builtin-report.c 31.3 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
struct report {
56
	struct perf_tool	tool;
57
	struct perf_session	*session;
58
	bool			use_tui, use_gtk, use_stdio;
59 60 61
	bool			show_full_info;
	bool			show_threads;
	bool			inverted_callchain;
62
	bool			mem_mode;
63 64
	bool			header;
	bool			header_only;
65
	bool			nonany_branch_mode;
66
	int			max_stack;
67 68 69
	struct perf_read_values	show_threads_values;
	const char		*pretty_printing_style;
	const char		*cpu_list;
70
	const char		*symbol_filter_str;
71 72
	const char		*time_str;
	struct perf_time_interval ptime;
73
	float			min_percent;
74
	u64			nr_entries;
75
	u64			queue_size;
76
	int			socket_filter;
77
	DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS);
78
	struct branch_type_stat	brtype_stat;
79
};
80

81
static int report__config(const char *var, const char *value, void *cb)
82
{
83 84
	struct report *rep = cb;

85 86 87 88
	if (!strcmp(var, "report.group")) {
		symbol_conf.event_group = perf_config_bool(var, value);
		return 0;
	}
89
	if (!strcmp(var, "report.percent-limit")) {
90 91 92 93
		double pcnt = strtof(value, NULL);

		rep->min_percent = pcnt;
		callchain_param.min_percent = pcnt;
94 95
		return 0;
	}
96 97 98 99
	if (!strcmp(var, "report.children")) {
		symbol_conf.cumulate_callchain = perf_config_bool(var, value);
		return 0;
	}
100 101 102
	if (!strcmp(var, "report.queue-size"))
		return perf_config_u64(&rep->queue_size, var, value);

103 104 105 106
	if (!strcmp(var, "report.sort_order")) {
		default_sort_order = strdup(value);
		return 0;
	}
107

108
	return 0;
109 110
}

111 112 113 114 115 116 117 118
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;
119
	struct perf_sample *sample = iter->sample;
120 121 122 123 124 125
	struct mem_info *mi;
	struct branch_info *bi;

	if (!ui__has_annotation())
		return 0;

126
	hist__account_cycles(sample->branch_stack, al, sample,
127 128
			     rep->nonany_branch_mode);

129 130
	if (sort__mode == SORT_MODE__BRANCH) {
		bi = he->branch_info;
131
		err = addr_map_symbol__inc_samples(&bi->from, sample, evsel->idx);
132 133 134
		if (err)
			goto out;

135
		err = addr_map_symbol__inc_samples(&bi->to, sample, evsel->idx);
136 137 138

	} else if (rep->mem_mode) {
		mi = he->mem_info;
139
		err = addr_map_symbol__inc_samples(&mi->daddr, sample, evsel->idx);
140 141 142
		if (err)
			goto out;

143
		err = hist_entry__inc_addr_samples(he, sample, evsel->idx, al->addr);
144 145 146

	} else if (symbol_conf.cumulate_callchain) {
		if (single)
147
			err = hist_entry__inc_addr_samples(he, sample, evsel->idx,
148 149
							   al->addr);
	} else {
150
		err = hist_entry__inc_addr_samples(he, sample, evsel->idx, al->addr);
151 152 153 154
	}

out:
	return err;
155 156
}

157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172
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;

	bi = he->branch_info;
	branch_type_count(&rep->brtype_stat, &bi->flags,
			  bi->from.addr, bi->to.addr);

	return 0;
}

173
static int process_sample_event(struct perf_tool *tool,
174
				union perf_event *event,
175
				struct perf_sample *sample,
176
				struct perf_evsel *evsel,
177
				struct machine *machine)
178
{
179
	struct report *rep = container_of(tool, struct report, tool);
180
	struct addr_location al;
181
	struct hist_entry_iter iter = {
182 183
		.evsel 			= evsel,
		.sample 		= sample,
184
		.hide_unresolved 	= symbol_conf.hide_unresolved,
185
		.add_entry_cb 		= hist_iter__report_callback,
186
	};
187
	int ret = 0;
188

189 190 191
	if (perf_time__skip_sample(&rep->ptime, sample->time))
		return 0;

192
	if (machine__resolve(machine, &al, sample) < 0) {
193 194
		pr_debug("problem processing %d event, skipping it.\n",
			 event->header.type);
195 196
		return -1;
	}
197

198
	if (symbol_conf.hide_unresolved && al.sym == NULL)
199
		goto out_put;
200

201
	if (rep->cpu_list && !test_bit(sample->cpu, rep->cpu_bitmap))
202
		goto out_put;
203

204 205 206 207 208 209 210
	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;
211 212

		iter.add_entry_cb = hist_iter__branch_callback;
213
		iter.ops = &hist_iter_branch;
214
	} else if (rep->mem_mode) {
215
		iter.ops = &hist_iter_mem;
216
	} else if (symbol_conf.cumulate_callchain) {
217
		iter.ops = &hist_iter_cumulative;
218
	} else {
219
		iter.ops = &hist_iter_normal;
220
	}
221 222 223 224

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

225
	ret = hist_entry_iter__add(&iter, &al, rep->max_stack, rep);
226 227
	if (ret < 0)
		pr_debug("problem adding hist entry, skipping event\n");
228 229
out_put:
	addr_location__put(&al);
230
	return ret;
231
}
I
Ingo Molnar 已提交
232

233
static int process_read_event(struct perf_tool *tool,
234
			      union perf_event *event,
235
			      struct perf_sample *sample __maybe_unused,
236
			      struct perf_evsel *evsel,
237
			      struct machine *machine __maybe_unused)
238
{
239
	struct report *rep = container_of(tool, struct report, tool);
240

241
	if (rep->show_threads) {
242
		const char *name = evsel ? perf_evsel__name(evsel) : "unknown";
243
		int err = perf_read_values_add_value(&rep->show_threads_values,
244
					   event->read.pid, event->read.tid,
245
					   evsel->idx,
246 247
					   name,
					   event->read.value);
248 249 250

		if (err)
			return err;
251 252
	}

253 254 255
	return 0;
}

256
/* For pipe mode, sample_type is not currently set */
257
static int report__setup_sample_type(struct report *rep)
258
{
259 260
	struct perf_session *session = rep->session;
	u64 sample_type = perf_evlist__combined_sample_type(session->evlist);
261
	bool is_pipe = perf_data__is_pipe(session->data);
262

263 264 265 266 267 268
	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;

269 270 271
	if (session->itrace_synth_opts->last_branch)
		sample_type |= PERF_SAMPLE_BRANCH_STACK;

272
	if (!is_pipe && !(sample_type & PERF_SAMPLE_CALLCHAIN)) {
273
		if (perf_hpp_list.parent) {
274
			ui__error("Selected --sort parent, but no "
275 276
				    "callchain data. Did you call "
				    "'perf record' without -g?\n");
277
			return -EINVAL;
278
		}
279 280 281 282 283
		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");
284
			return -1;
285
		}
286
	} else if (!callchain_param.enabled &&
287
		   callchain_param.mode != CHAIN_NONE &&
288
		   !symbol_conf.use_callchain) {
289
			symbol_conf.use_callchain = true;
290
			if (callchain_register_param(&callchain_param) < 0) {
291
				ui__error("Can't register callchain params.\n");
292
				return -EINVAL;
293
			}
294 295
	}

296 297 298 299 300 301 302 303
	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();
		}
	}

304
	if (sort__mode == SORT_MODE__BRANCH) {
305
		if (!is_pipe &&
306
		    !(sample_type & PERF_SAMPLE_BRANCH_STACK)) {
307 308
			ui__error("Selected -b but no branch data. "
				  "Did you call perf record without -b?\n");
309 310 311 312
			return -1;
		}
	}

313 314 315 316
	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;
317 318
		else if (sample_type & PERF_SAMPLE_BRANCH_STACK)
			callchain_param.record_mode = CALLCHAIN_LBR;
319 320 321
		else
			callchain_param.record_mode = CALLCHAIN_FP;
	}
322 323 324 325 326 327

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

328 329
	return 0;
}
330

331
static void sig_handler(int sig __maybe_unused)
332 333 334 335
{
	session_done = 1;
}

336
static size_t hists__fprintf_nr_sample_events(struct hists *hists, struct report *rep,
337 338 339 340
					      const char *evname, FILE *fp)
{
	size_t ret;
	char unit;
341 342 343
	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);
344 345
	char buf[512];
	size_t size = sizeof(buf);
346
	int socked_id = hists->socket_filter;
347

348 349 350
	if (quiet)
		return 0;

351 352 353 354 355
	if (symbol_conf.filter_relative) {
		nr_samples = hists->stats.nr_non_filtered_samples;
		nr_events = hists->stats.total_non_filtered_period;
	}

356
	if (perf_evsel__is_group_event(evsel)) {
357 358 359 360 361 362
		struct perf_evsel *pos;

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

		for_each_group_member(pos, evsel) {
363 364
			const struct hists *pos_hists = evsel__hists(pos);

365
			if (symbol_conf.filter_relative) {
366 367
				nr_samples += pos_hists->stats.nr_non_filtered_samples;
				nr_events += pos_hists->stats.total_non_filtered_period;
368
			} else {
369 370
				nr_samples += pos_hists->stats.nr_events[PERF_RECORD_SAMPLE];
				nr_events += pos_hists->stats.total_period;
371
			}
372 373
		}
	}
374

375 376
	nr_samples = convert_unit(nr_samples, &unit);
	ret = fprintf(fp, "# Samples: %lu%c", nr_samples, unit);
377
	if (evname != NULL)
378 379
		ret += fprintf(fp, " of event '%s'", evname);

380 381 382 383 384
	if (symbol_conf.show_ref_callgraph &&
	    strstr(evname, "call-graph=no")) {
		ret += fprintf(fp, ", show reference callgraph");
	}

385 386
	if (rep->mem_mode) {
		ret += fprintf(fp, "\n# Total weight : %" PRIu64, nr_events);
387
		ret += fprintf(fp, "\n# Sort order   : %s", sort_order ? : default_mem_sort_order);
388 389
	} else
		ret += fprintf(fp, "\n# Event count (approx.): %" PRIu64, nr_events);
390

391 392
	if (socked_id > -1)
		ret += fprintf(fp, "\n# Processor Socket: %d", socked_id);
393

394 395 396
	return ret + fprintf(fp, "\n#\n");
}

397
static int perf_evlist__tty_browse_hists(struct perf_evlist *evlist,
398
					 struct report *rep,
399
					 const char *help)
400
{
401
	struct perf_evsel *pos;
402

403 404 405 406 407
	if (!quiet) {
		fprintf(stdout, "#\n# Total Lost Samples: %" PRIu64 "\n#\n",
			evlist->stats.total_lost_samples);
	}

408
	evlist__for_each_entry(evlist, pos) {
409
		struct hists *hists = evsel__hists(pos);
410
		const char *evname = perf_evsel__name(pos);
411

412 413 414 415
		if (symbol_conf.event_group &&
		    !perf_evsel__is_group_leader(pos))
			continue;

416
		hists__fprintf_nr_sample_events(hists, rep, evname, stdout);
417
		hists__fprintf(hists, !quiet, 0, 0, rep->min_percent, stdout,
418 419
			       symbol_conf.use_callchain ||
			       symbol_conf.show_branchflag_count);
420 421 422
		fprintf(stdout, "\n\n");
	}

423
	if (!quiet)
424 425
		fprintf(stdout, "#\n# (%s)\n#\n", help);

426 427 428 429 430
	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);
431 432
	}

433 434 435
	if (sort__mode == SORT_MODE__BRANCH)
		branch_type_stat_display(stdout, &rep->brtype_stat);

436 437 438
	return 0;
}

439 440
static void report__warn_kptr_restrict(const struct report *rep)
{
441
	struct map *kernel_map = machine__kernel_map(&rep->session->machines.host);
442
	struct kmap *kernel_kmap = kernel_map ? map__kmap(kernel_map) : NULL;
443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467

	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);
	}
}

468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487
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;
488 489 490 491 492 493 494 495
	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!";
	}
496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519

	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;
}

520
static int report__collapse_hists(struct report *rep)
521 522 523
{
	struct ui_progress prog;
	struct perf_evsel *pos;
524
	int ret = 0;
525

526
	ui_progress__init(&prog, rep->nr_entries, "Merging related events...");
527

528
	evlist__for_each_entry(rep->session->evlist, pos) {
529
		struct hists *hists = evsel__hists(pos);
530 531 532 533

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

534 535
		hists->socket_filter = rep->socket_filter;

536 537 538
		ret = hists__collapse_resort(hists, &prog);
		if (ret < 0)
			break;
539 540 541 542

		/* Non-group events are considered as leader */
		if (symbol_conf.event_group &&
		    !perf_evsel__is_group_leader(pos)) {
543
			struct hists *leader_hists = evsel__hists(pos->leader);
544 545 546 547 548 549 550

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

	ui_progress__finish();
551
	return ret;
552 553
}

554 555 556 557 558 559 560
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...");

561
	evlist__for_each_entry(rep->session->evlist, pos)
562
		perf_evsel__output_resort(pos, &prog);
563 564 565 566

	ui_progress__finish();
}

567
static int __cmd_report(struct report *rep)
568
{
569
	int ret;
570
	struct perf_session *session = rep->session;
571
	struct perf_evsel *pos;
572
	struct perf_data *data = session->data;
573

574 575
	signal(SIGINT, sig_handler);

576 577 578
	if (rep->cpu_list) {
		ret = perf_session__cpu_bitmap(session, rep->cpu_list,
					       rep->cpu_bitmap);
579 580
		if (ret) {
			ui__error("failed to set cpu bitmap\n");
581
			return ret;
582
		}
583
		session->itrace_synth_opts->cpu_bitmap = rep->cpu_bitmap;
584 585
	}

586 587 588 589 590
	if (rep->show_threads) {
		ret = perf_read_values_init(&rep->show_threads_values);
		if (ret)
			return ret;
	}
591

592
	ret = report__setup_sample_type(rep);
593 594
	if (ret) {
		/* report__setup_sample_type() already showed error message */
595
		return ret;
596
	}
597

598
	ret = perf_session__process_events(session);
599 600
	if (ret) {
		ui__error("failed to process sample\n");
601
		return ret;
602
	}
603

604
	report__warn_kptr_restrict(rep);
605

606
	evlist__for_each_entry(session->evlist, pos)
607 608
		rep->nr_entries += evsel__hists(pos)->nr_entries;

609 610 611
	if (use_browser == 0) {
		if (verbose > 3)
			perf_session__fprintf(session, stdout);
612

613 614
		if (verbose > 2)
			perf_session__fprintf_dsos(session, stdout);
615

616 617
		if (dump_trace) {
			perf_session__fprintf_nr_events(session, stdout);
618
			perf_evlist__fprintf_nr_events(session->evlist, stdout);
619 620
			return 0;
		}
621 622
	}

623 624 625 626 627
	ret = report__collapse_hists(rep);
	if (ret) {
		ui__error("failed to process hist entry\n");
		return ret;
	}
628

629 630 631
	if (session_done())
		return 0;

632 633 634 635 636
	/*
	 * recalculate number of entries after collapsing since it
	 * might be changed during the collapse phase.
	 */
	rep->nr_entries = 0;
637
	evlist__for_each_entry(session->evlist, pos)
638 639
		rep->nr_entries += evsel__hists(pos)->nr_entries;

640
	if (rep->nr_entries == 0) {
J
Jiri Olsa 已提交
641
		ui__error("The %s file has no samples!\n", data->file.path);
642
		return 0;
643 644
	}

645
	report__output_resort(rep);
646

647
	return report__browse_hists(rep);
648 649
}

650
static int
651
report_parse_callchain_opt(const struct option *opt, const char *arg, int unset)
652
{
653
	struct callchain_param *callchain = opt->value;
654

655
	callchain->enabled = !unset;
656 657 658 659
	/*
	 * --no-call-graph
	 */
	if (unset) {
660 661
		symbol_conf.use_callchain = false;
		callchain->mode = CHAIN_NONE;
662 663 664
		return 0;
	}

665
	return parse_callchain_report_opt(arg);
666 667
}

668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685
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;
}

686
static int
687
parse_branch_mode(const struct option *opt,
688
		  const char *str __maybe_unused, int unset)
689
{
690 691 692
	int *branch_mode = opt->value;

	*branch_mode = !unset;
693 694 695
	return 0;
}

696 697 698 699
static int
parse_percent_limit(const struct option *opt, const char *str,
		    int unset __maybe_unused)
{
700
	struct report *rep = opt->value;
701
	double pcnt = strtof(str, NULL);
702

703 704
	rep->min_percent = pcnt;
	callchain_param.min_percent = pcnt;
705 706 707
	return 0;
}

708
#define CALLCHAIN_DEFAULT_OPT  "graph,0.5,caller,function,percent"
709 710 711 712

const char report_callchain_help[] = "Display call graph (stack chain/backtrace):\n\n"
				     CALLCHAIN_REPORT_HELP
				     "\n\t\t\t\tDefault: " CALLCHAIN_DEFAULT_OPT;
713

714
int cmd_report(int argc, const char **argv)
715
{
716
	struct perf_session *session;
717
	struct itrace_synth_opts itrace_synth_opts = { .set = 0, };
718
	struct stat st;
719
	bool has_br_stack = false;
720
	int branch_mode = -1;
721
	bool branch_call_mode = false;
722
	char callchain_default_opt[] = CALLCHAIN_DEFAULT_OPT;
723
	const char * const report_usage[] = {
N
Namhyung Kim 已提交
724
		"perf report [<options>]",
725 726
		NULL
	};
727
	struct report report = {
728
		.tool = {
729 730
			.sample		 = process_sample_event,
			.mmap		 = perf_event__process_mmap,
731
			.mmap2		 = perf_event__process_mmap2,
732
			.comm		 = perf_event__process_comm,
733
			.namespaces	 = perf_event__process_namespaces,
734 735
			.exit		 = perf_event__process_exit,
			.fork		 = perf_event__process_fork,
736 737 738 739 740
			.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,
741 742 743
			.id_index	 = perf_event__process_id_index,
			.auxtrace_info	 = perf_event__process_auxtrace_info,
			.auxtrace	 = perf_event__process_auxtrace,
744
			.feature	 = perf_event__process_feature,
745
			.ordered_events	 = true,
746 747
			.ordering_requires_timestamps = true,
		},
748
		.max_stack		 = PERF_MAX_STACK_DEPTH,
749
		.pretty_printing_style	 = "normal",
750
		.socket_filter		 = -1,
751 752
	};
	const struct option options[] = {
753
	OPT_STRING('i', "input", &input_name, "file",
754
		    "input file name"),
755
	OPT_INCR('v', "verbose", &verbose,
756
		    "be more verbose (show symbol address, etc)"),
757
	OPT_BOOLEAN('q', "quiet", &quiet, "Do not show any message"),
758 759
	OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
		    "dump raw trace in ASCII"),
760 761
	OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name,
		   "file", "vmlinux pathname"),
762 763
	OPT_STRING(0, "kallsyms", &symbol_conf.kallsyms_name,
		   "file", "kallsyms pathname"),
764
	OPT_BOOLEAN('f', "force", &symbol_conf.force, "don't complain, do it"),
765
	OPT_BOOLEAN('m', "modules", &symbol_conf.use_modules,
766
		    "load module symbols - WARNING: use only with -k and LIVE kernel"),
767
	OPT_BOOLEAN('n', "show-nr-samples", &symbol_conf.show_nr_samples,
768
		    "Show a column with the number of samples"),
769
	OPT_BOOLEAN('T', "threads", &report.show_threads,
770
		    "Show per-thread event counters"),
771
	OPT_STRING(0, "pretty", &report.pretty_printing_style, "key",
772
		   "pretty printing style key: normal raw"),
773
	OPT_BOOLEAN(0, "tui", &report.use_tui, "Use the TUI interface"),
774
	OPT_BOOLEAN(0, "gtk", &report.use_gtk, "Use the GTK2 interface"),
775 776
	OPT_BOOLEAN(0, "stdio", &report.use_stdio,
		    "Use the stdio interface"),
777 778 779
	OPT_BOOLEAN(0, "header", &report.header, "Show data header."),
	OPT_BOOLEAN(0, "header-only", &report.header_only,
		    "Show only data header."),
780
	OPT_STRING('s', "sort", &sort_order, "key[,key2...]",
781 782
		   "sort by key(s): pid, comm, dso, symbol, parent, cpu, srcline, ..."
		   " Please refer the man page for the complete list."),
783 784
	OPT_STRING('F', "fields", &field_order, "key[,keys...]",
		   "output field(s): overhead, period, sample plus all of sort keys"),
785
	OPT_BOOLEAN(0, "show-cpu-utilization", &symbol_conf.show_cpu_utilization,
786
		    "Show sample percentage for different cpu modes"),
787 788
	OPT_BOOLEAN_FLAG(0, "showcpuutilization", &symbol_conf.show_cpu_utilization,
		    "Show sample percentage for different cpu modes", PARSE_OPT_HIDDEN),
789 790
	OPT_STRING('p', "parent", &parent_pattern, "regex",
		   "regex filter to identify parent, see: '--sort parent'"),
791
	OPT_BOOLEAN('x', "exclude-other", &symbol_conf.exclude_other,
792
		    "Only display entries with parent-match"),
793
	OPT_CALLBACK_DEFAULT('g', "call-graph", &callchain_param,
794
			     "print_type,threshold[,print_limit],order,sort_key[,branch],value",
795 796
			     report_callchain_help, &report_parse_callchain_opt,
			     callchain_default_opt),
797 798
	OPT_BOOLEAN(0, "children", &symbol_conf.cumulate_callchain,
		    "Accumulate callchains of children and show total overhead as well"),
799 800 801
	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. "
802
		    "Default: kernel.perf_event_max_stack or " __stringify(PERF_MAX_STACK_DEPTH)),
803 804
	OPT_BOOLEAN('G', "inverted", &report.inverted_callchain,
		    "alias for inverted call graph"),
805 806 807
	OPT_CALLBACK(0, "ignore-callees", NULL, "regex",
		   "ignore callees of these functions in call graphs",
		   report_parse_ignore_callees_opt),
808
	OPT_STRING('d', "dsos", &symbol_conf.dso_list_str, "dso[,dso...]",
809
		   "only consider symbols in these dsos"),
810
	OPT_STRING('c', "comms", &symbol_conf.comm_list_str, "comm[,comm...]",
811
		   "only consider symbols in these comms"),
812 813 814 815
	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"),
816
	OPT_STRING('S', "symbols", &symbol_conf.sym_list_str, "symbol[,symbol...]",
817
		   "only consider these symbols"),
818 819
	OPT_STRING(0, "symbol-filter", &report.symbol_filter_str, "filter",
		   "only show symbols that (partially) match with this filter"),
820
	OPT_STRING('w', "column-widths", &symbol_conf.col_width_list_str,
821 822
		   "width[,width...]",
		   "don't try to adjust column width, use these fixed values"),
823
	OPT_STRING_NOEMPTY('t', "field-separator", &symbol_conf.field_sep, "separator",
824 825
		   "separator for columns, no spaces will be added between "
		   "columns '.' is reserved."),
826
	OPT_BOOLEAN('U', "hide-unresolved", &symbol_conf.hide_unresolved,
827
		    "Only display entries resolved to a symbol"),
828 829 830
	OPT_CALLBACK(0, "symfs", NULL, "directory",
		     "Look for files with symbols relative to this directory",
		     symbol__config_symfs),
831
	OPT_STRING('C', "cpu", &report.cpu_list, "cpu",
832 833
		   "list of cpus to profile"),
	OPT_BOOLEAN('I', "show-info", &report.show_full_info,
834
		    "Display extended information about perf.data file"),
835 836 837 838
	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)"),
839 840
	OPT_STRING('M', "disassembler-style", &disassembler_style, "disassembler style",
		   "Specify disassembler style (e.g. -M intel for intel syntax)"),
841 842
	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
843 844
	OPT_BOOLEAN(0, "group", &symbol_conf.event_group,
		    "Show event group information together"),
845
	OPT_CALLBACK_NOOPT('b', "branch-stack", &branch_mode, "",
846 847 848 849
		    "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"),
850 851
	OPT_STRING(0, "objdump", &objdump_path, "path",
		   "objdump binary to use for disassembly and annotations"),
852 853
	OPT_BOOLEAN(0, "demangle", &symbol_conf.demangle,
		    "Disable symbol demangling"),
854 855
	OPT_BOOLEAN(0, "demangle-kernel", &symbol_conf.demangle_kernel,
		    "Enable kernel symbol demangling"),
856
	OPT_BOOLEAN(0, "mem-mode", &report.mem_mode, "mem access profile"),
857 858
	OPT_CALLBACK(0, "percent-limit", &report, "percent",
		     "Don't show entries under that percent", parse_percent_limit),
859
	OPT_CALLBACK(0, "percentage", NULL, "relative|absolute",
N
Namhyung Kim 已提交
860
		     "how to display percentage of filtered entries", parse_filter_percentage),
861 862 863
	OPT_CALLBACK_OPTARG(0, "itrace", &itrace_synth_opts, NULL, "opts",
			    "Instruction Tracing options",
			    itrace_parse_synth_opts),
864 865
	OPT_BOOLEAN(0, "full-source-path", &srcline_full_filename,
			"Show full source file name path for source lines"),
866 867
	OPT_BOOLEAN(0, "show-ref-call-graph", &symbol_conf.show_ref_callgraph,
		    "Show callgraph from reference event"),
868 869
	OPT_INTEGER(0, "socket-filter", &report.socket_filter,
		    "only show processor socket that match with this filter"),
870 871
	OPT_BOOLEAN(0, "raw-trace", &symbol_conf.raw_trace,
		    "Show raw trace event output (do not use print fmt or plugins)"),
872 873
	OPT_BOOLEAN(0, "hierarchy", &symbol_conf.report_hierarchy,
		    "Show entries in a hierarchy"),
874 875 876
	OPT_CALLBACK_DEFAULT(0, "stdio-color", NULL, "mode",
			     "'always' (default), 'never' or 'auto' only applicable to --stdio mode",
			     stdio__config_color, "always"),
877 878
	OPT_STRING(0, "time", &report.time_str, "str",
		   "Time span of interest (start,stop)"),
J
Jin Yao 已提交
879 880
	OPT_BOOLEAN(0, "inline", &symbol_conf.inline_name,
		    "Show inline function"),
881
	OPT_END()
882
	};
883
	struct perf_data data = {
884 885
		.mode  = PERF_DATA_MODE_READ,
	};
886 887 888 889
	int ret = hists__init();

	if (ret < 0)
		return ret;
890

891 892 893
	ret = perf_config(report__config, &report);
	if (ret)
		return ret;
894

895
	argc = parse_options(argc, argv, options, report_usage, 0);
896 897 898 899 900 901 902 903 904 905
	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];
	}
906

907 908 909
	if (quiet)
		perf_quiet_option();

910 911 912 913 914 915 916 917 918 919 920
	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;
	}

921
	if (report.use_stdio)
922
		use_browser = 0;
923
	else if (report.use_tui)
924
		use_browser = 1;
925 926
	else if (report.use_gtk)
		use_browser = 2;
927

928
	if (report.inverted_callchain)
929
		callchain_param.order = ORDER_CALLER;
930 931
	if (symbol_conf.cumulate_callchain && !callchain_param.order_set)
		callchain_param.order = ORDER_CALLER;
932

933 934 935 936
	if (itrace_synth_opts.callchain &&
	    (int)itrace_synth_opts.callchain_sz > report.max_stack)
		report.max_stack = itrace_synth_opts.callchain_sz;

937
	if (!input_name || !strlen(input_name)) {
938
		if (!fstat(STDIN_FILENO, &st) && S_ISFIFO(st.st_mode))
939
			input_name = "-";
940
		else
941
			input_name = "perf.data";
942
	}
943

J
Jiri Olsa 已提交
944 945
	data.file.path = input_name;
	data.force     = symbol_conf.force;
946

947
repeat:
948
	session = perf_session__new(&data, false, &report.tool);
949
	if (session == NULL)
950
		return -1;
951

952 953 954 955 956
	if (report.queue_size) {
		ordered_events__set_alloc_size(&session->ordered_events,
					       report.queue_size);
	}

957 958
	session->itrace_synth_opts = &itrace_synth_opts;

959 960 961 962
	report.session = session;

	has_br_stack = perf_header__has_feat(&session->header,
					     HEADER_BRANCH_STACK);
963

964 965 966
	if (itrace_synth_opts.last_branch)
		has_br_stack = true;

967 968 969
	if (has_br_stack && branch_call_mode)
		symbol_conf.show_branchflag_count = true;

970 971
	memset(&report.brtype_stat, 0, sizeof(struct branch_type_stat));

972 973 974 975 976 977
	/*
	 * 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) &&
978
	    !branch_call_mode) {
979
		sort__mode = SORT_MODE__BRANCH;
980 981
		symbol_conf.cumulate_callchain = false;
	}
982
	if (branch_call_mode) {
983
		callchain_param.key = CCKEY_ADDRESS;
984 985 986 987 988 989
		callchain_param.branch_callstack = 1;
		symbol_conf.use_callchain = true;
		callchain_register_param(&callchain_param);
		if (sort_order == NULL)
			sort_order = "srcline,symbol,dso";
	}
990

991
	if (report.mem_mode) {
992
		if (sort__mode == SORT_MODE__BRANCH) {
993
			pr_err("branch and mem mode incompatible\n");
994 995
			goto error;
		}
996
		sort__mode = SORT_MODE__MEMORY;
997
		symbol_conf.cumulate_callchain = false;
998
	}
999

1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010
	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;
		}

1011
		perf_hpp_list.need_collapse = true;
1012 1013
	}

1014 1015
	/* Force tty output for header output and per-thread stat. */
	if (report.header || report.header_only || report.show_threads)
1016
		use_browser = 0;
1017 1018 1019 1020
	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;
1021

1022 1023
	if (strcmp(input_name, "-") != 0)
		setup_browser(true);
1024
	else
1025 1026
		use_browser = 0;

1027 1028 1029 1030 1031 1032 1033 1034 1035
	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;
	}

1036
	if ((report.header || report.header_only) && !quiet) {
1037 1038
		perf_session__fprintf_info(session, stdout,
					   report.show_full_info);
1039 1040 1041 1042
		if (report.header_only) {
			ret = 0;
			goto error;
		}
1043
	} else if (use_browser == 0 && !quiet) {
1044 1045 1046 1047
		fputs("# To display the perf.data header info, please use --header/--header-only options.\n#\n",
		      stdout);
	}

1048
	/*
1049
	 * Only in the TUI browser we are doing integrated annotation,
1050 1051 1052
	 * so don't allocate extra space that won't be used in the stdio
	 * implementation.
	 */
1053
	if (ui__has_annotation()) {
1054 1055 1056
		ret = symbol__annotation_init();
		if (ret < 0)
			goto error;
1057 1058 1059 1060 1061
		/*
 		 * For searching by name on the "Browse map details".
 		 * providing it only in verbose mode not to bloat too
 		 * much struct symbol.
 		 */
1062
		if (verbose > 0) {
1063 1064 1065 1066 1067 1068 1069 1070 1071 1072
			/*
			 * 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;
		}
	}
1073

1074
	if (symbol__init(&session->header.env) < 0)
1075
		goto error;
1076

1077 1078 1079 1080 1081
	if (perf_time__parse_str(&report.ptime, report.time_str) != 0) {
		pr_err("Invalid time string\n");
		return -EINVAL;
	}

1082
	sort__setup_elide(stdout);
1083

1084
	ret = __cmd_report(&report);
1085 1086 1087 1088 1089 1090
	if (ret == K_SWITCH_INPUT_DATA) {
		perf_session__delete(session);
		goto repeat;
	} else
		ret = 0;

1091 1092 1093
error:
	perf_session__delete(session);
	return ret;
1094
}