xref: /openbmc/linux/tools/perf/builtin-sched.c (revision d623f60d)
1 // SPDX-License-Identifier: GPL-2.0
2 #include "builtin.h"
3 #include "perf.h"
4 
5 #include "util/util.h"
6 #include "util/evlist.h"
7 #include "util/cache.h"
8 #include "util/evsel.h"
9 #include "util/symbol.h"
10 #include "util/thread.h"
11 #include "util/header.h"
12 #include "util/session.h"
13 #include "util/tool.h"
14 #include "util/cloexec.h"
15 #include "util/thread_map.h"
16 #include "util/color.h"
17 #include "util/stat.h"
18 #include "util/callchain.h"
19 #include "util/time-utils.h"
20 
21 #include <subcmd/parse-options.h>
22 #include "util/trace-event.h"
23 
24 #include "util/debug.h"
25 
26 #include <linux/kernel.h>
27 #include <linux/log2.h>
28 #include <sys/prctl.h>
29 #include <sys/resource.h>
30 #include <inttypes.h>
31 
32 #include <errno.h>
33 #include <semaphore.h>
34 #include <pthread.h>
35 #include <math.h>
36 #include <api/fs/fs.h>
37 #include <linux/time64.h>
38 
39 #include "sane_ctype.h"
40 
41 #define PR_SET_NAME		15               /* Set process name */
42 #define MAX_CPUS		4096
43 #define COMM_LEN		20
44 #define SYM_LEN			129
45 #define MAX_PID			1024000
46 
47 struct sched_atom;
48 
49 struct task_desc {
50 	unsigned long		nr;
51 	unsigned long		pid;
52 	char			comm[COMM_LEN];
53 
54 	unsigned long		nr_events;
55 	unsigned long		curr_event;
56 	struct sched_atom	**atoms;
57 
58 	pthread_t		thread;
59 	sem_t			sleep_sem;
60 
61 	sem_t			ready_for_work;
62 	sem_t			work_done_sem;
63 
64 	u64			cpu_usage;
65 };
66 
67 enum sched_event_type {
68 	SCHED_EVENT_RUN,
69 	SCHED_EVENT_SLEEP,
70 	SCHED_EVENT_WAKEUP,
71 	SCHED_EVENT_MIGRATION,
72 };
73 
74 struct sched_atom {
75 	enum sched_event_type	type;
76 	int			specific_wait;
77 	u64			timestamp;
78 	u64			duration;
79 	unsigned long		nr;
80 	sem_t			*wait_sem;
81 	struct task_desc	*wakee;
82 };
83 
84 #define TASK_STATE_TO_CHAR_STR "RSDTtZXxKWP"
85 
86 /* task state bitmask, copied from include/linux/sched.h */
87 #define TASK_RUNNING		0
88 #define TASK_INTERRUPTIBLE	1
89 #define TASK_UNINTERRUPTIBLE	2
90 #define __TASK_STOPPED		4
91 #define __TASK_TRACED		8
92 /* in tsk->exit_state */
93 #define EXIT_DEAD		16
94 #define EXIT_ZOMBIE		32
95 #define EXIT_TRACE		(EXIT_ZOMBIE | EXIT_DEAD)
96 /* in tsk->state again */
97 #define TASK_DEAD		64
98 #define TASK_WAKEKILL		128
99 #define TASK_WAKING		256
100 #define TASK_PARKED		512
101 
102 enum thread_state {
103 	THREAD_SLEEPING = 0,
104 	THREAD_WAIT_CPU,
105 	THREAD_SCHED_IN,
106 	THREAD_IGNORE
107 };
108 
109 struct work_atom {
110 	struct list_head	list;
111 	enum thread_state	state;
112 	u64			sched_out_time;
113 	u64			wake_up_time;
114 	u64			sched_in_time;
115 	u64			runtime;
116 };
117 
118 struct work_atoms {
119 	struct list_head	work_list;
120 	struct thread		*thread;
121 	struct rb_node		node;
122 	u64			max_lat;
123 	u64			max_lat_at;
124 	u64			total_lat;
125 	u64			nb_atoms;
126 	u64			total_runtime;
127 	int			num_merged;
128 };
129 
130 typedef int (*sort_fn_t)(struct work_atoms *, struct work_atoms *);
131 
132 struct perf_sched;
133 
134 struct trace_sched_handler {
135 	int (*switch_event)(struct perf_sched *sched, struct perf_evsel *evsel,
136 			    struct perf_sample *sample, struct machine *machine);
137 
138 	int (*runtime_event)(struct perf_sched *sched, struct perf_evsel *evsel,
139 			     struct perf_sample *sample, struct machine *machine);
140 
141 	int (*wakeup_event)(struct perf_sched *sched, struct perf_evsel *evsel,
142 			    struct perf_sample *sample, struct machine *machine);
143 
144 	/* PERF_RECORD_FORK event, not sched_process_fork tracepoint */
145 	int (*fork_event)(struct perf_sched *sched, union perf_event *event,
146 			  struct machine *machine);
147 
148 	int (*migrate_task_event)(struct perf_sched *sched,
149 				  struct perf_evsel *evsel,
150 				  struct perf_sample *sample,
151 				  struct machine *machine);
152 };
153 
154 #define COLOR_PIDS PERF_COLOR_BLUE
155 #define COLOR_CPUS PERF_COLOR_BG_RED
156 
157 struct perf_sched_map {
158 	DECLARE_BITMAP(comp_cpus_mask, MAX_CPUS);
159 	int			*comp_cpus;
160 	bool			 comp;
161 	struct thread_map	*color_pids;
162 	const char		*color_pids_str;
163 	struct cpu_map		*color_cpus;
164 	const char		*color_cpus_str;
165 	struct cpu_map		*cpus;
166 	const char		*cpus_str;
167 };
168 
169 struct perf_sched {
170 	struct perf_tool tool;
171 	const char	 *sort_order;
172 	unsigned long	 nr_tasks;
173 	struct task_desc **pid_to_task;
174 	struct task_desc **tasks;
175 	const struct trace_sched_handler *tp_handler;
176 	pthread_mutex_t	 start_work_mutex;
177 	pthread_mutex_t	 work_done_wait_mutex;
178 	int		 profile_cpu;
179 /*
180  * Track the current task - that way we can know whether there's any
181  * weird events, such as a task being switched away that is not current.
182  */
183 	int		 max_cpu;
184 	u32		 curr_pid[MAX_CPUS];
185 	struct thread	 *curr_thread[MAX_CPUS];
186 	char		 next_shortname1;
187 	char		 next_shortname2;
188 	unsigned int	 replay_repeat;
189 	unsigned long	 nr_run_events;
190 	unsigned long	 nr_sleep_events;
191 	unsigned long	 nr_wakeup_events;
192 	unsigned long	 nr_sleep_corrections;
193 	unsigned long	 nr_run_events_optimized;
194 	unsigned long	 targetless_wakeups;
195 	unsigned long	 multitarget_wakeups;
196 	unsigned long	 nr_runs;
197 	unsigned long	 nr_timestamps;
198 	unsigned long	 nr_unordered_timestamps;
199 	unsigned long	 nr_context_switch_bugs;
200 	unsigned long	 nr_events;
201 	unsigned long	 nr_lost_chunks;
202 	unsigned long	 nr_lost_events;
203 	u64		 run_measurement_overhead;
204 	u64		 sleep_measurement_overhead;
205 	u64		 start_time;
206 	u64		 cpu_usage;
207 	u64		 runavg_cpu_usage;
208 	u64		 parent_cpu_usage;
209 	u64		 runavg_parent_cpu_usage;
210 	u64		 sum_runtime;
211 	u64		 sum_fluct;
212 	u64		 run_avg;
213 	u64		 all_runtime;
214 	u64		 all_count;
215 	u64		 cpu_last_switched[MAX_CPUS];
216 	struct rb_root	 atom_root, sorted_atom_root, merged_atom_root;
217 	struct list_head sort_list, cmp_pid;
218 	bool force;
219 	bool skip_merge;
220 	struct perf_sched_map map;
221 
222 	/* options for timehist command */
223 	bool		summary;
224 	bool		summary_only;
225 	bool		idle_hist;
226 	bool		show_callchain;
227 	unsigned int	max_stack;
228 	bool		show_cpu_visual;
229 	bool		show_wakeups;
230 	bool		show_next;
231 	bool		show_migrations;
232 	bool		show_state;
233 	u64		skipped_samples;
234 	const char	*time_str;
235 	struct perf_time_interval ptime;
236 	struct perf_time_interval hist_time;
237 };
238 
239 /* per thread run time data */
240 struct thread_runtime {
241 	u64 last_time;      /* time of previous sched in/out event */
242 	u64 dt_run;         /* run time */
243 	u64 dt_sleep;       /* time between CPU access by sleep (off cpu) */
244 	u64 dt_iowait;      /* time between CPU access by iowait (off cpu) */
245 	u64 dt_preempt;     /* time between CPU access by preempt (off cpu) */
246 	u64 dt_delay;       /* time between wakeup and sched-in */
247 	u64 ready_to_run;   /* time of wakeup */
248 
249 	struct stats run_stats;
250 	u64 total_run_time;
251 	u64 total_sleep_time;
252 	u64 total_iowait_time;
253 	u64 total_preempt_time;
254 	u64 total_delay_time;
255 
256 	int last_state;
257 
258 	char shortname[3];
259 	bool comm_changed;
260 
261 	u64 migrations;
262 };
263 
264 /* per event run time data */
265 struct evsel_runtime {
266 	u64 *last_time; /* time this event was last seen per cpu */
267 	u32 ncpu;       /* highest cpu slot allocated */
268 };
269 
270 /* per cpu idle time data */
271 struct idle_thread_runtime {
272 	struct thread_runtime	tr;
273 	struct thread		*last_thread;
274 	struct rb_root		sorted_root;
275 	struct callchain_root	callchain;
276 	struct callchain_cursor	cursor;
277 };
278 
279 /* track idle times per cpu */
280 static struct thread **idle_threads;
281 static int idle_max_cpu;
282 static char idle_comm[] = "<idle>";
283 
284 static u64 get_nsecs(void)
285 {
286 	struct timespec ts;
287 
288 	clock_gettime(CLOCK_MONOTONIC, &ts);
289 
290 	return ts.tv_sec * NSEC_PER_SEC + ts.tv_nsec;
291 }
292 
293 static void burn_nsecs(struct perf_sched *sched, u64 nsecs)
294 {
295 	u64 T0 = get_nsecs(), T1;
296 
297 	do {
298 		T1 = get_nsecs();
299 	} while (T1 + sched->run_measurement_overhead < T0 + nsecs);
300 }
301 
302 static void sleep_nsecs(u64 nsecs)
303 {
304 	struct timespec ts;
305 
306 	ts.tv_nsec = nsecs % 999999999;
307 	ts.tv_sec = nsecs / 999999999;
308 
309 	nanosleep(&ts, NULL);
310 }
311 
312 static void calibrate_run_measurement_overhead(struct perf_sched *sched)
313 {
314 	u64 T0, T1, delta, min_delta = NSEC_PER_SEC;
315 	int i;
316 
317 	for (i = 0; i < 10; i++) {
318 		T0 = get_nsecs();
319 		burn_nsecs(sched, 0);
320 		T1 = get_nsecs();
321 		delta = T1-T0;
322 		min_delta = min(min_delta, delta);
323 	}
324 	sched->run_measurement_overhead = min_delta;
325 
326 	printf("run measurement overhead: %" PRIu64 " nsecs\n", min_delta);
327 }
328 
329 static void calibrate_sleep_measurement_overhead(struct perf_sched *sched)
330 {
331 	u64 T0, T1, delta, min_delta = NSEC_PER_SEC;
332 	int i;
333 
334 	for (i = 0; i < 10; i++) {
335 		T0 = get_nsecs();
336 		sleep_nsecs(10000);
337 		T1 = get_nsecs();
338 		delta = T1-T0;
339 		min_delta = min(min_delta, delta);
340 	}
341 	min_delta -= 10000;
342 	sched->sleep_measurement_overhead = min_delta;
343 
344 	printf("sleep measurement overhead: %" PRIu64 " nsecs\n", min_delta);
345 }
346 
347 static struct sched_atom *
348 get_new_event(struct task_desc *task, u64 timestamp)
349 {
350 	struct sched_atom *event = zalloc(sizeof(*event));
351 	unsigned long idx = task->nr_events;
352 	size_t size;
353 
354 	event->timestamp = timestamp;
355 	event->nr = idx;
356 
357 	task->nr_events++;
358 	size = sizeof(struct sched_atom *) * task->nr_events;
359 	task->atoms = realloc(task->atoms, size);
360 	BUG_ON(!task->atoms);
361 
362 	task->atoms[idx] = event;
363 
364 	return event;
365 }
366 
367 static struct sched_atom *last_event(struct task_desc *task)
368 {
369 	if (!task->nr_events)
370 		return NULL;
371 
372 	return task->atoms[task->nr_events - 1];
373 }
374 
375 static void add_sched_event_run(struct perf_sched *sched, struct task_desc *task,
376 				u64 timestamp, u64 duration)
377 {
378 	struct sched_atom *event, *curr_event = last_event(task);
379 
380 	/*
381 	 * optimize an existing RUN event by merging this one
382 	 * to it:
383 	 */
384 	if (curr_event && curr_event->type == SCHED_EVENT_RUN) {
385 		sched->nr_run_events_optimized++;
386 		curr_event->duration += duration;
387 		return;
388 	}
389 
390 	event = get_new_event(task, timestamp);
391 
392 	event->type = SCHED_EVENT_RUN;
393 	event->duration = duration;
394 
395 	sched->nr_run_events++;
396 }
397 
398 static void add_sched_event_wakeup(struct perf_sched *sched, struct task_desc *task,
399 				   u64 timestamp, struct task_desc *wakee)
400 {
401 	struct sched_atom *event, *wakee_event;
402 
403 	event = get_new_event(task, timestamp);
404 	event->type = SCHED_EVENT_WAKEUP;
405 	event->wakee = wakee;
406 
407 	wakee_event = last_event(wakee);
408 	if (!wakee_event || wakee_event->type != SCHED_EVENT_SLEEP) {
409 		sched->targetless_wakeups++;
410 		return;
411 	}
412 	if (wakee_event->wait_sem) {
413 		sched->multitarget_wakeups++;
414 		return;
415 	}
416 
417 	wakee_event->wait_sem = zalloc(sizeof(*wakee_event->wait_sem));
418 	sem_init(wakee_event->wait_sem, 0, 0);
419 	wakee_event->specific_wait = 1;
420 	event->wait_sem = wakee_event->wait_sem;
421 
422 	sched->nr_wakeup_events++;
423 }
424 
425 static void add_sched_event_sleep(struct perf_sched *sched, struct task_desc *task,
426 				  u64 timestamp, u64 task_state __maybe_unused)
427 {
428 	struct sched_atom *event = get_new_event(task, timestamp);
429 
430 	event->type = SCHED_EVENT_SLEEP;
431 
432 	sched->nr_sleep_events++;
433 }
434 
435 static struct task_desc *register_pid(struct perf_sched *sched,
436 				      unsigned long pid, const char *comm)
437 {
438 	struct task_desc *task;
439 	static int pid_max;
440 
441 	if (sched->pid_to_task == NULL) {
442 		if (sysctl__read_int("kernel/pid_max", &pid_max) < 0)
443 			pid_max = MAX_PID;
444 		BUG_ON((sched->pid_to_task = calloc(pid_max, sizeof(struct task_desc *))) == NULL);
445 	}
446 	if (pid >= (unsigned long)pid_max) {
447 		BUG_ON((sched->pid_to_task = realloc(sched->pid_to_task, (pid + 1) *
448 			sizeof(struct task_desc *))) == NULL);
449 		while (pid >= (unsigned long)pid_max)
450 			sched->pid_to_task[pid_max++] = NULL;
451 	}
452 
453 	task = sched->pid_to_task[pid];
454 
455 	if (task)
456 		return task;
457 
458 	task = zalloc(sizeof(*task));
459 	task->pid = pid;
460 	task->nr = sched->nr_tasks;
461 	strcpy(task->comm, comm);
462 	/*
463 	 * every task starts in sleeping state - this gets ignored
464 	 * if there's no wakeup pointing to this sleep state:
465 	 */
466 	add_sched_event_sleep(sched, task, 0, 0);
467 
468 	sched->pid_to_task[pid] = task;
469 	sched->nr_tasks++;
470 	sched->tasks = realloc(sched->tasks, sched->nr_tasks * sizeof(struct task_desc *));
471 	BUG_ON(!sched->tasks);
472 	sched->tasks[task->nr] = task;
473 
474 	if (verbose > 0)
475 		printf("registered task #%ld, PID %ld (%s)\n", sched->nr_tasks, pid, comm);
476 
477 	return task;
478 }
479 
480 
481 static void print_task_traces(struct perf_sched *sched)
482 {
483 	struct task_desc *task;
484 	unsigned long i;
485 
486 	for (i = 0; i < sched->nr_tasks; i++) {
487 		task = sched->tasks[i];
488 		printf("task %6ld (%20s:%10ld), nr_events: %ld\n",
489 			task->nr, task->comm, task->pid, task->nr_events);
490 	}
491 }
492 
493 static void add_cross_task_wakeups(struct perf_sched *sched)
494 {
495 	struct task_desc *task1, *task2;
496 	unsigned long i, j;
497 
498 	for (i = 0; i < sched->nr_tasks; i++) {
499 		task1 = sched->tasks[i];
500 		j = i + 1;
501 		if (j == sched->nr_tasks)
502 			j = 0;
503 		task2 = sched->tasks[j];
504 		add_sched_event_wakeup(sched, task1, 0, task2);
505 	}
506 }
507 
508 static void perf_sched__process_event(struct perf_sched *sched,
509 				      struct sched_atom *atom)
510 {
511 	int ret = 0;
512 
513 	switch (atom->type) {
514 		case SCHED_EVENT_RUN:
515 			burn_nsecs(sched, atom->duration);
516 			break;
517 		case SCHED_EVENT_SLEEP:
518 			if (atom->wait_sem)
519 				ret = sem_wait(atom->wait_sem);
520 			BUG_ON(ret);
521 			break;
522 		case SCHED_EVENT_WAKEUP:
523 			if (atom->wait_sem)
524 				ret = sem_post(atom->wait_sem);
525 			BUG_ON(ret);
526 			break;
527 		case SCHED_EVENT_MIGRATION:
528 			break;
529 		default:
530 			BUG_ON(1);
531 	}
532 }
533 
534 static u64 get_cpu_usage_nsec_parent(void)
535 {
536 	struct rusage ru;
537 	u64 sum;
538 	int err;
539 
540 	err = getrusage(RUSAGE_SELF, &ru);
541 	BUG_ON(err);
542 
543 	sum =  ru.ru_utime.tv_sec * NSEC_PER_SEC + ru.ru_utime.tv_usec * NSEC_PER_USEC;
544 	sum += ru.ru_stime.tv_sec * NSEC_PER_SEC + ru.ru_stime.tv_usec * NSEC_PER_USEC;
545 
546 	return sum;
547 }
548 
549 static int self_open_counters(struct perf_sched *sched, unsigned long cur_task)
550 {
551 	struct perf_event_attr attr;
552 	char sbuf[STRERR_BUFSIZE], info[STRERR_BUFSIZE];
553 	int fd;
554 	struct rlimit limit;
555 	bool need_privilege = false;
556 
557 	memset(&attr, 0, sizeof(attr));
558 
559 	attr.type = PERF_TYPE_SOFTWARE;
560 	attr.config = PERF_COUNT_SW_TASK_CLOCK;
561 
562 force_again:
563 	fd = sys_perf_event_open(&attr, 0, -1, -1,
564 				 perf_event_open_cloexec_flag());
565 
566 	if (fd < 0) {
567 		if (errno == EMFILE) {
568 			if (sched->force) {
569 				BUG_ON(getrlimit(RLIMIT_NOFILE, &limit) == -1);
570 				limit.rlim_cur += sched->nr_tasks - cur_task;
571 				if (limit.rlim_cur > limit.rlim_max) {
572 					limit.rlim_max = limit.rlim_cur;
573 					need_privilege = true;
574 				}
575 				if (setrlimit(RLIMIT_NOFILE, &limit) == -1) {
576 					if (need_privilege && errno == EPERM)
577 						strcpy(info, "Need privilege\n");
578 				} else
579 					goto force_again;
580 			} else
581 				strcpy(info, "Have a try with -f option\n");
582 		}
583 		pr_err("Error: sys_perf_event_open() syscall returned "
584 		       "with %d (%s)\n%s", fd,
585 		       str_error_r(errno, sbuf, sizeof(sbuf)), info);
586 		exit(EXIT_FAILURE);
587 	}
588 	return fd;
589 }
590 
591 static u64 get_cpu_usage_nsec_self(int fd)
592 {
593 	u64 runtime;
594 	int ret;
595 
596 	ret = read(fd, &runtime, sizeof(runtime));
597 	BUG_ON(ret != sizeof(runtime));
598 
599 	return runtime;
600 }
601 
602 struct sched_thread_parms {
603 	struct task_desc  *task;
604 	struct perf_sched *sched;
605 	int fd;
606 };
607 
608 static void *thread_func(void *ctx)
609 {
610 	struct sched_thread_parms *parms = ctx;
611 	struct task_desc *this_task = parms->task;
612 	struct perf_sched *sched = parms->sched;
613 	u64 cpu_usage_0, cpu_usage_1;
614 	unsigned long i, ret;
615 	char comm2[22];
616 	int fd = parms->fd;
617 
618 	zfree(&parms);
619 
620 	sprintf(comm2, ":%s", this_task->comm);
621 	prctl(PR_SET_NAME, comm2);
622 	if (fd < 0)
623 		return NULL;
624 again:
625 	ret = sem_post(&this_task->ready_for_work);
626 	BUG_ON(ret);
627 	ret = pthread_mutex_lock(&sched->start_work_mutex);
628 	BUG_ON(ret);
629 	ret = pthread_mutex_unlock(&sched->start_work_mutex);
630 	BUG_ON(ret);
631 
632 	cpu_usage_0 = get_cpu_usage_nsec_self(fd);
633 
634 	for (i = 0; i < this_task->nr_events; i++) {
635 		this_task->curr_event = i;
636 		perf_sched__process_event(sched, this_task->atoms[i]);
637 	}
638 
639 	cpu_usage_1 = get_cpu_usage_nsec_self(fd);
640 	this_task->cpu_usage = cpu_usage_1 - cpu_usage_0;
641 	ret = sem_post(&this_task->work_done_sem);
642 	BUG_ON(ret);
643 
644 	ret = pthread_mutex_lock(&sched->work_done_wait_mutex);
645 	BUG_ON(ret);
646 	ret = pthread_mutex_unlock(&sched->work_done_wait_mutex);
647 	BUG_ON(ret);
648 
649 	goto again;
650 }
651 
652 static void create_tasks(struct perf_sched *sched)
653 {
654 	struct task_desc *task;
655 	pthread_attr_t attr;
656 	unsigned long i;
657 	int err;
658 
659 	err = pthread_attr_init(&attr);
660 	BUG_ON(err);
661 	err = pthread_attr_setstacksize(&attr,
662 			(size_t) max(16 * 1024, PTHREAD_STACK_MIN));
663 	BUG_ON(err);
664 	err = pthread_mutex_lock(&sched->start_work_mutex);
665 	BUG_ON(err);
666 	err = pthread_mutex_lock(&sched->work_done_wait_mutex);
667 	BUG_ON(err);
668 	for (i = 0; i < sched->nr_tasks; i++) {
669 		struct sched_thread_parms *parms = malloc(sizeof(*parms));
670 		BUG_ON(parms == NULL);
671 		parms->task = task = sched->tasks[i];
672 		parms->sched = sched;
673 		parms->fd = self_open_counters(sched, i);
674 		sem_init(&task->sleep_sem, 0, 0);
675 		sem_init(&task->ready_for_work, 0, 0);
676 		sem_init(&task->work_done_sem, 0, 0);
677 		task->curr_event = 0;
678 		err = pthread_create(&task->thread, &attr, thread_func, parms);
679 		BUG_ON(err);
680 	}
681 }
682 
683 static void wait_for_tasks(struct perf_sched *sched)
684 {
685 	u64 cpu_usage_0, cpu_usage_1;
686 	struct task_desc *task;
687 	unsigned long i, ret;
688 
689 	sched->start_time = get_nsecs();
690 	sched->cpu_usage = 0;
691 	pthread_mutex_unlock(&sched->work_done_wait_mutex);
692 
693 	for (i = 0; i < sched->nr_tasks; i++) {
694 		task = sched->tasks[i];
695 		ret = sem_wait(&task->ready_for_work);
696 		BUG_ON(ret);
697 		sem_init(&task->ready_for_work, 0, 0);
698 	}
699 	ret = pthread_mutex_lock(&sched->work_done_wait_mutex);
700 	BUG_ON(ret);
701 
702 	cpu_usage_0 = get_cpu_usage_nsec_parent();
703 
704 	pthread_mutex_unlock(&sched->start_work_mutex);
705 
706 	for (i = 0; i < sched->nr_tasks; i++) {
707 		task = sched->tasks[i];
708 		ret = sem_wait(&task->work_done_sem);
709 		BUG_ON(ret);
710 		sem_init(&task->work_done_sem, 0, 0);
711 		sched->cpu_usage += task->cpu_usage;
712 		task->cpu_usage = 0;
713 	}
714 
715 	cpu_usage_1 = get_cpu_usage_nsec_parent();
716 	if (!sched->runavg_cpu_usage)
717 		sched->runavg_cpu_usage = sched->cpu_usage;
718 	sched->runavg_cpu_usage = (sched->runavg_cpu_usage * (sched->replay_repeat - 1) + sched->cpu_usage) / sched->replay_repeat;
719 
720 	sched->parent_cpu_usage = cpu_usage_1 - cpu_usage_0;
721 	if (!sched->runavg_parent_cpu_usage)
722 		sched->runavg_parent_cpu_usage = sched->parent_cpu_usage;
723 	sched->runavg_parent_cpu_usage = (sched->runavg_parent_cpu_usage * (sched->replay_repeat - 1) +
724 					 sched->parent_cpu_usage)/sched->replay_repeat;
725 
726 	ret = pthread_mutex_lock(&sched->start_work_mutex);
727 	BUG_ON(ret);
728 
729 	for (i = 0; i < sched->nr_tasks; i++) {
730 		task = sched->tasks[i];
731 		sem_init(&task->sleep_sem, 0, 0);
732 		task->curr_event = 0;
733 	}
734 }
735 
736 static void run_one_test(struct perf_sched *sched)
737 {
738 	u64 T0, T1, delta, avg_delta, fluct;
739 
740 	T0 = get_nsecs();
741 	wait_for_tasks(sched);
742 	T1 = get_nsecs();
743 
744 	delta = T1 - T0;
745 	sched->sum_runtime += delta;
746 	sched->nr_runs++;
747 
748 	avg_delta = sched->sum_runtime / sched->nr_runs;
749 	if (delta < avg_delta)
750 		fluct = avg_delta - delta;
751 	else
752 		fluct = delta - avg_delta;
753 	sched->sum_fluct += fluct;
754 	if (!sched->run_avg)
755 		sched->run_avg = delta;
756 	sched->run_avg = (sched->run_avg * (sched->replay_repeat - 1) + delta) / sched->replay_repeat;
757 
758 	printf("#%-3ld: %0.3f, ", sched->nr_runs, (double)delta / NSEC_PER_MSEC);
759 
760 	printf("ravg: %0.2f, ", (double)sched->run_avg / NSEC_PER_MSEC);
761 
762 	printf("cpu: %0.2f / %0.2f",
763 		(double)sched->cpu_usage / NSEC_PER_MSEC, (double)sched->runavg_cpu_usage / NSEC_PER_MSEC);
764 
765 #if 0
766 	/*
767 	 * rusage statistics done by the parent, these are less
768 	 * accurate than the sched->sum_exec_runtime based statistics:
769 	 */
770 	printf(" [%0.2f / %0.2f]",
771 		(double)sched->parent_cpu_usage / NSEC_PER_MSEC,
772 		(double)sched->runavg_parent_cpu_usage / NSEC_PER_MSEC);
773 #endif
774 
775 	printf("\n");
776 
777 	if (sched->nr_sleep_corrections)
778 		printf(" (%ld sleep corrections)\n", sched->nr_sleep_corrections);
779 	sched->nr_sleep_corrections = 0;
780 }
781 
782 static void test_calibrations(struct perf_sched *sched)
783 {
784 	u64 T0, T1;
785 
786 	T0 = get_nsecs();
787 	burn_nsecs(sched, NSEC_PER_MSEC);
788 	T1 = get_nsecs();
789 
790 	printf("the run test took %" PRIu64 " nsecs\n", T1 - T0);
791 
792 	T0 = get_nsecs();
793 	sleep_nsecs(NSEC_PER_MSEC);
794 	T1 = get_nsecs();
795 
796 	printf("the sleep test took %" PRIu64 " nsecs\n", T1 - T0);
797 }
798 
799 static int
800 replay_wakeup_event(struct perf_sched *sched,
801 		    struct perf_evsel *evsel, struct perf_sample *sample,
802 		    struct machine *machine __maybe_unused)
803 {
804 	const char *comm = perf_evsel__strval(evsel, sample, "comm");
805 	const u32 pid	 = perf_evsel__intval(evsel, sample, "pid");
806 	struct task_desc *waker, *wakee;
807 
808 	if (verbose > 0) {
809 		printf("sched_wakeup event %p\n", evsel);
810 
811 		printf(" ... pid %d woke up %s/%d\n", sample->tid, comm, pid);
812 	}
813 
814 	waker = register_pid(sched, sample->tid, "<unknown>");
815 	wakee = register_pid(sched, pid, comm);
816 
817 	add_sched_event_wakeup(sched, waker, sample->time, wakee);
818 	return 0;
819 }
820 
821 static int replay_switch_event(struct perf_sched *sched,
822 			       struct perf_evsel *evsel,
823 			       struct perf_sample *sample,
824 			       struct machine *machine __maybe_unused)
825 {
826 	const char *prev_comm  = perf_evsel__strval(evsel, sample, "prev_comm"),
827 		   *next_comm  = perf_evsel__strval(evsel, sample, "next_comm");
828 	const u32 prev_pid = perf_evsel__intval(evsel, sample, "prev_pid"),
829 		  next_pid = perf_evsel__intval(evsel, sample, "next_pid");
830 	const u64 prev_state = perf_evsel__intval(evsel, sample, "prev_state");
831 	struct task_desc *prev, __maybe_unused *next;
832 	u64 timestamp0, timestamp = sample->time;
833 	int cpu = sample->cpu;
834 	s64 delta;
835 
836 	if (verbose > 0)
837 		printf("sched_switch event %p\n", evsel);
838 
839 	if (cpu >= MAX_CPUS || cpu < 0)
840 		return 0;
841 
842 	timestamp0 = sched->cpu_last_switched[cpu];
843 	if (timestamp0)
844 		delta = timestamp - timestamp0;
845 	else
846 		delta = 0;
847 
848 	if (delta < 0) {
849 		pr_err("hm, delta: %" PRIu64 " < 0 ?\n", delta);
850 		return -1;
851 	}
852 
853 	pr_debug(" ... switch from %s/%d to %s/%d [ran %" PRIu64 " nsecs]\n",
854 		 prev_comm, prev_pid, next_comm, next_pid, delta);
855 
856 	prev = register_pid(sched, prev_pid, prev_comm);
857 	next = register_pid(sched, next_pid, next_comm);
858 
859 	sched->cpu_last_switched[cpu] = timestamp;
860 
861 	add_sched_event_run(sched, prev, timestamp, delta);
862 	add_sched_event_sleep(sched, prev, timestamp, prev_state);
863 
864 	return 0;
865 }
866 
867 static int replay_fork_event(struct perf_sched *sched,
868 			     union perf_event *event,
869 			     struct machine *machine)
870 {
871 	struct thread *child, *parent;
872 
873 	child = machine__findnew_thread(machine, event->fork.pid,
874 					event->fork.tid);
875 	parent = machine__findnew_thread(machine, event->fork.ppid,
876 					 event->fork.ptid);
877 
878 	if (child == NULL || parent == NULL) {
879 		pr_debug("thread does not exist on fork event: child %p, parent %p\n",
880 				 child, parent);
881 		goto out_put;
882 	}
883 
884 	if (verbose > 0) {
885 		printf("fork event\n");
886 		printf("... parent: %s/%d\n", thread__comm_str(parent), parent->tid);
887 		printf("...  child: %s/%d\n", thread__comm_str(child), child->tid);
888 	}
889 
890 	register_pid(sched, parent->tid, thread__comm_str(parent));
891 	register_pid(sched, child->tid, thread__comm_str(child));
892 out_put:
893 	thread__put(child);
894 	thread__put(parent);
895 	return 0;
896 }
897 
898 struct sort_dimension {
899 	const char		*name;
900 	sort_fn_t		cmp;
901 	struct list_head	list;
902 };
903 
904 /*
905  * handle runtime stats saved per thread
906  */
907 static struct thread_runtime *thread__init_runtime(struct thread *thread)
908 {
909 	struct thread_runtime *r;
910 
911 	r = zalloc(sizeof(struct thread_runtime));
912 	if (!r)
913 		return NULL;
914 
915 	init_stats(&r->run_stats);
916 	thread__set_priv(thread, r);
917 
918 	return r;
919 }
920 
921 static struct thread_runtime *thread__get_runtime(struct thread *thread)
922 {
923 	struct thread_runtime *tr;
924 
925 	tr = thread__priv(thread);
926 	if (tr == NULL) {
927 		tr = thread__init_runtime(thread);
928 		if (tr == NULL)
929 			pr_debug("Failed to malloc memory for runtime data.\n");
930 	}
931 
932 	return tr;
933 }
934 
935 static int
936 thread_lat_cmp(struct list_head *list, struct work_atoms *l, struct work_atoms *r)
937 {
938 	struct sort_dimension *sort;
939 	int ret = 0;
940 
941 	BUG_ON(list_empty(list));
942 
943 	list_for_each_entry(sort, list, list) {
944 		ret = sort->cmp(l, r);
945 		if (ret)
946 			return ret;
947 	}
948 
949 	return ret;
950 }
951 
952 static struct work_atoms *
953 thread_atoms_search(struct rb_root *root, struct thread *thread,
954 			 struct list_head *sort_list)
955 {
956 	struct rb_node *node = root->rb_node;
957 	struct work_atoms key = { .thread = thread };
958 
959 	while (node) {
960 		struct work_atoms *atoms;
961 		int cmp;
962 
963 		atoms = container_of(node, struct work_atoms, node);
964 
965 		cmp = thread_lat_cmp(sort_list, &key, atoms);
966 		if (cmp > 0)
967 			node = node->rb_left;
968 		else if (cmp < 0)
969 			node = node->rb_right;
970 		else {
971 			BUG_ON(thread != atoms->thread);
972 			return atoms;
973 		}
974 	}
975 	return NULL;
976 }
977 
978 static void
979 __thread_latency_insert(struct rb_root *root, struct work_atoms *data,
980 			 struct list_head *sort_list)
981 {
982 	struct rb_node **new = &(root->rb_node), *parent = NULL;
983 
984 	while (*new) {
985 		struct work_atoms *this;
986 		int cmp;
987 
988 		this = container_of(*new, struct work_atoms, node);
989 		parent = *new;
990 
991 		cmp = thread_lat_cmp(sort_list, data, this);
992 
993 		if (cmp > 0)
994 			new = &((*new)->rb_left);
995 		else
996 			new = &((*new)->rb_right);
997 	}
998 
999 	rb_link_node(&data->node, parent, new);
1000 	rb_insert_color(&data->node, root);
1001 }
1002 
1003 static int thread_atoms_insert(struct perf_sched *sched, struct thread *thread)
1004 {
1005 	struct work_atoms *atoms = zalloc(sizeof(*atoms));
1006 	if (!atoms) {
1007 		pr_err("No memory at %s\n", __func__);
1008 		return -1;
1009 	}
1010 
1011 	atoms->thread = thread__get(thread);
1012 	INIT_LIST_HEAD(&atoms->work_list);
1013 	__thread_latency_insert(&sched->atom_root, atoms, &sched->cmp_pid);
1014 	return 0;
1015 }
1016 
1017 static char sched_out_state(u64 prev_state)
1018 {
1019 	const char *str = TASK_STATE_TO_CHAR_STR;
1020 
1021 	return str[prev_state];
1022 }
1023 
1024 static int
1025 add_sched_out_event(struct work_atoms *atoms,
1026 		    char run_state,
1027 		    u64 timestamp)
1028 {
1029 	struct work_atom *atom = zalloc(sizeof(*atom));
1030 	if (!atom) {
1031 		pr_err("Non memory at %s", __func__);
1032 		return -1;
1033 	}
1034 
1035 	atom->sched_out_time = timestamp;
1036 
1037 	if (run_state == 'R') {
1038 		atom->state = THREAD_WAIT_CPU;
1039 		atom->wake_up_time = atom->sched_out_time;
1040 	}
1041 
1042 	list_add_tail(&atom->list, &atoms->work_list);
1043 	return 0;
1044 }
1045 
1046 static void
1047 add_runtime_event(struct work_atoms *atoms, u64 delta,
1048 		  u64 timestamp __maybe_unused)
1049 {
1050 	struct work_atom *atom;
1051 
1052 	BUG_ON(list_empty(&atoms->work_list));
1053 
1054 	atom = list_entry(atoms->work_list.prev, struct work_atom, list);
1055 
1056 	atom->runtime += delta;
1057 	atoms->total_runtime += delta;
1058 }
1059 
1060 static void
1061 add_sched_in_event(struct work_atoms *atoms, u64 timestamp)
1062 {
1063 	struct work_atom *atom;
1064 	u64 delta;
1065 
1066 	if (list_empty(&atoms->work_list))
1067 		return;
1068 
1069 	atom = list_entry(atoms->work_list.prev, struct work_atom, list);
1070 
1071 	if (atom->state != THREAD_WAIT_CPU)
1072 		return;
1073 
1074 	if (timestamp < atom->wake_up_time) {
1075 		atom->state = THREAD_IGNORE;
1076 		return;
1077 	}
1078 
1079 	atom->state = THREAD_SCHED_IN;
1080 	atom->sched_in_time = timestamp;
1081 
1082 	delta = atom->sched_in_time - atom->wake_up_time;
1083 	atoms->total_lat += delta;
1084 	if (delta > atoms->max_lat) {
1085 		atoms->max_lat = delta;
1086 		atoms->max_lat_at = timestamp;
1087 	}
1088 	atoms->nb_atoms++;
1089 }
1090 
1091 static int latency_switch_event(struct perf_sched *sched,
1092 				struct perf_evsel *evsel,
1093 				struct perf_sample *sample,
1094 				struct machine *machine)
1095 {
1096 	const u32 prev_pid = perf_evsel__intval(evsel, sample, "prev_pid"),
1097 		  next_pid = perf_evsel__intval(evsel, sample, "next_pid");
1098 	const u64 prev_state = perf_evsel__intval(evsel, sample, "prev_state");
1099 	struct work_atoms *out_events, *in_events;
1100 	struct thread *sched_out, *sched_in;
1101 	u64 timestamp0, timestamp = sample->time;
1102 	int cpu = sample->cpu, err = -1;
1103 	s64 delta;
1104 
1105 	BUG_ON(cpu >= MAX_CPUS || cpu < 0);
1106 
1107 	timestamp0 = sched->cpu_last_switched[cpu];
1108 	sched->cpu_last_switched[cpu] = timestamp;
1109 	if (timestamp0)
1110 		delta = timestamp - timestamp0;
1111 	else
1112 		delta = 0;
1113 
1114 	if (delta < 0) {
1115 		pr_err("hm, delta: %" PRIu64 " < 0 ?\n", delta);
1116 		return -1;
1117 	}
1118 
1119 	sched_out = machine__findnew_thread(machine, -1, prev_pid);
1120 	sched_in = machine__findnew_thread(machine, -1, next_pid);
1121 	if (sched_out == NULL || sched_in == NULL)
1122 		goto out_put;
1123 
1124 	out_events = thread_atoms_search(&sched->atom_root, sched_out, &sched->cmp_pid);
1125 	if (!out_events) {
1126 		if (thread_atoms_insert(sched, sched_out))
1127 			goto out_put;
1128 		out_events = thread_atoms_search(&sched->atom_root, sched_out, &sched->cmp_pid);
1129 		if (!out_events) {
1130 			pr_err("out-event: Internal tree error");
1131 			goto out_put;
1132 		}
1133 	}
1134 	if (add_sched_out_event(out_events, sched_out_state(prev_state), timestamp))
1135 		return -1;
1136 
1137 	in_events = thread_atoms_search(&sched->atom_root, sched_in, &sched->cmp_pid);
1138 	if (!in_events) {
1139 		if (thread_atoms_insert(sched, sched_in))
1140 			goto out_put;
1141 		in_events = thread_atoms_search(&sched->atom_root, sched_in, &sched->cmp_pid);
1142 		if (!in_events) {
1143 			pr_err("in-event: Internal tree error");
1144 			goto out_put;
1145 		}
1146 		/*
1147 		 * Take came in we have not heard about yet,
1148 		 * add in an initial atom in runnable state:
1149 		 */
1150 		if (add_sched_out_event(in_events, 'R', timestamp))
1151 			goto out_put;
1152 	}
1153 	add_sched_in_event(in_events, timestamp);
1154 	err = 0;
1155 out_put:
1156 	thread__put(sched_out);
1157 	thread__put(sched_in);
1158 	return err;
1159 }
1160 
1161 static int latency_runtime_event(struct perf_sched *sched,
1162 				 struct perf_evsel *evsel,
1163 				 struct perf_sample *sample,
1164 				 struct machine *machine)
1165 {
1166 	const u32 pid	   = perf_evsel__intval(evsel, sample, "pid");
1167 	const u64 runtime  = perf_evsel__intval(evsel, sample, "runtime");
1168 	struct thread *thread = machine__findnew_thread(machine, -1, pid);
1169 	struct work_atoms *atoms = thread_atoms_search(&sched->atom_root, thread, &sched->cmp_pid);
1170 	u64 timestamp = sample->time;
1171 	int cpu = sample->cpu, err = -1;
1172 
1173 	if (thread == NULL)
1174 		return -1;
1175 
1176 	BUG_ON(cpu >= MAX_CPUS || cpu < 0);
1177 	if (!atoms) {
1178 		if (thread_atoms_insert(sched, thread))
1179 			goto out_put;
1180 		atoms = thread_atoms_search(&sched->atom_root, thread, &sched->cmp_pid);
1181 		if (!atoms) {
1182 			pr_err("in-event: Internal tree error");
1183 			goto out_put;
1184 		}
1185 		if (add_sched_out_event(atoms, 'R', timestamp))
1186 			goto out_put;
1187 	}
1188 
1189 	add_runtime_event(atoms, runtime, timestamp);
1190 	err = 0;
1191 out_put:
1192 	thread__put(thread);
1193 	return err;
1194 }
1195 
1196 static int latency_wakeup_event(struct perf_sched *sched,
1197 				struct perf_evsel *evsel,
1198 				struct perf_sample *sample,
1199 				struct machine *machine)
1200 {
1201 	const u32 pid	  = perf_evsel__intval(evsel, sample, "pid");
1202 	struct work_atoms *atoms;
1203 	struct work_atom *atom;
1204 	struct thread *wakee;
1205 	u64 timestamp = sample->time;
1206 	int err = -1;
1207 
1208 	wakee = machine__findnew_thread(machine, -1, pid);
1209 	if (wakee == NULL)
1210 		return -1;
1211 	atoms = thread_atoms_search(&sched->atom_root, wakee, &sched->cmp_pid);
1212 	if (!atoms) {
1213 		if (thread_atoms_insert(sched, wakee))
1214 			goto out_put;
1215 		atoms = thread_atoms_search(&sched->atom_root, wakee, &sched->cmp_pid);
1216 		if (!atoms) {
1217 			pr_err("wakeup-event: Internal tree error");
1218 			goto out_put;
1219 		}
1220 		if (add_sched_out_event(atoms, 'S', timestamp))
1221 			goto out_put;
1222 	}
1223 
1224 	BUG_ON(list_empty(&atoms->work_list));
1225 
1226 	atom = list_entry(atoms->work_list.prev, struct work_atom, list);
1227 
1228 	/*
1229 	 * As we do not guarantee the wakeup event happens when
1230 	 * task is out of run queue, also may happen when task is
1231 	 * on run queue and wakeup only change ->state to TASK_RUNNING,
1232 	 * then we should not set the ->wake_up_time when wake up a
1233 	 * task which is on run queue.
1234 	 *
1235 	 * You WILL be missing events if you've recorded only
1236 	 * one CPU, or are only looking at only one, so don't
1237 	 * skip in this case.
1238 	 */
1239 	if (sched->profile_cpu == -1 && atom->state != THREAD_SLEEPING)
1240 		goto out_ok;
1241 
1242 	sched->nr_timestamps++;
1243 	if (atom->sched_out_time > timestamp) {
1244 		sched->nr_unordered_timestamps++;
1245 		goto out_ok;
1246 	}
1247 
1248 	atom->state = THREAD_WAIT_CPU;
1249 	atom->wake_up_time = timestamp;
1250 out_ok:
1251 	err = 0;
1252 out_put:
1253 	thread__put(wakee);
1254 	return err;
1255 }
1256 
1257 static int latency_migrate_task_event(struct perf_sched *sched,
1258 				      struct perf_evsel *evsel,
1259 				      struct perf_sample *sample,
1260 				      struct machine *machine)
1261 {
1262 	const u32 pid = perf_evsel__intval(evsel, sample, "pid");
1263 	u64 timestamp = sample->time;
1264 	struct work_atoms *atoms;
1265 	struct work_atom *atom;
1266 	struct thread *migrant;
1267 	int err = -1;
1268 
1269 	/*
1270 	 * Only need to worry about migration when profiling one CPU.
1271 	 */
1272 	if (sched->profile_cpu == -1)
1273 		return 0;
1274 
1275 	migrant = machine__findnew_thread(machine, -1, pid);
1276 	if (migrant == NULL)
1277 		return -1;
1278 	atoms = thread_atoms_search(&sched->atom_root, migrant, &sched->cmp_pid);
1279 	if (!atoms) {
1280 		if (thread_atoms_insert(sched, migrant))
1281 			goto out_put;
1282 		register_pid(sched, migrant->tid, thread__comm_str(migrant));
1283 		atoms = thread_atoms_search(&sched->atom_root, migrant, &sched->cmp_pid);
1284 		if (!atoms) {
1285 			pr_err("migration-event: Internal tree error");
1286 			goto out_put;
1287 		}
1288 		if (add_sched_out_event(atoms, 'R', timestamp))
1289 			goto out_put;
1290 	}
1291 
1292 	BUG_ON(list_empty(&atoms->work_list));
1293 
1294 	atom = list_entry(atoms->work_list.prev, struct work_atom, list);
1295 	atom->sched_in_time = atom->sched_out_time = atom->wake_up_time = timestamp;
1296 
1297 	sched->nr_timestamps++;
1298 
1299 	if (atom->sched_out_time > timestamp)
1300 		sched->nr_unordered_timestamps++;
1301 	err = 0;
1302 out_put:
1303 	thread__put(migrant);
1304 	return err;
1305 }
1306 
1307 static void output_lat_thread(struct perf_sched *sched, struct work_atoms *work_list)
1308 {
1309 	int i;
1310 	int ret;
1311 	u64 avg;
1312 	char max_lat_at[32];
1313 
1314 	if (!work_list->nb_atoms)
1315 		return;
1316 	/*
1317 	 * Ignore idle threads:
1318 	 */
1319 	if (!strcmp(thread__comm_str(work_list->thread), "swapper"))
1320 		return;
1321 
1322 	sched->all_runtime += work_list->total_runtime;
1323 	sched->all_count   += work_list->nb_atoms;
1324 
1325 	if (work_list->num_merged > 1)
1326 		ret = printf("  %s:(%d) ", thread__comm_str(work_list->thread), work_list->num_merged);
1327 	else
1328 		ret = printf("  %s:%d ", thread__comm_str(work_list->thread), work_list->thread->tid);
1329 
1330 	for (i = 0; i < 24 - ret; i++)
1331 		printf(" ");
1332 
1333 	avg = work_list->total_lat / work_list->nb_atoms;
1334 	timestamp__scnprintf_usec(work_list->max_lat_at, max_lat_at, sizeof(max_lat_at));
1335 
1336 	printf("|%11.3f ms |%9" PRIu64 " | avg:%9.3f ms | max:%9.3f ms | max at: %13s s\n",
1337 	      (double)work_list->total_runtime / NSEC_PER_MSEC,
1338 		 work_list->nb_atoms, (double)avg / NSEC_PER_MSEC,
1339 		 (double)work_list->max_lat / NSEC_PER_MSEC,
1340 		 max_lat_at);
1341 }
1342 
1343 static int pid_cmp(struct work_atoms *l, struct work_atoms *r)
1344 {
1345 	if (l->thread == r->thread)
1346 		return 0;
1347 	if (l->thread->tid < r->thread->tid)
1348 		return -1;
1349 	if (l->thread->tid > r->thread->tid)
1350 		return 1;
1351 	return (int)(l->thread - r->thread);
1352 }
1353 
1354 static int avg_cmp(struct work_atoms *l, struct work_atoms *r)
1355 {
1356 	u64 avgl, avgr;
1357 
1358 	if (!l->nb_atoms)
1359 		return -1;
1360 
1361 	if (!r->nb_atoms)
1362 		return 1;
1363 
1364 	avgl = l->total_lat / l->nb_atoms;
1365 	avgr = r->total_lat / r->nb_atoms;
1366 
1367 	if (avgl < avgr)
1368 		return -1;
1369 	if (avgl > avgr)
1370 		return 1;
1371 
1372 	return 0;
1373 }
1374 
1375 static int max_cmp(struct work_atoms *l, struct work_atoms *r)
1376 {
1377 	if (l->max_lat < r->max_lat)
1378 		return -1;
1379 	if (l->max_lat > r->max_lat)
1380 		return 1;
1381 
1382 	return 0;
1383 }
1384 
1385 static int switch_cmp(struct work_atoms *l, struct work_atoms *r)
1386 {
1387 	if (l->nb_atoms < r->nb_atoms)
1388 		return -1;
1389 	if (l->nb_atoms > r->nb_atoms)
1390 		return 1;
1391 
1392 	return 0;
1393 }
1394 
1395 static int runtime_cmp(struct work_atoms *l, struct work_atoms *r)
1396 {
1397 	if (l->total_runtime < r->total_runtime)
1398 		return -1;
1399 	if (l->total_runtime > r->total_runtime)
1400 		return 1;
1401 
1402 	return 0;
1403 }
1404 
1405 static int sort_dimension__add(const char *tok, struct list_head *list)
1406 {
1407 	size_t i;
1408 	static struct sort_dimension avg_sort_dimension = {
1409 		.name = "avg",
1410 		.cmp  = avg_cmp,
1411 	};
1412 	static struct sort_dimension max_sort_dimension = {
1413 		.name = "max",
1414 		.cmp  = max_cmp,
1415 	};
1416 	static struct sort_dimension pid_sort_dimension = {
1417 		.name = "pid",
1418 		.cmp  = pid_cmp,
1419 	};
1420 	static struct sort_dimension runtime_sort_dimension = {
1421 		.name = "runtime",
1422 		.cmp  = runtime_cmp,
1423 	};
1424 	static struct sort_dimension switch_sort_dimension = {
1425 		.name = "switch",
1426 		.cmp  = switch_cmp,
1427 	};
1428 	struct sort_dimension *available_sorts[] = {
1429 		&pid_sort_dimension,
1430 		&avg_sort_dimension,
1431 		&max_sort_dimension,
1432 		&switch_sort_dimension,
1433 		&runtime_sort_dimension,
1434 	};
1435 
1436 	for (i = 0; i < ARRAY_SIZE(available_sorts); i++) {
1437 		if (!strcmp(available_sorts[i]->name, tok)) {
1438 			list_add_tail(&available_sorts[i]->list, list);
1439 
1440 			return 0;
1441 		}
1442 	}
1443 
1444 	return -1;
1445 }
1446 
1447 static void perf_sched__sort_lat(struct perf_sched *sched)
1448 {
1449 	struct rb_node *node;
1450 	struct rb_root *root = &sched->atom_root;
1451 again:
1452 	for (;;) {
1453 		struct work_atoms *data;
1454 		node = rb_first(root);
1455 		if (!node)
1456 			break;
1457 
1458 		rb_erase(node, root);
1459 		data = rb_entry(node, struct work_atoms, node);
1460 		__thread_latency_insert(&sched->sorted_atom_root, data, &sched->sort_list);
1461 	}
1462 	if (root == &sched->atom_root) {
1463 		root = &sched->merged_atom_root;
1464 		goto again;
1465 	}
1466 }
1467 
1468 static int process_sched_wakeup_event(struct perf_tool *tool,
1469 				      struct perf_evsel *evsel,
1470 				      struct perf_sample *sample,
1471 				      struct machine *machine)
1472 {
1473 	struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
1474 
1475 	if (sched->tp_handler->wakeup_event)
1476 		return sched->tp_handler->wakeup_event(sched, evsel, sample, machine);
1477 
1478 	return 0;
1479 }
1480 
1481 union map_priv {
1482 	void	*ptr;
1483 	bool	 color;
1484 };
1485 
1486 static bool thread__has_color(struct thread *thread)
1487 {
1488 	union map_priv priv = {
1489 		.ptr = thread__priv(thread),
1490 	};
1491 
1492 	return priv.color;
1493 }
1494 
1495 static struct thread*
1496 map__findnew_thread(struct perf_sched *sched, struct machine *machine, pid_t pid, pid_t tid)
1497 {
1498 	struct thread *thread = machine__findnew_thread(machine, pid, tid);
1499 	union map_priv priv = {
1500 		.color = false,
1501 	};
1502 
1503 	if (!sched->map.color_pids || !thread || thread__priv(thread))
1504 		return thread;
1505 
1506 	if (thread_map__has(sched->map.color_pids, tid))
1507 		priv.color = true;
1508 
1509 	thread__set_priv(thread, priv.ptr);
1510 	return thread;
1511 }
1512 
1513 static int map_switch_event(struct perf_sched *sched, struct perf_evsel *evsel,
1514 			    struct perf_sample *sample, struct machine *machine)
1515 {
1516 	const u32 next_pid = perf_evsel__intval(evsel, sample, "next_pid");
1517 	struct thread *sched_in;
1518 	struct thread_runtime *tr;
1519 	int new_shortname;
1520 	u64 timestamp0, timestamp = sample->time;
1521 	s64 delta;
1522 	int i, this_cpu = sample->cpu;
1523 	int cpus_nr;
1524 	bool new_cpu = false;
1525 	const char *color = PERF_COLOR_NORMAL;
1526 	char stimestamp[32];
1527 
1528 	BUG_ON(this_cpu >= MAX_CPUS || this_cpu < 0);
1529 
1530 	if (this_cpu > sched->max_cpu)
1531 		sched->max_cpu = this_cpu;
1532 
1533 	if (sched->map.comp) {
1534 		cpus_nr = bitmap_weight(sched->map.comp_cpus_mask, MAX_CPUS);
1535 		if (!test_and_set_bit(this_cpu, sched->map.comp_cpus_mask)) {
1536 			sched->map.comp_cpus[cpus_nr++] = this_cpu;
1537 			new_cpu = true;
1538 		}
1539 	} else
1540 		cpus_nr = sched->max_cpu;
1541 
1542 	timestamp0 = sched->cpu_last_switched[this_cpu];
1543 	sched->cpu_last_switched[this_cpu] = timestamp;
1544 	if (timestamp0)
1545 		delta = timestamp - timestamp0;
1546 	else
1547 		delta = 0;
1548 
1549 	if (delta < 0) {
1550 		pr_err("hm, delta: %" PRIu64 " < 0 ?\n", delta);
1551 		return -1;
1552 	}
1553 
1554 	sched_in = map__findnew_thread(sched, machine, -1, next_pid);
1555 	if (sched_in == NULL)
1556 		return -1;
1557 
1558 	tr = thread__get_runtime(sched_in);
1559 	if (tr == NULL) {
1560 		thread__put(sched_in);
1561 		return -1;
1562 	}
1563 
1564 	sched->curr_thread[this_cpu] = thread__get(sched_in);
1565 
1566 	printf("  ");
1567 
1568 	new_shortname = 0;
1569 	if (!tr->shortname[0]) {
1570 		if (!strcmp(thread__comm_str(sched_in), "swapper")) {
1571 			/*
1572 			 * Don't allocate a letter-number for swapper:0
1573 			 * as a shortname. Instead, we use '.' for it.
1574 			 */
1575 			tr->shortname[0] = '.';
1576 			tr->shortname[1] = ' ';
1577 		} else {
1578 			tr->shortname[0] = sched->next_shortname1;
1579 			tr->shortname[1] = sched->next_shortname2;
1580 
1581 			if (sched->next_shortname1 < 'Z') {
1582 				sched->next_shortname1++;
1583 			} else {
1584 				sched->next_shortname1 = 'A';
1585 				if (sched->next_shortname2 < '9')
1586 					sched->next_shortname2++;
1587 				else
1588 					sched->next_shortname2 = '0';
1589 			}
1590 		}
1591 		new_shortname = 1;
1592 	}
1593 
1594 	for (i = 0; i < cpus_nr; i++) {
1595 		int cpu = sched->map.comp ? sched->map.comp_cpus[i] : i;
1596 		struct thread *curr_thread = sched->curr_thread[cpu];
1597 		struct thread_runtime *curr_tr;
1598 		const char *pid_color = color;
1599 		const char *cpu_color = color;
1600 
1601 		if (curr_thread && thread__has_color(curr_thread))
1602 			pid_color = COLOR_PIDS;
1603 
1604 		if (sched->map.cpus && !cpu_map__has(sched->map.cpus, cpu))
1605 			continue;
1606 
1607 		if (sched->map.color_cpus && cpu_map__has(sched->map.color_cpus, cpu))
1608 			cpu_color = COLOR_CPUS;
1609 
1610 		if (cpu != this_cpu)
1611 			color_fprintf(stdout, color, " ");
1612 		else
1613 			color_fprintf(stdout, cpu_color, "*");
1614 
1615 		if (sched->curr_thread[cpu]) {
1616 			curr_tr = thread__get_runtime(sched->curr_thread[cpu]);
1617 			if (curr_tr == NULL) {
1618 				thread__put(sched_in);
1619 				return -1;
1620 			}
1621 			color_fprintf(stdout, pid_color, "%2s ", curr_tr->shortname);
1622 		} else
1623 			color_fprintf(stdout, color, "   ");
1624 	}
1625 
1626 	if (sched->map.cpus && !cpu_map__has(sched->map.cpus, this_cpu))
1627 		goto out;
1628 
1629 	timestamp__scnprintf_usec(timestamp, stimestamp, sizeof(stimestamp));
1630 	color_fprintf(stdout, color, "  %12s secs ", stimestamp);
1631 	if (new_shortname || tr->comm_changed || (verbose > 0 && sched_in->tid)) {
1632 		const char *pid_color = color;
1633 
1634 		if (thread__has_color(sched_in))
1635 			pid_color = COLOR_PIDS;
1636 
1637 		color_fprintf(stdout, pid_color, "%s => %s:%d",
1638 		       tr->shortname, thread__comm_str(sched_in), sched_in->tid);
1639 		tr->comm_changed = false;
1640 	}
1641 
1642 	if (sched->map.comp && new_cpu)
1643 		color_fprintf(stdout, color, " (CPU %d)", this_cpu);
1644 
1645 out:
1646 	color_fprintf(stdout, color, "\n");
1647 
1648 	thread__put(sched_in);
1649 
1650 	return 0;
1651 }
1652 
1653 static int process_sched_switch_event(struct perf_tool *tool,
1654 				      struct perf_evsel *evsel,
1655 				      struct perf_sample *sample,
1656 				      struct machine *machine)
1657 {
1658 	struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
1659 	int this_cpu = sample->cpu, err = 0;
1660 	u32 prev_pid = perf_evsel__intval(evsel, sample, "prev_pid"),
1661 	    next_pid = perf_evsel__intval(evsel, sample, "next_pid");
1662 
1663 	if (sched->curr_pid[this_cpu] != (u32)-1) {
1664 		/*
1665 		 * Are we trying to switch away a PID that is
1666 		 * not current?
1667 		 */
1668 		if (sched->curr_pid[this_cpu] != prev_pid)
1669 			sched->nr_context_switch_bugs++;
1670 	}
1671 
1672 	if (sched->tp_handler->switch_event)
1673 		err = sched->tp_handler->switch_event(sched, evsel, sample, machine);
1674 
1675 	sched->curr_pid[this_cpu] = next_pid;
1676 	return err;
1677 }
1678 
1679 static int process_sched_runtime_event(struct perf_tool *tool,
1680 				       struct perf_evsel *evsel,
1681 				       struct perf_sample *sample,
1682 				       struct machine *machine)
1683 {
1684 	struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
1685 
1686 	if (sched->tp_handler->runtime_event)
1687 		return sched->tp_handler->runtime_event(sched, evsel, sample, machine);
1688 
1689 	return 0;
1690 }
1691 
1692 static int perf_sched__process_fork_event(struct perf_tool *tool,
1693 					  union perf_event *event,
1694 					  struct perf_sample *sample,
1695 					  struct machine *machine)
1696 {
1697 	struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
1698 
1699 	/* run the fork event through the perf machineruy */
1700 	perf_event__process_fork(tool, event, sample, machine);
1701 
1702 	/* and then run additional processing needed for this command */
1703 	if (sched->tp_handler->fork_event)
1704 		return sched->tp_handler->fork_event(sched, event, machine);
1705 
1706 	return 0;
1707 }
1708 
1709 static int process_sched_migrate_task_event(struct perf_tool *tool,
1710 					    struct perf_evsel *evsel,
1711 					    struct perf_sample *sample,
1712 					    struct machine *machine)
1713 {
1714 	struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
1715 
1716 	if (sched->tp_handler->migrate_task_event)
1717 		return sched->tp_handler->migrate_task_event(sched, evsel, sample, machine);
1718 
1719 	return 0;
1720 }
1721 
1722 typedef int (*tracepoint_handler)(struct perf_tool *tool,
1723 				  struct perf_evsel *evsel,
1724 				  struct perf_sample *sample,
1725 				  struct machine *machine);
1726 
1727 static int perf_sched__process_tracepoint_sample(struct perf_tool *tool __maybe_unused,
1728 						 union perf_event *event __maybe_unused,
1729 						 struct perf_sample *sample,
1730 						 struct perf_evsel *evsel,
1731 						 struct machine *machine)
1732 {
1733 	int err = 0;
1734 
1735 	if (evsel->handler != NULL) {
1736 		tracepoint_handler f = evsel->handler;
1737 		err = f(tool, evsel, sample, machine);
1738 	}
1739 
1740 	return err;
1741 }
1742 
1743 static int perf_sched__process_comm(struct perf_tool *tool __maybe_unused,
1744 				    union perf_event *event,
1745 				    struct perf_sample *sample,
1746 				    struct machine *machine)
1747 {
1748 	struct thread *thread;
1749 	struct thread_runtime *tr;
1750 	int err;
1751 
1752 	err = perf_event__process_comm(tool, event, sample, machine);
1753 	if (err)
1754 		return err;
1755 
1756 	thread = machine__find_thread(machine, sample->pid, sample->tid);
1757 	if (!thread) {
1758 		pr_err("Internal error: can't find thread\n");
1759 		return -1;
1760 	}
1761 
1762 	tr = thread__get_runtime(thread);
1763 	if (tr == NULL) {
1764 		thread__put(thread);
1765 		return -1;
1766 	}
1767 
1768 	tr->comm_changed = true;
1769 	thread__put(thread);
1770 
1771 	return 0;
1772 }
1773 
1774 static int perf_sched__read_events(struct perf_sched *sched)
1775 {
1776 	const struct perf_evsel_str_handler handlers[] = {
1777 		{ "sched:sched_switch",	      process_sched_switch_event, },
1778 		{ "sched:sched_stat_runtime", process_sched_runtime_event, },
1779 		{ "sched:sched_wakeup",	      process_sched_wakeup_event, },
1780 		{ "sched:sched_wakeup_new",   process_sched_wakeup_event, },
1781 		{ "sched:sched_migrate_task", process_sched_migrate_task_event, },
1782 	};
1783 	struct perf_session *session;
1784 	struct perf_data data = {
1785 		.file      = {
1786 			.path = input_name,
1787 		},
1788 		.mode      = PERF_DATA_MODE_READ,
1789 		.force     = sched->force,
1790 	};
1791 	int rc = -1;
1792 
1793 	session = perf_session__new(&data, false, &sched->tool);
1794 	if (session == NULL) {
1795 		pr_debug("No Memory for session\n");
1796 		return -1;
1797 	}
1798 
1799 	symbol__init(&session->header.env);
1800 
1801 	if (perf_session__set_tracepoints_handlers(session, handlers))
1802 		goto out_delete;
1803 
1804 	if (perf_session__has_traces(session, "record -R")) {
1805 		int err = perf_session__process_events(session);
1806 		if (err) {
1807 			pr_err("Failed to process events, error %d", err);
1808 			goto out_delete;
1809 		}
1810 
1811 		sched->nr_events      = session->evlist->stats.nr_events[0];
1812 		sched->nr_lost_events = session->evlist->stats.total_lost;
1813 		sched->nr_lost_chunks = session->evlist->stats.nr_events[PERF_RECORD_LOST];
1814 	}
1815 
1816 	rc = 0;
1817 out_delete:
1818 	perf_session__delete(session);
1819 	return rc;
1820 }
1821 
1822 /*
1823  * scheduling times are printed as msec.usec
1824  */
1825 static inline void print_sched_time(unsigned long long nsecs, int width)
1826 {
1827 	unsigned long msecs;
1828 	unsigned long usecs;
1829 
1830 	msecs  = nsecs / NSEC_PER_MSEC;
1831 	nsecs -= msecs * NSEC_PER_MSEC;
1832 	usecs  = nsecs / NSEC_PER_USEC;
1833 	printf("%*lu.%03lu ", width, msecs, usecs);
1834 }
1835 
1836 /*
1837  * returns runtime data for event, allocating memory for it the
1838  * first time it is used.
1839  */
1840 static struct evsel_runtime *perf_evsel__get_runtime(struct perf_evsel *evsel)
1841 {
1842 	struct evsel_runtime *r = evsel->priv;
1843 
1844 	if (r == NULL) {
1845 		r = zalloc(sizeof(struct evsel_runtime));
1846 		evsel->priv = r;
1847 	}
1848 
1849 	return r;
1850 }
1851 
1852 /*
1853  * save last time event was seen per cpu
1854  */
1855 static void perf_evsel__save_time(struct perf_evsel *evsel,
1856 				  u64 timestamp, u32 cpu)
1857 {
1858 	struct evsel_runtime *r = perf_evsel__get_runtime(evsel);
1859 
1860 	if (r == NULL)
1861 		return;
1862 
1863 	if ((cpu >= r->ncpu) || (r->last_time == NULL)) {
1864 		int i, n = __roundup_pow_of_two(cpu+1);
1865 		void *p = r->last_time;
1866 
1867 		p = realloc(r->last_time, n * sizeof(u64));
1868 		if (!p)
1869 			return;
1870 
1871 		r->last_time = p;
1872 		for (i = r->ncpu; i < n; ++i)
1873 			r->last_time[i] = (u64) 0;
1874 
1875 		r->ncpu = n;
1876 	}
1877 
1878 	r->last_time[cpu] = timestamp;
1879 }
1880 
1881 /* returns last time this event was seen on the given cpu */
1882 static u64 perf_evsel__get_time(struct perf_evsel *evsel, u32 cpu)
1883 {
1884 	struct evsel_runtime *r = perf_evsel__get_runtime(evsel);
1885 
1886 	if ((r == NULL) || (r->last_time == NULL) || (cpu >= r->ncpu))
1887 		return 0;
1888 
1889 	return r->last_time[cpu];
1890 }
1891 
1892 static int comm_width = 30;
1893 
1894 static char *timehist_get_commstr(struct thread *thread)
1895 {
1896 	static char str[32];
1897 	const char *comm = thread__comm_str(thread);
1898 	pid_t tid = thread->tid;
1899 	pid_t pid = thread->pid_;
1900 	int n;
1901 
1902 	if (pid == 0)
1903 		n = scnprintf(str, sizeof(str), "%s", comm);
1904 
1905 	else if (tid != pid)
1906 		n = scnprintf(str, sizeof(str), "%s[%d/%d]", comm, tid, pid);
1907 
1908 	else
1909 		n = scnprintf(str, sizeof(str), "%s[%d]", comm, tid);
1910 
1911 	if (n > comm_width)
1912 		comm_width = n;
1913 
1914 	return str;
1915 }
1916 
1917 static void timehist_header(struct perf_sched *sched)
1918 {
1919 	u32 ncpus = sched->max_cpu + 1;
1920 	u32 i, j;
1921 
1922 	printf("%15s %6s ", "time", "cpu");
1923 
1924 	if (sched->show_cpu_visual) {
1925 		printf(" ");
1926 		for (i = 0, j = 0; i < ncpus; ++i) {
1927 			printf("%x", j++);
1928 			if (j > 15)
1929 				j = 0;
1930 		}
1931 		printf(" ");
1932 	}
1933 
1934 	printf(" %-*s  %9s  %9s  %9s", comm_width,
1935 		"task name", "wait time", "sch delay", "run time");
1936 
1937 	if (sched->show_state)
1938 		printf("  %s", "state");
1939 
1940 	printf("\n");
1941 
1942 	/*
1943 	 * units row
1944 	 */
1945 	printf("%15s %-6s ", "", "");
1946 
1947 	if (sched->show_cpu_visual)
1948 		printf(" %*s ", ncpus, "");
1949 
1950 	printf(" %-*s  %9s  %9s  %9s", comm_width,
1951 	       "[tid/pid]", "(msec)", "(msec)", "(msec)");
1952 
1953 	if (sched->show_state)
1954 		printf("  %5s", "");
1955 
1956 	printf("\n");
1957 
1958 	/*
1959 	 * separator
1960 	 */
1961 	printf("%.15s %.6s ", graph_dotted_line, graph_dotted_line);
1962 
1963 	if (sched->show_cpu_visual)
1964 		printf(" %.*s ", ncpus, graph_dotted_line);
1965 
1966 	printf(" %.*s  %.9s  %.9s  %.9s", comm_width,
1967 		graph_dotted_line, graph_dotted_line, graph_dotted_line,
1968 		graph_dotted_line);
1969 
1970 	if (sched->show_state)
1971 		printf("  %.5s", graph_dotted_line);
1972 
1973 	printf("\n");
1974 }
1975 
1976 static char task_state_char(struct thread *thread, int state)
1977 {
1978 	static const char state_to_char[] = TASK_STATE_TO_CHAR_STR;
1979 	unsigned bit = state ? ffs(state) : 0;
1980 
1981 	/* 'I' for idle */
1982 	if (thread->tid == 0)
1983 		return 'I';
1984 
1985 	return bit < sizeof(state_to_char) - 1 ? state_to_char[bit] : '?';
1986 }
1987 
1988 static void timehist_print_sample(struct perf_sched *sched,
1989 				  struct perf_evsel *evsel,
1990 				  struct perf_sample *sample,
1991 				  struct addr_location *al,
1992 				  struct thread *thread,
1993 				  u64 t, int state)
1994 {
1995 	struct thread_runtime *tr = thread__priv(thread);
1996 	const char *next_comm = perf_evsel__strval(evsel, sample, "next_comm");
1997 	const u32 next_pid = perf_evsel__intval(evsel, sample, "next_pid");
1998 	u32 max_cpus = sched->max_cpu + 1;
1999 	char tstr[64];
2000 	char nstr[30];
2001 	u64 wait_time;
2002 
2003 	timestamp__scnprintf_usec(t, tstr, sizeof(tstr));
2004 	printf("%15s [%04d] ", tstr, sample->cpu);
2005 
2006 	if (sched->show_cpu_visual) {
2007 		u32 i;
2008 		char c;
2009 
2010 		printf(" ");
2011 		for (i = 0; i < max_cpus; ++i) {
2012 			/* flag idle times with 'i'; others are sched events */
2013 			if (i == sample->cpu)
2014 				c = (thread->tid == 0) ? 'i' : 's';
2015 			else
2016 				c = ' ';
2017 			printf("%c", c);
2018 		}
2019 		printf(" ");
2020 	}
2021 
2022 	printf(" %-*s ", comm_width, timehist_get_commstr(thread));
2023 
2024 	wait_time = tr->dt_sleep + tr->dt_iowait + tr->dt_preempt;
2025 	print_sched_time(wait_time, 6);
2026 
2027 	print_sched_time(tr->dt_delay, 6);
2028 	print_sched_time(tr->dt_run, 6);
2029 
2030 	if (sched->show_state)
2031 		printf(" %5c ", task_state_char(thread, state));
2032 
2033 	if (sched->show_next) {
2034 		snprintf(nstr, sizeof(nstr), "next: %s[%d]", next_comm, next_pid);
2035 		printf(" %-*s", comm_width, nstr);
2036 	}
2037 
2038 	if (sched->show_wakeups && !sched->show_next)
2039 		printf("  %-*s", comm_width, "");
2040 
2041 	if (thread->tid == 0)
2042 		goto out;
2043 
2044 	if (sched->show_callchain)
2045 		printf("  ");
2046 
2047 	sample__fprintf_sym(sample, al, 0,
2048 			    EVSEL__PRINT_SYM | EVSEL__PRINT_ONELINE |
2049 			    EVSEL__PRINT_CALLCHAIN_ARROW |
2050 			    EVSEL__PRINT_SKIP_IGNORED,
2051 			    &callchain_cursor, stdout);
2052 
2053 out:
2054 	printf("\n");
2055 }
2056 
2057 /*
2058  * Explanation of delta-time stats:
2059  *
2060  *            t = time of current schedule out event
2061  *        tprev = time of previous sched out event
2062  *                also time of schedule-in event for current task
2063  *    last_time = time of last sched change event for current task
2064  *                (i.e, time process was last scheduled out)
2065  * ready_to_run = time of wakeup for current task
2066  *
2067  * -----|------------|------------|------------|------
2068  *    last         ready        tprev          t
2069  *    time         to run
2070  *
2071  *      |-------- dt_wait --------|
2072  *                   |- dt_delay -|-- dt_run --|
2073  *
2074  *   dt_run = run time of current task
2075  *  dt_wait = time between last schedule out event for task and tprev
2076  *            represents time spent off the cpu
2077  * dt_delay = time between wakeup and schedule-in of task
2078  */
2079 
2080 static void timehist_update_runtime_stats(struct thread_runtime *r,
2081 					 u64 t, u64 tprev)
2082 {
2083 	r->dt_delay   = 0;
2084 	r->dt_sleep   = 0;
2085 	r->dt_iowait  = 0;
2086 	r->dt_preempt = 0;
2087 	r->dt_run     = 0;
2088 
2089 	if (tprev) {
2090 		r->dt_run = t - tprev;
2091 		if (r->ready_to_run) {
2092 			if (r->ready_to_run > tprev)
2093 				pr_debug("time travel: wakeup time for task > previous sched_switch event\n");
2094 			else
2095 				r->dt_delay = tprev - r->ready_to_run;
2096 		}
2097 
2098 		if (r->last_time > tprev)
2099 			pr_debug("time travel: last sched out time for task > previous sched_switch event\n");
2100 		else if (r->last_time) {
2101 			u64 dt_wait = tprev - r->last_time;
2102 
2103 			if (r->last_state == TASK_RUNNING)
2104 				r->dt_preempt = dt_wait;
2105 			else if (r->last_state == TASK_UNINTERRUPTIBLE)
2106 				r->dt_iowait = dt_wait;
2107 			else
2108 				r->dt_sleep = dt_wait;
2109 		}
2110 	}
2111 
2112 	update_stats(&r->run_stats, r->dt_run);
2113 
2114 	r->total_run_time     += r->dt_run;
2115 	r->total_delay_time   += r->dt_delay;
2116 	r->total_sleep_time   += r->dt_sleep;
2117 	r->total_iowait_time  += r->dt_iowait;
2118 	r->total_preempt_time += r->dt_preempt;
2119 }
2120 
2121 static bool is_idle_sample(struct perf_sample *sample,
2122 			   struct perf_evsel *evsel)
2123 {
2124 	/* pid 0 == swapper == idle task */
2125 	if (strcmp(perf_evsel__name(evsel), "sched:sched_switch") == 0)
2126 		return perf_evsel__intval(evsel, sample, "prev_pid") == 0;
2127 
2128 	return sample->pid == 0;
2129 }
2130 
2131 static void save_task_callchain(struct perf_sched *sched,
2132 				struct perf_sample *sample,
2133 				struct perf_evsel *evsel,
2134 				struct machine *machine)
2135 {
2136 	struct callchain_cursor *cursor = &callchain_cursor;
2137 	struct thread *thread;
2138 
2139 	/* want main thread for process - has maps */
2140 	thread = machine__findnew_thread(machine, sample->pid, sample->pid);
2141 	if (thread == NULL) {
2142 		pr_debug("Failed to get thread for pid %d.\n", sample->pid);
2143 		return;
2144 	}
2145 
2146 	if (!sched->show_callchain || sample->callchain == NULL)
2147 		return;
2148 
2149 	if (thread__resolve_callchain(thread, cursor, evsel, sample,
2150 				      NULL, NULL, sched->max_stack + 2) != 0) {
2151 		if (verbose > 0)
2152 			pr_err("Failed to resolve callchain. Skipping\n");
2153 
2154 		return;
2155 	}
2156 
2157 	callchain_cursor_commit(cursor);
2158 
2159 	while (true) {
2160 		struct callchain_cursor_node *node;
2161 		struct symbol *sym;
2162 
2163 		node = callchain_cursor_current(cursor);
2164 		if (node == NULL)
2165 			break;
2166 
2167 		sym = node->sym;
2168 		if (sym) {
2169 			if (!strcmp(sym->name, "schedule") ||
2170 			    !strcmp(sym->name, "__schedule") ||
2171 			    !strcmp(sym->name, "preempt_schedule"))
2172 				sym->ignore = 1;
2173 		}
2174 
2175 		callchain_cursor_advance(cursor);
2176 	}
2177 }
2178 
2179 static int init_idle_thread(struct thread *thread)
2180 {
2181 	struct idle_thread_runtime *itr;
2182 
2183 	thread__set_comm(thread, idle_comm, 0);
2184 
2185 	itr = zalloc(sizeof(*itr));
2186 	if (itr == NULL)
2187 		return -ENOMEM;
2188 
2189 	init_stats(&itr->tr.run_stats);
2190 	callchain_init(&itr->callchain);
2191 	callchain_cursor_reset(&itr->cursor);
2192 	thread__set_priv(thread, itr);
2193 
2194 	return 0;
2195 }
2196 
2197 /*
2198  * Track idle stats per cpu by maintaining a local thread
2199  * struct for the idle task on each cpu.
2200  */
2201 static int init_idle_threads(int ncpu)
2202 {
2203 	int i, ret;
2204 
2205 	idle_threads = zalloc(ncpu * sizeof(struct thread *));
2206 	if (!idle_threads)
2207 		return -ENOMEM;
2208 
2209 	idle_max_cpu = ncpu;
2210 
2211 	/* allocate the actual thread struct if needed */
2212 	for (i = 0; i < ncpu; ++i) {
2213 		idle_threads[i] = thread__new(0, 0);
2214 		if (idle_threads[i] == NULL)
2215 			return -ENOMEM;
2216 
2217 		ret = init_idle_thread(idle_threads[i]);
2218 		if (ret < 0)
2219 			return ret;
2220 	}
2221 
2222 	return 0;
2223 }
2224 
2225 static void free_idle_threads(void)
2226 {
2227 	int i;
2228 
2229 	if (idle_threads == NULL)
2230 		return;
2231 
2232 	for (i = 0; i < idle_max_cpu; ++i) {
2233 		if ((idle_threads[i]))
2234 			thread__delete(idle_threads[i]);
2235 	}
2236 
2237 	free(idle_threads);
2238 }
2239 
2240 static struct thread *get_idle_thread(int cpu)
2241 {
2242 	/*
2243 	 * expand/allocate array of pointers to local thread
2244 	 * structs if needed
2245 	 */
2246 	if ((cpu >= idle_max_cpu) || (idle_threads == NULL)) {
2247 		int i, j = __roundup_pow_of_two(cpu+1);
2248 		void *p;
2249 
2250 		p = realloc(idle_threads, j * sizeof(struct thread *));
2251 		if (!p)
2252 			return NULL;
2253 
2254 		idle_threads = (struct thread **) p;
2255 		for (i = idle_max_cpu; i < j; ++i)
2256 			idle_threads[i] = NULL;
2257 
2258 		idle_max_cpu = j;
2259 	}
2260 
2261 	/* allocate a new thread struct if needed */
2262 	if (idle_threads[cpu] == NULL) {
2263 		idle_threads[cpu] = thread__new(0, 0);
2264 
2265 		if (idle_threads[cpu]) {
2266 			if (init_idle_thread(idle_threads[cpu]) < 0)
2267 				return NULL;
2268 		}
2269 	}
2270 
2271 	return idle_threads[cpu];
2272 }
2273 
2274 static void save_idle_callchain(struct perf_sched *sched,
2275 				struct idle_thread_runtime *itr,
2276 				struct perf_sample *sample)
2277 {
2278 	if (!sched->show_callchain || sample->callchain == NULL)
2279 		return;
2280 
2281 	callchain_cursor__copy(&itr->cursor, &callchain_cursor);
2282 }
2283 
2284 static struct thread *timehist_get_thread(struct perf_sched *sched,
2285 					  struct perf_sample *sample,
2286 					  struct machine *machine,
2287 					  struct perf_evsel *evsel)
2288 {
2289 	struct thread *thread;
2290 
2291 	if (is_idle_sample(sample, evsel)) {
2292 		thread = get_idle_thread(sample->cpu);
2293 		if (thread == NULL)
2294 			pr_err("Failed to get idle thread for cpu %d.\n", sample->cpu);
2295 
2296 	} else {
2297 		/* there were samples with tid 0 but non-zero pid */
2298 		thread = machine__findnew_thread(machine, sample->pid,
2299 						 sample->tid ?: sample->pid);
2300 		if (thread == NULL) {
2301 			pr_debug("Failed to get thread for tid %d. skipping sample.\n",
2302 				 sample->tid);
2303 		}
2304 
2305 		save_task_callchain(sched, sample, evsel, machine);
2306 		if (sched->idle_hist) {
2307 			struct thread *idle;
2308 			struct idle_thread_runtime *itr;
2309 
2310 			idle = get_idle_thread(sample->cpu);
2311 			if (idle == NULL) {
2312 				pr_err("Failed to get idle thread for cpu %d.\n", sample->cpu);
2313 				return NULL;
2314 			}
2315 
2316 			itr = thread__priv(idle);
2317 			if (itr == NULL)
2318 				return NULL;
2319 
2320 			itr->last_thread = thread;
2321 
2322 			/* copy task callchain when entering to idle */
2323 			if (perf_evsel__intval(evsel, sample, "next_pid") == 0)
2324 				save_idle_callchain(sched, itr, sample);
2325 		}
2326 	}
2327 
2328 	return thread;
2329 }
2330 
2331 static bool timehist_skip_sample(struct perf_sched *sched,
2332 				 struct thread *thread,
2333 				 struct perf_evsel *evsel,
2334 				 struct perf_sample *sample)
2335 {
2336 	bool rc = false;
2337 
2338 	if (thread__is_filtered(thread)) {
2339 		rc = true;
2340 		sched->skipped_samples++;
2341 	}
2342 
2343 	if (sched->idle_hist) {
2344 		if (strcmp(perf_evsel__name(evsel), "sched:sched_switch"))
2345 			rc = true;
2346 		else if (perf_evsel__intval(evsel, sample, "prev_pid") != 0 &&
2347 			 perf_evsel__intval(evsel, sample, "next_pid") != 0)
2348 			rc = true;
2349 	}
2350 
2351 	return rc;
2352 }
2353 
2354 static void timehist_print_wakeup_event(struct perf_sched *sched,
2355 					struct perf_evsel *evsel,
2356 					struct perf_sample *sample,
2357 					struct machine *machine,
2358 					struct thread *awakened)
2359 {
2360 	struct thread *thread;
2361 	char tstr[64];
2362 
2363 	thread = machine__findnew_thread(machine, sample->pid, sample->tid);
2364 	if (thread == NULL)
2365 		return;
2366 
2367 	/* show wakeup unless both awakee and awaker are filtered */
2368 	if (timehist_skip_sample(sched, thread, evsel, sample) &&
2369 	    timehist_skip_sample(sched, awakened, evsel, sample)) {
2370 		return;
2371 	}
2372 
2373 	timestamp__scnprintf_usec(sample->time, tstr, sizeof(tstr));
2374 	printf("%15s [%04d] ", tstr, sample->cpu);
2375 	if (sched->show_cpu_visual)
2376 		printf(" %*s ", sched->max_cpu + 1, "");
2377 
2378 	printf(" %-*s ", comm_width, timehist_get_commstr(thread));
2379 
2380 	/* dt spacer */
2381 	printf("  %9s  %9s  %9s ", "", "", "");
2382 
2383 	printf("awakened: %s", timehist_get_commstr(awakened));
2384 
2385 	printf("\n");
2386 }
2387 
2388 static int timehist_sched_wakeup_event(struct perf_tool *tool,
2389 				       union perf_event *event __maybe_unused,
2390 				       struct perf_evsel *evsel,
2391 				       struct perf_sample *sample,
2392 				       struct machine *machine)
2393 {
2394 	struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
2395 	struct thread *thread;
2396 	struct thread_runtime *tr = NULL;
2397 	/* want pid of awakened task not pid in sample */
2398 	const u32 pid = perf_evsel__intval(evsel, sample, "pid");
2399 
2400 	thread = machine__findnew_thread(machine, 0, pid);
2401 	if (thread == NULL)
2402 		return -1;
2403 
2404 	tr = thread__get_runtime(thread);
2405 	if (tr == NULL)
2406 		return -1;
2407 
2408 	if (tr->ready_to_run == 0)
2409 		tr->ready_to_run = sample->time;
2410 
2411 	/* show wakeups if requested */
2412 	if (sched->show_wakeups &&
2413 	    !perf_time__skip_sample(&sched->ptime, sample->time))
2414 		timehist_print_wakeup_event(sched, evsel, sample, machine, thread);
2415 
2416 	return 0;
2417 }
2418 
2419 static void timehist_print_migration_event(struct perf_sched *sched,
2420 					struct perf_evsel *evsel,
2421 					struct perf_sample *sample,
2422 					struct machine *machine,
2423 					struct thread *migrated)
2424 {
2425 	struct thread *thread;
2426 	char tstr[64];
2427 	u32 max_cpus = sched->max_cpu + 1;
2428 	u32 ocpu, dcpu;
2429 
2430 	if (sched->summary_only)
2431 		return;
2432 
2433 	max_cpus = sched->max_cpu + 1;
2434 	ocpu = perf_evsel__intval(evsel, sample, "orig_cpu");
2435 	dcpu = perf_evsel__intval(evsel, sample, "dest_cpu");
2436 
2437 	thread = machine__findnew_thread(machine, sample->pid, sample->tid);
2438 	if (thread == NULL)
2439 		return;
2440 
2441 	if (timehist_skip_sample(sched, thread, evsel, sample) &&
2442 	    timehist_skip_sample(sched, migrated, evsel, sample)) {
2443 		return;
2444 	}
2445 
2446 	timestamp__scnprintf_usec(sample->time, tstr, sizeof(tstr));
2447 	printf("%15s [%04d] ", tstr, sample->cpu);
2448 
2449 	if (sched->show_cpu_visual) {
2450 		u32 i;
2451 		char c;
2452 
2453 		printf("  ");
2454 		for (i = 0; i < max_cpus; ++i) {
2455 			c = (i == sample->cpu) ? 'm' : ' ';
2456 			printf("%c", c);
2457 		}
2458 		printf("  ");
2459 	}
2460 
2461 	printf(" %-*s ", comm_width, timehist_get_commstr(thread));
2462 
2463 	/* dt spacer */
2464 	printf("  %9s  %9s  %9s ", "", "", "");
2465 
2466 	printf("migrated: %s", timehist_get_commstr(migrated));
2467 	printf(" cpu %d => %d", ocpu, dcpu);
2468 
2469 	printf("\n");
2470 }
2471 
2472 static int timehist_migrate_task_event(struct perf_tool *tool,
2473 				       union perf_event *event __maybe_unused,
2474 				       struct perf_evsel *evsel,
2475 				       struct perf_sample *sample,
2476 				       struct machine *machine)
2477 {
2478 	struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
2479 	struct thread *thread;
2480 	struct thread_runtime *tr = NULL;
2481 	/* want pid of migrated task not pid in sample */
2482 	const u32 pid = perf_evsel__intval(evsel, sample, "pid");
2483 
2484 	thread = machine__findnew_thread(machine, 0, pid);
2485 	if (thread == NULL)
2486 		return -1;
2487 
2488 	tr = thread__get_runtime(thread);
2489 	if (tr == NULL)
2490 		return -1;
2491 
2492 	tr->migrations++;
2493 
2494 	/* show migrations if requested */
2495 	timehist_print_migration_event(sched, evsel, sample, machine, thread);
2496 
2497 	return 0;
2498 }
2499 
2500 static int timehist_sched_change_event(struct perf_tool *tool,
2501 				       union perf_event *event,
2502 				       struct perf_evsel *evsel,
2503 				       struct perf_sample *sample,
2504 				       struct machine *machine)
2505 {
2506 	struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
2507 	struct perf_time_interval *ptime = &sched->ptime;
2508 	struct addr_location al;
2509 	struct thread *thread;
2510 	struct thread_runtime *tr = NULL;
2511 	u64 tprev, t = sample->time;
2512 	int rc = 0;
2513 	int state = perf_evsel__intval(evsel, sample, "prev_state");
2514 
2515 
2516 	if (machine__resolve(machine, &al, sample) < 0) {
2517 		pr_err("problem processing %d event. skipping it\n",
2518 		       event->header.type);
2519 		rc = -1;
2520 		goto out;
2521 	}
2522 
2523 	thread = timehist_get_thread(sched, sample, machine, evsel);
2524 	if (thread == NULL) {
2525 		rc = -1;
2526 		goto out;
2527 	}
2528 
2529 	if (timehist_skip_sample(sched, thread, evsel, sample))
2530 		goto out;
2531 
2532 	tr = thread__get_runtime(thread);
2533 	if (tr == NULL) {
2534 		rc = -1;
2535 		goto out;
2536 	}
2537 
2538 	tprev = perf_evsel__get_time(evsel, sample->cpu);
2539 
2540 	/*
2541 	 * If start time given:
2542 	 * - sample time is under window user cares about - skip sample
2543 	 * - tprev is under window user cares about  - reset to start of window
2544 	 */
2545 	if (ptime->start && ptime->start > t)
2546 		goto out;
2547 
2548 	if (tprev && ptime->start > tprev)
2549 		tprev = ptime->start;
2550 
2551 	/*
2552 	 * If end time given:
2553 	 * - previous sched event is out of window - we are done
2554 	 * - sample time is beyond window user cares about - reset it
2555 	 *   to close out stats for time window interest
2556 	 */
2557 	if (ptime->end) {
2558 		if (tprev > ptime->end)
2559 			goto out;
2560 
2561 		if (t > ptime->end)
2562 			t = ptime->end;
2563 	}
2564 
2565 	if (!sched->idle_hist || thread->tid == 0) {
2566 		timehist_update_runtime_stats(tr, t, tprev);
2567 
2568 		if (sched->idle_hist) {
2569 			struct idle_thread_runtime *itr = (void *)tr;
2570 			struct thread_runtime *last_tr;
2571 
2572 			BUG_ON(thread->tid != 0);
2573 
2574 			if (itr->last_thread == NULL)
2575 				goto out;
2576 
2577 			/* add current idle time as last thread's runtime */
2578 			last_tr = thread__get_runtime(itr->last_thread);
2579 			if (last_tr == NULL)
2580 				goto out;
2581 
2582 			timehist_update_runtime_stats(last_tr, t, tprev);
2583 			/*
2584 			 * remove delta time of last thread as it's not updated
2585 			 * and otherwise it will show an invalid value next
2586 			 * time.  we only care total run time and run stat.
2587 			 */
2588 			last_tr->dt_run = 0;
2589 			last_tr->dt_delay = 0;
2590 			last_tr->dt_sleep = 0;
2591 			last_tr->dt_iowait = 0;
2592 			last_tr->dt_preempt = 0;
2593 
2594 			if (itr->cursor.nr)
2595 				callchain_append(&itr->callchain, &itr->cursor, t - tprev);
2596 
2597 			itr->last_thread = NULL;
2598 		}
2599 	}
2600 
2601 	if (!sched->summary_only)
2602 		timehist_print_sample(sched, evsel, sample, &al, thread, t, state);
2603 
2604 out:
2605 	if (sched->hist_time.start == 0 && t >= ptime->start)
2606 		sched->hist_time.start = t;
2607 	if (ptime->end == 0 || t <= ptime->end)
2608 		sched->hist_time.end = t;
2609 
2610 	if (tr) {
2611 		/* time of this sched_switch event becomes last time task seen */
2612 		tr->last_time = sample->time;
2613 
2614 		/* last state is used to determine where to account wait time */
2615 		tr->last_state = state;
2616 
2617 		/* sched out event for task so reset ready to run time */
2618 		tr->ready_to_run = 0;
2619 	}
2620 
2621 	perf_evsel__save_time(evsel, sample->time, sample->cpu);
2622 
2623 	return rc;
2624 }
2625 
2626 static int timehist_sched_switch_event(struct perf_tool *tool,
2627 			     union perf_event *event,
2628 			     struct perf_evsel *evsel,
2629 			     struct perf_sample *sample,
2630 			     struct machine *machine __maybe_unused)
2631 {
2632 	return timehist_sched_change_event(tool, event, evsel, sample, machine);
2633 }
2634 
2635 static int process_lost(struct perf_tool *tool __maybe_unused,
2636 			union perf_event *event,
2637 			struct perf_sample *sample,
2638 			struct machine *machine __maybe_unused)
2639 {
2640 	char tstr[64];
2641 
2642 	timestamp__scnprintf_usec(sample->time, tstr, sizeof(tstr));
2643 	printf("%15s ", tstr);
2644 	printf("lost %" PRIu64 " events on cpu %d\n", event->lost.lost, sample->cpu);
2645 
2646 	return 0;
2647 }
2648 
2649 
2650 static void print_thread_runtime(struct thread *t,
2651 				 struct thread_runtime *r)
2652 {
2653 	double mean = avg_stats(&r->run_stats);
2654 	float stddev;
2655 
2656 	printf("%*s   %5d  %9" PRIu64 " ",
2657 	       comm_width, timehist_get_commstr(t), t->ppid,
2658 	       (u64) r->run_stats.n);
2659 
2660 	print_sched_time(r->total_run_time, 8);
2661 	stddev = rel_stddev_stats(stddev_stats(&r->run_stats), mean);
2662 	print_sched_time(r->run_stats.min, 6);
2663 	printf(" ");
2664 	print_sched_time((u64) mean, 6);
2665 	printf(" ");
2666 	print_sched_time(r->run_stats.max, 6);
2667 	printf("  ");
2668 	printf("%5.2f", stddev);
2669 	printf("   %5" PRIu64, r->migrations);
2670 	printf("\n");
2671 }
2672 
2673 static void print_thread_waittime(struct thread *t,
2674 				  struct thread_runtime *r)
2675 {
2676 	printf("%*s   %5d  %9" PRIu64 " ",
2677 	       comm_width, timehist_get_commstr(t), t->ppid,
2678 	       (u64) r->run_stats.n);
2679 
2680 	print_sched_time(r->total_run_time, 8);
2681 	print_sched_time(r->total_sleep_time, 6);
2682 	printf(" ");
2683 	print_sched_time(r->total_iowait_time, 6);
2684 	printf(" ");
2685 	print_sched_time(r->total_preempt_time, 6);
2686 	printf(" ");
2687 	print_sched_time(r->total_delay_time, 6);
2688 	printf("\n");
2689 }
2690 
2691 struct total_run_stats {
2692 	struct perf_sched *sched;
2693 	u64  sched_count;
2694 	u64  task_count;
2695 	u64  total_run_time;
2696 };
2697 
2698 static int __show_thread_runtime(struct thread *t, void *priv)
2699 {
2700 	struct total_run_stats *stats = priv;
2701 	struct thread_runtime *r;
2702 
2703 	if (thread__is_filtered(t))
2704 		return 0;
2705 
2706 	r = thread__priv(t);
2707 	if (r && r->run_stats.n) {
2708 		stats->task_count++;
2709 		stats->sched_count += r->run_stats.n;
2710 		stats->total_run_time += r->total_run_time;
2711 
2712 		if (stats->sched->show_state)
2713 			print_thread_waittime(t, r);
2714 		else
2715 			print_thread_runtime(t, r);
2716 	}
2717 
2718 	return 0;
2719 }
2720 
2721 static int show_thread_runtime(struct thread *t, void *priv)
2722 {
2723 	if (t->dead)
2724 		return 0;
2725 
2726 	return __show_thread_runtime(t, priv);
2727 }
2728 
2729 static int show_deadthread_runtime(struct thread *t, void *priv)
2730 {
2731 	if (!t->dead)
2732 		return 0;
2733 
2734 	return __show_thread_runtime(t, priv);
2735 }
2736 
2737 static size_t callchain__fprintf_folded(FILE *fp, struct callchain_node *node)
2738 {
2739 	const char *sep = " <- ";
2740 	struct callchain_list *chain;
2741 	size_t ret = 0;
2742 	char bf[1024];
2743 	bool first;
2744 
2745 	if (node == NULL)
2746 		return 0;
2747 
2748 	ret = callchain__fprintf_folded(fp, node->parent);
2749 	first = (ret == 0);
2750 
2751 	list_for_each_entry(chain, &node->val, list) {
2752 		if (chain->ip >= PERF_CONTEXT_MAX)
2753 			continue;
2754 		if (chain->ms.sym && chain->ms.sym->ignore)
2755 			continue;
2756 		ret += fprintf(fp, "%s%s", first ? "" : sep,
2757 			       callchain_list__sym_name(chain, bf, sizeof(bf),
2758 							false));
2759 		first = false;
2760 	}
2761 
2762 	return ret;
2763 }
2764 
2765 static size_t timehist_print_idlehist_callchain(struct rb_root *root)
2766 {
2767 	size_t ret = 0;
2768 	FILE *fp = stdout;
2769 	struct callchain_node *chain;
2770 	struct rb_node *rb_node = rb_first(root);
2771 
2772 	printf("  %16s  %8s  %s\n", "Idle time (msec)", "Count", "Callchains");
2773 	printf("  %.16s  %.8s  %.50s\n", graph_dotted_line, graph_dotted_line,
2774 	       graph_dotted_line);
2775 
2776 	while (rb_node) {
2777 		chain = rb_entry(rb_node, struct callchain_node, rb_node);
2778 		rb_node = rb_next(rb_node);
2779 
2780 		ret += fprintf(fp, "  ");
2781 		print_sched_time(chain->hit, 12);
2782 		ret += 16;  /* print_sched_time returns 2nd arg + 4 */
2783 		ret += fprintf(fp, " %8d  ", chain->count);
2784 		ret += callchain__fprintf_folded(fp, chain);
2785 		ret += fprintf(fp, "\n");
2786 	}
2787 
2788 	return ret;
2789 }
2790 
2791 static void timehist_print_summary(struct perf_sched *sched,
2792 				   struct perf_session *session)
2793 {
2794 	struct machine *m = &session->machines.host;
2795 	struct total_run_stats totals;
2796 	u64 task_count;
2797 	struct thread *t;
2798 	struct thread_runtime *r;
2799 	int i;
2800 	u64 hist_time = sched->hist_time.end - sched->hist_time.start;
2801 
2802 	memset(&totals, 0, sizeof(totals));
2803 	totals.sched = sched;
2804 
2805 	if (sched->idle_hist) {
2806 		printf("\nIdle-time summary\n");
2807 		printf("%*s  parent  sched-out  ", comm_width, "comm");
2808 		printf("  idle-time   min-idle    avg-idle    max-idle  stddev  migrations\n");
2809 	} else if (sched->show_state) {
2810 		printf("\nWait-time summary\n");
2811 		printf("%*s  parent   sched-in  ", comm_width, "comm");
2812 		printf("   run-time      sleep      iowait     preempt       delay\n");
2813 	} else {
2814 		printf("\nRuntime summary\n");
2815 		printf("%*s  parent   sched-in  ", comm_width, "comm");
2816 		printf("   run-time    min-run     avg-run     max-run  stddev  migrations\n");
2817 	}
2818 	printf("%*s            (count)  ", comm_width, "");
2819 	printf("     (msec)     (msec)      (msec)      (msec)       %s\n",
2820 	       sched->show_state ? "(msec)" : "%");
2821 	printf("%.117s\n", graph_dotted_line);
2822 
2823 	machine__for_each_thread(m, show_thread_runtime, &totals);
2824 	task_count = totals.task_count;
2825 	if (!task_count)
2826 		printf("<no still running tasks>\n");
2827 
2828 	printf("\nTerminated tasks:\n");
2829 	machine__for_each_thread(m, show_deadthread_runtime, &totals);
2830 	if (task_count == totals.task_count)
2831 		printf("<no terminated tasks>\n");
2832 
2833 	/* CPU idle stats not tracked when samples were skipped */
2834 	if (sched->skipped_samples && !sched->idle_hist)
2835 		return;
2836 
2837 	printf("\nIdle stats:\n");
2838 	for (i = 0; i < idle_max_cpu; ++i) {
2839 		t = idle_threads[i];
2840 		if (!t)
2841 			continue;
2842 
2843 		r = thread__priv(t);
2844 		if (r && r->run_stats.n) {
2845 			totals.sched_count += r->run_stats.n;
2846 			printf("    CPU %2d idle for ", i);
2847 			print_sched_time(r->total_run_time, 6);
2848 			printf(" msec  (%6.2f%%)\n", 100.0 * r->total_run_time / hist_time);
2849 		} else
2850 			printf("    CPU %2d idle entire time window\n", i);
2851 	}
2852 
2853 	if (sched->idle_hist && sched->show_callchain) {
2854 		callchain_param.mode  = CHAIN_FOLDED;
2855 		callchain_param.value = CCVAL_PERIOD;
2856 
2857 		callchain_register_param(&callchain_param);
2858 
2859 		printf("\nIdle stats by callchain:\n");
2860 		for (i = 0; i < idle_max_cpu; ++i) {
2861 			struct idle_thread_runtime *itr;
2862 
2863 			t = idle_threads[i];
2864 			if (!t)
2865 				continue;
2866 
2867 			itr = thread__priv(t);
2868 			if (itr == NULL)
2869 				continue;
2870 
2871 			callchain_param.sort(&itr->sorted_root, &itr->callchain,
2872 					     0, &callchain_param);
2873 
2874 			printf("  CPU %2d:", i);
2875 			print_sched_time(itr->tr.total_run_time, 6);
2876 			printf(" msec\n");
2877 			timehist_print_idlehist_callchain(&itr->sorted_root);
2878 			printf("\n");
2879 		}
2880 	}
2881 
2882 	printf("\n"
2883 	       "    Total number of unique tasks: %" PRIu64 "\n"
2884 	       "Total number of context switches: %" PRIu64 "\n",
2885 	       totals.task_count, totals.sched_count);
2886 
2887 	printf("           Total run time (msec): ");
2888 	print_sched_time(totals.total_run_time, 2);
2889 	printf("\n");
2890 
2891 	printf("    Total scheduling time (msec): ");
2892 	print_sched_time(hist_time, 2);
2893 	printf(" (x %d)\n", sched->max_cpu);
2894 }
2895 
2896 typedef int (*sched_handler)(struct perf_tool *tool,
2897 			  union perf_event *event,
2898 			  struct perf_evsel *evsel,
2899 			  struct perf_sample *sample,
2900 			  struct machine *machine);
2901 
2902 static int perf_timehist__process_sample(struct perf_tool *tool,
2903 					 union perf_event *event,
2904 					 struct perf_sample *sample,
2905 					 struct perf_evsel *evsel,
2906 					 struct machine *machine)
2907 {
2908 	struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
2909 	int err = 0;
2910 	int this_cpu = sample->cpu;
2911 
2912 	if (this_cpu > sched->max_cpu)
2913 		sched->max_cpu = this_cpu;
2914 
2915 	if (evsel->handler != NULL) {
2916 		sched_handler f = evsel->handler;
2917 
2918 		err = f(tool, event, evsel, sample, machine);
2919 	}
2920 
2921 	return err;
2922 }
2923 
2924 static int timehist_check_attr(struct perf_sched *sched,
2925 			       struct perf_evlist *evlist)
2926 {
2927 	struct perf_evsel *evsel;
2928 	struct evsel_runtime *er;
2929 
2930 	list_for_each_entry(evsel, &evlist->entries, node) {
2931 		er = perf_evsel__get_runtime(evsel);
2932 		if (er == NULL) {
2933 			pr_err("Failed to allocate memory for evsel runtime data\n");
2934 			return -1;
2935 		}
2936 
2937 		if (sched->show_callchain && !evsel__has_callchain(evsel)) {
2938 			pr_info("Samples do not have callchains.\n");
2939 			sched->show_callchain = 0;
2940 			symbol_conf.use_callchain = 0;
2941 		}
2942 	}
2943 
2944 	return 0;
2945 }
2946 
2947 static int perf_sched__timehist(struct perf_sched *sched)
2948 {
2949 	const struct perf_evsel_str_handler handlers[] = {
2950 		{ "sched:sched_switch",       timehist_sched_switch_event, },
2951 		{ "sched:sched_wakeup",	      timehist_sched_wakeup_event, },
2952 		{ "sched:sched_wakeup_new",   timehist_sched_wakeup_event, },
2953 	};
2954 	const struct perf_evsel_str_handler migrate_handlers[] = {
2955 		{ "sched:sched_migrate_task", timehist_migrate_task_event, },
2956 	};
2957 	struct perf_data data = {
2958 		.file      = {
2959 			.path = input_name,
2960 		},
2961 		.mode      = PERF_DATA_MODE_READ,
2962 		.force     = sched->force,
2963 	};
2964 
2965 	struct perf_session *session;
2966 	struct perf_evlist *evlist;
2967 	int err = -1;
2968 
2969 	/*
2970 	 * event handlers for timehist option
2971 	 */
2972 	sched->tool.sample	 = perf_timehist__process_sample;
2973 	sched->tool.mmap	 = perf_event__process_mmap;
2974 	sched->tool.comm	 = perf_event__process_comm;
2975 	sched->tool.exit	 = perf_event__process_exit;
2976 	sched->tool.fork	 = perf_event__process_fork;
2977 	sched->tool.lost	 = process_lost;
2978 	sched->tool.attr	 = perf_event__process_attr;
2979 	sched->tool.tracing_data = perf_event__process_tracing_data;
2980 	sched->tool.build_id	 = perf_event__process_build_id;
2981 
2982 	sched->tool.ordered_events = true;
2983 	sched->tool.ordering_requires_timestamps = true;
2984 
2985 	symbol_conf.use_callchain = sched->show_callchain;
2986 
2987 	session = perf_session__new(&data, false, &sched->tool);
2988 	if (session == NULL)
2989 		return -ENOMEM;
2990 
2991 	evlist = session->evlist;
2992 
2993 	symbol__init(&session->header.env);
2994 
2995 	if (perf_time__parse_str(&sched->ptime, sched->time_str) != 0) {
2996 		pr_err("Invalid time string\n");
2997 		return -EINVAL;
2998 	}
2999 
3000 	if (timehist_check_attr(sched, evlist) != 0)
3001 		goto out;
3002 
3003 	setup_pager();
3004 
3005 	/* setup per-evsel handlers */
3006 	if (perf_session__set_tracepoints_handlers(session, handlers))
3007 		goto out;
3008 
3009 	/* sched_switch event at a minimum needs to exist */
3010 	if (!perf_evlist__find_tracepoint_by_name(session->evlist,
3011 						  "sched:sched_switch")) {
3012 		pr_err("No sched_switch events found. Have you run 'perf sched record'?\n");
3013 		goto out;
3014 	}
3015 
3016 	if (sched->show_migrations &&
3017 	    perf_session__set_tracepoints_handlers(session, migrate_handlers))
3018 		goto out;
3019 
3020 	/* pre-allocate struct for per-CPU idle stats */
3021 	sched->max_cpu = session->header.env.nr_cpus_online;
3022 	if (sched->max_cpu == 0)
3023 		sched->max_cpu = 4;
3024 	if (init_idle_threads(sched->max_cpu))
3025 		goto out;
3026 
3027 	/* summary_only implies summary option, but don't overwrite summary if set */
3028 	if (sched->summary_only)
3029 		sched->summary = sched->summary_only;
3030 
3031 	if (!sched->summary_only)
3032 		timehist_header(sched);
3033 
3034 	err = perf_session__process_events(session);
3035 	if (err) {
3036 		pr_err("Failed to process events, error %d", err);
3037 		goto out;
3038 	}
3039 
3040 	sched->nr_events      = evlist->stats.nr_events[0];
3041 	sched->nr_lost_events = evlist->stats.total_lost;
3042 	sched->nr_lost_chunks = evlist->stats.nr_events[PERF_RECORD_LOST];
3043 
3044 	if (sched->summary)
3045 		timehist_print_summary(sched, session);
3046 
3047 out:
3048 	free_idle_threads();
3049 	perf_session__delete(session);
3050 
3051 	return err;
3052 }
3053 
3054 
3055 static void print_bad_events(struct perf_sched *sched)
3056 {
3057 	if (sched->nr_unordered_timestamps && sched->nr_timestamps) {
3058 		printf("  INFO: %.3f%% unordered timestamps (%ld out of %ld)\n",
3059 			(double)sched->nr_unordered_timestamps/(double)sched->nr_timestamps*100.0,
3060 			sched->nr_unordered_timestamps, sched->nr_timestamps);
3061 	}
3062 	if (sched->nr_lost_events && sched->nr_events) {
3063 		printf("  INFO: %.3f%% lost events (%ld out of %ld, in %ld chunks)\n",
3064 			(double)sched->nr_lost_events/(double)sched->nr_events * 100.0,
3065 			sched->nr_lost_events, sched->nr_events, sched->nr_lost_chunks);
3066 	}
3067 	if (sched->nr_context_switch_bugs && sched->nr_timestamps) {
3068 		printf("  INFO: %.3f%% context switch bugs (%ld out of %ld)",
3069 			(double)sched->nr_context_switch_bugs/(double)sched->nr_timestamps*100.0,
3070 			sched->nr_context_switch_bugs, sched->nr_timestamps);
3071 		if (sched->nr_lost_events)
3072 			printf(" (due to lost events?)");
3073 		printf("\n");
3074 	}
3075 }
3076 
3077 static void __merge_work_atoms(struct rb_root *root, struct work_atoms *data)
3078 {
3079 	struct rb_node **new = &(root->rb_node), *parent = NULL;
3080 	struct work_atoms *this;
3081 	const char *comm = thread__comm_str(data->thread), *this_comm;
3082 
3083 	while (*new) {
3084 		int cmp;
3085 
3086 		this = container_of(*new, struct work_atoms, node);
3087 		parent = *new;
3088 
3089 		this_comm = thread__comm_str(this->thread);
3090 		cmp = strcmp(comm, this_comm);
3091 		if (cmp > 0) {
3092 			new = &((*new)->rb_left);
3093 		} else if (cmp < 0) {
3094 			new = &((*new)->rb_right);
3095 		} else {
3096 			this->num_merged++;
3097 			this->total_runtime += data->total_runtime;
3098 			this->nb_atoms += data->nb_atoms;
3099 			this->total_lat += data->total_lat;
3100 			list_splice(&data->work_list, &this->work_list);
3101 			if (this->max_lat < data->max_lat) {
3102 				this->max_lat = data->max_lat;
3103 				this->max_lat_at = data->max_lat_at;
3104 			}
3105 			zfree(&data);
3106 			return;
3107 		}
3108 	}
3109 
3110 	data->num_merged++;
3111 	rb_link_node(&data->node, parent, new);
3112 	rb_insert_color(&data->node, root);
3113 }
3114 
3115 static void perf_sched__merge_lat(struct perf_sched *sched)
3116 {
3117 	struct work_atoms *data;
3118 	struct rb_node *node;
3119 
3120 	if (sched->skip_merge)
3121 		return;
3122 
3123 	while ((node = rb_first(&sched->atom_root))) {
3124 		rb_erase(node, &sched->atom_root);
3125 		data = rb_entry(node, struct work_atoms, node);
3126 		__merge_work_atoms(&sched->merged_atom_root, data);
3127 	}
3128 }
3129 
3130 static int perf_sched__lat(struct perf_sched *sched)
3131 {
3132 	struct rb_node *next;
3133 
3134 	setup_pager();
3135 
3136 	if (perf_sched__read_events(sched))
3137 		return -1;
3138 
3139 	perf_sched__merge_lat(sched);
3140 	perf_sched__sort_lat(sched);
3141 
3142 	printf("\n -----------------------------------------------------------------------------------------------------------------\n");
3143 	printf("  Task                  |   Runtime ms  | Switches | Average delay ms | Maximum delay ms | Maximum delay at       |\n");
3144 	printf(" -----------------------------------------------------------------------------------------------------------------\n");
3145 
3146 	next = rb_first(&sched->sorted_atom_root);
3147 
3148 	while (next) {
3149 		struct work_atoms *work_list;
3150 
3151 		work_list = rb_entry(next, struct work_atoms, node);
3152 		output_lat_thread(sched, work_list);
3153 		next = rb_next(next);
3154 		thread__zput(work_list->thread);
3155 	}
3156 
3157 	printf(" -----------------------------------------------------------------------------------------------------------------\n");
3158 	printf("  TOTAL:                |%11.3f ms |%9" PRIu64 " |\n",
3159 		(double)sched->all_runtime / NSEC_PER_MSEC, sched->all_count);
3160 
3161 	printf(" ---------------------------------------------------\n");
3162 
3163 	print_bad_events(sched);
3164 	printf("\n");
3165 
3166 	return 0;
3167 }
3168 
3169 static int setup_map_cpus(struct perf_sched *sched)
3170 {
3171 	struct cpu_map *map;
3172 
3173 	sched->max_cpu  = sysconf(_SC_NPROCESSORS_CONF);
3174 
3175 	if (sched->map.comp) {
3176 		sched->map.comp_cpus = zalloc(sched->max_cpu * sizeof(int));
3177 		if (!sched->map.comp_cpus)
3178 			return -1;
3179 	}
3180 
3181 	if (!sched->map.cpus_str)
3182 		return 0;
3183 
3184 	map = cpu_map__new(sched->map.cpus_str);
3185 	if (!map) {
3186 		pr_err("failed to get cpus map from %s\n", sched->map.cpus_str);
3187 		return -1;
3188 	}
3189 
3190 	sched->map.cpus = map;
3191 	return 0;
3192 }
3193 
3194 static int setup_color_pids(struct perf_sched *sched)
3195 {
3196 	struct thread_map *map;
3197 
3198 	if (!sched->map.color_pids_str)
3199 		return 0;
3200 
3201 	map = thread_map__new_by_tid_str(sched->map.color_pids_str);
3202 	if (!map) {
3203 		pr_err("failed to get thread map from %s\n", sched->map.color_pids_str);
3204 		return -1;
3205 	}
3206 
3207 	sched->map.color_pids = map;
3208 	return 0;
3209 }
3210 
3211 static int setup_color_cpus(struct perf_sched *sched)
3212 {
3213 	struct cpu_map *map;
3214 
3215 	if (!sched->map.color_cpus_str)
3216 		return 0;
3217 
3218 	map = cpu_map__new(sched->map.color_cpus_str);
3219 	if (!map) {
3220 		pr_err("failed to get thread map from %s\n", sched->map.color_cpus_str);
3221 		return -1;
3222 	}
3223 
3224 	sched->map.color_cpus = map;
3225 	return 0;
3226 }
3227 
3228 static int perf_sched__map(struct perf_sched *sched)
3229 {
3230 	if (setup_map_cpus(sched))
3231 		return -1;
3232 
3233 	if (setup_color_pids(sched))
3234 		return -1;
3235 
3236 	if (setup_color_cpus(sched))
3237 		return -1;
3238 
3239 	setup_pager();
3240 	if (perf_sched__read_events(sched))
3241 		return -1;
3242 	print_bad_events(sched);
3243 	return 0;
3244 }
3245 
3246 static int perf_sched__replay(struct perf_sched *sched)
3247 {
3248 	unsigned long i;
3249 
3250 	calibrate_run_measurement_overhead(sched);
3251 	calibrate_sleep_measurement_overhead(sched);
3252 
3253 	test_calibrations(sched);
3254 
3255 	if (perf_sched__read_events(sched))
3256 		return -1;
3257 
3258 	printf("nr_run_events:        %ld\n", sched->nr_run_events);
3259 	printf("nr_sleep_events:      %ld\n", sched->nr_sleep_events);
3260 	printf("nr_wakeup_events:     %ld\n", sched->nr_wakeup_events);
3261 
3262 	if (sched->targetless_wakeups)
3263 		printf("target-less wakeups:  %ld\n", sched->targetless_wakeups);
3264 	if (sched->multitarget_wakeups)
3265 		printf("multi-target wakeups: %ld\n", sched->multitarget_wakeups);
3266 	if (sched->nr_run_events_optimized)
3267 		printf("run atoms optimized: %ld\n",
3268 			sched->nr_run_events_optimized);
3269 
3270 	print_task_traces(sched);
3271 	add_cross_task_wakeups(sched);
3272 
3273 	create_tasks(sched);
3274 	printf("------------------------------------------------------------\n");
3275 	for (i = 0; i < sched->replay_repeat; i++)
3276 		run_one_test(sched);
3277 
3278 	return 0;
3279 }
3280 
3281 static void setup_sorting(struct perf_sched *sched, const struct option *options,
3282 			  const char * const usage_msg[])
3283 {
3284 	char *tmp, *tok, *str = strdup(sched->sort_order);
3285 
3286 	for (tok = strtok_r(str, ", ", &tmp);
3287 			tok; tok = strtok_r(NULL, ", ", &tmp)) {
3288 		if (sort_dimension__add(tok, &sched->sort_list) < 0) {
3289 			usage_with_options_msg(usage_msg, options,
3290 					"Unknown --sort key: `%s'", tok);
3291 		}
3292 	}
3293 
3294 	free(str);
3295 
3296 	sort_dimension__add("pid", &sched->cmp_pid);
3297 }
3298 
3299 static int __cmd_record(int argc, const char **argv)
3300 {
3301 	unsigned int rec_argc, i, j;
3302 	const char **rec_argv;
3303 	const char * const record_args[] = {
3304 		"record",
3305 		"-a",
3306 		"-R",
3307 		"-m", "1024",
3308 		"-c", "1",
3309 		"-e", "sched:sched_switch",
3310 		"-e", "sched:sched_stat_wait",
3311 		"-e", "sched:sched_stat_sleep",
3312 		"-e", "sched:sched_stat_iowait",
3313 		"-e", "sched:sched_stat_runtime",
3314 		"-e", "sched:sched_process_fork",
3315 		"-e", "sched:sched_wakeup",
3316 		"-e", "sched:sched_wakeup_new",
3317 		"-e", "sched:sched_migrate_task",
3318 	};
3319 
3320 	rec_argc = ARRAY_SIZE(record_args) + argc - 1;
3321 	rec_argv = calloc(rec_argc + 1, sizeof(char *));
3322 
3323 	if (rec_argv == NULL)
3324 		return -ENOMEM;
3325 
3326 	for (i = 0; i < ARRAY_SIZE(record_args); i++)
3327 		rec_argv[i] = strdup(record_args[i]);
3328 
3329 	for (j = 1; j < (unsigned int)argc; j++, i++)
3330 		rec_argv[i] = argv[j];
3331 
3332 	BUG_ON(i != rec_argc);
3333 
3334 	return cmd_record(i, rec_argv);
3335 }
3336 
3337 int cmd_sched(int argc, const char **argv)
3338 {
3339 	const char default_sort_order[] = "avg, max, switch, runtime";
3340 	struct perf_sched sched = {
3341 		.tool = {
3342 			.sample		 = perf_sched__process_tracepoint_sample,
3343 			.comm		 = perf_sched__process_comm,
3344 			.namespaces	 = perf_event__process_namespaces,
3345 			.lost		 = perf_event__process_lost,
3346 			.fork		 = perf_sched__process_fork_event,
3347 			.ordered_events = true,
3348 		},
3349 		.cmp_pid	      = LIST_HEAD_INIT(sched.cmp_pid),
3350 		.sort_list	      = LIST_HEAD_INIT(sched.sort_list),
3351 		.start_work_mutex     = PTHREAD_MUTEX_INITIALIZER,
3352 		.work_done_wait_mutex = PTHREAD_MUTEX_INITIALIZER,
3353 		.sort_order	      = default_sort_order,
3354 		.replay_repeat	      = 10,
3355 		.profile_cpu	      = -1,
3356 		.next_shortname1      = 'A',
3357 		.next_shortname2      = '0',
3358 		.skip_merge           = 0,
3359 		.show_callchain	      = 1,
3360 		.max_stack            = 5,
3361 	};
3362 	const struct option sched_options[] = {
3363 	OPT_STRING('i', "input", &input_name, "file",
3364 		    "input file name"),
3365 	OPT_INCR('v', "verbose", &verbose,
3366 		    "be more verbose (show symbol address, etc)"),
3367 	OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
3368 		    "dump raw trace in ASCII"),
3369 	OPT_BOOLEAN('f', "force", &sched.force, "don't complain, do it"),
3370 	OPT_END()
3371 	};
3372 	const struct option latency_options[] = {
3373 	OPT_STRING('s', "sort", &sched.sort_order, "key[,key2...]",
3374 		   "sort by key(s): runtime, switch, avg, max"),
3375 	OPT_INTEGER('C', "CPU", &sched.profile_cpu,
3376 		    "CPU to profile on"),
3377 	OPT_BOOLEAN('p', "pids", &sched.skip_merge,
3378 		    "latency stats per pid instead of per comm"),
3379 	OPT_PARENT(sched_options)
3380 	};
3381 	const struct option replay_options[] = {
3382 	OPT_UINTEGER('r', "repeat", &sched.replay_repeat,
3383 		     "repeat the workload replay N times (-1: infinite)"),
3384 	OPT_PARENT(sched_options)
3385 	};
3386 	const struct option map_options[] = {
3387 	OPT_BOOLEAN(0, "compact", &sched.map.comp,
3388 		    "map output in compact mode"),
3389 	OPT_STRING(0, "color-pids", &sched.map.color_pids_str, "pids",
3390 		   "highlight given pids in map"),
3391 	OPT_STRING(0, "color-cpus", &sched.map.color_cpus_str, "cpus",
3392                     "highlight given CPUs in map"),
3393 	OPT_STRING(0, "cpus", &sched.map.cpus_str, "cpus",
3394                     "display given CPUs in map"),
3395 	OPT_PARENT(sched_options)
3396 	};
3397 	const struct option timehist_options[] = {
3398 	OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name,
3399 		   "file", "vmlinux pathname"),
3400 	OPT_STRING(0, "kallsyms", &symbol_conf.kallsyms_name,
3401 		   "file", "kallsyms pathname"),
3402 	OPT_BOOLEAN('g', "call-graph", &sched.show_callchain,
3403 		    "Display call chains if present (default on)"),
3404 	OPT_UINTEGER(0, "max-stack", &sched.max_stack,
3405 		   "Maximum number of functions to display backtrace."),
3406 	OPT_STRING(0, "symfs", &symbol_conf.symfs, "directory",
3407 		    "Look for files with symbols relative to this directory"),
3408 	OPT_BOOLEAN('s', "summary", &sched.summary_only,
3409 		    "Show only syscall summary with statistics"),
3410 	OPT_BOOLEAN('S', "with-summary", &sched.summary,
3411 		    "Show all syscalls and summary with statistics"),
3412 	OPT_BOOLEAN('w', "wakeups", &sched.show_wakeups, "Show wakeup events"),
3413 	OPT_BOOLEAN('n', "next", &sched.show_next, "Show next task"),
3414 	OPT_BOOLEAN('M', "migrations", &sched.show_migrations, "Show migration events"),
3415 	OPT_BOOLEAN('V', "cpu-visual", &sched.show_cpu_visual, "Add CPU visual"),
3416 	OPT_BOOLEAN('I', "idle-hist", &sched.idle_hist, "Show idle events only"),
3417 	OPT_STRING(0, "time", &sched.time_str, "str",
3418 		   "Time span for analysis (start,stop)"),
3419 	OPT_BOOLEAN(0, "state", &sched.show_state, "Show task state when sched-out"),
3420 	OPT_STRING('p', "pid", &symbol_conf.pid_list_str, "pid[,pid...]",
3421 		   "analyze events only for given process id(s)"),
3422 	OPT_STRING('t', "tid", &symbol_conf.tid_list_str, "tid[,tid...]",
3423 		   "analyze events only for given thread id(s)"),
3424 	OPT_PARENT(sched_options)
3425 	};
3426 
3427 	const char * const latency_usage[] = {
3428 		"perf sched latency [<options>]",
3429 		NULL
3430 	};
3431 	const char * const replay_usage[] = {
3432 		"perf sched replay [<options>]",
3433 		NULL
3434 	};
3435 	const char * const map_usage[] = {
3436 		"perf sched map [<options>]",
3437 		NULL
3438 	};
3439 	const char * const timehist_usage[] = {
3440 		"perf sched timehist [<options>]",
3441 		NULL
3442 	};
3443 	const char *const sched_subcommands[] = { "record", "latency", "map",
3444 						  "replay", "script",
3445 						  "timehist", NULL };
3446 	const char *sched_usage[] = {
3447 		NULL,
3448 		NULL
3449 	};
3450 	struct trace_sched_handler lat_ops  = {
3451 		.wakeup_event	    = latency_wakeup_event,
3452 		.switch_event	    = latency_switch_event,
3453 		.runtime_event	    = latency_runtime_event,
3454 		.migrate_task_event = latency_migrate_task_event,
3455 	};
3456 	struct trace_sched_handler map_ops  = {
3457 		.switch_event	    = map_switch_event,
3458 	};
3459 	struct trace_sched_handler replay_ops  = {
3460 		.wakeup_event	    = replay_wakeup_event,
3461 		.switch_event	    = replay_switch_event,
3462 		.fork_event	    = replay_fork_event,
3463 	};
3464 	unsigned int i;
3465 
3466 	for (i = 0; i < ARRAY_SIZE(sched.curr_pid); i++)
3467 		sched.curr_pid[i] = -1;
3468 
3469 	argc = parse_options_subcommand(argc, argv, sched_options, sched_subcommands,
3470 					sched_usage, PARSE_OPT_STOP_AT_NON_OPTION);
3471 	if (!argc)
3472 		usage_with_options(sched_usage, sched_options);
3473 
3474 	/*
3475 	 * Aliased to 'perf script' for now:
3476 	 */
3477 	if (!strcmp(argv[0], "script"))
3478 		return cmd_script(argc, argv);
3479 
3480 	if (!strncmp(argv[0], "rec", 3)) {
3481 		return __cmd_record(argc, argv);
3482 	} else if (!strncmp(argv[0], "lat", 3)) {
3483 		sched.tp_handler = &lat_ops;
3484 		if (argc > 1) {
3485 			argc = parse_options(argc, argv, latency_options, latency_usage, 0);
3486 			if (argc)
3487 				usage_with_options(latency_usage, latency_options);
3488 		}
3489 		setup_sorting(&sched, latency_options, latency_usage);
3490 		return perf_sched__lat(&sched);
3491 	} else if (!strcmp(argv[0], "map")) {
3492 		if (argc) {
3493 			argc = parse_options(argc, argv, map_options, map_usage, 0);
3494 			if (argc)
3495 				usage_with_options(map_usage, map_options);
3496 		}
3497 		sched.tp_handler = &map_ops;
3498 		setup_sorting(&sched, latency_options, latency_usage);
3499 		return perf_sched__map(&sched);
3500 	} else if (!strncmp(argv[0], "rep", 3)) {
3501 		sched.tp_handler = &replay_ops;
3502 		if (argc) {
3503 			argc = parse_options(argc, argv, replay_options, replay_usage, 0);
3504 			if (argc)
3505 				usage_with_options(replay_usage, replay_options);
3506 		}
3507 		return perf_sched__replay(&sched);
3508 	} else if (!strcmp(argv[0], "timehist")) {
3509 		if (argc) {
3510 			argc = parse_options(argc, argv, timehist_options,
3511 					     timehist_usage, 0);
3512 			if (argc)
3513 				usage_with_options(timehist_usage, timehist_options);
3514 		}
3515 		if ((sched.show_wakeups || sched.show_next) &&
3516 		    sched.summary_only) {
3517 			pr_err(" Error: -s and -[n|w] are mutually exclusive.\n");
3518 			parse_options_usage(timehist_usage, timehist_options, "s", true);
3519 			if (sched.show_wakeups)
3520 				parse_options_usage(NULL, timehist_options, "w", true);
3521 			if (sched.show_next)
3522 				parse_options_usage(NULL, timehist_options, "n", true);
3523 			return -EINVAL;
3524 		}
3525 
3526 		return perf_sched__timehist(&sched);
3527 	} else {
3528 		usage_with_options(sched_usage, sched_options);
3529 	}
3530 
3531 	return 0;
3532 }
3533