builtin-report.c 30.6 KB
Newer Older
1 2 3 4 5 6 7
/*
 * builtin-report.c
 *
 * Builtin report command: Analyze the perf.data input file,
 * look up and read DSOs and symbol information and display
 * a histogram of results, along various sorting keys.
 */
8
#include "builtin.h"
9

10
#include "util/util.h"
11
#include "util/config.h"
12

13
#include "util/annotate.h"
14
#include "util/color.h"
15
#include <linux/list.h>
16
#include <linux/rbtree.h>
17
#include "util/symbol.h"
18
#include "util/callchain.h"
19
#include "util/values.h"
20

21
#include "perf.h"
22
#include "util/debug.h"
23 24
#include "util/evlist.h"
#include "util/evsel.h"
25
#include "util/header.h"
26
#include "util/session.h"
27
#include "util/tool.h"
28

29
#include <subcmd/parse-options.h>
30
#include <subcmd/exec-cmd.h>
31 32
#include "util/parse-events.h"

33
#include "util/thread.h"
34
#include "util/sort.h"
35
#include "util/hist.h"
36
#include "util/data.h"
37
#include "arch/common.h"
38
#include "util/time-utils.h"
39
#include "util/auxtrace.h"
40
#include "util/units.h"
41

42
#include <dlfcn.h>
43
#include <errno.h>
44
#include <inttypes.h>
45
#include <regex.h>
46
#include <signal.h>
47
#include <linux/bitmap.h>
48
#include <linux/stringify.h>
49 50 51
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
52

53
struct report {
54
	struct perf_tool	tool;
55
	struct perf_session	*session;
56
	bool			use_tui, use_gtk, use_stdio;
57 58 59
	bool			show_full_info;
	bool			show_threads;
	bool			inverted_callchain;
60
	bool			mem_mode;
61 62
	bool			header;
	bool			header_only;
63
	bool			nonany_branch_mode;
64
	int			max_stack;
65 66 67
	struct perf_read_values	show_threads_values;
	const char		*pretty_printing_style;
	const char		*cpu_list;
68
	const char		*symbol_filter_str;
69 70
	const char		*time_str;
	struct perf_time_interval ptime;
71
	float			min_percent;
72
	u64			nr_entries;
73
	u64			queue_size;
74
	int			socket_filter;
75
	DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS);
76
};
77

78
static int report__config(const char *var, const char *value, void *cb)
79
{
80 81
	struct report *rep = cb;

82 83 84 85
	if (!strcmp(var, "report.group")) {
		symbol_conf.event_group = perf_config_bool(var, value);
		return 0;
	}
86
	if (!strcmp(var, "report.percent-limit")) {
87 88 89 90
		double pcnt = strtof(value, NULL);

		rep->min_percent = pcnt;
		callchain_param.min_percent = pcnt;
91 92
		return 0;
	}
93 94 95 96
	if (!strcmp(var, "report.children")) {
		symbol_conf.cumulate_callchain = perf_config_bool(var, value);
		return 0;
	}
97 98 99
	if (!strcmp(var, "report.queue-size"))
		return perf_config_u64(&rep->queue_size, var, value);

100 101 102 103
	if (!strcmp(var, "report.sort_order")) {
		default_sort_order = strdup(value);
		return 0;
	}
104

105
	return 0;
106 107
}

108 109 110 111 112 113 114 115 116 117 118 119 120 121
static int hist_iter__report_callback(struct hist_entry_iter *iter,
				      struct addr_location *al, bool single,
				      void *arg)
{
	int err = 0;
	struct report *rep = arg;
	struct hist_entry *he = iter->he;
	struct perf_evsel *evsel = iter->evsel;
	struct mem_info *mi;
	struct branch_info *bi;

	if (!ui__has_annotation())
		return 0;

122 123 124
	hist__account_cycles(iter->sample->branch_stack, al, iter->sample,
			     rep->nonany_branch_mode);

125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150
	if (sort__mode == SORT_MODE__BRANCH) {
		bi = he->branch_info;
		err = addr_map_symbol__inc_samples(&bi->from, evsel->idx);
		if (err)
			goto out;

		err = addr_map_symbol__inc_samples(&bi->to, evsel->idx);

	} else if (rep->mem_mode) {
		mi = he->mem_info;
		err = addr_map_symbol__inc_samples(&mi->daddr, evsel->idx);
		if (err)
			goto out;

		err = hist_entry__inc_addr_samples(he, evsel->idx, al->addr);

	} else if (symbol_conf.cumulate_callchain) {
		if (single)
			err = hist_entry__inc_addr_samples(he, evsel->idx,
							   al->addr);
	} else {
		err = hist_entry__inc_addr_samples(he, evsel->idx, al->addr);
	}

out:
	return err;
151 152
}

153
static int process_sample_event(struct perf_tool *tool,
154
				union perf_event *event,
155
				struct perf_sample *sample,
156
				struct perf_evsel *evsel,
157
				struct machine *machine)
158
{
159
	struct report *rep = container_of(tool, struct report, tool);
160
	struct addr_location al;
161
	struct hist_entry_iter iter = {
162 163
		.evsel 			= evsel,
		.sample 		= sample,
164
		.hide_unresolved 	= symbol_conf.hide_unresolved,
165
		.add_entry_cb 		= hist_iter__report_callback,
166
	};
167
	int ret = 0;
168

169 170 171
	if (perf_time__skip_sample(&rep->ptime, sample->time))
		return 0;

172
	if (machine__resolve(machine, &al, sample) < 0) {
173 174
		pr_debug("problem processing %d event, skipping it.\n",
			 event->header.type);
175 176
		return -1;
	}
177

178
	if (symbol_conf.hide_unresolved && al.sym == NULL)
179
		goto out_put;
180

181
	if (rep->cpu_list && !test_bit(sample->cpu, rep->cpu_bitmap))
182
		goto out_put;
183

184 185 186 187 188 189 190
	if (sort__mode == SORT_MODE__BRANCH) {
		/*
		 * A non-synthesized event might not have a branch stack if
		 * branch stacks have been synthesized (using itrace options).
		 */
		if (!sample->branch_stack)
			goto out_put;
191
		iter.ops = &hist_iter_branch;
192
	} else if (rep->mem_mode) {
193
		iter.ops = &hist_iter_mem;
194
	} else if (symbol_conf.cumulate_callchain) {
195
		iter.ops = &hist_iter_cumulative;
196
	} else {
197
		iter.ops = &hist_iter_normal;
198
	}
199 200 201 202

	if (al.map != NULL)
		al.map->dso->hit = 1;

203
	ret = hist_entry_iter__add(&iter, &al, rep->max_stack, rep);
204 205
	if (ret < 0)
		pr_debug("problem adding hist entry, skipping event\n");
206 207
out_put:
	addr_location__put(&al);
208
	return ret;
209
}
I
Ingo Molnar 已提交
210

211
static int process_read_event(struct perf_tool *tool,
212
			      union perf_event *event,
213
			      struct perf_sample *sample __maybe_unused,
214
			      struct perf_evsel *evsel,
215
			      struct machine *machine __maybe_unused)
216
{
217
	struct report *rep = container_of(tool, struct report, tool);
218

219
	if (rep->show_threads) {
220
		const char *name = evsel ? perf_evsel__name(evsel) : "unknown";
221
		int err = perf_read_values_add_value(&rep->show_threads_values,
222 223 224 225
					   event->read.pid, event->read.tid,
					   event->read.id,
					   name,
					   event->read.value);
226 227 228

		if (err)
			return err;
229 230
	}

231
	dump_printf(": %d %d %s %" PRIu64 "\n", event->read.pid, event->read.tid,
232
		    evsel ? perf_evsel__name(evsel) : "FAIL",
233
		    event->read.value);
234 235 236 237

	return 0;
}

238
/* For pipe mode, sample_type is not currently set */
239
static int report__setup_sample_type(struct report *rep)
240
{
241 242 243
	struct perf_session *session = rep->session;
	u64 sample_type = perf_evlist__combined_sample_type(session->evlist);
	bool is_pipe = perf_data_file__is_pipe(session->file);
244

245 246 247 248 249 250
	if (session->itrace_synth_opts->callchain ||
	    (!is_pipe &&
	     perf_header__has_feat(&session->header, HEADER_AUXTRACE) &&
	     !session->itrace_synth_opts->set))
		sample_type |= PERF_SAMPLE_CALLCHAIN;

251 252 253
	if (session->itrace_synth_opts->last_branch)
		sample_type |= PERF_SAMPLE_BRANCH_STACK;

254
	if (!is_pipe && !(sample_type & PERF_SAMPLE_CALLCHAIN)) {
255
		if (perf_hpp_list.parent) {
256
			ui__error("Selected --sort parent, but no "
257 258
				    "callchain data. Did you call "
				    "'perf record' without -g?\n");
259
			return -EINVAL;
260
		}
261
		if (symbol_conf.use_callchain) {
262 263 264
			ui__error("Selected -g or --branch-history but no "
				  "callchain data. Did\n"
				  "you call 'perf record' without -g?\n");
265
			return -1;
266
		}
267
	} else if (!callchain_param.enabled &&
268
		   callchain_param.mode != CHAIN_NONE &&
269
		   !symbol_conf.use_callchain) {
270
			symbol_conf.use_callchain = true;
271
			if (callchain_register_param(&callchain_param) < 0) {
272
				ui__error("Can't register callchain params.\n");
273
				return -EINVAL;
274
			}
275 276
	}

277 278 279 280 281 282 283 284
	if (symbol_conf.cumulate_callchain) {
		/* Silently ignore if callchain is missing */
		if (!(sample_type & PERF_SAMPLE_CALLCHAIN)) {
			symbol_conf.cumulate_callchain = false;
			perf_hpp__cancel_cumulate();
		}
	}

285
	if (sort__mode == SORT_MODE__BRANCH) {
286
		if (!is_pipe &&
287
		    !(sample_type & PERF_SAMPLE_BRANCH_STACK)) {
288 289
			ui__error("Selected -b but no branch data. "
				  "Did you call perf record without -b?\n");
290 291 292 293
			return -1;
		}
	}

294 295 296 297
	if (symbol_conf.use_callchain || symbol_conf.cumulate_callchain) {
		if ((sample_type & PERF_SAMPLE_REGS_USER) &&
		    (sample_type & PERF_SAMPLE_STACK_USER))
			callchain_param.record_mode = CALLCHAIN_DWARF;
298 299
		else if (sample_type & PERF_SAMPLE_BRANCH_STACK)
			callchain_param.record_mode = CALLCHAIN_LBR;
300 301 302
		else
			callchain_param.record_mode = CALLCHAIN_FP;
	}
303 304 305 306 307 308

	/* ??? handle more cases than just ANY? */
	if (!(perf_evlist__combined_branch_type(session->evlist) &
				PERF_SAMPLE_BRANCH_ANY))
		rep->nonany_branch_mode = true;

309 310
	return 0;
}
311

312
static void sig_handler(int sig __maybe_unused)
313 314 315 316
{
	session_done = 1;
}

317
static size_t hists__fprintf_nr_sample_events(struct hists *hists, struct report *rep,
318 319 320 321
					      const char *evname, FILE *fp)
{
	size_t ret;
	char unit;
322 323 324
	unsigned long nr_samples = hists->stats.nr_events[PERF_RECORD_SAMPLE];
	u64 nr_events = hists->stats.total_period;
	struct perf_evsel *evsel = hists_to_evsel(hists);
325 326
	char buf[512];
	size_t size = sizeof(buf);
327
	int socked_id = hists->socket_filter;
328

329 330 331
	if (quiet)
		return 0;

332 333 334 335 336
	if (symbol_conf.filter_relative) {
		nr_samples = hists->stats.nr_non_filtered_samples;
		nr_events = hists->stats.total_non_filtered_period;
	}

337
	if (perf_evsel__is_group_event(evsel)) {
338 339 340 341 342 343
		struct perf_evsel *pos;

		perf_evsel__group_desc(evsel, buf, size);
		evname = buf;

		for_each_group_member(pos, evsel) {
344 345
			const struct hists *pos_hists = evsel__hists(pos);

346
			if (symbol_conf.filter_relative) {
347 348
				nr_samples += pos_hists->stats.nr_non_filtered_samples;
				nr_events += pos_hists->stats.total_non_filtered_period;
349
			} else {
350 351
				nr_samples += pos_hists->stats.nr_events[PERF_RECORD_SAMPLE];
				nr_events += pos_hists->stats.total_period;
352
			}
353 354
		}
	}
355

356 357
	nr_samples = convert_unit(nr_samples, &unit);
	ret = fprintf(fp, "# Samples: %lu%c", nr_samples, unit);
358
	if (evname != NULL)
359 360
		ret += fprintf(fp, " of event '%s'", evname);

361 362 363 364 365
	if (symbol_conf.show_ref_callgraph &&
	    strstr(evname, "call-graph=no")) {
		ret += fprintf(fp, ", show reference callgraph");
	}

366 367
	if (rep->mem_mode) {
		ret += fprintf(fp, "\n# Total weight : %" PRIu64, nr_events);
368
		ret += fprintf(fp, "\n# Sort order   : %s", sort_order ? : default_mem_sort_order);
369 370
	} else
		ret += fprintf(fp, "\n# Event count (approx.): %" PRIu64, nr_events);
371

372 373
	if (socked_id > -1)
		ret += fprintf(fp, "\n# Processor Socket: %d", socked_id);
374

375 376 377
	return ret + fprintf(fp, "\n#\n");
}

378
static int perf_evlist__tty_browse_hists(struct perf_evlist *evlist,
379
					 struct report *rep,
380
					 const char *help)
381
{
382
	struct perf_evsel *pos;
383

384 385 386 387 388
	if (!quiet) {
		fprintf(stdout, "#\n# Total Lost Samples: %" PRIu64 "\n#\n",
			evlist->stats.total_lost_samples);
	}

389
	evlist__for_each_entry(evlist, pos) {
390
		struct hists *hists = evsel__hists(pos);
391
		const char *evname = perf_evsel__name(pos);
392

393 394 395 396
		if (symbol_conf.event_group &&
		    !perf_evsel__is_group_leader(pos))
			continue;

397
		hists__fprintf_nr_sample_events(hists, rep, evname, stdout);
398
		hists__fprintf(hists, !quiet, 0, 0, rep->min_percent, stdout,
399
			       symbol_conf.use_callchain);
400 401 402
		fprintf(stdout, "\n\n");
	}

403
	if (!quiet)
404 405
		fprintf(stdout, "#\n# (%s)\n#\n", help);

406 407 408 409 410
	if (rep->show_threads) {
		bool style = !strcmp(rep->pretty_printing_style, "raw");
		perf_read_values_display(stdout, &rep->show_threads_values,
					 style);
		perf_read_values_destroy(&rep->show_threads_values);
411 412 413 414 415
	}

	return 0;
}

416 417
static void report__warn_kptr_restrict(const struct report *rep)
{
418
	struct map *kernel_map = machine__kernel_map(&rep->session->machines.host);
419
	struct kmap *kernel_kmap = kernel_map ? map__kmap(kernel_map) : NULL;
420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444

	if (kernel_map == NULL ||
	    (kernel_map->dso->hit &&
	     (kernel_kmap->ref_reloc_sym == NULL ||
	      kernel_kmap->ref_reloc_sym->addr == 0))) {
		const char *desc =
		    "As no suitable kallsyms nor vmlinux was found, kernel samples\n"
		    "can't be resolved.";

		if (kernel_map) {
			const struct dso *kdso = kernel_map->dso;
			if (!RB_EMPTY_ROOT(&kdso->symbols[MAP__FUNCTION])) {
				desc = "If some relocation was applied (e.g. "
				       "kexec) symbols may be misresolved.";
			}
		}

		ui__warning(
"Kernel address maps (/proc/{kallsyms,modules}) were restricted.\n\n"
"Check /proc/sys/kernel/kptr_restrict before running 'perf record'.\n\n%s\n\n"
"Samples in kernel modules can't be resolved as well.\n\n",
		desc);
	}
}

445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464
static int report__gtk_browse_hists(struct report *rep, const char *help)
{
	int (*hist_browser)(struct perf_evlist *evlist, const char *help,
			    struct hist_browser_timer *timer, float min_pcnt);

	hist_browser = dlsym(perf_gtk_handle, "perf_evlist__gtk_browse_hists");

	if (hist_browser == NULL) {
		ui__error("GTK browser not found!\n");
		return -1;
	}

	return hist_browser(rep->session->evlist, help, NULL, rep->min_percent);
}

static int report__browse_hists(struct report *rep)
{
	int ret;
	struct perf_session *session = rep->session;
	struct perf_evlist *evlist = session->evlist;
465 466 467 468 469 470 471 472
	const char *help = perf_tip(system_path(TIPDIR));

	if (help == NULL) {
		/* fallback for people who don't install perf ;-) */
		help = perf_tip(DOCDIR);
		if (help == NULL)
			help = "Cannot load tips.txt file, please install perf!";
	}
473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496

	switch (use_browser) {
	case 1:
		ret = perf_evlist__tui_browse_hists(evlist, help, NULL,
						    rep->min_percent,
						    &session->header.env);
		/*
		 * Usually "ret" is the last pressed key, and we only
		 * care if the key notifies us to switch data file.
		 */
		if (ret != K_SWITCH_INPUT_DATA)
			ret = 0;
		break;
	case 2:
		ret = report__gtk_browse_hists(rep, help);
		break;
	default:
		ret = perf_evlist__tty_browse_hists(evlist, rep, help);
		break;
	}

	return ret;
}

497
static int report__collapse_hists(struct report *rep)
498 499 500
{
	struct ui_progress prog;
	struct perf_evsel *pos;
501
	int ret = 0;
502

503
	ui_progress__init(&prog, rep->nr_entries, "Merging related events...");
504

505
	evlist__for_each_entry(rep->session->evlist, pos) {
506
		struct hists *hists = evsel__hists(pos);
507 508 509 510

		if (pos->idx == 0)
			hists->symbol_filter_str = rep->symbol_filter_str;

511 512
		hists->socket_filter = rep->socket_filter;

513 514 515
		ret = hists__collapse_resort(hists, &prog);
		if (ret < 0)
			break;
516 517 518 519

		/* Non-group events are considered as leader */
		if (symbol_conf.event_group &&
		    !perf_evsel__is_group_leader(pos)) {
520
			struct hists *leader_hists = evsel__hists(pos->leader);
521 522 523 524 525 526 527

			hists__match(leader_hists, hists);
			hists__link(leader_hists, hists);
		}
	}

	ui_progress__finish();
528
	return ret;
529 530
}

531 532 533 534 535 536 537
static void report__output_resort(struct report *rep)
{
	struct ui_progress prog;
	struct perf_evsel *pos;

	ui_progress__init(&prog, rep->nr_entries, "Sorting events for output...");

538
	evlist__for_each_entry(rep->session->evlist, pos)
539
		perf_evsel__output_resort(pos, &prog);
540 541 542 543

	ui_progress__finish();
}

544
static int __cmd_report(struct report *rep)
545
{
546
	int ret;
547
	struct perf_session *session = rep->session;
548
	struct perf_evsel *pos;
549
	struct perf_data_file *file = session->file;
550

551 552
	signal(SIGINT, sig_handler);

553 554 555
	if (rep->cpu_list) {
		ret = perf_session__cpu_bitmap(session, rep->cpu_list,
					       rep->cpu_bitmap);
556 557
		if (ret) {
			ui__error("failed to set cpu bitmap\n");
558
			return ret;
559
		}
560
		session->itrace_synth_opts->cpu_bitmap = rep->cpu_bitmap;
561 562
	}

563 564 565 566 567
	if (rep->show_threads) {
		ret = perf_read_values_init(&rep->show_threads_values);
		if (ret)
			return ret;
	}
568

569
	ret = report__setup_sample_type(rep);
570 571
	if (ret) {
		/* report__setup_sample_type() already showed error message */
572
		return ret;
573
	}
574

575
	ret = perf_session__process_events(session);
576 577
	if (ret) {
		ui__error("failed to process sample\n");
578
		return ret;
579
	}
580

581
	report__warn_kptr_restrict(rep);
582

583
	evlist__for_each_entry(session->evlist, pos)
584 585
		rep->nr_entries += evsel__hists(pos)->nr_entries;

586 587 588
	if (use_browser == 0) {
		if (verbose > 3)
			perf_session__fprintf(session, stdout);
589

590 591
		if (verbose > 2)
			perf_session__fprintf_dsos(session, stdout);
592

593 594
		if (dump_trace) {
			perf_session__fprintf_nr_events(session, stdout);
595
			perf_evlist__fprintf_nr_events(session->evlist, stdout);
596 597
			return 0;
		}
598 599
	}

600 601 602 603 604
	ret = report__collapse_hists(rep);
	if (ret) {
		ui__error("failed to process hist entry\n");
		return ret;
	}
605

606 607 608
	if (session_done())
		return 0;

609 610 611 612 613
	/*
	 * recalculate number of entries after collapsing since it
	 * might be changed during the collapse phase.
	 */
	rep->nr_entries = 0;
614
	evlist__for_each_entry(session->evlist, pos)
615 616
		rep->nr_entries += evsel__hists(pos)->nr_entries;

617
	if (rep->nr_entries == 0) {
618
		ui__error("The %s file has no samples!\n", file->path);
619
		return 0;
620 621
	}

622
	report__output_resort(rep);
623

624
	return report__browse_hists(rep);
625 626
}

627
static int
628
report_parse_callchain_opt(const struct option *opt, const char *arg, int unset)
629
{
630
	struct callchain_param *callchain = opt->value;
631

632
	callchain->enabled = !unset;
633 634 635 636
	/*
	 * --no-call-graph
	 */
	if (unset) {
637 638
		symbol_conf.use_callchain = false;
		callchain->mode = CHAIN_NONE;
639 640 641
		return 0;
	}

642
	return parse_callchain_report_opt(arg);
643 644
}

645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662
int
report_parse_ignore_callees_opt(const struct option *opt __maybe_unused,
				const char *arg, int unset __maybe_unused)
{
	if (arg) {
		int err = regcomp(&ignore_callees_regex, arg, REG_EXTENDED);
		if (err) {
			char buf[BUFSIZ];
			regerror(err, &ignore_callees_regex, buf, sizeof(buf));
			pr_err("Invalid --ignore-callees regex: %s\n%s", arg, buf);
			return -1;
		}
		have_ignore_callees = 1;
	}

	return 0;
}

663
static int
664
parse_branch_mode(const struct option *opt,
665
		  const char *str __maybe_unused, int unset)
666
{
667 668 669
	int *branch_mode = opt->value;

	*branch_mode = !unset;
670 671 672
	return 0;
}

673 674 675 676
static int
parse_percent_limit(const struct option *opt, const char *str,
		    int unset __maybe_unused)
{
677
	struct report *rep = opt->value;
678
	double pcnt = strtof(str, NULL);
679

680 681
	rep->min_percent = pcnt;
	callchain_param.min_percent = pcnt;
682 683 684
	return 0;
}

685
#define CALLCHAIN_DEFAULT_OPT  "graph,0.5,caller,function,percent"
686 687 688 689

const char report_callchain_help[] = "Display call graph (stack chain/backtrace):\n\n"
				     CALLCHAIN_REPORT_HELP
				     "\n\t\t\t\tDefault: " CALLCHAIN_DEFAULT_OPT;
690

691
int cmd_report(int argc, const char **argv)
692
{
693
	struct perf_session *session;
694
	struct itrace_synth_opts itrace_synth_opts = { .set = 0, };
695
	struct stat st;
696
	bool has_br_stack = false;
697
	int branch_mode = -1;
698
	bool branch_call_mode = false;
699
	char callchain_default_opt[] = CALLCHAIN_DEFAULT_OPT;
700
	const char * const report_usage[] = {
N
Namhyung Kim 已提交
701
		"perf report [<options>]",
702 703
		NULL
	};
704
	struct report report = {
705
		.tool = {
706 707
			.sample		 = process_sample_event,
			.mmap		 = perf_event__process_mmap,
708
			.mmap2		 = perf_event__process_mmap2,
709
			.comm		 = perf_event__process_comm,
710
			.namespaces	 = perf_event__process_namespaces,
711 712
			.exit		 = perf_event__process_exit,
			.fork		 = perf_event__process_fork,
713 714 715 716 717
			.lost		 = perf_event__process_lost,
			.read		 = process_read_event,
			.attr		 = perf_event__process_attr,
			.tracing_data	 = perf_event__process_tracing_data,
			.build_id	 = perf_event__process_build_id,
718 719 720
			.id_index	 = perf_event__process_id_index,
			.auxtrace_info	 = perf_event__process_auxtrace_info,
			.auxtrace	 = perf_event__process_auxtrace,
721
			.feature	 = perf_event__process_feature,
722
			.ordered_events	 = true,
723 724
			.ordering_requires_timestamps = true,
		},
725
		.max_stack		 = PERF_MAX_STACK_DEPTH,
726
		.pretty_printing_style	 = "normal",
727
		.socket_filter		 = -1,
728 729
	};
	const struct option options[] = {
730
	OPT_STRING('i', "input", &input_name, "file",
731
		    "input file name"),
732
	OPT_INCR('v', "verbose", &verbose,
733
		    "be more verbose (show symbol address, etc)"),
734
	OPT_BOOLEAN('q', "quiet", &quiet, "Do not show any message"),
735 736
	OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
		    "dump raw trace in ASCII"),
737 738
	OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name,
		   "file", "vmlinux pathname"),
739 740
	OPT_STRING(0, "kallsyms", &symbol_conf.kallsyms_name,
		   "file", "kallsyms pathname"),
741
	OPT_BOOLEAN('f', "force", &symbol_conf.force, "don't complain, do it"),
742
	OPT_BOOLEAN('m', "modules", &symbol_conf.use_modules,
743
		    "load module symbols - WARNING: use only with -k and LIVE kernel"),
744
	OPT_BOOLEAN('n', "show-nr-samples", &symbol_conf.show_nr_samples,
745
		    "Show a column with the number of samples"),
746
	OPT_BOOLEAN('T', "threads", &report.show_threads,
747
		    "Show per-thread event counters"),
748
	OPT_STRING(0, "pretty", &report.pretty_printing_style, "key",
749
		   "pretty printing style key: normal raw"),
750
	OPT_BOOLEAN(0, "tui", &report.use_tui, "Use the TUI interface"),
751
	OPT_BOOLEAN(0, "gtk", &report.use_gtk, "Use the GTK2 interface"),
752 753
	OPT_BOOLEAN(0, "stdio", &report.use_stdio,
		    "Use the stdio interface"),
754 755 756
	OPT_BOOLEAN(0, "header", &report.header, "Show data header."),
	OPT_BOOLEAN(0, "header-only", &report.header_only,
		    "Show only data header."),
757
	OPT_STRING('s', "sort", &sort_order, "key[,key2...]",
758 759
		   "sort by key(s): pid, comm, dso, symbol, parent, cpu, srcline, ..."
		   " Please refer the man page for the complete list."),
760 761
	OPT_STRING('F', "fields", &field_order, "key[,keys...]",
		   "output field(s): overhead, period, sample plus all of sort keys"),
762
	OPT_BOOLEAN(0, "show-cpu-utilization", &symbol_conf.show_cpu_utilization,
763
		    "Show sample percentage for different cpu modes"),
764 765
	OPT_BOOLEAN_FLAG(0, "showcpuutilization", &symbol_conf.show_cpu_utilization,
		    "Show sample percentage for different cpu modes", PARSE_OPT_HIDDEN),
766 767
	OPT_STRING('p', "parent", &parent_pattern, "regex",
		   "regex filter to identify parent, see: '--sort parent'"),
768
	OPT_BOOLEAN('x', "exclude-other", &symbol_conf.exclude_other,
769
		    "Only display entries with parent-match"),
770
	OPT_CALLBACK_DEFAULT('g', "call-graph", &callchain_param,
771
			     "print_type,threshold[,print_limit],order,sort_key[,branch],value",
772 773
			     report_callchain_help, &report_parse_callchain_opt,
			     callchain_default_opt),
774 775
	OPT_BOOLEAN(0, "children", &symbol_conf.cumulate_callchain,
		    "Accumulate callchains of children and show total overhead as well"),
776 777 778
	OPT_INTEGER(0, "max-stack", &report.max_stack,
		    "Set the maximum stack depth when parsing the callchain, "
		    "anything beyond the specified depth will be ignored. "
779
		    "Default: kernel.perf_event_max_stack or " __stringify(PERF_MAX_STACK_DEPTH)),
780 781
	OPT_BOOLEAN('G', "inverted", &report.inverted_callchain,
		    "alias for inverted call graph"),
782 783 784
	OPT_CALLBACK(0, "ignore-callees", NULL, "regex",
		   "ignore callees of these functions in call graphs",
		   report_parse_ignore_callees_opt),
785
	OPT_STRING('d', "dsos", &symbol_conf.dso_list_str, "dso[,dso...]",
786
		   "only consider symbols in these dsos"),
787
	OPT_STRING('c', "comms", &symbol_conf.comm_list_str, "comm[,comm...]",
788
		   "only consider symbols in these comms"),
789 790 791 792
	OPT_STRING(0, "pid", &symbol_conf.pid_list_str, "pid[,pid...]",
		   "only consider symbols in these pids"),
	OPT_STRING(0, "tid", &symbol_conf.tid_list_str, "tid[,tid...]",
		   "only consider symbols in these tids"),
793
	OPT_STRING('S', "symbols", &symbol_conf.sym_list_str, "symbol[,symbol...]",
794
		   "only consider these symbols"),
795 796
	OPT_STRING(0, "symbol-filter", &report.symbol_filter_str, "filter",
		   "only show symbols that (partially) match with this filter"),
797
	OPT_STRING('w', "column-widths", &symbol_conf.col_width_list_str,
798 799
		   "width[,width...]",
		   "don't try to adjust column width, use these fixed values"),
800
	OPT_STRING_NOEMPTY('t', "field-separator", &symbol_conf.field_sep, "separator",
801 802
		   "separator for columns, no spaces will be added between "
		   "columns '.' is reserved."),
803
	OPT_BOOLEAN('U', "hide-unresolved", &symbol_conf.hide_unresolved,
804
		    "Only display entries resolved to a symbol"),
805 806 807
	OPT_CALLBACK(0, "symfs", NULL, "directory",
		     "Look for files with symbols relative to this directory",
		     symbol__config_symfs),
808
	OPT_STRING('C', "cpu", &report.cpu_list, "cpu",
809 810
		   "list of cpus to profile"),
	OPT_BOOLEAN('I', "show-info", &report.show_full_info,
811
		    "Display extended information about perf.data file"),
812 813 814 815
	OPT_BOOLEAN(0, "source", &symbol_conf.annotate_src,
		    "Interleave source code with assembly code (default)"),
	OPT_BOOLEAN(0, "asm-raw", &symbol_conf.annotate_asm_raw,
		    "Display raw encoding of assembly instructions (default)"),
816 817
	OPT_STRING('M', "disassembler-style", &disassembler_style, "disassembler style",
		   "Specify disassembler style (e.g. -M intel for intel syntax)"),
818 819
	OPT_BOOLEAN(0, "show-total-period", &symbol_conf.show_total_period,
		    "Show a column with the sum of periods"),
N
Namhyung Kim 已提交
820 821
	OPT_BOOLEAN(0, "group", &symbol_conf.event_group,
		    "Show event group information together"),
822
	OPT_CALLBACK_NOOPT('b', "branch-stack", &branch_mode, "",
823 824 825 826
		    "use branch records for per branch histogram filling",
		    parse_branch_mode),
	OPT_BOOLEAN(0, "branch-history", &branch_call_mode,
		    "add last branch records to call history"),
827 828
	OPT_STRING(0, "objdump", &objdump_path, "path",
		   "objdump binary to use for disassembly and annotations"),
829 830
	OPT_BOOLEAN(0, "demangle", &symbol_conf.demangle,
		    "Disable symbol demangling"),
831 832
	OPT_BOOLEAN(0, "demangle-kernel", &symbol_conf.demangle_kernel,
		    "Enable kernel symbol demangling"),
833
	OPT_BOOLEAN(0, "mem-mode", &report.mem_mode, "mem access profile"),
834 835
	OPT_CALLBACK(0, "percent-limit", &report, "percent",
		     "Don't show entries under that percent", parse_percent_limit),
836
	OPT_CALLBACK(0, "percentage", NULL, "relative|absolute",
N
Namhyung Kim 已提交
837
		     "how to display percentage of filtered entries", parse_filter_percentage),
838 839 840
	OPT_CALLBACK_OPTARG(0, "itrace", &itrace_synth_opts, NULL, "opts",
			    "Instruction Tracing options",
			    itrace_parse_synth_opts),
841 842
	OPT_BOOLEAN(0, "full-source-path", &srcline_full_filename,
			"Show full source file name path for source lines"),
843 844
	OPT_BOOLEAN(0, "show-ref-call-graph", &symbol_conf.show_ref_callgraph,
		    "Show callgraph from reference event"),
845 846
	OPT_INTEGER(0, "socket-filter", &report.socket_filter,
		    "only show processor socket that match with this filter"),
847 848
	OPT_BOOLEAN(0, "raw-trace", &symbol_conf.raw_trace,
		    "Show raw trace event output (do not use print fmt or plugins)"),
849 850
	OPT_BOOLEAN(0, "hierarchy", &symbol_conf.report_hierarchy,
		    "Show entries in a hierarchy"),
851 852 853
	OPT_CALLBACK_DEFAULT(0, "stdio-color", NULL, "mode",
			     "'always' (default), 'never' or 'auto' only applicable to --stdio mode",
			     stdio__config_color, "always"),
854 855
	OPT_STRING(0, "time", &report.time_str, "str",
		   "Time span of interest (start,stop)"),
J
Jin Yao 已提交
856 857
	OPT_BOOLEAN(0, "inline", &symbol_conf.inline_name,
		    "Show inline function"),
858
	OPT_END()
859
	};
860 861 862
	struct perf_data_file file = {
		.mode  = PERF_DATA_MODE_READ,
	};
863 864 865 866
	int ret = hists__init();

	if (ret < 0)
		return ret;
867

868 869 870
	ret = perf_config(report__config, &report);
	if (ret)
		return ret;
871

872
	argc = parse_options(argc, argv, options, report_usage, 0);
873 874 875 876 877 878 879 880 881 882
	if (argc) {
		/*
		 * Special case: if there's an argument left then assume that
		 * it's a symbol filter:
		 */
		if (argc > 1)
			usage_with_options(report_usage, options);

		report.symbol_filter_str = argv[0];
	}
883

884 885 886
	if (quiet)
		perf_quiet_option();

887 888 889 890 891 892 893 894 895 896 897
	if (symbol_conf.vmlinux_name &&
	    access(symbol_conf.vmlinux_name, R_OK)) {
		pr_err("Invalid file: %s\n", symbol_conf.vmlinux_name);
		return -EINVAL;
	}
	if (symbol_conf.kallsyms_name &&
	    access(symbol_conf.kallsyms_name, R_OK)) {
		pr_err("Invalid file: %s\n", symbol_conf.kallsyms_name);
		return -EINVAL;
	}

898
	if (report.use_stdio)
899
		use_browser = 0;
900
	else if (report.use_tui)
901
		use_browser = 1;
902 903
	else if (report.use_gtk)
		use_browser = 2;
904

905
	if (report.inverted_callchain)
906
		callchain_param.order = ORDER_CALLER;
907 908
	if (symbol_conf.cumulate_callchain && !callchain_param.order_set)
		callchain_param.order = ORDER_CALLER;
909

910 911 912 913
	if (itrace_synth_opts.callchain &&
	    (int)itrace_synth_opts.callchain_sz > report.max_stack)
		report.max_stack = itrace_synth_opts.callchain_sz;

914
	if (!input_name || !strlen(input_name)) {
915
		if (!fstat(STDIN_FILENO, &st) && S_ISFIFO(st.st_mode))
916
			input_name = "-";
917
		else
918
			input_name = "perf.data";
919
	}
920

921
	file.path  = input_name;
922
	file.force = symbol_conf.force;
923

924
repeat:
925
	session = perf_session__new(&file, false, &report.tool);
926
	if (session == NULL)
927
		return -1;
928

929 930 931 932 933
	if (report.queue_size) {
		ordered_events__set_alloc_size(&session->ordered_events,
					       report.queue_size);
	}

934 935
	session->itrace_synth_opts = &itrace_synth_opts;

936 937 938 939
	report.session = session;

	has_br_stack = perf_header__has_feat(&session->header,
					     HEADER_BRANCH_STACK);
940

941 942 943
	if (itrace_synth_opts.last_branch)
		has_br_stack = true;

944 945 946
	if (has_br_stack && branch_call_mode)
		symbol_conf.show_branchflag_count = true;

947 948 949 950 951 952
	/*
	 * Branch mode is a tristate:
	 * -1 means default, so decide based on the file having branch data.
	 * 0/1 means the user chose a mode.
	 */
	if (((branch_mode == -1 && has_br_stack) || branch_mode == 1) &&
953
	    !branch_call_mode) {
954
		sort__mode = SORT_MODE__BRANCH;
955 956
		symbol_conf.cumulate_callchain = false;
	}
957
	if (branch_call_mode) {
958
		callchain_param.key = CCKEY_ADDRESS;
959 960 961 962 963 964
		callchain_param.branch_callstack = 1;
		symbol_conf.use_callchain = true;
		callchain_register_param(&callchain_param);
		if (sort_order == NULL)
			sort_order = "srcline,symbol,dso";
	}
965

966
	if (report.mem_mode) {
967
		if (sort__mode == SORT_MODE__BRANCH) {
968
			pr_err("branch and mem mode incompatible\n");
969 970
			goto error;
		}
971
		sort__mode = SORT_MODE__MEMORY;
972
		symbol_conf.cumulate_callchain = false;
973
	}
974

975 976 977 978 979 980 981 982 983 984 985
	if (symbol_conf.report_hierarchy) {
		/* disable incompatible options */
		symbol_conf.cumulate_callchain = false;

		if (field_order) {
			pr_err("Error: --hierarchy and --fields options cannot be used together\n");
			parse_options_usage(report_usage, options, "F", 1);
			parse_options_usage(NULL, options, "hierarchy", 0);
			goto error;
		}

986
		perf_hpp_list.need_collapse = true;
987 988
	}

989 990
	/* Force tty output for header output and per-thread stat. */
	if (report.header || report.header_only || report.show_threads)
991
		use_browser = 0;
992 993 994 995
	if (report.header || report.header_only)
		report.tool.show_feat_hdr = SHOW_FEAT_HEADER;
	if (report.show_full_info)
		report.tool.show_feat_hdr = SHOW_FEAT_HEADER_FULL_INFO;
996

997 998
	if (strcmp(input_name, "-") != 0)
		setup_browser(true);
999
	else
1000 1001
		use_browser = 0;

1002 1003 1004 1005 1006 1007 1008 1009 1010
	if (setup_sorting(session->evlist) < 0) {
		if (sort_order)
			parse_options_usage(report_usage, options, "s", 1);
		if (field_order)
			parse_options_usage(sort_order ? NULL : report_usage,
					    options, "F", 1);
		goto error;
	}

1011
	if ((report.header || report.header_only) && !quiet) {
1012 1013
		perf_session__fprintf_info(session, stdout,
					   report.show_full_info);
1014 1015 1016 1017
		if (report.header_only) {
			ret = 0;
			goto error;
		}
1018
	} else if (use_browser == 0 && !quiet) {
1019 1020 1021 1022
		fputs("# To display the perf.data header info, please use --header/--header-only options.\n#\n",
		      stdout);
	}

1023
	/*
1024
	 * Only in the TUI browser we are doing integrated annotation,
1025 1026 1027
	 * so don't allocate extra space that won't be used in the stdio
	 * implementation.
	 */
1028
	if (ui__has_annotation()) {
1029 1030 1031
		ret = symbol__annotation_init();
		if (ret < 0)
			goto error;
1032 1033 1034 1035 1036
		/*
 		 * For searching by name on the "Browse map details".
 		 * providing it only in verbose mode not to bloat too
 		 * much struct symbol.
 		 */
1037
		if (verbose > 0) {
1038 1039 1040 1041 1042 1043 1044 1045 1046 1047
			/*
			 * XXX: Need to provide a less kludgy way to ask for
			 * more space per symbol, the u32 is for the index on
			 * the ui browser.
			 * See symbol__browser_index.
			 */
			symbol_conf.priv_size += sizeof(u32);
			symbol_conf.sort_by_name = true;
		}
	}
1048

1049
	if (symbol__init(&session->header.env) < 0)
1050
		goto error;
1051

1052 1053 1054 1055 1056
	if (perf_time__parse_str(&report.ptime, report.time_str) != 0) {
		pr_err("Invalid time string\n");
		return -EINVAL;
	}

1057
	sort__setup_elide(stdout);
1058

1059
	ret = __cmd_report(&report);
1060 1061 1062 1063 1064 1065
	if (ret == K_SWITCH_INPUT_DATA) {
		perf_session__delete(session);
		goto repeat;
	} else
		ret = 0;

1066 1067 1068
error:
	perf_session__delete(session);
	return ret;
1069
}