builtin-report.c 32.8 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
2 3 4 5 6 7 8
/*
 * 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.
 */
9
#include "builtin.h"
10

11
#include "util/util.h"
12
#include "util/config.h"
13

14
#include "util/annotate.h"
15
#include "util/color.h"
16
#include <linux/list.h>
17
#include <linux/rbtree.h>
18
#include "util/symbol.h"
19
#include "util/callchain.h"
20
#include "util/values.h"
21

22
#include "perf.h"
23
#include "util/debug.h"
24 25
#include "util/evlist.h"
#include "util/evsel.h"
26
#include "util/header.h"
27
#include "util/session.h"
28
#include "util/tool.h"
29

30
#include <subcmd/parse-options.h>
31
#include <subcmd/exec-cmd.h>
32 33
#include "util/parse-events.h"

34
#include "util/thread.h"
35
#include "util/sort.h"
36
#include "util/hist.h"
37
#include "util/data.h"
38
#include "arch/common.h"
39
#include "util/time-utils.h"
40
#include "util/auxtrace.h"
41
#include "util/units.h"
42
#include "util/branch.h"
43

44
#include <dlfcn.h>
45
#include <errno.h>
46
#include <inttypes.h>
47
#include <regex.h>
48
#include <signal.h>
49
#include <linux/bitmap.h>
50
#include <linux/stringify.h>
51 52 53
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
54

55 56
#define PTIME_RANGE_MAX	10

57
struct report {
58
	struct perf_tool	tool;
59
	struct perf_session	*session;
60
	bool			use_tui, use_gtk, use_stdio;
61 62 63
	bool			show_full_info;
	bool			show_threads;
	bool			inverted_callchain;
64
	bool			mem_mode;
65
	bool			stats_mode;
66 67
	bool			header;
	bool			header_only;
68
	bool			nonany_branch_mode;
69
	int			max_stack;
70 71 72
	struct perf_read_values	show_threads_values;
	const char		*pretty_printing_style;
	const char		*cpu_list;
73
	const char		*symbol_filter_str;
74
	const char		*time_str;
75 76
	struct perf_time_interval ptime_range[PTIME_RANGE_MAX];
	int			range_num;
77
	float			min_percent;
78
	u64			nr_entries;
79
	u64			queue_size;
80
	int			socket_filter;
81
	DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS);
82
	struct branch_type_stat	brtype_stat;
83
};
84

85
static int report__config(const char *var, const char *value, void *cb)
86
{
87 88
	struct report *rep = cb;

89 90 91 92
	if (!strcmp(var, "report.group")) {
		symbol_conf.event_group = perf_config_bool(var, value);
		return 0;
	}
93
	if (!strcmp(var, "report.percent-limit")) {
94 95 96 97
		double pcnt = strtof(value, NULL);

		rep->min_percent = pcnt;
		callchain_param.min_percent = pcnt;
98 99
		return 0;
	}
100 101 102 103
	if (!strcmp(var, "report.children")) {
		symbol_conf.cumulate_callchain = perf_config_bool(var, value);
		return 0;
	}
104 105 106
	if (!strcmp(var, "report.queue-size"))
		return perf_config_u64(&rep->queue_size, var, value);

107 108 109 110
	if (!strcmp(var, "report.sort_order")) {
		default_sort_order = strdup(value);
		return 0;
	}
111

112
	return 0;
113 114
}

115 116 117 118 119 120 121 122
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;
123
	struct perf_sample *sample = iter->sample;
124 125 126 127 128 129
	struct mem_info *mi;
	struct branch_info *bi;

	if (!ui__has_annotation())
		return 0;

130
	hist__account_cycles(sample->branch_stack, al, sample,
131 132
			     rep->nonany_branch_mode);

133 134
	if (sort__mode == SORT_MODE__BRANCH) {
		bi = he->branch_info;
135
		err = addr_map_symbol__inc_samples(&bi->from, sample, evsel->idx);
136 137 138
		if (err)
			goto out;

139
		err = addr_map_symbol__inc_samples(&bi->to, sample, evsel->idx);
140 141 142

	} else if (rep->mem_mode) {
		mi = he->mem_info;
143
		err = addr_map_symbol__inc_samples(&mi->daddr, sample, evsel->idx);
144 145 146
		if (err)
			goto out;

147
		err = hist_entry__inc_addr_samples(he, sample, evsel->idx, al->addr);
148 149 150

	} else if (symbol_conf.cumulate_callchain) {
		if (single)
151
			err = hist_entry__inc_addr_samples(he, sample, evsel->idx,
152 153
							   al->addr);
	} else {
154
		err = hist_entry__inc_addr_samples(he, sample, evsel->idx, al->addr);
155 156 157 158
	}

out:
	return err;
159 160
}

161 162 163 164 165 166 167 168
static int hist_iter__branch_callback(struct hist_entry_iter *iter,
				      struct addr_location *al __maybe_unused,
				      bool single __maybe_unused,
				      void *arg)
{
	struct hist_entry *he = iter->he;
	struct report *rep = arg;
	struct branch_info *bi;
169 170 171 172 173 174 175 176 177
	struct perf_sample *sample = iter->sample;
	struct perf_evsel *evsel = iter->evsel;
	int err;

	if (!ui__has_annotation())
		return 0;

	hist__account_cycles(sample->branch_stack, al, sample,
			     rep->nonany_branch_mode);
178 179

	bi = he->branch_info;
180 181 182 183 184 185
	err = addr_map_symbol__inc_samples(&bi->from, sample, evsel->idx);
	if (err)
		goto out;

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

186 187 188
	branch_type_count(&rep->brtype_stat, &bi->flags,
			  bi->from.addr, bi->to.addr);

189 190
out:
	return err;
191 192
}

193
static int process_sample_event(struct perf_tool *tool,
194
				union perf_event *event,
195
				struct perf_sample *sample,
196
				struct perf_evsel *evsel,
197
				struct machine *machine)
198
{
199
	struct report *rep = container_of(tool, struct report, tool);
200
	struct addr_location al;
201
	struct hist_entry_iter iter = {
202 203
		.evsel 			= evsel,
		.sample 		= sample,
204
		.hide_unresolved 	= symbol_conf.hide_unresolved,
205
		.add_entry_cb 		= hist_iter__report_callback,
206
	};
207
	int ret = 0;
208

209 210
	if (perf_time__ranges_skip_sample(rep->ptime_range, rep->range_num,
					  sample->time)) {
211
		return 0;
212
	}
213

214
	if (machine__resolve(machine, &al, sample) < 0) {
215 216
		pr_debug("problem processing %d event, skipping it.\n",
			 event->header.type);
217 218
		return -1;
	}
219

220
	if (symbol_conf.hide_unresolved && al.sym == NULL)
221
		goto out_put;
222

223
	if (rep->cpu_list && !test_bit(sample->cpu, rep->cpu_bitmap))
224
		goto out_put;
225

226 227 228 229 230 231 232
	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;
233 234

		iter.add_entry_cb = hist_iter__branch_callback;
235
		iter.ops = &hist_iter_branch;
236
	} else if (rep->mem_mode) {
237
		iter.ops = &hist_iter_mem;
238
	} else if (symbol_conf.cumulate_callchain) {
239
		iter.ops = &hist_iter_cumulative;
240
	} else {
241
		iter.ops = &hist_iter_normal;
242
	}
243 244 245 246

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

247
	ret = hist_entry_iter__add(&iter, &al, rep->max_stack, rep);
248 249
	if (ret < 0)
		pr_debug("problem adding hist entry, skipping event\n");
250 251
out_put:
	addr_location__put(&al);
252
	return ret;
253
}
I
Ingo Molnar 已提交
254

255
static int process_read_event(struct perf_tool *tool,
256
			      union perf_event *event,
257
			      struct perf_sample *sample __maybe_unused,
258
			      struct perf_evsel *evsel,
259
			      struct machine *machine __maybe_unused)
260
{
261
	struct report *rep = container_of(tool, struct report, tool);
262

263
	if (rep->show_threads) {
264
		const char *name = evsel ? perf_evsel__name(evsel) : "unknown";
265
		int err = perf_read_values_add_value(&rep->show_threads_values,
266
					   event->read.pid, event->read.tid,
267
					   evsel->idx,
268 269
					   name,
					   event->read.value);
270 271 272

		if (err)
			return err;
273 274
	}

275 276 277
	return 0;
}

278
/* For pipe mode, sample_type is not currently set */
279
static int report__setup_sample_type(struct report *rep)
280
{
281 282
	struct perf_session *session = rep->session;
	u64 sample_type = perf_evlist__combined_sample_type(session->evlist);
283
	bool is_pipe = perf_data__is_pipe(session->data);
284

285 286 287 288 289 290
	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;

291 292 293
	if (session->itrace_synth_opts->last_branch)
		sample_type |= PERF_SAMPLE_BRANCH_STACK;

294
	if (!is_pipe && !(sample_type & PERF_SAMPLE_CALLCHAIN)) {
295
		if (perf_hpp_list.parent) {
296
			ui__error("Selected --sort parent, but no "
297 298
				    "callchain data. Did you call "
				    "'perf record' without -g?\n");
299
			return -EINVAL;
300
		}
301 302 303 304 305
		if (symbol_conf.use_callchain &&
			!symbol_conf.show_branchflag_count) {
			ui__error("Selected -g or --branch-history.\n"
				  "But no callchain or branch data.\n"
				  "Did you call 'perf record' without -g or -b?\n");
306
			return -1;
307
		}
308
	} else if (!callchain_param.enabled &&
309
		   callchain_param.mode != CHAIN_NONE &&
310
		   !symbol_conf.use_callchain) {
311
			symbol_conf.use_callchain = true;
312
			if (callchain_register_param(&callchain_param) < 0) {
313
				ui__error("Can't register callchain params.\n");
314
				return -EINVAL;
315
			}
316 317
	}

318 319 320 321 322 323 324 325
	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();
		}
	}

326
	if (sort__mode == SORT_MODE__BRANCH) {
327
		if (!is_pipe &&
328
		    !(sample_type & PERF_SAMPLE_BRANCH_STACK)) {
329 330
			ui__error("Selected -b but no branch data. "
				  "Did you call perf record without -b?\n");
331 332 333 334
			return -1;
		}
	}

335 336 337 338
	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;
339 340
		else if (sample_type & PERF_SAMPLE_BRANCH_STACK)
			callchain_param.record_mode = CALLCHAIN_LBR;
341 342 343
		else
			callchain_param.record_mode = CALLCHAIN_FP;
	}
344 345 346 347 348 349

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

350 351
	return 0;
}
352

353
static void sig_handler(int sig __maybe_unused)
354 355 356 357
{
	session_done = 1;
}

358
static size_t hists__fprintf_nr_sample_events(struct hists *hists, struct report *rep,
359 360 361 362
					      const char *evname, FILE *fp)
{
	size_t ret;
	char unit;
363 364 365
	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);
366 367
	char buf[512];
	size_t size = sizeof(buf);
368
	int socked_id = hists->socket_filter;
369

370 371 372
	if (quiet)
		return 0;

373 374 375 376 377
	if (symbol_conf.filter_relative) {
		nr_samples = hists->stats.nr_non_filtered_samples;
		nr_events = hists->stats.total_non_filtered_period;
	}

378
	if (perf_evsel__is_group_event(evsel)) {
379 380 381 382 383 384
		struct perf_evsel *pos;

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

		for_each_group_member(pos, evsel) {
385 386
			const struct hists *pos_hists = evsel__hists(pos);

387
			if (symbol_conf.filter_relative) {
388 389
				nr_samples += pos_hists->stats.nr_non_filtered_samples;
				nr_events += pos_hists->stats.total_non_filtered_period;
390
			} else {
391 392
				nr_samples += pos_hists->stats.nr_events[PERF_RECORD_SAMPLE];
				nr_events += pos_hists->stats.total_period;
393
			}
394 395
		}
	}
396

397 398
	nr_samples = convert_unit(nr_samples, &unit);
	ret = fprintf(fp, "# Samples: %lu%c", nr_samples, unit);
399
	if (evname != NULL)
400 401
		ret += fprintf(fp, " of event '%s'", evname);

402 403 404 405 406
	if (symbol_conf.show_ref_callgraph &&
	    strstr(evname, "call-graph=no")) {
		ret += fprintf(fp, ", show reference callgraph");
	}

407 408
	if (rep->mem_mode) {
		ret += fprintf(fp, "\n# Total weight : %" PRIu64, nr_events);
409
		ret += fprintf(fp, "\n# Sort order   : %s", sort_order ? : default_mem_sort_order);
410 411
	} else
		ret += fprintf(fp, "\n# Event count (approx.): %" PRIu64, nr_events);
412

413 414
	if (socked_id > -1)
		ret += fprintf(fp, "\n# Processor Socket: %d", socked_id);
415

416 417 418
	return ret + fprintf(fp, "\n#\n");
}

419
static int perf_evlist__tty_browse_hists(struct perf_evlist *evlist,
420
					 struct report *rep,
421
					 const char *help)
422
{
423
	struct perf_evsel *pos;
424

425 426 427 428 429
	if (!quiet) {
		fprintf(stdout, "#\n# Total Lost Samples: %" PRIu64 "\n#\n",
			evlist->stats.total_lost_samples);
	}

430
	evlist__for_each_entry(evlist, pos) {
431
		struct hists *hists = evsel__hists(pos);
432
		const char *evname = perf_evsel__name(pos);
433

434 435 436 437
		if (symbol_conf.event_group &&
		    !perf_evsel__is_group_leader(pos))
			continue;

438
		hists__fprintf_nr_sample_events(hists, rep, evname, stdout);
439
		hists__fprintf(hists, !quiet, 0, 0, rep->min_percent, stdout,
440 441
			       symbol_conf.use_callchain ||
			       symbol_conf.show_branchflag_count);
442 443 444
		fprintf(stdout, "\n\n");
	}

445
	if (!quiet)
446 447
		fprintf(stdout, "#\n# (%s)\n#\n", help);

448 449 450 451 452
	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);
453 454
	}

455 456 457
	if (sort__mode == SORT_MODE__BRANCH)
		branch_type_stat_display(stdout, &rep->brtype_stat);

458 459 460
	return 0;
}

461 462
static void report__warn_kptr_restrict(const struct report *rep)
{
463
	struct map *kernel_map = machine__kernel_map(&rep->session->machines.host);
464
	struct kmap *kernel_kmap = kernel_map ? map__kmap(kernel_map) : NULL;
465

466 467 468
	if (perf_evlist__exclude_kernel(rep->session->evlist))
		return;

469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492
	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);
	}
}

493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512
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;
513 514 515 516 517 518 519 520
	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!";
	}
521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544

	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;
}

545
static int report__collapse_hists(struct report *rep)
546 547 548
{
	struct ui_progress prog;
	struct perf_evsel *pos;
549
	int ret = 0;
550

551
	ui_progress__init(&prog, rep->nr_entries, "Merging related events...");
552

553
	evlist__for_each_entry(rep->session->evlist, pos) {
554
		struct hists *hists = evsel__hists(pos);
555 556 557 558

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

559 560
		hists->socket_filter = rep->socket_filter;

561 562 563
		ret = hists__collapse_resort(hists, &prog);
		if (ret < 0)
			break;
564 565 566 567

		/* Non-group events are considered as leader */
		if (symbol_conf.event_group &&
		    !perf_evsel__is_group_leader(pos)) {
568
			struct hists *leader_hists = evsel__hists(pos->leader);
569 570 571 572 573 574 575

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

	ui_progress__finish();
576
	return ret;
577 578
}

579 580 581 582 583 584 585
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...");

586
	evlist__for_each_entry(rep->session->evlist, pos)
587
		perf_evsel__output_resort(pos, &prog);
588 589 590 591

	ui_progress__finish();
}

592 593 594 595 596 597 598 599 600 601 602 603 604 605
static void stats_setup(struct report *rep)
{
	memset(&rep->tool, 0, sizeof(rep->tool));
	rep->tool.no_warn = true;
}

static int stats_print(struct report *rep)
{
	struct perf_session *session = rep->session;

	perf_session__fprintf_nr_events(session, stdout);
	return 0;
}

606
static int __cmd_report(struct report *rep)
607
{
608
	int ret;
609
	struct perf_session *session = rep->session;
610
	struct perf_evsel *pos;
611
	struct perf_data *data = session->data;
612

613 614
	signal(SIGINT, sig_handler);

615 616 617
	if (rep->cpu_list) {
		ret = perf_session__cpu_bitmap(session, rep->cpu_list,
					       rep->cpu_bitmap);
618 619
		if (ret) {
			ui__error("failed to set cpu bitmap\n");
620
			return ret;
621
		}
622
		session->itrace_synth_opts->cpu_bitmap = rep->cpu_bitmap;
623 624
	}

625 626 627 628 629
	if (rep->show_threads) {
		ret = perf_read_values_init(&rep->show_threads_values);
		if (ret)
			return ret;
	}
630

631
	ret = report__setup_sample_type(rep);
632 633
	if (ret) {
		/* report__setup_sample_type() already showed error message */
634
		return ret;
635
	}
636

637 638 639
	if (rep->stats_mode)
		stats_setup(rep);

640
	ret = perf_session__process_events(session);
641 642
	if (ret) {
		ui__error("failed to process sample\n");
643
		return ret;
644
	}
645

646 647 648
	if (rep->stats_mode)
		return stats_print(rep);

649
	report__warn_kptr_restrict(rep);
650

651
	evlist__for_each_entry(session->evlist, pos)
652 653
		rep->nr_entries += evsel__hists(pos)->nr_entries;

654 655 656
	if (use_browser == 0) {
		if (verbose > 3)
			perf_session__fprintf(session, stdout);
657

658 659
		if (verbose > 2)
			perf_session__fprintf_dsos(session, stdout);
660

661 662
		if (dump_trace) {
			perf_session__fprintf_nr_events(session, stdout);
663
			perf_evlist__fprintf_nr_events(session->evlist, stdout);
664 665
			return 0;
		}
666 667
	}

668 669 670 671 672
	ret = report__collapse_hists(rep);
	if (ret) {
		ui__error("failed to process hist entry\n");
		return ret;
	}
673

674 675 676
	if (session_done())
		return 0;

677 678 679 680 681
	/*
	 * recalculate number of entries after collapsing since it
	 * might be changed during the collapse phase.
	 */
	rep->nr_entries = 0;
682
	evlist__for_each_entry(session->evlist, pos)
683 684
		rep->nr_entries += evsel__hists(pos)->nr_entries;

685
	if (rep->nr_entries == 0) {
J
Jiri Olsa 已提交
686
		ui__error("The %s file has no samples!\n", data->file.path);
687
		return 0;
688 689
	}

690
	report__output_resort(rep);
691

692
	return report__browse_hists(rep);
693 694
}

695
static int
696
report_parse_callchain_opt(const struct option *opt, const char *arg, int unset)
697
{
698
	struct callchain_param *callchain = opt->value;
699

700
	callchain->enabled = !unset;
701 702 703 704
	/*
	 * --no-call-graph
	 */
	if (unset) {
705 706
		symbol_conf.use_callchain = false;
		callchain->mode = CHAIN_NONE;
707 708 709
		return 0;
	}

710
	return parse_callchain_report_opt(arg);
711 712
}

713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730
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;
}

731
static int
732
parse_branch_mode(const struct option *opt,
733
		  const char *str __maybe_unused, int unset)
734
{
735 736 737
	int *branch_mode = opt->value;

	*branch_mode = !unset;
738 739 740
	return 0;
}

741 742 743 744
static int
parse_percent_limit(const struct option *opt, const char *str,
		    int unset __maybe_unused)
{
745
	struct report *rep = opt->value;
746
	double pcnt = strtof(str, NULL);
747

748 749
	rep->min_percent = pcnt;
	callchain_param.min_percent = pcnt;
750 751 752
	return 0;
}

753
#define CALLCHAIN_DEFAULT_OPT  "graph,0.5,caller,function,percent"
754 755 756 757

const char report_callchain_help[] = "Display call graph (stack chain/backtrace):\n\n"
				     CALLCHAIN_REPORT_HELP
				     "\n\t\t\t\tDefault: " CALLCHAIN_DEFAULT_OPT;
758

759
int cmd_report(int argc, const char **argv)
760
{
761
	struct perf_session *session;
762
	struct itrace_synth_opts itrace_synth_opts = { .set = 0, };
763
	struct stat st;
764
	bool has_br_stack = false;
765
	int branch_mode = -1;
766
	bool branch_call_mode = false;
767
	char callchain_default_opt[] = CALLCHAIN_DEFAULT_OPT;
768
	const char * const report_usage[] = {
N
Namhyung Kim 已提交
769
		"perf report [<options>]",
770 771
		NULL
	};
772
	struct report report = {
773
		.tool = {
774 775
			.sample		 = process_sample_event,
			.mmap		 = perf_event__process_mmap,
776
			.mmap2		 = perf_event__process_mmap2,
777
			.comm		 = perf_event__process_comm,
778
			.namespaces	 = perf_event__process_namespaces,
779 780
			.exit		 = perf_event__process_exit,
			.fork		 = perf_event__process_fork,
781 782 783 784 785
			.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,
786 787 788
			.id_index	 = perf_event__process_id_index,
			.auxtrace_info	 = perf_event__process_auxtrace_info,
			.auxtrace	 = perf_event__process_auxtrace,
789
			.feature	 = perf_event__process_feature,
790
			.ordered_events	 = true,
791 792
			.ordering_requires_timestamps = true,
		},
793
		.max_stack		 = PERF_MAX_STACK_DEPTH,
794
		.pretty_printing_style	 = "normal",
795
		.socket_filter		 = -1,
796 797
	};
	const struct option options[] = {
798
	OPT_STRING('i', "input", &input_name, "file",
799
		    "input file name"),
800
	OPT_INCR('v', "verbose", &verbose,
801
		    "be more verbose (show symbol address, etc)"),
802
	OPT_BOOLEAN('q', "quiet", &quiet, "Do not show any message"),
803 804
	OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
		    "dump raw trace in ASCII"),
805
	OPT_BOOLEAN(0, "stats", &report.stats_mode, "Display event stats"),
806 807
	OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name,
		   "file", "vmlinux pathname"),
808 809
	OPT_STRING(0, "kallsyms", &symbol_conf.kallsyms_name,
		   "file", "kallsyms pathname"),
810
	OPT_BOOLEAN('f', "force", &symbol_conf.force, "don't complain, do it"),
811
	OPT_BOOLEAN('m', "modules", &symbol_conf.use_modules,
812
		    "load module symbols - WARNING: use only with -k and LIVE kernel"),
813
	OPT_BOOLEAN('n', "show-nr-samples", &symbol_conf.show_nr_samples,
814
		    "Show a column with the number of samples"),
815
	OPT_BOOLEAN('T', "threads", &report.show_threads,
816
		    "Show per-thread event counters"),
817
	OPT_STRING(0, "pretty", &report.pretty_printing_style, "key",
818
		   "pretty printing style key: normal raw"),
819
	OPT_BOOLEAN(0, "tui", &report.use_tui, "Use the TUI interface"),
820
	OPT_BOOLEAN(0, "gtk", &report.use_gtk, "Use the GTK2 interface"),
821 822
	OPT_BOOLEAN(0, "stdio", &report.use_stdio,
		    "Use the stdio interface"),
823 824 825
	OPT_BOOLEAN(0, "header", &report.header, "Show data header."),
	OPT_BOOLEAN(0, "header-only", &report.header_only,
		    "Show only data header."),
826
	OPT_STRING('s', "sort", &sort_order, "key[,key2...]",
827 828
		   "sort by key(s): pid, comm, dso, symbol, parent, cpu, srcline, ..."
		   " Please refer the man page for the complete list."),
829 830
	OPT_STRING('F', "fields", &field_order, "key[,keys...]",
		   "output field(s): overhead, period, sample plus all of sort keys"),
831
	OPT_BOOLEAN(0, "show-cpu-utilization", &symbol_conf.show_cpu_utilization,
832
		    "Show sample percentage for different cpu modes"),
833 834
	OPT_BOOLEAN_FLAG(0, "showcpuutilization", &symbol_conf.show_cpu_utilization,
		    "Show sample percentage for different cpu modes", PARSE_OPT_HIDDEN),
835 836
	OPT_STRING('p', "parent", &parent_pattern, "regex",
		   "regex filter to identify parent, see: '--sort parent'"),
837
	OPT_BOOLEAN('x', "exclude-other", &symbol_conf.exclude_other,
838
		    "Only display entries with parent-match"),
839
	OPT_CALLBACK_DEFAULT('g', "call-graph", &callchain_param,
840
			     "print_type,threshold[,print_limit],order,sort_key[,branch],value",
841 842
			     report_callchain_help, &report_parse_callchain_opt,
			     callchain_default_opt),
843 844
	OPT_BOOLEAN(0, "children", &symbol_conf.cumulate_callchain,
		    "Accumulate callchains of children and show total overhead as well"),
845 846 847
	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. "
848
		    "Default: kernel.perf_event_max_stack or " __stringify(PERF_MAX_STACK_DEPTH)),
849 850
	OPT_BOOLEAN('G', "inverted", &report.inverted_callchain,
		    "alias for inverted call graph"),
851 852 853
	OPT_CALLBACK(0, "ignore-callees", NULL, "regex",
		   "ignore callees of these functions in call graphs",
		   report_parse_ignore_callees_opt),
854
	OPT_STRING('d', "dsos", &symbol_conf.dso_list_str, "dso[,dso...]",
855
		   "only consider symbols in these dsos"),
856
	OPT_STRING('c', "comms", &symbol_conf.comm_list_str, "comm[,comm...]",
857
		   "only consider symbols in these comms"),
858 859 860 861
	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"),
862
	OPT_STRING('S', "symbols", &symbol_conf.sym_list_str, "symbol[,symbol...]",
863
		   "only consider these symbols"),
864 865
	OPT_STRING(0, "symbol-filter", &report.symbol_filter_str, "filter",
		   "only show symbols that (partially) match with this filter"),
866
	OPT_STRING('w', "column-widths", &symbol_conf.col_width_list_str,
867 868
		   "width[,width...]",
		   "don't try to adjust column width, use these fixed values"),
869
	OPT_STRING_NOEMPTY('t', "field-separator", &symbol_conf.field_sep, "separator",
870 871
		   "separator for columns, no spaces will be added between "
		   "columns '.' is reserved."),
872
	OPT_BOOLEAN('U', "hide-unresolved", &symbol_conf.hide_unresolved,
873
		    "Only display entries resolved to a symbol"),
874 875 876
	OPT_CALLBACK(0, "symfs", NULL, "directory",
		     "Look for files with symbols relative to this directory",
		     symbol__config_symfs),
877
	OPT_STRING('C', "cpu", &report.cpu_list, "cpu",
878 879
		   "list of cpus to profile"),
	OPT_BOOLEAN('I', "show-info", &report.show_full_info,
880
		    "Display extended information about perf.data file"),
881 882 883 884
	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)"),
885 886
	OPT_STRING('M', "disassembler-style", &disassembler_style, "disassembler style",
		   "Specify disassembler style (e.g. -M intel for intel syntax)"),
887 888
	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
889 890
	OPT_BOOLEAN(0, "group", &symbol_conf.event_group,
		    "Show event group information together"),
891
	OPT_CALLBACK_NOOPT('b', "branch-stack", &branch_mode, "",
892 893 894 895
		    "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"),
896 897
	OPT_STRING(0, "objdump", &objdump_path, "path",
		   "objdump binary to use for disassembly and annotations"),
898 899
	OPT_BOOLEAN(0, "demangle", &symbol_conf.demangle,
		    "Disable symbol demangling"),
900 901
	OPT_BOOLEAN(0, "demangle-kernel", &symbol_conf.demangle_kernel,
		    "Enable kernel symbol demangling"),
902
	OPT_BOOLEAN(0, "mem-mode", &report.mem_mode, "mem access profile"),
903 904
	OPT_CALLBACK(0, "percent-limit", &report, "percent",
		     "Don't show entries under that percent", parse_percent_limit),
905
	OPT_CALLBACK(0, "percentage", NULL, "relative|absolute",
N
Namhyung Kim 已提交
906
		     "how to display percentage of filtered entries", parse_filter_percentage),
907 908 909
	OPT_CALLBACK_OPTARG(0, "itrace", &itrace_synth_opts, NULL, "opts",
			    "Instruction Tracing options",
			    itrace_parse_synth_opts),
910 911
	OPT_BOOLEAN(0, "full-source-path", &srcline_full_filename,
			"Show full source file name path for source lines"),
912 913
	OPT_BOOLEAN(0, "show-ref-call-graph", &symbol_conf.show_ref_callgraph,
		    "Show callgraph from reference event"),
914 915
	OPT_INTEGER(0, "socket-filter", &report.socket_filter,
		    "only show processor socket that match with this filter"),
916 917
	OPT_BOOLEAN(0, "raw-trace", &symbol_conf.raw_trace,
		    "Show raw trace event output (do not use print fmt or plugins)"),
918 919
	OPT_BOOLEAN(0, "hierarchy", &symbol_conf.report_hierarchy,
		    "Show entries in a hierarchy"),
920 921 922
	OPT_CALLBACK_DEFAULT(0, "stdio-color", NULL, "mode",
			     "'always' (default), 'never' or 'auto' only applicable to --stdio mode",
			     stdio__config_color, "always"),
923 924
	OPT_STRING(0, "time", &report.time_str, "str",
		   "Time span of interest (start,stop)"),
J
Jin Yao 已提交
925 926
	OPT_BOOLEAN(0, "inline", &symbol_conf.inline_name,
		    "Show inline function"),
927
	OPT_END()
928
	};
929
	struct perf_data data = {
930 931
		.mode  = PERF_DATA_MODE_READ,
	};
932 933 934 935
	int ret = hists__init();

	if (ret < 0)
		return ret;
936

937 938 939
	ret = perf_config(report__config, &report);
	if (ret)
		return ret;
940

941
	argc = parse_options(argc, argv, options, report_usage, 0);
942 943 944 945 946 947 948 949 950 951
	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];
	}
952

953 954 955
	if (quiet)
		perf_quiet_option();

956 957 958 959 960 961 962 963 964 965 966
	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;
	}

967
	if (report.inverted_callchain)
968
		callchain_param.order = ORDER_CALLER;
969 970
	if (symbol_conf.cumulate_callchain && !callchain_param.order_set)
		callchain_param.order = ORDER_CALLER;
971

972 973 974 975
	if (itrace_synth_opts.callchain &&
	    (int)itrace_synth_opts.callchain_sz > report.max_stack)
		report.max_stack = itrace_synth_opts.callchain_sz;

976
	if (!input_name || !strlen(input_name)) {
977
		if (!fstat(STDIN_FILENO, &st) && S_ISFIFO(st.st_mode))
978
			input_name = "-";
979
		else
980
			input_name = "perf.data";
981
	}
982

J
Jiri Olsa 已提交
983 984
	data.file.path = input_name;
	data.force     = symbol_conf.force;
985

986
repeat:
987
	session = perf_session__new(&data, false, &report.tool);
988
	if (session == NULL)
989
		return -1;
990

991 992 993 994 995
	if (report.queue_size) {
		ordered_events__set_alloc_size(&session->ordered_events,
					       report.queue_size);
	}

996 997
	session->itrace_synth_opts = &itrace_synth_opts;

998 999 1000 1001
	report.session = session;

	has_br_stack = perf_header__has_feat(&session->header,
					     HEADER_BRANCH_STACK);
1002

1003 1004 1005
	if (itrace_synth_opts.last_branch)
		has_br_stack = true;

1006 1007 1008
	if (has_br_stack && branch_call_mode)
		symbol_conf.show_branchflag_count = true;

1009 1010
	memset(&report.brtype_stat, 0, sizeof(struct branch_type_stat));

1011 1012 1013 1014 1015 1016
	/*
	 * 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) &&
1017
	    !branch_call_mode) {
1018
		sort__mode = SORT_MODE__BRANCH;
1019 1020
		symbol_conf.cumulate_callchain = false;
	}
1021
	if (branch_call_mode) {
1022
		callchain_param.key = CCKEY_ADDRESS;
1023 1024 1025 1026 1027 1028
		callchain_param.branch_callstack = 1;
		symbol_conf.use_callchain = true;
		callchain_register_param(&callchain_param);
		if (sort_order == NULL)
			sort_order = "srcline,symbol,dso";
	}
1029

1030
	if (report.mem_mode) {
1031
		if (sort__mode == SORT_MODE__BRANCH) {
1032
			pr_err("branch and mem mode incompatible\n");
1033 1034
			goto error;
		}
1035
		sort__mode = SORT_MODE__MEMORY;
1036
		symbol_conf.cumulate_callchain = false;
1037
	}
1038

1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049
	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;
		}

1050
		perf_hpp_list.need_collapse = true;
1051 1052
	}

1053 1054 1055 1056 1057 1058 1059
	if (report.use_stdio)
		use_browser = 0;
	else if (report.use_tui)
		use_browser = 1;
	else if (report.use_gtk)
		use_browser = 2;

1060 1061
	/* Force tty output for header output and per-thread stat. */
	if (report.header || report.header_only || report.show_threads)
1062
		use_browser = 0;
1063 1064 1065 1066
	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;
1067 1068
	if (report.stats_mode)
		use_browser = 0;
1069

1070 1071
	if (strcmp(input_name, "-") != 0)
		setup_browser(true);
1072
	else
1073 1074
		use_browser = 0;

1075 1076 1077 1078 1079 1080 1081 1082 1083
	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;
	}

1084
	if ((report.header || report.header_only) && !quiet) {
1085 1086
		perf_session__fprintf_info(session, stdout,
					   report.show_full_info);
1087 1088 1089 1090
		if (report.header_only) {
			ret = 0;
			goto error;
		}
1091
	} else if (use_browser == 0 && !quiet && !report.stats_mode) {
1092 1093 1094 1095
		fputs("# To display the perf.data header info, please use --header/--header-only options.\n#\n",
		      stdout);
	}

1096
	/*
1097
	 * Only in the TUI browser we are doing integrated annotation,
1098 1099 1100
	 * so don't allocate extra space that won't be used in the stdio
	 * implementation.
	 */
1101
	if (ui__has_annotation()) {
1102 1103 1104
		ret = symbol__annotation_init();
		if (ret < 0)
			goto error;
1105 1106 1107 1108 1109
		/*
 		 * For searching by name on the "Browse map details".
 		 * providing it only in verbose mode not to bloat too
 		 * much struct symbol.
 		 */
1110
		if (verbose > 0) {
1111 1112 1113 1114 1115 1116 1117 1118 1119 1120
			/*
			 * 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;
		}
	}
1121

1122
	if (symbol__init(&session->header.env) < 0)
1123
		goto error;
1124

1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143
	if (perf_time__parse_str(report.ptime_range, report.time_str) != 0) {
		if (session->evlist->first_sample_time == 0 &&
		    session->evlist->last_sample_time == 0) {
			pr_err("No first/last sample time in perf data\n");
			return -EINVAL;
		}

		report.range_num = perf_time__percent_parse_str(
					report.ptime_range, PTIME_RANGE_MAX,
					report.time_str,
					session->evlist->first_sample_time,
					session->evlist->last_sample_time);

		if (report.range_num < 0) {
			pr_err("Invalid time string\n");
			return -EINVAL;
		}
	} else {
		report.range_num = 1;
1144 1145
	}

1146
	sort__setup_elide(stdout);
1147

1148
	ret = __cmd_report(&report);
1149 1150 1151 1152 1153 1154
	if (ret == K_SWITCH_INPUT_DATA) {
		perf_session__delete(session);
		goto repeat;
	} else
		ret = 0;

1155 1156 1157
error:
	perf_session__delete(session);
	return ret;
1158
}