Merge branch 'pm-cpuidle'

* pm-cpuidle:
  cpuidle: Add a kerneldoc comment to cpuidle_use_deepest_state()
  cpuidle: fix improper return value on error
  intel_idle: Convert to hotplug state machine
  intel_idle: Remove superfluous SMP fuction call
  MAINTAINERS: Add Jacob Pan as a new intel_idle maintainer
  MAINTAINERS: Add bug tracking system location entries for cpuidle
  x86/intel_idle: Add Knights Mill CPUID
  x86/intel_idle: Add CPU model 0x4a (Atom Z34xx series)
  thermal/intel_powerclamp: stop sched tick in forced idle
  thermal/intel_powerclamp: Convert to CPU hotplug state
  thermal/intel_powerclamp: Convert the kthread to kthread worker API
  thermal/intel_powerclamp: Remove duplicated code that starts the kthread
  sched/idle: Add support for tasks that inject idle
  cpuidle: Allow enforcing deepest idle state selection
  cpuidle/powernv: staticise powernv_idle_driver
  cpuidle: dt: assign ->enter_freeze to same as ->enter callback function
  cpuidle: governors: Remove remaining old module code

9 files changed, 301 insertions, 251 deletions

diff --git a/drivers/cpuidle/cpuidle-powernv.c b/drivers/cpuidle/cpuidle-powernv.c
index 7fe442ca38f4..0835a37a5f3a 100644
--- a/drivers/cpuidle/cpuidle-powernv.c
+++ b/drivers/cpuidle/cpuidle-powernv.c
@@ -22,7 +22,7 @@
 
 #define POWERNV_THRESHOLD_LATENCY_NS 200000
 
-struct cpuidle_driver powernv_idle_driver = {
+static struct cpuidle_driver powernv_idle_driver = {
 	.name             = "powernv_idle",
 	.owner            = THIS_MODULE,
 };
diff --git a/drivers/cpuidle/cpuidle.c b/drivers/cpuidle/cpuidle.c
index c73207abb5a4..62810ff3b00f 100644
--- a/drivers/cpuidle/cpuidle.c
+++ b/drivers/cpuidle/cpuidle.c
@@ -97,7 +97,23 @@ static int find_deepest_state(struct cpuidle_driver *drv,
 	return ret;
 }
 
-#ifdef CONFIG_SUSPEND
+/**
+ * cpuidle_use_deepest_state - Set/clear governor override flag.
+ * @enable: New value of the flag.
+ *
+ * Set/unset the current CPU to use the deepest idle state (override governors
+ * going forward if set).
+ */
+void cpuidle_use_deepest_state(bool enable)
+{
+	struct cpuidle_device *dev;
+
+	preempt_disable();
+	dev = cpuidle_get_device();
+	dev->use_deepest_state = enable;
+	preempt_enable();
+}
+
 /**
  * cpuidle_find_deepest_state - Find the deepest available idle state.
  * @drv: cpuidle driver for the given CPU.
@@ -109,6 +125,7 @@ int cpuidle_find_deepest_state(struct cpuidle_driver *drv,
 	return find_deepest_state(drv, dev, UINT_MAX, 0, false);
 }
 
+#ifdef CONFIG_SUSPEND
 static void enter_freeze_proper(struct cpuidle_driver *drv,
 				struct cpuidle_device *dev, int index)
 {
diff --git a/drivers/cpuidle/dt_idle_states.c b/drivers/cpuidle/dt_idle_states.c
index a5c111b67f37..ffca4fc0061d 100644
--- a/drivers/cpuidle/dt_idle_states.c
+++ b/drivers/cpuidle/dt_idle_states.c
@@ -38,6 +38,12 @@ static int init_state_node(struct cpuidle_state *idle_state,
 	 * state enter function.
 	 */
 	idle_state->enter = match_id->data;
+	/*
+	 * Since this is not a "coupled" state, it's safe to assume interrupts
+	 * won't be enabled when it exits allowing the tick to be frozen
+	 * safely. So enter() can be also enter_freeze() callback.
+	 */
+	idle_state->enter_freeze = match_id->data;
 
 	err = of_property_read_u32(state_node, "wakeup-latency-us",
 				   &idle_state->exit_latency);
diff --git a/drivers/cpuidle/governor.c b/drivers/cpuidle/governor.c
index fb9f511cca23..4e78263e34a4 100644
--- a/drivers/cpuidle/governor.c
+++ b/drivers/cpuidle/governor.c
@@ -9,7 +9,6 @@
  */
 
 #include <linux/mutex.h>
-#include <linux/module.h>
 #include <linux/cpuidle.h>
 
 #include "cpuidle.h"
@@ -53,14 +52,11 @@ int cpuidle_switch_governor(struct cpuidle_governor *gov)
 	if (cpuidle_curr_governor) {
 		list_for_each_entry(dev, &cpuidle_detected_devices, device_list)
 			cpuidle_disable_device(dev);
-		module_put(cpuidle_curr_governor->owner);
 	}
 
 	cpuidle_curr_governor = gov;
 
 	if (gov) {
-		if (!try_module_get(cpuidle_curr_governor->owner))
-			return -EINVAL;
 		list_for_each_entry(dev, &cpuidle_detected_devices, device_list)
 			cpuidle_enable_device(dev);
 		cpuidle_install_idle_handler();
diff --git a/drivers/cpuidle/governors/ladder.c b/drivers/cpuidle/governors/ladder.c
index 63bd5a403e22..fe8f08948fcb 100644
--- a/drivers/cpuidle/governors/ladder.c
+++ b/drivers/cpuidle/governors/ladder.c
@@ -15,7 +15,6 @@
 #include <linux/kernel.h>
 #include <linux/cpuidle.h>
 #include <linux/pm_qos.h>
-#include <linux/module.h>
 #include <linux/jiffies.h>
 #include <linux/tick.h>
 
@@ -177,7 +176,6 @@ static struct cpuidle_governor ladder_governor = {
 	.enable =	ladder_enable_device,
 	.select =	ladder_select_state,
 	.reflect =	ladder_reflect,
-	.owner =	THIS_MODULE,
 };
 
 /**
diff --git a/drivers/cpuidle/governors/menu.c b/drivers/cpuidle/governors/menu.c
index 03d38c291de6..d9b5b9398a0f 100644
--- a/drivers/cpuidle/governors/menu.c
+++ b/drivers/cpuidle/governors/menu.c
@@ -19,7 +19,6 @@
 #include <linux/tick.h>
 #include <linux/sched.h>
 #include <linux/math64.h>
-#include <linux/module.h>
 
 /*
  * Please note when changing the tuning values:
@@ -484,7 +483,6 @@ static struct cpuidle_governor menu_governor = {
 	.enable =	menu_enable_device,
 	.select =	menu_select,
 	.reflect =	menu_reflect,
-	.owner =	THIS_MODULE,
 };
 
 /**
diff --git a/drivers/cpuidle/sysfs.c b/drivers/cpuidle/sysfs.c
index 832a2c3f01ff..c5adc8c9ac43 100644
--- a/drivers/cpuidle/sysfs.c
+++ b/drivers/cpuidle/sysfs.c
@@ -403,8 +403,10 @@ static int cpuidle_add_state_sysfs(struct cpuidle_device *device)
 	/* state statistics */
 	for (i = 0; i < drv->state_count; i++) {
 		kobj = kzalloc(sizeof(struct cpuidle_state_kobj), GFP_KERNEL);
-		if (!kobj)
+		if (!kobj) {
+			ret = -ENOMEM;
 			goto error_state;
+		}
 		kobj->state = &drv->states[i];
 		kobj->state_usage = &device->states_usage[i];
 		init_completion(&kobj->kobj_unregister);
diff --git a/drivers/idle/intel_idle.c b/drivers/idle/intel_idle.c
index 4466a2f969d7..7d8ea3d5fda6 100644
--- a/drivers/idle/intel_idle.c
+++ b/drivers/idle/intel_idle.c
@@ -98,8 +98,6 @@ static int intel_idle(struct cpuidle_device *dev,
 			struct cpuidle_driver *drv, int index);
 static void intel_idle_freeze(struct cpuidle_device *dev,
 			      struct cpuidle_driver *drv, int index);
-static int intel_idle_cpu_init(int cpu);
-
 static struct cpuidle_state *cpuidle_state_table;
 
 /*
@@ -724,6 +722,50 @@ static struct cpuidle_state atom_cstates[] = {
 	{
 		.enter = NULL }
 };
+static struct cpuidle_state tangier_cstates[] = {
+	{
+		.name = "C1-TNG",
+		.desc = "MWAIT 0x00",
+		.flags = MWAIT2flg(0x00),
+		.exit_latency = 1,
+		.target_residency = 4,
+		.enter = &intel_idle,
+		.enter_freeze = intel_idle_freeze, },
+	{
+		.name = "C4-TNG",
+		.desc = "MWAIT 0x30",
+		.flags = MWAIT2flg(0x30) | CPUIDLE_FLAG_TLB_FLUSHED,
+		.exit_latency = 100,
+		.target_residency = 400,
+		.enter = &intel_idle,
+		.enter_freeze = intel_idle_freeze, },
+	{
+		.name = "C6-TNG",
+		.desc = "MWAIT 0x52",
+		.flags = MWAIT2flg(0x52) | CPUIDLE_FLAG_TLB_FLUSHED,
+		.exit_latency = 140,
+		.target_residency = 560,
+		.enter = &intel_idle,
+		.enter_freeze = intel_idle_freeze, },
+	{
+		.name = "C7-TNG",
+		.desc = "MWAIT 0x60",
+		.flags = MWAIT2flg(0x60) | CPUIDLE_FLAG_TLB_FLUSHED,
+		.exit_latency = 1200,
+		.target_residency = 4000,
+		.enter = &intel_idle,
+		.enter_freeze = intel_idle_freeze, },
+	{
+		.name = "C9-TNG",
+		.desc = "MWAIT 0x64",
+		.flags = MWAIT2flg(0x64) | CPUIDLE_FLAG_TLB_FLUSHED,
+		.exit_latency = 10000,
+		.target_residency = 20000,
+		.enter = &intel_idle,
+		.enter_freeze = intel_idle_freeze, },
+	{
+		.enter = NULL }
+};
 static struct cpuidle_state avn_cstates[] = {
 	{
 		.name = "C1-AVN",
@@ -907,51 +949,15 @@ static void intel_idle_freeze(struct cpuidle_device *dev,
 	mwait_idle_with_hints(eax, ecx);
 }
 
-static void __setup_broadcast_timer(void *arg)
+static void __setup_broadcast_timer(bool on)
 {
-	unsigned long on = (unsigned long)arg;
-
 	if (on)
 		tick_broadcast_enable();
 	else
 		tick_broadcast_disable();
 }
 
-static int cpu_hotplug_notify(struct notifier_block *n,
-			      unsigned long action, void *hcpu)
-{
-	int hotcpu = (unsigned long)hcpu;
-	struct cpuidle_device *dev;
-
-	switch (action & ~CPU_TASKS_FROZEN) {
-	case CPU_ONLINE:
-
-		if (lapic_timer_reliable_states != LAPIC_TIMER_ALWAYS_RELIABLE)
-			smp_call_function_single(hotcpu, __setup_broadcast_timer,
-						 (void *)true, 1);
-
-		/*
-		 * Some systems can hotplug a cpu at runtime after
-		 * the kernel has booted, we have to initialize the
-		 * driver in this case
-		 */
-		dev = per_cpu_ptr(intel_idle_cpuidle_devices, hotcpu);
-		if (dev->registered)
-			break;
-
-		if (intel_idle_cpu_init(hotcpu))
-			return NOTIFY_BAD;
-
-		break;
-	}
-	return NOTIFY_OK;
-}
-
-static struct notifier_block cpu_hotplug_notifier = {
-	.notifier_call = cpu_hotplug_notify,
-};
-
-static void auto_demotion_disable(void *dummy)
+static void auto_demotion_disable(void)
 {
 	unsigned long long msr_bits;
 
@@ -959,7 +965,7 @@ static void auto_demotion_disable(void *dummy)
 	msr_bits &= ~(icpu->auto_demotion_disable_flags);
 	wrmsrl(MSR_NHM_SNB_PKG_CST_CFG_CTL, msr_bits);
 }
-static void c1e_promotion_disable(void *dummy)
+static void c1e_promotion_disable(void)
 {
 	unsigned long long msr_bits;
 
@@ -978,6 +984,10 @@ static const struct idle_cpu idle_cpu_atom = {
 	.state_table = atom_cstates,
 };
 
+static const struct idle_cpu idle_cpu_tangier = {
+	.state_table = tangier_cstates,
+};
+
 static const struct idle_cpu idle_cpu_lincroft = {
 	.state_table = atom_cstates,
 	.auto_demotion_disable_flags = ATM_LNC_C6_AUTO_DEMOTE,
@@ -1066,6 +1076,7 @@ static const struct x86_cpu_id intel_idle_ids[] __initconst = {
 	ICPU(INTEL_FAM6_SANDYBRIDGE_X,		idle_cpu_snb),
 	ICPU(INTEL_FAM6_ATOM_CEDARVIEW,		idle_cpu_atom),
 	ICPU(INTEL_FAM6_ATOM_SILVERMONT1,	idle_cpu_byt),
+	ICPU(INTEL_FAM6_ATOM_MERRIFIELD,	idle_cpu_tangier),
 	ICPU(INTEL_FAM6_ATOM_AIRMONT,		idle_cpu_cht),
 	ICPU(INTEL_FAM6_IVYBRIDGE,		idle_cpu_ivb),
 	ICPU(INTEL_FAM6_IVYBRIDGE_X,		idle_cpu_ivt),
@@ -1084,6 +1095,7 @@ static const struct x86_cpu_id intel_idle_ids[] __initconst = {
 	ICPU(INTEL_FAM6_KABYLAKE_DESKTOP,	idle_cpu_skl),
 	ICPU(INTEL_FAM6_SKYLAKE_X,		idle_cpu_skx),
 	ICPU(INTEL_FAM6_XEON_PHI_KNL,		idle_cpu_knl),
+	ICPU(INTEL_FAM6_XEON_PHI_KNM,		idle_cpu_knl),
 	ICPU(INTEL_FAM6_ATOM_GOLDMONT,		idle_cpu_bxt),
 	ICPU(INTEL_FAM6_ATOM_DENVERTON,		idle_cpu_dnv),
 	{}
@@ -1373,12 +1385,11 @@ static void __init intel_idle_cpuidle_driver_init(void)
  * allocate, initialize, register cpuidle_devices
  * @cpu: cpu/core to initialize
  */
-static int intel_idle_cpu_init(int cpu)
+static int intel_idle_cpu_init(unsigned int cpu)
 {
 	struct cpuidle_device *dev;
 
 	dev = per_cpu_ptr(intel_idle_cpuidle_devices, cpu);
-
 	dev->cpu = cpu;
 
 	if (cpuidle_register_device(dev)) {
@@ -1387,17 +1398,36 @@ static int intel_idle_cpu_init(int cpu)
 	}
 
 	if (icpu->auto_demotion_disable_flags)
-		smp_call_function_single(cpu, auto_demotion_disable, NULL, 1);
+		auto_demotion_disable();
 
 	if (icpu->disable_promotion_to_c1e)
-		smp_call_function_single(cpu, c1e_promotion_disable, NULL, 1);
+		c1e_promotion_disable();
+
+	return 0;
+}
+
+static int intel_idle_cpu_online(unsigned int cpu)
+{
+	struct cpuidle_device *dev;
+
+	if (lapic_timer_reliable_states != LAPIC_TIMER_ALWAYS_RELIABLE)
+		__setup_broadcast_timer(true);
+
+	/*
+	 * Some systems can hotplug a cpu at runtime after
+	 * the kernel has booted, we have to initialize the
+	 * driver in this case
+	 */
+	dev = per_cpu_ptr(intel_idle_cpuidle_devices, cpu);
+	if (!dev->registered)
+		return intel_idle_cpu_init(cpu);
 
 	return 0;
 }
 
 static int __init intel_idle_init(void)
 {
-	int retval, i;
+	int retval;
 
 	/* Do not load intel_idle at all for now if idle= is passed */
 	if (boot_option_idle_override != IDLE_NO_OVERRIDE)
@@ -1417,35 +1447,29 @@ static int __init intel_idle_init(void)
 		struct cpuidle_driver *drv = cpuidle_get_driver();
 		printk(KERN_DEBUG PREFIX "intel_idle yielding to %s",
 			drv ? drv->name : "none");
-		free_percpu(intel_idle_cpuidle_devices);
-		return retval;
+		goto init_driver_fail;
 	}
 
-	cpu_notifier_register_begin();
-
-	for_each_online_cpu(i) {
-		retval = intel_idle_cpu_init(i);
-		if (retval) {
-			intel_idle_cpuidle_devices_uninit();
-			cpu_notifier_register_done();
-			cpuidle_unregister_driver(&intel_idle_driver);
-			free_percpu(intel_idle_cpuidle_devices);
-			return retval;
-		}
-	}
-	__register_cpu_notifier(&cpu_hotplug_notifier);
-
 	if (boot_cpu_has(X86_FEATURE_ARAT))	/* Always Reliable APIC Timer */
 		lapic_timer_reliable_states = LAPIC_TIMER_ALWAYS_RELIABLE;
-	else
-		on_each_cpu(__setup_broadcast_timer, (void *)true, 1);
 
-	cpu_notifier_register_done();
+	retval = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "idle/intel:online",
+				   intel_idle_cpu_online, NULL);
+	if (retval < 0)
+		goto hp_setup_fail;
 
 	pr_debug(PREFIX "lapic_timer_reliable_states 0x%x\n",
 		lapic_timer_reliable_states);
 
 	return 0;
+
+hp_setup_fail:
+	intel_idle_cpuidle_devices_uninit();
+	cpuidle_unregister_driver(&intel_idle_driver);
+init_driver_fail:
+	free_percpu(intel_idle_cpuidle_devices);
+	return retval;
+
 }
 device_initcall(intel_idle_init);
 
diff --git a/drivers/thermal/intel_powerclamp.c b/drivers/thermal/intel_powerclamp.c
index afada655f861..83e697186410 100644
--- a/drivers/thermal/intel_powerclamp.c
+++ b/drivers/thermal/intel_powerclamp.c
@@ -43,7 +43,6 @@
 #include <linux/kernel.h>
 #include <linux/delay.h>
 #include <linux/kthread.h>
-#include <linux/freezer.h>
 #include <linux/cpu.h>
 #include <linux/thermal.h>
 #include <linux/slab.h>
@@ -86,11 +85,26 @@ static unsigned int control_cpu; /* The cpu assigned to collect stat and update
 				  */
 static bool clamping;
 
+static const struct sched_param sparam = {
+	.sched_priority = MAX_USER_RT_PRIO / 2,
+};
+struct powerclamp_worker_data {
+	struct kthread_worker *worker;
+	struct kthread_work balancing_work;
+	struct kthread_delayed_work idle_injection_work;
+	unsigned int cpu;
+	unsigned int count;
+	unsigned int guard;
+	unsigned int window_size_now;
+	unsigned int target_ratio;
+	unsigned int duration_jiffies;
+	bool clamping;
+};
 
-static struct task_struct * __percpu *powerclamp_thread;
+static struct powerclamp_worker_data * __percpu worker_data;
 static struct thermal_cooling_device *cooling_dev;
 static unsigned long *cpu_clamping_mask;  /* bit map for tracking per cpu
-					   * clamping thread
+					   * clamping kthread worker
 					   */
 
 static unsigned int duration;
@@ -262,11 +276,6 @@ static u64 pkg_state_counter(void)
 	return count;
 }
 
-static void noop_timer(unsigned long foo)
-{
-	/* empty... just the fact that we get the interrupt wakes us up */
-}
-
 static unsigned int get_compensation(int ratio)
 {
 	unsigned int comp = 0;
@@ -368,103 +377,79 @@ static bool powerclamp_adjust_controls(unsigned int target_ratio,
 	return set_target_ratio + guard <= current_ratio;
 }
 
-static int clamp_thread(void *arg)
+static void clamp_balancing_func(struct kthread_work *work)
 {
-	int cpunr = (unsigned long)arg;
-	DEFINE_TIMER(wakeup_timer, noop_timer, 0, 0);
-	static const struct sched_param param = {
-		.sched_priority = MAX_USER_RT_PRIO/2,
-	};
-	unsigned int count = 0;
-	unsigned int target_ratio;
+	struct powerclamp_worker_data *w_data;
+	int sleeptime;
+	unsigned long target_jiffies;
+	unsigned int compensated_ratio;
+	int interval; /* jiffies to sleep for each attempt */
 
-	set_bit(cpunr, cpu_clamping_mask);
-	set_freezable();
-	init_timer_on_stack(&wakeup_timer);
-	sched_setscheduler(current, SCHED_FIFO, &param);
-
-	while (true == clamping && !kthread_should_stop() &&
-		cpu_online(cpunr)) {
-		int sleeptime;
-		unsigned long target_jiffies;
-		unsigned int guard;
-		unsigned int compensated_ratio;
-		int interval; /* jiffies to sleep for each attempt */
-		unsigned int duration_jiffies = msecs_to_jiffies(duration);
-		unsigned int window_size_now;
-
-		try_to_freeze();
-		/*
-		 * make sure user selected ratio does not take effect until
-		 * the next round. adjust target_ratio if user has changed
-		 * target such that we can converge quickly.
-		 */
-		target_ratio = set_target_ratio;
-		guard = 1 + target_ratio/20;
-		window_size_now = window_size;
-		count++;
-
-		/*
-		 * systems may have different ability to enter package level
-		 * c-states, thus we need to compensate the injected idle ratio
-		 * to achieve the actual target reported by the HW.
-		 */
-		compensated_ratio = target_ratio +
-			get_compensation(target_ratio);
-		if (compensated_ratio <= 0)
-			compensated_ratio = 1;
-		interval = duration_jiffies * 100 / compensated_ratio;
-
-		/* align idle time */
-		target_jiffies = roundup(jiffies, interval);
-		sleeptime = target_jiffies - jiffies;
-		if (sleeptime <= 0)
-			sleeptime = 1;
-		schedule_timeout_interruptible(sleeptime);
-		/*
-		 * only elected controlling cpu can collect stats and update
-		 * control parameters.
-		 */
-		if (cpunr == control_cpu && !(count%window_size_now)) {
-			should_skip =
-				powerclamp_adjust_controls(target_ratio,
-							guard, window_size_now);
-			smp_mb();
-		}
+	w_data = container_of(work, struct powerclamp_worker_data,
+			      balancing_work);
 
-		if (should_skip)
-			continue;
-
-		target_jiffies = jiffies + duration_jiffies;
-		mod_timer(&wakeup_timer, target_jiffies);
-		if (unlikely(local_softirq_pending()))
-			continue;
-		/*
-		 * stop tick sched during idle time, interrupts are still
-		 * allowed. thus jiffies are updated properly.
-		 */
-		preempt_disable();
-		/* mwait until target jiffies is reached */
-		while (time_before(jiffies, target_jiffies)) {
-			unsigned long ecx = 1;
-			unsigned long eax = target_mwait;
-
-			/*
-			 * REVISIT: may call enter_idle() to notify drivers who
-			 * can save power during cpu idle. same for exit_idle()
-			 */
-			local_touch_nmi();
-			stop_critical_timings();
-			mwait_idle_with_hints(eax, ecx);
-			start_critical_timings();
-			atomic_inc(&idle_wakeup_counter);
-		}
-		preempt_enable();
+	/*
+	 * make sure user selected ratio does not take effect until
+	 * the next round. adjust target_ratio if user has changed
+	 * target such that we can converge quickly.
+	 */
+	w_data->target_ratio = READ_ONCE(set_target_ratio);
+	w_data->guard = 1 + w_data->target_ratio / 20;
+	w_data->window_size_now = window_size;
+	w_data->duration_jiffies = msecs_to_jiffies(duration);
+	w_data->count++;
+
+	/*
+	 * systems may have different ability to enter package level
+	 * c-states, thus we need to compensate the injected idle ratio
+	 * to achieve the actual target reported by the HW.
+	 */
+	compensated_ratio = w_data->target_ratio +
+		get_compensation(w_data->target_ratio);
+	if (compensated_ratio <= 0)
+		compensated_ratio = 1;
+	interval = w_data->duration_jiffies * 100 / compensated_ratio;
+
+	/* align idle time */
+	target_jiffies = roundup(jiffies, interval);
+	sleeptime = target_jiffies - jiffies;
+	if (sleeptime <= 0)
+		sleeptime = 1;
+
+	if (clamping && w_data->clamping && cpu_online(w_data->cpu))
+		kthread_queue_delayed_work(w_data->worker,
+					   &w_data->idle_injection_work,
+					   sleeptime);
+}
+
+static void clamp_idle_injection_func(struct kthread_work *work)
+{
+	struct powerclamp_worker_data *w_data;
+
+	w_data = container_of(work, struct powerclamp_worker_data,
+			      idle_injection_work.work);
+
+	/*
+	 * only elected controlling cpu can collect stats and update
+	 * control parameters.
+	 */
+	if (w_data->cpu == control_cpu &&
+	    !(w_data->count % w_data->window_size_now)) {
+		should_skip =
+			powerclamp_adjust_controls(w_data->target_ratio,
+						   w_data->guard,
+						   w_data->window_size_now);
+		smp_mb();
 	}
-	del_timer_sync(&wakeup_timer);
-	clear_bit(cpunr, cpu_clamping_mask);
 
-	return 0;
+	if (should_skip)
+		goto balance;
+
+	play_idle(jiffies_to_msecs(w_data->duration_jiffies));
+
+balance:
+	if (clamping && w_data->clamping && cpu_online(w_data->cpu))
+		kthread_queue_work(w_data->worker, &w_data->balancing_work);
 }
 
 /*
@@ -508,10 +493,60 @@ static void poll_pkg_cstate(struct work_struct *dummy)
 		schedule_delayed_work(&poll_pkg_cstate_work, HZ);
 }
 
+static void start_power_clamp_worker(unsigned long cpu)
+{
+	struct powerclamp_worker_data *w_data = per_cpu_ptr(worker_data, cpu);
+	struct kthread_worker *worker;
+
+	worker = kthread_create_worker_on_cpu(cpu, 0, "kidle_inject/%ld", cpu);
+	if (IS_ERR(worker))
+		return;
+
+	w_data->worker = worker;
+	w_data->count = 0;
+	w_data->cpu = cpu;
+	w_data->clamping = true;
+	set_bit(cpu, cpu_clamping_mask);
+	sched_setscheduler(worker->task, SCHED_FIFO, &sparam);
+	kthread_init_work(&w_data->balancing_work, clamp_balancing_func);
+	kthread_init_delayed_work(&w_data->idle_injection_work,
+				  clamp_idle_injection_func);
+	kthread_queue_work(w_data->worker, &w_data->balancing_work);
+}
+
+static void stop_power_clamp_worker(unsigned long cpu)
+{
+	struct powerclamp_worker_data *w_data = per_cpu_ptr(worker_data, cpu);
+
+	if (!w_data->worker)
+		return;
+
+	w_data->clamping = false;
+	/*
+	 * Make sure that all works that get queued after this point see
+	 * the clamping disabled. The counter part is not needed because
+	 * there is an implicit memory barrier when the queued work
+	 * is proceed.
+	 */
+	smp_wmb();
+	kthread_cancel_work_sync(&w_data->balancing_work);
+	kthread_cancel_delayed_work_sync(&w_data->idle_injection_work);
+	/*
+	 * The balancing work still might be queued here because
+	 * the handling of the "clapming" variable, cancel, and queue
+	 * operations are not synchronized via a lock. But it is not
+	 * a big deal. The balancing work is fast and destroy kthread
+	 * will wait for it.
+	 */
+	clear_bit(w_data->cpu, cpu_clamping_mask);
+	kthread_destroy_worker(w_data->worker);
+
+	w_data->worker = NULL;
+}
+
 static int start_power_clamp(void)
 {
 	unsigned long cpu;
-	struct task_struct *thread;
 
 	set_target_ratio = clamp(set_target_ratio, 0U, MAX_TARGET_RATIO - 1);
 	/* prevent cpu hotplug */
@@ -525,22 +560,9 @@ static int start_power_clamp(void)
 	clamping = true;
 	schedule_delayed_work(&poll_pkg_cstate_work, 0);
 
-	/* start one thread per online cpu */
+	/* start one kthread worker per online cpu */
 	for_each_online_cpu(cpu) {
-		struct task_struct **p =
-			per_cpu_ptr(powerclamp_thread, cpu);
-
-		thread = kthread_create_on_node(clamp_thread,
-						(void *) cpu,
-						cpu_to_node(cpu),
-						"kidle_inject/%ld", cpu);
-		/* bind to cpu here */
-		if (likely(!IS_ERR(thread))) {
-			kthread_bind(thread, cpu);
-			wake_up_process(thread);
-			*p = thread;
-		}
-
+		start_power_clamp_worker(cpu);
 	}
 	put_online_cpus();
 
@@ -550,71 +572,49 @@ static int start_power_clamp(void)
 static void end_power_clamp(void)
 {
 	int i;
-	struct task_struct *thread;
 
-	clamping = false;
 	/*
-	 * make clamping visible to other cpus and give per cpu clamping threads
-	 * sometime to exit, or gets killed later.
+	 * Block requeuing in all the kthread workers. They will flush and
+	 * stop faster.
 	 */
-	smp_mb();
-	msleep(20);
+	clamping = false;
 	if (bitmap_weight(cpu_clamping_mask, num_possible_cpus())) {
 		for_each_set_bit(i, cpu_clamping_mask, num_possible_cpus()) {
-			pr_debug("clamping thread for cpu %d alive, kill\n", i);
-			thread = *per_cpu_ptr(powerclamp_thread, i);
-			kthread_stop(thread);
+			pr_debug("clamping worker for cpu %d alive, destroy\n",
+				 i);
+			stop_power_clamp_worker(i);
 		}
 	}
 }
 
-static int powerclamp_cpu_callback(struct notifier_block *nfb,
-				unsigned long action, void *hcpu)
+static int powerclamp_cpu_online(unsigned int cpu)
 {
-	unsigned long cpu = (unsigned long)hcpu;
-	struct task_struct *thread;
-	struct task_struct **percpu_thread =
-		per_cpu_ptr(powerclamp_thread, cpu);
-
-	if (false == clamping)
-		goto exit_ok;
-
-	switch (action) {
-	case CPU_ONLINE:
-		thread = kthread_create_on_node(clamp_thread,
-						(void *) cpu,
-						cpu_to_node(cpu),
-						"kidle_inject/%lu", cpu);
-		if (likely(!IS_ERR(thread))) {
-			kthread_bind(thread, cpu);
-			wake_up_process(thread);
-			*percpu_thread = thread;
-		}
-		/* prefer BSP as controlling CPU */
-		if (cpu == 0) {
-			control_cpu = 0;
-			smp_mb();
-		}
-		break;
-	case CPU_DEAD:
-		if (test_bit(cpu, cpu_clamping_mask)) {
-			pr_err("cpu %lu dead but powerclamping thread is not\n",
-				cpu);
-			kthread_stop(*percpu_thread);
-		}
-		if (cpu == control_cpu) {
-			control_cpu = smp_processor_id();
-			smp_mb();
-		}
+	if (clamping == false)
+		return 0;
+	start_power_clamp_worker(cpu);
+	/* prefer BSP as controlling CPU */
+	if (cpu == 0) {
+		control_cpu = 0;
+		smp_mb();
 	}
-
-exit_ok:
-	return NOTIFY_OK;
+	return 0;
 }
 
-static struct notifier_block powerclamp_cpu_notifier = {
-	.notifier_call = powerclamp_cpu_callback,
-};
+static int powerclamp_cpu_predown(unsigned int cpu)
+{
+	if (clamping == false)
+		return 0;
+
+	stop_power_clamp_worker(cpu);
+	if (cpu != control_cpu)
+		return 0;
+
+	control_cpu = cpumask_first(cpu_online_mask);
+	if (control_cpu == cpu)
+		control_cpu = cpumask_next(cpu, cpu_online_mask);
+	smp_mb();
+	return 0;
+}
 
 static int powerclamp_get_max_state(struct thermal_cooling_device *cdev,
 				 unsigned long *state)
@@ -742,6 +742,8 @@ file_error:
 	debugfs_remove_recursive(debug_dir);
 }
 
+static enum cpuhp_state hp_state;
+
 static int __init powerclamp_init(void)
 {
 	int retval;
@@ -759,10 +761,17 @@ static int __init powerclamp_init(void)
 
 	/* set default limit, maybe adjusted during runtime based on feedback */
 	window_size = 2;
-	register_hotcpu_notifier(&powerclamp_cpu_notifier);
+	retval = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN,
+					   "thermal/intel_powerclamp:online",
+					   powerclamp_cpu_online,
+					   powerclamp_cpu_predown);
+	if (retval < 0)
+		goto exit_free;
+
+	hp_state = retval;
 
-	powerclamp_thread = alloc_percpu(struct task_struct *);
-	if (!powerclamp_thread) {
+	worker_data = alloc_percpu(struct powerclamp_worker_data);
+	if (!worker_data) {
 		retval = -ENOMEM;
 		goto exit_unregister;
 	}
@@ -782,9 +791,9 @@ static int __init powerclamp_init(void)
 	return 0;
 
 exit_free_thread:
-	free_percpu(powerclamp_thread);
+	free_percpu(worker_data);
 exit_unregister:
-	unregister_hotcpu_notifier(&powerclamp_cpu_notifier);
+	cpuhp_remove_state_nocalls(hp_state);
 exit_free:
 	kfree(cpu_clamping_mask);
 	return retval;
@@ -793,9 +802,9 @@ module_init(powerclamp_init);
 
 static void __exit powerclamp_exit(void)
 {
-	unregister_hotcpu_notifier(&powerclamp_cpu_notifier);
 	end_power_clamp();
-	free_percpu(powerclamp_thread);
+	cpuhp_remove_state_nocalls(hp_state);
+	free_percpu(worker_data);
 	thermal_cooling_device_unregister(cooling_dev);
 	kfree(cpu_clamping_mask);