?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
hwmon: (w83627ehf) Optimize multi-bank register access
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / tools / perf / builtin-sched.c
blob29acb894e035154432c708e1bf6ae7b7db5c2be1
1 #include "builtin.h"
2 #include "perf.h"
3
4 #include "util/util.h"
5 #include "util/cache.h"
6 #include "util/symbol.h"
7 #include "util/thread.h"
8 #include "util/header.h"
9 #include "util/session.h"
10
11 #include "util/parse-options.h"
12 #include "util/trace-event.h"
13
14 #include "util/debug.h"
15
16 #include <sys/prctl.h>
17
18 #include <semaphore.h>
19 #include <pthread.h>
20 #include <math.h>
21
22 static char const *input_name = "perf.data";
23
24 static char default_sort_order[] = "avg, max, switch, runtime";
25 static const char *sort_order = default_sort_order;
26
27 static int profile_cpu = -1;
28
29 #define PR_SET_NAME 15 /* Set process name */
30 #define MAX_CPUS 4096
31
32 static u64 run_measurement_overhead;
33 static u64 sleep_measurement_overhead;
34
35 #define COMM_LEN 20
36 #define SYM_LEN 129
37
38 #define MAX_PID 65536
39
40 static unsigned long nr_tasks;
41
42 struct sched_atom;
43
44 struct task_desc {
45 unsigned long nr;
46 unsigned long pid;
47 char comm[COMM_LEN];
48
49 unsigned long nr_events;
50 unsigned long curr_event;
51 struct sched_atom **atoms;
52
53 pthread_t thread;
54 sem_t sleep_sem;
55
56 sem_t ready_for_work;
57 sem_t work_done_sem;
58
59 u64 cpu_usage;
60 };
61
62 enum sched_event_type {
63 SCHED_EVENT_RUN,
64 SCHED_EVENT_SLEEP,
65 SCHED_EVENT_WAKEUP,
66 SCHED_EVENT_MIGRATION,
67 };
68
69 struct sched_atom {
70 enum sched_event_type type;
71 int specific_wait;
72 u64 timestamp;
73 u64 duration;
74 unsigned long nr;
75 sem_t *wait_sem;
76 struct task_desc *wakee;
77 };
78
79 static struct task_desc *pid_to_task[MAX_PID];
80
81 static struct task_desc **tasks;
82
83 static pthread_mutex_t start_work_mutex = PTHREAD_MUTEX_INITIALIZER;
84 static u64 start_time;
85
86 static pthread_mutex_t work_done_wait_mutex = PTHREAD_MUTEX_INITIALIZER;
87
88 static unsigned long nr_run_events;
89 static unsigned long nr_sleep_events;
90 static unsigned long nr_wakeup_events;
91
92 static unsigned long nr_sleep_corrections;
93 static unsigned long nr_run_events_optimized;
94
95 static unsigned long targetless_wakeups;
96 static unsigned long multitarget_wakeups;
97
98 static u64 cpu_usage;
99 static u64 runavg_cpu_usage;
100 static u64 parent_cpu_usage;
101 static u64 runavg_parent_cpu_usage;
102
103 static unsigned long nr_runs;
104 static u64 sum_runtime;
105 static u64 sum_fluct;
106 static u64 run_avg;
107
108 static unsigned int replay_repeat = 10;
109 static unsigned long nr_timestamps;
110 static unsigned long nr_unordered_timestamps;
111 static unsigned long nr_state_machine_bugs;
112 static unsigned long nr_context_switch_bugs;
113 static unsigned long nr_events;
114 static unsigned long nr_lost_chunks;
115 static unsigned long nr_lost_events;
116
117 #define TASK_STATE_TO_CHAR_STR "RSDTtZX"
118
119 enum thread_state {
120 THREAD_SLEEPING = 0,
121 THREAD_WAIT_CPU,
122 THREAD_SCHED_IN,
123 THREAD_IGNORE
124 };
125
126 struct work_atom {
127 struct list_head list;
128 enum thread_state state;
129 u64 sched_out_time;
130 u64 wake_up_time;
131 u64 sched_in_time;
132 u64 runtime;
133 };
134
135 struct work_atoms {
136 struct list_head work_list;
137 struct thread *thread;
138 struct rb_node node;
139 u64 max_lat;
140 u64 max_lat_at;
141 u64 total_lat;
142 u64 nb_atoms;
143 u64 total_runtime;
144 };
145
146 typedef int (*sort_fn_t)(struct work_atoms *, struct work_atoms *);
147
148 static struct rb_root atom_root, sorted_atom_root;
149
150 static u64 all_runtime;
151 static u64 all_count;
152
153
154 static u64 get_nsecs(void)
155 {
156 struct timespec ts;
157
158 clock_gettime(CLOCK_MONOTONIC, &ts);
159
160 return ts.tv_sec * 1000000000ULL + ts.tv_nsec;
161 }
162
163 static void burn_nsecs(u64 nsecs)
164 {
165 u64 T0 = get_nsecs(), T1;
166
167 do {
168 T1 = get_nsecs();
169 } while (T1 + run_measurement_overhead < T0 + nsecs);
170 }
171
172 static void sleep_nsecs(u64 nsecs)
173 {
174 struct timespec ts;
175
176 ts.tv_nsec = nsecs % 999999999;
177 ts.tv_sec = nsecs / 999999999;
178
179 nanosleep(&ts, NULL);
180 }
181
182 static void calibrate_run_measurement_overhead(void)
183 {
184 u64 T0, T1, delta, min_delta = 1000000000ULL;
185 int i;
186
187 for (i = 0; i < 10; i++) {
188 T0 = get_nsecs();
189 burn_nsecs(0);
190 T1 = get_nsecs();
191 delta = T1-T0;
192 min_delta = min(min_delta, delta);
193 }
194 run_measurement_overhead = min_delta;
195
196 printf("run measurement overhead: %" PRIu64 " nsecs\n", min_delta);
197 }
198
199 static void calibrate_sleep_measurement_overhead(void)
200 {
201 u64 T0, T1, delta, min_delta = 1000000000ULL;
202 int i;
203
204 for (i = 0; i < 10; i++) {
205 T0 = get_nsecs();
206 sleep_nsecs(10000);
207 T1 = get_nsecs();
208 delta = T1-T0;
209 min_delta = min(min_delta, delta);
210 }
211 min_delta -= 10000;
212 sleep_measurement_overhead = min_delta;
213
214 printf("sleep measurement overhead: %" PRIu64 " nsecs\n", min_delta);
215 }
216
217 static struct sched_atom *
218 get_new_event(struct task_desc *task, u64 timestamp)
219 {
220 struct sched_atom *event = zalloc(sizeof(*event));
221 unsigned long idx = task->nr_events;
222 size_t size;
223
224 event->timestamp = timestamp;
225 event->nr = idx;
226
227 task->nr_events++;
228 size = sizeof(struct sched_atom *) * task->nr_events;
229 task->atoms = realloc(task->atoms, size);
230 BUG_ON(!task->atoms);
231
232 task->atoms[idx] = event;
233
234 return event;
235 }
236
237 static struct sched_atom *last_event(struct task_desc *task)
238 {
239 if (!task->nr_events)
240 return NULL;
241
242 return task->atoms[task->nr_events - 1];
243 }
244
245 static void
246 add_sched_event_run(struct task_desc *task, u64 timestamp, u64 duration)
247 {
248 struct sched_atom *event, *curr_event = last_event(task);
249
250 /*
251 * optimize an existing RUN event by merging this one
252 * to it:
253 */
254 if (curr_event && curr_event->type == SCHED_EVENT_RUN) {
255 nr_run_events_optimized++;
256 curr_event->duration += duration;
257 return;
258 }
259
260 event = get_new_event(task, timestamp);
261
262 event->type = SCHED_EVENT_RUN;
263 event->duration = duration;
264
265 nr_run_events++;
266 }
267
268 static void
269 add_sched_event_wakeup(struct task_desc *task, u64 timestamp,
270 struct task_desc *wakee)
271 {
272 struct sched_atom *event, *wakee_event;
273
274 event = get_new_event(task, timestamp);
275 event->type = SCHED_EVENT_WAKEUP;
276 event->wakee = wakee;
277
278 wakee_event = last_event(wakee);
279 if (!wakee_event || wakee_event->type != SCHED_EVENT_SLEEP) {
280 targetless_wakeups++;
281 return;
282 }
283 if (wakee_event->wait_sem) {
284 multitarget_wakeups++;
285 return;
286 }
287
288 wakee_event->wait_sem = zalloc(sizeof(*wakee_event->wait_sem));
289 sem_init(wakee_event->wait_sem, 0, 0);
290 wakee_event->specific_wait = 1;
291 event->wait_sem = wakee_event->wait_sem;
292
293 nr_wakeup_events++;
294 }
295
296 static void
297 add_sched_event_sleep(struct task_desc *task, u64 timestamp,
298 u64 task_state __used)
299 {
300 struct sched_atom *event = get_new_event(task, timestamp);
301
302 event->type = SCHED_EVENT_SLEEP;
303
304 nr_sleep_events++;
305 }
306
307 static struct task_desc *register_pid(unsigned long pid, const char *comm)
308 {
309 struct task_desc *task;
310
311 BUG_ON(pid >= MAX_PID);
312
313 task = pid_to_task[pid];
314
315 if (task)
316 return task;
317
318 task = zalloc(sizeof(*task));
319 task->pid = pid;
320 task->nr = nr_tasks;
321 strcpy(task->comm, comm);
322 /*
323 * every task starts in sleeping state - this gets ignored
324 * if there's no wakeup pointing to this sleep state:
325 */
326 add_sched_event_sleep(task, 0, 0);
327
328 pid_to_task[pid] = task;
329 nr_tasks++;
330 tasks = realloc(tasks, nr_tasks*sizeof(struct task_task *));
331 BUG_ON(!tasks);
332 tasks[task->nr] = task;
333
334 if (verbose)
335 printf("registered task #%ld, PID %ld (%s)\n", nr_tasks, pid, comm);
336
337 return task;
338 }
339
340
341 static void print_task_traces(void)
342 {
343 struct task_desc *task;
344 unsigned long i;
345
346 for (i = 0; i < nr_tasks; i++) {
347 task = tasks[i];
348 printf("task %6ld (%20s:%10ld), nr_events: %ld\n",
349 task->nr, task->comm, task->pid, task->nr_events);
350 }
351 }
352
353 static void add_cross_task_wakeups(void)
354 {
355 struct task_desc *task1, *task2;
356 unsigned long i, j;
357
358 for (i = 0; i < nr_tasks; i++) {
359 task1 = tasks[i];
360 j = i + 1;
361 if (j == nr_tasks)
362 j = 0;
363 task2 = tasks[j];
364 add_sched_event_wakeup(task1, 0, task2);
365 }
366 }
367
368 static void
369 process_sched_event(struct task_desc *this_task __used, struct sched_atom *atom)
370 {
371 int ret = 0;
372 u64 now;
373 long long delta;
374
375 now = get_nsecs();
376 delta = start_time + atom->timestamp - now;
377
378 switch (atom->type) {
379 case SCHED_EVENT_RUN:
380 burn_nsecs(atom->duration);
381 break;
382 case SCHED_EVENT_SLEEP:
383 if (atom->wait_sem)
384 ret = sem_wait(atom->wait_sem);
385 BUG_ON(ret);
386 break;
387 case SCHED_EVENT_WAKEUP:
388 if (atom->wait_sem)
389 ret = sem_post(atom->wait_sem);
390 BUG_ON(ret);
391 break;
392 case SCHED_EVENT_MIGRATION:
393 break;
394 default:
395 BUG_ON(1);
396 }
397 }
398
399 static u64 get_cpu_usage_nsec_parent(void)
400 {
401 struct rusage ru;
402 u64 sum;
403 int err;
404
405 err = getrusage(RUSAGE_SELF, &ru);
406 BUG_ON(err);
407
408 sum = ru.ru_utime.tv_sec*1e9 + ru.ru_utime.tv_usec*1e3;
409 sum += ru.ru_stime.tv_sec*1e9 + ru.ru_stime.tv_usec*1e3;
410
411 return sum;
412 }
413
414 static int self_open_counters(void)
415 {
416 struct perf_event_attr attr;
417 int fd;
418
419 memset(&attr, 0, sizeof(attr));
420
421 attr.type = PERF_TYPE_SOFTWARE;
422 attr.config = PERF_COUNT_SW_TASK_CLOCK;
423
424 fd = sys_perf_event_open(&attr, 0, -1, -1, 0);
425
426 if (fd < 0)
427 die("Error: sys_perf_event_open() syscall returned"
428 "with %d (%s)\n", fd, strerror(errno));
429 return fd;
430 }
431
432 static u64 get_cpu_usage_nsec_self(int fd)
433 {
434 u64 runtime;
435 int ret;
436
437 ret = read(fd, &runtime, sizeof(runtime));
438 BUG_ON(ret != sizeof(runtime));
439
440 return runtime;
441 }
442
443 static void *thread_func(void *ctx)
444 {
445 struct task_desc *this_task = ctx;
446 u64 cpu_usage_0, cpu_usage_1;
447 unsigned long i, ret;
448 char comm2[22];
449 int fd;
450
451 sprintf(comm2, ":%s", this_task->comm);
452 prctl(PR_SET_NAME, comm2);
453 fd = self_open_counters();
454
455 again:
456 ret = sem_post(&this_task->ready_for_work);
457 BUG_ON(ret);
458 ret = pthread_mutex_lock(&start_work_mutex);
459 BUG_ON(ret);
460 ret = pthread_mutex_unlock(&start_work_mutex);
461 BUG_ON(ret);
462
463 cpu_usage_0 = get_cpu_usage_nsec_self(fd);
464
465 for (i = 0; i < this_task->nr_events; i++) {
466 this_task->curr_event = i;
467 process_sched_event(this_task, this_task->atoms[i]);
468 }
469
470 cpu_usage_1 = get_cpu_usage_nsec_self(fd);
471 this_task->cpu_usage = cpu_usage_1 - cpu_usage_0;
472 ret = sem_post(&this_task->work_done_sem);
473 BUG_ON(ret);
474
475 ret = pthread_mutex_lock(&work_done_wait_mutex);
476 BUG_ON(ret);
477 ret = pthread_mutex_unlock(&work_done_wait_mutex);
478 BUG_ON(ret);
479
480 goto again;
481 }
482
483 static void create_tasks(void)
484 {
485 struct task_desc *task;
486 pthread_attr_t attr;
487 unsigned long i;
488 int err;
489
490 err = pthread_attr_init(&attr);
491 BUG_ON(err);
492 err = pthread_attr_setstacksize(&attr,
493 (size_t) max(16 * 1024, PTHREAD_STACK_MIN));
494 BUG_ON(err);
495 err = pthread_mutex_lock(&start_work_mutex);
496 BUG_ON(err);
497 err = pthread_mutex_lock(&work_done_wait_mutex);
498 BUG_ON(err);
499 for (i = 0; i < nr_tasks; i++) {
500 task = tasks[i];
501 sem_init(&task->sleep_sem, 0, 0);
502 sem_init(&task->ready_for_work, 0, 0);
503 sem_init(&task->work_done_sem, 0, 0);
504 task->curr_event = 0;
505 err = pthread_create(&task->thread, &attr, thread_func, task);
506 BUG_ON(err);
507 }
508 }
509
510 static void wait_for_tasks(void)
511 {
512 u64 cpu_usage_0, cpu_usage_1;
513 struct task_desc *task;
514 unsigned long i, ret;
515
516 start_time = get_nsecs();
517 cpu_usage = 0;
518 pthread_mutex_unlock(&work_done_wait_mutex);
519
520 for (i = 0; i < nr_tasks; i++) {
521 task = tasks[i];
522 ret = sem_wait(&task->ready_for_work);
523 BUG_ON(ret);
524 sem_init(&task->ready_for_work, 0, 0);
525 }
526 ret = pthread_mutex_lock(&work_done_wait_mutex);
527 BUG_ON(ret);
528
529 cpu_usage_0 = get_cpu_usage_nsec_parent();
530
531 pthread_mutex_unlock(&start_work_mutex);
532
533 for (i = 0; i < nr_tasks; i++) {
534 task = tasks[i];
535 ret = sem_wait(&task->work_done_sem);
536 BUG_ON(ret);
537 sem_init(&task->work_done_sem, 0, 0);
538 cpu_usage += task->cpu_usage;
539 task->cpu_usage = 0;
540 }
541
542 cpu_usage_1 = get_cpu_usage_nsec_parent();
543 if (!runavg_cpu_usage)
544 runavg_cpu_usage = cpu_usage;
545 runavg_cpu_usage = (runavg_cpu_usage*9 + cpu_usage)/10;
546
547 parent_cpu_usage = cpu_usage_1 - cpu_usage_0;
548 if (!runavg_parent_cpu_usage)
549 runavg_parent_cpu_usage = parent_cpu_usage;
550 runavg_parent_cpu_usage = (runavg_parent_cpu_usage*9 +
551 parent_cpu_usage)/10;
552
553 ret = pthread_mutex_lock(&start_work_mutex);
554 BUG_ON(ret);
555
556 for (i = 0; i < nr_tasks; i++) {
557 task = tasks[i];
558 sem_init(&task->sleep_sem, 0, 0);
559 task->curr_event = 0;
560 }
561 }
562
563 static void run_one_test(void)
564 {
565 u64 T0, T1, delta, avg_delta, fluct, std_dev;
566
567 T0 = get_nsecs();
568 wait_for_tasks();
569 T1 = get_nsecs();
570
571 delta = T1 - T0;
572 sum_runtime += delta;
573 nr_runs++;
574
575 avg_delta = sum_runtime / nr_runs;
576 if (delta < avg_delta)
577 fluct = avg_delta - delta;
578 else
579 fluct = delta - avg_delta;
580 sum_fluct += fluct;
581 std_dev = sum_fluct / nr_runs / sqrt(nr_runs);
582 if (!run_avg)
583 run_avg = delta;
584 run_avg = (run_avg*9 + delta)/10;
585
586 printf("#%-3ld: %0.3f, ",
587 nr_runs, (double)delta/1000000.0);
588
589 printf("ravg: %0.2f, ",
590 (double)run_avg/1e6);
591
592 printf("cpu: %0.2f / %0.2f",
593 (double)cpu_usage/1e6, (double)runavg_cpu_usage/1e6);
594
595 #if 0
596 /*
597 * rusage statistics done by the parent, these are less
598 * accurate than the sum_exec_runtime based statistics:
599 */
600 printf(" [%0.2f / %0.2f]",
601 (double)parent_cpu_usage/1e6,
602 (double)runavg_parent_cpu_usage/1e6);
603 #endif
604
605 printf("\n");
606
607 if (nr_sleep_corrections)
608 printf(" (%ld sleep corrections)\n", nr_sleep_corrections);
609 nr_sleep_corrections = 0;
610 }
611
612 static void test_calibrations(void)
613 {
614 u64 T0, T1;
615
616 T0 = get_nsecs();
617 burn_nsecs(1e6);
618 T1 = get_nsecs();
619
620 printf("the run test took %" PRIu64 " nsecs\n", T1 - T0);
621
622 T0 = get_nsecs();
623 sleep_nsecs(1e6);
624 T1 = get_nsecs();
625
626 printf("the sleep test took %" PRIu64 " nsecs\n", T1 - T0);
627 }
628
629 #define FILL_FIELD(ptr, field, event, data) \
630 ptr.field = (typeof(ptr.field)) raw_field_value(event, #field, data)
631
632 #define FILL_ARRAY(ptr, array, event, data) \
633 do { \
634 void *__array = raw_field_ptr(event, #array, data); \
635 memcpy(ptr.array, __array, sizeof(ptr.array)); \
636 } while(0)
637
638 #define FILL_COMMON_FIELDS(ptr, event, data) \
639 do { \
640 FILL_FIELD(ptr, common_type, event, data); \
641 FILL_FIELD(ptr, common_flags, event, data); \
642 FILL_FIELD(ptr, common_preempt_count, event, data); \
643 FILL_FIELD(ptr, common_pid, event, data); \
644 FILL_FIELD(ptr, common_tgid, event, data); \
645 } while (0)
646
647
648
649 struct trace_switch_event {
650 u32 size;
651
652 u16 common_type;
653 u8 common_flags;
654 u8 common_preempt_count;
655 u32 common_pid;
656 u32 common_tgid;
657
658 char prev_comm[16];
659 u32 prev_pid;
660 u32 prev_prio;
661 u64 prev_state;
662 char next_comm[16];
663 u32 next_pid;
664 u32 next_prio;
665 };
666
667 struct trace_runtime_event {
668 u32 size;
669
670 u16 common_type;
671 u8 common_flags;
672 u8 common_preempt_count;
673 u32 common_pid;
674 u32 common_tgid;
675
676 char comm[16];
677 u32 pid;
678 u64 runtime;
679 u64 vruntime;
680 };
681
682 struct trace_wakeup_event {
683 u32 size;
684
685 u16 common_type;
686 u8 common_flags;
687 u8 common_preempt_count;
688 u32 common_pid;
689 u32 common_tgid;
690
691 char comm[16];
692 u32 pid;
693
694 u32 prio;
695 u32 success;
696 u32 cpu;
697 };
698
699 struct trace_fork_event {
700 u32 size;
701
702 u16 common_type;
703 u8 common_flags;
704 u8 common_preempt_count;
705 u32 common_pid;
706 u32 common_tgid;
707
708 char parent_comm[16];
709 u32 parent_pid;
710 char child_comm[16];
711 u32 child_pid;
712 };
713
714 struct trace_migrate_task_event {
715 u32 size;
716
717 u16 common_type;
718 u8 common_flags;
719 u8 common_preempt_count;
720 u32 common_pid;
721 u32 common_tgid;
722
723 char comm[16];
724 u32 pid;
725
726 u32 prio;
727 u32 cpu;
728 };
729
730 struct trace_sched_handler {
731 void (*switch_event)(struct trace_switch_event *,
732 struct perf_session *,
733 struct event *,
734 int cpu,
735 u64 timestamp,
736 struct thread *thread);
737
738 void (*runtime_event)(struct trace_runtime_event *,
739 struct perf_session *,
740 struct event *,
741 int cpu,
742 u64 timestamp,
743 struct thread *thread);
744
745 void (*wakeup_event)(struct trace_wakeup_event *,
746 struct perf_session *,
747 struct event *,
748 int cpu,
749 u64 timestamp,
750 struct thread *thread);
751
752 void (*fork_event)(struct trace_fork_event *,
753 struct event *,
754 int cpu,
755 u64 timestamp,
756 struct thread *thread);
757
758 void (*migrate_task_event)(struct trace_migrate_task_event *,
759 struct perf_session *session,
760 struct event *,
761 int cpu,
762 u64 timestamp,
763 struct thread *thread);
764 };
765
766
767 static void
768 replay_wakeup_event(struct trace_wakeup_event *wakeup_event,
769 struct perf_session *session __used,
770 struct event *event,
771 int cpu __used,
772 u64 timestamp __used,
773 struct thread *thread __used)
774 {
775 struct task_desc *waker, *wakee;
776
777 if (verbose) {
778 printf("sched_wakeup event %p\n", event);
779
780 printf(" ... pid %d woke up %s/%d\n",
781 wakeup_event->common_pid,
782 wakeup_event->comm,
783 wakeup_event->pid);
784 }
785
786 waker = register_pid(wakeup_event->common_pid, "<unknown>");
787 wakee = register_pid(wakeup_event->pid, wakeup_event->comm);
788
789 add_sched_event_wakeup(waker, timestamp, wakee);
790 }
791
792 static u64 cpu_last_switched[MAX_CPUS];
793
794 static void
795 replay_switch_event(struct trace_switch_event *switch_event,
796 struct perf_session *session __used,
797 struct event *event,
798 int cpu,
799 u64 timestamp,
800 struct thread *thread __used)
801 {
802 struct task_desc *prev, *next;
803 u64 timestamp0;
804 s64 delta;
805
806 if (verbose)
807 printf("sched_switch event %p\n", event);
808
809 if (cpu >= MAX_CPUS || cpu < 0)
810 return;
811
812 timestamp0 = cpu_last_switched[cpu];
813 if (timestamp0)
814 delta = timestamp - timestamp0;
815 else
816 delta = 0;
817
818 if (delta < 0)
819 die("hm, delta: %" PRIu64 " < 0 ?\n", delta);
820
821 if (verbose) {
822 printf(" ... switch from %s/%d to %s/%d [ran %" PRIu64 " nsecs]\n",
823 switch_event->prev_comm, switch_event->prev_pid,
824 switch_event->next_comm, switch_event->next_pid,
825 delta);
826 }
827
828 prev = register_pid(switch_event->prev_pid, switch_event->prev_comm);
829 next = register_pid(switch_event->next_pid, switch_event->next_comm);
830
831 cpu_last_switched[cpu] = timestamp;
832
833 add_sched_event_run(prev, timestamp, delta);
834 add_sched_event_sleep(prev, timestamp, switch_event->prev_state);
835 }
836
837
838 static void
839 replay_fork_event(struct trace_fork_event *fork_event,
840 struct event *event,
841 int cpu __used,
842 u64 timestamp __used,
843 struct thread *thread __used)
844 {
845 if (verbose) {
846 printf("sched_fork event %p\n", event);
847 printf("... parent: %s/%d\n", fork_event->parent_comm, fork_event->parent_pid);
848 printf("... child: %s/%d\n", fork_event->child_comm, fork_event->child_pid);
849 }
850 register_pid(fork_event->parent_pid, fork_event->parent_comm);
851 register_pid(fork_event->child_pid, fork_event->child_comm);
852 }
853
854 static struct trace_sched_handler replay_ops = {
855 .wakeup_event = replay_wakeup_event,
856 .switch_event = replay_switch_event,
857 .fork_event = replay_fork_event,
858 };
859
860 struct sort_dimension {
861 const char *name;
862 sort_fn_t cmp;
863 struct list_head list;
864 };
865
866 static LIST_HEAD(cmp_pid);
867
868 static int
869 thread_lat_cmp(struct list_head *list, struct work_atoms *l, struct work_atoms *r)
870 {
871 struct sort_dimension *sort;
872 int ret = 0;
873
874 BUG_ON(list_empty(list));
875
876 list_for_each_entry(sort, list, list) {
877 ret = sort->cmp(l, r);
878 if (ret)
879 return ret;
880 }
881
882 return ret;
883 }
884
885 static struct work_atoms *
886 thread_atoms_search(struct rb_root *root, struct thread *thread,
887 struct list_head *sort_list)
888 {
889 struct rb_node *node = root->rb_node;
890 struct work_atoms key = { .thread = thread };
891
892 while (node) {
893 struct work_atoms *atoms;
894 int cmp;
895
896 atoms = container_of(node, struct work_atoms, node);
897
898 cmp = thread_lat_cmp(sort_list, &key, atoms);
899 if (cmp > 0)
900 node = node->rb_left;
901 else if (cmp < 0)
902 node = node->rb_right;
903 else {
904 BUG_ON(thread != atoms->thread);
905 return atoms;
906 }
907 }
908 return NULL;
909 }
910
911 static void
912 __thread_latency_insert(struct rb_root *root, struct work_atoms *data,
913 struct list_head *sort_list)
914 {
915 struct rb_node **new = &(root->rb_node), *parent = NULL;
916
917 while (*new) {
918 struct work_atoms *this;
919 int cmp;
920
921 this = container_of(*new, struct work_atoms, node);
922 parent = *new;
923
924 cmp = thread_lat_cmp(sort_list, data, this);
925
926 if (cmp > 0)
927 new = &((*new)->rb_left);
928 else
929 new = &((*new)->rb_right);
930 }
931
932 rb_link_node(&data->node, parent, new);
933 rb_insert_color(&data->node, root);
934 }
935
936 static void thread_atoms_insert(struct thread *thread)
937 {
938 struct work_atoms *atoms = zalloc(sizeof(*atoms));
939 if (!atoms)
940 die("No memory");
941
942 atoms->thread = thread;
943 INIT_LIST_HEAD(&atoms->work_list);
944 __thread_latency_insert(&atom_root, atoms, &cmp_pid);
945 }
946
947 static void
948 latency_fork_event(struct trace_fork_event *fork_event __used,
949 struct event *event __used,
950 int cpu __used,
951 u64 timestamp __used,
952 struct thread *thread __used)
953 {
954 /* should insert the newcomer */
955 }
956
957 __used
958 static char sched_out_state(struct trace_switch_event *switch_event)
959 {
960 const char *str = TASK_STATE_TO_CHAR_STR;
961
962 return str[switch_event->prev_state];
963 }
964
965 static void
966 add_sched_out_event(struct work_atoms *atoms,
967 char run_state,
968 u64 timestamp)
969 {
970 struct work_atom *atom = zalloc(sizeof(*atom));
971 if (!atom)
972 die("Non memory");
973
974 atom->sched_out_time = timestamp;
975
976 if (run_state == 'R') {
977 atom->state = THREAD_WAIT_CPU;
978 atom->wake_up_time = atom->sched_out_time;
979 }
980
981 list_add_tail(&atom->list, &atoms->work_list);
982 }
983
984 static void
985 add_runtime_event(struct work_atoms *atoms, u64 delta, u64 timestamp __used)
986 {
987 struct work_atom *atom;
988
989 BUG_ON(list_empty(&atoms->work_list));
990
991 atom = list_entry(atoms->work_list.prev, struct work_atom, list);
992
993 atom->runtime += delta;
994 atoms->total_runtime += delta;
995 }
996
997 static void
998 add_sched_in_event(struct work_atoms *atoms, u64 timestamp)
999 {
1000 struct work_atom *atom;
1001 u64 delta;
1002
1003 if (list_empty(&atoms->work_list))
1004 return;
1005
1006 atom = list_entry(atoms->work_list.prev, struct work_atom, list);
1007
1008 if (atom->state != THREAD_WAIT_CPU)
1009 return;
1010
1011 if (timestamp < atom->wake_up_time) {
1012 atom->state = THREAD_IGNORE;
1013 return;
1014 }
1015
1016 atom->state = THREAD_SCHED_IN;
1017 atom->sched_in_time = timestamp;
1018
1019 delta = atom->sched_in_time - atom->wake_up_time;
1020 atoms->total_lat += delta;
1021 if (delta > atoms->max_lat) {
1022 atoms->max_lat = delta;
1023 atoms->max_lat_at = timestamp;
1024 }
1025 atoms->nb_atoms++;
1026 }
1027
1028 static void
1029 latency_switch_event(struct trace_switch_event *switch_event,
1030 struct perf_session *session,
1031 struct event *event __used,
1032 int cpu,
1033 u64 timestamp,
1034 struct thread *thread __used)
1035 {
1036 struct work_atoms *out_events, *in_events;
1037 struct thread *sched_out, *sched_in;
1038 u64 timestamp0;
1039 s64 delta;
1040
1041 BUG_ON(cpu >= MAX_CPUS || cpu < 0);
1042
1043 timestamp0 = cpu_last_switched[cpu];
1044 cpu_last_switched[cpu] = timestamp;
1045 if (timestamp0)
1046 delta = timestamp - timestamp0;
1047 else
1048 delta = 0;
1049
1050 if (delta < 0)
1051 die("hm, delta: %" PRIu64 " < 0 ?\n", delta);
1052
1053
1054 sched_out = perf_session__findnew(session, switch_event->prev_pid);
1055 sched_in = perf_session__findnew(session, switch_event->next_pid);
1056
1057 out_events = thread_atoms_search(&atom_root, sched_out, &cmp_pid);
1058 if (!out_events) {
1059 thread_atoms_insert(sched_out);
1060 out_events = thread_atoms_search(&atom_root, sched_out, &cmp_pid);
1061 if (!out_events)
1062 die("out-event: Internal tree error");
1063 }
1064 add_sched_out_event(out_events, sched_out_state(switch_event), timestamp);
1065
1066 in_events = thread_atoms_search(&atom_root, sched_in, &cmp_pid);
1067 if (!in_events) {
1068 thread_atoms_insert(sched_in);
1069 in_events = thread_atoms_search(&atom_root, sched_in, &cmp_pid);
1070 if (!in_events)
1071 die("in-event: Internal tree error");
1072 /*
1073 * Take came in we have not heard about yet,
1074 * add in an initial atom in runnable state:
1075 */
1076 add_sched_out_event(in_events, 'R', timestamp);
1077 }
1078 add_sched_in_event(in_events, timestamp);
1079 }
1080
1081 static void
1082 latency_runtime_event(struct trace_runtime_event *runtime_event,
1083 struct perf_session *session,
1084 struct event *event __used,
1085 int cpu,
1086 u64 timestamp,
1087 struct thread *this_thread __used)
1088 {
1089 struct thread *thread = perf_session__findnew(session, runtime_event->pid);
1090 struct work_atoms *atoms = thread_atoms_search(&atom_root, thread, &cmp_pid);
1091
1092 BUG_ON(cpu >= MAX_CPUS || cpu < 0);
1093 if (!atoms) {
1094 thread_atoms_insert(thread);
1095 atoms = thread_atoms_search(&atom_root, thread, &cmp_pid);
1096 if (!atoms)
1097 die("in-event: Internal tree error");
1098 add_sched_out_event(atoms, 'R', timestamp);
1099 }
1100
1101 add_runtime_event(atoms, runtime_event->runtime, timestamp);
1102 }
1103
1104 static void
1105 latency_wakeup_event(struct trace_wakeup_event *wakeup_event,
1106 struct perf_session *session,
1107 struct event *__event __used,
1108 int cpu __used,
1109 u64 timestamp,
1110 struct thread *thread __used)
1111 {
1112 struct work_atoms *atoms;
1113 struct work_atom *atom;
1114 struct thread *wakee;
1115
1116 /* Note for later, it may be interesting to observe the failing cases */
1117 if (!wakeup_event->success)
1118 return;
1119
1120 wakee = perf_session__findnew(session, wakeup_event->pid);
1121 atoms = thread_atoms_search(&atom_root, wakee, &cmp_pid);
1122 if (!atoms) {
1123 thread_atoms_insert(wakee);
1124 atoms = thread_atoms_search(&atom_root, wakee, &cmp_pid);
1125 if (!atoms)
1126 die("wakeup-event: Internal tree error");
1127 add_sched_out_event(atoms, 'S', timestamp);
1128 }
1129
1130 BUG_ON(list_empty(&atoms->work_list));
1131
1132 atom = list_entry(atoms->work_list.prev, struct work_atom, list);
1133
1134 /*
1135 * You WILL be missing events if you've recorded only
1136 * one CPU, or are only looking at only one, so don't
1137 * make useless noise.
1138 */
1139 if (profile_cpu == -1 && atom->state != THREAD_SLEEPING)
1140 nr_state_machine_bugs++;
1141
1142 nr_timestamps++;
1143 if (atom->sched_out_time > timestamp) {
1144 nr_unordered_timestamps++;
1145 return;
1146 }
1147
1148 atom->state = THREAD_WAIT_CPU;
1149 atom->wake_up_time = timestamp;
1150 }
1151
1152 static void
1153 latency_migrate_task_event(struct trace_migrate_task_event *migrate_task_event,
1154 struct perf_session *session,
1155 struct event *__event __used,
1156 int cpu __used,
1157 u64 timestamp,
1158 struct thread *thread __used)
1159 {
1160 struct work_atoms *atoms;
1161 struct work_atom *atom;
1162 struct thread *migrant;
1163
1164 /*
1165 * Only need to worry about migration when profiling one CPU.
1166 */
1167 if (profile_cpu == -1)
1168 return;
1169
1170 migrant = perf_session__findnew(session, migrate_task_event->pid);
1171 atoms = thread_atoms_search(&atom_root, migrant, &cmp_pid);
1172 if (!atoms) {
1173 thread_atoms_insert(migrant);
1174 register_pid(migrant->pid, migrant->comm);
1175 atoms = thread_atoms_search(&atom_root, migrant, &cmp_pid);
1176 if (!atoms)
1177 die("migration-event: Internal tree error");
1178 add_sched_out_event(atoms, 'R', timestamp);
1179 }
1180
1181 BUG_ON(list_empty(&atoms->work_list));
1182
1183 atom = list_entry(atoms->work_list.prev, struct work_atom, list);
1184 atom->sched_in_time = atom->sched_out_time = atom->wake_up_time = timestamp;
1185
1186 nr_timestamps++;
1187
1188 if (atom->sched_out_time > timestamp)
1189 nr_unordered_timestamps++;
1190 }
1191
1192 static struct trace_sched_handler lat_ops = {
1193 .wakeup_event = latency_wakeup_event,
1194 .switch_event = latency_switch_event,
1195 .runtime_event = latency_runtime_event,
1196 .fork_event = latency_fork_event,
1197 .migrate_task_event = latency_migrate_task_event,
1198 };
1199
1200 static void output_lat_thread(struct work_atoms *work_list)
1201 {
1202 int i;
1203 int ret;
1204 u64 avg;
1205
1206 if (!work_list->nb_atoms)
1207 return;
1208 /*
1209 * Ignore idle threads:
1210 */
1211 if (!strcmp(work_list->thread->comm, "swapper"))
1212 return;
1213
1214 all_runtime += work_list->total_runtime;
1215 all_count += work_list->nb_atoms;
1216
1217 ret = printf(" %s:%d ", work_list->thread->comm, work_list->thread->pid);
1218
1219 for (i = 0; i < 24 - ret; i++)
1220 printf(" ");
1221
1222 avg = work_list->total_lat / work_list->nb_atoms;
1223
1224 printf("|%11.3f ms |%9" PRIu64 " | avg:%9.3f ms | max:%9.3f ms | max at: %9.6f s\n",
1225 (double)work_list->total_runtime / 1e6,
1226 work_list->nb_atoms, (double)avg / 1e6,
1227 (double)work_list->max_lat / 1e6,
1228 (double)work_list->max_lat_at / 1e9);
1229 }
1230
1231 static int pid_cmp(struct work_atoms *l, struct work_atoms *r)
1232 {
1233 if (l->thread->pid < r->thread->pid)
1234 return -1;
1235 if (l->thread->pid > r->thread->pid)
1236 return 1;
1237
1238 return 0;
1239 }
1240
1241 static struct sort_dimension pid_sort_dimension = {
1242 .name = "pid",
1243 .cmp = pid_cmp,
1244 };
1245
1246 static int avg_cmp(struct work_atoms *l, struct work_atoms *r)
1247 {
1248 u64 avgl, avgr;
1249
1250 if (!l->nb_atoms)
1251 return -1;
1252
1253 if (!r->nb_atoms)
1254 return 1;
1255
1256 avgl = l->total_lat / l->nb_atoms;
1257 avgr = r->total_lat / r->nb_atoms;
1258
1259 if (avgl < avgr)
1260 return -1;
1261 if (avgl > avgr)
1262 return 1;
1263
1264 return 0;
1265 }
1266
1267 static struct sort_dimension avg_sort_dimension = {
1268 .name = "avg",
1269 .cmp = avg_cmp,
1270 };
1271
1272 static int max_cmp(struct work_atoms *l, struct work_atoms *r)
1273 {
1274 if (l->max_lat < r->max_lat)
1275 return -1;
1276 if (l->max_lat > r->max_lat)
1277 return 1;
1278
1279 return 0;
1280 }
1281
1282 static struct sort_dimension max_sort_dimension = {
1283 .name = "max",
1284 .cmp = max_cmp,
1285 };
1286
1287 static int switch_cmp(struct work_atoms *l, struct work_atoms *r)
1288 {
1289 if (l->nb_atoms < r->nb_atoms)
1290 return -1;
1291 if (l->nb_atoms > r->nb_atoms)
1292 return 1;
1293
1294 return 0;
1295 }
1296
1297 static struct sort_dimension switch_sort_dimension = {
1298 .name = "switch",
1299 .cmp = switch_cmp,
1300 };
1301
1302 static int runtime_cmp(struct work_atoms *l, struct work_atoms *r)
1303 {
1304 if (l->total_runtime < r->total_runtime)
1305 return -1;
1306 if (l->total_runtime > r->total_runtime)
1307 return 1;
1308
1309 return 0;
1310 }
1311
1312 static struct sort_dimension runtime_sort_dimension = {
1313 .name = "runtime",
1314 .cmp = runtime_cmp,
1315 };
1316
1317 static struct sort_dimension *available_sorts[] = {
1318 &pid_sort_dimension,
1319 &avg_sort_dimension,
1320 &max_sort_dimension,
1321 &switch_sort_dimension,
1322 &runtime_sort_dimension,
1323 };
1324
1325 #define NB_AVAILABLE_SORTS (int)(sizeof(available_sorts) / sizeof(struct sort_dimension *))
1326
1327 static LIST_HEAD(sort_list);
1328
1329 static int sort_dimension__add(const char *tok, struct list_head *list)
1330 {
1331 int i;
1332
1333 for (i = 0; i < NB_AVAILABLE_SORTS; i++) {
1334 if (!strcmp(available_sorts[i]->name, tok)) {
1335 list_add_tail(&available_sorts[i]->list, list);
1336
1337 return 0;
1338 }
1339 }
1340
1341 return -1;
1342 }
1343
1344 static void setup_sorting(void);
1345
1346 static void sort_lat(void)
1347 {
1348 struct rb_node *node;
1349
1350 for (;;) {
1351 struct work_atoms *data;
1352 node = rb_first(&atom_root);
1353 if (!node)
1354 break;
1355
1356 rb_erase(node, &atom_root);
1357 data = rb_entry(node, struct work_atoms, node);
1358 __thread_latency_insert(&sorted_atom_root, data, &sort_list);
1359 }
1360 }
1361
1362 static struct trace_sched_handler *trace_handler;
1363
1364 static void
1365 process_sched_wakeup_event(void *data, struct perf_session *session,
1366 struct event *event,
1367 int cpu __used,
1368 u64 timestamp __used,
1369 struct thread *thread __used)
1370 {
1371 struct trace_wakeup_event wakeup_event;
1372
1373 FILL_COMMON_FIELDS(wakeup_event, event, data);
1374
1375 FILL_ARRAY(wakeup_event, comm, event, data);
1376 FILL_FIELD(wakeup_event, pid, event, data);
1377 FILL_FIELD(wakeup_event, prio, event, data);
1378 FILL_FIELD(wakeup_event, success, event, data);
1379 FILL_FIELD(wakeup_event, cpu, event, data);
1380
1381 if (trace_handler->wakeup_event)
1382 trace_handler->wakeup_event(&wakeup_event, session, event,
1383 cpu, timestamp, thread);
1384 }
1385
1386 /*
1387 * Track the current task - that way we can know whether there's any
1388 * weird events, such as a task being switched away that is not current.
1389 */
1390 static int max_cpu;
1391
1392 static u32 curr_pid[MAX_CPUS] = { [0 ... MAX_CPUS-1] = -1 };
1393
1394 static struct thread *curr_thread[MAX_CPUS];
1395
1396 static char next_shortname1 = 'A';
1397 static char next_shortname2 = '0';
1398
1399 static void
1400 map_switch_event(struct trace_switch_event *switch_event,
1401 struct perf_session *session,
1402 struct event *event __used,
1403 int this_cpu,
1404 u64 timestamp,
1405 struct thread *thread __used)
1406 {
1407 struct thread *sched_out, *sched_in;
1408 int new_shortname;
1409 u64 timestamp0;
1410 s64 delta;
1411 int cpu;
1412
1413 BUG_ON(this_cpu >= MAX_CPUS || this_cpu < 0);
1414
1415 if (this_cpu > max_cpu)
1416 max_cpu = this_cpu;
1417
1418 timestamp0 = cpu_last_switched[this_cpu];
1419 cpu_last_switched[this_cpu] = timestamp;
1420 if (timestamp0)
1421 delta = timestamp - timestamp0;
1422 else
1423 delta = 0;
1424
1425 if (delta < 0)
1426 die("hm, delta: %" PRIu64 " < 0 ?\n", delta);
1427
1428
1429 sched_out = perf_session__findnew(session, switch_event->prev_pid);
1430 sched_in = perf_session__findnew(session, switch_event->next_pid);
1431
1432 curr_thread[this_cpu] = sched_in;
1433
1434 printf(" ");
1435
1436 new_shortname = 0;
1437 if (!sched_in->shortname[0]) {
1438 sched_in->shortname[0] = next_shortname1;
1439 sched_in->shortname[1] = next_shortname2;
1440
1441 if (next_shortname1 < 'Z') {
1442 next_shortname1++;
1443 } else {
1444 next_shortname1='A';
1445 if (next_shortname2 < '9') {
1446 next_shortname2++;
1447 } else {
1448 next_shortname2='0';
1449 }
1450 }
1451 new_shortname = 1;
1452 }
1453
1454 for (cpu = 0; cpu <= max_cpu; cpu++) {
1455 if (cpu != this_cpu)
1456 printf(" ");
1457 else
1458 printf("*");
1459
1460 if (curr_thread[cpu]) {
1461 if (curr_thread[cpu]->pid)
1462 printf("%2s ", curr_thread[cpu]->shortname);
1463 else
1464 printf(". ");
1465 } else
1466 printf(" ");
1467 }
1468
1469 printf(" %12.6f secs ", (double)timestamp/1e9);
1470 if (new_shortname) {
1471 printf("%s => %s:%d\n",
1472 sched_in->shortname, sched_in->comm, sched_in->pid);
1473 } else {
1474 printf("\n");
1475 }
1476 }
1477
1478
1479 static void
1480 process_sched_switch_event(void *data, struct perf_session *session,
1481 struct event *event,
1482 int this_cpu,
1483 u64 timestamp __used,
1484 struct thread *thread __used)
1485 {
1486 struct trace_switch_event switch_event;
1487
1488 FILL_COMMON_FIELDS(switch_event, event, data);
1489
1490 FILL_ARRAY(switch_event, prev_comm, event, data);
1491 FILL_FIELD(switch_event, prev_pid, event, data);
1492 FILL_FIELD(switch_event, prev_prio, event, data);
1493 FILL_FIELD(switch_event, prev_state, event, data);
1494 FILL_ARRAY(switch_event, next_comm, event, data);
1495 FILL_FIELD(switch_event, next_pid, event, data);
1496 FILL_FIELD(switch_event, next_prio, event, data);
1497
1498 if (curr_pid[this_cpu] != (u32)-1) {
1499 /*
1500 * Are we trying to switch away a PID that is
1501 * not current?
1502 */
1503 if (curr_pid[this_cpu] != switch_event.prev_pid)
1504 nr_context_switch_bugs++;
1505 }
1506 if (trace_handler->switch_event)
1507 trace_handler->switch_event(&switch_event, session, event,
1508 this_cpu, timestamp, thread);
1509
1510 curr_pid[this_cpu] = switch_event.next_pid;
1511 }
1512
1513 static void
1514 process_sched_runtime_event(void *data, struct perf_session *session,
1515 struct event *event,
1516 int cpu __used,
1517 u64 timestamp __used,
1518 struct thread *thread __used)
1519 {
1520 struct trace_runtime_event runtime_event;
1521
1522 FILL_ARRAY(runtime_event, comm, event, data);
1523 FILL_FIELD(runtime_event, pid, event, data);
1524 FILL_FIELD(runtime_event, runtime, event, data);
1525 FILL_FIELD(runtime_event, vruntime, event, data);
1526
1527 if (trace_handler->runtime_event)
1528 trace_handler->runtime_event(&runtime_event, session, event, cpu, timestamp, thread);
1529 }
1530
1531 static void
1532 process_sched_fork_event(void *data,
1533 struct event *event,
1534 int cpu __used,
1535 u64 timestamp __used,
1536 struct thread *thread __used)
1537 {
1538 struct trace_fork_event fork_event;
1539
1540 FILL_COMMON_FIELDS(fork_event, event, data);
1541
1542 FILL_ARRAY(fork_event, parent_comm, event, data);
1543 FILL_FIELD(fork_event, parent_pid, event, data);
1544 FILL_ARRAY(fork_event, child_comm, event, data);
1545 FILL_FIELD(fork_event, child_pid, event, data);
1546
1547 if (trace_handler->fork_event)
1548 trace_handler->fork_event(&fork_event, event,
1549 cpu, timestamp, thread);
1550 }
1551
1552 static void
1553 process_sched_exit_event(struct event *event,
1554 int cpu __used,
1555 u64 timestamp __used,
1556 struct thread *thread __used)
1557 {
1558 if (verbose)
1559 printf("sched_exit event %p\n", event);
1560 }
1561
1562 static void
1563 process_sched_migrate_task_event(void *data, struct perf_session *session,
1564 struct event *event,
1565 int cpu __used,
1566 u64 timestamp __used,
1567 struct thread *thread __used)
1568 {
1569 struct trace_migrate_task_event migrate_task_event;
1570
1571 FILL_COMMON_FIELDS(migrate_task_event, event, data);
1572
1573 FILL_ARRAY(migrate_task_event, comm, event, data);
1574 FILL_FIELD(migrate_task_event, pid, event, data);
1575 FILL_FIELD(migrate_task_event, prio, event, data);
1576 FILL_FIELD(migrate_task_event, cpu, event, data);
1577
1578 if (trace_handler->migrate_task_event)
1579 trace_handler->migrate_task_event(&migrate_task_event, session,
1580 event, cpu, timestamp, thread);
1581 }
1582
1583 static void
1584 process_raw_event(event_t *raw_event __used, struct perf_session *session,
1585 void *data, int cpu, u64 timestamp, struct thread *thread)
1586 {
1587 struct event *event;
1588 int type;
1589
1590
1591 type = trace_parse_common_type(data);
1592 event = trace_find_event(type);
1593
1594 if (!strcmp(event->name, "sched_switch"))
1595 process_sched_switch_event(data, session, event, cpu, timestamp, thread);
1596 if (!strcmp(event->name, "sched_stat_runtime"))
1597 process_sched_runtime_event(data, session, event, cpu, timestamp, thread);
1598 if (!strcmp(event->name, "sched_wakeup"))
1599 process_sched_wakeup_event(data, session, event, cpu, timestamp, thread);
1600 if (!strcmp(event->name, "sched_wakeup_new"))
1601 process_sched_wakeup_event(data, session, event, cpu, timestamp, thread);
1602 if (!strcmp(event->name, "sched_process_fork"))
1603 process_sched_fork_event(data, event, cpu, timestamp, thread);
1604 if (!strcmp(event->name, "sched_process_exit"))
1605 process_sched_exit_event(event, cpu, timestamp, thread);
1606 if (!strcmp(event->name, "sched_migrate_task"))
1607 process_sched_migrate_task_event(data, session, event, cpu, timestamp, thread);
1608 }
1609
1610 static int process_sample_event(event_t *event, struct sample_data *sample,
1611 struct perf_session *session)
1612 {
1613 struct thread *thread;
1614
1615 if (!(session->sample_type & PERF_SAMPLE_RAW))
1616 return 0;
1617
1618 thread = perf_session__findnew(session, sample->pid);
1619 if (thread == NULL) {
1620 pr_debug("problem processing %d event, skipping it.\n",
1621 event->header.type);
1622 return -1;
1623 }
1624
1625 dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid);
1626
1627 if (profile_cpu != -1 && profile_cpu != (int)sample->cpu)
1628 return 0;
1629
1630 process_raw_event(event, session, sample->raw_data, sample->cpu,
1631 sample->time, thread);
1632
1633 return 0;
1634 }
1635
1636 static struct perf_event_ops event_ops = {
1637 .sample = process_sample_event,
1638 .comm = event__process_comm,
1639 .lost = event__process_lost,
1640 .fork = event__process_task,
1641 .ordered_samples = true,
1642 };
1643
1644 static int read_events(void)
1645 {
1646 int err = -EINVAL;
1647 struct perf_session *session = perf_session__new(input_name, O_RDONLY,
1648 0, false, &event_ops);
1649 if (session == NULL)
1650 return -ENOMEM;
1651
1652 if (perf_session__has_traces(session, "record -R")) {
1653 err = perf_session__process_events(session, &event_ops);
1654 nr_events = session->hists.stats.nr_events[0];
1655 nr_lost_events = session->hists.stats.total_lost;
1656 nr_lost_chunks = session->hists.stats.nr_events[PERF_RECORD_LOST];
1657 }
1658
1659 perf_session__delete(session);
1660 return err;
1661 }
1662
1663 static void print_bad_events(void)
1664 {
1665 if (nr_unordered_timestamps && nr_timestamps) {
1666 printf(" INFO: %.3f%% unordered timestamps (%ld out of %ld)\n",
1667 (double)nr_unordered_timestamps/(double)nr_timestamps*100.0,
1668 nr_unordered_timestamps, nr_timestamps);
1669 }
1670 if (nr_lost_events && nr_events) {
1671 printf(" INFO: %.3f%% lost events (%ld out of %ld, in %ld chunks)\n",
1672 (double)nr_lost_events/(double)nr_events*100.0,
1673 nr_lost_events, nr_events, nr_lost_chunks);
1674 }
1675 if (nr_state_machine_bugs && nr_timestamps) {
1676 printf(" INFO: %.3f%% state machine bugs (%ld out of %ld)",
1677 (double)nr_state_machine_bugs/(double)nr_timestamps*100.0,
1678 nr_state_machine_bugs, nr_timestamps);
1679 if (nr_lost_events)
1680 printf(" (due to lost events?)");
1681 printf("\n");
1682 }
1683 if (nr_context_switch_bugs && nr_timestamps) {
1684 printf(" INFO: %.3f%% context switch bugs (%ld out of %ld)",
1685 (double)nr_context_switch_bugs/(double)nr_timestamps*100.0,
1686 nr_context_switch_bugs, nr_timestamps);
1687 if (nr_lost_events)
1688 printf(" (due to lost events?)");
1689 printf("\n");
1690 }
1691 }
1692
1693 static void __cmd_lat(void)
1694 {
1695 struct rb_node *next;
1696
1697 setup_pager();
1698 read_events();
1699 sort_lat();
1700
1701 printf("\n ---------------------------------------------------------------------------------------------------------------\n");
1702 printf(" Task | Runtime ms | Switches | Average delay ms | Maximum delay ms | Maximum delay at |\n");
1703 printf(" ---------------------------------------------------------------------------------------------------------------\n");
1704
1705 next = rb_first(&sorted_atom_root);
1706
1707 while (next) {
1708 struct work_atoms *work_list;
1709
1710 work_list = rb_entry(next, struct work_atoms, node);
1711 output_lat_thread(work_list);
1712 next = rb_next(next);
1713 }
1714
1715 printf(" -----------------------------------------------------------------------------------------\n");
1716 printf(" TOTAL: |%11.3f ms |%9" PRIu64 " |\n",
1717 (double)all_runtime/1e6, all_count);
1718
1719 printf(" ---------------------------------------------------\n");
1720
1721 print_bad_events();
1722 printf("\n");
1723
1724 }
1725
1726 static struct trace_sched_handler map_ops = {
1727 .wakeup_event = NULL,
1728 .switch_event = map_switch_event,
1729 .runtime_event = NULL,
1730 .fork_event = NULL,
1731 };
1732
1733 static void __cmd_map(void)
1734 {
1735 max_cpu = sysconf(_SC_NPROCESSORS_CONF);
1736
1737 setup_pager();
1738 read_events();
1739 print_bad_events();
1740 }
1741
1742 static void __cmd_replay(void)
1743 {
1744 unsigned long i;
1745
1746 calibrate_run_measurement_overhead();
1747 calibrate_sleep_measurement_overhead();
1748
1749 test_calibrations();
1750
1751 read_events();
1752
1753 printf("nr_run_events: %ld\n", nr_run_events);
1754 printf("nr_sleep_events: %ld\n", nr_sleep_events);
1755 printf("nr_wakeup_events: %ld\n", nr_wakeup_events);
1756
1757 if (targetless_wakeups)
1758 printf("target-less wakeups: %ld\n", targetless_wakeups);
1759 if (multitarget_wakeups)
1760 printf("multi-target wakeups: %ld\n", multitarget_wakeups);
1761 if (nr_run_events_optimized)
1762 printf("run atoms optimized: %ld\n",
1763 nr_run_events_optimized);
1764
1765 print_task_traces();
1766 add_cross_task_wakeups();
1767
1768 create_tasks();
1769 printf("------------------------------------------------------------\n");
1770 for (i = 0; i < replay_repeat; i++)
1771 run_one_test();
1772 }
1773
1774
1775 static const char * const sched_usage[] = {
1776 "perf sched [<options>] {record|latency|map|replay|trace}",
1777 NULL
1778 };
1779
1780 static const struct option sched_options[] = {
1781 OPT_STRING('i', "input", &input_name, "file",
1782 "input file name"),
1783 OPT_INCR('v', "verbose", &verbose,
1784 "be more verbose (show symbol address, etc)"),
1785 OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
1786 "dump raw trace in ASCII"),
1787 OPT_END()
1788 };
1789
1790 static const char * const latency_usage[] = {
1791 "perf sched latency [<options>]",
1792 NULL
1793 };
1794
1795 static const struct option latency_options[] = {
1796 OPT_STRING('s', "sort", &sort_order, "key[,key2...]",
1797 "sort by key(s): runtime, switch, avg, max"),
1798 OPT_INCR('v', "verbose", &verbose,
1799 "be more verbose (show symbol address, etc)"),
1800 OPT_INTEGER('C', "CPU", &profile_cpu,
1801 "CPU to profile on"),
1802 OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
1803 "dump raw trace in ASCII"),
1804 OPT_END()
1805 };
1806
1807 static const char * const replay_usage[] = {
1808 "perf sched replay [<options>]",
1809 NULL
1810 };
1811
1812 static const struct option replay_options[] = {
1813 OPT_UINTEGER('r', "repeat", &replay_repeat,
1814 "repeat the workload replay N times (-1: infinite)"),
1815 OPT_INCR('v', "verbose", &verbose,
1816 "be more verbose (show symbol address, etc)"),
1817 OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
1818 "dump raw trace in ASCII"),
1819 OPT_END()
1820 };
1821
1822 static void setup_sorting(void)
1823 {
1824 char *tmp, *tok, *str = strdup(sort_order);
1825
1826 for (tok = strtok_r(str, ", ", &tmp);
1827 tok; tok = strtok_r(NULL, ", ", &tmp)) {
1828 if (sort_dimension__add(tok, &sort_list) < 0) {
1829 error("Unknown --sort key: `%s'", tok);
1830 usage_with_options(latency_usage, latency_options);
1831 }
1832 }
1833
1834 free(str);
1835
1836 sort_dimension__add("pid", &cmp_pid);
1837 }
1838
1839 static const char *record_args[] = {
1840 "record",
1841 "-a",
1842 "-R",
1843 "-f",
1844 "-m", "1024",
1845 "-c", "1",
1846 "-e", "sched:sched_switch",
1847 "-e", "sched:sched_stat_wait",
1848 "-e", "sched:sched_stat_sleep",
1849 "-e", "sched:sched_stat_iowait",
1850 "-e", "sched:sched_stat_runtime",
1851 "-e", "sched:sched_process_exit",
1852 "-e", "sched:sched_process_fork",
1853 "-e", "sched:sched_wakeup",
1854 "-e", "sched:sched_migrate_task",
1855 };
1856
1857 static int __cmd_record(int argc, const char **argv)
1858 {
1859 unsigned int rec_argc, i, j;
1860 const char **rec_argv;
1861
1862 rec_argc = ARRAY_SIZE(record_args) + argc - 1;
1863 rec_argv = calloc(rec_argc + 1, sizeof(char *));
1864
1865 if (rec_argv == NULL)
1866 return -ENOMEM;
1867
1868 for (i = 0; i < ARRAY_SIZE(record_args); i++)
1869 rec_argv[i] = strdup(record_args[i]);
1870
1871 for (j = 1; j < (unsigned int)argc; j++, i++)
1872 rec_argv[i] = argv[j];
1873
1874 BUG_ON(i != rec_argc);
1875
1876 return cmd_record(i, rec_argv, NULL);
1877 }
1878
1879 int cmd_sched(int argc, const char **argv, const char *prefix __used)
1880 {
1881 argc = parse_options(argc, argv, sched_options, sched_usage,
1882 PARSE_OPT_STOP_AT_NON_OPTION);
1883 if (!argc)
1884 usage_with_options(sched_usage, sched_options);
1885
1886 /*
1887 * Aliased to 'perf script' for now:
1888 */
1889 if (!strcmp(argv[0], "script"))
1890 return cmd_script(argc, argv, prefix);
1891
1892 symbol__init();
1893 if (!strncmp(argv[0], "rec", 3)) {
1894 return __cmd_record(argc, argv);
1895 } else if (!strncmp(argv[0], "lat", 3)) {
1896 trace_handler = &lat_ops;
1897 if (argc > 1) {
1898 argc = parse_options(argc, argv, latency_options, latency_usage, 0);
1899 if (argc)
1900 usage_with_options(latency_usage, latency_options);
1901 }
1902 setup_sorting();
1903 __cmd_lat();
1904 } else if (!strcmp(argv[0], "map")) {
1905 trace_handler = &map_ops;
1906 setup_sorting();
1907 __cmd_map();
1908 } else if (!strncmp(argv[0], "rep", 3)) {
1909 trace_handler = &replay_ops;
1910 if (argc) {
1911 argc = parse_options(argc, argv, replay_options, replay_usage, 0);
1912 if (argc)
1913 usage_with_options(replay_usage, replay_options);
1914 }
1915 __cmd_replay();
1916 } else {
1917 usage_with_options(sched_usage, sched_options);
1918 }
1919
1920 return 0;
1921 }
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree