symbol.c 26.5 KB
Newer Older
1 2
#include "util.h"
#include "../perf.h"
3
#include "string.h"
4
#include "symbol.h"
5
#include "thread.h"
6

7 8
#include "debug.h"

9 10 11
#include <libelf.h>
#include <gelf.h>
#include <elf.h>
12
#include <sys/utsname.h>
P
Peter Zijlstra 已提交
13

14 15
const char *sym_hist_filter;

16 17 18 19 20 21 22
enum dso_origin {
	DSO__ORIG_KERNEL = 0,
	DSO__ORIG_JAVA_JIT,
	DSO__ORIG_FEDORA,
	DSO__ORIG_UBUNTU,
	DSO__ORIG_BUILDID,
	DSO__ORIG_DSO,
23
	DSO__ORIG_KMODULE,
24 25 26
	DSO__ORIG_NOT_FOUND,
};

27 28 29
static void dsos__add(struct dso *dso);
static struct dso *dsos__find(const char *name);

30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67
static struct rb_root kernel_maps;

static void dso__set_symbols_end(struct dso *self)
{
	struct rb_node *nd, *prevnd = rb_first(&self->syms);

	if (prevnd == NULL)
		return;

	for (nd = rb_next(prevnd); nd; nd = rb_next(nd)) {
		struct symbol *prev = rb_entry(prevnd, struct symbol, rb_node),
			      *curr = rb_entry(nd, struct symbol, rb_node);

		if (prev->end == prev->start)
			prev->end = curr->start - 1;
		prevnd = nd;
	}
}

static void kernel_maps__fixup_sym_end(void)
{
	struct map *prev, *curr;
	struct rb_node *nd, *prevnd = rb_first(&kernel_maps);

	if (prevnd == NULL)
		return;

	curr = rb_entry(prevnd, struct map, rb_node);
	dso__set_symbols_end(curr->dso);

	for (nd = rb_next(prevnd); nd; nd = rb_next(nd)) {
		prev = curr;
		curr = rb_entry(nd, struct map, rb_node);
		prev->end = curr->start - 1;
		dso__set_symbols_end(curr->dso);
	}
}

68 69
static struct symbol *symbol__new(u64 start, u64 len, const char *name,
				  unsigned int priv_size, int v)
70
{
71
	size_t namelen = strlen(name) + 1;
72
	struct symbol *self = calloc(1, priv_size + sizeof(*self) + namelen);
73

74 75 76
	if (!self)
		return NULL;

77
	if (v >= 2)
78 79
		printf("new symbol: %016Lx [%08lx]: %s, hist: %p\n",
			start, (unsigned long)len, name, self->hist);
80 81 82 83 84

	self->hist = NULL;
	self->hist_sum = 0;

	if (sym_hist_filter && !strcmp(name, sym_hist_filter))
85
		self->hist = calloc(sizeof(u64), len);
86 87 88 89

	if (priv_size) {
		memset(self, 0, priv_size);
		self = ((void *)self) + priv_size;
90
	}
91
	self->start = start;
92
	self->end   = len ? start + len - 1 : start;
93
	memcpy(self->name, name, namelen);
94 95 96 97

	return self;
}

98
static void symbol__delete(struct symbol *self, unsigned int priv_size)
99
{
100
	free(((void *)self) - priv_size);
101 102 103 104
}

static size_t symbol__fprintf(struct symbol *self, FILE *fp)
{
105
	return fprintf(fp, " %llx-%llx %s\n",
106 107 108
		       self->start, self->end, self->name);
}

109
struct dso *dso__new(const char *name, unsigned int sym_priv_size)
110 111 112 113 114
{
	struct dso *self = malloc(sizeof(*self) + strlen(name) + 1);

	if (self != NULL) {
		strcpy(self->name, name);
115 116
		self->long_name = self->name;
		self->short_name = self->name;
117
		self->syms = RB_ROOT;
118
		self->sym_priv_size = sym_priv_size;
P
Peter Zijlstra 已提交
119
		self->find_symbol = dso__find_symbol;
120
		self->slen_calculated = 0;
121
		self->origin = DSO__ORIG_NOT_FOUND;
122 123 124 125 126 127 128 129 130 131 132 133 134
	}

	return self;
}

static void dso__delete_symbols(struct dso *self)
{
	struct symbol *pos;
	struct rb_node *next = rb_first(&self->syms);

	while (next) {
		pos = rb_entry(next, struct symbol, rb_node);
		next = rb_next(&pos->rb_node);
135
		rb_erase(&pos->rb_node, &self->syms);
136
		symbol__delete(pos, self->sym_priv_size);
137 138 139 140 141 142
	}
}

void dso__delete(struct dso *self)
{
	dso__delete_symbols(self);
143 144
	if (self->long_name != self->name)
		free(self->long_name);
145 146 147 148 149 150 151
	free(self);
}

static void dso__insert_symbol(struct dso *self, struct symbol *sym)
{
	struct rb_node **p = &self->syms.rb_node;
	struct rb_node *parent = NULL;
152
	const u64 ip = sym->start;
153 154 155 156 157 158 159 160 161 162 163 164 165 166
	struct symbol *s;

	while (*p != NULL) {
		parent = *p;
		s = rb_entry(parent, struct symbol, rb_node);
		if (ip < s->start)
			p = &(*p)->rb_left;
		else
			p = &(*p)->rb_right;
	}
	rb_link_node(&sym->rb_node, parent, p);
	rb_insert_color(&sym->rb_node, &self->syms);
}

167
struct symbol *dso__find_symbol(struct dso *self, u64 ip)
168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
{
	struct rb_node *n;

	if (self == NULL)
		return NULL;

	n = self->syms.rb_node;

	while (n) {
		struct symbol *s = rb_entry(n, struct symbol, rb_node);

		if (ip < s->start)
			n = n->rb_left;
		else if (ip > s->end)
			n = n->rb_right;
		else
			return s;
	}

	return NULL;
}

size_t dso__fprintf(struct dso *self, FILE *fp)
{
192
	size_t ret = fprintf(fp, "dso: %s\n", self->long_name);
193 194 195 196 197 198 199 200 201 202

	struct rb_node *nd;
	for (nd = rb_first(&self->syms); nd; nd = rb_next(nd)) {
		struct symbol *pos = rb_entry(nd, struct symbol, rb_node);
		ret += symbol__fprintf(pos, fp);
	}

	return ret;
}

203
static int maps__load_kallsyms(symbol_filter_t filter, int use_modules, int v)
204
{
205
	struct map *map = kernel_map;
206 207 208
	char *line = NULL;
	size_t n;
	FILE *file = fopen("/proc/kallsyms", "r");
209
	int count = 0;
210 211 212 213 214

	if (file == NULL)
		goto out_failure;

	while (!feof(file)) {
215
		u64 start;
216 217 218
		struct symbol *sym;
		int line_len, len;
		char symbol_type;
219
		char *module, *symbol_name;
220 221 222 223 224 225 226 227 228 229

		line_len = getline(&line, &n, file);
		if (line_len < 0)
			break;

		if (!line)
			goto out_failure;

		line[--line_len] = '\0'; /* \n */

230
		len = hex2u64(line, &start);
231 232 233 234 235 236 237 238 239 240 241

		len++;
		if (len + 2 >= line_len)
			continue;

		symbol_type = toupper(line[len]);
		/*
		 * We're interested only in code ('T'ext)
		 */
		if (symbol_type != 'T' && symbol_type != 'W')
			continue;
242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268

		symbol_name = line + len + 2;
		module = strchr(symbol_name, '\t');
		if (module) {
			char *module_name_end;

			if (!use_modules)
				continue;
			*module = '\0';
			module = strchr(module + 1, '[');
			if (!module)
				continue;
			module_name_end = strchr(module + 1, ']');
			if (!module_name_end)
				continue;
			*(module_name_end + 1) = '\0';
			if (strcmp(map->dso->name, module)) {
				map = kernel_maps__find_by_dso_name(module);
				if (!map) {
					fputs("/proc/{kallsyms,modules} "
					      "inconsistency!\n", stderr);
					return -1;
				}
			}
			start = map->map_ip(map, start);
		} else
			map = kernel_map;
269 270 271
		/*
		 * Well fix up the end later, when we have all sorted.
		 */
272 273
		sym = symbol__new(start, 0, symbol_name,
				  map->dso->sym_priv_size, v);
274 275 276 277

		if (sym == NULL)
			goto out_delete_line;

278
		if (filter && filter(map, sym))
279
			symbol__delete(sym, map->dso->sym_priv_size);
280
		else {
281
			dso__insert_symbol(map->dso, sym);
282 283
			count++;
		}
284 285 286 287 288
	}

	free(line);
	fclose(file);

289
	return count;
290 291 292 293 294 295 296

out_delete_line:
	free(line);
out_failure:
	return -1;
}

297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314
static size_t kernel_maps__fprintf(FILE *fp)
{
	size_t printed = fprintf(stderr, "Kernel maps:\n");
	struct rb_node *nd;

	printed += map__fprintf(kernel_map, fp);
	printed += dso__fprintf(kernel_map->dso, fp);

	for (nd = rb_first(&kernel_maps); nd; nd = rb_next(nd)) {
		struct map *pos = rb_entry(nd, struct map, rb_node);

		printed += map__fprintf(pos, fp);
		printed += dso__fprintf(pos->dso, fp);
	}

	return printed + fprintf(stderr, "END kernel maps\n");
}

315 316
static int dso__load_perf_map(struct dso *self, struct map *map,
			      symbol_filter_t filter, int v)
317 318 319 320 321 322
{
	char *line = NULL;
	size_t n;
	FILE *file;
	int nr_syms = 0;

323
	file = fopen(self->long_name, "r");
324 325 326 327
	if (file == NULL)
		goto out_failure;

	while (!feof(file)) {
328
		u64 start, size;
329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353
		struct symbol *sym;
		int line_len, len;

		line_len = getline(&line, &n, file);
		if (line_len < 0)
			break;

		if (!line)
			goto out_failure;

		line[--line_len] = '\0'; /* \n */

		len = hex2u64(line, &start);

		len++;
		if (len + 2 >= line_len)
			continue;

		len += hex2u64(line + len, &size);

		len++;
		if (len + 2 >= line_len)
			continue;

		sym = symbol__new(start, size, line + len,
354
				  self->sym_priv_size, v);
355 356 357 358

		if (sym == NULL)
			goto out_delete_line;

359
		if (filter && filter(map, sym))
360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377
			symbol__delete(sym, self->sym_priv_size);
		else {
			dso__insert_symbol(self, sym);
			nr_syms++;
		}
	}

	free(line);
	fclose(file);

	return nr_syms;

out_delete_line:
	free(line);
out_failure:
	return -1;
}

378 379 380 381
/**
 * elf_symtab__for_each_symbol - iterate thru all the symbols
 *
 * @self: struct elf_symtab instance to iterate
382
 * @idx: uint32_t idx
383 384
 * @sym: GElf_Sym iterator
 */
385 386 387 388
#define elf_symtab__for_each_symbol(syms, nr_syms, idx, sym) \
	for (idx = 0, gelf_getsym(syms, idx, &sym);\
	     idx < nr_syms; \
	     idx++, gelf_getsym(syms, idx, &sym))
389 390 391 392 393 394 395 396 397 398 399 400 401 402

static inline uint8_t elf_sym__type(const GElf_Sym *sym)
{
	return GELF_ST_TYPE(sym->st_info);
}

static inline int elf_sym__is_function(const GElf_Sym *sym)
{
	return elf_sym__type(sym) == STT_FUNC &&
	       sym->st_name != 0 &&
	       sym->st_shndx != SHN_UNDEF &&
	       sym->st_size != 0;
}

403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422
static inline int elf_sym__is_label(const GElf_Sym *sym)
{
	return elf_sym__type(sym) == STT_NOTYPE &&
		sym->st_name != 0 &&
		sym->st_shndx != SHN_UNDEF &&
		sym->st_shndx != SHN_ABS;
}

static inline const char *elf_sec__name(const GElf_Shdr *shdr,
					const Elf_Data *secstrs)
{
	return secstrs->d_buf + shdr->sh_name;
}

static inline int elf_sec__is_text(const GElf_Shdr *shdr,
					const Elf_Data *secstrs)
{
	return strstr(elf_sec__name(shdr, secstrs), "text") != NULL;
}

423 424 425 426 427 428 429 430
static inline const char *elf_sym__name(const GElf_Sym *sym,
					const Elf_Data *symstrs)
{
	return symstrs->d_buf + sym->st_name;
}

static Elf_Scn *elf_section_by_name(Elf *elf, GElf_Ehdr *ep,
				    GElf_Shdr *shp, const char *name,
431
				    size_t *idx)
432 433 434 435 436 437 438 439 440 441
{
	Elf_Scn *sec = NULL;
	size_t cnt = 1;

	while ((sec = elf_nextscn(elf, sec)) != NULL) {
		char *str;

		gelf_getshdr(sec, shp);
		str = elf_strptr(elf, ep->e_shstrndx, shp->sh_name);
		if (!strcmp(name, str)) {
442 443
			if (idx)
				*idx = cnt;
444 445 446 447 448 449 450 451
			break;
		}
		++cnt;
	}

	return sec;
}

452 453 454 455 456 457 458 459 460 461
#define elf_section__for_each_rel(reldata, pos, pos_mem, idx, nr_entries) \
	for (idx = 0, pos = gelf_getrel(reldata, 0, &pos_mem); \
	     idx < nr_entries; \
	     ++idx, pos = gelf_getrel(reldata, idx, &pos_mem))

#define elf_section__for_each_rela(reldata, pos, pos_mem, idx, nr_entries) \
	for (idx = 0, pos = gelf_getrela(reldata, 0, &pos_mem); \
	     idx < nr_entries; \
	     ++idx, pos = gelf_getrela(reldata, idx, &pos_mem))

462 463 464 465 466 467 468
/*
 * We need to check if we have a .dynsym, so that we can handle the
 * .plt, synthesizing its symbols, that aren't on the symtabs (be it
 * .dynsym or .symtab).
 * And always look at the original dso, not at debuginfo packages, that
 * have the PLT data stripped out (shdr_rel_plt.sh_type == SHT_NOBITS).
 */
469
static int dso__synthesize_plt_symbols(struct  dso *self, int v)
470 471 472
{
	uint32_t nr_rel_entries, idx;
	GElf_Sym sym;
473
	u64 plt_offset;
474 475
	GElf_Shdr shdr_plt;
	struct symbol *f;
476
	GElf_Shdr shdr_rel_plt, shdr_dynsym;
477
	Elf_Data *reldata, *syms, *symstrs;
478 479 480
	Elf_Scn *scn_plt_rel, *scn_symstrs, *scn_dynsym;
	size_t dynsym_idx;
	GElf_Ehdr ehdr;
481
	char sympltname[1024];
482 483 484
	Elf *elf;
	int nr = 0, symidx, fd, err = 0;

485
	fd = open(self->long_name, O_RDONLY);
486 487 488 489 490 491 492 493 494 495 496 497 498 499
	if (fd < 0)
		goto out;

	elf = elf_begin(fd, ELF_C_READ_MMAP, NULL);
	if (elf == NULL)
		goto out_close;

	if (gelf_getehdr(elf, &ehdr) == NULL)
		goto out_elf_end;

	scn_dynsym = elf_section_by_name(elf, &ehdr, &shdr_dynsym,
					 ".dynsym", &dynsym_idx);
	if (scn_dynsym == NULL)
		goto out_elf_end;
500

501
	scn_plt_rel = elf_section_by_name(elf, &ehdr, &shdr_rel_plt,
502 503
					  ".rela.plt", NULL);
	if (scn_plt_rel == NULL) {
504
		scn_plt_rel = elf_section_by_name(elf, &ehdr, &shdr_rel_plt,
505 506
						  ".rel.plt", NULL);
		if (scn_plt_rel == NULL)
507
			goto out_elf_end;
508 509
	}

510 511
	err = -1;

512
	if (shdr_rel_plt.sh_link != dynsym_idx)
513
		goto out_elf_end;
514

515 516
	if (elf_section_by_name(elf, &ehdr, &shdr_plt, ".plt", NULL) == NULL)
		goto out_elf_end;
517 518

	/*
519
	 * Fetch the relocation section to find the idxes to the GOT
520 521 522 523
	 * and the symbols in the .dynsym they refer to.
	 */
	reldata = elf_getdata(scn_plt_rel, NULL);
	if (reldata == NULL)
524
		goto out_elf_end;
525 526 527

	syms = elf_getdata(scn_dynsym, NULL);
	if (syms == NULL)
528
		goto out_elf_end;
529

530
	scn_symstrs = elf_getscn(elf, shdr_dynsym.sh_link);
531
	if (scn_symstrs == NULL)
532
		goto out_elf_end;
533 534 535

	symstrs = elf_getdata(scn_symstrs, NULL);
	if (symstrs == NULL)
536
		goto out_elf_end;
537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552

	nr_rel_entries = shdr_rel_plt.sh_size / shdr_rel_plt.sh_entsize;
	plt_offset = shdr_plt.sh_offset;

	if (shdr_rel_plt.sh_type == SHT_RELA) {
		GElf_Rela pos_mem, *pos;

		elf_section__for_each_rela(reldata, pos, pos_mem, idx,
					   nr_rel_entries) {
			symidx = GELF_R_SYM(pos->r_info);
			plt_offset += shdr_plt.sh_entsize;
			gelf_getsym(syms, symidx, &sym);
			snprintf(sympltname, sizeof(sympltname),
				 "%s@plt", elf_sym__name(&sym, symstrs));

			f = symbol__new(plt_offset, shdr_plt.sh_entsize,
553
					sympltname, self->sym_priv_size, v);
554
			if (!f)
555
				goto out_elf_end;
556 557 558 559 560 561 562 563 564 565 566 567 568 569 570

			dso__insert_symbol(self, f);
			++nr;
		}
	} else if (shdr_rel_plt.sh_type == SHT_REL) {
		GElf_Rel pos_mem, *pos;
		elf_section__for_each_rel(reldata, pos, pos_mem, idx,
					  nr_rel_entries) {
			symidx = GELF_R_SYM(pos->r_info);
			plt_offset += shdr_plt.sh_entsize;
			gelf_getsym(syms, symidx, &sym);
			snprintf(sympltname, sizeof(sympltname),
				 "%s@plt", elf_sym__name(&sym, symstrs));

			f = symbol__new(plt_offset, shdr_plt.sh_entsize,
571
					sympltname, self->sym_priv_size, v);
572
			if (!f)
573
				goto out_elf_end;
574 575 576 577 578 579

			dso__insert_symbol(self, f);
			++nr;
		}
	}

580 581 582 583 584 585 586 587 588 589
	err = 0;
out_elf_end:
	elf_end(elf);
out_close:
	close(fd);

	if (err == 0)
		return nr;
out:
	fprintf(stderr, "%s: problems reading %s PLT info.\n",
590
		__func__, self->long_name);
591
	return 0;
592 593
}

594 595 596
static int dso__load_sym(struct dso *self, struct map *map, const char *name,
			 int fd, symbol_filter_t filter, int kernel,
			 int kmodule, int v)
597
{
598
	Elf_Data *symstrs, *secstrs;
599 600
	uint32_t nr_syms;
	int err = -1;
601
	uint32_t idx;
602 603 604 605
	GElf_Ehdr ehdr;
	GElf_Shdr shdr;
	Elf_Data *syms;
	GElf_Sym sym;
606
	Elf_Scn *sec, *sec_strndx;
607
	Elf *elf;
608
	int nr = 0;
609 610 611

	elf = elf_begin(fd, ELF_C_READ_MMAP, NULL);
	if (elf == NULL) {
612
		if (v)
613 614
			fprintf(stderr, "%s: cannot read %s ELF file.\n",
				__func__, name);
615 616 617 618
		goto out_close;
	}

	if (gelf_getehdr(elf, &ehdr) == NULL) {
619
		if (v)
620
			fprintf(stderr, "%s: cannot get elf header.\n", __func__);
621 622 623 624
		goto out_elf_end;
	}

	sec = elf_section_by_name(elf, &ehdr, &shdr, ".symtab", NULL);
625
	if (sec == NULL) {
626 627
		sec = elf_section_by_name(elf, &ehdr, &shdr, ".dynsym", NULL);
		if (sec == NULL)
628 629
			goto out_elf_end;
	}
630 631 632 633 634 635 636 637 638 639 640 641 642

	syms = elf_getdata(sec, NULL);
	if (syms == NULL)
		goto out_elf_end;

	sec = elf_getscn(elf, shdr.sh_link);
	if (sec == NULL)
		goto out_elf_end;

	symstrs = elf_getdata(sec, NULL);
	if (symstrs == NULL)
		goto out_elf_end;

643 644 645 646 647
	sec_strndx = elf_getscn(elf, ehdr.e_shstrndx);
	if (sec_strndx == NULL)
		goto out_elf_end;

	secstrs = elf_getdata(sec_strndx, NULL);
S
Stoyan Gaydarov 已提交
648
	if (secstrs == NULL)
649 650
		goto out_elf_end;

651 652
	nr_syms = shdr.sh_size / shdr.sh_entsize;

653
	memset(&sym, 0, sizeof(sym));
654 655
	if (!kernel) {
		self->adjust_symbols = (ehdr.e_type == ET_EXEC ||
656 657 658
				elf_section_by_name(elf, &ehdr, &shdr,
						     ".gnu.prelink_undo",
						     NULL) != NULL);
659 660
	} else self->adjust_symbols = 0;

661
	elf_symtab__for_each_symbol(syms, nr_syms, idx, sym) {
662
		struct symbol *f;
663
		const char *elf_name;
664
		char *demangled;
665 666
		int is_label = elf_sym__is_label(&sym);
		const char *section_name;
667
		u64 sh_offset = 0;
668

669
		if (!is_label && !elf_sym__is_function(&sym))
670 671 672 673 674 675 676
			continue;

		sec = elf_getscn(elf, sym.st_shndx);
		if (!sec)
			goto out_elf_end;

		gelf_getshdr(sec, &shdr);
677 678 679 680 681

		if (is_label && !elf_sec__is_text(&shdr, secstrs))
			continue;

		section_name = elf_sec__name(&shdr, secstrs);
682

683 684 685 686 687
		if ((kernel || kmodule)) {
			if (strstr(section_name, ".init"))
				sh_offset = shdr.sh_offset;
		}

688
		if (self->adjust_symbols) {
689
			if (v >= 2)
690 691
				printf("adjusting symbol: st_value: %Lx sh_addr: %Lx sh_offset: %Lx\n",
					(u64)sym.st_value, (u64)shdr.sh_addr, (u64)shdr.sh_offset);
692

693
			sym.st_value -= shdr.sh_addr - shdr.sh_offset;
694
		}
695 696 697 698 699
		/*
		 * We need to figure out if the object was created from C++ sources
		 * DWARF DW_compile_unit has this, but we don't always have access
		 * to it...
		 */
700 701
		elf_name = elf_sym__name(&sym, symstrs);
		demangled = bfd_demangle(NULL, elf_name, DMGL_PARAMS | DMGL_ANSI);
702
		if (demangled != NULL)
703
			elf_name = demangled;
704

705
		f = symbol__new(sym.st_value + sh_offset, sym.st_size, elf_name,
706
				self->sym_priv_size, v);
707
		free(demangled);
708 709 710
		if (!f)
			goto out_elf_end;

711
		if (filter && filter(map, f))
712 713 714 715 716
			symbol__delete(f, self->sym_priv_size);
		else {
			dso__insert_symbol(self, f);
			nr++;
		}
717 718 719 720 721 722 723 724 725
	}

	err = nr;
out_elf_end:
	elf_end(elf);
out_close:
	return err;
}

726 727
#define BUILD_ID_SIZE 128

728
static char *dso__read_build_id(struct dso *self, int v)
729 730 731 732 733 734 735 736 737
{
	int i;
	GElf_Ehdr ehdr;
	GElf_Shdr shdr;
	Elf_Data *build_id_data;
	Elf_Scn *sec;
	char *build_id = NULL, *bid;
	unsigned char *raw;
	Elf *elf;
738
	int fd = open(self->long_name, O_RDONLY);
739 740 741 742 743 744

	if (fd < 0)
		goto out;

	elf = elf_begin(fd, ELF_C_READ_MMAP, NULL);
	if (elf == NULL) {
745
		if (v)
746
			fprintf(stderr, "%s: cannot read %s ELF file.\n",
747
				__func__, self->long_name);
748 749 750 751
		goto out_close;
	}

	if (gelf_getehdr(elf, &ehdr) == NULL) {
752
		if (v)
753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774
			fprintf(stderr, "%s: cannot get elf header.\n", __func__);
		goto out_elf_end;
	}

	sec = elf_section_by_name(elf, &ehdr, &shdr, ".note.gnu.build-id", NULL);
	if (sec == NULL)
		goto out_elf_end;

	build_id_data = elf_getdata(sec, NULL);
	if (build_id_data == NULL)
		goto out_elf_end;
	build_id = malloc(BUILD_ID_SIZE);
	if (build_id == NULL)
		goto out_elf_end;
	raw = build_id_data->d_buf + 16;
	bid = build_id;

	for (i = 0; i < 20; ++i) {
		sprintf(bid, "%02x", *raw);
		++raw;
		bid += 2;
	}
775
	if (v >= 2)
776
		printf("%s(%s): %s\n", __func__, self->long_name, build_id);
777 778 779 780 781 782 783 784
out_elf_end:
	elf_end(elf);
out_close:
	close(fd);
out:
	return build_id;
}

785 786 787 788 789 790 791 792 793
char dso__symtab_origin(const struct dso *self)
{
	static const char origin[] = {
		[DSO__ORIG_KERNEL] =   'k',
		[DSO__ORIG_JAVA_JIT] = 'j',
		[DSO__ORIG_FEDORA] =   'f',
		[DSO__ORIG_UBUNTU] =   'u',
		[DSO__ORIG_BUILDID] =  'b',
		[DSO__ORIG_DSO] =      'd',
794
		[DSO__ORIG_KMODULE] =  'K',
795 796 797 798 799 800 801
	};

	if (self == NULL || self->origin == DSO__ORIG_NOT_FOUND)
		return '!';
	return origin[self->origin];
}

802
int dso__load(struct dso *self, struct map *map, symbol_filter_t filter, int v)
803
{
804 805
	int size = PATH_MAX;
	char *name = malloc(size), *build_id = NULL;
806 807 808 809 810 811
	int ret = -1;
	int fd;

	if (!name)
		return -1;

812
	self->adjust_symbols = 0;
813

814
	if (strncmp(self->name, "/tmp/perf-", 10) == 0) {
815
		ret = dso__load_perf_map(self, map, filter, v);
816 817 818 819 820 821
		self->origin = ret > 0 ? DSO__ORIG_JAVA_JIT :
					 DSO__ORIG_NOT_FOUND;
		return ret;
	}

	self->origin = DSO__ORIG_FEDORA - 1;
822

823 824
more:
	do {
825 826 827
		self->origin++;
		switch (self->origin) {
		case DSO__ORIG_FEDORA:
828 829
			snprintf(name, size, "/usr/lib/debug%s.debug",
				 self->long_name);
830
			break;
831
		case DSO__ORIG_UBUNTU:
832 833
			snprintf(name, size, "/usr/lib/debug%s",
				 self->long_name);
834
			break;
835
		case DSO__ORIG_BUILDID:
836
			build_id = dso__read_build_id(self, v);
837 838 839 840 841 842 843
			if (build_id != NULL) {
				snprintf(name, size,
					 "/usr/lib/debug/.build-id/%.2s/%s.debug",
					build_id, build_id + 2);
				free(build_id);
				break;
			}
844
			self->origin++;
845
			/* Fall thru */
846
		case DSO__ORIG_DSO:
847
			snprintf(name, size, "%s", self->long_name);
848 849 850 851 852 853 854 855 856
			break;

		default:
			goto out;
		}

		fd = open(name, O_RDONLY);
	} while (fd < 0);

857
	ret = dso__load_sym(self, map, name, fd, filter, 0, 0, v);
858 859 860 861 862 863 864 865
	close(fd);

	/*
	 * Some people seem to have debuginfo files _WITHOUT_ debug info!?!?
	 */
	if (!ret)
		goto more;

866
	if (ret > 0) {
867
		int nr_plt = dso__synthesize_plt_symbols(self, v);
868 869 870
		if (nr_plt > 0)
			ret += nr_plt;
	}
871 872
out:
	free(name);
873 874
	if (ret < 0 && strstr(self->name, " (deleted)") != NULL)
		return 0;
875 876 877
	return ret;
}

878 879 880
struct map *kernel_map;

static void kernel_maps__insert(struct map *map)
881
{
882 883
	maps__insert(&kernel_maps, map);
}
884

885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902
struct symbol *kernel_maps__find_symbol(u64 ip, struct map **mapp)
{
	/*
	 * We can't have kernel_map in kernel_maps because it spans an address
	 * space that includes the modules. The right way to fix this is to
	 * create several maps, so that we don't have overlapping ranges with
	 * modules. For now lets look first on the kernel dso.
	 */
	struct map *map = maps__find(&kernel_maps, ip);
	struct symbol *sym;

	if (map) {
		ip = map->map_ip(map, ip);
		sym = map->dso->find_symbol(map->dso, ip);
	} else {
		map = kernel_map;
		sym = map->dso->find_symbol(map->dso, ip);
	}
903

904 905
	if (mapp)
		*mapp = map;
906

907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932
	return sym;
}

struct map *kernel_maps__find_by_dso_name(const char *name)
{
	struct rb_node *nd;

	for (nd = rb_first(&kernel_maps); nd; nd = rb_next(nd)) {
		struct map *map = rb_entry(nd, struct map, rb_node);

		if (map->dso && strcmp(map->dso->name, name) == 0)
			return map;
	}

	return NULL;
}

static int dso__load_module_sym(struct dso *self, struct map *map,
				symbol_filter_t filter, int v)
{
	int err = 0, fd = open(self->long_name, O_RDONLY);

	if (fd < 0) {
		if (v)
			fprintf(stderr, "%s: cannot open %s\n",
				__func__, self->long_name);
933
		return err;
934
	}
935

936
	err = dso__load_sym(self, map, self->long_name, fd, filter, 0, 1, v);
937 938 939 940 941
	close(fd);

	return err;
}

942 943
static int dsos__load_modules_sym_dir(char *dirname,
				      symbol_filter_t filter, int v)
944
{
945 946 947
	struct dirent *dent;
	int nr_symbols = 0, err;
	DIR *dir = opendir(dirname);
948

949 950 951 952 953 954
	if (!dir) {
		if (v)
			fprintf(stderr, "%s: cannot open %s dir\n", __func__,
				dirname);
		return -1;
	}
955

956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002
	while ((dent = readdir(dir)) != NULL) {
		char path[PATH_MAX];

		if (dent->d_type == DT_DIR) {
			if (!strcmp(dent->d_name, ".") ||
			    !strcmp(dent->d_name, ".."))
				continue;

			snprintf(path, sizeof(path), "%s/%s",
				 dirname, dent->d_name);
			err = dsos__load_modules_sym_dir(path, filter, v);
			if (err < 0)
				goto failure;
		} else {
			char *dot = strrchr(dent->d_name, '.'),
			     dso_name[PATH_MAX];
			struct map *map;
			struct rb_node *last;

			if (dot == NULL || strcmp(dot, ".ko"))
				continue;
			snprintf(dso_name, sizeof(dso_name), "[%.*s]",
				 (int)(dot - dent->d_name), dent->d_name);

			map = kernel_maps__find_by_dso_name(dso_name);
			if (map == NULL)
				continue;

			snprintf(path, sizeof(path), "%s/%s",
				 dirname, dent->d_name);

			map->dso->long_name = strdup(path);
			if (map->dso->long_name == NULL)
				goto failure;

			err = dso__load_module_sym(map->dso, map, filter, v);
			if (err < 0)
				goto failure;
			last = rb_last(&map->dso->syms);
			if (last) {
				struct symbol *sym;
				sym = rb_entry(last, struct symbol, rb_node);
				map->end = map->start + sym->end;
			}
		}
		nr_symbols += err;
	}
1003

1004 1005 1006 1007 1008
	return nr_symbols;
failure:
	closedir(dir);
	return -1;
}
1009

1010 1011 1012 1013
static int dsos__load_modules_sym(symbol_filter_t filter, int v)
{
	struct utsname uts;
	char modules_path[PATH_MAX];
1014

1015 1016
	if (uname(&uts) < 0)
		return -1;
1017

1018 1019
	snprintf(modules_path, sizeof(modules_path), "/lib/modules/%s/kernel",
		 uts.release);
1020

1021
	return dsos__load_modules_sym_dir(modules_path, filter, v);
1022 1023
}

1024 1025 1026 1027 1028 1029
/*
 * Constructor variant for modules (where we know from /proc/modules where
 * they are loaded) and for vmlinux, where only after we load all the
 * symbols we'll know where it starts and ends.
 */
static struct map *map__new2(u64 start, struct dso *dso)
1030
{
1031
	struct map *self = malloc(sizeof(*self));
1032

1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047
	if (self != NULL) {
		self->start = start;
		/*
		 * Will be filled after we load all the symbols
		 */
		self->end = 0;

		self->pgoff = 0;
		self->dso = dso;
		self->map_ip = map__map_ip;
		RB_CLEAR_NODE(&self->rb_node);
	}
	return self;
}

1048
static int dsos__load_modules(unsigned int sym_priv_size)
1049 1050 1051 1052 1053
{
	char *line = NULL;
	size_t n;
	FILE *file = fopen("/proc/modules", "r");
	struct map *map;
1054

1055 1056
	if (file == NULL)
		return -1;
1057

1058 1059 1060 1061 1062 1063
	while (!feof(file)) {
		char name[PATH_MAX];
		u64 start;
		struct dso *dso;
		char *sep;
		int line_len;
1064

1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095
		line_len = getline(&line, &n, file);
		if (line_len < 0)
			break;

		if (!line)
			goto out_failure;

		line[--line_len] = '\0'; /* \n */

		sep = strrchr(line, 'x');
		if (sep == NULL)
			continue;

		hex2u64(sep + 1, &start);

		sep = strchr(line, ' ');
		if (sep == NULL)
			continue;

		*sep = '\0';

		snprintf(name, sizeof(name), "[%s]", line);
		dso = dso__new(name, sym_priv_size);

		if (dso == NULL)
			goto out_delete_line;

		map = map__new2(start, dso);
		if (map == NULL) {
			dso__delete(dso);
			goto out_delete_line;
1096
		}
1097 1098 1099 1100

		dso->origin = DSO__ORIG_KMODULE;
		kernel_maps__insert(map);
		dsos__add(dso);
1101
	}
1102 1103 1104 1105

	free(line);
	fclose(file);

1106
	return 0;
1107 1108 1109 1110 1111

out_delete_line:
	free(line);
out_failure:
	return -1;
1112 1113
}

1114 1115
static int dso__load_vmlinux(struct dso *self, struct map *map,
			     const char *vmlinux,
1116
			     symbol_filter_t filter, int v)
1117 1118 1119 1120 1121 1122
{
	int err, fd = open(vmlinux, O_RDONLY);

	if (fd < 0)
		return -1;

1123
	err = dso__load_sym(self, map, self->long_name, fd, filter, 1, 0, v);
1124

1125 1126 1127 1128 1129
	close(fd);

	return err;
}

1130 1131
int dsos__load_kernel(const char *vmlinux, unsigned int sym_priv_size,
		      symbol_filter_t filter, int v, int use_modules)
1132 1133
{
	int err = -1;
1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144
	struct dso *dso = dso__new(vmlinux, sym_priv_size);

	if (dso == NULL)
		return -1;

	dso->short_name = "[kernel]";
	kernel_map = map__new2(0, dso);
	if (kernel_map == NULL)
		goto out_delete_dso;

	kernel_map->map_ip = vdso__map_ip;
1145

1146 1147 1148 1149 1150 1151
	if (use_modules && dsos__load_modules(sym_priv_size) < 0) {
		fprintf(stderr, "Failed to load list of modules in use! "
				"Continuing...\n");
		use_modules = 0;
	}

1152
	if (vmlinux) {
1153
		err = dso__load_vmlinux(dso, kernel_map, vmlinux, filter, v);
1154
		if (err > 0 && use_modules) {
1155
			int syms = dsos__load_modules_sym(filter, v);
1156

1157 1158 1159 1160 1161
			if (syms < 0)
				fprintf(stderr, "Failed to read module symbols!"
					" Continuing...\n");
			else
				err += syms;
1162
		}
1163
	}
1164

1165
	if (err <= 0)
1166
		err = maps__load_kallsyms(filter, use_modules, v);
1167 1168 1169 1170

	if (err > 0) {
		struct rb_node *node = rb_first(&dso->syms);
		struct symbol *sym = rb_entry(node, struct symbol, rb_node);
1171 1172 1173 1174 1175 1176
		/*
		 * Now that we have all sorted out, just set the ->end of all
		 * symbols that still don't have it.
		 */
		dso__set_symbols_end(dso);
		kernel_maps__fixup_sym_end();
1177

1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188
		kernel_map->start = sym->start;
		node = rb_last(&dso->syms);
		sym = rb_entry(node, struct symbol, rb_node);
		kernel_map->end = sym->end;

		dso->origin = DSO__ORIG_KERNEL;
		/*
		 * XXX See kernel_maps__find_symbol comment
		 * kernel_maps__insert(kernel_map)
		 */
		dsos__add(dso);
1189 1190 1191

		if (v > 0)
			kernel_maps__fprintf(stderr);
1192
	}
1193

1194
	return err;
1195 1196 1197 1198

out_delete_dso:
	dso__delete(dso);
	return -1;
1199 1200
}

1201 1202 1203
LIST_HEAD(dsos);
struct dso	*vdso;

1204
const char	*vmlinux_name = "vmlinux";
1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233
int		modules;

static void dsos__add(struct dso *dso)
{
	list_add_tail(&dso->node, &dsos);
}

static struct dso *dsos__find(const char *name)
{
	struct dso *pos;

	list_for_each_entry(pos, &dsos, node)
		if (strcmp(pos->name, name) == 0)
			return pos;
	return NULL;
}

struct dso *dsos__findnew(const char *name)
{
	struct dso *dso = dsos__find(name);
	int nr;

	if (dso)
		return dso;

	dso = dso__new(name, 0);
	if (!dso)
		goto out_delete_dso;

1234
	nr = dso__load(dso, NULL, NULL, verbose);
1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260
	if (nr < 0) {
		eprintf("Failed to open: %s\n", name);
		goto out_delete_dso;
	}
	if (!nr)
		eprintf("No symbols found in: %s, maybe install a debug package?\n", name);

	dsos__add(dso);

	return dso;

out_delete_dso:
	dso__delete(dso);
	return NULL;
}

void dsos__fprintf(FILE *fp)
{
	struct dso *pos;

	list_for_each_entry(pos, &dsos, node)
		dso__fprintf(pos, fp);
}

int load_kernel(void)
{
1261
	if (dsos__load_kernel(vmlinux_name, 0, NULL, verbose, modules) <= 0)
1262 1263 1264 1265 1266 1267 1268 1269
		return -1;

	vdso = dso__new("[vdso]", 0);
	if (!vdso)
		return -1;

	dsos__add(vdso);

1270
	return 0;
1271 1272
}

1273 1274 1275 1276
void symbol__init(void)
{
	elf_version(EV_CURRENT);
}