cpufreq.c 65.9 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5
/*
 *  linux/drivers/cpufreq/cpufreq.c
 *
 *  Copyright (C) 2001 Russell King
 *            (C) 2002 - 2003 Dominik Brodowski <linux@brodo.de>
6
 *            (C) 2013 Viresh Kumar <viresh.kumar@linaro.org>
L
Linus Torvalds 已提交
7
 *
8
 *  Oct 2005 - Ashok Raj <ashok.raj@intel.com>
9
 *	Added handling for CPU hotplug
10 11
 *  Feb 2006 - Jacob Shin <jacob.shin@amd.com>
 *	Fix handling for CPU hotplug -- affected CPUs
12
 *
L
Linus Torvalds 已提交
13 14 15 16 17
 * 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.
 */

V
Viresh Kumar 已提交
18 19
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

20
#include <linux/cpu.h>
L
Linus Torvalds 已提交
21 22 23
#include <linux/cpufreq.h>
#include <linux/delay.h>
#include <linux/device.h>
24 25 26
#include <linux/init.h>
#include <linux/kernel_stat.h>
#include <linux/module.h>
27
#include <linux/mutex.h>
28
#include <linux/slab.h>
29
#include <linux/suspend.h>
30
#include <linux/syscore_ops.h>
31
#include <linux/tick.h>
32 33
#include <trace/events/power.h>

V
Viresh Kumar 已提交
34
static LIST_HEAD(cpufreq_policy_list);
35 36 37 38 39 40 41

static inline bool policy_is_inactive(struct cpufreq_policy *policy)
{
	return cpumask_empty(policy->cpus);
}

/* Macros to iterate over CPU policies */
42 43 44
#define for_each_suitable_policy(__policy, __active)			 \
	list_for_each_entry(__policy, &cpufreq_policy_list, policy_list) \
		if ((__active) == !policy_is_inactive(__policy))
45 46 47 48 49 50 51

#define for_each_active_policy(__policy)		\
	for_each_suitable_policy(__policy, true)
#define for_each_inactive_policy(__policy)		\
	for_each_suitable_policy(__policy, false)

#define for_each_policy(__policy)			\
V
Viresh Kumar 已提交
52 53
	list_for_each_entry(__policy, &cpufreq_policy_list, policy_list)

54 55 56 57 58
/* Iterate over governors */
static LIST_HEAD(cpufreq_governor_list);
#define for_each_governor(__governor)				\
	list_for_each_entry(__governor, &cpufreq_governor_list, governor_list)

L
Linus Torvalds 已提交
59
/**
D
Dave Jones 已提交
60
 * The "cpufreq driver" - the arch- or hardware-dependent low
L
Linus Torvalds 已提交
61 62 63
 * level driver of CPUFreq support, and its spinlock. This lock
 * also protects the cpufreq_cpu_data array.
 */
64
static struct cpufreq_driver *cpufreq_driver;
65
static DEFINE_PER_CPU(struct cpufreq_policy *, cpufreq_cpu_data);
66 67
static DEFINE_RWLOCK(cpufreq_driver_lock);

68 69
/* Flag to suspend/resume CPUFreq governors */
static bool cpufreq_suspended;
L
Linus Torvalds 已提交
70

71 72 73 74 75
static inline bool has_target(void)
{
	return cpufreq_driver->target_index || cpufreq_driver->target;
}

L
Linus Torvalds 已提交
76
/* internal prototypes */
77
static int cpufreq_governor(struct cpufreq_policy *policy, unsigned int event);
78
static unsigned int __cpufreq_get(struct cpufreq_policy *policy);
79
static int cpufreq_start_governor(struct cpufreq_policy *policy);
L
Linus Torvalds 已提交
80

81
static inline void cpufreq_exit_governor(struct cpufreq_policy *policy)
82
{
83
	(void)cpufreq_governor(policy, CPUFREQ_GOV_POLICY_EXIT);
84
}
L
Linus Torvalds 已提交
85

86 87 88 89 90
static inline void cpufreq_stop_governor(struct cpufreq_policy *policy)
{
	(void)cpufreq_governor(policy, CPUFREQ_GOV_STOP);
}

L
Linus Torvalds 已提交
91
/**
92 93
 * Two notifier lists: the "policy" list is involved in the
 * validation process for a new CPU frequency policy; the
L
Linus Torvalds 已提交
94 95 96 97
 * "transition" list for kernel code that needs to handle
 * changes to devices when the CPU clock speed changes.
 * The mutex locks both lists.
 */
98
static BLOCKING_NOTIFIER_HEAD(cpufreq_policy_notifier_list);
99
static struct srcu_notifier_head cpufreq_transition_notifier_list;
L
Linus Torvalds 已提交
100

101
static bool init_cpufreq_transition_notifier_list_called;
102 103 104
static int __init init_cpufreq_transition_notifier_list(void)
{
	srcu_init_notifier_head(&cpufreq_transition_notifier_list);
105
	init_cpufreq_transition_notifier_list_called = true;
106 107
	return 0;
}
108
pure_initcall(init_cpufreq_transition_notifier_list);
L
Linus Torvalds 已提交
109

110
static int off __read_mostly;
111
static int cpufreq_disabled(void)
112 113 114 115 116 117 118
{
	return off;
}
void disable_cpufreq(void)
{
	off = 1;
}
119
static DEFINE_MUTEX(cpufreq_governor_mutex);
L
Linus Torvalds 已提交
120

121 122
bool have_governor_per_policy(void)
{
123
	return !!(cpufreq_driver->flags & CPUFREQ_HAVE_GOVERNOR_PER_POLICY);
124
}
125
EXPORT_SYMBOL_GPL(have_governor_per_policy);
126

127 128 129 130 131 132 133 134 135
struct kobject *get_governor_parent_kobj(struct cpufreq_policy *policy)
{
	if (have_governor_per_policy())
		return &policy->kobj;
	else
		return cpufreq_global_kobject;
}
EXPORT_SYMBOL_GPL(get_governor_parent_kobj);

136 137 138 139 140 141 142 143 144
struct cpufreq_frequency_table *cpufreq_frequency_get_table(unsigned int cpu)
{
	struct cpufreq_policy *policy = per_cpu(cpufreq_cpu_data, cpu);

	return policy && !policy_is_inactive(policy) ?
		policy->freq_table : NULL;
}
EXPORT_SYMBOL_GPL(cpufreq_frequency_get_table);

145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179
static inline u64 get_cpu_idle_time_jiffy(unsigned int cpu, u64 *wall)
{
	u64 idle_time;
	u64 cur_wall_time;
	u64 busy_time;

	cur_wall_time = jiffies64_to_cputime64(get_jiffies_64());

	busy_time = kcpustat_cpu(cpu).cpustat[CPUTIME_USER];
	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM];
	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_IRQ];
	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SOFTIRQ];
	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_STEAL];
	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_NICE];

	idle_time = cur_wall_time - busy_time;
	if (wall)
		*wall = cputime_to_usecs(cur_wall_time);

	return cputime_to_usecs(idle_time);
}

u64 get_cpu_idle_time(unsigned int cpu, u64 *wall, int io_busy)
{
	u64 idle_time = get_cpu_idle_time_us(cpu, io_busy ? wall : NULL);

	if (idle_time == -1ULL)
		return get_cpu_idle_time_jiffy(cpu, wall);
	else if (!io_busy)
		idle_time += get_cpu_iowait_time_us(cpu, wall);

	return idle_time;
}
EXPORT_SYMBOL_GPL(get_cpu_idle_time);

180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201
/*
 * This is a generic cpufreq init() routine which can be used by cpufreq
 * drivers of SMP systems. It will do following:
 * - validate & show freq table passed
 * - set policies transition latency
 * - policy->cpus with all possible CPUs
 */
int cpufreq_generic_init(struct cpufreq_policy *policy,
		struct cpufreq_frequency_table *table,
		unsigned int transition_latency)
{
	int ret;

	ret = cpufreq_table_validate_and_show(policy, table);
	if (ret) {
		pr_err("%s: invalid frequency table: %d\n", __func__, ret);
		return ret;
	}

	policy->cpuinfo.transition_latency = transition_latency;

	/*
202
	 * The driver only supports the SMP configuration where all processors
203 204 205 206 207 208 209 210
	 * share the clock and voltage and clock.
	 */
	cpumask_setall(policy->cpus);

	return 0;
}
EXPORT_SYMBOL_GPL(cpufreq_generic_init);

211
struct cpufreq_policy *cpufreq_cpu_get_raw(unsigned int cpu)
212 213 214
{
	struct cpufreq_policy *policy = per_cpu(cpufreq_cpu_data, cpu);

215 216
	return policy && cpumask_test_cpu(cpu, policy->cpus) ? policy : NULL;
}
217
EXPORT_SYMBOL_GPL(cpufreq_cpu_get_raw);
218 219 220 221 222

unsigned int cpufreq_generic_get(unsigned int cpu)
{
	struct cpufreq_policy *policy = cpufreq_cpu_get_raw(cpu);

223
	if (!policy || IS_ERR(policy->clk)) {
224 225
		pr_err("%s: No %s associated to cpu: %d\n",
		       __func__, policy ? "clk" : "policy", cpu);
226 227 228 229 230 231 232
		return 0;
	}

	return clk_get_rate(policy->clk) / 1000;
}
EXPORT_SYMBOL_GPL(cpufreq_generic_get);

233 234 235 236 237 238 239 240 241 242 243 244 245
/**
 * cpufreq_cpu_get: returns policy for a cpu and marks it busy.
 *
 * @cpu: cpu to find policy for.
 *
 * This returns policy for 'cpu', returns NULL if it doesn't exist.
 * It also increments the kobject reference count to mark it busy and so would
 * require a corresponding call to cpufreq_cpu_put() to decrement it back.
 * If corresponding call cpufreq_cpu_put() isn't made, the policy wouldn't be
 * freed as that depends on the kobj count.
 *
 * Return: A valid policy on success, otherwise NULL on failure.
 */
246
struct cpufreq_policy *cpufreq_cpu_get(unsigned int cpu)
L
Linus Torvalds 已提交
247
{
248
	struct cpufreq_policy *policy = NULL;
L
Linus Torvalds 已提交
249 250
	unsigned long flags;

251
	if (WARN_ON(cpu >= nr_cpu_ids))
252 253
		return NULL;

L
Linus Torvalds 已提交
254
	/* get the cpufreq driver */
255
	read_lock_irqsave(&cpufreq_driver_lock, flags);
L
Linus Torvalds 已提交
256

257 258
	if (cpufreq_driver) {
		/* get the CPU */
259
		policy = cpufreq_cpu_get_raw(cpu);
260 261 262
		if (policy)
			kobject_get(&policy->kobj);
	}
L
Linus Torvalds 已提交
263

264
	read_unlock_irqrestore(&cpufreq_driver_lock, flags);
L
Linus Torvalds 已提交
265

266
	return policy;
267
}
L
Linus Torvalds 已提交
268 269
EXPORT_SYMBOL_GPL(cpufreq_cpu_get);

270 271 272 273 274 275 276 277
/**
 * cpufreq_cpu_put: Decrements the usage count of a policy
 *
 * @policy: policy earlier returned by cpufreq_cpu_get().
 *
 * This decrements the kobject reference count incremented earlier by calling
 * cpufreq_cpu_get().
 */
278
void cpufreq_cpu_put(struct cpufreq_policy *policy)
L
Linus Torvalds 已提交
279
{
280
	kobject_put(&policy->kobj);
L
Linus Torvalds 已提交
281 282 283 284 285 286 287 288 289 290 291 292
}
EXPORT_SYMBOL_GPL(cpufreq_cpu_put);

/*********************************************************************
 *            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
293
 * systems as each CPU might be scaled differently. So, use the arch
L
Linus Torvalds 已提交
294 295
 * per-CPU loops_per_jiffy value wherever possible.
 */
296
static void adjust_jiffies(unsigned long val, struct cpufreq_freqs *ci)
L
Linus Torvalds 已提交
297
{
298 299 300 301
#ifndef CONFIG_SMP
	static unsigned long l_p_j_ref;
	static unsigned int l_p_j_ref_freq;

L
Linus Torvalds 已提交
302 303 304 305 306 307
	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;
308 309
		pr_debug("saving %lu as reference value for loops_per_jiffy; freq is %u kHz\n",
			 l_p_j_ref, l_p_j_ref_freq);
L
Linus Torvalds 已提交
310
	}
311
	if (val == CPUFREQ_POSTCHANGE && ci->old != ci->new) {
312 313
		loops_per_jiffy = cpufreq_scale(l_p_j_ref, l_p_j_ref_freq,
								ci->new);
314 315
		pr_debug("scaling loops_per_jiffy to %lu for frequency %u kHz\n",
			 loops_per_jiffy, ci->new);
L
Linus Torvalds 已提交
316 317
	}
#endif
318
}
L
Linus Torvalds 已提交
319

320
static void __cpufreq_notify_transition(struct cpufreq_policy *policy,
321
		struct cpufreq_freqs *freqs, unsigned int state)
L
Linus Torvalds 已提交
322 323 324
{
	BUG_ON(irqs_disabled());

325 326 327
	if (cpufreq_disabled())
		return;

328
	freqs->flags = cpufreq_driver->flags;
329
	pr_debug("notification %u of frequency transition to %u kHz\n",
330
		 state, freqs->new);
L
Linus Torvalds 已提交
331 332

	switch (state) {
333

L
Linus Torvalds 已提交
334
	case CPUFREQ_PRECHANGE:
335
		/* detect if the driver reported a value as "old frequency"
336 337
		 * which is not equal to what the cpufreq core thinks is
		 * "old frequency".
L
Linus Torvalds 已提交
338
		 */
339
		if (!(cpufreq_driver->flags & CPUFREQ_CONST_LOOPS)) {
340 341
			if ((policy) && (policy->cpu == freqs->cpu) &&
			    (policy->cur) && (policy->cur != freqs->old)) {
342 343
				pr_debug("Warning: CPU frequency is %u, cpufreq assumed %u kHz\n",
					 freqs->old, policy->cur);
344
				freqs->old = policy->cur;
L
Linus Torvalds 已提交
345 346
			}
		}
347
		srcu_notifier_call_chain(&cpufreq_transition_notifier_list,
348
				CPUFREQ_PRECHANGE, freqs);
L
Linus Torvalds 已提交
349 350
		adjust_jiffies(CPUFREQ_PRECHANGE, freqs);
		break;
351

L
Linus Torvalds 已提交
352 353
	case CPUFREQ_POSTCHANGE:
		adjust_jiffies(CPUFREQ_POSTCHANGE, freqs);
354 355
		pr_debug("FREQ: %lu - CPU: %lu\n",
			 (unsigned long)freqs->new, (unsigned long)freqs->cpu);
356
		trace_cpu_frequency(freqs->new, freqs->cpu);
357
		srcu_notifier_call_chain(&cpufreq_transition_notifier_list,
358
				CPUFREQ_POSTCHANGE, freqs);
359 360
		if (likely(policy) && likely(policy->cpu == freqs->cpu))
			policy->cur = freqs->new;
L
Linus Torvalds 已提交
361 362 363
		break;
	}
}
364

365 366 367 368 369 370 371 372
/**
 * 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.
 */
373
static void cpufreq_notify_transition(struct cpufreq_policy *policy,
374 375 376 377 378
		struct cpufreq_freqs *freqs, unsigned int state)
{
	for_each_cpu(freqs->cpu, policy->cpus)
		__cpufreq_notify_transition(policy, freqs, state);
}
L
Linus Torvalds 已提交
379

380
/* Do post notifications when there are chances that transition has failed */
381
static void cpufreq_notify_post_transition(struct cpufreq_policy *policy,
382 383 384 385 386 387 388 389 390 391 392
		struct cpufreq_freqs *freqs, int transition_failed)
{
	cpufreq_notify_transition(policy, freqs, CPUFREQ_POSTCHANGE);
	if (!transition_failed)
		return;

	swap(freqs->old, freqs->new);
	cpufreq_notify_transition(policy, freqs, CPUFREQ_PRECHANGE);
	cpufreq_notify_transition(policy, freqs, CPUFREQ_POSTCHANGE);
}

393 394 395
void cpufreq_freq_transition_begin(struct cpufreq_policy *policy,
		struct cpufreq_freqs *freqs)
{
396 397 398 399 400 401 402 403 404 405 406 407

	/*
	 * Catch double invocations of _begin() which lead to self-deadlock.
	 * ASYNC_NOTIFICATION drivers are left out because the cpufreq core
	 * doesn't invoke _begin() on their behalf, and hence the chances of
	 * double invocations are very low. Moreover, there are scenarios
	 * where these checks can emit false-positive warnings in these
	 * drivers; so we avoid that by skipping them altogether.
	 */
	WARN_ON(!(cpufreq_driver->flags & CPUFREQ_ASYNC_NOTIFICATION)
				&& current == policy->transition_task);

408 409 410 411 412 413 414 415 416 417 418
wait:
	wait_event(policy->transition_wait, !policy->transition_ongoing);

	spin_lock(&policy->transition_lock);

	if (unlikely(policy->transition_ongoing)) {
		spin_unlock(&policy->transition_lock);
		goto wait;
	}

	policy->transition_ongoing = true;
419
	policy->transition_task = current;
420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435

	spin_unlock(&policy->transition_lock);

	cpufreq_notify_transition(policy, freqs, CPUFREQ_PRECHANGE);
}
EXPORT_SYMBOL_GPL(cpufreq_freq_transition_begin);

void cpufreq_freq_transition_end(struct cpufreq_policy *policy,
		struct cpufreq_freqs *freqs, int transition_failed)
{
	if (unlikely(WARN_ON(!policy->transition_ongoing)))
		return;

	cpufreq_notify_post_transition(policy, freqs, transition_failed);

	policy->transition_ongoing = false;
436
	policy->transition_task = NULL;
437 438 439 440 441

	wake_up(&policy->transition_wait);
}
EXPORT_SYMBOL_GPL(cpufreq_freq_transition_end);

442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493
/*
 * Fast frequency switching status count.  Positive means "enabled", negative
 * means "disabled" and 0 means "not decided yet".
 */
static int cpufreq_fast_switch_count;
static DEFINE_MUTEX(cpufreq_fast_switch_lock);

static void cpufreq_list_transition_notifiers(void)
{
	struct notifier_block *nb;

	pr_info("Registered transition notifiers:\n");

	mutex_lock(&cpufreq_transition_notifier_list.mutex);

	for (nb = cpufreq_transition_notifier_list.head; nb; nb = nb->next)
		pr_info("%pF\n", nb->notifier_call);

	mutex_unlock(&cpufreq_transition_notifier_list.mutex);
}

/**
 * cpufreq_enable_fast_switch - Enable fast frequency switching for policy.
 * @policy: cpufreq policy to enable fast frequency switching for.
 *
 * Try to enable fast frequency switching for @policy.
 *
 * The attempt will fail if there is at least one transition notifier registered
 * at this point, as fast frequency switching is quite fundamentally at odds
 * with transition notifiers.  Thus if successful, it will make registration of
 * transition notifiers fail going forward.
 */
void cpufreq_enable_fast_switch(struct cpufreq_policy *policy)
{
	lockdep_assert_held(&policy->rwsem);

	if (!policy->fast_switch_possible)
		return;

	mutex_lock(&cpufreq_fast_switch_lock);
	if (cpufreq_fast_switch_count >= 0) {
		cpufreq_fast_switch_count++;
		policy->fast_switch_enabled = true;
	} else {
		pr_warn("CPU%u: Fast frequency switching not enabled\n",
			policy->cpu);
		cpufreq_list_transition_notifiers();
	}
	mutex_unlock(&cpufreq_fast_switch_lock);
}
EXPORT_SYMBOL_GPL(cpufreq_enable_fast_switch);

494 495 496 497 498
/**
 * cpufreq_disable_fast_switch - Disable fast frequency switching for policy.
 * @policy: cpufreq policy to disable fast frequency switching for.
 */
void cpufreq_disable_fast_switch(struct cpufreq_policy *policy)
499 500 501 502 503 504 505 506 507
{
	mutex_lock(&cpufreq_fast_switch_lock);
	if (policy->fast_switch_enabled) {
		policy->fast_switch_enabled = false;
		if (!WARN_ON(cpufreq_fast_switch_count <= 0))
			cpufreq_fast_switch_count--;
	}
	mutex_unlock(&cpufreq_fast_switch_lock);
}
508
EXPORT_SYMBOL_GPL(cpufreq_disable_fast_switch);
L
Linus Torvalds 已提交
509 510 511 512

/*********************************************************************
 *                          SYSFS INTERFACE                          *
 *********************************************************************/
513
static ssize_t show_boost(struct kobject *kobj,
514 515 516 517 518 519 520 521 522 523 524 525 526 527 528
				 struct attribute *attr, char *buf)
{
	return sprintf(buf, "%d\n", cpufreq_driver->boost_enabled);
}

static ssize_t store_boost(struct kobject *kobj, struct attribute *attr,
				  const char *buf, size_t count)
{
	int ret, enable;

	ret = sscanf(buf, "%d", &enable);
	if (ret != 1 || enable < 0 || enable > 1)
		return -EINVAL;

	if (cpufreq_boost_trigger_state(enable)) {
529 530
		pr_err("%s: Cannot %s BOOST!\n",
		       __func__, enable ? "enable" : "disable");
531 532 533
		return -EINVAL;
	}

534 535
	pr_debug("%s: cpufreq BOOST %s\n",
		 __func__, enable ? "enabled" : "disabled");
536 537 538 539

	return count;
}
define_one_global_rw(boost);
L
Linus Torvalds 已提交
540

541
static struct cpufreq_governor *find_governor(const char *str_governor)
542 543 544
{
	struct cpufreq_governor *t;

545
	for_each_governor(t)
546
		if (!strncasecmp(str_governor, t->name, CPUFREQ_NAME_LEN))
547 548 549 550 551
			return t;

	return NULL;
}

L
Linus Torvalds 已提交
552 553 554
/**
 * cpufreq_parse_governor - parse a governor string
 */
D
Dave Jones 已提交
555
static int cpufreq_parse_governor(char *str_governor, unsigned int *policy,
L
Linus Torvalds 已提交
556 557
				struct cpufreq_governor **governor)
{
558
	int err = -EINVAL;
559 560

	if (cpufreq_driver->setpolicy) {
561
		if (!strncasecmp(str_governor, "performance", CPUFREQ_NAME_LEN)) {
L
Linus Torvalds 已提交
562
			*policy = CPUFREQ_POLICY_PERFORMANCE;
563
			err = 0;
564
		} else if (!strncasecmp(str_governor, "powersave",
565
						CPUFREQ_NAME_LEN)) {
L
Linus Torvalds 已提交
566
			*policy = CPUFREQ_POLICY_POWERSAVE;
567
			err = 0;
L
Linus Torvalds 已提交
568
		}
569
	} else {
L
Linus Torvalds 已提交
570
		struct cpufreq_governor *t;
571

572
		mutex_lock(&cpufreq_governor_mutex);
573

574
		t = find_governor(str_governor);
575

576
		if (t == NULL) {
577
			int ret;
578

579 580 581
			mutex_unlock(&cpufreq_governor_mutex);
			ret = request_module("cpufreq_%s", str_governor);
			mutex_lock(&cpufreq_governor_mutex);
582

583
			if (ret == 0)
584
				t = find_governor(str_governor);
585 586
		}

587 588 589
		if (t != NULL) {
			*governor = t;
			err = 0;
L
Linus Torvalds 已提交
590
		}
591

592
		mutex_unlock(&cpufreq_governor_mutex);
L
Linus Torvalds 已提交
593
	}
594
	return err;
L
Linus Torvalds 已提交
595 596 597
}

/**
598 599
 * cpufreq_per_cpu_attr_read() / show_##file_name() -
 * print out cpufreq information
L
Linus Torvalds 已提交
600 601 602 603 604
 *
 * Write out information from cpufreq_driver->policy[cpu]; object must be
 * "unsigned int".
 */

605 606
#define show_one(file_name, object)			\
static ssize_t show_##file_name				\
D
Dave Jones 已提交
607
(struct cpufreq_policy *policy, char *buf)		\
608
{							\
609
	return sprintf(buf, "%u\n", policy->object);	\
L
Linus Torvalds 已提交
610 611 612 613
}

show_one(cpuinfo_min_freq, cpuinfo.min_freq);
show_one(cpuinfo_max_freq, cpuinfo.max_freq);
614
show_one(cpuinfo_transition_latency, cpuinfo.transition_latency);
L
Linus Torvalds 已提交
615 616
show_one(scaling_min_freq, min);
show_one(scaling_max_freq, max);
617

618
static ssize_t show_scaling_cur_freq(struct cpufreq_policy *policy, char *buf)
619 620 621 622 623 624 625 626 627
{
	ssize_t ret;

	if (cpufreq_driver && cpufreq_driver->setpolicy && cpufreq_driver->get)
		ret = sprintf(buf, "%u\n", cpufreq_driver->get(policy->cpu));
	else
		ret = sprintf(buf, "%u\n", policy->cur);
	return ret;
}
L
Linus Torvalds 已提交
628

629
static int cpufreq_set_policy(struct cpufreq_policy *policy,
630
				struct cpufreq_policy *new_policy);
631

L
Linus Torvalds 已提交
632 633 634 635 636
/**
 * cpufreq_per_cpu_attr_write() / store_##file_name() - sysfs write access
 */
#define store_one(file_name, object)			\
static ssize_t store_##file_name					\
D
Dave Jones 已提交
637
(struct cpufreq_policy *policy, const char *buf, size_t count)		\
L
Linus Torvalds 已提交
638
{									\
639
	int ret, temp;							\
L
Linus Torvalds 已提交
640 641
	struct cpufreq_policy new_policy;				\
									\
642
	memcpy(&new_policy, policy, sizeof(*policy));			\
L
Linus Torvalds 已提交
643
									\
644
	ret = sscanf(buf, "%u", &new_policy.object);			\
L
Linus Torvalds 已提交
645 646 647
	if (ret != 1)							\
		return -EINVAL;						\
									\
648
	temp = new_policy.object;					\
649
	ret = cpufreq_set_policy(policy, &new_policy);		\
650 651
	if (!ret)							\
		policy->user_policy.object = temp;			\
L
Linus Torvalds 已提交
652 653 654 655
									\
	return ret ? ret : count;					\
}

656 657
store_one(scaling_min_freq, min);
store_one(scaling_max_freq, max);
L
Linus Torvalds 已提交
658 659 660 661

/**
 * show_cpuinfo_cur_freq - current CPU frequency as detected by hardware
 */
D
Dave Jones 已提交
662 663
static ssize_t show_cpuinfo_cur_freq(struct cpufreq_policy *policy,
					char *buf)
L
Linus Torvalds 已提交
664
{
665
	unsigned int cur_freq = __cpufreq_get(policy);
L
Linus Torvalds 已提交
666 667 668 669 670 671 672 673
	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
 */
D
Dave Jones 已提交
674
static ssize_t show_scaling_governor(struct cpufreq_policy *policy, char *buf)
L
Linus Torvalds 已提交
675
{
676
	if (policy->policy == CPUFREQ_POLICY_POWERSAVE)
L
Linus Torvalds 已提交
677 678 679 680
		return sprintf(buf, "powersave\n");
	else if (policy->policy == CPUFREQ_POLICY_PERFORMANCE)
		return sprintf(buf, "performance\n");
	else if (policy->governor)
681
		return scnprintf(buf, CPUFREQ_NAME_PLEN, "%s\n",
682
				policy->governor->name);
L
Linus Torvalds 已提交
683 684 685 686 687 688
	return -EINVAL;
}

/**
 * store_scaling_governor - store policy for the specified CPU
 */
D
Dave Jones 已提交
689 690
static ssize_t store_scaling_governor(struct cpufreq_policy *policy,
					const char *buf, size_t count)
L
Linus Torvalds 已提交
691
{
692
	int ret;
L
Linus Torvalds 已提交
693 694 695
	char	str_governor[16];
	struct cpufreq_policy new_policy;

696
	memcpy(&new_policy, policy, sizeof(*policy));
L
Linus Torvalds 已提交
697

698
	ret = sscanf(buf, "%15s", str_governor);
L
Linus Torvalds 已提交
699 700 701
	if (ret != 1)
		return -EINVAL;

702 703
	if (cpufreq_parse_governor(str_governor, &new_policy.policy,
						&new_policy.governor))
L
Linus Torvalds 已提交
704 705
		return -EINVAL;

706
	ret = cpufreq_set_policy(policy, &new_policy);
707
	return ret ? ret : count;
L
Linus Torvalds 已提交
708 709 710 711 712
}

/**
 * show_scaling_driver - show the cpufreq driver currently loaded
 */
D
Dave Jones 已提交
713
static ssize_t show_scaling_driver(struct cpufreq_policy *policy, char *buf)
L
Linus Torvalds 已提交
714
{
715
	return scnprintf(buf, CPUFREQ_NAME_PLEN, "%s\n", cpufreq_driver->name);
L
Linus Torvalds 已提交
716 717 718 719 720
}

/**
 * show_scaling_available_governors - show the available CPUfreq governors
 */
D
Dave Jones 已提交
721 722
static ssize_t show_scaling_available_governors(struct cpufreq_policy *policy,
						char *buf)
L
Linus Torvalds 已提交
723 724 725 726
{
	ssize_t i = 0;
	struct cpufreq_governor *t;

727
	if (!has_target()) {
L
Linus Torvalds 已提交
728 729 730 731
		i += sprintf(buf, "performance powersave");
		goto out;
	}

732
	for_each_governor(t) {
733 734
		if (i >= (ssize_t) ((PAGE_SIZE / sizeof(char))
		    - (CPUFREQ_NAME_LEN + 2)))
L
Linus Torvalds 已提交
735
			goto out;
736
		i += scnprintf(&buf[i], CPUFREQ_NAME_PLEN, "%s ", t->name);
L
Linus Torvalds 已提交
737
	}
738
out:
L
Linus Torvalds 已提交
739 740 741
	i += sprintf(&buf[i], "\n");
	return i;
}
742

743
ssize_t cpufreq_show_cpus(const struct cpumask *mask, char *buf)
L
Linus Torvalds 已提交
744 745 746 747
{
	ssize_t i = 0;
	unsigned int cpu;

748
	for_each_cpu(cpu, mask) {
L
Linus Torvalds 已提交
749 750 751 752
		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))
753
			break;
L
Linus Torvalds 已提交
754 755 756 757
	}
	i += sprintf(&buf[i], "\n");
	return i;
}
758
EXPORT_SYMBOL_GPL(cpufreq_show_cpus);
L
Linus Torvalds 已提交
759

760 761 762 763 764 765
/**
 * 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)
{
766
	return cpufreq_show_cpus(policy->related_cpus, buf);
767 768 769 770 771 772 773
}

/**
 * show_affected_cpus - show the CPUs affected by each transition
 */
static ssize_t show_affected_cpus(struct cpufreq_policy *policy, char *buf)
{
774
	return cpufreq_show_cpus(policy->cpus, buf);
775 776
}

777
static ssize_t store_scaling_setspeed(struct cpufreq_policy *policy,
D
Dave Jones 已提交
778
					const char *buf, size_t count)
779 780 781 782
{
	unsigned int freq = 0;
	unsigned int ret;

783
	if (!policy->governor || !policy->governor->store_setspeed)
784 785 786 787 788 789 790 791 792 793 794 795 796
		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)
{
797
	if (!policy->governor || !policy->governor->show_setspeed)
798 799 800 801
		return sprintf(buf, "<unsupported>\n");

	return policy->governor->show_setspeed(policy, buf);
}
L
Linus Torvalds 已提交
802

803
/**
804
 * show_bios_limit - show the current cpufreq HW/BIOS limitation
805 806 807 808 809
 */
static ssize_t show_bios_limit(struct cpufreq_policy *policy, char *buf)
{
	unsigned int limit;
	int ret;
810 811
	if (cpufreq_driver->bios_limit) {
		ret = cpufreq_driver->bios_limit(policy->cpu, &limit);
812 813 814 815 816 817
		if (!ret)
			return sprintf(buf, "%u\n", limit);
	}
	return sprintf(buf, "%u\n", policy->cpuinfo.max_freq);
}

818 819 820 821 822 823 824 825 826 827 828 829 830 831
cpufreq_freq_attr_ro_perm(cpuinfo_cur_freq, 0400);
cpufreq_freq_attr_ro(cpuinfo_min_freq);
cpufreq_freq_attr_ro(cpuinfo_max_freq);
cpufreq_freq_attr_ro(cpuinfo_transition_latency);
cpufreq_freq_attr_ro(scaling_available_governors);
cpufreq_freq_attr_ro(scaling_driver);
cpufreq_freq_attr_ro(scaling_cur_freq);
cpufreq_freq_attr_ro(bios_limit);
cpufreq_freq_attr_ro(related_cpus);
cpufreq_freq_attr_ro(affected_cpus);
cpufreq_freq_attr_rw(scaling_min_freq);
cpufreq_freq_attr_rw(scaling_max_freq);
cpufreq_freq_attr_rw(scaling_governor);
cpufreq_freq_attr_rw(scaling_setspeed);
L
Linus Torvalds 已提交
832

D
Dave Jones 已提交
833
static struct attribute *default_attrs[] = {
L
Linus Torvalds 已提交
834 835
	&cpuinfo_min_freq.attr,
	&cpuinfo_max_freq.attr,
836
	&cpuinfo_transition_latency.attr,
L
Linus Torvalds 已提交
837 838 839
	&scaling_min_freq.attr,
	&scaling_max_freq.attr,
	&affected_cpus.attr,
840
	&related_cpus.attr,
L
Linus Torvalds 已提交
841 842 843
	&scaling_governor.attr,
	&scaling_driver.attr,
	&scaling_available_governors.attr,
844
	&scaling_setspeed.attr,
L
Linus Torvalds 已提交
845 846 847
	NULL
};

848 849
#define to_policy(k) container_of(k, struct cpufreq_policy, kobj)
#define to_attr(a) container_of(a, struct freq_attr, attr)
L
Linus Torvalds 已提交
850

851
static ssize_t show(struct kobject *kobj, struct attribute *attr, char *buf)
L
Linus Torvalds 已提交
852
{
D
Dave Jones 已提交
853 854
	struct cpufreq_policy *policy = to_policy(kobj);
	struct freq_attr *fattr = to_attr(attr);
855
	ssize_t ret;
856

857
	down_read(&policy->rwsem);
858
	ret = fattr->show(policy, buf);
859
	up_read(&policy->rwsem);
860

L
Linus Torvalds 已提交
861 862 863
	return ret;
}

D
Dave Jones 已提交
864 865
static ssize_t store(struct kobject *kobj, struct attribute *attr,
		     const char *buf, size_t count)
L
Linus Torvalds 已提交
866
{
D
Dave Jones 已提交
867 868
	struct cpufreq_policy *policy = to_policy(kobj);
	struct freq_attr *fattr = to_attr(attr);
869
	ssize_t ret = -EINVAL;
870

871 872
	get_online_cpus();

873 874
	if (cpu_online(policy->cpu)) {
		down_write(&policy->rwsem);
875
		ret = fattr->store(policy, buf, count);
876 877
		up_write(&policy->rwsem);
	}
878

879 880
	put_online_cpus();

L
Linus Torvalds 已提交
881 882 883
	return ret;
}

D
Dave Jones 已提交
884
static void cpufreq_sysfs_release(struct kobject *kobj)
L
Linus Torvalds 已提交
885
{
D
Dave Jones 已提交
886
	struct cpufreq_policy *policy = to_policy(kobj);
887
	pr_debug("last reference is dropped\n");
L
Linus Torvalds 已提交
888 889 890
	complete(&policy->kobj_unregister);
}

891
static const struct sysfs_ops sysfs_ops = {
L
Linus Torvalds 已提交
892 893 894 895 896 897 898 899 900 901
	.show	= show,
	.store	= store,
};

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

902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931
static int add_cpu_dev_symlink(struct cpufreq_policy *policy, int cpu)
{
	struct device *cpu_dev;

	pr_debug("%s: Adding symlink for CPU: %u\n", __func__, cpu);

	if (!policy)
		return 0;

	cpu_dev = get_cpu_device(cpu);
	if (WARN_ON(!cpu_dev))
		return 0;

	return sysfs_create_link(&cpu_dev->kobj, &policy->kobj, "cpufreq");
}

static void remove_cpu_dev_symlink(struct cpufreq_policy *policy, int cpu)
{
	struct device *cpu_dev;

	pr_debug("%s: Removing symlink for CPU: %u\n", __func__, cpu);

	cpu_dev = get_cpu_device(cpu);
	if (WARN_ON(!cpu_dev))
		return;

	sysfs_remove_link(&cpu_dev->kobj, "cpufreq");
}

/* Add/remove symlinks for all related CPUs */
932
static int cpufreq_add_dev_symlink(struct cpufreq_policy *policy)
933 934 935 936
{
	unsigned int j;
	int ret = 0;

937
	/* Some related CPUs might not be present (physically hotplugged) */
938
	for_each_cpu(j, policy->real_cpus) {
939
		ret = add_cpu_dev_symlink(policy, j);
940 941
		if (ret)
			break;
942
	}
943

944 945 946
	return ret;
}

947 948 949 950 951
static void cpufreq_remove_dev_symlink(struct cpufreq_policy *policy)
{
	unsigned int j;

	/* Some related CPUs might not be present (physically hotplugged) */
952
	for_each_cpu(j, policy->real_cpus)
953 954 955
		remove_cpu_dev_symlink(policy, j);
}

956
static int cpufreq_add_dev_interface(struct cpufreq_policy *policy)
957 958 959 960 961
{
	struct freq_attr **drv_attr;
	int ret = 0;

	/* set up files for this cpu device */
962
	drv_attr = cpufreq_driver->attr;
V
Viresh Kumar 已提交
963
	while (drv_attr && *drv_attr) {
964 965
		ret = sysfs_create_file(&policy->kobj, &((*drv_attr)->attr));
		if (ret)
966
			return ret;
967 968
		drv_attr++;
	}
969
	if (cpufreq_driver->get) {
970 971
		ret = sysfs_create_file(&policy->kobj, &cpuinfo_cur_freq.attr);
		if (ret)
972
			return ret;
973
	}
974 975 976

	ret = sysfs_create_file(&policy->kobj, &scaling_cur_freq.attr);
	if (ret)
977
		return ret;
978

979
	if (cpufreq_driver->bios_limit) {
980 981
		ret = sysfs_create_file(&policy->kobj, &bios_limit.attr);
		if (ret)
982
			return ret;
983
	}
984

985
	return cpufreq_add_dev_symlink(policy);
986 987
}

988 989 990 991 992
__weak struct cpufreq_governor *cpufreq_default_governor(void)
{
	return NULL;
}

993
static int cpufreq_init_policy(struct cpufreq_policy *policy)
994
{
995
	struct cpufreq_governor *gov = NULL;
996 997
	struct cpufreq_policy new_policy;

998
	memcpy(&new_policy, policy, sizeof(*policy));
999

1000
	/* Update governor of new_policy to the governor used before hotplug */
1001
	gov = find_governor(policy->last_governor);
1002
	if (gov) {
1003 1004
		pr_debug("Restoring governor %s for cpu %d\n",
				policy->governor->name, policy->cpu);
1005 1006 1007 1008 1009
	} else {
		gov = cpufreq_default_governor();
		if (!gov)
			return -ENODATA;
	}
1010 1011 1012

	new_policy.governor = gov;

1013 1014 1015 1016 1017 1018 1019 1020
	/* Use the default policy if there is no last_policy. */
	if (cpufreq_driver->setpolicy) {
		if (policy->last_policy)
			new_policy.policy = policy->last_policy;
		else
			cpufreq_parse_governor(gov->name, &new_policy.policy,
					       NULL);
	}
1021
	/* set default policy */
1022
	return cpufreq_set_policy(policy, &new_policy);
1023 1024
}

1025
static int cpufreq_add_policy_cpu(struct cpufreq_policy *policy, unsigned int cpu)
1026
{
1027
	int ret = 0;
1028

1029 1030 1031 1032
	/* Has this CPU been taken care of already? */
	if (cpumask_test_cpu(cpu, policy->cpus))
		return 0;

1033
	down_write(&policy->rwsem);
1034 1035
	if (has_target())
		cpufreq_stop_governor(policy);
1036 1037

	cpumask_set_cpu(cpu, policy->cpus);
V
Viresh Kumar 已提交
1038

1039
	if (has_target()) {
1040
		ret = cpufreq_start_governor(policy);
1041
		if (ret)
1042
			pr_err("%s: Failed to start governor\n", __func__);
1043
	}
1044 1045
	up_write(&policy->rwsem);
	return ret;
1046
}
L
Linus Torvalds 已提交
1047

1048 1049 1050 1051 1052 1053 1054
static void handle_update(struct work_struct *work)
{
	struct cpufreq_policy *policy =
		container_of(work, struct cpufreq_policy, update);
	unsigned int cpu = policy->cpu;
	pr_debug("handle_update for cpu %u called\n", cpu);
	cpufreq_update_policy(cpu);
1055
}
L
Linus Torvalds 已提交
1056

1057
static struct cpufreq_policy *cpufreq_policy_alloc(unsigned int cpu)
1058
{
1059
	struct device *dev = get_cpu_device(cpu);
1060
	struct cpufreq_policy *policy;
1061
	int ret;
1062

1063 1064 1065
	if (WARN_ON(!dev))
		return NULL;

1066 1067 1068 1069 1070 1071 1072 1073 1074 1075
	policy = kzalloc(sizeof(*policy), GFP_KERNEL);
	if (!policy)
		return NULL;

	if (!alloc_cpumask_var(&policy->cpus, GFP_KERNEL))
		goto err_free_policy;

	if (!zalloc_cpumask_var(&policy->related_cpus, GFP_KERNEL))
		goto err_free_cpumask;

1076 1077 1078
	if (!zalloc_cpumask_var(&policy->real_cpus, GFP_KERNEL))
		goto err_free_rcpumask;

1079 1080 1081 1082 1083 1084 1085
	ret = kobject_init_and_add(&policy->kobj, &ktype_cpufreq,
				   cpufreq_global_kobject, "policy%u", cpu);
	if (ret) {
		pr_err("%s: failed to init policy->kobj: %d\n", __func__, ret);
		goto err_free_real_cpus;
	}

1086
	INIT_LIST_HEAD(&policy->policy_list);
1087
	init_rwsem(&policy->rwsem);
1088 1089
	spin_lock_init(&policy->transition_lock);
	init_waitqueue_head(&policy->transition_wait);
1090 1091
	init_completion(&policy->kobj_unregister);
	INIT_WORK(&policy->update, handle_update);
1092

1093
	policy->cpu = cpu;
1094 1095
	return policy;

1096 1097
err_free_real_cpus:
	free_cpumask_var(policy->real_cpus);
1098 1099
err_free_rcpumask:
	free_cpumask_var(policy->related_cpus);
1100 1101 1102 1103 1104 1105 1106 1107
err_free_cpumask:
	free_cpumask_var(policy->cpus);
err_free_policy:
	kfree(policy);

	return NULL;
}

1108
static void cpufreq_policy_put_kobj(struct cpufreq_policy *policy, bool notify)
1109 1110 1111 1112
{
	struct kobject *kobj;
	struct completion *cmp;

1113 1114 1115
	if (notify)
		blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
					     CPUFREQ_REMOVE_POLICY, policy);
1116

1117 1118
	down_write(&policy->rwsem);
	cpufreq_remove_dev_symlink(policy);
1119 1120
	kobj = &policy->kobj;
	cmp = &policy->kobj_unregister;
1121
	up_write(&policy->rwsem);
1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133
	kobject_put(kobj);

	/*
	 * We need to make sure that the underlying kobj is
	 * actually not referenced anymore by anybody before we
	 * proceed with unloading.
	 */
	pr_debug("waiting for dropping of refcount\n");
	wait_for_completion(cmp);
	pr_debug("wait complete\n");
}

1134
static void cpufreq_policy_free(struct cpufreq_policy *policy, bool notify)
1135
{
1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146
	unsigned long flags;
	int cpu;

	/* Remove policy from list */
	write_lock_irqsave(&cpufreq_driver_lock, flags);
	list_del(&policy->policy_list);

	for_each_cpu(cpu, policy->related_cpus)
		per_cpu(cpufreq_cpu_data, cpu) = NULL;
	write_unlock_irqrestore(&cpufreq_driver_lock, flags);

1147
	cpufreq_policy_put_kobj(policy, notify);
1148
	free_cpumask_var(policy->real_cpus);
1149 1150 1151 1152 1153
	free_cpumask_var(policy->related_cpus);
	free_cpumask_var(policy->cpus);
	kfree(policy);
}

1154
static int cpufreq_online(unsigned int cpu)
L
Linus Torvalds 已提交
1155
{
1156
	struct cpufreq_policy *policy;
1157
	bool new_policy;
L
Linus Torvalds 已提交
1158
	unsigned long flags;
1159 1160
	unsigned int j;
	int ret;
1161

1162
	pr_debug("%s: bringing CPU%u online\n", __func__, cpu);
1163

1164
	/* Check if this CPU already has a policy to manage it */
1165
	policy = per_cpu(cpufreq_cpu_data, cpu);
1166
	if (policy) {
1167
		WARN_ON(!cpumask_test_cpu(cpu, policy->related_cpus));
1168
		if (!policy_is_inactive(policy))
1169
			return cpufreq_add_policy_cpu(policy, cpu);
L
Linus Torvalds 已提交
1170

1171
		/* This is the only online CPU for the policy.  Start over. */
1172
		new_policy = false;
1173 1174 1175 1176 1177
		down_write(&policy->rwsem);
		policy->cpu = cpu;
		policy->governor = NULL;
		up_write(&policy->rwsem);
	} else {
1178
		new_policy = true;
1179
		policy = cpufreq_policy_alloc(cpu);
1180
		if (!policy)
1181
			return -ENOMEM;
1182
	}
1183

1184
	cpumask_copy(policy->cpus, cpumask_of(cpu));
L
Linus Torvalds 已提交
1185 1186 1187 1188

	/* call driver. From then on the cpufreq must be able
	 * to accept all calls to ->verify and ->setpolicy for this CPU
	 */
1189
	ret = cpufreq_driver->init(policy);
L
Linus Torvalds 已提交
1190
	if (ret) {
1191
		pr_debug("initialization failed\n");
1192
		goto out_free_policy;
L
Linus Torvalds 已提交
1193
	}
V
Viresh Kumar 已提交
1194

1195 1196
	down_write(&policy->rwsem);

1197
	if (new_policy) {
1198
		/* related_cpus should at least include policy->cpus. */
1199
		cpumask_copy(policy->related_cpus, policy->cpus);
1200
		/* Remember CPUs present at the policy creation time. */
1201
		cpumask_and(policy->real_cpus, policy->cpus, cpu_present_mask);
1202
	}
1203

1204 1205 1206 1207 1208 1209
	/*
	 * affected cpus must always be the one, which are online. We aren't
	 * managing offline cpus here.
	 */
	cpumask_and(policy->cpus, policy->cpus, cpu_online_mask);

1210
	if (new_policy) {
1211 1212
		policy->user_policy.min = policy->min;
		policy->user_policy.max = policy->max;
1213

1214 1215 1216 1217 1218
		write_lock_irqsave(&cpufreq_driver_lock, flags);
		for_each_cpu(j, policy->related_cpus)
			per_cpu(cpufreq_cpu_data, j) = policy;
		write_unlock_irqrestore(&cpufreq_driver_lock, flags);
	}
1219

1220
	if (cpufreq_driver->get && !cpufreq_driver->setpolicy) {
1221 1222 1223
		policy->cur = cpufreq_driver->get(policy->cpu);
		if (!policy->cur) {
			pr_err("%s: ->get() failed\n", __func__);
1224
			goto out_exit_policy;
1225 1226 1227
		}
	}

1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267
	/*
	 * Sometimes boot loaders set CPU frequency to a value outside of
	 * frequency table present with cpufreq core. In such cases CPU might be
	 * unstable if it has to run on that frequency for long duration of time
	 * and so its better to set it to a frequency which is specified in
	 * freq-table. This also makes cpufreq stats inconsistent as
	 * cpufreq-stats would fail to register because current frequency of CPU
	 * isn't found in freq-table.
	 *
	 * Because we don't want this change to effect boot process badly, we go
	 * for the next freq which is >= policy->cur ('cur' must be set by now,
	 * otherwise we will end up setting freq to lowest of the table as 'cur'
	 * is initialized to zero).
	 *
	 * We are passing target-freq as "policy->cur - 1" otherwise
	 * __cpufreq_driver_target() would simply fail, as policy->cur will be
	 * equal to target-freq.
	 */
	if ((cpufreq_driver->flags & CPUFREQ_NEED_INITIAL_FREQ_CHECK)
	    && has_target()) {
		/* Are we running at unknown frequency ? */
		ret = cpufreq_frequency_table_get_index(policy, policy->cur);
		if (ret == -EINVAL) {
			/* Warn user and fix it */
			pr_warn("%s: CPU%d: Running at unlisted freq: %u KHz\n",
				__func__, policy->cpu, policy->cur);
			ret = __cpufreq_driver_target(policy, policy->cur - 1,
				CPUFREQ_RELATION_L);

			/*
			 * Reaching here after boot in a few seconds may not
			 * mean that system will remain stable at "unknown"
			 * frequency for longer duration. Hence, a BUG_ON().
			 */
			BUG_ON(ret);
			pr_warn("%s: CPU%d: Unlisted initial frequency changed to: %u KHz\n",
				__func__, policy->cpu, policy->cur);
		}
	}

1268 1269 1270
	blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
				     CPUFREQ_START, policy);

1271
	if (new_policy) {
1272
		ret = cpufreq_add_dev_interface(policy);
1273
		if (ret)
1274
			goto out_exit_policy;
1275 1276
		blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
				CPUFREQ_CREATE_POLICY, policy);
1277

1278 1279 1280 1281
		write_lock_irqsave(&cpufreq_driver_lock, flags);
		list_add(&policy->policy_list, &cpufreq_policy_list);
		write_unlock_irqrestore(&cpufreq_driver_lock, flags);
	}
1282

1283 1284 1285 1286
	ret = cpufreq_init_policy(policy);
	if (ret) {
		pr_err("%s: Failed to initialize policy for cpu: %d (%d)\n",
		       __func__, cpu, ret);
1287 1288 1289
		/* cpufreq_policy_free() will notify based on this */
		new_policy = false;
		goto out_exit_policy;
1290
	}
1291

1292
	up_write(&policy->rwsem);
1293

1294
	kobject_uevent(&policy->kobj, KOBJ_ADD);
1295 1296 1297 1298 1299

	/* Callback for handling stuff after policy is ready */
	if (cpufreq_driver->ready)
		cpufreq_driver->ready(policy);

1300
	pr_debug("initialization complete\n");
1301

L
Linus Torvalds 已提交
1302 1303
	return 0;

1304
out_exit_policy:
1305 1306
	up_write(&policy->rwsem);

1307 1308
	if (cpufreq_driver->exit)
		cpufreq_driver->exit(policy);
1309
out_free_policy:
1310
	cpufreq_policy_free(policy, !new_policy);
L
Linus Torvalds 已提交
1311 1312 1313
	return ret;
}

1314 1315 1316 1317 1318 1319 1320
/**
 * cpufreq_add_dev - the cpufreq interface for a CPU device.
 * @dev: CPU device.
 * @sif: Subsystem interface structure pointer (not used)
 */
static int cpufreq_add_dev(struct device *dev, struct subsys_interface *sif)
{
1321
	struct cpufreq_policy *policy;
1322 1323 1324 1325
	unsigned cpu = dev->id;

	dev_dbg(dev, "%s: adding CPU%u\n", __func__, cpu);

1326 1327
	if (cpu_online(cpu))
		return cpufreq_online(cpu);
1328

1329 1330 1331 1332 1333 1334 1335 1336
	/*
	 * A hotplug notifier will follow and we will handle it as CPU online
	 * then.  For now, just create the sysfs link, unless there is no policy
	 * or the link is already present.
	 */
	policy = per_cpu(cpufreq_cpu_data, cpu);
	if (!policy || cpumask_test_and_set_cpu(cpu, policy->real_cpus))
		return 0;
1337

1338
	return add_cpu_dev_symlink(policy, cpu);
L
Linus Torvalds 已提交
1339 1340
}

1341
static void cpufreq_offline(unsigned int cpu)
L
Linus Torvalds 已提交
1342
{
1343
	struct cpufreq_policy *policy;
1344
	int ret;
L
Linus Torvalds 已提交
1345

1346
	pr_debug("%s: unregistering CPU %u\n", __func__, cpu);
L
Linus Torvalds 已提交
1347

1348
	policy = cpufreq_cpu_get_raw(cpu);
1349
	if (!policy) {
1350
		pr_debug("%s: No cpu_data found\n", __func__);
1351
		return;
L
Linus Torvalds 已提交
1352 1353
	}

1354
	down_write(&policy->rwsem);
1355 1356
	if (has_target())
		cpufreq_stop_governor(policy);
L
Linus Torvalds 已提交
1357

1358
	cpumask_clear_cpu(cpu, policy->cpus);
1359

1360 1361 1362 1363
	if (policy_is_inactive(policy)) {
		if (has_target())
			strncpy(policy->last_governor, policy->governor->name,
				CPUFREQ_NAME_LEN);
1364 1365
		else
			policy->last_policy = policy->policy;
1366 1367 1368 1369
	} else if (cpu == policy->cpu) {
		/* Nominate new CPU */
		policy->cpu = cpumask_any(policy->cpus);
	}
1370

1371 1372 1373
	/* Start governor again for active policy */
	if (!policy_is_inactive(policy)) {
		if (has_target()) {
1374
			ret = cpufreq_start_governor(policy);
1375 1376 1377
			if (ret)
				pr_err("%s: Failed to start governor\n", __func__);
		}
1378

1379
		goto unlock;
1380 1381
	}

1382 1383
	if (cpufreq_driver->stop_cpu)
		cpufreq_driver->stop_cpu(policy);
1384

1385 1386
	if (has_target())
		cpufreq_exit_governor(policy);
L
Linus Torvalds 已提交
1387

1388 1389 1390 1391 1392
	/*
	 * Perform the ->exit() even during light-weight tear-down,
	 * since this is a core component, and is essential for the
	 * subsequent light-weight ->init() to succeed.
	 */
1393
	if (cpufreq_driver->exit) {
1394
		cpufreq_driver->exit(policy);
1395 1396
		policy->freq_table = NULL;
	}
1397 1398 1399

unlock:
	up_write(&policy->rwsem);
L
Linus Torvalds 已提交
1400 1401
}

1402
/**
1403
 * cpufreq_remove_dev - remove a CPU device
1404 1405 1406
 *
 * Removes the cpufreq interface for a CPU device.
 */
1407
static void cpufreq_remove_dev(struct device *dev, struct subsys_interface *sif)
1408
{
1409
	unsigned int cpu = dev->id;
1410
	struct cpufreq_policy *policy = per_cpu(cpufreq_cpu_data, cpu);
1411

1412
	if (!policy)
1413
		return;
1414

1415 1416
	if (cpu_online(cpu))
		cpufreq_offline(cpu);
1417

1418
	cpumask_clear_cpu(cpu, policy->real_cpus);
1419
	remove_cpu_dev_symlink(policy, cpu);
1420

1421
	if (cpumask_empty(policy->real_cpus))
1422
		cpufreq_policy_free(policy, true);
1423 1424
}

L
Linus Torvalds 已提交
1425
/**
1426 1427
 *	cpufreq_out_of_sync - If actual and saved CPU frequency differs, we're
 *	in deep trouble.
1428
 *	@policy: policy managing CPUs
L
Linus Torvalds 已提交
1429 1430
 *	@new_freq: CPU frequency the CPU actually runs at
 *
1431 1432
 *	We adjust to current frequency first, and need to clean up later.
 *	So either call to cpufreq_update_policy() or schedule handle_update()).
L
Linus Torvalds 已提交
1433
 */
1434
static void cpufreq_out_of_sync(struct cpufreq_policy *policy,
1435
				unsigned int new_freq)
L
Linus Torvalds 已提交
1436 1437
{
	struct cpufreq_freqs freqs;
1438

1439
	pr_debug("Warning: CPU frequency out of sync: cpufreq and timing core thinks of %u, is %u kHz\n",
1440
		 policy->cur, new_freq);
L
Linus Torvalds 已提交
1441

1442
	freqs.old = policy->cur;
L
Linus Torvalds 已提交
1443
	freqs.new = new_freq;
1444

1445 1446
	cpufreq_freq_transition_begin(policy, &freqs);
	cpufreq_freq_transition_end(policy, &freqs, 0);
L
Linus Torvalds 已提交
1447 1448
}

1449
/**
D
Dhaval Giani 已提交
1450
 * cpufreq_quick_get - get the CPU frequency (in kHz) from policy->cur
1451 1452 1453 1454 1455 1456 1457
 * @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)
{
1458
	struct cpufreq_policy *policy;
1459
	unsigned int ret_freq = 0;
1460
	unsigned long flags;
1461

1462 1463 1464 1465 1466 1467 1468 1469 1470
	read_lock_irqsave(&cpufreq_driver_lock, flags);

	if (cpufreq_driver && cpufreq_driver->setpolicy && cpufreq_driver->get) {
		ret_freq = cpufreq_driver->get(cpu);
		read_unlock_irqrestore(&cpufreq_driver_lock, flags);
		return ret_freq;
	}

	read_unlock_irqrestore(&cpufreq_driver_lock, flags);
1471 1472

	policy = cpufreq_cpu_get(cpu);
1473
	if (policy) {
1474
		ret_freq = policy->cur;
1475 1476 1477
		cpufreq_cpu_put(policy);
	}

D
Dave Jones 已提交
1478
	return ret_freq;
1479 1480 1481
}
EXPORT_SYMBOL(cpufreq_quick_get);

1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501
/**
 * cpufreq_quick_get_max - get the max reported CPU frequency for this CPU
 * @cpu: CPU number
 *
 * Just return the max possible frequency for a given CPU.
 */
unsigned int cpufreq_quick_get_max(unsigned int cpu)
{
	struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);
	unsigned int ret_freq = 0;

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

	return ret_freq;
}
EXPORT_SYMBOL(cpufreq_quick_get_max);

1502
static unsigned int __cpufreq_get(struct cpufreq_policy *policy)
L
Linus Torvalds 已提交
1503
{
1504
	unsigned int ret_freq = 0;
1505

1506
	if (!cpufreq_driver->get)
D
Dave Jones 已提交
1507
		return ret_freq;
L
Linus Torvalds 已提交
1508

1509
	ret_freq = cpufreq_driver->get(policy->cpu);
L
Linus Torvalds 已提交
1510

1511 1512 1513 1514 1515 1516
	/*
	 * Updating inactive policies is invalid, so avoid doing that.  Also
	 * if fast frequency switching is used with the given policy, the check
	 * against policy->cur is pointless, so skip it in that case too.
	 */
	if (unlikely(policy_is_inactive(policy)) || policy->fast_switch_enabled)
1517 1518
		return ret_freq;

1519
	if (ret_freq && policy->cur &&
1520
		!(cpufreq_driver->flags & CPUFREQ_CONST_LOOPS)) {
1521 1522 1523
		/* verify no discrepancy between actual and
					saved value exists */
		if (unlikely(ret_freq != policy->cur)) {
1524
			cpufreq_out_of_sync(policy, ret_freq);
L
Linus Torvalds 已提交
1525 1526 1527 1528
			schedule_work(&policy->update);
		}
	}

D
Dave Jones 已提交
1529
	return ret_freq;
1530
}
L
Linus Torvalds 已提交
1531

1532 1533 1534 1535 1536 1537 1538 1539
/**
 * 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)
{
1540
	struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);
1541 1542
	unsigned int ret_freq = 0;

1543 1544
	if (policy) {
		down_read(&policy->rwsem);
1545
		ret_freq = __cpufreq_get(policy);
1546
		up_read(&policy->rwsem);
1547

1548 1549
		cpufreq_cpu_put(policy);
	}
1550

D
Dave Jones 已提交
1551
	return ret_freq;
L
Linus Torvalds 已提交
1552 1553 1554
}
EXPORT_SYMBOL(cpufreq_get);

1555 1556 1557 1558
static unsigned int cpufreq_update_current_freq(struct cpufreq_policy *policy)
{
	unsigned int new_freq;

1559 1560 1561
	if (cpufreq_suspended)
		return 0;

1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575
	new_freq = cpufreq_driver->get(policy->cpu);
	if (!new_freq)
		return 0;

	if (!policy->cur) {
		pr_debug("cpufreq: Driver did not initialize current freq\n");
		policy->cur = new_freq;
	} else if (policy->cur != new_freq && has_target()) {
		cpufreq_out_of_sync(policy, new_freq);
	}

	return new_freq;
}

1576 1577 1578 1579 1580
static struct subsys_interface cpufreq_interface = {
	.name		= "cpufreq",
	.subsys		= &cpu_subsys,
	.add_dev	= cpufreq_add_dev,
	.remove_dev	= cpufreq_remove_dev,
1581 1582
};

1583 1584 1585 1586 1587 1588 1589 1590 1591
/*
 * In case platform wants some specific frequency to be configured
 * during suspend..
 */
int cpufreq_generic_suspend(struct cpufreq_policy *policy)
{
	int ret;

	if (!policy->suspend_freq) {
1592 1593
		pr_debug("%s: suspend_freq not defined\n", __func__);
		return 0;
1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608
	}

	pr_debug("%s: Setting suspend-freq: %u\n", __func__,
			policy->suspend_freq);

	ret = __cpufreq_driver_target(policy, policy->suspend_freq,
			CPUFREQ_RELATION_H);
	if (ret)
		pr_err("%s: unable to set suspend-freq: %u. err: %d\n",
				__func__, policy->suspend_freq, ret);

	return ret;
}
EXPORT_SYMBOL(cpufreq_generic_suspend);

1609
/**
1610
 * cpufreq_suspend() - Suspend CPUFreq governors
1611
 *
1612 1613 1614 1615
 * Called during system wide Suspend/Hibernate cycles for suspending governors
 * as some platforms can't change frequency after this point in suspend cycle.
 * Because some of the devices (like: i2c, regulators, etc) they use for
 * changing frequency are suspended quickly after this point.
1616
 */
1617
void cpufreq_suspend(void)
1618
{
1619
	struct cpufreq_policy *policy;
1620

1621 1622
	if (!cpufreq_driver)
		return;
1623

1624
	if (!has_target() && !cpufreq_driver->suspend)
1625
		goto suspend;
1626

1627 1628
	pr_debug("%s: Suspending Governors\n", __func__);

1629
	for_each_active_policy(policy) {
1630 1631
		if (has_target()) {
			down_write(&policy->rwsem);
1632
			cpufreq_stop_governor(policy);
1633 1634 1635 1636
			up_write(&policy->rwsem);
		}

		if (cpufreq_driver->suspend && cpufreq_driver->suspend(policy))
1637 1638
			pr_err("%s: Failed to suspend driver: %p\n", __func__,
				policy);
1639
	}
1640 1641 1642

suspend:
	cpufreq_suspended = true;
1643 1644
}

L
Linus Torvalds 已提交
1645
/**
1646
 * cpufreq_resume() - Resume CPUFreq governors
L
Linus Torvalds 已提交
1647
 *
1648 1649
 * Called during system wide Suspend/Hibernate cycle for resuming governors that
 * are suspended with cpufreq_suspend().
L
Linus Torvalds 已提交
1650
 */
1651
void cpufreq_resume(void)
L
Linus Torvalds 已提交
1652
{
1653
	struct cpufreq_policy *policy;
1654
	int ret;
L
Linus Torvalds 已提交
1655

1656 1657
	if (!cpufreq_driver)
		return;
L
Linus Torvalds 已提交
1658

1659 1660
	cpufreq_suspended = false;

1661
	if (!has_target() && !cpufreq_driver->resume)
1662
		return;
L
Linus Torvalds 已提交
1663

1664
	pr_debug("%s: Resuming Governors\n", __func__);
L
Linus Torvalds 已提交
1665

1666
	for_each_active_policy(policy) {
1667
		if (cpufreq_driver->resume && cpufreq_driver->resume(policy)) {
1668 1669
			pr_err("%s: Failed to resume driver: %p\n", __func__,
				policy);
1670
		} else if (has_target()) {
1671
			down_write(&policy->rwsem);
1672
			ret = cpufreq_start_governor(policy);
1673 1674 1675 1676 1677 1678
			up_write(&policy->rwsem);

			if (ret)
				pr_err("%s: Failed to start governor for policy: %p\n",
				       __func__, policy);
		}
1679 1680
	}
}
L
Linus Torvalds 已提交
1681

1682 1683 1684 1685 1686 1687 1688 1689
/**
 *	cpufreq_get_current_driver - return current driver's name
 *
 *	Return the name string of the currently loaded cpufreq driver
 *	or NULL, if none.
 */
const char *cpufreq_get_current_driver(void)
{
1690 1691 1692 1693
	if (cpufreq_driver)
		return cpufreq_driver->name;

	return NULL;
1694 1695
}
EXPORT_SYMBOL_GPL(cpufreq_get_current_driver);
L
Linus Torvalds 已提交
1696

1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711
/**
 *	cpufreq_get_driver_data - return current driver data
 *
 *	Return the private data of the currently loaded cpufreq
 *	driver, or NULL if no cpufreq driver is loaded.
 */
void *cpufreq_get_driver_data(void)
{
	if (cpufreq_driver)
		return cpufreq_driver->driver_data;

	return NULL;
}
EXPORT_SYMBOL_GPL(cpufreq_get_driver_data);

L
Linus Torvalds 已提交
1712 1713 1714 1715 1716 1717 1718 1719 1720
/*********************************************************************
 *                     NOTIFIER LISTS INTERFACE                      *
 *********************************************************************/

/**
 *	cpufreq_register_notifier - register a driver with cpufreq
 *	@nb: notifier function to register
 *      @list: CPUFREQ_TRANSITION_NOTIFIER or CPUFREQ_POLICY_NOTIFIER
 *
1721
 *	Add a driver to one of two lists: either a list of drivers that
L
Linus Torvalds 已提交
1722 1723 1724 1725 1726
 *      are notified about clock rate changes (once before and once after
 *      the transition), or a list of drivers that are notified about
 *      changes in cpufreq policy.
 *
 *	This function may sleep, and has the same return conditions as
1727
 *	blocking_notifier_chain_register.
L
Linus Torvalds 已提交
1728 1729 1730 1731 1732
 */
int cpufreq_register_notifier(struct notifier_block *nb, unsigned int list)
{
	int ret;

1733 1734 1735
	if (cpufreq_disabled())
		return -EINVAL;

1736 1737
	WARN_ON(!init_cpufreq_transition_notifier_list_called);

L
Linus Torvalds 已提交
1738 1739
	switch (list) {
	case CPUFREQ_TRANSITION_NOTIFIER:
1740 1741 1742 1743 1744 1745
		mutex_lock(&cpufreq_fast_switch_lock);

		if (cpufreq_fast_switch_count > 0) {
			mutex_unlock(&cpufreq_fast_switch_lock);
			return -EBUSY;
		}
1746
		ret = srcu_notifier_chain_register(
1747
				&cpufreq_transition_notifier_list, nb);
1748 1749 1750 1751
		if (!ret)
			cpufreq_fast_switch_count--;

		mutex_unlock(&cpufreq_fast_switch_lock);
L
Linus Torvalds 已提交
1752 1753
		break;
	case CPUFREQ_POLICY_NOTIFIER:
1754 1755
		ret = blocking_notifier_chain_register(
				&cpufreq_policy_notifier_list, nb);
L
Linus Torvalds 已提交
1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767
		break;
	default:
		ret = -EINVAL;
	}

	return ret;
}
EXPORT_SYMBOL(cpufreq_register_notifier);

/**
 *	cpufreq_unregister_notifier - unregister a driver with cpufreq
 *	@nb: notifier block to be unregistered
1768
 *	@list: CPUFREQ_TRANSITION_NOTIFIER or CPUFREQ_POLICY_NOTIFIER
L
Linus Torvalds 已提交
1769 1770 1771 1772
 *
 *	Remove a driver from the CPU frequency notifier list.
 *
 *	This function may sleep, and has the same return conditions as
1773
 *	blocking_notifier_chain_unregister.
L
Linus Torvalds 已提交
1774 1775 1776 1777 1778
 */
int cpufreq_unregister_notifier(struct notifier_block *nb, unsigned int list)
{
	int ret;

1779 1780 1781
	if (cpufreq_disabled())
		return -EINVAL;

L
Linus Torvalds 已提交
1782 1783
	switch (list) {
	case CPUFREQ_TRANSITION_NOTIFIER:
1784 1785
		mutex_lock(&cpufreq_fast_switch_lock);

1786
		ret = srcu_notifier_chain_unregister(
1787
				&cpufreq_transition_notifier_list, nb);
1788 1789 1790 1791
		if (!ret && !WARN_ON(cpufreq_fast_switch_count >= 0))
			cpufreq_fast_switch_count++;

		mutex_unlock(&cpufreq_fast_switch_lock);
L
Linus Torvalds 已提交
1792 1793
		break;
	case CPUFREQ_POLICY_NOTIFIER:
1794 1795
		ret = blocking_notifier_chain_unregister(
				&cpufreq_policy_notifier_list, nb);
L
Linus Torvalds 已提交
1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809
		break;
	default:
		ret = -EINVAL;
	}

	return ret;
}
EXPORT_SYMBOL(cpufreq_unregister_notifier);


/*********************************************************************
 *                              GOVERNORS                            *
 *********************************************************************/

1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834
/**
 * cpufreq_driver_fast_switch - Carry out a fast CPU frequency switch.
 * @policy: cpufreq policy to switch the frequency for.
 * @target_freq: New frequency to set (may be approximate).
 *
 * Carry out a fast frequency switch without sleeping.
 *
 * The driver's ->fast_switch() callback invoked by this function must be
 * suitable for being called from within RCU-sched read-side critical sections
 * and it is expected to select the minimum available frequency greater than or
 * equal to @target_freq (CPUFREQ_RELATION_L).
 *
 * This function must not be called if policy->fast_switch_enabled is unset.
 *
 * Governors calling this function must guarantee that it will never be invoked
 * twice in parallel for the same policy and that it will never be called in
 * parallel with either ->target() or ->target_index() for the same policy.
 *
 * If CPUFREQ_ENTRY_INVALID is returned by the driver's ->fast_switch()
 * callback to indicate an error condition, the hardware configuration must be
 * preserved.
 */
unsigned int cpufreq_driver_fast_switch(struct cpufreq_policy *policy,
					unsigned int target_freq)
{
1835
	target_freq = clamp_val(target_freq, policy->min, policy->max);
1836 1837 1838 1839 1840

	return cpufreq_driver->fast_switch(policy, target_freq);
}
EXPORT_SYMBOL_GPL(cpufreq_driver_fast_switch);

1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866
/* Must set freqs->new to intermediate frequency */
static int __target_intermediate(struct cpufreq_policy *policy,
				 struct cpufreq_freqs *freqs, int index)
{
	int ret;

	freqs->new = cpufreq_driver->get_intermediate(policy, index);

	/* We don't need to switch to intermediate freq */
	if (!freqs->new)
		return 0;

	pr_debug("%s: cpu: %d, switching to intermediate freq: oldfreq: %u, intermediate freq: %u\n",
		 __func__, policy->cpu, freqs->old, freqs->new);

	cpufreq_freq_transition_begin(policy, freqs);
	ret = cpufreq_driver->target_intermediate(policy, index);
	cpufreq_freq_transition_end(policy, freqs, ret);

	if (ret)
		pr_err("%s: Failed to change to intermediate frequency: %d\n",
		       __func__, ret);

	return ret;
}

1867 1868 1869
static int __target_index(struct cpufreq_policy *policy,
			  struct cpufreq_frequency_table *freq_table, int index)
{
1870 1871
	struct cpufreq_freqs freqs = {.old = policy->cur, .flags = 0};
	unsigned int intermediate_freq = 0;
1872 1873 1874 1875 1876
	int retval = -EINVAL;
	bool notify;

	notify = !(cpufreq_driver->flags & CPUFREQ_ASYNC_NOTIFICATION);
	if (notify) {
1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887
		/* Handle switching to intermediate frequency */
		if (cpufreq_driver->get_intermediate) {
			retval = __target_intermediate(policy, &freqs, index);
			if (retval)
				return retval;

			intermediate_freq = freqs.new;
			/* Set old freq to intermediate */
			if (intermediate_freq)
				freqs.old = freqs.new;
		}
1888

1889
		freqs.new = freq_table[index].frequency;
1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900
		pr_debug("%s: cpu: %d, oldfreq: %u, new freq: %u\n",
			 __func__, policy->cpu, freqs.old, freqs.new);

		cpufreq_freq_transition_begin(policy, &freqs);
	}

	retval = cpufreq_driver->target_index(policy, index);
	if (retval)
		pr_err("%s: Failed to change cpu frequency: %d\n", __func__,
		       retval);

1901
	if (notify) {
1902 1903
		cpufreq_freq_transition_end(policy, &freqs, retval);

1904 1905 1906 1907
		/*
		 * Failed after setting to intermediate freq? Driver should have
		 * reverted back to initial frequency and so should we. Check
		 * here for intermediate_freq instead of get_intermediate, in
1908
		 * case we haven't switched to intermediate freq at all.
1909 1910 1911 1912 1913 1914 1915 1916 1917
		 */
		if (unlikely(retval && intermediate_freq)) {
			freqs.old = intermediate_freq;
			freqs.new = policy->restore_freq;
			cpufreq_freq_transition_begin(policy, &freqs);
			cpufreq_freq_transition_end(policy, &freqs, 0);
		}
	}

1918 1919 1920
	return retval;
}

L
Linus Torvalds 已提交
1921 1922 1923 1924
int __cpufreq_driver_target(struct cpufreq_policy *policy,
			    unsigned int target_freq,
			    unsigned int relation)
{
1925
	unsigned int old_target_freq = target_freq;
1926 1927
	struct cpufreq_frequency_table *freq_table;
	int index, retval;
1928

1929 1930 1931
	if (cpufreq_disabled())
		return -ENODEV;

1932 1933 1934 1935 1936 1937 1938
	/* Make sure that target_freq is within supported range */
	if (target_freq > policy->max)
		target_freq = policy->max;
	if (target_freq < policy->min)
		target_freq = policy->min;

	pr_debug("target for CPU %u: %u kHz, relation %u, requested %u kHz\n",
1939
		 policy->cpu, target_freq, relation, old_target_freq);
1940

1941 1942 1943 1944 1945 1946
	/*
	 * This might look like a redundant call as we are checking it again
	 * after finding index. But it is left intentionally for cases where
	 * exactly same freq is called again and so we can save on few function
	 * calls.
	 */
1947 1948 1949
	if (target_freq == policy->cur)
		return 0;

1950 1951 1952
	/* Save last value to restore later on errors */
	policy->restore_freq = policy->cur;

1953
	if (cpufreq_driver->target)
1954
		return cpufreq_driver->target(policy, target_freq, relation);
1955

1956 1957
	if (!cpufreq_driver->target_index)
		return -EINVAL;
1958

1959 1960 1961 1962 1963
	freq_table = cpufreq_frequency_get_table(policy->cpu);
	if (unlikely(!freq_table)) {
		pr_err("%s: Unable to find freq_table\n", __func__);
		return -EINVAL;
	}
1964

1965 1966 1967 1968 1969
	retval = cpufreq_frequency_table_target(policy, freq_table, target_freq,
						relation, &index);
	if (unlikely(retval)) {
		pr_err("%s: Unable to find matching freq\n", __func__);
		return retval;
1970 1971
	}

1972 1973 1974 1975
	if (freq_table[index].frequency == policy->cur)
		return 0;

	return __target_index(policy, freq_table, index);
L
Linus Torvalds 已提交
1976 1977 1978 1979 1980 1981 1982
}
EXPORT_SYMBOL_GPL(__cpufreq_driver_target);

int cpufreq_driver_target(struct cpufreq_policy *policy,
			  unsigned int target_freq,
			  unsigned int relation)
{
1983
	int ret = -EINVAL;
L
Linus Torvalds 已提交
1984

1985
	down_write(&policy->rwsem);
L
Linus Torvalds 已提交
1986 1987 1988

	ret = __cpufreq_driver_target(policy, target_freq, relation);

1989
	up_write(&policy->rwsem);
L
Linus Torvalds 已提交
1990 1991 1992 1993 1994

	return ret;
}
EXPORT_SYMBOL_GPL(cpufreq_driver_target);

1995 1996 1997 1998 1999
__weak struct cpufreq_governor *cpufreq_fallback_governor(void)
{
	return NULL;
}

2000
static int cpufreq_governor(struct cpufreq_policy *policy, unsigned int event)
L
Linus Torvalds 已提交
2001
{
2002
	int ret;
2003

2004 2005 2006
	/* Don't start any governor operations if we are entering suspend */
	if (cpufreq_suspended)
		return 0;
2007 2008 2009 2010 2011 2012
	/*
	 * Governor might not be initiated here if ACPI _PPC changed
	 * notification happened, so check it.
	 */
	if (!policy->governor)
		return -EINVAL;
2013

2014 2015 2016
	if (policy->governor->max_transition_latency &&
	    policy->cpuinfo.transition_latency >
	    policy->governor->max_transition_latency) {
2017 2018 2019
		struct cpufreq_governor *gov = cpufreq_fallback_governor();

		if (gov) {
2020 2021
			pr_warn("%s governor failed, too long transition latency of HW, fallback to %s governor\n",
				policy->governor->name, gov->name);
2022
			policy->governor = gov;
2023 2024
		} else {
			return -EINVAL;
2025
		}
2026
	}
L
Linus Torvalds 已提交
2027

2028 2029 2030
	if (event == CPUFREQ_GOV_POLICY_INIT)
		if (!try_module_get(policy->governor->owner))
			return -EINVAL;
L
Linus Torvalds 已提交
2031

2032
	pr_debug("%s: for CPU %u, event %u\n", __func__, policy->cpu, event);
2033

L
Linus Torvalds 已提交
2034 2035
	ret = policy->governor->governor(policy, event);

2036 2037 2038 2039
	if (event == CPUFREQ_GOV_POLICY_INIT) {
		if (ret)
			module_put(policy->governor->owner);
		else
2040
			policy->governor->initialized++;
2041 2042
	} else if (event == CPUFREQ_GOV_POLICY_EXIT) {
		policy->governor->initialized--;
L
Linus Torvalds 已提交
2043
		module_put(policy->governor->owner);
2044
	}
L
Linus Torvalds 已提交
2045 2046 2047 2048

	return ret;
}

2049 2050 2051 2052
static int cpufreq_start_governor(struct cpufreq_policy *policy)
{
	int ret;

2053 2054 2055
	if (cpufreq_driver->get && !cpufreq_driver->setpolicy)
		cpufreq_update_current_freq(policy);

2056 2057 2058 2059
	ret = cpufreq_governor(policy, CPUFREQ_GOV_START);
	return ret ? ret : cpufreq_governor(policy, CPUFREQ_GOV_LIMITS);
}

L
Linus Torvalds 已提交
2060 2061
int cpufreq_register_governor(struct cpufreq_governor *governor)
{
2062
	int err;
L
Linus Torvalds 已提交
2063 2064 2065 2066

	if (!governor)
		return -EINVAL;

2067 2068 2069
	if (cpufreq_disabled())
		return -ENODEV;

2070
	mutex_lock(&cpufreq_governor_mutex);
2071

2072
	governor->initialized = 0;
2073
	err = -EBUSY;
2074
	if (!find_governor(governor->name)) {
2075 2076
		err = 0;
		list_add(&governor->governor_list, &cpufreq_governor_list);
L
Linus Torvalds 已提交
2077 2078
	}

2079
	mutex_unlock(&cpufreq_governor_mutex);
2080
	return err;
L
Linus Torvalds 已提交
2081 2082 2083 2084 2085
}
EXPORT_SYMBOL_GPL(cpufreq_register_governor);

void cpufreq_unregister_governor(struct cpufreq_governor *governor)
{
2086 2087
	struct cpufreq_policy *policy;
	unsigned long flags;
2088

L
Linus Torvalds 已提交
2089 2090 2091
	if (!governor)
		return;

2092 2093 2094
	if (cpufreq_disabled())
		return;

2095 2096 2097
	/* clear last_governor for all inactive policies */
	read_lock_irqsave(&cpufreq_driver_lock, flags);
	for_each_inactive_policy(policy) {
2098 2099
		if (!strcmp(policy->last_governor, governor->name)) {
			policy->governor = NULL;
2100
			strcpy(policy->last_governor, "\0");
2101
		}
2102
	}
2103
	read_unlock_irqrestore(&cpufreq_driver_lock, flags);
2104

2105
	mutex_lock(&cpufreq_governor_mutex);
L
Linus Torvalds 已提交
2106
	list_del(&governor->governor_list);
2107
	mutex_unlock(&cpufreq_governor_mutex);
L
Linus Torvalds 已提交
2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118
	return;
}
EXPORT_SYMBOL_GPL(cpufreq_unregister_governor);


/*********************************************************************
 *                          POLICY INTERFACE                         *
 *********************************************************************/

/**
 * cpufreq_get_policy - get the current cpufreq_policy
2119 2120
 * @policy: struct cpufreq_policy into which the current cpufreq_policy
 *	is written
L
Linus Torvalds 已提交
2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133
 *
 * Reads the current cpufreq policy.
 */
int cpufreq_get_policy(struct cpufreq_policy *policy, unsigned int cpu)
{
	struct cpufreq_policy *cpu_policy;
	if (!policy)
		return -EINVAL;

	cpu_policy = cpufreq_cpu_get(cpu);
	if (!cpu_policy)
		return -EINVAL;

2134
	memcpy(policy, cpu_policy, sizeof(*policy));
L
Linus Torvalds 已提交
2135 2136 2137 2138 2139 2140

	cpufreq_cpu_put(cpu_policy);
	return 0;
}
EXPORT_SYMBOL(cpufreq_get_policy);

2141
/*
2142 2143
 * policy : current policy.
 * new_policy: policy to be set.
2144
 */
2145
static int cpufreq_set_policy(struct cpufreq_policy *policy,
2146
				struct cpufreq_policy *new_policy)
L
Linus Torvalds 已提交
2147
{
2148 2149
	struct cpufreq_governor *old_gov;
	int ret;
L
Linus Torvalds 已提交
2150

2151 2152
	pr_debug("setting new policy for CPU %u: %u - %u kHz\n",
		 new_policy->cpu, new_policy->min, new_policy->max);
L
Linus Torvalds 已提交
2153

2154
	memcpy(&new_policy->cpuinfo, &policy->cpuinfo, sizeof(policy->cpuinfo));
L
Linus Torvalds 已提交
2155

2156 2157 2158 2159 2160
	/*
	* This check works well when we store new min/max freq attributes,
	* because new_policy is a copy of policy with one field updated.
	*/
	if (new_policy->min > new_policy->max)
2161
		return -EINVAL;
2162

L
Linus Torvalds 已提交
2163
	/* verify the cpu speed can be set within this limit */
2164
	ret = cpufreq_driver->verify(new_policy);
L
Linus Torvalds 已提交
2165
	if (ret)
2166
		return ret;
L
Linus Torvalds 已提交
2167 2168

	/* adjust if necessary - all reasons */
2169
	blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
2170
			CPUFREQ_ADJUST, new_policy);
L
Linus Torvalds 已提交
2171

2172 2173 2174 2175
	/*
	 * verify the cpu speed can be set within this limit, which might be
	 * different to the first one
	 */
2176
	ret = cpufreq_driver->verify(new_policy);
2177
	if (ret)
2178
		return ret;
L
Linus Torvalds 已提交
2179 2180

	/* notification of the new policy */
2181
	blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
2182
			CPUFREQ_NOTIFY, new_policy);
L
Linus Torvalds 已提交
2183

2184 2185
	policy->min = new_policy->min;
	policy->max = new_policy->max;
L
Linus Torvalds 已提交
2186

2187
	pr_debug("new min and max freqs are %u - %u kHz\n",
2188
		 policy->min, policy->max);
L
Linus Torvalds 已提交
2189

2190
	if (cpufreq_driver->setpolicy) {
2191
		policy->policy = new_policy->policy;
2192
		pr_debug("setting range\n");
2193 2194
		return cpufreq_driver->setpolicy(new_policy);
	}
L
Linus Torvalds 已提交
2195

2196 2197 2198 2199
	if (new_policy->governor == policy->governor) {
		pr_debug("cpufreq: governor limits update\n");
		return cpufreq_governor(policy, CPUFREQ_GOV_LIMITS);
	}
2200

2201 2202 2203 2204 2205 2206
	pr_debug("governor switch\n");

	/* save old, working values */
	old_gov = policy->governor;
	/* end old governor */
	if (old_gov) {
2207
		cpufreq_stop_governor(policy);
2208
		cpufreq_exit_governor(policy);
L
Linus Torvalds 已提交
2209 2210
	}

2211 2212
	/* start new governor */
	policy->governor = new_policy->governor;
2213
	ret = cpufreq_governor(policy, CPUFREQ_GOV_POLICY_INIT);
2214
	if (!ret) {
2215 2216 2217 2218 2219
		ret = cpufreq_start_governor(policy);
		if (!ret) {
			pr_debug("cpufreq: governor change\n");
			return 0;
		}
2220
		cpufreq_exit_governor(policy);
2221 2222 2223 2224 2225 2226
	}

	/* new governor failed, so re-start old one */
	pr_debug("starting governor %s failed\n", policy->governor->name);
	if (old_gov) {
		policy->governor = old_gov;
2227
		if (cpufreq_governor(policy, CPUFREQ_GOV_POLICY_INIT))
2228 2229
			policy->governor = NULL;
		else
2230
			cpufreq_start_governor(policy);
2231 2232
	}

2233
	return ret;
L
Linus Torvalds 已提交
2234 2235 2236 2237 2238 2239
}

/**
 *	cpufreq_update_policy - re-evaluate an existing cpufreq policy
 *	@cpu: CPU which shall be re-evaluated
 *
L
Lucas De Marchi 已提交
2240
 *	Useful for policy notifiers which have different necessities
L
Linus Torvalds 已提交
2241 2242 2243 2244
 *	at different times.
 */
int cpufreq_update_policy(unsigned int cpu)
{
2245 2246
	struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);
	struct cpufreq_policy new_policy;
2247
	int ret;
L
Linus Torvalds 已提交
2248

2249 2250
	if (!policy)
		return -ENODEV;
L
Linus Torvalds 已提交
2251

2252
	down_write(&policy->rwsem);
L
Linus Torvalds 已提交
2253

2254
	pr_debug("updating policy for CPU %u\n", cpu);
2255
	memcpy(&new_policy, policy, sizeof(*policy));
2256 2257
	new_policy.min = policy->user_policy.min;
	new_policy.max = policy->user_policy.max;
L
Linus Torvalds 已提交
2258

2259 2260 2261 2262
	/*
	 * BIOS might change freq behind our back
	 * -> ask driver for current freq and notify governors about a change
	 */
2263
	if (cpufreq_driver->get && !cpufreq_driver->setpolicy) {
2264
		new_policy.cur = cpufreq_update_current_freq(policy);
2265 2266
		if (WARN_ON(!new_policy.cur)) {
			ret = -EIO;
2267
			goto unlock;
2268
		}
2269 2270
	}

2271
	ret = cpufreq_set_policy(policy, &new_policy);
L
Linus Torvalds 已提交
2272

2273
unlock:
2274
	up_write(&policy->rwsem);
2275

2276
	cpufreq_cpu_put(policy);
L
Linus Torvalds 已提交
2277 2278 2279 2280
	return ret;
}
EXPORT_SYMBOL(cpufreq_update_policy);

2281
static int cpufreq_cpu_callback(struct notifier_block *nfb,
2282 2283 2284 2285
					unsigned long action, void *hcpu)
{
	unsigned int cpu = (unsigned long)hcpu;

2286 2287
	switch (action & ~CPU_TASKS_FROZEN) {
	case CPU_ONLINE:
2288
	case CPU_DOWN_FAILED:
2289 2290
		cpufreq_online(cpu);
		break;
2291

2292
	case CPU_DOWN_PREPARE:
2293
		cpufreq_offline(cpu);
2294
		break;
2295 2296 2297 2298
	}
	return NOTIFY_OK;
}

2299
static struct notifier_block __refdata cpufreq_cpu_notifier = {
2300
	.notifier_call = cpufreq_cpu_callback,
2301
};
L
Linus Torvalds 已提交
2302

2303 2304 2305 2306 2307 2308 2309 2310 2311
/*********************************************************************
 *               BOOST						     *
 *********************************************************************/
static int cpufreq_boost_set_sw(int state)
{
	struct cpufreq_frequency_table *freq_table;
	struct cpufreq_policy *policy;
	int ret = -EINVAL;

2312
	for_each_active_policy(policy) {
2313 2314 2315 2316 2317 2318 2319 2320 2321
		freq_table = cpufreq_frequency_get_table(policy->cpu);
		if (freq_table) {
			ret = cpufreq_frequency_table_cpuinfo(policy,
							freq_table);
			if (ret) {
				pr_err("%s: Policy frequency update failed\n",
				       __func__);
				break;
			}
2322 2323

			down_write(&policy->rwsem);
2324
			policy->user_policy.max = policy->max;
2325
			cpufreq_governor(policy, CPUFREQ_GOV_LIMITS);
2326
			up_write(&policy->rwsem);
2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350
		}
	}

	return ret;
}

int cpufreq_boost_trigger_state(int state)
{
	unsigned long flags;
	int ret = 0;

	if (cpufreq_driver->boost_enabled == state)
		return 0;

	write_lock_irqsave(&cpufreq_driver_lock, flags);
	cpufreq_driver->boost_enabled = state;
	write_unlock_irqrestore(&cpufreq_driver_lock, flags);

	ret = cpufreq_driver->set_boost(state);
	if (ret) {
		write_lock_irqsave(&cpufreq_driver_lock, flags);
		cpufreq_driver->boost_enabled = !state;
		write_unlock_irqrestore(&cpufreq_driver_lock, flags);

2351 2352
		pr_err("%s: Cannot %s BOOST\n",
		       __func__, state ? "enable" : "disable");
2353 2354 2355 2356 2357
	}

	return ret;
}

2358
static bool cpufreq_boost_supported(void)
2359
{
2360
	return likely(cpufreq_driver) && cpufreq_driver->set_boost;
2361 2362
}

2363 2364 2365 2366
static int create_boost_sysfs_file(void)
{
	int ret;

2367
	ret = sysfs_create_file(cpufreq_global_kobject, &boost.attr);
2368 2369 2370 2371 2372 2373 2374 2375 2376 2377
	if (ret)
		pr_err("%s: cannot register global BOOST sysfs file\n",
		       __func__);

	return ret;
}

static void remove_boost_sysfs_file(void)
{
	if (cpufreq_boost_supported())
2378
		sysfs_remove_file(cpufreq_global_kobject, &boost.attr);
2379 2380 2381 2382 2383 2384 2385 2386 2387 2388
}

int cpufreq_enable_boost_support(void)
{
	if (!cpufreq_driver)
		return -EINVAL;

	if (cpufreq_boost_supported())
		return 0;

2389
	cpufreq_driver->set_boost = cpufreq_boost_set_sw;
2390 2391 2392 2393 2394 2395

	/* This will get removed on driver unregister */
	return create_boost_sysfs_file();
}
EXPORT_SYMBOL_GPL(cpufreq_enable_boost_support);

2396 2397 2398 2399 2400 2401
int cpufreq_boost_enabled(void)
{
	return cpufreq_driver->boost_enabled;
}
EXPORT_SYMBOL_GPL(cpufreq_boost_enabled);

L
Linus Torvalds 已提交
2402 2403 2404 2405 2406 2407 2408 2409 2410
/*********************************************************************
 *               REGISTER / UNREGISTER CPUFREQ DRIVER                *
 *********************************************************************/

/**
 * cpufreq_register_driver - register a CPU Frequency driver
 * @driver_data: A struct cpufreq_driver containing the values#
 * submitted by the CPU Frequency driver.
 *
2411
 * Registers a CPU Frequency driver to this core code. This code
2412
 * returns zero on success, -EEXIST when another driver got here first
2413
 * (and isn't unregistered in the meantime).
L
Linus Torvalds 已提交
2414 2415
 *
 */
2416
int cpufreq_register_driver(struct cpufreq_driver *driver_data)
L
Linus Torvalds 已提交
2417 2418 2419 2420
{
	unsigned long flags;
	int ret;

2421 2422 2423
	if (cpufreq_disabled())
		return -ENODEV;

L
Linus Torvalds 已提交
2424
	if (!driver_data || !driver_data->verify || !driver_data->init ||
2425
	    !(driver_data->setpolicy || driver_data->target_index ||
2426 2427
		    driver_data->target) ||
	     (driver_data->setpolicy && (driver_data->target_index ||
2428 2429
		    driver_data->target)) ||
	     (!!driver_data->get_intermediate != !!driver_data->target_intermediate))
L
Linus Torvalds 已提交
2430 2431
		return -EINVAL;

2432
	pr_debug("trying to register driver %s\n", driver_data->name);
L
Linus Torvalds 已提交
2433

2434 2435 2436
	/* Protect against concurrent CPU online/offline. */
	get_online_cpus();

2437
	write_lock_irqsave(&cpufreq_driver_lock, flags);
2438
	if (cpufreq_driver) {
2439
		write_unlock_irqrestore(&cpufreq_driver_lock, flags);
2440 2441
		ret = -EEXIST;
		goto out;
L
Linus Torvalds 已提交
2442
	}
2443
	cpufreq_driver = driver_data;
2444
	write_unlock_irqrestore(&cpufreq_driver_lock, flags);
L
Linus Torvalds 已提交
2445

2446 2447 2448
	if (driver_data->setpolicy)
		driver_data->flags |= CPUFREQ_CONST_LOOPS;

2449 2450 2451 2452 2453
	if (cpufreq_boost_supported()) {
		ret = create_boost_sysfs_file();
		if (ret)
			goto err_null_driver;
	}
2454

2455
	ret = subsys_interface_register(&cpufreq_interface);
2456
	if (ret)
2457
		goto err_boost_unreg;
L
Linus Torvalds 已提交
2458

2459 2460
	if (!(cpufreq_driver->flags & CPUFREQ_STICKY) &&
	    list_empty(&cpufreq_policy_list)) {
L
Linus Torvalds 已提交
2461
		/* if all ->init() calls failed, unregister */
2462 2463 2464
		pr_debug("%s: No CPU initialized for driver %s\n", __func__,
			 driver_data->name);
		goto err_if_unreg;
L
Linus Torvalds 已提交
2465 2466
	}

2467
	register_hotcpu_notifier(&cpufreq_cpu_notifier);
2468
	pr_debug("driver %s up and running\n", driver_data->name);
2469
	goto out;
2470

2471 2472
err_if_unreg:
	subsys_interface_unregister(&cpufreq_interface);
2473
err_boost_unreg:
2474
	remove_boost_sysfs_file();
2475
err_null_driver:
2476
	write_lock_irqsave(&cpufreq_driver_lock, flags);
2477
	cpufreq_driver = NULL;
2478
	write_unlock_irqrestore(&cpufreq_driver_lock, flags);
2479 2480 2481
out:
	put_online_cpus();
	return ret;
L
Linus Torvalds 已提交
2482 2483 2484 2485 2486 2487
}
EXPORT_SYMBOL_GPL(cpufreq_register_driver);

/**
 * cpufreq_unregister_driver - unregister the current CPUFreq driver
 *
2488
 * Unregister the current CPUFreq driver. Only call this if you have
L
Linus Torvalds 已提交
2489 2490 2491 2492
 * the right to do so, i.e. if you have succeeded in initialising before!
 * Returns zero if successful, and -EINVAL if the cpufreq_driver is
 * currently not initialised.
 */
2493
int cpufreq_unregister_driver(struct cpufreq_driver *driver)
L
Linus Torvalds 已提交
2494 2495 2496
{
	unsigned long flags;

2497
	if (!cpufreq_driver || (driver != cpufreq_driver))
L
Linus Torvalds 已提交
2498 2499
		return -EINVAL;

2500
	pr_debug("unregistering driver %s\n", driver->name);
L
Linus Torvalds 已提交
2501

2502 2503
	/* Protect against concurrent cpu hotplug */
	get_online_cpus();
2504
	subsys_interface_unregister(&cpufreq_interface);
2505
	remove_boost_sysfs_file();
2506
	unregister_hotcpu_notifier(&cpufreq_cpu_notifier);
L
Linus Torvalds 已提交
2507

2508
	write_lock_irqsave(&cpufreq_driver_lock, flags);
2509

2510
	cpufreq_driver = NULL;
2511

2512
	write_unlock_irqrestore(&cpufreq_driver_lock, flags);
2513
	put_online_cpus();
L
Linus Torvalds 已提交
2514 2515 2516 2517

	return 0;
}
EXPORT_SYMBOL_GPL(cpufreq_unregister_driver);
2518

2519 2520 2521 2522 2523 2524 2525 2526
/*
 * Stop cpufreq at shutdown to make sure it isn't holding any locks
 * or mutexes when secondary CPUs are halted.
 */
static struct syscore_ops cpufreq_syscore_ops = {
	.shutdown = cpufreq_suspend,
};

2527 2528 2529
struct kobject *cpufreq_global_kobject;
EXPORT_SYMBOL(cpufreq_global_kobject);

2530 2531
static int __init cpufreq_core_init(void)
{
2532 2533 2534
	if (cpufreq_disabled())
		return -ENODEV;

2535
	cpufreq_global_kobject = kobject_create_and_add("cpufreq", &cpu_subsys.dev_root->kobj);
2536 2537
	BUG_ON(!cpufreq_global_kobject);

2538 2539
	register_syscore_ops(&cpufreq_syscore_ops);

2540 2541 2542
	return 0;
}
core_initcall(cpufreq_core_init);