Commit | Line | Data |
---|---|---|
c40a57e5 AB |
1 | /* |
2 | * so-info.c | |
3 | * | |
4 | * Babeltrace - Executable and Shared Object Debug Info Reader | |
5 | * | |
6 | * Copyright 2015 Antoine Busque <abusque@efficios.com> | |
7 | * | |
8 | * Author: Antoine Busque <abusque@efficios.com> | |
9 | * | |
10 | * Permission is hereby granted, free of charge, to any person obtaining a copy | |
11 | * of this software and associated documentation files (the "Software"), to deal | |
12 | * in the Software without restriction, including without limitation the rights | |
13 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | |
14 | * copies of the Software, and to permit persons to whom the Software is | |
15 | * furnished to do so, subject to the following conditions: | |
16 | * | |
17 | * The above copyright notice and this permission notice shall be included in | |
18 | * all copies or substantial portions of the Software. | |
19 | * | |
20 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
21 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
22 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE | |
23 | * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
24 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
25 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE | |
26 | * SOFTWARE. | |
27 | */ | |
28 | ||
29 | #include <fcntl.h> | |
30 | #include <math.h> | |
31 | #include <libgen.h> | |
32 | #include <stdio.h> | |
33 | #include <inttypes.h> | |
34 | #include <stdlib.h> | |
35 | #include <string.h> | |
36 | #include <unistd.h> | |
37 | #include <dwarf.h> | |
38 | #include <glib.h> | |
39 | #include <babeltrace/dwarf.h> | |
40 | #include <babeltrace/so-info.h> | |
41 | #include <babeltrace/crc32.h> | |
42 | #include <babeltrace/babeltrace-internal.h> | |
55cd033d | 43 | #include <babeltrace/utils.h> |
c40a57e5 AB |
44 | |
45 | /* | |
46 | * An address printed in hex is at most 20 bytes (16 for 64-bits + | |
47 | * leading 0x + optional leading '+' if addr is an offset + null | |
48 | * character). | |
49 | */ | |
50 | #define ADDR_STR_LEN 20 | |
51 | ||
52 | BT_HIDDEN | |
53 | int so_info_init(void) | |
54 | { | |
55 | int ret = 0; | |
56 | ||
57 | if (elf_version(EV_CURRENT) == EV_NONE) { | |
58 | fprintf(stderr, "ELF library initialization failed: %s\n", | |
59 | elf_errmsg(-1)); | |
60 | ret = -1; | |
61 | } | |
62 | ||
63 | return ret; | |
64 | } | |
65 | ||
66 | BT_HIDDEN | |
67 | struct so_info *so_info_create(const char *path, uint64_t low_addr, | |
9f2b13ca | 68 | uint64_t memsz, bool is_pic) |
c40a57e5 AB |
69 | { |
70 | struct so_info *so = NULL; | |
c40a57e5 AB |
71 | |
72 | if (!path) { | |
73 | goto error; | |
74 | } | |
75 | ||
76 | so = g_new0(struct so_info, 1); | |
77 | if (!so) { | |
78 | goto error; | |
79 | } | |
80 | ||
81 | so->elf_path = strdup(path); | |
82 | if (!so->elf_path) { | |
83 | goto error; | |
84 | } | |
85 | ||
9f2b13ca | 86 | so->is_pic = is_pic; |
c40a57e5 AB |
87 | so->memsz = memsz; |
88 | so->low_addr = low_addr; | |
89 | so->high_addr = so->low_addr + so->memsz; | |
90 | ||
c40a57e5 AB |
91 | return so; |
92 | ||
93 | error: | |
c40a57e5 AB |
94 | so_info_destroy(so); |
95 | return NULL; | |
96 | } | |
97 | ||
98 | BT_HIDDEN | |
99 | void so_info_destroy(struct so_info *so) | |
100 | { | |
101 | if (!so) { | |
102 | return; | |
103 | } | |
104 | ||
105 | dwarf_end(so->dwarf_info); | |
106 | ||
107 | free(so->elf_path); | |
108 | free(so->dwarf_path); | |
109 | free(so->build_id); | |
110 | free(so->dbg_link_filename); | |
111 | ||
112 | elf_end(so->elf_file); | |
113 | ||
114 | close(so->elf_fd); | |
115 | close(so->dwarf_fd); | |
116 | ||
117 | g_free(so); | |
118 | } | |
119 | ||
49824faa | 120 | |
c40a57e5 AB |
121 | BT_HIDDEN |
122 | int so_info_set_build_id(struct so_info *so, uint8_t *build_id, | |
123 | size_t build_id_len) | |
124 | { | |
125 | if (!so || !build_id) { | |
126 | goto error; | |
127 | } | |
128 | ||
129 | so->build_id = malloc(build_id_len); | |
130 | if (!so->build_id) { | |
131 | goto error; | |
132 | } | |
133 | ||
134 | memcpy(so->build_id, build_id, build_id_len); | |
135 | so->build_id_len = build_id_len; | |
136 | ||
137 | /* | |
138 | * Reset the is_elf_only flag in case it had been set | |
139 | * previously, because we might find separate debug info using | |
140 | * the new build id information. | |
141 | */ | |
142 | so->is_elf_only = false; | |
143 | ||
144 | return 0; | |
145 | ||
146 | error: | |
147 | ||
148 | return -1; | |
149 | } | |
150 | ||
151 | BT_HIDDEN | |
152 | int so_info_set_debug_link(struct so_info *so, char *filename, uint32_t crc) | |
153 | { | |
154 | if (!so || !filename) { | |
155 | goto error; | |
156 | } | |
157 | ||
158 | so->dbg_link_filename = strdup(filename); | |
159 | if (!so->dbg_link_filename) { | |
160 | goto error; | |
161 | } | |
162 | ||
163 | so->dbg_link_crc = crc; | |
164 | ||
165 | /* | |
166 | * Reset the is_elf_only flag in case it had been set | |
167 | * previously, because we might find separate debug info using | |
168 | * the new build id information. | |
169 | */ | |
170 | so->is_elf_only = false; | |
171 | ||
172 | return 0; | |
173 | ||
174 | error: | |
175 | ||
176 | return -1; | |
177 | } | |
178 | ||
179 | /** | |
180 | * Tries to read DWARF info from the location given by path, and | |
181 | * attach it to the given so_info instance if it exists. | |
182 | * | |
183 | * @param so so_info instance for which to set DWARF info | |
184 | * @param path Presumed location of the DWARF info | |
185 | * @returns 0 on success, -1 on failure | |
186 | */ | |
187 | static | |
188 | int so_info_set_dwarf_info_from_path(struct so_info *so, char *path) | |
189 | { | |
190 | int fd = -1, ret = 0; | |
191 | struct bt_dwarf_cu *cu = NULL; | |
192 | Dwarf *dwarf_info = NULL; | |
193 | ||
194 | if (!so || !path) { | |
195 | goto error; | |
196 | } | |
197 | ||
198 | fd = open(path, O_RDONLY); | |
199 | if (fd < 0) { | |
200 | goto error; | |
201 | } | |
202 | ||
203 | dwarf_info = dwarf_begin(fd, DWARF_C_READ); | |
204 | if (!dwarf_info) { | |
205 | goto error; | |
206 | } | |
207 | ||
208 | /* | |
209 | * Check if the dwarf info has any CU. If not, the SO's object | |
210 | * file contains no DWARF info. | |
211 | */ | |
212 | cu = bt_dwarf_cu_create(dwarf_info); | |
213 | if (!cu) { | |
214 | goto error; | |
215 | } | |
216 | ||
217 | ret = bt_dwarf_cu_next(cu); | |
218 | if (ret) { | |
219 | goto error; | |
220 | } | |
221 | ||
222 | so->dwarf_fd = fd; | |
223 | so->dwarf_path = strdup(path); | |
224 | if (!so->dwarf_path) { | |
225 | goto error; | |
226 | } | |
227 | so->dwarf_info = dwarf_info; | |
228 | free(cu); | |
229 | ||
230 | return 0; | |
231 | ||
232 | error: | |
233 | close(fd); | |
234 | dwarf_end(dwarf_info); | |
235 | g_free(dwarf_info); | |
236 | free(cu); | |
237 | ||
238 | return -1; | |
239 | } | |
240 | ||
241 | /** | |
242 | * Try to set the dwarf_info for a given so_info instance via the | |
243 | * build ID method. | |
244 | * | |
245 | * @param so so_info instance for which to retrieve the | |
246 | * DWARF info via build ID | |
247 | * @returns 0 on success (i.e. dwarf_info set), -1 on failure | |
248 | */ | |
249 | static | |
250 | int so_info_set_dwarf_info_build_id(struct so_info *so) | |
251 | { | |
252 | int i = 0, ret = 0, dbg_dir_trailing_slash = 0; | |
253 | char *path = NULL, *build_id_file = NULL; | |
254 | const char *dbg_dir = NULL; | |
255 | size_t build_id_file_len, path_len; | |
256 | ||
257 | if (!so || !so->build_id) { | |
258 | goto error; | |
259 | } | |
260 | ||
05984e0c | 261 | dbg_dir = opt_debug_info_dir ? : DEFAULT_DEBUG_DIR; |
c40a57e5 AB |
262 | |
263 | dbg_dir_trailing_slash = dbg_dir[strlen(dbg_dir) - 1] == '/'; | |
264 | ||
265 | /* 2 characters per byte printed in hex, +2 for '/' and '\0' */ | |
266 | build_id_file_len = (2 * so->build_id_len) + 2; | |
267 | build_id_file = malloc(build_id_file_len); | |
268 | if (!build_id_file) { | |
269 | goto error; | |
270 | } | |
271 | ||
272 | snprintf(build_id_file, 4, "%02x/", so->build_id[0]); | |
273 | for (i = 1; i < so->build_id_len; ++i) { | |
274 | int path_idx = 3 + 2 * (i - 1); | |
275 | ||
276 | snprintf(&build_id_file[path_idx], 3, "%02x", so->build_id[i]); | |
277 | } | |
278 | ||
279 | path_len = strlen(dbg_dir) + strlen(BUILD_ID_SUBDIR) + | |
280 | strlen(build_id_file) + strlen(BUILD_ID_SUFFIX) + 1; | |
281 | if (!dbg_dir_trailing_slash) { | |
282 | path_len += 1; | |
283 | } | |
284 | ||
285 | path = malloc(path_len); | |
286 | if (!path) { | |
287 | goto error; | |
288 | } | |
289 | ||
290 | strcpy(path, dbg_dir); | |
291 | if (!dbg_dir_trailing_slash) { | |
292 | strcat(path, "/"); | |
293 | } | |
294 | strcat(path, BUILD_ID_SUBDIR); | |
295 | strcat(path, build_id_file); | |
296 | strcat(path, BUILD_ID_SUFFIX); | |
297 | ||
298 | ret = so_info_set_dwarf_info_from_path(so, path); | |
299 | if (ret) { | |
300 | goto error; | |
301 | } | |
302 | ||
303 | goto end; | |
304 | ||
305 | error: | |
306 | ret = -1; | |
307 | end: | |
308 | free(build_id_file); | |
309 | free(path); | |
310 | ||
311 | return ret; | |
312 | } | |
313 | ||
314 | /** | |
315 | * Tests whether the file located at path exists and has the expected | |
316 | * checksum. | |
317 | * | |
318 | * This predicate is used when looking up separate debug info via the | |
319 | * GNU debuglink method. The expected crc can be found .gnu_debuglink | |
320 | * section in the original ELF file, along with the filename for the | |
321 | * file containing the debug info. | |
322 | * | |
323 | * @param path Full path at which to look for the debug file | |
324 | * @param crc Expected checksum for the debug file | |
325 | * @returns 1 if the file exists and has the correct checksum, | |
326 | * 0 otherwise | |
327 | */ | |
328 | static | |
329 | int is_valid_debug_file(char *path, uint32_t crc) | |
330 | { | |
331 | int ret = 0, fd = -1; | |
332 | uint32_t _crc = 0; | |
333 | ||
334 | if (!path) { | |
335 | goto end; | |
336 | } | |
337 | ||
338 | fd = open(path, O_RDONLY); | |
339 | if (fd < 0) { | |
340 | goto end; | |
341 | } | |
342 | ||
343 | ret = crc32(fd, &_crc); | |
344 | if (ret) { | |
345 | ret = 0; | |
346 | goto end; | |
347 | } | |
348 | ||
349 | ret = (crc == _crc); | |
350 | ||
351 | end: | |
352 | close(fd); | |
353 | return ret; | |
354 | } | |
355 | ||
356 | /** | |
357 | * Try to set the dwarf_info for a given so_info instance via the | |
358 | * build ID method. | |
359 | * | |
360 | * @param so so_info instance for which to retrieve the | |
361 | * DWARF info via debug link | |
362 | * @returns 0 on success (i.e. dwarf_info set), -1 on failure | |
363 | */ | |
364 | static | |
365 | int so_info_set_dwarf_info_debug_link(struct so_info *so) | |
366 | { | |
367 | int ret = 0; | |
368 | const char *dbg_dir = NULL; | |
369 | char *dir_name = NULL, *so_dir = NULL, *path = NULL; | |
370 | size_t max_path_len = 0; | |
371 | ||
372 | if (!so || !so->dbg_link_filename) { | |
373 | goto error; | |
374 | } | |
375 | ||
05984e0c | 376 | dbg_dir = opt_debug_info_dir ? : DEFAULT_DEBUG_DIR; |
c40a57e5 AB |
377 | |
378 | dir_name = dirname(so->elf_path); | |
379 | if (!dir_name) { | |
380 | goto error; | |
381 | } | |
382 | ||
383 | /* so_dir is just dir_name with a trailing slash */ | |
384 | so_dir = malloc(strlen(dir_name) + 2); | |
385 | if (!so_dir) { | |
386 | goto error; | |
387 | } | |
388 | ||
389 | strcpy(so_dir, dir_name); | |
390 | strcat(so_dir, "/"); | |
391 | ||
392 | max_path_len = strlen(dbg_dir) + strlen(so_dir) + | |
393 | strlen(DEBUG_SUBDIR) + strlen(so->dbg_link_filename) | |
394 | + 1; | |
395 | path = malloc(max_path_len); | |
396 | if (!path) { | |
397 | goto error; | |
398 | } | |
399 | ||
400 | /* First look in the SO's dir */ | |
401 | strcpy(path, so_dir); | |
402 | strcat(path, so->dbg_link_filename); | |
403 | ||
404 | if (is_valid_debug_file(path, so->dbg_link_crc)) { | |
405 | goto found; | |
406 | } | |
407 | ||
408 | /* If not found, look in .debug subdir */ | |
409 | strcpy(path, so_dir); | |
410 | strcat(path, DEBUG_SUBDIR); | |
411 | strcat(path, so->dbg_link_filename); | |
412 | ||
413 | if (is_valid_debug_file(path, so->dbg_link_crc)) { | |
414 | goto found; | |
415 | } | |
416 | ||
417 | /* Lastly, look under the global debug directory */ | |
418 | strcpy(path, dbg_dir); | |
419 | strcat(path, so_dir); | |
420 | strcat(path, so->dbg_link_filename); | |
421 | ||
422 | if (is_valid_debug_file(path, so->dbg_link_crc)) { | |
423 | goto found; | |
424 | } | |
425 | ||
426 | error: | |
427 | ret = -1; | |
428 | end: | |
429 | free(path); | |
430 | free(so_dir); | |
431 | ||
432 | return ret; | |
433 | ||
434 | found: | |
435 | ret = so_info_set_dwarf_info_from_path(so, path); | |
436 | if (ret) { | |
437 | goto error; | |
438 | } | |
439 | ||
440 | goto end; | |
441 | } | |
442 | ||
443 | /** | |
444 | * Initialize the DWARF info for a given executable. | |
445 | * | |
446 | * @param so so_info instance | |
447 | * @returns 0 on success, -1 on failure | |
448 | */ | |
449 | static | |
450 | int so_info_set_dwarf_info(struct so_info *so) | |
451 | { | |
452 | int ret = 0; | |
453 | ||
454 | if (!so) { | |
455 | goto error; | |
456 | } | |
457 | ||
458 | /* First try to set the DWARF info from the ELF file */ | |
459 | ret = so_info_set_dwarf_info_from_path(so, so->elf_path); | |
460 | if (!ret) { | |
461 | goto end; | |
462 | } | |
463 | ||
464 | /* | |
465 | * If that fails, try to find separate debug info via build ID | |
466 | * and debug link. | |
467 | */ | |
468 | ret = so_info_set_dwarf_info_build_id(so); | |
469 | if (!ret) { | |
470 | goto end; | |
471 | } | |
472 | ||
473 | ret = so_info_set_dwarf_info_debug_link(so); | |
474 | if (!ret) { | |
475 | goto end; | |
476 | } | |
477 | ||
478 | error: | |
479 | ret = -1; | |
480 | end: | |
481 | return ret; | |
482 | } | |
483 | ||
49824faa AB |
484 | /** |
485 | * Initialize the ELF file for a given executable. | |
486 | * | |
487 | * @param so so_info instance | |
488 | * @returns 0 on success, -1 on failure | |
489 | */ | |
490 | static | |
491 | int so_info_set_elf_file(struct so_info *so) | |
492 | { | |
493 | int elf_fd; | |
f99dc219 | 494 | Elf *elf_file = NULL; |
49824faa AB |
495 | |
496 | if (!so) { | |
497 | goto error; | |
498 | } | |
499 | ||
500 | elf_fd = open(so->elf_path, O_RDONLY); | |
501 | if (elf_fd < 0) { | |
502 | fprintf(stderr, "Failed to open %s\n", so->elf_path); | |
503 | goto error; | |
504 | } | |
505 | ||
506 | elf_file = elf_begin(elf_fd, ELF_C_READ, NULL); | |
507 | if (!elf_file) { | |
508 | fprintf(stderr, "elf_begin failed: %s\n", elf_errmsg(-1)); | |
509 | goto error; | |
510 | } | |
511 | ||
512 | if (elf_kind(elf_file) != ELF_K_ELF) { | |
513 | fprintf(stderr, "Error: %s is not an ELF object\n", | |
514 | so->elf_path); | |
515 | goto error; | |
516 | } | |
517 | ||
518 | so->elf_fd = elf_fd; | |
519 | so->elf_file = elf_file; | |
520 | return 0; | |
521 | ||
522 | error: | |
523 | close(elf_fd); | |
524 | elf_end(elf_file); | |
525 | return -1; | |
526 | } | |
527 | ||
528 | ||
c40a57e5 AB |
529 | BT_HIDDEN |
530 | void source_location_destroy(struct source_location *src_loc) | |
531 | { | |
532 | if (!src_loc) { | |
533 | return; | |
534 | } | |
535 | ||
536 | free(src_loc->filename); | |
537 | g_free(src_loc); | |
538 | } | |
539 | ||
540 | /** | |
541 | * Try to find the symbol closest to an address within a given ELF | |
542 | * section. | |
543 | * | |
544 | * Only function symbols are taken into account. The symbol's address | |
545 | * must precede `addr`. A symbol with a closer address might exist | |
546 | * after `addr` but is irrelevant because it cannot encompass `addr`. | |
547 | * | |
548 | * On success, if found, the out parameters `sym` and `shdr` are | |
549 | * set. On failure or if none are found, they remain unchanged. | |
550 | * | |
551 | * @param scn ELF section in which to look for the address | |
552 | * @param addr Virtual memory address for which to find the | |
553 | * nearest function symbol | |
554 | * @param sym Out parameter, the nearest function symbol | |
555 | * @param shdr Out parameter, the section header for scn | |
556 | * @returns 0 on success, -1 on failure | |
557 | */ | |
558 | static | |
559 | int so_info_get_nearest_symbol_from_section(Elf_Scn *scn, uint64_t addr, | |
560 | GElf_Sym **sym, GElf_Shdr **shdr) | |
561 | { | |
562 | int i; | |
563 | size_t symbol_count; | |
564 | Elf_Data *data = NULL; | |
565 | GElf_Shdr *_shdr = NULL; | |
566 | GElf_Sym *nearest_sym = NULL; | |
567 | ||
568 | if (!scn || !sym || !shdr) { | |
569 | goto error; | |
570 | } | |
571 | ||
572 | _shdr = g_new0(GElf_Shdr, 1); | |
573 | if (!_shdr) { | |
574 | goto error; | |
575 | } | |
576 | ||
577 | _shdr = gelf_getshdr(scn, _shdr); | |
578 | if (!_shdr) { | |
579 | goto error; | |
580 | } | |
581 | ||
582 | if (_shdr->sh_type != SHT_SYMTAB) { | |
583 | /* | |
584 | * We are only interested in symbol table (symtab) | |
585 | * sections, skip this one. | |
586 | */ | |
587 | goto end; | |
588 | } | |
589 | ||
590 | data = elf_getdata(scn, NULL); | |
591 | if (!data) { | |
592 | goto error; | |
593 | } | |
594 | ||
595 | symbol_count = _shdr->sh_size / _shdr->sh_entsize; | |
596 | ||
597 | for (i = 0; i < symbol_count; ++i) { | |
598 | GElf_Sym *cur_sym = NULL; | |
599 | ||
600 | cur_sym = g_new0(GElf_Sym, 1); | |
601 | if (!cur_sym) { | |
602 | goto error; | |
603 | } | |
604 | cur_sym = gelf_getsym(data, i, cur_sym); | |
605 | if (!cur_sym) { | |
606 | goto error; | |
607 | } | |
608 | if (GELF_ST_TYPE(cur_sym->st_info) != STT_FUNC) { | |
609 | /* We're only interested in the functions. */ | |
610 | g_free(cur_sym); | |
611 | continue; | |
612 | } | |
613 | ||
614 | if (cur_sym->st_value <= addr && | |
615 | (!nearest_sym || | |
616 | cur_sym->st_value > nearest_sym->st_value)) { | |
617 | g_free(nearest_sym); | |
618 | nearest_sym = cur_sym; | |
619 | } else { | |
620 | g_free(cur_sym); | |
621 | } | |
622 | } | |
623 | ||
624 | end: | |
625 | if (nearest_sym) { | |
626 | *sym = nearest_sym; | |
627 | *shdr = _shdr; | |
628 | } else { | |
629 | g_free(_shdr); | |
630 | } | |
631 | ||
632 | return 0; | |
633 | ||
634 | error: | |
635 | g_free(nearest_sym); | |
636 | g_free(_shdr); | |
637 | return -1; | |
638 | } | |
639 | ||
640 | /** | |
641 | * Get the name of the function containing a given address within an | |
642 | * executable using ELF symbols. | |
643 | * | |
644 | * The function name is in fact the name of the nearest ELF symbol, | |
645 | * followed by the offset in bytes between the address and the symbol | |
646 | * (in hex), separated by a '+' character. | |
647 | * | |
648 | * If found, the out parameter `func_name` is set on success. On failure, | |
649 | * it remains unchanged. | |
650 | * | |
651 | * @param so so_info instance for the executable containing | |
652 | * the address | |
653 | * @param addr Virtual memory address for which to find the | |
654 | * function name | |
655 | * @param func_name Out parameter, the function name | |
656 | * @returns 0 on success, -1 on failure | |
657 | */ | |
658 | static | |
659 | int so_info_lookup_elf_function_name(struct so_info *so, uint64_t addr, | |
660 | char **func_name) | |
661 | { | |
662 | /* | |
663 | * TODO (possible optimisation): if an ELF has no symtab | |
664 | * section, it has been stripped. Therefore, it would be wise | |
665 | * to store a flag indicating the stripped status after the | |
666 | * first iteration to prevent subsequent ones. | |
667 | */ | |
668 | int ret = 0; | |
669 | Elf_Scn *scn = NULL; | |
670 | GElf_Sym *sym = NULL; | |
671 | GElf_Shdr *shdr = NULL; | |
672 | char *sym_name = NULL; | |
673 | char *_func_name = NULL; | |
674 | char offset_str[ADDR_STR_LEN]; | |
675 | ||
49824faa AB |
676 | /* Set ELF file if it hasn't been accessed yet. */ |
677 | if (!so->elf_file) { | |
678 | ret = so_info_set_elf_file(so); | |
679 | if (ret) { | |
680 | /* Failed to set ELF file. */ | |
681 | goto error; | |
682 | } | |
683 | } | |
684 | ||
c40a57e5 AB |
685 | scn = elf_nextscn(so->elf_file, scn); |
686 | if (!scn) { | |
687 | goto error; | |
688 | } | |
689 | ||
690 | while (scn && !sym) { | |
691 | ret = so_info_get_nearest_symbol_from_section( | |
692 | scn, addr, &sym, &shdr); | |
693 | if (ret) { | |
694 | goto error; | |
695 | } | |
696 | ||
697 | scn = elf_nextscn(so->elf_file, scn); | |
698 | } | |
699 | ||
700 | if (sym) { | |
701 | sym_name = elf_strptr(so->elf_file, shdr->sh_link, | |
702 | sym->st_name); | |
703 | if (!sym_name) { | |
704 | goto error; | |
705 | } | |
706 | ||
707 | snprintf(offset_str, ADDR_STR_LEN, "+%#0" PRIx64, | |
708 | addr - sym->st_value); | |
709 | _func_name = malloc(strlen(sym_name) + ADDR_STR_LEN); | |
710 | if (!_func_name) { | |
711 | goto error; | |
712 | } | |
713 | ||
714 | strcpy(_func_name, sym_name); | |
715 | strcat(_func_name, offset_str); | |
716 | *func_name = _func_name; | |
717 | } | |
718 | ||
719 | g_free(shdr); | |
720 | g_free(sym); | |
721 | return 0; | |
722 | ||
723 | error: | |
724 | g_free(shdr); | |
725 | g_free(sym); | |
726 | free(_func_name); | |
727 | return -1; | |
728 | } | |
729 | ||
730 | /** | |
731 | * Get the name of the function containing a given address within a | |
732 | * given compile unit (CU). | |
733 | * | |
734 | * If found, the out parameter `func_name` is set on success. On | |
735 | * failure, it remains unchanged. | |
736 | * | |
737 | * @param cu bt_dwarf_cu instance which may contain the address | |
738 | * @param addr Virtual memory address for which to find the | |
739 | * function name | |
740 | * @param func_name Out parameter, the function name | |
741 | * @returns 0 on success, -1 on failure | |
742 | */ | |
743 | static | |
744 | int so_info_lookup_cu_function_name(struct bt_dwarf_cu *cu, uint64_t addr, | |
745 | char **func_name) | |
746 | { | |
747 | int ret = 0, found = 0; | |
748 | char *_func_name = NULL; | |
749 | struct bt_dwarf_die *die = NULL; | |
750 | ||
751 | if (!cu || !func_name) { | |
752 | goto error; | |
753 | } | |
754 | ||
755 | die = bt_dwarf_die_create(cu); | |
756 | if (!die) { | |
757 | goto error; | |
758 | } | |
759 | ||
760 | while (bt_dwarf_die_next(die) == 0) { | |
761 | int tag; | |
762 | ||
763 | ret = bt_dwarf_die_get_tag(die, &tag); | |
764 | if (ret) { | |
765 | goto error; | |
766 | } | |
767 | ||
768 | if (tag == DW_TAG_subprogram) { | |
769 | ret = bt_dwarf_die_contains_addr(die, addr, &found); | |
770 | if (ret) { | |
771 | goto error; | |
772 | } | |
773 | ||
774 | if (found) { | |
775 | break; | |
776 | } | |
777 | } | |
778 | } | |
779 | ||
780 | if (found) { | |
781 | ret = bt_dwarf_die_get_name(die, &_func_name); | |
782 | if (ret) { | |
783 | goto error; | |
784 | } | |
785 | ||
786 | *func_name = _func_name; | |
787 | } | |
788 | ||
789 | bt_dwarf_die_destroy(die); | |
790 | return 0; | |
791 | ||
792 | error: | |
793 | bt_dwarf_die_destroy(die); | |
794 | return -1; | |
795 | } | |
796 | ||
797 | /** | |
798 | * Get the name of the function containing a given address within an | |
799 | * executable using DWARF debug info. | |
800 | * | |
801 | * If found, the out parameter `func_name` is set on success. On | |
802 | * failure, it remains unchanged. | |
803 | * | |
804 | * @param so so_info instance for the executable containing | |
805 | * the address | |
806 | * @param addr Virtual memory address for which to find the | |
807 | * function name | |
808 | * @param func_name Out parameter, the function name | |
809 | * @returns 0 on success, -1 on failure | |
810 | */ | |
811 | static | |
812 | int so_info_lookup_dwarf_function_name(struct so_info *so, uint64_t addr, | |
813 | char **func_name) | |
814 | { | |
815 | int ret = 0; | |
816 | char *_func_name = NULL; | |
817 | struct bt_dwarf_cu *cu = NULL; | |
818 | ||
819 | if (!so || !func_name) { | |
820 | goto error; | |
821 | } | |
822 | ||
823 | cu = bt_dwarf_cu_create(so->dwarf_info); | |
824 | if (!cu) { | |
825 | goto error; | |
826 | } | |
827 | ||
828 | while (bt_dwarf_cu_next(cu) == 0) { | |
829 | ret = so_info_lookup_cu_function_name(cu, addr, &_func_name); | |
830 | if (ret) { | |
831 | goto error; | |
832 | } | |
833 | ||
834 | if (_func_name) { | |
835 | break; | |
836 | } | |
837 | } | |
838 | ||
839 | if (_func_name) { | |
840 | *func_name = _func_name; | |
841 | } | |
842 | ||
843 | bt_dwarf_cu_destroy(cu); | |
844 | return 0; | |
845 | ||
846 | error: | |
847 | bt_dwarf_cu_destroy(cu); | |
848 | return -1; | |
849 | } | |
850 | ||
851 | BT_HIDDEN | |
36ae9941 | 852 | int so_info_lookup_function_name(struct so_info *so, uint64_t addr, |
c40a57e5 AB |
853 | char **func_name) |
854 | { | |
855 | int ret = 0; | |
856 | char *_func_name = NULL; | |
857 | ||
858 | if (!so || !func_name) { | |
859 | goto error; | |
860 | } | |
861 | ||
862 | /* Set DWARF info if it hasn't been accessed yet. */ | |
863 | if (!so->dwarf_info && !so->is_elf_only) { | |
864 | ret = so_info_set_dwarf_info(so); | |
865 | if (ret) { | |
866 | /* Failed to set DWARF info, fallback to ELF. */ | |
867 | so->is_elf_only = true; | |
868 | } | |
869 | } | |
870 | ||
36ae9941 | 871 | if (!so_info_has_address(so, addr)) { |
c40a57e5 AB |
872 | goto error; |
873 | } | |
874 | ||
875 | /* | |
876 | * Addresses in ELF and DWARF are relative to base address for | |
877 | * PIC, so make the address argument relative too if needed. | |
878 | */ | |
36ae9941 AB |
879 | if (so->is_pic) { |
880 | addr -= so->low_addr; | |
881 | } | |
882 | ||
c40a57e5 | 883 | if (so->is_elf_only) { |
36ae9941 | 884 | ret = so_info_lookup_elf_function_name(so, addr, &_func_name); |
c40a57e5 | 885 | } else { |
36ae9941 | 886 | ret = so_info_lookup_dwarf_function_name(so, addr, &_func_name); |
c40a57e5 AB |
887 | } |
888 | ||
36ae9941 | 889 | if (ret || !_func_name) { |
c40a57e5 AB |
890 | goto error; |
891 | } | |
892 | ||
36ae9941 AB |
893 | *func_name = _func_name; |
894 | return 0; | |
55cd033d | 895 | |
36ae9941 AB |
896 | error: |
897 | return -1; | |
898 | } | |
899 | ||
900 | BT_HIDDEN | |
901 | int so_info_get_bin_loc(struct so_info *so, uint64_t addr, char **bin_loc) | |
902 | { | |
903 | int ret = 0; | |
904 | char *_bin_loc = NULL; | |
905 | ||
906 | if (!so || !bin_loc) { | |
907 | goto error; | |
c40a57e5 AB |
908 | } |
909 | ||
36ae9941 AB |
910 | if (so->is_pic) { |
911 | addr -= so->low_addr; | |
912 | ret = asprintf(&_bin_loc, "+%#0" PRIx64, addr); | |
913 | } else { | |
914 | ret = asprintf(&_bin_loc, "@%#0" PRIx64, addr); | |
915 | } | |
916 | ||
917 | if (ret == -1 || !_bin_loc) { | |
918 | goto error; | |
919 | } | |
920 | ||
921 | *bin_loc = _bin_loc; | |
c40a57e5 AB |
922 | return 0; |
923 | ||
924 | error: | |
925 | return -1; | |
926 | } | |
927 | ||
928 | /** | |
929 | * Predicate used to determine whether the children of a given DIE | |
930 | * contain a specific address. | |
931 | * | |
932 | * More specifically, the parameter `die` is expected to be a | |
933 | * subprogram (function) DIE, and this predicate tells whether any | |
934 | * subroutines are inlined within this function and would contain | |
935 | * `addr`. | |
936 | * | |
937 | * Do note that this function advances the position of `die`. If the | |
938 | * address is found within one of its children, `die` will be pointing | |
939 | * to that child upon returning from the function, allowing to extract | |
940 | * the information deemed necessary. | |
941 | * | |
942 | * @param die The parent DIE in whose children the address will be | |
943 | * looked for | |
944 | * @param addr The address for which to look for in the DIEs | |
945 | * @returns Returns 1 if the address was found, 0 if not | |
946 | */ | |
947 | static | |
948 | int so_info_child_die_has_address(struct bt_dwarf_die *die, uint64_t addr) | |
949 | { | |
950 | int ret = 0, contains = 0; | |
951 | ||
952 | if (!die) { | |
953 | goto error; | |
954 | } | |
955 | ||
956 | ret = bt_dwarf_die_child(die); | |
957 | if (ret) { | |
958 | goto error; | |
959 | } | |
960 | ||
961 | do { | |
962 | int tag; | |
963 | ||
964 | ret = bt_dwarf_die_get_tag(die, &tag); | |
965 | if (ret) { | |
966 | goto error; | |
967 | } | |
968 | ||
969 | if (tag == DW_TAG_inlined_subroutine) { | |
970 | ret = bt_dwarf_die_contains_addr(die, addr, &contains); | |
971 | if (ret) { | |
972 | goto error; | |
973 | } | |
974 | ||
975 | if (contains) { | |
976 | ret = 1; | |
977 | goto end; | |
978 | } | |
979 | } | |
980 | } while (bt_dwarf_die_next(die) == 0); | |
981 | ||
982 | end: | |
983 | return ret; | |
984 | ||
985 | error: | |
986 | ret = 0; | |
987 | goto end; | |
988 | } | |
989 | ||
990 | /** | |
991 | * Lookup the source location for a given address within a CU, making | |
992 | * the assumption that it is contained within an inline routine in a | |
993 | * function. | |
994 | * | |
995 | * @param cu bt_dwarf_cu instance in which to look for the address | |
996 | * @param addr The address for which to look for | |
997 | * @param src_loc Out parameter, the source location (filename and | |
998 | * line number) for the address | |
999 | * @returns 0 on success, -1 on failure | |
1000 | */ | |
1001 | static | |
1002 | int so_info_lookup_cu_src_loc_inl(struct bt_dwarf_cu *cu, uint64_t addr, | |
1003 | struct source_location **src_loc) | |
1004 | { | |
1005 | int ret = 0, found = 0; | |
1006 | struct bt_dwarf_die *die = NULL; | |
1007 | struct source_location *_src_loc = NULL; | |
1008 | ||
1009 | if (!cu || !src_loc) { | |
1010 | goto error; | |
1011 | } | |
1012 | ||
1013 | die = bt_dwarf_die_create(cu); | |
1014 | if (!die) { | |
1015 | goto error; | |
1016 | } | |
1017 | ||
1018 | while (bt_dwarf_die_next(die) == 0) { | |
1019 | int tag; | |
1020 | ||
1021 | ret = bt_dwarf_die_get_tag(die, &tag); | |
1022 | if (ret) { | |
1023 | goto error; | |
1024 | } | |
1025 | ||
1026 | if (tag == DW_TAG_subprogram) { | |
1027 | int contains = 0; | |
1028 | ||
1029 | ret = bt_dwarf_die_contains_addr(die, addr, &contains); | |
1030 | if (ret) { | |
1031 | goto error; | |
1032 | } | |
1033 | ||
1034 | if (contains) { | |
1035 | /* | |
1036 | * Try to find an inlined subroutine | |
1037 | * child of this DIE containing addr. | |
1038 | */ | |
1039 | found = so_info_child_die_has_address( | |
1040 | die, addr); | |
1041 | goto end; | |
1042 | } | |
1043 | } | |
1044 | } | |
1045 | ||
1046 | end: | |
1047 | if (found) { | |
1048 | char *filename = NULL; | |
1049 | uint64_t line_no; | |
1050 | ||
1051 | _src_loc = g_new0(struct source_location, 1); | |
1052 | if (!_src_loc) { | |
1053 | goto error; | |
1054 | } | |
1055 | ||
1056 | ret = bt_dwarf_die_get_call_file(die, &filename); | |
1057 | if (ret) { | |
1058 | goto error; | |
1059 | } | |
1060 | ret = bt_dwarf_die_get_call_line(die, &line_no); | |
1061 | if (ret) { | |
1062 | free(filename); | |
1063 | goto error; | |
1064 | } | |
1065 | ||
1066 | _src_loc->filename = filename; | |
1067 | _src_loc->line_no = line_no; | |
1068 | *src_loc = _src_loc; | |
1069 | } | |
1070 | ||
1071 | bt_dwarf_die_destroy(die); | |
1072 | return 0; | |
1073 | ||
1074 | error: | |
1075 | source_location_destroy(_src_loc); | |
1076 | bt_dwarf_die_destroy(die); | |
1077 | return -1; | |
1078 | } | |
1079 | ||
1080 | /** | |
1081 | * Lookup the source location for a given address within a CU, | |
1082 | * assuming that it is contained within an inlined function. | |
1083 | * | |
1084 | * A source location can be found regardless of inlining status for | |
1085 | * this method, but in the case of an inlined function, the returned | |
1086 | * source location will point not to the callsite but rather to the | |
1087 | * definition site of the inline function. | |
1088 | * | |
1089 | * @param cu bt_dwarf_cu instance in which to look for the address | |
1090 | * @param addr The address for which to look for | |
1091 | * @param src_loc Out parameter, the source location (filename and | |
1092 | * line number) for the address | |
1093 | * @returns 0 on success, -1 on failure | |
1094 | */ | |
1095 | static | |
1096 | int so_info_lookup_cu_src_loc_no_inl(struct bt_dwarf_cu *cu, uint64_t addr, | |
1097 | struct source_location **src_loc) | |
1098 | { | |
1099 | struct source_location *_src_loc = NULL; | |
1100 | struct bt_dwarf_die *die = NULL; | |
1101 | const char *filename = NULL; | |
1102 | Dwarf_Line *line = NULL; | |
1103 | Dwarf_Addr line_addr; | |
1104 | int ret, line_no; | |
1105 | ||
1106 | if (!cu || !src_loc) { | |
1107 | goto error; | |
1108 | } | |
1109 | ||
1110 | die = bt_dwarf_die_create(cu); | |
1111 | if (!die) { | |
1112 | goto error; | |
1113 | } | |
1114 | ||
1115 | line = dwarf_getsrc_die(die->dwarf_die, addr); | |
1116 | if (!line) { | |
1117 | goto error; | |
1118 | } | |
1119 | ||
1120 | ret = dwarf_lineaddr(line, &line_addr); | |
1121 | if (ret) { | |
1122 | goto error; | |
1123 | } | |
1124 | ||
1125 | filename = dwarf_linesrc(line, NULL, NULL); | |
1126 | if (!filename) { | |
1127 | goto error; | |
1128 | } | |
1129 | ||
1130 | if (addr == line_addr) { | |
1131 | _src_loc = g_new0(struct source_location, 1); | |
1132 | if (!_src_loc) { | |
1133 | goto error; | |
1134 | } | |
1135 | ||
1136 | ret = dwarf_lineno(line, &line_no); | |
1137 | if (ret) { | |
1138 | goto error; | |
1139 | } | |
1140 | ||
1141 | _src_loc->line_no = line_no; | |
1142 | _src_loc->filename = strdup(filename); | |
1143 | } | |
1144 | ||
1145 | bt_dwarf_die_destroy(die); | |
1146 | ||
1147 | if (_src_loc) { | |
1148 | *src_loc = _src_loc; | |
1149 | } | |
1150 | ||
1151 | return 0; | |
1152 | ||
1153 | error: | |
1154 | source_location_destroy(_src_loc); | |
1155 | bt_dwarf_die_destroy(die); | |
1156 | return -1; | |
1157 | } | |
1158 | ||
1159 | /** | |
1160 | * Get the source location (file name and line number) for a given | |
1161 | * address within a compile unit (CU). | |
1162 | * | |
1163 | * On success, the out parameter `src_loc` is set if found. On | |
1164 | * failure, it remains unchanged. | |
1165 | * | |
1166 | * @param so bt_dwarf_cu instance for the compile unit which | |
1167 | * may contain the address | |
1168 | * @param addr Virtual memory address for which to find the | |
1169 | * source location | |
1170 | * @param src_loc Out parameter, the source location | |
1171 | * @returns 0 on success, -1 on failure | |
1172 | */ | |
1173 | static | |
1174 | int so_info_lookup_cu_src_loc(struct bt_dwarf_cu *cu, uint64_t addr, | |
1175 | struct source_location **src_loc) | |
1176 | { | |
1177 | int ret = 0; | |
1178 | struct source_location *_src_loc = NULL; | |
1179 | ||
1180 | if (!cu || !src_loc) { | |
1181 | goto error; | |
1182 | } | |
1183 | ||
1184 | ret = so_info_lookup_cu_src_loc_inl(cu, addr, &_src_loc); | |
1185 | if (ret) { | |
1186 | goto error; | |
1187 | } | |
1188 | ||
1189 | if (_src_loc) { | |
1190 | goto end; | |
1191 | } | |
1192 | ||
1193 | ret = so_info_lookup_cu_src_loc_no_inl(cu, addr, &_src_loc); | |
1194 | if (ret) { | |
1195 | goto error; | |
1196 | } | |
1197 | ||
1198 | if (_src_loc) { | |
1199 | goto end; | |
1200 | } | |
1201 | ||
1202 | end: | |
1203 | if (_src_loc) { | |
1204 | *src_loc = _src_loc; | |
1205 | } | |
1206 | ||
1207 | return 0; | |
1208 | ||
1209 | error: | |
1210 | source_location_destroy(_src_loc); | |
1211 | return -1; | |
1212 | } | |
1213 | ||
1214 | BT_HIDDEN | |
1215 | int so_info_lookup_source_location(struct so_info *so, uint64_t addr, | |
1216 | struct source_location **src_loc) | |
1217 | { | |
1218 | struct bt_dwarf_cu *cu = NULL; | |
1219 | struct source_location *_src_loc = NULL; | |
1220 | ||
1221 | if (!so || !src_loc) { | |
1222 | goto error; | |
1223 | } | |
1224 | ||
1225 | /* Set DWARF info if it hasn't been accessed yet. */ | |
1226 | if (!so->dwarf_info && !so->is_elf_only) { | |
1227 | if (so_info_set_dwarf_info(so)) { | |
1228 | /* Failed to set DWARF info. */ | |
1229 | so->is_elf_only = true; | |
1230 | } | |
1231 | } | |
1232 | ||
1233 | if (so->is_elf_only) { | |
1234 | /* We cannot lookup source location without DWARF info. */ | |
1235 | goto error; | |
1236 | } | |
1237 | ||
1238 | if (!so_info_has_address(so, addr)) { | |
1239 | goto error; | |
1240 | } | |
1241 | ||
1242 | /* | |
1243 | * Addresses in ELF and DWARF are relative to base address for | |
1244 | * PIC, so make the address argument relative too if needed. | |
1245 | */ | |
1246 | if (so->is_pic) { | |
1247 | addr -= so->low_addr; | |
1248 | } | |
1249 | ||
1250 | cu = bt_dwarf_cu_create(so->dwarf_info); | |
1251 | if (!cu) { | |
1252 | goto error; | |
1253 | } | |
1254 | ||
1255 | while (bt_dwarf_cu_next(cu) == 0) { | |
1256 | int ret; | |
1257 | ||
1258 | ret = so_info_lookup_cu_src_loc(cu, addr, &_src_loc); | |
1259 | if (ret) { | |
1260 | goto error; | |
1261 | } | |
1262 | ||
1263 | if (_src_loc) { | |
1264 | break; | |
1265 | } | |
1266 | } | |
1267 | ||
1268 | bt_dwarf_cu_destroy(cu); | |
1269 | if (_src_loc) { | |
1270 | *src_loc = _src_loc; | |
1271 | } | |
1272 | ||
1273 | return 0; | |
1274 | ||
1275 | error: | |
1276 | source_location_destroy(_src_loc); | |
1277 | bt_dwarf_cu_destroy(cu); | |
1278 | return -1; | |
1279 | } |