builtin-record.c 29.8 KB
Newer Older
I
Ingo Molnar 已提交
1
/*
2 3 4 5 6
 * builtin-record.c
 *
 * Builtin record command: Record the profile of a workload
 * (or a CPU, or a PID) into the perf.data output file - for
 * later analysis via perf report.
I
Ingo Molnar 已提交
7
 */
8
#include "builtin.h"
9 10 11

#include "perf.h"

12
#include "util/build-id.h"
13
#include "util/util.h"
14
#include "util/parse-options.h"
15
#include "util/parse-events.h"
16

17
#include "util/callchain.h"
18
#include "util/cgroup.h"
19
#include "util/header.h"
20
#include "util/event.h"
21
#include "util/evlist.h"
22
#include "util/evsel.h"
23
#include "util/debug.h"
24
#include "util/session.h"
25
#include "util/tool.h"
26
#include "util/symbol.h"
27
#include "util/cpumap.h"
28
#include "util/thread_map.h"
29
#include "util/data.h"
30
#include "util/auxtrace.h"
31
#include "util/parse-branch-options.h"
32

33
#include <unistd.h>
34
#include <sched.h>
35
#include <sys/mman.h>
36

37

38
struct record {
39
	struct perf_tool	tool;
40
	struct record_opts	opts;
41
	u64			bytes_written;
42
	struct perf_data_file	file;
43
	struct auxtrace_record	*itr;
44 45 46 47 48 49 50
	struct perf_evlist	*evlist;
	struct perf_session	*session;
	const char		*progname;
	int			realtime_prio;
	bool			no_buildid;
	bool			no_buildid_cache;
	long			samples;
51
};
52

53
static int record__write(struct record *rec, void *bf, size_t size)
54
{
55
	if (perf_data_file__write(rec->session->file, bf, size) < 0) {
56 57
		pr_err("failed to write perf data, error: %m\n");
		return -1;
58
	}
59

60
	rec->bytes_written += size;
61
	return 0;
62 63
}

64
static int process_synthesized_event(struct perf_tool *tool,
65
				     union perf_event *event,
66 67
				     struct perf_sample *sample __maybe_unused,
				     struct machine *machine __maybe_unused)
68
{
69 70
	struct record *rec = container_of(tool, struct record, tool);
	return record__write(rec, event, event->header.size);
71 72
}

73
static int record__mmap_read(struct record *rec, int idx)
74
{
75
	struct perf_mmap *md = &rec->evlist->mmap[idx];
76 77
	u64 head = perf_mmap__read_head(md);
	u64 old = md->prev;
J
Jiri Olsa 已提交
78
	unsigned char *data = md->base + page_size;
79 80
	unsigned long size;
	void *buf;
81
	int rc = 0;
82

83
	if (old == head)
84
		return 0;
85

86
	rec->samples++;
87 88 89 90 91 92 93

	size = head - old;

	if ((old & md->mask) + size != (head & md->mask)) {
		buf = &data[old & md->mask];
		size = md->mask + 1 - (old & md->mask);
		old += size;
94

95
		if (record__write(rec, buf, size) < 0) {
96 97 98
			rc = -1;
			goto out;
		}
99 100 101 102 103
	}

	buf = &data[old & md->mask];
	size = head - old;
	old += size;
104

105
	if (record__write(rec, buf, size) < 0) {
106 107 108
		rc = -1;
		goto out;
	}
109 110

	md->prev = old;
111
	perf_evlist__mmap_consume(rec->evlist, idx);
112 113
out:
	return rc;
114 115
}

116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141
static volatile int done;
static volatile int signr = -1;
static volatile int child_finished;
static volatile int auxtrace_snapshot_enabled;
static volatile int auxtrace_snapshot_err;
static volatile int auxtrace_record__snapshot_started;

static void sig_handler(int sig)
{
	if (sig == SIGCHLD)
		child_finished = 1;
	else
		signr = sig;

	done = 1;
}

static void record__sig_exit(void)
{
	if (signr == -1)
		return;

	signal(signr, SIG_DFL);
	raise(signr);
}

142 143
#ifdef HAVE_AUXTRACE_SUPPORT

144 145 146 147 148
static int record__process_auxtrace(struct perf_tool *tool,
				    union perf_event *event, void *data1,
				    size_t len1, void *data2, size_t len2)
{
	struct record *rec = container_of(tool, struct record, tool);
149
	struct perf_data_file *file = &rec->file;
150 151 152
	size_t padding;
	u8 pad[8] = {0};

153 154 155 156 157 158 159 160 161 162 163 164 165 166
	if (!perf_data_file__is_pipe(file)) {
		off_t file_offset;
		int fd = perf_data_file__fd(file);
		int err;

		file_offset = lseek(fd, 0, SEEK_CUR);
		if (file_offset == -1)
			return -1;
		err = auxtrace_index__auxtrace_event(&rec->session->auxtrace_index,
						     event, file_offset);
		if (err)
			return err;
	}

167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196
	/* event.auxtrace.size includes padding, see __auxtrace_mmap__read() */
	padding = (len1 + len2) & 7;
	if (padding)
		padding = 8 - padding;

	record__write(rec, event, event->header.size);
	record__write(rec, data1, len1);
	if (len2)
		record__write(rec, data2, len2);
	record__write(rec, &pad, padding);

	return 0;
}

static int record__auxtrace_mmap_read(struct record *rec,
				      struct auxtrace_mmap *mm)
{
	int ret;

	ret = auxtrace_mmap__read(mm, rec->itr, &rec->tool,
				  record__process_auxtrace);
	if (ret < 0)
		return ret;

	if (ret)
		rec->samples++;

	return 0;
}

197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246
static int record__auxtrace_mmap_read_snapshot(struct record *rec,
					       struct auxtrace_mmap *mm)
{
	int ret;

	ret = auxtrace_mmap__read_snapshot(mm, rec->itr, &rec->tool,
					   record__process_auxtrace,
					   rec->opts.auxtrace_snapshot_size);
	if (ret < 0)
		return ret;

	if (ret)
		rec->samples++;

	return 0;
}

static int record__auxtrace_read_snapshot_all(struct record *rec)
{
	int i;
	int rc = 0;

	for (i = 0; i < rec->evlist->nr_mmaps; i++) {
		struct auxtrace_mmap *mm =
				&rec->evlist->mmap[i].auxtrace_mmap;

		if (!mm->base)
			continue;

		if (record__auxtrace_mmap_read_snapshot(rec, mm) != 0) {
			rc = -1;
			goto out;
		}
	}
out:
	return rc;
}

static void record__read_auxtrace_snapshot(struct record *rec)
{
	pr_debug("Recording AUX area tracing snapshot\n");
	if (record__auxtrace_read_snapshot_all(rec) < 0) {
		auxtrace_snapshot_err = -1;
	} else {
		auxtrace_snapshot_err = auxtrace_record__snapshot_finish(rec->itr);
		if (!auxtrace_snapshot_err)
			auxtrace_snapshot_enabled = 1;
	}
}

247 248 249 250 251 252 253 254 255
#else

static inline
int record__auxtrace_mmap_read(struct record *rec __maybe_unused,
			       struct auxtrace_mmap *mm __maybe_unused)
{
	return 0;
}

256 257
static inline
void record__read_auxtrace_snapshot(struct record *rec __maybe_unused)
258
{
259 260
}

261 262
static inline
int auxtrace_record__snapshot_start(struct auxtrace_record *itr __maybe_unused)
263
{
264
	return 0;
265 266
}

267 268
#endif

269
static int record__open(struct record *rec)
270
{
271
	char msg[512];
272
	struct perf_evsel *pos;
273 274
	struct perf_evlist *evlist = rec->evlist;
	struct perf_session *session = rec->session;
275
	struct record_opts *opts = &rec->opts;
276
	int rc = 0;
277

278
	perf_evlist__config(evlist, opts);
279

280
	evlist__for_each(evlist, pos) {
281
try_again:
282
		if (perf_evsel__open(pos, evlist->cpus, evlist->threads) < 0) {
283
			if (perf_evsel__fallback(pos, errno, msg, sizeof(msg))) {
284
				if (verbose)
285
					ui__warning("%s\n", msg);
286 287
				goto try_again;
			}
288

289 290 291 292
			rc = -errno;
			perf_evsel__open_strerror(pos, &opts->target,
						  errno, msg, sizeof(msg));
			ui__error("%s\n", msg);
293
			goto out;
L
Li Zefan 已提交
294 295
		}
	}
296

297 298 299
	if (perf_evlist__apply_filters(evlist, &pos)) {
		error("failed to set filter \"%s\" on event %s with %d (%s)\n",
			pos->filter, perf_evsel__name(pos), errno,
300
			strerror_r(errno, msg, sizeof(msg)));
301 302
		rc = -1;
		goto out;
303 304
	}

305
	if (perf_evlist__mmap_ex(evlist, opts->mmap_pages, false,
306 307
				 opts->auxtrace_mmap_pages,
				 opts->auxtrace_snapshot_mode) < 0) {
308 309 310 311 312
		if (errno == EPERM) {
			pr_err("Permission error mapping pages.\n"
			       "Consider increasing "
			       "/proc/sys/kernel/perf_event_mlock_kb,\n"
			       "or try again with a smaller value of -m/--mmap_pages.\n"
313 314
			       "(current value: %u,%u)\n",
			       opts->mmap_pages, opts->auxtrace_mmap_pages);
315 316
			rc = -errno;
		} else {
317 318
			pr_err("failed to mmap with %d (%s)\n", errno,
				strerror_r(errno, msg, sizeof(msg)));
319 320 321
			rc = -errno;
		}
		goto out;
322
	}
323

324
	session->evlist = evlist;
325
	perf_session__set_id_hdr_size(session);
326 327
out:
	return rc;
328 329
}

330 331 332 333 334 335 336 337 338 339 340 341 342
static int process_sample_event(struct perf_tool *tool,
				union perf_event *event,
				struct perf_sample *sample,
				struct perf_evsel *evsel,
				struct machine *machine)
{
	struct record *rec = container_of(tool, struct record, tool);

	rec->samples++;

	return build_id__mark_dso_hit(tool, event, sample, evsel, machine);
}

343
static int process_buildids(struct record *rec)
344
{
345 346
	struct perf_data_file *file  = &rec->file;
	struct perf_session *session = rec->session;
347

348
	u64 size = lseek(perf_data_file__fd(file), 0, SEEK_CUR);
349 350 351
	if (size == 0)
		return 0;

352 353
	file->size = size;

354 355 356 357 358 359 360 361 362 363 364
	/*
	 * During this process, it'll load kernel map and replace the
	 * dso->long_name to a real pathname it found.  In this case
	 * we prefer the vmlinux path like
	 *   /lib/modules/3.16.4/build/vmlinux
	 *
	 * rather than build-id path (in debug directory).
	 *   $HOME/.debug/.build-id/f0/6e17aa50adf4d00b88925e03775de107611551
	 */
	symbol_conf.ignore_vmlinux_buildid = true;

365
	return perf_session__process_events(session);
366 367
}

368
static void perf_event__synthesize_guest_os(struct machine *machine, void *data)
369 370
{
	int err;
371
	struct perf_tool *tool = data;
372 373 374 375 376 377 378 379
	/*
	 *As for guest kernel when processing subcommand record&report,
	 *we arrange module mmap prior to guest kernel mmap and trigger
	 *a preload dso because default guest module symbols are loaded
	 *from guest kallsyms instead of /lib/modules/XXX/XXX. This
	 *method is used to avoid symbol missing when the first addr is
	 *in module instead of in guest kernel.
	 */
380
	err = perf_event__synthesize_modules(tool, process_synthesized_event,
381
					     machine);
382 383
	if (err < 0)
		pr_err("Couldn't record guest kernel [%d]'s reference"
384
		       " relocation symbol.\n", machine->pid);
385 386 387 388 389

	/*
	 * We use _stext for guest kernel because guest kernel's /proc/kallsyms
	 * have no _text sometimes.
	 */
390
	err = perf_event__synthesize_kernel_mmap(tool, process_synthesized_event,
391
						 machine);
392 393
	if (err < 0)
		pr_err("Couldn't record guest kernel [%d]'s reference"
394
		       " relocation symbol.\n", machine->pid);
395 396
}

397 398 399 400 401
static struct perf_event_header finished_round_event = {
	.size = sizeof(struct perf_event_header),
	.type = PERF_RECORD_FINISHED_ROUND,
};

402
static int record__mmap_read_all(struct record *rec)
403
{
404
	u64 bytes_written = rec->bytes_written;
405
	int i;
406
	int rc = 0;
407

408
	for (i = 0; i < rec->evlist->nr_mmaps; i++) {
409 410
		struct auxtrace_mmap *mm = &rec->evlist->mmap[i].auxtrace_mmap;

411
		if (rec->evlist->mmap[i].base) {
412
			if (record__mmap_read(rec, i) != 0) {
413 414 415 416
				rc = -1;
				goto out;
			}
		}
417

418
		if (mm->base && !rec->opts.auxtrace_snapshot_mode &&
419 420 421 422
		    record__auxtrace_mmap_read(rec, mm) != 0) {
			rc = -1;
			goto out;
		}
423 424
	}

425 426 427 428 429 430
	/*
	 * Mark the round finished in case we wrote
	 * at least one event.
	 */
	if (bytes_written != rec->bytes_written)
		rc = record__write(rec, &finished_round_event, sizeof(finished_round_event));
431 432 433

out:
	return rc;
434 435
}

436
static void record__init_features(struct record *rec)
437 438 439 440 441 442 443 444 445 446
{
	struct perf_session *session = rec->session;
	int feat;

	for (feat = HEADER_FIRST_FEATURE; feat < HEADER_LAST_FEATURE; feat++)
		perf_header__set_feat(&session->header, feat);

	if (rec->no_buildid)
		perf_header__clear_feat(&session->header, HEADER_BUILD_ID);

447
	if (!have_tracepoints(&rec->evlist->entries))
448 449 450 451
		perf_header__clear_feat(&session->header, HEADER_TRACING_DATA);

	if (!rec->opts.branch_stack)
		perf_header__clear_feat(&session->header, HEADER_BRANCH_STACK);
452 453 454

	if (!rec->opts.full_auxtrace)
		perf_header__clear_feat(&session->header, HEADER_AUXTRACE);
455 456
}

457 458 459 460 461 462 463
static volatile int workload_exec_errno;

/*
 * perf_evlist__prepare_workload will send a SIGUSR1
 * if the fork fails, since we asked by setting its
 * want_signal to true.
 */
464 465
static void workload_exec_failed_signal(int signo __maybe_unused,
					siginfo_t *info,
466 467 468 469 470 471 472
					void *ucontext __maybe_unused)
{
	workload_exec_errno = info->si_value.sival_int;
	done = 1;
	child_finished = 1;
}

473 474
static void snapshot_sig_handler(int sig);

475
static int __cmd_record(struct record *rec, int argc, const char **argv)
476
{
477
	int err;
478
	int status = 0;
479
	unsigned long waking = 0;
480
	const bool forks = argc > 0;
481
	struct machine *machine;
482
	struct perf_tool *tool = &rec->tool;
483
	struct record_opts *opts = &rec->opts;
484
	struct perf_data_file *file = &rec->file;
485
	struct perf_session *session;
486
	bool disabled = false, draining = false;
487
	int fd;
488

489
	rec->progname = argv[0];
490

491
	atexit(record__sig_exit);
492 493
	signal(SIGCHLD, sig_handler);
	signal(SIGINT, sig_handler);
494
	signal(SIGTERM, sig_handler);
495 496 497 498
	if (rec->opts.auxtrace_snapshot_mode)
		signal(SIGUSR2, snapshot_sig_handler);
	else
		signal(SIGUSR2, SIG_IGN);
499

500
	session = perf_session__new(file, false, tool);
501
	if (session == NULL) {
A
Adrien BAK 已提交
502
		pr_err("Perf session creation failed.\n");
503 504 505
		return -1;
	}

506
	fd = perf_data_file__fd(file);
507 508
	rec->session = session;

509
	record__init_features(rec);
510

511
	if (forks) {
512
		err = perf_evlist__prepare_workload(rec->evlist, &opts->target,
513
						    argv, file->is_pipe,
514
						    workload_exec_failed_signal);
515 516
		if (err < 0) {
			pr_err("Couldn't run the workload!\n");
517
			status = err;
518
			goto out_delete_session;
519 520 521
		}
	}

522
	if (record__open(rec) != 0) {
523
		err = -1;
524
		goto out_child;
525
	}
526

527
	if (!rec->evlist->nr_groups)
528 529
		perf_header__clear_feat(&session->header, HEADER_GROUP_DESC);

530
	if (file->is_pipe) {
531
		err = perf_header__write_pipe(fd);
532
		if (err < 0)
533
			goto out_child;
534
	} else {
535
		err = perf_session__write_header(session, rec->evlist, fd, false);
536
		if (err < 0)
537
			goto out_child;
538 539
	}

540
	if (!rec->no_buildid
541
	    && !perf_header__has_feat(&session->header, HEADER_BUILD_ID)) {
542
		pr_err("Couldn't generate buildids. "
543
		       "Use --no-buildid to profile anyway.\n");
544
		err = -1;
545
		goto out_child;
546 547
	}

548
	machine = &session->machines.host;
549

550
	if (file->is_pipe) {
551
		err = perf_event__synthesize_attrs(tool, session,
552
						   process_synthesized_event);
553 554
		if (err < 0) {
			pr_err("Couldn't synthesize attrs.\n");
555
			goto out_child;
556
		}
557

558
		if (have_tracepoints(&rec->evlist->entries)) {
559 560 561 562 563 564 565 566
			/*
			 * FIXME err <= 0 here actually means that
			 * there were no tracepoints so its not really
			 * an error, just that we don't need to
			 * synthesize anything.  We really have to
			 * return this more properly and also
			 * propagate errors that now are calling die()
			 */
567
			err = perf_event__synthesize_tracing_data(tool,	fd, rec->evlist,
568
								  process_synthesized_event);
569 570
			if (err <= 0) {
				pr_err("Couldn't record tracing data.\n");
571
				goto out_child;
572
			}
573
			rec->bytes_written += err;
574
		}
575 576
	}

577 578 579 580 581 582 583
	if (rec->opts.full_auxtrace) {
		err = perf_event__synthesize_auxtrace_info(rec->itr, tool,
					session, process_synthesized_event);
		if (err)
			goto out_delete_session;
	}

584
	err = perf_event__synthesize_kernel_mmap(tool, process_synthesized_event,
585
						 machine);
586 587 588 589
	if (err < 0)
		pr_err("Couldn't record kernel reference relocation symbol\n"
		       "Symbol resolution may be skewed if relocation was used (e.g. kexec).\n"
		       "Check /proc/kallsyms permission or run as root.\n");
590

591
	err = perf_event__synthesize_modules(tool, process_synthesized_event,
592
					     machine);
593 594 595 596 597
	if (err < 0)
		pr_err("Couldn't record kernel module information.\n"
		       "Symbol resolution may be skewed if relocation was used (e.g. kexec).\n"
		       "Check /proc/modules permission or run as root.\n");

598
	if (perf_guest) {
599 600
		machines__process_guests(&session->machines,
					 perf_event__synthesize_guest_os, tool);
601
	}
602

603
	err = __machine__synthesize_threads(machine, tool, &opts->target, rec->evlist->threads,
604
					    process_synthesized_event, opts->sample_address);
605
	if (err != 0)
606
		goto out_child;
607

608
	if (rec->realtime_prio) {
609 610
		struct sched_param param;

611
		param.sched_priority = rec->realtime_prio;
612
		if (sched_setscheduler(0, SCHED_FIFO, &param)) {
613
			pr_err("Could not set realtime priority.\n");
614
			err = -1;
615
			goto out_child;
616 617 618
		}
	}

619 620 621 622 623
	/*
	 * When perf is starting the traced process, all the events
	 * (apart from group members) have enable_on_exec=1 set,
	 * so don't spoil it by prematurely enabling them.
	 */
624
	if (!target__none(&opts->target) && !opts->initial_delay)
625
		perf_evlist__enable(rec->evlist);
626

627 628 629
	/*
	 * Let the child rip
	 */
630
	if (forks)
631
		perf_evlist__start_workload(rec->evlist);
632

633 634 635 636 637
	if (opts->initial_delay) {
		usleep(opts->initial_delay * 1000);
		perf_evlist__enable(rec->evlist);
	}

638
	auxtrace_snapshot_enabled = 1;
639
	for (;;) {
640
		int hits = rec->samples;
641

642
		if (record__mmap_read_all(rec) < 0) {
643
			auxtrace_snapshot_enabled = 0;
644
			err = -1;
645
			goto out_child;
646
		}
647

648 649 650 651 652 653 654 655 656 657 658
		if (auxtrace_record__snapshot_started) {
			auxtrace_record__snapshot_started = 0;
			if (!auxtrace_snapshot_err)
				record__read_auxtrace_snapshot(rec);
			if (auxtrace_snapshot_err) {
				pr_err("AUX area tracing snapshot failed\n");
				err = -1;
				goto out_child;
			}
		}

659
		if (hits == rec->samples) {
660
			if (done || draining)
661
				break;
662
			err = perf_evlist__poll(rec->evlist, -1);
663 664 665 666 667
			/*
			 * Propagate error, only if there's any. Ignore positive
			 * number of returned events and interrupt error.
			 */
			if (err > 0 || (err < 0 && errno == EINTR))
668
				err = 0;
669
			waking++;
670 671 672

			if (perf_evlist__filter_pollfd(rec->evlist, POLLERR | POLLHUP) == 0)
				draining = true;
673 674
		}

675 676 677 678 679
		/*
		 * When perf is starting the traced process, at the end events
		 * die with the process and we wait for that. Thus no need to
		 * disable events in this case.
		 */
680
		if (done && !disabled && !target__none(&opts->target)) {
681
			auxtrace_snapshot_enabled = 0;
682
			perf_evlist__disable(rec->evlist);
683 684
			disabled = true;
		}
685
	}
686
	auxtrace_snapshot_enabled = 0;
687

688
	if (forks && workload_exec_errno) {
689
		char msg[STRERR_BUFSIZE];
690 691 692
		const char *emsg = strerror_r(workload_exec_errno, msg, sizeof(msg));
		pr_err("Workload failed: %s\n", emsg);
		err = -1;
693
		goto out_child;
694 695
	}

696
	if (!quiet)
697
		fprintf(stderr, "[ perf record: Woken up %ld times to write data ]\n", waking);
698

699 700 701
out_child:
	if (forks) {
		int exit_status;
702

703 704 705 706 707 708 709 710 711 712 713 714 715 716
		if (!child_finished)
			kill(rec->evlist->workload.pid, SIGTERM);

		wait(&exit_status);

		if (err < 0)
			status = err;
		else if (WIFEXITED(exit_status))
			status = WEXITSTATUS(exit_status);
		else if (WIFSIGNALED(exit_status))
			signr = WTERMSIG(exit_status);
	} else
		status = err;

717 718 719
	/* this will be recalculated during process_buildids() */
	rec->samples = 0;

720 721 722
	if (!err && !file->is_pipe) {
		rec->session->header.data_size += rec->bytes_written;

723
		if (!rec->no_buildid) {
724
			process_buildids(rec);
725 726 727 728 729 730 731 732
			/*
			 * We take all buildids when the file contains
			 * AUX area tracing data because we do not decode the
			 * trace because it would take too long.
			 */
			if (rec->opts.full_auxtrace)
				dsos__hit_all(rec->session);
		}
733
		perf_session__write_header(rec->session, rec->evlist, fd, true);
734
	}
735

736 737 738
	if (!err && !quiet) {
		char samples[128];

739
		if (rec->samples && !rec->opts.full_auxtrace)
740 741 742 743 744 745 746 747 748 749
			scnprintf(samples, sizeof(samples),
				  " (%" PRIu64 " samples)", rec->samples);
		else
			samples[0] = '\0';

		fprintf(stderr,	"[ perf record: Captured and wrote %.3f MB %s%s ]\n",
			perf_data_file__size(file) / 1024.0 / 1024.0,
			file->path, samples);
	}

750 751
out_delete_session:
	perf_session__delete(session);
752
	return status;
753
}
754

755
static void callchain_debug(void)
J
Jiri Olsa 已提交
756
{
757
	static const char *str[CALLCHAIN_MAX] = { "NONE", "FP", "DWARF", "LBR" };
758

759
	pr_debug("callchain: type %s\n", str[callchain_param.record_mode]);
760

761
	if (callchain_param.record_mode == CALLCHAIN_DWARF)
J
Jiri Olsa 已提交
762
		pr_debug("callchain: stack dump size %d\n",
763
			 callchain_param.dump_size);
J
Jiri Olsa 已提交
764 765
}

766
int record_parse_callchain_opt(const struct option *opt __maybe_unused,
J
Jiri Olsa 已提交
767 768 769 770 771
			       const char *arg,
			       int unset)
{
	int ret;

772
	callchain_param.enabled = !unset;
773

J
Jiri Olsa 已提交
774 775
	/* --no-call-graph */
	if (unset) {
776
		callchain_param.record_mode = CALLCHAIN_NONE;
J
Jiri Olsa 已提交
777 778 779 780
		pr_debug("callchain: disabled\n");
		return 0;
	}

781
	ret = parse_callchain_record_opt(arg);
782
	if (!ret)
783
		callchain_debug();
784 785 786 787

	return ret;
}

788
int record_callchain_opt(const struct option *opt __maybe_unused,
J
Jiri Olsa 已提交
789 790 791
			 const char *arg __maybe_unused,
			 int unset __maybe_unused)
{
792
	callchain_param.enabled = true;
J
Jiri Olsa 已提交
793

794 795
	if (callchain_param.record_mode == CALLCHAIN_NONE)
		callchain_param.record_mode = CALLCHAIN_FP;
796

797
	callchain_debug();
J
Jiri Olsa 已提交
798 799 800
	return 0;
}

801 802 803
static int perf_record_config(const char *var, const char *value, void *cb)
{
	if (!strcmp(var, "record.call-graph"))
804
		var = "call-graph.record-mode"; /* fall-through */
805 806 807 808

	return perf_default_config(var, value, cb);
}

809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892
struct clockid_map {
	const char *name;
	int clockid;
};

#define CLOCKID_MAP(n, c)	\
	{ .name = n, .clockid = (c), }

#define CLOCKID_END	{ .name = NULL, }


/*
 * Add the missing ones, we need to build on many distros...
 */
#ifndef CLOCK_MONOTONIC_RAW
#define CLOCK_MONOTONIC_RAW 4
#endif
#ifndef CLOCK_BOOTTIME
#define CLOCK_BOOTTIME 7
#endif
#ifndef CLOCK_TAI
#define CLOCK_TAI 11
#endif

static const struct clockid_map clockids[] = {
	/* available for all events, NMI safe */
	CLOCKID_MAP("monotonic", CLOCK_MONOTONIC),
	CLOCKID_MAP("monotonic_raw", CLOCK_MONOTONIC_RAW),

	/* available for some events */
	CLOCKID_MAP("realtime", CLOCK_REALTIME),
	CLOCKID_MAP("boottime", CLOCK_BOOTTIME),
	CLOCKID_MAP("tai", CLOCK_TAI),

	/* available for the lazy */
	CLOCKID_MAP("mono", CLOCK_MONOTONIC),
	CLOCKID_MAP("raw", CLOCK_MONOTONIC_RAW),
	CLOCKID_MAP("real", CLOCK_REALTIME),
	CLOCKID_MAP("boot", CLOCK_BOOTTIME),

	CLOCKID_END,
};

static int parse_clockid(const struct option *opt, const char *str, int unset)
{
	struct record_opts *opts = (struct record_opts *)opt->value;
	const struct clockid_map *cm;
	const char *ostr = str;

	if (unset) {
		opts->use_clockid = 0;
		return 0;
	}

	/* no arg passed */
	if (!str)
		return 0;

	/* no setting it twice */
	if (opts->use_clockid)
		return -1;

	opts->use_clockid = true;

	/* if its a number, we're done */
	if (sscanf(str, "%d", &opts->clockid) == 1)
		return 0;

	/* allow a "CLOCK_" prefix to the name */
	if (!strncasecmp(str, "CLOCK_", 6))
		str += 6;

	for (cm = clockids; cm->name; cm++) {
		if (!strcasecmp(str, cm->name)) {
			opts->clockid = cm->clockid;
			return 0;
		}
	}

	opts->use_clockid = false;
	ui__warning("unknown clockid %s, check man page\n", ostr);
	return -1;
}

893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935
static int record__parse_mmap_pages(const struct option *opt,
				    const char *str,
				    int unset __maybe_unused)
{
	struct record_opts *opts = opt->value;
	char *s, *p;
	unsigned int mmap_pages;
	int ret;

	if (!str)
		return -EINVAL;

	s = strdup(str);
	if (!s)
		return -ENOMEM;

	p = strchr(s, ',');
	if (p)
		*p = '\0';

	if (*s) {
		ret = __perf_evlist__parse_mmap_pages(&mmap_pages, s);
		if (ret)
			goto out_free;
		opts->mmap_pages = mmap_pages;
	}

	if (!p) {
		ret = 0;
		goto out_free;
	}

	ret = __perf_evlist__parse_mmap_pages(&mmap_pages, p + 1);
	if (ret)
		goto out_free;

	opts->auxtrace_mmap_pages = mmap_pages;

out_free:
	free(s);
	return ret;
}

936
static const char * const __record_usage[] = {
937 938
	"perf record [<options>] [<command>]",
	"perf record [<options>] -- <command> [<options>]",
939 940
	NULL
};
941
const char * const *record_usage = __record_usage;
942

943
/*
944 945
 * XXX Ideally would be local to cmd_record() and passed to a record__new
 * because we need to have access to it in record__exit, that is called
946 947 948 949 950 951 952
 * after cmd_record() exits, but since record_options need to be accessible to
 * builtin-script, leave it here.
 *
 * At least we don't ouch it in all the other functions here directly.
 *
 * Just say no to tons of global variables, sigh.
 */
953
static struct record record = {
954
	.opts = {
955
		.sample_time	     = true,
956 957 958
		.mmap_pages	     = UINT_MAX,
		.user_freq	     = UINT_MAX,
		.user_interval	     = ULLONG_MAX,
959
		.freq		     = 4000,
N
Namhyung Kim 已提交
960 961
		.target		     = {
			.uses_mmap   = true,
962
			.default_per_cpu = true,
N
Namhyung Kim 已提交
963
		},
964
	},
965 966 967 968 969 970 971
	.tool = {
		.sample		= process_sample_event,
		.fork		= perf_event__process_fork,
		.comm		= perf_event__process_comm,
		.mmap		= perf_event__process_mmap,
		.mmap2		= perf_event__process_mmap2,
	},
972
};
973

J
Jiri Olsa 已提交
974
#define CALLCHAIN_HELP "setup and enables call-graph (stack chain/backtrace) recording: "
975

976
#ifdef HAVE_DWARF_UNWIND_SUPPORT
977
const char record_callchain_help[] = CALLCHAIN_HELP "fp dwarf lbr";
978
#else
979
const char record_callchain_help[] = CALLCHAIN_HELP "fp lbr";
980 981
#endif

982 983 984
/*
 * XXX Will stay a global variable till we fix builtin-script.c to stop messing
 * with it and switch to use the library functions in perf_evlist that came
985
 * from builtin-record.c, i.e. use record_opts,
986 987 988
 * perf_evlist__prepare_workload, etc instead of fork+exec'in 'perf record',
 * using pipes, etc.
 */
989
struct option __record_options[] = {
990
	OPT_CALLBACK('e', "event", &record.evlist, "event",
991
		     "event selector. use 'perf list' to list available events",
992
		     parse_events_option),
993
	OPT_CALLBACK(0, "filter", &record.evlist, "filter",
L
Li Zefan 已提交
994
		     "event filter", parse_filter),
995
	OPT_STRING('p', "pid", &record.opts.target.pid, "pid",
996
		    "record events on existing process id"),
997
	OPT_STRING('t', "tid", &record.opts.target.tid, "tid",
998
		    "record events on existing thread id"),
999
	OPT_INTEGER('r', "realtime", &record.realtime_prio,
1000
		    "collect data with this RT SCHED_FIFO priority"),
1001
	OPT_BOOLEAN(0, "no-buffering", &record.opts.no_buffering,
1002
		    "collect data without buffering"),
1003
	OPT_BOOLEAN('R', "raw-samples", &record.opts.raw_samples,
1004
		    "collect raw sample records from all opened counters"),
1005
	OPT_BOOLEAN('a', "all-cpus", &record.opts.target.system_wide,
1006
			    "system-wide collection from all CPUs"),
1007
	OPT_STRING('C', "cpu", &record.opts.target.cpu_list, "cpu",
1008
		    "list of cpus to monitor"),
1009
	OPT_U64('c', "count", &record.opts.user_interval, "event period to sample"),
1010
	OPT_STRING('o', "output", &record.file.path, "file",
I
Ingo Molnar 已提交
1011
		    "output file name"),
1012 1013 1014
	OPT_BOOLEAN_SET('i', "no-inherit", &record.opts.no_inherit,
			&record.opts.no_inherit_set,
			"child tasks do not inherit counters"),
1015
	OPT_UINTEGER('F', "freq", &record.opts.user_freq, "profile at this frequency"),
1016 1017 1018
	OPT_CALLBACK('m', "mmap-pages", &record.opts, "pages[,pages]",
		     "number of mmap data pages and AUX area tracing mmap pages",
		     record__parse_mmap_pages),
1019
	OPT_BOOLEAN(0, "group", &record.opts.group,
1020
		    "put the counters into a counter group"),
J
Jiri Olsa 已提交
1021 1022 1023 1024 1025 1026
	OPT_CALLBACK_NOOPT('g', NULL, &record.opts,
			   NULL, "enables call-graph recording" ,
			   &record_callchain_opt),
	OPT_CALLBACK(0, "call-graph", &record.opts,
		     "mode[,dump_size]", record_callchain_help,
		     &record_parse_callchain_opt),
1027
	OPT_INCR('v', "verbose", &verbose,
1028
		    "be more verbose (show counter open errors, etc)"),
1029
	OPT_BOOLEAN('q', "quiet", &quiet, "don't print any message"),
1030
	OPT_BOOLEAN('s', "stat", &record.opts.inherit_stat,
1031
		    "per thread counts"),
1032
	OPT_BOOLEAN('d', "data", &record.opts.sample_address,
1033
		    "Sample addresses"),
1034
	OPT_BOOLEAN('T', "timestamp", &record.opts.sample_time, "Sample timestamps"),
1035
	OPT_BOOLEAN('P', "period", &record.opts.period, "Sample period"),
1036
	OPT_BOOLEAN('n', "no-samples", &record.opts.no_samples,
1037
		    "don't sample"),
1038
	OPT_BOOLEAN('N', "no-buildid-cache", &record.no_buildid_cache,
1039
		    "do not update the buildid cache"),
1040
	OPT_BOOLEAN('B', "no-buildid", &record.no_buildid,
1041
		    "do not collect buildids in perf.data"),
1042
	OPT_CALLBACK('G', "cgroup", &record.evlist, "name",
S
Stephane Eranian 已提交
1043 1044
		     "monitor event in cgroup name only",
		     parse_cgroups),
1045
	OPT_UINTEGER('D', "delay", &record.opts.initial_delay,
1046
		  "ms to wait before starting measurement after program start"),
1047 1048
	OPT_STRING('u', "uid", &record.opts.target.uid_str, "user",
		   "user to profile"),
1049 1050 1051 1052 1053 1054 1055

	OPT_CALLBACK_NOOPT('b', "branch-any", &record.opts.branch_stack,
		     "branch any", "sample any taken branches",
		     parse_branch_stack),

	OPT_CALLBACK('j', "branch-filter", &record.opts.branch_stack,
		     "branch filter mask", "branch stack filter modes",
1056
		     parse_branch_stack),
1057 1058
	OPT_BOOLEAN('W', "weight", &record.opts.sample_weight,
		    "sample by weight (on special events only)"),
1059 1060
	OPT_BOOLEAN(0, "transaction", &record.opts.sample_transaction,
		    "sample transaction flags (special events only)"),
1061 1062
	OPT_BOOLEAN(0, "per-thread", &record.opts.target.per_thread,
		    "use per-thread mmaps"),
1063 1064
	OPT_BOOLEAN('I', "intr-regs", &record.opts.sample_intr_regs,
		    "Sample machine registers on interrupt"),
1065 1066
	OPT_BOOLEAN(0, "running-time", &record.opts.running_time,
		    "Record running/enabled time of read (:S) events"),
1067 1068 1069
	OPT_CALLBACK('k', "clockid", &record.opts,
	"clockid", "clockid to use for events, see clock_gettime()",
	parse_clockid),
1070 1071
	OPT_STRING_OPTARG('S', "snapshot", &record.opts.auxtrace_snapshot_opts,
			  "opts", "AUX area tracing Snapshot Mode", ""),
1072 1073 1074
	OPT_END()
};

1075 1076
struct option *record_options = __record_options;

1077
int cmd_record(int argc, const char **argv, const char *prefix __maybe_unused)
1078
{
1079
	int err;
1080
	struct record *rec = &record;
1081
	char errbuf[BUFSIZ];
1082

1083 1084
	rec->evlist = perf_evlist__new();
	if (rec->evlist == NULL)
1085 1086
		return -ENOMEM;

1087 1088
	perf_config(perf_record_config, rec);

1089
	argc = parse_options(argc, argv, record_options, record_usage,
1090
			    PARSE_OPT_STOP_AT_NON_OPTION);
1091
	if (!argc && target__none(&rec->opts.target))
1092
		usage_with_options(record_usage, record_options);
1093

1094
	if (nr_cgroups && !rec->opts.target.system_wide) {
1095 1096
		ui__error("cgroup monitoring only available in"
			  " system-wide mode\n");
S
Stephane Eranian 已提交
1097 1098 1099
		usage_with_options(record_usage, record_options);
	}

1100 1101 1102 1103 1104 1105
	if (!rec->itr) {
		rec->itr = auxtrace_record__init(rec->evlist, &err);
		if (err)
			return err;
	}

1106 1107 1108 1109 1110
	err = auxtrace_parse_snapshot_options(rec->itr, &rec->opts,
					      rec->opts.auxtrace_snapshot_opts);
	if (err)
		return err;

1111 1112
	err = -ENOMEM;

1113
	symbol__init(NULL);
1114

1115
	if (symbol_conf.kptr_restrict)
1116 1117 1118 1119 1120 1121 1122 1123
		pr_warning(
"WARNING: Kernel address maps (/proc/{kallsyms,modules}) are restricted,\n"
"check /proc/sys/kernel/kptr_restrict.\n\n"
"Samples in kernel functions may not be resolved if a suitable vmlinux\n"
"file is not found in the buildid cache or in the vmlinux path.\n\n"
"Samples in kernel modules won't be resolved at all.\n\n"
"If some relocation was applied (e.g. kexec) symbols may be misresolved\n"
"even with a suitable vmlinux or kallsyms file.\n\n");
1124

1125
	if (rec->no_buildid_cache || rec->no_buildid)
1126
		disable_buildid_cache();
1127

1128 1129
	if (rec->evlist->nr_entries == 0 &&
	    perf_evlist__add_default(rec->evlist) < 0) {
1130 1131
		pr_err("Not enough memory for event selector list\n");
		goto out_symbol_exit;
1132
	}
1133

1134 1135 1136
	if (rec->opts.target.tid && !rec->opts.no_inherit_set)
		rec->opts.no_inherit = true;

1137
	err = target__validate(&rec->opts.target);
1138
	if (err) {
1139
		target__strerror(&rec->opts.target, err, errbuf, BUFSIZ);
1140 1141 1142
		ui__warning("%s", errbuf);
	}

1143
	err = target__parse_uid(&rec->opts.target);
1144 1145
	if (err) {
		int saved_errno = errno;
1146

1147
		target__strerror(&rec->opts.target, err, errbuf, BUFSIZ);
1148
		ui__error("%s", errbuf);
1149 1150

		err = -saved_errno;
1151
		goto out_symbol_exit;
1152
	}
1153

1154
	err = -ENOMEM;
1155
	if (perf_evlist__create_maps(rec->evlist, &rec->opts.target) < 0)
1156
		usage_with_options(record_usage, record_options);
1157

1158 1159 1160 1161
	err = auxtrace_record__options(rec->itr, rec->evlist, &rec->opts);
	if (err)
		goto out_symbol_exit;

1162
	if (record_opts__config(&rec->opts)) {
1163
		err = -EINVAL;
1164
		goto out_symbol_exit;
1165 1166
	}

1167
	err = __cmd_record(&record, argc, argv);
1168
out_symbol_exit:
1169
	perf_evlist__delete(rec->evlist);
1170
	symbol__exit();
1171
	auxtrace_record__free(rec->itr);
1172
	return err;
1173
}
1174 1175 1176 1177 1178 1179 1180 1181 1182

static void snapshot_sig_handler(int sig __maybe_unused)
{
	if (!auxtrace_snapshot_enabled)
		return;
	auxtrace_snapshot_enabled = 0;
	auxtrace_snapshot_err = auxtrace_record__snapshot_start(record.itr);
	auxtrace_record__snapshot_started = 1;
}