_sample_time(struct cpufreq_interactive_tunables
+		*tunables, const char *buf, size_t count)
+{
+	int ret;
+	unsigned long val;
+
+	ret = kstrtoul(buf, 0, &val);
+	if (ret < 0)
+		return ret;
+	tunables->min_sample_time = val;
+	return count;
+}
+
+static ssize_t show_timer_rate(struct cpufreq_interactive_tunables *tunables,
+		char *buf)
+{
+	return sprintf(buf, "%lu\n", tunables->timer_rate);
+}
+
+static ssize_t store_timer_rate(struct cpufreq_interactive_tunables *tunables,
+		const char *buf, size_t count)
+{
+	int ret;
+	unsigned long val, val_round;
+
+	ret = kstrtoul(buf, 0, &val);
+	if (ret < 0)
+		return ret;
+
+	val_round = jiffies_to_usecs(usecs_to_jiffies(val));
+	if (val != val_round)
+		pr_warn("timer_rate not aligned to jiffy. Rounded up to %lu\n",
+			val_round);
+
+	tunables->timer_rate = val_round;
+	return count;
+}
+
+static ssize_t show_timer_slack(struct cpufreq_interactive_tunables *tunables,
+		char *buf)
+{
+	return sprintf(buf, "%d\n", tunables->timer_slack_val);
+}
+
+static ssize_t store_timer_slack(struct cpufreq_interactive_tunables *tunables,
+		const char *buf, size_t count)
+{
+	int ret;
+	unsigned long val;
+
+	ret = kstrtol(buf, 10, &val);
+	if (ret < 0)
+		return ret;
+
+	tunables->timer_slack_val = val;
+	return count;
+}
+
+static ssize_t show_boost(struct cpufreq_interactive_tunables *tunables,
+			  char *buf)
+{
+	return sprintf(buf, "%d\n", tunables->boost_val);
+}
+
+static ssize_t store_boost(struct cpufreq_interactive_tunables *tunables,
+			   const char *buf, size_t count)
+{
+	int ret;
+	unsigned long val;
+
+	ret = kstrtoul(buf, 0, &val);
+	if (ret < 0)
+		return ret;
+
+	tunables->boost_val = val;
+
+	if (tunables->boost_val) {
+		trace_cpufreq_interactive_boost("on");
+		if (!tunables->boosted)
+			cpufreq_interactive_boost(tunables);
+	} else {
+		tunables->boostpulse_endtime = ktime_to_us(ktime_get());
+		trace_cpufreq_interactive_unboost("off");
+	}
+
+	return count;
+}
+
+static ssize_t store_boostpulse(struct cpufreq_interactive_tunables *tunables,
+				const char *buf, size_t count)
+{
+	int ret;
+	unsigned long val;
+
+	ret = kstrtoul(buf, 0, &val);
+	if (ret < 0)
+		return ret;
+
+	tunables->boostpulse_endtime = ktime_to_us(ktime_get()) +
+		tunables->boostpulse_duration_val;
+	trace_cpufreq_interactive_boost("pulse");
+	if (!tunables->boosted)
+		cpufreq_interactive_boost(tunables);
+	return count;
+}
+
+static ssize_t show_boostpulse_duration(struct cpufreq_interactive_tunables
+		*tunables, char *buf)
+{
+	return sprintf(buf, "%d\n", tunables->boostpulse_duration_val);
+}
+
+static ssize_t store_boostpulse_duration(struct cpufreq_interactive_tunables
+		*tunables, const char *buf, size_t count)
+{
+	int ret;
+	unsigned long val;
+
+	ret = kstrtoul(buf, 0, &val);
+	if (ret < 0)
+		return ret;
+
+	tunables->boostpulse_duration_val = val;
+	return count;
+}
+
+static ssize_t show_io_is_busy(struct cpufreq_interactive_tunables *tunables,
+		char *buf)
+{
+	return sprintf(buf, "%u\n", tunables->io_is_busy);
+}
+
+static ssize_t store_io_is_busy(struct cpufreq_interactive_tunables *tunables,
+		const char *buf, size_t count)
+{
+	int ret;
+	unsigned long val;
+
+	ret = kstrtoul(buf, 0, &val);
+	if (ret < 0)
+		return ret;
+	tunables->io_is_busy = val;
+	return count;
+}
+
+/*
+ * Create show/store routines
+ * - sys: One governor instance for complete SYSTEM
+ * - pol: One governor instance per struct cpufreq_policy
+ */
+#define show_gov_pol_sys(file_name)					\
+static ssize_t show_##file_name##_gov_sys				\
+(struct kobject *kobj, struct attribute *attr, char *buf)		\
+{									\
+	return show_##file_name(common_tunables, buf);			\
+}									\
+									\
+static ssize_t show_##file_name##_gov_pol				\
+(struct cpufreq_policy *policy, char *buf)				\
+{									\
+	return show_##file_name(policy->governor_data, buf);		\
+}
+
+#define store_gov_pol_sys(file_name)					\
+static ssize_t store_##file_name##_gov_sys				\
+(struct kobject *kobj, struct attribute *attr, const char *buf,		\
+	size_t count)							\
+{									\
+	return store_##file_name(common_tunables, buf, count);		\
+}									\
+									\
+static ssize_t store_##file_name##_gov_pol				\
+(struct cpufreq_policy *policy, const char *buf, size_t count)		\
+{									\
+	return store_##file_name(policy->governor_data, buf, count);	\
+}
+
+#define show_store_gov_pol_sys(file_name)				\
+show_gov_pol_sys(file_name);						\
+store_gov_pol_sys(file_name)
+
+show_store_gov_pol_sys(target_loads);
+show_store_gov_pol_sys(above_hispeed_delay);
+show_store_gov_pol_sys(hispeed_freq);
+show_store_gov_pol_sys(go_hispeed_load);
+show_store_gov_pol_sys(min_sample_time);
+show_store_gov_pol_sys(timer_rate);
+show_store_gov_pol_sys(timer_slack);
+show_store_gov_pol_sys(boost);
+store_gov_pol_sys(boostpulse);
+show_store_gov_pol_sys(boostpulse_duration);
+show_store_gov_pol_sys(io_is_busy);
+
+#define gov_sys_attr_rw(_name)						\
+static struct global_attr _name##_gov_sys =				\
+__ATTR(_name, 0644, show_##_name##_gov_sys, store_##_name##_gov_sys)
+
+#define gov_pol_attr_rw(_name)						\
+static struct freq_attr _name##_gov_pol =				\
+__ATTR(_name, 0644, show_##_name##_gov_pol, store_##_name##_gov_pol)
+
+#define gov_sys_pol_attr_rw(_name)					\
+	gov_sys_attr_rw(_name);						\
+	gov_pol_attr_rw(_name)
+
+gov_sys_pol_attr_rw(target_loads);
+gov_sys_pol_attr_rw(above_hispeed_delay);
+gov_sys_pol_attr_rw(hispeed_freq);
+gov_sys_pol_attr_rw(go_hispeed_load);
+gov_sys_pol_attr_rw(min_sample_time);
+gov_sys_pol_attr_rw(timer_rate);
+gov_sys_pol_attr_rw(timer_slack);
+gov_sys_pol_attr_rw(boost);
+gov_sys_pol_attr_rw(boostpulse_duration);
+gov_sys_pol_attr_rw(io_is_busy);
+
+static struct global_attr boostpulse_gov_sys =
+	__ATTR(boostpulse, 0200, NULL, store_boostpulse_gov_sys);
+
+static struct freq_attr boostpulse_gov_pol =
+	__ATTR(boostpulse, 0200, NULL, store_boostpulse_gov_pol);
+
+/* One Governor instance for entire system */
+static struct attribute *interactive_attributes_gov_sys[] = {
+	&target_loads_gov_sys.attr,
+	&above_hispeed_delay_gov_sys.attr,
+	&hispeed_freq_gov_sys.attr,
+	&go_hispeed_load_gov_sys.attr,
+	&min_sample_time_gov_sys.attr,
+	&timer_rate_gov_sys.attr,
+	&timer_slack_gov_sys.attr,
+	&boost_gov_sys.attr,
+	&boostpulse_gov_sys.attr,
+	&boostpulse_duration_gov_sys.attr,
+	&io_is_busy_gov_sys.attr,
+	NULL,
+};
+
+static struct attribute_group interactive_attr_group_gov_sys = {
+	.attrs = interactive_attributes_gov_sys,
+	.name = "interactive",
+};
+
+/* Per policy governor instance */
+static struct attribute *interactive_attributes_gov_pol[] = {
+	&target_loads_gov_pol.attr,
+	&above_hispeed_delay_gov_pol.attr,
+	&hispeed_freq_gov_pol.attr,
+	&go_hispeed_load_gov_pol.attr,
+	&min_sample_time_gov_pol.attr,
+	&timer_rate_gov_pol.attr,
+	&timer_slack_gov_pol.attr,
+	&boost_gov_pol.attr,
+	&boostpulse_gov_pol.attr,
+	&boostpulse_duration_gov_pol.attr,
+	&io_is_busy_gov_pol.attr,
+	NULL,
+};
+
+static struct attribute_group interactive_attr_group_gov_pol = {
+	.attrs = interactive_attributes_gov_pol,
+	.name = "interactive",
+};
+
+static struct attribute_group *get_sysfs_attr(void)
+{
+	if (have_governor_per_policy())
+		return &interactive_attr_group_gov_pol;
+	else
+		return &interactive_attr_group_gov_sys;
+}
+
+static int cpufreq_interactive_idle_notifier(struct notifier_block *nb,
+					     unsigned long val,
+					     void *data)
+{
+	if (val == IDLE_END)
+		cpufreq_interactive_idle_end();
+
+	return 0;
+}
+
+static struct notifier_block cpufreq_interactive_idle_nb = {
+	.notifier_call = cpufreq_interactive_idle_notifier,
+};
+
+static int cpufreq_governor_interactive(struct cpufreq_policy *policy,
+		unsigned int event)
+{
+	int rc;
+	unsigned int j;
+	struct cpufreq_interactive_cpuinfo *pcpu;
+	struct cpufreq_frequency_table *freq_table;
+	struct cpufreq_interactive_tunables *tunables;
+	unsigned long flags;
+
+	if (have_governor_per_policy())
+		tunables = policy->governor_data;
+	else
+		tunables = common_tunables;
+
+	WARN_ON(!tunables && (event != CPUFREQ_GOV_POLICY_INIT));
+
+	switch (event) {
+	case CPUFREQ_GOV_POLICY_INIT:
+		if (have_governor_per_policy()) {
+			WARN_ON(tunables);
+		} else if (tunables) {
+			tunables->usage_count++;
+			policy->governor_data = tunables;
+			return 0;
+		}
+
+		tunables = kzalloc(sizeof(*tunables), GFP_KERNEL);
+		if (!tunables) {
+			pr_err("%s: POLICY_INIT: kzalloc failed\n", __func__);
+			return -ENOMEM;
+		}
+
+		tunables->usage_count = 1;
+		tunables->above_hispeed_delay = default_above_hispeed_delay;
+		tunables->nabove_hispeed_delay =
+			ARRAY_SIZE(default_above_hispeed_delay);
+		tunables->go_hispeed_load = DEFAULT_GO_HISPEED_LOAD;
+		tunables->target_loads = default_target_loads;
+		tunables->ntarget_loads = ARRAY_SIZE(default_target_loads);
+		tunables->min_sample_time = DEFAULT_MIN_SAMPLE_TIME;
+		tunables->timer_rate = DEFAULT_TIMER_RATE;
+		tunables->boostpulse_duration_val = DEFAULT_MIN_SAMPLE_TIME;
+		tunables->timer_slack_val = DEFAULT_TIMER_SLACK;
+
+		spin_lock_init(&tunables->target_loads_lock);
+		spin_lock_init(&tunables->above_hispeed_delay_lock);
+
+		policy->governor_data = tunables;
+		if (!have_governor_per_policy()) {
+			common_tunables = tunables;
+			WARN_ON(cpufreq_get_global_kobject());
+		}
+
+		rc = sysfs_create_group(get_governor_parent_kobj(policy),
+				get_sysfs_attr());
+		if (rc) {
+			kfree(tunables);
+			policy->governor_data = NULL;
+			if (!have_governor_per_policy()) {
+				common_tunables = NULL;
+				cpufreq_put_global_kobject();
+			}
+			return rc;
+		}
+
+		if (!policy->governor->initialized) {
+			idle_notifier_register(&cpufreq_interactive_idle_nb);
+			cpufreq_register_notifier(&cpufreq_notifier_block,
+					CPUFREQ_TRANSITION_NOTIFIER);
+		}
+
+		break;
+
+	case CPUFREQ_GOV_POLICY_EXIT:
+		if (!--tunables->usage_count) {
+			if (policy->governor->initialized == 1) {
+				cpufreq_unregister_notifier(&cpufreq_notifier_block,
+						CPUFREQ_TRANSITION_NOTIFIER);
+				idle_notifier_unregister(&cpufreq_interactive_idle_nb);
+			}
+
+			sysfs_remove_group(get_governor_parent_kobj(policy),
+					get_sysfs_attr());
+
+			if (!have_governor_per_policy())
+				cpufreq_put_global_kobject();
+
+			kfree(tunables);
+			common_tunables = NULL;
+		}
+
+		policy->governor_data = NULL;
+		break;
+
+	case CPUFREQ_GOV_START:
+		mutex_lock(&gov_lock);
+
+		freq_table = cpufreq_frequency_get_table(policy->cpu);
+		if (!tunables->hispeed_freq)
+			tunables->hispeed_freq = policy->max;
+
+		for_each_cpu(j, policy->cpus) {
+			pcpu = &per_cpu(cpuinfo, j);
+			pcpu->policy = policy;
+			pcpu->target_freq = policy->cur;
+			pcpu->freq_table = freq_table;
+			pcpu->floor_freq = pcpu->target_freq;
+			pcpu->pol_floor_val_time =
+				ktime_to_us(ktime_get());
+			pcpu->loc_floor_val_time = pcpu->pol_floor_val_time;
+			pcpu->pol_hispeed_val_time = pcpu->pol_floor_val_time;
+			pcpu->loc_hispeed_val_time = pcpu->pol_floor_val_time;
+			down_write(&pcpu->enable_sem);
+			del_timer_sync(&pcpu->cpu_timer);
+			del_timer_sync(&pcpu->cpu_slack_timer);
+			cpufreq_interactive_timer_start(tunables, j);
+			pcpu->governor_enabled = 1;
+			up_write(&pcpu->enable_sem);
+		}
+
+		mutex_unlock(&gov_lock);
+		break;
+
+	case CPUFREQ_GOV_STOP:
+		mutex_lock(&gov_lock);
+		for_each_cpu(j, policy->cpus) {
+			pcpu = &per_cpu(cpuinfo, j);
+			down_write(&pcpu->enable_sem);
+			pcpu->governor_enabled = 0;
+			del_timer_sync(&pcpu->cpu_timer);
+			del_timer_sync(&pcpu->cpu_slack_timer);
+			up_write(&pcpu->enable_sem);
+		}
+
+		mutex_unlock(&gov_lock);
+		break;
+
+	case CPUFREQ_GOV_LIMITS:
+		if (policy->max < policy->cur)
+			__cpufreq_driver_target(policy,
+					policy->max, CPUFREQ_RELATION_H);
+		else if (policy->min > policy->cur)
+			__cpufreq_driver_target(policy,
+					policy->min, CPUFREQ_RELATION_L);
+		for_each_cpu(j, policy->cpus) {
+			pcpu = &per_cpu(cpuinfo, j);
+
+			down_read(&pcpu->enable_sem);
+			if (pcpu->governor_enabled == 0) {
+				up_read(&pcpu->enable_sem);
+				continue;
+			}
+
+			spin_lock_irqsave(&pcpu->target_freq_lock, flags);
+			if (policy->max < pcpu->target_freq)
+				pcpu->target_freq = policy->max;
+			else if (policy->min > pcpu->target_freq)
+				pcpu->target_freq = policy->min;
+
+			spin_unlock_irqrestore(&pcpu->target_freq_lock, flags);
+			up_read(&pcpu->enable_sem);
+		}
+		break;
+	}
+	return 0;
+}
+
+#ifndef CONFIG_CPU_FREQ_DEFAULT_GOV_INTERACTIVE
+static
+#endif
+struct cpufreq_governor cpufreq_gov_interactive = {
+	.name = "interactive",
+	.governor = cpufreq_governor_interactive,
+	.max_transition_latency = 10000000,
+	.owner = THIS_MODULE,
+};
+
+static void cpufreq_interactive_nop_timer(unsigned long data)
+{
+}
+
+static int __init cpufreq_interactive_init(void)
+{
+	unsigned int i;
+	struct cpufreq_interactive_cpuinfo *pcpu;
+	struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 };
+
+	/* Initalize per-cpu timers */
+	for_each_possible_cpu(i) {
+		pcpu = &per_cpu(cpuinfo, i);
+		init_timer_deferrable(&pcpu->cpu_timer);
+		pcpu->cpu_timer.function = cpufreq_interactive_timer;
+		pcpu->cpu_timer.data = i;
+		init_timer(&pcpu->cpu_slack_timer);
+		pcpu->cpu_slack_timer.function = cpufreq_interactive_nop_timer;
+		spin_lock_init(&pcpu->load_lock);
+		spin_lock_init(&pcpu->target_freq_lock);
+		init_rwsem(&pcpu->enable_sem);
+	}
+
+	spin_lock_init(&speedchange_cpumask_lock);
+	mutex_init(&gov_lock);
+	speedchange_task =
+		kthread_create(cpufreq_interactive_speedchange_task, NULL,
+			       "cfinteractive");
+	if (IS_ERR(speedchange_task))
+		return PTR_ERR(speedchange_task);
+
+	sched_setscheduler_nocheck(speedchange_task, SCHED_FIFO, &param);
+	get_task_struct(speedchange_task);
+
+	/* NB: wake up so the thread does not look hung to the freezer */
+	wake_up_process(speedchange_task);
+
+	return cpufreq_register_governor(&cpufreq_gov_interactive);
+}
+
+#ifdef CONFIG_CPU_FREQ_DEFAULT_GOV_INTERACTIVE
+fs_initcall(cpufreq_interactive_init);
+#else
+module_init(cpufreq_interactive_init);
+#endif
+
+static void __exit cpufreq_interactive_exit(void)
+{
+	cpufreq_unregister_governor(&cpufreq_gov_interactive);
+	kthread_stop(speedchange_task);
+	put_task_struct(speedchange_task);
+}
+
+module_exit(cpufreq_interactive_exit);
+
+MODULE_AUTHOR("Mike Chan <mike@android.com>");
+MODULE_DESCRIPTION("'cpufreq_interactive' - A cpufreq governor for "
+	"Latency sensitive workloads");
+MODULE_LICENSE("GPL");
diff -Naur a/drivers/cpufreq/Kconfig b/drivers/cpufreq/Kconfig
--- a/drivers/cpufreq/Kconfig	2016-05-01 23:54:23.733232399 +1000
+++ b/drivers/cpufreq/Kconfig	2016-05-01 23:54:24.938251516 +1000
@@ -102,6 +102,16 @@
 	  Be aware that not all cpufreq drivers support the conservative
 	  governor. If unsure have a look at the help section of the
 	  driver. Fallback governor will be the performance governor.
+
+config CPU_FREQ_DEFAULT_GOV_INTERACTIVE
+	bool "interactive"
+	select CPU_FREQ_GOV_INTERACTIVE
+	help
+	  Use the CPUFreq governor 'interactive' as default. This allows
+	  you to get a full dynamic cpu frequency capable system by simply
+	  loading your cpufreq low-level hardware driver, using the
+	  'interactive' governor for latency-sensitive workloads.
+
 endchoice
 
 config CPU_FREQ_GOV_PERFORMANCE
@@ -157,6 +167,20 @@
 
 	  For details, take a look at linux/Documentation/cpu-freq.
 
+	  If in doubt, say N.
+
+config CPU_FREQ_GOV_INTERACTIVE
+	bool "'interactive' cpufreq policy governor"
+	help
+	  'interactive' - This driver adds a dynamic cpufreq policy governor
+	  designed for latency-sensitive workloads.
+
+	  This governor attempts to reduce the latency of clock
+	  increases so that the system is more responsive to
+	  interactive workloads.
+
+	  For details, take a look at linux/Documentation/cpu-freq.
+
 	  If in doubt, say N.
 
 config CPU_FREQ_GOV_CONSERVATIVE
diff -Naur a/drivers/cpufreq/Makefile b/drivers/cpufreq/Makefile
--- a/drivers/cpufreq/Makefile	2016-05-01 23:54:27.993299985 +1000
+++ b/drivers/cpufreq/Makefile	2016-05-01 23:54:30.633341873 +1000
@@ -10,6 +10,7 @@
 obj-$(CONFIG_CPU_FREQ_GOV_USERSPACE)	+= cpufreq_userspace.o
 obj-$(CONFIG_CPU_FREQ_GOV_ONDEMAND)	+= cpufreq_ondemand.o
 obj-$(CONFIG_CPU_FREQ_GOV_CONSERVATIVE)	+= cpufreq_conservative.o
+obj-$(CONFIG_CPU_FREQ_GOV_INTERACTIVE)	+= cpufreq_interactive.o
 obj-$(CONFIG_CPU_FREQ_GOV_COMMON)		+= cpufreq_governor.o
 
 obj-$(CONFIG_CPUFREQ_DT)		+= cpufreq-dt.o
diff -Naur a/include/linux/cpufreq.h b/include/linux/cpufreq.h
--- a/include/linux/cpufreq.h	2016-05-01 23:54:33.950061168 +1000
+++ b/include/linux/cpufreq.h	2016-05-01 23:54:34.855075531 +1000
@@ -487,6 +487,9 @@
 #elif defined(CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE)
 extern struct cpufreq_governor cpufreq_gov_conservative;
 #define CPUFREQ_DEFAULT_GOVERNOR	(&cpufreq_gov_conservative)
+#elif defined(CONFIG_CPU_FREQ_DEFAULT_GOV_INTERACTIVE)
+extern struct cpufreq_governor cpufreq_gov_interactive;
+#define CPUFREQ_DEFAULT_GOVERNOR	(&cpufreq_gov_interactive)
 #endif
 
 /*********************************************************************
diff -Naur a/include/linux/cpu.h b/include/linux/cpu.h
--- a/include/linux/cpu.h	2016-05-01 23:54:00.184525619 +1000
+++ b/include/linux/cpu.h	2016-05-01 23:54:08.216319629 +1000
@@ -290,4 +290,11 @@
 bool cpu_report_death(void);
 #endif /* #ifdef CONFIG_HOTPLUG_CPU */
 
+#define IDLE_START 1
+#define IDLE_END 2
+
+void idle_notifier_register(struct notifier_block *n);
+void idle_notifier_unregister(struct notifier_block *n);
+void idle_notifier_call_chain(unsigned long val);
+
 #endif /* _LINUX_CPU_H_ */
diff -Naur a/include/trace/events/cpufreq_interactive.h b/include/trace/events/cpufreq_interactive.h
--- a/include/trace/events/cpufreq_interactive.h	1970-01-01 10:00:00.000000000 +1000
+++ b/include/trace/events/cpufreq_interactive.h	2016-05-01 23:54:14.238081791 +1000
@@ -0,0 +1,112 @@
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM cpufreq_interactive
+
+#if !defined(_TRACE_CPUFREQ_INTERACTIVE_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_CPUFREQ_INTERACTIVE_H
+
+#include <linux/tracepoint.h>
+
+DECLARE_EVENT_CLASS(set,
+	TP_PROTO(u32 cpu_id, unsigned long targfreq,
+	         unsigned long actualfreq),
+	TP_ARGS(cpu_id, targfreq, actualfreq),
+
+	TP_STRUCT__entry(
+	    __field(          u32, cpu_id    )
+	    __field(unsigned long, targfreq   )
+	    __field(unsigned long, actualfreq )
+	   ),
+
+	TP_fast_assign(
+	    __entry->cpu_id = (u32) cpu_id;
+	    __entry->targfreq = targfreq;
+	    __entry->actualfreq = actualfreq;
+	),
+
+	TP_printk("cpu=%u targ=%lu actual=%lu",
+	      __entry->cpu_id, __entry->targfreq,
+	      __entry->actualfreq)
+);
+
+DEFINE_EVENT(set, cpufreq_interactive_setspeed,
+	TP_PROTO(u32 cpu_id, unsigned long targfreq,
+	     unsigned long actualfreq),
+	TP_ARGS(cpu_id, targfreq, actualfreq)
+);
+
+DECLARE_EVENT_CLASS(loadeval,
+	    TP_PROTO(unsigned long cpu_id, unsigned long load,
+		     unsigned long curtarg, unsigned long curactual,
+		     unsigned long newtarg),
+		    TP_ARGS(cpu_id, load, curtarg, curactual, newtarg),
+
+	    TP_STRUCT__entry(
+		    __field(unsigned long, cpu_id    )
+		    __field(unsigned long, load      )
+		    __field(unsigned long, curtarg   )
+		    __field(unsigned long, curactual )
+		    __field(unsigned long, newtarg   )
+	    ),
+
+	    TP_fast_assign(
+		    __entry->cpu_id = cpu_id;
+		    __entry->load = load;
+		    __entry->curtarg = curtarg;
+		    __entry->curactual = curactual;
+		    __entry->newtarg = newtarg;
+	    ),
+
+	    TP_printk("cpu=%lu load=%lu cur=%lu actual=%lu targ=%lu",
+		      __entry->cpu_id, __entry->load, __entry->curtarg,
+		      __entry->curactual, __entry->newtarg)
+);
+
+DEFINE_EVENT(loadeval, cpufreq_interactive_target,
+	    TP_PROTO(unsigned long cpu_id, unsigned long load,
+		     unsigned long curtarg, unsigned long curactual,
+		     unsigned long newtarg),
+	    TP_ARGS(cpu_id, load, curtarg, curactual, newtarg)
+);
+
+DEFINE_EVENT(loadeval, cpufreq_interactive_already,
+	    TP_PROTO(unsigned long cpu_id, unsigned long load,
+		     unsigned long curtarg, unsigned long curactual,
+		     unsigned long newtarg),
+	    TP_ARGS(cpu_id, load, curtarg, curactual, newtarg)
+);
+
+DEFINE_EVENT(loadeval, cpufreq_interactive_notyet,
+	    TP_PROTO(unsigned long cpu_id, unsigned long load,
+		     unsigned long curtarg, unsigned long curactual,
+		     unsigned long newtarg),
+	    TP_ARGS(cpu_id, load, curtarg, curactual, newtarg)
+);
+
+TRACE_EVENT(cpufreq_interactive_boost,
+	    TP_PROTO(const char *s),
+	    TP_ARGS(s),
+	    TP_STRUCT__entry(
+		    __string(s, s)
+	    ),
+	    TP_fast_assign(
+		    __assign_str(s, s);
+	    ),
+	    TP_printk("%s", __get_str(s))
+);
+
+TRACE_EVENT(cpufreq_interactive_unboost,
+	    TP_PROTO(const char *s),
+	    TP_ARGS(s),
+	    TP_STRUCT__entry(
+		    __string(s, s)
+	    ),
+	    TP_fast_assign(
+		    __assign_str(s, s);
+	    ),
+	    TP_printk("%s", __get_str(s))
+);
+
+#endif /* _TRACE_CPUFREQ_INTERACTIVE_H */
+
+/* This part must be outside protection */
+#include <trace/define_trace.h>
diff -Naur a/kernel/cpu.c b/kernel/cpu.c
--- a/kernel/cpu.c	2016-05-01 23:54:03.506244948 +1000
+++ b/kernel/cpu.c	2016-05-01 23:54:08.216319629 +1000
@@ -827,3 +827,23 @@
 {
 	cpumask_copy(to_cpumask(cpu_online_bits), src);
 }
+
+static ATOMIC_NOTIFIER_HEAD(idle_notifier);
+
+void idle_notifier_register(struct notifier_block *n)
+{
+	atomic_notifier_chain_register(&idle_notifier, n);
+}
+EXPORT_SYMBOL_GPL(idle_notifier_register);
+
+void idle_notifier_unregister(struct notifier_block *n)
+{
+	atomic_notifier_chain_unregister(&idle_notifier, n);
+}
+EXPORT_SYMBOL_GPL(idle_notifier_unregister);
+
+void idle_notifier_call_chain(unsigned long val)
+{
+	atomic_notifier_call_chain(&idle_notifier, val, NULL);
+}
+EXPORT_SYMBOL_GPL(idle_notifier_call_chain);