xref: /linux/drivers/thermal/thermal_core.c (revision 5a558f369ef89c6fd8170ee1137274fcc08517ae)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  thermal.c - Generic Thermal Management Sysfs support.
4  *
5  *  Copyright (C) 2008 Intel Corp
6  *  Copyright (C) 2008 Zhang Rui <rui.zhang@intel.com>
7  *  Copyright (C) 2008 Sujith Thomas <sujith.thomas@intel.com>
8  */
9 
10 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
11 
12 #include <linux/device.h>
13 #include <linux/err.h>
14 #include <linux/export.h>
15 #include <linux/slab.h>
16 #include <linux/kdev_t.h>
17 #include <linux/idr.h>
18 #include <linux/list_sort.h>
19 #include <linux/thermal.h>
20 #include <linux/reboot.h>
21 #include <linux/string.h>
22 #include <linux/of.h>
23 #include <linux/suspend.h>
24 
25 #define CREATE_TRACE_POINTS
26 #include "thermal_trace.h"
27 
28 #include "thermal_core.h"
29 #include "thermal_hwmon.h"
30 
31 static DEFINE_IDA(thermal_tz_ida);
32 static DEFINE_IDA(thermal_cdev_ida);
33 
34 static LIST_HEAD(thermal_tz_list);
35 static LIST_HEAD(thermal_cdev_list);
36 static LIST_HEAD(thermal_governor_list);
37 
38 static DEFINE_MUTEX(thermal_list_lock);
39 static DEFINE_MUTEX(thermal_governor_lock);
40 
41 static struct thermal_governor *def_governor;
42 
43 /*
44  * Governor section: set of functions to handle thermal governors
45  *
46  * Functions to help in the life cycle of thermal governors within
47  * the thermal core and by the thermal governor code.
48  */
49 
50 static struct thermal_governor *__find_governor(const char *name)
51 {
52 	struct thermal_governor *pos;
53 
54 	if (!name || !name[0])
55 		return def_governor;
56 
57 	list_for_each_entry(pos, &thermal_governor_list, governor_list)
58 		if (!strncasecmp(name, pos->name, THERMAL_NAME_LENGTH))
59 			return pos;
60 
61 	return NULL;
62 }
63 
64 /**
65  * bind_previous_governor() - bind the previous governor of the thermal zone
66  * @tz:		a valid pointer to a struct thermal_zone_device
67  * @failed_gov_name:	the name of the governor that failed to register
68  *
69  * Register the previous governor of the thermal zone after a new
70  * governor has failed to be bound.
71  */
72 static void bind_previous_governor(struct thermal_zone_device *tz,
73 				   const char *failed_gov_name)
74 {
75 	if (tz->governor && tz->governor->bind_to_tz) {
76 		if (tz->governor->bind_to_tz(tz)) {
77 			dev_err(&tz->device,
78 				"governor %s failed to bind and the previous one (%s) failed to bind again, thermal zone %s has no governor\n",
79 				failed_gov_name, tz->governor->name, tz->type);
80 			tz->governor = NULL;
81 		}
82 	}
83 }
84 
85 /**
86  * thermal_set_governor() - Switch to another governor
87  * @tz:		a valid pointer to a struct thermal_zone_device
88  * @new_gov:	pointer to the new governor
89  *
90  * Change the governor of thermal zone @tz.
91  *
92  * Return: 0 on success, an error if the new governor's bind_to_tz() failed.
93  */
94 static int thermal_set_governor(struct thermal_zone_device *tz,
95 				struct thermal_governor *new_gov)
96 {
97 	int ret = 0;
98 
99 	if (tz->governor && tz->governor->unbind_from_tz)
100 		tz->governor->unbind_from_tz(tz);
101 
102 	if (new_gov && new_gov->bind_to_tz) {
103 		ret = new_gov->bind_to_tz(tz);
104 		if (ret) {
105 			bind_previous_governor(tz, new_gov->name);
106 
107 			return ret;
108 		}
109 	}
110 
111 	tz->governor = new_gov;
112 
113 	return ret;
114 }
115 
116 int thermal_register_governor(struct thermal_governor *governor)
117 {
118 	int err;
119 	const char *name;
120 	struct thermal_zone_device *pos;
121 
122 	if (!governor)
123 		return -EINVAL;
124 
125 	mutex_lock(&thermal_governor_lock);
126 
127 	err = -EBUSY;
128 	if (!__find_governor(governor->name)) {
129 		bool match_default;
130 
131 		err = 0;
132 		list_add(&governor->governor_list, &thermal_governor_list);
133 		match_default = !strncmp(governor->name,
134 					 DEFAULT_THERMAL_GOVERNOR,
135 					 THERMAL_NAME_LENGTH);
136 
137 		if (!def_governor && match_default)
138 			def_governor = governor;
139 	}
140 
141 	mutex_lock(&thermal_list_lock);
142 
143 	list_for_each_entry(pos, &thermal_tz_list, node) {
144 		/*
145 		 * only thermal zones with specified tz->tzp->governor_name
146 		 * may run with tz->govenor unset
147 		 */
148 		if (pos->governor)
149 			continue;
150 
151 		name = pos->tzp->governor_name;
152 
153 		if (!strncasecmp(name, governor->name, THERMAL_NAME_LENGTH)) {
154 			int ret;
155 
156 			ret = thermal_set_governor(pos, governor);
157 			if (ret)
158 				dev_err(&pos->device,
159 					"Failed to set governor %s for thermal zone %s: %d\n",
160 					governor->name, pos->type, ret);
161 		}
162 	}
163 
164 	mutex_unlock(&thermal_list_lock);
165 	mutex_unlock(&thermal_governor_lock);
166 
167 	return err;
168 }
169 
170 void thermal_unregister_governor(struct thermal_governor *governor)
171 {
172 	struct thermal_zone_device *pos;
173 
174 	if (!governor)
175 		return;
176 
177 	mutex_lock(&thermal_governor_lock);
178 
179 	if (!__find_governor(governor->name))
180 		goto exit;
181 
182 	mutex_lock(&thermal_list_lock);
183 
184 	list_for_each_entry(pos, &thermal_tz_list, node) {
185 		if (!strncasecmp(pos->governor->name, governor->name,
186 				 THERMAL_NAME_LENGTH))
187 			thermal_set_governor(pos, NULL);
188 	}
189 
190 	mutex_unlock(&thermal_list_lock);
191 	list_del(&governor->governor_list);
192 exit:
193 	mutex_unlock(&thermal_governor_lock);
194 }
195 
196 int thermal_zone_device_set_policy(struct thermal_zone_device *tz,
197 				   char *policy)
198 {
199 	struct thermal_governor *gov;
200 	int ret = -EINVAL;
201 
202 	mutex_lock(&thermal_governor_lock);
203 	mutex_lock(&tz->lock);
204 
205 	gov = __find_governor(strim(policy));
206 	if (!gov)
207 		goto exit;
208 
209 	ret = thermal_set_governor(tz, gov);
210 
211 exit:
212 	mutex_unlock(&tz->lock);
213 	mutex_unlock(&thermal_governor_lock);
214 
215 	thermal_notify_tz_gov_change(tz, policy);
216 
217 	return ret;
218 }
219 
220 int thermal_build_list_of_policies(char *buf)
221 {
222 	struct thermal_governor *pos;
223 	ssize_t count = 0;
224 
225 	mutex_lock(&thermal_governor_lock);
226 
227 	list_for_each_entry(pos, &thermal_governor_list, governor_list) {
228 		count += sysfs_emit_at(buf, count, "%s ", pos->name);
229 	}
230 	count += sysfs_emit_at(buf, count, "\n");
231 
232 	mutex_unlock(&thermal_governor_lock);
233 
234 	return count;
235 }
236 
237 static void __init thermal_unregister_governors(void)
238 {
239 	struct thermal_governor **governor;
240 
241 	for_each_governor_table(governor)
242 		thermal_unregister_governor(*governor);
243 }
244 
245 static int __init thermal_register_governors(void)
246 {
247 	int ret = 0;
248 	struct thermal_governor **governor;
249 
250 	for_each_governor_table(governor) {
251 		ret = thermal_register_governor(*governor);
252 		if (ret) {
253 			pr_err("Failed to register governor: '%s'",
254 			       (*governor)->name);
255 			break;
256 		}
257 
258 		pr_info("Registered thermal governor '%s'",
259 			(*governor)->name);
260 	}
261 
262 	if (ret) {
263 		struct thermal_governor **gov;
264 
265 		for_each_governor_table(gov) {
266 			if (gov == governor)
267 				break;
268 			thermal_unregister_governor(*gov);
269 		}
270 	}
271 
272 	return ret;
273 }
274 
275 /*
276  * Zone update section: main control loop applied to each zone while monitoring
277  * in polling mode. The monitoring is done using a workqueue.
278  * Same update may be done on a zone by calling thermal_zone_device_update().
279  *
280  * An update means:
281  * - Non-critical trips will invoke the governor responsible for that zone;
282  * - Hot trips will produce a notification to userspace;
283  * - Critical trip point will cause a system shutdown.
284  */
285 static void thermal_zone_device_set_polling(struct thermal_zone_device *tz,
286 					    unsigned long delay)
287 {
288 	if (delay)
289 		mod_delayed_work(system_freezable_power_efficient_wq,
290 				 &tz->poll_queue, delay);
291 	else
292 		cancel_delayed_work(&tz->poll_queue);
293 }
294 
295 static void monitor_thermal_zone(struct thermal_zone_device *tz)
296 {
297 	if (tz->mode != THERMAL_DEVICE_ENABLED)
298 		thermal_zone_device_set_polling(tz, 0);
299 	else if (tz->passive > 0)
300 		thermal_zone_device_set_polling(tz, tz->passive_delay_jiffies);
301 	else if (tz->polling_delay_jiffies)
302 		thermal_zone_device_set_polling(tz, tz->polling_delay_jiffies);
303 }
304 
305 static struct thermal_governor *thermal_get_tz_governor(struct thermal_zone_device *tz)
306 {
307 	if (tz->governor)
308 		return tz->governor;
309 
310 	return def_governor;
311 }
312 
313 void thermal_governor_update_tz(struct thermal_zone_device *tz,
314 				enum thermal_notify_event reason)
315 {
316 	if (!tz->governor || !tz->governor->update_tz)
317 		return;
318 
319 	tz->governor->update_tz(tz, reason);
320 }
321 
322 static void thermal_zone_device_halt(struct thermal_zone_device *tz, bool shutdown)
323 {
324 	/*
325 	 * poweroff_delay_ms must be a carefully profiled positive value.
326 	 * Its a must for forced_emergency_poweroff_work to be scheduled.
327 	 */
328 	int poweroff_delay_ms = CONFIG_THERMAL_EMERGENCY_POWEROFF_DELAY_MS;
329 	const char *msg = "Temperature too high";
330 
331 	dev_emerg(&tz->device, "%s: critical temperature reached\n", tz->type);
332 
333 	if (shutdown)
334 		hw_protection_shutdown(msg, poweroff_delay_ms);
335 	else
336 		hw_protection_reboot(msg, poweroff_delay_ms);
337 }
338 
339 void thermal_zone_device_critical(struct thermal_zone_device *tz)
340 {
341 	thermal_zone_device_halt(tz, true);
342 }
343 EXPORT_SYMBOL(thermal_zone_device_critical);
344 
345 void thermal_zone_device_critical_reboot(struct thermal_zone_device *tz)
346 {
347 	thermal_zone_device_halt(tz, false);
348 }
349 
350 static void handle_critical_trips(struct thermal_zone_device *tz,
351 				  const struct thermal_trip *trip)
352 {
353 	trace_thermal_zone_trip(tz, thermal_zone_trip_id(tz, trip), trip->type);
354 
355 	if (trip->type == THERMAL_TRIP_CRITICAL)
356 		tz->ops.critical(tz);
357 	else if (tz->ops.hot)
358 		tz->ops.hot(tz);
359 }
360 
361 static void handle_thermal_trip(struct thermal_zone_device *tz,
362 				struct thermal_trip_desc *td,
363 				struct list_head *way_up_list,
364 				struct list_head *way_down_list)
365 {
366 	const struct thermal_trip *trip = &td->trip;
367 	int old_threshold;
368 
369 	if (trip->temperature == THERMAL_TEMP_INVALID)
370 		return;
371 
372 	/*
373 	 * If the trip temperature or hysteresis has been updated recently,
374 	 * the threshold needs to be computed again using the new values.
375 	 * However, its initial value still reflects the old ones and that
376 	 * is what needs to be compared with the previous zone temperature
377 	 * to decide which action to take.
378 	 */
379 	old_threshold = td->threshold;
380 	td->threshold = trip->temperature;
381 
382 	if (tz->last_temperature >= old_threshold &&
383 	    tz->last_temperature != THERMAL_TEMP_INVALID) {
384 		/*
385 		 * Mitigation is under way, so it needs to stop if the zone
386 		 * temperature falls below the low temperature of the trip.
387 		 * In that case, the trip temperature becomes the new threshold.
388 		 */
389 		if (tz->temperature < trip->temperature - trip->hysteresis) {
390 			list_add(&td->notify_list_node, way_down_list);
391 			td->notify_temp = trip->temperature - trip->hysteresis;
392 
393 			if (trip->type == THERMAL_TRIP_PASSIVE) {
394 				tz->passive--;
395 				WARN_ON(tz->passive < 0);
396 			}
397 		} else {
398 			td->threshold -= trip->hysteresis;
399 		}
400 	} else if (tz->temperature >= trip->temperature) {
401 		/*
402 		 * There is no mitigation under way, so it needs to be started
403 		 * if the zone temperature exceeds the trip one.  The new
404 		 * threshold is then set to the low temperature of the trip.
405 		 */
406 		list_add_tail(&td->notify_list_node, way_up_list);
407 		td->notify_temp = trip->temperature;
408 		td->threshold -= trip->hysteresis;
409 
410 		if (trip->type == THERMAL_TRIP_PASSIVE)
411 			tz->passive++;
412 		else if (trip->type == THERMAL_TRIP_CRITICAL ||
413 			 trip->type == THERMAL_TRIP_HOT)
414 			handle_critical_trips(tz, trip);
415 	}
416 }
417 
418 static void update_temperature(struct thermal_zone_device *tz)
419 {
420 	int temp, ret;
421 
422 	ret = __thermal_zone_get_temp(tz, &temp);
423 	if (ret) {
424 		if (ret != -EAGAIN)
425 			dev_warn(&tz->device,
426 				 "failed to read out thermal zone (%d)\n",
427 				 ret);
428 		return;
429 	}
430 
431 	tz->last_temperature = tz->temperature;
432 	tz->temperature = temp;
433 
434 	trace_thermal_temperature(tz);
435 
436 	thermal_genl_sampling_temp(tz->id, temp);
437 }
438 
439 static void thermal_zone_device_check(struct work_struct *work)
440 {
441 	struct thermal_zone_device *tz = container_of(work, struct
442 						      thermal_zone_device,
443 						      poll_queue.work);
444 	thermal_zone_device_update(tz, THERMAL_EVENT_UNSPECIFIED);
445 }
446 
447 static void thermal_zone_device_init(struct thermal_zone_device *tz)
448 {
449 	struct thermal_instance *pos;
450 
451 	INIT_DELAYED_WORK(&tz->poll_queue, thermal_zone_device_check);
452 
453 	tz->temperature = THERMAL_TEMP_INVALID;
454 	tz->passive = 0;
455 	tz->prev_low_trip = -INT_MAX;
456 	tz->prev_high_trip = INT_MAX;
457 	list_for_each_entry(pos, &tz->thermal_instances, tz_node)
458 		pos->initialized = false;
459 }
460 
461 static void thermal_governor_trip_crossed(struct thermal_governor *governor,
462 					  struct thermal_zone_device *tz,
463 					  const struct thermal_trip *trip,
464 					  bool crossed_up)
465 {
466 	if (governor->trip_crossed)
467 		governor->trip_crossed(tz, trip, crossed_up);
468 }
469 
470 static void thermal_trip_crossed(struct thermal_zone_device *tz,
471 				 const struct thermal_trip *trip,
472 				 struct thermal_governor *governor,
473 				 bool crossed_up)
474 {
475 	if (crossed_up) {
476 		thermal_notify_tz_trip_up(tz, trip);
477 		thermal_debug_tz_trip_up(tz, trip);
478 	} else {
479 		thermal_notify_tz_trip_down(tz, trip);
480 		thermal_debug_tz_trip_down(tz, trip);
481 	}
482 	thermal_governor_trip_crossed(governor, tz, trip, crossed_up);
483 }
484 
485 static int thermal_trip_notify_cmp(void *ascending, const struct list_head *a,
486 				   const struct list_head *b)
487 {
488 	struct thermal_trip_desc *tda = container_of(a, struct thermal_trip_desc,
489 						     notify_list_node);
490 	struct thermal_trip_desc *tdb = container_of(b, struct thermal_trip_desc,
491 						     notify_list_node);
492 	int ret = tdb->notify_temp - tda->notify_temp;
493 
494 	return ascending ? ret : -ret;
495 }
496 
497 void __thermal_zone_device_update(struct thermal_zone_device *tz,
498 				  enum thermal_notify_event event)
499 {
500 	struct thermal_governor *governor = thermal_get_tz_governor(tz);
501 	struct thermal_trip_desc *td;
502 	LIST_HEAD(way_down_list);
503 	LIST_HEAD(way_up_list);
504 
505 	if (tz->suspended)
506 		return;
507 
508 	if (!thermal_zone_device_is_enabled(tz))
509 		return;
510 
511 	update_temperature(tz);
512 
513 	if (tz->temperature == THERMAL_TEMP_INVALID)
514 		return;
515 
516 	__thermal_zone_set_trips(tz);
517 
518 	tz->notify_event = event;
519 
520 	for_each_trip_desc(tz, td)
521 		handle_thermal_trip(tz, td, &way_up_list, &way_down_list);
522 
523 	list_sort(&way_up_list, &way_up_list, thermal_trip_notify_cmp);
524 	list_for_each_entry(td, &way_up_list, notify_list_node)
525 		thermal_trip_crossed(tz, &td->trip, governor, true);
526 
527 	list_sort(NULL, &way_down_list, thermal_trip_notify_cmp);
528 	list_for_each_entry(td, &way_down_list, notify_list_node)
529 		thermal_trip_crossed(tz, &td->trip, governor, false);
530 
531 	if (governor->manage)
532 		governor->manage(tz);
533 
534 	thermal_debug_update_trip_stats(tz);
535 
536 	monitor_thermal_zone(tz);
537 }
538 
539 static int thermal_zone_device_set_mode(struct thermal_zone_device *tz,
540 					enum thermal_device_mode mode)
541 {
542 	int ret = 0;
543 
544 	mutex_lock(&tz->lock);
545 
546 	/* do nothing if mode isn't changing */
547 	if (mode == tz->mode) {
548 		mutex_unlock(&tz->lock);
549 
550 		return ret;
551 	}
552 
553 	if (tz->ops.change_mode)
554 		ret = tz->ops.change_mode(tz, mode);
555 
556 	if (!ret)
557 		tz->mode = mode;
558 
559 	__thermal_zone_device_update(tz, THERMAL_EVENT_UNSPECIFIED);
560 
561 	mutex_unlock(&tz->lock);
562 
563 	if (mode == THERMAL_DEVICE_ENABLED)
564 		thermal_notify_tz_enable(tz);
565 	else
566 		thermal_notify_tz_disable(tz);
567 
568 	return ret;
569 }
570 
571 int thermal_zone_device_enable(struct thermal_zone_device *tz)
572 {
573 	return thermal_zone_device_set_mode(tz, THERMAL_DEVICE_ENABLED);
574 }
575 EXPORT_SYMBOL_GPL(thermal_zone_device_enable);
576 
577 int thermal_zone_device_disable(struct thermal_zone_device *tz)
578 {
579 	return thermal_zone_device_set_mode(tz, THERMAL_DEVICE_DISABLED);
580 }
581 EXPORT_SYMBOL_GPL(thermal_zone_device_disable);
582 
583 int thermal_zone_device_is_enabled(struct thermal_zone_device *tz)
584 {
585 	lockdep_assert_held(&tz->lock);
586 
587 	return tz->mode == THERMAL_DEVICE_ENABLED;
588 }
589 
590 static bool thermal_zone_is_present(struct thermal_zone_device *tz)
591 {
592 	return !list_empty(&tz->node);
593 }
594 
595 void thermal_zone_device_update(struct thermal_zone_device *tz,
596 				enum thermal_notify_event event)
597 {
598 	mutex_lock(&tz->lock);
599 	if (thermal_zone_is_present(tz))
600 		__thermal_zone_device_update(tz, event);
601 	mutex_unlock(&tz->lock);
602 }
603 EXPORT_SYMBOL_GPL(thermal_zone_device_update);
604 
605 void thermal_zone_trip_down(struct thermal_zone_device *tz,
606 			    const struct thermal_trip *trip)
607 {
608 	thermal_trip_crossed(tz, trip, thermal_get_tz_governor(tz), false);
609 }
610 
611 int for_each_thermal_governor(int (*cb)(struct thermal_governor *, void *),
612 			      void *data)
613 {
614 	struct thermal_governor *gov;
615 	int ret = 0;
616 
617 	mutex_lock(&thermal_governor_lock);
618 	list_for_each_entry(gov, &thermal_governor_list, governor_list) {
619 		ret = cb(gov, data);
620 		if (ret)
621 			break;
622 	}
623 	mutex_unlock(&thermal_governor_lock);
624 
625 	return ret;
626 }
627 
628 int for_each_thermal_cooling_device(int (*cb)(struct thermal_cooling_device *,
629 					      void *), void *data)
630 {
631 	struct thermal_cooling_device *cdev;
632 	int ret = 0;
633 
634 	mutex_lock(&thermal_list_lock);
635 	list_for_each_entry(cdev, &thermal_cdev_list, node) {
636 		ret = cb(cdev, data);
637 		if (ret)
638 			break;
639 	}
640 	mutex_unlock(&thermal_list_lock);
641 
642 	return ret;
643 }
644 
645 int for_each_thermal_zone(int (*cb)(struct thermal_zone_device *, void *),
646 			  void *data)
647 {
648 	struct thermal_zone_device *tz;
649 	int ret = 0;
650 
651 	mutex_lock(&thermal_list_lock);
652 	list_for_each_entry(tz, &thermal_tz_list, node) {
653 		ret = cb(tz, data);
654 		if (ret)
655 			break;
656 	}
657 	mutex_unlock(&thermal_list_lock);
658 
659 	return ret;
660 }
661 
662 struct thermal_zone_device *thermal_zone_get_by_id(int id)
663 {
664 	struct thermal_zone_device *tz, *match = NULL;
665 
666 	mutex_lock(&thermal_list_lock);
667 	list_for_each_entry(tz, &thermal_tz_list, node) {
668 		if (tz->id == id) {
669 			match = tz;
670 			break;
671 		}
672 	}
673 	mutex_unlock(&thermal_list_lock);
674 
675 	return match;
676 }
677 
678 /*
679  * Device management section: cooling devices, zones devices, and binding
680  *
681  * Set of functions provided by the thermal core for:
682  * - cooling devices lifecycle: registration, unregistration,
683  *				binding, and unbinding.
684  * - thermal zone devices lifecycle: registration, unregistration,
685  *				     binding, and unbinding.
686  */
687 
688 /**
689  * thermal_bind_cdev_to_trip - bind a cooling device to a thermal zone
690  * @tz:		pointer to struct thermal_zone_device
691  * @trip:	trip point the cooling devices is associated with in this zone.
692  * @cdev:	pointer to struct thermal_cooling_device
693  * @upper:	the Maximum cooling state for this trip point.
694  *		THERMAL_NO_LIMIT means no upper limit,
695  *		and the cooling device can be in max_state.
696  * @lower:	the Minimum cooling state can be used for this trip point.
697  *		THERMAL_NO_LIMIT means no lower limit,
698  *		and the cooling device can be in cooling state 0.
699  * @weight:	The weight of the cooling device to be bound to the
700  *		thermal zone. Use THERMAL_WEIGHT_DEFAULT for the
701  *		default value
702  *
703  * This interface function bind a thermal cooling device to the certain trip
704  * point of a thermal zone device.
705  * This function is usually called in the thermal zone device .bind callback.
706  *
707  * Return: 0 on success, the proper error value otherwise.
708  */
709 int thermal_bind_cdev_to_trip(struct thermal_zone_device *tz,
710 				     const struct thermal_trip *trip,
711 				     struct thermal_cooling_device *cdev,
712 				     unsigned long upper, unsigned long lower,
713 				     unsigned int weight)
714 {
715 	struct thermal_instance *dev;
716 	struct thermal_instance *pos;
717 	struct thermal_zone_device *pos1;
718 	struct thermal_cooling_device *pos2;
719 	bool upper_no_limit;
720 	int result;
721 
722 	list_for_each_entry(pos1, &thermal_tz_list, node) {
723 		if (pos1 == tz)
724 			break;
725 	}
726 	list_for_each_entry(pos2, &thermal_cdev_list, node) {
727 		if (pos2 == cdev)
728 			break;
729 	}
730 
731 	if (tz != pos1 || cdev != pos2)
732 		return -EINVAL;
733 
734 	/* lower default 0, upper default max_state */
735 	lower = lower == THERMAL_NO_LIMIT ? 0 : lower;
736 
737 	if (upper == THERMAL_NO_LIMIT) {
738 		upper = cdev->max_state;
739 		upper_no_limit = true;
740 	} else {
741 		upper_no_limit = false;
742 	}
743 
744 	if (lower > upper || upper > cdev->max_state)
745 		return -EINVAL;
746 
747 	dev = kzalloc(sizeof(*dev), GFP_KERNEL);
748 	if (!dev)
749 		return -ENOMEM;
750 	dev->tz = tz;
751 	dev->cdev = cdev;
752 	dev->trip = trip;
753 	dev->upper = upper;
754 	dev->upper_no_limit = upper_no_limit;
755 	dev->lower = lower;
756 	dev->target = THERMAL_NO_TARGET;
757 	dev->weight = weight;
758 
759 	result = ida_alloc(&tz->ida, GFP_KERNEL);
760 	if (result < 0)
761 		goto free_mem;
762 
763 	dev->id = result;
764 	sprintf(dev->name, "cdev%d", dev->id);
765 	result =
766 	    sysfs_create_link(&tz->device.kobj, &cdev->device.kobj, dev->name);
767 	if (result)
768 		goto release_ida;
769 
770 	snprintf(dev->attr_name, sizeof(dev->attr_name), "cdev%d_trip_point",
771 		 dev->id);
772 	sysfs_attr_init(&dev->attr.attr);
773 	dev->attr.attr.name = dev->attr_name;
774 	dev->attr.attr.mode = 0444;
775 	dev->attr.show = trip_point_show;
776 	result = device_create_file(&tz->device, &dev->attr);
777 	if (result)
778 		goto remove_symbol_link;
779 
780 	snprintf(dev->weight_attr_name, sizeof(dev->weight_attr_name),
781 		 "cdev%d_weight", dev->id);
782 	sysfs_attr_init(&dev->weight_attr.attr);
783 	dev->weight_attr.attr.name = dev->weight_attr_name;
784 	dev->weight_attr.attr.mode = S_IWUSR | S_IRUGO;
785 	dev->weight_attr.show = weight_show;
786 	dev->weight_attr.store = weight_store;
787 	result = device_create_file(&tz->device, &dev->weight_attr);
788 	if (result)
789 		goto remove_trip_file;
790 
791 	mutex_lock(&tz->lock);
792 	mutex_lock(&cdev->lock);
793 	list_for_each_entry(pos, &tz->thermal_instances, tz_node)
794 		if (pos->tz == tz && pos->trip == trip && pos->cdev == cdev) {
795 			result = -EEXIST;
796 			break;
797 		}
798 	if (!result) {
799 		list_add_tail(&dev->tz_node, &tz->thermal_instances);
800 		list_add_tail(&dev->cdev_node, &cdev->thermal_instances);
801 		atomic_set(&tz->need_update, 1);
802 
803 		thermal_governor_update_tz(tz, THERMAL_TZ_BIND_CDEV);
804 	}
805 	mutex_unlock(&cdev->lock);
806 	mutex_unlock(&tz->lock);
807 
808 	if (!result)
809 		return 0;
810 
811 	device_remove_file(&tz->device, &dev->weight_attr);
812 remove_trip_file:
813 	device_remove_file(&tz->device, &dev->attr);
814 remove_symbol_link:
815 	sysfs_remove_link(&tz->device.kobj, dev->name);
816 release_ida:
817 	ida_free(&tz->ida, dev->id);
818 free_mem:
819 	kfree(dev);
820 	return result;
821 }
822 EXPORT_SYMBOL_GPL(thermal_bind_cdev_to_trip);
823 
824 int thermal_zone_bind_cooling_device(struct thermal_zone_device *tz,
825 				     int trip_index,
826 				     struct thermal_cooling_device *cdev,
827 				     unsigned long upper, unsigned long lower,
828 				     unsigned int weight)
829 {
830 	if (trip_index < 0 || trip_index >= tz->num_trips)
831 		return -EINVAL;
832 
833 	return thermal_bind_cdev_to_trip(tz, &tz->trips[trip_index].trip, cdev,
834 					 upper, lower, weight);
835 }
836 EXPORT_SYMBOL_GPL(thermal_zone_bind_cooling_device);
837 
838 /**
839  * thermal_unbind_cdev_from_trip - unbind a cooling device from a thermal zone.
840  * @tz:		pointer to a struct thermal_zone_device.
841  * @trip:	trip point the cooling devices is associated with in this zone.
842  * @cdev:	pointer to a struct thermal_cooling_device.
843  *
844  * This interface function unbind a thermal cooling device from the certain
845  * trip point of a thermal zone device.
846  * This function is usually called in the thermal zone device .unbind callback.
847  *
848  * Return: 0 on success, the proper error value otherwise.
849  */
850 int thermal_unbind_cdev_from_trip(struct thermal_zone_device *tz,
851 				  const struct thermal_trip *trip,
852 				  struct thermal_cooling_device *cdev)
853 {
854 	struct thermal_instance *pos, *next;
855 
856 	mutex_lock(&tz->lock);
857 	mutex_lock(&cdev->lock);
858 	list_for_each_entry_safe(pos, next, &tz->thermal_instances, tz_node) {
859 		if (pos->tz == tz && pos->trip == trip && pos->cdev == cdev) {
860 			list_del(&pos->tz_node);
861 			list_del(&pos->cdev_node);
862 
863 			thermal_governor_update_tz(tz, THERMAL_TZ_UNBIND_CDEV);
864 
865 			mutex_unlock(&cdev->lock);
866 			mutex_unlock(&tz->lock);
867 			goto unbind;
868 		}
869 	}
870 	mutex_unlock(&cdev->lock);
871 	mutex_unlock(&tz->lock);
872 
873 	return -ENODEV;
874 
875 unbind:
876 	device_remove_file(&tz->device, &pos->weight_attr);
877 	device_remove_file(&tz->device, &pos->attr);
878 	sysfs_remove_link(&tz->device.kobj, pos->name);
879 	ida_free(&tz->ida, pos->id);
880 	kfree(pos);
881 	return 0;
882 }
883 EXPORT_SYMBOL_GPL(thermal_unbind_cdev_from_trip);
884 
885 int thermal_zone_unbind_cooling_device(struct thermal_zone_device *tz,
886 				       int trip_index,
887 				       struct thermal_cooling_device *cdev)
888 {
889 	if (trip_index < 0 || trip_index >= tz->num_trips)
890 		return -EINVAL;
891 
892 	return thermal_unbind_cdev_from_trip(tz, &tz->trips[trip_index].trip, cdev);
893 }
894 EXPORT_SYMBOL_GPL(thermal_zone_unbind_cooling_device);
895 
896 static void thermal_release(struct device *dev)
897 {
898 	struct thermal_zone_device *tz;
899 	struct thermal_cooling_device *cdev;
900 
901 	if (!strncmp(dev_name(dev), "thermal_zone",
902 		     sizeof("thermal_zone") - 1)) {
903 		tz = to_thermal_zone(dev);
904 		thermal_zone_destroy_device_groups(tz);
905 		mutex_destroy(&tz->lock);
906 		complete(&tz->removal);
907 	} else if (!strncmp(dev_name(dev), "cooling_device",
908 			    sizeof("cooling_device") - 1)) {
909 		cdev = to_cooling_device(dev);
910 		thermal_cooling_device_destroy_sysfs(cdev);
911 		kfree_const(cdev->type);
912 		ida_free(&thermal_cdev_ida, cdev->id);
913 		kfree(cdev);
914 	}
915 }
916 
917 static struct class *thermal_class;
918 
919 static inline
920 void print_bind_err_msg(struct thermal_zone_device *tz,
921 			struct thermal_cooling_device *cdev, int ret)
922 {
923 	dev_err(&tz->device, "binding zone %s with cdev %s failed:%d\n",
924 		tz->type, cdev->type, ret);
925 }
926 
927 static void bind_cdev(struct thermal_cooling_device *cdev)
928 {
929 	int ret;
930 	struct thermal_zone_device *pos = NULL;
931 
932 	list_for_each_entry(pos, &thermal_tz_list, node) {
933 		if (pos->ops.bind) {
934 			ret = pos->ops.bind(pos, cdev);
935 			if (ret)
936 				print_bind_err_msg(pos, cdev, ret);
937 		}
938 	}
939 }
940 
941 /**
942  * __thermal_cooling_device_register() - register a new thermal cooling device
943  * @np:		a pointer to a device tree node.
944  * @type:	the thermal cooling device type.
945  * @devdata:	device private data.
946  * @ops:		standard thermal cooling devices callbacks.
947  *
948  * This interface function adds a new thermal cooling device (fan/processor/...)
949  * to /sys/class/thermal/ folder as cooling_device[0-*]. It tries to bind itself
950  * to all the thermal zone devices registered at the same time.
951  * It also gives the opportunity to link the cooling device to a device tree
952  * node, so that it can be bound to a thermal zone created out of device tree.
953  *
954  * Return: a pointer to the created struct thermal_cooling_device or an
955  * ERR_PTR. Caller must check return value with IS_ERR*() helpers.
956  */
957 static struct thermal_cooling_device *
958 __thermal_cooling_device_register(struct device_node *np,
959 				  const char *type, void *devdata,
960 				  const struct thermal_cooling_device_ops *ops)
961 {
962 	struct thermal_cooling_device *cdev;
963 	struct thermal_zone_device *pos = NULL;
964 	unsigned long current_state;
965 	int id, ret;
966 
967 	if (!ops || !ops->get_max_state || !ops->get_cur_state ||
968 	    !ops->set_cur_state)
969 		return ERR_PTR(-EINVAL);
970 
971 	if (!thermal_class)
972 		return ERR_PTR(-ENODEV);
973 
974 	cdev = kzalloc(sizeof(*cdev), GFP_KERNEL);
975 	if (!cdev)
976 		return ERR_PTR(-ENOMEM);
977 
978 	ret = ida_alloc(&thermal_cdev_ida, GFP_KERNEL);
979 	if (ret < 0)
980 		goto out_kfree_cdev;
981 	cdev->id = ret;
982 	id = ret;
983 
984 	cdev->type = kstrdup_const(type ? type : "", GFP_KERNEL);
985 	if (!cdev->type) {
986 		ret = -ENOMEM;
987 		goto out_ida_remove;
988 	}
989 
990 	mutex_init(&cdev->lock);
991 	INIT_LIST_HEAD(&cdev->thermal_instances);
992 	cdev->np = np;
993 	cdev->ops = ops;
994 	cdev->updated = false;
995 	cdev->device.class = thermal_class;
996 	cdev->devdata = devdata;
997 
998 	ret = cdev->ops->get_max_state(cdev, &cdev->max_state);
999 	if (ret)
1000 		goto out_cdev_type;
1001 
1002 	ret = cdev->ops->get_cur_state(cdev, &current_state);
1003 	if (ret)
1004 		goto out_cdev_type;
1005 
1006 	thermal_cooling_device_setup_sysfs(cdev);
1007 
1008 	ret = dev_set_name(&cdev->device, "cooling_device%d", cdev->id);
1009 	if (ret)
1010 		goto out_cooling_dev;
1011 
1012 	ret = device_register(&cdev->device);
1013 	if (ret) {
1014 		/* thermal_release() handles rest of the cleanup */
1015 		put_device(&cdev->device);
1016 		return ERR_PTR(ret);
1017 	}
1018 
1019 	thermal_debug_cdev_add(cdev, current_state);
1020 
1021 	/* Add 'this' new cdev to the global cdev list */
1022 	mutex_lock(&thermal_list_lock);
1023 
1024 	list_add(&cdev->node, &thermal_cdev_list);
1025 
1026 	/* Update binding information for 'this' new cdev */
1027 	bind_cdev(cdev);
1028 
1029 	list_for_each_entry(pos, &thermal_tz_list, node)
1030 		if (atomic_cmpxchg(&pos->need_update, 1, 0))
1031 			thermal_zone_device_update(pos,
1032 						   THERMAL_EVENT_UNSPECIFIED);
1033 
1034 	mutex_unlock(&thermal_list_lock);
1035 
1036 	return cdev;
1037 
1038 out_cooling_dev:
1039 	thermal_cooling_device_destroy_sysfs(cdev);
1040 out_cdev_type:
1041 	kfree_const(cdev->type);
1042 out_ida_remove:
1043 	ida_free(&thermal_cdev_ida, id);
1044 out_kfree_cdev:
1045 	kfree(cdev);
1046 	return ERR_PTR(ret);
1047 }
1048 
1049 /**
1050  * thermal_cooling_device_register() - register a new thermal cooling device
1051  * @type:	the thermal cooling device type.
1052  * @devdata:	device private data.
1053  * @ops:		standard thermal cooling devices callbacks.
1054  *
1055  * This interface function adds a new thermal cooling device (fan/processor/...)
1056  * to /sys/class/thermal/ folder as cooling_device[0-*]. It tries to bind itself
1057  * to all the thermal zone devices registered at the same time.
1058  *
1059  * Return: a pointer to the created struct thermal_cooling_device or an
1060  * ERR_PTR. Caller must check return value with IS_ERR*() helpers.
1061  */
1062 struct thermal_cooling_device *
1063 thermal_cooling_device_register(const char *type, void *devdata,
1064 				const struct thermal_cooling_device_ops *ops)
1065 {
1066 	return __thermal_cooling_device_register(NULL, type, devdata, ops);
1067 }
1068 EXPORT_SYMBOL_GPL(thermal_cooling_device_register);
1069 
1070 /**
1071  * thermal_of_cooling_device_register() - register an OF thermal cooling device
1072  * @np:		a pointer to a device tree node.
1073  * @type:	the thermal cooling device type.
1074  * @devdata:	device private data.
1075  * @ops:		standard thermal cooling devices callbacks.
1076  *
1077  * This function will register a cooling device with device tree node reference.
1078  * This interface function adds a new thermal cooling device (fan/processor/...)
1079  * to /sys/class/thermal/ folder as cooling_device[0-*]. It tries to bind itself
1080  * to all the thermal zone devices registered at the same time.
1081  *
1082  * Return: a pointer to the created struct thermal_cooling_device or an
1083  * ERR_PTR. Caller must check return value with IS_ERR*() helpers.
1084  */
1085 struct thermal_cooling_device *
1086 thermal_of_cooling_device_register(struct device_node *np,
1087 				   const char *type, void *devdata,
1088 				   const struct thermal_cooling_device_ops *ops)
1089 {
1090 	return __thermal_cooling_device_register(np, type, devdata, ops);
1091 }
1092 EXPORT_SYMBOL_GPL(thermal_of_cooling_device_register);
1093 
1094 static void thermal_cooling_device_release(struct device *dev, void *res)
1095 {
1096 	thermal_cooling_device_unregister(
1097 				*(struct thermal_cooling_device **)res);
1098 }
1099 
1100 /**
1101  * devm_thermal_of_cooling_device_register() - register an OF thermal cooling
1102  *					       device
1103  * @dev:	a valid struct device pointer of a sensor device.
1104  * @np:		a pointer to a device tree node.
1105  * @type:	the thermal cooling device type.
1106  * @devdata:	device private data.
1107  * @ops:	standard thermal cooling devices callbacks.
1108  *
1109  * This function will register a cooling device with device tree node reference.
1110  * This interface function adds a new thermal cooling device (fan/processor/...)
1111  * to /sys/class/thermal/ folder as cooling_device[0-*]. It tries to bind itself
1112  * to all the thermal zone devices registered at the same time.
1113  *
1114  * Return: a pointer to the created struct thermal_cooling_device or an
1115  * ERR_PTR. Caller must check return value with IS_ERR*() helpers.
1116  */
1117 struct thermal_cooling_device *
1118 devm_thermal_of_cooling_device_register(struct device *dev,
1119 				struct device_node *np,
1120 				char *type, void *devdata,
1121 				const struct thermal_cooling_device_ops *ops)
1122 {
1123 	struct thermal_cooling_device **ptr, *tcd;
1124 
1125 	ptr = devres_alloc(thermal_cooling_device_release, sizeof(*ptr),
1126 			   GFP_KERNEL);
1127 	if (!ptr)
1128 		return ERR_PTR(-ENOMEM);
1129 
1130 	tcd = __thermal_cooling_device_register(np, type, devdata, ops);
1131 	if (IS_ERR(tcd)) {
1132 		devres_free(ptr);
1133 		return tcd;
1134 	}
1135 
1136 	*ptr = tcd;
1137 	devres_add(dev, ptr);
1138 
1139 	return tcd;
1140 }
1141 EXPORT_SYMBOL_GPL(devm_thermal_of_cooling_device_register);
1142 
1143 static bool thermal_cooling_device_present(struct thermal_cooling_device *cdev)
1144 {
1145 	struct thermal_cooling_device *pos = NULL;
1146 
1147 	list_for_each_entry(pos, &thermal_cdev_list, node) {
1148 		if (pos == cdev)
1149 			return true;
1150 	}
1151 
1152 	return false;
1153 }
1154 
1155 /**
1156  * thermal_cooling_device_update - Update a cooling device object
1157  * @cdev: Target cooling device.
1158  *
1159  * Update @cdev to reflect a change of the underlying hardware or platform.
1160  *
1161  * Must be called when the maximum cooling state of @cdev becomes invalid and so
1162  * its .get_max_state() callback needs to be run to produce the new maximum
1163  * cooling state value.
1164  */
1165 void thermal_cooling_device_update(struct thermal_cooling_device *cdev)
1166 {
1167 	struct thermal_instance *ti;
1168 	unsigned long state;
1169 
1170 	if (IS_ERR_OR_NULL(cdev))
1171 		return;
1172 
1173 	/*
1174 	 * Hold thermal_list_lock throughout the update to prevent the device
1175 	 * from going away while being updated.
1176 	 */
1177 	mutex_lock(&thermal_list_lock);
1178 
1179 	if (!thermal_cooling_device_present(cdev))
1180 		goto unlock_list;
1181 
1182 	/*
1183 	 * Update under the cdev lock to prevent the state from being set beyond
1184 	 * the new limit concurrently.
1185 	 */
1186 	mutex_lock(&cdev->lock);
1187 
1188 	if (cdev->ops->get_max_state(cdev, &cdev->max_state))
1189 		goto unlock;
1190 
1191 	thermal_cooling_device_stats_reinit(cdev);
1192 
1193 	list_for_each_entry(ti, &cdev->thermal_instances, cdev_node) {
1194 		if (ti->upper == cdev->max_state)
1195 			continue;
1196 
1197 		if (ti->upper < cdev->max_state) {
1198 			if (ti->upper_no_limit)
1199 				ti->upper = cdev->max_state;
1200 
1201 			continue;
1202 		}
1203 
1204 		ti->upper = cdev->max_state;
1205 		if (ti->lower > ti->upper)
1206 			ti->lower = ti->upper;
1207 
1208 		if (ti->target == THERMAL_NO_TARGET)
1209 			continue;
1210 
1211 		if (ti->target > ti->upper)
1212 			ti->target = ti->upper;
1213 	}
1214 
1215 	if (cdev->ops->get_cur_state(cdev, &state) || state > cdev->max_state)
1216 		goto unlock;
1217 
1218 	thermal_cooling_device_stats_update(cdev, state);
1219 
1220 unlock:
1221 	mutex_unlock(&cdev->lock);
1222 
1223 unlock_list:
1224 	mutex_unlock(&thermal_list_lock);
1225 }
1226 EXPORT_SYMBOL_GPL(thermal_cooling_device_update);
1227 
1228 /**
1229  * thermal_cooling_device_unregister - removes a thermal cooling device
1230  * @cdev:	the thermal cooling device to remove.
1231  *
1232  * thermal_cooling_device_unregister() must be called when a registered
1233  * thermal cooling device is no longer needed.
1234  */
1235 void thermal_cooling_device_unregister(struct thermal_cooling_device *cdev)
1236 {
1237 	struct thermal_zone_device *tz;
1238 
1239 	if (!cdev)
1240 		return;
1241 
1242 	thermal_debug_cdev_remove(cdev);
1243 
1244 	mutex_lock(&thermal_list_lock);
1245 
1246 	if (!thermal_cooling_device_present(cdev)) {
1247 		mutex_unlock(&thermal_list_lock);
1248 		return;
1249 	}
1250 
1251 	list_del(&cdev->node);
1252 
1253 	/* Unbind all thermal zones associated with 'this' cdev */
1254 	list_for_each_entry(tz, &thermal_tz_list, node) {
1255 		if (tz->ops.unbind)
1256 			tz->ops.unbind(tz, cdev);
1257 	}
1258 
1259 	mutex_unlock(&thermal_list_lock);
1260 
1261 	device_unregister(&cdev->device);
1262 }
1263 EXPORT_SYMBOL_GPL(thermal_cooling_device_unregister);
1264 
1265 static void bind_tz(struct thermal_zone_device *tz)
1266 {
1267 	int ret;
1268 	struct thermal_cooling_device *pos = NULL;
1269 
1270 	if (!tz->ops.bind)
1271 		return;
1272 
1273 	mutex_lock(&thermal_list_lock);
1274 
1275 	list_for_each_entry(pos, &thermal_cdev_list, node) {
1276 		ret = tz->ops.bind(tz, pos);
1277 		if (ret)
1278 			print_bind_err_msg(tz, pos, ret);
1279 	}
1280 
1281 	mutex_unlock(&thermal_list_lock);
1282 }
1283 
1284 static void thermal_set_delay_jiffies(unsigned long *delay_jiffies, int delay_ms)
1285 {
1286 	*delay_jiffies = msecs_to_jiffies(delay_ms);
1287 	if (delay_ms > 1000)
1288 		*delay_jiffies = round_jiffies(*delay_jiffies);
1289 }
1290 
1291 int thermal_zone_get_crit_temp(struct thermal_zone_device *tz, int *temp)
1292 {
1293 	const struct thermal_trip_desc *td;
1294 	int ret = -EINVAL;
1295 
1296 	if (tz->ops.get_crit_temp)
1297 		return tz->ops.get_crit_temp(tz, temp);
1298 
1299 	mutex_lock(&tz->lock);
1300 
1301 	for_each_trip_desc(tz, td) {
1302 		const struct thermal_trip *trip = &td->trip;
1303 
1304 		if (trip->type == THERMAL_TRIP_CRITICAL) {
1305 			*temp = trip->temperature;
1306 			ret = 0;
1307 			break;
1308 		}
1309 	}
1310 
1311 	mutex_unlock(&tz->lock);
1312 
1313 	return ret;
1314 }
1315 EXPORT_SYMBOL_GPL(thermal_zone_get_crit_temp);
1316 
1317 /**
1318  * thermal_zone_device_register_with_trips() - register a new thermal zone device
1319  * @type:	the thermal zone device type
1320  * @trips:	a pointer to an array of thermal trips
1321  * @num_trips:	the number of trip points the thermal zone support
1322  * @devdata:	private device data
1323  * @ops:	standard thermal zone device callbacks
1324  * @tzp:	thermal zone platform parameters
1325  * @passive_delay: number of milliseconds to wait between polls when
1326  *		   performing passive cooling
1327  * @polling_delay: number of milliseconds to wait between polls when checking
1328  *		   whether trip points have been crossed (0 for interrupt
1329  *		   driven systems)
1330  *
1331  * This interface function adds a new thermal zone device (sensor) to
1332  * /sys/class/thermal folder as thermal_zone[0-*]. It tries to bind all the
1333  * thermal cooling devices registered at the same time.
1334  * thermal_zone_device_unregister() must be called when the device is no
1335  * longer needed. The passive cooling depends on the .get_trend() return value.
1336  *
1337  * Return: a pointer to the created struct thermal_zone_device or an
1338  * in case of error, an ERR_PTR. Caller must check return value with
1339  * IS_ERR*() helpers.
1340  */
1341 struct thermal_zone_device *
1342 thermal_zone_device_register_with_trips(const char *type,
1343 					const struct thermal_trip *trips,
1344 					int num_trips, void *devdata,
1345 					const struct thermal_zone_device_ops *ops,
1346 					const struct thermal_zone_params *tzp,
1347 					int passive_delay, int polling_delay)
1348 {
1349 	const struct thermal_trip *trip = trips;
1350 	struct thermal_zone_device *tz;
1351 	struct thermal_trip_desc *td;
1352 	int id;
1353 	int result;
1354 	struct thermal_governor *governor;
1355 
1356 	if (!type || strlen(type) == 0) {
1357 		pr_err("No thermal zone type defined\n");
1358 		return ERR_PTR(-EINVAL);
1359 	}
1360 
1361 	if (strlen(type) >= THERMAL_NAME_LENGTH) {
1362 		pr_err("Thermal zone name (%s) too long, should be under %d chars\n",
1363 		       type, THERMAL_NAME_LENGTH);
1364 		return ERR_PTR(-EINVAL);
1365 	}
1366 
1367 	if (num_trips < 0) {
1368 		pr_err("Incorrect number of thermal trips\n");
1369 		return ERR_PTR(-EINVAL);
1370 	}
1371 
1372 	if (!ops || !ops->get_temp) {
1373 		pr_err("Thermal zone device ops not defined\n");
1374 		return ERR_PTR(-EINVAL);
1375 	}
1376 
1377 	if (num_trips > 0 && !trips)
1378 		return ERR_PTR(-EINVAL);
1379 
1380 	if (!thermal_class)
1381 		return ERR_PTR(-ENODEV);
1382 
1383 	tz = kzalloc(struct_size(tz, trips, num_trips), GFP_KERNEL);
1384 	if (!tz)
1385 		return ERR_PTR(-ENOMEM);
1386 
1387 	if (tzp) {
1388 		tz->tzp = kmemdup(tzp, sizeof(*tzp), GFP_KERNEL);
1389 		if (!tz->tzp) {
1390 			result = -ENOMEM;
1391 			goto free_tz;
1392 		}
1393 	}
1394 
1395 	INIT_LIST_HEAD(&tz->thermal_instances);
1396 	INIT_LIST_HEAD(&tz->node);
1397 	ida_init(&tz->ida);
1398 	mutex_init(&tz->lock);
1399 	init_completion(&tz->removal);
1400 	id = ida_alloc(&thermal_tz_ida, GFP_KERNEL);
1401 	if (id < 0) {
1402 		result = id;
1403 		goto free_tzp;
1404 	}
1405 
1406 	tz->id = id;
1407 	strscpy(tz->type, type, sizeof(tz->type));
1408 
1409 	tz->ops = *ops;
1410 	if (!tz->ops.critical)
1411 		tz->ops.critical = thermal_zone_device_critical;
1412 
1413 	tz->device.class = thermal_class;
1414 	tz->devdata = devdata;
1415 	tz->num_trips = num_trips;
1416 	for_each_trip_desc(tz, td) {
1417 		td->trip = *trip++;
1418 		/*
1419 		 * Mark all thresholds as invalid to start with even though
1420 		 * this only matters for the trips that start as invalid and
1421 		 * become valid later.
1422 		 */
1423 		td->threshold = INT_MAX;
1424 	}
1425 
1426 	thermal_set_delay_jiffies(&tz->passive_delay_jiffies, passive_delay);
1427 	thermal_set_delay_jiffies(&tz->polling_delay_jiffies, polling_delay);
1428 
1429 	/* sys I/F */
1430 	/* Add nodes that are always present via .groups */
1431 	result = thermal_zone_create_device_groups(tz);
1432 	if (result)
1433 		goto remove_id;
1434 
1435 	/* A new thermal zone needs to be updated anyway. */
1436 	atomic_set(&tz->need_update, 1);
1437 
1438 	result = dev_set_name(&tz->device, "thermal_zone%d", tz->id);
1439 	if (result) {
1440 		thermal_zone_destroy_device_groups(tz);
1441 		goto remove_id;
1442 	}
1443 	result = device_register(&tz->device);
1444 	if (result)
1445 		goto release_device;
1446 
1447 	/* Update 'this' zone's governor information */
1448 	mutex_lock(&thermal_governor_lock);
1449 
1450 	if (tz->tzp)
1451 		governor = __find_governor(tz->tzp->governor_name);
1452 	else
1453 		governor = def_governor;
1454 
1455 	result = thermal_set_governor(tz, governor);
1456 	if (result) {
1457 		mutex_unlock(&thermal_governor_lock);
1458 		goto unregister;
1459 	}
1460 
1461 	mutex_unlock(&thermal_governor_lock);
1462 
1463 	if (!tz->tzp || !tz->tzp->no_hwmon) {
1464 		result = thermal_add_hwmon_sysfs(tz);
1465 		if (result)
1466 			goto unregister;
1467 	}
1468 
1469 	mutex_lock(&thermal_list_lock);
1470 	mutex_lock(&tz->lock);
1471 	list_add_tail(&tz->node, &thermal_tz_list);
1472 	mutex_unlock(&tz->lock);
1473 	mutex_unlock(&thermal_list_lock);
1474 
1475 	/* Bind cooling devices for this zone */
1476 	bind_tz(tz);
1477 
1478 	thermal_zone_device_init(tz);
1479 	/* Update the new thermal zone and mark it as already updated. */
1480 	if (atomic_cmpxchg(&tz->need_update, 1, 0))
1481 		thermal_zone_device_update(tz, THERMAL_EVENT_UNSPECIFIED);
1482 
1483 	thermal_notify_tz_create(tz);
1484 
1485 	thermal_debug_tz_add(tz);
1486 
1487 	return tz;
1488 
1489 unregister:
1490 	device_del(&tz->device);
1491 release_device:
1492 	put_device(&tz->device);
1493 remove_id:
1494 	ida_free(&thermal_tz_ida, id);
1495 free_tzp:
1496 	kfree(tz->tzp);
1497 free_tz:
1498 	kfree(tz);
1499 	return ERR_PTR(result);
1500 }
1501 EXPORT_SYMBOL_GPL(thermal_zone_device_register_with_trips);
1502 
1503 struct thermal_zone_device *thermal_tripless_zone_device_register(
1504 					const char *type,
1505 					void *devdata,
1506 					const struct thermal_zone_device_ops *ops,
1507 					const struct thermal_zone_params *tzp)
1508 {
1509 	return thermal_zone_device_register_with_trips(type, NULL, 0, devdata,
1510 						       ops, tzp, 0, 0);
1511 }
1512 EXPORT_SYMBOL_GPL(thermal_tripless_zone_device_register);
1513 
1514 void *thermal_zone_device_priv(struct thermal_zone_device *tzd)
1515 {
1516 	return tzd->devdata;
1517 }
1518 EXPORT_SYMBOL_GPL(thermal_zone_device_priv);
1519 
1520 const char *thermal_zone_device_type(struct thermal_zone_device *tzd)
1521 {
1522 	return tzd->type;
1523 }
1524 EXPORT_SYMBOL_GPL(thermal_zone_device_type);
1525 
1526 int thermal_zone_device_id(struct thermal_zone_device *tzd)
1527 {
1528 	return tzd->id;
1529 }
1530 EXPORT_SYMBOL_GPL(thermal_zone_device_id);
1531 
1532 struct device *thermal_zone_device(struct thermal_zone_device *tzd)
1533 {
1534 	return &tzd->device;
1535 }
1536 EXPORT_SYMBOL_GPL(thermal_zone_device);
1537 
1538 /**
1539  * thermal_zone_device_unregister - removes the registered thermal zone device
1540  * @tz: the thermal zone device to remove
1541  */
1542 void thermal_zone_device_unregister(struct thermal_zone_device *tz)
1543 {
1544 	struct thermal_cooling_device *cdev;
1545 	struct thermal_zone_device *pos = NULL;
1546 
1547 	if (!tz)
1548 		return;
1549 
1550 	thermal_debug_tz_remove(tz);
1551 
1552 	mutex_lock(&thermal_list_lock);
1553 	list_for_each_entry(pos, &thermal_tz_list, node)
1554 		if (pos == tz)
1555 			break;
1556 	if (pos != tz) {
1557 		/* thermal zone device not found */
1558 		mutex_unlock(&thermal_list_lock);
1559 		return;
1560 	}
1561 
1562 	mutex_lock(&tz->lock);
1563 	list_del(&tz->node);
1564 	mutex_unlock(&tz->lock);
1565 
1566 	/* Unbind all cdevs associated with 'this' thermal zone */
1567 	list_for_each_entry(cdev, &thermal_cdev_list, node)
1568 		if (tz->ops.unbind)
1569 			tz->ops.unbind(tz, cdev);
1570 
1571 	mutex_unlock(&thermal_list_lock);
1572 
1573 	cancel_delayed_work_sync(&tz->poll_queue);
1574 
1575 	thermal_set_governor(tz, NULL);
1576 
1577 	thermal_remove_hwmon_sysfs(tz);
1578 	ida_free(&thermal_tz_ida, tz->id);
1579 	ida_destroy(&tz->ida);
1580 
1581 	device_del(&tz->device);
1582 
1583 	kfree(tz->tzp);
1584 
1585 	put_device(&tz->device);
1586 
1587 	thermal_notify_tz_delete(tz);
1588 
1589 	wait_for_completion(&tz->removal);
1590 	kfree(tz);
1591 }
1592 EXPORT_SYMBOL_GPL(thermal_zone_device_unregister);
1593 
1594 /**
1595  * thermal_zone_get_zone_by_name() - search for a zone and returns its ref
1596  * @name: thermal zone name to fetch the temperature
1597  *
1598  * When only one zone is found with the passed name, returns a reference to it.
1599  *
1600  * Return: On success returns a reference to an unique thermal zone with
1601  * matching name equals to @name, an ERR_PTR otherwise (-EINVAL for invalid
1602  * paramenters, -ENODEV for not found and -EEXIST for multiple matches).
1603  */
1604 struct thermal_zone_device *thermal_zone_get_zone_by_name(const char *name)
1605 {
1606 	struct thermal_zone_device *pos = NULL, *ref = ERR_PTR(-EINVAL);
1607 	unsigned int found = 0;
1608 
1609 	if (!name)
1610 		goto exit;
1611 
1612 	mutex_lock(&thermal_list_lock);
1613 	list_for_each_entry(pos, &thermal_tz_list, node)
1614 		if (!strncasecmp(name, pos->type, THERMAL_NAME_LENGTH)) {
1615 			found++;
1616 			ref = pos;
1617 		}
1618 	mutex_unlock(&thermal_list_lock);
1619 
1620 	/* nothing has been found, thus an error code for it */
1621 	if (found == 0)
1622 		ref = ERR_PTR(-ENODEV);
1623 	else if (found > 1)
1624 	/* Success only when an unique zone is found */
1625 		ref = ERR_PTR(-EEXIST);
1626 
1627 exit:
1628 	return ref;
1629 }
1630 EXPORT_SYMBOL_GPL(thermal_zone_get_zone_by_name);
1631 
1632 static void thermal_zone_device_resume(struct work_struct *work)
1633 {
1634 	struct thermal_zone_device *tz;
1635 
1636 	tz = container_of(work, struct thermal_zone_device, poll_queue.work);
1637 
1638 	mutex_lock(&tz->lock);
1639 
1640 	tz->suspended = false;
1641 
1642 	thermal_zone_device_init(tz);
1643 	__thermal_zone_device_update(tz, THERMAL_EVENT_UNSPECIFIED);
1644 
1645 	mutex_unlock(&tz->lock);
1646 }
1647 
1648 static int thermal_pm_notify(struct notifier_block *nb,
1649 			     unsigned long mode, void *_unused)
1650 {
1651 	struct thermal_zone_device *tz;
1652 
1653 	switch (mode) {
1654 	case PM_HIBERNATION_PREPARE:
1655 	case PM_RESTORE_PREPARE:
1656 	case PM_SUSPEND_PREPARE:
1657 		mutex_lock(&thermal_list_lock);
1658 
1659 		list_for_each_entry(tz, &thermal_tz_list, node) {
1660 			mutex_lock(&tz->lock);
1661 
1662 			tz->suspended = true;
1663 
1664 			mutex_unlock(&tz->lock);
1665 		}
1666 
1667 		mutex_unlock(&thermal_list_lock);
1668 		break;
1669 	case PM_POST_HIBERNATION:
1670 	case PM_POST_RESTORE:
1671 	case PM_POST_SUSPEND:
1672 		mutex_lock(&thermal_list_lock);
1673 
1674 		list_for_each_entry(tz, &thermal_tz_list, node) {
1675 			mutex_lock(&tz->lock);
1676 
1677 			cancel_delayed_work(&tz->poll_queue);
1678 
1679 			/*
1680 			 * Replace the work function with the resume one, which
1681 			 * will restore the original work function and schedule
1682 			 * the polling work if needed.
1683 			 */
1684 			INIT_DELAYED_WORK(&tz->poll_queue,
1685 					  thermal_zone_device_resume);
1686 			/* Queue up the work without a delay. */
1687 			mod_delayed_work(system_freezable_power_efficient_wq,
1688 					 &tz->poll_queue, 0);
1689 
1690 			mutex_unlock(&tz->lock);
1691 		}
1692 
1693 		mutex_unlock(&thermal_list_lock);
1694 		break;
1695 	default:
1696 		break;
1697 	}
1698 	return 0;
1699 }
1700 
1701 static struct notifier_block thermal_pm_nb = {
1702 	.notifier_call = thermal_pm_notify,
1703 };
1704 
1705 static int __init thermal_init(void)
1706 {
1707 	int result;
1708 
1709 	thermal_debug_init();
1710 
1711 	result = thermal_netlink_init();
1712 	if (result)
1713 		goto error;
1714 
1715 	result = thermal_register_governors();
1716 	if (result)
1717 		goto unregister_netlink;
1718 
1719 	thermal_class = kzalloc(sizeof(*thermal_class), GFP_KERNEL);
1720 	if (!thermal_class) {
1721 		result = -ENOMEM;
1722 		goto unregister_governors;
1723 	}
1724 
1725 	thermal_class->name = "thermal";
1726 	thermal_class->dev_release = thermal_release;
1727 
1728 	result = class_register(thermal_class);
1729 	if (result) {
1730 		kfree(thermal_class);
1731 		thermal_class = NULL;
1732 		goto unregister_governors;
1733 	}
1734 
1735 	result = register_pm_notifier(&thermal_pm_nb);
1736 	if (result)
1737 		pr_warn("Thermal: Can not register suspend notifier, return %d\n",
1738 			result);
1739 
1740 	return 0;
1741 
1742 unregister_governors:
1743 	thermal_unregister_governors();
1744 unregister_netlink:
1745 	thermal_netlink_exit();
1746 error:
1747 	mutex_destroy(&thermal_list_lock);
1748 	mutex_destroy(&thermal_governor_lock);
1749 	return result;
1750 }
1751 postcore_initcall(thermal_init);
1752