cpufreq.c

/*
 *  linux/drivers/cpufreq/cpufreq.c
 *
 *  Copyright (C) 2001 Russell King
 *            (C) 2002 - 2003 Dominik Brodowski <linux@brodo.de>
 *
 *  Oct 2005 - Ashok Raj <ashok.raj@intel.com>
 *	Added handling for CPU hotplug
 *  Feb 2006 - Jacob Shin <jacob.shin@amd.com>
 *	Fix handling for CPU hotplug -- affected CPUs
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 *
 */

#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/notifier.h>
#include <linux/cpufreq.h>
#include <linux/delay.h>
#include <linux/interrupt.h>
#include <linux/spinlock.h>
#include <linux/device.h>
#include <linux/slab.h>
#include <linux/cpu.h>
#include <linux/completion.h>
#include <linux/mutex.h>

#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_CORE, \
						"cpufreq-core", msg)

/**
 * The "cpufreq driver" - the arch- or hardware-dependent low
 * level driver of CPUFreq support, and its spinlock. This lock
 * also protects the cpufreq_cpu_data array.
 */
static struct cpufreq_driver *cpufreq_driver;
static DEFINE_PER_CPU(struct cpufreq_policy *, cpufreq_cpu_data);
#ifdef CONFIG_HOTPLUG_CPU
/* This one keeps track of the previously set governor of a removed CPU */
static DEFINE_PER_CPU(struct cpufreq_governor *, cpufreq_cpu_governor);
#endif
static DEFINE_SPINLOCK(cpufreq_driver_lock);

/*
 * cpu_policy_rwsem is a per CPU reader-writer semaphore designed to cure
 * all cpufreq/hotplug/workqueue/etc related lock issues.
 *
 * The rules for this semaphore:
 * - Any routine that wants to read from the policy structure will
 *   do a down_read on this semaphore.
 * - Any routine that will write to the policy structure and/or may take away
 *   the policy altogether (eg. CPU hotplug), will hold this lock in write
 *   mode before doing so.
 *
 * Additional rules:
 * - All holders of the lock should check to make sure that the CPU they
 *   are concerned with are online after they get the lock.
 * - Governor routines that can be called in cpufreq hotplug path should not
 *   take this sem as top level hotplug notifier handler takes this.
 */
static DEFINE_PER_CPU(int, policy_cpu);
static DEFINE_PER_CPU(struct rw_semaphore, cpu_policy_rwsem);

#define lock_policy_rwsem(mode, cpu)					\
int lock_policy_rwsem_##mode						\
(int cpu)								\
{									\
	int policy_cpu = per_cpu(policy_cpu, cpu);			\
	BUG_ON(policy_cpu == -1);					\
	down_##mode(&per_cpu(cpu_policy_rwsem, policy_cpu));		\
	if (unlikely(!cpu_online(cpu))) {				\
		up_##mode(&per_cpu(cpu_policy_rwsem, policy_cpu));	\
		return -1;						\
	}								\
									\
	return 0;							\
}

lock_policy_rwsem(read, cpu);
EXPORT_SYMBOL_GPL(lock_policy_rwsem_read);

lock_policy_rwsem(write, cpu);
EXPORT_SYMBOL_GPL(lock_policy_rwsem_write);

void unlock_policy_rwsem_read(int cpu)
{
	int policy_cpu = per_cpu(policy_cpu, cpu);
	BUG_ON(policy_cpu == -1);
	up_read(&per_cpu(cpu_policy_rwsem, policy_cpu));
}
EXPORT_SYMBOL_GPL(unlock_policy_rwsem_read);

void unlock_policy_rwsem_write(int cpu)
{
	int policy_cpu = per_cpu(policy_cpu, cpu);
	BUG_ON(policy_cpu == -1);
	up_write(&per_cpu(cpu_policy_rwsem, policy_cpu));
}
EXPORT_SYMBOL_GPL(unlock_policy_rwsem_write);


/* internal prototypes */
static int __cpufreq_governor(struct cpufreq_policy *policy,
		unsigned int event);
static unsigned int __cpufreq_get(unsigned int cpu);
static void handle_update(struct work_struct *work);

/**
 * Two notifier lists: the "policy" list is involved in the
 * validation process for a new CPU frequency policy; the
 * "transition" list for kernel code that needs to handle
 * changes to devices when the CPU clock speed changes.
 * The mutex locks both lists.
 */
static BLOCKING_NOTIFIER_HEAD(cpufreq_policy_notifier_list);
static struct srcu_notifier_head cpufreq_transition_notifier_list;

static bool init_cpufreq_transition_notifier_list_called;
static int __init init_cpufreq_transition_notifier_list(void)
{
	srcu_init_notifier_head(&cpufreq_transition_notifier_list);
	init_cpufreq_transition_notifier_list_called = true;
	return 0;
}
pure_initcall(init_cpufreq_transition_notifier_list);

static LIST_HEAD(cpufreq_governor_list);
static DEFINE_MUTEX(cpufreq_governor_mutex);

struct cpufreq_policy *cpufreq_cpu_get(unsigned int cpu)
{
	struct cpufreq_policy *data;
	unsigned long flags;

	if (cpu >= nr_cpu_ids)
		goto err_out;

	/* get the cpufreq driver */
	spin_lock_irqsave(&cpufreq_driver_lock, flags);

	if (!cpufreq_driver)
		goto err_out_unlock;

	if (!try_module_get(cpufreq_driver->owner))
		goto err_out_unlock;


	/* get the CPU */
	data = per_cpu(cpufreq_cpu_data, cpu);

	if (!data)
		goto err_out_put_module;

	if (!kobject_get(&data->kobj))
		goto err_out_put_module;

	spin_unlock_irqrestore(&cpufreq_driver_lock, flags);
	return data;

err_out_put_module:
	module_put(cpufreq_driver->owner);
err_out_unlock:
	spin_unlock_irqrestore(&cpufreq_driver_lock, flags);
err_out:
	return NULL;
}
EXPORT_SYMBOL_GPL(cpufreq_cpu_get);


void cpufreq_cpu_put(struct cpufreq_policy *data)
{
	kobject_put(&data->kobj);
	module_put(cpufreq_driver->owner);
}
EXPORT_SYMBOL_GPL(cpufreq_cpu_put);


/*********************************************************************
 *                     UNIFIED DEBUG HELPERS                         *
 *********************************************************************/
#ifdef CONFIG_CPU_FREQ_DEBUG

/* what part(s) of the CPUfreq subsystem are debugged? */
static unsigned int debug;

/* is the debug output ratelimit'ed using printk_ratelimit? User can
 * set or modify this value.
 */
static unsigned int debug_ratelimit = 1;

/* is the printk_ratelimit'ing enabled? It's enabled after a successful
 * loading of a cpufreq driver, temporarily disabled when a new policy
 * is set, and disabled upon cpufreq driver removal
 */
static unsigned int disable_ratelimit = 1;
static DEFINE_SPINLOCK(disable_ratelimit_lock);

static void cpufreq_debug_enable_ratelimit(void)
{
	unsigned long flags;

	spin_lock_irqsave(&disable_ratelimit_lock, flags);
	if (disable_ratelimit)
		disable_ratelimit--;
	spin_unlock_irqrestore(&disable_ratelimit_lock, flags);
}

static void cpufreq_debug_disable_ratelimit(void)
{
	unsigned long flags;

	spin_lock_irqsave(&disable_ratelimit_lock, flags);
	disable_ratelimit++;
	spin_unlock_irqrestore(&disable_ratelimit_lock, flags);
}

void cpufreq_debug_printk(unsigned int type, const char *prefix,
			const char *fmt, ...)
{
	char s[256];
	va_list args;
	unsigned int len;
	unsigned long flags;

	WARN_ON(!prefix);
	if (type & debug) {
		spin_lock_irqsave(&disable_ratelimit_lock, flags);
		if (!disable_ratelimit && debug_ratelimit
					&& !printk_ratelimit()) {
			spin_unlock_irqrestore(&disable_ratelimit_lock, flags);
			return;
		}
		spin_unlock_irqrestore(&disable_ratelimit_lock, flags);

		len = snprintf(s, 256, KERN_DEBUG "%s: ", prefix);

		va_start(args, fmt);
		len += vsnprintf(&s[len], (256 - len), fmt, args);
		va_end(args);

		printk(s);

		WARN_ON(len < 5);
	}
}
EXPORT_SYMBOL(cpufreq_debug_printk);


module_param(debug, uint, 0644);
MODULE_PARM_DESC(debug, "CPUfreq debugging: add 1 to debug core,"
			" 2 to debug drivers, and 4 to debug governors.");

module_param(debug_ratelimit, uint, 0644);
MODULE_PARM_DESC(debug_ratelimit, "CPUfreq debugging:"
					" set to 0 to disable ratelimiting.");

#else /* !CONFIG_CPU_FREQ_DEBUG */

static inline void cpufreq_debug_enable_ratelimit(void) { return; }
static inline void cpufreq_debug_disable_ratelimit(void) { return; }

#endif /* CONFIG_CPU_FREQ_DEBUG */


/*********************************************************************
 *            EXTERNALLY AFFECTING FREQUENCY CHANGES                 *
 *********************************************************************/

/**
 * adjust_jiffies - adjust the system "loops_per_jiffy"
 *
 * This function alters the system "loops_per_jiffy" for the clock
 * speed change. Note that loops_per_jiffy cannot be updated on SMP
 * systems as each CPU might be scaled differently. So, use the arch
 * per-CPU loops_per_jiffy value wherever possible.
 */
#ifndef CONFIG_SMP
static unsigned long l_p_j_ref;
static unsigned int  l_p_j_ref_freq;

static void adjust_jiffies(unsigned long val, struct cpufreq_freqs *ci)
{
	if (ci->flags & CPUFREQ_CONST_LOOPS)
		return;

	if (!l_p_j_ref_freq) {
		l_p_j_ref = loops_per_jiffy;
		l_p_j_ref_freq = ci->old;
		dprintk("saving %lu as reference value for loops_per_jiffy; "
			"freq is %u kHz\n", l_p_j_ref, l_p_j_ref_freq);
	}
	if ((val == CPUFREQ_PRECHANGE  && ci->old < ci->new) ||
	    (val == CPUFREQ_POSTCHANGE && ci->old > ci->new) ||
	    (val == CPUFREQ_RESUMECHANGE || val == CPUFREQ_SUSPENDCHANGE)) {
		loops_per_jiffy = cpufreq_scale(l_p_j_ref, l_p_j_ref_freq,
								ci->new);
		dprintk("scaling loops_per_jiffy to %lu "
			"for frequency %u kHz\n", loops_per_jiffy, ci->new);
	}
}
#else
static inline void adjust_jiffies(unsigned long val, struct cpufreq_freqs *ci)
{
	return;
}
#endif


/**
 * cpufreq_notify_transition - call notifier chain and adjust_jiffies
 * on frequency transition.
 *
 * This function calls the transition notifiers and the "adjust_jiffies"
 * function. It is called twice on all CPU frequency changes that have
 * external effects.
 */
void cpufreq_notify_transition(struct cpufreq_freqs *freqs, unsigned int state)
{
	struct cpufreq_policy *policy;

	BUG_ON(irqs_disabled());

	freqs->flags = cpufreq_driver->flags;
	dprintk("notification %u of frequency transition to %u kHz\n",
		state, freqs->new);

	policy = per_cpu(cpufreq_cpu_data, freqs->cpu);
	switch (state) {

	case CPUFREQ_PRECHANGE:
		/* detect if the driver reported a value as "old frequency"
		 * which is not equal to what the cpufreq core thinks is
		 * "old frequency".
		 */
		if (!(cpufreq_driver->flags & CPUFREQ_CONST_LOOPS)) {
			if ((policy) && (policy->cpu == freqs->cpu) &&
			    (policy->cur) && (policy->cur != freqs->old)) {
				dprintk("Warning: CPU frequency is"
					" %u, cpufreq assumed %u kHz.\n",
					freqs->old, policy->cur);
				freqs->old = policy->cur;
			}
		}
		srcu_notifier_call_chain(&cpufreq_transition_notifier_list,
				CPUFREQ_PRECHANGE, freqs);
		adjust_jiffies(CPUFREQ_PRECHANGE, freqs);
		break;

	case CPUFREQ_POSTCHANGE:
		adjust_jiffies(CPUFREQ_POSTCHANGE, freqs);
		srcu_notifier_call_chain(&cpufreq_transition_notifier_list,
				CPUFREQ_POSTCHANGE, freqs);
		if (likely(policy) && likely(policy->cpu == freqs->cpu))
			policy->cur = freqs->new;
		break;
	}
}
EXPORT_SYMBOL_GPL(cpufreq_notify_transition);



/*********************************************************************
 *                          SYSFS INTERFACE                          *
 *********************************************************************/

static struct cpufreq_governor *__find_governor(const char *str_governor)
{
	struct cpufreq_governor *t;

	list_for_each_entry(t, &cpufreq_governor_list, governor_list)
		if (!strnicmp(str_governor, t->name, CPUFREQ_NAME_LEN))
			return t;

	return NULL;
}

/**
 * cpufreq_parse_governor - parse a governor string
 */
static int cpufreq_parse_governor(char *str_governor, unsigned int *policy,
				struct cpufreq_governor **governor)
{
	int err = -EINVAL;

	if (!cpufreq_driver)
		goto out;

	if (cpufreq_driver->setpolicy) {
		if (!strnicmp(str_governor, "performance", CPUFREQ_NAME_LEN)) {
			*policy = CPUFREQ_POLICY_PERFORMANCE;
			err = 0;
		} else if (!strnicmp(str_governor, "powersave",
						CPUFREQ_NAME_LEN)) {
			*policy = CPUFREQ_POLICY_POWERSAVE;
			err = 0;
		}
	} else if (cpufreq_driver->target) {
		struct cpufreq_governor *t;

		mutex_lock(&cpufreq_governor_mutex);

		t = __find_governor(str_governor);

		if (t == NULL) {
			char *name = kasprintf(GFP_KERNEL, "cpufreq_%s",
								str_governor);

			if (name) {
				int ret;

				mutex_unlock(&cpufreq_governor_mutex);
				ret = request_module("%s", name);
				mutex_lock(&cpufreq_governor_mutex);

				if (ret == 0)
					t = __find_governor(str_governor);
			}

			kfree(name);
		}

		if (t != NULL) {
			*governor = t;
			err = 0;
		}

		mutex_unlock(&cpufreq_governor_mutex);
	}
out:
	return err;
}


/**
 * cpufreq_per_cpu_attr_read() / show_##file_name() -
 * print out cpufreq information
 *
 * Write out information from cpufreq_driver->policy[cpu]; object must be
 * "unsigned int".
 */

#define show_one(file_name, object)			\
static ssize_t show_##file_name				\
(struct cpufreq_policy *policy, char *buf)		\
{							\
	return sprintf(buf, "%u\n", policy->object);	\
}

show_one(cpuinfo_min_freq, cpuinfo.min_freq);
show_one(cpuinfo_max_freq, cpuinfo.max_freq);
show_one(cpuinfo_transition_latency, cpuinfo.transition_latency);
show_one(scaling_min_freq, min);
show_one(scaling_max_freq, max);
show_one(scaling_cur_freq, cur);

static int __cpufreq_set_policy(struct cpufreq_policy *data,
				struct cpufreq_policy *policy);

/**
 * cpufreq_per_cpu_attr_write() / store_##file_name() - sysfs write access
 */
#define store_one(file_name, object)			\
static ssize_t store_##file_name					\
(struct cpufreq_policy *policy, const char *buf, size_t count)		\
{									\
	unsigned int ret = -EINVAL;					\
	struct cpufreq_policy new_policy;				\
									\
	ret = cpufreq_get_policy(&new_policy, policy->cpu);		\
	if (ret)							\
		return -EINVAL;						\
									\
	ret = sscanf(buf, "%u", &new_policy.object);			\
	if (ret != 1)							\
		return -EINVAL;						\
									\
	ret = __cpufreq_set_policy(policy, &new_policy);		\
	policy->user_policy.object = policy->object;			\
									\
	return ret ? ret : count;					\
}

store_one(scaling_min_freq, min);
store_one(scaling_max_freq, max);

/**
 * show_cpuinfo_cur_freq - current CPU frequency as detected by hardware
 */
static ssize_t show_cpuinfo_cur_freq(struct cpufreq_policy *policy,
					char *buf)
{
	unsigned int cur_freq = __cpufreq_get(policy->cpu);
	if (!cur_freq)
		return sprintf(buf, "<unknown>");
	return sprintf(buf, "%u\n", cur_freq);
}


/**
 * show_scaling_governor - show the current policy for the specified CPU
 */
static ssize_t show_scaling_governor(struct cpufreq_policy *policy, char *buf)
{
	if (policy->policy == CPUFREQ_POLICY_POWERSAVE)
		return sprintf(buf, "powersave\n");
	else if (policy->policy == CPUFREQ_POLICY_PERFORMANCE)
		return sprintf(buf, "performance\n");
	else if (policy->governor)
		return scnprintf(buf, CPUFREQ_NAME_LEN, "%s\n",
				policy->governor->name);
	return -EINVAL;
}


/**
 * store_scaling_governor - store policy for the specified CPU
 */
static ssize_t store_scaling_governor(struct cpufreq_policy *policy,
					const char *buf, size_t count)
{
	unsigned int ret = -EINVAL;
	char	str_governor[16];
	struct cpufreq_policy new_policy;

	ret = cpufreq_get_policy(&new_policy, policy->cpu);
	if (ret)
		return ret;

	ret = sscanf(buf, "%15s", str_governor);
	if (ret != 1)
		return -EINVAL;

	if (cpufreq_parse_governor(str_governor, &new_policy.policy,
						&new_policy.governor))
		return -EINVAL;

	/* Do not use cpufreq_set_policy here or the user_policy.max
	   will be wrongly overridden */
	ret = __cpufreq_set_policy(policy, &new_policy);

	policy->user_policy.policy = policy->policy;
	policy->user_policy.governor = policy->governor;

	if (ret)
		return ret;
	else
		return count;
}

/**
 * show_scaling_driver - show the cpufreq driver currently loaded
 */
static ssize_t show_scaling_driver(struct cpufreq_policy *policy, char *buf)
{
	return scnprintf(buf, CPUFREQ_NAME_LEN, "%s\n", cpufreq_driver->name);
}

/**
 * show_scaling_available_governors - show the available CPUfreq governors
 */
static ssize_t show_scaling_available_governors(struct cpufreq_policy *policy,
						char *buf)
{
	ssize_t i = 0;
	struct cpufreq_governor *t;

	if (!cpufreq_driver->target) {
		i += sprintf(buf, "performance powersave");
		goto out;
	}

	list_for_each_entry(t, &cpufreq_governor_list, governor_list) {
		if (i >= (ssize_t) ((PAGE_SIZE / sizeof(char))
		    - (CPUFREQ_NAME_LEN + 2)))
			goto out;
		i += scnprintf(&buf[i], CPUFREQ_NAME_LEN, "%s ", t->name);
	}
out:
	i += sprintf(&buf[i], "\n");
	return i;
}

static ssize_t show_cpus(const struct cpumask *mask, char *buf)
{
	ssize_t i = 0;
	unsigned int cpu;

	for_each_cpu(cpu, mask) {
		if (i)
			i += scnprintf(&buf[i], (PAGE_SIZE - i - 2), " ");
		i += scnprintf(&buf[i], (PAGE_SIZE - i - 2), "%u", cpu);
		if (i >= (PAGE_SIZE - 5))
			break;
	}
	i += sprintf(&buf[i], "\n");
	return i;
}

/**
 * show_related_cpus - show the CPUs affected by each transition even if
 * hw coordination is in use
 */
static ssize_t show_related_cpus(struct cpufreq_policy *policy, char *buf)
{
	if (cpumask_empty(policy->related_cpus))
		return show_cpus(policy->cpus, buf);
	return show_cpus(policy->related_cpus, buf);
}

/**
 * show_affected_cpus - show the CPUs affected by each transition
 */
static ssize_t show_affected_cpus(struct cpufreq_policy *policy, char *buf)
{
	return show_cpus(policy->cpus, buf);
}

static ssize_t store_scaling_setspeed(struct cpufreq_policy *policy,
					const char *buf, size_t count)
{
	unsigned int freq = 0;
	unsigned int ret;

	if (!policy->governor || !policy->governor->store_setspeed)
		return -EINVAL;

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

	policy->governor->store_setspeed(policy, freq);

	return count;
}

static ssize_t show_scaling_setspeed(struct cpufreq_policy *policy, char *buf)
{
	if (!policy->governor || !policy->governor->show_setspeed)
		return sprintf(buf, "<unsupported>\n");

	return policy->governor->show_setspeed(policy, buf);
}

#define define_one_ro(_name) \
static struct freq_attr _name = \
__ATTR(_name, 0444, show_##_name, NULL)

#define define_one_ro0400(_name) \
static struct freq_attr _name = \
__ATTR(_name, 0400, show_##_name, NULL)

#define define_one_rw(_name) \
static struct freq_attr _name = \
__ATTR(_name, 0644, show_##_name, store_##_name)

define_one_ro0400(cpuinfo_cur_freq);
define_one_ro(cpuinfo_min_freq);
define_one_ro(cpuinfo_max_freq);
define_one_ro(cpuinfo_transition_latency);
define_one_ro(scaling_available_governors);
define_one_ro(scaling_driver);
define_one_ro(scaling_cur_freq);
define_one_ro(related_cpus);
define_one_ro(affected_cpus);
define_one_rw(scaling_min_freq);
define_one_rw(scaling_max_freq);
define_one_rw(scaling_governor);
define_one_rw(scaling_setspeed);

static struct attribute *default_attrs[] = {
	&cpuinfo_min_freq.attr,
	&cpuinfo_max_freq.attr,
	&cpuinfo_transition_latency.attr,
	&scaling_min_freq.attr,
	&scaling_max_freq.attr,
	&affected_cpus.attr,
	&related_cpus.attr,
	&scaling_governor.attr,
	&scaling_driver.attr,
	&scaling_available_governors.attr,
	&scaling_setspeed.attr,
	NULL
};

#define to_policy(k) container_of(k, struct cpufreq_policy, kobj)
#define to_attr(a) container_of(a, struct freq_attr, attr)

static ssize_t show(struct kobject *kobj, struct attribute *attr, char *buf)
{
	struct cpufreq_policy *policy = to_policy(kobj);
	struct freq_attr *fattr = to_attr(attr);
	ssize_t ret = -EINVAL;
	policy = cpufreq_cpu_get(policy->cpu);
	if (!policy)
		goto no_policy;

	if (lock_policy_rwsem_read(policy->cpu) < 0)
		goto fail;

	if (fattr->show)
		ret = fattr->show(policy, buf);
	else
		ret = -EIO;

	unlock_policy_rwsem_read(policy->cpu);
fail:
	cpufreq_cpu_put(policy);
no_policy:
	return ret;
}

static ssize_t store(struct kobject *kobj, struct attribute *attr,
		     const char *buf, size_t count)
{
	struct cpufreq_policy *policy = to_policy(kobj);
	struct freq_attr *fattr = to_attr(attr);
	ssize_t ret = -EINVAL;
	policy = cpufreq_cpu_get(policy->cpu);
	if (!policy)
		goto no_policy;

	if (lock_policy_rwsem_write(policy->cpu) < 0)
		goto fail;

	if (fattr->store)
		ret = fattr->store(policy, buf, count);
	else
		ret = -EIO;

	unlock_policy_rwsem_write(policy->cpu);
fail:
	cpufreq_cpu_put(policy);
no_policy:
	return ret;
}

static void cpufreq_sysfs_release(struct kobject *kobj)
{
	struct cpufreq_policy *policy = to_policy(kobj);
	dprintk("last reference is dropped\n");
	complete(&policy->kobj_unregister);
}

static struct sysfs_ops sysfs_ops = {
	.show	= show,
	.store	= store,
};

static struct kobj_type ktype_cpufreq = {
	.sysfs_ops	= &sysfs_ops,
	.default_attrs	= default_attrs,
	.release	= cpufreq_sysfs_release,
};


/**
 * cpufreq_add_dev - add a CPU device
 *
 * Adds the cpufreq interface for a CPU device.
 */
static int cpufreq_add_dev(struct sys_device *sys_dev)
{
	unsigned int cpu = sys_dev->id;
	int ret = 0;
	struct cpufreq_policy new_policy;
	struct cpufreq_policy *policy;
	struct freq_attr **drv_attr;
	struct sys_device *cpu_sys_dev;
	unsigned long flags;
	unsigned int j;
#ifdef CONFIG_SMP
	struct cpufreq_policy *managed_policy;
#endif

	if (cpu_is_offline(cpu))
		return 0;

	cpufreq_debug_disable_ratelimit();
	dprintk("adding CPU %u\n", cpu);

#ifdef CONFIG_SMP
	/* check whether a different CPU already registered this
	 * CPU because it is in the same boat. */
	policy = cpufreq_cpu_get(cpu);
	if (unlikely(policy)) {
		cpufreq_cpu_put(policy);
		cpufreq_debug_enable_ratelimit();
		return 0;
	}
#endif

	if (!try_module_get(cpufreq_driver->owner)) {
		ret = -EINVAL;
		goto module_out;
	}

	policy = kzalloc(sizeof(struct cpufreq_policy), GFP_KERNEL);
	if (!policy) {
		ret = -ENOMEM;
		goto nomem_out;
	}
	if (!alloc_cpumask_var(&policy->cpus, GFP_KERNEL)) {
		kfree(policy);
		ret = -ENOMEM;
		goto nomem_out;
	}
	if (!zalloc_cpumask_var(&policy->related_cpus, GFP_KERNEL)) {
		free_cpumask_var(policy->cpus);
		kfree(policy);
		ret = -ENOMEM;
		goto nomem_out;
	}

	policy->cpu = cpu;
	cpumask_copy(policy->cpus, cpumask_of(cpu));

	/* Initially set CPU itself as the policy_cpu */
	per_cpu(policy_cpu, cpu) = cpu;
	lock_policy_rwsem_write(cpu);

	init_completion(&policy->kobj_unregister);
	INIT_WORK(&policy->update, handle_update);

	/* Set governor before ->init, so that driver could check it */
	policy->governor = CPUFREQ_DEFAULT_GOVERNOR;
	/* call driver. From then on the cpufreq must be able
	 * to accept all calls to ->verify and ->setpolicy for this CPU
	 */
	ret = cpufreq_driver->init(policy);
	if (ret) {
		dprintk("initialization failed\n");
		goto err_out;
	}
	policy->user_policy.min = policy->min;
	policy->user_policy.max = policy->max;

	blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
				     CPUFREQ_START, policy);

#ifdef CONFIG_SMP

#ifdef CONFIG_HOTPLUG_CPU
	if (per_cpu(cpufreq_cpu_governor, cpu)) {
		policy->governor = per_cpu(cpufreq_cpu_governor, cpu);
		dprintk("Restoring governor %s for cpu %d\n",
		       policy->governor->name, cpu);
	}
#endif

	for_each_cpu(j, policy->cpus) {
		if (cpu == j)
			continue;

		/* Check for existing affected CPUs.
		 * They may not be aware of it due to CPU Hotplug.
		 */
		managed_policy = cpufreq_cpu_get(j);		/* FIXME: Where is this released?  What about error paths? */
		if (unlikely(managed_policy)) {

			/* Set proper policy_cpu */
			unlock_policy_rwsem_write(cpu);
			per_cpu(policy_cpu, cpu) = managed_policy->cpu;

			if (lock_policy_rwsem_write(cpu) < 0)
				goto err_out_driver_exit;

			spin_lock_irqsave(&cpufreq_driver_lock, flags);
			cpumask_copy(managed_policy->cpus, policy->cpus);
			per_cpu(cpufreq_cpu_data, cpu) = managed_policy;
			spin_unlock_irqrestore(&cpufreq_driver_lock, flags);

			dprintk("CPU already managed, adding link\n");
			ret = sysfs_create_link(&sys_dev->kobj,
						&managed_policy->kobj,
						"cpufreq");
			if (ret)
				goto err_out_driver_exit;

			cpufreq_debug_enable_ratelimit();
			ret = 0;
			goto err_out_driver_exit; /* call driver->exit() */
		}
	}
#endif
	memcpy(&new_policy, policy, sizeof(struct cpufreq_policy));

	/* prepare interface data */
	ret = kobject_init_and_add(&policy->kobj, &ktype_cpufreq, &sys_dev->kobj,
				   "cpufreq");
	if (ret)
		goto err_out_driver_exit;

	/* set up files for this cpu device */
	drv_attr = cpufreq_driver->attr;
	while ((drv_attr) && (*drv_attr)) {
		ret = sysfs_create_file(&policy->kobj, &((*drv_attr)->attr));
		if (ret)
			goto err_out_driver_exit;
		drv_attr++;
	}
	if (cpufreq_driver->get) {
		ret = sysfs_create_file(&policy->kobj, &cpuinfo_cur_freq.attr);
		if (ret)
			goto err_out_driver_exit;
	}
	if (cpufreq_driver->target) {
		ret = sysfs_create_file(&policy->kobj, &scaling_cur_freq.attr);
		if (ret)
			goto err_out_driver_exit;
	}

	spin_lock_irqsave(&cpufreq_driver_lock, flags);
	for_each_cpu(j, policy->cpus) {
		per_cpu(cpufreq_cpu_data, j) = policy;
		per_cpu(policy_cpu, j) = policy->cpu;
	}
	spin_unlock_irqrestore(&cpufreq_driver_lock, flags);

	/* symlink affected CPUs */
	for_each_cpu(j, policy->cpus) {
		if (j == cpu)
			continue;
		if (!cpu_online(j))
			continue;

		dprintk("CPU %u already managed, adding link\n", j);
		cpufreq_cpu_get(cpu);
		cpu_sys_dev = get_cpu_sysdev(j);
		ret = sysfs_create_link(&cpu_sys_dev->kobj, &policy->kobj,
					"cpufreq");
		if (ret)
			goto err_out_unregister;
	}

	policy->governor = NULL; /* to assure that the starting sequence is
				  * run in cpufreq_set_policy */

	/* set default policy */
	ret = __cpufreq_set_policy(policy, &new_policy);
	policy->user_policy.policy = policy->policy;
	policy->user_policy.governor = policy->governor;

	if (ret) {
		dprintk("setting policy failed\n");
		goto err_out_unregister;
	}

	unlock_policy_rwsem_write(cpu);

	kobject_uevent(&policy->kobj, KOBJ_ADD);
	module_put(cpufreq_driver->owner);
	dprintk("initialization complete\n");
	cpufreq_debug_enable_ratelimit();

	return 0;


err_out_unregister:
	spin_lock_irqsave(&cpufreq_driver_lock, flags);
	for_each_cpu(j, policy->cpus)
		per_cpu(cpufreq_cpu_data, j) = NULL;
	spin_unlock_irqrestore(&cpufreq_driver_lock, flags);

	kobject_put(&policy->kobj);
	wait_for_completion(&policy->kobj_unregister);

err_out_driver_exit:
	if (cpufreq_driver->exit)
		cpufreq_driver->exit(policy);

err_out:
	unlock_policy_rwsem_write(cpu);
	kfree(policy);

nomem_out:
	module_put(cpufreq_driver->owner);
module_out:
	cpufreq_debug_enable_ratelimit();
	return ret;
}


/**
 * __cpufreq_remove_dev - remove a CPU device
 *
 * Removes the cpufreq interface for a CPU device.
 * Caller should already have policy_rwsem in write mode for this CPU.
 * This routine frees the rwsem before returning.
 */
static int __cpufreq_remove_dev(struct sys_device *sys_dev)
{
	unsigned int cpu = sys_dev->id;
	unsigned long flags;
	struct cpufreq_policy *data;
#ifdef CONFIG_SMP
	struct sys_device *cpu_sys_dev;
	unsigned int j;
#endif

	cpufreq_debug_disable_ratelimit();
	dprintk("unregistering CPU %u\n", cpu);

	spin_lock_irqsave(&cpufreq_driver_lock, flags);
	data = per_cpu(cpufreq_cpu_data, cpu);

	if (!data) {
		spin_unlock_irqrestore(&cpufreq_driver_lock, flags);
		cpufreq_debug_enable_ratelimit();
		unlock_policy_rwsem_write(cpu);
		return -EINVAL;
	}
	per_cpu(cpufreq_cpu_data, cpu) = NULL;


#ifdef CONFIG_SMP
	/* if this isn't the CPU which is the parent of the kobj, we
	 * only need to unlink, put and exit
	 */
	if (unlikely(cpu != data->cpu)) {
		dprintk("removing link\n");
		cpumask_clear_cpu(cpu, data->cpus);
		spin_unlock_irqrestore(&cpufreq_driver_lock, flags);
		sysfs_remove_link(&sys_dev->kobj, "cpufreq");
		cpufreq_cpu_put(data);
		cpufreq_debug_enable_ratelimit();
		unlock_policy_rwsem_write(cpu);
		return 0;
	}
#endif

#ifdef CONFIG_SMP

#ifdef CONFIG_HOTPLUG_CPU
	per_cpu(cpufreq_cpu_governor, cpu) = data->governor;
#endif

	/* if we have other CPUs still registered, we need to unlink them,
	 * or else wait_for_completion below will lock up. Clean the
	 * per_cpu(cpufreq_cpu_data) while holding the lock, and remove
	 * the sysfs links afterwards.
	 */
	if (unlikely(cpumask_weight(data->cpus) > 1)) {
		for_each_cpu(j, data->cpus) {
			if (j == cpu)
				continue;
			per_cpu(cpufreq_cpu_data, j) = NULL;
		}
	}

	spin_unlock_irqrestore(&cpufreq_driver_lock, flags);

	if (unlikely(cpumask_weight(data->cpus) > 1)) {
		for_each_cpu(j, data->cpus) {
			if (j == cpu)
				continue;
			dprintk("removing link for cpu %u\n", j);
#ifdef CONFIG_HOTPLUG_CPU
			per_cpu(cpufreq_cpu_governor, j) = data->governor;
#endif
			cpu_sys_dev = get_cpu_sysdev(j);
			sysfs_remove_link(&cpu_sys_dev->kobj, "cpufreq");
			cpufreq_cpu_put(data);
		}
	}
#else
	spin_unlock_irqrestore(&cpufreq_driver_lock, flags);
#endif

	unlock_policy_rwsem_write(cpu);

	if (cpufreq_driver->target)
		__cpufreq_governor(data, CPUFREQ_GOV_STOP);

	kobject_put(&data->kobj);

	/* we need to make sure that the underlying kobj is actually
	 * not referenced anymore by anybody before we proceed with
	 * unloading.
	 */
	dprintk("waiting for dropping of refcount\n");
	wait_for_completion(&data->kobj_unregister);
	dprintk("wait complete\n");

	if (cpufreq_driver->exit)
		cpufreq_driver->exit(data);

	free_cpumask_var(data->related_cpus);
	free_cpumask_var(data->cpus);
	kfree(data);
	per_cpu(cpufreq_cpu_data, cpu) = NULL;

	cpufreq_debug_enable_ratelimit();
	return 0;
}


static int cpufreq_remove_dev(struct sys_device *sys_dev)
{
	unsigned int cpu = sys_dev->id;
	int retval;

	if (cpu_is_offline(cpu))
		return 0;

	if (unlikely(lock_policy_rwsem_write(cpu)))
		BUG();

	retval = __cpufreq_remove_dev(sys_dev);
	return retval;
}


static void handle_update(struct work_struct *work)
{
	struct cpufreq_policy *policy =
		container_of(work, struct cpufreq_policy, update);
	unsigned int cpu = policy->cpu;
	dprintk("handle_update for cpu %u called\n", cpu);
	cpufreq_update_policy(cpu);
}

/**
 *	cpufreq_out_of_sync - If actual and saved CPU frequency differs, we're in deep trouble.
 *	@cpu: cpu number
 *	@old_freq: CPU frequency the kernel thinks the CPU runs at
 *	@new_freq: CPU frequency the CPU actually runs at
 *
 *	We adjust to current frequency first, and need to clean up later.
 *	So either call to cpufreq_update_policy() or schedule handle_update()).
 */
static void cpufreq_out_of_sync(unsigned int cpu, unsigned int old_freq,
				unsigned int new_freq)
{
	struct cpufreq_freqs freqs;

	dprintk("Warning: CPU frequency out of sync: cpufreq and timing "
	       "core thinks of %u, is %u kHz.\n", old_freq, new_freq);

	freqs.cpu = cpu;
	freqs.old = old_freq;
	freqs.new = new_freq;
	cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
	cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE);
}


/**
 * cpufreq_quick_get - get the CPU frequency (in kHz) from policy->cur
 * @cpu: CPU number
 *
 * This is the last known freq, without actually getting it from the driver.
 * Return value will be same as what is shown in scaling_cur_freq in sysfs.
 */
unsigned int cpufreq_quick_get(unsigned int cpu)
{
	struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);
	unsigned int ret_freq = 0;

	if (policy) {
		ret_freq = policy->cur;
		cpufreq_cpu_put(policy);
	}

	return ret_freq;
}
EXPORT_SYMBOL(cpufreq_quick_get);


static unsigned int __cpufreq_get(unsigned int cpu)
{
	struct cpufreq_policy *policy = per_cpu(cpufreq_cpu_data, cpu);
	unsigned int ret_freq = 0;

	if (!cpufreq_driver->get)
		return ret_freq;

	ret_freq = cpufreq_driver->get(cpu);

	if (ret_freq && policy->cur &&
		!(cpufreq_driver->flags & CPUFREQ_CONST_LOOPS)) {
		/* verify no discrepancy between actual and
					saved value exists */
		if (unlikely(ret_freq != policy->cur)) {
			cpufreq_out_of_sync(cpu, policy->cur, ret_freq);
			schedule_work(&policy->update);
		}
	}

	return ret_freq;
}

/**
 * cpufreq_get - get the current CPU frequency (in kHz)
 * @cpu: CPU number
 *
 * Get the CPU current (static) CPU frequency
 */
unsigned int cpufreq_get(unsigned int cpu)
{
	unsigned int ret_freq = 0;
	struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);

	if (!policy)
		goto out;

	if (unlikely(lock_policy_rwsem_read(cpu)))
		goto out_policy;<