Linux Audio

Check our new training course

Loading...
Note: File does not exist in v6.8.
  1// SPDX-License-Identifier: GPL-2.0
  2/*
  3 *  linux/drivers/thermal/cpu_cooling.c
  4 *
  5 *  Copyright (C) 2012	Samsung Electronics Co., Ltd(http://www.samsung.com)
  6 *
  7 *  Copyright (C) 2012-2018 Linaro Limited.
  8 *
  9 *  Authors:	Amit Daniel <amit.kachhap@linaro.org>
 10 *		Viresh Kumar <viresh.kumar@linaro.org>
 11 *
 12 */
 13#include <linux/module.h>
 14#include <linux/thermal.h>
 15#include <linux/cpufreq.h>
 16#include <linux/err.h>
 17#include <linux/idr.h>
 18#include <linux/pm_opp.h>
 19#include <linux/pm_qos.h>
 20#include <linux/slab.h>
 21#include <linux/cpu.h>
 22#include <linux/cpu_cooling.h>
 23
 24#include <trace/events/thermal.h>
 25
 26/*
 27 * Cooling state <-> CPUFreq frequency
 28 *
 29 * Cooling states are translated to frequencies throughout this driver and this
 30 * is the relation between them.
 31 *
 32 * Highest cooling state corresponds to lowest possible frequency.
 33 *
 34 * i.e.
 35 *	level 0 --> 1st Max Freq
 36 *	level 1 --> 2nd Max Freq
 37 *	...
 38 */
 39
 40/**
 41 * struct freq_table - frequency table along with power entries
 42 * @frequency:	frequency in KHz
 43 * @power:	power in mW
 44 *
 45 * This structure is built when the cooling device registers and helps
 46 * in translating frequency to power and vice versa.
 47 */
 48struct freq_table {
 49	u32 frequency;
 50	u32 power;
 51};
 52
 53/**
 54 * struct time_in_idle - Idle time stats
 55 * @time: previous reading of the absolute time that this cpu was idle
 56 * @timestamp: wall time of the last invocation of get_cpu_idle_time_us()
 57 */
 58struct time_in_idle {
 59	u64 time;
 60	u64 timestamp;
 61};
 62
 63/**
 64 * struct cpufreq_cooling_device - data for cooling device with cpufreq
 65 * @id: unique integer value corresponding to each cpufreq_cooling_device
 66 *	registered.
 67 * @last_load: load measured by the latest call to cpufreq_get_requested_power()
 68 * @cpufreq_state: integer value representing the current state of cpufreq
 69 *	cooling	devices.
 70 * @max_level: maximum cooling level. One less than total number of valid
 71 *	cpufreq frequencies.
 72 * @freq_table: Freq table in descending order of frequencies
 73 * @cdev: thermal_cooling_device pointer to keep track of the
 74 *	registered cooling device.
 75 * @policy: cpufreq policy.
 76 * @node: list_head to link all cpufreq_cooling_device together.
 77 * @idle_time: idle time stats
 78 *
 79 * This structure is required for keeping information of each registered
 80 * cpufreq_cooling_device.
 81 */
 82struct cpufreq_cooling_device {
 83	int id;
 84	u32 last_load;
 85	unsigned int cpufreq_state;
 86	unsigned int max_level;
 87	struct freq_table *freq_table;	/* In descending order */
 88	struct cpufreq_policy *policy;
 89	struct list_head node;
 90	struct time_in_idle *idle_time;
 91	struct freq_qos_request qos_req;
 92};
 93
 94static DEFINE_IDA(cpufreq_ida);
 95static DEFINE_MUTEX(cooling_list_lock);
 96static LIST_HEAD(cpufreq_cdev_list);
 97
 98/* Below code defines functions to be used for cpufreq as cooling device */
 99
100/**
101 * get_level: Find the level for a particular frequency
102 * @cpufreq_cdev: cpufreq_cdev for which the property is required
103 * @freq: Frequency
104 *
105 * Return: level corresponding to the frequency.
106 */
107static unsigned long get_level(struct cpufreq_cooling_device *cpufreq_cdev,
108			       unsigned int freq)
109{
110	struct freq_table *freq_table = cpufreq_cdev->freq_table;
111	unsigned long level;
112
113	for (level = 1; level <= cpufreq_cdev->max_level; level++)
114		if (freq > freq_table[level].frequency)
115			break;
116
117	return level - 1;
118}
119
120/**
121 * update_freq_table() - Update the freq table with power numbers
122 * @cpufreq_cdev:	the cpufreq cooling device in which to update the table
123 * @capacitance: dynamic power coefficient for these cpus
124 *
125 * Update the freq table with power numbers.  This table will be used in
126 * cpu_power_to_freq() and cpu_freq_to_power() to convert between power and
127 * frequency efficiently.  Power is stored in mW, frequency in KHz.  The
128 * resulting table is in descending order.
129 *
130 * Return: 0 on success, -EINVAL if there are no OPPs for any CPUs,
131 * or -ENOMEM if we run out of memory.
132 */
133static int update_freq_table(struct cpufreq_cooling_device *cpufreq_cdev,
134			     u32 capacitance)
135{
136	struct freq_table *freq_table = cpufreq_cdev->freq_table;
137	struct dev_pm_opp *opp;
138	struct device *dev = NULL;
139	int num_opps = 0, cpu = cpufreq_cdev->policy->cpu, i;
140
141	dev = get_cpu_device(cpu);
142	if (unlikely(!dev)) {
143		pr_warn("No cpu device for cpu %d\n", cpu);
144		return -ENODEV;
145	}
146
147	num_opps = dev_pm_opp_get_opp_count(dev);
148	if (num_opps < 0)
149		return num_opps;
150
151	/*
152	 * The cpufreq table is also built from the OPP table and so the count
153	 * should match.
154	 */
155	if (num_opps != cpufreq_cdev->max_level + 1) {
156		dev_warn(dev, "Number of OPPs not matching with max_levels\n");
157		return -EINVAL;
158	}
159
160	for (i = 0; i <= cpufreq_cdev->max_level; i++) {
161		unsigned long freq = freq_table[i].frequency * 1000;
162		u32 freq_mhz = freq_table[i].frequency / 1000;
163		u64 power;
164		u32 voltage_mv;
165
166		/*
167		 * Find ceil frequency as 'freq' may be slightly lower than OPP
168		 * freq due to truncation while converting to kHz.
169		 */
170		opp = dev_pm_opp_find_freq_ceil(dev, &freq);
171		if (IS_ERR(opp)) {
172			dev_err(dev, "failed to get opp for %lu frequency\n",
173				freq);
174			return -EINVAL;
175		}
176
177		voltage_mv = dev_pm_opp_get_voltage(opp) / 1000;
178		dev_pm_opp_put(opp);
179
180		/*
181		 * Do the multiplication with MHz and millivolt so as
182		 * to not overflow.
183		 */
184		power = (u64)capacitance * freq_mhz * voltage_mv * voltage_mv;
185		do_div(power, 1000000000);
186
187		/* power is stored in mW */
188		freq_table[i].power = power;
189	}
190
191	return 0;
192}
193
194static u32 cpu_freq_to_power(struct cpufreq_cooling_device *cpufreq_cdev,
195			     u32 freq)
196{
197	int i;
198	struct freq_table *freq_table = cpufreq_cdev->freq_table;
199
200	for (i = 1; i <= cpufreq_cdev->max_level; i++)
201		if (freq > freq_table[i].frequency)
202			break;
203
204	return freq_table[i - 1].power;
205}
206
207static u32 cpu_power_to_freq(struct cpufreq_cooling_device *cpufreq_cdev,
208			     u32 power)
209{
210	int i;
211	struct freq_table *freq_table = cpufreq_cdev->freq_table;
212
213	for (i = 1; i <= cpufreq_cdev->max_level; i++)
214		if (power > freq_table[i].power)
215			break;
216
217	return freq_table[i - 1].frequency;
218}
219
220/**
221 * get_load() - get load for a cpu since last updated
222 * @cpufreq_cdev:	&struct cpufreq_cooling_device for this cpu
223 * @cpu:	cpu number
224 * @cpu_idx:	index of the cpu in time_in_idle*
225 *
226 * Return: The average load of cpu @cpu in percentage since this
227 * function was last called.
228 */
229static u32 get_load(struct cpufreq_cooling_device *cpufreq_cdev, int cpu,
230		    int cpu_idx)
231{
232	u32 load;
233	u64 now, now_idle, delta_time, delta_idle;
234	struct time_in_idle *idle_time = &cpufreq_cdev->idle_time[cpu_idx];
235
236	now_idle = get_cpu_idle_time(cpu, &now, 0);
237	delta_idle = now_idle - idle_time->time;
238	delta_time = now - idle_time->timestamp;
239
240	if (delta_time <= delta_idle)
241		load = 0;
242	else
243		load = div64_u64(100 * (delta_time - delta_idle), delta_time);
244
245	idle_time->time = now_idle;
246	idle_time->timestamp = now;
247
248	return load;
249}
250
251/**
252 * get_dynamic_power() - calculate the dynamic power
253 * @cpufreq_cdev:	&cpufreq_cooling_device for this cdev
254 * @freq:	current frequency
255 *
256 * Return: the dynamic power consumed by the cpus described by
257 * @cpufreq_cdev.
258 */
259static u32 get_dynamic_power(struct cpufreq_cooling_device *cpufreq_cdev,
260			     unsigned long freq)
261{
262	u32 raw_cpu_power;
263
264	raw_cpu_power = cpu_freq_to_power(cpufreq_cdev, freq);
265	return (raw_cpu_power * cpufreq_cdev->last_load) / 100;
266}
267
268/* cpufreq cooling device callback functions are defined below */
269
270/**
271 * cpufreq_get_max_state - callback function to get the max cooling state.
272 * @cdev: thermal cooling device pointer.
273 * @state: fill this variable with the max cooling state.
274 *
275 * Callback for the thermal cooling device to return the cpufreq
276 * max cooling state.
277 *
278 * Return: 0 on success, an error code otherwise.
279 */
280static int cpufreq_get_max_state(struct thermal_cooling_device *cdev,
281				 unsigned long *state)
282{
283	struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata;
284
285	*state = cpufreq_cdev->max_level;
286	return 0;
287}
288
289/**
290 * cpufreq_get_cur_state - callback function to get the current cooling state.
291 * @cdev: thermal cooling device pointer.
292 * @state: fill this variable with the current cooling state.
293 *
294 * Callback for the thermal cooling device to return the cpufreq
295 * current cooling state.
296 *
297 * Return: 0 on success, an error code otherwise.
298 */
299static int cpufreq_get_cur_state(struct thermal_cooling_device *cdev,
300				 unsigned long *state)
301{
302	struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata;
303
304	*state = cpufreq_cdev->cpufreq_state;
305
306	return 0;
307}
308
309/**
310 * cpufreq_set_cur_state - callback function to set the current cooling state.
311 * @cdev: thermal cooling device pointer.
312 * @state: set this variable to the current cooling state.
313 *
314 * Callback for the thermal cooling device to change the cpufreq
315 * current cooling state.
316 *
317 * Return: 0 on success, an error code otherwise.
318 */
319static int cpufreq_set_cur_state(struct thermal_cooling_device *cdev,
320				 unsigned long state)
321{
322	struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata;
323
324	/* Request state should be less than max_level */
325	if (WARN_ON(state > cpufreq_cdev->max_level))
326		return -EINVAL;
327
328	/* Check if the old cooling action is same as new cooling action */
329	if (cpufreq_cdev->cpufreq_state == state)
330		return 0;
331
332	cpufreq_cdev->cpufreq_state = state;
333
334	return freq_qos_update_request(&cpufreq_cdev->qos_req,
335				cpufreq_cdev->freq_table[state].frequency);
336}
337
338/**
339 * cpufreq_get_requested_power() - get the current power
340 * @cdev:	&thermal_cooling_device pointer
341 * @tz:		a valid thermal zone device pointer
342 * @power:	pointer in which to store the resulting power
343 *
344 * Calculate the current power consumption of the cpus in milliwatts
345 * and store it in @power.  This function should actually calculate
346 * the requested power, but it's hard to get the frequency that
347 * cpufreq would have assigned if there were no thermal limits.
348 * Instead, we calculate the current power on the assumption that the
349 * immediate future will look like the immediate past.
350 *
351 * We use the current frequency and the average load since this
352 * function was last called.  In reality, there could have been
353 * multiple opps since this function was last called and that affects
354 * the load calculation.  While it's not perfectly accurate, this
355 * simplification is good enough and works.  REVISIT this, as more
356 * complex code may be needed if experiments show that it's not
357 * accurate enough.
358 *
359 * Return: 0 on success, -E* if getting the static power failed.
360 */
361static int cpufreq_get_requested_power(struct thermal_cooling_device *cdev,
362				       struct thermal_zone_device *tz,
363				       u32 *power)
364{
365	unsigned long freq;
366	int i = 0, cpu;
367	u32 total_load = 0;
368	struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata;
369	struct cpufreq_policy *policy = cpufreq_cdev->policy;
370	u32 *load_cpu = NULL;
371
372	freq = cpufreq_quick_get(policy->cpu);
373
374	if (trace_thermal_power_cpu_get_power_enabled()) {
375		u32 ncpus = cpumask_weight(policy->related_cpus);
376
377		load_cpu = kcalloc(ncpus, sizeof(*load_cpu), GFP_KERNEL);
378	}
379
380	for_each_cpu(cpu, policy->related_cpus) {
381		u32 load;
382
383		if (cpu_online(cpu))
384			load = get_load(cpufreq_cdev, cpu, i);
385		else
386			load = 0;
387
388		total_load += load;
389		if (load_cpu)
390			load_cpu[i] = load;
391
392		i++;
393	}
394
395	cpufreq_cdev->last_load = total_load;
396
397	*power = get_dynamic_power(cpufreq_cdev, freq);
398
399	if (load_cpu) {
400		trace_thermal_power_cpu_get_power(policy->related_cpus, freq,
401						  load_cpu, i, *power);
402
403		kfree(load_cpu);
404	}
405
406	return 0;
407}
408
409/**
410 * cpufreq_state2power() - convert a cpu cdev state to power consumed
411 * @cdev:	&thermal_cooling_device pointer
412 * @tz:		a valid thermal zone device pointer
413 * @state:	cooling device state to be converted
414 * @power:	pointer in which to store the resulting power
415 *
416 * Convert cooling device state @state into power consumption in
417 * milliwatts assuming 100% load.  Store the calculated power in
418 * @power.
419 *
420 * Return: 0 on success, -EINVAL if the cooling device state could not
421 * be converted into a frequency or other -E* if there was an error
422 * when calculating the static power.
423 */
424static int cpufreq_state2power(struct thermal_cooling_device *cdev,
425			       struct thermal_zone_device *tz,
426			       unsigned long state, u32 *power)
427{
428	unsigned int freq, num_cpus;
429	struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata;
430
431	/* Request state should be less than max_level */
432	if (WARN_ON(state > cpufreq_cdev->max_level))
433		return -EINVAL;
434
435	num_cpus = cpumask_weight(cpufreq_cdev->policy->cpus);
436
437	freq = cpufreq_cdev->freq_table[state].frequency;
438	*power = cpu_freq_to_power(cpufreq_cdev, freq) * num_cpus;
439
440	return 0;
441}
442
443/**
444 * cpufreq_power2state() - convert power to a cooling device state
445 * @cdev:	&thermal_cooling_device pointer
446 * @tz:		a valid thermal zone device pointer
447 * @power:	power in milliwatts to be converted
448 * @state:	pointer in which to store the resulting state
449 *
450 * Calculate a cooling device state for the cpus described by @cdev
451 * that would allow them to consume at most @power mW and store it in
452 * @state.  Note that this calculation depends on external factors
453 * such as the cpu load or the current static power.  Calling this
454 * function with the same power as input can yield different cooling
455 * device states depending on those external factors.
456 *
457 * Return: 0 on success, -ENODEV if no cpus are online or -EINVAL if
458 * the calculated frequency could not be converted to a valid state.
459 * The latter should not happen unless the frequencies available to
460 * cpufreq have changed since the initialization of the cpu cooling
461 * device.
462 */
463static int cpufreq_power2state(struct thermal_cooling_device *cdev,
464			       struct thermal_zone_device *tz, u32 power,
465			       unsigned long *state)
466{
467	unsigned int target_freq;
468	u32 last_load, normalised_power;
469	struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata;
470	struct cpufreq_policy *policy = cpufreq_cdev->policy;
471
472	last_load = cpufreq_cdev->last_load ?: 1;
473	normalised_power = (power * 100) / last_load;
474	target_freq = cpu_power_to_freq(cpufreq_cdev, normalised_power);
475
476	*state = get_level(cpufreq_cdev, target_freq);
477	trace_thermal_power_cpu_limit(policy->related_cpus, target_freq, *state,
478				      power);
479	return 0;
480}
481
482/* Bind cpufreq callbacks to thermal cooling device ops */
483
484static struct thermal_cooling_device_ops cpufreq_cooling_ops = {
485	.get_max_state = cpufreq_get_max_state,
486	.get_cur_state = cpufreq_get_cur_state,
487	.set_cur_state = cpufreq_set_cur_state,
488};
489
490static struct thermal_cooling_device_ops cpufreq_power_cooling_ops = {
491	.get_max_state		= cpufreq_get_max_state,
492	.get_cur_state		= cpufreq_get_cur_state,
493	.set_cur_state		= cpufreq_set_cur_state,
494	.get_requested_power	= cpufreq_get_requested_power,
495	.state2power		= cpufreq_state2power,
496	.power2state		= cpufreq_power2state,
497};
498
499static unsigned int find_next_max(struct cpufreq_frequency_table *table,
500				  unsigned int prev_max)
501{
502	struct cpufreq_frequency_table *pos;
503	unsigned int max = 0;
504
505	cpufreq_for_each_valid_entry(pos, table) {
506		if (pos->frequency > max && pos->frequency < prev_max)
507			max = pos->frequency;
508	}
509
510	return max;
511}
512
513/**
514 * __cpufreq_cooling_register - helper function to create cpufreq cooling device
515 * @np: a valid struct device_node to the cooling device device tree node
516 * @policy: cpufreq policy
517 * Normally this should be same as cpufreq policy->related_cpus.
518 * @capacitance: dynamic power coefficient for these cpus
519 *
520 * This interface function registers the cpufreq cooling device with the name
521 * "thermal-cpufreq-%x". This api can support multiple instances of cpufreq
522 * cooling devices. It also gives the opportunity to link the cooling device
523 * with a device tree node, in order to bind it via the thermal DT code.
524 *
525 * Return: a valid struct thermal_cooling_device pointer on success,
526 * on failure, it returns a corresponding ERR_PTR().
527 */
528static struct thermal_cooling_device *
529__cpufreq_cooling_register(struct device_node *np,
530			struct cpufreq_policy *policy, u32 capacitance)
531{
532	struct thermal_cooling_device *cdev;
533	struct cpufreq_cooling_device *cpufreq_cdev;
534	char dev_name[THERMAL_NAME_LENGTH];
535	unsigned int freq, i, num_cpus;
536	struct device *dev;
537	int ret;
538	struct thermal_cooling_device_ops *cooling_ops;
539
540	dev = get_cpu_device(policy->cpu);
541	if (unlikely(!dev)) {
542		pr_warn("No cpu device for cpu %d\n", policy->cpu);
543		return ERR_PTR(-ENODEV);
544	}
545
546
547	if (IS_ERR_OR_NULL(policy)) {
548		pr_err("%s: cpufreq policy isn't valid: %p\n", __func__, policy);
549		return ERR_PTR(-EINVAL);
550	}
551
552	i = cpufreq_table_count_valid_entries(policy);
553	if (!i) {
554		pr_debug("%s: CPUFreq table not found or has no valid entries\n",
555			 __func__);
556		return ERR_PTR(-ENODEV);
557	}
558
559	cpufreq_cdev = kzalloc(sizeof(*cpufreq_cdev), GFP_KERNEL);
560	if (!cpufreq_cdev)
561		return ERR_PTR(-ENOMEM);
562
563	cpufreq_cdev->policy = policy;
564	num_cpus = cpumask_weight(policy->related_cpus);
565	cpufreq_cdev->idle_time = kcalloc(num_cpus,
566					 sizeof(*cpufreq_cdev->idle_time),
567					 GFP_KERNEL);
568	if (!cpufreq_cdev->idle_time) {
569		cdev = ERR_PTR(-ENOMEM);
570		goto free_cdev;
571	}
572
573	/* max_level is an index, not a counter */
574	cpufreq_cdev->max_level = i - 1;
575
576	cpufreq_cdev->freq_table = kmalloc_array(i,
577					sizeof(*cpufreq_cdev->freq_table),
578					GFP_KERNEL);
579	if (!cpufreq_cdev->freq_table) {
580		cdev = ERR_PTR(-ENOMEM);
581		goto free_idle_time;
582	}
583
584	ret = ida_simple_get(&cpufreq_ida, 0, 0, GFP_KERNEL);
585	if (ret < 0) {
586		cdev = ERR_PTR(ret);
587		goto free_table;
588	}
589	cpufreq_cdev->id = ret;
590
591	snprintf(dev_name, sizeof(dev_name), "thermal-cpufreq-%d",
592		 cpufreq_cdev->id);
593
594	/* Fill freq-table in descending order of frequencies */
595	for (i = 0, freq = -1; i <= cpufreq_cdev->max_level; i++) {
596		freq = find_next_max(policy->freq_table, freq);
597		cpufreq_cdev->freq_table[i].frequency = freq;
598
599		/* Warn for duplicate entries */
600		if (!freq)
601			pr_warn("%s: table has duplicate entries\n", __func__);
602		else
603			pr_debug("%s: freq:%u KHz\n", __func__, freq);
604	}
605
606	if (capacitance) {
607		ret = update_freq_table(cpufreq_cdev, capacitance);
608		if (ret) {
609			cdev = ERR_PTR(ret);
610			goto remove_ida;
611		}
612
613		cooling_ops = &cpufreq_power_cooling_ops;
614	} else {
615		cooling_ops = &cpufreq_cooling_ops;
616	}
617
618	ret = freq_qos_add_request(&policy->constraints,
619				   &cpufreq_cdev->qos_req, FREQ_QOS_MAX,
620				   cpufreq_cdev->freq_table[0].frequency);
621	if (ret < 0) {
622		pr_err("%s: Failed to add freq constraint (%d)\n", __func__,
623		       ret);
624		cdev = ERR_PTR(ret);
625		goto remove_ida;
626	}
627
628	cdev = thermal_of_cooling_device_register(np, dev_name, cpufreq_cdev,
629						  cooling_ops);
630	if (IS_ERR(cdev))
631		goto remove_qos_req;
632
633	mutex_lock(&cooling_list_lock);
634	list_add(&cpufreq_cdev->node, &cpufreq_cdev_list);
635	mutex_unlock(&cooling_list_lock);
636
637	return cdev;
638
639remove_qos_req:
640	freq_qos_remove_request(&cpufreq_cdev->qos_req);
641remove_ida:
642	ida_simple_remove(&cpufreq_ida, cpufreq_cdev->id);
643free_table:
644	kfree(cpufreq_cdev->freq_table);
645free_idle_time:
646	kfree(cpufreq_cdev->idle_time);
647free_cdev:
648	kfree(cpufreq_cdev);
649	return cdev;
650}
651
652/**
653 * cpufreq_cooling_register - function to create cpufreq cooling device.
654 * @policy: cpufreq policy
655 *
656 * This interface function registers the cpufreq cooling device with the name
657 * "thermal-cpufreq-%x". This api can support multiple instances of cpufreq
658 * cooling devices.
659 *
660 * Return: a valid struct thermal_cooling_device pointer on success,
661 * on failure, it returns a corresponding ERR_PTR().
662 */
663struct thermal_cooling_device *
664cpufreq_cooling_register(struct cpufreq_policy *policy)
665{
666	return __cpufreq_cooling_register(NULL, policy, 0);
667}
668EXPORT_SYMBOL_GPL(cpufreq_cooling_register);
669
670/**
671 * of_cpufreq_cooling_register - function to create cpufreq cooling device.
672 * @policy: cpufreq policy
673 *
674 * This interface function registers the cpufreq cooling device with the name
675 * "thermal-cpufreq-%x". This api can support multiple instances of cpufreq
676 * cooling devices. Using this API, the cpufreq cooling device will be
677 * linked to the device tree node provided.
678 *
679 * Using this function, the cooling device will implement the power
680 * extensions by using a simple cpu power model.  The cpus must have
681 * registered their OPPs using the OPP library.
682 *
683 * It also takes into account, if property present in policy CPU node, the
684 * static power consumed by the cpu.
685 *
686 * Return: a valid struct thermal_cooling_device pointer on success,
687 * and NULL on failure.
688 */
689struct thermal_cooling_device *
690of_cpufreq_cooling_register(struct cpufreq_policy *policy)
691{
692	struct device_node *np = of_get_cpu_node(policy->cpu, NULL);
693	struct thermal_cooling_device *cdev = NULL;
694	u32 capacitance = 0;
695
696	if (!np) {
697		pr_err("cpu_cooling: OF node not available for cpu%d\n",
698		       policy->cpu);
699		return NULL;
700	}
701
702	if (of_find_property(np, "#cooling-cells", NULL)) {
703		of_property_read_u32(np, "dynamic-power-coefficient",
704				     &capacitance);
705
706		cdev = __cpufreq_cooling_register(np, policy, capacitance);
707		if (IS_ERR(cdev)) {
708			pr_err("cpu_cooling: cpu%d failed to register as cooling device: %ld\n",
709			       policy->cpu, PTR_ERR(cdev));
710			cdev = NULL;
711		}
712	}
713
714	of_node_put(np);
715	return cdev;
716}
717EXPORT_SYMBOL_GPL(of_cpufreq_cooling_register);
718
719/**
720 * cpufreq_cooling_unregister - function to remove cpufreq cooling device.
721 * @cdev: thermal cooling device pointer.
722 *
723 * This interface function unregisters the "thermal-cpufreq-%x" cooling device.
724 */
725void cpufreq_cooling_unregister(struct thermal_cooling_device *cdev)
726{
727	struct cpufreq_cooling_device *cpufreq_cdev;
728
729	if (!cdev)
730		return;
731
732	cpufreq_cdev = cdev->devdata;
733
734	mutex_lock(&cooling_list_lock);
735	list_del(&cpufreq_cdev->node);
736	mutex_unlock(&cooling_list_lock);
737
738	thermal_cooling_device_unregister(cdev);
739	freq_qos_remove_request(&cpufreq_cdev->qos_req);
740	ida_simple_remove(&cpufreq_ida, cpufreq_cdev->id);
741	kfree(cpufreq_cdev->idle_time);
742	kfree(cpufreq_cdev->freq_table);
743	kfree(cpufreq_cdev);
744}
745EXPORT_SYMBOL_GPL(cpufreq_cooling_unregister);