]>
Commit | Line | Data |
---|---|---|
1 | /* | |
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/pm_opp.h> | |
30 | #include <linux/slab.h> | |
31 | #include <linux/cpu.h> | |
32 | #include <linux/cpu_cooling.h> | |
33 | ||
34 | #include <trace/events/thermal.h> | |
35 | ||
36 | /* | |
37 | * Cooling state <-> CPUFreq frequency | |
38 | * | |
39 | * Cooling states are translated to frequencies throughout this driver and this | |
40 | * is the relation between them. | |
41 | * | |
42 | * Highest cooling state corresponds to lowest possible frequency. | |
43 | * | |
44 | * i.e. | |
45 | * level 0 --> 1st Max Freq | |
46 | * level 1 --> 2nd Max Freq | |
47 | * ... | |
48 | */ | |
49 | ||
50 | /** | |
51 | * struct power_table - frequency to power conversion | |
52 | * @frequency: frequency in KHz | |
53 | * @power: power in mW | |
54 | * | |
55 | * This structure is built when the cooling device registers and helps | |
56 | * in translating frequency to power and viceversa. | |
57 | */ | |
58 | struct power_table { | |
59 | u32 frequency; | |
60 | u32 power; | |
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 | * @cool_dev: thermal_cooling_device pointer to keep track of the | |
68 | * registered cooling device. | |
69 | * @cpufreq_state: integer value representing the current state of cpufreq | |
70 | * cooling devices. | |
71 | * @clipped_freq: integer value representing the absolute value of the clipped | |
72 | * frequency. | |
73 | * @max_level: maximum cooling level. One less than total number of valid | |
74 | * cpufreq frequencies. | |
75 | * @allowed_cpus: all the cpus involved for this cpufreq_cooling_device. | |
76 | * @node: list_head to link all cpufreq_cooling_device together. | |
77 | * @last_load: load measured by the latest call to cpufreq_get_requested_power() | |
78 | * @time_in_idle: previous reading of the absolute time that this cpu was idle | |
79 | * @time_in_idle_timestamp: wall time of the last invocation of | |
80 | * get_cpu_idle_time_us() | |
81 | * @dyn_power_table: array of struct power_table for frequency to power | |
82 | * conversion, sorted in ascending order. | |
83 | * @dyn_power_table_entries: number of entries in the @dyn_power_table array | |
84 | * @cpu_dev: the first cpu_device from @allowed_cpus that has OPPs registered | |
85 | * @plat_get_static_power: callback to calculate the static power | |
86 | * | |
87 | * This structure is required for keeping information of each registered | |
88 | * cpufreq_cooling_device. | |
89 | */ | |
90 | struct cpufreq_cooling_device { | |
91 | int id; | |
92 | struct thermal_cooling_device *cool_dev; | |
93 | unsigned int cpufreq_state; | |
94 | unsigned int clipped_freq; | |
95 | unsigned int max_level; | |
96 | unsigned int *freq_table; /* In descending order */ | |
97 | struct cpumask allowed_cpus; | |
98 | struct list_head node; | |
99 | u32 last_load; | |
100 | u64 *time_in_idle; | |
101 | u64 *time_in_idle_timestamp; | |
102 | struct power_table *dyn_power_table; | |
103 | int dyn_power_table_entries; | |
104 | struct device *cpu_dev; | |
105 | get_static_t plat_get_static_power; | |
106 | }; | |
107 | static DEFINE_IDR(cpufreq_idr); | |
108 | static DEFINE_MUTEX(cooling_cpufreq_lock); | |
109 | ||
110 | static unsigned int cpufreq_dev_count; | |
111 | ||
112 | static DEFINE_MUTEX(cooling_list_lock); | |
113 | static LIST_HEAD(cpufreq_dev_list); | |
114 | ||
115 | /** | |
116 | * get_idr - function to get a unique id. | |
117 | * @idr: struct idr * handle used to create a id. | |
118 | * @id: int * value generated by this function. | |
119 | * | |
120 | * This function will populate @id with an unique | |
121 | * id, using the idr API. | |
122 | * | |
123 | * Return: 0 on success, an error code on failure. | |
124 | */ | |
125 | static int get_idr(struct idr *idr, int *id) | |
126 | { | |
127 | int ret; | |
128 | ||
129 | mutex_lock(&cooling_cpufreq_lock); | |
130 | ret = idr_alloc(idr, NULL, 0, 0, GFP_KERNEL); | |
131 | mutex_unlock(&cooling_cpufreq_lock); | |
132 | if (unlikely(ret < 0)) | |
133 | return ret; | |
134 | *id = ret; | |
135 | ||
136 | return 0; | |
137 | } | |
138 | ||
139 | /** | |
140 | * release_idr - function to free the unique id. | |
141 | * @idr: struct idr * handle used for creating the id. | |
142 | * @id: int value representing the unique id. | |
143 | */ | |
144 | static void release_idr(struct idr *idr, int id) | |
145 | { | |
146 | mutex_lock(&cooling_cpufreq_lock); | |
147 | idr_remove(idr, id); | |
148 | mutex_unlock(&cooling_cpufreq_lock); | |
149 | } | |
150 | ||
151 | /* Below code defines functions to be used for cpufreq as cooling device */ | |
152 | ||
153 | /** | |
154 | * get_level: Find the level for a particular frequency | |
155 | * @cpufreq_dev: cpufreq_dev for which the property is required | |
156 | * @freq: Frequency | |
157 | * | |
158 | * Return: level on success, THERMAL_CSTATE_INVALID on error. | |
159 | */ | |
160 | static unsigned long get_level(struct cpufreq_cooling_device *cpufreq_dev, | |
161 | unsigned int freq) | |
162 | { | |
163 | unsigned long level; | |
164 | ||
165 | for (level = 0; level <= cpufreq_dev->max_level; level++) { | |
166 | if (freq == cpufreq_dev->freq_table[level]) | |
167 | return level; | |
168 | ||
169 | if (freq > cpufreq_dev->freq_table[level]) | |
170 | break; | |
171 | } | |
172 | ||
173 | return THERMAL_CSTATE_INVALID; | |
174 | } | |
175 | ||
176 | /** | |
177 | * cpufreq_cooling_get_level - for a given cpu, return the cooling level. | |
178 | * @cpu: cpu for which the level is required | |
179 | * @freq: the frequency of interest | |
180 | * | |
181 | * This function will match the cooling level corresponding to the | |
182 | * requested @freq and return it. | |
183 | * | |
184 | * Return: The matched cooling level on success or THERMAL_CSTATE_INVALID | |
185 | * otherwise. | |
186 | */ | |
187 | unsigned long cpufreq_cooling_get_level(unsigned int cpu, unsigned int freq) | |
188 | { | |
189 | struct cpufreq_cooling_device *cpufreq_dev; | |
190 | ||
191 | mutex_lock(&cooling_list_lock); | |
192 | list_for_each_entry(cpufreq_dev, &cpufreq_dev_list, node) { | |
193 | if (cpumask_test_cpu(cpu, &cpufreq_dev->allowed_cpus)) { | |
194 | mutex_unlock(&cooling_list_lock); | |
195 | return get_level(cpufreq_dev, freq); | |
196 | } | |
197 | } | |
198 | mutex_unlock(&cooling_list_lock); | |
199 | ||
200 | pr_err("%s: cpu:%d not part of any cooling device\n", __func__, cpu); | |
201 | return THERMAL_CSTATE_INVALID; | |
202 | } | |
203 | EXPORT_SYMBOL_GPL(cpufreq_cooling_get_level); | |
204 | ||
205 | /** | |
206 | * cpufreq_thermal_notifier - notifier callback for cpufreq policy change. | |
207 | * @nb: struct notifier_block * with callback info. | |
208 | * @event: value showing cpufreq event for which this function invoked. | |
209 | * @data: callback-specific data | |
210 | * | |
211 | * Callback to hijack the notification on cpufreq policy transition. | |
212 | * Every time there is a change in policy, we will intercept and | |
213 | * update the cpufreq policy with thermal constraints. | |
214 | * | |
215 | * Return: 0 (success) | |
216 | */ | |
217 | static int cpufreq_thermal_notifier(struct notifier_block *nb, | |
218 | unsigned long event, void *data) | |
219 | { | |
220 | struct cpufreq_policy *policy = data; | |
221 | unsigned long clipped_freq; | |
222 | struct cpufreq_cooling_device *cpufreq_dev; | |
223 | ||
224 | if (event != CPUFREQ_ADJUST) | |
225 | return NOTIFY_DONE; | |
226 | ||
227 | mutex_lock(&cooling_list_lock); | |
228 | list_for_each_entry(cpufreq_dev, &cpufreq_dev_list, node) { | |
229 | if (!cpumask_test_cpu(policy->cpu, &cpufreq_dev->allowed_cpus)) | |
230 | continue; | |
231 | ||
232 | /* | |
233 | * policy->max is the maximum allowed frequency defined by user | |
234 | * and clipped_freq is the maximum that thermal constraints | |
235 | * allow. | |
236 | * | |
237 | * If clipped_freq is lower than policy->max, then we need to | |
238 | * readjust policy->max. | |
239 | * | |
240 | * But, if clipped_freq is greater than policy->max, we don't | |
241 | * need to do anything. | |
242 | */ | |
243 | clipped_freq = cpufreq_dev->clipped_freq; | |
244 | ||
245 | if (policy->max > clipped_freq) | |
246 | cpufreq_verify_within_limits(policy, 0, clipped_freq); | |
247 | break; | |
248 | } | |
249 | mutex_unlock(&cooling_list_lock); | |
250 | ||
251 | return NOTIFY_OK; | |
252 | } | |
253 | ||
254 | /** | |
255 | * build_dyn_power_table() - create a dynamic power to frequency table | |
256 | * @cpufreq_device: the cpufreq cooling device in which to store the table | |
257 | * @capacitance: dynamic power coefficient for these cpus | |
258 | * | |
259 | * Build a dynamic power to frequency table for this cpu and store it | |
260 | * in @cpufreq_device. This table will be used in cpu_power_to_freq() and | |
261 | * cpu_freq_to_power() to convert between power and frequency | |
262 | * efficiently. Power is stored in mW, frequency in KHz. The | |
263 | * resulting table is in ascending order. | |
264 | * | |
265 | * Return: 0 on success, -EINVAL if there are no OPPs for any CPUs, | |
266 | * -ENOMEM if we run out of memory or -EAGAIN if an OPP was | |
267 | * added/enabled while the function was executing. | |
268 | */ | |
269 | static int build_dyn_power_table(struct cpufreq_cooling_device *cpufreq_device, | |
270 | u32 capacitance) | |
271 | { | |
272 | struct power_table *power_table; | |
273 | struct dev_pm_opp *opp; | |
274 | struct device *dev = NULL; | |
275 | int num_opps = 0, cpu, i, ret = 0; | |
276 | unsigned long freq; | |
277 | ||
278 | for_each_cpu(cpu, &cpufreq_device->allowed_cpus) { | |
279 | dev = get_cpu_device(cpu); | |
280 | if (!dev) { | |
281 | dev_warn(&cpufreq_device->cool_dev->device, | |
282 | "No cpu device for cpu %d\n", cpu); | |
283 | continue; | |
284 | } | |
285 | ||
286 | num_opps = dev_pm_opp_get_opp_count(dev); | |
287 | if (num_opps > 0) | |
288 | break; | |
289 | else if (num_opps < 0) | |
290 | return num_opps; | |
291 | } | |
292 | ||
293 | if (num_opps == 0) | |
294 | return -EINVAL; | |
295 | ||
296 | power_table = kcalloc(num_opps, sizeof(*power_table), GFP_KERNEL); | |
297 | if (!power_table) | |
298 | return -ENOMEM; | |
299 | ||
300 | for (freq = 0, i = 0; | |
301 | opp = dev_pm_opp_find_freq_ceil(dev, &freq), !IS_ERR(opp); | |
302 | freq++, i++) { | |
303 | u32 freq_mhz, voltage_mv; | |
304 | u64 power; | |
305 | ||
306 | if (i >= num_opps) { | |
307 | ret = -EAGAIN; | |
308 | goto free_power_table; | |
309 | } | |
310 | ||
311 | freq_mhz = freq / 1000000; | |
312 | voltage_mv = dev_pm_opp_get_voltage(opp) / 1000; | |
313 | dev_pm_opp_put(opp); | |
314 | ||
315 | /* | |
316 | * Do the multiplication with MHz and millivolt so as | |
317 | * to not overflow. | |
318 | */ | |
319 | power = (u64)capacitance * freq_mhz * voltage_mv * voltage_mv; | |
320 | do_div(power, 1000000000); | |
321 | ||
322 | /* frequency is stored in power_table in KHz */ | |
323 | power_table[i].frequency = freq / 1000; | |
324 | ||
325 | /* power is stored in mW */ | |
326 | power_table[i].power = power; | |
327 | } | |
328 | ||
329 | if (i != num_opps) { | |
330 | ret = PTR_ERR(opp); | |
331 | goto free_power_table; | |
332 | } | |
333 | ||
334 | cpufreq_device->cpu_dev = dev; | |
335 | cpufreq_device->dyn_power_table = power_table; | |
336 | cpufreq_device->dyn_power_table_entries = i; | |
337 | ||
338 | return 0; | |
339 | ||
340 | free_power_table: | |
341 | kfree(power_table); | |
342 | ||
343 | return ret; | |
344 | } | |
345 | ||
346 | static u32 cpu_freq_to_power(struct cpufreq_cooling_device *cpufreq_device, | |
347 | u32 freq) | |
348 | { | |
349 | int i; | |
350 | struct power_table *pt = cpufreq_device->dyn_power_table; | |
351 | ||
352 | for (i = 1; i < cpufreq_device->dyn_power_table_entries; i++) | |
353 | if (freq < pt[i].frequency) | |
354 | break; | |
355 | ||
356 | return pt[i - 1].power; | |
357 | } | |
358 | ||
359 | static u32 cpu_power_to_freq(struct cpufreq_cooling_device *cpufreq_device, | |
360 | u32 power) | |
361 | { | |
362 | int i; | |
363 | struct power_table *pt = cpufreq_device->dyn_power_table; | |
364 | ||
365 | for (i = 1; i < cpufreq_device->dyn_power_table_entries; i++) | |
366 | if (power < pt[i].power) | |
367 | break; | |
368 | ||
369 | return pt[i - 1].frequency; | |
370 | } | |
371 | ||
372 | /** | |
373 | * get_load() - get load for a cpu since last updated | |
374 | * @cpufreq_device: &struct cpufreq_cooling_device for this cpu | |
375 | * @cpu: cpu number | |
376 | * @cpu_idx: index of the cpu in cpufreq_device->allowed_cpus | |
377 | * | |
378 | * Return: The average load of cpu @cpu in percentage since this | |
379 | * function was last called. | |
380 | */ | |
381 | static u32 get_load(struct cpufreq_cooling_device *cpufreq_device, int cpu, | |
382 | int cpu_idx) | |
383 | { | |
384 | u32 load; | |
385 | u64 now, now_idle, delta_time, delta_idle; | |
386 | ||
387 | now_idle = get_cpu_idle_time(cpu, &now, 0); | |
388 | delta_idle = now_idle - cpufreq_device->time_in_idle[cpu_idx]; | |
389 | delta_time = now - cpufreq_device->time_in_idle_timestamp[cpu_idx]; | |
390 | ||
391 | if (delta_time <= delta_idle) | |
392 | load = 0; | |
393 | else | |
394 | load = div64_u64(100 * (delta_time - delta_idle), delta_time); | |
395 | ||
396 | cpufreq_device->time_in_idle[cpu_idx] = now_idle; | |
397 | cpufreq_device->time_in_idle_timestamp[cpu_idx] = now; | |
398 | ||
399 | return load; | |
400 | } | |
401 | ||
402 | /** | |
403 | * get_static_power() - calculate the static power consumed by the cpus | |
404 | * @cpufreq_device: struct &cpufreq_cooling_device for this cpu cdev | |
405 | * @tz: thermal zone device in which we're operating | |
406 | * @freq: frequency in KHz | |
407 | * @power: pointer in which to store the calculated static power | |
408 | * | |
409 | * Calculate the static power consumed by the cpus described by | |
410 | * @cpu_actor running at frequency @freq. This function relies on a | |
411 | * platform specific function that should have been provided when the | |
412 | * actor was registered. If it wasn't, the static power is assumed to | |
413 | * be negligible. The calculated static power is stored in @power. | |
414 | * | |
415 | * Return: 0 on success, -E* on failure. | |
416 | */ | |
417 | static int get_static_power(struct cpufreq_cooling_device *cpufreq_device, | |
418 | struct thermal_zone_device *tz, unsigned long freq, | |
419 | u32 *power) | |
420 | { | |
421 | struct dev_pm_opp *opp; | |
422 | unsigned long voltage; | |
423 | struct cpumask *cpumask = &cpufreq_device->allowed_cpus; | |
424 | unsigned long freq_hz = freq * 1000; | |
425 | ||
426 | if (!cpufreq_device->plat_get_static_power || | |
427 | !cpufreq_device->cpu_dev) { | |
428 | *power = 0; | |
429 | return 0; | |
430 | } | |
431 | ||
432 | opp = dev_pm_opp_find_freq_exact(cpufreq_device->cpu_dev, freq_hz, | |
433 | true); | |
434 | voltage = dev_pm_opp_get_voltage(opp); | |
435 | dev_pm_opp_put(opp); | |
436 | ||
437 | if (voltage == 0) { | |
438 | dev_warn_ratelimited(cpufreq_device->cpu_dev, | |
439 | "Failed to get voltage for frequency %lu: %ld\n", | |
440 | freq_hz, IS_ERR(opp) ? PTR_ERR(opp) : 0); | |
441 | return -EINVAL; | |
442 | } | |
443 | ||
444 | return cpufreq_device->plat_get_static_power(cpumask, tz->passive_delay, | |
445 | voltage, power); | |
446 | } | |
447 | ||
448 | /** | |
449 | * get_dynamic_power() - calculate the dynamic power | |
450 | * @cpufreq_device: &cpufreq_cooling_device for this cdev | |
451 | * @freq: current frequency | |
452 | * | |
453 | * Return: the dynamic power consumed by the cpus described by | |
454 | * @cpufreq_device. | |
455 | */ | |
456 | static u32 get_dynamic_power(struct cpufreq_cooling_device *cpufreq_device, | |
457 | unsigned long freq) | |
458 | { | |
459 | u32 raw_cpu_power; | |
460 | ||
461 | raw_cpu_power = cpu_freq_to_power(cpufreq_device, freq); | |
462 | return (raw_cpu_power * cpufreq_device->last_load) / 100; | |
463 | } | |
464 | ||
465 | /* cpufreq cooling device callback functions are defined below */ | |
466 | ||
467 | /** | |
468 | * cpufreq_get_max_state - callback function to get the max cooling state. | |
469 | * @cdev: thermal cooling device pointer. | |
470 | * @state: fill this variable with the max cooling state. | |
471 | * | |
472 | * Callback for the thermal cooling device to return the cpufreq | |
473 | * max cooling state. | |
474 | * | |
475 | * Return: 0 on success, an error code otherwise. | |
476 | */ | |
477 | static int cpufreq_get_max_state(struct thermal_cooling_device *cdev, | |
478 | unsigned long *state) | |
479 | { | |
480 | struct cpufreq_cooling_device *cpufreq_device = cdev->devdata; | |
481 | ||
482 | *state = cpufreq_device->max_level; | |
483 | return 0; | |
484 | } | |
485 | ||
486 | /** | |
487 | * cpufreq_get_cur_state - callback function to get the current cooling state. | |
488 | * @cdev: thermal cooling device pointer. | |
489 | * @state: fill this variable with the current cooling state. | |
490 | * | |
491 | * Callback for the thermal cooling device to return the cpufreq | |
492 | * current cooling state. | |
493 | * | |
494 | * Return: 0 on success, an error code otherwise. | |
495 | */ | |
496 | static int cpufreq_get_cur_state(struct thermal_cooling_device *cdev, | |
497 | unsigned long *state) | |
498 | { | |
499 | struct cpufreq_cooling_device *cpufreq_device = cdev->devdata; | |
500 | ||
501 | *state = cpufreq_device->cpufreq_state; | |
502 | ||
503 | return 0; | |
504 | } | |
505 | ||
506 | /** | |
507 | * cpufreq_set_cur_state - callback function to set the current cooling state. | |
508 | * @cdev: thermal cooling device pointer. | |
509 | * @state: set this variable to the current cooling state. | |
510 | * | |
511 | * Callback for the thermal cooling device to change the cpufreq | |
512 | * current cooling state. | |
513 | * | |
514 | * Return: 0 on success, an error code otherwise. | |
515 | */ | |
516 | static int cpufreq_set_cur_state(struct thermal_cooling_device *cdev, | |
517 | unsigned long state) | |
518 | { | |
519 | struct cpufreq_cooling_device *cpufreq_device = cdev->devdata; | |
520 | unsigned int cpu = cpumask_any(&cpufreq_device->allowed_cpus); | |
521 | unsigned int clip_freq; | |
522 | ||
523 | /* Request state should be less than max_level */ | |
524 | if (WARN_ON(state > cpufreq_device->max_level)) | |
525 | return -EINVAL; | |
526 | ||
527 | /* Check if the old cooling action is same as new cooling action */ | |
528 | if (cpufreq_device->cpufreq_state == state) | |
529 | return 0; | |
530 | ||
531 | clip_freq = cpufreq_device->freq_table[state]; | |
532 | cpufreq_device->cpufreq_state = state; | |
533 | cpufreq_device->clipped_freq = clip_freq; | |
534 | ||
535 | cpufreq_update_policy(cpu); | |
536 | ||
537 | return 0; | |
538 | } | |
539 | ||
540 | /** | |
541 | * cpufreq_get_requested_power() - get the current power | |
542 | * @cdev: &thermal_cooling_device pointer | |
543 | * @tz: a valid thermal zone device pointer | |
544 | * @power: pointer in which to store the resulting power | |
545 | * | |
546 | * Calculate the current power consumption of the cpus in milliwatts | |
547 | * and store it in @power. This function should actually calculate | |
548 | * the requested power, but it's hard to get the frequency that | |
549 | * cpufreq would have assigned if there were no thermal limits. | |
550 | * Instead, we calculate the current power on the assumption that the | |
551 | * immediate future will look like the immediate past. | |
552 | * | |
553 | * We use the current frequency and the average load since this | |
554 | * function was last called. In reality, there could have been | |
555 | * multiple opps since this function was last called and that affects | |
556 | * the load calculation. While it's not perfectly accurate, this | |
557 | * simplification is good enough and works. REVISIT this, as more | |
558 | * complex code may be needed if experiments show that it's not | |
559 | * accurate enough. | |
560 | * | |
561 | * Return: 0 on success, -E* if getting the static power failed. | |
562 | */ | |
563 | static int cpufreq_get_requested_power(struct thermal_cooling_device *cdev, | |
564 | struct thermal_zone_device *tz, | |
565 | u32 *power) | |
566 | { | |
567 | unsigned long freq; | |
568 | int i = 0, cpu, ret; | |
569 | u32 static_power, dynamic_power, total_load = 0; | |
570 | struct cpufreq_cooling_device *cpufreq_device = cdev->devdata; | |
571 | u32 *load_cpu = NULL; | |
572 | ||
573 | cpu = cpumask_any_and(&cpufreq_device->allowed_cpus, cpu_online_mask); | |
574 | ||
575 | /* | |
576 | * All the CPUs are offline, thus the requested power by | |
577 | * the cdev is 0 | |
578 | */ | |
579 | if (cpu >= nr_cpu_ids) { | |
580 | *power = 0; | |
581 | return 0; | |
582 | } | |
583 | ||
584 | freq = cpufreq_quick_get(cpu); | |
585 | ||
586 | if (trace_thermal_power_cpu_get_power_enabled()) { | |
587 | u32 ncpus = cpumask_weight(&cpufreq_device->allowed_cpus); | |
588 | ||
589 | load_cpu = kcalloc(ncpus, sizeof(*load_cpu), GFP_KERNEL); | |
590 | } | |
591 | ||
592 | for_each_cpu(cpu, &cpufreq_device->allowed_cpus) { | |
593 | u32 load; | |
594 | ||
595 | if (cpu_online(cpu)) | |
596 | load = get_load(cpufreq_device, cpu, i); | |
597 | else | |
598 | load = 0; | |
599 | ||
600 | total_load += load; | |
601 | if (trace_thermal_power_cpu_limit_enabled() && load_cpu) | |
602 | load_cpu[i] = load; | |
603 | ||
604 | i++; | |
605 | } | |
606 | ||
607 | cpufreq_device->last_load = total_load; | |
608 | ||
609 | dynamic_power = get_dynamic_power(cpufreq_device, freq); | |
610 | ret = get_static_power(cpufreq_device, tz, freq, &static_power); | |
611 | if (ret) { | |
612 | kfree(load_cpu); | |
613 | return ret; | |
614 | } | |
615 | ||
616 | if (load_cpu) { | |
617 | trace_thermal_power_cpu_get_power( | |
618 | &cpufreq_device->allowed_cpus, | |
619 | freq, load_cpu, i, dynamic_power, static_power); | |
620 | ||
621 | kfree(load_cpu); | |
622 | } | |
623 | ||
624 | *power = static_power + dynamic_power; | |
625 | return 0; | |
626 | } | |
627 | ||
628 | /** | |
629 | * cpufreq_state2power() - convert a cpu cdev state to power consumed | |
630 | * @cdev: &thermal_cooling_device pointer | |
631 | * @tz: a valid thermal zone device pointer | |
632 | * @state: cooling device state to be converted | |
633 | * @power: pointer in which to store the resulting power | |
634 | * | |
635 | * Convert cooling device state @state into power consumption in | |
636 | * milliwatts assuming 100% load. Store the calculated power in | |
637 | * @power. | |
638 | * | |
639 | * Return: 0 on success, -EINVAL if the cooling device state could not | |
640 | * be converted into a frequency or other -E* if there was an error | |
641 | * when calculating the static power. | |
642 | */ | |
643 | static int cpufreq_state2power(struct thermal_cooling_device *cdev, | |
644 | struct thermal_zone_device *tz, | |
645 | unsigned long state, u32 *power) | |
646 | { | |
647 | unsigned int freq, num_cpus; | |
648 | cpumask_t cpumask; | |
649 | u32 static_power, dynamic_power; | |
650 | int ret; | |
651 | struct cpufreq_cooling_device *cpufreq_device = cdev->devdata; | |
652 | ||
653 | cpumask_and(&cpumask, &cpufreq_device->allowed_cpus, cpu_online_mask); | |
654 | num_cpus = cpumask_weight(&cpumask); | |
655 | ||
656 | /* None of our cpus are online, so no power */ | |
657 | if (num_cpus == 0) { | |
658 | *power = 0; | |
659 | return 0; | |
660 | } | |
661 | ||
662 | freq = cpufreq_device->freq_table[state]; | |
663 | if (!freq) | |
664 | return -EINVAL; | |
665 | ||
666 | dynamic_power = cpu_freq_to_power(cpufreq_device, freq) * num_cpus; | |
667 | ret = get_static_power(cpufreq_device, tz, freq, &static_power); | |
668 | if (ret) | |
669 | return ret; | |
670 | ||
671 | *power = static_power + dynamic_power; | |
672 | return 0; | |
673 | } | |
674 | ||
675 | /** | |
676 | * cpufreq_power2state() - convert power to a cooling device state | |
677 | * @cdev: &thermal_cooling_device pointer | |
678 | * @tz: a valid thermal zone device pointer | |
679 | * @power: power in milliwatts to be converted | |
680 | * @state: pointer in which to store the resulting state | |
681 | * | |
682 | * Calculate a cooling device state for the cpus described by @cdev | |
683 | * that would allow them to consume at most @power mW and store it in | |
684 | * @state. Note that this calculation depends on external factors | |
685 | * such as the cpu load or the current static power. Calling this | |
686 | * function with the same power as input can yield different cooling | |
687 | * device states depending on those external factors. | |
688 | * | |
689 | * Return: 0 on success, -ENODEV if no cpus are online or -EINVAL if | |
690 | * the calculated frequency could not be converted to a valid state. | |
691 | * The latter should not happen unless the frequencies available to | |
692 | * cpufreq have changed since the initialization of the cpu cooling | |
693 | * device. | |
694 | */ | |
695 | static int cpufreq_power2state(struct thermal_cooling_device *cdev, | |
696 | struct thermal_zone_device *tz, u32 power, | |
697 | unsigned long *state) | |
698 | { | |
699 | unsigned int cpu, cur_freq, target_freq; | |
700 | int ret; | |
701 | s32 dyn_power; | |
702 | u32 last_load, normalised_power, static_power; | |
703 | struct cpufreq_cooling_device *cpufreq_device = cdev->devdata; | |
704 | ||
705 | cpu = cpumask_any_and(&cpufreq_device->allowed_cpus, cpu_online_mask); | |
706 | ||
707 | /* None of our cpus are online */ | |
708 | if (cpu >= nr_cpu_ids) | |
709 | return -ENODEV; | |
710 | ||
711 | cur_freq = cpufreq_quick_get(cpu); | |
712 | ret = get_static_power(cpufreq_device, tz, cur_freq, &static_power); | |
713 | if (ret) | |
714 | return ret; | |
715 | ||
716 | dyn_power = power - static_power; | |
717 | dyn_power = dyn_power > 0 ? dyn_power : 0; | |
718 | last_load = cpufreq_device->last_load ?: 1; | |
719 | normalised_power = (dyn_power * 100) / last_load; | |
720 | target_freq = cpu_power_to_freq(cpufreq_device, normalised_power); | |
721 | ||
722 | *state = cpufreq_cooling_get_level(cpu, target_freq); | |
723 | if (*state == THERMAL_CSTATE_INVALID) { | |
724 | dev_warn_ratelimited(&cdev->device, | |
725 | "Failed to convert %dKHz for cpu %d into a cdev state\n", | |
726 | target_freq, cpu); | |
727 | return -EINVAL; | |
728 | } | |
729 | ||
730 | trace_thermal_power_cpu_limit(&cpufreq_device->allowed_cpus, | |
731 | target_freq, *state, power); | |
732 | return 0; | |
733 | } | |
734 | ||
735 | /* Bind cpufreq callbacks to thermal cooling device ops */ | |
736 | ||
737 | static struct thermal_cooling_device_ops cpufreq_cooling_ops = { | |
738 | .get_max_state = cpufreq_get_max_state, | |
739 | .get_cur_state = cpufreq_get_cur_state, | |
740 | .set_cur_state = cpufreq_set_cur_state, | |
741 | }; | |
742 | ||
743 | static struct thermal_cooling_device_ops cpufreq_power_cooling_ops = { | |
744 | .get_max_state = cpufreq_get_max_state, | |
745 | .get_cur_state = cpufreq_get_cur_state, | |
746 | .set_cur_state = cpufreq_set_cur_state, | |
747 | .get_requested_power = cpufreq_get_requested_power, | |
748 | .state2power = cpufreq_state2power, | |
749 | .power2state = cpufreq_power2state, | |
750 | }; | |
751 | ||
752 | /* Notifier for cpufreq policy change */ | |
753 | static struct notifier_block thermal_cpufreq_notifier_block = { | |
754 | .notifier_call = cpufreq_thermal_notifier, | |
755 | }; | |
756 | ||
757 | static unsigned int find_next_max(struct cpufreq_frequency_table *table, | |
758 | unsigned int prev_max) | |
759 | { | |
760 | struct cpufreq_frequency_table *pos; | |
761 | unsigned int max = 0; | |
762 | ||
763 | cpufreq_for_each_valid_entry(pos, table) { | |
764 | if (pos->frequency > max && pos->frequency < prev_max) | |
765 | max = pos->frequency; | |
766 | } | |
767 | ||
768 | return max; | |
769 | } | |
770 | ||
771 | /** | |
772 | * __cpufreq_cooling_register - helper function to create cpufreq cooling device | |
773 | * @np: a valid struct device_node to the cooling device device tree node | |
774 | * @clip_cpus: cpumask of cpus where the frequency constraints will happen. | |
775 | * Normally this should be same as cpufreq policy->related_cpus. | |
776 | * @capacitance: dynamic power coefficient for these cpus | |
777 | * @plat_static_func: function to calculate the static power consumed by these | |
778 | * cpus (optional) | |
779 | * | |
780 | * This interface function registers the cpufreq cooling device with the name | |
781 | * "thermal-cpufreq-%x". This api can support multiple instances of cpufreq | |
782 | * cooling devices. It also gives the opportunity to link the cooling device | |
783 | * with a device tree node, in order to bind it via the thermal DT code. | |
784 | * | |
785 | * Return: a valid struct thermal_cooling_device pointer on success, | |
786 | * on failure, it returns a corresponding ERR_PTR(). | |
787 | */ | |
788 | static struct thermal_cooling_device * | |
789 | __cpufreq_cooling_register(struct device_node *np, | |
790 | const struct cpumask *clip_cpus, u32 capacitance, | |
791 | get_static_t plat_static_func) | |
792 | { | |
793 | struct cpufreq_policy *policy; | |
794 | struct thermal_cooling_device *cool_dev; | |
795 | struct cpufreq_cooling_device *cpufreq_dev; | |
796 | char dev_name[THERMAL_NAME_LENGTH]; | |
797 | struct cpufreq_frequency_table *pos, *table; | |
798 | struct cpumask temp_mask; | |
799 | unsigned int freq, i, num_cpus; | |
800 | int ret; | |
801 | struct thermal_cooling_device_ops *cooling_ops; | |
802 | ||
803 | cpumask_and(&temp_mask, clip_cpus, cpu_online_mask); | |
804 | policy = cpufreq_cpu_get(cpumask_first(&temp_mask)); | |
805 | if (!policy) { | |
806 | pr_debug("%s: CPUFreq policy not found\n", __func__); | |
807 | return ERR_PTR(-EPROBE_DEFER); | |
808 | } | |
809 | ||
810 | table = policy->freq_table; | |
811 | if (!table) { | |
812 | pr_debug("%s: CPUFreq table not found\n", __func__); | |
813 | cool_dev = ERR_PTR(-ENODEV); | |
814 | goto put_policy; | |
815 | } | |
816 | ||
817 | cpufreq_dev = kzalloc(sizeof(*cpufreq_dev), GFP_KERNEL); | |
818 | if (!cpufreq_dev) { | |
819 | cool_dev = ERR_PTR(-ENOMEM); | |
820 | goto put_policy; | |
821 | } | |
822 | ||
823 | num_cpus = cpumask_weight(clip_cpus); | |
824 | cpufreq_dev->time_in_idle = kcalloc(num_cpus, | |
825 | sizeof(*cpufreq_dev->time_in_idle), | |
826 | GFP_KERNEL); | |
827 | if (!cpufreq_dev->time_in_idle) { | |
828 | cool_dev = ERR_PTR(-ENOMEM); | |
829 | goto free_cdev; | |
830 | } | |
831 | ||
832 | cpufreq_dev->time_in_idle_timestamp = | |
833 | kcalloc(num_cpus, sizeof(*cpufreq_dev->time_in_idle_timestamp), | |
834 | GFP_KERNEL); | |
835 | if (!cpufreq_dev->time_in_idle_timestamp) { | |
836 | cool_dev = ERR_PTR(-ENOMEM); | |
837 | goto free_time_in_idle; | |
838 | } | |
839 | ||
840 | /* Find max levels */ | |
841 | cpufreq_for_each_valid_entry(pos, table) | |
842 | cpufreq_dev->max_level++; | |
843 | ||
844 | cpufreq_dev->freq_table = kmalloc(sizeof(*cpufreq_dev->freq_table) * | |
845 | cpufreq_dev->max_level, GFP_KERNEL); | |
846 | if (!cpufreq_dev->freq_table) { | |
847 | cool_dev = ERR_PTR(-ENOMEM); | |
848 | goto free_time_in_idle_timestamp; | |
849 | } | |
850 | ||
851 | /* max_level is an index, not a counter */ | |
852 | cpufreq_dev->max_level--; | |
853 | ||
854 | cpumask_copy(&cpufreq_dev->allowed_cpus, clip_cpus); | |
855 | ||
856 | if (capacitance) { | |
857 | cpufreq_dev->plat_get_static_power = plat_static_func; | |
858 | ||
859 | ret = build_dyn_power_table(cpufreq_dev, capacitance); | |
860 | if (ret) { | |
861 | cool_dev = ERR_PTR(ret); | |
862 | goto free_table; | |
863 | } | |
864 | ||
865 | cooling_ops = &cpufreq_power_cooling_ops; | |
866 | } else { | |
867 | cooling_ops = &cpufreq_cooling_ops; | |
868 | } | |
869 | ||
870 | ret = get_idr(&cpufreq_idr, &cpufreq_dev->id); | |
871 | if (ret) { | |
872 | cool_dev = ERR_PTR(ret); | |
873 | goto free_power_table; | |
874 | } | |
875 | ||
876 | /* Fill freq-table in descending order of frequencies */ | |
877 | for (i = 0, freq = -1; i <= cpufreq_dev->max_level; i++) { | |
878 | freq = find_next_max(table, freq); | |
879 | cpufreq_dev->freq_table[i] = freq; | |
880 | ||
881 | /* Warn for duplicate entries */ | |
882 | if (!freq) | |
883 | pr_warn("%s: table has duplicate entries\n", __func__); | |
884 | else | |
885 | pr_debug("%s: freq:%u KHz\n", __func__, freq); | |
886 | } | |
887 | ||
888 | snprintf(dev_name, sizeof(dev_name), "thermal-cpufreq-%d", | |
889 | cpufreq_dev->id); | |
890 | ||
891 | cool_dev = thermal_of_cooling_device_register(np, dev_name, cpufreq_dev, | |
892 | cooling_ops); | |
893 | if (IS_ERR(cool_dev)) | |
894 | goto remove_idr; | |
895 | ||
896 | cpufreq_dev->clipped_freq = cpufreq_dev->freq_table[0]; | |
897 | cpufreq_dev->cool_dev = cool_dev; | |
898 | ||
899 | mutex_lock(&cooling_cpufreq_lock); | |
900 | ||
901 | mutex_lock(&cooling_list_lock); | |
902 | list_add(&cpufreq_dev->node, &cpufreq_dev_list); | |
903 | mutex_unlock(&cooling_list_lock); | |
904 | ||
905 | /* Register the notifier for first cpufreq cooling device */ | |
906 | if (!cpufreq_dev_count++) | |
907 | cpufreq_register_notifier(&thermal_cpufreq_notifier_block, | |
908 | CPUFREQ_POLICY_NOTIFIER); | |
909 | mutex_unlock(&cooling_cpufreq_lock); | |
910 | ||
911 | goto put_policy; | |
912 | ||
913 | remove_idr: | |
914 | release_idr(&cpufreq_idr, cpufreq_dev->id); | |
915 | free_power_table: | |
916 | kfree(cpufreq_dev->dyn_power_table); | |
917 | free_table: | |
918 | kfree(cpufreq_dev->freq_table); | |
919 | free_time_in_idle_timestamp: | |
920 | kfree(cpufreq_dev->time_in_idle_timestamp); | |
921 | free_time_in_idle: | |
922 | kfree(cpufreq_dev->time_in_idle); | |
923 | free_cdev: | |
924 | kfree(cpufreq_dev); | |
925 | put_policy: | |
926 | cpufreq_cpu_put(policy); | |
927 | ||
928 | return cool_dev; | |
929 | } | |
930 | ||
931 | /** | |
932 | * cpufreq_cooling_register - function to create cpufreq cooling device. | |
933 | * @clip_cpus: cpumask of cpus where the frequency constraints will happen. | |
934 | * | |
935 | * This interface function registers the cpufreq cooling device with the name | |
936 | * "thermal-cpufreq-%x". This api can support multiple instances of cpufreq | |
937 | * cooling devices. | |
938 | * | |
939 | * Return: a valid struct thermal_cooling_device pointer on success, | |
940 | * on failure, it returns a corresponding ERR_PTR(). | |
941 | */ | |
942 | struct thermal_cooling_device * | |
943 | cpufreq_cooling_register(const struct cpumask *clip_cpus) | |
944 | { | |
945 | return __cpufreq_cooling_register(NULL, clip_cpus, 0, NULL); | |
946 | } | |
947 | EXPORT_SYMBOL_GPL(cpufreq_cooling_register); | |
948 | ||
949 | /** | |
950 | * of_cpufreq_cooling_register - function to create cpufreq cooling device. | |
951 | * @np: a valid struct device_node to the cooling device device tree node | |
952 | * @clip_cpus: cpumask of cpus where the frequency constraints will happen. | |
953 | * | |
954 | * This interface function registers the cpufreq cooling device with the name | |
955 | * "thermal-cpufreq-%x". This api can support multiple instances of cpufreq | |
956 | * cooling devices. Using this API, the cpufreq cooling device will be | |
957 | * linked to the device tree node provided. | |
958 | * | |
959 | * Return: a valid struct thermal_cooling_device pointer on success, | |
960 | * on failure, it returns a corresponding ERR_PTR(). | |
961 | */ | |
962 | struct thermal_cooling_device * | |
963 | of_cpufreq_cooling_register(struct device_node *np, | |
964 | const struct cpumask *clip_cpus) | |
965 | { | |
966 | if (!np) | |
967 | return ERR_PTR(-EINVAL); | |
968 | ||
969 | return __cpufreq_cooling_register(np, clip_cpus, 0, NULL); | |
970 | } | |
971 | EXPORT_SYMBOL_GPL(of_cpufreq_cooling_register); | |
972 | ||
973 | /** | |
974 | * cpufreq_power_cooling_register() - create cpufreq cooling device with power extensions | |
975 | * @clip_cpus: cpumask of cpus where the frequency constraints will happen | |
976 | * @capacitance: dynamic power coefficient for these cpus | |
977 | * @plat_static_func: function to calculate the static power consumed by these | |
978 | * cpus (optional) | |
979 | * | |
980 | * This interface function registers the cpufreq cooling device with | |
981 | * the name "thermal-cpufreq-%x". This api can support multiple | |
982 | * instances of cpufreq cooling devices. Using this function, the | |
983 | * cooling device will implement the power extensions by using a | |
984 | * simple cpu power model. The cpus must have registered their OPPs | |
985 | * using the OPP library. | |
986 | * | |
987 | * An optional @plat_static_func may be provided to calculate the | |
988 | * static power consumed by these cpus. If the platform's static | |
989 | * power consumption is unknown or negligible, make it NULL. | |
990 | * | |
991 | * Return: a valid struct thermal_cooling_device pointer on success, | |
992 | * on failure, it returns a corresponding ERR_PTR(). | |
993 | */ | |
994 | struct thermal_cooling_device * | |
995 | cpufreq_power_cooling_register(const struct cpumask *clip_cpus, u32 capacitance, | |
996 | get_static_t plat_static_func) | |
997 | { | |
998 | return __cpufreq_cooling_register(NULL, clip_cpus, capacitance, | |
999 | plat_static_func); | |
1000 | } | |
1001 | EXPORT_SYMBOL(cpufreq_power_cooling_register); | |
1002 | ||
1003 | /** | |
1004 | * of_cpufreq_power_cooling_register() - create cpufreq cooling device with power extensions | |
1005 | * @np: a valid struct device_node to the cooling device device tree node | |
1006 | * @clip_cpus: cpumask of cpus where the frequency constraints will happen | |
1007 | * @capacitance: dynamic power coefficient for these cpus | |
1008 | * @plat_static_func: function to calculate the static power consumed by these | |
1009 | * cpus (optional) | |
1010 | * | |
1011 | * This interface function registers the cpufreq cooling device with | |
1012 | * the name "thermal-cpufreq-%x". This api can support multiple | |
1013 | * instances of cpufreq cooling devices. Using this API, the cpufreq | |
1014 | * cooling device will be linked to the device tree node provided. | |
1015 | * Using this function, the cooling device will implement the power | |
1016 | * extensions by using a simple cpu power model. The cpus must have | |
1017 | * registered their OPPs using the OPP library. | |
1018 | * | |
1019 | * An optional @plat_static_func may be provided to calculate the | |
1020 | * static power consumed by these cpus. If the platform's static | |
1021 | * power consumption is unknown or negligible, make it NULL. | |
1022 | * | |
1023 | * Return: a valid struct thermal_cooling_device pointer on success, | |
1024 | * on failure, it returns a corresponding ERR_PTR(). | |
1025 | */ | |
1026 | struct thermal_cooling_device * | |
1027 | of_cpufreq_power_cooling_register(struct device_node *np, | |
1028 | const struct cpumask *clip_cpus, | |
1029 | u32 capacitance, | |
1030 | get_static_t plat_static_func) | |
1031 | { | |
1032 | if (!np) | |
1033 | return ERR_PTR(-EINVAL); | |
1034 | ||
1035 | return __cpufreq_cooling_register(np, clip_cpus, capacitance, | |
1036 | plat_static_func); | |
1037 | } | |
1038 | EXPORT_SYMBOL(of_cpufreq_power_cooling_register); | |
1039 | ||
1040 | /** | |
1041 | * cpufreq_cooling_unregister - function to remove cpufreq cooling device. | |
1042 | * @cdev: thermal cooling device pointer. | |
1043 | * | |
1044 | * This interface function unregisters the "thermal-cpufreq-%x" cooling device. | |
1045 | */ | |
1046 | void cpufreq_cooling_unregister(struct thermal_cooling_device *cdev) | |
1047 | { | |
1048 | struct cpufreq_cooling_device *cpufreq_dev; | |
1049 | ||
1050 | if (!cdev) | |
1051 | return; | |
1052 | ||
1053 | cpufreq_dev = cdev->devdata; | |
1054 | ||
1055 | /* Unregister the notifier for the last cpufreq cooling device */ | |
1056 | mutex_lock(&cooling_cpufreq_lock); | |
1057 | if (!--cpufreq_dev_count) | |
1058 | cpufreq_unregister_notifier(&thermal_cpufreq_notifier_block, | |
1059 | CPUFREQ_POLICY_NOTIFIER); | |
1060 | ||
1061 | mutex_lock(&cooling_list_lock); | |
1062 | list_del(&cpufreq_dev->node); | |
1063 | mutex_unlock(&cooling_list_lock); | |
1064 | ||
1065 | mutex_unlock(&cooling_cpufreq_lock); | |
1066 | ||
1067 | thermal_cooling_device_unregister(cpufreq_dev->cool_dev); | |
1068 | release_idr(&cpufreq_idr, cpufreq_dev->id); | |
1069 | kfree(cpufreq_dev->dyn_power_table); | |
1070 | kfree(cpufreq_dev->time_in_idle_timestamp); | |
1071 | kfree(cpufreq_dev->time_in_idle); | |
1072 | kfree(cpufreq_dev->freq_table); | |
1073 | kfree(cpufreq_dev); | |
1074 | } | |
1075 | EXPORT_SYMBOL_GPL(cpufreq_cooling_unregister); |