builtin-report.c 26.3 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/cache.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 31 32

#include "util/parse-options.h"
#include "util/parse-events.h"

33
#include "util/thread.h"
34
#include "util/sort.h"
35
#include "util/hist.h"
36
#include "arch/common.h"
37

38 39
#include <linux/bitmap.h>

40
struct perf_report {
41
	struct perf_tool	tool;
42
	struct perf_session	*session;
43
	bool			force, use_tui, use_gtk, use_stdio;
44 45 46 47 48
	bool			hide_unresolved;
	bool			dont_use_callchains;
	bool			show_full_info;
	bool			show_threads;
	bool			inverted_callchain;
49
	bool			mem_mode;
50 51 52 53
	struct perf_read_values	show_threads_values;
	const char		*pretty_printing_style;
	symbol_filter_t		annotate_init;
	const char		*cpu_list;
54
	const char		*symbol_filter_str;
55
	float			min_percent;
56
	DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS);
57
};
58

59 60 61 62 63 64
static int perf_report_config(const char *var, const char *value, void *cb)
{
	if (!strcmp(var, "report.group")) {
		symbol_conf.event_group = perf_config_bool(var, value);
		return 0;
	}
65 66 67 68 69
	if (!strcmp(var, "report.percent-limit")) {
		struct perf_report *rep = cb;
		rep->min_percent = strtof(value, NULL);
		return 0;
	}
70 71 72 73

	return perf_default_config(var, value, cb);
}

74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119
static int perf_report__add_mem_hist_entry(struct perf_tool *tool,
					   struct addr_location *al,
					   struct perf_sample *sample,
					   struct perf_evsel *evsel,
					   struct machine *machine,
					   union perf_event *event)
{
	struct perf_report *rep = container_of(tool, struct perf_report, tool);
	struct symbol *parent = NULL;
	u8 cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK;
	int err = 0;
	struct hist_entry *he;
	struct mem_info *mi, *mx;
	uint64_t cost;

	if ((sort__has_parent || symbol_conf.use_callchain) &&
	    sample->callchain) {
		err = machine__resolve_callchain(machine, evsel, al->thread,
						 sample, &parent);
		if (err)
			return err;
	}

	mi = machine__resolve_mem(machine, al->thread, sample, cpumode);
	if (!mi)
		return -ENOMEM;

	if (rep->hide_unresolved && !al->sym)
		return 0;

	cost = sample->weight;
	if (!cost)
		cost = 1;

	/*
	 * must pass period=weight in order to get the correct
	 * sorting from hists__collapse_resort() which is solely
	 * based on periods. We want sorting be done on nr_events * weight
	 * and this is indirectly achieved by passing period=weight here
	 * and the he_stat__add_period() function.
	 */
	he = __hists__add_mem_entry(&evsel->hists, al, parent, mi, cost, cost);
	if (!he)
		return -ENOMEM;

	/*
120
	 * In the TUI browser, we are doing integrated annotation,
121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166
	 * so we don't allocate the extra space needed because the stdio
	 * code will not use it.
	 */
	if (sort__has_sym && he->ms.sym && use_browser > 0) {
		struct annotation *notes = symbol__annotation(he->ms.sym);

		assert(evsel != NULL);

		if (notes->src == NULL && symbol__alloc_hist(he->ms.sym) < 0)
			goto out;

		err = hist_entry__inc_addr_samples(he, evsel->idx, al->addr);
		if (err)
			goto out;
	}

	if (sort__has_sym && he->mem_info->daddr.sym && use_browser > 0) {
		struct annotation *notes;

		mx = he->mem_info;

		notes = symbol__annotation(mx->daddr.sym);
		if (notes->src == NULL && symbol__alloc_hist(mx->daddr.sym) < 0)
			goto out;

		err = symbol__inc_addr_samples(mx->daddr.sym,
					       mx->daddr.map,
					       evsel->idx,
					       mx->daddr.al_addr);
		if (err)
			goto out;
	}

	evsel->hists.stats.total_period += cost;
	hists__inc_nr_events(&evsel->hists, PERF_RECORD_SAMPLE);
	err = 0;

	if (symbol_conf.use_callchain) {
		err = callchain_append(he->callchain,
				       &callchain_cursor,
				       sample->period);
	}
out:
	return err;
}

167 168 169 170 171 172 173 174 175 176 177
static int perf_report__add_branch_hist_entry(struct perf_tool *tool,
					struct addr_location *al,
					struct perf_sample *sample,
					struct perf_evsel *evsel,
				      struct machine *machine)
{
	struct perf_report *rep = container_of(tool, struct perf_report, tool);
	struct symbol *parent = NULL;
	int err = 0;
	unsigned i;
	struct hist_entry *he;
178
	struct branch_info *bi, *bx;
179 180 181

	if ((sort__has_parent || symbol_conf.use_callchain)
	    && sample->callchain) {
182 183
		err = machine__resolve_callchain(machine, evsel, al->thread,
						 sample, &parent);
184 185 186 187 188 189 190 191 192 193 194 195
		if (err)
			return err;
	}

	bi = machine__resolve_bstack(machine, al->thread,
				     sample->branch_stack);
	if (!bi)
		return -ENOMEM;

	for (i = 0; i < sample->branch_stack->nr; i++) {
		if (rep->hide_unresolved && !(bi[i].from.sym && bi[i].to.sym))
			continue;
196 197 198

		err = -ENOMEM;

199 200 201 202 203
		/*
		 * The report shows the percentage of total branches captured
		 * and not events sampled. Thus we use a pseudo period of 1.
		 */
		he = __hists__add_branch_entry(&evsel->hists, al, parent,
204
				&bi[i], 1, 1);
205
		if (he) {
206 207
			struct annotation *notes;
			bx = he->branch_info;
208
			if (bx->from.sym && use_browser == 1 && sort__has_sym) {
209 210 211 212 213 214 215 216 217 218 219 220 221
				notes = symbol__annotation(bx->from.sym);
				if (!notes->src
				    && symbol__alloc_hist(bx->from.sym) < 0)
					goto out;

				err = symbol__inc_addr_samples(bx->from.sym,
							       bx->from.map,
							       evsel->idx,
							       bx->from.al_addr);
				if (err)
					goto out;
			}

222
			if (bx->to.sym && use_browser == 1 && sort__has_sym) {
223 224 225 226 227 228 229 230 231 232 233 234
				notes = symbol__annotation(bx->to.sym);
				if (!notes->src
				    && symbol__alloc_hist(bx->to.sym) < 0)
					goto out;

				err = symbol__inc_addr_samples(bx->to.sym,
							       bx->to.map,
							       evsel->idx,
							       bx->to.al_addr);
				if (err)
					goto out;
			}
235 236 237
			evsel->hists.stats.total_period += 1;
			hists__inc_nr_events(&evsel->hists, PERF_RECORD_SAMPLE);
		} else
238
			goto out;
239
	}
240
	err = 0;
241
out:
242
	free(bi);
243 244 245
	return err;
}

246 247 248 249
static int perf_evsel__add_hist_entry(struct perf_evsel *evsel,
				      struct addr_location *al,
				      struct perf_sample *sample,
				      struct machine *machine)
250
{
251
	struct symbol *parent = NULL;
252
	int err = 0;
253 254
	struct hist_entry *he;

255
	if ((sort__has_parent || symbol_conf.use_callchain) && sample->callchain) {
256 257
		err = machine__resolve_callchain(machine, evsel, al->thread,
						 sample, &parent);
258 259
		if (err)
			return err;
260
	}
261

262 263
	he = __hists__add_entry(&evsel->hists, al, parent, sample->period,
					sample->weight);
264
	if (he == NULL)
265 266
		return -ENOMEM;

267
	if (symbol_conf.use_callchain) {
268
		err = callchain_append(he->callchain,
269
				       &callchain_cursor,
270
				       sample->period);
271
		if (err)
272
			return err;
273 274
	}
	/*
275
	 * Only in the TUI browser we are doing integrated annotation,
276 277 278
	 * so we don't allocated the extra space needed because the stdio
	 * code will not use it.
	 */
279
	if (he->ms.sym != NULL && use_browser == 1 && sort__has_sym) {
280
		struct annotation *notes = symbol__annotation(he->ms.sym);
281 282 283 284

		assert(evsel != NULL);

		err = -ENOMEM;
285
		if (notes->src == NULL && symbol__alloc_hist(he->ms.sym) < 0)
286 287 288
			goto out;

		err = hist_entry__inc_addr_samples(he, evsel->idx, al->addr);
289
	}
290

291 292 293
	evsel->hists.stats.total_period += sample->period;
	hists__inc_nr_events(&evsel->hists, PERF_RECORD_SAMPLE);
out:
294
	return err;
295 296
}

297

298
static int process_sample_event(struct perf_tool *tool,
299
				union perf_event *event,
300
				struct perf_sample *sample,
301
				struct perf_evsel *evsel,
302
				struct machine *machine)
303
{
304
	struct perf_report *rep = container_of(tool, struct perf_report, tool);
305
	struct addr_location al;
306
	int ret;
307

308
	if (perf_event__preprocess_sample(event, machine, &al, sample,
309
					  rep->annotate_init) < 0) {
310
		fprintf(stderr, "problem processing %d event, skipping it.\n",
311 312 313
			event->header.type);
		return -1;
	}
314

315
	if (al.filtered || (rep->hide_unresolved && al.sym == NULL))
316
		return 0;
317

318
	if (rep->cpu_list && !test_bit(sample->cpu, rep->cpu_bitmap))
319 320
		return 0;

321
	if (sort__mode == SORT_MODE__BRANCH) {
322 323 324
		ret = perf_report__add_branch_hist_entry(tool, &al, sample,
							 evsel, machine);
		if (ret < 0)
325
			pr_debug("problem adding lbr entry, skipping event\n");
326
	} else if (rep->mem_mode == 1) {
327 328 329
		ret = perf_report__add_mem_hist_entry(tool, &al, sample,
						      evsel, machine, event);
		if (ret < 0)
330
			pr_debug("problem adding mem entry, skipping event\n");
331 332 333
	} else {
		if (al.map != NULL)
			al.map->dso->hit = 1;
334

335 336
		ret = perf_evsel__add_hist_entry(evsel, &al, sample, machine);
		if (ret < 0)
337
			pr_debug("problem incrementing symbol period, skipping event\n");
338
	}
339
	return ret;
340
}
I
Ingo Molnar 已提交
341

342
static int process_read_event(struct perf_tool *tool,
343
			      union perf_event *event,
344
			      struct perf_sample *sample __maybe_unused,
345
			      struct perf_evsel *evsel,
346
			      struct machine *machine __maybe_unused)
347
{
348
	struct perf_report *rep = container_of(tool, struct perf_report, tool);
349

350
	if (rep->show_threads) {
351
		const char *name = evsel ? perf_evsel__name(evsel) : "unknown";
352
		perf_read_values_add_value(&rep->show_threads_values,
353 354 355 356 357 358
					   event->read.pid, event->read.tid,
					   event->read.id,
					   name,
					   event->read.value);
	}

359
	dump_printf(": %d %d %s %" PRIu64 "\n", event->read.pid, event->read.tid,
360
		    evsel ? perf_evsel__name(evsel) : "FAIL",
361
		    event->read.value);
362 363 364 365

	return 0;
}

366
/* For pipe mode, sample_type is not currently set */
367
static int perf_report__setup_sample_type(struct perf_report *rep)
368
{
369
	struct perf_session *self = rep->session;
370
	u64 sample_type = perf_evlist__sample_type(self->evlist);
371

372
	if (!self->fd_pipe && !(sample_type & PERF_SAMPLE_CALLCHAIN)) {
373
		if (sort__has_parent) {
374
			ui__error("Selected --sort parent, but no "
375 376
				    "callchain data. Did you call "
				    "'perf record' without -g?\n");
377
			return -EINVAL;
378
		}
379
		if (symbol_conf.use_callchain) {
380
			ui__error("Selected -g but no callchain data. Did "
381
				    "you call 'perf record' without -g?\n");
382
			return -1;
383
		}
384 385
	} else if (!rep->dont_use_callchains &&
		   callchain_param.mode != CHAIN_NONE &&
386
		   !symbol_conf.use_callchain) {
387
			symbol_conf.use_callchain = true;
388
			if (callchain_register_param(&callchain_param) < 0) {
389
				ui__error("Can't register callchain params.\n");
390
				return -EINVAL;
391
			}
392 393
	}

394
	if (sort__mode == SORT_MODE__BRANCH) {
395
		if (!self->fd_pipe &&
396
		    !(sample_type & PERF_SAMPLE_BRANCH_STACK)) {
397 398
			ui__error("Selected -b but no branch data. "
				  "Did you call perf record without -b?\n");
399 400 401 402
			return -1;
		}
	}

403 404
	return 0;
}
405

406 407
extern volatile int session_done;

408
static void sig_handler(int sig __maybe_unused)
409 410 411 412
{
	session_done = 1;
}

413 414
static size_t hists__fprintf_nr_sample_events(struct perf_report *rep,
					      struct hists *self,
415 416 417 418
					      const char *evname, FILE *fp)
{
	size_t ret;
	char unit;
419 420
	unsigned long nr_samples = self->stats.nr_events[PERF_RECORD_SAMPLE];
	u64 nr_events = self->stats.total_period;
421 422 423 424
	struct perf_evsel *evsel = hists_to_evsel(self);
	char buf[512];
	size_t size = sizeof(buf);

425
	if (perf_evsel__is_group_event(evsel)) {
426 427 428 429 430 431 432 433 434 435
		struct perf_evsel *pos;

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

		for_each_group_member(pos, evsel) {
			nr_samples += pos->hists.stats.nr_events[PERF_RECORD_SAMPLE];
			nr_events += pos->hists.stats.total_period;
		}
	}
436

437 438
	nr_samples = convert_unit(nr_samples, &unit);
	ret = fprintf(fp, "# Samples: %lu%c", nr_samples, unit);
439
	if (evname != NULL)
440 441
		ret += fprintf(fp, " of event '%s'", evname);

442 443 444 445 446
	if (rep->mem_mode) {
		ret += fprintf(fp, "\n# Total weight : %" PRIu64, nr_events);
		ret += fprintf(fp, "\n# Sort order   : %s", sort_order);
	} else
		ret += fprintf(fp, "\n# Event count (approx.): %" PRIu64, nr_events);
447 448 449
	return ret + fprintf(fp, "\n#\n");
}

450
static int perf_evlist__tty_browse_hists(struct perf_evlist *evlist,
451
					 struct perf_report *rep,
452
					 const char *help)
453
{
454
	struct perf_evsel *pos;
455

456 457
	list_for_each_entry(pos, &evlist->entries, node) {
		struct hists *hists = &pos->hists;
458
		const char *evname = perf_evsel__name(pos);
459

460 461 462 463
		if (symbol_conf.event_group &&
		    !perf_evsel__is_group_leader(pos))
			continue;

464
		hists__fprintf_nr_sample_events(rep, hists, evname, stdout);
465
		hists__fprintf(hists, true, 0, 0, rep->min_percent, stdout);
466 467 468 469 470 471 472
		fprintf(stdout, "\n\n");
	}

	if (sort_order == default_sort_order &&
	    parent_pattern == default_parent_pattern) {
		fprintf(stdout, "#\n# (%s)\n#\n", help);

473 474 475
		if (rep->show_threads) {
			bool style = !strcmp(rep->pretty_printing_style, "raw");
			perf_read_values_display(stdout, &rep->show_threads_values,
476
						 style);
477
			perf_read_values_destroy(&rep->show_threads_values);
478 479 480 481 482 483
		}
	}

	return 0;
}

484
static int __cmd_report(struct perf_report *rep)
485
{
486
	int ret = -EINVAL;
487
	u64 nr_samples;
488
	struct perf_session *session = rep->session;
489
	struct perf_evsel *pos;
490 491
	struct map *kernel_map;
	struct kmap *kernel_kmap;
492
	const char *help = "For a higher level overview, try: perf report --sort comm,dso";
493

494 495
	signal(SIGINT, sig_handler);

496 497 498
	if (rep->cpu_list) {
		ret = perf_session__cpu_bitmap(session, rep->cpu_list,
					       rep->cpu_bitmap);
499 500 501 502
		if (ret)
			goto out_delete;
	}

503
	if (use_browser <= 0)
504
		perf_session__fprintf_info(session, stdout, rep->show_full_info);
505

506 507
	if (rep->show_threads)
		perf_read_values_init(&rep->show_threads_values);
508

509
	ret = perf_report__setup_sample_type(rep);
510 511 512
	if (ret)
		goto out_delete;

513
	ret = perf_session__process_events(session, &rep->tool);
514
	if (ret)
515
		goto out_delete;
516

517
	kernel_map = session->machines.host.vmlinux_maps[MAP__FUNCTION];
518 519 520 521 522
	kernel_kmap = map__kmap(kernel_map);
	if (kernel_map == NULL ||
	    (kernel_map->dso->hit &&
	     (kernel_kmap->ref_reloc_sym == NULL ||
	      kernel_kmap->ref_reloc_sym->addr == 0))) {
523 524 525 526 527 528 529 530 531 532 533
		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.";
			}
		}
534

535 536 537 538
		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",
539
		desc);
540 541
	}

542
	if (verbose > 3)
543
		perf_session__fprintf(session, stdout);
544

545
	if (verbose > 2)
546
		perf_session__fprintf_dsos(session, stdout);
547

548 549 550 551 552
	if (dump_trace) {
		perf_session__fprintf_nr_events(session, stdout);
		goto out_delete;
	}

553 554 555
	nr_samples = 0;
	list_for_each_entry(pos, &session->evlist->entries, node) {
		struct hists *hists = &pos->hists;
556

557 558 559
		if (pos->idx == 0)
			hists->symbol_filter_str = rep->symbol_filter_str;

560
		hists__collapse_resort(hists);
561
		nr_samples += hists->stats.nr_events[PERF_RECORD_SAMPLE];
562 563 564 565 566 567 568 569 570

		/* Non-group events are considered as leader */
		if (symbol_conf.event_group &&
		    !perf_evsel__is_group_leader(pos)) {
			struct hists *leader_hists = &pos->leader->hists;

			hists__match(leader_hists, hists);
			hists__link(leader_hists, hists);
		}
571 572 573
	}

	if (nr_samples == 0) {
574
		ui__error("The %s file has no samples!\n", session->filename);
575
		goto out_delete;
576 577
	}

578 579 580
	list_for_each_entry(pos, &session->evlist->entries, node)
		hists__output_resort(&pos->hists);

581
	if (use_browser > 0) {
582
		if (use_browser == 1) {
583
			ret = perf_evlist__tui_browse_hists(session->evlist,
584 585
							help, NULL,
							rep->min_percent,
586 587 588 589 590 591 592 593
							&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;

594 595
		} else if (use_browser == 2) {
			perf_evlist__gtk_browse_hists(session->evlist, help,
596
						      NULL, rep->min_percent);
597
		}
598
	} else
599
		perf_evlist__tty_browse_hists(session->evlist, rep, help);
600

601
out_delete:
602 603 604 605 606 607 608 609 610 611 612 613
	/*
	 * Speed up the exit process, for large files this can
	 * take quite a while.
	 *
	 * XXX Enable this when using valgrind or if we ever
	 * librarize this command.
	 *
	 * Also experiment with obstacks to see how much speed
	 * up we'll get here.
	 *
 	 * perf_session__delete(session);
 	 */
614
	return ret;
615 616
}

617
static int
618
parse_callchain_opt(const struct option *opt, const char *arg, int unset)
619
{
620
	struct perf_report *rep = (struct perf_report *)opt->value;
621
	char *tok, *tok2;
622 623
	char *endptr;

624 625 626 627
	/*
	 * --no-call-graph
	 */
	if (unset) {
628
		rep->dont_use_callchains = true;
629 630 631
		return 0;
	}

632
	symbol_conf.use_callchain = true;
633 634 635 636

	if (!arg)
		return 0;

637 638 639 640 641 642
	tok = strtok((char *)arg, ",");
	if (!tok)
		return -1;

	/* get the output mode */
	if (!strncmp(tok, "graph", strlen(arg)))
643
		callchain_param.mode = CHAIN_GRAPH_ABS;
644

645
	else if (!strncmp(tok, "flat", strlen(arg)))
646 647 648 649 650
		callchain_param.mode = CHAIN_FLAT;

	else if (!strncmp(tok, "fractal", strlen(arg)))
		callchain_param.mode = CHAIN_GRAPH_REL;

651 652
	else if (!strncmp(tok, "none", strlen(arg))) {
		callchain_param.mode = CHAIN_NONE;
653
		symbol_conf.use_callchain = false;
654 655 656 657

		return 0;
	}

658 659 660
	else
		return -1;

661 662 663
	/* get the min percentage */
	tok = strtok(NULL, ",");
	if (!tok)
664
		goto setup;
665

666
	callchain_param.min_percent = strtod(tok, &endptr);
667 668 669
	if (tok == endptr)
		return -1;

670 671 672 673 674 675
	/* get the print limit */
	tok2 = strtok(NULL, ",");
	if (!tok2)
		goto setup;

	if (tok2[0] != 'c') {
676
		callchain_param.print_limit = strtoul(tok2, &endptr, 0);
677 678 679 680 681 682 683 684 685 686 687 688
		tok2 = strtok(NULL, ",");
		if (!tok2)
			goto setup;
	}

	/* get the call chain order */
	if (!strcmp(tok2, "caller"))
		callchain_param.order = ORDER_CALLER;
	else if (!strcmp(tok2, "callee"))
		callchain_param.order = ORDER_CALLEE;
	else
		return -1;
689
setup:
690
	if (callchain_register_param(&callchain_param) < 0) {
691 692 693
		fprintf(stderr, "Can't register callchain params\n");
		return -1;
	}
694 695 696
	return 0;
}

697
static int
698 699
parse_branch_mode(const struct option *opt __maybe_unused,
		  const char *str __maybe_unused, int unset)
700
{
701 702 703
	int *branch_mode = opt->value;

	*branch_mode = !unset;
704 705 706
	return 0;
}

707 708 709 710 711 712 713 714 715 716
static int
parse_percent_limit(const struct option *opt, const char *str,
		    int unset __maybe_unused)
{
	struct perf_report *rep = opt->value;

	rep->min_percent = strtof(str, NULL);
	return 0;
}

717
int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused)
718
{
719
	struct perf_session *session;
720
	struct stat st;
721
	bool has_br_stack = false;
722
	int branch_mode = -1;
723
	int ret = -1;
724 725
	char callchain_default_opt[] = "fractal,0.5,callee";
	const char * const report_usage[] = {
N
Namhyung Kim 已提交
726
		"perf report [<options>]",
727 728 729
		NULL
	};
	struct perf_report report = {
730
		.tool = {
731 732 733
			.sample		 = process_sample_event,
			.mmap		 = perf_event__process_mmap,
			.comm		 = perf_event__process_comm,
734 735
			.exit		 = perf_event__process_exit,
			.fork		 = perf_event__process_fork,
736 737 738 739 740 741 742 743 744 745 746 747
			.lost		 = perf_event__process_lost,
			.read		 = process_read_event,
			.attr		 = perf_event__process_attr,
			.event_type	 = perf_event__process_event_type,
			.tracing_data	 = perf_event__process_tracing_data,
			.build_id	 = perf_event__process_build_id,
			.ordered_samples = true,
			.ordering_requires_timestamps = true,
		},
		.pretty_printing_style	 = "normal",
	};
	const struct option options[] = {
748
	OPT_STRING('i', "input", &input_name, "file",
749
		    "input file name"),
750
	OPT_INCR('v', "verbose", &verbose,
751
		    "be more verbose (show symbol address, etc)"),
752 753
	OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
		    "dump raw trace in ASCII"),
754 755
	OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name,
		   "file", "vmlinux pathname"),
756 757
	OPT_STRING(0, "kallsyms", &symbol_conf.kallsyms_name,
		   "file", "kallsyms pathname"),
758
	OPT_BOOLEAN('f', "force", &report.force, "don't complain, do it"),
759
	OPT_BOOLEAN('m', "modules", &symbol_conf.use_modules,
760
		    "load module symbols - WARNING: use only with -k and LIVE kernel"),
761
	OPT_BOOLEAN('n', "show-nr-samples", &symbol_conf.show_nr_samples,
762
		    "Show a column with the number of samples"),
763
	OPT_BOOLEAN('T', "threads", &report.show_threads,
764
		    "Show per-thread event counters"),
765
	OPT_STRING(0, "pretty", &report.pretty_printing_style, "key",
766
		   "pretty printing style key: normal raw"),
767
	OPT_BOOLEAN(0, "tui", &report.use_tui, "Use the TUI interface"),
768
	OPT_BOOLEAN(0, "gtk", &report.use_gtk, "Use the GTK2 interface"),
769 770
	OPT_BOOLEAN(0, "stdio", &report.use_stdio,
		    "Use the stdio interface"),
771
	OPT_STRING('s', "sort", &sort_order, "key[,key2...]",
772
		   "sort by key(s): pid, comm, dso, symbol, parent, cpu, srcline,"
773
		   " dso_to, dso_from, symbol_to, symbol_from, mispredict,"
774 775
		   " weight, local_weight, mem, symbol_daddr, dso_daddr, tlb, "
		   "snoop, locked"),
776 777
	OPT_BOOLEAN(0, "showcpuutilization", &symbol_conf.show_cpu_utilization,
		    "Show sample percentage for different cpu modes"),
778 779
	OPT_STRING('p', "parent", &parent_pattern, "regex",
		   "regex filter to identify parent, see: '--sort parent'"),
780
	OPT_BOOLEAN('x', "exclude-other", &symbol_conf.exclude_other,
781
		    "Only display entries with parent-match"),
782 783
	OPT_CALLBACK_DEFAULT('g', "call-graph", &report, "output_type,min_percent[,print_limit],call_order",
		     "Display callchains using output_type (graph, flat, fractal, or none) , min percent threshold, optional print limit and callchain order. "
784
		     "Default: fractal,0.5,callee", &parse_callchain_opt, callchain_default_opt),
785 786
	OPT_BOOLEAN('G', "inverted", &report.inverted_callchain,
		    "alias for inverted call graph"),
787
	OPT_STRING('d', "dsos", &symbol_conf.dso_list_str, "dso[,dso...]",
788
		   "only consider symbols in these dsos"),
789
	OPT_STRING('c', "comms", &symbol_conf.comm_list_str, "comm[,comm...]",
790
		   "only consider symbols in these comms"),
791
	OPT_STRING('S', "symbols", &symbol_conf.sym_list_str, "symbol[,symbol...]",
792
		   "only consider these symbols"),
793 794
	OPT_STRING(0, "symbol-filter", &report.symbol_filter_str, "filter",
		   "only show symbols that (partially) match with this filter"),
795
	OPT_STRING('w', "column-widths", &symbol_conf.col_width_list_str,
796 797
		   "width[,width...]",
		   "don't try to adjust column width, use these fixed values"),
798
	OPT_STRING('t', "field-separator", &symbol_conf.field_sep, "separator",
799 800
		   "separator for columns, no spaces will be added between "
		   "columns '.' is reserved."),
801
	OPT_BOOLEAN('U', "hide-unresolved", &report.hide_unresolved,
802
		    "Only display entries resolved to a symbol"),
803 804
	OPT_STRING(0, "symfs", &symbol_conf.symfs, "directory",
		    "Look for files with symbols relative to this directory"),
805
	OPT_STRING('C', "cpu", &report.cpu_list, "cpu",
806 807
		   "list of cpus to profile"),
	OPT_BOOLEAN('I', "show-info", &report.show_full_info,
808
		    "Display extended information about perf.data file"),
809 810 811 812
	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)"),
813 814
	OPT_STRING('M', "disassembler-style", &disassembler_style, "disassembler style",
		   "Specify disassembler style (e.g. -M intel for intel syntax)"),
815 816
	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
817 818
	OPT_BOOLEAN(0, "group", &symbol_conf.event_group,
		    "Show event group information together"),
819
	OPT_CALLBACK_NOOPT('b', "branch-stack", &branch_mode, "",
820
		    "use branch records for histogram filling", parse_branch_mode),
821 822
	OPT_STRING(0, "objdump", &objdump_path, "path",
		   "objdump binary to use for disassembly and annotations"),
823 824
	OPT_BOOLEAN(0, "demangle", &symbol_conf.demangle,
		    "Disable symbol demangling"),
825
	OPT_BOOLEAN(0, "mem-mode", &report.mem_mode, "mem access profile"),
826 827
	OPT_CALLBACK(0, "percent-limit", &report, "percent",
		     "Don't show entries under that percent", parse_percent_limit),
828
	OPT_END()
829
	};
830

831
	perf_config(perf_report_config, &report);
832

833 834
	argc = parse_options(argc, argv, options, report_usage, 0);

835
	if (report.use_stdio)
836
		use_browser = 0;
837
	else if (report.use_tui)
838
		use_browser = 1;
839 840
	else if (report.use_gtk)
		use_browser = 2;
841

842
	if (report.inverted_callchain)
843 844
		callchain_param.order = ORDER_CALLER;

845
	if (!input_name || !strlen(input_name)) {
846
		if (!fstat(STDIN_FILENO, &st) && S_ISFIFO(st.st_mode))
847
			input_name = "-";
848
		else
849
			input_name = "perf.data";
850
	}
851 852 853 854 855 856 857 858 859 860

	if (strcmp(input_name, "-") != 0)
		setup_browser(true);
	else {
		use_browser = 0;
		perf_hpp__column_enable(PERF_HPP__OVERHEAD);
		perf_hpp__init();
	}

repeat:
861
	session = perf_session__new(input_name, O_RDONLY,
862 863 864 865 866 867 868 869
				    report.force, false, &report.tool);
	if (session == NULL)
		return -ENOMEM;

	report.session = session;

	has_br_stack = perf_header__has_feat(&session->header,
					     HEADER_BRANCH_STACK);
870

871 872
	if (branch_mode == -1 && has_br_stack)
		sort__mode = SORT_MODE__BRANCH;
873

874 875
	/* sort__mode could be NORMAL if --no-branch-stack */
	if (sort__mode == SORT_MODE__BRANCH) {
876
		/*
877 878
		 * if no sort_order is provided, then specify
		 * branch-mode specific order
879 880 881 882 883
		 */
		if (sort_order == default_sort_order)
			sort_order = "comm,dso_from,symbol_from,"
				     "dso_to,symbol_to";

884
	}
885
	if (report.mem_mode) {
886
		if (sort__mode == SORT_MODE__BRANCH) {
887 888 889
			fprintf(stderr, "branch and mem mode incompatible\n");
			goto error;
		}
890 891
		sort__mode = SORT_MODE__MEMORY;

892 893 894 895 896 897 898
		/*
		 * if no sort_order is provided, then specify
		 * branch-mode specific order
		 */
		if (sort_order == default_sort_order)
			sort_order = "local_weight,mem,sym,dso,symbol_daddr,dso_daddr,snoop,tlb,locked";
	}
899

900 901
	if (setup_sorting() < 0)
		usage_with_options(report_usage, options);
902

903
	/*
904
	 * Only in the TUI browser we are doing integrated annotation,
905 906 907
	 * so don't allocate extra space that won't be used in the stdio
	 * implementation.
	 */
908
	if (use_browser == 1 && sort__has_sym) {
909
		symbol_conf.priv_size = sizeof(struct annotation);
910
		report.annotate_init  = symbol__annotate_init;
911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926
		/*
 		 * 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;
		}
	}
927

928
	if (symbol__init() < 0)
929
		goto error;
930

931
	if (parent_pattern != default_parent_pattern) {
932
		if (sort_dimension__add("parent") < 0)
933
			goto error;
934
	}
935

936 937 938 939 940 941 942 943 944 945
	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];
	}
946

947
	sort__setup_elide(stdout);
948

949
	ret = __cmd_report(&report);
950 951 952 953 954 955
	if (ret == K_SWITCH_INPUT_DATA) {
		perf_session__delete(session);
		goto repeat;
	} else
		ret = 0;

956 957 958
error:
	perf_session__delete(session);
	return ret;
959
}