cpuidle.c 12.6 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

#include "cpuidle.h"

DEFINE_PER_CPU(struct cpuidle_device *, cpuidle_devices);
27
DEFINE_PER_CPU(struct cpuidle_device, cpuidle_dev);
28 29 30 31 32

DEFINE_MUTEX(cpuidle_lock);
LIST_HEAD(cpuidle_detected_devices);

static int enabled_devices;
33
static int off __read_mostly;
L
Len Brown 已提交
34
static int initialized __read_mostly;
35 36 37 38 39

int cpuidle_disabled(void)
{
	return off;
}
40 41 42 43
void disable_cpuidle(void)
{
	off = 1;
}
44

45 46
static int __cpuidle_register_device(struct cpuidle_device *dev);

47 48 49
/**
 * cpuidle_play_dead - cpu off-lining
 *
50
 * Returns in case of an error or no driver
51 52 53 54
 */
int cpuidle_play_dead(void)
{
	struct cpuidle_device *dev = __this_cpu_read(cpuidle_devices);
D
Daniel Lezcano 已提交
55
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
56
	int i;
57

58 59 60
	if (!drv)
		return -ENODEV;

61
	/* Find lowest-power state that supports long-term idle */
62 63 64
	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);
65 66 67 68

	return -ENODEV;
}

69 70 71 72 73 74 75
/**
 * 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,
76
			int index)
77 78 79
{
	int entered_state;

80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
	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;
97 98 99 100 101 102

	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.
		 */
103
		dev->states_usage[entered_state].time += dev->last_residency;
104 105 106 107 108 109 110 111
		dev->states_usage[entered_state].usage++;
	} else {
		dev->last_residency = 0;
	}

	return entered_state;
}

112 113 114 115
/**
 * cpuidle_idle_call - the main idle loop
 *
 * NOTE: no locks or semaphores should be used here
L
Len Brown 已提交
116
 * return non-zero on failure
117
 */
L
Len Brown 已提交
118
int cpuidle_idle_call(void)
119
{
120
	struct cpuidle_device *dev = __this_cpu_read(cpuidle_devices);
D
Daniel Lezcano 已提交
121
	struct cpuidle_driver *drv;
122
	int next_state, entered_state;
123

L
Len Brown 已提交
124 125 126 127 128 129
	if (off)
		return -ENODEV;

	if (!initialized)
		return -ENODEV;

130
	/* check if the device is ready */
L
Len Brown 已提交
131 132
	if (!dev || !dev->enabled)
		return -EBUSY;
133

D
Daniel Lezcano 已提交
134 135
	drv = cpuidle_get_cpu_driver(dev);

136
	/* ask the governor for the next state */
137
	next_state = cpuidle_curr_governor->select(drv, dev);
138
	if (need_resched()) {
139 140 141 142
		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);
143
		local_irq_enable();
L
Len Brown 已提交
144
		return 0;
145 146
	}

147
	trace_cpu_idle_rcuidle(next_state, dev->cpu);
148

149 150 151 152
	if (drv->states[next_state].flags & CPUIDLE_FLAG_TIMER_STOP)
		clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_ENTER,
				   &dev->cpu);

153 154 155 156 157
	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);
158

159 160 161 162
	if (drv->states[next_state].flags & CPUIDLE_FLAG_TIMER_STOP)
		clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_EXIT,
				   &dev->cpu);

163
	trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, dev->cpu);
164

165 166
	/* give the governor an opportunity to reflect on the outcome */
	if (cpuidle_curr_governor->reflect)
167
		cpuidle_curr_governor->reflect(dev, entered_state);
L
Len Brown 已提交
168 169

	return 0;
170 171 172 173 174 175 176
}

/**
 * cpuidle_install_idle_handler - installs the cpuidle idle loop handler
 */
void cpuidle_install_idle_handler(void)
{
L
Len Brown 已提交
177
	if (enabled_devices) {
178 179
		/* Make sure all changes finished before we switch to new idle */
		smp_wmb();
L
Len Brown 已提交
180
		initialized = 1;
181 182 183 184 185 186 187 188
	}
}

/**
 * cpuidle_uninstall_idle_handler - uninstalls the cpuidle idle loop handler
 */
void cpuidle_uninstall_idle_handler(void)
{
L
Len Brown 已提交
189 190
	if (enabled_devices) {
		initialized = 0;
191
		kick_all_cpus_sync();
192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216
	}
}

/**
 * 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);

217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232
/* 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);
}

233
#ifdef CONFIG_ARCH_HAS_CPU_RELAX
234 235
static int poll_idle(struct cpuidle_device *dev,
		struct cpuidle_driver *drv, int index)
236 237 238 239 240 241 242 243 244 245 246 247 248 249
{
	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;

250 251 252
	dev->last_residency = (int) diff;

	return index;
253 254
}

255
static void poll_idle_init(struct cpuidle_driver *drv)
256
{
257
	struct cpuidle_state *state = &drv->states[0];
258

259
	snprintf(state->name, CPUIDLE_NAME_LEN, "POLL");
260 261 262 263
	snprintf(state->desc, CPUIDLE_DESC_LEN, "CPUIDLE CORE POLL IDLE");
	state->exit_latency = 0;
	state->target_residency = 0;
	state->power_usage = -1;
264
	state->flags = 0;
265
	state->enter = poll_idle;
266
	state->disabled = false;
267 268
}
#else
269
static void poll_idle_init(struct cpuidle_driver *drv) {}
270 271
#endif /* CONFIG_ARCH_HAS_CPU_RELAX */

272 273 274 275 276 277 278 279 280 281
/**
 * 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
282
	struct cpuidle_driver *drv;
283

284 285 286
	if (!dev)
		return -EINVAL;

287 288
	if (dev->enabled)
		return 0;
D
Daniel Lezcano 已提交
289 290 291

	drv = cpuidle_get_cpu_driver(dev);

292
	if (!drv || !cpuidle_curr_governor)
293
		return -EIO;
D
Daniel Lezcano 已提交
294

295
	if (!dev->state_count)
296
		dev->state_count = drv->state_count;
297

298 299 300 301 302 303
	if (dev->registered == 0) {
		ret = __cpuidle_register_device(dev);
		if (ret)
			return ret;
	}

304
	poll_idle_init(drv);
305

D
Daniel Lezcano 已提交
306 307
	ret = cpuidle_add_device_sysfs(dev);
	if (ret)
308 309 310
		return ret;

	if (cpuidle_curr_governor->enable &&
311
	    (ret = cpuidle_curr_governor->enable(drv, dev)))
312 313 314
		goto fail_sysfs;

	for (i = 0; i < dev->state_count; i++) {
315 316
		dev->states_usage[i].usage = 0;
		dev->states_usage[i].time = 0;
317 318 319 320 321 322 323 324 325 326 327
	}
	dev->last_residency = 0;

	smp_wmb();

	dev->enabled = 1;

	enabled_devices++;
	return 0;

fail_sysfs:
D
Daniel Lezcano 已提交
328
	cpuidle_remove_device_sysfs(dev);
329 330 331 332 333 334 335 336 337 338 339 340 341 342 343

	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
344 345
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);

346
	if (!dev || !dev->enabled)
347
		return;
D
Daniel Lezcano 已提交
348 349

	if (!drv || !cpuidle_curr_governor)
350 351 352 353 354
		return;

	dev->enabled = 0;

	if (cpuidle_curr_governor->disable)
D
Daniel Lezcano 已提交
355
		cpuidle_curr_governor->disable(drv, dev);
356

D
Daniel Lezcano 已提交
357
	cpuidle_remove_device_sysfs(dev);
358 359 360 361 362 363
	enabled_devices--;
}

EXPORT_SYMBOL_GPL(cpuidle_disable_device);

/**
364 365
 * __cpuidle_register_device - internal register function called before register
 * and enable routines
366
 * @dev: the cpu
367 368
 *
 * cpuidle_lock mutex must be held before this is called
369
 */
370
static int __cpuidle_register_device(struct cpuidle_device *dev)
371 372
{
	int ret;
D
Daniel Lezcano 已提交
373
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
374

D
Daniel Lezcano 已提交
375
	if (!try_module_get(drv->owner))
376 377 378 379
		return -EINVAL;

	per_cpu(cpuidle_devices, dev->cpu) = dev;
	list_add(&dev->device_list, &cpuidle_detected_devices);
380
	ret = cpuidle_add_sysfs(dev);
381 382
	if (ret)
		goto err_sysfs;
383

384 385 386
	ret = cpuidle_coupled_register_device(dev);
	if (ret)
		goto err_coupled;
387

388 389
	dev->registered = 1;
	return 0;
390

391
err_coupled:
392
	cpuidle_remove_sysfs(dev);
393 394 395
err_sysfs:
	list_del(&dev->device_list);
	per_cpu(cpuidle_devices, dev->cpu) = NULL;
D
Daniel Lezcano 已提交
396
	module_put(drv->owner);
397
	return ret;
398 399 400 401 402 403 404 405 406 407
}

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

408 409 410
	if (!dev)
		return -EINVAL;

411 412 413 414 415 416 417
	mutex_lock(&cpuidle_lock);

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

418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434
	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
435
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
436

437 438 439
	if (dev->registered == 0)
		return;

440 441 442 443
	cpuidle_pause_and_lock();

	cpuidle_disable_device(dev);

444
	cpuidle_remove_sysfs(dev);
445 446 447
	list_del(&dev->device_list);
	per_cpu(cpuidle_devices, dev->cpu) = NULL;

448 449
	cpuidle_coupled_unregister_device(dev);

450 451
	cpuidle_resume_and_unlock();

D
Daniel Lezcano 已提交
452
	module_put(drv->owner);
453 454 455 456
}

EXPORT_SYMBOL_GPL(cpuidle_unregister_device);

D
Daniel Lezcano 已提交
457
/**
458 459 460 461 462 463 464 465 466 467 468
 * cpuidle_unregister: unregister a driver and the devices. This function
 * can be used only if the driver has been previously registered through
 * the cpuidle_register function.
 *
 * @drv: a valid pointer to a struct cpuidle_driver
 */
void cpuidle_unregister(struct cpuidle_driver *drv)
{
	int cpu;
	struct cpuidle_device *device;

469
	for_each_cpu(cpu, drv->cpumask) {
470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
		device = &per_cpu(cpuidle_dev, cpu);
		cpuidle_unregister_device(device);
	}

	cpuidle_unregister_driver(drv);
}
EXPORT_SYMBOL_GPL(cpuidle_unregister);

/**
 * cpuidle_register: registers the driver and the cpu devices with the
 * coupled_cpus passed as parameter. This function is used for all common
 * initialization pattern there are in the arch specific drivers. The
 * devices is globally defined in this file.
 *
 * @drv         : a valid pointer to a struct cpuidle_driver
 * @coupled_cpus: a cpumask for the coupled states
 *
 * Returns 0 on success, < 0 otherwise
 */
int cpuidle_register(struct cpuidle_driver *drv,
		     const struct cpumask *const coupled_cpus)
{
	int ret, cpu;
	struct cpuidle_device *device;

	ret = cpuidle_register_driver(drv);
	if (ret) {
		pr_err("failed to register cpuidle driver\n");
		return ret;
	}

501
	for_each_cpu(cpu, drv->cpumask) {
502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527
		device = &per_cpu(cpuidle_dev, cpu);
		device->cpu = cpu;

#ifdef CONFIG_ARCH_NEEDS_CPU_IDLE_COUPLED
		/*
		 * On multiplatform for ARM, the coupled idle states could
		 * enabled in the kernel even if the cpuidle driver does not
		 * use it. Note, coupled_cpus is a struct copy.
		 */
		if (coupled_cpus)
			device->coupled_cpus = *coupled_cpus;
#endif
		ret = cpuidle_register_device(device);
		if (!ret)
			continue;

		pr_err("Failed to register cpuidle device for cpu%d\n", cpu);

		cpuidle_unregister(drv);
		break;
	}

	return ret;
}
EXPORT_SYMBOL_GPL(cpuidle_register);

528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543
#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)
{
544
	smp_call_function(smp_callback, NULL, 1);
545 546 547 548 549 550 551
	return NOTIFY_OK;
}

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

M
Mark Gross 已提交
552 553 554 555
static inline void latency_notifier_init(struct notifier_block *n)
{
	pm_qos_add_notifier(PM_QOS_CPU_DMA_LATENCY, n);
}
556 557 558 559 560 561 562 563 564 565 566 567 568 569

#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;

570 571 572
	if (cpuidle_disabled())
		return -ENODEV;

573
	ret = cpuidle_add_interface(cpu_subsys.dev_root);
574 575 576 577 578 579 580 581
	if (ret)
		return ret;

	latency_notifier_init(&cpuidle_latency_notifier);

	return 0;
}

582
module_param(off, int, 0444);
583
core_initcall(cpuidle_init);