cpuidle.c 14.8 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 <linux/suspend.h>
23
#include <linux/tick.h>
24
#include <trace/events/power.h>
25 26 27 28

#include "cpuidle.h"

DEFINE_PER_CPU(struct cpuidle_device *, cpuidle_devices);
29
DEFINE_PER_CPU(struct cpuidle_device, cpuidle_dev);
30 31 32 33 34

DEFINE_MUTEX(cpuidle_lock);
LIST_HEAD(cpuidle_detected_devices);

static int enabled_devices;
35
static int off __read_mostly;
L
Len Brown 已提交
36
static int initialized __read_mostly;
37 38 39 40 41

int cpuidle_disabled(void)
{
	return off;
}
42 43 44 45
void disable_cpuidle(void)
{
	off = 1;
}
46

47 48
bool cpuidle_not_available(struct cpuidle_driver *drv,
			   struct cpuidle_device *dev)
49 50 51 52
{
	return off || !initialized || !drv || !dev || !dev->enabled;
}

53 54 55
/**
 * cpuidle_play_dead - cpu off-lining
 *
56
 * Returns in case of an error or no driver
57 58 59 60
 */
int cpuidle_play_dead(void)
{
	struct cpuidle_device *dev = __this_cpu_read(cpuidle_devices);
D
Daniel Lezcano 已提交
61
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
62
	int i;
63

64 65 66
	if (!drv)
		return -ENODEV;

67
	/* Find lowest-power state that supports long-term idle */
68
	for (i = drv->state_count - 1; i >= 0; i--)
69 70
		if (drv->states[i].enter_dead)
			return drv->states[i].enter_dead(dev, i);
71 72 73 74

	return -ENODEV;
}

75
static int find_deepest_state(struct cpuidle_driver *drv,
76 77 78 79
			      struct cpuidle_device *dev,
			      unsigned int max_latency,
			      unsigned int forbidden_flags,
			      bool freeze)
80 81
{
	unsigned int latency_req = 0;
82
	int i, ret = -ENXIO;
83

84
	for (i = 0; i < drv->state_count; i++) {
85 86 87
		struct cpuidle_state *s = &drv->states[i];
		struct cpuidle_state_usage *su = &dev->states_usage[i];

88
		if (s->disabled || su->disable || s->exit_latency <= latency_req
89 90
		    || s->exit_latency > max_latency
		    || (s->flags & forbidden_flags)
91
		    || (freeze && !s->enter_freeze))
92 93 94 95 96 97 98 99
			continue;

		latency_req = s->exit_latency;
		ret = i;
	}
	return ret;
}

100
#ifdef CONFIG_SUSPEND
101 102 103 104 105 106 107 108
/**
 * cpuidle_find_deepest_state - Find the deepest available idle state.
 * @drv: cpuidle driver for the given CPU.
 * @dev: cpuidle device for the given CPU.
 */
int cpuidle_find_deepest_state(struct cpuidle_driver *drv,
			       struct cpuidle_device *dev)
{
109
	return find_deepest_state(drv, dev, UINT_MAX, 0, false);
110 111
}

112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130
static void enter_freeze_proper(struct cpuidle_driver *drv,
				struct cpuidle_device *dev, int index)
{
	tick_freeze();
	/*
	 * The state used here cannot be a "coupled" one, because the "coupled"
	 * cpuidle mechanism enables interrupts and doing that with timekeeping
	 * suspended is generally unsafe.
	 */
	drv->states[index].enter_freeze(dev, drv, index);
	WARN_ON(!irqs_disabled());
	/*
	 * timekeeping_resume() that will be called by tick_unfreeze() for the
	 * last CPU executing it calls functions containing RCU read-side
	 * critical sections, so tell RCU about that.
	 */
	RCU_NONIDLE(tick_unfreeze());
}

131 132
/**
 * cpuidle_enter_freeze - Enter an idle state suitable for suspend-to-idle.
133 134
 * @drv: cpuidle driver for the given CPU.
 * @dev: cpuidle device for the given CPU.
135
 *
136
 * If there are states with the ->enter_freeze callback, find the deepest of
137
 * them and enter it with frozen tick.
138
 */
139
int cpuidle_enter_freeze(struct cpuidle_driver *drv, struct cpuidle_device *dev)
140 141 142
{
	int index;

143 144 145 146 147
	/*
	 * Find the deepest state with ->enter_freeze present, which guarantees
	 * that interrupts won't be enabled when it exits and allows the tick to
	 * be frozen safely.
	 */
148
	index = find_deepest_state(drv, dev, UINT_MAX, 0, true);
149
	if (index >= 0)
150 151
		enter_freeze_proper(drv, dev, index);

152
	return index;
153
}
154
#endif /* CONFIG_SUSPEND */
155

156 157 158 159
/**
 * cpuidle_enter_state - enter the state and update stats
 * @dev: cpuidle device for this cpu
 * @drv: cpuidle driver for this cpu
160
 * @index: index into the states table in @drv of the state to enter
161 162
 */
int cpuidle_enter_state(struct cpuidle_device *dev, struct cpuidle_driver *drv,
163
			int index)
164 165 166
{
	int entered_state;

167
	struct cpuidle_state *target_state = &drv->states[index];
168
	bool broadcast = !!(target_state->flags & CPUIDLE_FLAG_TIMER_STOP);
169 170 171
	ktime_t time_start, time_end;
	s64 diff;

172 173 174 175 176
	/*
	 * Tell the time framework to switch to a broadcast timer because our
	 * local timer will be shut down.  If a local timer is used from another
	 * CPU as a broadcast timer, this call may fail if it is not available.
	 */
177
	if (broadcast && tick_broadcast_enter()) {
178 179 180 181 182 183 184
		index = find_deepest_state(drv, dev, target_state->exit_latency,
					   CPUIDLE_FLAG_TIMER_STOP, false);
		if (index < 0) {
			default_idle_call();
			return -EBUSY;
		}
		target_state = &drv->states[index];
185
	}
186

187 188 189
	/* Take note of the planned idle state. */
	sched_idle_set_state(target_state);

190
	trace_cpu_idle_rcuidle(index, dev->cpu);
191 192 193 194 195
	time_start = ktime_get();

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

	time_end = ktime_get();
196
	trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, dev->cpu);
197

198 199 200
	/* The cpu is no longer idle or about to enter idle. */
	sched_idle_set_state(NULL);

201 202 203 204 205 206 207
	if (broadcast) {
		if (WARN_ON_ONCE(!irqs_disabled()))
			local_irq_disable();

		tick_broadcast_exit();
	}

208 209
	if (!cpuidle_state_is_coupled(dev, drv, entered_state))
		local_irq_enable();
210 211 212 213 214 215

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

	dev->last_residency = (int) diff;
216 217 218 219 220 221

	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.
		 */
222
		dev->states_usage[entered_state].time += dev->last_residency;
223 224 225 226 227 228 229 230
		dev->states_usage[entered_state].usage++;
	} else {
		dev->last_residency = 0;
	}

	return entered_state;
}

231
/**
232 233 234 235
 * cpuidle_select - ask the cpuidle framework to choose an idle state
 *
 * @drv: the cpuidle driver
 * @dev: the cpuidle device
236
 *
237
 * Returns the index of the idle state.
238
 */
239
int cpuidle_select(struct cpuidle_driver *drv, struct cpuidle_device *dev)
240
{
241 242
	return cpuidle_curr_governor->select(drv, dev);
}
243

244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260
/**
 * cpuidle_enter - enter into the specified idle state
 *
 * @drv:   the cpuidle driver tied with the cpu
 * @dev:   the cpuidle device
 * @index: the index in the idle state table
 *
 * Returns the index in the idle state, < 0 in case of error.
 * The error code depends on the backend driver
 */
int cpuidle_enter(struct cpuidle_driver *drv, struct cpuidle_device *dev,
		  int index)
{
	if (cpuidle_state_is_coupled(dev, drv, index))
		return cpuidle_enter_state_coupled(dev, drv, index);
	return cpuidle_enter_state(dev, drv, index);
}
261

262 263 264 265 266 267 268 269 270 271
/**
 * cpuidle_reflect - tell the underlying governor what was the state
 * we were in
 *
 * @dev  : the cpuidle device
 * @index: the index in the idle state table
 *
 */
void cpuidle_reflect(struct cpuidle_device *dev, int index)
{
272
	if (cpuidle_curr_governor->reflect && index >= 0)
273
		cpuidle_curr_governor->reflect(dev, index);
274 275 276 277 278 279 280
}

/**
 * cpuidle_install_idle_handler - installs the cpuidle idle loop handler
 */
void cpuidle_install_idle_handler(void)
{
L
Len Brown 已提交
281
	if (enabled_devices) {
282 283
		/* Make sure all changes finished before we switch to new idle */
		smp_wmb();
L
Len Brown 已提交
284
		initialized = 1;
285 286 287 288 289 290 291 292
	}
}

/**
 * cpuidle_uninstall_idle_handler - uninstalls the cpuidle idle loop handler
 */
void cpuidle_uninstall_idle_handler(void)
{
L
Len Brown 已提交
293 294
	if (enabled_devices) {
		initialized = 0;
295
		wake_up_all_idle_cpus();
296
	}
297 298 299 300 301 302

	/*
	 * Make sure external observers (such as the scheduler)
	 * are done looking at pointed idle states.
	 */
	synchronize_rcu();
303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326
}

/**
 * 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);

327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342
/* 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);
}

343 344 345 346 347 348 349 350 351
/**
 * 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)
{
352
	int ret;
D
Daniel Lezcano 已提交
353
	struct cpuidle_driver *drv;
354

355 356 357
	if (!dev)
		return -EINVAL;

358 359
	if (dev->enabled)
		return 0;
D
Daniel Lezcano 已提交
360 361 362

	drv = cpuidle_get_cpu_driver(dev);

363
	if (!drv || !cpuidle_curr_governor)
364
		return -EIO;
D
Daniel Lezcano 已提交
365

366 367 368
	if (!dev->registered)
		return -EINVAL;

D
Daniel Lezcano 已提交
369 370
	ret = cpuidle_add_device_sysfs(dev);
	if (ret)
371 372 373
		return ret;

	if (cpuidle_curr_governor->enable &&
374
	    (ret = cpuidle_curr_governor->enable(drv, dev)))
375 376 377 378 379 380 381 382 383 384
		goto fail_sysfs;

	smp_wmb();

	dev->enabled = 1;

	enabled_devices++;
	return 0;

fail_sysfs:
D
Daniel Lezcano 已提交
385
	cpuidle_remove_device_sysfs(dev);
386 387 388 389 390 391 392 393 394 395 396 397 398 399 400

	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
401 402
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);

403
	if (!dev || !dev->enabled)
404
		return;
D
Daniel Lezcano 已提交
405 406

	if (!drv || !cpuidle_curr_governor)
407 408 409 410 411
		return;

	dev->enabled = 0;

	if (cpuidle_curr_governor->disable)
D
Daniel Lezcano 已提交
412
		cpuidle_curr_governor->disable(drv, dev);
413

D
Daniel Lezcano 已提交
414
	cpuidle_remove_device_sysfs(dev);
415 416 417 418 419
	enabled_devices--;
}

EXPORT_SYMBOL_GPL(cpuidle_disable_device);

420 421 422 423 424 425 426 427 428
static void __cpuidle_unregister_device(struct cpuidle_device *dev)
{
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);

	list_del(&dev->device_list);
	per_cpu(cpuidle_devices, dev->cpu) = NULL;
	module_put(drv->owner);
}

429
static void __cpuidle_device_init(struct cpuidle_device *dev)
430 431 432 433 434
{
	memset(dev->states_usage, 0, sizeof(dev->states_usage));
	dev->last_residency = 0;
}

435
/**
436 437
 * __cpuidle_register_device - internal register function called before register
 * and enable routines
438
 * @dev: the cpu
439 440
 *
 * cpuidle_lock mutex must be held before this is called
441
 */
442
static int __cpuidle_register_device(struct cpuidle_device *dev)
443 444
{
	int ret;
D
Daniel Lezcano 已提交
445
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
446

D
Daniel Lezcano 已提交
447
	if (!try_module_get(drv->owner))
448 449 450 451 452
		return -EINVAL;

	per_cpu(cpuidle_devices, dev->cpu) = dev;
	list_add(&dev->device_list, &cpuidle_detected_devices);

453
	ret = cpuidle_coupled_register_device(dev);
454
	if (ret)
455
		__cpuidle_unregister_device(dev);
456 457
	else
		dev->registered = 1;
458

459
	return ret;
460 461 462 463 464 465 466 467
}

/**
 * cpuidle_register_device - registers a CPU's idle PM feature
 * @dev: the cpu
 */
int cpuidle_register_device(struct cpuidle_device *dev)
{
468
	int ret = -EBUSY;
469

470 471 472
	if (!dev)
		return -EINVAL;

473 474
	mutex_lock(&cpuidle_lock);

475 476 477
	if (dev->registered)
		goto out_unlock;

478
	__cpuidle_device_init(dev);
479

480 481 482 483 484 485 486
	ret = __cpuidle_register_device(dev);
	if (ret)
		goto out_unlock;

	ret = cpuidle_add_sysfs(dev);
	if (ret)
		goto out_unregister;
487

488
	ret = cpuidle_enable_device(dev);
489 490
	if (ret)
		goto out_sysfs;
491

492 493
	cpuidle_install_idle_handler();

494
out_unlock:
495 496
	mutex_unlock(&cpuidle_lock);

497 498 499 500 501 502 503
	return ret;

out_sysfs:
	cpuidle_remove_sysfs(dev);
out_unregister:
	__cpuidle_unregister_device(dev);
	goto out_unlock;
504 505 506 507 508 509 510 511 512 513
}

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)
{
514
	if (!dev || dev->registered == 0)
515 516
		return;

517 518 519 520
	cpuidle_pause_and_lock();

	cpuidle_disable_device(dev);

521
	cpuidle_remove_sysfs(dev);
522 523

	__cpuidle_unregister_device(dev);
524

525 526
	cpuidle_coupled_unregister_device(dev);

527 528 529 530 531
	cpuidle_resume_and_unlock();
}

EXPORT_SYMBOL_GPL(cpuidle_unregister_device);

D
Daniel Lezcano 已提交
532
/**
533 534 535 536 537 538 539 540 541 542 543
 * 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;

544
	for_each_cpu(cpu, drv->cpumask) {
545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
		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;
	}

576
	for_each_cpu(cpu, drv->cpumask) {
577 578 579 580 581
		device = &per_cpu(cpuidle_dev, cpu);
		device->cpu = cpu;

#ifdef CONFIG_ARCH_NEEDS_CPU_IDLE_COUPLED
		/*
582
		 * On multiplatform for ARM, the coupled idle states could be
583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602
		 * 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);

603 604 605 606 607 608 609 610 611 612 613
#ifdef CONFIG_SMP

/*
 * 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)
{
614
	wake_up_all_idle_cpus();
615 616 617 618 619 620 621
	return NOTIFY_OK;
}

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

M
Mark Gross 已提交
622 623 624 625
static inline void latency_notifier_init(struct notifier_block *n)
{
	pm_qos_add_notifier(PM_QOS_CPU_DMA_LATENCY, n);
}
626 627 628 629 630 631 632 633 634 635 636 637 638 639

#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;

640 641 642
	if (cpuidle_disabled())
		return -ENODEV;

643
	ret = cpuidle_add_interface(cpu_subsys.dev_root);
644 645 646 647 648 649 650 651
	if (ret)
		return ret;

	latency_notifier_init(&cpuidle_latency_notifier);

	return 0;
}

652
module_param(off, int, 0444);
653
core_initcall(cpuidle_init);