Loading...
1#include "builtin.h"
2#include "perf.h"
3
4#include "util/util.h"
5#include "util/evlist.h"
6#include "util/cache.h"
7#include "util/evsel.h"
8#include "util/symbol.h"
9#include "util/thread.h"
10#include "util/header.h"
11#include "util/session.h"
12#include "util/tool.h"
13
14#include "util/parse-options.h"
15#include "util/trace-event.h"
16
17#include "util/debug.h"
18
19#include <sys/prctl.h>
20#include <sys/resource.h>
21
22#include <semaphore.h>
23#include <pthread.h>
24#include <math.h>
25
26#define PR_SET_NAME 15 /* Set process name */
27#define MAX_CPUS 4096
28#define COMM_LEN 20
29#define SYM_LEN 129
30#define MAX_PID 65536
31
32struct sched_atom;
33
34struct task_desc {
35 unsigned long nr;
36 unsigned long pid;
37 char comm[COMM_LEN];
38
39 unsigned long nr_events;
40 unsigned long curr_event;
41 struct sched_atom **atoms;
42
43 pthread_t thread;
44 sem_t sleep_sem;
45
46 sem_t ready_for_work;
47 sem_t work_done_sem;
48
49 u64 cpu_usage;
50};
51
52enum sched_event_type {
53 SCHED_EVENT_RUN,
54 SCHED_EVENT_SLEEP,
55 SCHED_EVENT_WAKEUP,
56 SCHED_EVENT_MIGRATION,
57};
58
59struct sched_atom {
60 enum sched_event_type type;
61 int specific_wait;
62 u64 timestamp;
63 u64 duration;
64 unsigned long nr;
65 sem_t *wait_sem;
66 struct task_desc *wakee;
67};
68
69#define TASK_STATE_TO_CHAR_STR "RSDTtZX"
70
71enum thread_state {
72 THREAD_SLEEPING = 0,
73 THREAD_WAIT_CPU,
74 THREAD_SCHED_IN,
75 THREAD_IGNORE
76};
77
78struct work_atom {
79 struct list_head list;
80 enum thread_state state;
81 u64 sched_out_time;
82 u64 wake_up_time;
83 u64 sched_in_time;
84 u64 runtime;
85};
86
87struct work_atoms {
88 struct list_head work_list;
89 struct thread *thread;
90 struct rb_node node;
91 u64 max_lat;
92 u64 max_lat_at;
93 u64 total_lat;
94 u64 nb_atoms;
95 u64 total_runtime;
96};
97
98typedef int (*sort_fn_t)(struct work_atoms *, struct work_atoms *);
99
100struct perf_sched;
101
102struct trace_sched_handler {
103 int (*switch_event)(struct perf_sched *sched, struct perf_evsel *evsel,
104 struct perf_sample *sample, struct machine *machine);
105
106 int (*runtime_event)(struct perf_sched *sched, struct perf_evsel *evsel,
107 struct perf_sample *sample, struct machine *machine);
108
109 int (*wakeup_event)(struct perf_sched *sched, struct perf_evsel *evsel,
110 struct perf_sample *sample, struct machine *machine);
111
112 /* PERF_RECORD_FORK event, not sched_process_fork tracepoint */
113 int (*fork_event)(struct perf_sched *sched, union perf_event *event,
114 struct machine *machine);
115
116 int (*migrate_task_event)(struct perf_sched *sched,
117 struct perf_evsel *evsel,
118 struct perf_sample *sample,
119 struct machine *machine);
120};
121
122struct perf_sched {
123 struct perf_tool tool;
124 const char *sort_order;
125 unsigned long nr_tasks;
126 struct task_desc *pid_to_task[MAX_PID];
127 struct task_desc **tasks;
128 const struct trace_sched_handler *tp_handler;
129 pthread_mutex_t start_work_mutex;
130 pthread_mutex_t work_done_wait_mutex;
131 int profile_cpu;
132/*
133 * Track the current task - that way we can know whether there's any
134 * weird events, such as a task being switched away that is not current.
135 */
136 int max_cpu;
137 u32 curr_pid[MAX_CPUS];
138 struct thread *curr_thread[MAX_CPUS];
139 char next_shortname1;
140 char next_shortname2;
141 unsigned int replay_repeat;
142 unsigned long nr_run_events;
143 unsigned long nr_sleep_events;
144 unsigned long nr_wakeup_events;
145 unsigned long nr_sleep_corrections;
146 unsigned long nr_run_events_optimized;
147 unsigned long targetless_wakeups;
148 unsigned long multitarget_wakeups;
149 unsigned long nr_runs;
150 unsigned long nr_timestamps;
151 unsigned long nr_unordered_timestamps;
152 unsigned long nr_state_machine_bugs;
153 unsigned long nr_context_switch_bugs;
154 unsigned long nr_events;
155 unsigned long nr_lost_chunks;
156 unsigned long nr_lost_events;
157 u64 run_measurement_overhead;
158 u64 sleep_measurement_overhead;
159 u64 start_time;
160 u64 cpu_usage;
161 u64 runavg_cpu_usage;
162 u64 parent_cpu_usage;
163 u64 runavg_parent_cpu_usage;
164 u64 sum_runtime;
165 u64 sum_fluct;
166 u64 run_avg;
167 u64 all_runtime;
168 u64 all_count;
169 u64 cpu_last_switched[MAX_CPUS];
170 struct rb_root atom_root, sorted_atom_root;
171 struct list_head sort_list, cmp_pid;
172};
173
174static u64 get_nsecs(void)
175{
176 struct timespec ts;
177
178 clock_gettime(CLOCK_MONOTONIC, &ts);
179
180 return ts.tv_sec * 1000000000ULL + ts.tv_nsec;
181}
182
183static void burn_nsecs(struct perf_sched *sched, u64 nsecs)
184{
185 u64 T0 = get_nsecs(), T1;
186
187 do {
188 T1 = get_nsecs();
189 } while (T1 + sched->run_measurement_overhead < T0 + nsecs);
190}
191
192static void sleep_nsecs(u64 nsecs)
193{
194 struct timespec ts;
195
196 ts.tv_nsec = nsecs % 999999999;
197 ts.tv_sec = nsecs / 999999999;
198
199 nanosleep(&ts, NULL);
200}
201
202static void calibrate_run_measurement_overhead(struct perf_sched *sched)
203{
204 u64 T0, T1, delta, min_delta = 1000000000ULL;
205 int i;
206
207 for (i = 0; i < 10; i++) {
208 T0 = get_nsecs();
209 burn_nsecs(sched, 0);
210 T1 = get_nsecs();
211 delta = T1-T0;
212 min_delta = min(min_delta, delta);
213 }
214 sched->run_measurement_overhead = min_delta;
215
216 printf("run measurement overhead: %" PRIu64 " nsecs\n", min_delta);
217}
218
219static void calibrate_sleep_measurement_overhead(struct perf_sched *sched)
220{
221 u64 T0, T1, delta, min_delta = 1000000000ULL;
222 int i;
223
224 for (i = 0; i < 10; i++) {
225 T0 = get_nsecs();
226 sleep_nsecs(10000);
227 T1 = get_nsecs();
228 delta = T1-T0;
229 min_delta = min(min_delta, delta);
230 }
231 min_delta -= 10000;
232 sched->sleep_measurement_overhead = min_delta;
233
234 printf("sleep measurement overhead: %" PRIu64 " nsecs\n", min_delta);
235}
236
237static struct sched_atom *
238get_new_event(struct task_desc *task, u64 timestamp)
239{
240 struct sched_atom *event = zalloc(sizeof(*event));
241 unsigned long idx = task->nr_events;
242 size_t size;
243
244 event->timestamp = timestamp;
245 event->nr = idx;
246
247 task->nr_events++;
248 size = sizeof(struct sched_atom *) * task->nr_events;
249 task->atoms = realloc(task->atoms, size);
250 BUG_ON(!task->atoms);
251
252 task->atoms[idx] = event;
253
254 return event;
255}
256
257static struct sched_atom *last_event(struct task_desc *task)
258{
259 if (!task->nr_events)
260 return NULL;
261
262 return task->atoms[task->nr_events - 1];
263}
264
265static void add_sched_event_run(struct perf_sched *sched, struct task_desc *task,
266 u64 timestamp, u64 duration)
267{
268 struct sched_atom *event, *curr_event = last_event(task);
269
270 /*
271 * optimize an existing RUN event by merging this one
272 * to it:
273 */
274 if (curr_event && curr_event->type == SCHED_EVENT_RUN) {
275 sched->nr_run_events_optimized++;
276 curr_event->duration += duration;
277 return;
278 }
279
280 event = get_new_event(task, timestamp);
281
282 event->type = SCHED_EVENT_RUN;
283 event->duration = duration;
284
285 sched->nr_run_events++;
286}
287
288static void add_sched_event_wakeup(struct perf_sched *sched, struct task_desc *task,
289 u64 timestamp, struct task_desc *wakee)
290{
291 struct sched_atom *event, *wakee_event;
292
293 event = get_new_event(task, timestamp);
294 event->type = SCHED_EVENT_WAKEUP;
295 event->wakee = wakee;
296
297 wakee_event = last_event(wakee);
298 if (!wakee_event || wakee_event->type != SCHED_EVENT_SLEEP) {
299 sched->targetless_wakeups++;
300 return;
301 }
302 if (wakee_event->wait_sem) {
303 sched->multitarget_wakeups++;
304 return;
305 }
306
307 wakee_event->wait_sem = zalloc(sizeof(*wakee_event->wait_sem));
308 sem_init(wakee_event->wait_sem, 0, 0);
309 wakee_event->specific_wait = 1;
310 event->wait_sem = wakee_event->wait_sem;
311
312 sched->nr_wakeup_events++;
313}
314
315static void add_sched_event_sleep(struct perf_sched *sched, struct task_desc *task,
316 u64 timestamp, u64 task_state __maybe_unused)
317{
318 struct sched_atom *event = get_new_event(task, timestamp);
319
320 event->type = SCHED_EVENT_SLEEP;
321
322 sched->nr_sleep_events++;
323}
324
325static struct task_desc *register_pid(struct perf_sched *sched,
326 unsigned long pid, const char *comm)
327{
328 struct task_desc *task;
329
330 BUG_ON(pid >= MAX_PID);
331
332 task = sched->pid_to_task[pid];
333
334 if (task)
335 return task;
336
337 task = zalloc(sizeof(*task));
338 task->pid = pid;
339 task->nr = sched->nr_tasks;
340 strcpy(task->comm, comm);
341 /*
342 * every task starts in sleeping state - this gets ignored
343 * if there's no wakeup pointing to this sleep state:
344 */
345 add_sched_event_sleep(sched, task, 0, 0);
346
347 sched->pid_to_task[pid] = task;
348 sched->nr_tasks++;
349 sched->tasks = realloc(sched->tasks, sched->nr_tasks * sizeof(struct task_task *));
350 BUG_ON(!sched->tasks);
351 sched->tasks[task->nr] = task;
352
353 if (verbose)
354 printf("registered task #%ld, PID %ld (%s)\n", sched->nr_tasks, pid, comm);
355
356 return task;
357}
358
359
360static void print_task_traces(struct perf_sched *sched)
361{
362 struct task_desc *task;
363 unsigned long i;
364
365 for (i = 0; i < sched->nr_tasks; i++) {
366 task = sched->tasks[i];
367 printf("task %6ld (%20s:%10ld), nr_events: %ld\n",
368 task->nr, task->comm, task->pid, task->nr_events);
369 }
370}
371
372static void add_cross_task_wakeups(struct perf_sched *sched)
373{
374 struct task_desc *task1, *task2;
375 unsigned long i, j;
376
377 for (i = 0; i < sched->nr_tasks; i++) {
378 task1 = sched->tasks[i];
379 j = i + 1;
380 if (j == sched->nr_tasks)
381 j = 0;
382 task2 = sched->tasks[j];
383 add_sched_event_wakeup(sched, task1, 0, task2);
384 }
385}
386
387static void perf_sched__process_event(struct perf_sched *sched,
388 struct sched_atom *atom)
389{
390 int ret = 0;
391
392 switch (atom->type) {
393 case SCHED_EVENT_RUN:
394 burn_nsecs(sched, atom->duration);
395 break;
396 case SCHED_EVENT_SLEEP:
397 if (atom->wait_sem)
398 ret = sem_wait(atom->wait_sem);
399 BUG_ON(ret);
400 break;
401 case SCHED_EVENT_WAKEUP:
402 if (atom->wait_sem)
403 ret = sem_post(atom->wait_sem);
404 BUG_ON(ret);
405 break;
406 case SCHED_EVENT_MIGRATION:
407 break;
408 default:
409 BUG_ON(1);
410 }
411}
412
413static u64 get_cpu_usage_nsec_parent(void)
414{
415 struct rusage ru;
416 u64 sum;
417 int err;
418
419 err = getrusage(RUSAGE_SELF, &ru);
420 BUG_ON(err);
421
422 sum = ru.ru_utime.tv_sec*1e9 + ru.ru_utime.tv_usec*1e3;
423 sum += ru.ru_stime.tv_sec*1e9 + ru.ru_stime.tv_usec*1e3;
424
425 return sum;
426}
427
428static int self_open_counters(void)
429{
430 struct perf_event_attr attr;
431 int fd;
432
433 memset(&attr, 0, sizeof(attr));
434
435 attr.type = PERF_TYPE_SOFTWARE;
436 attr.config = PERF_COUNT_SW_TASK_CLOCK;
437
438 fd = sys_perf_event_open(&attr, 0, -1, -1, 0);
439
440 if (fd < 0)
441 pr_err("Error: sys_perf_event_open() syscall returned "
442 "with %d (%s)\n", fd, strerror(errno));
443 return fd;
444}
445
446static u64 get_cpu_usage_nsec_self(int fd)
447{
448 u64 runtime;
449 int ret;
450
451 ret = read(fd, &runtime, sizeof(runtime));
452 BUG_ON(ret != sizeof(runtime));
453
454 return runtime;
455}
456
457struct sched_thread_parms {
458 struct task_desc *task;
459 struct perf_sched *sched;
460};
461
462static void *thread_func(void *ctx)
463{
464 struct sched_thread_parms *parms = ctx;
465 struct task_desc *this_task = parms->task;
466 struct perf_sched *sched = parms->sched;
467 u64 cpu_usage_0, cpu_usage_1;
468 unsigned long i, ret;
469 char comm2[22];
470 int fd;
471
472 zfree(&parms);
473
474 sprintf(comm2, ":%s", this_task->comm);
475 prctl(PR_SET_NAME, comm2);
476 fd = self_open_counters();
477 if (fd < 0)
478 return NULL;
479again:
480 ret = sem_post(&this_task->ready_for_work);
481 BUG_ON(ret);
482 ret = pthread_mutex_lock(&sched->start_work_mutex);
483 BUG_ON(ret);
484 ret = pthread_mutex_unlock(&sched->start_work_mutex);
485 BUG_ON(ret);
486
487 cpu_usage_0 = get_cpu_usage_nsec_self(fd);
488
489 for (i = 0; i < this_task->nr_events; i++) {
490 this_task->curr_event = i;
491 perf_sched__process_event(sched, this_task->atoms[i]);
492 }
493
494 cpu_usage_1 = get_cpu_usage_nsec_self(fd);
495 this_task->cpu_usage = cpu_usage_1 - cpu_usage_0;
496 ret = sem_post(&this_task->work_done_sem);
497 BUG_ON(ret);
498
499 ret = pthread_mutex_lock(&sched->work_done_wait_mutex);
500 BUG_ON(ret);
501 ret = pthread_mutex_unlock(&sched->work_done_wait_mutex);
502 BUG_ON(ret);
503
504 goto again;
505}
506
507static void create_tasks(struct perf_sched *sched)
508{
509 struct task_desc *task;
510 pthread_attr_t attr;
511 unsigned long i;
512 int err;
513
514 err = pthread_attr_init(&attr);
515 BUG_ON(err);
516 err = pthread_attr_setstacksize(&attr,
517 (size_t) max(16 * 1024, PTHREAD_STACK_MIN));
518 BUG_ON(err);
519 err = pthread_mutex_lock(&sched->start_work_mutex);
520 BUG_ON(err);
521 err = pthread_mutex_lock(&sched->work_done_wait_mutex);
522 BUG_ON(err);
523 for (i = 0; i < sched->nr_tasks; i++) {
524 struct sched_thread_parms *parms = malloc(sizeof(*parms));
525 BUG_ON(parms == NULL);
526 parms->task = task = sched->tasks[i];
527 parms->sched = sched;
528 sem_init(&task->sleep_sem, 0, 0);
529 sem_init(&task->ready_for_work, 0, 0);
530 sem_init(&task->work_done_sem, 0, 0);
531 task->curr_event = 0;
532 err = pthread_create(&task->thread, &attr, thread_func, parms);
533 BUG_ON(err);
534 }
535}
536
537static void wait_for_tasks(struct perf_sched *sched)
538{
539 u64 cpu_usage_0, cpu_usage_1;
540 struct task_desc *task;
541 unsigned long i, ret;
542
543 sched->start_time = get_nsecs();
544 sched->cpu_usage = 0;
545 pthread_mutex_unlock(&sched->work_done_wait_mutex);
546
547 for (i = 0; i < sched->nr_tasks; i++) {
548 task = sched->tasks[i];
549 ret = sem_wait(&task->ready_for_work);
550 BUG_ON(ret);
551 sem_init(&task->ready_for_work, 0, 0);
552 }
553 ret = pthread_mutex_lock(&sched->work_done_wait_mutex);
554 BUG_ON(ret);
555
556 cpu_usage_0 = get_cpu_usage_nsec_parent();
557
558 pthread_mutex_unlock(&sched->start_work_mutex);
559
560 for (i = 0; i < sched->nr_tasks; i++) {
561 task = sched->tasks[i];
562 ret = sem_wait(&task->work_done_sem);
563 BUG_ON(ret);
564 sem_init(&task->work_done_sem, 0, 0);
565 sched->cpu_usage += task->cpu_usage;
566 task->cpu_usage = 0;
567 }
568
569 cpu_usage_1 = get_cpu_usage_nsec_parent();
570 if (!sched->runavg_cpu_usage)
571 sched->runavg_cpu_usage = sched->cpu_usage;
572 sched->runavg_cpu_usage = (sched->runavg_cpu_usage * 9 + sched->cpu_usage) / 10;
573
574 sched->parent_cpu_usage = cpu_usage_1 - cpu_usage_0;
575 if (!sched->runavg_parent_cpu_usage)
576 sched->runavg_parent_cpu_usage = sched->parent_cpu_usage;
577 sched->runavg_parent_cpu_usage = (sched->runavg_parent_cpu_usage * 9 +
578 sched->parent_cpu_usage)/10;
579
580 ret = pthread_mutex_lock(&sched->start_work_mutex);
581 BUG_ON(ret);
582
583 for (i = 0; i < sched->nr_tasks; i++) {
584 task = sched->tasks[i];
585 sem_init(&task->sleep_sem, 0, 0);
586 task->curr_event = 0;
587 }
588}
589
590static void run_one_test(struct perf_sched *sched)
591{
592 u64 T0, T1, delta, avg_delta, fluct;
593
594 T0 = get_nsecs();
595 wait_for_tasks(sched);
596 T1 = get_nsecs();
597
598 delta = T1 - T0;
599 sched->sum_runtime += delta;
600 sched->nr_runs++;
601
602 avg_delta = sched->sum_runtime / sched->nr_runs;
603 if (delta < avg_delta)
604 fluct = avg_delta - delta;
605 else
606 fluct = delta - avg_delta;
607 sched->sum_fluct += fluct;
608 if (!sched->run_avg)
609 sched->run_avg = delta;
610 sched->run_avg = (sched->run_avg * 9 + delta) / 10;
611
612 printf("#%-3ld: %0.3f, ", sched->nr_runs, (double)delta / 1000000.0);
613
614 printf("ravg: %0.2f, ", (double)sched->run_avg / 1e6);
615
616 printf("cpu: %0.2f / %0.2f",
617 (double)sched->cpu_usage / 1e6, (double)sched->runavg_cpu_usage / 1e6);
618
619#if 0
620 /*
621 * rusage statistics done by the parent, these are less
622 * accurate than the sched->sum_exec_runtime based statistics:
623 */
624 printf(" [%0.2f / %0.2f]",
625 (double)sched->parent_cpu_usage/1e6,
626 (double)sched->runavg_parent_cpu_usage/1e6);
627#endif
628
629 printf("\n");
630
631 if (sched->nr_sleep_corrections)
632 printf(" (%ld sleep corrections)\n", sched->nr_sleep_corrections);
633 sched->nr_sleep_corrections = 0;
634}
635
636static void test_calibrations(struct perf_sched *sched)
637{
638 u64 T0, T1;
639
640 T0 = get_nsecs();
641 burn_nsecs(sched, 1e6);
642 T1 = get_nsecs();
643
644 printf("the run test took %" PRIu64 " nsecs\n", T1 - T0);
645
646 T0 = get_nsecs();
647 sleep_nsecs(1e6);
648 T1 = get_nsecs();
649
650 printf("the sleep test took %" PRIu64 " nsecs\n", T1 - T0);
651}
652
653static int
654replay_wakeup_event(struct perf_sched *sched,
655 struct perf_evsel *evsel, struct perf_sample *sample,
656 struct machine *machine __maybe_unused)
657{
658 const char *comm = perf_evsel__strval(evsel, sample, "comm");
659 const u32 pid = perf_evsel__intval(evsel, sample, "pid");
660 struct task_desc *waker, *wakee;
661
662 if (verbose) {
663 printf("sched_wakeup event %p\n", evsel);
664
665 printf(" ... pid %d woke up %s/%d\n", sample->tid, comm, pid);
666 }
667
668 waker = register_pid(sched, sample->tid, "<unknown>");
669 wakee = register_pid(sched, pid, comm);
670
671 add_sched_event_wakeup(sched, waker, sample->time, wakee);
672 return 0;
673}
674
675static int replay_switch_event(struct perf_sched *sched,
676 struct perf_evsel *evsel,
677 struct perf_sample *sample,
678 struct machine *machine __maybe_unused)
679{
680 const char *prev_comm = perf_evsel__strval(evsel, sample, "prev_comm"),
681 *next_comm = perf_evsel__strval(evsel, sample, "next_comm");
682 const u32 prev_pid = perf_evsel__intval(evsel, sample, "prev_pid"),
683 next_pid = perf_evsel__intval(evsel, sample, "next_pid");
684 const u64 prev_state = perf_evsel__intval(evsel, sample, "prev_state");
685 struct task_desc *prev, __maybe_unused *next;
686 u64 timestamp0, timestamp = sample->time;
687 int cpu = sample->cpu;
688 s64 delta;
689
690 if (verbose)
691 printf("sched_switch event %p\n", evsel);
692
693 if (cpu >= MAX_CPUS || cpu < 0)
694 return 0;
695
696 timestamp0 = sched->cpu_last_switched[cpu];
697 if (timestamp0)
698 delta = timestamp - timestamp0;
699 else
700 delta = 0;
701
702 if (delta < 0) {
703 pr_err("hm, delta: %" PRIu64 " < 0 ?\n", delta);
704 return -1;
705 }
706
707 pr_debug(" ... switch from %s/%d to %s/%d [ran %" PRIu64 " nsecs]\n",
708 prev_comm, prev_pid, next_comm, next_pid, delta);
709
710 prev = register_pid(sched, prev_pid, prev_comm);
711 next = register_pid(sched, next_pid, next_comm);
712
713 sched->cpu_last_switched[cpu] = timestamp;
714
715 add_sched_event_run(sched, prev, timestamp, delta);
716 add_sched_event_sleep(sched, prev, timestamp, prev_state);
717
718 return 0;
719}
720
721static int replay_fork_event(struct perf_sched *sched,
722 union perf_event *event,
723 struct machine *machine)
724{
725 struct thread *child, *parent;
726
727 child = machine__findnew_thread(machine, event->fork.pid,
728 event->fork.tid);
729 parent = machine__findnew_thread(machine, event->fork.ppid,
730 event->fork.ptid);
731
732 if (child == NULL || parent == NULL) {
733 pr_debug("thread does not exist on fork event: child %p, parent %p\n",
734 child, parent);
735 return 0;
736 }
737
738 if (verbose) {
739 printf("fork event\n");
740 printf("... parent: %s/%d\n", thread__comm_str(parent), parent->tid);
741 printf("... child: %s/%d\n", thread__comm_str(child), child->tid);
742 }
743
744 register_pid(sched, parent->tid, thread__comm_str(parent));
745 register_pid(sched, child->tid, thread__comm_str(child));
746 return 0;
747}
748
749struct sort_dimension {
750 const char *name;
751 sort_fn_t cmp;
752 struct list_head list;
753};
754
755static int
756thread_lat_cmp(struct list_head *list, struct work_atoms *l, struct work_atoms *r)
757{
758 struct sort_dimension *sort;
759 int ret = 0;
760
761 BUG_ON(list_empty(list));
762
763 list_for_each_entry(sort, list, list) {
764 ret = sort->cmp(l, r);
765 if (ret)
766 return ret;
767 }
768
769 return ret;
770}
771
772static struct work_atoms *
773thread_atoms_search(struct rb_root *root, struct thread *thread,
774 struct list_head *sort_list)
775{
776 struct rb_node *node = root->rb_node;
777 struct work_atoms key = { .thread = thread };
778
779 while (node) {
780 struct work_atoms *atoms;
781 int cmp;
782
783 atoms = container_of(node, struct work_atoms, node);
784
785 cmp = thread_lat_cmp(sort_list, &key, atoms);
786 if (cmp > 0)
787 node = node->rb_left;
788 else if (cmp < 0)
789 node = node->rb_right;
790 else {
791 BUG_ON(thread != atoms->thread);
792 return atoms;
793 }
794 }
795 return NULL;
796}
797
798static void
799__thread_latency_insert(struct rb_root *root, struct work_atoms *data,
800 struct list_head *sort_list)
801{
802 struct rb_node **new = &(root->rb_node), *parent = NULL;
803
804 while (*new) {
805 struct work_atoms *this;
806 int cmp;
807
808 this = container_of(*new, struct work_atoms, node);
809 parent = *new;
810
811 cmp = thread_lat_cmp(sort_list, data, this);
812
813 if (cmp > 0)
814 new = &((*new)->rb_left);
815 else
816 new = &((*new)->rb_right);
817 }
818
819 rb_link_node(&data->node, parent, new);
820 rb_insert_color(&data->node, root);
821}
822
823static int thread_atoms_insert(struct perf_sched *sched, struct thread *thread)
824{
825 struct work_atoms *atoms = zalloc(sizeof(*atoms));
826 if (!atoms) {
827 pr_err("No memory at %s\n", __func__);
828 return -1;
829 }
830
831 atoms->thread = thread;
832 INIT_LIST_HEAD(&atoms->work_list);
833 __thread_latency_insert(&sched->atom_root, atoms, &sched->cmp_pid);
834 return 0;
835}
836
837static char sched_out_state(u64 prev_state)
838{
839 const char *str = TASK_STATE_TO_CHAR_STR;
840
841 return str[prev_state];
842}
843
844static int
845add_sched_out_event(struct work_atoms *atoms,
846 char run_state,
847 u64 timestamp)
848{
849 struct work_atom *atom = zalloc(sizeof(*atom));
850 if (!atom) {
851 pr_err("Non memory at %s", __func__);
852 return -1;
853 }
854
855 atom->sched_out_time = timestamp;
856
857 if (run_state == 'R') {
858 atom->state = THREAD_WAIT_CPU;
859 atom->wake_up_time = atom->sched_out_time;
860 }
861
862 list_add_tail(&atom->list, &atoms->work_list);
863 return 0;
864}
865
866static void
867add_runtime_event(struct work_atoms *atoms, u64 delta,
868 u64 timestamp __maybe_unused)
869{
870 struct work_atom *atom;
871
872 BUG_ON(list_empty(&atoms->work_list));
873
874 atom = list_entry(atoms->work_list.prev, struct work_atom, list);
875
876 atom->runtime += delta;
877 atoms->total_runtime += delta;
878}
879
880static void
881add_sched_in_event(struct work_atoms *atoms, u64 timestamp)
882{
883 struct work_atom *atom;
884 u64 delta;
885
886 if (list_empty(&atoms->work_list))
887 return;
888
889 atom = list_entry(atoms->work_list.prev, struct work_atom, list);
890
891 if (atom->state != THREAD_WAIT_CPU)
892 return;
893
894 if (timestamp < atom->wake_up_time) {
895 atom->state = THREAD_IGNORE;
896 return;
897 }
898
899 atom->state = THREAD_SCHED_IN;
900 atom->sched_in_time = timestamp;
901
902 delta = atom->sched_in_time - atom->wake_up_time;
903 atoms->total_lat += delta;
904 if (delta > atoms->max_lat) {
905 atoms->max_lat = delta;
906 atoms->max_lat_at = timestamp;
907 }
908 atoms->nb_atoms++;
909}
910
911static int latency_switch_event(struct perf_sched *sched,
912 struct perf_evsel *evsel,
913 struct perf_sample *sample,
914 struct machine *machine)
915{
916 const u32 prev_pid = perf_evsel__intval(evsel, sample, "prev_pid"),
917 next_pid = perf_evsel__intval(evsel, sample, "next_pid");
918 const u64 prev_state = perf_evsel__intval(evsel, sample, "prev_state");
919 struct work_atoms *out_events, *in_events;
920 struct thread *sched_out, *sched_in;
921 u64 timestamp0, timestamp = sample->time;
922 int cpu = sample->cpu;
923 s64 delta;
924
925 BUG_ON(cpu >= MAX_CPUS || cpu < 0);
926
927 timestamp0 = sched->cpu_last_switched[cpu];
928 sched->cpu_last_switched[cpu] = timestamp;
929 if (timestamp0)
930 delta = timestamp - timestamp0;
931 else
932 delta = 0;
933
934 if (delta < 0) {
935 pr_err("hm, delta: %" PRIu64 " < 0 ?\n", delta);
936 return -1;
937 }
938
939 sched_out = machine__findnew_thread(machine, 0, prev_pid);
940 sched_in = machine__findnew_thread(machine, 0, next_pid);
941
942 out_events = thread_atoms_search(&sched->atom_root, sched_out, &sched->cmp_pid);
943 if (!out_events) {
944 if (thread_atoms_insert(sched, sched_out))
945 return -1;
946 out_events = thread_atoms_search(&sched->atom_root, sched_out, &sched->cmp_pid);
947 if (!out_events) {
948 pr_err("out-event: Internal tree error");
949 return -1;
950 }
951 }
952 if (add_sched_out_event(out_events, sched_out_state(prev_state), timestamp))
953 return -1;
954
955 in_events = thread_atoms_search(&sched->atom_root, sched_in, &sched->cmp_pid);
956 if (!in_events) {
957 if (thread_atoms_insert(sched, sched_in))
958 return -1;
959 in_events = thread_atoms_search(&sched->atom_root, sched_in, &sched->cmp_pid);
960 if (!in_events) {
961 pr_err("in-event: Internal tree error");
962 return -1;
963 }
964 /*
965 * Take came in we have not heard about yet,
966 * add in an initial atom in runnable state:
967 */
968 if (add_sched_out_event(in_events, 'R', timestamp))
969 return -1;
970 }
971 add_sched_in_event(in_events, timestamp);
972
973 return 0;
974}
975
976static int latency_runtime_event(struct perf_sched *sched,
977 struct perf_evsel *evsel,
978 struct perf_sample *sample,
979 struct machine *machine)
980{
981 const u32 pid = perf_evsel__intval(evsel, sample, "pid");
982 const u64 runtime = perf_evsel__intval(evsel, sample, "runtime");
983 struct thread *thread = machine__findnew_thread(machine, 0, pid);
984 struct work_atoms *atoms = thread_atoms_search(&sched->atom_root, thread, &sched->cmp_pid);
985 u64 timestamp = sample->time;
986 int cpu = sample->cpu;
987
988 BUG_ON(cpu >= MAX_CPUS || cpu < 0);
989 if (!atoms) {
990 if (thread_atoms_insert(sched, thread))
991 return -1;
992 atoms = thread_atoms_search(&sched->atom_root, thread, &sched->cmp_pid);
993 if (!atoms) {
994 pr_err("in-event: Internal tree error");
995 return -1;
996 }
997 if (add_sched_out_event(atoms, 'R', timestamp))
998 return -1;
999 }
1000
1001 add_runtime_event(atoms, runtime, timestamp);
1002 return 0;
1003}
1004
1005static int latency_wakeup_event(struct perf_sched *sched,
1006 struct perf_evsel *evsel,
1007 struct perf_sample *sample,
1008 struct machine *machine)
1009{
1010 const u32 pid = perf_evsel__intval(evsel, sample, "pid"),
1011 success = perf_evsel__intval(evsel, sample, "success");
1012 struct work_atoms *atoms;
1013 struct work_atom *atom;
1014 struct thread *wakee;
1015 u64 timestamp = sample->time;
1016
1017 /* Note for later, it may be interesting to observe the failing cases */
1018 if (!success)
1019 return 0;
1020
1021 wakee = machine__findnew_thread(machine, 0, pid);
1022 atoms = thread_atoms_search(&sched->atom_root, wakee, &sched->cmp_pid);
1023 if (!atoms) {
1024 if (thread_atoms_insert(sched, wakee))
1025 return -1;
1026 atoms = thread_atoms_search(&sched->atom_root, wakee, &sched->cmp_pid);
1027 if (!atoms) {
1028 pr_err("wakeup-event: Internal tree error");
1029 return -1;
1030 }
1031 if (add_sched_out_event(atoms, 'S', timestamp))
1032 return -1;
1033 }
1034
1035 BUG_ON(list_empty(&atoms->work_list));
1036
1037 atom = list_entry(atoms->work_list.prev, struct work_atom, list);
1038
1039 /*
1040 * You WILL be missing events if you've recorded only
1041 * one CPU, or are only looking at only one, so don't
1042 * make useless noise.
1043 */
1044 if (sched->profile_cpu == -1 && atom->state != THREAD_SLEEPING)
1045 sched->nr_state_machine_bugs++;
1046
1047 sched->nr_timestamps++;
1048 if (atom->sched_out_time > timestamp) {
1049 sched->nr_unordered_timestamps++;
1050 return 0;
1051 }
1052
1053 atom->state = THREAD_WAIT_CPU;
1054 atom->wake_up_time = timestamp;
1055 return 0;
1056}
1057
1058static int latency_migrate_task_event(struct perf_sched *sched,
1059 struct perf_evsel *evsel,
1060 struct perf_sample *sample,
1061 struct machine *machine)
1062{
1063 const u32 pid = perf_evsel__intval(evsel, sample, "pid");
1064 u64 timestamp = sample->time;
1065 struct work_atoms *atoms;
1066 struct work_atom *atom;
1067 struct thread *migrant;
1068
1069 /*
1070 * Only need to worry about migration when profiling one CPU.
1071 */
1072 if (sched->profile_cpu == -1)
1073 return 0;
1074
1075 migrant = machine__findnew_thread(machine, 0, pid);
1076 atoms = thread_atoms_search(&sched->atom_root, migrant, &sched->cmp_pid);
1077 if (!atoms) {
1078 if (thread_atoms_insert(sched, migrant))
1079 return -1;
1080 register_pid(sched, migrant->tid, thread__comm_str(migrant));
1081 atoms = thread_atoms_search(&sched->atom_root, migrant, &sched->cmp_pid);
1082 if (!atoms) {
1083 pr_err("migration-event: Internal tree error");
1084 return -1;
1085 }
1086 if (add_sched_out_event(atoms, 'R', timestamp))
1087 return -1;
1088 }
1089
1090 BUG_ON(list_empty(&atoms->work_list));
1091
1092 atom = list_entry(atoms->work_list.prev, struct work_atom, list);
1093 atom->sched_in_time = atom->sched_out_time = atom->wake_up_time = timestamp;
1094
1095 sched->nr_timestamps++;
1096
1097 if (atom->sched_out_time > timestamp)
1098 sched->nr_unordered_timestamps++;
1099
1100 return 0;
1101}
1102
1103static void output_lat_thread(struct perf_sched *sched, struct work_atoms *work_list)
1104{
1105 int i;
1106 int ret;
1107 u64 avg;
1108
1109 if (!work_list->nb_atoms)
1110 return;
1111 /*
1112 * Ignore idle threads:
1113 */
1114 if (!strcmp(thread__comm_str(work_list->thread), "swapper"))
1115 return;
1116
1117 sched->all_runtime += work_list->total_runtime;
1118 sched->all_count += work_list->nb_atoms;
1119
1120 ret = printf(" %s:%d ", thread__comm_str(work_list->thread), work_list->thread->tid);
1121
1122 for (i = 0; i < 24 - ret; i++)
1123 printf(" ");
1124
1125 avg = work_list->total_lat / work_list->nb_atoms;
1126
1127 printf("|%11.3f ms |%9" PRIu64 " | avg:%9.3f ms | max:%9.3f ms | max at: %13.6f s\n",
1128 (double)work_list->total_runtime / 1e6,
1129 work_list->nb_atoms, (double)avg / 1e6,
1130 (double)work_list->max_lat / 1e6,
1131 (double)work_list->max_lat_at / 1e9);
1132}
1133
1134static int pid_cmp(struct work_atoms *l, struct work_atoms *r)
1135{
1136 if (l->thread->tid < r->thread->tid)
1137 return -1;
1138 if (l->thread->tid > r->thread->tid)
1139 return 1;
1140
1141 return 0;
1142}
1143
1144static int avg_cmp(struct work_atoms *l, struct work_atoms *r)
1145{
1146 u64 avgl, avgr;
1147
1148 if (!l->nb_atoms)
1149 return -1;
1150
1151 if (!r->nb_atoms)
1152 return 1;
1153
1154 avgl = l->total_lat / l->nb_atoms;
1155 avgr = r->total_lat / r->nb_atoms;
1156
1157 if (avgl < avgr)
1158 return -1;
1159 if (avgl > avgr)
1160 return 1;
1161
1162 return 0;
1163}
1164
1165static int max_cmp(struct work_atoms *l, struct work_atoms *r)
1166{
1167 if (l->max_lat < r->max_lat)
1168 return -1;
1169 if (l->max_lat > r->max_lat)
1170 return 1;
1171
1172 return 0;
1173}
1174
1175static int switch_cmp(struct work_atoms *l, struct work_atoms *r)
1176{
1177 if (l->nb_atoms < r->nb_atoms)
1178 return -1;
1179 if (l->nb_atoms > r->nb_atoms)
1180 return 1;
1181
1182 return 0;
1183}
1184
1185static int runtime_cmp(struct work_atoms *l, struct work_atoms *r)
1186{
1187 if (l->total_runtime < r->total_runtime)
1188 return -1;
1189 if (l->total_runtime > r->total_runtime)
1190 return 1;
1191
1192 return 0;
1193}
1194
1195static int sort_dimension__add(const char *tok, struct list_head *list)
1196{
1197 size_t i;
1198 static struct sort_dimension avg_sort_dimension = {
1199 .name = "avg",
1200 .cmp = avg_cmp,
1201 };
1202 static struct sort_dimension max_sort_dimension = {
1203 .name = "max",
1204 .cmp = max_cmp,
1205 };
1206 static struct sort_dimension pid_sort_dimension = {
1207 .name = "pid",
1208 .cmp = pid_cmp,
1209 };
1210 static struct sort_dimension runtime_sort_dimension = {
1211 .name = "runtime",
1212 .cmp = runtime_cmp,
1213 };
1214 static struct sort_dimension switch_sort_dimension = {
1215 .name = "switch",
1216 .cmp = switch_cmp,
1217 };
1218 struct sort_dimension *available_sorts[] = {
1219 &pid_sort_dimension,
1220 &avg_sort_dimension,
1221 &max_sort_dimension,
1222 &switch_sort_dimension,
1223 &runtime_sort_dimension,
1224 };
1225
1226 for (i = 0; i < ARRAY_SIZE(available_sorts); i++) {
1227 if (!strcmp(available_sorts[i]->name, tok)) {
1228 list_add_tail(&available_sorts[i]->list, list);
1229
1230 return 0;
1231 }
1232 }
1233
1234 return -1;
1235}
1236
1237static void perf_sched__sort_lat(struct perf_sched *sched)
1238{
1239 struct rb_node *node;
1240
1241 for (;;) {
1242 struct work_atoms *data;
1243 node = rb_first(&sched->atom_root);
1244 if (!node)
1245 break;
1246
1247 rb_erase(node, &sched->atom_root);
1248 data = rb_entry(node, struct work_atoms, node);
1249 __thread_latency_insert(&sched->sorted_atom_root, data, &sched->sort_list);
1250 }
1251}
1252
1253static int process_sched_wakeup_event(struct perf_tool *tool,
1254 struct perf_evsel *evsel,
1255 struct perf_sample *sample,
1256 struct machine *machine)
1257{
1258 struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
1259
1260 if (sched->tp_handler->wakeup_event)
1261 return sched->tp_handler->wakeup_event(sched, evsel, sample, machine);
1262
1263 return 0;
1264}
1265
1266static int map_switch_event(struct perf_sched *sched, struct perf_evsel *evsel,
1267 struct perf_sample *sample, struct machine *machine)
1268{
1269 const u32 prev_pid = perf_evsel__intval(evsel, sample, "prev_pid"),
1270 next_pid = perf_evsel__intval(evsel, sample, "next_pid");
1271 struct thread *sched_out __maybe_unused, *sched_in;
1272 int new_shortname;
1273 u64 timestamp0, timestamp = sample->time;
1274 s64 delta;
1275 int cpu, this_cpu = sample->cpu;
1276
1277 BUG_ON(this_cpu >= MAX_CPUS || this_cpu < 0);
1278
1279 if (this_cpu > sched->max_cpu)
1280 sched->max_cpu = this_cpu;
1281
1282 timestamp0 = sched->cpu_last_switched[this_cpu];
1283 sched->cpu_last_switched[this_cpu] = timestamp;
1284 if (timestamp0)
1285 delta = timestamp - timestamp0;
1286 else
1287 delta = 0;
1288
1289 if (delta < 0) {
1290 pr_err("hm, delta: %" PRIu64 " < 0 ?\n", delta);
1291 return -1;
1292 }
1293
1294 sched_out = machine__findnew_thread(machine, 0, prev_pid);
1295 sched_in = machine__findnew_thread(machine, 0, next_pid);
1296
1297 sched->curr_thread[this_cpu] = sched_in;
1298
1299 printf(" ");
1300
1301 new_shortname = 0;
1302 if (!sched_in->shortname[0]) {
1303 sched_in->shortname[0] = sched->next_shortname1;
1304 sched_in->shortname[1] = sched->next_shortname2;
1305
1306 if (sched->next_shortname1 < 'Z') {
1307 sched->next_shortname1++;
1308 } else {
1309 sched->next_shortname1='A';
1310 if (sched->next_shortname2 < '9') {
1311 sched->next_shortname2++;
1312 } else {
1313 sched->next_shortname2='0';
1314 }
1315 }
1316 new_shortname = 1;
1317 }
1318
1319 for (cpu = 0; cpu <= sched->max_cpu; cpu++) {
1320 if (cpu != this_cpu)
1321 printf(" ");
1322 else
1323 printf("*");
1324
1325 if (sched->curr_thread[cpu]) {
1326 if (sched->curr_thread[cpu]->tid)
1327 printf("%2s ", sched->curr_thread[cpu]->shortname);
1328 else
1329 printf(". ");
1330 } else
1331 printf(" ");
1332 }
1333
1334 printf(" %12.6f secs ", (double)timestamp/1e9);
1335 if (new_shortname) {
1336 printf("%s => %s:%d\n",
1337 sched_in->shortname, thread__comm_str(sched_in), sched_in->tid);
1338 } else {
1339 printf("\n");
1340 }
1341
1342 return 0;
1343}
1344
1345static int process_sched_switch_event(struct perf_tool *tool,
1346 struct perf_evsel *evsel,
1347 struct perf_sample *sample,
1348 struct machine *machine)
1349{
1350 struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
1351 int this_cpu = sample->cpu, err = 0;
1352 u32 prev_pid = perf_evsel__intval(evsel, sample, "prev_pid"),
1353 next_pid = perf_evsel__intval(evsel, sample, "next_pid");
1354
1355 if (sched->curr_pid[this_cpu] != (u32)-1) {
1356 /*
1357 * Are we trying to switch away a PID that is
1358 * not current?
1359 */
1360 if (sched->curr_pid[this_cpu] != prev_pid)
1361 sched->nr_context_switch_bugs++;
1362 }
1363
1364 if (sched->tp_handler->switch_event)
1365 err = sched->tp_handler->switch_event(sched, evsel, sample, machine);
1366
1367 sched->curr_pid[this_cpu] = next_pid;
1368 return err;
1369}
1370
1371static int process_sched_runtime_event(struct perf_tool *tool,
1372 struct perf_evsel *evsel,
1373 struct perf_sample *sample,
1374 struct machine *machine)
1375{
1376 struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
1377
1378 if (sched->tp_handler->runtime_event)
1379 return sched->tp_handler->runtime_event(sched, evsel, sample, machine);
1380
1381 return 0;
1382}
1383
1384static int perf_sched__process_fork_event(struct perf_tool *tool,
1385 union perf_event *event,
1386 struct perf_sample *sample,
1387 struct machine *machine)
1388{
1389 struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
1390
1391 /* run the fork event through the perf machineruy */
1392 perf_event__process_fork(tool, event, sample, machine);
1393
1394 /* and then run additional processing needed for this command */
1395 if (sched->tp_handler->fork_event)
1396 return sched->tp_handler->fork_event(sched, event, machine);
1397
1398 return 0;
1399}
1400
1401static int process_sched_migrate_task_event(struct perf_tool *tool,
1402 struct perf_evsel *evsel,
1403 struct perf_sample *sample,
1404 struct machine *machine)
1405{
1406 struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
1407
1408 if (sched->tp_handler->migrate_task_event)
1409 return sched->tp_handler->migrate_task_event(sched, evsel, sample, machine);
1410
1411 return 0;
1412}
1413
1414typedef int (*tracepoint_handler)(struct perf_tool *tool,
1415 struct perf_evsel *evsel,
1416 struct perf_sample *sample,
1417 struct machine *machine);
1418
1419static int perf_sched__process_tracepoint_sample(struct perf_tool *tool __maybe_unused,
1420 union perf_event *event __maybe_unused,
1421 struct perf_sample *sample,
1422 struct perf_evsel *evsel,
1423 struct machine *machine)
1424{
1425 int err = 0;
1426
1427 evsel->hists.stats.total_period += sample->period;
1428 hists__inc_nr_events(&evsel->hists, PERF_RECORD_SAMPLE);
1429
1430 if (evsel->handler != NULL) {
1431 tracepoint_handler f = evsel->handler;
1432 err = f(tool, evsel, sample, machine);
1433 }
1434
1435 return err;
1436}
1437
1438static int perf_sched__read_events(struct perf_sched *sched,
1439 struct perf_session **psession)
1440{
1441 const struct perf_evsel_str_handler handlers[] = {
1442 { "sched:sched_switch", process_sched_switch_event, },
1443 { "sched:sched_stat_runtime", process_sched_runtime_event, },
1444 { "sched:sched_wakeup", process_sched_wakeup_event, },
1445 { "sched:sched_wakeup_new", process_sched_wakeup_event, },
1446 { "sched:sched_migrate_task", process_sched_migrate_task_event, },
1447 };
1448 struct perf_session *session;
1449 struct perf_data_file file = {
1450 .path = input_name,
1451 .mode = PERF_DATA_MODE_READ,
1452 };
1453
1454 session = perf_session__new(&file, false, &sched->tool);
1455 if (session == NULL) {
1456 pr_debug("No Memory for session\n");
1457 return -1;
1458 }
1459
1460 if (perf_session__set_tracepoints_handlers(session, handlers))
1461 goto out_delete;
1462
1463 if (perf_session__has_traces(session, "record -R")) {
1464 int err = perf_session__process_events(session, &sched->tool);
1465 if (err) {
1466 pr_err("Failed to process events, error %d", err);
1467 goto out_delete;
1468 }
1469
1470 sched->nr_events = session->stats.nr_events[0];
1471 sched->nr_lost_events = session->stats.total_lost;
1472 sched->nr_lost_chunks = session->stats.nr_events[PERF_RECORD_LOST];
1473 }
1474
1475 if (psession)
1476 *psession = session;
1477 else
1478 perf_session__delete(session);
1479
1480 return 0;
1481
1482out_delete:
1483 perf_session__delete(session);
1484 return -1;
1485}
1486
1487static void print_bad_events(struct perf_sched *sched)
1488{
1489 if (sched->nr_unordered_timestamps && sched->nr_timestamps) {
1490 printf(" INFO: %.3f%% unordered timestamps (%ld out of %ld)\n",
1491 (double)sched->nr_unordered_timestamps/(double)sched->nr_timestamps*100.0,
1492 sched->nr_unordered_timestamps, sched->nr_timestamps);
1493 }
1494 if (sched->nr_lost_events && sched->nr_events) {
1495 printf(" INFO: %.3f%% lost events (%ld out of %ld, in %ld chunks)\n",
1496 (double)sched->nr_lost_events/(double)sched->nr_events * 100.0,
1497 sched->nr_lost_events, sched->nr_events, sched->nr_lost_chunks);
1498 }
1499 if (sched->nr_state_machine_bugs && sched->nr_timestamps) {
1500 printf(" INFO: %.3f%% state machine bugs (%ld out of %ld)",
1501 (double)sched->nr_state_machine_bugs/(double)sched->nr_timestamps*100.0,
1502 sched->nr_state_machine_bugs, sched->nr_timestamps);
1503 if (sched->nr_lost_events)
1504 printf(" (due to lost events?)");
1505 printf("\n");
1506 }
1507 if (sched->nr_context_switch_bugs && sched->nr_timestamps) {
1508 printf(" INFO: %.3f%% context switch bugs (%ld out of %ld)",
1509 (double)sched->nr_context_switch_bugs/(double)sched->nr_timestamps*100.0,
1510 sched->nr_context_switch_bugs, sched->nr_timestamps);
1511 if (sched->nr_lost_events)
1512 printf(" (due to lost events?)");
1513 printf("\n");
1514 }
1515}
1516
1517static int perf_sched__lat(struct perf_sched *sched)
1518{
1519 struct rb_node *next;
1520 struct perf_session *session;
1521
1522 setup_pager();
1523
1524 /* save session -- references to threads are held in work_list */
1525 if (perf_sched__read_events(sched, &session))
1526 return -1;
1527
1528 perf_sched__sort_lat(sched);
1529
1530 printf("\n -----------------------------------------------------------------------------------------------------------------\n");
1531 printf(" Task | Runtime ms | Switches | Average delay ms | Maximum delay ms | Maximum delay at |\n");
1532 printf(" -----------------------------------------------------------------------------------------------------------------\n");
1533
1534 next = rb_first(&sched->sorted_atom_root);
1535
1536 while (next) {
1537 struct work_atoms *work_list;
1538
1539 work_list = rb_entry(next, struct work_atoms, node);
1540 output_lat_thread(sched, work_list);
1541 next = rb_next(next);
1542 }
1543
1544 printf(" -----------------------------------------------------------------------------------------------------------------\n");
1545 printf(" TOTAL: |%11.3f ms |%9" PRIu64 " |\n",
1546 (double)sched->all_runtime / 1e6, sched->all_count);
1547
1548 printf(" ---------------------------------------------------\n");
1549
1550 print_bad_events(sched);
1551 printf("\n");
1552
1553 perf_session__delete(session);
1554 return 0;
1555}
1556
1557static int perf_sched__map(struct perf_sched *sched)
1558{
1559 sched->max_cpu = sysconf(_SC_NPROCESSORS_CONF);
1560
1561 setup_pager();
1562 if (perf_sched__read_events(sched, NULL))
1563 return -1;
1564 print_bad_events(sched);
1565 return 0;
1566}
1567
1568static int perf_sched__replay(struct perf_sched *sched)
1569{
1570 unsigned long i;
1571
1572 calibrate_run_measurement_overhead(sched);
1573 calibrate_sleep_measurement_overhead(sched);
1574
1575 test_calibrations(sched);
1576
1577 if (perf_sched__read_events(sched, NULL))
1578 return -1;
1579
1580 printf("nr_run_events: %ld\n", sched->nr_run_events);
1581 printf("nr_sleep_events: %ld\n", sched->nr_sleep_events);
1582 printf("nr_wakeup_events: %ld\n", sched->nr_wakeup_events);
1583
1584 if (sched->targetless_wakeups)
1585 printf("target-less wakeups: %ld\n", sched->targetless_wakeups);
1586 if (sched->multitarget_wakeups)
1587 printf("multi-target wakeups: %ld\n", sched->multitarget_wakeups);
1588 if (sched->nr_run_events_optimized)
1589 printf("run atoms optimized: %ld\n",
1590 sched->nr_run_events_optimized);
1591
1592 print_task_traces(sched);
1593 add_cross_task_wakeups(sched);
1594
1595 create_tasks(sched);
1596 printf("------------------------------------------------------------\n");
1597 for (i = 0; i < sched->replay_repeat; i++)
1598 run_one_test(sched);
1599
1600 return 0;
1601}
1602
1603static void setup_sorting(struct perf_sched *sched, const struct option *options,
1604 const char * const usage_msg[])
1605{
1606 char *tmp, *tok, *str = strdup(sched->sort_order);
1607
1608 for (tok = strtok_r(str, ", ", &tmp);
1609 tok; tok = strtok_r(NULL, ", ", &tmp)) {
1610 if (sort_dimension__add(tok, &sched->sort_list) < 0) {
1611 error("Unknown --sort key: `%s'", tok);
1612 usage_with_options(usage_msg, options);
1613 }
1614 }
1615
1616 free(str);
1617
1618 sort_dimension__add("pid", &sched->cmp_pid);
1619}
1620
1621static int __cmd_record(int argc, const char **argv)
1622{
1623 unsigned int rec_argc, i, j;
1624 const char **rec_argv;
1625 const char * const record_args[] = {
1626 "record",
1627 "-a",
1628 "-R",
1629 "-m", "1024",
1630 "-c", "1",
1631 "-e", "sched:sched_switch",
1632 "-e", "sched:sched_stat_wait",
1633 "-e", "sched:sched_stat_sleep",
1634 "-e", "sched:sched_stat_iowait",
1635 "-e", "sched:sched_stat_runtime",
1636 "-e", "sched:sched_process_fork",
1637 "-e", "sched:sched_wakeup",
1638 "-e", "sched:sched_migrate_task",
1639 };
1640
1641 rec_argc = ARRAY_SIZE(record_args) + argc - 1;
1642 rec_argv = calloc(rec_argc + 1, sizeof(char *));
1643
1644 if (rec_argv == NULL)
1645 return -ENOMEM;
1646
1647 for (i = 0; i < ARRAY_SIZE(record_args); i++)
1648 rec_argv[i] = strdup(record_args[i]);
1649
1650 for (j = 1; j < (unsigned int)argc; j++, i++)
1651 rec_argv[i] = argv[j];
1652
1653 BUG_ON(i != rec_argc);
1654
1655 return cmd_record(i, rec_argv, NULL);
1656}
1657
1658int cmd_sched(int argc, const char **argv, const char *prefix __maybe_unused)
1659{
1660 const char default_sort_order[] = "avg, max, switch, runtime";
1661 struct perf_sched sched = {
1662 .tool = {
1663 .sample = perf_sched__process_tracepoint_sample,
1664 .comm = perf_event__process_comm,
1665 .lost = perf_event__process_lost,
1666 .fork = perf_sched__process_fork_event,
1667 .ordered_samples = true,
1668 },
1669 .cmp_pid = LIST_HEAD_INIT(sched.cmp_pid),
1670 .sort_list = LIST_HEAD_INIT(sched.sort_list),
1671 .start_work_mutex = PTHREAD_MUTEX_INITIALIZER,
1672 .work_done_wait_mutex = PTHREAD_MUTEX_INITIALIZER,
1673 .sort_order = default_sort_order,
1674 .replay_repeat = 10,
1675 .profile_cpu = -1,
1676 .next_shortname1 = 'A',
1677 .next_shortname2 = '0',
1678 };
1679 const struct option latency_options[] = {
1680 OPT_STRING('s', "sort", &sched.sort_order, "key[,key2...]",
1681 "sort by key(s): runtime, switch, avg, max"),
1682 OPT_INCR('v', "verbose", &verbose,
1683 "be more verbose (show symbol address, etc)"),
1684 OPT_INTEGER('C', "CPU", &sched.profile_cpu,
1685 "CPU to profile on"),
1686 OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
1687 "dump raw trace in ASCII"),
1688 OPT_END()
1689 };
1690 const struct option replay_options[] = {
1691 OPT_UINTEGER('r', "repeat", &sched.replay_repeat,
1692 "repeat the workload replay N times (-1: infinite)"),
1693 OPT_INCR('v', "verbose", &verbose,
1694 "be more verbose (show symbol address, etc)"),
1695 OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
1696 "dump raw trace in ASCII"),
1697 OPT_END()
1698 };
1699 const struct option sched_options[] = {
1700 OPT_STRING('i', "input", &input_name, "file",
1701 "input file name"),
1702 OPT_INCR('v', "verbose", &verbose,
1703 "be more verbose (show symbol address, etc)"),
1704 OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
1705 "dump raw trace in ASCII"),
1706 OPT_END()
1707 };
1708 const char * const latency_usage[] = {
1709 "perf sched latency [<options>]",
1710 NULL
1711 };
1712 const char * const replay_usage[] = {
1713 "perf sched replay [<options>]",
1714 NULL
1715 };
1716 const char * const sched_usage[] = {
1717 "perf sched [<options>] {record|latency|map|replay|script}",
1718 NULL
1719 };
1720 struct trace_sched_handler lat_ops = {
1721 .wakeup_event = latency_wakeup_event,
1722 .switch_event = latency_switch_event,
1723 .runtime_event = latency_runtime_event,
1724 .migrate_task_event = latency_migrate_task_event,
1725 };
1726 struct trace_sched_handler map_ops = {
1727 .switch_event = map_switch_event,
1728 };
1729 struct trace_sched_handler replay_ops = {
1730 .wakeup_event = replay_wakeup_event,
1731 .switch_event = replay_switch_event,
1732 .fork_event = replay_fork_event,
1733 };
1734 unsigned int i;
1735
1736 for (i = 0; i < ARRAY_SIZE(sched.curr_pid); i++)
1737 sched.curr_pid[i] = -1;
1738
1739 argc = parse_options(argc, argv, sched_options, sched_usage,
1740 PARSE_OPT_STOP_AT_NON_OPTION);
1741 if (!argc)
1742 usage_with_options(sched_usage, sched_options);
1743
1744 /*
1745 * Aliased to 'perf script' for now:
1746 */
1747 if (!strcmp(argv[0], "script"))
1748 return cmd_script(argc, argv, prefix);
1749
1750 symbol__init();
1751 if (!strncmp(argv[0], "rec", 3)) {
1752 return __cmd_record(argc, argv);
1753 } else if (!strncmp(argv[0], "lat", 3)) {
1754 sched.tp_handler = &lat_ops;
1755 if (argc > 1) {
1756 argc = parse_options(argc, argv, latency_options, latency_usage, 0);
1757 if (argc)
1758 usage_with_options(latency_usage, latency_options);
1759 }
1760 setup_sorting(&sched, latency_options, latency_usage);
1761 return perf_sched__lat(&sched);
1762 } else if (!strcmp(argv[0], "map")) {
1763 sched.tp_handler = &map_ops;
1764 setup_sorting(&sched, latency_options, latency_usage);
1765 return perf_sched__map(&sched);
1766 } else if (!strncmp(argv[0], "rep", 3)) {
1767 sched.tp_handler = &replay_ops;
1768 if (argc) {
1769 argc = parse_options(argc, argv, replay_options, replay_usage, 0);
1770 if (argc)
1771 usage_with_options(replay_usage, replay_options);
1772 }
1773 return perf_sched__replay(&sched);
1774 } else {
1775 usage_with_options(sched_usage, sched_options);
1776 }
1777
1778 return 0;
1779}
1// SPDX-License-Identifier: GPL-2.0
2#include "builtin.h"
3#include "perf.h"
4#include "perf-sys.h"
5
6#include "util/cpumap.h"
7#include "util/evlist.h"
8#include "util/evsel.h"
9#include "util/evsel_fprintf.h"
10#include "util/symbol.h"
11#include "util/thread.h"
12#include "util/header.h"
13#include "util/session.h"
14#include "util/tool.h"
15#include "util/cloexec.h"
16#include "util/thread_map.h"
17#include "util/color.h"
18#include "util/stat.h"
19#include "util/string2.h"
20#include "util/callchain.h"
21#include "util/time-utils.h"
22
23#include <subcmd/pager.h>
24#include <subcmd/parse-options.h>
25#include "util/trace-event.h"
26
27#include "util/debug.h"
28#include "util/event.h"
29
30#include <linux/kernel.h>
31#include <linux/log2.h>
32#include <linux/zalloc.h>
33#include <sys/prctl.h>
34#include <sys/resource.h>
35#include <inttypes.h>
36
37#include <errno.h>
38#include <semaphore.h>
39#include <pthread.h>
40#include <math.h>
41#include <api/fs/fs.h>
42#include <perf/cpumap.h>
43#include <linux/time64.h>
44#include <linux/err.h>
45
46#include <linux/ctype.h>
47
48#define PR_SET_NAME 15 /* Set process name */
49#define MAX_CPUS 4096
50#define COMM_LEN 20
51#define SYM_LEN 129
52#define MAX_PID 1024000
53
54static const char *cpu_list;
55static DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS);
56
57struct sched_atom;
58
59struct task_desc {
60 unsigned long nr;
61 unsigned long pid;
62 char comm[COMM_LEN];
63
64 unsigned long nr_events;
65 unsigned long curr_event;
66 struct sched_atom **atoms;
67
68 pthread_t thread;
69 sem_t sleep_sem;
70
71 sem_t ready_for_work;
72 sem_t work_done_sem;
73
74 u64 cpu_usage;
75};
76
77enum sched_event_type {
78 SCHED_EVENT_RUN,
79 SCHED_EVENT_SLEEP,
80 SCHED_EVENT_WAKEUP,
81 SCHED_EVENT_MIGRATION,
82};
83
84struct sched_atom {
85 enum sched_event_type type;
86 int specific_wait;
87 u64 timestamp;
88 u64 duration;
89 unsigned long nr;
90 sem_t *wait_sem;
91 struct task_desc *wakee;
92};
93
94#define TASK_STATE_TO_CHAR_STR "RSDTtZXxKWP"
95
96/* task state bitmask, copied from include/linux/sched.h */
97#define TASK_RUNNING 0
98#define TASK_INTERRUPTIBLE 1
99#define TASK_UNINTERRUPTIBLE 2
100#define __TASK_STOPPED 4
101#define __TASK_TRACED 8
102/* in tsk->exit_state */
103#define EXIT_DEAD 16
104#define EXIT_ZOMBIE 32
105#define EXIT_TRACE (EXIT_ZOMBIE | EXIT_DEAD)
106/* in tsk->state again */
107#define TASK_DEAD 64
108#define TASK_WAKEKILL 128
109#define TASK_WAKING 256
110#define TASK_PARKED 512
111
112enum thread_state {
113 THREAD_SLEEPING = 0,
114 THREAD_WAIT_CPU,
115 THREAD_SCHED_IN,
116 THREAD_IGNORE
117};
118
119struct work_atom {
120 struct list_head list;
121 enum thread_state state;
122 u64 sched_out_time;
123 u64 wake_up_time;
124 u64 sched_in_time;
125 u64 runtime;
126};
127
128struct work_atoms {
129 struct list_head work_list;
130 struct thread *thread;
131 struct rb_node node;
132 u64 max_lat;
133 u64 max_lat_start;
134 u64 max_lat_end;
135 u64 total_lat;
136 u64 nb_atoms;
137 u64 total_runtime;
138 int num_merged;
139};
140
141typedef int (*sort_fn_t)(struct work_atoms *, struct work_atoms *);
142
143struct perf_sched;
144
145struct trace_sched_handler {
146 int (*switch_event)(struct perf_sched *sched, struct evsel *evsel,
147 struct perf_sample *sample, struct machine *machine);
148
149 int (*runtime_event)(struct perf_sched *sched, struct evsel *evsel,
150 struct perf_sample *sample, struct machine *machine);
151
152 int (*wakeup_event)(struct perf_sched *sched, struct evsel *evsel,
153 struct perf_sample *sample, struct machine *machine);
154
155 /* PERF_RECORD_FORK event, not sched_process_fork tracepoint */
156 int (*fork_event)(struct perf_sched *sched, union perf_event *event,
157 struct machine *machine);
158
159 int (*migrate_task_event)(struct perf_sched *sched,
160 struct evsel *evsel,
161 struct perf_sample *sample,
162 struct machine *machine);
163};
164
165#define COLOR_PIDS PERF_COLOR_BLUE
166#define COLOR_CPUS PERF_COLOR_BG_RED
167
168struct perf_sched_map {
169 DECLARE_BITMAP(comp_cpus_mask, MAX_CPUS);
170 int *comp_cpus;
171 bool comp;
172 struct perf_thread_map *color_pids;
173 const char *color_pids_str;
174 struct perf_cpu_map *color_cpus;
175 const char *color_cpus_str;
176 struct perf_cpu_map *cpus;
177 const char *cpus_str;
178};
179
180struct perf_sched {
181 struct perf_tool tool;
182 const char *sort_order;
183 unsigned long nr_tasks;
184 struct task_desc **pid_to_task;
185 struct task_desc **tasks;
186 const struct trace_sched_handler *tp_handler;
187 pthread_mutex_t start_work_mutex;
188 pthread_mutex_t work_done_wait_mutex;
189 int profile_cpu;
190/*
191 * Track the current task - that way we can know whether there's any
192 * weird events, such as a task being switched away that is not current.
193 */
194 int max_cpu;
195 u32 curr_pid[MAX_CPUS];
196 struct thread *curr_thread[MAX_CPUS];
197 char next_shortname1;
198 char next_shortname2;
199 unsigned int replay_repeat;
200 unsigned long nr_run_events;
201 unsigned long nr_sleep_events;
202 unsigned long nr_wakeup_events;
203 unsigned long nr_sleep_corrections;
204 unsigned long nr_run_events_optimized;
205 unsigned long targetless_wakeups;
206 unsigned long multitarget_wakeups;
207 unsigned long nr_runs;
208 unsigned long nr_timestamps;
209 unsigned long nr_unordered_timestamps;
210 unsigned long nr_context_switch_bugs;
211 unsigned long nr_events;
212 unsigned long nr_lost_chunks;
213 unsigned long nr_lost_events;
214 u64 run_measurement_overhead;
215 u64 sleep_measurement_overhead;
216 u64 start_time;
217 u64 cpu_usage;
218 u64 runavg_cpu_usage;
219 u64 parent_cpu_usage;
220 u64 runavg_parent_cpu_usage;
221 u64 sum_runtime;
222 u64 sum_fluct;
223 u64 run_avg;
224 u64 all_runtime;
225 u64 all_count;
226 u64 cpu_last_switched[MAX_CPUS];
227 struct rb_root_cached atom_root, sorted_atom_root, merged_atom_root;
228 struct list_head sort_list, cmp_pid;
229 bool force;
230 bool skip_merge;
231 struct perf_sched_map map;
232
233 /* options for timehist command */
234 bool summary;
235 bool summary_only;
236 bool idle_hist;
237 bool show_callchain;
238 unsigned int max_stack;
239 bool show_cpu_visual;
240 bool show_wakeups;
241 bool show_next;
242 bool show_migrations;
243 bool show_state;
244 u64 skipped_samples;
245 const char *time_str;
246 struct perf_time_interval ptime;
247 struct perf_time_interval hist_time;
248};
249
250/* per thread run time data */
251struct thread_runtime {
252 u64 last_time; /* time of previous sched in/out event */
253 u64 dt_run; /* run time */
254 u64 dt_sleep; /* time between CPU access by sleep (off cpu) */
255 u64 dt_iowait; /* time between CPU access by iowait (off cpu) */
256 u64 dt_preempt; /* time between CPU access by preempt (off cpu) */
257 u64 dt_delay; /* time between wakeup and sched-in */
258 u64 ready_to_run; /* time of wakeup */
259
260 struct stats run_stats;
261 u64 total_run_time;
262 u64 total_sleep_time;
263 u64 total_iowait_time;
264 u64 total_preempt_time;
265 u64 total_delay_time;
266
267 int last_state;
268
269 char shortname[3];
270 bool comm_changed;
271
272 u64 migrations;
273};
274
275/* per event run time data */
276struct evsel_runtime {
277 u64 *last_time; /* time this event was last seen per cpu */
278 u32 ncpu; /* highest cpu slot allocated */
279};
280
281/* per cpu idle time data */
282struct idle_thread_runtime {
283 struct thread_runtime tr;
284 struct thread *last_thread;
285 struct rb_root_cached sorted_root;
286 struct callchain_root callchain;
287 struct callchain_cursor cursor;
288};
289
290/* track idle times per cpu */
291static struct thread **idle_threads;
292static int idle_max_cpu;
293static char idle_comm[] = "<idle>";
294
295static u64 get_nsecs(void)
296{
297 struct timespec ts;
298
299 clock_gettime(CLOCK_MONOTONIC, &ts);
300
301 return ts.tv_sec * NSEC_PER_SEC + ts.tv_nsec;
302}
303
304static void burn_nsecs(struct perf_sched *sched, u64 nsecs)
305{
306 u64 T0 = get_nsecs(), T1;
307
308 do {
309 T1 = get_nsecs();
310 } while (T1 + sched->run_measurement_overhead < T0 + nsecs);
311}
312
313static void sleep_nsecs(u64 nsecs)
314{
315 struct timespec ts;
316
317 ts.tv_nsec = nsecs % 999999999;
318 ts.tv_sec = nsecs / 999999999;
319
320 nanosleep(&ts, NULL);
321}
322
323static void calibrate_run_measurement_overhead(struct perf_sched *sched)
324{
325 u64 T0, T1, delta, min_delta = NSEC_PER_SEC;
326 int i;
327
328 for (i = 0; i < 10; i++) {
329 T0 = get_nsecs();
330 burn_nsecs(sched, 0);
331 T1 = get_nsecs();
332 delta = T1-T0;
333 min_delta = min(min_delta, delta);
334 }
335 sched->run_measurement_overhead = min_delta;
336
337 printf("run measurement overhead: %" PRIu64 " nsecs\n", min_delta);
338}
339
340static void calibrate_sleep_measurement_overhead(struct perf_sched *sched)
341{
342 u64 T0, T1, delta, min_delta = NSEC_PER_SEC;
343 int i;
344
345 for (i = 0; i < 10; i++) {
346 T0 = get_nsecs();
347 sleep_nsecs(10000);
348 T1 = get_nsecs();
349 delta = T1-T0;
350 min_delta = min(min_delta, delta);
351 }
352 min_delta -= 10000;
353 sched->sleep_measurement_overhead = min_delta;
354
355 printf("sleep measurement overhead: %" PRIu64 " nsecs\n", min_delta);
356}
357
358static struct sched_atom *
359get_new_event(struct task_desc *task, u64 timestamp)
360{
361 struct sched_atom *event = zalloc(sizeof(*event));
362 unsigned long idx = task->nr_events;
363 size_t size;
364
365 event->timestamp = timestamp;
366 event->nr = idx;
367
368 task->nr_events++;
369 size = sizeof(struct sched_atom *) * task->nr_events;
370 task->atoms = realloc(task->atoms, size);
371 BUG_ON(!task->atoms);
372
373 task->atoms[idx] = event;
374
375 return event;
376}
377
378static struct sched_atom *last_event(struct task_desc *task)
379{
380 if (!task->nr_events)
381 return NULL;
382
383 return task->atoms[task->nr_events - 1];
384}
385
386static void add_sched_event_run(struct perf_sched *sched, struct task_desc *task,
387 u64 timestamp, u64 duration)
388{
389 struct sched_atom *event, *curr_event = last_event(task);
390
391 /*
392 * optimize an existing RUN event by merging this one
393 * to it:
394 */
395 if (curr_event && curr_event->type == SCHED_EVENT_RUN) {
396 sched->nr_run_events_optimized++;
397 curr_event->duration += duration;
398 return;
399 }
400
401 event = get_new_event(task, timestamp);
402
403 event->type = SCHED_EVENT_RUN;
404 event->duration = duration;
405
406 sched->nr_run_events++;
407}
408
409static void add_sched_event_wakeup(struct perf_sched *sched, struct task_desc *task,
410 u64 timestamp, struct task_desc *wakee)
411{
412 struct sched_atom *event, *wakee_event;
413
414 event = get_new_event(task, timestamp);
415 event->type = SCHED_EVENT_WAKEUP;
416 event->wakee = wakee;
417
418 wakee_event = last_event(wakee);
419 if (!wakee_event || wakee_event->type != SCHED_EVENT_SLEEP) {
420 sched->targetless_wakeups++;
421 return;
422 }
423 if (wakee_event->wait_sem) {
424 sched->multitarget_wakeups++;
425 return;
426 }
427
428 wakee_event->wait_sem = zalloc(sizeof(*wakee_event->wait_sem));
429 sem_init(wakee_event->wait_sem, 0, 0);
430 wakee_event->specific_wait = 1;
431 event->wait_sem = wakee_event->wait_sem;
432
433 sched->nr_wakeup_events++;
434}
435
436static void add_sched_event_sleep(struct perf_sched *sched, struct task_desc *task,
437 u64 timestamp, u64 task_state __maybe_unused)
438{
439 struct sched_atom *event = get_new_event(task, timestamp);
440
441 event->type = SCHED_EVENT_SLEEP;
442
443 sched->nr_sleep_events++;
444}
445
446static struct task_desc *register_pid(struct perf_sched *sched,
447 unsigned long pid, const char *comm)
448{
449 struct task_desc *task;
450 static int pid_max;
451
452 if (sched->pid_to_task == NULL) {
453 if (sysctl__read_int("kernel/pid_max", &pid_max) < 0)
454 pid_max = MAX_PID;
455 BUG_ON((sched->pid_to_task = calloc(pid_max, sizeof(struct task_desc *))) == NULL);
456 }
457 if (pid >= (unsigned long)pid_max) {
458 BUG_ON((sched->pid_to_task = realloc(sched->pid_to_task, (pid + 1) *
459 sizeof(struct task_desc *))) == NULL);
460 while (pid >= (unsigned long)pid_max)
461 sched->pid_to_task[pid_max++] = NULL;
462 }
463
464 task = sched->pid_to_task[pid];
465
466 if (task)
467 return task;
468
469 task = zalloc(sizeof(*task));
470 task->pid = pid;
471 task->nr = sched->nr_tasks;
472 strcpy(task->comm, comm);
473 /*
474 * every task starts in sleeping state - this gets ignored
475 * if there's no wakeup pointing to this sleep state:
476 */
477 add_sched_event_sleep(sched, task, 0, 0);
478
479 sched->pid_to_task[pid] = task;
480 sched->nr_tasks++;
481 sched->tasks = realloc(sched->tasks, sched->nr_tasks * sizeof(struct task_desc *));
482 BUG_ON(!sched->tasks);
483 sched->tasks[task->nr] = task;
484
485 if (verbose > 0)
486 printf("registered task #%ld, PID %ld (%s)\n", sched->nr_tasks, pid, comm);
487
488 return task;
489}
490
491
492static void print_task_traces(struct perf_sched *sched)
493{
494 struct task_desc *task;
495 unsigned long i;
496
497 for (i = 0; i < sched->nr_tasks; i++) {
498 task = sched->tasks[i];
499 printf("task %6ld (%20s:%10ld), nr_events: %ld\n",
500 task->nr, task->comm, task->pid, task->nr_events);
501 }
502}
503
504static void add_cross_task_wakeups(struct perf_sched *sched)
505{
506 struct task_desc *task1, *task2;
507 unsigned long i, j;
508
509 for (i = 0; i < sched->nr_tasks; i++) {
510 task1 = sched->tasks[i];
511 j = i + 1;
512 if (j == sched->nr_tasks)
513 j = 0;
514 task2 = sched->tasks[j];
515 add_sched_event_wakeup(sched, task1, 0, task2);
516 }
517}
518
519static void perf_sched__process_event(struct perf_sched *sched,
520 struct sched_atom *atom)
521{
522 int ret = 0;
523
524 switch (atom->type) {
525 case SCHED_EVENT_RUN:
526 burn_nsecs(sched, atom->duration);
527 break;
528 case SCHED_EVENT_SLEEP:
529 if (atom->wait_sem)
530 ret = sem_wait(atom->wait_sem);
531 BUG_ON(ret);
532 break;
533 case SCHED_EVENT_WAKEUP:
534 if (atom->wait_sem)
535 ret = sem_post(atom->wait_sem);
536 BUG_ON(ret);
537 break;
538 case SCHED_EVENT_MIGRATION:
539 break;
540 default:
541 BUG_ON(1);
542 }
543}
544
545static u64 get_cpu_usage_nsec_parent(void)
546{
547 struct rusage ru;
548 u64 sum;
549 int err;
550
551 err = getrusage(RUSAGE_SELF, &ru);
552 BUG_ON(err);
553
554 sum = ru.ru_utime.tv_sec * NSEC_PER_SEC + ru.ru_utime.tv_usec * NSEC_PER_USEC;
555 sum += ru.ru_stime.tv_sec * NSEC_PER_SEC + ru.ru_stime.tv_usec * NSEC_PER_USEC;
556
557 return sum;
558}
559
560static int self_open_counters(struct perf_sched *sched, unsigned long cur_task)
561{
562 struct perf_event_attr attr;
563 char sbuf[STRERR_BUFSIZE], info[STRERR_BUFSIZE];
564 int fd;
565 struct rlimit limit;
566 bool need_privilege = false;
567
568 memset(&attr, 0, sizeof(attr));
569
570 attr.type = PERF_TYPE_SOFTWARE;
571 attr.config = PERF_COUNT_SW_TASK_CLOCK;
572
573force_again:
574 fd = sys_perf_event_open(&attr, 0, -1, -1,
575 perf_event_open_cloexec_flag());
576
577 if (fd < 0) {
578 if (errno == EMFILE) {
579 if (sched->force) {
580 BUG_ON(getrlimit(RLIMIT_NOFILE, &limit) == -1);
581 limit.rlim_cur += sched->nr_tasks - cur_task;
582 if (limit.rlim_cur > limit.rlim_max) {
583 limit.rlim_max = limit.rlim_cur;
584 need_privilege = true;
585 }
586 if (setrlimit(RLIMIT_NOFILE, &limit) == -1) {
587 if (need_privilege && errno == EPERM)
588 strcpy(info, "Need privilege\n");
589 } else
590 goto force_again;
591 } else
592 strcpy(info, "Have a try with -f option\n");
593 }
594 pr_err("Error: sys_perf_event_open() syscall returned "
595 "with %d (%s)\n%s", fd,
596 str_error_r(errno, sbuf, sizeof(sbuf)), info);
597 exit(EXIT_FAILURE);
598 }
599 return fd;
600}
601
602static u64 get_cpu_usage_nsec_self(int fd)
603{
604 u64 runtime;
605 int ret;
606
607 ret = read(fd, &runtime, sizeof(runtime));
608 BUG_ON(ret != sizeof(runtime));
609
610 return runtime;
611}
612
613struct sched_thread_parms {
614 struct task_desc *task;
615 struct perf_sched *sched;
616 int fd;
617};
618
619static void *thread_func(void *ctx)
620{
621 struct sched_thread_parms *parms = ctx;
622 struct task_desc *this_task = parms->task;
623 struct perf_sched *sched = parms->sched;
624 u64 cpu_usage_0, cpu_usage_1;
625 unsigned long i, ret;
626 char comm2[22];
627 int fd = parms->fd;
628
629 zfree(&parms);
630
631 sprintf(comm2, ":%s", this_task->comm);
632 prctl(PR_SET_NAME, comm2);
633 if (fd < 0)
634 return NULL;
635again:
636 ret = sem_post(&this_task->ready_for_work);
637 BUG_ON(ret);
638 ret = pthread_mutex_lock(&sched->start_work_mutex);
639 BUG_ON(ret);
640 ret = pthread_mutex_unlock(&sched->start_work_mutex);
641 BUG_ON(ret);
642
643 cpu_usage_0 = get_cpu_usage_nsec_self(fd);
644
645 for (i = 0; i < this_task->nr_events; i++) {
646 this_task->curr_event = i;
647 perf_sched__process_event(sched, this_task->atoms[i]);
648 }
649
650 cpu_usage_1 = get_cpu_usage_nsec_self(fd);
651 this_task->cpu_usage = cpu_usage_1 - cpu_usage_0;
652 ret = sem_post(&this_task->work_done_sem);
653 BUG_ON(ret);
654
655 ret = pthread_mutex_lock(&sched->work_done_wait_mutex);
656 BUG_ON(ret);
657 ret = pthread_mutex_unlock(&sched->work_done_wait_mutex);
658 BUG_ON(ret);
659
660 goto again;
661}
662
663static void create_tasks(struct perf_sched *sched)
664{
665 struct task_desc *task;
666 pthread_attr_t attr;
667 unsigned long i;
668 int err;
669
670 err = pthread_attr_init(&attr);
671 BUG_ON(err);
672 err = pthread_attr_setstacksize(&attr,
673 (size_t) max(16 * 1024, (int)PTHREAD_STACK_MIN));
674 BUG_ON(err);
675 err = pthread_mutex_lock(&sched->start_work_mutex);
676 BUG_ON(err);
677 err = pthread_mutex_lock(&sched->work_done_wait_mutex);
678 BUG_ON(err);
679 for (i = 0; i < sched->nr_tasks; i++) {
680 struct sched_thread_parms *parms = malloc(sizeof(*parms));
681 BUG_ON(parms == NULL);
682 parms->task = task = sched->tasks[i];
683 parms->sched = sched;
684 parms->fd = self_open_counters(sched, i);
685 sem_init(&task->sleep_sem, 0, 0);
686 sem_init(&task->ready_for_work, 0, 0);
687 sem_init(&task->work_done_sem, 0, 0);
688 task->curr_event = 0;
689 err = pthread_create(&task->thread, &attr, thread_func, parms);
690 BUG_ON(err);
691 }
692}
693
694static void wait_for_tasks(struct perf_sched *sched)
695{
696 u64 cpu_usage_0, cpu_usage_1;
697 struct task_desc *task;
698 unsigned long i, ret;
699
700 sched->start_time = get_nsecs();
701 sched->cpu_usage = 0;
702 pthread_mutex_unlock(&sched->work_done_wait_mutex);
703
704 for (i = 0; i < sched->nr_tasks; i++) {
705 task = sched->tasks[i];
706 ret = sem_wait(&task->ready_for_work);
707 BUG_ON(ret);
708 sem_init(&task->ready_for_work, 0, 0);
709 }
710 ret = pthread_mutex_lock(&sched->work_done_wait_mutex);
711 BUG_ON(ret);
712
713 cpu_usage_0 = get_cpu_usage_nsec_parent();
714
715 pthread_mutex_unlock(&sched->start_work_mutex);
716
717 for (i = 0; i < sched->nr_tasks; i++) {
718 task = sched->tasks[i];
719 ret = sem_wait(&task->work_done_sem);
720 BUG_ON(ret);
721 sem_init(&task->work_done_sem, 0, 0);
722 sched->cpu_usage += task->cpu_usage;
723 task->cpu_usage = 0;
724 }
725
726 cpu_usage_1 = get_cpu_usage_nsec_parent();
727 if (!sched->runavg_cpu_usage)
728 sched->runavg_cpu_usage = sched->cpu_usage;
729 sched->runavg_cpu_usage = (sched->runavg_cpu_usage * (sched->replay_repeat - 1) + sched->cpu_usage) / sched->replay_repeat;
730
731 sched->parent_cpu_usage = cpu_usage_1 - cpu_usage_0;
732 if (!sched->runavg_parent_cpu_usage)
733 sched->runavg_parent_cpu_usage = sched->parent_cpu_usage;
734 sched->runavg_parent_cpu_usage = (sched->runavg_parent_cpu_usage * (sched->replay_repeat - 1) +
735 sched->parent_cpu_usage)/sched->replay_repeat;
736
737 ret = pthread_mutex_lock(&sched->start_work_mutex);
738 BUG_ON(ret);
739
740 for (i = 0; i < sched->nr_tasks; i++) {
741 task = sched->tasks[i];
742 sem_init(&task->sleep_sem, 0, 0);
743 task->curr_event = 0;
744 }
745}
746
747static void run_one_test(struct perf_sched *sched)
748{
749 u64 T0, T1, delta, avg_delta, fluct;
750
751 T0 = get_nsecs();
752 wait_for_tasks(sched);
753 T1 = get_nsecs();
754
755 delta = T1 - T0;
756 sched->sum_runtime += delta;
757 sched->nr_runs++;
758
759 avg_delta = sched->sum_runtime / sched->nr_runs;
760 if (delta < avg_delta)
761 fluct = avg_delta - delta;
762 else
763 fluct = delta - avg_delta;
764 sched->sum_fluct += fluct;
765 if (!sched->run_avg)
766 sched->run_avg = delta;
767 sched->run_avg = (sched->run_avg * (sched->replay_repeat - 1) + delta) / sched->replay_repeat;
768
769 printf("#%-3ld: %0.3f, ", sched->nr_runs, (double)delta / NSEC_PER_MSEC);
770
771 printf("ravg: %0.2f, ", (double)sched->run_avg / NSEC_PER_MSEC);
772
773 printf("cpu: %0.2f / %0.2f",
774 (double)sched->cpu_usage / NSEC_PER_MSEC, (double)sched->runavg_cpu_usage / NSEC_PER_MSEC);
775
776#if 0
777 /*
778 * rusage statistics done by the parent, these are less
779 * accurate than the sched->sum_exec_runtime based statistics:
780 */
781 printf(" [%0.2f / %0.2f]",
782 (double)sched->parent_cpu_usage / NSEC_PER_MSEC,
783 (double)sched->runavg_parent_cpu_usage / NSEC_PER_MSEC);
784#endif
785
786 printf("\n");
787
788 if (sched->nr_sleep_corrections)
789 printf(" (%ld sleep corrections)\n", sched->nr_sleep_corrections);
790 sched->nr_sleep_corrections = 0;
791}
792
793static void test_calibrations(struct perf_sched *sched)
794{
795 u64 T0, T1;
796
797 T0 = get_nsecs();
798 burn_nsecs(sched, NSEC_PER_MSEC);
799 T1 = get_nsecs();
800
801 printf("the run test took %" PRIu64 " nsecs\n", T1 - T0);
802
803 T0 = get_nsecs();
804 sleep_nsecs(NSEC_PER_MSEC);
805 T1 = get_nsecs();
806
807 printf("the sleep test took %" PRIu64 " nsecs\n", T1 - T0);
808}
809
810static int
811replay_wakeup_event(struct perf_sched *sched,
812 struct evsel *evsel, struct perf_sample *sample,
813 struct machine *machine __maybe_unused)
814{
815 const char *comm = evsel__strval(evsel, sample, "comm");
816 const u32 pid = evsel__intval(evsel, sample, "pid");
817 struct task_desc *waker, *wakee;
818
819 if (verbose > 0) {
820 printf("sched_wakeup event %p\n", evsel);
821
822 printf(" ... pid %d woke up %s/%d\n", sample->tid, comm, pid);
823 }
824
825 waker = register_pid(sched, sample->tid, "<unknown>");
826 wakee = register_pid(sched, pid, comm);
827
828 add_sched_event_wakeup(sched, waker, sample->time, wakee);
829 return 0;
830}
831
832static int replay_switch_event(struct perf_sched *sched,
833 struct evsel *evsel,
834 struct perf_sample *sample,
835 struct machine *machine __maybe_unused)
836{
837 const char *prev_comm = evsel__strval(evsel, sample, "prev_comm"),
838 *next_comm = evsel__strval(evsel, sample, "next_comm");
839 const u32 prev_pid = evsel__intval(evsel, sample, "prev_pid"),
840 next_pid = evsel__intval(evsel, sample, "next_pid");
841 const u64 prev_state = evsel__intval(evsel, sample, "prev_state");
842 struct task_desc *prev, __maybe_unused *next;
843 u64 timestamp0, timestamp = sample->time;
844 int cpu = sample->cpu;
845 s64 delta;
846
847 if (verbose > 0)
848 printf("sched_switch event %p\n", evsel);
849
850 if (cpu >= MAX_CPUS || cpu < 0)
851 return 0;
852
853 timestamp0 = sched->cpu_last_switched[cpu];
854 if (timestamp0)
855 delta = timestamp - timestamp0;
856 else
857 delta = 0;
858
859 if (delta < 0) {
860 pr_err("hm, delta: %" PRIu64 " < 0 ?\n", delta);
861 return -1;
862 }
863
864 pr_debug(" ... switch from %s/%d to %s/%d [ran %" PRIu64 " nsecs]\n",
865 prev_comm, prev_pid, next_comm, next_pid, delta);
866
867 prev = register_pid(sched, prev_pid, prev_comm);
868 next = register_pid(sched, next_pid, next_comm);
869
870 sched->cpu_last_switched[cpu] = timestamp;
871
872 add_sched_event_run(sched, prev, timestamp, delta);
873 add_sched_event_sleep(sched, prev, timestamp, prev_state);
874
875 return 0;
876}
877
878static int replay_fork_event(struct perf_sched *sched,
879 union perf_event *event,
880 struct machine *machine)
881{
882 struct thread *child, *parent;
883
884 child = machine__findnew_thread(machine, event->fork.pid,
885 event->fork.tid);
886 parent = machine__findnew_thread(machine, event->fork.ppid,
887 event->fork.ptid);
888
889 if (child == NULL || parent == NULL) {
890 pr_debug("thread does not exist on fork event: child %p, parent %p\n",
891 child, parent);
892 goto out_put;
893 }
894
895 if (verbose > 0) {
896 printf("fork event\n");
897 printf("... parent: %s/%d\n", thread__comm_str(parent), parent->tid);
898 printf("... child: %s/%d\n", thread__comm_str(child), child->tid);
899 }
900
901 register_pid(sched, parent->tid, thread__comm_str(parent));
902 register_pid(sched, child->tid, thread__comm_str(child));
903out_put:
904 thread__put(child);
905 thread__put(parent);
906 return 0;
907}
908
909struct sort_dimension {
910 const char *name;
911 sort_fn_t cmp;
912 struct list_head list;
913};
914
915/*
916 * handle runtime stats saved per thread
917 */
918static struct thread_runtime *thread__init_runtime(struct thread *thread)
919{
920 struct thread_runtime *r;
921
922 r = zalloc(sizeof(struct thread_runtime));
923 if (!r)
924 return NULL;
925
926 init_stats(&r->run_stats);
927 thread__set_priv(thread, r);
928
929 return r;
930}
931
932static struct thread_runtime *thread__get_runtime(struct thread *thread)
933{
934 struct thread_runtime *tr;
935
936 tr = thread__priv(thread);
937 if (tr == NULL) {
938 tr = thread__init_runtime(thread);
939 if (tr == NULL)
940 pr_debug("Failed to malloc memory for runtime data.\n");
941 }
942
943 return tr;
944}
945
946static int
947thread_lat_cmp(struct list_head *list, struct work_atoms *l, struct work_atoms *r)
948{
949 struct sort_dimension *sort;
950 int ret = 0;
951
952 BUG_ON(list_empty(list));
953
954 list_for_each_entry(sort, list, list) {
955 ret = sort->cmp(l, r);
956 if (ret)
957 return ret;
958 }
959
960 return ret;
961}
962
963static struct work_atoms *
964thread_atoms_search(struct rb_root_cached *root, struct thread *thread,
965 struct list_head *sort_list)
966{
967 struct rb_node *node = root->rb_root.rb_node;
968 struct work_atoms key = { .thread = thread };
969
970 while (node) {
971 struct work_atoms *atoms;
972 int cmp;
973
974 atoms = container_of(node, struct work_atoms, node);
975
976 cmp = thread_lat_cmp(sort_list, &key, atoms);
977 if (cmp > 0)
978 node = node->rb_left;
979 else if (cmp < 0)
980 node = node->rb_right;
981 else {
982 BUG_ON(thread != atoms->thread);
983 return atoms;
984 }
985 }
986 return NULL;
987}
988
989static void
990__thread_latency_insert(struct rb_root_cached *root, struct work_atoms *data,
991 struct list_head *sort_list)
992{
993 struct rb_node **new = &(root->rb_root.rb_node), *parent = NULL;
994 bool leftmost = true;
995
996 while (*new) {
997 struct work_atoms *this;
998 int cmp;
999
1000 this = container_of(*new, struct work_atoms, node);
1001 parent = *new;
1002
1003 cmp = thread_lat_cmp(sort_list, data, this);
1004
1005 if (cmp > 0)
1006 new = &((*new)->rb_left);
1007 else {
1008 new = &((*new)->rb_right);
1009 leftmost = false;
1010 }
1011 }
1012
1013 rb_link_node(&data->node, parent, new);
1014 rb_insert_color_cached(&data->node, root, leftmost);
1015}
1016
1017static int thread_atoms_insert(struct perf_sched *sched, struct thread *thread)
1018{
1019 struct work_atoms *atoms = zalloc(sizeof(*atoms));
1020 if (!atoms) {
1021 pr_err("No memory at %s\n", __func__);
1022 return -1;
1023 }
1024
1025 atoms->thread = thread__get(thread);
1026 INIT_LIST_HEAD(&atoms->work_list);
1027 __thread_latency_insert(&sched->atom_root, atoms, &sched->cmp_pid);
1028 return 0;
1029}
1030
1031static char sched_out_state(u64 prev_state)
1032{
1033 const char *str = TASK_STATE_TO_CHAR_STR;
1034
1035 return str[prev_state];
1036}
1037
1038static int
1039add_sched_out_event(struct work_atoms *atoms,
1040 char run_state,
1041 u64 timestamp)
1042{
1043 struct work_atom *atom = zalloc(sizeof(*atom));
1044 if (!atom) {
1045 pr_err("Non memory at %s", __func__);
1046 return -1;
1047 }
1048
1049 atom->sched_out_time = timestamp;
1050
1051 if (run_state == 'R') {
1052 atom->state = THREAD_WAIT_CPU;
1053 atom->wake_up_time = atom->sched_out_time;
1054 }
1055
1056 list_add_tail(&atom->list, &atoms->work_list);
1057 return 0;
1058}
1059
1060static void
1061add_runtime_event(struct work_atoms *atoms, u64 delta,
1062 u64 timestamp __maybe_unused)
1063{
1064 struct work_atom *atom;
1065
1066 BUG_ON(list_empty(&atoms->work_list));
1067
1068 atom = list_entry(atoms->work_list.prev, struct work_atom, list);
1069
1070 atom->runtime += delta;
1071 atoms->total_runtime += delta;
1072}
1073
1074static void
1075add_sched_in_event(struct work_atoms *atoms, u64 timestamp)
1076{
1077 struct work_atom *atom;
1078 u64 delta;
1079
1080 if (list_empty(&atoms->work_list))
1081 return;
1082
1083 atom = list_entry(atoms->work_list.prev, struct work_atom, list);
1084
1085 if (atom->state != THREAD_WAIT_CPU)
1086 return;
1087
1088 if (timestamp < atom->wake_up_time) {
1089 atom->state = THREAD_IGNORE;
1090 return;
1091 }
1092
1093 atom->state = THREAD_SCHED_IN;
1094 atom->sched_in_time = timestamp;
1095
1096 delta = atom->sched_in_time - atom->wake_up_time;
1097 atoms->total_lat += delta;
1098 if (delta > atoms->max_lat) {
1099 atoms->max_lat = delta;
1100 atoms->max_lat_start = atom->wake_up_time;
1101 atoms->max_lat_end = timestamp;
1102 }
1103 atoms->nb_atoms++;
1104}
1105
1106static int latency_switch_event(struct perf_sched *sched,
1107 struct evsel *evsel,
1108 struct perf_sample *sample,
1109 struct machine *machine)
1110{
1111 const u32 prev_pid = evsel__intval(evsel, sample, "prev_pid"),
1112 next_pid = evsel__intval(evsel, sample, "next_pid");
1113 const u64 prev_state = evsel__intval(evsel, sample, "prev_state");
1114 struct work_atoms *out_events, *in_events;
1115 struct thread *sched_out, *sched_in;
1116 u64 timestamp0, timestamp = sample->time;
1117 int cpu = sample->cpu, err = -1;
1118 s64 delta;
1119
1120 BUG_ON(cpu >= MAX_CPUS || cpu < 0);
1121
1122 timestamp0 = sched->cpu_last_switched[cpu];
1123 sched->cpu_last_switched[cpu] = timestamp;
1124 if (timestamp0)
1125 delta = timestamp - timestamp0;
1126 else
1127 delta = 0;
1128
1129 if (delta < 0) {
1130 pr_err("hm, delta: %" PRIu64 " < 0 ?\n", delta);
1131 return -1;
1132 }
1133
1134 sched_out = machine__findnew_thread(machine, -1, prev_pid);
1135 sched_in = machine__findnew_thread(machine, -1, next_pid);
1136 if (sched_out == NULL || sched_in == NULL)
1137 goto out_put;
1138
1139 out_events = thread_atoms_search(&sched->atom_root, sched_out, &sched->cmp_pid);
1140 if (!out_events) {
1141 if (thread_atoms_insert(sched, sched_out))
1142 goto out_put;
1143 out_events = thread_atoms_search(&sched->atom_root, sched_out, &sched->cmp_pid);
1144 if (!out_events) {
1145 pr_err("out-event: Internal tree error");
1146 goto out_put;
1147 }
1148 }
1149 if (add_sched_out_event(out_events, sched_out_state(prev_state), timestamp))
1150 return -1;
1151
1152 in_events = thread_atoms_search(&sched->atom_root, sched_in, &sched->cmp_pid);
1153 if (!in_events) {
1154 if (thread_atoms_insert(sched, sched_in))
1155 goto out_put;
1156 in_events = thread_atoms_search(&sched->atom_root, sched_in, &sched->cmp_pid);
1157 if (!in_events) {
1158 pr_err("in-event: Internal tree error");
1159 goto out_put;
1160 }
1161 /*
1162 * Take came in we have not heard about yet,
1163 * add in an initial atom in runnable state:
1164 */
1165 if (add_sched_out_event(in_events, 'R', timestamp))
1166 goto out_put;
1167 }
1168 add_sched_in_event(in_events, timestamp);
1169 err = 0;
1170out_put:
1171 thread__put(sched_out);
1172 thread__put(sched_in);
1173 return err;
1174}
1175
1176static int latency_runtime_event(struct perf_sched *sched,
1177 struct evsel *evsel,
1178 struct perf_sample *sample,
1179 struct machine *machine)
1180{
1181 const u32 pid = evsel__intval(evsel, sample, "pid");
1182 const u64 runtime = evsel__intval(evsel, sample, "runtime");
1183 struct thread *thread = machine__findnew_thread(machine, -1, pid);
1184 struct work_atoms *atoms = thread_atoms_search(&sched->atom_root, thread, &sched->cmp_pid);
1185 u64 timestamp = sample->time;
1186 int cpu = sample->cpu, err = -1;
1187
1188 if (thread == NULL)
1189 return -1;
1190
1191 BUG_ON(cpu >= MAX_CPUS || cpu < 0);
1192 if (!atoms) {
1193 if (thread_atoms_insert(sched, thread))
1194 goto out_put;
1195 atoms = thread_atoms_search(&sched->atom_root, thread, &sched->cmp_pid);
1196 if (!atoms) {
1197 pr_err("in-event: Internal tree error");
1198 goto out_put;
1199 }
1200 if (add_sched_out_event(atoms, 'R', timestamp))
1201 goto out_put;
1202 }
1203
1204 add_runtime_event(atoms, runtime, timestamp);
1205 err = 0;
1206out_put:
1207 thread__put(thread);
1208 return err;
1209}
1210
1211static int latency_wakeup_event(struct perf_sched *sched,
1212 struct evsel *evsel,
1213 struct perf_sample *sample,
1214 struct machine *machine)
1215{
1216 const u32 pid = evsel__intval(evsel, sample, "pid");
1217 struct work_atoms *atoms;
1218 struct work_atom *atom;
1219 struct thread *wakee;
1220 u64 timestamp = sample->time;
1221 int err = -1;
1222
1223 wakee = machine__findnew_thread(machine, -1, pid);
1224 if (wakee == NULL)
1225 return -1;
1226 atoms = thread_atoms_search(&sched->atom_root, wakee, &sched->cmp_pid);
1227 if (!atoms) {
1228 if (thread_atoms_insert(sched, wakee))
1229 goto out_put;
1230 atoms = thread_atoms_search(&sched->atom_root, wakee, &sched->cmp_pid);
1231 if (!atoms) {
1232 pr_err("wakeup-event: Internal tree error");
1233 goto out_put;
1234 }
1235 if (add_sched_out_event(atoms, 'S', timestamp))
1236 goto out_put;
1237 }
1238
1239 BUG_ON(list_empty(&atoms->work_list));
1240
1241 atom = list_entry(atoms->work_list.prev, struct work_atom, list);
1242
1243 /*
1244 * As we do not guarantee the wakeup event happens when
1245 * task is out of run queue, also may happen when task is
1246 * on run queue and wakeup only change ->state to TASK_RUNNING,
1247 * then we should not set the ->wake_up_time when wake up a
1248 * task which is on run queue.
1249 *
1250 * You WILL be missing events if you've recorded only
1251 * one CPU, or are only looking at only one, so don't
1252 * skip in this case.
1253 */
1254 if (sched->profile_cpu == -1 && atom->state != THREAD_SLEEPING)
1255 goto out_ok;
1256
1257 sched->nr_timestamps++;
1258 if (atom->sched_out_time > timestamp) {
1259 sched->nr_unordered_timestamps++;
1260 goto out_ok;
1261 }
1262
1263 atom->state = THREAD_WAIT_CPU;
1264 atom->wake_up_time = timestamp;
1265out_ok:
1266 err = 0;
1267out_put:
1268 thread__put(wakee);
1269 return err;
1270}
1271
1272static int latency_migrate_task_event(struct perf_sched *sched,
1273 struct evsel *evsel,
1274 struct perf_sample *sample,
1275 struct machine *machine)
1276{
1277 const u32 pid = evsel__intval(evsel, sample, "pid");
1278 u64 timestamp = sample->time;
1279 struct work_atoms *atoms;
1280 struct work_atom *atom;
1281 struct thread *migrant;
1282 int err = -1;
1283
1284 /*
1285 * Only need to worry about migration when profiling one CPU.
1286 */
1287 if (sched->profile_cpu == -1)
1288 return 0;
1289
1290 migrant = machine__findnew_thread(machine, -1, pid);
1291 if (migrant == NULL)
1292 return -1;
1293 atoms = thread_atoms_search(&sched->atom_root, migrant, &sched->cmp_pid);
1294 if (!atoms) {
1295 if (thread_atoms_insert(sched, migrant))
1296 goto out_put;
1297 register_pid(sched, migrant->tid, thread__comm_str(migrant));
1298 atoms = thread_atoms_search(&sched->atom_root, migrant, &sched->cmp_pid);
1299 if (!atoms) {
1300 pr_err("migration-event: Internal tree error");
1301 goto out_put;
1302 }
1303 if (add_sched_out_event(atoms, 'R', timestamp))
1304 goto out_put;
1305 }
1306
1307 BUG_ON(list_empty(&atoms->work_list));
1308
1309 atom = list_entry(atoms->work_list.prev, struct work_atom, list);
1310 atom->sched_in_time = atom->sched_out_time = atom->wake_up_time = timestamp;
1311
1312 sched->nr_timestamps++;
1313
1314 if (atom->sched_out_time > timestamp)
1315 sched->nr_unordered_timestamps++;
1316 err = 0;
1317out_put:
1318 thread__put(migrant);
1319 return err;
1320}
1321
1322static void output_lat_thread(struct perf_sched *sched, struct work_atoms *work_list)
1323{
1324 int i;
1325 int ret;
1326 u64 avg;
1327 char max_lat_start[32], max_lat_end[32];
1328
1329 if (!work_list->nb_atoms)
1330 return;
1331 /*
1332 * Ignore idle threads:
1333 */
1334 if (!strcmp(thread__comm_str(work_list->thread), "swapper"))
1335 return;
1336
1337 sched->all_runtime += work_list->total_runtime;
1338 sched->all_count += work_list->nb_atoms;
1339
1340 if (work_list->num_merged > 1)
1341 ret = printf(" %s:(%d) ", thread__comm_str(work_list->thread), work_list->num_merged);
1342 else
1343 ret = printf(" %s:%d ", thread__comm_str(work_list->thread), work_list->thread->tid);
1344
1345 for (i = 0; i < 24 - ret; i++)
1346 printf(" ");
1347
1348 avg = work_list->total_lat / work_list->nb_atoms;
1349 timestamp__scnprintf_usec(work_list->max_lat_start, max_lat_start, sizeof(max_lat_start));
1350 timestamp__scnprintf_usec(work_list->max_lat_end, max_lat_end, sizeof(max_lat_end));
1351
1352 printf("|%11.3f ms |%9" PRIu64 " | avg:%8.3f ms | max:%8.3f ms | max start: %12s s | max end: %12s s\n",
1353 (double)work_list->total_runtime / NSEC_PER_MSEC,
1354 work_list->nb_atoms, (double)avg / NSEC_PER_MSEC,
1355 (double)work_list->max_lat / NSEC_PER_MSEC,
1356 max_lat_start, max_lat_end);
1357}
1358
1359static int pid_cmp(struct work_atoms *l, struct work_atoms *r)
1360{
1361 if (l->thread == r->thread)
1362 return 0;
1363 if (l->thread->tid < r->thread->tid)
1364 return -1;
1365 if (l->thread->tid > r->thread->tid)
1366 return 1;
1367 return (int)(l->thread - r->thread);
1368}
1369
1370static int avg_cmp(struct work_atoms *l, struct work_atoms *r)
1371{
1372 u64 avgl, avgr;
1373
1374 if (!l->nb_atoms)
1375 return -1;
1376
1377 if (!r->nb_atoms)
1378 return 1;
1379
1380 avgl = l->total_lat / l->nb_atoms;
1381 avgr = r->total_lat / r->nb_atoms;
1382
1383 if (avgl < avgr)
1384 return -1;
1385 if (avgl > avgr)
1386 return 1;
1387
1388 return 0;
1389}
1390
1391static int max_cmp(struct work_atoms *l, struct work_atoms *r)
1392{
1393 if (l->max_lat < r->max_lat)
1394 return -1;
1395 if (l->max_lat > r->max_lat)
1396 return 1;
1397
1398 return 0;
1399}
1400
1401static int switch_cmp(struct work_atoms *l, struct work_atoms *r)
1402{
1403 if (l->nb_atoms < r->nb_atoms)
1404 return -1;
1405 if (l->nb_atoms > r->nb_atoms)
1406 return 1;
1407
1408 return 0;
1409}
1410
1411static int runtime_cmp(struct work_atoms *l, struct work_atoms *r)
1412{
1413 if (l->total_runtime < r->total_runtime)
1414 return -1;
1415 if (l->total_runtime > r->total_runtime)
1416 return 1;
1417
1418 return 0;
1419}
1420
1421static int sort_dimension__add(const char *tok, struct list_head *list)
1422{
1423 size_t i;
1424 static struct sort_dimension avg_sort_dimension = {
1425 .name = "avg",
1426 .cmp = avg_cmp,
1427 };
1428 static struct sort_dimension max_sort_dimension = {
1429 .name = "max",
1430 .cmp = max_cmp,
1431 };
1432 static struct sort_dimension pid_sort_dimension = {
1433 .name = "pid",
1434 .cmp = pid_cmp,
1435 };
1436 static struct sort_dimension runtime_sort_dimension = {
1437 .name = "runtime",
1438 .cmp = runtime_cmp,
1439 };
1440 static struct sort_dimension switch_sort_dimension = {
1441 .name = "switch",
1442 .cmp = switch_cmp,
1443 };
1444 struct sort_dimension *available_sorts[] = {
1445 &pid_sort_dimension,
1446 &avg_sort_dimension,
1447 &max_sort_dimension,
1448 &switch_sort_dimension,
1449 &runtime_sort_dimension,
1450 };
1451
1452 for (i = 0; i < ARRAY_SIZE(available_sorts); i++) {
1453 if (!strcmp(available_sorts[i]->name, tok)) {
1454 list_add_tail(&available_sorts[i]->list, list);
1455
1456 return 0;
1457 }
1458 }
1459
1460 return -1;
1461}
1462
1463static void perf_sched__sort_lat(struct perf_sched *sched)
1464{
1465 struct rb_node *node;
1466 struct rb_root_cached *root = &sched->atom_root;
1467again:
1468 for (;;) {
1469 struct work_atoms *data;
1470 node = rb_first_cached(root);
1471 if (!node)
1472 break;
1473
1474 rb_erase_cached(node, root);
1475 data = rb_entry(node, struct work_atoms, node);
1476 __thread_latency_insert(&sched->sorted_atom_root, data, &sched->sort_list);
1477 }
1478 if (root == &sched->atom_root) {
1479 root = &sched->merged_atom_root;
1480 goto again;
1481 }
1482}
1483
1484static int process_sched_wakeup_event(struct perf_tool *tool,
1485 struct evsel *evsel,
1486 struct perf_sample *sample,
1487 struct machine *machine)
1488{
1489 struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
1490
1491 if (sched->tp_handler->wakeup_event)
1492 return sched->tp_handler->wakeup_event(sched, evsel, sample, machine);
1493
1494 return 0;
1495}
1496
1497union map_priv {
1498 void *ptr;
1499 bool color;
1500};
1501
1502static bool thread__has_color(struct thread *thread)
1503{
1504 union map_priv priv = {
1505 .ptr = thread__priv(thread),
1506 };
1507
1508 return priv.color;
1509}
1510
1511static struct thread*
1512map__findnew_thread(struct perf_sched *sched, struct machine *machine, pid_t pid, pid_t tid)
1513{
1514 struct thread *thread = machine__findnew_thread(machine, pid, tid);
1515 union map_priv priv = {
1516 .color = false,
1517 };
1518
1519 if (!sched->map.color_pids || !thread || thread__priv(thread))
1520 return thread;
1521
1522 if (thread_map__has(sched->map.color_pids, tid))
1523 priv.color = true;
1524
1525 thread__set_priv(thread, priv.ptr);
1526 return thread;
1527}
1528
1529static int map_switch_event(struct perf_sched *sched, struct evsel *evsel,
1530 struct perf_sample *sample, struct machine *machine)
1531{
1532 const u32 next_pid = evsel__intval(evsel, sample, "next_pid");
1533 struct thread *sched_in;
1534 struct thread_runtime *tr;
1535 int new_shortname;
1536 u64 timestamp0, timestamp = sample->time;
1537 s64 delta;
1538 int i, this_cpu = sample->cpu;
1539 int cpus_nr;
1540 bool new_cpu = false;
1541 const char *color = PERF_COLOR_NORMAL;
1542 char stimestamp[32];
1543
1544 BUG_ON(this_cpu >= MAX_CPUS || this_cpu < 0);
1545
1546 if (this_cpu > sched->max_cpu)
1547 sched->max_cpu = this_cpu;
1548
1549 if (sched->map.comp) {
1550 cpus_nr = bitmap_weight(sched->map.comp_cpus_mask, MAX_CPUS);
1551 if (!test_and_set_bit(this_cpu, sched->map.comp_cpus_mask)) {
1552 sched->map.comp_cpus[cpus_nr++] = this_cpu;
1553 new_cpu = true;
1554 }
1555 } else
1556 cpus_nr = sched->max_cpu;
1557
1558 timestamp0 = sched->cpu_last_switched[this_cpu];
1559 sched->cpu_last_switched[this_cpu] = timestamp;
1560 if (timestamp0)
1561 delta = timestamp - timestamp0;
1562 else
1563 delta = 0;
1564
1565 if (delta < 0) {
1566 pr_err("hm, delta: %" PRIu64 " < 0 ?\n", delta);
1567 return -1;
1568 }
1569
1570 sched_in = map__findnew_thread(sched, machine, -1, next_pid);
1571 if (sched_in == NULL)
1572 return -1;
1573
1574 tr = thread__get_runtime(sched_in);
1575 if (tr == NULL) {
1576 thread__put(sched_in);
1577 return -1;
1578 }
1579
1580 sched->curr_thread[this_cpu] = thread__get(sched_in);
1581
1582 printf(" ");
1583
1584 new_shortname = 0;
1585 if (!tr->shortname[0]) {
1586 if (!strcmp(thread__comm_str(sched_in), "swapper")) {
1587 /*
1588 * Don't allocate a letter-number for swapper:0
1589 * as a shortname. Instead, we use '.' for it.
1590 */
1591 tr->shortname[0] = '.';
1592 tr->shortname[1] = ' ';
1593 } else {
1594 tr->shortname[0] = sched->next_shortname1;
1595 tr->shortname[1] = sched->next_shortname2;
1596
1597 if (sched->next_shortname1 < 'Z') {
1598 sched->next_shortname1++;
1599 } else {
1600 sched->next_shortname1 = 'A';
1601 if (sched->next_shortname2 < '9')
1602 sched->next_shortname2++;
1603 else
1604 sched->next_shortname2 = '0';
1605 }
1606 }
1607 new_shortname = 1;
1608 }
1609
1610 for (i = 0; i < cpus_nr; i++) {
1611 int cpu = sched->map.comp ? sched->map.comp_cpus[i] : i;
1612 struct thread *curr_thread = sched->curr_thread[cpu];
1613 struct thread_runtime *curr_tr;
1614 const char *pid_color = color;
1615 const char *cpu_color = color;
1616
1617 if (curr_thread && thread__has_color(curr_thread))
1618 pid_color = COLOR_PIDS;
1619
1620 if (sched->map.cpus && !cpu_map__has(sched->map.cpus, cpu))
1621 continue;
1622
1623 if (sched->map.color_cpus && cpu_map__has(sched->map.color_cpus, cpu))
1624 cpu_color = COLOR_CPUS;
1625
1626 if (cpu != this_cpu)
1627 color_fprintf(stdout, color, " ");
1628 else
1629 color_fprintf(stdout, cpu_color, "*");
1630
1631 if (sched->curr_thread[cpu]) {
1632 curr_tr = thread__get_runtime(sched->curr_thread[cpu]);
1633 if (curr_tr == NULL) {
1634 thread__put(sched_in);
1635 return -1;
1636 }
1637 color_fprintf(stdout, pid_color, "%2s ", curr_tr->shortname);
1638 } else
1639 color_fprintf(stdout, color, " ");
1640 }
1641
1642 if (sched->map.cpus && !cpu_map__has(sched->map.cpus, this_cpu))
1643 goto out;
1644
1645 timestamp__scnprintf_usec(timestamp, stimestamp, sizeof(stimestamp));
1646 color_fprintf(stdout, color, " %12s secs ", stimestamp);
1647 if (new_shortname || tr->comm_changed || (verbose > 0 && sched_in->tid)) {
1648 const char *pid_color = color;
1649
1650 if (thread__has_color(sched_in))
1651 pid_color = COLOR_PIDS;
1652
1653 color_fprintf(stdout, pid_color, "%s => %s:%d",
1654 tr->shortname, thread__comm_str(sched_in), sched_in->tid);
1655 tr->comm_changed = false;
1656 }
1657
1658 if (sched->map.comp && new_cpu)
1659 color_fprintf(stdout, color, " (CPU %d)", this_cpu);
1660
1661out:
1662 color_fprintf(stdout, color, "\n");
1663
1664 thread__put(sched_in);
1665
1666 return 0;
1667}
1668
1669static int process_sched_switch_event(struct perf_tool *tool,
1670 struct evsel *evsel,
1671 struct perf_sample *sample,
1672 struct machine *machine)
1673{
1674 struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
1675 int this_cpu = sample->cpu, err = 0;
1676 u32 prev_pid = evsel__intval(evsel, sample, "prev_pid"),
1677 next_pid = evsel__intval(evsel, sample, "next_pid");
1678
1679 if (sched->curr_pid[this_cpu] != (u32)-1) {
1680 /*
1681 * Are we trying to switch away a PID that is
1682 * not current?
1683 */
1684 if (sched->curr_pid[this_cpu] != prev_pid)
1685 sched->nr_context_switch_bugs++;
1686 }
1687
1688 if (sched->tp_handler->switch_event)
1689 err = sched->tp_handler->switch_event(sched, evsel, sample, machine);
1690
1691 sched->curr_pid[this_cpu] = next_pid;
1692 return err;
1693}
1694
1695static int process_sched_runtime_event(struct perf_tool *tool,
1696 struct evsel *evsel,
1697 struct perf_sample *sample,
1698 struct machine *machine)
1699{
1700 struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
1701
1702 if (sched->tp_handler->runtime_event)
1703 return sched->tp_handler->runtime_event(sched, evsel, sample, machine);
1704
1705 return 0;
1706}
1707
1708static int perf_sched__process_fork_event(struct perf_tool *tool,
1709 union perf_event *event,
1710 struct perf_sample *sample,
1711 struct machine *machine)
1712{
1713 struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
1714
1715 /* run the fork event through the perf machinery */
1716 perf_event__process_fork(tool, event, sample, machine);
1717
1718 /* and then run additional processing needed for this command */
1719 if (sched->tp_handler->fork_event)
1720 return sched->tp_handler->fork_event(sched, event, machine);
1721
1722 return 0;
1723}
1724
1725static int process_sched_migrate_task_event(struct perf_tool *tool,
1726 struct evsel *evsel,
1727 struct perf_sample *sample,
1728 struct machine *machine)
1729{
1730 struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
1731
1732 if (sched->tp_handler->migrate_task_event)
1733 return sched->tp_handler->migrate_task_event(sched, evsel, sample, machine);
1734
1735 return 0;
1736}
1737
1738typedef int (*tracepoint_handler)(struct perf_tool *tool,
1739 struct evsel *evsel,
1740 struct perf_sample *sample,
1741 struct machine *machine);
1742
1743static int perf_sched__process_tracepoint_sample(struct perf_tool *tool __maybe_unused,
1744 union perf_event *event __maybe_unused,
1745 struct perf_sample *sample,
1746 struct evsel *evsel,
1747 struct machine *machine)
1748{
1749 int err = 0;
1750
1751 if (evsel->handler != NULL) {
1752 tracepoint_handler f = evsel->handler;
1753 err = f(tool, evsel, sample, machine);
1754 }
1755
1756 return err;
1757}
1758
1759static int perf_sched__process_comm(struct perf_tool *tool __maybe_unused,
1760 union perf_event *event,
1761 struct perf_sample *sample,
1762 struct machine *machine)
1763{
1764 struct thread *thread;
1765 struct thread_runtime *tr;
1766 int err;
1767
1768 err = perf_event__process_comm(tool, event, sample, machine);
1769 if (err)
1770 return err;
1771
1772 thread = machine__find_thread(machine, sample->pid, sample->tid);
1773 if (!thread) {
1774 pr_err("Internal error: can't find thread\n");
1775 return -1;
1776 }
1777
1778 tr = thread__get_runtime(thread);
1779 if (tr == NULL) {
1780 thread__put(thread);
1781 return -1;
1782 }
1783
1784 tr->comm_changed = true;
1785 thread__put(thread);
1786
1787 return 0;
1788}
1789
1790static int perf_sched__read_events(struct perf_sched *sched)
1791{
1792 const struct evsel_str_handler handlers[] = {
1793 { "sched:sched_switch", process_sched_switch_event, },
1794 { "sched:sched_stat_runtime", process_sched_runtime_event, },
1795 { "sched:sched_wakeup", process_sched_wakeup_event, },
1796 { "sched:sched_wakeup_new", process_sched_wakeup_event, },
1797 { "sched:sched_migrate_task", process_sched_migrate_task_event, },
1798 };
1799 struct perf_session *session;
1800 struct perf_data data = {
1801 .path = input_name,
1802 .mode = PERF_DATA_MODE_READ,
1803 .force = sched->force,
1804 };
1805 int rc = -1;
1806
1807 session = perf_session__new(&data, false, &sched->tool);
1808 if (IS_ERR(session)) {
1809 pr_debug("Error creating perf session");
1810 return PTR_ERR(session);
1811 }
1812
1813 symbol__init(&session->header.env);
1814
1815 if (perf_session__set_tracepoints_handlers(session, handlers))
1816 goto out_delete;
1817
1818 if (perf_session__has_traces(session, "record -R")) {
1819 int err = perf_session__process_events(session);
1820 if (err) {
1821 pr_err("Failed to process events, error %d", err);
1822 goto out_delete;
1823 }
1824
1825 sched->nr_events = session->evlist->stats.nr_events[0];
1826 sched->nr_lost_events = session->evlist->stats.total_lost;
1827 sched->nr_lost_chunks = session->evlist->stats.nr_events[PERF_RECORD_LOST];
1828 }
1829
1830 rc = 0;
1831out_delete:
1832 perf_session__delete(session);
1833 return rc;
1834}
1835
1836/*
1837 * scheduling times are printed as msec.usec
1838 */
1839static inline void print_sched_time(unsigned long long nsecs, int width)
1840{
1841 unsigned long msecs;
1842 unsigned long usecs;
1843
1844 msecs = nsecs / NSEC_PER_MSEC;
1845 nsecs -= msecs * NSEC_PER_MSEC;
1846 usecs = nsecs / NSEC_PER_USEC;
1847 printf("%*lu.%03lu ", width, msecs, usecs);
1848}
1849
1850/*
1851 * returns runtime data for event, allocating memory for it the
1852 * first time it is used.
1853 */
1854static struct evsel_runtime *evsel__get_runtime(struct evsel *evsel)
1855{
1856 struct evsel_runtime *r = evsel->priv;
1857
1858 if (r == NULL) {
1859 r = zalloc(sizeof(struct evsel_runtime));
1860 evsel->priv = r;
1861 }
1862
1863 return r;
1864}
1865
1866/*
1867 * save last time event was seen per cpu
1868 */
1869static void evsel__save_time(struct evsel *evsel, u64 timestamp, u32 cpu)
1870{
1871 struct evsel_runtime *r = evsel__get_runtime(evsel);
1872
1873 if (r == NULL)
1874 return;
1875
1876 if ((cpu >= r->ncpu) || (r->last_time == NULL)) {
1877 int i, n = __roundup_pow_of_two(cpu+1);
1878 void *p = r->last_time;
1879
1880 p = realloc(r->last_time, n * sizeof(u64));
1881 if (!p)
1882 return;
1883
1884 r->last_time = p;
1885 for (i = r->ncpu; i < n; ++i)
1886 r->last_time[i] = (u64) 0;
1887
1888 r->ncpu = n;
1889 }
1890
1891 r->last_time[cpu] = timestamp;
1892}
1893
1894/* returns last time this event was seen on the given cpu */
1895static u64 evsel__get_time(struct evsel *evsel, u32 cpu)
1896{
1897 struct evsel_runtime *r = evsel__get_runtime(evsel);
1898
1899 if ((r == NULL) || (r->last_time == NULL) || (cpu >= r->ncpu))
1900 return 0;
1901
1902 return r->last_time[cpu];
1903}
1904
1905static int comm_width = 30;
1906
1907static char *timehist_get_commstr(struct thread *thread)
1908{
1909 static char str[32];
1910 const char *comm = thread__comm_str(thread);
1911 pid_t tid = thread->tid;
1912 pid_t pid = thread->pid_;
1913 int n;
1914
1915 if (pid == 0)
1916 n = scnprintf(str, sizeof(str), "%s", comm);
1917
1918 else if (tid != pid)
1919 n = scnprintf(str, sizeof(str), "%s[%d/%d]", comm, tid, pid);
1920
1921 else
1922 n = scnprintf(str, sizeof(str), "%s[%d]", comm, tid);
1923
1924 if (n > comm_width)
1925 comm_width = n;
1926
1927 return str;
1928}
1929
1930static void timehist_header(struct perf_sched *sched)
1931{
1932 u32 ncpus = sched->max_cpu + 1;
1933 u32 i, j;
1934
1935 printf("%15s %6s ", "time", "cpu");
1936
1937 if (sched->show_cpu_visual) {
1938 printf(" ");
1939 for (i = 0, j = 0; i < ncpus; ++i) {
1940 printf("%x", j++);
1941 if (j > 15)
1942 j = 0;
1943 }
1944 printf(" ");
1945 }
1946
1947 printf(" %-*s %9s %9s %9s", comm_width,
1948 "task name", "wait time", "sch delay", "run time");
1949
1950 if (sched->show_state)
1951 printf(" %s", "state");
1952
1953 printf("\n");
1954
1955 /*
1956 * units row
1957 */
1958 printf("%15s %-6s ", "", "");
1959
1960 if (sched->show_cpu_visual)
1961 printf(" %*s ", ncpus, "");
1962
1963 printf(" %-*s %9s %9s %9s", comm_width,
1964 "[tid/pid]", "(msec)", "(msec)", "(msec)");
1965
1966 if (sched->show_state)
1967 printf(" %5s", "");
1968
1969 printf("\n");
1970
1971 /*
1972 * separator
1973 */
1974 printf("%.15s %.6s ", graph_dotted_line, graph_dotted_line);
1975
1976 if (sched->show_cpu_visual)
1977 printf(" %.*s ", ncpus, graph_dotted_line);
1978
1979 printf(" %.*s %.9s %.9s %.9s", comm_width,
1980 graph_dotted_line, graph_dotted_line, graph_dotted_line,
1981 graph_dotted_line);
1982
1983 if (sched->show_state)
1984 printf(" %.5s", graph_dotted_line);
1985
1986 printf("\n");
1987}
1988
1989static char task_state_char(struct thread *thread, int state)
1990{
1991 static const char state_to_char[] = TASK_STATE_TO_CHAR_STR;
1992 unsigned bit = state ? ffs(state) : 0;
1993
1994 /* 'I' for idle */
1995 if (thread->tid == 0)
1996 return 'I';
1997
1998 return bit < sizeof(state_to_char) - 1 ? state_to_char[bit] : '?';
1999}
2000
2001static void timehist_print_sample(struct perf_sched *sched,
2002 struct evsel *evsel,
2003 struct perf_sample *sample,
2004 struct addr_location *al,
2005 struct thread *thread,
2006 u64 t, int state)
2007{
2008 struct thread_runtime *tr = thread__priv(thread);
2009 const char *next_comm = evsel__strval(evsel, sample, "next_comm");
2010 const u32 next_pid = evsel__intval(evsel, sample, "next_pid");
2011 u32 max_cpus = sched->max_cpu + 1;
2012 char tstr[64];
2013 char nstr[30];
2014 u64 wait_time;
2015
2016 if (cpu_list && !test_bit(sample->cpu, cpu_bitmap))
2017 return;
2018
2019 timestamp__scnprintf_usec(t, tstr, sizeof(tstr));
2020 printf("%15s [%04d] ", tstr, sample->cpu);
2021
2022 if (sched->show_cpu_visual) {
2023 u32 i;
2024 char c;
2025
2026 printf(" ");
2027 for (i = 0; i < max_cpus; ++i) {
2028 /* flag idle times with 'i'; others are sched events */
2029 if (i == sample->cpu)
2030 c = (thread->tid == 0) ? 'i' : 's';
2031 else
2032 c = ' ';
2033 printf("%c", c);
2034 }
2035 printf(" ");
2036 }
2037
2038 printf(" %-*s ", comm_width, timehist_get_commstr(thread));
2039
2040 wait_time = tr->dt_sleep + tr->dt_iowait + tr->dt_preempt;
2041 print_sched_time(wait_time, 6);
2042
2043 print_sched_time(tr->dt_delay, 6);
2044 print_sched_time(tr->dt_run, 6);
2045
2046 if (sched->show_state)
2047 printf(" %5c ", task_state_char(thread, state));
2048
2049 if (sched->show_next) {
2050 snprintf(nstr, sizeof(nstr), "next: %s[%d]", next_comm, next_pid);
2051 printf(" %-*s", comm_width, nstr);
2052 }
2053
2054 if (sched->show_wakeups && !sched->show_next)
2055 printf(" %-*s", comm_width, "");
2056
2057 if (thread->tid == 0)
2058 goto out;
2059
2060 if (sched->show_callchain)
2061 printf(" ");
2062
2063 sample__fprintf_sym(sample, al, 0,
2064 EVSEL__PRINT_SYM | EVSEL__PRINT_ONELINE |
2065 EVSEL__PRINT_CALLCHAIN_ARROW |
2066 EVSEL__PRINT_SKIP_IGNORED,
2067 &callchain_cursor, symbol_conf.bt_stop_list, stdout);
2068
2069out:
2070 printf("\n");
2071}
2072
2073/*
2074 * Explanation of delta-time stats:
2075 *
2076 * t = time of current schedule out event
2077 * tprev = time of previous sched out event
2078 * also time of schedule-in event for current task
2079 * last_time = time of last sched change event for current task
2080 * (i.e, time process was last scheduled out)
2081 * ready_to_run = time of wakeup for current task
2082 *
2083 * -----|------------|------------|------------|------
2084 * last ready tprev t
2085 * time to run
2086 *
2087 * |-------- dt_wait --------|
2088 * |- dt_delay -|-- dt_run --|
2089 *
2090 * dt_run = run time of current task
2091 * dt_wait = time between last schedule out event for task and tprev
2092 * represents time spent off the cpu
2093 * dt_delay = time between wakeup and schedule-in of task
2094 */
2095
2096static void timehist_update_runtime_stats(struct thread_runtime *r,
2097 u64 t, u64 tprev)
2098{
2099 r->dt_delay = 0;
2100 r->dt_sleep = 0;
2101 r->dt_iowait = 0;
2102 r->dt_preempt = 0;
2103 r->dt_run = 0;
2104
2105 if (tprev) {
2106 r->dt_run = t - tprev;
2107 if (r->ready_to_run) {
2108 if (r->ready_to_run > tprev)
2109 pr_debug("time travel: wakeup time for task > previous sched_switch event\n");
2110 else
2111 r->dt_delay = tprev - r->ready_to_run;
2112 }
2113
2114 if (r->last_time > tprev)
2115 pr_debug("time travel: last sched out time for task > previous sched_switch event\n");
2116 else if (r->last_time) {
2117 u64 dt_wait = tprev - r->last_time;
2118
2119 if (r->last_state == TASK_RUNNING)
2120 r->dt_preempt = dt_wait;
2121 else if (r->last_state == TASK_UNINTERRUPTIBLE)
2122 r->dt_iowait = dt_wait;
2123 else
2124 r->dt_sleep = dt_wait;
2125 }
2126 }
2127
2128 update_stats(&r->run_stats, r->dt_run);
2129
2130 r->total_run_time += r->dt_run;
2131 r->total_delay_time += r->dt_delay;
2132 r->total_sleep_time += r->dt_sleep;
2133 r->total_iowait_time += r->dt_iowait;
2134 r->total_preempt_time += r->dt_preempt;
2135}
2136
2137static bool is_idle_sample(struct perf_sample *sample,
2138 struct evsel *evsel)
2139{
2140 /* pid 0 == swapper == idle task */
2141 if (strcmp(evsel__name(evsel), "sched:sched_switch") == 0)
2142 return evsel__intval(evsel, sample, "prev_pid") == 0;
2143
2144 return sample->pid == 0;
2145}
2146
2147static void save_task_callchain(struct perf_sched *sched,
2148 struct perf_sample *sample,
2149 struct evsel *evsel,
2150 struct machine *machine)
2151{
2152 struct callchain_cursor *cursor = &callchain_cursor;
2153 struct thread *thread;
2154
2155 /* want main thread for process - has maps */
2156 thread = machine__findnew_thread(machine, sample->pid, sample->pid);
2157 if (thread == NULL) {
2158 pr_debug("Failed to get thread for pid %d.\n", sample->pid);
2159 return;
2160 }
2161
2162 if (!sched->show_callchain || sample->callchain == NULL)
2163 return;
2164
2165 if (thread__resolve_callchain(thread, cursor, evsel, sample,
2166 NULL, NULL, sched->max_stack + 2) != 0) {
2167 if (verbose > 0)
2168 pr_err("Failed to resolve callchain. Skipping\n");
2169
2170 return;
2171 }
2172
2173 callchain_cursor_commit(cursor);
2174
2175 while (true) {
2176 struct callchain_cursor_node *node;
2177 struct symbol *sym;
2178
2179 node = callchain_cursor_current(cursor);
2180 if (node == NULL)
2181 break;
2182
2183 sym = node->ms.sym;
2184 if (sym) {
2185 if (!strcmp(sym->name, "schedule") ||
2186 !strcmp(sym->name, "__schedule") ||
2187 !strcmp(sym->name, "preempt_schedule"))
2188 sym->ignore = 1;
2189 }
2190
2191 callchain_cursor_advance(cursor);
2192 }
2193}
2194
2195static int init_idle_thread(struct thread *thread)
2196{
2197 struct idle_thread_runtime *itr;
2198
2199 thread__set_comm(thread, idle_comm, 0);
2200
2201 itr = zalloc(sizeof(*itr));
2202 if (itr == NULL)
2203 return -ENOMEM;
2204
2205 init_stats(&itr->tr.run_stats);
2206 callchain_init(&itr->callchain);
2207 callchain_cursor_reset(&itr->cursor);
2208 thread__set_priv(thread, itr);
2209
2210 return 0;
2211}
2212
2213/*
2214 * Track idle stats per cpu by maintaining a local thread
2215 * struct for the idle task on each cpu.
2216 */
2217static int init_idle_threads(int ncpu)
2218{
2219 int i, ret;
2220
2221 idle_threads = zalloc(ncpu * sizeof(struct thread *));
2222 if (!idle_threads)
2223 return -ENOMEM;
2224
2225 idle_max_cpu = ncpu;
2226
2227 /* allocate the actual thread struct if needed */
2228 for (i = 0; i < ncpu; ++i) {
2229 idle_threads[i] = thread__new(0, 0);
2230 if (idle_threads[i] == NULL)
2231 return -ENOMEM;
2232
2233 ret = init_idle_thread(idle_threads[i]);
2234 if (ret < 0)
2235 return ret;
2236 }
2237
2238 return 0;
2239}
2240
2241static void free_idle_threads(void)
2242{
2243 int i;
2244
2245 if (idle_threads == NULL)
2246 return;
2247
2248 for (i = 0; i < idle_max_cpu; ++i) {
2249 if ((idle_threads[i]))
2250 thread__delete(idle_threads[i]);
2251 }
2252
2253 free(idle_threads);
2254}
2255
2256static struct thread *get_idle_thread(int cpu)
2257{
2258 /*
2259 * expand/allocate array of pointers to local thread
2260 * structs if needed
2261 */
2262 if ((cpu >= idle_max_cpu) || (idle_threads == NULL)) {
2263 int i, j = __roundup_pow_of_two(cpu+1);
2264 void *p;
2265
2266 p = realloc(idle_threads, j * sizeof(struct thread *));
2267 if (!p)
2268 return NULL;
2269
2270 idle_threads = (struct thread **) p;
2271 for (i = idle_max_cpu; i < j; ++i)
2272 idle_threads[i] = NULL;
2273
2274 idle_max_cpu = j;
2275 }
2276
2277 /* allocate a new thread struct if needed */
2278 if (idle_threads[cpu] == NULL) {
2279 idle_threads[cpu] = thread__new(0, 0);
2280
2281 if (idle_threads[cpu]) {
2282 if (init_idle_thread(idle_threads[cpu]) < 0)
2283 return NULL;
2284 }
2285 }
2286
2287 return idle_threads[cpu];
2288}
2289
2290static void save_idle_callchain(struct perf_sched *sched,
2291 struct idle_thread_runtime *itr,
2292 struct perf_sample *sample)
2293{
2294 if (!sched->show_callchain || sample->callchain == NULL)
2295 return;
2296
2297 callchain_cursor__copy(&itr->cursor, &callchain_cursor);
2298}
2299
2300static struct thread *timehist_get_thread(struct perf_sched *sched,
2301 struct perf_sample *sample,
2302 struct machine *machine,
2303 struct evsel *evsel)
2304{
2305 struct thread *thread;
2306
2307 if (is_idle_sample(sample, evsel)) {
2308 thread = get_idle_thread(sample->cpu);
2309 if (thread == NULL)
2310 pr_err("Failed to get idle thread for cpu %d.\n", sample->cpu);
2311
2312 } else {
2313 /* there were samples with tid 0 but non-zero pid */
2314 thread = machine__findnew_thread(machine, sample->pid,
2315 sample->tid ?: sample->pid);
2316 if (thread == NULL) {
2317 pr_debug("Failed to get thread for tid %d. skipping sample.\n",
2318 sample->tid);
2319 }
2320
2321 save_task_callchain(sched, sample, evsel, machine);
2322 if (sched->idle_hist) {
2323 struct thread *idle;
2324 struct idle_thread_runtime *itr;
2325
2326 idle = get_idle_thread(sample->cpu);
2327 if (idle == NULL) {
2328 pr_err("Failed to get idle thread for cpu %d.\n", sample->cpu);
2329 return NULL;
2330 }
2331
2332 itr = thread__priv(idle);
2333 if (itr == NULL)
2334 return NULL;
2335
2336 itr->last_thread = thread;
2337
2338 /* copy task callchain when entering to idle */
2339 if (evsel__intval(evsel, sample, "next_pid") == 0)
2340 save_idle_callchain(sched, itr, sample);
2341 }
2342 }
2343
2344 return thread;
2345}
2346
2347static bool timehist_skip_sample(struct perf_sched *sched,
2348 struct thread *thread,
2349 struct evsel *evsel,
2350 struct perf_sample *sample)
2351{
2352 bool rc = false;
2353
2354 if (thread__is_filtered(thread)) {
2355 rc = true;
2356 sched->skipped_samples++;
2357 }
2358
2359 if (sched->idle_hist) {
2360 if (strcmp(evsel__name(evsel), "sched:sched_switch"))
2361 rc = true;
2362 else if (evsel__intval(evsel, sample, "prev_pid") != 0 &&
2363 evsel__intval(evsel, sample, "next_pid") != 0)
2364 rc = true;
2365 }
2366
2367 return rc;
2368}
2369
2370static void timehist_print_wakeup_event(struct perf_sched *sched,
2371 struct evsel *evsel,
2372 struct perf_sample *sample,
2373 struct machine *machine,
2374 struct thread *awakened)
2375{
2376 struct thread *thread;
2377 char tstr[64];
2378
2379 thread = machine__findnew_thread(machine, sample->pid, sample->tid);
2380 if (thread == NULL)
2381 return;
2382
2383 /* show wakeup unless both awakee and awaker are filtered */
2384 if (timehist_skip_sample(sched, thread, evsel, sample) &&
2385 timehist_skip_sample(sched, awakened, evsel, sample)) {
2386 return;
2387 }
2388
2389 timestamp__scnprintf_usec(sample->time, tstr, sizeof(tstr));
2390 printf("%15s [%04d] ", tstr, sample->cpu);
2391 if (sched->show_cpu_visual)
2392 printf(" %*s ", sched->max_cpu + 1, "");
2393
2394 printf(" %-*s ", comm_width, timehist_get_commstr(thread));
2395
2396 /* dt spacer */
2397 printf(" %9s %9s %9s ", "", "", "");
2398
2399 printf("awakened: %s", timehist_get_commstr(awakened));
2400
2401 printf("\n");
2402}
2403
2404static int timehist_sched_wakeup_ignore(struct perf_tool *tool __maybe_unused,
2405 union perf_event *event __maybe_unused,
2406 struct evsel *evsel __maybe_unused,
2407 struct perf_sample *sample __maybe_unused,
2408 struct machine *machine __maybe_unused)
2409{
2410 return 0;
2411}
2412
2413static int timehist_sched_wakeup_event(struct perf_tool *tool,
2414 union perf_event *event __maybe_unused,
2415 struct evsel *evsel,
2416 struct perf_sample *sample,
2417 struct machine *machine)
2418{
2419 struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
2420 struct thread *thread;
2421 struct thread_runtime *tr = NULL;
2422 /* want pid of awakened task not pid in sample */
2423 const u32 pid = evsel__intval(evsel, sample, "pid");
2424
2425 thread = machine__findnew_thread(machine, 0, pid);
2426 if (thread == NULL)
2427 return -1;
2428
2429 tr = thread__get_runtime(thread);
2430 if (tr == NULL)
2431 return -1;
2432
2433 if (tr->ready_to_run == 0)
2434 tr->ready_to_run = sample->time;
2435
2436 /* show wakeups if requested */
2437 if (sched->show_wakeups &&
2438 !perf_time__skip_sample(&sched->ptime, sample->time))
2439 timehist_print_wakeup_event(sched, evsel, sample, machine, thread);
2440
2441 return 0;
2442}
2443
2444static void timehist_print_migration_event(struct perf_sched *sched,
2445 struct evsel *evsel,
2446 struct perf_sample *sample,
2447 struct machine *machine,
2448 struct thread *migrated)
2449{
2450 struct thread *thread;
2451 char tstr[64];
2452 u32 max_cpus = sched->max_cpu + 1;
2453 u32 ocpu, dcpu;
2454
2455 if (sched->summary_only)
2456 return;
2457
2458 max_cpus = sched->max_cpu + 1;
2459 ocpu = evsel__intval(evsel, sample, "orig_cpu");
2460 dcpu = evsel__intval(evsel, sample, "dest_cpu");
2461
2462 thread = machine__findnew_thread(machine, sample->pid, sample->tid);
2463 if (thread == NULL)
2464 return;
2465
2466 if (timehist_skip_sample(sched, thread, evsel, sample) &&
2467 timehist_skip_sample(sched, migrated, evsel, sample)) {
2468 return;
2469 }
2470
2471 timestamp__scnprintf_usec(sample->time, tstr, sizeof(tstr));
2472 printf("%15s [%04d] ", tstr, sample->cpu);
2473
2474 if (sched->show_cpu_visual) {
2475 u32 i;
2476 char c;
2477
2478 printf(" ");
2479 for (i = 0; i < max_cpus; ++i) {
2480 c = (i == sample->cpu) ? 'm' : ' ';
2481 printf("%c", c);
2482 }
2483 printf(" ");
2484 }
2485
2486 printf(" %-*s ", comm_width, timehist_get_commstr(thread));
2487
2488 /* dt spacer */
2489 printf(" %9s %9s %9s ", "", "", "");
2490
2491 printf("migrated: %s", timehist_get_commstr(migrated));
2492 printf(" cpu %d => %d", ocpu, dcpu);
2493
2494 printf("\n");
2495}
2496
2497static int timehist_migrate_task_event(struct perf_tool *tool,
2498 union perf_event *event __maybe_unused,
2499 struct evsel *evsel,
2500 struct perf_sample *sample,
2501 struct machine *machine)
2502{
2503 struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
2504 struct thread *thread;
2505 struct thread_runtime *tr = NULL;
2506 /* want pid of migrated task not pid in sample */
2507 const u32 pid = evsel__intval(evsel, sample, "pid");
2508
2509 thread = machine__findnew_thread(machine, 0, pid);
2510 if (thread == NULL)
2511 return -1;
2512
2513 tr = thread__get_runtime(thread);
2514 if (tr == NULL)
2515 return -1;
2516
2517 tr->migrations++;
2518
2519 /* show migrations if requested */
2520 timehist_print_migration_event(sched, evsel, sample, machine, thread);
2521
2522 return 0;
2523}
2524
2525static int timehist_sched_change_event(struct perf_tool *tool,
2526 union perf_event *event,
2527 struct evsel *evsel,
2528 struct perf_sample *sample,
2529 struct machine *machine)
2530{
2531 struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
2532 struct perf_time_interval *ptime = &sched->ptime;
2533 struct addr_location al;
2534 struct thread *thread;
2535 struct thread_runtime *tr = NULL;
2536 u64 tprev, t = sample->time;
2537 int rc = 0;
2538 int state = evsel__intval(evsel, sample, "prev_state");
2539
2540 if (machine__resolve(machine, &al, sample) < 0) {
2541 pr_err("problem processing %d event. skipping it\n",
2542 event->header.type);
2543 rc = -1;
2544 goto out;
2545 }
2546
2547 thread = timehist_get_thread(sched, sample, machine, evsel);
2548 if (thread == NULL) {
2549 rc = -1;
2550 goto out;
2551 }
2552
2553 if (timehist_skip_sample(sched, thread, evsel, sample))
2554 goto out;
2555
2556 tr = thread__get_runtime(thread);
2557 if (tr == NULL) {
2558 rc = -1;
2559 goto out;
2560 }
2561
2562 tprev = evsel__get_time(evsel, sample->cpu);
2563
2564 /*
2565 * If start time given:
2566 * - sample time is under window user cares about - skip sample
2567 * - tprev is under window user cares about - reset to start of window
2568 */
2569 if (ptime->start && ptime->start > t)
2570 goto out;
2571
2572 if (tprev && ptime->start > tprev)
2573 tprev = ptime->start;
2574
2575 /*
2576 * If end time given:
2577 * - previous sched event is out of window - we are done
2578 * - sample time is beyond window user cares about - reset it
2579 * to close out stats for time window interest
2580 */
2581 if (ptime->end) {
2582 if (tprev > ptime->end)
2583 goto out;
2584
2585 if (t > ptime->end)
2586 t = ptime->end;
2587 }
2588
2589 if (!sched->idle_hist || thread->tid == 0) {
2590 if (!cpu_list || test_bit(sample->cpu, cpu_bitmap))
2591 timehist_update_runtime_stats(tr, t, tprev);
2592
2593 if (sched->idle_hist) {
2594 struct idle_thread_runtime *itr = (void *)tr;
2595 struct thread_runtime *last_tr;
2596
2597 BUG_ON(thread->tid != 0);
2598
2599 if (itr->last_thread == NULL)
2600 goto out;
2601
2602 /* add current idle time as last thread's runtime */
2603 last_tr = thread__get_runtime(itr->last_thread);
2604 if (last_tr == NULL)
2605 goto out;
2606
2607 timehist_update_runtime_stats(last_tr, t, tprev);
2608 /*
2609 * remove delta time of last thread as it's not updated
2610 * and otherwise it will show an invalid value next
2611 * time. we only care total run time and run stat.
2612 */
2613 last_tr->dt_run = 0;
2614 last_tr->dt_delay = 0;
2615 last_tr->dt_sleep = 0;
2616 last_tr->dt_iowait = 0;
2617 last_tr->dt_preempt = 0;
2618
2619 if (itr->cursor.nr)
2620 callchain_append(&itr->callchain, &itr->cursor, t - tprev);
2621
2622 itr->last_thread = NULL;
2623 }
2624 }
2625
2626 if (!sched->summary_only)
2627 timehist_print_sample(sched, evsel, sample, &al, thread, t, state);
2628
2629out:
2630 if (sched->hist_time.start == 0 && t >= ptime->start)
2631 sched->hist_time.start = t;
2632 if (ptime->end == 0 || t <= ptime->end)
2633 sched->hist_time.end = t;
2634
2635 if (tr) {
2636 /* time of this sched_switch event becomes last time task seen */
2637 tr->last_time = sample->time;
2638
2639 /* last state is used to determine where to account wait time */
2640 tr->last_state = state;
2641
2642 /* sched out event for task so reset ready to run time */
2643 tr->ready_to_run = 0;
2644 }
2645
2646 evsel__save_time(evsel, sample->time, sample->cpu);
2647
2648 return rc;
2649}
2650
2651static int timehist_sched_switch_event(struct perf_tool *tool,
2652 union perf_event *event,
2653 struct evsel *evsel,
2654 struct perf_sample *sample,
2655 struct machine *machine __maybe_unused)
2656{
2657 return timehist_sched_change_event(tool, event, evsel, sample, machine);
2658}
2659
2660static int process_lost(struct perf_tool *tool __maybe_unused,
2661 union perf_event *event,
2662 struct perf_sample *sample,
2663 struct machine *machine __maybe_unused)
2664{
2665 char tstr[64];
2666
2667 timestamp__scnprintf_usec(sample->time, tstr, sizeof(tstr));
2668 printf("%15s ", tstr);
2669 printf("lost %" PRI_lu64 " events on cpu %d\n", event->lost.lost, sample->cpu);
2670
2671 return 0;
2672}
2673
2674
2675static void print_thread_runtime(struct thread *t,
2676 struct thread_runtime *r)
2677{
2678 double mean = avg_stats(&r->run_stats);
2679 float stddev;
2680
2681 printf("%*s %5d %9" PRIu64 " ",
2682 comm_width, timehist_get_commstr(t), t->ppid,
2683 (u64) r->run_stats.n);
2684
2685 print_sched_time(r->total_run_time, 8);
2686 stddev = rel_stddev_stats(stddev_stats(&r->run_stats), mean);
2687 print_sched_time(r->run_stats.min, 6);
2688 printf(" ");
2689 print_sched_time((u64) mean, 6);
2690 printf(" ");
2691 print_sched_time(r->run_stats.max, 6);
2692 printf(" ");
2693 printf("%5.2f", stddev);
2694 printf(" %5" PRIu64, r->migrations);
2695 printf("\n");
2696}
2697
2698static void print_thread_waittime(struct thread *t,
2699 struct thread_runtime *r)
2700{
2701 printf("%*s %5d %9" PRIu64 " ",
2702 comm_width, timehist_get_commstr(t), t->ppid,
2703 (u64) r->run_stats.n);
2704
2705 print_sched_time(r->total_run_time, 8);
2706 print_sched_time(r->total_sleep_time, 6);
2707 printf(" ");
2708 print_sched_time(r->total_iowait_time, 6);
2709 printf(" ");
2710 print_sched_time(r->total_preempt_time, 6);
2711 printf(" ");
2712 print_sched_time(r->total_delay_time, 6);
2713 printf("\n");
2714}
2715
2716struct total_run_stats {
2717 struct perf_sched *sched;
2718 u64 sched_count;
2719 u64 task_count;
2720 u64 total_run_time;
2721};
2722
2723static int __show_thread_runtime(struct thread *t, void *priv)
2724{
2725 struct total_run_stats *stats = priv;
2726 struct thread_runtime *r;
2727
2728 if (thread__is_filtered(t))
2729 return 0;
2730
2731 r = thread__priv(t);
2732 if (r && r->run_stats.n) {
2733 stats->task_count++;
2734 stats->sched_count += r->run_stats.n;
2735 stats->total_run_time += r->total_run_time;
2736
2737 if (stats->sched->show_state)
2738 print_thread_waittime(t, r);
2739 else
2740 print_thread_runtime(t, r);
2741 }
2742
2743 return 0;
2744}
2745
2746static int show_thread_runtime(struct thread *t, void *priv)
2747{
2748 if (t->dead)
2749 return 0;
2750
2751 return __show_thread_runtime(t, priv);
2752}
2753
2754static int show_deadthread_runtime(struct thread *t, void *priv)
2755{
2756 if (!t->dead)
2757 return 0;
2758
2759 return __show_thread_runtime(t, priv);
2760}
2761
2762static size_t callchain__fprintf_folded(FILE *fp, struct callchain_node *node)
2763{
2764 const char *sep = " <- ";
2765 struct callchain_list *chain;
2766 size_t ret = 0;
2767 char bf[1024];
2768 bool first;
2769
2770 if (node == NULL)
2771 return 0;
2772
2773 ret = callchain__fprintf_folded(fp, node->parent);
2774 first = (ret == 0);
2775
2776 list_for_each_entry(chain, &node->val, list) {
2777 if (chain->ip >= PERF_CONTEXT_MAX)
2778 continue;
2779 if (chain->ms.sym && chain->ms.sym->ignore)
2780 continue;
2781 ret += fprintf(fp, "%s%s", first ? "" : sep,
2782 callchain_list__sym_name(chain, bf, sizeof(bf),
2783 false));
2784 first = false;
2785 }
2786
2787 return ret;
2788}
2789
2790static size_t timehist_print_idlehist_callchain(struct rb_root_cached *root)
2791{
2792 size_t ret = 0;
2793 FILE *fp = stdout;
2794 struct callchain_node *chain;
2795 struct rb_node *rb_node = rb_first_cached(root);
2796
2797 printf(" %16s %8s %s\n", "Idle time (msec)", "Count", "Callchains");
2798 printf(" %.16s %.8s %.50s\n", graph_dotted_line, graph_dotted_line,
2799 graph_dotted_line);
2800
2801 while (rb_node) {
2802 chain = rb_entry(rb_node, struct callchain_node, rb_node);
2803 rb_node = rb_next(rb_node);
2804
2805 ret += fprintf(fp, " ");
2806 print_sched_time(chain->hit, 12);
2807 ret += 16; /* print_sched_time returns 2nd arg + 4 */
2808 ret += fprintf(fp, " %8d ", chain->count);
2809 ret += callchain__fprintf_folded(fp, chain);
2810 ret += fprintf(fp, "\n");
2811 }
2812
2813 return ret;
2814}
2815
2816static void timehist_print_summary(struct perf_sched *sched,
2817 struct perf_session *session)
2818{
2819 struct machine *m = &session->machines.host;
2820 struct total_run_stats totals;
2821 u64 task_count;
2822 struct thread *t;
2823 struct thread_runtime *r;
2824 int i;
2825 u64 hist_time = sched->hist_time.end - sched->hist_time.start;
2826
2827 memset(&totals, 0, sizeof(totals));
2828 totals.sched = sched;
2829
2830 if (sched->idle_hist) {
2831 printf("\nIdle-time summary\n");
2832 printf("%*s parent sched-out ", comm_width, "comm");
2833 printf(" idle-time min-idle avg-idle max-idle stddev migrations\n");
2834 } else if (sched->show_state) {
2835 printf("\nWait-time summary\n");
2836 printf("%*s parent sched-in ", comm_width, "comm");
2837 printf(" run-time sleep iowait preempt delay\n");
2838 } else {
2839 printf("\nRuntime summary\n");
2840 printf("%*s parent sched-in ", comm_width, "comm");
2841 printf(" run-time min-run avg-run max-run stddev migrations\n");
2842 }
2843 printf("%*s (count) ", comm_width, "");
2844 printf(" (msec) (msec) (msec) (msec) %s\n",
2845 sched->show_state ? "(msec)" : "%");
2846 printf("%.117s\n", graph_dotted_line);
2847
2848 machine__for_each_thread(m, show_thread_runtime, &totals);
2849 task_count = totals.task_count;
2850 if (!task_count)
2851 printf("<no still running tasks>\n");
2852
2853 printf("\nTerminated tasks:\n");
2854 machine__for_each_thread(m, show_deadthread_runtime, &totals);
2855 if (task_count == totals.task_count)
2856 printf("<no terminated tasks>\n");
2857
2858 /* CPU idle stats not tracked when samples were skipped */
2859 if (sched->skipped_samples && !sched->idle_hist)
2860 return;
2861
2862 printf("\nIdle stats:\n");
2863 for (i = 0; i < idle_max_cpu; ++i) {
2864 if (cpu_list && !test_bit(i, cpu_bitmap))
2865 continue;
2866
2867 t = idle_threads[i];
2868 if (!t)
2869 continue;
2870
2871 r = thread__priv(t);
2872 if (r && r->run_stats.n) {
2873 totals.sched_count += r->run_stats.n;
2874 printf(" CPU %2d idle for ", i);
2875 print_sched_time(r->total_run_time, 6);
2876 printf(" msec (%6.2f%%)\n", 100.0 * r->total_run_time / hist_time);
2877 } else
2878 printf(" CPU %2d idle entire time window\n", i);
2879 }
2880
2881 if (sched->idle_hist && sched->show_callchain) {
2882 callchain_param.mode = CHAIN_FOLDED;
2883 callchain_param.value = CCVAL_PERIOD;
2884
2885 callchain_register_param(&callchain_param);
2886
2887 printf("\nIdle stats by callchain:\n");
2888 for (i = 0; i < idle_max_cpu; ++i) {
2889 struct idle_thread_runtime *itr;
2890
2891 t = idle_threads[i];
2892 if (!t)
2893 continue;
2894
2895 itr = thread__priv(t);
2896 if (itr == NULL)
2897 continue;
2898
2899 callchain_param.sort(&itr->sorted_root.rb_root, &itr->callchain,
2900 0, &callchain_param);
2901
2902 printf(" CPU %2d:", i);
2903 print_sched_time(itr->tr.total_run_time, 6);
2904 printf(" msec\n");
2905 timehist_print_idlehist_callchain(&itr->sorted_root);
2906 printf("\n");
2907 }
2908 }
2909
2910 printf("\n"
2911 " Total number of unique tasks: %" PRIu64 "\n"
2912 "Total number of context switches: %" PRIu64 "\n",
2913 totals.task_count, totals.sched_count);
2914
2915 printf(" Total run time (msec): ");
2916 print_sched_time(totals.total_run_time, 2);
2917 printf("\n");
2918
2919 printf(" Total scheduling time (msec): ");
2920 print_sched_time(hist_time, 2);
2921 printf(" (x %d)\n", sched->max_cpu);
2922}
2923
2924typedef int (*sched_handler)(struct perf_tool *tool,
2925 union perf_event *event,
2926 struct evsel *evsel,
2927 struct perf_sample *sample,
2928 struct machine *machine);
2929
2930static int perf_timehist__process_sample(struct perf_tool *tool,
2931 union perf_event *event,
2932 struct perf_sample *sample,
2933 struct evsel *evsel,
2934 struct machine *machine)
2935{
2936 struct perf_sched *sched = container_of(tool, struct perf_sched, tool);
2937 int err = 0;
2938 int this_cpu = sample->cpu;
2939
2940 if (this_cpu > sched->max_cpu)
2941 sched->max_cpu = this_cpu;
2942
2943 if (evsel->handler != NULL) {
2944 sched_handler f = evsel->handler;
2945
2946 err = f(tool, event, evsel, sample, machine);
2947 }
2948
2949 return err;
2950}
2951
2952static int timehist_check_attr(struct perf_sched *sched,
2953 struct evlist *evlist)
2954{
2955 struct evsel *evsel;
2956 struct evsel_runtime *er;
2957
2958 list_for_each_entry(evsel, &evlist->core.entries, core.node) {
2959 er = evsel__get_runtime(evsel);
2960 if (er == NULL) {
2961 pr_err("Failed to allocate memory for evsel runtime data\n");
2962 return -1;
2963 }
2964
2965 if (sched->show_callchain && !evsel__has_callchain(evsel)) {
2966 pr_info("Samples do not have callchains.\n");
2967 sched->show_callchain = 0;
2968 symbol_conf.use_callchain = 0;
2969 }
2970 }
2971
2972 return 0;
2973}
2974
2975static int perf_sched__timehist(struct perf_sched *sched)
2976{
2977 struct evsel_str_handler handlers[] = {
2978 { "sched:sched_switch", timehist_sched_switch_event, },
2979 { "sched:sched_wakeup", timehist_sched_wakeup_event, },
2980 { "sched:sched_waking", timehist_sched_wakeup_event, },
2981 { "sched:sched_wakeup_new", timehist_sched_wakeup_event, },
2982 };
2983 const struct evsel_str_handler migrate_handlers[] = {
2984 { "sched:sched_migrate_task", timehist_migrate_task_event, },
2985 };
2986 struct perf_data data = {
2987 .path = input_name,
2988 .mode = PERF_DATA_MODE_READ,
2989 .force = sched->force,
2990 };
2991
2992 struct perf_session *session;
2993 struct evlist *evlist;
2994 int err = -1;
2995
2996 /*
2997 * event handlers for timehist option
2998 */
2999 sched->tool.sample = perf_timehist__process_sample;
3000 sched->tool.mmap = perf_event__process_mmap;
3001 sched->tool.comm = perf_event__process_comm;
3002 sched->tool.exit = perf_event__process_exit;
3003 sched->tool.fork = perf_event__process_fork;
3004 sched->tool.lost = process_lost;
3005 sched->tool.attr = perf_event__process_attr;
3006 sched->tool.tracing_data = perf_event__process_tracing_data;
3007 sched->tool.build_id = perf_event__process_build_id;
3008
3009 sched->tool.ordered_events = true;
3010 sched->tool.ordering_requires_timestamps = true;
3011
3012 symbol_conf.use_callchain = sched->show_callchain;
3013
3014 session = perf_session__new(&data, false, &sched->tool);
3015 if (IS_ERR(session))
3016 return PTR_ERR(session);
3017
3018 if (cpu_list) {
3019 err = perf_session__cpu_bitmap(session, cpu_list, cpu_bitmap);
3020 if (err < 0)
3021 goto out;
3022 }
3023
3024 evlist = session->evlist;
3025
3026 symbol__init(&session->header.env);
3027
3028 if (perf_time__parse_str(&sched->ptime, sched->time_str) != 0) {
3029 pr_err("Invalid time string\n");
3030 return -EINVAL;
3031 }
3032
3033 if (timehist_check_attr(sched, evlist) != 0)
3034 goto out;
3035
3036 setup_pager();
3037
3038 /* prefer sched_waking if it is captured */
3039 if (evlist__find_tracepoint_by_name(session->evlist, "sched:sched_waking"))
3040 handlers[1].handler = timehist_sched_wakeup_ignore;
3041
3042 /* setup per-evsel handlers */
3043 if (perf_session__set_tracepoints_handlers(session, handlers))
3044 goto out;
3045
3046 /* sched_switch event at a minimum needs to exist */
3047 if (!evlist__find_tracepoint_by_name(session->evlist, "sched:sched_switch")) {
3048 pr_err("No sched_switch events found. Have you run 'perf sched record'?\n");
3049 goto out;
3050 }
3051
3052 if (sched->show_migrations &&
3053 perf_session__set_tracepoints_handlers(session, migrate_handlers))
3054 goto out;
3055
3056 /* pre-allocate struct for per-CPU idle stats */
3057 sched->max_cpu = session->header.env.nr_cpus_online;
3058 if (sched->max_cpu == 0)
3059 sched->max_cpu = 4;
3060 if (init_idle_threads(sched->max_cpu))
3061 goto out;
3062
3063 /* summary_only implies summary option, but don't overwrite summary if set */
3064 if (sched->summary_only)
3065 sched->summary = sched->summary_only;
3066
3067 if (!sched->summary_only)
3068 timehist_header(sched);
3069
3070 err = perf_session__process_events(session);
3071 if (err) {
3072 pr_err("Failed to process events, error %d", err);
3073 goto out;
3074 }
3075
3076 sched->nr_events = evlist->stats.nr_events[0];
3077 sched->nr_lost_events = evlist->stats.total_lost;
3078 sched->nr_lost_chunks = evlist->stats.nr_events[PERF_RECORD_LOST];
3079
3080 if (sched->summary)
3081 timehist_print_summary(sched, session);
3082
3083out:
3084 free_idle_threads();
3085 perf_session__delete(session);
3086
3087 return err;
3088}
3089
3090
3091static void print_bad_events(struct perf_sched *sched)
3092{
3093 if (sched->nr_unordered_timestamps && sched->nr_timestamps) {
3094 printf(" INFO: %.3f%% unordered timestamps (%ld out of %ld)\n",
3095 (double)sched->nr_unordered_timestamps/(double)sched->nr_timestamps*100.0,
3096 sched->nr_unordered_timestamps, sched->nr_timestamps);
3097 }
3098 if (sched->nr_lost_events && sched->nr_events) {
3099 printf(" INFO: %.3f%% lost events (%ld out of %ld, in %ld chunks)\n",
3100 (double)sched->nr_lost_events/(double)sched->nr_events * 100.0,
3101 sched->nr_lost_events, sched->nr_events, sched->nr_lost_chunks);
3102 }
3103 if (sched->nr_context_switch_bugs && sched->nr_timestamps) {
3104 printf(" INFO: %.3f%% context switch bugs (%ld out of %ld)",
3105 (double)sched->nr_context_switch_bugs/(double)sched->nr_timestamps*100.0,
3106 sched->nr_context_switch_bugs, sched->nr_timestamps);
3107 if (sched->nr_lost_events)
3108 printf(" (due to lost events?)");
3109 printf("\n");
3110 }
3111}
3112
3113static void __merge_work_atoms(struct rb_root_cached *root, struct work_atoms *data)
3114{
3115 struct rb_node **new = &(root->rb_root.rb_node), *parent = NULL;
3116 struct work_atoms *this;
3117 const char *comm = thread__comm_str(data->thread), *this_comm;
3118 bool leftmost = true;
3119
3120 while (*new) {
3121 int cmp;
3122
3123 this = container_of(*new, struct work_atoms, node);
3124 parent = *new;
3125
3126 this_comm = thread__comm_str(this->thread);
3127 cmp = strcmp(comm, this_comm);
3128 if (cmp > 0) {
3129 new = &((*new)->rb_left);
3130 } else if (cmp < 0) {
3131 new = &((*new)->rb_right);
3132 leftmost = false;
3133 } else {
3134 this->num_merged++;
3135 this->total_runtime += data->total_runtime;
3136 this->nb_atoms += data->nb_atoms;
3137 this->total_lat += data->total_lat;
3138 list_splice(&data->work_list, &this->work_list);
3139 if (this->max_lat < data->max_lat) {
3140 this->max_lat = data->max_lat;
3141 this->max_lat_start = data->max_lat_start;
3142 this->max_lat_end = data->max_lat_end;
3143 }
3144 zfree(&data);
3145 return;
3146 }
3147 }
3148
3149 data->num_merged++;
3150 rb_link_node(&data->node, parent, new);
3151 rb_insert_color_cached(&data->node, root, leftmost);
3152}
3153
3154static void perf_sched__merge_lat(struct perf_sched *sched)
3155{
3156 struct work_atoms *data;
3157 struct rb_node *node;
3158
3159 if (sched->skip_merge)
3160 return;
3161
3162 while ((node = rb_first_cached(&sched->atom_root))) {
3163 rb_erase_cached(node, &sched->atom_root);
3164 data = rb_entry(node, struct work_atoms, node);
3165 __merge_work_atoms(&sched->merged_atom_root, data);
3166 }
3167}
3168
3169static int perf_sched__lat(struct perf_sched *sched)
3170{
3171 struct rb_node *next;
3172
3173 setup_pager();
3174
3175 if (perf_sched__read_events(sched))
3176 return -1;
3177
3178 perf_sched__merge_lat(sched);
3179 perf_sched__sort_lat(sched);
3180
3181 printf("\n -------------------------------------------------------------------------------------------------------------------------------------------\n");
3182 printf(" Task | Runtime ms | Switches | Avg delay ms | Max delay ms | Max delay start | Max delay end |\n");
3183 printf(" -------------------------------------------------------------------------------------------------------------------------------------------\n");
3184
3185 next = rb_first_cached(&sched->sorted_atom_root);
3186
3187 while (next) {
3188 struct work_atoms *work_list;
3189
3190 work_list = rb_entry(next, struct work_atoms, node);
3191 output_lat_thread(sched, work_list);
3192 next = rb_next(next);
3193 thread__zput(work_list->thread);
3194 }
3195
3196 printf(" -----------------------------------------------------------------------------------------------------------------\n");
3197 printf(" TOTAL: |%11.3f ms |%9" PRIu64 " |\n",
3198 (double)sched->all_runtime / NSEC_PER_MSEC, sched->all_count);
3199
3200 printf(" ---------------------------------------------------\n");
3201
3202 print_bad_events(sched);
3203 printf("\n");
3204
3205 return 0;
3206}
3207
3208static int setup_map_cpus(struct perf_sched *sched)
3209{
3210 struct perf_cpu_map *map;
3211
3212 sched->max_cpu = sysconf(_SC_NPROCESSORS_CONF);
3213
3214 if (sched->map.comp) {
3215 sched->map.comp_cpus = zalloc(sched->max_cpu * sizeof(int));
3216 if (!sched->map.comp_cpus)
3217 return -1;
3218 }
3219
3220 if (!sched->map.cpus_str)
3221 return 0;
3222
3223 map = perf_cpu_map__new(sched->map.cpus_str);
3224 if (!map) {
3225 pr_err("failed to get cpus map from %s\n", sched->map.cpus_str);
3226 return -1;
3227 }
3228
3229 sched->map.cpus = map;
3230 return 0;
3231}
3232
3233static int setup_color_pids(struct perf_sched *sched)
3234{
3235 struct perf_thread_map *map;
3236
3237 if (!sched->map.color_pids_str)
3238 return 0;
3239
3240 map = thread_map__new_by_tid_str(sched->map.color_pids_str);
3241 if (!map) {
3242 pr_err("failed to get thread map from %s\n", sched->map.color_pids_str);
3243 return -1;
3244 }
3245
3246 sched->map.color_pids = map;
3247 return 0;
3248}
3249
3250static int setup_color_cpus(struct perf_sched *sched)
3251{
3252 struct perf_cpu_map *map;
3253
3254 if (!sched->map.color_cpus_str)
3255 return 0;
3256
3257 map = perf_cpu_map__new(sched->map.color_cpus_str);
3258 if (!map) {
3259 pr_err("failed to get thread map from %s\n", sched->map.color_cpus_str);
3260 return -1;
3261 }
3262
3263 sched->map.color_cpus = map;
3264 return 0;
3265}
3266
3267static int perf_sched__map(struct perf_sched *sched)
3268{
3269 if (setup_map_cpus(sched))
3270 return -1;
3271
3272 if (setup_color_pids(sched))
3273 return -1;
3274
3275 if (setup_color_cpus(sched))
3276 return -1;
3277
3278 setup_pager();
3279 if (perf_sched__read_events(sched))
3280 return -1;
3281 print_bad_events(sched);
3282 return 0;
3283}
3284
3285static int perf_sched__replay(struct perf_sched *sched)
3286{
3287 unsigned long i;
3288
3289 calibrate_run_measurement_overhead(sched);
3290 calibrate_sleep_measurement_overhead(sched);
3291
3292 test_calibrations(sched);
3293
3294 if (perf_sched__read_events(sched))
3295 return -1;
3296
3297 printf("nr_run_events: %ld\n", sched->nr_run_events);
3298 printf("nr_sleep_events: %ld\n", sched->nr_sleep_events);
3299 printf("nr_wakeup_events: %ld\n", sched->nr_wakeup_events);
3300
3301 if (sched->targetless_wakeups)
3302 printf("target-less wakeups: %ld\n", sched->targetless_wakeups);
3303 if (sched->multitarget_wakeups)
3304 printf("multi-target wakeups: %ld\n", sched->multitarget_wakeups);
3305 if (sched->nr_run_events_optimized)
3306 printf("run atoms optimized: %ld\n",
3307 sched->nr_run_events_optimized);
3308
3309 print_task_traces(sched);
3310 add_cross_task_wakeups(sched);
3311
3312 create_tasks(sched);
3313 printf("------------------------------------------------------------\n");
3314 for (i = 0; i < sched->replay_repeat; i++)
3315 run_one_test(sched);
3316
3317 return 0;
3318}
3319
3320static void setup_sorting(struct perf_sched *sched, const struct option *options,
3321 const char * const usage_msg[])
3322{
3323 char *tmp, *tok, *str = strdup(sched->sort_order);
3324
3325 for (tok = strtok_r(str, ", ", &tmp);
3326 tok; tok = strtok_r(NULL, ", ", &tmp)) {
3327 if (sort_dimension__add(tok, &sched->sort_list) < 0) {
3328 usage_with_options_msg(usage_msg, options,
3329 "Unknown --sort key: `%s'", tok);
3330 }
3331 }
3332
3333 free(str);
3334
3335 sort_dimension__add("pid", &sched->cmp_pid);
3336}
3337
3338static bool schedstat_events_exposed(void)
3339{
3340 /*
3341 * Select "sched:sched_stat_wait" event to check
3342 * whether schedstat tracepoints are exposed.
3343 */
3344 return IS_ERR(trace_event__tp_format("sched", "sched_stat_wait")) ?
3345 false : true;
3346}
3347
3348static int __cmd_record(int argc, const char **argv)
3349{
3350 unsigned int rec_argc, i, j;
3351 const char **rec_argv;
3352 const char * const record_args[] = {
3353 "record",
3354 "-a",
3355 "-R",
3356 "-m", "1024",
3357 "-c", "1",
3358 "-e", "sched:sched_switch",
3359 "-e", "sched:sched_stat_runtime",
3360 "-e", "sched:sched_process_fork",
3361 "-e", "sched:sched_wakeup_new",
3362 "-e", "sched:sched_migrate_task",
3363 };
3364
3365 /*
3366 * The tracepoints trace_sched_stat_{wait, sleep, iowait}
3367 * are not exposed to user if CONFIG_SCHEDSTATS is not set,
3368 * to prevent "perf sched record" execution failure, determine
3369 * whether to record schedstat events according to actual situation.
3370 */
3371 const char * const schedstat_args[] = {
3372 "-e", "sched:sched_stat_wait",
3373 "-e", "sched:sched_stat_sleep",
3374 "-e", "sched:sched_stat_iowait",
3375 };
3376 unsigned int schedstat_argc = schedstat_events_exposed() ?
3377 ARRAY_SIZE(schedstat_args) : 0;
3378
3379 struct tep_event *waking_event;
3380
3381 /*
3382 * +2 for either "-e", "sched:sched_wakeup" or
3383 * "-e", "sched:sched_waking"
3384 */
3385 rec_argc = ARRAY_SIZE(record_args) + 2 + schedstat_argc + argc - 1;
3386 rec_argv = calloc(rec_argc + 1, sizeof(char *));
3387
3388 if (rec_argv == NULL)
3389 return -ENOMEM;
3390
3391 for (i = 0; i < ARRAY_SIZE(record_args); i++)
3392 rec_argv[i] = strdup(record_args[i]);
3393
3394 rec_argv[i++] = "-e";
3395 waking_event = trace_event__tp_format("sched", "sched_waking");
3396 if (!IS_ERR(waking_event))
3397 rec_argv[i++] = strdup("sched:sched_waking");
3398 else
3399 rec_argv[i++] = strdup("sched:sched_wakeup");
3400
3401 for (j = 0; j < schedstat_argc; j++)
3402 rec_argv[i++] = strdup(schedstat_args[j]);
3403
3404 for (j = 1; j < (unsigned int)argc; j++, i++)
3405 rec_argv[i] = argv[j];
3406
3407 BUG_ON(i != rec_argc);
3408
3409 return cmd_record(i, rec_argv);
3410}
3411
3412int cmd_sched(int argc, const char **argv)
3413{
3414 static const char default_sort_order[] = "avg, max, switch, runtime";
3415 struct perf_sched sched = {
3416 .tool = {
3417 .sample = perf_sched__process_tracepoint_sample,
3418 .comm = perf_sched__process_comm,
3419 .namespaces = perf_event__process_namespaces,
3420 .lost = perf_event__process_lost,
3421 .fork = perf_sched__process_fork_event,
3422 .ordered_events = true,
3423 },
3424 .cmp_pid = LIST_HEAD_INIT(sched.cmp_pid),
3425 .sort_list = LIST_HEAD_INIT(sched.sort_list),
3426 .start_work_mutex = PTHREAD_MUTEX_INITIALIZER,
3427 .work_done_wait_mutex = PTHREAD_MUTEX_INITIALIZER,
3428 .sort_order = default_sort_order,
3429 .replay_repeat = 10,
3430 .profile_cpu = -1,
3431 .next_shortname1 = 'A',
3432 .next_shortname2 = '0',
3433 .skip_merge = 0,
3434 .show_callchain = 1,
3435 .max_stack = 5,
3436 };
3437 const struct option sched_options[] = {
3438 OPT_STRING('i', "input", &input_name, "file",
3439 "input file name"),
3440 OPT_INCR('v', "verbose", &verbose,
3441 "be more verbose (show symbol address, etc)"),
3442 OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
3443 "dump raw trace in ASCII"),
3444 OPT_BOOLEAN('f', "force", &sched.force, "don't complain, do it"),
3445 OPT_END()
3446 };
3447 const struct option latency_options[] = {
3448 OPT_STRING('s', "sort", &sched.sort_order, "key[,key2...]",
3449 "sort by key(s): runtime, switch, avg, max"),
3450 OPT_INTEGER('C', "CPU", &sched.profile_cpu,
3451 "CPU to profile on"),
3452 OPT_BOOLEAN('p', "pids", &sched.skip_merge,
3453 "latency stats per pid instead of per comm"),
3454 OPT_PARENT(sched_options)
3455 };
3456 const struct option replay_options[] = {
3457 OPT_UINTEGER('r', "repeat", &sched.replay_repeat,
3458 "repeat the workload replay N times (-1: infinite)"),
3459 OPT_PARENT(sched_options)
3460 };
3461 const struct option map_options[] = {
3462 OPT_BOOLEAN(0, "compact", &sched.map.comp,
3463 "map output in compact mode"),
3464 OPT_STRING(0, "color-pids", &sched.map.color_pids_str, "pids",
3465 "highlight given pids in map"),
3466 OPT_STRING(0, "color-cpus", &sched.map.color_cpus_str, "cpus",
3467 "highlight given CPUs in map"),
3468 OPT_STRING(0, "cpus", &sched.map.cpus_str, "cpus",
3469 "display given CPUs in map"),
3470 OPT_PARENT(sched_options)
3471 };
3472 const struct option timehist_options[] = {
3473 OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name,
3474 "file", "vmlinux pathname"),
3475 OPT_STRING(0, "kallsyms", &symbol_conf.kallsyms_name,
3476 "file", "kallsyms pathname"),
3477 OPT_BOOLEAN('g', "call-graph", &sched.show_callchain,
3478 "Display call chains if present (default on)"),
3479 OPT_UINTEGER(0, "max-stack", &sched.max_stack,
3480 "Maximum number of functions to display backtrace."),
3481 OPT_STRING(0, "symfs", &symbol_conf.symfs, "directory",
3482 "Look for files with symbols relative to this directory"),
3483 OPT_BOOLEAN('s', "summary", &sched.summary_only,
3484 "Show only syscall summary with statistics"),
3485 OPT_BOOLEAN('S', "with-summary", &sched.summary,
3486 "Show all syscalls and summary with statistics"),
3487 OPT_BOOLEAN('w', "wakeups", &sched.show_wakeups, "Show wakeup events"),
3488 OPT_BOOLEAN('n', "next", &sched.show_next, "Show next task"),
3489 OPT_BOOLEAN('M', "migrations", &sched.show_migrations, "Show migration events"),
3490 OPT_BOOLEAN('V', "cpu-visual", &sched.show_cpu_visual, "Add CPU visual"),
3491 OPT_BOOLEAN('I', "idle-hist", &sched.idle_hist, "Show idle events only"),
3492 OPT_STRING(0, "time", &sched.time_str, "str",
3493 "Time span for analysis (start,stop)"),
3494 OPT_BOOLEAN(0, "state", &sched.show_state, "Show task state when sched-out"),
3495 OPT_STRING('p', "pid", &symbol_conf.pid_list_str, "pid[,pid...]",
3496 "analyze events only for given process id(s)"),
3497 OPT_STRING('t', "tid", &symbol_conf.tid_list_str, "tid[,tid...]",
3498 "analyze events only for given thread id(s)"),
3499 OPT_STRING('C', "cpu", &cpu_list, "cpu", "list of cpus to profile"),
3500 OPT_PARENT(sched_options)
3501 };
3502
3503 const char * const latency_usage[] = {
3504 "perf sched latency [<options>]",
3505 NULL
3506 };
3507 const char * const replay_usage[] = {
3508 "perf sched replay [<options>]",
3509 NULL
3510 };
3511 const char * const map_usage[] = {
3512 "perf sched map [<options>]",
3513 NULL
3514 };
3515 const char * const timehist_usage[] = {
3516 "perf sched timehist [<options>]",
3517 NULL
3518 };
3519 const char *const sched_subcommands[] = { "record", "latency", "map",
3520 "replay", "script",
3521 "timehist", NULL };
3522 const char *sched_usage[] = {
3523 NULL,
3524 NULL
3525 };
3526 struct trace_sched_handler lat_ops = {
3527 .wakeup_event = latency_wakeup_event,
3528 .switch_event = latency_switch_event,
3529 .runtime_event = latency_runtime_event,
3530 .migrate_task_event = latency_migrate_task_event,
3531 };
3532 struct trace_sched_handler map_ops = {
3533 .switch_event = map_switch_event,
3534 };
3535 struct trace_sched_handler replay_ops = {
3536 .wakeup_event = replay_wakeup_event,
3537 .switch_event = replay_switch_event,
3538 .fork_event = replay_fork_event,
3539 };
3540 unsigned int i;
3541
3542 for (i = 0; i < ARRAY_SIZE(sched.curr_pid); i++)
3543 sched.curr_pid[i] = -1;
3544
3545 argc = parse_options_subcommand(argc, argv, sched_options, sched_subcommands,
3546 sched_usage, PARSE_OPT_STOP_AT_NON_OPTION);
3547 if (!argc)
3548 usage_with_options(sched_usage, sched_options);
3549
3550 /*
3551 * Aliased to 'perf script' for now:
3552 */
3553 if (!strcmp(argv[0], "script"))
3554 return cmd_script(argc, argv);
3555
3556 if (!strncmp(argv[0], "rec", 3)) {
3557 return __cmd_record(argc, argv);
3558 } else if (!strncmp(argv[0], "lat", 3)) {
3559 sched.tp_handler = &lat_ops;
3560 if (argc > 1) {
3561 argc = parse_options(argc, argv, latency_options, latency_usage, 0);
3562 if (argc)
3563 usage_with_options(latency_usage, latency_options);
3564 }
3565 setup_sorting(&sched, latency_options, latency_usage);
3566 return perf_sched__lat(&sched);
3567 } else if (!strcmp(argv[0], "map")) {
3568 if (argc) {
3569 argc = parse_options(argc, argv, map_options, map_usage, 0);
3570 if (argc)
3571 usage_with_options(map_usage, map_options);
3572 }
3573 sched.tp_handler = &map_ops;
3574 setup_sorting(&sched, latency_options, latency_usage);
3575 return perf_sched__map(&sched);
3576 } else if (!strncmp(argv[0], "rep", 3)) {
3577 sched.tp_handler = &replay_ops;
3578 if (argc) {
3579 argc = parse_options(argc, argv, replay_options, replay_usage, 0);
3580 if (argc)
3581 usage_with_options(replay_usage, replay_options);
3582 }
3583 return perf_sched__replay(&sched);
3584 } else if (!strcmp(argv[0], "timehist")) {
3585 if (argc) {
3586 argc = parse_options(argc, argv, timehist_options,
3587 timehist_usage, 0);
3588 if (argc)
3589 usage_with_options(timehist_usage, timehist_options);
3590 }
3591 if ((sched.show_wakeups || sched.show_next) &&
3592 sched.summary_only) {
3593 pr_err(" Error: -s and -[n|w] are mutually exclusive.\n");
3594 parse_options_usage(timehist_usage, timehist_options, "s", true);
3595 if (sched.show_wakeups)
3596 parse_options_usage(NULL, timehist_options, "w", true);
3597 if (sched.show_next)
3598 parse_options_usage(NULL, timehist_options, "n", true);
3599 return -EINVAL;
3600 }
3601
3602 return perf_sched__timehist(&sched);
3603 } else {
3604 usage_with_options(sched_usage, sched_options);
3605 }
3606
3607 return 0;
3608}