cpuidle.c 11.6 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
	int i;
73

74 75 76
	if (!drv)
		return -ENODEV;

77
	/* Find lowest-power state that supports long-term idle */
78 79 80
	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);
81 82 83 84

	return -ENODEV;
}

85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102
/**
 * 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.
		 */
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
	}

147 148
	trace_power_start_rcuidle(POWER_CSTATE, next_state, dev->cpu);
	trace_cpu_idle_rcuidle(next_state, dev->cpu);
149

150 151 152 153 154
	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);
155

156 157
	trace_power_end_rcuidle(dev->cpu);
	trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, dev->cpu);
158

159 160
	/* give the governor an opportunity to reflect on the outcome */
	if (cpuidle_curr_governor->reflect)
161
		cpuidle_curr_governor->reflect(dev, entered_state);
L
Len Brown 已提交
162 163

	return 0;
164 165 166 167 168 169 170
}

/**
 * cpuidle_install_idle_handler - installs the cpuidle idle loop handler
 */
void cpuidle_install_idle_handler(void)
{
L
Len Brown 已提交
171
	if (enabled_devices) {
172 173
		/* Make sure all changes finished before we switch to new idle */
		smp_wmb();
L
Len Brown 已提交
174
		initialized = 1;
175 176 177 178 179 180 181 182
	}
}

/**
 * cpuidle_uninstall_idle_handler - uninstalls the cpuidle idle loop handler
 */
void cpuidle_uninstall_idle_handler(void)
{
L
Len Brown 已提交
183 184
	if (enabled_devices) {
		initialized = 0;
185
		kick_all_cpus_sync();
186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210
	}
}

/**
 * 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);

211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226
/* 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);
}

227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257
/**
 * 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;
}

258
#ifdef CONFIG_ARCH_HAS_CPU_RELAX
259 260
static int poll_idle(struct cpuidle_device *dev,
		struct cpuidle_driver *drv, int index)
261 262 263 264 265 266 267 268 269 270 271 272 273 274
{
	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;

275 276 277
	dev->last_residency = (int) diff;

	return index;
278 279
}

280
static void poll_idle_init(struct cpuidle_driver *drv)
281
{
282
	struct cpuidle_state *state = &drv->states[0];
283

284
	snprintf(state->name, CPUIDLE_NAME_LEN, "POLL");
285 286 287 288
	snprintf(state->desc, CPUIDLE_DESC_LEN, "CPUIDLE CORE POLL IDLE");
	state->exit_latency = 0;
	state->target_residency = 0;
	state->power_usage = -1;
289
	state->flags = 0;
290
	state->enter = poll_idle;
291
	state->disabled = false;
292 293
}
#else
294
static void poll_idle_init(struct cpuidle_driver *drv) {}
295 296
#endif /* CONFIG_ARCH_HAS_CPU_RELAX */

297 298 299 300 301 302 303 304 305 306
/**
 * 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
307
	struct cpuidle_driver *drv;
308

309 310 311
	if (!dev)
		return -EINVAL;

312 313
	if (dev->enabled)
		return 0;
D
Daniel Lezcano 已提交
314 315 316

	drv = cpuidle_get_cpu_driver(dev);

317
	if (!drv || !cpuidle_curr_governor)
318
		return -EIO;
D
Daniel Lezcano 已提交
319

320
	if (!dev->state_count)
321
		dev->state_count = drv->state_count;
322

323 324 325 326 327 328
	if (dev->registered == 0) {
		ret = __cpuidle_register_device(dev);
		if (ret)
			return ret;
	}

329 330 331 332
	cpuidle_enter_ops = drv->en_core_tk_irqen ?
		cpuidle_enter_tk : cpuidle_enter;

	poll_idle_init(drv);
333

D
Daniel Lezcano 已提交
334 335
	ret = cpuidle_add_device_sysfs(dev);
	if (ret)
336 337 338
		return ret;

	if (cpuidle_curr_governor->enable &&
339
	    (ret = cpuidle_curr_governor->enable(drv, dev)))
340 341 342
		goto fail_sysfs;

	for (i = 0; i < dev->state_count; i++) {
343 344
		dev->states_usage[i].usage = 0;
		dev->states_usage[i].time = 0;
345 346 347 348 349 350 351 352 353 354 355
	}
	dev->last_residency = 0;

	smp_wmb();

	dev->enabled = 1;

	enabled_devices++;
	return 0;

fail_sysfs:
D
Daniel Lezcano 已提交
356
	cpuidle_remove_device_sysfs(dev);
357 358 359 360 361 362 363 364 365 366 367 368 369 370 371

	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
372 373
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);

374
	if (!dev || !dev->enabled)
375
		return;
D
Daniel Lezcano 已提交
376 377

	if (!drv || !cpuidle_curr_governor)
378 379 380 381 382
		return;

	dev->enabled = 0;

	if (cpuidle_curr_governor->disable)
D
Daniel Lezcano 已提交
383
		cpuidle_curr_governor->disable(drv, dev);
384

D
Daniel Lezcano 已提交
385
	cpuidle_remove_device_sysfs(dev);
386 387 388 389 390 391
	enabled_devices--;
}

EXPORT_SYMBOL_GPL(cpuidle_disable_device);

/**
392 393
 * __cpuidle_register_device - internal register function called before register
 * and enable routines
394
 * @dev: the cpu
395 396
 *
 * cpuidle_lock mutex must be held before this is called
397
 */
398
static int __cpuidle_register_device(struct cpuidle_device *dev)
399 400
{
	int ret;
D
Daniel Lezcano 已提交
401
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
402

D
Daniel Lezcano 已提交
403
	if (!try_module_get(drv->owner))
404 405 406 407
		return -EINVAL;

	per_cpu(cpuidle_devices, dev->cpu) = dev;
	list_add(&dev->device_list, &cpuidle_detected_devices);
408
	ret = cpuidle_add_sysfs(dev);
409 410
	if (ret)
		goto err_sysfs;
411

412 413 414
	ret = cpuidle_coupled_register_device(dev);
	if (ret)
		goto err_coupled;
415

416 417
	dev->registered = 1;
	return 0;
418

419
err_coupled:
420
	cpuidle_remove_sysfs(dev);
421 422 423
err_sysfs:
	list_del(&dev->device_list);
	per_cpu(cpuidle_devices, dev->cpu) = NULL;
D
Daniel Lezcano 已提交
424
	module_put(drv->owner);
425
	return ret;
426 427 428 429 430 431 432 433 434 435
}

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

436 437 438
	if (!dev)
		return -EINVAL;

439 440 441 442 443 444 445
	mutex_lock(&cpuidle_lock);

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

446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462
	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
463
	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
464

465 466 467
	if (dev->registered == 0)
		return;

468 469 470 471
	cpuidle_pause_and_lock();

	cpuidle_disable_device(dev);

472
	cpuidle_remove_sysfs(dev);
473 474 475
	list_del(&dev->device_list);
	per_cpu(cpuidle_devices, dev->cpu) = NULL;

476 477
	cpuidle_coupled_unregister_device(dev);

478 479
	cpuidle_resume_and_unlock();

D
Daniel Lezcano 已提交
480
	module_put(drv->owner);
481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
}

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)
{
501
	smp_call_function(smp_callback, NULL, 1);
502 503 504 505 506 507 508
	return NOTIFY_OK;
}

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

M
Mark Gross 已提交
509 510 511 512
static inline void latency_notifier_init(struct notifier_block *n)
{
	pm_qos_add_notifier(PM_QOS_CPU_DMA_LATENCY, n);
}
513 514 515 516 517 518 519 520 521 522 523 524 525 526

#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;

527 528 529
	if (cpuidle_disabled())
		return -ENODEV;

530
	ret = cpuidle_add_interface(cpu_subsys.dev_root);
531 532 533 534 535 536 537 538
	if (ret)
		return ret;

	latency_notifier_init(&cpuidle_latency_notifier);

	return 0;
}

539
module_param(off, int, 0444);
540
core_initcall(cpuidle_init);