cpuidle.c 10.7 KB
Newer Older
1 2 3 4 5 6 7 8 9 10
/*
 * cpuidle.c - core cpuidle infrastructure
 *
 * (C) 2006-2007 Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
 *               Shaohua Li <shaohua.li@intel.com>
 *               Adam Belay <abelay@novell.com>
 *
 * This code is licenced under the GPL.
 */

11
#include <linux/clockchips.h>
12 13 14 15
#include <linux/kernel.h>
#include <linux/mutex.h>
#include <linux/sched.h>
#include <linux/notifier.h>
16
#include <linux/pm_qos.h>
17 18
#include <linux/cpu.h>
#include <linux/cpuidle.h>
19
#include <linux/ktime.h>
20
#include <linux/hrtimer.h>
21
#include <linux/module.h>
22
#include <trace/events/power.h>
23 24 25 26 27 28 29 30 31

#include "cpuidle.h"

DEFINE_PER_CPU(struct cpuidle_device *, cpuidle_devices);

DEFINE_MUTEX(cpuidle_lock);
LIST_HEAD(cpuidle_detected_devices);

static int enabled_devices;
32
static int off __read_mostly;
L
Len Brown 已提交
33
static int initialized __read_mostly;
34 35 36 37 38

int cpuidle_disabled(void)
{
	return off;
}
39 40 41 42
void disable_cpuidle(void)
{
	off = 1;
}
43

44 45
static int __cpuidle_register_device(struct cpuidle_device *dev);

46 47 48
/**
 * cpuidle_play_dead - cpu off-lining
 *
49
 * Returns in case of an error or no driver
50 51 52 53
 */
int cpuidle_play_dead(void)
{
	struct cpuidle_device *dev = __this_cpu_read(cpuidle_devices);
D
Daniel Lezcano 已提交
54
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
55
	int i;
56

57 58 59
	if (!drv)
		return -ENODEV;

60
	/* Find lowest-power state that supports long-term idle */
61 62 63
	for (i = drv->state_count - 1; i >= CPUIDLE_DRIVER_STATE_START; i--)
		if (drv->states[i].enter_dead)
			return drv->states[i].enter_dead(dev, i);
64 65 66 67

	return -ENODEV;
}

68 69 70 71 72 73 74
/**
 * cpuidle_enter_state - enter the state and update stats
 * @dev: cpuidle device for this cpu
 * @drv: cpuidle driver for this cpu
 * @next_state: index into drv->states of the state to enter
 */
int cpuidle_enter_state(struct cpuidle_device *dev, struct cpuidle_driver *drv,
75
			int index)
76 77 78
{
	int entered_state;

79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
	struct cpuidle_state *target_state = &drv->states[index];
	ktime_t time_start, time_end;
	s64 diff;

	time_start = ktime_get();

	entered_state = target_state->enter(dev, drv, index);

	time_end = ktime_get();

	local_irq_enable();

	diff = ktime_to_us(ktime_sub(time_end, time_start));
	if (diff > INT_MAX)
		diff = INT_MAX;

	dev->last_residency = (int) diff;
96 97 98 99 100 101

	if (entered_state >= 0) {
		/* Update cpuidle counters */
		/* This can be moved to within driver enter routine
		 * but that results in multiple copies of same code.
		 */
102
		dev->states_usage[entered_state].time += dev->last_residency;
103 104 105 106 107 108 109 110
		dev->states_usage[entered_state].usage++;
	} else {
		dev->last_residency = 0;
	}

	return entered_state;
}

111 112 113 114
/**
 * cpuidle_idle_call - the main idle loop
 *
 * NOTE: no locks or semaphores should be used here
L
Len Brown 已提交
115
 * return non-zero on failure
116
 */
L
Len Brown 已提交
117
int cpuidle_idle_call(void)
118
{
119
	struct cpuidle_device *dev = __this_cpu_read(cpuidle_devices);
D
Daniel Lezcano 已提交
120
	struct cpuidle_driver *drv;
121
	int next_state, entered_state;
122

L
Len Brown 已提交
123 124 125 126 127 128
	if (off)
		return -ENODEV;

	if (!initialized)
		return -ENODEV;

129
	/* check if the device is ready */
L
Len Brown 已提交
130 131
	if (!dev || !dev->enabled)
		return -EBUSY;
132

D
Daniel Lezcano 已提交
133 134
	drv = cpuidle_get_cpu_driver(dev);

135
	/* ask the governor for the next state */
136
	next_state = cpuidle_curr_governor->select(drv, dev);
137
	if (need_resched()) {
138 139 140 141
		dev->last_residency = 0;
		/* give the governor an opportunity to reflect on the outcome */
		if (cpuidle_curr_governor->reflect)
			cpuidle_curr_governor->reflect(dev, next_state);
142
		local_irq_enable();
L
Len Brown 已提交
143
		return 0;
144 145
	}

146
	trace_cpu_idle_rcuidle(next_state, dev->cpu);
147

148 149 150 151
	if (drv->states[next_state].flags & CPUIDLE_FLAG_TIMER_STOP)
		clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_ENTER,
				   &dev->cpu);

152 153 154 155 156
	if (cpuidle_state_is_coupled(dev, drv, next_state))
		entered_state = cpuidle_enter_state_coupled(dev, drv,
							    next_state);
	else
		entered_state = cpuidle_enter_state(dev, drv, next_state);
157

158 159 160 161
	if (drv->states[next_state].flags & CPUIDLE_FLAG_TIMER_STOP)
		clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_EXIT,
				   &dev->cpu);

162
	trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, dev->cpu);
163

164 165
	/* give the governor an opportunity to reflect on the outcome */
	if (cpuidle_curr_governor->reflect)
166
		cpuidle_curr_governor->reflect(dev, entered_state);
L
Len Brown 已提交
167 168

	return 0;
169 170 171 172 173 174 175
}

/**
 * cpuidle_install_idle_handler - installs the cpuidle idle loop handler
 */
void cpuidle_install_idle_handler(void)
{
L
Len Brown 已提交
176
	if (enabled_devices) {
177 178
		/* Make sure all changes finished before we switch to new idle */
		smp_wmb();
L
Len Brown 已提交
179
		initialized = 1;
180 181 182 183 184 185 186 187
	}
}

/**
 * cpuidle_uninstall_idle_handler - uninstalls the cpuidle idle loop handler
 */
void cpuidle_uninstall_idle_handler(void)
{
L
Len Brown 已提交
188 189
	if (enabled_devices) {
		initialized = 0;
190
		kick_all_cpus_sync();
191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215
	}
}

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

EXPORT_SYMBOL_GPL(cpuidle_pause_and_lock);

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

EXPORT_SYMBOL_GPL(cpuidle_resume_and_unlock);

216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231
/* Currently used in suspend/resume path to suspend cpuidle */
void cpuidle_pause(void)
{
	mutex_lock(&cpuidle_lock);
	cpuidle_uninstall_idle_handler();
	mutex_unlock(&cpuidle_lock);
}

/* Currently used in suspend/resume path to resume cpuidle */
void cpuidle_resume(void)
{
	mutex_lock(&cpuidle_lock);
	cpuidle_install_idle_handler();
	mutex_unlock(&cpuidle_lock);
}

232
#ifdef CONFIG_ARCH_HAS_CPU_RELAX
233 234
static int poll_idle(struct cpuidle_device *dev,
		struct cpuidle_driver *drv, int index)
235 236 237 238 239 240 241 242 243 244 245 246 247 248
{
	ktime_t	t1, t2;
	s64 diff;

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

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

249 250 251
	dev->last_residency = (int) diff;

	return index;
252 253
}

254
static void poll_idle_init(struct cpuidle_driver *drv)
255
{
256
	struct cpuidle_state *state = &drv->states[0];
257

258
	snprintf(state->name, CPUIDLE_NAME_LEN, "POLL");
259 260 261 262
	snprintf(state->desc, CPUIDLE_DESC_LEN, "CPUIDLE CORE POLL IDLE");
	state->exit_latency = 0;
	state->target_residency = 0;
	state->power_usage = -1;
263
	state->flags = 0;
264
	state->enter = poll_idle;
265
	state->disabled = false;
266 267
}
#else
268
static void poll_idle_init(struct cpuidle_driver *drv) {}
269 270
#endif /* CONFIG_ARCH_HAS_CPU_RELAX */

271 272 273 274 275 276 277 278 279 280
/**
 * cpuidle_enable_device - enables idle PM for a CPU
 * @dev: the CPU
 *
 * This function must be called between cpuidle_pause_and_lock and
 * cpuidle_resume_and_unlock when used externally.
 */
int cpuidle_enable_device(struct cpuidle_device *dev)
{
	int ret, i;
D
Daniel Lezcano 已提交
281
	struct cpuidle_driver *drv;
282

283 284 285
	if (!dev)
		return -EINVAL;

286 287
	if (dev->enabled)
		return 0;
D
Daniel Lezcano 已提交
288 289 290

	drv = cpuidle_get_cpu_driver(dev);

291
	if (!drv || !cpuidle_curr_governor)
292
		return -EIO;
D
Daniel Lezcano 已提交
293

294
	if (!dev->state_count)
295
		dev->state_count = drv->state_count;
296

297 298 299 300 301 302
	if (dev->registered == 0) {
		ret = __cpuidle_register_device(dev);
		if (ret)
			return ret;
	}

303
	poll_idle_init(drv);
304

D
Daniel Lezcano 已提交
305 306
	ret = cpuidle_add_device_sysfs(dev);
	if (ret)
307 308 309
		return ret;

	if (cpuidle_curr_governor->enable &&
310
	    (ret = cpuidle_curr_governor->enable(drv, dev)))
311 312 313
		goto fail_sysfs;

	for (i = 0; i < dev->state_count; i++) {
314 315
		dev->states_usage[i].usage = 0;
		dev->states_usage[i].time = 0;
316 317 318 319 320 321 322 323 324 325 326
	}
	dev->last_residency = 0;

	smp_wmb();

	dev->enabled = 1;

	enabled_devices++;
	return 0;

fail_sysfs:
D
Daniel Lezcano 已提交
327
	cpuidle_remove_device_sysfs(dev);
328 329 330 331 332 333 334 335 336 337 338 339 340 341 342

	return ret;
}

EXPORT_SYMBOL_GPL(cpuidle_enable_device);

/**
 * cpuidle_disable_device - disables idle PM for a CPU
 * @dev: the CPU
 *
 * This function must be called between cpuidle_pause_and_lock and
 * cpuidle_resume_and_unlock when used externally.
 */
void cpuidle_disable_device(struct cpuidle_device *dev)
{
D
Daniel Lezcano 已提交
343 344
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);

345
	if (!dev || !dev->enabled)
346
		return;
D
Daniel Lezcano 已提交
347 348

	if (!drv || !cpuidle_curr_governor)
349 350 351 352 353
		return;

	dev->enabled = 0;

	if (cpuidle_curr_governor->disable)
D
Daniel Lezcano 已提交
354
		cpuidle_curr_governor->disable(drv, dev);
355

D
Daniel Lezcano 已提交
356
	cpuidle_remove_device_sysfs(dev);
357 358 359 360 361 362
	enabled_devices--;
}

EXPORT_SYMBOL_GPL(cpuidle_disable_device);

/**
363 364
 * __cpuidle_register_device - internal register function called before register
 * and enable routines
365
 * @dev: the cpu
366 367
 *
 * cpuidle_lock mutex must be held before this is called
368
 */
369
static int __cpuidle_register_device(struct cpuidle_device *dev)
370 371
{
	int ret;
D
Daniel Lezcano 已提交
372
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
373

D
Daniel Lezcano 已提交
374
	if (!try_module_get(drv->owner))
375 376 377 378
		return -EINVAL;

	per_cpu(cpuidle_devices, dev->cpu) = dev;
	list_add(&dev->device_list, &cpuidle_detected_devices);
379
	ret = cpuidle_add_sysfs(dev);
380 381
	if (ret)
		goto err_sysfs;
382

383 384 385
	ret = cpuidle_coupled_register_device(dev);
	if (ret)
		goto err_coupled;
386

387 388
	dev->registered = 1;
	return 0;
389

390
err_coupled:
391
	cpuidle_remove_sysfs(dev);
392 393 394
err_sysfs:
	list_del(&dev->device_list);
	per_cpu(cpuidle_devices, dev->cpu) = NULL;
D
Daniel Lezcano 已提交
395
	module_put(drv->owner);
396
	return ret;
397 398 399 400 401 402 403 404 405 406
}

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

407 408 409
	if (!dev)
		return -EINVAL;

410 411 412 413 414 415 416
	mutex_lock(&cpuidle_lock);

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

417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433
	cpuidle_enable_device(dev);
	cpuidle_install_idle_handler();

	mutex_unlock(&cpuidle_lock);

	return 0;

}

EXPORT_SYMBOL_GPL(cpuidle_register_device);

/**
 * cpuidle_unregister_device - unregisters a CPU's idle PM feature
 * @dev: the cpu
 */
void cpuidle_unregister_device(struct cpuidle_device *dev)
{
D
Daniel Lezcano 已提交
434
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
435

436 437 438
	if (dev->registered == 0)
		return;

439 440 441 442
	cpuidle_pause_and_lock();

	cpuidle_disable_device(dev);

443
	cpuidle_remove_sysfs(dev);
444 445 446
	list_del(&dev->device_list);
	per_cpu(cpuidle_devices, dev->cpu) = NULL;

447 448
	cpuidle_coupled_unregister_device(dev);

449 450
	cpuidle_resume_and_unlock();

D
Daniel Lezcano 已提交
451
	module_put(drv->owner);
452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471
}

EXPORT_SYMBOL_GPL(cpuidle_unregister_device);

#ifdef CONFIG_SMP

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

/*
 * This function gets called when a part of the kernel has a new latency
 * requirement.  This means we need to get all processors out of their C-state,
 * and then recalculate a new suitable C-state. Just do a cross-cpu IPI; that
 * wakes them all right up.
 */
static int cpuidle_latency_notify(struct notifier_block *b,
		unsigned long l, void *v)
{
472
	smp_call_function(smp_callback, NULL, 1);
473 474 475 476 477 478 479
	return NOTIFY_OK;
}

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

M
Mark Gross 已提交
480 481 482 483
static inline void latency_notifier_init(struct notifier_block *n)
{
	pm_qos_add_notifier(PM_QOS_CPU_DMA_LATENCY, n);
}
484 485 486 487 488 489 490 491 492 493 494 495 496 497

#else /* CONFIG_SMP */

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

#endif /* CONFIG_SMP */

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

498 499 500
	if (cpuidle_disabled())
		return -ENODEV;

501
	ret = cpuidle_add_interface(cpu_subsys.dev_root);
502 503 504 505 506 507 508 509
	if (ret)
		return ret;

	latency_notifier_init(&cpuidle_latency_notifier);

	return 0;
}

510
module_param(off, int, 0444);
511
core_initcall(cpuidle_init);