builtin-report.c 29.3 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
	return 0;
94 95
}

96 97 98 99 100 101 102 103 104 105 106 107 108 109
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;

110 111 112
	hist__account_cycles(iter->sample->branch_stack, al, iter->sample,
			     rep->nonany_branch_mode);

113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138
	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;
139 140
}

141
static int process_sample_event(struct perf_tool *tool,
142
				union perf_event *event,
143
				struct perf_sample *sample,
144
				struct perf_evsel *evsel,
145
				struct machine *machine)
146
{
147
	struct report *rep = container_of(tool, struct report, tool);
148
	struct addr_location al;
149
	struct hist_entry_iter iter = {
150 151
		.evsel 			= evsel,
		.sample 		= sample,
152
		.hide_unresolved 	= symbol_conf.hide_unresolved,
153
		.add_entry_cb 		= hist_iter__report_callback,
154
	};
155
	int ret = 0;
156

157
	if (machine__resolve(machine, &al, sample) < 0) {
158 159
		pr_debug("problem processing %d event, skipping it.\n",
			 event->header.type);
160 161
		return -1;
	}
162

163
	if (symbol_conf.hide_unresolved && al.sym == NULL)
164
		goto out_put;
165

166
	if (rep->cpu_list && !test_bit(sample->cpu, rep->cpu_bitmap))
167
		goto out_put;
168

169 170 171 172 173 174 175
	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;
176
		iter.ops = &hist_iter_branch;
177
	} else if (rep->mem_mode) {
178
		iter.ops = &hist_iter_mem;
179
	} else if (symbol_conf.cumulate_callchain) {
180
		iter.ops = &hist_iter_cumulative;
181
	} else {
182
		iter.ops = &hist_iter_normal;
183
	}
184 185 186 187

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

188
	ret = hist_entry_iter__add(&iter, &al, rep->max_stack, rep);
189 190
	if (ret < 0)
		pr_debug("problem adding hist entry, skipping event\n");
191 192
out_put:
	addr_location__put(&al);
193
	return ret;
194
}
I
Ingo Molnar 已提交
195

196
static int process_read_event(struct perf_tool *tool,
197
			      union perf_event *event,
198
			      struct perf_sample *sample __maybe_unused,
199
			      struct perf_evsel *evsel,
200
			      struct machine *machine __maybe_unused)
201
{
202
	struct report *rep = container_of(tool, struct report, tool);
203

204
	if (rep->show_threads) {
205
		const char *name = evsel ? perf_evsel__name(evsel) : "unknown";
206
		perf_read_values_add_value(&rep->show_threads_values,
207 208 209 210 211 212
					   event->read.pid, event->read.tid,
					   event->read.id,
					   name,
					   event->read.value);
	}

213
	dump_printf(": %d %d %s %" PRIu64 "\n", event->read.pid, event->read.tid,
214
		    evsel ? perf_evsel__name(evsel) : "FAIL",
215
		    event->read.value);
216 217 218 219

	return 0;
}

220
/* For pipe mode, sample_type is not currently set */
221
static int report__setup_sample_type(struct report *rep)
222
{
223 224 225
	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);
226

227 228 229 230 231 232
	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;

233 234 235
	if (session->itrace_synth_opts->last_branch)
		sample_type |= PERF_SAMPLE_BRANCH_STACK;

236
	if (!is_pipe && !(sample_type & PERF_SAMPLE_CALLCHAIN)) {
237
		if (perf_hpp_list.parent) {
238
			ui__error("Selected --sort parent, but no "
239 240
				    "callchain data. Did you call "
				    "'perf record' without -g?\n");
241
			return -EINVAL;
242
		}
243
		if (symbol_conf.use_callchain) {
244 245 246
			ui__error("Selected -g or --branch-history but no "
				  "callchain data. Did\n"
				  "you call 'perf record' without -g?\n");
247
			return -1;
248
		}
249
	} else if (!callchain_param.enabled &&
250
		   callchain_param.mode != CHAIN_NONE &&
251
		   !symbol_conf.use_callchain) {
252
			symbol_conf.use_callchain = true;
253
			if (callchain_register_param(&callchain_param) < 0) {
254
				ui__error("Can't register callchain params.\n");
255
				return -EINVAL;
256
			}
257 258
	}

259 260 261 262 263 264 265 266
	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();
		}
	}

267
	if (sort__mode == SORT_MODE__BRANCH) {
268
		if (!is_pipe &&
269
		    !(sample_type & PERF_SAMPLE_BRANCH_STACK)) {
270 271
			ui__error("Selected -b but no branch data. "
				  "Did you call perf record without -b?\n");
272 273 274 275
			return -1;
		}
	}

276 277 278 279
	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;
280 281
		else if (sample_type & PERF_SAMPLE_BRANCH_STACK)
			callchain_param.record_mode = CALLCHAIN_LBR;
282 283 284
		else
			callchain_param.record_mode = CALLCHAIN_FP;
	}
285 286 287 288 289 290

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

291 292
	return 0;
}
293

294
static void sig_handler(int sig __maybe_unused)
295 296 297 298
{
	session_done = 1;
}

299
static size_t hists__fprintf_nr_sample_events(struct hists *hists, struct report *rep,
300 301 302 303
					      const char *evname, FILE *fp)
{
	size_t ret;
	char unit;
304 305 306
	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);
307 308
	char buf[512];
	size_t size = sizeof(buf);
309
	int socked_id = hists->socket_filter;
310

311 312 313 314 315
	if (symbol_conf.filter_relative) {
		nr_samples = hists->stats.nr_non_filtered_samples;
		nr_events = hists->stats.total_non_filtered_period;
	}

316
	if (perf_evsel__is_group_event(evsel)) {
317 318 319 320 321 322
		struct perf_evsel *pos;

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

		for_each_group_member(pos, evsel) {
323 324
			const struct hists *pos_hists = evsel__hists(pos);

325
			if (symbol_conf.filter_relative) {
326 327
				nr_samples += pos_hists->stats.nr_non_filtered_samples;
				nr_events += pos_hists->stats.total_non_filtered_period;
328
			} else {
329 330
				nr_samples += pos_hists->stats.nr_events[PERF_RECORD_SAMPLE];
				nr_events += pos_hists->stats.total_period;
331
			}
332 333
		}
	}
334

335 336
	nr_samples = convert_unit(nr_samples, &unit);
	ret = fprintf(fp, "# Samples: %lu%c", nr_samples, unit);
337
	if (evname != NULL)
338 339
		ret += fprintf(fp, " of event '%s'", evname);

340 341 342 343 344
	if (symbol_conf.show_ref_callgraph &&
	    strstr(evname, "call-graph=no")) {
		ret += fprintf(fp, ", show reference callgraph");
	}

345 346
	if (rep->mem_mode) {
		ret += fprintf(fp, "\n# Total weight : %" PRIu64, nr_events);
347
		ret += fprintf(fp, "\n# Sort order   : %s", sort_order ? : default_mem_sort_order);
348 349
	} else
		ret += fprintf(fp, "\n# Event count (approx.): %" PRIu64, nr_events);
350

351 352
	if (socked_id > -1)
		ret += fprintf(fp, "\n# Processor Socket: %d", socked_id);
353

354 355 356
	return ret + fprintf(fp, "\n#\n");
}

357
static int perf_evlist__tty_browse_hists(struct perf_evlist *evlist,
358
					 struct report *rep,
359
					 const char *help)
360
{
361
	struct perf_evsel *pos;
362

363
	fprintf(stdout, "#\n# Total Lost Samples: %" PRIu64 "\n#\n", evlist->stats.total_lost_samples);
364
	evlist__for_each_entry(evlist, pos) {
365
		struct hists *hists = evsel__hists(pos);
366
		const char *evname = perf_evsel__name(pos);
367

368 369 370 371
		if (symbol_conf.event_group &&
		    !perf_evsel__is_group_leader(pos))
			continue;

372
		hists__fprintf_nr_sample_events(hists, rep, evname, stdout);
373 374
		hists__fprintf(hists, true, 0, 0, rep->min_percent, stdout,
			       symbol_conf.use_callchain);
375 376 377
		fprintf(stdout, "\n\n");
	}

378
	if (sort_order == NULL &&
379
	    parent_pattern == default_parent_pattern)
380 381
		fprintf(stdout, "#\n# (%s)\n#\n", help);

382 383 384 385 386
	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);
387 388 389 390 391
	}

	return 0;
}

392 393
static void report__warn_kptr_restrict(const struct report *rep)
{
394
	struct map *kernel_map = machine__kernel_map(&rep->session->machines.host);
395
	struct kmap *kernel_kmap = kernel_map ? map__kmap(kernel_map) : NULL;
396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420

	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);
	}
}

421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440
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;
441 442 443 444 445 446 447 448
	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!";
	}
449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472

	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;
}

473
static int report__collapse_hists(struct report *rep)
474 475 476
{
	struct ui_progress prog;
	struct perf_evsel *pos;
477
	int ret = 0;
478

479
	ui_progress__init(&prog, rep->nr_entries, "Merging related events...");
480

481
	evlist__for_each_entry(rep->session->evlist, pos) {
482
		struct hists *hists = evsel__hists(pos);
483 484 485 486

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

487 488
		hists->socket_filter = rep->socket_filter;

489 490 491
		ret = hists__collapse_resort(hists, &prog);
		if (ret < 0)
			break;
492 493 494 495

		/* Non-group events are considered as leader */
		if (symbol_conf.event_group &&
		    !perf_evsel__is_group_leader(pos)) {
496
			struct hists *leader_hists = evsel__hists(pos->leader);
497 498 499 500 501 502 503

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

	ui_progress__finish();
504
	return ret;
505 506
}

507 508 509 510 511 512 513
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...");

514
	evlist__for_each_entry(rep->session->evlist, pos)
515
		perf_evsel__output_resort(pos, &prog);
516 517 518 519

	ui_progress__finish();
}

520
static int __cmd_report(struct report *rep)
521
{
522
	int ret;
523
	struct perf_session *session = rep->session;
524
	struct perf_evsel *pos;
525
	struct perf_data_file *file = session->file;
526

527 528
	signal(SIGINT, sig_handler);

529 530 531
	if (rep->cpu_list) {
		ret = perf_session__cpu_bitmap(session, rep->cpu_list,
					       rep->cpu_bitmap);
532 533
		if (ret) {
			ui__error("failed to set cpu bitmap\n");
534
			return ret;
535
		}
536 537
	}

538 539
	if (rep->show_threads)
		perf_read_values_init(&rep->show_threads_values);
540

541
	ret = report__setup_sample_type(rep);
542 543
	if (ret) {
		/* report__setup_sample_type() already showed error message */
544
		return ret;
545
	}
546

547
	ret = perf_session__process_events(session);
548 549
	if (ret) {
		ui__error("failed to process sample\n");
550
		return ret;
551
	}
552

553
	report__warn_kptr_restrict(rep);
554

555
	evlist__for_each_entry(session->evlist, pos)
556 557
		rep->nr_entries += evsel__hists(pos)->nr_entries;

558 559 560
	if (use_browser == 0) {
		if (verbose > 3)
			perf_session__fprintf(session, stdout);
561

562 563
		if (verbose > 2)
			perf_session__fprintf_dsos(session, stdout);
564

565 566
		if (dump_trace) {
			perf_session__fprintf_nr_events(session, stdout);
567
			perf_evlist__fprintf_nr_events(session->evlist, stdout);
568 569
			return 0;
		}
570 571
	}

572 573 574 575 576
	ret = report__collapse_hists(rep);
	if (ret) {
		ui__error("failed to process hist entry\n");
		return ret;
	}
577

578 579 580
	if (session_done())
		return 0;

581 582 583 584 585
	/*
	 * recalculate number of entries after collapsing since it
	 * might be changed during the collapse phase.
	 */
	rep->nr_entries = 0;
586
	evlist__for_each_entry(session->evlist, pos)
587 588
		rep->nr_entries += evsel__hists(pos)->nr_entries;

589
	if (rep->nr_entries == 0) {
590
		ui__error("The %s file has no samples!\n", file->path);
591
		return 0;
592 593
	}

594
	report__output_resort(rep);
595

596
	return report__browse_hists(rep);
597 598
}

599
static int
600
report_parse_callchain_opt(const struct option *opt, const char *arg, int unset)
601
{
602
	struct callchain_param *callchain = opt->value;
603

604
	callchain->enabled = !unset;
605 606 607 608
	/*
	 * --no-call-graph
	 */
	if (unset) {
609 610
		symbol_conf.use_callchain = false;
		callchain->mode = CHAIN_NONE;
611 612 613
		return 0;
	}

614
	return parse_callchain_report_opt(arg);
615 616
}

617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634
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;
}

635
static int
636 637
parse_branch_mode(const struct option *opt __maybe_unused,
		  const char *str __maybe_unused, int unset)
638
{
639 640 641
	int *branch_mode = opt->value;

	*branch_mode = !unset;
642 643 644
	return 0;
}

645 646 647 648
static int
parse_percent_limit(const struct option *opt, const char *str,
		    int unset __maybe_unused)
{
649
	struct report *rep = opt->value;
650
	double pcnt = strtof(str, NULL);
651

652 653
	rep->min_percent = pcnt;
	callchain_param.min_percent = pcnt;
654 655 656
	return 0;
}

657
#define CALLCHAIN_DEFAULT_OPT  "graph,0.5,caller,function,percent"
658 659 660 661

const char report_callchain_help[] = "Display call graph (stack chain/backtrace):\n\n"
				     CALLCHAIN_REPORT_HELP
				     "\n\t\t\t\tDefault: " CALLCHAIN_DEFAULT_OPT;
662

663
int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused)
664
{
665
	struct perf_session *session;
666
	struct itrace_synth_opts itrace_synth_opts = { .set = 0, };
667
	struct stat st;
668
	bool has_br_stack = false;
669
	int branch_mode = -1;
670
	bool branch_call_mode = false;
671
	char callchain_default_opt[] = CALLCHAIN_DEFAULT_OPT;
672
	const char * const report_usage[] = {
N
Namhyung Kim 已提交
673
		"perf report [<options>]",
674 675
		NULL
	};
676
	struct report report = {
677
		.tool = {
678 679
			.sample		 = process_sample_event,
			.mmap		 = perf_event__process_mmap,
680
			.mmap2		 = perf_event__process_mmap2,
681
			.comm		 = perf_event__process_comm,
682 683
			.exit		 = perf_event__process_exit,
			.fork		 = perf_event__process_fork,
684 685 686 687 688
			.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,
689 690 691
			.id_index	 = perf_event__process_id_index,
			.auxtrace_info	 = perf_event__process_auxtrace_info,
			.auxtrace	 = perf_event__process_auxtrace,
692
			.ordered_events	 = true,
693 694
			.ordering_requires_timestamps = true,
		},
695
		.max_stack		 = PERF_MAX_STACK_DEPTH,
696
		.pretty_printing_style	 = "normal",
697
		.socket_filter		 = -1,
698 699
	};
	const struct option options[] = {
700
	OPT_STRING('i', "input", &input_name, "file",
701
		    "input file name"),
702
	OPT_INCR('v', "verbose", &verbose,
703
		    "be more verbose (show symbol address, etc)"),
704 705
	OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
		    "dump raw trace in ASCII"),
706 707
	OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name,
		   "file", "vmlinux pathname"),
708 709
	OPT_STRING(0, "kallsyms", &symbol_conf.kallsyms_name,
		   "file", "kallsyms pathname"),
710
	OPT_BOOLEAN('f', "force", &symbol_conf.force, "don't complain, do it"),
711
	OPT_BOOLEAN('m', "modules", &symbol_conf.use_modules,
712
		    "load module symbols - WARNING: use only with -k and LIVE kernel"),
713
	OPT_BOOLEAN('n', "show-nr-samples", &symbol_conf.show_nr_samples,
714
		    "Show a column with the number of samples"),
715
	OPT_BOOLEAN('T', "threads", &report.show_threads,
716
		    "Show per-thread event counters"),
717
	OPT_STRING(0, "pretty", &report.pretty_printing_style, "key",
718
		   "pretty printing style key: normal raw"),
719
	OPT_BOOLEAN(0, "tui", &report.use_tui, "Use the TUI interface"),
720
	OPT_BOOLEAN(0, "gtk", &report.use_gtk, "Use the GTK2 interface"),
721 722
	OPT_BOOLEAN(0, "stdio", &report.use_stdio,
		    "Use the stdio interface"),
723 724 725
	OPT_BOOLEAN(0, "header", &report.header, "Show data header."),
	OPT_BOOLEAN(0, "header-only", &report.header_only,
		    "Show only data header."),
726
	OPT_STRING('s', "sort", &sort_order, "key[,key2...]",
727 728
		   "sort by key(s): pid, comm, dso, symbol, parent, cpu, srcline, ..."
		   " Please refer the man page for the complete list."),
729 730
	OPT_STRING('F', "fields", &field_order, "key[,keys...]",
		   "output field(s): overhead, period, sample plus all of sort keys"),
731
	OPT_BOOLEAN(0, "show-cpu-utilization", &symbol_conf.show_cpu_utilization,
732
		    "Show sample percentage for different cpu modes"),
733 734
	OPT_BOOLEAN_FLAG(0, "showcpuutilization", &symbol_conf.show_cpu_utilization,
		    "Show sample percentage for different cpu modes", PARSE_OPT_HIDDEN),
735 736
	OPT_STRING('p', "parent", &parent_pattern, "regex",
		   "regex filter to identify parent, see: '--sort parent'"),
737
	OPT_BOOLEAN('x', "exclude-other", &symbol_conf.exclude_other,
738
		    "Only display entries with parent-match"),
739
	OPT_CALLBACK_DEFAULT('g', "call-graph", &callchain_param,
740
			     "print_type,threshold[,print_limit],order,sort_key[,branch],value",
741 742
			     report_callchain_help, &report_parse_callchain_opt,
			     callchain_default_opt),
743 744
	OPT_BOOLEAN(0, "children", &symbol_conf.cumulate_callchain,
		    "Accumulate callchains of children and show total overhead as well"),
745 746 747
	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. "
748
		    "Default: kernel.perf_event_max_stack or " __stringify(PERF_MAX_STACK_DEPTH)),
749 750
	OPT_BOOLEAN('G', "inverted", &report.inverted_callchain,
		    "alias for inverted call graph"),
751 752 753
	OPT_CALLBACK(0, "ignore-callees", NULL, "regex",
		   "ignore callees of these functions in call graphs",
		   report_parse_ignore_callees_opt),
754
	OPT_STRING('d', "dsos", &symbol_conf.dso_list_str, "dso[,dso...]",
755
		   "only consider symbols in these dsos"),
756
	OPT_STRING('c', "comms", &symbol_conf.comm_list_str, "comm[,comm...]",
757
		   "only consider symbols in these comms"),
758 759 760 761
	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"),
762
	OPT_STRING('S', "symbols", &symbol_conf.sym_list_str, "symbol[,symbol...]",
763
		   "only consider these symbols"),
764 765
	OPT_STRING(0, "symbol-filter", &report.symbol_filter_str, "filter",
		   "only show symbols that (partially) match with this filter"),
766
	OPT_STRING('w', "column-widths", &symbol_conf.col_width_list_str,
767 768
		   "width[,width...]",
		   "don't try to adjust column width, use these fixed values"),
769
	OPT_STRING_NOEMPTY('t', "field-separator", &symbol_conf.field_sep, "separator",
770 771
		   "separator for columns, no spaces will be added between "
		   "columns '.' is reserved."),
772
	OPT_BOOLEAN('U', "hide-unresolved", &symbol_conf.hide_unresolved,
773
		    "Only display entries resolved to a symbol"),
774 775 776
	OPT_CALLBACK(0, "symfs", NULL, "directory",
		     "Look for files with symbols relative to this directory",
		     symbol__config_symfs),
777
	OPT_STRING('C', "cpu", &report.cpu_list, "cpu",
778 779
		   "list of cpus to profile"),
	OPT_BOOLEAN('I', "show-info", &report.show_full_info,
780
		    "Display extended information about perf.data file"),
781 782 783 784
	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)"),
785 786
	OPT_STRING('M', "disassembler-style", &disassembler_style, "disassembler style",
		   "Specify disassembler style (e.g. -M intel for intel syntax)"),
787 788
	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
789 790
	OPT_BOOLEAN(0, "group", &symbol_conf.event_group,
		    "Show event group information together"),
791
	OPT_CALLBACK_NOOPT('b', "branch-stack", &branch_mode, "",
792 793 794 795
		    "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"),
796 797
	OPT_STRING(0, "objdump", &objdump_path, "path",
		   "objdump binary to use for disassembly and annotations"),
798 799
	OPT_BOOLEAN(0, "demangle", &symbol_conf.demangle,
		    "Disable symbol demangling"),
800 801
	OPT_BOOLEAN(0, "demangle-kernel", &symbol_conf.demangle_kernel,
		    "Enable kernel symbol demangling"),
802
	OPT_BOOLEAN(0, "mem-mode", &report.mem_mode, "mem access profile"),
803 804
	OPT_CALLBACK(0, "percent-limit", &report, "percent",
		     "Don't show entries under that percent", parse_percent_limit),
805
	OPT_CALLBACK(0, "percentage", NULL, "relative|absolute",
N
Namhyung Kim 已提交
806
		     "how to display percentage of filtered entries", parse_filter_percentage),
807 808 809
	OPT_CALLBACK_OPTARG(0, "itrace", &itrace_synth_opts, NULL, "opts",
			    "Instruction Tracing options",
			    itrace_parse_synth_opts),
810 811
	OPT_BOOLEAN(0, "full-source-path", &srcline_full_filename,
			"Show full source file name path for source lines"),
812 813
	OPT_BOOLEAN(0, "show-ref-call-graph", &symbol_conf.show_ref_callgraph,
		    "Show callgraph from reference event"),
814 815
	OPT_INTEGER(0, "socket-filter", &report.socket_filter,
		    "only show processor socket that match with this filter"),
816 817
	OPT_BOOLEAN(0, "raw-trace", &symbol_conf.raw_trace,
		    "Show raw trace event output (do not use print fmt or plugins)"),
818 819
	OPT_BOOLEAN(0, "hierarchy", &symbol_conf.report_hierarchy,
		    "Show entries in a hierarchy"),
820
	OPT_END()
821
	};
822 823 824
	struct perf_data_file file = {
		.mode  = PERF_DATA_MODE_READ,
	};
825 826 827 828
	int ret = hists__init();

	if (ret < 0)
		return ret;
829

830
	perf_config(report__config, &report);
831

832
	argc = parse_options(argc, argv, options, report_usage, 0);
833 834 835 836 837 838 839 840 841 842
	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];
	}
843

844 845 846 847 848 849 850 851 852 853 854
	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;
	}

855
	if (report.use_stdio)
856
		use_browser = 0;
857
	else if (report.use_tui)
858
		use_browser = 1;
859 860
	else if (report.use_gtk)
		use_browser = 2;
861

862
	if (report.inverted_callchain)
863
		callchain_param.order = ORDER_CALLER;
864 865
	if (symbol_conf.cumulate_callchain && !callchain_param.order_set)
		callchain_param.order = ORDER_CALLER;
866

867 868 869 870
	if (itrace_synth_opts.callchain &&
	    (int)itrace_synth_opts.callchain_sz > report.max_stack)
		report.max_stack = itrace_synth_opts.callchain_sz;

871
	if (!input_name || !strlen(input_name)) {
872
		if (!fstat(STDIN_FILENO, &st) && S_ISFIFO(st.st_mode))
873
			input_name = "-";
874
		else
875
			input_name = "perf.data";
876
	}
877

878
	file.path  = input_name;
879
	file.force = symbol_conf.force;
880

881
repeat:
882
	session = perf_session__new(&file, false, &report.tool);
883
	if (session == NULL)
884
		return -1;
885

886 887 888 889 890
	if (report.queue_size) {
		ordered_events__set_alloc_size(&session->ordered_events,
					       report.queue_size);
	}

891 892
	session->itrace_synth_opts = &itrace_synth_opts;

893 894 895 896
	report.session = session;

	has_br_stack = perf_header__has_feat(&session->header,
					     HEADER_BRANCH_STACK);
897

898 899 900
	if (itrace_synth_opts.last_branch)
		has_br_stack = true;

901 902 903 904 905 906
	/*
	 * 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) &&
907
	    !branch_call_mode) {
908
		sort__mode = SORT_MODE__BRANCH;
909 910
		symbol_conf.cumulate_callchain = false;
	}
911
	if (branch_call_mode) {
912
		callchain_param.key = CCKEY_ADDRESS;
913 914 915 916 917 918
		callchain_param.branch_callstack = 1;
		symbol_conf.use_callchain = true;
		callchain_register_param(&callchain_param);
		if (sort_order == NULL)
			sort_order = "srcline,symbol,dso";
	}
919

920
	if (report.mem_mode) {
921
		if (sort__mode == SORT_MODE__BRANCH) {
922
			pr_err("branch and mem mode incompatible\n");
923 924
			goto error;
		}
925
		sort__mode = SORT_MODE__MEMORY;
926
		symbol_conf.cumulate_callchain = false;
927
	}
928

929 930 931 932 933 934 935 936 937 938 939 940
	if (symbol_conf.report_hierarchy) {
		/* disable incompatible options */
		symbol_conf.event_group = false;
		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;
		}

941
		perf_hpp_list.need_collapse = true;
942 943
	}

944 945
	/* Force tty output for header output and per-thread stat. */
	if (report.header || report.header_only || report.show_threads)
946 947
		use_browser = 0;

948 949
	if (strcmp(input_name, "-") != 0)
		setup_browser(true);
950
	else
951 952
		use_browser = 0;

953 954 955 956 957 958 959 960 961
	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;
	}

962 963 964
	if (report.header || report.header_only) {
		perf_session__fprintf_info(session, stdout,
					   report.show_full_info);
965 966 967 968
		if (report.header_only) {
			ret = 0;
			goto error;
		}
969 970 971 972 973
	} else if (use_browser == 0) {
		fputs("# To display the perf.data header info, please use --header/--header-only options.\n#\n",
		      stdout);
	}

974
	/*
975
	 * Only in the TUI browser we are doing integrated annotation,
976 977 978
	 * so don't allocate extra space that won't be used in the stdio
	 * implementation.
	 */
979
	if (ui__has_annotation()) {
980
		symbol_conf.priv_size = sizeof(struct annotation);
981 982
		machines__set_symbol_filter(&session->machines,
					    symbol__annotate_init);
983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998
		/*
 		 * 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;
		}
	}
999

1000
	if (symbol__init(&session->header.env) < 0)
1001
		goto error;
1002

1003
	sort__setup_elide(stdout);
1004

1005
	ret = __cmd_report(&report);
1006 1007 1008 1009 1010 1011
	if (ret == K_SWITCH_INPUT_DATA) {
		perf_session__delete(session);
		goto repeat;
	} else
		ret = 0;

1012 1013 1014
error:
	perf_session__delete(session);
	return ret;
1015
}