xref: /linux/tools/perf/builtin-lock.c (revision 24bce201d79807b668bf9d9e0aca801c5c0d5f78)
1 // SPDX-License-Identifier: GPL-2.0
2 #include <errno.h>
3 #include <inttypes.h>
4 #include "builtin.h"
5 #include "perf.h"
6 
7 #include "util/evlist.h" // for struct evsel_str_handler
8 #include "util/evsel.h"
9 #include "util/symbol.h"
10 #include "util/thread.h"
11 #include "util/header.h"
12 
13 #include <subcmd/pager.h>
14 #include <subcmd/parse-options.h>
15 #include "util/trace-event.h"
16 
17 #include "util/debug.h"
18 #include "util/session.h"
19 #include "util/tool.h"
20 #include "util/data.h"
21 #include "util/string2.h"
22 
23 #include <sys/types.h>
24 #include <sys/prctl.h>
25 #include <semaphore.h>
26 #include <pthread.h>
27 #include <math.h>
28 #include <limits.h>
29 
30 #include <linux/list.h>
31 #include <linux/hash.h>
32 #include <linux/kernel.h>
33 #include <linux/zalloc.h>
34 #include <linux/err.h>
35 
36 static struct perf_session *session;
37 
38 /* based on kernel/lockdep.c */
39 #define LOCKHASH_BITS		12
40 #define LOCKHASH_SIZE		(1UL << LOCKHASH_BITS)
41 
42 static struct hlist_head lockhash_table[LOCKHASH_SIZE];
43 
44 #define __lockhashfn(key)	hash_long((unsigned long)key, LOCKHASH_BITS)
45 #define lockhashentry(key)	(lockhash_table + __lockhashfn((key)))
46 
47 struct lock_stat {
48 	struct hlist_node	hash_entry;
49 	struct rb_node		rb;		/* used for sorting */
50 
51 	u64			addr;		/* address of lockdep_map, used as ID */
52 	char			*name;		/* for strcpy(), we cannot use const */
53 
54 	unsigned int		nr_acquire;
55 	unsigned int		nr_acquired;
56 	unsigned int		nr_contended;
57 	unsigned int		nr_release;
58 
59 	unsigned int		nr_readlock;
60 	unsigned int		nr_trylock;
61 
62 	/* these times are in nano sec. */
63 	u64                     avg_wait_time;
64 	u64			wait_time_total;
65 	u64			wait_time_min;
66 	u64			wait_time_max;
67 
68 	int			broken; /* flag of blacklist */
69 	int			combined;
70 };
71 
72 /*
73  * States of lock_seq_stat
74  *
75  * UNINITIALIZED is required for detecting first event of acquire.
76  * As the nature of lock events, there is no guarantee
77  * that the first event for the locks are acquire,
78  * it can be acquired, contended or release.
79  */
80 #define SEQ_STATE_UNINITIALIZED      0	       /* initial state */
81 #define SEQ_STATE_RELEASED	1
82 #define SEQ_STATE_ACQUIRING	2
83 #define SEQ_STATE_ACQUIRED	3
84 #define SEQ_STATE_READ_ACQUIRED	4
85 #define SEQ_STATE_CONTENDED	5
86 
87 /*
88  * MAX_LOCK_DEPTH
89  * Imported from include/linux/sched.h.
90  * Should this be synchronized?
91  */
92 #define MAX_LOCK_DEPTH 48
93 
94 /*
95  * struct lock_seq_stat:
96  * Place to put on state of one lock sequence
97  * 1) acquire -> acquired -> release
98  * 2) acquire -> contended -> acquired -> release
99  * 3) acquire (with read or try) -> release
100  * 4) Are there other patterns?
101  */
102 struct lock_seq_stat {
103 	struct list_head        list;
104 	int			state;
105 	u64			prev_event_time;
106 	u64                     addr;
107 
108 	int                     read_count;
109 };
110 
111 struct thread_stat {
112 	struct rb_node		rb;
113 
114 	u32                     tid;
115 	struct list_head        seq_list;
116 };
117 
118 static struct rb_root		thread_stats;
119 
120 static bool combine_locks;
121 static bool show_thread_stats;
122 
123 static struct thread_stat *thread_stat_find(u32 tid)
124 {
125 	struct rb_node *node;
126 	struct thread_stat *st;
127 
128 	node = thread_stats.rb_node;
129 	while (node) {
130 		st = container_of(node, struct thread_stat, rb);
131 		if (st->tid == tid)
132 			return st;
133 		else if (tid < st->tid)
134 			node = node->rb_left;
135 		else
136 			node = node->rb_right;
137 	}
138 
139 	return NULL;
140 }
141 
142 static void thread_stat_insert(struct thread_stat *new)
143 {
144 	struct rb_node **rb = &thread_stats.rb_node;
145 	struct rb_node *parent = NULL;
146 	struct thread_stat *p;
147 
148 	while (*rb) {
149 		p = container_of(*rb, struct thread_stat, rb);
150 		parent = *rb;
151 
152 		if (new->tid < p->tid)
153 			rb = &(*rb)->rb_left;
154 		else if (new->tid > p->tid)
155 			rb = &(*rb)->rb_right;
156 		else
157 			BUG_ON("inserting invalid thread_stat\n");
158 	}
159 
160 	rb_link_node(&new->rb, parent, rb);
161 	rb_insert_color(&new->rb, &thread_stats);
162 }
163 
164 static struct thread_stat *thread_stat_findnew_after_first(u32 tid)
165 {
166 	struct thread_stat *st;
167 
168 	st = thread_stat_find(tid);
169 	if (st)
170 		return st;
171 
172 	st = zalloc(sizeof(struct thread_stat));
173 	if (!st) {
174 		pr_err("memory allocation failed\n");
175 		return NULL;
176 	}
177 
178 	st->tid = tid;
179 	INIT_LIST_HEAD(&st->seq_list);
180 
181 	thread_stat_insert(st);
182 
183 	return st;
184 }
185 
186 static struct thread_stat *thread_stat_findnew_first(u32 tid);
187 static struct thread_stat *(*thread_stat_findnew)(u32 tid) =
188 	thread_stat_findnew_first;
189 
190 static struct thread_stat *thread_stat_findnew_first(u32 tid)
191 {
192 	struct thread_stat *st;
193 
194 	st = zalloc(sizeof(struct thread_stat));
195 	if (!st) {
196 		pr_err("memory allocation failed\n");
197 		return NULL;
198 	}
199 	st->tid = tid;
200 	INIT_LIST_HEAD(&st->seq_list);
201 
202 	rb_link_node(&st->rb, NULL, &thread_stats.rb_node);
203 	rb_insert_color(&st->rb, &thread_stats);
204 
205 	thread_stat_findnew = thread_stat_findnew_after_first;
206 	return st;
207 }
208 
209 /* build simple key function one is bigger than two */
210 #define SINGLE_KEY(member)						\
211 	static int lock_stat_key_ ## member(struct lock_stat *one,	\
212 					 struct lock_stat *two)		\
213 	{								\
214 		return one->member > two->member;			\
215 	}
216 
217 SINGLE_KEY(nr_acquired)
218 SINGLE_KEY(nr_contended)
219 SINGLE_KEY(avg_wait_time)
220 SINGLE_KEY(wait_time_total)
221 SINGLE_KEY(wait_time_max)
222 
223 static int lock_stat_key_wait_time_min(struct lock_stat *one,
224 					struct lock_stat *two)
225 {
226 	u64 s1 = one->wait_time_min;
227 	u64 s2 = two->wait_time_min;
228 	if (s1 == ULLONG_MAX)
229 		s1 = 0;
230 	if (s2 == ULLONG_MAX)
231 		s2 = 0;
232 	return s1 > s2;
233 }
234 
235 struct lock_key {
236 	/*
237 	 * name: the value for specify by user
238 	 * this should be simpler than raw name of member
239 	 * e.g. nr_acquired -> acquired, wait_time_total -> wait_total
240 	 */
241 	const char		*name;
242 	/* header: the string printed on the header line */
243 	const char		*header;
244 	/* len: the printing width of the field */
245 	int			len;
246 	/* key: a pointer to function to compare two lock stats for sorting */
247 	int			(*key)(struct lock_stat*, struct lock_stat*);
248 	/* print: a pointer to function to print a given lock stats */
249 	void			(*print)(struct lock_key*, struct lock_stat*);
250 	/* list: list entry to link this */
251 	struct list_head	list;
252 };
253 
254 #define PRINT_KEY(member)						\
255 static void lock_stat_key_print_ ## member(struct lock_key *key,	\
256 					   struct lock_stat *ls)	\
257 {									\
258 	pr_info("%*llu", key->len, (unsigned long long)ls->member);	\
259 }
260 
261 PRINT_KEY(nr_acquired)
262 PRINT_KEY(nr_contended)
263 PRINT_KEY(avg_wait_time)
264 PRINT_KEY(wait_time_total)
265 PRINT_KEY(wait_time_max)
266 
267 static void lock_stat_key_print_wait_time_min(struct lock_key *key,
268 					      struct lock_stat *ls)
269 {
270 	u64 wait_time = ls->wait_time_min;
271 
272 	if (wait_time == ULLONG_MAX)
273 		wait_time = 0;
274 
275 	pr_info("%*"PRIu64, key->len, wait_time);
276 }
277 
278 
279 static const char		*sort_key = "acquired";
280 
281 static int			(*compare)(struct lock_stat *, struct lock_stat *);
282 
283 static struct rb_root		sorted; /* place to store intermediate data */
284 static struct rb_root		result;	/* place to store sorted data */
285 
286 static LIST_HEAD(lock_keys);
287 static const char		*output_fields;
288 
289 #define DEF_KEY_LOCK(name, header, fn_suffix, len)			\
290 	{ #name, header, len, lock_stat_key_ ## fn_suffix, lock_stat_key_print_ ## fn_suffix, {} }
291 struct lock_key keys[] = {
292 	DEF_KEY_LOCK(acquired, "acquired", nr_acquired, 10),
293 	DEF_KEY_LOCK(contended, "contended", nr_contended, 10),
294 	DEF_KEY_LOCK(avg_wait, "avg wait (ns)", avg_wait_time, 15),
295 	DEF_KEY_LOCK(wait_total, "total wait (ns)", wait_time_total, 15),
296 	DEF_KEY_LOCK(wait_max, "max wait (ns)", wait_time_max, 15),
297 	DEF_KEY_LOCK(wait_min, "min wait (ns)", wait_time_min, 15),
298 
299 	/* extra comparisons much complicated should be here */
300 	{ }
301 };
302 
303 static int select_key(void)
304 {
305 	int i;
306 
307 	for (i = 0; keys[i].name; i++) {
308 		if (!strcmp(keys[i].name, sort_key)) {
309 			compare = keys[i].key;
310 
311 			/* selected key should be in the output fields */
312 			if (list_empty(&keys[i].list))
313 				list_add_tail(&keys[i].list, &lock_keys);
314 
315 			return 0;
316 		}
317 	}
318 
319 	pr_err("Unknown compare key: %s\n", sort_key);
320 	return -1;
321 }
322 
323 static int add_output_field(struct list_head *head, char *name)
324 {
325 	int i;
326 
327 	for (i = 0; keys[i].name; i++) {
328 		if (strcmp(keys[i].name, name))
329 			continue;
330 
331 		/* prevent double link */
332 		if (list_empty(&keys[i].list))
333 			list_add_tail(&keys[i].list, head);
334 
335 		return 0;
336 	}
337 
338 	pr_err("Unknown output field: %s\n", name);
339 	return -1;
340 }
341 
342 static int setup_output_field(const char *str)
343 {
344 	char *tok, *tmp, *orig;
345 	int i, ret = 0;
346 
347 	/* no output field given: use all of them */
348 	if (str == NULL) {
349 		for (i = 0; keys[i].name; i++)
350 			list_add_tail(&keys[i].list, &lock_keys);
351 		return 0;
352 	}
353 
354 	for (i = 0; keys[i].name; i++)
355 		INIT_LIST_HEAD(&keys[i].list);
356 
357 	orig = tmp = strdup(str);
358 	if (orig == NULL)
359 		return -ENOMEM;
360 
361 	while ((tok = strsep(&tmp, ",")) != NULL){
362 		ret = add_output_field(&lock_keys, tok);
363 		if (ret < 0)
364 			break;
365 	}
366 	free(orig);
367 
368 	return ret;
369 }
370 
371 static void combine_lock_stats(struct lock_stat *st)
372 {
373 	struct rb_node **rb = &sorted.rb_node;
374 	struct rb_node *parent = NULL;
375 	struct lock_stat *p;
376 	int ret;
377 
378 	while (*rb) {
379 		p = container_of(*rb, struct lock_stat, rb);
380 		parent = *rb;
381 
382 		if (st->name && p->name)
383 			ret = strcmp(st->name, p->name);
384 		else
385 			ret = !!st->name - !!p->name;
386 
387 		if (ret == 0) {
388 			p->nr_acquired += st->nr_acquired;
389 			p->nr_contended += st->nr_contended;
390 			p->wait_time_total += st->wait_time_total;
391 
392 			if (p->nr_contended)
393 				p->avg_wait_time = p->wait_time_total / p->nr_contended;
394 
395 			if (p->wait_time_min > st->wait_time_min)
396 				p->wait_time_min = st->wait_time_min;
397 			if (p->wait_time_max < st->wait_time_max)
398 				p->wait_time_max = st->wait_time_max;
399 
400 			p->broken |= st->broken;
401 			st->combined = 1;
402 			return;
403 		}
404 
405 		if (ret < 0)
406 			rb = &(*rb)->rb_left;
407 		else
408 			rb = &(*rb)->rb_right;
409 	}
410 
411 	rb_link_node(&st->rb, parent, rb);
412 	rb_insert_color(&st->rb, &sorted);
413 }
414 
415 static void insert_to_result(struct lock_stat *st,
416 			     int (*bigger)(struct lock_stat *, struct lock_stat *))
417 {
418 	struct rb_node **rb = &result.rb_node;
419 	struct rb_node *parent = NULL;
420 	struct lock_stat *p;
421 
422 	if (combine_locks && st->combined)
423 		return;
424 
425 	while (*rb) {
426 		p = container_of(*rb, struct lock_stat, rb);
427 		parent = *rb;
428 
429 		if (bigger(st, p))
430 			rb = &(*rb)->rb_left;
431 		else
432 			rb = &(*rb)->rb_right;
433 	}
434 
435 	rb_link_node(&st->rb, parent, rb);
436 	rb_insert_color(&st->rb, &result);
437 }
438 
439 /* returns left most element of result, and erase it */
440 static struct lock_stat *pop_from_result(void)
441 {
442 	struct rb_node *node = result.rb_node;
443 
444 	if (!node)
445 		return NULL;
446 
447 	while (node->rb_left)
448 		node = node->rb_left;
449 
450 	rb_erase(node, &result);
451 	return container_of(node, struct lock_stat, rb);
452 }
453 
454 static struct lock_stat *lock_stat_findnew(u64 addr, const char *name)
455 {
456 	struct hlist_head *entry = lockhashentry(addr);
457 	struct lock_stat *ret, *new;
458 
459 	hlist_for_each_entry(ret, entry, hash_entry) {
460 		if (ret->addr == addr)
461 			return ret;
462 	}
463 
464 	new = zalloc(sizeof(struct lock_stat));
465 	if (!new)
466 		goto alloc_failed;
467 
468 	new->addr = addr;
469 	new->name = zalloc(sizeof(char) * strlen(name) + 1);
470 	if (!new->name) {
471 		free(new);
472 		goto alloc_failed;
473 	}
474 
475 	strcpy(new->name, name);
476 	new->wait_time_min = ULLONG_MAX;
477 
478 	hlist_add_head(&new->hash_entry, entry);
479 	return new;
480 
481 alloc_failed:
482 	pr_err("memory allocation failed\n");
483 	return NULL;
484 }
485 
486 struct trace_lock_handler {
487 	int (*acquire_event)(struct evsel *evsel,
488 			     struct perf_sample *sample);
489 
490 	int (*acquired_event)(struct evsel *evsel,
491 			      struct perf_sample *sample);
492 
493 	int (*contended_event)(struct evsel *evsel,
494 			       struct perf_sample *sample);
495 
496 	int (*release_event)(struct evsel *evsel,
497 			     struct perf_sample *sample);
498 };
499 
500 static struct lock_seq_stat *get_seq(struct thread_stat *ts, u64 addr)
501 {
502 	struct lock_seq_stat *seq;
503 
504 	list_for_each_entry(seq, &ts->seq_list, list) {
505 		if (seq->addr == addr)
506 			return seq;
507 	}
508 
509 	seq = zalloc(sizeof(struct lock_seq_stat));
510 	if (!seq) {
511 		pr_err("memory allocation failed\n");
512 		return NULL;
513 	}
514 	seq->state = SEQ_STATE_UNINITIALIZED;
515 	seq->addr = addr;
516 
517 	list_add(&seq->list, &ts->seq_list);
518 	return seq;
519 }
520 
521 enum broken_state {
522 	BROKEN_ACQUIRE,
523 	BROKEN_ACQUIRED,
524 	BROKEN_CONTENDED,
525 	BROKEN_RELEASE,
526 	BROKEN_MAX,
527 };
528 
529 static int bad_hist[BROKEN_MAX];
530 
531 enum acquire_flags {
532 	TRY_LOCK = 1,
533 	READ_LOCK = 2,
534 };
535 
536 static int report_lock_acquire_event(struct evsel *evsel,
537 				     struct perf_sample *sample)
538 {
539 	struct lock_stat *ls;
540 	struct thread_stat *ts;
541 	struct lock_seq_stat *seq;
542 	const char *name = evsel__strval(evsel, sample, "name");
543 	u64 addr = evsel__intval(evsel, sample, "lockdep_addr");
544 	int flag = evsel__intval(evsel, sample, "flags");
545 
546 	/* abuse ls->addr for tid */
547 	if (show_thread_stats)
548 		addr = sample->tid;
549 
550 	ls = lock_stat_findnew(addr, name);
551 	if (!ls)
552 		return -ENOMEM;
553 
554 	ts = thread_stat_findnew(sample->tid);
555 	if (!ts)
556 		return -ENOMEM;
557 
558 	seq = get_seq(ts, addr);
559 	if (!seq)
560 		return -ENOMEM;
561 
562 	switch (seq->state) {
563 	case SEQ_STATE_UNINITIALIZED:
564 	case SEQ_STATE_RELEASED:
565 		if (!flag) {
566 			seq->state = SEQ_STATE_ACQUIRING;
567 		} else {
568 			if (flag & TRY_LOCK)
569 				ls->nr_trylock++;
570 			if (flag & READ_LOCK)
571 				ls->nr_readlock++;
572 			seq->state = SEQ_STATE_READ_ACQUIRED;
573 			seq->read_count = 1;
574 			ls->nr_acquired++;
575 		}
576 		break;
577 	case SEQ_STATE_READ_ACQUIRED:
578 		if (flag & READ_LOCK) {
579 			seq->read_count++;
580 			ls->nr_acquired++;
581 			goto end;
582 		} else {
583 			goto broken;
584 		}
585 		break;
586 	case SEQ_STATE_ACQUIRED:
587 	case SEQ_STATE_ACQUIRING:
588 	case SEQ_STATE_CONTENDED:
589 broken:
590 		/* broken lock sequence */
591 		if (!ls->broken) {
592 			ls->broken = 1;
593 			bad_hist[BROKEN_ACQUIRE]++;
594 		}
595 		list_del_init(&seq->list);
596 		free(seq);
597 		goto end;
598 	default:
599 		BUG_ON("Unknown state of lock sequence found!\n");
600 		break;
601 	}
602 
603 	ls->nr_acquire++;
604 	seq->prev_event_time = sample->time;
605 end:
606 	return 0;
607 }
608 
609 static int report_lock_acquired_event(struct evsel *evsel,
610 				      struct perf_sample *sample)
611 {
612 	struct lock_stat *ls;
613 	struct thread_stat *ts;
614 	struct lock_seq_stat *seq;
615 	u64 contended_term;
616 	const char *name = evsel__strval(evsel, sample, "name");
617 	u64 addr = evsel__intval(evsel, sample, "lockdep_addr");
618 
619 	if (show_thread_stats)
620 		addr = sample->tid;
621 
622 	ls = lock_stat_findnew(addr, name);
623 	if (!ls)
624 		return -ENOMEM;
625 
626 	ts = thread_stat_findnew(sample->tid);
627 	if (!ts)
628 		return -ENOMEM;
629 
630 	seq = get_seq(ts, addr);
631 	if (!seq)
632 		return -ENOMEM;
633 
634 	switch (seq->state) {
635 	case SEQ_STATE_UNINITIALIZED:
636 		/* orphan event, do nothing */
637 		return 0;
638 	case SEQ_STATE_ACQUIRING:
639 		break;
640 	case SEQ_STATE_CONTENDED:
641 		contended_term = sample->time - seq->prev_event_time;
642 		ls->wait_time_total += contended_term;
643 		if (contended_term < ls->wait_time_min)
644 			ls->wait_time_min = contended_term;
645 		if (ls->wait_time_max < contended_term)
646 			ls->wait_time_max = contended_term;
647 		break;
648 	case SEQ_STATE_RELEASED:
649 	case SEQ_STATE_ACQUIRED:
650 	case SEQ_STATE_READ_ACQUIRED:
651 		/* broken lock sequence */
652 		if (!ls->broken) {
653 			ls->broken = 1;
654 			bad_hist[BROKEN_ACQUIRED]++;
655 		}
656 		list_del_init(&seq->list);
657 		free(seq);
658 		goto end;
659 	default:
660 		BUG_ON("Unknown state of lock sequence found!\n");
661 		break;
662 	}
663 
664 	seq->state = SEQ_STATE_ACQUIRED;
665 	ls->nr_acquired++;
666 	ls->avg_wait_time = ls->nr_contended ? ls->wait_time_total/ls->nr_contended : 0;
667 	seq->prev_event_time = sample->time;
668 end:
669 	return 0;
670 }
671 
672 static int report_lock_contended_event(struct evsel *evsel,
673 				       struct perf_sample *sample)
674 {
675 	struct lock_stat *ls;
676 	struct thread_stat *ts;
677 	struct lock_seq_stat *seq;
678 	const char *name = evsel__strval(evsel, sample, "name");
679 	u64 addr = evsel__intval(evsel, sample, "lockdep_addr");
680 
681 	if (show_thread_stats)
682 		addr = sample->tid;
683 
684 	ls = lock_stat_findnew(addr, name);
685 	if (!ls)
686 		return -ENOMEM;
687 
688 	ts = thread_stat_findnew(sample->tid);
689 	if (!ts)
690 		return -ENOMEM;
691 
692 	seq = get_seq(ts, addr);
693 	if (!seq)
694 		return -ENOMEM;
695 
696 	switch (seq->state) {
697 	case SEQ_STATE_UNINITIALIZED:
698 		/* orphan event, do nothing */
699 		return 0;
700 	case SEQ_STATE_ACQUIRING:
701 		break;
702 	case SEQ_STATE_RELEASED:
703 	case SEQ_STATE_ACQUIRED:
704 	case SEQ_STATE_READ_ACQUIRED:
705 	case SEQ_STATE_CONTENDED:
706 		/* broken lock sequence */
707 		if (!ls->broken) {
708 			ls->broken = 1;
709 			bad_hist[BROKEN_CONTENDED]++;
710 		}
711 		list_del_init(&seq->list);
712 		free(seq);
713 		goto end;
714 	default:
715 		BUG_ON("Unknown state of lock sequence found!\n");
716 		break;
717 	}
718 
719 	seq->state = SEQ_STATE_CONTENDED;
720 	ls->nr_contended++;
721 	ls->avg_wait_time = ls->wait_time_total/ls->nr_contended;
722 	seq->prev_event_time = sample->time;
723 end:
724 	return 0;
725 }
726 
727 static int report_lock_release_event(struct evsel *evsel,
728 				     struct perf_sample *sample)
729 {
730 	struct lock_stat *ls;
731 	struct thread_stat *ts;
732 	struct lock_seq_stat *seq;
733 	const char *name = evsel__strval(evsel, sample, "name");
734 	u64 addr = evsel__intval(evsel, sample, "lockdep_addr");
735 
736 	if (show_thread_stats)
737 		addr = sample->tid;
738 
739 	ls = lock_stat_findnew(addr, name);
740 	if (!ls)
741 		return -ENOMEM;
742 
743 	ts = thread_stat_findnew(sample->tid);
744 	if (!ts)
745 		return -ENOMEM;
746 
747 	seq = get_seq(ts, addr);
748 	if (!seq)
749 		return -ENOMEM;
750 
751 	switch (seq->state) {
752 	case SEQ_STATE_UNINITIALIZED:
753 		goto end;
754 	case SEQ_STATE_ACQUIRED:
755 		break;
756 	case SEQ_STATE_READ_ACQUIRED:
757 		seq->read_count--;
758 		BUG_ON(seq->read_count < 0);
759 		if (seq->read_count) {
760 			ls->nr_release++;
761 			goto end;
762 		}
763 		break;
764 	case SEQ_STATE_ACQUIRING:
765 	case SEQ_STATE_CONTENDED:
766 	case SEQ_STATE_RELEASED:
767 		/* broken lock sequence */
768 		if (!ls->broken) {
769 			ls->broken = 1;
770 			bad_hist[BROKEN_RELEASE]++;
771 		}
772 		goto free_seq;
773 	default:
774 		BUG_ON("Unknown state of lock sequence found!\n");
775 		break;
776 	}
777 
778 	ls->nr_release++;
779 free_seq:
780 	list_del_init(&seq->list);
781 	free(seq);
782 end:
783 	return 0;
784 }
785 
786 /* lock oriented handlers */
787 /* TODO: handlers for CPU oriented, thread oriented */
788 static struct trace_lock_handler report_lock_ops  = {
789 	.acquire_event		= report_lock_acquire_event,
790 	.acquired_event		= report_lock_acquired_event,
791 	.contended_event	= report_lock_contended_event,
792 	.release_event		= report_lock_release_event,
793 };
794 
795 static struct trace_lock_handler *trace_handler;
796 
797 static int evsel__process_lock_acquire(struct evsel *evsel, struct perf_sample *sample)
798 {
799 	if (trace_handler->acquire_event)
800 		return trace_handler->acquire_event(evsel, sample);
801 	return 0;
802 }
803 
804 static int evsel__process_lock_acquired(struct evsel *evsel, struct perf_sample *sample)
805 {
806 	if (trace_handler->acquired_event)
807 		return trace_handler->acquired_event(evsel, sample);
808 	return 0;
809 }
810 
811 static int evsel__process_lock_contended(struct evsel *evsel, struct perf_sample *sample)
812 {
813 	if (trace_handler->contended_event)
814 		return trace_handler->contended_event(evsel, sample);
815 	return 0;
816 }
817 
818 static int evsel__process_lock_release(struct evsel *evsel, struct perf_sample *sample)
819 {
820 	if (trace_handler->release_event)
821 		return trace_handler->release_event(evsel, sample);
822 	return 0;
823 }
824 
825 static void print_bad_events(int bad, int total)
826 {
827 	/* Output for debug, this have to be removed */
828 	int i;
829 	const char *name[4] =
830 		{ "acquire", "acquired", "contended", "release" };
831 
832 	pr_info("\n=== output for debug===\n\n");
833 	pr_info("bad: %d, total: %d\n", bad, total);
834 	pr_info("bad rate: %.2f %%\n", (double)bad / (double)total * 100);
835 	pr_info("histogram of events caused bad sequence\n");
836 	for (i = 0; i < BROKEN_MAX; i++)
837 		pr_info(" %10s: %d\n", name[i], bad_hist[i]);
838 }
839 
840 /* TODO: various way to print, coloring, nano or milli sec */
841 static void print_result(void)
842 {
843 	struct lock_stat *st;
844 	struct lock_key *key;
845 	char cut_name[20];
846 	int bad, total;
847 
848 	pr_info("%20s ", "Name");
849 	list_for_each_entry(key, &lock_keys, list)
850 		pr_info("%*s ", key->len, key->header);
851 	pr_info("\n\n");
852 
853 	bad = total = 0;
854 	while ((st = pop_from_result())) {
855 		total++;
856 		if (st->broken)
857 			bad++;
858 		if (!st->nr_acquired)
859 			continue;
860 
861 		bzero(cut_name, 20);
862 
863 		if (strlen(st->name) < 20) {
864 			/* output raw name */
865 			const char *name = st->name;
866 
867 			if (show_thread_stats) {
868 				struct thread *t;
869 
870 				/* st->addr contains tid of thread */
871 				t = perf_session__findnew(session, st->addr);
872 				name = thread__comm_str(t);
873 			}
874 
875 			pr_info("%20s ", name);
876 		} else {
877 			strncpy(cut_name, st->name, 16);
878 			cut_name[16] = '.';
879 			cut_name[17] = '.';
880 			cut_name[18] = '.';
881 			cut_name[19] = '\0';
882 			/* cut off name for saving output style */
883 			pr_info("%20s ", cut_name);
884 		}
885 
886 		list_for_each_entry(key, &lock_keys, list) {
887 			key->print(key, st);
888 			pr_info(" ");
889 		}
890 		pr_info("\n");
891 	}
892 
893 	print_bad_events(bad, total);
894 }
895 
896 static bool info_threads, info_map;
897 
898 static void dump_threads(void)
899 {
900 	struct thread_stat *st;
901 	struct rb_node *node;
902 	struct thread *t;
903 
904 	pr_info("%10s: comm\n", "Thread ID");
905 
906 	node = rb_first(&thread_stats);
907 	while (node) {
908 		st = container_of(node, struct thread_stat, rb);
909 		t = perf_session__findnew(session, st->tid);
910 		pr_info("%10d: %s\n", st->tid, thread__comm_str(t));
911 		node = rb_next(node);
912 		thread__put(t);
913 	}
914 }
915 
916 static int compare_maps(struct lock_stat *a, struct lock_stat *b)
917 {
918 	int ret;
919 
920 	if (a->name && b->name)
921 		ret = strcmp(a->name, b->name);
922 	else
923 		ret = !!a->name - !!b->name;
924 
925 	if (!ret)
926 		return a->addr < b->addr;
927 	else
928 		return ret < 0;
929 }
930 
931 static void dump_map(void)
932 {
933 	unsigned int i;
934 	struct lock_stat *st;
935 
936 	pr_info("Address of instance: name of class\n");
937 	for (i = 0; i < LOCKHASH_SIZE; i++) {
938 		hlist_for_each_entry(st, &lockhash_table[i], hash_entry) {
939 			insert_to_result(st, compare_maps);
940 		}
941 	}
942 
943 	while ((st = pop_from_result()))
944 		pr_info(" %#llx: %s\n", (unsigned long long)st->addr, st->name);
945 }
946 
947 static int dump_info(void)
948 {
949 	int rc = 0;
950 
951 	if (info_threads)
952 		dump_threads();
953 	else if (info_map)
954 		dump_map();
955 	else {
956 		rc = -1;
957 		pr_err("Unknown type of information\n");
958 	}
959 
960 	return rc;
961 }
962 
963 typedef int (*tracepoint_handler)(struct evsel *evsel,
964 				  struct perf_sample *sample);
965 
966 static int process_sample_event(struct perf_tool *tool __maybe_unused,
967 				union perf_event *event,
968 				struct perf_sample *sample,
969 				struct evsel *evsel,
970 				struct machine *machine)
971 {
972 	int err = 0;
973 	struct thread *thread = machine__findnew_thread(machine, sample->pid,
974 							sample->tid);
975 
976 	if (thread == NULL) {
977 		pr_debug("problem processing %d event, skipping it.\n",
978 			event->header.type);
979 		return -1;
980 	}
981 
982 	if (evsel->handler != NULL) {
983 		tracepoint_handler f = evsel->handler;
984 		err = f(evsel, sample);
985 	}
986 
987 	thread__put(thread);
988 
989 	return err;
990 }
991 
992 static void combine_result(void)
993 {
994 	unsigned int i;
995 	struct lock_stat *st;
996 
997 	if (!combine_locks)
998 		return;
999 
1000 	for (i = 0; i < LOCKHASH_SIZE; i++) {
1001 		hlist_for_each_entry(st, &lockhash_table[i], hash_entry) {
1002 			combine_lock_stats(st);
1003 		}
1004 	}
1005 }
1006 
1007 static void sort_result(void)
1008 {
1009 	unsigned int i;
1010 	struct lock_stat *st;
1011 
1012 	for (i = 0; i < LOCKHASH_SIZE; i++) {
1013 		hlist_for_each_entry(st, &lockhash_table[i], hash_entry) {
1014 			insert_to_result(st, compare);
1015 		}
1016 	}
1017 }
1018 
1019 static const struct evsel_str_handler lock_tracepoints[] = {
1020 	{ "lock:lock_acquire",	 evsel__process_lock_acquire,   }, /* CONFIG_LOCKDEP */
1021 	{ "lock:lock_acquired",	 evsel__process_lock_acquired,  }, /* CONFIG_LOCKDEP, CONFIG_LOCK_STAT */
1022 	{ "lock:lock_contended", evsel__process_lock_contended, }, /* CONFIG_LOCKDEP, CONFIG_LOCK_STAT */
1023 	{ "lock:lock_release",	 evsel__process_lock_release,   }, /* CONFIG_LOCKDEP */
1024 };
1025 
1026 static bool force;
1027 
1028 static int __cmd_report(bool display_info)
1029 {
1030 	int err = -EINVAL;
1031 	struct perf_tool eops = {
1032 		.sample		 = process_sample_event,
1033 		.comm		 = perf_event__process_comm,
1034 		.namespaces	 = perf_event__process_namespaces,
1035 		.ordered_events	 = true,
1036 	};
1037 	struct perf_data data = {
1038 		.path  = input_name,
1039 		.mode  = PERF_DATA_MODE_READ,
1040 		.force = force,
1041 	};
1042 
1043 	session = perf_session__new(&data, &eops);
1044 	if (IS_ERR(session)) {
1045 		pr_err("Initializing perf session failed\n");
1046 		return PTR_ERR(session);
1047 	}
1048 
1049 	symbol__init(&session->header.env);
1050 
1051 	if (!perf_session__has_traces(session, "lock record"))
1052 		goto out_delete;
1053 
1054 	if (perf_session__set_tracepoints_handlers(session, lock_tracepoints)) {
1055 		pr_err("Initializing perf session tracepoint handlers failed\n");
1056 		goto out_delete;
1057 	}
1058 
1059 	if (setup_output_field(output_fields))
1060 		goto out_delete;
1061 
1062 	if (select_key())
1063 		goto out_delete;
1064 
1065 	err = perf_session__process_events(session);
1066 	if (err)
1067 		goto out_delete;
1068 
1069 	setup_pager();
1070 	if (display_info) /* used for info subcommand */
1071 		err = dump_info();
1072 	else {
1073 		combine_result();
1074 		sort_result();
1075 		print_result();
1076 	}
1077 
1078 out_delete:
1079 	perf_session__delete(session);
1080 	return err;
1081 }
1082 
1083 static int __cmd_record(int argc, const char **argv)
1084 {
1085 	const char *record_args[] = {
1086 		"record", "-R", "-m", "1024", "-c", "1", "--synth", "task",
1087 	};
1088 	unsigned int rec_argc, i, j, ret;
1089 	const char **rec_argv;
1090 
1091 	for (i = 0; i < ARRAY_SIZE(lock_tracepoints); i++) {
1092 		if (!is_valid_tracepoint(lock_tracepoints[i].name)) {
1093 				pr_err("tracepoint %s is not enabled. "
1094 				       "Are CONFIG_LOCKDEP and CONFIG_LOCK_STAT enabled?\n",
1095 				       lock_tracepoints[i].name);
1096 				return 1;
1097 		}
1098 	}
1099 
1100 	rec_argc = ARRAY_SIZE(record_args) + argc - 1;
1101 	/* factor of 2 is for -e in front of each tracepoint */
1102 	rec_argc += 2 * ARRAY_SIZE(lock_tracepoints);
1103 
1104 	rec_argv = calloc(rec_argc + 1, sizeof(char *));
1105 	if (!rec_argv)
1106 		return -ENOMEM;
1107 
1108 	for (i = 0; i < ARRAY_SIZE(record_args); i++)
1109 		rec_argv[i] = strdup(record_args[i]);
1110 
1111 	for (j = 0; j < ARRAY_SIZE(lock_tracepoints); j++) {
1112 		rec_argv[i++] = "-e";
1113 		rec_argv[i++] = strdup(lock_tracepoints[j].name);
1114 	}
1115 
1116 	for (j = 1; j < (unsigned int)argc; j++, i++)
1117 		rec_argv[i] = argv[j];
1118 
1119 	BUG_ON(i != rec_argc);
1120 
1121 	ret = cmd_record(i, rec_argv);
1122 	free(rec_argv);
1123 	return ret;
1124 }
1125 
1126 int cmd_lock(int argc, const char **argv)
1127 {
1128 	const struct option lock_options[] = {
1129 	OPT_STRING('i', "input", &input_name, "file", "input file name"),
1130 	OPT_INCR('v', "verbose", &verbose, "be more verbose (show symbol address, etc)"),
1131 	OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace, "dump raw trace in ASCII"),
1132 	OPT_BOOLEAN('f', "force", &force, "don't complain, do it"),
1133 	OPT_END()
1134 	};
1135 
1136 	const struct option info_options[] = {
1137 	OPT_BOOLEAN('t', "threads", &info_threads,
1138 		    "dump thread list in perf.data"),
1139 	OPT_BOOLEAN('m', "map", &info_map,
1140 		    "map of lock instances (address:name table)"),
1141 	OPT_PARENT(lock_options)
1142 	};
1143 
1144 	const struct option report_options[] = {
1145 	OPT_STRING('k', "key", &sort_key, "acquired",
1146 		    "key for sorting (acquired / contended / avg_wait / wait_total / wait_max / wait_min)"),
1147 	OPT_STRING('F', "field", &output_fields, NULL,
1148 		    "output fields (acquired / contended / avg_wait / wait_total / wait_max / wait_min)"),
1149 	/* TODO: type */
1150 	OPT_BOOLEAN('c', "combine-locks", &combine_locks,
1151 		    "combine locks in the same class"),
1152 	OPT_BOOLEAN('t', "threads", &show_thread_stats,
1153 		    "show per-thread lock stats"),
1154 	OPT_PARENT(lock_options)
1155 	};
1156 
1157 	const char * const info_usage[] = {
1158 		"perf lock info [<options>]",
1159 		NULL
1160 	};
1161 	const char *const lock_subcommands[] = { "record", "report", "script",
1162 						 "info", NULL };
1163 	const char *lock_usage[] = {
1164 		NULL,
1165 		NULL
1166 	};
1167 	const char * const report_usage[] = {
1168 		"perf lock report [<options>]",
1169 		NULL
1170 	};
1171 	unsigned int i;
1172 	int rc = 0;
1173 
1174 	for (i = 0; i < LOCKHASH_SIZE; i++)
1175 		INIT_HLIST_HEAD(lockhash_table + i);
1176 
1177 	argc = parse_options_subcommand(argc, argv, lock_options, lock_subcommands,
1178 					lock_usage, PARSE_OPT_STOP_AT_NON_OPTION);
1179 	if (!argc)
1180 		usage_with_options(lock_usage, lock_options);
1181 
1182 	if (strlen(argv[0]) > 2 && strstarts("record", argv[0])) {
1183 		return __cmd_record(argc, argv);
1184 	} else if (strlen(argv[0]) > 2 && strstarts("report", argv[0])) {
1185 		trace_handler = &report_lock_ops;
1186 		if (argc) {
1187 			argc = parse_options(argc, argv,
1188 					     report_options, report_usage, 0);
1189 			if (argc)
1190 				usage_with_options(report_usage, report_options);
1191 		}
1192 		rc = __cmd_report(false);
1193 	} else if (!strcmp(argv[0], "script")) {
1194 		/* Aliased to 'perf script' */
1195 		return cmd_script(argc, argv);
1196 	} else if (!strcmp(argv[0], "info")) {
1197 		if (argc) {
1198 			argc = parse_options(argc, argv,
1199 					     info_options, info_usage, 0);
1200 			if (argc)
1201 				usage_with_options(info_usage, info_options);
1202 		}
1203 		/* recycling report_lock_ops */
1204 		trace_handler = &report_lock_ops;
1205 		rc = __cmd_report(true);
1206 	} else {
1207 		usage_with_options(lock_usage, lock_options);
1208 	}
1209 
1210 	return rc;
1211 }
1212