xref: /linux/drivers/thermal/thermal_debugfs.c (revision 6ae4e48b67eaf8a140ac3a6f0f8ea2627a174aca)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright 2023 Linaro Limited
4  *
5  * Author: Daniel Lezcano <daniel.lezcano@linaro.org>
6  *
7  * Thermal subsystem debug support
8  */
9 #include <linux/debugfs.h>
10 #include <linux/ktime.h>
11 #include <linux/list.h>
12 #include <linux/minmax.h>
13 #include <linux/mutex.h>
14 #include <linux/thermal.h>
15 
16 #include "thermal_core.h"
17 
18 static struct dentry *d_root;
19 static struct dentry *d_cdev;
20 static struct dentry *d_tz;
21 
22 /*
23  * Length of the string containing the thermal zone id or the cooling
24  * device id, including the ending nul character. We can reasonably
25  * assume there won't be more than 256 thermal zones as the maximum
26  * observed today is around 32.
27  */
28 #define IDSLENGTH 4
29 
30 /*
31  * The cooling device transition list is stored in a hash table where
32  * the size is CDEVSTATS_HASH_SIZE. The majority of cooling devices
33  * have dozen of states but some can have much more, so a hash table
34  * is more adequate in this case, because the cost of browsing the entire
35  * list when storing the transitions may not be negligible.
36  */
37 #define CDEVSTATS_HASH_SIZE 16
38 
39 /**
40  * struct cdev_debugfs - per cooling device statistics structure
41  * A cooling device can have a high number of states. Showing the
42  * transitions on a matrix based representation can be overkill given
43  * most of the transitions won't happen and we end up with a matrix
44  * filled with zero. Instead, we show the transitions which actually
45  * happened.
46  *
47  * Every transition updates the current_state and the timestamp. The
48  * transitions and the durations are stored in lists.
49  *
50  * @total: the number of transitions for this cooling device
51  * @current_state: the current cooling device state
52  * @timestamp: the state change timestamp
53  * @transitions: an array of lists containing the state transitions
54  * @durations: an array of lists containing the residencies of each state
55  */
56 struct cdev_debugfs {
57 	u32 total;
58 	int current_state;
59 	ktime_t timestamp;
60 	struct list_head transitions[CDEVSTATS_HASH_SIZE];
61 	struct list_head durations[CDEVSTATS_HASH_SIZE];
62 };
63 
64 /**
65  * struct cdev_record - Common structure for cooling device entry
66  *
67  * The following common structure allows to store the information
68  * related to the transitions and to the state residencies. They are
69  * identified with a id which is associated to a value. It is used as
70  * nodes for the "transitions" and "durations" above.
71  *
72  * @node: node to insert the structure in a list
73  * @id: identifier of the value which can be a state or a transition
74  * @residency: a ktime_t representing a state residency duration
75  * @count: a number of occurrences
76  */
77 struct cdev_record {
78 	struct list_head node;
79 	int id;
80 	union {
81                 ktime_t residency;
82                 u64 count;
83         };
84 };
85 
86 /**
87  * struct trip_stats - Thermal trip statistics
88  *
89  * The trip_stats structure has the relevant information to show the
90  * statistics related to temperature going above a trip point.
91  *
92  * @timestamp: the trip crossing timestamp
93  * @duration: total time when the zone temperature was above the trip point
94  * @trip_temp: trip temperature at mitigation start
95  * @trip_hyst: trip hysteresis at mitigation start
96  * @count: the number of times the zone temperature was above the trip point
97  * @min: minimum recorded temperature above the trip point
98  * @avg: average temperature above the trip point
99  */
100 struct trip_stats {
101 	ktime_t timestamp;
102 	ktime_t duration;
103 	int trip_temp;
104 	int trip_hyst;
105 	int count;
106 	int min;
107 	int avg;
108 };
109 
110 /**
111  * struct tz_episode - A mitigation episode information
112  *
113  * The tz_episode structure describes a mitigation episode. A
114  * mitigation episode begins the trip point with the lower temperature
115  * is crossed the way up and ends when it is crossed the way
116  * down. During this episode we can have multiple trip points crossed
117  * the way up and down if there are multiple trip described in the
118  * firmware after the lowest temperature trip point.
119  *
120  * @timestamp: first trip point crossed the way up
121  * @duration: total duration of the mitigation episode
122  * @node: a list element to be added to the list of tz events
123  * @max_temp: maximum zone temperature during this episode
124  * @trip_stats: per trip point statistics, flexible array
125  */
126 struct tz_episode {
127 	ktime_t timestamp;
128 	ktime_t duration;
129 	struct list_head node;
130 	int max_temp;
131 	struct trip_stats trip_stats[];
132 };
133 
134 /**
135  * struct tz_debugfs - Store all mitigation episodes for a thermal zone
136  *
137  * The tz_debugfs structure contains the list of the mitigation
138  * episodes and has to track which trip point has been crossed in
139  * order to handle correctly nested trip point mitigation episodes.
140  *
141  * We keep the history of the trip point crossed in an array and as we
142  * can go back and forth inside this history, eg. trip 0,1,2,1,2,1,0,
143  * we keep track of the current position in the history array.
144  *
145  * @tz_episodes: a list of thermal mitigation episodes
146  * @tz: thermal zone this object belongs to
147  * @trips_crossed: an array of trip points crossed by id
148  * @nr_trips: the number of trip points currently being crossed
149  */
150 struct tz_debugfs {
151 	struct list_head tz_episodes;
152 	struct thermal_zone_device *tz;
153 	int *trips_crossed;
154 	int nr_trips;
155 };
156 
157 /**
158  * struct thermal_debugfs - High level structure for a thermal object in debugfs
159  *
160  * The thermal_debugfs structure is the common structure used by the
161  * cooling device or the thermal zone to store the statistics.
162  *
163  * @d_top: top directory of the thermal object directory
164  * @lock: per object lock to protect the internals
165  *
166  * @cdev_dbg: a cooling device debug structure
167  * @tz_dbg: a thermal zone debug structure
168  */
169 struct thermal_debugfs {
170 	struct dentry *d_top;
171 	struct mutex lock;
172 	union {
173 		struct cdev_debugfs cdev_dbg;
174 		struct tz_debugfs tz_dbg;
175 	};
176 };
177 
thermal_debug_init(void)178 void thermal_debug_init(void)
179 {
180 	d_root = debugfs_create_dir("thermal", NULL);
181 	if (IS_ERR(d_root))
182 		return;
183 
184 	d_cdev = debugfs_create_dir("cooling_devices", d_root);
185 	if (IS_ERR(d_cdev))
186 		return;
187 
188 	d_tz = debugfs_create_dir("thermal_zones", d_root);
189 }
190 
thermal_debugfs_add_id(struct dentry * d,int id)191 static struct thermal_debugfs *thermal_debugfs_add_id(struct dentry *d, int id)
192 {
193 	struct thermal_debugfs *thermal_dbg;
194 	char ids[IDSLENGTH];
195 
196 	thermal_dbg = kzalloc(sizeof(*thermal_dbg), GFP_KERNEL);
197 	if (!thermal_dbg)
198 		return NULL;
199 
200 	mutex_init(&thermal_dbg->lock);
201 
202 	snprintf(ids, IDSLENGTH, "%d", id);
203 
204 	thermal_dbg->d_top = debugfs_create_dir(ids, d);
205 	if (IS_ERR(thermal_dbg->d_top)) {
206 		kfree(thermal_dbg);
207 		return NULL;
208 	}
209 
210 	return thermal_dbg;
211 }
212 
thermal_debugfs_remove_id(struct thermal_debugfs * thermal_dbg)213 static void thermal_debugfs_remove_id(struct thermal_debugfs *thermal_dbg)
214 {
215 	if (!thermal_dbg)
216 		return;
217 
218 	debugfs_remove(thermal_dbg->d_top);
219 
220 	kfree(thermal_dbg);
221 }
222 
223 static struct cdev_record *
thermal_debugfs_cdev_record_alloc(struct thermal_debugfs * thermal_dbg,struct list_head * lists,int id)224 thermal_debugfs_cdev_record_alloc(struct thermal_debugfs *thermal_dbg,
225 				  struct list_head *lists, int id)
226 {
227 	struct cdev_record *cdev_record;
228 
229 	cdev_record = kzalloc(sizeof(*cdev_record), GFP_KERNEL);
230 	if (!cdev_record)
231 		return NULL;
232 
233 	cdev_record->id = id;
234 	INIT_LIST_HEAD(&cdev_record->node);
235 	list_add_tail(&cdev_record->node,
236 		      &lists[cdev_record->id % CDEVSTATS_HASH_SIZE]);
237 
238 	return cdev_record;
239 }
240 
241 static struct cdev_record *
thermal_debugfs_cdev_record_find(struct thermal_debugfs * thermal_dbg,struct list_head * lists,int id)242 thermal_debugfs_cdev_record_find(struct thermal_debugfs *thermal_dbg,
243 				 struct list_head *lists, int id)
244 {
245 	struct cdev_record *entry;
246 
247 	list_for_each_entry(entry, &lists[id % CDEVSTATS_HASH_SIZE], node)
248 		if (entry->id == id)
249 			return entry;
250 
251 	return NULL;
252 }
253 
254 static struct cdev_record *
thermal_debugfs_cdev_record_get(struct thermal_debugfs * thermal_dbg,struct list_head * lists,int id)255 thermal_debugfs_cdev_record_get(struct thermal_debugfs *thermal_dbg,
256 				struct list_head *lists, int id)
257 {
258 	struct cdev_record *cdev_record;
259 
260 	cdev_record = thermal_debugfs_cdev_record_find(thermal_dbg, lists, id);
261 	if (cdev_record)
262 		return cdev_record;
263 
264 	return thermal_debugfs_cdev_record_alloc(thermal_dbg, lists, id);
265 }
266 
thermal_debugfs_cdev_clear(struct cdev_debugfs * cdev_dbg)267 static void thermal_debugfs_cdev_clear(struct cdev_debugfs *cdev_dbg)
268 {
269 	int i;
270 	struct cdev_record *entry, *tmp;
271 
272 	for (i = 0; i < CDEVSTATS_HASH_SIZE; i++) {
273 
274 		list_for_each_entry_safe(entry, tmp,
275 					 &cdev_dbg->transitions[i], node) {
276 			list_del(&entry->node);
277 			kfree(entry);
278 		}
279 
280 		list_for_each_entry_safe(entry, tmp,
281 					 &cdev_dbg->durations[i], node) {
282 			list_del(&entry->node);
283 			kfree(entry);
284 		}
285 	}
286 
287 	cdev_dbg->total = 0;
288 }
289 
cdev_seq_start(struct seq_file * s,loff_t * pos)290 static void *cdev_seq_start(struct seq_file *s, loff_t *pos)
291 {
292 	struct thermal_debugfs *thermal_dbg = s->private;
293 
294 	mutex_lock(&thermal_dbg->lock);
295 
296 	return (*pos < CDEVSTATS_HASH_SIZE) ? pos : NULL;
297 }
298 
cdev_seq_next(struct seq_file * s,void * v,loff_t * pos)299 static void *cdev_seq_next(struct seq_file *s, void *v, loff_t *pos)
300 {
301 	(*pos)++;
302 
303 	return (*pos < CDEVSTATS_HASH_SIZE) ? pos : NULL;
304 }
305 
cdev_seq_stop(struct seq_file * s,void * v)306 static void cdev_seq_stop(struct seq_file *s, void *v)
307 {
308 	struct thermal_debugfs *thermal_dbg = s->private;
309 
310 	mutex_unlock(&thermal_dbg->lock);
311 }
312 
cdev_tt_seq_show(struct seq_file * s,void * v)313 static int cdev_tt_seq_show(struct seq_file *s, void *v)
314 {
315 	struct thermal_debugfs *thermal_dbg = s->private;
316 	struct cdev_debugfs *cdev_dbg = &thermal_dbg->cdev_dbg;
317 	struct list_head *transitions = cdev_dbg->transitions;
318 	struct cdev_record *entry;
319 	int i = *(loff_t *)v;
320 
321 	if (!i)
322 		seq_puts(s, "Transition\tOccurences\n");
323 
324 	list_for_each_entry(entry, &transitions[i], node) {
325 		/*
326 		 * Assuming maximum cdev states is 1024, the longer
327 		 * string for a transition would be "1024->1024\0"
328 		 */
329 		char buffer[11];
330 
331 		snprintf(buffer, ARRAY_SIZE(buffer), "%d->%d",
332 			 entry->id >> 16, entry->id & 0xFFFF);
333 
334 		seq_printf(s, "%-10s\t%-10llu\n", buffer, entry->count);
335 	}
336 
337 	return 0;
338 }
339 
340 static const struct seq_operations tt_sops = {
341 	.start = cdev_seq_start,
342 	.next = cdev_seq_next,
343 	.stop = cdev_seq_stop,
344 	.show = cdev_tt_seq_show,
345 };
346 
347 DEFINE_SEQ_ATTRIBUTE(tt);
348 
cdev_dt_seq_show(struct seq_file * s,void * v)349 static int cdev_dt_seq_show(struct seq_file *s, void *v)
350 {
351 	struct thermal_debugfs *thermal_dbg = s->private;
352 	struct cdev_debugfs *cdev_dbg = &thermal_dbg->cdev_dbg;
353 	struct list_head *durations = cdev_dbg->durations;
354 	struct cdev_record *entry;
355 	int i = *(loff_t *)v;
356 
357 	if (!i)
358 		seq_puts(s, "State\tResidency\n");
359 
360 	list_for_each_entry(entry, &durations[i], node) {
361 		s64 duration = ktime_to_ms(entry->residency);
362 
363 		if (entry->id == cdev_dbg->current_state)
364 			duration += ktime_ms_delta(ktime_get(),
365 						   cdev_dbg->timestamp);
366 
367 		seq_printf(s, "%-5d\t%-10llu\n", entry->id, duration);
368 	}
369 
370 	return 0;
371 }
372 
373 static const struct seq_operations dt_sops = {
374 	.start = cdev_seq_start,
375 	.next = cdev_seq_next,
376 	.stop = cdev_seq_stop,
377 	.show = cdev_dt_seq_show,
378 };
379 
380 DEFINE_SEQ_ATTRIBUTE(dt);
381 
cdev_clear_set(void * data,u64 val)382 static int cdev_clear_set(void *data, u64 val)
383 {
384 	struct thermal_debugfs *thermal_dbg = data;
385 
386 	if (!val)
387 		return -EINVAL;
388 
389 	mutex_lock(&thermal_dbg->lock);
390 
391 	thermal_debugfs_cdev_clear(&thermal_dbg->cdev_dbg);
392 
393 	mutex_unlock(&thermal_dbg->lock);
394 
395 	return 0;
396 }
397 
398 DEFINE_DEBUGFS_ATTRIBUTE(cdev_clear_fops, NULL, cdev_clear_set, "%llu\n");
399 
400 /**
401  * thermal_debug_cdev_state_update - Update a cooling device state change
402  *
403  * Computes a transition and the duration of the previous state residency.
404  *
405  * @cdev : a pointer to a cooling device
406  * @new_state: an integer corresponding to the new cooling device state
407  */
thermal_debug_cdev_state_update(const struct thermal_cooling_device * cdev,int new_state)408 void thermal_debug_cdev_state_update(const struct thermal_cooling_device *cdev,
409 				     int new_state)
410 {
411 	struct thermal_debugfs *thermal_dbg = cdev->debugfs;
412 	struct cdev_debugfs *cdev_dbg;
413 	struct cdev_record *cdev_record;
414 	int transition, old_state;
415 
416 	if (!thermal_dbg || (thermal_dbg->cdev_dbg.current_state == new_state))
417 		return;
418 
419 	mutex_lock(&thermal_dbg->lock);
420 
421 	cdev_dbg = &thermal_dbg->cdev_dbg;
422 
423 	old_state = cdev_dbg->current_state;
424 
425 	/*
426 	 * Get the old state information in the durations list. If
427 	 * this one does not exist, a new allocated one will be
428 	 * returned. Recompute the total duration in the old state and
429 	 * get a new timestamp for the new state.
430 	 */
431 	cdev_record = thermal_debugfs_cdev_record_get(thermal_dbg,
432 						      cdev_dbg->durations,
433 						      old_state);
434 	if (cdev_record) {
435 		ktime_t now = ktime_get();
436 		ktime_t delta = ktime_sub(now, cdev_dbg->timestamp);
437 		cdev_record->residency = ktime_add(cdev_record->residency, delta);
438 		cdev_dbg->timestamp = now;
439 	}
440 
441 	cdev_dbg->current_state = new_state;
442 
443 	/*
444 	 * Create a record for the new state if it is not there, so its
445 	 * duration will be printed by cdev_dt_seq_show() as expected if it
446 	 * runs before the next state transition.
447 	 */
448 	thermal_debugfs_cdev_record_get(thermal_dbg, cdev_dbg->durations, new_state);
449 
450 	transition = (old_state << 16) | new_state;
451 
452 	/*
453 	 * Get the transition in the transitions list. If this one
454 	 * does not exist, a new allocated one will be returned.
455 	 * Increment the occurrence of this transition which is stored
456 	 * in the value field.
457 	 */
458 	cdev_record = thermal_debugfs_cdev_record_get(thermal_dbg,
459 						      cdev_dbg->transitions,
460 						      transition);
461 	if (cdev_record)
462 		cdev_record->count++;
463 
464 	cdev_dbg->total++;
465 
466 	mutex_unlock(&thermal_dbg->lock);
467 }
468 
469 /**
470  * thermal_debug_cdev_add - Add a cooling device debugfs entry
471  *
472  * Allocates a cooling device object for debug, initializes the
473  * statistics and create the entries in sysfs.
474  * @cdev: a pointer to a cooling device
475  * @state: current state of the cooling device
476  */
thermal_debug_cdev_add(struct thermal_cooling_device * cdev,int state)477 void thermal_debug_cdev_add(struct thermal_cooling_device *cdev, int state)
478 {
479 	struct thermal_debugfs *thermal_dbg;
480 	struct cdev_debugfs *cdev_dbg;
481 	int i;
482 
483 	thermal_dbg = thermal_debugfs_add_id(d_cdev, cdev->id);
484 	if (!thermal_dbg)
485 		return;
486 
487 	cdev_dbg = &thermal_dbg->cdev_dbg;
488 
489 	for (i = 0; i < CDEVSTATS_HASH_SIZE; i++) {
490 		INIT_LIST_HEAD(&cdev_dbg->transitions[i]);
491 		INIT_LIST_HEAD(&cdev_dbg->durations[i]);
492 	}
493 
494 	cdev_dbg->current_state = state;
495 	cdev_dbg->timestamp = ktime_get();
496 
497 	/*
498 	 * Create a record for the initial cooling device state, so its
499 	 * duration will be printed by cdev_dt_seq_show() as expected if it
500 	 * runs before the first state transition.
501 	 */
502 	thermal_debugfs_cdev_record_get(thermal_dbg, cdev_dbg->durations, state);
503 
504 	debugfs_create_file("trans_table", 0400, thermal_dbg->d_top,
505 			    thermal_dbg, &tt_fops);
506 
507 	debugfs_create_file("time_in_state_ms", 0400, thermal_dbg->d_top,
508 			    thermal_dbg, &dt_fops);
509 
510 	debugfs_create_file("clear", 0200, thermal_dbg->d_top,
511 			    thermal_dbg, &cdev_clear_fops);
512 
513 	debugfs_create_u32("total_trans", 0400, thermal_dbg->d_top,
514 			   &cdev_dbg->total);
515 
516 	cdev->debugfs = thermal_dbg;
517 }
518 
519 /**
520  * thermal_debug_cdev_remove - Remove a cooling device debugfs entry
521  *
522  * Frees the statistics memory data and remove the debugfs entry
523  *
524  * @cdev: a pointer to a cooling device
525  */
thermal_debug_cdev_remove(struct thermal_cooling_device * cdev)526 void thermal_debug_cdev_remove(struct thermal_cooling_device *cdev)
527 {
528 	struct thermal_debugfs *thermal_dbg;
529 
530 	mutex_lock(&cdev->lock);
531 
532 	thermal_dbg = cdev->debugfs;
533 	if (!thermal_dbg) {
534 		mutex_unlock(&cdev->lock);
535 		return;
536 	}
537 
538 	cdev->debugfs = NULL;
539 
540 	mutex_unlock(&cdev->lock);
541 
542 	mutex_lock(&thermal_dbg->lock);
543 
544 	thermal_debugfs_cdev_clear(&thermal_dbg->cdev_dbg);
545 
546 	mutex_unlock(&thermal_dbg->lock);
547 
548 	thermal_debugfs_remove_id(thermal_dbg);
549 }
550 
thermal_debugfs_tz_event_alloc(struct thermal_zone_device * tz,ktime_t now)551 static struct tz_episode *thermal_debugfs_tz_event_alloc(struct thermal_zone_device *tz,
552 							ktime_t now)
553 {
554 	struct tz_episode *tze;
555 	int i;
556 
557 	tze = kzalloc(struct_size(tze, trip_stats, tz->num_trips), GFP_KERNEL);
558 	if (!tze)
559 		return NULL;
560 
561 	INIT_LIST_HEAD(&tze->node);
562 	tze->timestamp = now;
563 	tze->duration = KTIME_MIN;
564 	tze->max_temp = INT_MIN;
565 
566 	for (i = 0; i < tz->num_trips; i++) {
567 		tze->trip_stats[i].trip_temp = THERMAL_TEMP_INVALID;
568 		tze->trip_stats[i].min = INT_MAX;
569 	}
570 
571 	return tze;
572 }
573 
thermal_debug_tz_trip_up(struct thermal_zone_device * tz,const struct thermal_trip * trip)574 void thermal_debug_tz_trip_up(struct thermal_zone_device *tz,
575 			      const struct thermal_trip *trip)
576 {
577 	struct thermal_debugfs *thermal_dbg = tz->debugfs;
578 	int trip_id = thermal_zone_trip_id(tz, trip);
579 	ktime_t now = ktime_get();
580 	struct trip_stats *trip_stats;
581 	struct tz_debugfs *tz_dbg;
582 	struct tz_episode *tze;
583 
584 	if (!thermal_dbg)
585 		return;
586 
587 	tz_dbg = &thermal_dbg->tz_dbg;
588 
589 	mutex_lock(&thermal_dbg->lock);
590 
591 	/*
592 	 * The mitigation is starting. A mitigation can contain
593 	 * several episodes where each of them is related to a
594 	 * temperature crossing a trip point. The episodes are
595 	 * nested. That means when the temperature is crossing the
596 	 * first trip point, the duration begins to be measured. If
597 	 * the temperature continues to increase and reaches the
598 	 * second trip point, the duration of the first trip must be
599 	 * also accumulated.
600 	 *
601 	 * eg.
602 	 *
603 	 * temp
604 	 *   ^
605 	 *   |             --------
606 	 * trip 2         /        \         ------
607 	 *   |           /|        |\      /|      |\
608 	 * trip 1       / |        | `----  |      | \
609 	 *   |         /| |        |        |      | |\
610 	 * trip 0     / | |        |        |      | | \
611 	 *   |       /| | |        |        |      | | |\
612 	 *   |      / | | |        |        |      | | | `--
613 	 *   |     /  | | |        |        |      | | |
614 	 *   |-----   | | |        |        |      | | |
615 	 *   |        | | |        |        |      | | |
616 	 *    --------|-|-|--------|--------|------|-|-|------------------> time
617 	 *            | | |<--t2-->|        |<-t2'>| | |
618 	 *            | |                            | |
619 	 *            | |<------------t1------------>| |
620 	 *            |                                |
621 	 *            |<-------------t0--------------->|
622 	 *
623 	 */
624 	if (!tz_dbg->nr_trips) {
625 		tze = thermal_debugfs_tz_event_alloc(tz, now);
626 		if (!tze)
627 			goto unlock;
628 
629 		list_add(&tze->node, &tz_dbg->tz_episodes);
630 	}
631 
632 	/*
633 	 * Each time a trip point is crossed the way up, the trip_id
634 	 * is stored in the trip_crossed array and the nr_trips is
635 	 * incremented. A nr_trips equal to zero means we are entering
636 	 * a mitigation episode.
637 	 *
638 	 * The trip ids may not be in the ascending order but the
639 	 * result in the array trips_crossed will be in the ascending
640 	 * temperature order. The function detecting when a trip point
641 	 * is crossed the way down will handle the very rare case when
642 	 * the trip points may have been reordered during this
643 	 * mitigation episode.
644 	 */
645 	tz_dbg->trips_crossed[tz_dbg->nr_trips++] = trip_id;
646 
647 	tze = list_first_entry(&tz_dbg->tz_episodes, struct tz_episode, node);
648 	trip_stats = &tze->trip_stats[trip_id];
649 	trip_stats->trip_temp = trip->temperature;
650 	trip_stats->trip_hyst = trip->hysteresis;
651 	trip_stats->timestamp = now;
652 
653 unlock:
654 	mutex_unlock(&thermal_dbg->lock);
655 }
656 
tz_episode_close_trip(struct tz_episode * tze,int trip_id,ktime_t now)657 static void tz_episode_close_trip(struct tz_episode *tze, int trip_id, ktime_t now)
658 {
659 	struct trip_stats *trip_stats = &tze->trip_stats[trip_id];
660 	ktime_t delta = ktime_sub(now, trip_stats->timestamp);
661 
662 	trip_stats->duration = ktime_add(delta, trip_stats->duration);
663 	/* Mark the end of mitigation for this trip point. */
664 	trip_stats->timestamp = KTIME_MAX;
665 }
666 
thermal_debug_tz_trip_down(struct thermal_zone_device * tz,const struct thermal_trip * trip)667 void thermal_debug_tz_trip_down(struct thermal_zone_device *tz,
668 				const struct thermal_trip *trip)
669 {
670 	struct thermal_debugfs *thermal_dbg = tz->debugfs;
671 	int trip_id = thermal_zone_trip_id(tz, trip);
672 	ktime_t now = ktime_get();
673 	struct tz_episode *tze;
674 	struct tz_debugfs *tz_dbg;
675 	int i;
676 
677 	if (!thermal_dbg)
678 		return;
679 
680 	tz_dbg = &thermal_dbg->tz_dbg;
681 
682 	mutex_lock(&thermal_dbg->lock);
683 
684 	/*
685 	 * The temperature crosses the way down but there was not
686 	 * mitigation detected before. That may happen when the
687 	 * temperature is greater than a trip point when registering a
688 	 * thermal zone, which is a common use case as the kernel has
689 	 * no mitigation mechanism yet at boot time.
690 	 */
691 	if (!tz_dbg->nr_trips)
692 		goto out;
693 
694 	for (i = tz_dbg->nr_trips - 1; i >= 0; i--) {
695 		if (tz_dbg->trips_crossed[i] == trip_id)
696 			break;
697 	}
698 
699 	if (i < 0)
700 		goto out;
701 
702 	tz_dbg->nr_trips--;
703 
704 	if (i < tz_dbg->nr_trips)
705 		tz_dbg->trips_crossed[i] = tz_dbg->trips_crossed[tz_dbg->nr_trips];
706 
707 	tze = list_first_entry(&tz_dbg->tz_episodes, struct tz_episode, node);
708 
709 	tz_episode_close_trip(tze, trip_id, now);
710 
711 	/*
712 	 * This event closes the mitigation as we are crossing the
713 	 * last trip point the way down.
714 	 */
715 	if (!tz_dbg->nr_trips)
716 		tze->duration = ktime_sub(now, tze->timestamp);
717 
718 out:
719 	mutex_unlock(&thermal_dbg->lock);
720 }
721 
thermal_debug_update_trip_stats(struct thermal_zone_device * tz)722 void thermal_debug_update_trip_stats(struct thermal_zone_device *tz)
723 {
724 	struct thermal_debugfs *thermal_dbg = tz->debugfs;
725 	struct tz_debugfs *tz_dbg;
726 	struct tz_episode *tze;
727 	int i;
728 
729 	if (!thermal_dbg)
730 		return;
731 
732 	tz_dbg = &thermal_dbg->tz_dbg;
733 
734 	mutex_lock(&thermal_dbg->lock);
735 
736 	if (!tz_dbg->nr_trips)
737 		goto out;
738 
739 	tze = list_first_entry(&tz_dbg->tz_episodes, struct tz_episode, node);
740 
741 	if (tz->temperature > tze->max_temp)
742 		tze->max_temp = tz->temperature;
743 
744 	for (i = 0; i < tz_dbg->nr_trips; i++) {
745 		int trip_id = tz_dbg->trips_crossed[i];
746 		struct trip_stats *trip_stats = &tze->trip_stats[trip_id];
747 
748 		trip_stats->min = min(trip_stats->min, tz->temperature);
749 		trip_stats->avg += (tz->temperature - trip_stats->avg) /
750 					++trip_stats->count;
751 	}
752 out:
753 	mutex_unlock(&thermal_dbg->lock);
754 }
755 
tze_seq_start(struct seq_file * s,loff_t * pos)756 static void *tze_seq_start(struct seq_file *s, loff_t *pos)
757 {
758 	struct thermal_debugfs *thermal_dbg = s->private;
759 	struct tz_debugfs *tz_dbg = &thermal_dbg->tz_dbg;
760 
761 	mutex_lock(&thermal_dbg->lock);
762 
763 	return seq_list_start(&tz_dbg->tz_episodes, *pos);
764 }
765 
tze_seq_next(struct seq_file * s,void * v,loff_t * pos)766 static void *tze_seq_next(struct seq_file *s, void *v, loff_t *pos)
767 {
768 	struct thermal_debugfs *thermal_dbg = s->private;
769 	struct tz_debugfs *tz_dbg = &thermal_dbg->tz_dbg;
770 
771 	return seq_list_next(v, &tz_dbg->tz_episodes, pos);
772 }
773 
tze_seq_stop(struct seq_file * s,void * v)774 static void tze_seq_stop(struct seq_file *s, void *v)
775 {
776 	struct thermal_debugfs *thermal_dbg = s->private;
777 
778 	mutex_unlock(&thermal_dbg->lock);
779 }
780 
tze_seq_show(struct seq_file * s,void * v)781 static int tze_seq_show(struct seq_file *s, void *v)
782 {
783 	struct thermal_debugfs *thermal_dbg = s->private;
784 	struct thermal_zone_device *tz = thermal_dbg->tz_dbg.tz;
785 	struct thermal_trip_desc *td;
786 	struct tz_episode *tze;
787 	u64 duration_ms;
788 	int trip_id;
789 	char c;
790 
791 	tze = list_entry((struct list_head *)v, struct tz_episode, node);
792 
793 	if (tze->duration == KTIME_MIN) {
794 		/* Mitigation in progress. */
795 		duration_ms = ktime_to_ms(ktime_sub(ktime_get(), tze->timestamp));
796 		c = '>';
797 	} else {
798 		duration_ms = ktime_to_ms(tze->duration);
799 		c = '=';
800 	}
801 
802 	seq_printf(s, ",-Mitigation at %llums, duration%c%llums, max. temp=%dm°C\n",
803 		   ktime_to_ms(tze->timestamp), c, duration_ms, tze->max_temp);
804 
805 	seq_printf(s, "| trip |     type | temp(m°C) | hyst(m°C) | duration(ms) |  avg(m°C) |  min(m°C) |\n");
806 
807 	for_each_trip_desc(tz, td) {
808 		const struct thermal_trip *trip = &td->trip;
809 		struct trip_stats *trip_stats;
810 
811 		/*
812 		 * There is no possible mitigation happening at the
813 		 * critical trip point, so the stats will be always
814 		 * zero, skip this trip point
815 		 */
816 		if (trip->type == THERMAL_TRIP_CRITICAL)
817 			continue;
818 
819 		trip_id = thermal_zone_trip_id(tz, trip);
820 		trip_stats = &tze->trip_stats[trip_id];
821 
822 		/* Skip trips without any stats. */
823 		if (trip_stats->trip_temp == THERMAL_TEMP_INVALID)
824 			continue;
825 
826 		if (trip_stats->timestamp != KTIME_MAX) {
827 			/* Mitigation in progress. */
828 			ktime_t delta = ktime_sub(ktime_get(),
829 						  trip_stats->timestamp);
830 
831 			delta = ktime_add(delta, trip_stats->duration);
832 			duration_ms = ktime_to_ms(delta);
833 			c = '>';
834 		} else {
835 			duration_ms = ktime_to_ms(trip_stats->duration);
836 			c = ' ';
837 		}
838 
839 		seq_printf(s, "| %*d | %*s | %*d | %*d | %c%*lld | %*d | %*d |\n",
840 			   4 , trip_id,
841 			   8, thermal_trip_type_name(trip->type),
842 			   9, trip_stats->trip_temp,
843 			   9, trip_stats->trip_hyst,
844 			   c, 11, duration_ms,
845 			   9, trip_stats->avg,
846 			   9, trip_stats->min);
847 	}
848 
849 	return 0;
850 }
851 
852 static const struct seq_operations tze_sops = {
853 	.start = tze_seq_start,
854 	.next = tze_seq_next,
855 	.stop = tze_seq_stop,
856 	.show = tze_seq_show,
857 };
858 
859 DEFINE_SEQ_ATTRIBUTE(tze);
860 
thermal_debug_tz_add(struct thermal_zone_device * tz)861 void thermal_debug_tz_add(struct thermal_zone_device *tz)
862 {
863 	struct thermal_debugfs *thermal_dbg;
864 	struct tz_debugfs *tz_dbg;
865 
866 	thermal_dbg = thermal_debugfs_add_id(d_tz, tz->id);
867 	if (!thermal_dbg)
868 		return;
869 
870 	tz_dbg = &thermal_dbg->tz_dbg;
871 
872 	tz_dbg->tz = tz;
873 
874 	tz_dbg->trips_crossed = kzalloc(sizeof(int) * tz->num_trips, GFP_KERNEL);
875 	if (!tz_dbg->trips_crossed) {
876 		thermal_debugfs_remove_id(thermal_dbg);
877 		return;
878 	}
879 
880 	INIT_LIST_HEAD(&tz_dbg->tz_episodes);
881 
882 	debugfs_create_file("mitigations", 0400, thermal_dbg->d_top,
883 			    thermal_dbg, &tze_fops);
884 
885 	tz->debugfs = thermal_dbg;
886 }
887 
thermal_debug_tz_remove(struct thermal_zone_device * tz)888 void thermal_debug_tz_remove(struct thermal_zone_device *tz)
889 {
890 	struct thermal_debugfs *thermal_dbg;
891 	struct tz_episode *tze, *tmp;
892 	struct tz_debugfs *tz_dbg;
893 	int *trips_crossed;
894 
895 	mutex_lock(&tz->lock);
896 
897 	thermal_dbg = tz->debugfs;
898 	if (!thermal_dbg) {
899 		mutex_unlock(&tz->lock);
900 		return;
901 	}
902 
903 	tz->debugfs = NULL;
904 
905 	mutex_unlock(&tz->lock);
906 
907 	tz_dbg = &thermal_dbg->tz_dbg;
908 
909 	mutex_lock(&thermal_dbg->lock);
910 
911 	trips_crossed = tz_dbg->trips_crossed;
912 
913 	list_for_each_entry_safe(tze, tmp, &tz_dbg->tz_episodes, node) {
914 		list_del(&tze->node);
915 		kfree(tze);
916 	}
917 
918 	mutex_unlock(&thermal_dbg->lock);
919 
920 	thermal_debugfs_remove_id(thermal_dbg);
921 	kfree(trips_crossed);
922 }
923 
thermal_debug_tz_resume(struct thermal_zone_device * tz)924 void thermal_debug_tz_resume(struct thermal_zone_device *tz)
925 {
926 	struct thermal_debugfs *thermal_dbg = tz->debugfs;
927 	ktime_t now = ktime_get();
928 	struct tz_debugfs *tz_dbg;
929 	struct tz_episode *tze;
930 	int i;
931 
932 	if (!thermal_dbg)
933 		return;
934 
935 	mutex_lock(&thermal_dbg->lock);
936 
937 	tz_dbg = &thermal_dbg->tz_dbg;
938 
939 	if (!tz_dbg->nr_trips)
940 		goto out;
941 
942 	/*
943 	 * A mitigation episode was in progress before the preceding system
944 	 * suspend transition, so close it because the zone handling is starting
945 	 * over from scratch.
946 	 */
947 	tze = list_first_entry(&tz_dbg->tz_episodes, struct tz_episode, node);
948 
949 	for (i = 0; i < tz_dbg->nr_trips; i++)
950 		tz_episode_close_trip(tze, tz_dbg->trips_crossed[i], now);
951 
952 	tze->duration = ktime_sub(now, tze->timestamp);
953 
954 	tz_dbg->nr_trips = 0;
955 
956 out:
957 	mutex_unlock(&thermal_dbg->lock);
958 }
959