Commit | Line | Data |
---|---|---|
02361418 ADK |
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 | * | |
73904cbc VK |
7 | * Copyright (C) 2014 Viresh Kumar <viresh.kumar@linaro.org> |
8 | * | |
02361418 ADK |
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 | */ | |
02361418 ADK |
25 | #include <linux/module.h> |
26 | #include <linux/thermal.h> | |
02361418 ADK |
27 | #include <linux/cpufreq.h> |
28 | #include <linux/err.h> | |
ae606089 | 29 | #include <linux/idr.h> |
c36cf071 | 30 | #include <linux/pm_opp.h> |
02361418 ADK |
31 | #include <linux/slab.h> |
32 | #include <linux/cpu.h> | |
33 | #include <linux/cpu_cooling.h> | |
34 | ||
6828a471 JM |
35 | #include <trace/events/thermal.h> |
36 | ||
07d888d8 VK |
37 | /* |
38 | * Cooling state <-> CPUFreq frequency | |
39 | * | |
40 | * Cooling states are translated to frequencies throughout this driver and this | |
41 | * is the relation between them. | |
42 | * | |
43 | * Highest cooling state corresponds to lowest possible frequency. | |
44 | * | |
45 | * i.e. | |
46 | * level 0 --> 1st Max Freq | |
47 | * level 1 --> 2nd Max Freq | |
48 | * ... | |
49 | */ | |
50 | ||
c36cf071 | 51 | /** |
349d39dc | 52 | * struct freq_table - frequency table along with power entries |
c36cf071 JM |
53 | * @frequency: frequency in KHz |
54 | * @power: power in mW | |
55 | * | |
56 | * This structure is built when the cooling device registers and helps | |
349d39dc | 57 | * in translating frequency to power and vice versa. |
c36cf071 | 58 | */ |
349d39dc | 59 | struct freq_table { |
c36cf071 JM |
60 | u32 frequency; |
61 | u32 power; | |
62 | }; | |
63 | ||
81ee14da VK |
64 | /** |
65 | * struct time_in_idle - Idle time stats | |
66 | * @time: previous reading of the absolute time that this cpu was idle | |
67 | * @timestamp: wall time of the last invocation of get_cpu_idle_time_us() | |
68 | */ | |
69 | struct time_in_idle { | |
70 | u64 time; | |
71 | u64 timestamp; | |
72 | }; | |
73 | ||
02361418 | 74 | /** |
3b3c0748 | 75 | * struct cpufreq_cooling_device - data for cooling device with cpufreq |
02361418 ADK |
76 | * @id: unique integer value corresponding to each cpufreq_cooling_device |
77 | * registered. | |
d72b4015 | 78 | * @last_load: load measured by the latest call to cpufreq_get_requested_power() |
02361418 ADK |
79 | * @cpufreq_state: integer value representing the current state of cpufreq |
80 | * cooling devices. | |
59f0d218 | 81 | * @clipped_freq: integer value representing the absolute value of the clipped |
02361418 | 82 | * frequency. |
dcc6c7fd VK |
83 | * @max_level: maximum cooling level. One less than total number of valid |
84 | * cpufreq frequencies. | |
d72b4015 VK |
85 | * @freq_table: Freq table in descending order of frequencies |
86 | * @cdev: thermal_cooling_device pointer to keep track of the | |
87 | * registered cooling device. | |
88 | * @policy: cpufreq policy. | |
fc4de356 | 89 | * @node: list_head to link all cpufreq_cooling_device together. |
81ee14da | 90 | * @idle_time: idle time stats |
02361418 | 91 | * |
beca6053 VK |
92 | * This structure is required for keeping information of each registered |
93 | * cpufreq_cooling_device. | |
02361418 ADK |
94 | */ |
95 | struct cpufreq_cooling_device { | |
96 | int id; | |
d72b4015 | 97 | u32 last_load; |
02361418 | 98 | unsigned int cpufreq_state; |
59f0d218 | 99 | unsigned int clipped_freq; |
dcc6c7fd | 100 | unsigned int max_level; |
349d39dc | 101 | struct freq_table *freq_table; /* In descending order */ |
d72b4015 VK |
102 | struct thermal_cooling_device *cdev; |
103 | struct cpufreq_policy *policy; | |
2dcd851f | 104 | struct list_head node; |
81ee14da | 105 | struct time_in_idle *idle_time; |
02361418 | 106 | }; |
02361418 | 107 | |
fb8ea308 | 108 | static DEFINE_IDA(cpufreq_ida); |
02373d7c | 109 | static DEFINE_MUTEX(cooling_list_lock); |
1dea432a | 110 | static LIST_HEAD(cpufreq_cdev_list); |
02361418 | 111 | |
02361418 ADK |
112 | /* Below code defines functions to be used for cpufreq as cooling device */ |
113 | ||
114 | /** | |
4843c4a1 | 115 | * get_level: Find the level for a particular frequency |
1dea432a | 116 | * @cpufreq_cdev: cpufreq_cdev for which the property is required |
4843c4a1 | 117 | * @freq: Frequency |
82b9ee40 | 118 | * |
da27f69d | 119 | * Return: level corresponding to the frequency. |
02361418 | 120 | */ |
1dea432a | 121 | static unsigned long get_level(struct cpufreq_cooling_device *cpufreq_cdev, |
4843c4a1 | 122 | unsigned int freq) |
02361418 | 123 | { |
da27f69d | 124 | struct freq_table *freq_table = cpufreq_cdev->freq_table; |
4843c4a1 | 125 | unsigned long level; |
a116776f | 126 | |
da27f69d VK |
127 | for (level = 1; level <= cpufreq_cdev->max_level; level++) |
128 | if (freq > freq_table[level].frequency) | |
4843c4a1 | 129 | break; |
02361418 | 130 | |
da27f69d | 131 | return level - 1; |
fc35b35c ZR |
132 | } |
133 | ||
02361418 ADK |
134 | /** |
135 | * cpufreq_thermal_notifier - notifier callback for cpufreq policy change. | |
136 | * @nb: struct notifier_block * with callback info. | |
137 | * @event: value showing cpufreq event for which this function invoked. | |
138 | * @data: callback-specific data | |
bab30554 | 139 | * |
9746b6e7 | 140 | * Callback to hijack the notification on cpufreq policy transition. |
bab30554 EV |
141 | * Every time there is a change in policy, we will intercept and |
142 | * update the cpufreq policy with thermal constraints. | |
143 | * | |
144 | * Return: 0 (success) | |
02361418 ADK |
145 | */ |
146 | static int cpufreq_thermal_notifier(struct notifier_block *nb, | |
5fda7f68 | 147 | unsigned long event, void *data) |
02361418 ADK |
148 | { |
149 | struct cpufreq_policy *policy = data; | |
abcbcc25 | 150 | unsigned long clipped_freq; |
1dea432a | 151 | struct cpufreq_cooling_device *cpufreq_cdev; |
02361418 | 152 | |
a24af233 VK |
153 | if (event != CPUFREQ_ADJUST) |
154 | return NOTIFY_DONE; | |
02361418 | 155 | |
a24af233 | 156 | mutex_lock(&cooling_list_lock); |
1dea432a | 157 | list_for_each_entry(cpufreq_cdev, &cpufreq_cdev_list, node) { |
ba76dd9d VK |
158 | /* |
159 | * A new copy of the policy is sent to the notifier and can't | |
160 | * compare that directly. | |
161 | */ | |
162 | if (policy->cpu != cpufreq_cdev->policy->cpu) | |
a24af233 | 163 | continue; |
c36cf071 | 164 | |
1afb9c53 VK |
165 | /* |
166 | * policy->max is the maximum allowed frequency defined by user | |
167 | * and clipped_freq is the maximum that thermal constraints | |
168 | * allow. | |
169 | * | |
170 | * If clipped_freq is lower than policy->max, then we need to | |
171 | * readjust policy->max. | |
172 | * | |
173 | * But, if clipped_freq is greater than policy->max, we don't | |
174 | * need to do anything. | |
175 | */ | |
1dea432a | 176 | clipped_freq = cpufreq_cdev->clipped_freq; |
c36cf071 | 177 | |
1afb9c53 | 178 | if (policy->max > clipped_freq) |
abcbcc25 | 179 | cpufreq_verify_within_limits(policy, 0, clipped_freq); |
c36cf071 | 180 | break; |
c36cf071 | 181 | } |
a24af233 | 182 | mutex_unlock(&cooling_list_lock); |
c36cf071 JM |
183 | |
184 | return NOTIFY_OK; | |
185 | } | |
186 | ||
187 | /** | |
349d39dc VK |
188 | * update_freq_table() - Update the freq table with power numbers |
189 | * @cpufreq_cdev: the cpufreq cooling device in which to update the table | |
c36cf071 JM |
190 | * @capacitance: dynamic power coefficient for these cpus |
191 | * | |
349d39dc VK |
192 | * Update the freq table with power numbers. This table will be used in |
193 | * cpu_power_to_freq() and cpu_freq_to_power() to convert between power and | |
194 | * frequency efficiently. Power is stored in mW, frequency in KHz. The | |
195 | * resulting table is in descending order. | |
c36cf071 | 196 | * |
459ac375 | 197 | * Return: 0 on success, -EINVAL if there are no OPPs for any CPUs, |
349d39dc | 198 | * or -ENOMEM if we run out of memory. |
c36cf071 | 199 | */ |
349d39dc VK |
200 | static int update_freq_table(struct cpufreq_cooling_device *cpufreq_cdev, |
201 | u32 capacitance) | |
c36cf071 | 202 | { |
349d39dc | 203 | struct freq_table *freq_table = cpufreq_cdev->freq_table; |
c36cf071 JM |
204 | struct dev_pm_opp *opp; |
205 | struct device *dev = NULL; | |
349d39dc | 206 | int num_opps = 0, cpu = cpufreq_cdev->policy->cpu, i; |
c36cf071 | 207 | |
02bacb21 VK |
208 | dev = get_cpu_device(cpu); |
209 | if (unlikely(!dev)) { | |
210 | dev_warn(&cpufreq_cdev->cdev->device, | |
211 | "No cpu device for cpu %d\n", cpu); | |
212 | return -ENODEV; | |
c36cf071 | 213 | } |
02361418 | 214 | |
02bacb21 VK |
215 | num_opps = dev_pm_opp_get_opp_count(dev); |
216 | if (num_opps < 0) | |
217 | return num_opps; | |
218 | ||
349d39dc VK |
219 | /* |
220 | * The cpufreq table is also built from the OPP table and so the count | |
221 | * should match. | |
222 | */ | |
223 | if (num_opps != cpufreq_cdev->max_level + 1) { | |
224 | dev_warn(dev, "Number of OPPs not matching with max_levels\n"); | |
459ac375 | 225 | return -EINVAL; |
349d39dc | 226 | } |
02361418 | 227 | |
349d39dc VK |
228 | for (i = 0; i <= cpufreq_cdev->max_level; i++) { |
229 | unsigned long freq = freq_table[i].frequency * 1000; | |
230 | u32 freq_mhz = freq_table[i].frequency / 1000; | |
c36cf071 | 231 | u64 power; |
349d39dc | 232 | u32 voltage_mv; |
c36cf071 | 233 | |
349d39dc VK |
234 | /* |
235 | * Find ceil frequency as 'freq' may be slightly lower than OPP | |
236 | * freq due to truncation while converting to kHz. | |
237 | */ | |
238 | opp = dev_pm_opp_find_freq_ceil(dev, &freq); | |
239 | if (IS_ERR(opp)) { | |
240 | dev_err(dev, "failed to get opp for %lu frequency\n", | |
241 | freq); | |
242 | return -EINVAL; | |
459ac375 JM |
243 | } |
244 | ||
c36cf071 | 245 | voltage_mv = dev_pm_opp_get_voltage(opp) / 1000; |
8a31d9d9 | 246 | dev_pm_opp_put(opp); |
c36cf071 JM |
247 | |
248 | /* | |
249 | * Do the multiplication with MHz and millivolt so as | |
250 | * to not overflow. | |
251 | */ | |
252 | power = (u64)capacitance * freq_mhz * voltage_mv * voltage_mv; | |
253 | do_div(power, 1000000000); | |
254 | ||
c36cf071 | 255 | /* power is stored in mW */ |
349d39dc | 256 | freq_table[i].power = power; |
eba4f88d | 257 | } |
c36cf071 | 258 | |
459ac375 | 259 | return 0; |
c36cf071 JM |
260 | } |
261 | ||
1dea432a | 262 | static u32 cpu_freq_to_power(struct cpufreq_cooling_device *cpufreq_cdev, |
c36cf071 JM |
263 | u32 freq) |
264 | { | |
265 | int i; | |
349d39dc | 266 | struct freq_table *freq_table = cpufreq_cdev->freq_table; |
c36cf071 | 267 | |
349d39dc VK |
268 | for (i = 1; i <= cpufreq_cdev->max_level; i++) |
269 | if (freq > freq_table[i].frequency) | |
c36cf071 JM |
270 | break; |
271 | ||
349d39dc | 272 | return freq_table[i - 1].power; |
c36cf071 JM |
273 | } |
274 | ||
1dea432a | 275 | static u32 cpu_power_to_freq(struct cpufreq_cooling_device *cpufreq_cdev, |
c36cf071 JM |
276 | u32 power) |
277 | { | |
278 | int i; | |
349d39dc | 279 | struct freq_table *freq_table = cpufreq_cdev->freq_table; |
c36cf071 | 280 | |
349d39dc VK |
281 | for (i = 1; i <= cpufreq_cdev->max_level; i++) |
282 | if (power > freq_table[i].power) | |
c36cf071 JM |
283 | break; |
284 | ||
349d39dc | 285 | return freq_table[i - 1].frequency; |
c36cf071 JM |
286 | } |
287 | ||
288 | /** | |
289 | * get_load() - get load for a cpu since last updated | |
1dea432a | 290 | * @cpufreq_cdev: &struct cpufreq_cooling_device for this cpu |
c36cf071 | 291 | * @cpu: cpu number |
ba76dd9d | 292 | * @cpu_idx: index of the cpu in time_in_idle* |
c36cf071 JM |
293 | * |
294 | * Return: The average load of cpu @cpu in percentage since this | |
295 | * function was last called. | |
296 | */ | |
1dea432a | 297 | static u32 get_load(struct cpufreq_cooling_device *cpufreq_cdev, int cpu, |
a53b8394 | 298 | int cpu_idx) |
c36cf071 JM |
299 | { |
300 | u32 load; | |
301 | u64 now, now_idle, delta_time, delta_idle; | |
81ee14da | 302 | struct time_in_idle *idle_time = &cpufreq_cdev->idle_time[cpu_idx]; |
c36cf071 JM |
303 | |
304 | now_idle = get_cpu_idle_time(cpu, &now, 0); | |
81ee14da VK |
305 | delta_idle = now_idle - idle_time->time; |
306 | delta_time = now - idle_time->timestamp; | |
c36cf071 JM |
307 | |
308 | if (delta_time <= delta_idle) | |
309 | load = 0; | |
310 | else | |
311 | load = div64_u64(100 * (delta_time - delta_idle), delta_time); | |
312 | ||
81ee14da VK |
313 | idle_time->time = now_idle; |
314 | idle_time->timestamp = now; | |
c36cf071 JM |
315 | |
316 | return load; | |
317 | } | |
318 | ||
c36cf071 JM |
319 | /** |
320 | * get_dynamic_power() - calculate the dynamic power | |
1dea432a | 321 | * @cpufreq_cdev: &cpufreq_cooling_device for this cdev |
c36cf071 JM |
322 | * @freq: current frequency |
323 | * | |
324 | * Return: the dynamic power consumed by the cpus described by | |
1dea432a | 325 | * @cpufreq_cdev. |
c36cf071 | 326 | */ |
1dea432a | 327 | static u32 get_dynamic_power(struct cpufreq_cooling_device *cpufreq_cdev, |
c36cf071 JM |
328 | unsigned long freq) |
329 | { | |
330 | u32 raw_cpu_power; | |
331 | ||
1dea432a VK |
332 | raw_cpu_power = cpu_freq_to_power(cpufreq_cdev, freq); |
333 | return (raw_cpu_power * cpufreq_cdev->last_load) / 100; | |
02361418 ADK |
334 | } |
335 | ||
1b9e3526 | 336 | /* cpufreq cooling device callback functions are defined below */ |
02361418 ADK |
337 | |
338 | /** | |
339 | * cpufreq_get_max_state - callback function to get the max cooling state. | |
340 | * @cdev: thermal cooling device pointer. | |
341 | * @state: fill this variable with the max cooling state. | |
62c00421 EV |
342 | * |
343 | * Callback for the thermal cooling device to return the cpufreq | |
344 | * max cooling state. | |
345 | * | |
346 | * Return: 0 on success, an error code otherwise. | |
02361418 ADK |
347 | */ |
348 | static int cpufreq_get_max_state(struct thermal_cooling_device *cdev, | |
349 | unsigned long *state) | |
350 | { | |
1dea432a | 351 | struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata; |
9c51b05a | 352 | |
1dea432a | 353 | *state = cpufreq_cdev->max_level; |
dcc6c7fd | 354 | return 0; |
02361418 ADK |
355 | } |
356 | ||
357 | /** | |
358 | * cpufreq_get_cur_state - callback function to get the current cooling state. | |
359 | * @cdev: thermal cooling device pointer. | |
360 | * @state: fill this variable with the current cooling state. | |
3672552d EV |
361 | * |
362 | * Callback for the thermal cooling device to return the cpufreq | |
363 | * current cooling state. | |
364 | * | |
365 | * Return: 0 on success, an error code otherwise. | |
02361418 ADK |
366 | */ |
367 | static int cpufreq_get_cur_state(struct thermal_cooling_device *cdev, | |
368 | unsigned long *state) | |
369 | { | |
1dea432a | 370 | struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata; |
02361418 | 371 | |
1dea432a | 372 | *state = cpufreq_cdev->cpufreq_state; |
79491e53 | 373 | |
160b7d80 | 374 | return 0; |
02361418 ADK |
375 | } |
376 | ||
377 | /** | |
378 | * cpufreq_set_cur_state - callback function to set the current cooling state. | |
379 | * @cdev: thermal cooling device pointer. | |
380 | * @state: set this variable to the current cooling state. | |
56e05fdb EV |
381 | * |
382 | * Callback for the thermal cooling device to change the cpufreq | |
383 | * current cooling state. | |
384 | * | |
385 | * Return: 0 on success, an error code otherwise. | |
02361418 ADK |
386 | */ |
387 | static int cpufreq_set_cur_state(struct thermal_cooling_device *cdev, | |
388 | unsigned long state) | |
389 | { | |
1dea432a | 390 | struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata; |
5194fe46 | 391 | unsigned int clip_freq; |
4843c4a1 VK |
392 | |
393 | /* Request state should be less than max_level */ | |
1dea432a | 394 | if (WARN_ON(state > cpufreq_cdev->max_level)) |
4843c4a1 | 395 | return -EINVAL; |
5194fe46 VK |
396 | |
397 | /* Check if the old cooling action is same as new cooling action */ | |
1dea432a | 398 | if (cpufreq_cdev->cpufreq_state == state) |
5194fe46 | 399 | return 0; |
02361418 | 400 | |
349d39dc | 401 | clip_freq = cpufreq_cdev->freq_table[state].frequency; |
1dea432a VK |
402 | cpufreq_cdev->cpufreq_state = state; |
403 | cpufreq_cdev->clipped_freq = clip_freq; | |
5194fe46 | 404 | |
ba76dd9d | 405 | cpufreq_update_policy(cpufreq_cdev->policy->cpu); |
5194fe46 VK |
406 | |
407 | return 0; | |
02361418 ADK |
408 | } |
409 | ||
c36cf071 JM |
410 | /** |
411 | * cpufreq_get_requested_power() - get the current power | |
412 | * @cdev: &thermal_cooling_device pointer | |
413 | * @tz: a valid thermal zone device pointer | |
414 | * @power: pointer in which to store the resulting power | |
415 | * | |
416 | * Calculate the current power consumption of the cpus in milliwatts | |
417 | * and store it in @power. This function should actually calculate | |
418 | * the requested power, but it's hard to get the frequency that | |
419 | * cpufreq would have assigned if there were no thermal limits. | |
420 | * Instead, we calculate the current power on the assumption that the | |
421 | * immediate future will look like the immediate past. | |
422 | * | |
423 | * We use the current frequency and the average load since this | |
424 | * function was last called. In reality, there could have been | |
425 | * multiple opps since this function was last called and that affects | |
426 | * the load calculation. While it's not perfectly accurate, this | |
427 | * simplification is good enough and works. REVISIT this, as more | |
428 | * complex code may be needed if experiments show that it's not | |
429 | * accurate enough. | |
430 | * | |
431 | * Return: 0 on success, -E* if getting the static power failed. | |
432 | */ | |
433 | static int cpufreq_get_requested_power(struct thermal_cooling_device *cdev, | |
434 | struct thermal_zone_device *tz, | |
435 | u32 *power) | |
436 | { | |
437 | unsigned long freq; | |
84fe2cab VK |
438 | int i = 0, cpu; |
439 | u32 total_load = 0; | |
1dea432a | 440 | struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata; |
ba76dd9d | 441 | struct cpufreq_policy *policy = cpufreq_cdev->policy; |
6828a471 | 442 | u32 *load_cpu = NULL; |
c36cf071 | 443 | |
ba76dd9d | 444 | freq = cpufreq_quick_get(policy->cpu); |
c36cf071 | 445 | |
6828a471 | 446 | if (trace_thermal_power_cpu_get_power_enabled()) { |
ba76dd9d | 447 | u32 ncpus = cpumask_weight(policy->related_cpus); |
6828a471 | 448 | |
a71544cd | 449 | load_cpu = kcalloc(ncpus, sizeof(*load_cpu), GFP_KERNEL); |
6828a471 JM |
450 | } |
451 | ||
ba76dd9d | 452 | for_each_cpu(cpu, policy->related_cpus) { |
c36cf071 JM |
453 | u32 load; |
454 | ||
455 | if (cpu_online(cpu)) | |
1dea432a | 456 | load = get_load(cpufreq_cdev, cpu, i); |
c36cf071 JM |
457 | else |
458 | load = 0; | |
459 | ||
460 | total_load += load; | |
6828a471 JM |
461 | if (trace_thermal_power_cpu_limit_enabled() && load_cpu) |
462 | load_cpu[i] = load; | |
463 | ||
464 | i++; | |
c36cf071 JM |
465 | } |
466 | ||
1dea432a | 467 | cpufreq_cdev->last_load = total_load; |
c36cf071 | 468 | |
84fe2cab | 469 | *power = get_dynamic_power(cpufreq_cdev, freq); |
6828a471 JM |
470 | |
471 | if (load_cpu) { | |
ba76dd9d | 472 | trace_thermal_power_cpu_get_power(policy->related_cpus, freq, |
84fe2cab | 473 | load_cpu, i, *power); |
6828a471 | 474 | |
a71544cd | 475 | kfree(load_cpu); |
6828a471 | 476 | } |
c36cf071 | 477 | |
c36cf071 JM |
478 | return 0; |
479 | } | |
480 | ||
481 | /** | |
482 | * cpufreq_state2power() - convert a cpu cdev state to power consumed | |
483 | * @cdev: &thermal_cooling_device pointer | |
484 | * @tz: a valid thermal zone device pointer | |
485 | * @state: cooling device state to be converted | |
486 | * @power: pointer in which to store the resulting power | |
487 | * | |
488 | * Convert cooling device state @state into power consumption in | |
489 | * milliwatts assuming 100% load. Store the calculated power in | |
490 | * @power. | |
491 | * | |
492 | * Return: 0 on success, -EINVAL if the cooling device state could not | |
493 | * be converted into a frequency or other -E* if there was an error | |
494 | * when calculating the static power. | |
495 | */ | |
496 | static int cpufreq_state2power(struct thermal_cooling_device *cdev, | |
497 | struct thermal_zone_device *tz, | |
498 | unsigned long state, u32 *power) | |
499 | { | |
500 | unsigned int freq, num_cpus; | |
1dea432a | 501 | struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata; |
c36cf071 | 502 | |
cb1b6318 VK |
503 | /* Request state should be less than max_level */ |
504 | if (WARN_ON(state > cpufreq_cdev->max_level)) | |
505 | return -EINVAL; | |
506 | ||
ba76dd9d | 507 | num_cpus = cpumask_weight(cpufreq_cdev->policy->cpus); |
c36cf071 | 508 | |
349d39dc | 509 | freq = cpufreq_cdev->freq_table[state].frequency; |
84fe2cab | 510 | *power = cpu_freq_to_power(cpufreq_cdev, freq) * num_cpus; |
c36cf071 | 511 | |
84fe2cab | 512 | return 0; |
c36cf071 JM |
513 | } |
514 | ||
515 | /** | |
516 | * cpufreq_power2state() - convert power to a cooling device state | |
517 | * @cdev: &thermal_cooling_device pointer | |
518 | * @tz: a valid thermal zone device pointer | |
519 | * @power: power in milliwatts to be converted | |
520 | * @state: pointer in which to store the resulting state | |
521 | * | |
522 | * Calculate a cooling device state for the cpus described by @cdev | |
523 | * that would allow them to consume at most @power mW and store it in | |
524 | * @state. Note that this calculation depends on external factors | |
525 | * such as the cpu load or the current static power. Calling this | |
526 | * function with the same power as input can yield different cooling | |
527 | * device states depending on those external factors. | |
528 | * | |
529 | * Return: 0 on success, -ENODEV if no cpus are online or -EINVAL if | |
530 | * the calculated frequency could not be converted to a valid state. | |
531 | * The latter should not happen unless the frequencies available to | |
532 | * cpufreq have changed since the initialization of the cpu cooling | |
533 | * device. | |
534 | */ | |
535 | static int cpufreq_power2state(struct thermal_cooling_device *cdev, | |
536 | struct thermal_zone_device *tz, u32 power, | |
537 | unsigned long *state) | |
538 | { | |
ba76dd9d | 539 | unsigned int cur_freq, target_freq; |
84fe2cab | 540 | u32 last_load, normalised_power; |
1dea432a | 541 | struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata; |
ba76dd9d | 542 | struct cpufreq_policy *policy = cpufreq_cdev->policy; |
c36cf071 | 543 | |
ba76dd9d | 544 | cur_freq = cpufreq_quick_get(policy->cpu); |
84fe2cab | 545 | power = power > 0 ? power : 0; |
1dea432a | 546 | last_load = cpufreq_cdev->last_load ?: 1; |
84fe2cab | 547 | normalised_power = (power * 100) / last_load; |
1dea432a | 548 | target_freq = cpu_power_to_freq(cpufreq_cdev, normalised_power); |
c36cf071 | 549 | |
3e08b2df | 550 | *state = get_level(cpufreq_cdev, target_freq); |
ba76dd9d VK |
551 | trace_thermal_power_cpu_limit(policy->related_cpus, target_freq, *state, |
552 | power); | |
c36cf071 JM |
553 | return 0; |
554 | } | |
555 | ||
02361418 | 556 | /* Bind cpufreq callbacks to thermal cooling device ops */ |
a305a438 | 557 | |
c36cf071 | 558 | static struct thermal_cooling_device_ops cpufreq_cooling_ops = { |
02361418 ADK |
559 | .get_max_state = cpufreq_get_max_state, |
560 | .get_cur_state = cpufreq_get_cur_state, | |
561 | .set_cur_state = cpufreq_set_cur_state, | |
562 | }; | |
563 | ||
a305a438 BJ |
564 | static struct thermal_cooling_device_ops cpufreq_power_cooling_ops = { |
565 | .get_max_state = cpufreq_get_max_state, | |
566 | .get_cur_state = cpufreq_get_cur_state, | |
567 | .set_cur_state = cpufreq_set_cur_state, | |
568 | .get_requested_power = cpufreq_get_requested_power, | |
569 | .state2power = cpufreq_state2power, | |
570 | .power2state = cpufreq_power2state, | |
571 | }; | |
572 | ||
02361418 ADK |
573 | /* Notifier for cpufreq policy change */ |
574 | static struct notifier_block thermal_cpufreq_notifier_block = { | |
575 | .notifier_call = cpufreq_thermal_notifier, | |
576 | }; | |
577 | ||
f6859014 VK |
578 | static unsigned int find_next_max(struct cpufreq_frequency_table *table, |
579 | unsigned int prev_max) | |
580 | { | |
581 | struct cpufreq_frequency_table *pos; | |
582 | unsigned int max = 0; | |
583 | ||
584 | cpufreq_for_each_valid_entry(pos, table) { | |
585 | if (pos->frequency > max && pos->frequency < prev_max) | |
586 | max = pos->frequency; | |
587 | } | |
588 | ||
589 | return max; | |
590 | } | |
591 | ||
02361418 | 592 | /** |
39d99cff EV |
593 | * __cpufreq_cooling_register - helper function to create cpufreq cooling device |
594 | * @np: a valid struct device_node to the cooling device device tree node | |
4d753aa7 | 595 | * @policy: cpufreq policy |
405fb825 | 596 | * Normally this should be same as cpufreq policy->related_cpus. |
c36cf071 | 597 | * @capacitance: dynamic power coefficient for these cpus |
12cb08ba EV |
598 | * |
599 | * This interface function registers the cpufreq cooling device with the name | |
600 | * "thermal-cpufreq-%x". This api can support multiple instances of cpufreq | |
39d99cff EV |
601 | * cooling devices. It also gives the opportunity to link the cooling device |
602 | * with a device tree node, in order to bind it via the thermal DT code. | |
12cb08ba EV |
603 | * |
604 | * Return: a valid struct thermal_cooling_device pointer on success, | |
605 | * on failure, it returns a corresponding ERR_PTR(). | |
02361418 | 606 | */ |
39d99cff EV |
607 | static struct thermal_cooling_device * |
608 | __cpufreq_cooling_register(struct device_node *np, | |
84fe2cab | 609 | struct cpufreq_policy *policy, u32 capacitance) |
02361418 | 610 | { |
04bdbdf9 | 611 | struct thermal_cooling_device *cdev; |
1dea432a | 612 | struct cpufreq_cooling_device *cpufreq_cdev; |
02361418 | 613 | char dev_name[THERMAL_NAME_LENGTH]; |
c36cf071 | 614 | unsigned int freq, i, num_cpus; |
405fb825 | 615 | int ret; |
a305a438 | 616 | struct thermal_cooling_device_ops *cooling_ops; |
088db931 | 617 | bool first; |
02361418 | 618 | |
4d753aa7 | 619 | if (IS_ERR_OR_NULL(policy)) { |
b2fd708f | 620 | pr_err("%s: cpufreq policy isn't valid: %p\n", __func__, policy); |
4d753aa7 | 621 | return ERR_PTR(-EINVAL); |
f8bfc116 VK |
622 | } |
623 | ||
55d85293 VK |
624 | i = cpufreq_table_count_valid_entries(policy); |
625 | if (!i) { | |
626 | pr_debug("%s: CPUFreq table not found or has no valid entries\n", | |
627 | __func__); | |
4d753aa7 | 628 | return ERR_PTR(-ENODEV); |
02361418 | 629 | } |
0f1be51c | 630 | |
1dea432a | 631 | cpufreq_cdev = kzalloc(sizeof(*cpufreq_cdev), GFP_KERNEL); |
4d753aa7 VK |
632 | if (!cpufreq_cdev) |
633 | return ERR_PTR(-ENOMEM); | |
02361418 | 634 | |
b12b6519 | 635 | cpufreq_cdev->policy = policy; |
4d753aa7 | 636 | num_cpus = cpumask_weight(policy->related_cpus); |
81ee14da VK |
637 | cpufreq_cdev->idle_time = kcalloc(num_cpus, |
638 | sizeof(*cpufreq_cdev->idle_time), | |
639 | GFP_KERNEL); | |
640 | if (!cpufreq_cdev->idle_time) { | |
04bdbdf9 | 641 | cdev = ERR_PTR(-ENOMEM); |
c36cf071 JM |
642 | goto free_cdev; |
643 | } | |
644 | ||
55d85293 VK |
645 | /* max_level is an index, not a counter */ |
646 | cpufreq_cdev->max_level = i - 1; | |
dcc6c7fd | 647 | |
f19b1a17 VK |
648 | cpufreq_cdev->freq_table = kmalloc_array(i, |
649 | sizeof(*cpufreq_cdev->freq_table), | |
650 | GFP_KERNEL); | |
1dea432a | 651 | if (!cpufreq_cdev->freq_table) { |
04bdbdf9 | 652 | cdev = ERR_PTR(-ENOMEM); |
81ee14da | 653 | goto free_idle_time; |
f6859014 VK |
654 | } |
655 | ||
ae606089 MW |
656 | ret = ida_simple_get(&cpufreq_ida, 0, 0, GFP_KERNEL); |
657 | if (ret < 0) { | |
04bdbdf9 | 658 | cdev = ERR_PTR(ret); |
349d39dc | 659 | goto free_table; |
02361418 | 660 | } |
1dea432a | 661 | cpufreq_cdev->id = ret; |
02361418 | 662 | |
349d39dc VK |
663 | snprintf(dev_name, sizeof(dev_name), "thermal-cpufreq-%d", |
664 | cpufreq_cdev->id); | |
665 | ||
f6859014 | 666 | /* Fill freq-table in descending order of frequencies */ |
1dea432a | 667 | for (i = 0, freq = -1; i <= cpufreq_cdev->max_level; i++) { |
55d85293 | 668 | freq = find_next_max(policy->freq_table, freq); |
349d39dc | 669 | cpufreq_cdev->freq_table[i].frequency = freq; |
f6859014 VK |
670 | |
671 | /* Warn for duplicate entries */ | |
672 | if (!freq) | |
673 | pr_warn("%s: table has duplicate entries\n", __func__); | |
674 | else | |
675 | pr_debug("%s: freq:%u KHz\n", __func__, freq); | |
02361418 | 676 | } |
f6859014 | 677 | |
349d39dc | 678 | if (capacitance) { |
349d39dc VK |
679 | ret = update_freq_table(cpufreq_cdev, capacitance); |
680 | if (ret) { | |
681 | cdev = ERR_PTR(ret); | |
682 | goto remove_ida; | |
683 | } | |
684 | ||
685 | cooling_ops = &cpufreq_power_cooling_ops; | |
686 | } else { | |
687 | cooling_ops = &cpufreq_cooling_ops; | |
688 | } | |
f840ab18 | 689 | |
04bdbdf9 VK |
690 | cdev = thermal_of_cooling_device_register(np, dev_name, cpufreq_cdev, |
691 | cooling_ops); | |
692 | if (IS_ERR(cdev)) | |
ae606089 | 693 | goto remove_ida; |
f840ab18 | 694 | |
349d39dc | 695 | cpufreq_cdev->clipped_freq = cpufreq_cdev->freq_table[0].frequency; |
04bdbdf9 | 696 | cpufreq_cdev->cdev = cdev; |
92e615ec | 697 | |
02373d7c | 698 | mutex_lock(&cooling_list_lock); |
088db931 | 699 | /* Register the notifier for first cpufreq cooling device */ |
1dea432a VK |
700 | first = list_empty(&cpufreq_cdev_list); |
701 | list_add(&cpufreq_cdev->node, &cpufreq_cdev_list); | |
088db931 | 702 | mutex_unlock(&cooling_list_lock); |
02373d7c | 703 | |
088db931 | 704 | if (first) |
02361418 | 705 | cpufreq_register_notifier(&thermal_cpufreq_notifier_block, |
5fda7f68 | 706 | CPUFREQ_POLICY_NOTIFIER); |
79491e53 | 707 | |
4d753aa7 | 708 | return cdev; |
730abe06 | 709 | |
ae606089 | 710 | remove_ida: |
1dea432a | 711 | ida_simple_remove(&cpufreq_ida, cpufreq_cdev->id); |
f6859014 | 712 | free_table: |
1dea432a | 713 | kfree(cpufreq_cdev->freq_table); |
81ee14da VK |
714 | free_idle_time: |
715 | kfree(cpufreq_cdev->idle_time); | |
730abe06 | 716 | free_cdev: |
1dea432a | 717 | kfree(cpufreq_cdev); |
04bdbdf9 | 718 | return cdev; |
02361418 | 719 | } |
39d99cff EV |
720 | |
721 | /** | |
722 | * cpufreq_cooling_register - function to create cpufreq cooling device. | |
4d753aa7 | 723 | * @policy: cpufreq policy |
39d99cff EV |
724 | * |
725 | * This interface function registers the cpufreq cooling device with the name | |
726 | * "thermal-cpufreq-%x". This api can support multiple instances of cpufreq | |
727 | * cooling devices. | |
728 | * | |
729 | * Return: a valid struct thermal_cooling_device pointer on success, | |
730 | * on failure, it returns a corresponding ERR_PTR(). | |
731 | */ | |
732 | struct thermal_cooling_device * | |
4d753aa7 | 733 | cpufreq_cooling_register(struct cpufreq_policy *policy) |
39d99cff | 734 | { |
84fe2cab | 735 | return __cpufreq_cooling_register(NULL, policy, 0); |
39d99cff | 736 | } |
243dbd9c | 737 | EXPORT_SYMBOL_GPL(cpufreq_cooling_register); |
02361418 | 738 | |
39d99cff EV |
739 | /** |
740 | * of_cpufreq_cooling_register - function to create cpufreq cooling device. | |
4d753aa7 | 741 | * @policy: cpufreq policy |
39d99cff EV |
742 | * |
743 | * This interface function registers the cpufreq cooling device with the name | |
744 | * "thermal-cpufreq-%x". This api can support multiple instances of cpufreq | |
745 | * cooling devices. Using this API, the cpufreq cooling device will be | |
746 | * linked to the device tree node provided. | |
747 | * | |
c36cf071 JM |
748 | * Using this function, the cooling device will implement the power |
749 | * extensions by using a simple cpu power model. The cpus must have | |
750 | * registered their OPPs using the OPP library. | |
751 | * | |
f5f263fe VK |
752 | * It also takes into account, if property present in policy CPU node, the |
753 | * static power consumed by the cpu. | |
c36cf071 JM |
754 | * |
755 | * Return: a valid struct thermal_cooling_device pointer on success, | |
f5f263fe | 756 | * and NULL on failure. |
c36cf071 JM |
757 | */ |
758 | struct thermal_cooling_device * | |
3ebb62ff | 759 | of_cpufreq_cooling_register(struct cpufreq_policy *policy) |
c36cf071 | 760 | { |
f5f263fe VK |
761 | struct device_node *np = of_get_cpu_node(policy->cpu, NULL); |
762 | struct thermal_cooling_device *cdev = NULL; | |
763 | u32 capacitance = 0; | |
764 | ||
765 | if (!np) { | |
766 | pr_err("cpu_cooling: OF node not available for cpu%d\n", | |
767 | policy->cpu); | |
768 | return NULL; | |
769 | } | |
c36cf071 | 770 | |
f5f263fe VK |
771 | if (of_find_property(np, "#cooling-cells", NULL)) { |
772 | of_property_read_u32(np, "dynamic-power-coefficient", | |
773 | &capacitance); | |
774 | ||
84fe2cab | 775 | cdev = __cpufreq_cooling_register(np, policy, capacitance); |
f5f263fe | 776 | if (IS_ERR(cdev)) { |
bf78f133 | 777 | pr_err("cpu_cooling: cpu%d failed to register as cooling device: %ld\n", |
f5f263fe VK |
778 | policy->cpu, PTR_ERR(cdev)); |
779 | cdev = NULL; | |
780 | } | |
781 | } | |
782 | ||
783 | of_node_put(np); | |
784 | return cdev; | |
c36cf071 | 785 | } |
3ebb62ff | 786 | EXPORT_SYMBOL_GPL(of_cpufreq_cooling_register); |
c36cf071 | 787 | |
02361418 ADK |
788 | /** |
789 | * cpufreq_cooling_unregister - function to remove cpufreq cooling device. | |
790 | * @cdev: thermal cooling device pointer. | |
135266b4 EV |
791 | * |
792 | * This interface function unregisters the "thermal-cpufreq-%x" cooling device. | |
02361418 ADK |
793 | */ |
794 | void cpufreq_cooling_unregister(struct thermal_cooling_device *cdev) | |
795 | { | |
1dea432a | 796 | struct cpufreq_cooling_device *cpufreq_cdev; |
088db931 | 797 | bool last; |
02361418 | 798 | |
50e66c7e EV |
799 | if (!cdev) |
800 | return; | |
801 | ||
1dea432a | 802 | cpufreq_cdev = cdev->devdata; |
02361418 | 803 | |
ae606089 | 804 | mutex_lock(&cooling_list_lock); |
1dea432a | 805 | list_del(&cpufreq_cdev->node); |
02361418 | 806 | /* Unregister the notifier for the last cpufreq cooling device */ |
1dea432a | 807 | last = list_empty(&cpufreq_cdev_list); |
088db931 MW |
808 | mutex_unlock(&cooling_list_lock); |
809 | ||
810 | if (last) | |
02361418 | 811 | cpufreq_unregister_notifier(&thermal_cpufreq_notifier_block, |
5fda7f68 | 812 | CPUFREQ_POLICY_NOTIFIER); |
02373d7c | 813 | |
04bdbdf9 | 814 | thermal_cooling_device_unregister(cpufreq_cdev->cdev); |
1dea432a | 815 | ida_simple_remove(&cpufreq_ida, cpufreq_cdev->id); |
81ee14da | 816 | kfree(cpufreq_cdev->idle_time); |
1dea432a VK |
817 | kfree(cpufreq_cdev->freq_table); |
818 | kfree(cpufreq_cdev); | |
02361418 | 819 | } |
243dbd9c | 820 | EXPORT_SYMBOL_GPL(cpufreq_cooling_unregister); |