xref: /openbmc/linux/tools/perf/builtin-kwork.c (revision 185c8f33)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * builtin-kwork.c
4  *
5  * Copyright (c) 2022  Huawei Inc,  Yang Jihong <yangjihong1@huawei.com>
6  */
7 
8 #include "builtin.h"
9 
10 #include "util/data.h"
11 #include "util/evlist.h"
12 #include "util/evsel.h"
13 #include "util/header.h"
14 #include "util/kwork.h"
15 #include "util/debug.h"
16 #include "util/session.h"
17 #include "util/symbol.h"
18 #include "util/thread.h"
19 #include "util/string2.h"
20 #include "util/callchain.h"
21 #include "util/evsel_fprintf.h"
22 #include "util/util.h"
23 
24 #include <subcmd/pager.h>
25 #include <subcmd/parse-options.h>
26 #include <traceevent/event-parse.h>
27 
28 #include <errno.h>
29 #include <inttypes.h>
30 #include <signal.h>
31 #include <linux/err.h>
32 #include <linux/time64.h>
33 #include <linux/zalloc.h>
34 
35 /*
36  * report header elements width
37  */
38 #define PRINT_CPU_WIDTH 4
39 #define PRINT_COUNT_WIDTH 9
40 #define PRINT_RUNTIME_WIDTH 10
41 #define PRINT_LATENCY_WIDTH 10
42 #define PRINT_TIMESTAMP_WIDTH 17
43 #define PRINT_KWORK_NAME_WIDTH 30
44 #define RPINT_DECIMAL_WIDTH 3
45 #define PRINT_BRACKETPAIR_WIDTH 2
46 #define PRINT_TIME_UNIT_SEC_WIDTH 2
47 #define PRINT_TIME_UNIT_MESC_WIDTH 3
48 #define PRINT_RUNTIME_HEADER_WIDTH (PRINT_RUNTIME_WIDTH + PRINT_TIME_UNIT_MESC_WIDTH)
49 #define PRINT_LATENCY_HEADER_WIDTH (PRINT_LATENCY_WIDTH + PRINT_TIME_UNIT_MESC_WIDTH)
50 #define PRINT_TIMEHIST_CPU_WIDTH (PRINT_CPU_WIDTH + PRINT_BRACKETPAIR_WIDTH)
51 #define PRINT_TIMESTAMP_HEADER_WIDTH (PRINT_TIMESTAMP_WIDTH + PRINT_TIME_UNIT_SEC_WIDTH)
52 
53 struct sort_dimension {
54 	const char      *name;
55 	int             (*cmp)(struct kwork_work *l, struct kwork_work *r);
56 	struct          list_head list;
57 };
58 
59 static int id_cmp(struct kwork_work *l, struct kwork_work *r)
60 {
61 	if (l->cpu > r->cpu)
62 		return 1;
63 	if (l->cpu < r->cpu)
64 		return -1;
65 
66 	if (l->id > r->id)
67 		return 1;
68 	if (l->id < r->id)
69 		return -1;
70 
71 	return 0;
72 }
73 
74 static int count_cmp(struct kwork_work *l, struct kwork_work *r)
75 {
76 	if (l->nr_atoms > r->nr_atoms)
77 		return 1;
78 	if (l->nr_atoms < r->nr_atoms)
79 		return -1;
80 
81 	return 0;
82 }
83 
84 static int runtime_cmp(struct kwork_work *l, struct kwork_work *r)
85 {
86 	if (l->total_runtime > r->total_runtime)
87 		return 1;
88 	if (l->total_runtime < r->total_runtime)
89 		return -1;
90 
91 	return 0;
92 }
93 
94 static int max_runtime_cmp(struct kwork_work *l, struct kwork_work *r)
95 {
96 	if (l->max_runtime > r->max_runtime)
97 		return 1;
98 	if (l->max_runtime < r->max_runtime)
99 		return -1;
100 
101 	return 0;
102 }
103 
104 static int avg_latency_cmp(struct kwork_work *l, struct kwork_work *r)
105 {
106 	u64 avgl, avgr;
107 
108 	if (!r->nr_atoms)
109 		return 1;
110 	if (!l->nr_atoms)
111 		return -1;
112 
113 	avgl = l->total_latency / l->nr_atoms;
114 	avgr = r->total_latency / r->nr_atoms;
115 
116 	if (avgl > avgr)
117 		return 1;
118 	if (avgl < avgr)
119 		return -1;
120 
121 	return 0;
122 }
123 
124 static int max_latency_cmp(struct kwork_work *l, struct kwork_work *r)
125 {
126 	if (l->max_latency > r->max_latency)
127 		return 1;
128 	if (l->max_latency < r->max_latency)
129 		return -1;
130 
131 	return 0;
132 }
133 
134 static int sort_dimension__add(struct perf_kwork *kwork __maybe_unused,
135 			       const char *tok, struct list_head *list)
136 {
137 	size_t i;
138 	static struct sort_dimension max_sort_dimension = {
139 		.name = "max",
140 		.cmp  = max_runtime_cmp,
141 	};
142 	static struct sort_dimension id_sort_dimension = {
143 		.name = "id",
144 		.cmp  = id_cmp,
145 	};
146 	static struct sort_dimension runtime_sort_dimension = {
147 		.name = "runtime",
148 		.cmp  = runtime_cmp,
149 	};
150 	static struct sort_dimension count_sort_dimension = {
151 		.name = "count",
152 		.cmp  = count_cmp,
153 	};
154 	static struct sort_dimension avg_sort_dimension = {
155 		.name = "avg",
156 		.cmp  = avg_latency_cmp,
157 	};
158 	struct sort_dimension *available_sorts[] = {
159 		&id_sort_dimension,
160 		&max_sort_dimension,
161 		&count_sort_dimension,
162 		&runtime_sort_dimension,
163 		&avg_sort_dimension,
164 	};
165 
166 	if (kwork->report == KWORK_REPORT_LATENCY)
167 		max_sort_dimension.cmp = max_latency_cmp;
168 
169 	for (i = 0; i < ARRAY_SIZE(available_sorts); i++) {
170 		if (!strcmp(available_sorts[i]->name, tok)) {
171 			list_add_tail(&available_sorts[i]->list, list);
172 			return 0;
173 		}
174 	}
175 
176 	return -1;
177 }
178 
179 static void setup_sorting(struct perf_kwork *kwork,
180 			  const struct option *options,
181 			  const char * const usage_msg[])
182 {
183 	char *tmp, *tok, *str = strdup(kwork->sort_order);
184 
185 	for (tok = strtok_r(str, ", ", &tmp);
186 	     tok; tok = strtok_r(NULL, ", ", &tmp)) {
187 		if (sort_dimension__add(kwork, tok, &kwork->sort_list) < 0)
188 			usage_with_options_msg(usage_msg, options,
189 					       "Unknown --sort key: `%s'", tok);
190 	}
191 
192 	pr_debug("Sort order: %s\n", kwork->sort_order);
193 	free(str);
194 }
195 
196 static struct kwork_atom *atom_new(struct perf_kwork *kwork,
197 				   struct perf_sample *sample)
198 {
199 	unsigned long i;
200 	struct kwork_atom_page *page;
201 	struct kwork_atom *atom = NULL;
202 
203 	list_for_each_entry(page, &kwork->atom_page_list, list) {
204 		if (!bitmap_full(page->bitmap, NR_ATOM_PER_PAGE)) {
205 			i = find_first_zero_bit(page->bitmap, NR_ATOM_PER_PAGE);
206 			BUG_ON(i >= NR_ATOM_PER_PAGE);
207 			atom = &page->atoms[i];
208 			goto found_atom;
209 		}
210 	}
211 
212 	/*
213 	 * new page
214 	 */
215 	page = zalloc(sizeof(*page));
216 	if (page == NULL) {
217 		pr_err("Failed to zalloc kwork atom page\n");
218 		return NULL;
219 	}
220 
221 	i = 0;
222 	atom = &page->atoms[0];
223 	list_add_tail(&page->list, &kwork->atom_page_list);
224 
225 found_atom:
226 	__set_bit(i, page->bitmap);
227 	atom->time = sample->time;
228 	atom->prev = NULL;
229 	atom->page_addr = page;
230 	atom->bit_inpage = i;
231 	return atom;
232 }
233 
234 static void atom_free(struct kwork_atom *atom)
235 {
236 	if (atom->prev != NULL)
237 		atom_free(atom->prev);
238 
239 	__clear_bit(atom->bit_inpage,
240 		    ((struct kwork_atom_page *)atom->page_addr)->bitmap);
241 }
242 
243 static void atom_del(struct kwork_atom *atom)
244 {
245 	list_del(&atom->list);
246 	atom_free(atom);
247 }
248 
249 static int work_cmp(struct list_head *list,
250 		    struct kwork_work *l, struct kwork_work *r)
251 {
252 	int ret = 0;
253 	struct sort_dimension *sort;
254 
255 	BUG_ON(list_empty(list));
256 
257 	list_for_each_entry(sort, list, list) {
258 		ret = sort->cmp(l, r);
259 		if (ret)
260 			return ret;
261 	}
262 
263 	return ret;
264 }
265 
266 static struct kwork_work *work_search(struct rb_root_cached *root,
267 				      struct kwork_work *key,
268 				      struct list_head *sort_list)
269 {
270 	int cmp;
271 	struct kwork_work *work;
272 	struct rb_node *node = root->rb_root.rb_node;
273 
274 	while (node) {
275 		work = container_of(node, struct kwork_work, node);
276 		cmp = work_cmp(sort_list, key, work);
277 		if (cmp > 0)
278 			node = node->rb_left;
279 		else if (cmp < 0)
280 			node = node->rb_right;
281 		else {
282 			if (work->name == NULL)
283 				work->name = key->name;
284 			return work;
285 		}
286 	}
287 	return NULL;
288 }
289 
290 static void work_insert(struct rb_root_cached *root,
291 			struct kwork_work *key, struct list_head *sort_list)
292 {
293 	int cmp;
294 	bool leftmost = true;
295 	struct kwork_work *cur;
296 	struct rb_node **new = &(root->rb_root.rb_node), *parent = NULL;
297 
298 	while (*new) {
299 		cur = container_of(*new, struct kwork_work, node);
300 		parent = *new;
301 		cmp = work_cmp(sort_list, key, cur);
302 
303 		if (cmp > 0)
304 			new = &((*new)->rb_left);
305 		else {
306 			new = &((*new)->rb_right);
307 			leftmost = false;
308 		}
309 	}
310 
311 	rb_link_node(&key->node, parent, new);
312 	rb_insert_color_cached(&key->node, root, leftmost);
313 }
314 
315 static struct kwork_work *work_new(struct kwork_work *key)
316 {
317 	int i;
318 	struct kwork_work *work = zalloc(sizeof(*work));
319 
320 	if (work == NULL) {
321 		pr_err("Failed to zalloc kwork work\n");
322 		return NULL;
323 	}
324 
325 	for (i = 0; i < KWORK_TRACE_MAX; i++)
326 		INIT_LIST_HEAD(&work->atom_list[i]);
327 
328 	work->id = key->id;
329 	work->cpu = key->cpu;
330 	work->name = key->name;
331 	work->class = key->class;
332 	return work;
333 }
334 
335 static struct kwork_work *work_findnew(struct rb_root_cached *root,
336 				       struct kwork_work *key,
337 				       struct list_head *sort_list)
338 {
339 	struct kwork_work *work = work_search(root, key, sort_list);
340 
341 	if (work != NULL)
342 		return work;
343 
344 	work = work_new(key);
345 	if (work)
346 		work_insert(root, work, sort_list);
347 
348 	return work;
349 }
350 
351 static void profile_update_timespan(struct perf_kwork *kwork,
352 				    struct perf_sample *sample)
353 {
354 	if (!kwork->summary)
355 		return;
356 
357 	if ((kwork->timestart == 0) || (kwork->timestart > sample->time))
358 		kwork->timestart = sample->time;
359 
360 	if (kwork->timeend < sample->time)
361 		kwork->timeend = sample->time;
362 }
363 
364 static bool profile_event_match(struct perf_kwork *kwork,
365 				struct kwork_work *work,
366 				struct perf_sample *sample)
367 {
368 	int cpu = work->cpu;
369 	u64 time = sample->time;
370 	struct perf_time_interval *ptime = &kwork->ptime;
371 
372 	if ((kwork->cpu_list != NULL) && !test_bit(cpu, kwork->cpu_bitmap))
373 		return false;
374 
375 	if (((ptime->start != 0) && (ptime->start > time)) ||
376 	    ((ptime->end != 0) && (ptime->end < time)))
377 		return false;
378 
379 	if ((kwork->profile_name != NULL) &&
380 	    (work->name != NULL) &&
381 	    (strcmp(work->name, kwork->profile_name) != 0))
382 		return false;
383 
384 	profile_update_timespan(kwork, sample);
385 	return true;
386 }
387 
388 static int work_push_atom(struct perf_kwork *kwork,
389 			  struct kwork_class *class,
390 			  enum kwork_trace_type src_type,
391 			  enum kwork_trace_type dst_type,
392 			  struct evsel *evsel,
393 			  struct perf_sample *sample,
394 			  struct machine *machine,
395 			  struct kwork_work **ret_work)
396 {
397 	struct kwork_atom *atom, *dst_atom;
398 	struct kwork_work *work, key;
399 
400 	BUG_ON(class->work_init == NULL);
401 	class->work_init(class, &key, evsel, sample, machine);
402 
403 	atom = atom_new(kwork, sample);
404 	if (atom == NULL)
405 		return -1;
406 
407 	work = work_findnew(&class->work_root, &key, &kwork->cmp_id);
408 	if (work == NULL) {
409 		free(atom);
410 		return -1;
411 	}
412 
413 	if (!profile_event_match(kwork, work, sample))
414 		return 0;
415 
416 	if (dst_type < KWORK_TRACE_MAX) {
417 		dst_atom = list_last_entry_or_null(&work->atom_list[dst_type],
418 						   struct kwork_atom, list);
419 		if (dst_atom != NULL) {
420 			atom->prev = dst_atom;
421 			list_del(&dst_atom->list);
422 		}
423 	}
424 
425 	if (ret_work != NULL)
426 		*ret_work = work;
427 
428 	list_add_tail(&atom->list, &work->atom_list[src_type]);
429 
430 	return 0;
431 }
432 
433 static struct kwork_atom *work_pop_atom(struct perf_kwork *kwork,
434 					struct kwork_class *class,
435 					enum kwork_trace_type src_type,
436 					enum kwork_trace_type dst_type,
437 					struct evsel *evsel,
438 					struct perf_sample *sample,
439 					struct machine *machine,
440 					struct kwork_work **ret_work)
441 {
442 	struct kwork_atom *atom, *src_atom;
443 	struct kwork_work *work, key;
444 
445 	BUG_ON(class->work_init == NULL);
446 	class->work_init(class, &key, evsel, sample, machine);
447 
448 	work = work_findnew(&class->work_root, &key, &kwork->cmp_id);
449 	if (ret_work != NULL)
450 		*ret_work = work;
451 
452 	if (work == NULL)
453 		return NULL;
454 
455 	if (!profile_event_match(kwork, work, sample))
456 		return NULL;
457 
458 	atom = list_last_entry_or_null(&work->atom_list[dst_type],
459 				       struct kwork_atom, list);
460 	if (atom != NULL)
461 		return atom;
462 
463 	src_atom = atom_new(kwork, sample);
464 	if (src_atom != NULL)
465 		list_add_tail(&src_atom->list, &work->atom_list[src_type]);
466 	else {
467 		if (ret_work != NULL)
468 			*ret_work = NULL;
469 	}
470 
471 	return NULL;
472 }
473 
474 static void report_update_exit_event(struct kwork_work *work,
475 				     struct kwork_atom *atom,
476 				     struct perf_sample *sample)
477 {
478 	u64 delta;
479 	u64 exit_time = sample->time;
480 	u64 entry_time = atom->time;
481 
482 	if ((entry_time != 0) && (exit_time >= entry_time)) {
483 		delta = exit_time - entry_time;
484 		if ((delta > work->max_runtime) ||
485 		    (work->max_runtime == 0)) {
486 			work->max_runtime = delta;
487 			work->max_runtime_start = entry_time;
488 			work->max_runtime_end = exit_time;
489 		}
490 		work->total_runtime += delta;
491 		work->nr_atoms++;
492 	}
493 }
494 
495 static int report_entry_event(struct perf_kwork *kwork,
496 			      struct kwork_class *class,
497 			      struct evsel *evsel,
498 			      struct perf_sample *sample,
499 			      struct machine *machine)
500 {
501 	return work_push_atom(kwork, class, KWORK_TRACE_ENTRY,
502 			      KWORK_TRACE_MAX, evsel, sample,
503 			      machine, NULL);
504 }
505 
506 static int report_exit_event(struct perf_kwork *kwork,
507 			     struct kwork_class *class,
508 			     struct evsel *evsel,
509 			     struct perf_sample *sample,
510 			     struct machine *machine)
511 {
512 	struct kwork_atom *atom = NULL;
513 	struct kwork_work *work = NULL;
514 
515 	atom = work_pop_atom(kwork, class, KWORK_TRACE_EXIT,
516 			     KWORK_TRACE_ENTRY, evsel, sample,
517 			     machine, &work);
518 	if (work == NULL)
519 		return -1;
520 
521 	if (atom != NULL) {
522 		report_update_exit_event(work, atom, sample);
523 		atom_del(atom);
524 	}
525 
526 	return 0;
527 }
528 
529 static void latency_update_entry_event(struct kwork_work *work,
530 				       struct kwork_atom *atom,
531 				       struct perf_sample *sample)
532 {
533 	u64 delta;
534 	u64 entry_time = sample->time;
535 	u64 raise_time = atom->time;
536 
537 	if ((raise_time != 0) && (entry_time >= raise_time)) {
538 		delta = entry_time - raise_time;
539 		if ((delta > work->max_latency) ||
540 		    (work->max_latency == 0)) {
541 			work->max_latency = delta;
542 			work->max_latency_start = raise_time;
543 			work->max_latency_end = entry_time;
544 		}
545 		work->total_latency += delta;
546 		work->nr_atoms++;
547 	}
548 }
549 
550 static int latency_raise_event(struct perf_kwork *kwork,
551 			       struct kwork_class *class,
552 			       struct evsel *evsel,
553 			       struct perf_sample *sample,
554 			       struct machine *machine)
555 {
556 	return work_push_atom(kwork, class, KWORK_TRACE_RAISE,
557 			      KWORK_TRACE_MAX, evsel, sample,
558 			      machine, NULL);
559 }
560 
561 static int latency_entry_event(struct perf_kwork *kwork,
562 			       struct kwork_class *class,
563 			       struct evsel *evsel,
564 			       struct perf_sample *sample,
565 			       struct machine *machine)
566 {
567 	struct kwork_atom *atom = NULL;
568 	struct kwork_work *work = NULL;
569 
570 	atom = work_pop_atom(kwork, class, KWORK_TRACE_ENTRY,
571 			     KWORK_TRACE_RAISE, evsel, sample,
572 			     machine, &work);
573 	if (work == NULL)
574 		return -1;
575 
576 	if (atom != NULL) {
577 		latency_update_entry_event(work, atom, sample);
578 		atom_del(atom);
579 	}
580 
581 	return 0;
582 }
583 
584 static void timehist_save_callchain(struct perf_kwork *kwork,
585 				    struct perf_sample *sample,
586 				    struct evsel *evsel,
587 				    struct machine *machine)
588 {
589 	struct symbol *sym;
590 	struct thread *thread;
591 	struct callchain_cursor_node *node;
592 	struct callchain_cursor *cursor;
593 
594 	if (!kwork->show_callchain || sample->callchain == NULL)
595 		return;
596 
597 	/* want main thread for process - has maps */
598 	thread = machine__findnew_thread(machine, sample->pid, sample->pid);
599 	if (thread == NULL) {
600 		pr_debug("Failed to get thread for pid %d\n", sample->pid);
601 		return;
602 	}
603 
604 	cursor = get_tls_callchain_cursor();
605 
606 	if (thread__resolve_callchain(thread, cursor, evsel, sample,
607 				      NULL, NULL, kwork->max_stack + 2) != 0) {
608 		pr_debug("Failed to resolve callchain, skipping\n");
609 		goto out_put;
610 	}
611 
612 	callchain_cursor_commit(cursor);
613 
614 	while (true) {
615 		node = callchain_cursor_current(cursor);
616 		if (node == NULL)
617 			break;
618 
619 		sym = node->ms.sym;
620 		if (sym) {
621 			if (!strcmp(sym->name, "__softirqentry_text_start") ||
622 			    !strcmp(sym->name, "__do_softirq"))
623 				sym->ignore = 1;
624 		}
625 
626 		callchain_cursor_advance(cursor);
627 	}
628 
629 out_put:
630 	thread__put(thread);
631 }
632 
633 static void timehist_print_event(struct perf_kwork *kwork,
634 				 struct kwork_work *work,
635 				 struct kwork_atom *atom,
636 				 struct perf_sample *sample,
637 				 struct addr_location *al)
638 {
639 	char entrytime[32], exittime[32];
640 	char kwork_name[PRINT_KWORK_NAME_WIDTH];
641 
642 	/*
643 	 * runtime start
644 	 */
645 	timestamp__scnprintf_usec(atom->time,
646 				  entrytime, sizeof(entrytime));
647 	printf(" %*s ", PRINT_TIMESTAMP_WIDTH, entrytime);
648 
649 	/*
650 	 * runtime end
651 	 */
652 	timestamp__scnprintf_usec(sample->time,
653 				  exittime, sizeof(exittime));
654 	printf(" %*s ", PRINT_TIMESTAMP_WIDTH, exittime);
655 
656 	/*
657 	 * cpu
658 	 */
659 	printf(" [%0*d] ", PRINT_CPU_WIDTH, work->cpu);
660 
661 	/*
662 	 * kwork name
663 	 */
664 	if (work->class && work->class->work_name) {
665 		work->class->work_name(work, kwork_name,
666 				       PRINT_KWORK_NAME_WIDTH);
667 		printf(" %-*s ", PRINT_KWORK_NAME_WIDTH, kwork_name);
668 	} else
669 		printf(" %-*s ", PRINT_KWORK_NAME_WIDTH, "");
670 
671 	/*
672 	 *runtime
673 	 */
674 	printf(" %*.*f ",
675 	       PRINT_RUNTIME_WIDTH, RPINT_DECIMAL_WIDTH,
676 	       (double)(sample->time - atom->time) / NSEC_PER_MSEC);
677 
678 	/*
679 	 * delaytime
680 	 */
681 	if (atom->prev != NULL)
682 		printf(" %*.*f ", PRINT_LATENCY_WIDTH, RPINT_DECIMAL_WIDTH,
683 		       (double)(atom->time - atom->prev->time) / NSEC_PER_MSEC);
684 	else
685 		printf(" %*s ", PRINT_LATENCY_WIDTH, " ");
686 
687 	/*
688 	 * callchain
689 	 */
690 	if (kwork->show_callchain) {
691 		struct callchain_cursor *cursor = get_tls_callchain_cursor();
692 
693 		if (cursor == NULL)
694 			return;
695 
696 		printf(" ");
697 
698 		sample__fprintf_sym(sample, al, 0,
699 				    EVSEL__PRINT_SYM | EVSEL__PRINT_ONELINE |
700 				    EVSEL__PRINT_CALLCHAIN_ARROW |
701 				    EVSEL__PRINT_SKIP_IGNORED,
702 				    cursor, symbol_conf.bt_stop_list,
703 				    stdout);
704 	}
705 
706 	printf("\n");
707 }
708 
709 static int timehist_raise_event(struct perf_kwork *kwork,
710 				struct kwork_class *class,
711 				struct evsel *evsel,
712 				struct perf_sample *sample,
713 				struct machine *machine)
714 {
715 	return work_push_atom(kwork, class, KWORK_TRACE_RAISE,
716 			      KWORK_TRACE_MAX, evsel, sample,
717 			      machine, NULL);
718 }
719 
720 static int timehist_entry_event(struct perf_kwork *kwork,
721 				struct kwork_class *class,
722 				struct evsel *evsel,
723 				struct perf_sample *sample,
724 				struct machine *machine)
725 {
726 	int ret;
727 	struct kwork_work *work = NULL;
728 
729 	ret = work_push_atom(kwork, class, KWORK_TRACE_ENTRY,
730 			     KWORK_TRACE_RAISE, evsel, sample,
731 			     machine, &work);
732 	if (ret)
733 		return ret;
734 
735 	if (work != NULL)
736 		timehist_save_callchain(kwork, sample, evsel, machine);
737 
738 	return 0;
739 }
740 
741 static int timehist_exit_event(struct perf_kwork *kwork,
742 			       struct kwork_class *class,
743 			       struct evsel *evsel,
744 			       struct perf_sample *sample,
745 			       struct machine *machine)
746 {
747 	struct kwork_atom *atom = NULL;
748 	struct kwork_work *work = NULL;
749 	struct addr_location al;
750 	int ret = 0;
751 
752 	addr_location__init(&al);
753 	if (machine__resolve(machine, &al, sample) < 0) {
754 		pr_debug("Problem processing event, skipping it\n");
755 		ret = -1;
756 		goto out;
757 	}
758 
759 	atom = work_pop_atom(kwork, class, KWORK_TRACE_EXIT,
760 			     KWORK_TRACE_ENTRY, evsel, sample,
761 			     machine, &work);
762 	if (work == NULL) {
763 		ret = -1;
764 		goto out;
765 	}
766 
767 	if (atom != NULL) {
768 		work->nr_atoms++;
769 		timehist_print_event(kwork, work, atom, sample, &al);
770 		atom_del(atom);
771 	}
772 
773 out:
774 	addr_location__exit(&al);
775 	return ret;
776 }
777 
778 static struct kwork_class kwork_irq;
779 static int process_irq_handler_entry_event(struct perf_tool *tool,
780 					   struct evsel *evsel,
781 					   struct perf_sample *sample,
782 					   struct machine *machine)
783 {
784 	struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool);
785 
786 	if (kwork->tp_handler->entry_event)
787 		return kwork->tp_handler->entry_event(kwork, &kwork_irq,
788 						      evsel, sample, machine);
789 	return 0;
790 }
791 
792 static int process_irq_handler_exit_event(struct perf_tool *tool,
793 					  struct evsel *evsel,
794 					  struct perf_sample *sample,
795 					  struct machine *machine)
796 {
797 	struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool);
798 
799 	if (kwork->tp_handler->exit_event)
800 		return kwork->tp_handler->exit_event(kwork, &kwork_irq,
801 						     evsel, sample, machine);
802 	return 0;
803 }
804 
805 const struct evsel_str_handler irq_tp_handlers[] = {
806 	{ "irq:irq_handler_entry", process_irq_handler_entry_event, },
807 	{ "irq:irq_handler_exit",  process_irq_handler_exit_event,  },
808 };
809 
810 static int irq_class_init(struct kwork_class *class,
811 			  struct perf_session *session)
812 {
813 	if (perf_session__set_tracepoints_handlers(session, irq_tp_handlers)) {
814 		pr_err("Failed to set irq tracepoints handlers\n");
815 		return -1;
816 	}
817 
818 	class->work_root = RB_ROOT_CACHED;
819 	return 0;
820 }
821 
822 static void irq_work_init(struct kwork_class *class,
823 			  struct kwork_work *work,
824 			  struct evsel *evsel,
825 			  struct perf_sample *sample,
826 			  struct machine *machine __maybe_unused)
827 {
828 	work->class = class;
829 	work->cpu = sample->cpu;
830 	work->id = evsel__intval(evsel, sample, "irq");
831 	work->name = evsel__strval(evsel, sample, "name");
832 }
833 
834 static void irq_work_name(struct kwork_work *work, char *buf, int len)
835 {
836 	snprintf(buf, len, "%s:%" PRIu64 "", work->name, work->id);
837 }
838 
839 static struct kwork_class kwork_irq = {
840 	.name           = "irq",
841 	.type           = KWORK_CLASS_IRQ,
842 	.nr_tracepoints = 2,
843 	.tp_handlers    = irq_tp_handlers,
844 	.class_init     = irq_class_init,
845 	.work_init      = irq_work_init,
846 	.work_name      = irq_work_name,
847 };
848 
849 static struct kwork_class kwork_softirq;
850 static int process_softirq_raise_event(struct perf_tool *tool,
851 				       struct evsel *evsel,
852 				       struct perf_sample *sample,
853 				       struct machine *machine)
854 {
855 	struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool);
856 
857 	if (kwork->tp_handler->raise_event)
858 		return kwork->tp_handler->raise_event(kwork, &kwork_softirq,
859 						      evsel, sample, machine);
860 
861 	return 0;
862 }
863 
864 static int process_softirq_entry_event(struct perf_tool *tool,
865 				       struct evsel *evsel,
866 				       struct perf_sample *sample,
867 				       struct machine *machine)
868 {
869 	struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool);
870 
871 	if (kwork->tp_handler->entry_event)
872 		return kwork->tp_handler->entry_event(kwork, &kwork_softirq,
873 						      evsel, sample, machine);
874 
875 	return 0;
876 }
877 
878 static int process_softirq_exit_event(struct perf_tool *tool,
879 				      struct evsel *evsel,
880 				      struct perf_sample *sample,
881 				      struct machine *machine)
882 {
883 	struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool);
884 
885 	if (kwork->tp_handler->exit_event)
886 		return kwork->tp_handler->exit_event(kwork, &kwork_softirq,
887 						     evsel, sample, machine);
888 
889 	return 0;
890 }
891 
892 const struct evsel_str_handler softirq_tp_handlers[] = {
893 	{ "irq:softirq_raise", process_softirq_raise_event, },
894 	{ "irq:softirq_entry", process_softirq_entry_event, },
895 	{ "irq:softirq_exit",  process_softirq_exit_event,  },
896 };
897 
898 static int softirq_class_init(struct kwork_class *class,
899 			      struct perf_session *session)
900 {
901 	if (perf_session__set_tracepoints_handlers(session,
902 						   softirq_tp_handlers)) {
903 		pr_err("Failed to set softirq tracepoints handlers\n");
904 		return -1;
905 	}
906 
907 	class->work_root = RB_ROOT_CACHED;
908 	return 0;
909 }
910 
911 static char *evsel__softirq_name(struct evsel *evsel, u64 num)
912 {
913 	char *name = NULL;
914 	bool found = false;
915 	struct tep_print_flag_sym *sym = NULL;
916 	struct tep_print_arg *args = evsel->tp_format->print_fmt.args;
917 
918 	if ((args == NULL) || (args->next == NULL))
919 		return NULL;
920 
921 	/* skip softirq field: "REC->vec" */
922 	for (sym = args->next->symbol.symbols; sym != NULL; sym = sym->next) {
923 		if ((eval_flag(sym->value) == (unsigned long long)num) &&
924 		    (strlen(sym->str) != 0)) {
925 			found = true;
926 			break;
927 		}
928 	}
929 
930 	if (!found)
931 		return NULL;
932 
933 	name = strdup(sym->str);
934 	if (name == NULL) {
935 		pr_err("Failed to copy symbol name\n");
936 		return NULL;
937 	}
938 	return name;
939 }
940 
941 static void softirq_work_init(struct kwork_class *class,
942 			      struct kwork_work *work,
943 			      struct evsel *evsel,
944 			      struct perf_sample *sample,
945 			      struct machine *machine __maybe_unused)
946 {
947 	u64 num = evsel__intval(evsel, sample, "vec");
948 
949 	work->id = num;
950 	work->class = class;
951 	work->cpu = sample->cpu;
952 	work->name = evsel__softirq_name(evsel, num);
953 }
954 
955 static void softirq_work_name(struct kwork_work *work, char *buf, int len)
956 {
957 	snprintf(buf, len, "(s)%s:%" PRIu64 "", work->name, work->id);
958 }
959 
960 static struct kwork_class kwork_softirq = {
961 	.name           = "softirq",
962 	.type           = KWORK_CLASS_SOFTIRQ,
963 	.nr_tracepoints = 3,
964 	.tp_handlers    = softirq_tp_handlers,
965 	.class_init     = softirq_class_init,
966 	.work_init      = softirq_work_init,
967 	.work_name      = softirq_work_name,
968 };
969 
970 static struct kwork_class kwork_workqueue;
971 static int process_workqueue_activate_work_event(struct perf_tool *tool,
972 						 struct evsel *evsel,
973 						 struct perf_sample *sample,
974 						 struct machine *machine)
975 {
976 	struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool);
977 
978 	if (kwork->tp_handler->raise_event)
979 		return kwork->tp_handler->raise_event(kwork, &kwork_workqueue,
980 						    evsel, sample, machine);
981 
982 	return 0;
983 }
984 
985 static int process_workqueue_execute_start_event(struct perf_tool *tool,
986 						 struct evsel *evsel,
987 						 struct perf_sample *sample,
988 						 struct machine *machine)
989 {
990 	struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool);
991 
992 	if (kwork->tp_handler->entry_event)
993 		return kwork->tp_handler->entry_event(kwork, &kwork_workqueue,
994 						    evsel, sample, machine);
995 
996 	return 0;
997 }
998 
999 static int process_workqueue_execute_end_event(struct perf_tool *tool,
1000 					       struct evsel *evsel,
1001 					       struct perf_sample *sample,
1002 					       struct machine *machine)
1003 {
1004 	struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool);
1005 
1006 	if (kwork->tp_handler->exit_event)
1007 		return kwork->tp_handler->exit_event(kwork, &kwork_workqueue,
1008 						   evsel, sample, machine);
1009 
1010 	return 0;
1011 }
1012 
1013 const struct evsel_str_handler workqueue_tp_handlers[] = {
1014 	{ "workqueue:workqueue_activate_work", process_workqueue_activate_work_event, },
1015 	{ "workqueue:workqueue_execute_start", process_workqueue_execute_start_event, },
1016 	{ "workqueue:workqueue_execute_end",   process_workqueue_execute_end_event,   },
1017 };
1018 
1019 static int workqueue_class_init(struct kwork_class *class,
1020 				struct perf_session *session)
1021 {
1022 	if (perf_session__set_tracepoints_handlers(session,
1023 						   workqueue_tp_handlers)) {
1024 		pr_err("Failed to set workqueue tracepoints handlers\n");
1025 		return -1;
1026 	}
1027 
1028 	class->work_root = RB_ROOT_CACHED;
1029 	return 0;
1030 }
1031 
1032 static void workqueue_work_init(struct kwork_class *class,
1033 				struct kwork_work *work,
1034 				struct evsel *evsel,
1035 				struct perf_sample *sample,
1036 				struct machine *machine)
1037 {
1038 	char *modp = NULL;
1039 	unsigned long long function_addr = evsel__intval(evsel,
1040 							 sample, "function");
1041 
1042 	work->class = class;
1043 	work->cpu = sample->cpu;
1044 	work->id = evsel__intval(evsel, sample, "work");
1045 	work->name = function_addr == 0 ? NULL :
1046 		machine__resolve_kernel_addr(machine, &function_addr, &modp);
1047 }
1048 
1049 static void workqueue_work_name(struct kwork_work *work, char *buf, int len)
1050 {
1051 	if (work->name != NULL)
1052 		snprintf(buf, len, "(w)%s", work->name);
1053 	else
1054 		snprintf(buf, len, "(w)0x%" PRIx64, work->id);
1055 }
1056 
1057 static struct kwork_class kwork_workqueue = {
1058 	.name           = "workqueue",
1059 	.type           = KWORK_CLASS_WORKQUEUE,
1060 	.nr_tracepoints = 3,
1061 	.tp_handlers    = workqueue_tp_handlers,
1062 	.class_init     = workqueue_class_init,
1063 	.work_init      = workqueue_work_init,
1064 	.work_name      = workqueue_work_name,
1065 };
1066 
1067 static struct kwork_class *kwork_class_supported_list[KWORK_CLASS_MAX] = {
1068 	[KWORK_CLASS_IRQ]       = &kwork_irq,
1069 	[KWORK_CLASS_SOFTIRQ]   = &kwork_softirq,
1070 	[KWORK_CLASS_WORKQUEUE] = &kwork_workqueue,
1071 };
1072 
1073 static void print_separator(int len)
1074 {
1075 	printf(" %.*s\n", len, graph_dotted_line);
1076 }
1077 
1078 static int report_print_work(struct perf_kwork *kwork, struct kwork_work *work)
1079 {
1080 	int ret = 0;
1081 	char kwork_name[PRINT_KWORK_NAME_WIDTH];
1082 	char max_runtime_start[32], max_runtime_end[32];
1083 	char max_latency_start[32], max_latency_end[32];
1084 
1085 	printf(" ");
1086 
1087 	/*
1088 	 * kwork name
1089 	 */
1090 	if (work->class && work->class->work_name) {
1091 		work->class->work_name(work, kwork_name,
1092 				       PRINT_KWORK_NAME_WIDTH);
1093 		ret += printf(" %-*s |", PRINT_KWORK_NAME_WIDTH, kwork_name);
1094 	} else {
1095 		ret += printf(" %-*s |", PRINT_KWORK_NAME_WIDTH, "");
1096 	}
1097 
1098 	/*
1099 	 * cpu
1100 	 */
1101 	ret += printf(" %0*d |", PRINT_CPU_WIDTH, work->cpu);
1102 
1103 	/*
1104 	 * total runtime
1105 	 */
1106 	if (kwork->report == KWORK_REPORT_RUNTIME) {
1107 		ret += printf(" %*.*f ms |",
1108 			      PRINT_RUNTIME_WIDTH, RPINT_DECIMAL_WIDTH,
1109 			      (double)work->total_runtime / NSEC_PER_MSEC);
1110 	} else if (kwork->report == KWORK_REPORT_LATENCY) { // avg delay
1111 		ret += printf(" %*.*f ms |",
1112 			      PRINT_LATENCY_WIDTH, RPINT_DECIMAL_WIDTH,
1113 			      (double)work->total_latency /
1114 			      work->nr_atoms / NSEC_PER_MSEC);
1115 	}
1116 
1117 	/*
1118 	 * count
1119 	 */
1120 	ret += printf(" %*" PRIu64 " |", PRINT_COUNT_WIDTH, work->nr_atoms);
1121 
1122 	/*
1123 	 * max runtime, max runtime start, max runtime end
1124 	 */
1125 	if (kwork->report == KWORK_REPORT_RUNTIME) {
1126 		timestamp__scnprintf_usec(work->max_runtime_start,
1127 					  max_runtime_start,
1128 					  sizeof(max_runtime_start));
1129 		timestamp__scnprintf_usec(work->max_runtime_end,
1130 					  max_runtime_end,
1131 					  sizeof(max_runtime_end));
1132 		ret += printf(" %*.*f ms | %*s s | %*s s |",
1133 			      PRINT_RUNTIME_WIDTH, RPINT_DECIMAL_WIDTH,
1134 			      (double)work->max_runtime / NSEC_PER_MSEC,
1135 			      PRINT_TIMESTAMP_WIDTH, max_runtime_start,
1136 			      PRINT_TIMESTAMP_WIDTH, max_runtime_end);
1137 	}
1138 	/*
1139 	 * max delay, max delay start, max delay end
1140 	 */
1141 	else if (kwork->report == KWORK_REPORT_LATENCY) {
1142 		timestamp__scnprintf_usec(work->max_latency_start,
1143 					  max_latency_start,
1144 					  sizeof(max_latency_start));
1145 		timestamp__scnprintf_usec(work->max_latency_end,
1146 					  max_latency_end,
1147 					  sizeof(max_latency_end));
1148 		ret += printf(" %*.*f ms | %*s s | %*s s |",
1149 			      PRINT_LATENCY_WIDTH, RPINT_DECIMAL_WIDTH,
1150 			      (double)work->max_latency / NSEC_PER_MSEC,
1151 			      PRINT_TIMESTAMP_WIDTH, max_latency_start,
1152 			      PRINT_TIMESTAMP_WIDTH, max_latency_end);
1153 	}
1154 
1155 	printf("\n");
1156 	return ret;
1157 }
1158 
1159 static int report_print_header(struct perf_kwork *kwork)
1160 {
1161 	int ret;
1162 
1163 	printf("\n ");
1164 	ret = printf(" %-*s | %-*s |",
1165 		     PRINT_KWORK_NAME_WIDTH, "Kwork Name",
1166 		     PRINT_CPU_WIDTH, "Cpu");
1167 
1168 	if (kwork->report == KWORK_REPORT_RUNTIME) {
1169 		ret += printf(" %-*s |",
1170 			      PRINT_RUNTIME_HEADER_WIDTH, "Total Runtime");
1171 	} else if (kwork->report == KWORK_REPORT_LATENCY) {
1172 		ret += printf(" %-*s |",
1173 			      PRINT_LATENCY_HEADER_WIDTH, "Avg delay");
1174 	}
1175 
1176 	ret += printf(" %-*s |", PRINT_COUNT_WIDTH, "Count");
1177 
1178 	if (kwork->report == KWORK_REPORT_RUNTIME) {
1179 		ret += printf(" %-*s | %-*s | %-*s |",
1180 			      PRINT_RUNTIME_HEADER_WIDTH, "Max runtime",
1181 			      PRINT_TIMESTAMP_HEADER_WIDTH, "Max runtime start",
1182 			      PRINT_TIMESTAMP_HEADER_WIDTH, "Max runtime end");
1183 	} else if (kwork->report == KWORK_REPORT_LATENCY) {
1184 		ret += printf(" %-*s | %-*s | %-*s |",
1185 			      PRINT_LATENCY_HEADER_WIDTH, "Max delay",
1186 			      PRINT_TIMESTAMP_HEADER_WIDTH, "Max delay start",
1187 			      PRINT_TIMESTAMP_HEADER_WIDTH, "Max delay end");
1188 	}
1189 
1190 	printf("\n");
1191 	print_separator(ret);
1192 	return ret;
1193 }
1194 
1195 static void timehist_print_header(void)
1196 {
1197 	/*
1198 	 * header row
1199 	 */
1200 	printf(" %-*s  %-*s  %-*s  %-*s  %-*s  %-*s\n",
1201 	       PRINT_TIMESTAMP_WIDTH, "Runtime start",
1202 	       PRINT_TIMESTAMP_WIDTH, "Runtime end",
1203 	       PRINT_TIMEHIST_CPU_WIDTH, "Cpu",
1204 	       PRINT_KWORK_NAME_WIDTH, "Kwork name",
1205 	       PRINT_RUNTIME_WIDTH, "Runtime",
1206 	       PRINT_RUNTIME_WIDTH, "Delaytime");
1207 
1208 	/*
1209 	 * units row
1210 	 */
1211 	printf(" %-*s  %-*s  %-*s  %-*s  %-*s  %-*s\n",
1212 	       PRINT_TIMESTAMP_WIDTH, "",
1213 	       PRINT_TIMESTAMP_WIDTH, "",
1214 	       PRINT_TIMEHIST_CPU_WIDTH, "",
1215 	       PRINT_KWORK_NAME_WIDTH, "(TYPE)NAME:NUM",
1216 	       PRINT_RUNTIME_WIDTH, "(msec)",
1217 	       PRINT_RUNTIME_WIDTH, "(msec)");
1218 
1219 	/*
1220 	 * separator
1221 	 */
1222 	printf(" %.*s  %.*s  %.*s  %.*s  %.*s  %.*s\n",
1223 	       PRINT_TIMESTAMP_WIDTH, graph_dotted_line,
1224 	       PRINT_TIMESTAMP_WIDTH, graph_dotted_line,
1225 	       PRINT_TIMEHIST_CPU_WIDTH, graph_dotted_line,
1226 	       PRINT_KWORK_NAME_WIDTH, graph_dotted_line,
1227 	       PRINT_RUNTIME_WIDTH, graph_dotted_line,
1228 	       PRINT_RUNTIME_WIDTH, graph_dotted_line);
1229 }
1230 
1231 static void print_summary(struct perf_kwork *kwork)
1232 {
1233 	u64 time = kwork->timeend - kwork->timestart;
1234 
1235 	printf("  Total count            : %9" PRIu64 "\n", kwork->all_count);
1236 	printf("  Total runtime   (msec) : %9.3f (%.3f%% load average)\n",
1237 	       (double)kwork->all_runtime / NSEC_PER_MSEC,
1238 	       time == 0 ? 0 : (double)kwork->all_runtime / time);
1239 	printf("  Total time span (msec) : %9.3f\n",
1240 	       (double)time / NSEC_PER_MSEC);
1241 }
1242 
1243 static unsigned long long nr_list_entry(struct list_head *head)
1244 {
1245 	struct list_head *pos;
1246 	unsigned long long n = 0;
1247 
1248 	list_for_each(pos, head)
1249 		n++;
1250 
1251 	return n;
1252 }
1253 
1254 static void print_skipped_events(struct perf_kwork *kwork)
1255 {
1256 	int i;
1257 	const char *const kwork_event_str[] = {
1258 		[KWORK_TRACE_RAISE] = "raise",
1259 		[KWORK_TRACE_ENTRY] = "entry",
1260 		[KWORK_TRACE_EXIT]  = "exit",
1261 	};
1262 
1263 	if ((kwork->nr_skipped_events[KWORK_TRACE_MAX] != 0) &&
1264 	    (kwork->nr_events != 0)) {
1265 		printf("  INFO: %.3f%% skipped events (%" PRIu64 " including ",
1266 		       (double)kwork->nr_skipped_events[KWORK_TRACE_MAX] /
1267 		       (double)kwork->nr_events * 100.0,
1268 		       kwork->nr_skipped_events[KWORK_TRACE_MAX]);
1269 
1270 		for (i = 0; i < KWORK_TRACE_MAX; i++) {
1271 			printf("%" PRIu64 " %s%s",
1272 			       kwork->nr_skipped_events[i],
1273 			       kwork_event_str[i],
1274 			       (i == KWORK_TRACE_MAX - 1) ? ")\n" : ", ");
1275 		}
1276 	}
1277 
1278 	if (verbose > 0)
1279 		printf("  INFO: use %lld atom pages\n",
1280 		       nr_list_entry(&kwork->atom_page_list));
1281 }
1282 
1283 static void print_bad_events(struct perf_kwork *kwork)
1284 {
1285 	if ((kwork->nr_lost_events != 0) && (kwork->nr_events != 0)) {
1286 		printf("  INFO: %.3f%% lost events (%ld out of %ld, in %ld chunks)\n",
1287 		       (double)kwork->nr_lost_events /
1288 		       (double)kwork->nr_events * 100.0,
1289 		       kwork->nr_lost_events, kwork->nr_events,
1290 		       kwork->nr_lost_chunks);
1291 	}
1292 }
1293 
1294 static void work_sort(struct perf_kwork *kwork, struct kwork_class *class)
1295 {
1296 	struct rb_node *node;
1297 	struct kwork_work *data;
1298 	struct rb_root_cached *root = &class->work_root;
1299 
1300 	pr_debug("Sorting %s ...\n", class->name);
1301 	for (;;) {
1302 		node = rb_first_cached(root);
1303 		if (!node)
1304 			break;
1305 
1306 		rb_erase_cached(node, root);
1307 		data = rb_entry(node, struct kwork_work, node);
1308 		work_insert(&kwork->sorted_work_root,
1309 			       data, &kwork->sort_list);
1310 	}
1311 }
1312 
1313 static void perf_kwork__sort(struct perf_kwork *kwork)
1314 {
1315 	struct kwork_class *class;
1316 
1317 	list_for_each_entry(class, &kwork->class_list, list)
1318 		work_sort(kwork, class);
1319 }
1320 
1321 static int perf_kwork__check_config(struct perf_kwork *kwork,
1322 				    struct perf_session *session)
1323 {
1324 	int ret;
1325 	struct evsel *evsel;
1326 	struct kwork_class *class;
1327 
1328 	static struct trace_kwork_handler report_ops = {
1329 		.entry_event = report_entry_event,
1330 		.exit_event  = report_exit_event,
1331 	};
1332 	static struct trace_kwork_handler latency_ops = {
1333 		.raise_event = latency_raise_event,
1334 		.entry_event = latency_entry_event,
1335 	};
1336 	static struct trace_kwork_handler timehist_ops = {
1337 		.raise_event = timehist_raise_event,
1338 		.entry_event = timehist_entry_event,
1339 		.exit_event  = timehist_exit_event,
1340 	};
1341 
1342 	switch (kwork->report) {
1343 	case KWORK_REPORT_RUNTIME:
1344 		kwork->tp_handler = &report_ops;
1345 		break;
1346 	case KWORK_REPORT_LATENCY:
1347 		kwork->tp_handler = &latency_ops;
1348 		break;
1349 	case KWORK_REPORT_TIMEHIST:
1350 		kwork->tp_handler = &timehist_ops;
1351 		break;
1352 	default:
1353 		pr_debug("Invalid report type %d\n", kwork->report);
1354 		return -1;
1355 	}
1356 
1357 	list_for_each_entry(class, &kwork->class_list, list)
1358 		if ((class->class_init != NULL) &&
1359 		    (class->class_init(class, session) != 0))
1360 			return -1;
1361 
1362 	if (kwork->cpu_list != NULL) {
1363 		ret = perf_session__cpu_bitmap(session,
1364 					       kwork->cpu_list,
1365 					       kwork->cpu_bitmap);
1366 		if (ret < 0) {
1367 			pr_err("Invalid cpu bitmap\n");
1368 			return -1;
1369 		}
1370 	}
1371 
1372 	if (kwork->time_str != NULL) {
1373 		ret = perf_time__parse_str(&kwork->ptime, kwork->time_str);
1374 		if (ret != 0) {
1375 			pr_err("Invalid time span\n");
1376 			return -1;
1377 		}
1378 	}
1379 
1380 	list_for_each_entry(evsel, &session->evlist->core.entries, core.node) {
1381 		if (kwork->show_callchain && !evsel__has_callchain(evsel)) {
1382 			pr_debug("Samples do not have callchains\n");
1383 			kwork->show_callchain = 0;
1384 			symbol_conf.use_callchain = 0;
1385 		}
1386 	}
1387 
1388 	return 0;
1389 }
1390 
1391 static int perf_kwork__read_events(struct perf_kwork *kwork)
1392 {
1393 	int ret = -1;
1394 	struct perf_session *session = NULL;
1395 
1396 	struct perf_data data = {
1397 		.path  = input_name,
1398 		.mode  = PERF_DATA_MODE_READ,
1399 		.force = kwork->force,
1400 	};
1401 
1402 	session = perf_session__new(&data, &kwork->tool);
1403 	if (IS_ERR(session)) {
1404 		pr_debug("Error creating perf session\n");
1405 		return PTR_ERR(session);
1406 	}
1407 
1408 	symbol__init(&session->header.env);
1409 
1410 	if (perf_kwork__check_config(kwork, session) != 0)
1411 		goto out_delete;
1412 
1413 	if (session->tevent.pevent &&
1414 	    tep_set_function_resolver(session->tevent.pevent,
1415 				      machine__resolve_kernel_addr,
1416 				      &session->machines.host) < 0) {
1417 		pr_err("Failed to set libtraceevent function resolver\n");
1418 		goto out_delete;
1419 	}
1420 
1421 	if (kwork->report == KWORK_REPORT_TIMEHIST)
1422 		timehist_print_header();
1423 
1424 	ret = perf_session__process_events(session);
1425 	if (ret) {
1426 		pr_debug("Failed to process events, error %d\n", ret);
1427 		goto out_delete;
1428 	}
1429 
1430 	kwork->nr_events      = session->evlist->stats.nr_events[0];
1431 	kwork->nr_lost_events = session->evlist->stats.total_lost;
1432 	kwork->nr_lost_chunks = session->evlist->stats.nr_events[PERF_RECORD_LOST];
1433 
1434 out_delete:
1435 	perf_session__delete(session);
1436 	return ret;
1437 }
1438 
1439 static void process_skipped_events(struct perf_kwork *kwork,
1440 				   struct kwork_work *work)
1441 {
1442 	int i;
1443 	unsigned long long count;
1444 
1445 	for (i = 0; i < KWORK_TRACE_MAX; i++) {
1446 		count = nr_list_entry(&work->atom_list[i]);
1447 		kwork->nr_skipped_events[i] += count;
1448 		kwork->nr_skipped_events[KWORK_TRACE_MAX] += count;
1449 	}
1450 }
1451 
1452 struct kwork_work *perf_kwork_add_work(struct perf_kwork *kwork,
1453 				       struct kwork_class *class,
1454 				       struct kwork_work *key)
1455 {
1456 	struct kwork_work *work = NULL;
1457 
1458 	work = work_new(key);
1459 	if (work == NULL)
1460 		return NULL;
1461 
1462 	work_insert(&class->work_root, work, &kwork->cmp_id);
1463 	return work;
1464 }
1465 
1466 static void sig_handler(int sig)
1467 {
1468 	/*
1469 	 * Simply capture termination signal so that
1470 	 * the program can continue after pause returns
1471 	 */
1472 	pr_debug("Captuer signal %d\n", sig);
1473 }
1474 
1475 static int perf_kwork__report_bpf(struct perf_kwork *kwork)
1476 {
1477 	int ret;
1478 
1479 	signal(SIGINT, sig_handler);
1480 	signal(SIGTERM, sig_handler);
1481 
1482 	ret = perf_kwork__trace_prepare_bpf(kwork);
1483 	if (ret)
1484 		return -1;
1485 
1486 	printf("Starting trace, Hit <Ctrl+C> to stop and report\n");
1487 
1488 	perf_kwork__trace_start();
1489 
1490 	/*
1491 	 * a simple pause, wait here for stop signal
1492 	 */
1493 	pause();
1494 
1495 	perf_kwork__trace_finish();
1496 
1497 	perf_kwork__report_read_bpf(kwork);
1498 
1499 	perf_kwork__report_cleanup_bpf();
1500 
1501 	return 0;
1502 }
1503 
1504 static int perf_kwork__report(struct perf_kwork *kwork)
1505 {
1506 	int ret;
1507 	struct rb_node *next;
1508 	struct kwork_work *work;
1509 
1510 	if (kwork->use_bpf)
1511 		ret = perf_kwork__report_bpf(kwork);
1512 	else
1513 		ret = perf_kwork__read_events(kwork);
1514 
1515 	if (ret != 0)
1516 		return -1;
1517 
1518 	perf_kwork__sort(kwork);
1519 
1520 	setup_pager();
1521 
1522 	ret = report_print_header(kwork);
1523 	next = rb_first_cached(&kwork->sorted_work_root);
1524 	while (next) {
1525 		work = rb_entry(next, struct kwork_work, node);
1526 		process_skipped_events(kwork, work);
1527 
1528 		if (work->nr_atoms != 0) {
1529 			report_print_work(kwork, work);
1530 			if (kwork->summary) {
1531 				kwork->all_runtime += work->total_runtime;
1532 				kwork->all_count += work->nr_atoms;
1533 			}
1534 		}
1535 		next = rb_next(next);
1536 	}
1537 	print_separator(ret);
1538 
1539 	if (kwork->summary) {
1540 		print_summary(kwork);
1541 		print_separator(ret);
1542 	}
1543 
1544 	print_bad_events(kwork);
1545 	print_skipped_events(kwork);
1546 	printf("\n");
1547 
1548 	return 0;
1549 }
1550 
1551 typedef int (*tracepoint_handler)(struct perf_tool *tool,
1552 				  struct evsel *evsel,
1553 				  struct perf_sample *sample,
1554 				  struct machine *machine);
1555 
1556 static int perf_kwork__process_tracepoint_sample(struct perf_tool *tool,
1557 						 union perf_event *event __maybe_unused,
1558 						 struct perf_sample *sample,
1559 						 struct evsel *evsel,
1560 						 struct machine *machine)
1561 {
1562 	int err = 0;
1563 
1564 	if (evsel->handler != NULL) {
1565 		tracepoint_handler f = evsel->handler;
1566 
1567 		err = f(tool, evsel, sample, machine);
1568 	}
1569 
1570 	return err;
1571 }
1572 
1573 static int perf_kwork__timehist(struct perf_kwork *kwork)
1574 {
1575 	/*
1576 	 * event handlers for timehist option
1577 	 */
1578 	kwork->tool.comm	 = perf_event__process_comm;
1579 	kwork->tool.exit	 = perf_event__process_exit;
1580 	kwork->tool.fork	 = perf_event__process_fork;
1581 	kwork->tool.attr	 = perf_event__process_attr;
1582 	kwork->tool.tracing_data = perf_event__process_tracing_data;
1583 	kwork->tool.build_id	 = perf_event__process_build_id;
1584 	kwork->tool.ordered_events = true;
1585 	kwork->tool.ordering_requires_timestamps = true;
1586 	symbol_conf.use_callchain = kwork->show_callchain;
1587 
1588 	if (symbol__validate_sym_arguments()) {
1589 		pr_err("Failed to validate sym arguments\n");
1590 		return -1;
1591 	}
1592 
1593 	setup_pager();
1594 
1595 	return perf_kwork__read_events(kwork);
1596 }
1597 
1598 static void setup_event_list(struct perf_kwork *kwork,
1599 			     const struct option *options,
1600 			     const char * const usage_msg[])
1601 {
1602 	int i;
1603 	struct kwork_class *class;
1604 	char *tmp, *tok, *str;
1605 
1606 	if (kwork->event_list_str == NULL)
1607 		goto null_event_list_str;
1608 
1609 	str = strdup(kwork->event_list_str);
1610 	for (tok = strtok_r(str, ", ", &tmp);
1611 	     tok; tok = strtok_r(NULL, ", ", &tmp)) {
1612 		for (i = 0; i < KWORK_CLASS_MAX; i++) {
1613 			class = kwork_class_supported_list[i];
1614 			if (strcmp(tok, class->name) == 0) {
1615 				list_add_tail(&class->list, &kwork->class_list);
1616 				break;
1617 			}
1618 		}
1619 		if (i == KWORK_CLASS_MAX) {
1620 			usage_with_options_msg(usage_msg, options,
1621 					       "Unknown --event key: `%s'", tok);
1622 		}
1623 	}
1624 	free(str);
1625 
1626 null_event_list_str:
1627 	/*
1628 	 * config all kwork events if not specified
1629 	 */
1630 	if (list_empty(&kwork->class_list)) {
1631 		for (i = 0; i < KWORK_CLASS_MAX; i++) {
1632 			list_add_tail(&kwork_class_supported_list[i]->list,
1633 				      &kwork->class_list);
1634 		}
1635 	}
1636 
1637 	pr_debug("Config event list:");
1638 	list_for_each_entry(class, &kwork->class_list, list)
1639 		pr_debug(" %s", class->name);
1640 	pr_debug("\n");
1641 }
1642 
1643 static int perf_kwork__record(struct perf_kwork *kwork,
1644 			      int argc, const char **argv)
1645 {
1646 	const char **rec_argv;
1647 	unsigned int rec_argc, i, j;
1648 	struct kwork_class *class;
1649 
1650 	const char *const record_args[] = {
1651 		"record",
1652 		"-a",
1653 		"-R",
1654 		"-m", "1024",
1655 		"-c", "1",
1656 	};
1657 
1658 	rec_argc = ARRAY_SIZE(record_args) + argc - 1;
1659 
1660 	list_for_each_entry(class, &kwork->class_list, list)
1661 		rec_argc += 2 * class->nr_tracepoints;
1662 
1663 	rec_argv = calloc(rec_argc + 1, sizeof(char *));
1664 	if (rec_argv == NULL)
1665 		return -ENOMEM;
1666 
1667 	for (i = 0; i < ARRAY_SIZE(record_args); i++)
1668 		rec_argv[i] = strdup(record_args[i]);
1669 
1670 	list_for_each_entry(class, &kwork->class_list, list) {
1671 		for (j = 0; j < class->nr_tracepoints; j++) {
1672 			rec_argv[i++] = strdup("-e");
1673 			rec_argv[i++] = strdup(class->tp_handlers[j].name);
1674 		}
1675 	}
1676 
1677 	for (j = 1; j < (unsigned int)argc; j++, i++)
1678 		rec_argv[i] = argv[j];
1679 
1680 	BUG_ON(i != rec_argc);
1681 
1682 	pr_debug("record comm: ");
1683 	for (j = 0; j < rec_argc; j++)
1684 		pr_debug("%s ", rec_argv[j]);
1685 	pr_debug("\n");
1686 
1687 	return cmd_record(i, rec_argv);
1688 }
1689 
1690 int cmd_kwork(int argc, const char **argv)
1691 {
1692 	static struct perf_kwork kwork = {
1693 		.class_list          = LIST_HEAD_INIT(kwork.class_list),
1694 		.tool = {
1695 			.mmap    = perf_event__process_mmap,
1696 			.mmap2   = perf_event__process_mmap2,
1697 			.sample  = perf_kwork__process_tracepoint_sample,
1698 		},
1699 		.atom_page_list      = LIST_HEAD_INIT(kwork.atom_page_list),
1700 		.sort_list           = LIST_HEAD_INIT(kwork.sort_list),
1701 		.cmp_id              = LIST_HEAD_INIT(kwork.cmp_id),
1702 		.sorted_work_root    = RB_ROOT_CACHED,
1703 		.tp_handler          = NULL,
1704 		.profile_name        = NULL,
1705 		.cpu_list            = NULL,
1706 		.time_str            = NULL,
1707 		.force               = false,
1708 		.event_list_str      = NULL,
1709 		.summary             = false,
1710 		.sort_order          = NULL,
1711 		.show_callchain      = false,
1712 		.max_stack           = 5,
1713 		.timestart           = 0,
1714 		.timeend             = 0,
1715 		.nr_events           = 0,
1716 		.nr_lost_chunks      = 0,
1717 		.nr_lost_events      = 0,
1718 		.all_runtime         = 0,
1719 		.all_count           = 0,
1720 		.nr_skipped_events   = { 0 },
1721 	};
1722 	static const char default_report_sort_order[] = "runtime, max, count";
1723 	static const char default_latency_sort_order[] = "avg, max, count";
1724 	const struct option kwork_options[] = {
1725 	OPT_INCR('v', "verbose", &verbose,
1726 		 "be more verbose (show symbol address, etc)"),
1727 	OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
1728 		    "dump raw trace in ASCII"),
1729 	OPT_STRING('k', "kwork", &kwork.event_list_str, "kwork",
1730 		   "list of kwork to profile (irq, softirq, workqueue, etc)"),
1731 	OPT_BOOLEAN('f', "force", &kwork.force, "don't complain, do it"),
1732 	OPT_END()
1733 	};
1734 	const struct option report_options[] = {
1735 	OPT_STRING('s', "sort", &kwork.sort_order, "key[,key2...]",
1736 		   "sort by key(s): runtime, max, count"),
1737 	OPT_STRING('C', "cpu", &kwork.cpu_list, "cpu",
1738 		   "list of cpus to profile"),
1739 	OPT_STRING('n', "name", &kwork.profile_name, "name",
1740 		   "event name to profile"),
1741 	OPT_STRING(0, "time", &kwork.time_str, "str",
1742 		   "Time span for analysis (start,stop)"),
1743 	OPT_STRING('i', "input", &input_name, "file",
1744 		   "input file name"),
1745 	OPT_BOOLEAN('S', "with-summary", &kwork.summary,
1746 		    "Show summary with statistics"),
1747 #ifdef HAVE_BPF_SKEL
1748 	OPT_BOOLEAN('b', "use-bpf", &kwork.use_bpf,
1749 		    "Use BPF to measure kwork runtime"),
1750 #endif
1751 	OPT_PARENT(kwork_options)
1752 	};
1753 	const struct option latency_options[] = {
1754 	OPT_STRING('s', "sort", &kwork.sort_order, "key[,key2...]",
1755 		   "sort by key(s): avg, max, count"),
1756 	OPT_STRING('C', "cpu", &kwork.cpu_list, "cpu",
1757 		   "list of cpus to profile"),
1758 	OPT_STRING('n', "name", &kwork.profile_name, "name",
1759 		   "event name to profile"),
1760 	OPT_STRING(0, "time", &kwork.time_str, "str",
1761 		   "Time span for analysis (start,stop)"),
1762 	OPT_STRING('i', "input", &input_name, "file",
1763 		   "input file name"),
1764 #ifdef HAVE_BPF_SKEL
1765 	OPT_BOOLEAN('b', "use-bpf", &kwork.use_bpf,
1766 		    "Use BPF to measure kwork latency"),
1767 #endif
1768 	OPT_PARENT(kwork_options)
1769 	};
1770 	const struct option timehist_options[] = {
1771 	OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name,
1772 		   "file", "vmlinux pathname"),
1773 	OPT_STRING(0, "kallsyms", &symbol_conf.kallsyms_name,
1774 		   "file", "kallsyms pathname"),
1775 	OPT_BOOLEAN('g', "call-graph", &kwork.show_callchain,
1776 		    "Display call chains if present"),
1777 	OPT_UINTEGER(0, "max-stack", &kwork.max_stack,
1778 		   "Maximum number of functions to display backtrace."),
1779 	OPT_STRING(0, "symfs", &symbol_conf.symfs, "directory",
1780 		    "Look for files with symbols relative to this directory"),
1781 	OPT_STRING(0, "time", &kwork.time_str, "str",
1782 		   "Time span for analysis (start,stop)"),
1783 	OPT_STRING('C', "cpu", &kwork.cpu_list, "cpu",
1784 		   "list of cpus to profile"),
1785 	OPT_STRING('n', "name", &kwork.profile_name, "name",
1786 		   "event name to profile"),
1787 	OPT_STRING('i', "input", &input_name, "file",
1788 		   "input file name"),
1789 	OPT_PARENT(kwork_options)
1790 	};
1791 	const char *kwork_usage[] = {
1792 		NULL,
1793 		NULL
1794 	};
1795 	const char * const report_usage[] = {
1796 		"perf kwork report [<options>]",
1797 		NULL
1798 	};
1799 	const char * const latency_usage[] = {
1800 		"perf kwork latency [<options>]",
1801 		NULL
1802 	};
1803 	const char * const timehist_usage[] = {
1804 		"perf kwork timehist [<options>]",
1805 		NULL
1806 	};
1807 	const char *const kwork_subcommands[] = {
1808 		"record", "report", "latency", "timehist", NULL
1809 	};
1810 
1811 	argc = parse_options_subcommand(argc, argv, kwork_options,
1812 					kwork_subcommands, kwork_usage,
1813 					PARSE_OPT_STOP_AT_NON_OPTION);
1814 	if (!argc)
1815 		usage_with_options(kwork_usage, kwork_options);
1816 
1817 	setup_event_list(&kwork, kwork_options, kwork_usage);
1818 	sort_dimension__add(&kwork, "id", &kwork.cmp_id);
1819 
1820 	if (strlen(argv[0]) > 2 && strstarts("record", argv[0]))
1821 		return perf_kwork__record(&kwork, argc, argv);
1822 	else if (strlen(argv[0]) > 2 && strstarts("report", argv[0])) {
1823 		kwork.sort_order = default_report_sort_order;
1824 		if (argc > 1) {
1825 			argc = parse_options(argc, argv, report_options, report_usage, 0);
1826 			if (argc)
1827 				usage_with_options(report_usage, report_options);
1828 		}
1829 		kwork.report = KWORK_REPORT_RUNTIME;
1830 		setup_sorting(&kwork, report_options, report_usage);
1831 		return perf_kwork__report(&kwork);
1832 	} else if (strlen(argv[0]) > 2 && strstarts("latency", argv[0])) {
1833 		kwork.sort_order = default_latency_sort_order;
1834 		if (argc > 1) {
1835 			argc = parse_options(argc, argv, latency_options, latency_usage, 0);
1836 			if (argc)
1837 				usage_with_options(latency_usage, latency_options);
1838 		}
1839 		kwork.report = KWORK_REPORT_LATENCY;
1840 		setup_sorting(&kwork, latency_options, latency_usage);
1841 		return perf_kwork__report(&kwork);
1842 	} else if (strlen(argv[0]) > 2 && strstarts("timehist", argv[0])) {
1843 		if (argc > 1) {
1844 			argc = parse_options(argc, argv, timehist_options, timehist_usage, 0);
1845 			if (argc)
1846 				usage_with_options(timehist_usage, timehist_options);
1847 		}
1848 		kwork.report = KWORK_REPORT_TIMEHIST;
1849 		return perf_kwork__timehist(&kwork);
1850 	} else
1851 		usage_with_options(kwork_usage, kwork_options);
1852 
1853 	return 0;
1854 }
1855