[net-next-2.6.git] / drivers / cpuidle / cpuidle.c

/*
 * cpuidle.c - core cpuidle infrastructure
 *
 * (C) 2006-2007 Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
 *               Shaohua Li <shaohua.li@intel.com>
 *               Adam Belay <abelay@novell.com>
 *
 * This code is licenced under the GPL.
 */

#include <linux/kernel.h>
#include <linux/mutex.h>
#include <linux/sched.h>
#include <linux/notifier.h>
#include <linux/pm_qos_params.h>
#include <linux/cpu.h>
#include <linux/cpuidle.h>
#include <linux/ktime.h>
#include <linux/hrtimer.h>
#include <trace/events/power.h>

#include "cpuidle.h"

DEFINE_PER_CPU(struct cpuidle_device *, cpuidle_devices);

DEFINE_MUTEX(cpuidle_lock);
LIST_HEAD(cpuidle_detected_devices);
static void (*pm_idle_old)(void);

static int enabled_devices;

#if defined(CONFIG_ARCH_HAS_CPU_IDLE_WAIT)
static void cpuidle_kick_cpus(void)
{
	cpu_idle_wait();
}
#elif defined(CONFIG_SMP)
# error "Arch needs cpu_idle_wait() equivalent here"
#else /* !CONFIG_ARCH_HAS_CPU_IDLE_WAIT && !CONFIG_SMP */
static void cpuidle_kick_cpus(void) {}
#endif

static int __cpuidle_register_device(struct cpuidle_device *dev);

/**
 * cpuidle_idle_call - the main idle loop
 *
 * NOTE: no locks or semaphores should be used here
 */
static void cpuidle_idle_call(void)
{
	struct cpuidle_device *dev = __get_cpu_var(cpuidle_devices);
	struct cpuidle_state *target_state;
	int next_state;

	/* check if the device is ready */
	if (!dev || !dev->enabled) {
		if (pm_idle_old)
			pm_idle_old();
		else
#if defined(CONFIG_ARCH_HAS_DEFAULT_IDLE)
			default_idle();
#else
			local_irq_enable();
#endif
		return;
	}

#if 0
	/* shows regressions, re-enable for 2.6.29 */
	/*
	 * run any timers that can be run now, at this point
	 * before calculating the idle duration etc.
	 */
	hrtimer_peek_ahead_timers();
#endif

	/*
	 * Call the device's prepare function before calling the
	 * governor's select function.  ->prepare gives the device's
	 * cpuidle driver a chance to update any dynamic information
	 * of its cpuidle states for the current idle period, e.g.
	 * state availability, latencies, residencies, etc.
	 */
	if (dev->prepare)
		dev->prepare(dev);

	/* ask the governor for the next state */
	next_state = cpuidle_curr_governor->select(dev);
	if (need_resched()) {
		local_irq_enable();
		return;
	}

	target_state = &dev->states[next_state];

	/* enter the state and update stats */
	dev->last_state = target_state;
	dev->last_residency = target_state->enter(dev, target_state);
	if (dev->last_state)
		target_state = dev->last_state;

	target_state->time += (unsigned long long)dev->last_residency;
	target_state->usage++;

	/* give the governor an opportunity to reflect on the outcome */
	if (cpuidle_curr_governor->reflect)
		cpuidle_curr_governor->reflect(dev);
	trace_power_end(smp_processor_id());
}

/**
 * cpuidle_install_idle_handler - installs the cpuidle idle loop handler
 */
void cpuidle_install_idle_handler(void)
{
	if (enabled_devices && (pm_idle != cpuidle_idle_call)) {
		/* Make sure all changes finished before we switch to new idle */
		smp_wmb();
		pm_idle = cpuidle_idle_call;
	}
}

/**
 * cpuidle_uninstall_idle_handler - uninstalls the cpuidle idle loop handler
 */
void cpuidle_uninstall_idle_handler(void)
{
	if (enabled_devices && pm_idle_old && (pm_idle != pm_idle_old)) {
		pm_idle = pm_idle_old;
		cpuidle_kick_cpus();
	}
}

/**
 * cpuidle_pause_and_lock - temporarily disables CPUIDLE
 */
void cpuidle_pause_and_lock(void)
{
	mutex_lock(&cpuidle_lock);
	cpuidle_uninstall_idle_handler();
}

EXPORT_SYMBOL_GPL(cpuidle_pause_and_lock);

/**
 * cpuidle_resume_and_unlock - resumes CPUIDLE operation
 */
void cpuidle_resume_and_unlock(void)
{
	cpuidle_install_idle_handler();
	mutex_unlock(&cpuidle_lock);
}

EXPORT_SYMBOL_GPL(cpuidle_resume_and_unlock);

/**
 * cpuidle_enable_device - enables idle PM for a CPU
 * @dev: the CPU
 *
 * This function must be called between cpuidle_pause_and_lock and
 * cpuidle_resume_and_unlock when used externally.
 */
int cpuidle_enable_device(struct cpuidle_device *dev)
{
	int ret, i;

	if (dev->enabled)
		return 0;
	if (!cpuidle_get_driver() || !cpuidle_curr_governor)
		return -EIO;
	if (!dev->state_count)
		return -EINVAL;

	if (dev->registered == 0) {
		ret = __cpuidle_register_device(dev);
		if (ret)
			return ret;
	}

	if ((ret = cpuidle_add_state_sysfs(dev)))
		return ret;

	if (cpuidle_curr_governor->enable &&
	    (ret = cpuidle_curr_governor->enable(dev)))
		goto fail_sysfs;

	for (i = 0; i < dev->state_count; i++) {
		dev->states[i].usage = 0;
		dev->states[i].time = 0;
	}
	dev->last_residency = 0;
	dev->last_state = NULL;

	smp_wmb();

	dev->enabled = 1;

	enabled_devices++;
	return 0;

fail_sysfs:
	cpuidle_remove_state_sysfs(dev);

	return ret;
}

EXPORT_SYMBOL_GPL(cpuidle_enable_device);

/**
 * cpuidle_disable_device - disables idle PM for a CPU
 * @dev: the CPU
 *
 * This function must be called between cpuidle_pause_and_lock and
 * cpuidle_resume_and_unlock when used externally.
 */
void cpuidle_disable_device(struct cpuidle_device *dev)
{
	if (!dev->enabled)
		return;
	if (!cpuidle_get_driver() || !cpuidle_curr_governor)
		return;

	dev->enabled = 0;

	if (cpuidle_curr_governor->disable)
		cpuidle_curr_governor->disable(dev);

	cpuidle_remove_state_sysfs(dev);
	enabled_devices--;
}

EXPORT_SYMBOL_GPL(cpuidle_disable_device);

#ifdef CONFIG_ARCH_HAS_CPU_RELAX
static int poll_idle(struct cpuidle_device *dev, struct cpuidle_state *st)
{
	ktime_t	t1, t2;
	s64 diff;
	int ret;

	t1 = ktime_get();
	local_irq_enable();
	while (!need_resched())
		cpu_relax();

	t2 = ktime_get();
	diff = ktime_to_us(ktime_sub(t2, t1));
	if (diff > INT_MAX)
		diff = INT_MAX;

	ret = (int) diff;
	return ret;
}

static void poll_idle_init(struct cpuidle_device *dev)
{
	struct cpuidle_state *state = &dev->states[0];

	cpuidle_set_statedata(state, NULL);

	snprintf(state->name, CPUIDLE_NAME_LEN, "C0");
	snprintf(state->desc, CPUIDLE_DESC_LEN, "CPUIDLE CORE POLL IDLE");
	state->exit_latency = 0;
	state->target_residency = 0;
	state->power_usage = -1;
	state->flags = CPUIDLE_FLAG_POLL;
	state->enter = poll_idle;
}
#else
static void poll_idle_init(struct cpuidle_device *dev) {}
#endif /* CONFIG_ARCH_HAS_CPU_RELAX */

/**
 * __cpuidle_register_device - internal register function called before register
 * and enable routines
 * @dev: the cpu
 *
 * cpuidle_lock mutex must be held before this is called
 */
static int __cpuidle_register_device(struct cpuidle_device *dev)
{
	int ret;
	struct sys_device *sys_dev = get_cpu_sysdev((unsigned long)dev->cpu);
	struct cpuidle_driver *cpuidle_driver = cpuidle_get_driver();

	if (!sys_dev)
		return -EINVAL;
	if (!try_module_get(cpuidle_driver->owner))
		return -EINVAL;

	init_completion(&dev->kobj_unregister);

	poll_idle_init(dev);

	/*
	 * cpuidle driver should set the dev->power_specified bit
	 * before registering the device if the driver provides
	 * power_usage numbers.
	 *
	 * For those devices whose ->power_specified is not set,
	 * we fill in power_usage with decreasing values as the
	 * cpuidle code has an implicit assumption that state Cn
	 * uses less power than C(n-1).
	 *
	 * With CONFIG_ARCH_HAS_CPU_RELAX, C0 is already assigned
	 * an power value of -1.  So we use -2, -3, etc, for other
	 * c-states.
	 */
	if (!dev->power_specified) {
		int i;
		for (i = CPUIDLE_DRIVER_STATE_START; i < dev->state_count; i++)
			dev->states[i].power_usage = -1 - i;
	}

	per_cpu(cpuidle_devices, dev->cpu) = dev;
	list_add(&dev->device_list, &cpuidle_detected_devices);
	if ((ret = cpuidle_add_sysfs(sys_dev))) {
		module_put(cpuidle_driver->owner);
		return ret;
	}

	dev->registered = 1;
	return 0;
}

/**
 * cpuidle_register_device - registers a CPU's idle PM feature
 * @dev: the cpu
 */
int cpuidle_register_device(struct cpuidle_device *dev)
{
	int ret;

	mutex_lock(&cpuidle_lock);

	if ((ret = __cpuidle_register_device(dev))) {
		mutex_unlock(&cpuidle_lock);
		return ret;
	}

	cpuidle_enable_device(dev);
	cpuidle_install_idle_handler();

	mutex_unlock(&cpuidle_lock);

	return 0;

}

EXPORT_SYMBOL_GPL(cpuidle_register_device);

/**
 * cpuidle_unregister_device - unregisters a CPU's idle PM feature
 * @dev: the cpu
 */
void cpuidle_unregister_device(struct cpuidle_device *dev)
{
	struct sys_device *sys_dev = get_cpu_sysdev((unsigned long)dev->cpu);
	struct cpuidle_driver *cpuidle_driver = cpuidle_get_driver();

	if (dev->registered == 0)
		return;

	cpuidle_pause_and_lock();

	cpuidle_disable_device(dev);

	cpuidle_remove_sysfs(sys_dev);
	list_del(&dev->device_list);
	wait_for_completion(&dev->kobj_unregister);
	per_cpu(cpuidle_devices, dev->cpu) = NULL;

	cpuidle_resume_and_unlock();

	module_put(cpuidle_driver->owner);
}

EXPORT_SYMBOL_GPL(cpuidle_unregister_device);

#ifdef CONFIG_SMP

static void smp_callback(void *v)
{
	/* we already woke the CPU up, nothing more to do */
}

/*
 * This function gets called when a part of the kernel has a new latency
 * requirement.  This means we need to get all processors out of their C-state,
 * and then recalculate a new suitable C-state. Just do a cross-cpu IPI; that
 * wakes them all right up.
 */
static int cpuidle_latency_notify(struct notifier_block *b,
		unsigned long l, void *v)
{
	smp_call_function(smp_callback, NULL, 1);
	return NOTIFY_OK;
}

static struct notifier_block cpuidle_latency_notifier = {
	.notifier_call = cpuidle_latency_notify,
};

static inline void latency_notifier_init(struct notifier_block *n)
{
	pm_qos_add_notifier(PM_QOS_CPU_DMA_LATENCY, n);
}

#else /* CONFIG_SMP */

#define latency_notifier_init(x) do { } while (0)

#endif /* CONFIG_SMP */

/**
 * cpuidle_init - core initializer
 */
static int __init cpuidle_init(void)
{
	int ret;

	pm_idle_old = pm_idle;

	ret = cpuidle_add_class_sysfs(&cpu_sysdev_class);
	if (ret)
		return ret;

	latency_notifier_init(&cpuidle_latency_notifier);

	return 0;
}

core_initcall(cpuidle_init);
Commit	Line	Data
4f86d3a8 LB	1	/*
	2	* cpuidle.c - core cpuidle infrastructure
	3	*
	4	* (C) 2006-2007 Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
	5	* Shaohua Li <shaohua.li@intel.com>
	6	* Adam Belay <abelay@novell.com>
	7	*
	8	* This code is licenced under the GPL.
	9	*/
	10
	11	#include <linux/kernel.h>
	12	#include <linux/mutex.h>
	13	#include <linux/sched.h>
	14	#include <linux/notifier.h>
d82b3518	15	#include <linux/pm_qos_params.h>
4f86d3a8 LB	16	#include <linux/cpu.h>
4f86d3a8 LB	17	#include <linux/cpuidle.h>
9a0b8415	18	#include <linux/ktime.h>
2e94d1f7	19	#include <linux/hrtimer.h>
288f023e	20	#include <trace/events/power.h>
4f86d3a8 LB	21
	22	#include "cpuidle.h"
	23
	24	DEFINE_PER_CPU(struct cpuidle_device *, cpuidle_devices);
4f86d3a8 LB	25
	26	DEFINE_MUTEX(cpuidle_lock);
	27	LIST_HEAD(cpuidle_detected_devices);
	28	static void (*pm_idle_old)(void);
	29
	30	static int enabled_devices;
	31
a6869cc4 VP	32	#if defined(CONFIG_ARCH_HAS_CPU_IDLE_WAIT)
	33	static void cpuidle_kick_cpus(void)
	34	{
	35	cpu_idle_wait();
	36	}
	37	#elif defined(CONFIG_SMP)
	38	# error "Arch needs cpu_idle_wait() equivalent here"
	39	#else /* !CONFIG_ARCH_HAS_CPU_IDLE_WAIT && !CONFIG_SMP */
	40	static void cpuidle_kick_cpus(void) {}
	41	#endif
	42
dcb84f33 VP	43	static int __cpuidle_register_device(struct cpuidle_device *dev);
dcb84f33 VP	44
4f86d3a8 LB	45	/**
	46	* cpuidle_idle_call - the main idle loop
	47	*
	48	* NOTE: no locks or semaphores should be used here
	49	*/
	50	static void cpuidle_idle_call(void)
	51	{
	52	struct cpuidle_device *dev = __get_cpu_var(cpuidle_devices);
	53	struct cpuidle_state *target_state;
	54	int next_state;
	55
	56	/* check if the device is ready */
	57	if (!dev \|\| !dev->enabled) {
	58	if (pm_idle_old)
	59	pm_idle_old();
	60	else
89cedfef VP	61	#if defined(CONFIG_ARCH_HAS_DEFAULT_IDLE)
	62	default_idle();
	63	#else
4f86d3a8	64	local_irq_enable();
89cedfef	65	#endif
4f86d3a8 LB	66	return;
	67	}
	68
9a655837 AV	69	#if 0
9a655837 AV	70	/* shows regressions, re-enable for 2.6.29 */
2e94d1f7 AV	71	/*
	72	* run any timers that can be run now, at this point
	73	* before calculating the idle duration etc.
	74	*/
	75	hrtimer_peek_ahead_timers();
9a655837	76	#endif
71abbbf8 AL	77
	78	/*
	79	* Call the device's prepare function before calling the
	80	* governor's select function. ->prepare gives the device's
	81	* cpuidle driver a chance to update any dynamic information
	82	* of its cpuidle states for the current idle period, e.g.
	83	* state availability, latencies, residencies, etc.
	84	*/
	85	if (dev->prepare)
	86	dev->prepare(dev);
	87
4f86d3a8 LB	88	/* ask the governor for the next state */
4f86d3a8 LB	89	next_state = cpuidle_curr_governor->select(dev);
246eb7f0 KH	90	if (need_resched()) {
246eb7f0 KH	91	local_irq_enable();
4f86d3a8	92	return;
246eb7f0 KH	93	}
246eb7f0 KH	94
4f86d3a8 LB	95	target_state = &dev->states[next_state];
	96
	97	/* enter the state and update stats */
4f86d3a8	98	dev->last_state = target_state;
887e301a VP	99	dev->last_residency = target_state->enter(dev, target_state);
	100	if (dev->last_state)
	101	target_state = dev->last_state;
	102
8b78cf60	103	target_state->time += (unsigned long long)dev->last_residency;
4f86d3a8 LB	104	target_state->usage++;
	105
	106	/* give the governor an opportunity to reflect on the outcome */
	107	if (cpuidle_curr_governor->reflect)
	108	cpuidle_curr_governor->reflect(dev);
6f4f2723	109	trace_power_end(smp_processor_id());
4f86d3a8 LB	110	}
	111
	112	/**
	113	* cpuidle_install_idle_handler - installs the cpuidle idle loop handler
	114	*/
	115	void cpuidle_install_idle_handler(void)
	116	{
	117	if (enabled_devices && (pm_idle != cpuidle_idle_call)) {
	118	/* Make sure all changes finished before we switch to new idle */
	119	smp_wmb();
	120	pm_idle = cpuidle_idle_call;
	121	}
	122	}
	123
	124	/**
	125	* cpuidle_uninstall_idle_handler - uninstalls the cpuidle idle loop handler
	126	*/
	127	void cpuidle_uninstall_idle_handler(void)
	128	{
b032bf70	129	if (enabled_devices && pm_idle_old && (pm_idle != pm_idle_old)) {
4f86d3a8	130	pm_idle = pm_idle_old;
a6869cc4	131	cpuidle_kick_cpus();
4f86d3a8 LB	132	}
	133	}
	134
	135	/**
	136	* cpuidle_pause_and_lock - temporarily disables CPUIDLE
	137	*/
	138	void cpuidle_pause_and_lock(void)
	139	{
	140	mutex_lock(&cpuidle_lock);
	141	cpuidle_uninstall_idle_handler();
	142	}
	143
	144	EXPORT_SYMBOL_GPL(cpuidle_pause_and_lock);
	145
	146	/**
	147	* cpuidle_resume_and_unlock - resumes CPUIDLE operation
	148	*/
	149	void cpuidle_resume_and_unlock(void)
	150	{
	151	cpuidle_install_idle_handler();
	152	mutex_unlock(&cpuidle_lock);
	153	}
	154
	155	EXPORT_SYMBOL_GPL(cpuidle_resume_and_unlock);
	156
	157	/**
	158	* cpuidle_enable_device - enables idle PM for a CPU
	159	* @dev: the CPU
	160	*
	161	* This function must be called between cpuidle_pause_and_lock and
	162	* cpuidle_resume_and_unlock when used externally.
	163	*/
	164	int cpuidle_enable_device(struct cpuidle_device *dev)
	165	{
	166	int ret, i;
	167
	168	if (dev->enabled)
	169	return 0;
752138df	170	if (!cpuidle_get_driver() \|\| !cpuidle_curr_governor)
4f86d3a8 LB	171	return -EIO;
	172	if (!dev->state_count)
	173	return -EINVAL;
	174
dcb84f33 VP	175	if (dev->registered == 0) {
	176	ret = __cpuidle_register_device(dev);
	177	if (ret)
	178	return ret;
	179	}
	180
4f86d3a8 LB	181	if ((ret = cpuidle_add_state_sysfs(dev)))
	182	return ret;
	183
	184	if (cpuidle_curr_governor->enable &&
	185	(ret = cpuidle_curr_governor->enable(dev)))
	186	goto fail_sysfs;
	187
	188	for (i = 0; i < dev->state_count; i++) {
	189	dev->states[i].usage = 0;
	190	dev->states[i].time = 0;
	191	}
	192	dev->last_residency = 0;
	193	dev->last_state = NULL;
	194
	195	smp_wmb();
	196
	197	dev->enabled = 1;
	198
	199	enabled_devices++;
	200	return 0;
	201
	202	fail_sysfs:
	203	cpuidle_remove_state_sysfs(dev);
	204
	205	return ret;
	206	}
	207
	208	EXPORT_SYMBOL_GPL(cpuidle_enable_device);
	209
	210	/**
	211	* cpuidle_disable_device - disables idle PM for a CPU
	212	* @dev: the CPU
	213	*
	214	* This function must be called between cpuidle_pause_and_lock and
	215	* cpuidle_resume_and_unlock when used externally.
	216	*/
	217	void cpuidle_disable_device(struct cpuidle_device *dev)
	218	{
	219	if (!dev->enabled)
	220	return;
752138df	221	if (!cpuidle_get_driver() \|\| !cpuidle_curr_governor)
4f86d3a8 LB	222	return;
	223
	224	dev->enabled = 0;
	225
	226	if (cpuidle_curr_governor->disable)
	227	cpuidle_curr_governor->disable(dev);
	228
	229	cpuidle_remove_state_sysfs(dev);
	230	enabled_devices--;
	231	}
	232
	233	EXPORT_SYMBOL_GPL(cpuidle_disable_device);
	234
9a0b8415	235	#ifdef CONFIG_ARCH_HAS_CPU_RELAX
	236	static int poll_idle(struct cpuidle_device dev, struct cpuidle_state st)
	237	{
	238	ktime_t t1, t2;
	239	s64 diff;
	240	int ret;
	241
	242	t1 = ktime_get();
	243	local_irq_enable();
	244	while (!need_resched())
	245	cpu_relax();
	246
	247	t2 = ktime_get();
	248	diff = ktime_to_us(ktime_sub(t2, t1));
	249	if (diff > INT_MAX)
	250	diff = INT_MAX;
	251
	252	ret = (int) diff;
	253	return ret;
	254	}
	255
	256	static void poll_idle_init(struct cpuidle_device *dev)
	257	{
	258	struct cpuidle_state *state = &dev->states[0];
	259
	260	cpuidle_set_statedata(state, NULL);
	261
4fcb2fcd VP	262	snprintf(state->name, CPUIDLE_NAME_LEN, "C0");
4fcb2fcd VP	263	snprintf(state->desc, CPUIDLE_DESC_LEN, "CPUIDLE CORE POLL IDLE");
9a0b8415	264	state->exit_latency = 0;
	265	state->target_residency = 0;
	266	state->power_usage = -1;
8e92b660	267	state->flags = CPUIDLE_FLAG_POLL;
9a0b8415	268	state->enter = poll_idle;
	269	}
	270	#else
	271	static void poll_idle_init(struct cpuidle_device *dev) {}
	272	#endif /* CONFIG_ARCH_HAS_CPU_RELAX */
	273
4f86d3a8	274	/**
dcb84f33 VP	275	* __cpuidle_register_device - internal register function called before register
dcb84f33 VP	276	* and enable routines
4f86d3a8	277	* @dev: the cpu
dcb84f33 VP	278	*
dcb84f33 VP	279	* cpuidle_lock mutex must be held before this is called
4f86d3a8	280	*/
dcb84f33	281	static int __cpuidle_register_device(struct cpuidle_device *dev)
4f86d3a8 LB	282	{
	283	int ret;
	284	struct sys_device *sys_dev = get_cpu_sysdev((unsigned long)dev->cpu);
752138df	285	struct cpuidle_driver *cpuidle_driver = cpuidle_get_driver();
4f86d3a8 LB	286
	287	if (!sys_dev)
	288	return -EINVAL;
752138df	289	if (!try_module_get(cpuidle_driver->owner))
4f86d3a8 LB	290	return -EINVAL;
	291
	292	init_completion(&dev->kobj_unregister);
	293
9a0b8415	294	poll_idle_init(dev);
9a0b8415	295
71abbbf8 AL	296	/*
	297	* cpuidle driver should set the dev->power_specified bit
	298	* before registering the device if the driver provides
	299	* power_usage numbers.
	300	*
	301	* For those devices whose ->power_specified is not set,
	302	* we fill in power_usage with decreasing values as the
	303	* cpuidle code has an implicit assumption that state Cn
	304	* uses less power than C(n-1).
	305	*
	306	* With CONFIG_ARCH_HAS_CPU_RELAX, C0 is already assigned
	307	* an power value of -1. So we use -2, -3, etc, for other
	308	* c-states.
	309	*/
	310	if (!dev->power_specified) {
	311	int i;
	312	for (i = CPUIDLE_DRIVER_STATE_START; i < dev->state_count; i++)
	313	dev->states[i].power_usage = -1 - i;
	314	}
	315
4f86d3a8 LB	316	per_cpu(cpuidle_devices, dev->cpu) = dev;
	317	list_add(&dev->device_list, &cpuidle_detected_devices);
	318	if ((ret = cpuidle_add_sysfs(sys_dev))) {
752138df	319	module_put(cpuidle_driver->owner);
4f86d3a8 LB	320	return ret;
	321	}
	322
dcb84f33 VP	323	dev->registered = 1;
	324	return 0;
	325	}
	326
	327	/**
	328	* cpuidle_register_device - registers a CPU's idle PM feature
	329	* @dev: the cpu
	330	*/
	331	int cpuidle_register_device(struct cpuidle_device *dev)
	332	{
	333	int ret;
	334
	335	mutex_lock(&cpuidle_lock);
	336
	337	if ((ret = __cpuidle_register_device(dev))) {
	338	mutex_unlock(&cpuidle_lock);
	339	return ret;
	340	}
	341
4f86d3a8 LB	342	cpuidle_enable_device(dev);
	343	cpuidle_install_idle_handler();
	344
	345	mutex_unlock(&cpuidle_lock);
	346
	347	return 0;
	348
	349	}
	350
	351	EXPORT_SYMBOL_GPL(cpuidle_register_device);
	352
	353	/**
	354	* cpuidle_unregister_device - unregisters a CPU's idle PM feature
	355	* @dev: the cpu
	356	*/
	357	void cpuidle_unregister_device(struct cpuidle_device *dev)
	358	{
	359	struct sys_device *sys_dev = get_cpu_sysdev((unsigned long)dev->cpu);
752138df	360	struct cpuidle_driver *cpuidle_driver = cpuidle_get_driver();
4f86d3a8	361
dcb84f33 VP	362	if (dev->registered == 0)
	363	return;
	364
4f86d3a8 LB	365	cpuidle_pause_and_lock();
	366
	367	cpuidle_disable_device(dev);
	368
	369	cpuidle_remove_sysfs(sys_dev);
	370	list_del(&dev->device_list);
	371	wait_for_completion(&dev->kobj_unregister);
	372	per_cpu(cpuidle_devices, dev->cpu) = NULL;
	373
	374	cpuidle_resume_and_unlock();
	375
752138df	376	module_put(cpuidle_driver->owner);
4f86d3a8 LB	377	}
	378
	379	EXPORT_SYMBOL_GPL(cpuidle_unregister_device);
	380
	381	#ifdef CONFIG_SMP
	382
	383	static void smp_callback(void *v)
	384	{
	385	/* we already woke the CPU up, nothing more to do */
	386	}
	387
	388	/*
	389	* This function gets called when a part of the kernel has a new latency
	390	* requirement. This means we need to get all processors out of their C-state,
	391	* and then recalculate a new suitable C-state. Just do a cross-cpu IPI; that
	392	* wakes them all right up.
	393	*/
	394	static int cpuidle_latency_notify(struct notifier_block *b,
	395	unsigned long l, void *v)
	396	{
8691e5a8	397	smp_call_function(smp_callback, NULL, 1);
4f86d3a8 LB	398	return NOTIFY_OK;
	399	}
	400
	401	static struct notifier_block cpuidle_latency_notifier = {
	402	.notifier_call = cpuidle_latency_notify,
	403	};
	404
d82b3518 MG	405	static inline void latency_notifier_init(struct notifier_block *n)
	406	{
	407	pm_qos_add_notifier(PM_QOS_CPU_DMA_LATENCY, n);
	408	}
4f86d3a8 LB	409
	410	#else /* CONFIG_SMP */
	411
	412	#define latency_notifier_init(x) do { } while (0)
	413
	414	#endif /* CONFIG_SMP */
	415
	416	/**
	417	* cpuidle_init - core initializer
	418	*/
	419	static int __init cpuidle_init(void)
	420	{
	421	int ret;
	422
	423	pm_idle_old = pm_idle;
	424
	425	ret = cpuidle_add_class_sysfs(&cpu_sysdev_class);
	426	if (ret)
	427	return ret;
	428
	429	latency_notifier_init(&cpuidle_latency_notifier);
	430
	431	return 0;
	432	}
	433
	434	core_initcall(cpuidle_init);