builtin-report.c 29.5 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/strlist.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

40 41
#include "util/auxtrace.h"

42
#include <dlfcn.h>
43
#include <linux/bitmap.h>
44
#include <linux/stringify.h>
45

46
struct report {
47
	struct perf_tool	tool;
48
	struct perf_session	*session;
49
	bool			use_tui, use_gtk, use_stdio;
50 51 52
	bool			show_full_info;
	bool			show_threads;
	bool			inverted_callchain;
53
	bool			mem_mode;
54 55
	bool			header;
	bool			header_only;
56
	bool			nonany_branch_mode;
57
	int			max_stack;
58 59 60
	struct perf_read_values	show_threads_values;
	const char		*pretty_printing_style;
	const char		*cpu_list;
61
	const char		*symbol_filter_str;
62
	float			min_percent;
63
	u64			nr_entries;
64
	u64			queue_size;
65
	int			socket_filter;
66
	DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS);
67
};
68

69
static int report__config(const char *var, const char *value, void *cb)
70
{
71 72
	struct report *rep = cb;

73 74 75 76
	if (!strcmp(var, "report.group")) {
		symbol_conf.event_group = perf_config_bool(var, value);
		return 0;
	}
77
	if (!strcmp(var, "report.percent-limit")) {
78 79 80 81
		double pcnt = strtof(value, NULL);

		rep->min_percent = pcnt;
		callchain_param.min_percent = pcnt;
82 83
		return 0;
	}
84 85 86 87
	if (!strcmp(var, "report.children")) {
		symbol_conf.cumulate_callchain = perf_config_bool(var, value);
		return 0;
	}
88 89 90 91
	if (!strcmp(var, "report.queue-size")) {
		rep->queue_size = perf_config_u64(var, value);
		return 0;
	}
92 93 94 95
	if (!strcmp(var, "report.sort_order")) {
		default_sort_order = strdup(value);
		return 0;
	}
96

97
	return 0;
98 99
}

100 101 102 103 104 105 106 107 108 109 110 111 112 113
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;

114 115 116
	hist__account_cycles(iter->sample->branch_stack, al, iter->sample,
			     rep->nonany_branch_mode);

117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142
	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;
143 144
}

145
static int process_sample_event(struct perf_tool *tool,
146
				union perf_event *event,
147
				struct perf_sample *sample,
148
				struct perf_evsel *evsel,
149
				struct machine *machine)
150
{
151
	struct report *rep = container_of(tool, struct report, tool);
152
	struct addr_location al;
153
	struct hist_entry_iter iter = {
154 155
		.evsel 			= evsel,
		.sample 		= sample,
156
		.hide_unresolved 	= symbol_conf.hide_unresolved,
157
		.add_entry_cb 		= hist_iter__report_callback,
158
	};
159
	int ret = 0;
160

161
	if (machine__resolve(machine, &al, sample) < 0) {
162 163
		pr_debug("problem processing %d event, skipping it.\n",
			 event->header.type);
164 165
		return -1;
	}
166

167
	if (symbol_conf.hide_unresolved && al.sym == NULL)
168
		goto out_put;
169

170
	if (rep->cpu_list && !test_bit(sample->cpu, rep->cpu_bitmap))
171
		goto out_put;
172

173 174 175 176 177 178 179
	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;
180
		iter.ops = &hist_iter_branch;
181
	} else if (rep->mem_mode) {
182
		iter.ops = &hist_iter_mem;
183
	} else if (symbol_conf.cumulate_callchain) {
184
		iter.ops = &hist_iter_cumulative;
185
	} else {
186
		iter.ops = &hist_iter_normal;
187
	}
188 189 190 191

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

192
	ret = hist_entry_iter__add(&iter, &al, rep->max_stack, rep);
193 194
	if (ret < 0)
		pr_debug("problem adding hist entry, skipping event\n");
195 196
out_put:
	addr_location__put(&al);
197
	return ret;
198
}
I
Ingo Molnar 已提交
199

200
static int process_read_event(struct perf_tool *tool,
201
			      union perf_event *event,
202
			      struct perf_sample *sample __maybe_unused,
203
			      struct perf_evsel *evsel,
204
			      struct machine *machine __maybe_unused)
205
{
206
	struct report *rep = container_of(tool, struct report, tool);
207

208
	if (rep->show_threads) {
209
		const char *name = evsel ? perf_evsel__name(evsel) : "unknown";
210
		int err = perf_read_values_add_value(&rep->show_threads_values,
211 212 213 214
					   event->read.pid, event->read.tid,
					   event->read.id,
					   name,
					   event->read.value);
215 216 217

		if (err)
			return err;
218 219
	}

220
	dump_printf(": %d %d %s %" PRIu64 "\n", event->read.pid, event->read.tid,
221
		    evsel ? perf_evsel__name(evsel) : "FAIL",
222
		    event->read.value);
223 224 225 226

	return 0;
}

227
/* For pipe mode, sample_type is not currently set */
228
static int report__setup_sample_type(struct report *rep)
229
{
230 231 232
	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);
233

234 235 236 237 238 239
	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;

240 241 242
	if (session->itrace_synth_opts->last_branch)
		sample_type |= PERF_SAMPLE_BRANCH_STACK;

243
	if (!is_pipe && !(sample_type & PERF_SAMPLE_CALLCHAIN)) {
244
		if (perf_hpp_list.parent) {
245
			ui__error("Selected --sort parent, but no "
246 247
				    "callchain data. Did you call "
				    "'perf record' without -g?\n");
248
			return -EINVAL;
249
		}
250
		if (symbol_conf.use_callchain) {
251 252 253
			ui__error("Selected -g or --branch-history but no "
				  "callchain data. Did\n"
				  "you call 'perf record' without -g?\n");
254
			return -1;
255
		}
256
	} else if (!callchain_param.enabled &&
257
		   callchain_param.mode != CHAIN_NONE &&
258
		   !symbol_conf.use_callchain) {
259
			symbol_conf.use_callchain = true;
260
			if (callchain_register_param(&callchain_param) < 0) {
261
				ui__error("Can't register callchain params.\n");
262
				return -EINVAL;
263
			}
264 265
	}

266 267 268 269 270 271 272 273
	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();
		}
	}

274
	if (sort__mode == SORT_MODE__BRANCH) {
275
		if (!is_pipe &&
276
		    !(sample_type & PERF_SAMPLE_BRANCH_STACK)) {
277 278
			ui__error("Selected -b but no branch data. "
				  "Did you call perf record without -b?\n");
279 280 281 282
			return -1;
		}
	}

283 284 285 286
	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;
287 288
		else if (sample_type & PERF_SAMPLE_BRANCH_STACK)
			callchain_param.record_mode = CALLCHAIN_LBR;
289 290 291
		else
			callchain_param.record_mode = CALLCHAIN_FP;
	}
292 293 294 295 296 297

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

298 299
	return 0;
}
300

301
static void sig_handler(int sig __maybe_unused)
302 303 304 305
{
	session_done = 1;
}

306
static size_t hists__fprintf_nr_sample_events(struct hists *hists, struct report *rep,
307 308 309 310
					      const char *evname, FILE *fp)
{
	size_t ret;
	char unit;
311 312 313
	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);
314 315
	char buf[512];
	size_t size = sizeof(buf);
316
	int socked_id = hists->socket_filter;
317

318 319 320 321 322
	if (symbol_conf.filter_relative) {
		nr_samples = hists->stats.nr_non_filtered_samples;
		nr_events = hists->stats.total_non_filtered_period;
	}

323
	if (perf_evsel__is_group_event(evsel)) {
324 325 326 327 328 329
		struct perf_evsel *pos;

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

		for_each_group_member(pos, evsel) {
330 331
			const struct hists *pos_hists = evsel__hists(pos);

332
			if (symbol_conf.filter_relative) {
333 334
				nr_samples += pos_hists->stats.nr_non_filtered_samples;
				nr_events += pos_hists->stats.total_non_filtered_period;
335
			} else {
336 337
				nr_samples += pos_hists->stats.nr_events[PERF_RECORD_SAMPLE];
				nr_events += pos_hists->stats.total_period;
338
			}
339 340
		}
	}
341

342 343
	nr_samples = convert_unit(nr_samples, &unit);
	ret = fprintf(fp, "# Samples: %lu%c", nr_samples, unit);
344
	if (evname != NULL)
345 346
		ret += fprintf(fp, " of event '%s'", evname);

347 348 349 350 351
	if (symbol_conf.show_ref_callgraph &&
	    strstr(evname, "call-graph=no")) {
		ret += fprintf(fp, ", show reference callgraph");
	}

352 353
	if (rep->mem_mode) {
		ret += fprintf(fp, "\n# Total weight : %" PRIu64, nr_events);
354
		ret += fprintf(fp, "\n# Sort order   : %s", sort_order ? : default_mem_sort_order);
355 356
	} else
		ret += fprintf(fp, "\n# Event count (approx.): %" PRIu64, nr_events);
357

358 359
	if (socked_id > -1)
		ret += fprintf(fp, "\n# Processor Socket: %d", socked_id);
360

361 362 363
	return ret + fprintf(fp, "\n#\n");
}

364
static int perf_evlist__tty_browse_hists(struct perf_evlist *evlist,
365
					 struct report *rep,
366
					 const char *help)
367
{
368
	struct perf_evsel *pos;
369

370
	fprintf(stdout, "#\n# Total Lost Samples: %" PRIu64 "\n#\n", evlist->stats.total_lost_samples);
371
	evlist__for_each_entry(evlist, pos) {
372
		struct hists *hists = evsel__hists(pos);
373
		const char *evname = perf_evsel__name(pos);
374

375 376 377 378
		if (symbol_conf.event_group &&
		    !perf_evsel__is_group_leader(pos))
			continue;

379
		hists__fprintf_nr_sample_events(hists, rep, evname, stdout);
380 381
		hists__fprintf(hists, true, 0, 0, rep->min_percent, stdout,
			       symbol_conf.use_callchain);
382 383 384
		fprintf(stdout, "\n\n");
	}

385
	if (sort_order == NULL &&
386
	    parent_pattern == default_parent_pattern)
387 388
		fprintf(stdout, "#\n# (%s)\n#\n", help);

389 390 391 392 393
	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);
394 395 396 397 398
	}

	return 0;
}

399 400
static void report__warn_kptr_restrict(const struct report *rep)
{
401
	struct map *kernel_map = machine__kernel_map(&rep->session->machines.host);
402
	struct kmap *kernel_kmap = kernel_map ? map__kmap(kernel_map) : NULL;
403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427

	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);
	}
}

428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447
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;
448 449 450 451 452 453 454 455
	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!";
	}
456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479

	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;
}

480
static int report__collapse_hists(struct report *rep)
481 482 483
{
	struct ui_progress prog;
	struct perf_evsel *pos;
484
	int ret = 0;
485

486
	ui_progress__init(&prog, rep->nr_entries, "Merging related events...");
487

488
	evlist__for_each_entry(rep->session->evlist, pos) {
489
		struct hists *hists = evsel__hists(pos);
490 491 492 493

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

494 495
		hists->socket_filter = rep->socket_filter;

496 497 498
		ret = hists__collapse_resort(hists, &prog);
		if (ret < 0)
			break;
499 500 501 502

		/* Non-group events are considered as leader */
		if (symbol_conf.event_group &&
		    !perf_evsel__is_group_leader(pos)) {
503
			struct hists *leader_hists = evsel__hists(pos->leader);
504 505 506 507 508 509 510

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

	ui_progress__finish();
511
	return ret;
512 513
}

514 515 516 517 518 519 520
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...");

521
	evlist__for_each_entry(rep->session->evlist, pos)
522
		perf_evsel__output_resort(pos, &prog);
523 524 525 526

	ui_progress__finish();
}

527
static int __cmd_report(struct report *rep)
528
{
529
	int ret;
530
	struct perf_session *session = rep->session;
531
	struct perf_evsel *pos;
532
	struct perf_data_file *file = session->file;
533

534 535
	signal(SIGINT, sig_handler);

536 537 538
	if (rep->cpu_list) {
		ret = perf_session__cpu_bitmap(session, rep->cpu_list,
					       rep->cpu_bitmap);
539 540
		if (ret) {
			ui__error("failed to set cpu bitmap\n");
541
			return ret;
542
		}
543 544
	}

545 546 547 548 549
	if (rep->show_threads) {
		ret = perf_read_values_init(&rep->show_threads_values);
		if (ret)
			return ret;
	}
550

551
	ret = report__setup_sample_type(rep);
552 553
	if (ret) {
		/* report__setup_sample_type() already showed error message */
554
		return ret;
555
	}
556

557
	ret = perf_session__process_events(session);
558 559
	if (ret) {
		ui__error("failed to process sample\n");
560
		return ret;
561
	}
562

563
	report__warn_kptr_restrict(rep);
564

565
	evlist__for_each_entry(session->evlist, pos)
566 567
		rep->nr_entries += evsel__hists(pos)->nr_entries;

568 569 570
	if (use_browser == 0) {
		if (verbose > 3)
			perf_session__fprintf(session, stdout);
571

572 573
		if (verbose > 2)
			perf_session__fprintf_dsos(session, stdout);
574

575 576
		if (dump_trace) {
			perf_session__fprintf_nr_events(session, stdout);
577
			perf_evlist__fprintf_nr_events(session->evlist, stdout);
578 579
			return 0;
		}
580 581
	}

582 583 584 585 586
	ret = report__collapse_hists(rep);
	if (ret) {
		ui__error("failed to process hist entry\n");
		return ret;
	}
587

588 589 590
	if (session_done())
		return 0;

591 592 593 594 595
	/*
	 * recalculate number of entries after collapsing since it
	 * might be changed during the collapse phase.
	 */
	rep->nr_entries = 0;
596
	evlist__for_each_entry(session->evlist, pos)
597 598
		rep->nr_entries += evsel__hists(pos)->nr_entries;

599
	if (rep->nr_entries == 0) {
600
		ui__error("The %s file has no samples!\n", file->path);
601
		return 0;
602 603
	}

604
	report__output_resort(rep);
605

606
	return report__browse_hists(rep);
607 608
}

609
static int
610
report_parse_callchain_opt(const struct option *opt, const char *arg, int unset)
611
{
612
	struct callchain_param *callchain = opt->value;
613

614
	callchain->enabled = !unset;
615 616 617 618
	/*
	 * --no-call-graph
	 */
	if (unset) {
619 620
		symbol_conf.use_callchain = false;
		callchain->mode = CHAIN_NONE;
621 622 623
		return 0;
	}

624
	return parse_callchain_report_opt(arg);
625 626
}

627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644
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;
}

645
static int
646 647
parse_branch_mode(const struct option *opt __maybe_unused,
		  const char *str __maybe_unused, int unset)
648
{
649 650 651
	int *branch_mode = opt->value;

	*branch_mode = !unset;
652 653 654
	return 0;
}

655 656 657 658
static int
parse_percent_limit(const struct option *opt, const char *str,
		    int unset __maybe_unused)
{
659
	struct report *rep = opt->value;
660
	double pcnt = strtof(str, NULL);
661

662 663
	rep->min_percent = pcnt;
	callchain_param.min_percent = pcnt;
664 665 666
	return 0;
}

667
#define CALLCHAIN_DEFAULT_OPT  "graph,0.5,caller,function,percent"
668 669 670 671

const char report_callchain_help[] = "Display call graph (stack chain/backtrace):\n\n"
				     CALLCHAIN_REPORT_HELP
				     "\n\t\t\t\tDefault: " CALLCHAIN_DEFAULT_OPT;
672

673
int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused)
674
{
675
	struct perf_session *session;
676
	struct itrace_synth_opts itrace_synth_opts = { .set = 0, };
677
	struct stat st;
678
	bool has_br_stack = false;
679
	int branch_mode = -1;
680
	bool branch_call_mode = false;
681
	char callchain_default_opt[] = CALLCHAIN_DEFAULT_OPT;
682
	const char * const report_usage[] = {
N
Namhyung Kim 已提交
683
		"perf report [<options>]",
684 685
		NULL
	};
686
	struct report report = {
687
		.tool = {
688 689
			.sample		 = process_sample_event,
			.mmap		 = perf_event__process_mmap,
690
			.mmap2		 = perf_event__process_mmap2,
691
			.comm		 = perf_event__process_comm,
692 693
			.exit		 = perf_event__process_exit,
			.fork		 = perf_event__process_fork,
694 695 696 697 698
			.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,
699 700 701
			.id_index	 = perf_event__process_id_index,
			.auxtrace_info	 = perf_event__process_auxtrace_info,
			.auxtrace	 = perf_event__process_auxtrace,
702
			.ordered_events	 = true,
703 704
			.ordering_requires_timestamps = true,
		},
705
		.max_stack		 = PERF_MAX_STACK_DEPTH,
706
		.pretty_printing_style	 = "normal",
707
		.socket_filter		 = -1,
708 709
	};
	const struct option options[] = {
710
	OPT_STRING('i', "input", &input_name, "file",
711
		    "input file name"),
712
	OPT_INCR('v', "verbose", &verbose,
713
		    "be more verbose (show symbol address, etc)"),
714 715
	OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
		    "dump raw trace in ASCII"),
716 717
	OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name,
		   "file", "vmlinux pathname"),
718 719
	OPT_STRING(0, "kallsyms", &symbol_conf.kallsyms_name,
		   "file", "kallsyms pathname"),
720
	OPT_BOOLEAN('f', "force", &symbol_conf.force, "don't complain, do it"),
721
	OPT_BOOLEAN('m', "modules", &symbol_conf.use_modules,
722
		    "load module symbols - WARNING: use only with -k and LIVE kernel"),
723
	OPT_BOOLEAN('n', "show-nr-samples", &symbol_conf.show_nr_samples,
724
		    "Show a column with the number of samples"),
725
	OPT_BOOLEAN('T', "threads", &report.show_threads,
726
		    "Show per-thread event counters"),
727
	OPT_STRING(0, "pretty", &report.pretty_printing_style, "key",
728
		   "pretty printing style key: normal raw"),
729
	OPT_BOOLEAN(0, "tui", &report.use_tui, "Use the TUI interface"),
730
	OPT_BOOLEAN(0, "gtk", &report.use_gtk, "Use the GTK2 interface"),
731 732
	OPT_BOOLEAN(0, "stdio", &report.use_stdio,
		    "Use the stdio interface"),
733 734 735
	OPT_BOOLEAN(0, "header", &report.header, "Show data header."),
	OPT_BOOLEAN(0, "header-only", &report.header_only,
		    "Show only data header."),
736
	OPT_STRING('s', "sort", &sort_order, "key[,key2...]",
737 738
		   "sort by key(s): pid, comm, dso, symbol, parent, cpu, srcline, ..."
		   " Please refer the man page for the complete list."),
739 740
	OPT_STRING('F', "fields", &field_order, "key[,keys...]",
		   "output field(s): overhead, period, sample plus all of sort keys"),
741
	OPT_BOOLEAN(0, "show-cpu-utilization", &symbol_conf.show_cpu_utilization,
742
		    "Show sample percentage for different cpu modes"),
743 744
	OPT_BOOLEAN_FLAG(0, "showcpuutilization", &symbol_conf.show_cpu_utilization,
		    "Show sample percentage for different cpu modes", PARSE_OPT_HIDDEN),
745 746
	OPT_STRING('p', "parent", &parent_pattern, "regex",
		   "regex filter to identify parent, see: '--sort parent'"),
747
	OPT_BOOLEAN('x', "exclude-other", &symbol_conf.exclude_other,
748
		    "Only display entries with parent-match"),
749
	OPT_CALLBACK_DEFAULT('g', "call-graph", &callchain_param,
750
			     "print_type,threshold[,print_limit],order,sort_key[,branch],value",
751 752
			     report_callchain_help, &report_parse_callchain_opt,
			     callchain_default_opt),
753 754
	OPT_BOOLEAN(0, "children", &symbol_conf.cumulate_callchain,
		    "Accumulate callchains of children and show total overhead as well"),
755 756 757
	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. "
758
		    "Default: kernel.perf_event_max_stack or " __stringify(PERF_MAX_STACK_DEPTH)),
759 760
	OPT_BOOLEAN('G', "inverted", &report.inverted_callchain,
		    "alias for inverted call graph"),
761 762 763
	OPT_CALLBACK(0, "ignore-callees", NULL, "regex",
		   "ignore callees of these functions in call graphs",
		   report_parse_ignore_callees_opt),
764
	OPT_STRING('d', "dsos", &symbol_conf.dso_list_str, "dso[,dso...]",
765
		   "only consider symbols in these dsos"),
766
	OPT_STRING('c', "comms", &symbol_conf.comm_list_str, "comm[,comm...]",
767
		   "only consider symbols in these comms"),
768 769 770 771
	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"),
772
	OPT_STRING('S', "symbols", &symbol_conf.sym_list_str, "symbol[,symbol...]",
773
		   "only consider these symbols"),
774 775
	OPT_STRING(0, "symbol-filter", &report.symbol_filter_str, "filter",
		   "only show symbols that (partially) match with this filter"),
776
	OPT_STRING('w', "column-widths", &symbol_conf.col_width_list_str,
777 778
		   "width[,width...]",
		   "don't try to adjust column width, use these fixed values"),
779
	OPT_STRING_NOEMPTY('t', "field-separator", &symbol_conf.field_sep, "separator",
780 781
		   "separator for columns, no spaces will be added between "
		   "columns '.' is reserved."),
782
	OPT_BOOLEAN('U', "hide-unresolved", &symbol_conf.hide_unresolved,
783
		    "Only display entries resolved to a symbol"),
784 785 786
	OPT_CALLBACK(0, "symfs", NULL, "directory",
		     "Look for files with symbols relative to this directory",
		     symbol__config_symfs),
787
	OPT_STRING('C', "cpu", &report.cpu_list, "cpu",
788 789
		   "list of cpus to profile"),
	OPT_BOOLEAN('I', "show-info", &report.show_full_info,
790
		    "Display extended information about perf.data file"),
791 792 793 794
	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)"),
795 796
	OPT_STRING('M', "disassembler-style", &disassembler_style, "disassembler style",
		   "Specify disassembler style (e.g. -M intel for intel syntax)"),
797 798
	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
799 800
	OPT_BOOLEAN(0, "group", &symbol_conf.event_group,
		    "Show event group information together"),
801
	OPT_CALLBACK_NOOPT('b', "branch-stack", &branch_mode, "",
802 803 804 805
		    "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"),
806 807
	OPT_STRING(0, "objdump", &objdump_path, "path",
		   "objdump binary to use for disassembly and annotations"),
808 809
	OPT_BOOLEAN(0, "demangle", &symbol_conf.demangle,
		    "Disable symbol demangling"),
810 811
	OPT_BOOLEAN(0, "demangle-kernel", &symbol_conf.demangle_kernel,
		    "Enable kernel symbol demangling"),
812
	OPT_BOOLEAN(0, "mem-mode", &report.mem_mode, "mem access profile"),
813 814
	OPT_CALLBACK(0, "percent-limit", &report, "percent",
		     "Don't show entries under that percent", parse_percent_limit),
815
	OPT_CALLBACK(0, "percentage", NULL, "relative|absolute",
N
Namhyung Kim 已提交
816
		     "how to display percentage of filtered entries", parse_filter_percentage),
817 818 819
	OPT_CALLBACK_OPTARG(0, "itrace", &itrace_synth_opts, NULL, "opts",
			    "Instruction Tracing options",
			    itrace_parse_synth_opts),
820 821
	OPT_BOOLEAN(0, "full-source-path", &srcline_full_filename,
			"Show full source file name path for source lines"),
822 823
	OPT_BOOLEAN(0, "show-ref-call-graph", &symbol_conf.show_ref_callgraph,
		    "Show callgraph from reference event"),
824 825
	OPT_INTEGER(0, "socket-filter", &report.socket_filter,
		    "only show processor socket that match with this filter"),
826 827
	OPT_BOOLEAN(0, "raw-trace", &symbol_conf.raw_trace,
		    "Show raw trace event output (do not use print fmt or plugins)"),
828 829
	OPT_BOOLEAN(0, "hierarchy", &symbol_conf.report_hierarchy,
		    "Show entries in a hierarchy"),
830 831 832
	OPT_CALLBACK_DEFAULT(0, "stdio-color", NULL, "mode",
			     "'always' (default), 'never' or 'auto' only applicable to --stdio mode",
			     stdio__config_color, "always"),
833
	OPT_END()
834
	};
835 836 837
	struct perf_data_file file = {
		.mode  = PERF_DATA_MODE_READ,
	};
838 839 840 841
	int ret = hists__init();

	if (ret < 0)
		return ret;
842

843
	perf_config(report__config, &report);
844

845
	argc = parse_options(argc, argv, options, report_usage, 0);
846 847 848 849 850 851 852 853 854 855
	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];
	}
856

857 858 859 860 861 862 863 864 865 866 867
	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;
	}

868
	if (report.use_stdio)
869
		use_browser = 0;
870
	else if (report.use_tui)
871
		use_browser = 1;
872 873
	else if (report.use_gtk)
		use_browser = 2;
874

875
	if (report.inverted_callchain)
876
		callchain_param.order = ORDER_CALLER;
877 878
	if (symbol_conf.cumulate_callchain && !callchain_param.order_set)
		callchain_param.order = ORDER_CALLER;
879

880 881 882 883
	if (itrace_synth_opts.callchain &&
	    (int)itrace_synth_opts.callchain_sz > report.max_stack)
		report.max_stack = itrace_synth_opts.callchain_sz;

884
	if (!input_name || !strlen(input_name)) {
885
		if (!fstat(STDIN_FILENO, &st) && S_ISFIFO(st.st_mode))
886
			input_name = "-";
887
		else
888
			input_name = "perf.data";
889
	}
890

891
	file.path  = input_name;
892
	file.force = symbol_conf.force;
893

894
repeat:
895
	session = perf_session__new(&file, false, &report.tool);
896
	if (session == NULL)
897
		return -1;
898

899 900 901 902 903
	if (report.queue_size) {
		ordered_events__set_alloc_size(&session->ordered_events,
					       report.queue_size);
	}

904 905
	session->itrace_synth_opts = &itrace_synth_opts;

906 907 908 909
	report.session = session;

	has_br_stack = perf_header__has_feat(&session->header,
					     HEADER_BRANCH_STACK);
910

911 912 913
	if (itrace_synth_opts.last_branch)
		has_br_stack = true;

914 915 916 917 918 919
	/*
	 * 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) &&
920
	    !branch_call_mode) {
921
		sort__mode = SORT_MODE__BRANCH;
922 923
		symbol_conf.cumulate_callchain = false;
	}
924
	if (branch_call_mode) {
925
		callchain_param.key = CCKEY_ADDRESS;
926 927 928 929 930 931
		callchain_param.branch_callstack = 1;
		symbol_conf.use_callchain = true;
		callchain_register_param(&callchain_param);
		if (sort_order == NULL)
			sort_order = "srcline,symbol,dso";
	}
932

933
	if (report.mem_mode) {
934
		if (sort__mode == SORT_MODE__BRANCH) {
935
			pr_err("branch and mem mode incompatible\n");
936 937
			goto error;
		}
938
		sort__mode = SORT_MODE__MEMORY;
939
		symbol_conf.cumulate_callchain = false;
940
	}
941

942 943 944 945 946 947 948 949 950 951 952
	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;
		}

953
		perf_hpp_list.need_collapse = true;
954 955
	}

956 957
	/* Force tty output for header output and per-thread stat. */
	if (report.header || report.header_only || report.show_threads)
958 959
		use_browser = 0;

960 961
	if (strcmp(input_name, "-") != 0)
		setup_browser(true);
962
	else
963 964
		use_browser = 0;

965 966 967 968 969 970 971 972 973
	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;
	}

974 975 976
	if (report.header || report.header_only) {
		perf_session__fprintf_info(session, stdout,
					   report.show_full_info);
977 978 979 980
		if (report.header_only) {
			ret = 0;
			goto error;
		}
981 982 983 984 985
	} else if (use_browser == 0) {
		fputs("# To display the perf.data header info, please use --header/--header-only options.\n#\n",
		      stdout);
	}

986
	/*
987
	 * Only in the TUI browser we are doing integrated annotation,
988 989 990
	 * so don't allocate extra space that won't be used in the stdio
	 * implementation.
	 */
991
	if (ui__has_annotation()) {
992 993 994
		ret = symbol__annotation_init();
		if (ret < 0)
			goto error;
995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010
		/*
 		 * For searching by name on the "Browse map details".
 		 * providing it only in verbose mode not to bloat too
 		 * much struct symbol.
 		 */
		if (verbose) {
			/*
			 * 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;
		}
	}
1011

1012
	if (symbol__init(&session->header.env) < 0)
1013
		goto error;
1014

1015
	sort__setup_elide(stdout);
1016

1017
	ret = __cmd_report(&report);
1018 1019 1020 1021 1022 1023
	if (ret == K_SWITCH_INPUT_DATA) {
		perf_session__delete(session);
		goto repeat;
	} else
		ret = 0;

1024 1025 1026
error:
	perf_session__delete(session);
	return ret;
1027
}