cpuidle.c 11.8 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14
/*
 * 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.
 */

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

#include "cpuidle.h"

DEFINE_PER_CPU(struct cpuidle_device *, cpuidle_devices);

DEFINE_MUTEX(cpuidle_lock);
LIST_HEAD(cpuidle_detected_devices);

static int enabled_devices;
31
static int off __read_mostly;
L
Len Brown 已提交
32
static int initialized __read_mostly;
33 34 35 36 37

int cpuidle_disabled(void)
{
	return off;
}
38 39 40 41
void disable_cpuidle(void)
{
	off = 1;
}
42

43 44
static int __cpuidle_register_device(struct cpuidle_device *dev);

45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62
static inline int cpuidle_enter(struct cpuidle_device *dev,
				struct cpuidle_driver *drv, int index)
{
	struct cpuidle_state *target_state = &drv->states[index];
	return target_state->enter(dev, drv, index);
}

static inline int cpuidle_enter_tk(struct cpuidle_device *dev,
			       struct cpuidle_driver *drv, int index)
{
	return cpuidle_wrap_enter(dev, drv, index, cpuidle_enter);
}

typedef int (*cpuidle_enter_t)(struct cpuidle_device *dev,
			       struct cpuidle_driver *drv, int index);

static cpuidle_enter_t cpuidle_enter_ops;

63 64 65
/**
 * cpuidle_play_dead - cpu off-lining
 *
66
 * Returns in case of an error or no driver
67 68 69 70
 */
int cpuidle_play_dead(void)
{
	struct cpuidle_device *dev = __this_cpu_read(cpuidle_devices);
D
Daniel Lezcano 已提交
71
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
72 73 74
	int i, dead_state = -1;
	int power_usage = -1;

75 76 77
	if (!drv)
		return -ENODEV;

78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93
	/* Find lowest-power state that supports long-term idle */
	for (i = CPUIDLE_DRIVER_STATE_START; i < drv->state_count; i++) {
		struct cpuidle_state *s = &drv->states[i];

		if (s->power_usage < power_usage && s->enter_dead) {
			power_usage = s->power_usage;
			dead_state = i;
		}
	}

	if (dead_state != -1)
		return drv->states[dead_state].enter_dead(dev, dead_state);

	return -ENODEV;
}

94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121
/**
 * 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,
		int next_state)
{
	int entered_state;

	entered_state = cpuidle_enter_ops(dev, drv, next_state);

	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.
		 */
		dev->states_usage[entered_state].time +=
				(unsigned long long)dev->last_residency;
		dev->states_usage[entered_state].usage++;
	} else {
		dev->last_residency = 0;
	}

	return entered_state;
}

122 123 124 125
/**
 * cpuidle_idle_call - the main idle loop
 *
 * NOTE: no locks or semaphores should be used here
L
Len Brown 已提交
126
 * return non-zero on failure
127
 */
L
Len Brown 已提交
128
int cpuidle_idle_call(void)
129
{
130
	struct cpuidle_device *dev = __this_cpu_read(cpuidle_devices);
D
Daniel Lezcano 已提交
131
	struct cpuidle_driver *drv;
132
	int next_state, entered_state;
133

L
Len Brown 已提交
134 135 136 137 138 139
	if (off)
		return -ENODEV;

	if (!initialized)
		return -ENODEV;

140
	/* check if the device is ready */
L
Len Brown 已提交
141 142
	if (!dev || !dev->enabled)
		return -EBUSY;
143

D
Daniel Lezcano 已提交
144 145
	drv = cpuidle_get_cpu_driver(dev);

146
	/* ask the governor for the next state */
147
	next_state = cpuidle_curr_governor->select(drv, dev);
148
	if (need_resched()) {
149 150 151 152
		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);
153
		local_irq_enable();
L
Len Brown 已提交
154
		return 0;
155 156
	}

157 158
	trace_power_start_rcuidle(POWER_CSTATE, next_state, dev->cpu);
	trace_cpu_idle_rcuidle(next_state, dev->cpu);
159

160 161 162 163 164
	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);
165

166 167
	trace_power_end_rcuidle(dev->cpu);
	trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, dev->cpu);
168

169 170
	/* give the governor an opportunity to reflect on the outcome */
	if (cpuidle_curr_governor->reflect)
171
		cpuidle_curr_governor->reflect(dev, entered_state);
L
Len Brown 已提交
172 173

	return 0;
174 175 176 177 178 179 180
}

/**
 * cpuidle_install_idle_handler - installs the cpuidle idle loop handler
 */
void cpuidle_install_idle_handler(void)
{
L
Len Brown 已提交
181
	if (enabled_devices) {
182 183
		/* Make sure all changes finished before we switch to new idle */
		smp_wmb();
L
Len Brown 已提交
184
		initialized = 1;
185 186 187 188 189 190 191 192
	}
}

/**
 * cpuidle_uninstall_idle_handler - uninstalls the cpuidle idle loop handler
 */
void cpuidle_uninstall_idle_handler(void)
{
L
Len Brown 已提交
193 194
	if (enabled_devices) {
		initialized = 0;
195
		kick_all_cpus_sync();
196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220
	}
}

/**
 * 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);

221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236
/* 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);
}

237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267
/**
 * cpuidle_wrap_enter - performs timekeeping and irqen around enter function
 * @dev: pointer to a valid cpuidle_device object
 * @drv: pointer to a valid cpuidle_driver object
 * @index: index of the target cpuidle state.
 */
int cpuidle_wrap_enter(struct cpuidle_device *dev,
				struct cpuidle_driver *drv, int index,
				int (*enter)(struct cpuidle_device *dev,
					struct cpuidle_driver *drv, int index))
{
	ktime_t time_start, time_end;
	s64 diff;

	time_start = ktime_get();

	index = 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;

	return index;
}

268
#ifdef CONFIG_ARCH_HAS_CPU_RELAX
269 270
static int poll_idle(struct cpuidle_device *dev,
		struct cpuidle_driver *drv, int index)
271 272 273 274 275 276 277 278 279 280 281 282 283 284
{
	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;

285 286 287
	dev->last_residency = (int) diff;

	return index;
288 289
}

290
static void poll_idle_init(struct cpuidle_driver *drv)
291
{
292
	struct cpuidle_state *state = &drv->states[0];
293

294
	snprintf(state->name, CPUIDLE_NAME_LEN, "POLL");
295 296 297 298
	snprintf(state->desc, CPUIDLE_DESC_LEN, "CPUIDLE CORE POLL IDLE");
	state->exit_latency = 0;
	state->target_residency = 0;
	state->power_usage = -1;
299
	state->flags = 0;
300
	state->enter = poll_idle;
301
	state->disabled = false;
302 303
}
#else
304
static void poll_idle_init(struct cpuidle_driver *drv) {}
305 306
#endif /* CONFIG_ARCH_HAS_CPU_RELAX */

307 308 309 310 311 312 313 314 315 316
/**
 * 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
317
	struct cpuidle_driver *drv;
318

319 320 321
	if (!dev)
		return -EINVAL;

322 323
	if (dev->enabled)
		return 0;
D
Daniel Lezcano 已提交
324 325 326

	drv = cpuidle_get_cpu_driver(dev);

327
	if (!drv || !cpuidle_curr_governor)
328
		return -EIO;
D
Daniel Lezcano 已提交
329

330
	if (!dev->state_count)
331
		dev->state_count = drv->state_count;
332

333 334 335 336 337 338
	if (dev->registered == 0) {
		ret = __cpuidle_register_device(dev);
		if (ret)
			return ret;
	}

339 340 341 342
	cpuidle_enter_ops = drv->en_core_tk_irqen ?
		cpuidle_enter_tk : cpuidle_enter;

	poll_idle_init(drv);
343

D
Daniel Lezcano 已提交
344 345
	ret = cpuidle_add_device_sysfs(dev);
	if (ret)
346 347 348
		return ret;

	if (cpuidle_curr_governor->enable &&
349
	    (ret = cpuidle_curr_governor->enable(drv, dev)))
350 351 352
		goto fail_sysfs;

	for (i = 0; i < dev->state_count; i++) {
353 354
		dev->states_usage[i].usage = 0;
		dev->states_usage[i].time = 0;
355 356 357 358 359 360 361 362 363 364 365
	}
	dev->last_residency = 0;

	smp_wmb();

	dev->enabled = 1;

	enabled_devices++;
	return 0;

fail_sysfs:
D
Daniel Lezcano 已提交
366
	cpuidle_remove_device_sysfs(dev);
367 368 369 370 371 372 373 374 375 376 377 378 379 380 381

	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
382 383
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);

384
	if (!dev || !dev->enabled)
385
		return;
D
Daniel Lezcano 已提交
386 387

	if (!drv || !cpuidle_curr_governor)
388 389 390 391 392
		return;

	dev->enabled = 0;

	if (cpuidle_curr_governor->disable)
D
Daniel Lezcano 已提交
393
		cpuidle_curr_governor->disable(drv, dev);
394

D
Daniel Lezcano 已提交
395
	cpuidle_remove_device_sysfs(dev);
396 397 398 399 400 401
	enabled_devices--;
}

EXPORT_SYMBOL_GPL(cpuidle_disable_device);

/**
402 403
 * __cpuidle_register_device - internal register function called before register
 * and enable routines
404
 * @dev: the cpu
405 406
 *
 * cpuidle_lock mutex must be held before this is called
407
 */
408
static int __cpuidle_register_device(struct cpuidle_device *dev)
409 410
{
	int ret;
D
Daniel Lezcano 已提交
411
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
412

D
Daniel Lezcano 已提交
413
	if (!try_module_get(drv->owner))
414 415 416 417
		return -EINVAL;

	per_cpu(cpuidle_devices, dev->cpu) = dev;
	list_add(&dev->device_list, &cpuidle_detected_devices);
418
	ret = cpuidle_add_sysfs(dev);
419 420
	if (ret)
		goto err_sysfs;
421

422 423 424
	ret = cpuidle_coupled_register_device(dev);
	if (ret)
		goto err_coupled;
425

426 427
	dev->registered = 1;
	return 0;
428

429
err_coupled:
430
	cpuidle_remove_sysfs(dev);
431 432 433
err_sysfs:
	list_del(&dev->device_list);
	per_cpu(cpuidle_devices, dev->cpu) = NULL;
D
Daniel Lezcano 已提交
434
	module_put(drv->owner);
435
	return ret;
436 437 438 439 440 441 442 443 444 445
}

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

446 447 448
	if (!dev)
		return -EINVAL;

449 450 451 452 453 454 455
	mutex_lock(&cpuidle_lock);

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

456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472
	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
473
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
474

475 476 477
	if (dev->registered == 0)
		return;

478 479 480 481
	cpuidle_pause_and_lock();

	cpuidle_disable_device(dev);

482
	cpuidle_remove_sysfs(dev);
483 484 485
	list_del(&dev->device_list);
	per_cpu(cpuidle_devices, dev->cpu) = NULL;

486 487
	cpuidle_coupled_unregister_device(dev);

488 489
	cpuidle_resume_and_unlock();

D
Daniel Lezcano 已提交
490
	module_put(drv->owner);
491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510
}

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)
{
511
	smp_call_function(smp_callback, NULL, 1);
512 513 514 515 516 517 518
	return NOTIFY_OK;
}

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

M
Mark Gross 已提交
519 520 521 522
static inline void latency_notifier_init(struct notifier_block *n)
{
	pm_qos_add_notifier(PM_QOS_CPU_DMA_LATENCY, n);
}
523 524 525 526 527 528 529 530 531 532 533 534 535 536

#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;

537 538 539
	if (cpuidle_disabled())
		return -ENODEV;

540
	ret = cpuidle_add_interface(cpu_subsys.dev_root);
541 542 543 544 545 546 547 548
	if (ret)
		return ret;

	latency_notifier_init(&cpuidle_latency_notifier);

	return 0;
}

549
module_param(off, int, 0444);
550
core_initcall(cpuidle_init);