builtin-record.c 31.7 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/perf_regs.h"
31
#include "util/auxtrace.h"
32
#include "util/parse-branch-options.h"
33
#include "util/parse-regs-options.h"
34

35
#include <unistd.h>
36
#include <sched.h>
37
#include <sys/mman.h>
38

39

40
struct record {
41
	struct perf_tool	tool;
42
	struct record_opts	opts;
43
	u64			bytes_written;
44
	struct perf_data_file	file;
45
	struct auxtrace_record	*itr;
46 47 48 49 50 51 52
	struct perf_evlist	*evlist;
	struct perf_session	*session;
	const char		*progname;
	int			realtime_prio;
	bool			no_buildid;
	bool			no_buildid_cache;
	long			samples;
53
};
54

55
static int record__write(struct record *rec, void *bf, size_t size)
56
{
57
	if (perf_data_file__write(rec->session->file, bf, size) < 0) {
58 59
		pr_err("failed to write perf data, error: %m\n");
		return -1;
60
	}
61

62
	rec->bytes_written += size;
63
	return 0;
64 65
}

66
static int process_synthesized_event(struct perf_tool *tool,
67
				     union perf_event *event,
68 69
				     struct perf_sample *sample __maybe_unused,
				     struct machine *machine __maybe_unused)
70
{
71 72
	struct record *rec = container_of(tool, struct record, tool);
	return record__write(rec, event, event->header.size);
73 74
}

75
static int record__mmap_read(struct record *rec, int idx)
76
{
77
	struct perf_mmap *md = &rec->evlist->mmap[idx];
78 79
	u64 head = perf_mmap__read_head(md);
	u64 old = md->prev;
J
Jiri Olsa 已提交
80
	unsigned char *data = md->base + page_size;
81 82
	unsigned long size;
	void *buf;
83
	int rc = 0;
84

85
	if (old == head)
86
		return 0;
87

88
	rec->samples++;
89 90 91 92 93 94 95

	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;
96

97
		if (record__write(rec, buf, size) < 0) {
98 99 100
			rc = -1;
			goto out;
		}
101 102 103 104 105
	}

	buf = &data[old & md->mask];
	size = head - old;
	old += size;
106

107
	if (record__write(rec, buf, size) < 0) {
108 109 110
		rc = -1;
		goto out;
	}
111 112

	md->prev = old;
113
	perf_evlist__mmap_consume(rec->evlist, idx);
114 115
out:
	return rc;
116 117
}

118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143
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);
}

144 145
#ifdef HAVE_AUXTRACE_SUPPORT

146 147 148 149 150
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);
151
	struct perf_data_file *file = &rec->file;
152 153 154
	size_t padding;
	u8 pad[8] = {0};

155 156 157 158 159 160 161 162 163 164 165 166 167 168
	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;
	}

169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198
	/* 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;
}

199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248
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;
	}
}

249 250 251 252 253 254 255 256 257
#else

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

258 259
static inline
void record__read_auxtrace_snapshot(struct record *rec __maybe_unused)
260
{
261 262
}

263 264
static inline
int auxtrace_record__snapshot_start(struct auxtrace_record *itr __maybe_unused)
265
{
266
	return 0;
267 268
}

269 270
#endif

271
static int record__open(struct record *rec)
272
{
273
	char msg[512];
274
	struct perf_evsel *pos;
275 276
	struct perf_evlist *evlist = rec->evlist;
	struct perf_session *session = rec->session;
277
	struct record_opts *opts = &rec->opts;
278
	int rc = 0;
279

280
	perf_evlist__config(evlist, opts);
281

282
	evlist__for_each(evlist, pos) {
283
try_again:
284
		if (perf_evsel__open(pos, pos->cpus, pos->threads) < 0) {
285
			if (perf_evsel__fallback(pos, errno, msg, sizeof(msg))) {
286
				if (verbose)
287
					ui__warning("%s\n", msg);
288 289
				goto try_again;
			}
290

291 292 293 294
			rc = -errno;
			perf_evsel__open_strerror(pos, &opts->target,
						  errno, msg, sizeof(msg));
			ui__error("%s\n", msg);
295
			goto out;
L
Li Zefan 已提交
296 297
		}
	}
298

299 300 301
	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,
302
			strerror_r(errno, msg, sizeof(msg)));
303 304
		rc = -1;
		goto out;
305 306
	}

307
	if (perf_evlist__mmap_ex(evlist, opts->mmap_pages, false,
308 309
				 opts->auxtrace_mmap_pages,
				 opts->auxtrace_snapshot_mode) < 0) {
310 311 312 313 314
		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"
315 316
			       "(current value: %u,%u)\n",
			       opts->mmap_pages, opts->auxtrace_mmap_pages);
317 318
			rc = -errno;
		} else {
319 320
			pr_err("failed to mmap with %d (%s)\n", errno,
				strerror_r(errno, msg, sizeof(msg)));
321 322 323
			rc = -errno;
		}
		goto out;
324
	}
325

326
	session->evlist = evlist;
327
	perf_session__set_id_hdr_size(session);
328 329
out:
	return rc;
330 331
}

332 333 334 335 336 337 338 339 340 341 342 343 344
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);
}

345
static int process_buildids(struct record *rec)
346
{
347 348
	struct perf_data_file *file  = &rec->file;
	struct perf_session *session = rec->session;
349

350
	if (file->size == 0)
351 352
		return 0;

353 354 355 356 357 358 359 360 361 362 363
	/*
	 * 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;

364
	return perf_session__process_events(session);
365 366
}

367
static void perf_event__synthesize_guest_os(struct machine *machine, void *data)
368 369
{
	int err;
370
	struct perf_tool *tool = data;
371 372 373 374 375 376 377 378
	/*
	 *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.
	 */
379
	err = perf_event__synthesize_modules(tool, process_synthesized_event,
380
					     machine);
381 382
	if (err < 0)
		pr_err("Couldn't record guest kernel [%d]'s reference"
383
		       " relocation symbol.\n", machine->pid);
384 385 386 387 388

	/*
	 * We use _stext for guest kernel because guest kernel's /proc/kallsyms
	 * have no _text sometimes.
	 */
389
	err = perf_event__synthesize_kernel_mmap(tool, process_synthesized_event,
390
						 machine);
391 392
	if (err < 0)
		pr_err("Couldn't record guest kernel [%d]'s reference"
393
		       " relocation symbol.\n", machine->pid);
394 395
}

396 397 398 399 400
static struct perf_event_header finished_round_event = {
	.size = sizeof(struct perf_event_header),
	.type = PERF_RECORD_FINISHED_ROUND,
};

401
static int record__mmap_read_all(struct record *rec)
402
{
403
	u64 bytes_written = rec->bytes_written;
404
	int i;
405
	int rc = 0;
406

407
	for (i = 0; i < rec->evlist->nr_mmaps; i++) {
408 409
		struct auxtrace_mmap *mm = &rec->evlist->mmap[i].auxtrace_mmap;

410
		if (rec->evlist->mmap[i].base) {
411
			if (record__mmap_read(rec, i) != 0) {
412 413 414 415
				rc = -1;
				goto out;
			}
		}
416

417
		if (mm->base && !rec->opts.auxtrace_snapshot_mode &&
418 419 420 421
		    record__auxtrace_mmap_read(rec, mm) != 0) {
			rc = -1;
			goto out;
		}
422 423
	}

424 425 426 427 428 429
	/*
	 * 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));
430 431 432

out:
	return rc;
433 434
}

435
static void record__init_features(struct record *rec)
436 437 438 439 440 441 442 443 444 445
{
	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);

446
	if (!have_tracepoints(&rec->evlist->entries))
447 448 449 450
		perf_header__clear_feat(&session->header, HEADER_TRACING_DATA);

	if (!rec->opts.branch_stack)
		perf_header__clear_feat(&session->header, HEADER_BRANCH_STACK);
451 452 453

	if (!rec->opts.full_auxtrace)
		perf_header__clear_feat(&session->header, HEADER_AUXTRACE);
454 455
}

456 457 458 459 460 461 462
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.
 */
463 464
static void workload_exec_failed_signal(int signo __maybe_unused,
					siginfo_t *info,
465 466 467 468 469 470 471
					void *ucontext __maybe_unused)
{
	workload_exec_errno = info->si_value.sival_int;
	done = 1;
	child_finished = 1;
}

472 473
static void snapshot_sig_handler(int sig);

474
static int __cmd_record(struct record *rec, int argc, const char **argv)
475
{
476
	int err;
477
	int status = 0;
478
	unsigned long waking = 0;
479
	const bool forks = argc > 0;
480
	struct machine *machine;
481
	struct perf_tool *tool = &rec->tool;
482
	struct record_opts *opts = &rec->opts;
483
	struct perf_data_file *file = &rec->file;
484
	struct perf_session *session;
485
	bool disabled = false, draining = false;
486
	int fd;
487

488
	rec->progname = argv[0];
489

490
	atexit(record__sig_exit);
491 492
	signal(SIGCHLD, sig_handler);
	signal(SIGINT, sig_handler);
493
	signal(SIGTERM, sig_handler);
494 495 496 497
	if (rec->opts.auxtrace_snapshot_mode)
		signal(SIGUSR2, snapshot_sig_handler);
	else
		signal(SIGUSR2, SIG_IGN);
498

499
	session = perf_session__new(file, false, tool);
500
	if (session == NULL) {
A
Adrien BAK 已提交
501
		pr_err("Perf session creation failed.\n");
502 503 504
		return -1;
	}

505
	fd = perf_data_file__fd(file);
506 507
	rec->session = session;

508
	record__init_features(rec);
509

510
	if (forks) {
511
		err = perf_evlist__prepare_workload(rec->evlist, &opts->target,
512
						    argv, file->is_pipe,
513
						    workload_exec_failed_signal);
514 515
		if (err < 0) {
			pr_err("Couldn't run the workload!\n");
516
			status = err;
517
			goto out_delete_session;
518 519 520
		}
	}

521
	if (record__open(rec) != 0) {
522
		err = -1;
523
		goto out_child;
524
	}
525

526 527 528 529 530 531 532 533 534
	/*
	 * Normally perf_session__new would do this, but it doesn't have the
	 * evlist.
	 */
	if (rec->tool.ordered_events && !perf_evlist__sample_id_all(rec->evlist)) {
		pr_warning("WARNING: No sample_id_all support, falling back to unordered processing\n");
		rec->tool.ordered_events = false;
	}

535
	if (!rec->evlist->nr_groups)
536 537
		perf_header__clear_feat(&session->header, HEADER_GROUP_DESC);

538
	if (file->is_pipe) {
539
		err = perf_header__write_pipe(fd);
540
		if (err < 0)
541
			goto out_child;
542
	} else {
543
		err = perf_session__write_header(session, rec->evlist, fd, false);
544
		if (err < 0)
545
			goto out_child;
546 547
	}

548
	if (!rec->no_buildid
549
	    && !perf_header__has_feat(&session->header, HEADER_BUILD_ID)) {
550
		pr_err("Couldn't generate buildids. "
551
		       "Use --no-buildid to profile anyway.\n");
552
		err = -1;
553
		goto out_child;
554 555
	}

556
	machine = &session->machines.host;
557

558
	if (file->is_pipe) {
559
		err = perf_event__synthesize_attrs(tool, session,
560
						   process_synthesized_event);
561 562
		if (err < 0) {
			pr_err("Couldn't synthesize attrs.\n");
563
			goto out_child;
564
		}
565

566
		if (have_tracepoints(&rec->evlist->entries)) {
567 568 569 570 571 572 573 574
			/*
			 * 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()
			 */
575
			err = perf_event__synthesize_tracing_data(tool,	fd, rec->evlist,
576
								  process_synthesized_event);
577 578
			if (err <= 0) {
				pr_err("Couldn't record tracing data.\n");
579
				goto out_child;
580
			}
581
			rec->bytes_written += err;
582
		}
583 584
	}

585 586 587 588 589 590 591
	if (rec->opts.full_auxtrace) {
		err = perf_event__synthesize_auxtrace_info(rec->itr, tool,
					session, process_synthesized_event);
		if (err)
			goto out_delete_session;
	}

592
	err = perf_event__synthesize_kernel_mmap(tool, process_synthesized_event,
593
						 machine);
594 595 596 597
	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");
598

599
	err = perf_event__synthesize_modules(tool, process_synthesized_event,
600
					     machine);
601 602 603 604 605
	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");

606
	if (perf_guest) {
607 608
		machines__process_guests(&session->machines,
					 perf_event__synthesize_guest_os, tool);
609
	}
610

611
	err = __machine__synthesize_threads(machine, tool, &opts->target, rec->evlist->threads,
612 613
					    process_synthesized_event, opts->sample_address,
					    opts->proc_map_timeout);
614
	if (err != 0)
615
		goto out_child;
616

617
	if (rec->realtime_prio) {
618 619
		struct sched_param param;

620
		param.sched_priority = rec->realtime_prio;
621
		if (sched_setscheduler(0, SCHED_FIFO, &param)) {
622
			pr_err("Could not set realtime priority.\n");
623
			err = -1;
624
			goto out_child;
625 626 627
		}
	}

628 629 630 631 632
	/*
	 * 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.
	 */
633
	if (!target__none(&opts->target) && !opts->initial_delay)
634
		perf_evlist__enable(rec->evlist);
635

636 637 638
	/*
	 * Let the child rip
	 */
639
	if (forks) {
640 641 642 643 644 645 646 647
		union perf_event *event;

		event = malloc(sizeof(event->comm) + machine->id_hdr_size);
		if (event == NULL) {
			err = -ENOMEM;
			goto out_child;
		}

648 649 650 651 652 653
		/*
		 * Some H/W events are generated before COMM event
		 * which is emitted during exec(), so perf script
		 * cannot see a correct process name for those events.
		 * Synthesize COMM event to prevent it.
		 */
654
		perf_event__synthesize_comm(tool, event,
655 656 657
					    rec->evlist->workload.pid,
					    process_synthesized_event,
					    machine);
658
		free(event);
659

660
		perf_evlist__start_workload(rec->evlist);
661
	}
662

663 664 665 666 667
	if (opts->initial_delay) {
		usleep(opts->initial_delay * 1000);
		perf_evlist__enable(rec->evlist);
	}

668
	auxtrace_snapshot_enabled = 1;
669
	for (;;) {
670
		int hits = rec->samples;
671

672
		if (record__mmap_read_all(rec) < 0) {
673
			auxtrace_snapshot_enabled = 0;
674
			err = -1;
675
			goto out_child;
676
		}
677

678 679 680 681 682 683 684 685 686 687 688
		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;
			}
		}

689
		if (hits == rec->samples) {
690
			if (done || draining)
691
				break;
692
			err = perf_evlist__poll(rec->evlist, -1);
693 694 695 696 697
			/*
			 * Propagate error, only if there's any. Ignore positive
			 * number of returned events and interrupt error.
			 */
			if (err > 0 || (err < 0 && errno == EINTR))
698
				err = 0;
699
			waking++;
700 701 702

			if (perf_evlist__filter_pollfd(rec->evlist, POLLERR | POLLHUP) == 0)
				draining = true;
703 704
		}

705 706 707 708 709
		/*
		 * 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.
		 */
710
		if (done && !disabled && !target__none(&opts->target)) {
711
			auxtrace_snapshot_enabled = 0;
712
			perf_evlist__disable(rec->evlist);
713 714
			disabled = true;
		}
715
	}
716
	auxtrace_snapshot_enabled = 0;
717

718
	if (forks && workload_exec_errno) {
719
		char msg[STRERR_BUFSIZE];
720 721 722
		const char *emsg = strerror_r(workload_exec_errno, msg, sizeof(msg));
		pr_err("Workload failed: %s\n", emsg);
		err = -1;
723
		goto out_child;
724 725
	}

726
	if (!quiet)
727
		fprintf(stderr, "[ perf record: Woken up %ld times to write data ]\n", waking);
728

729 730 731
out_child:
	if (forks) {
		int exit_status;
732

733 734 735 736 737 738 739 740 741 742 743 744 745 746
		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;

747 748 749
	/* this will be recalculated during process_buildids() */
	rec->samples = 0;

750 751
	if (!err && !file->is_pipe) {
		rec->session->header.data_size += rec->bytes_written;
752
		file->size = lseek(perf_data_file__fd(file), 0, SEEK_CUR);
753

754
		if (!rec->no_buildid) {
755
			process_buildids(rec);
756 757 758 759 760 761 762 763
			/*
			 * 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);
		}
764
		perf_session__write_header(rec->session, rec->evlist, fd, true);
765
	}
766

767 768 769
	if (!err && !quiet) {
		char samples[128];

770
		if (rec->samples && !rec->opts.full_auxtrace)
771 772 773 774 775 776 777 778 779 780
			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);
	}

781 782
out_delete_session:
	perf_session__delete(session);
783
	return status;
784
}
785

786
static void callchain_debug(void)
J
Jiri Olsa 已提交
787
{
788
	static const char *str[CALLCHAIN_MAX] = { "NONE", "FP", "DWARF", "LBR" };
789

790
	pr_debug("callchain: type %s\n", str[callchain_param.record_mode]);
791

792
	if (callchain_param.record_mode == CALLCHAIN_DWARF)
J
Jiri Olsa 已提交
793
		pr_debug("callchain: stack dump size %d\n",
794
			 callchain_param.dump_size);
J
Jiri Olsa 已提交
795 796
}

797
int record_parse_callchain_opt(const struct option *opt,
J
Jiri Olsa 已提交
798 799 800 801
			       const char *arg,
			       int unset)
{
	int ret;
802
	struct record_opts *record = (struct record_opts *)opt->value;
J
Jiri Olsa 已提交
803

804
	record->callgraph_set = true;
805
	callchain_param.enabled = !unset;
806

J
Jiri Olsa 已提交
807 808
	/* --no-call-graph */
	if (unset) {
809
		callchain_param.record_mode = CALLCHAIN_NONE;
J
Jiri Olsa 已提交
810 811 812 813
		pr_debug("callchain: disabled\n");
		return 0;
	}

814
	ret = parse_callchain_record_opt(arg, &callchain_param);
815
	if (!ret)
816
		callchain_debug();
817 818 819 820

	return ret;
}

821
int record_callchain_opt(const struct option *opt,
J
Jiri Olsa 已提交
822 823 824
			 const char *arg __maybe_unused,
			 int unset __maybe_unused)
{
825 826 827
	struct record_opts *record = (struct record_opts *)opt->value;

	record->callgraph_set = true;
828
	callchain_param.enabled = true;
J
Jiri Olsa 已提交
829

830 831
	if (callchain_param.record_mode == CALLCHAIN_NONE)
		callchain_param.record_mode = CALLCHAIN_FP;
832

833
	callchain_debug();
J
Jiri Olsa 已提交
834 835 836
	return 0;
}

837 838 839
static int perf_record_config(const char *var, const char *value, void *cb)
{
	if (!strcmp(var, "record.call-graph"))
840
		var = "call-graph.record-mode"; /* fall-through */
841 842 843 844

	return perf_default_config(var, value, cb);
}

845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928
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;
}

929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971
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;
}

972
static const char * const __record_usage[] = {
973 974
	"perf record [<options>] [<command>]",
	"perf record [<options>] -- <command> [<options>]",
975 976
	NULL
};
977
const char * const *record_usage = __record_usage;
978

979
/*
980 981
 * 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
982 983 984 985 986 987 988
 * 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.
 */
989
static struct record record = {
990
	.opts = {
991
		.sample_time	     = true,
992 993 994
		.mmap_pages	     = UINT_MAX,
		.user_freq	     = UINT_MAX,
		.user_interval	     = ULLONG_MAX,
995
		.freq		     = 4000,
N
Namhyung Kim 已提交
996 997
		.target		     = {
			.uses_mmap   = true,
998
			.default_per_cpu = true,
N
Namhyung Kim 已提交
999
		},
1000
		.proc_map_timeout     = 500,
1001
	},
1002 1003 1004
	.tool = {
		.sample		= process_sample_event,
		.fork		= perf_event__process_fork,
1005
		.exit		= perf_event__process_exit,
1006 1007 1008
		.comm		= perf_event__process_comm,
		.mmap		= perf_event__process_mmap,
		.mmap2		= perf_event__process_mmap2,
1009
		.ordered_events	= true,
1010
	},
1011
};
1012

J
Jiri Olsa 已提交
1013
#define CALLCHAIN_HELP "setup and enables call-graph (stack chain/backtrace) recording: "
1014

1015
#ifdef HAVE_DWARF_UNWIND_SUPPORT
1016
const char record_callchain_help[] = CALLCHAIN_HELP "fp dwarf lbr";
1017
#else
1018
const char record_callchain_help[] = CALLCHAIN_HELP "fp lbr";
1019 1020
#endif

1021 1022 1023
/*
 * 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
1024
 * from builtin-record.c, i.e. use record_opts,
1025 1026 1027
 * perf_evlist__prepare_workload, etc instead of fork+exec'in 'perf record',
 * using pipes, etc.
 */
1028
struct option __record_options[] = {
1029
	OPT_CALLBACK('e', "event", &record.evlist, "event",
1030
		     "event selector. use 'perf list' to list available events",
1031
		     parse_events_option),
1032
	OPT_CALLBACK(0, "filter", &record.evlist, "filter",
L
Li Zefan 已提交
1033
		     "event filter", parse_filter),
1034 1035 1036
	OPT_CALLBACK_NOOPT(0, "exclude-perf", &record.evlist,
			   NULL, "don't record events from perf itself",
			   exclude_perf),
1037
	OPT_STRING('p', "pid", &record.opts.target.pid, "pid",
1038
		    "record events on existing process id"),
1039
	OPT_STRING('t', "tid", &record.opts.target.tid, "tid",
1040
		    "record events on existing thread id"),
1041
	OPT_INTEGER('r', "realtime", &record.realtime_prio,
1042
		    "collect data with this RT SCHED_FIFO priority"),
1043
	OPT_BOOLEAN(0, "no-buffering", &record.opts.no_buffering,
1044
		    "collect data without buffering"),
1045
	OPT_BOOLEAN('R', "raw-samples", &record.opts.raw_samples,
1046
		    "collect raw sample records from all opened counters"),
1047
	OPT_BOOLEAN('a', "all-cpus", &record.opts.target.system_wide,
1048
			    "system-wide collection from all CPUs"),
1049
	OPT_STRING('C', "cpu", &record.opts.target.cpu_list, "cpu",
1050
		    "list of cpus to monitor"),
1051
	OPT_U64('c', "count", &record.opts.user_interval, "event period to sample"),
1052
	OPT_STRING('o', "output", &record.file.path, "file",
I
Ingo Molnar 已提交
1053
		    "output file name"),
1054 1055 1056
	OPT_BOOLEAN_SET('i', "no-inherit", &record.opts.no_inherit,
			&record.opts.no_inherit_set,
			"child tasks do not inherit counters"),
1057
	OPT_UINTEGER('F', "freq", &record.opts.user_freq, "profile at this frequency"),
1058 1059 1060
	OPT_CALLBACK('m', "mmap-pages", &record.opts, "pages[,pages]",
		     "number of mmap data pages and AUX area tracing mmap pages",
		     record__parse_mmap_pages),
1061
	OPT_BOOLEAN(0, "group", &record.opts.group,
1062
		    "put the counters into a counter group"),
J
Jiri Olsa 已提交
1063 1064 1065 1066 1067 1068
	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),
1069
	OPT_INCR('v', "verbose", &verbose,
1070
		    "be more verbose (show counter open errors, etc)"),
1071
	OPT_BOOLEAN('q', "quiet", &quiet, "don't print any message"),
1072
	OPT_BOOLEAN('s', "stat", &record.opts.inherit_stat,
1073
		    "per thread counts"),
1074
	OPT_BOOLEAN('d', "data", &record.opts.sample_address, "Record the sample addresses"),
1075 1076 1077
	OPT_BOOLEAN_SET('T', "timestamp", &record.opts.sample_time,
			&record.opts.sample_time_set,
			"Record the sample timestamps"),
1078
	OPT_BOOLEAN('P', "period", &record.opts.period, "Record the sample period"),
1079
	OPT_BOOLEAN('n', "no-samples", &record.opts.no_samples,
1080
		    "don't sample"),
1081
	OPT_BOOLEAN('N', "no-buildid-cache", &record.no_buildid_cache,
1082
		    "do not update the buildid cache"),
1083
	OPT_BOOLEAN('B', "no-buildid", &record.no_buildid,
1084
		    "do not collect buildids in perf.data"),
1085
	OPT_CALLBACK('G', "cgroup", &record.evlist, "name",
S
Stephane Eranian 已提交
1086 1087
		     "monitor event in cgroup name only",
		     parse_cgroups),
1088
	OPT_UINTEGER('D', "delay", &record.opts.initial_delay,
1089
		  "ms to wait before starting measurement after program start"),
1090 1091
	OPT_STRING('u', "uid", &record.opts.target.uid_str, "user",
		   "user to profile"),
1092 1093 1094 1095 1096 1097 1098

	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",
1099
		     parse_branch_stack),
1100 1101
	OPT_BOOLEAN('W', "weight", &record.opts.sample_weight,
		    "sample by weight (on special events only)"),
1102 1103
	OPT_BOOLEAN(0, "transaction", &record.opts.sample_transaction,
		    "sample transaction flags (special events only)"),
1104 1105
	OPT_BOOLEAN(0, "per-thread", &record.opts.target.per_thread,
		    "use per-thread mmaps"),
1106 1107 1108
	OPT_CALLBACK_OPTARG('I', "intr-regs", &record.opts.sample_intr_regs, NULL, "any register",
		    "sample selected machine registers on interrupt,"
		    " use -I ? to list register names", parse_regs),
1109 1110
	OPT_BOOLEAN(0, "running-time", &record.opts.running_time,
		    "Record running/enabled time of read (:S) events"),
1111 1112 1113
	OPT_CALLBACK('k', "clockid", &record.opts,
	"clockid", "clockid to use for events, see clock_gettime()",
	parse_clockid),
1114 1115
	OPT_STRING_OPTARG('S', "snapshot", &record.opts.auxtrace_snapshot_opts,
			  "opts", "AUX area tracing Snapshot Mode", ""),
1116 1117
	OPT_UINTEGER(0, "proc-map-timeout", &record.opts.proc_map_timeout,
			"per thread proc mmap processing timeout in ms"),
1118 1119
	OPT_BOOLEAN(0, "switch-events", &record.opts.record_switch_events,
		    "Record context switch events"),
1120 1121 1122
	OPT_END()
};

1123 1124
struct option *record_options = __record_options;

1125
int cmd_record(int argc, const char **argv, const char *prefix __maybe_unused)
1126
{
1127
	int err;
1128
	struct record *rec = &record;
1129
	char errbuf[BUFSIZ];
1130

1131 1132
	rec->evlist = perf_evlist__new();
	if (rec->evlist == NULL)
1133 1134
		return -ENOMEM;

1135 1136
	perf_config(perf_record_config, rec);

1137
	argc = parse_options(argc, argv, record_options, record_usage,
1138
			    PARSE_OPT_STOP_AT_NON_OPTION);
1139
	if (!argc && target__none(&rec->opts.target))
1140
		usage_with_options(record_usage, record_options);
1141

1142
	if (nr_cgroups && !rec->opts.target.system_wide) {
1143 1144
		ui__error("cgroup monitoring only available in"
			  " system-wide mode\n");
S
Stephane Eranian 已提交
1145 1146
		usage_with_options(record_usage, record_options);
	}
1147 1148 1149 1150 1151
	if (rec->opts.record_switch_events &&
	    !perf_can_record_switch_events()) {
		ui__error("kernel does not support recording context switch events (--switch-events option)\n");
		usage_with_options(record_usage, record_options);
	}
S
Stephane Eranian 已提交
1152

1153 1154 1155 1156 1157 1158
	if (!rec->itr) {
		rec->itr = auxtrace_record__init(rec->evlist, &err);
		if (err)
			return err;
	}

1159 1160 1161 1162 1163
	err = auxtrace_parse_snapshot_options(rec->itr, &rec->opts,
					      rec->opts.auxtrace_snapshot_opts);
	if (err)
		return err;

1164 1165
	err = -ENOMEM;

1166
	symbol__init(NULL);
1167

1168
	if (symbol_conf.kptr_restrict)
1169 1170 1171 1172 1173 1174 1175 1176
		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");
1177

1178
	if (rec->no_buildid_cache || rec->no_buildid)
1179
		disable_buildid_cache();
1180

1181 1182
	if (rec->evlist->nr_entries == 0 &&
	    perf_evlist__add_default(rec->evlist) < 0) {
1183 1184
		pr_err("Not enough memory for event selector list\n");
		goto out_symbol_exit;
1185
	}
1186

1187 1188 1189
	if (rec->opts.target.tid && !rec->opts.no_inherit_set)
		rec->opts.no_inherit = true;

1190
	err = target__validate(&rec->opts.target);
1191
	if (err) {
1192
		target__strerror(&rec->opts.target, err, errbuf, BUFSIZ);
1193 1194 1195
		ui__warning("%s", errbuf);
	}

1196
	err = target__parse_uid(&rec->opts.target);
1197 1198
	if (err) {
		int saved_errno = errno;
1199

1200
		target__strerror(&rec->opts.target, err, errbuf, BUFSIZ);
1201
		ui__error("%s", errbuf);
1202 1203

		err = -saved_errno;
1204
		goto out_symbol_exit;
1205
	}
1206

1207
	err = -ENOMEM;
1208
	if (perf_evlist__create_maps(rec->evlist, &rec->opts.target) < 0)
1209
		usage_with_options(record_usage, record_options);
1210

1211 1212 1213 1214
	err = auxtrace_record__options(rec->itr, rec->evlist, &rec->opts);
	if (err)
		goto out_symbol_exit;

1215
	if (record_opts__config(&rec->opts)) {
1216
		err = -EINVAL;
1217
		goto out_symbol_exit;
1218 1219
	}

1220
	err = __cmd_record(&record, argc, argv);
1221
out_symbol_exit:
1222
	perf_evlist__delete(rec->evlist);
1223
	symbol__exit();
1224
	auxtrace_record__free(rec->itr);
1225
	return err;
1226
}
1227 1228 1229 1230 1231 1232 1233 1234 1235

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;
}