blob: 5c15a6357f0ff1f04a18d7539e6b387008cee603 [file] [log] [blame]
xjb04a4022021-11-25 15:01:52 +08001/*
2 * linux/drivers/thermal/cpu_cooling.c
3 *
4 * Copyright (C) 2012 Samsung Electronics Co., Ltd(http://www.samsung.com)
5 * Copyright (C) 2012 Amit Daniel <amit.kachhap@linaro.org>
6 *
7 * Copyright (C) 2014 Viresh Kumar <viresh.kumar@linaro.org>
8 *
9 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
10 * This program is free software; you can redistribute it and/or modify
11 * it under the terms of the GNU General Public License as published by
12 * the Free Software Foundation; version 2 of the License.
13 *
14 * This program is distributed in the hope that it will be useful, but
15 * WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * General Public License for more details.
18 *
19 * You should have received a copy of the GNU General Public License along
20 * with this program; if not, write to the Free Software Foundation, Inc.,
21 * 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA.
22 *
23 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
24 */
25#include <linux/module.h>
26#include <linux/thermal.h>
27#include <linux/cpufreq.h>
28#include <linux/err.h>
29#include <linux/idr.h>
30#include <linux/pm_opp.h>
31#include <linux/slab.h>
32#include <linux/cpu.h>
33#include <linux/cpu_cooling.h>
34#include <linux/energy_model.h>
35
36#include <trace/events/thermal.h>
37
38/*
39 * Cooling state <-> CPUFreq frequency
40 *
41 * Cooling states are translated to frequencies throughout this driver and this
42 * is the relation between them.
43 *
44 * Highest cooling state corresponds to lowest possible frequency.
45 *
46 * i.e.
47 * level 0 --> 1st Max Freq
48 * level 1 --> 2nd Max Freq
49 * ...
50 */
51
52/**
53 * struct time_in_idle - Idle time stats
54 * @time: previous reading of the absolute time that this cpu was idle
55 * @timestamp: wall time of the last invocation of get_cpu_idle_time_us()
56 */
57struct time_in_idle {
58 u64 time;
59 u64 timestamp;
60};
61
62/**
63 * struct cpufreq_cooling_device - data for cooling device with cpufreq
64 * @id: unique integer value corresponding to each cpufreq_cooling_device
65 * registered.
66 * @last_load: load measured by the latest call to cpufreq_get_requested_power()
67 * @cpufreq_state: integer value representing the current state of cpufreq
68 * cooling devices.
69 * @clipped_freq: integer value representing the absolute value of the clipped
70 * frequency.
71 * @max_level: maximum cooling level. One less than total number of valid
72 * cpufreq frequencies.
73 * @em: Reference on the Energy Model of the device
74 * @cdev: thermal_cooling_device pointer to keep track of the
75 * registered cooling device.
76 * @policy: cpufreq policy.
77 * @node: list_head to link all cpufreq_cooling_device together.
78 * @idle_time: idle time stats
79 *
80 * This structure is required for keeping information of each registered
81 * cpufreq_cooling_device.
82 */
83struct cpufreq_cooling_device {
84 int id;
85 u32 last_load;
86 unsigned int cpufreq_state;
87 unsigned int clipped_freq;
88 unsigned int max_level;
89 struct em_perf_domain *em;
90 struct thermal_cooling_device *cdev;
91 struct cpufreq_policy *policy;
92 struct list_head node;
93 struct time_in_idle *idle_time;
94};
95
96static DEFINE_IDA(cpufreq_ida);
97static DEFINE_MUTEX(cooling_list_lock);
98static LIST_HEAD(cpufreq_cdev_list);
99
100/* Below code defines functions to be used for cpufreq as cooling device */
101
102/**
103 * cpufreq_thermal_notifier - notifier callback for cpufreq policy change.
104 * @nb: struct notifier_block * with callback info.
105 * @event: value showing cpufreq event for which this function invoked.
106 * @data: callback-specific data
107 *
108 * Callback to hijack the notification on cpufreq policy transition.
109 * Every time there is a change in policy, we will intercept and
110 * update the cpufreq policy with thermal constraints.
111 *
112 * Return: 0 (success)
113 */
114static int cpufreq_thermal_notifier(struct notifier_block *nb,
115 unsigned long event, void *data)
116{
117 struct cpufreq_policy *policy = data;
118 unsigned long clipped_freq;
119 struct cpufreq_cooling_device *cpufreq_cdev;
120
121 if (event != CPUFREQ_ADJUST)
122 return NOTIFY_DONE;
123
124 mutex_lock(&cooling_list_lock);
125 list_for_each_entry(cpufreq_cdev, &cpufreq_cdev_list, node) {
126 /*
127 * A new copy of the policy is sent to the notifier and can't
128 * compare that directly.
129 */
130 if (policy->cpu != cpufreq_cdev->policy->cpu)
131 continue;
132
133 /*
134 * policy->max is the maximum allowed frequency defined by user
135 * and clipped_freq is the maximum that thermal constraints
136 * allow.
137 *
138 * If clipped_freq is lower than policy->max, then we need to
139 * readjust policy->max.
140 *
141 * But, if clipped_freq is greater than policy->max, we don't
142 * need to do anything.
143 */
144 clipped_freq = cpufreq_cdev->clipped_freq;
145
146 if (policy->max > clipped_freq)
147 cpufreq_verify_within_limits(policy, 0, clipped_freq);
148 break;
149 }
150 mutex_unlock(&cooling_list_lock);
151
152 return NOTIFY_OK;
153}
154
155#ifdef CONFIG_ENERGY_MODEL
156/**
157 * get_level: Find the level for a particular frequency
158 * @cpufreq_cdev: cpufreq_cdev for which the property is required
159 * @freq: Frequency
160 *
161 * Return: level corresponding to the frequency.
162 */
163static unsigned long get_level(struct cpufreq_cooling_device *cpufreq_cdev,
164 unsigned int freq)
165{
166 int i;
167
168 for (i = cpufreq_cdev->max_level - 1; i >= 0; i--) {
169 if (freq > cpufreq_cdev->em->table[i].frequency)
170 break;
171 }
172
173 return cpufreq_cdev->max_level - i - 1;
174}
175
176
177static u32 cpu_freq_to_power(struct cpufreq_cooling_device *cpufreq_cdev,
178 u32 freq)
179{
180 int i;
181
182 for (i = cpufreq_cdev->max_level - 1; i >= 0; i--) {
183 if (freq > cpufreq_cdev->em->table[i].frequency)
184 break;
185 }
186
187 return cpufreq_cdev->em->table[i + 1].power;
188}
189
190static u32 cpu_power_to_freq(struct cpufreq_cooling_device *cpufreq_cdev,
191 u32 power)
192{
193 int i;
194
195 for (i = cpufreq_cdev->max_level - 1; i >= 0; i--) {
196 if (power > cpufreq_cdev->em->table[i].power)
197 break;
198 }
199
200 return cpufreq_cdev->em->table[i + 1].frequency;
201}
202
203/**
204 * get_load() - get load for a cpu since last updated
205 * @cpufreq_cdev: &struct cpufreq_cooling_device for this cpu
206 * @cpu: cpu number
207 * @cpu_idx: index of the cpu in time_in_idle*
208 *
209 * Return: The average load of cpu @cpu in percentage since this
210 * function was last called.
211 */
212static u32 get_load(struct cpufreq_cooling_device *cpufreq_cdev, int cpu,
213 int cpu_idx)
214{
215 u32 load;
216 u64 now, now_idle, delta_time, delta_idle;
217 struct time_in_idle *idle_time = &cpufreq_cdev->idle_time[cpu_idx];
218
219 now_idle = get_cpu_idle_time(cpu, &now, 0);
220 delta_idle = now_idle - idle_time->time;
221 delta_time = now - idle_time->timestamp;
222
223 if (delta_time <= delta_idle)
224 load = 0;
225 else
226 load = div64_u64(100 * (delta_time - delta_idle), delta_time);
227
228 idle_time->time = now_idle;
229 idle_time->timestamp = now;
230
231 return load;
232}
233
234/**
235 * get_dynamic_power() - calculate the dynamic power
236 * @cpufreq_cdev: &cpufreq_cooling_device for this cdev
237 * @freq: current frequency
238 *
239 * Return: the dynamic power consumed by the cpus described by
240 * @cpufreq_cdev.
241 */
242static u32 get_dynamic_power(struct cpufreq_cooling_device *cpufreq_cdev,
243 unsigned long freq)
244{
245 u32 raw_cpu_power;
246
247 raw_cpu_power = cpu_freq_to_power(cpufreq_cdev, freq);
248 return (raw_cpu_power * cpufreq_cdev->last_load) / 100;
249}
250#endif
251
252/* cpufreq cooling device callback functions are defined below */
253
254/**
255 * cpufreq_get_max_state - callback function to get the max cooling state.
256 * @cdev: thermal cooling device pointer.
257 * @state: fill this variable with the max cooling state.
258 *
259 * Callback for the thermal cooling device to return the cpufreq
260 * max cooling state.
261 *
262 * Return: 0 on success, an error code otherwise.
263 */
264static int cpufreq_get_max_state(struct thermal_cooling_device *cdev,
265 unsigned long *state)
266{
267 struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata;
268
269 *state = cpufreq_cdev->max_level;
270 return 0;
271}
272
273/**
274 * cpufreq_get_cur_state - callback function to get the current cooling state.
275 * @cdev: thermal cooling device pointer.
276 * @state: fill this variable with the current cooling state.
277 *
278 * Callback for the thermal cooling device to return the cpufreq
279 * current cooling state.
280 *
281 * Return: 0 on success, an error code otherwise.
282 */
283static int cpufreq_get_cur_state(struct thermal_cooling_device *cdev,
284 unsigned long *state)
285{
286 struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata;
287
288 *state = cpufreq_cdev->cpufreq_state;
289
290 return 0;
291}
292
293static unsigned int get_state_freq(struct cpufreq_cooling_device *cpufreq_cdev,
294 unsigned long state)
295{
296 struct cpufreq_policy *policy;
297 unsigned long idx;
298
299#ifdef CONFIG_ENERGY_MODEL
300 /* Use the Energy Model table if available */
301 if (cpufreq_cdev->em) {
302 idx = cpufreq_cdev->max_level - state;
303 return cpufreq_cdev->em->table[idx].frequency;
304 }
305#endif
306
307 /* Otherwise, fallback on the CPUFreq table */
308 policy = cpufreq_cdev->policy;
309 if (policy->freq_table_sorted == CPUFREQ_TABLE_SORTED_ASCENDING)
310 idx = cpufreq_cdev->max_level - state;
311 else
312 idx = state;
313
314 return policy->freq_table[idx].frequency;
315}
316
317/**
318 * cpufreq_set_cur_state - callback function to set the current cooling state.
319 * @cdev: thermal cooling device pointer.
320 * @state: set this variable to the current cooling state.
321 *
322 * Callback for the thermal cooling device to change the cpufreq
323 * current cooling state.
324 *
325 * Return: 0 on success, an error code otherwise.
326 */
327static int cpufreq_set_cur_state(struct thermal_cooling_device *cdev,
328 unsigned long state)
329{
330 struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata;
331 unsigned int clip_freq;
332
333 /* Request state should be less than max_level */
334 if (WARN_ON(state > cpufreq_cdev->max_level))
335 return -EINVAL;
336
337 /* Check if the old cooling action is same as new cooling action */
338 if (cpufreq_cdev->cpufreq_state == state)
339 return 0;
340
341 clip_freq = get_state_freq(cpufreq_cdev, state);
342 cpufreq_cdev->cpufreq_state = state;
343 cpufreq_cdev->clipped_freq = clip_freq;
344
345 cpufreq_update_policy(cpufreq_cdev->policy->cpu);
346
347 return 0;
348}
349
350#ifdef CONFIG_ENERGY_MODEL
351/**
352 * cpufreq_get_requested_power() - get the current power
353 * @cdev: &thermal_cooling_device pointer
354 * @tz: a valid thermal zone device pointer
355 * @power: pointer in which to store the resulting power
356 *
357 * Calculate the current power consumption of the cpus in milliwatts
358 * and store it in @power. This function should actually calculate
359 * the requested power, but it's hard to get the frequency that
360 * cpufreq would have assigned if there were no thermal limits.
361 * Instead, we calculate the current power on the assumption that the
362 * immediate future will look like the immediate past.
363 *
364 * We use the current frequency and the average load since this
365 * function was last called. In reality, there could have been
366 * multiple opps since this function was last called and that affects
367 * the load calculation. While it's not perfectly accurate, this
368 * simplification is good enough and works. REVISIT this, as more
369 * complex code may be needed if experiments show that it's not
370 * accurate enough.
371 *
372 * Return: 0 on success, -E* if getting the static power failed.
373 */
374static int cpufreq_get_requested_power(struct thermal_cooling_device *cdev,
375 struct thermal_zone_device *tz,
376 u32 *power)
377{
378 unsigned long freq;
379 int i = 0, cpu;
380 u32 total_load = 0;
381 struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata;
382 struct cpufreq_policy *policy = cpufreq_cdev->policy;
383 u32 *load_cpu = NULL;
384
385 freq = cpufreq_quick_get(policy->cpu);
386
387 if (trace_thermal_power_cpu_get_power_enabled()) {
388 u32 ncpus = cpumask_weight(policy->related_cpus);
389
390 load_cpu = kcalloc(ncpus, sizeof(*load_cpu), GFP_KERNEL);
391 }
392
393 for_each_cpu(cpu, policy->related_cpus) {
394 u32 load;
395
396 if (cpu_online(cpu))
397 load = get_load(cpufreq_cdev, cpu, i);
398 else
399 load = 0;
400
401 total_load += load;
402 if (trace_thermal_power_cpu_limit_enabled() && load_cpu)
403 load_cpu[i] = load;
404
405 i++;
406 }
407
408 cpufreq_cdev->last_load = total_load;
409
410 *power = get_dynamic_power(cpufreq_cdev, freq);
411
412 if (load_cpu) {
413 trace_thermal_power_cpu_get_power(policy->related_cpus, freq,
414 load_cpu, i, *power);
415
416 kfree(load_cpu);
417 }
418
419 return 0;
420}
421
422/**
423 * cpufreq_state2power() - convert a cpu cdev state to power consumed
424 * @cdev: &thermal_cooling_device pointer
425 * @tz: a valid thermal zone device pointer
426 * @state: cooling device state to be converted
427 * @power: pointer in which to store the resulting power
428 *
429 * Convert cooling device state @state into power consumption in
430 * milliwatts assuming 100% load. Store the calculated power in
431 * @power.
432 *
433 * Return: 0 on success, -EINVAL if the cooling device state could not
434 * be converted into a frequency or other -E* if there was an error
435 * when calculating the static power.
436 */
437static int cpufreq_state2power(struct thermal_cooling_device *cdev,
438 struct thermal_zone_device *tz,
439 unsigned long state, u32 *power)
440{
441 unsigned int freq, num_cpus, idx;
442 struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata;
443
444 /* Request state should be less than max_level */
445 if (WARN_ON(state > cpufreq_cdev->max_level))
446 return -EINVAL;
447
448 num_cpus = cpumask_weight(cpufreq_cdev->policy->cpus);
449
450 idx = cpufreq_cdev->max_level - state;
451 freq = cpufreq_cdev->em->table[idx].frequency;
452 *power = cpu_freq_to_power(cpufreq_cdev, freq) * num_cpus;
453
454 return 0;
455}
456
457/**
458 * cpufreq_power2state() - convert power to a cooling device state
459 * @cdev: &thermal_cooling_device pointer
460 * @tz: a valid thermal zone device pointer
461 * @power: power in milliwatts to be converted
462 * @state: pointer in which to store the resulting state
463 *
464 * Calculate a cooling device state for the cpus described by @cdev
465 * that would allow them to consume at most @power mW and store it in
466 * @state. Note that this calculation depends on external factors
467 * such as the cpu load or the current static power. Calling this
468 * function with the same power as input can yield different cooling
469 * device states depending on those external factors.
470 *
471 * Return: 0 on success, -ENODEV if no cpus are online or -EINVAL if
472 * the calculated frequency could not be converted to a valid state.
473 * The latter should not happen unless the frequencies available to
474 * cpufreq have changed since the initialization of the cpu cooling
475 * device.
476 */
477static int cpufreq_power2state(struct thermal_cooling_device *cdev,
478 struct thermal_zone_device *tz, u32 power,
479 unsigned long *state)
480{
481 unsigned int cur_freq, target_freq;
482 u32 last_load, normalised_power;
483 struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata;
484 struct cpufreq_policy *policy = cpufreq_cdev->policy;
485
486 cur_freq = cpufreq_quick_get(policy->cpu);
487 power = power > 0 ? power : 0;
488 last_load = cpufreq_cdev->last_load ?: 1;
489 normalised_power = (power * 100) / last_load;
490 target_freq = cpu_power_to_freq(cpufreq_cdev, normalised_power);
491
492 *state = get_level(cpufreq_cdev, target_freq);
493 trace_thermal_power_cpu_limit(policy->related_cpus, target_freq, *state,
494 power);
495 return 0;
496}
497
498static struct thermal_cooling_device_ops cpufreq_power_cooling_ops = {
499 .get_max_state = cpufreq_get_max_state,
500 .get_cur_state = cpufreq_get_cur_state,
501 .set_cur_state = cpufreq_set_cur_state,
502 .get_requested_power = cpufreq_get_requested_power,
503 .state2power = cpufreq_state2power,
504 .power2state = cpufreq_power2state,
505};
506#endif
507
508/* Bind cpufreq callbacks to thermal cooling device ops */
509
510static struct thermal_cooling_device_ops cpufreq_cooling_ops = {
511 .get_max_state = cpufreq_get_max_state,
512 .get_cur_state = cpufreq_get_cur_state,
513 .set_cur_state = cpufreq_set_cur_state,
514};
515
516/* Notifier for cpufreq policy change */
517static struct notifier_block thermal_cpufreq_notifier_block = {
518 .notifier_call = cpufreq_thermal_notifier,
519};
520
521/**
522 * __cpufreq_cooling_register - helper function to create cpufreq cooling device
523 * @np: a valid struct device_node to the cooling device device tree node
524 * @policy: cpufreq policy
525 * Normally this should be same as cpufreq policy->related_cpus.
526 * @try_model: true if a power model should be used
527 *
528 * This interface function registers the cpufreq cooling device with the name
529 * "thermal-cpufreq-%x". This api can support multiple instances of cpufreq
530 * cooling devices. It also gives the opportunity to link the cooling device
531 * with a device tree node, in order to bind it via the thermal DT code.
532 *
533 * Return: a valid struct thermal_cooling_device pointer on success,
534 * on failure, it returns a corresponding ERR_PTR().
535 */
536static struct thermal_cooling_device *
537__cpufreq_cooling_register(struct device_node *np,
538 struct cpufreq_policy *policy, bool try_model)
539{
540 struct thermal_cooling_device *cdev;
541 struct cpufreq_cooling_device *cpufreq_cdev;
542 char dev_name[THERMAL_NAME_LENGTH];
543 unsigned int i, num_cpus;
544 int ret;
545 struct thermal_cooling_device_ops *cooling_ops;
546 bool first;
547
548 if (IS_ERR_OR_NULL(policy)) {
549 pr_err("%s: cpufreq policy isn't valid: %p\n", __func__, policy);
550 return ERR_PTR(-EINVAL);
551 }
552
553 i = cpufreq_table_count_valid_entries(policy);
554 if (!i) {
555 pr_debug("%s: CPUFreq table not found or has no valid entries\n",
556 __func__);
557 return ERR_PTR(-ENODEV);
558 }
559
560 cpufreq_cdev = kzalloc(sizeof(*cpufreq_cdev), GFP_KERNEL);
561 if (!cpufreq_cdev)
562 return ERR_PTR(-ENOMEM);
563
564 cpufreq_cdev->policy = policy;
565 num_cpus = cpumask_weight(policy->related_cpus);
566 cpufreq_cdev->idle_time = kcalloc(num_cpus,
567 sizeof(*cpufreq_cdev->idle_time),
568 GFP_KERNEL);
569 if (!cpufreq_cdev->idle_time) {
570 cdev = ERR_PTR(-ENOMEM);
571 goto free_cdev;
572 }
573
574 /* max_level is an index, not a counter */
575 cpufreq_cdev->max_level = i - 1;
576
577#ifdef CONFIG_ENERGY_MODEL
578 if (try_model) {
579 struct em_perf_domain *em = em_cpu_get(policy->cpu);
580
581 if (!em || !cpumask_equal(policy->related_cpus,
582 to_cpumask(em->cpus))) {
583 cdev = ERR_PTR(-EINVAL);
584 goto free_idle_time;
585 }
586 cpufreq_cdev->em = em;
587 cooling_ops = &cpufreq_power_cooling_ops;
588 } else
589#endif
590 cooling_ops = &cpufreq_cooling_ops;
591
592 ret = ida_simple_get(&cpufreq_ida, 0, 0, GFP_KERNEL);
593 if (ret < 0) {
594 cdev = ERR_PTR(ret);
595 goto free_idle_time;
596 }
597 cpufreq_cdev->id = ret;
598
599 snprintf(dev_name, sizeof(dev_name), "thermal-cpufreq-%d",
600 cpufreq_cdev->id);
601
602 cdev = thermal_of_cooling_device_register(np, dev_name, cpufreq_cdev,
603 cooling_ops);
604 if (IS_ERR(cdev))
605 goto remove_ida;
606
607 cpufreq_cdev->clipped_freq = get_state_freq(cpufreq_cdev, 0);
608 cpufreq_cdev->cdev = cdev;
609
610 mutex_lock(&cooling_list_lock);
611 /* Register the notifier for first cpufreq cooling device */
612 first = list_empty(&cpufreq_cdev_list);
613 list_add(&cpufreq_cdev->node, &cpufreq_cdev_list);
614 mutex_unlock(&cooling_list_lock);
615
616 if (first)
617 cpufreq_register_notifier(&thermal_cpufreq_notifier_block,
618 CPUFREQ_POLICY_NOTIFIER);
619
620 return cdev;
621
622remove_ida:
623 ida_simple_remove(&cpufreq_ida, cpufreq_cdev->id);
624free_idle_time:
625 kfree(cpufreq_cdev->idle_time);
626free_cdev:
627 kfree(cpufreq_cdev);
628 return cdev;
629}
630
631/**
632 * cpufreq_cooling_register - function to create cpufreq cooling device.
633 * @policy: cpufreq policy
634 *
635 * This interface function registers the cpufreq cooling device with the name
636 * "thermal-cpufreq-%x". This api can support multiple instances of cpufreq
637 * cooling devices.
638 *
639 * Return: a valid struct thermal_cooling_device pointer on success,
640 * on failure, it returns a corresponding ERR_PTR().
641 */
642struct thermal_cooling_device *
643cpufreq_cooling_register(struct cpufreq_policy *policy)
644{
645 return __cpufreq_cooling_register(NULL, policy, false);
646}
647EXPORT_SYMBOL_GPL(cpufreq_cooling_register);
648
649/**
650 * of_cpufreq_cooling_register - function to create cpufreq cooling device.
651 * @policy: cpufreq policy
652 *
653 * This interface function registers the cpufreq cooling device with the name
654 * "thermal-cpufreq-%x". This api can support multiple instances of cpufreq
655 * cooling devices. Using this API, the cpufreq cooling device will be
656 * linked to the device tree node provided.
657 *
658 * Using this function, the cooling device will implement the power
659 * extensions by using a simple cpu power model. The cpus must have
660 * registered their OPPs using the OPP library.
661 *
662 * It also takes into account, if property present in policy CPU node, the
663 * static power consumed by the cpu.
664 *
665 * Return: a valid struct thermal_cooling_device pointer on success,
666 * and NULL on failure.
667 */
668struct thermal_cooling_device *
669of_cpufreq_cooling_register(struct cpufreq_policy *policy)
670{
671 struct device_node *np = of_get_cpu_node(policy->cpu, NULL);
672 struct thermal_cooling_device *cdev = NULL;
673
674 if (!np) {
675 pr_err("cpu_cooling: OF node not available for cpu%d\n",
676 policy->cpu);
677 return NULL;
678 }
679
680 if (of_find_property(np, "#cooling-cells", NULL)) {
681 cdev = __cpufreq_cooling_register(np, policy, true);
682 if (IS_ERR(cdev)) {
683 pr_err("cpu_cooling: cpu%d is not running as cooling device: %ld\n",
684 policy->cpu, PTR_ERR(cdev));
685 cdev = NULL;
686 }
687 }
688
689 of_node_put(np);
690 return cdev;
691}
692EXPORT_SYMBOL_GPL(of_cpufreq_cooling_register);
693
694/**
695 * cpufreq_cooling_unregister - function to remove cpufreq cooling device.
696 * @cdev: thermal cooling device pointer.
697 *
698 * This interface function unregisters the "thermal-cpufreq-%x" cooling device.
699 */
700void cpufreq_cooling_unregister(struct thermal_cooling_device *cdev)
701{
702 struct cpufreq_cooling_device *cpufreq_cdev;
703 bool last;
704
705 if (!cdev)
706 return;
707
708 cpufreq_cdev = cdev->devdata;
709
710 mutex_lock(&cooling_list_lock);
711 list_del(&cpufreq_cdev->node);
712 /* Unregister the notifier for the last cpufreq cooling device */
713 last = list_empty(&cpufreq_cdev_list);
714 mutex_unlock(&cooling_list_lock);
715
716 if (last)
717 cpufreq_unregister_notifier(&thermal_cpufreq_notifier_block,
718 CPUFREQ_POLICY_NOTIFIER);
719
720 thermal_cooling_device_unregister(cpufreq_cdev->cdev);
721 ida_simple_remove(&cpufreq_ida, cpufreq_cdev->id);
722 kfree(cpufreq_cdev->idle_time);
723 kfree(cpufreq_cdev);
724}
725EXPORT_SYMBOL_GPL(cpufreq_cooling_unregister);