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 | ||
5cde0dc1 AB |
81 | if (opt_debug_info_target_prefix) { |
82 | so->elf_path = g_build_path("/", opt_debug_info_target_prefix, | |
83 | path, NULL); | |
84 | } else { | |
85 | so->elf_path = strdup(path); | |
86 | } | |
87 | ||
c40a57e5 AB |
88 | if (!so->elf_path) { |
89 | goto error; | |
90 | } | |
91 | ||
9f2b13ca | 92 | so->is_pic = is_pic; |
c40a57e5 AB |
93 | so->memsz = memsz; |
94 | so->low_addr = low_addr; | |
95 | so->high_addr = so->low_addr + so->memsz; | |
96 | ||
c40a57e5 AB |
97 | return so; |
98 | ||
99 | error: | |
c40a57e5 AB |
100 | so_info_destroy(so); |
101 | return NULL; | |
102 | } | |
103 | ||
104 | BT_HIDDEN | |
105 | void so_info_destroy(struct so_info *so) | |
106 | { | |
107 | if (!so) { | |
108 | return; | |
109 | } | |
110 | ||
111 | dwarf_end(so->dwarf_info); | |
112 | ||
113 | free(so->elf_path); | |
114 | free(so->dwarf_path); | |
115 | free(so->build_id); | |
116 | free(so->dbg_link_filename); | |
117 | ||
118 | elf_end(so->elf_file); | |
119 | ||
120 | close(so->elf_fd); | |
121 | close(so->dwarf_fd); | |
122 | ||
123 | g_free(so); | |
124 | } | |
125 | ||
49824faa | 126 | |
c40a57e5 AB |
127 | BT_HIDDEN |
128 | int so_info_set_build_id(struct so_info *so, uint8_t *build_id, | |
129 | size_t build_id_len) | |
130 | { | |
131 | if (!so || !build_id) { | |
132 | goto error; | |
133 | } | |
134 | ||
135 | so->build_id = malloc(build_id_len); | |
136 | if (!so->build_id) { | |
137 | goto error; | |
138 | } | |
139 | ||
140 | memcpy(so->build_id, build_id, build_id_len); | |
141 | so->build_id_len = build_id_len; | |
142 | ||
143 | /* | |
144 | * Reset the is_elf_only flag in case it had been set | |
145 | * previously, because we might find separate debug info using | |
146 | * the new build id information. | |
147 | */ | |
148 | so->is_elf_only = false; | |
149 | ||
150 | return 0; | |
151 | ||
152 | error: | |
153 | ||
154 | return -1; | |
155 | } | |
156 | ||
157 | BT_HIDDEN | |
158 | int so_info_set_debug_link(struct so_info *so, char *filename, uint32_t crc) | |
159 | { | |
160 | if (!so || !filename) { | |
161 | goto error; | |
162 | } | |
163 | ||
164 | so->dbg_link_filename = strdup(filename); | |
165 | if (!so->dbg_link_filename) { | |
166 | goto error; | |
167 | } | |
168 | ||
169 | so->dbg_link_crc = crc; | |
170 | ||
171 | /* | |
172 | * Reset the is_elf_only flag in case it had been set | |
173 | * previously, because we might find separate debug info using | |
174 | * the new build id information. | |
175 | */ | |
176 | so->is_elf_only = false; | |
177 | ||
178 | return 0; | |
179 | ||
180 | error: | |
181 | ||
182 | return -1; | |
183 | } | |
184 | ||
185 | /** | |
186 | * Tries to read DWARF info from the location given by path, and | |
187 | * attach it to the given so_info instance if it exists. | |
188 | * | |
189 | * @param so so_info instance for which to set DWARF info | |
190 | * @param path Presumed location of the DWARF info | |
191 | * @returns 0 on success, -1 on failure | |
192 | */ | |
193 | static | |
194 | int so_info_set_dwarf_info_from_path(struct so_info *so, char *path) | |
195 | { | |
196 | int fd = -1, ret = 0; | |
197 | struct bt_dwarf_cu *cu = NULL; | |
198 | Dwarf *dwarf_info = NULL; | |
199 | ||
200 | if (!so || !path) { | |
201 | goto error; | |
202 | } | |
203 | ||
204 | fd = open(path, O_RDONLY); | |
205 | if (fd < 0) { | |
206 | goto error; | |
207 | } | |
208 | ||
209 | dwarf_info = dwarf_begin(fd, DWARF_C_READ); | |
210 | if (!dwarf_info) { | |
211 | goto error; | |
212 | } | |
213 | ||
214 | /* | |
215 | * Check if the dwarf info has any CU. If not, the SO's object | |
216 | * file contains no DWARF info. | |
217 | */ | |
218 | cu = bt_dwarf_cu_create(dwarf_info); | |
219 | if (!cu) { | |
220 | goto error; | |
221 | } | |
222 | ||
223 | ret = bt_dwarf_cu_next(cu); | |
224 | if (ret) { | |
225 | goto error; | |
226 | } | |
227 | ||
228 | so->dwarf_fd = fd; | |
229 | so->dwarf_path = strdup(path); | |
230 | if (!so->dwarf_path) { | |
231 | goto error; | |
232 | } | |
233 | so->dwarf_info = dwarf_info; | |
234 | free(cu); | |
235 | ||
236 | return 0; | |
237 | ||
238 | error: | |
239 | close(fd); | |
240 | dwarf_end(dwarf_info); | |
241 | g_free(dwarf_info); | |
242 | free(cu); | |
243 | ||
244 | return -1; | |
245 | } | |
246 | ||
247 | /** | |
248 | * Try to set the dwarf_info for a given so_info instance via the | |
249 | * build ID method. | |
250 | * | |
251 | * @param so so_info instance for which to retrieve the | |
252 | * DWARF info via build ID | |
253 | * @returns 0 on success (i.e. dwarf_info set), -1 on failure | |
254 | */ | |
255 | static | |
256 | int so_info_set_dwarf_info_build_id(struct so_info *so) | |
257 | { | |
258 | int i = 0, ret = 0, dbg_dir_trailing_slash = 0; | |
259 | char *path = NULL, *build_id_file = NULL; | |
260 | const char *dbg_dir = NULL; | |
261 | size_t build_id_file_len, path_len; | |
262 | ||
263 | if (!so || !so->build_id) { | |
264 | goto error; | |
265 | } | |
266 | ||
05984e0c | 267 | dbg_dir = opt_debug_info_dir ? : DEFAULT_DEBUG_DIR; |
c40a57e5 AB |
268 | |
269 | dbg_dir_trailing_slash = dbg_dir[strlen(dbg_dir) - 1] == '/'; | |
270 | ||
271 | /* 2 characters per byte printed in hex, +2 for '/' and '\0' */ | |
272 | build_id_file_len = (2 * so->build_id_len) + 2; | |
273 | build_id_file = malloc(build_id_file_len); | |
274 | if (!build_id_file) { | |
275 | goto error; | |
276 | } | |
277 | ||
278 | snprintf(build_id_file, 4, "%02x/", so->build_id[0]); | |
279 | for (i = 1; i < so->build_id_len; ++i) { | |
280 | int path_idx = 3 + 2 * (i - 1); | |
281 | ||
282 | snprintf(&build_id_file[path_idx], 3, "%02x", so->build_id[i]); | |
283 | } | |
284 | ||
285 | path_len = strlen(dbg_dir) + strlen(BUILD_ID_SUBDIR) + | |
286 | strlen(build_id_file) + strlen(BUILD_ID_SUFFIX) + 1; | |
287 | if (!dbg_dir_trailing_slash) { | |
288 | path_len += 1; | |
289 | } | |
290 | ||
291 | path = malloc(path_len); | |
292 | if (!path) { | |
293 | goto error; | |
294 | } | |
295 | ||
296 | strcpy(path, dbg_dir); | |
297 | if (!dbg_dir_trailing_slash) { | |
298 | strcat(path, "/"); | |
299 | } | |
300 | strcat(path, BUILD_ID_SUBDIR); | |
301 | strcat(path, build_id_file); | |
302 | strcat(path, BUILD_ID_SUFFIX); | |
303 | ||
304 | ret = so_info_set_dwarf_info_from_path(so, path); | |
305 | if (ret) { | |
306 | goto error; | |
307 | } | |
308 | ||
309 | goto end; | |
310 | ||
311 | error: | |
312 | ret = -1; | |
313 | end: | |
314 | free(build_id_file); | |
315 | free(path); | |
316 | ||
317 | return ret; | |
318 | } | |
319 | ||
320 | /** | |
321 | * Tests whether the file located at path exists and has the expected | |
322 | * checksum. | |
323 | * | |
324 | * This predicate is used when looking up separate debug info via the | |
325 | * GNU debuglink method. The expected crc can be found .gnu_debuglink | |
326 | * section in the original ELF file, along with the filename for the | |
327 | * file containing the debug info. | |
328 | * | |
329 | * @param path Full path at which to look for the debug file | |
330 | * @param crc Expected checksum for the debug file | |
331 | * @returns 1 if the file exists and has the correct checksum, | |
332 | * 0 otherwise | |
333 | */ | |
334 | static | |
335 | int is_valid_debug_file(char *path, uint32_t crc) | |
336 | { | |
337 | int ret = 0, fd = -1; | |
338 | uint32_t _crc = 0; | |
339 | ||
340 | if (!path) { | |
341 | goto end; | |
342 | } | |
343 | ||
344 | fd = open(path, O_RDONLY); | |
345 | if (fd < 0) { | |
346 | goto end; | |
347 | } | |
348 | ||
349 | ret = crc32(fd, &_crc); | |
350 | if (ret) { | |
351 | ret = 0; | |
352 | goto end; | |
353 | } | |
354 | ||
355 | ret = (crc == _crc); | |
356 | ||
357 | end: | |
358 | close(fd); | |
359 | return ret; | |
360 | } | |
361 | ||
362 | /** | |
363 | * Try to set the dwarf_info for a given so_info instance via the | |
364 | * build ID method. | |
365 | * | |
366 | * @param so so_info instance for which to retrieve the | |
367 | * DWARF info via debug link | |
368 | * @returns 0 on success (i.e. dwarf_info set), -1 on failure | |
369 | */ | |
370 | static | |
371 | int so_info_set_dwarf_info_debug_link(struct so_info *so) | |
372 | { | |
373 | int ret = 0; | |
374 | const char *dbg_dir = NULL; | |
375 | char *dir_name = NULL, *so_dir = NULL, *path = NULL; | |
376 | size_t max_path_len = 0; | |
377 | ||
378 | if (!so || !so->dbg_link_filename) { | |
379 | goto error; | |
380 | } | |
381 | ||
05984e0c | 382 | dbg_dir = opt_debug_info_dir ? : DEFAULT_DEBUG_DIR; |
c40a57e5 AB |
383 | |
384 | dir_name = dirname(so->elf_path); | |
385 | if (!dir_name) { | |
386 | goto error; | |
387 | } | |
388 | ||
389 | /* so_dir is just dir_name with a trailing slash */ | |
390 | so_dir = malloc(strlen(dir_name) + 2); | |
391 | if (!so_dir) { | |
392 | goto error; | |
393 | } | |
394 | ||
395 | strcpy(so_dir, dir_name); | |
396 | strcat(so_dir, "/"); | |
397 | ||
398 | max_path_len = strlen(dbg_dir) + strlen(so_dir) + | |
399 | strlen(DEBUG_SUBDIR) + strlen(so->dbg_link_filename) | |
400 | + 1; | |
401 | path = malloc(max_path_len); | |
402 | if (!path) { | |
403 | goto error; | |
404 | } | |
405 | ||
406 | /* First look in the SO's dir */ | |
407 | strcpy(path, so_dir); | |
408 | strcat(path, so->dbg_link_filename); | |
409 | ||
410 | if (is_valid_debug_file(path, so->dbg_link_crc)) { | |
411 | goto found; | |
412 | } | |
413 | ||
414 | /* If not found, look in .debug subdir */ | |
415 | strcpy(path, so_dir); | |
416 | strcat(path, DEBUG_SUBDIR); | |
417 | strcat(path, so->dbg_link_filename); | |
418 | ||
419 | if (is_valid_debug_file(path, so->dbg_link_crc)) { | |
420 | goto found; | |
421 | } | |
422 | ||
423 | /* Lastly, look under the global debug directory */ | |
424 | strcpy(path, dbg_dir); | |
425 | strcat(path, so_dir); | |
426 | strcat(path, so->dbg_link_filename); | |
427 | ||
428 | if (is_valid_debug_file(path, so->dbg_link_crc)) { | |
429 | goto found; | |
430 | } | |
431 | ||
432 | error: | |
433 | ret = -1; | |
434 | end: | |
435 | free(path); | |
436 | free(so_dir); | |
437 | ||
438 | return ret; | |
439 | ||
440 | found: | |
441 | ret = so_info_set_dwarf_info_from_path(so, path); | |
442 | if (ret) { | |
443 | goto error; | |
444 | } | |
445 | ||
446 | goto end; | |
447 | } | |
448 | ||
449 | /** | |
450 | * Initialize the DWARF info for a given executable. | |
451 | * | |
452 | * @param so so_info instance | |
453 | * @returns 0 on success, -1 on failure | |
454 | */ | |
455 | static | |
456 | int so_info_set_dwarf_info(struct so_info *so) | |
457 | { | |
458 | int ret = 0; | |
459 | ||
460 | if (!so) { | |
461 | goto error; | |
462 | } | |
463 | ||
464 | /* First try to set the DWARF info from the ELF file */ | |
465 | ret = so_info_set_dwarf_info_from_path(so, so->elf_path); | |
466 | if (!ret) { | |
467 | goto end; | |
468 | } | |
469 | ||
470 | /* | |
471 | * If that fails, try to find separate debug info via build ID | |
472 | * and debug link. | |
473 | */ | |
474 | ret = so_info_set_dwarf_info_build_id(so); | |
475 | if (!ret) { | |
476 | goto end; | |
477 | } | |
478 | ||
479 | ret = so_info_set_dwarf_info_debug_link(so); | |
480 | if (!ret) { | |
481 | goto end; | |
482 | } | |
483 | ||
484 | error: | |
485 | ret = -1; | |
486 | end: | |
487 | return ret; | |
488 | } | |
489 | ||
49824faa AB |
490 | /** |
491 | * Initialize the ELF file for a given executable. | |
492 | * | |
493 | * @param so so_info instance | |
494 | * @returns 0 on success, -1 on failure | |
495 | */ | |
496 | static | |
497 | int so_info_set_elf_file(struct so_info *so) | |
498 | { | |
499 | int elf_fd; | |
f99dc219 | 500 | Elf *elf_file = NULL; |
49824faa AB |
501 | |
502 | if (!so) { | |
503 | goto error; | |
504 | } | |
505 | ||
506 | elf_fd = open(so->elf_path, O_RDONLY); | |
507 | if (elf_fd < 0) { | |
508 | fprintf(stderr, "Failed to open %s\n", so->elf_path); | |
509 | goto error; | |
510 | } | |
511 | ||
512 | elf_file = elf_begin(elf_fd, ELF_C_READ, NULL); | |
513 | if (!elf_file) { | |
514 | fprintf(stderr, "elf_begin failed: %s\n", elf_errmsg(-1)); | |
515 | goto error; | |
516 | } | |
517 | ||
518 | if (elf_kind(elf_file) != ELF_K_ELF) { | |
519 | fprintf(stderr, "Error: %s is not an ELF object\n", | |
520 | so->elf_path); | |
521 | goto error; | |
522 | } | |
523 | ||
524 | so->elf_fd = elf_fd; | |
525 | so->elf_file = elf_file; | |
526 | return 0; | |
527 | ||
528 | error: | |
529 | close(elf_fd); | |
530 | elf_end(elf_file); | |
531 | return -1; | |
532 | } | |
533 | ||
534 | ||
c40a57e5 AB |
535 | BT_HIDDEN |
536 | void source_location_destroy(struct source_location *src_loc) | |
537 | { | |
538 | if (!src_loc) { | |
539 | return; | |
540 | } | |
541 | ||
542 | free(src_loc->filename); | |
543 | g_free(src_loc); | |
544 | } | |
d2ac1099 AB |
545 | /** |
546 | * Append a string representation of an address offset to an existing | |
547 | * string. | |
548 | * | |
549 | * On success, the out parameter `result` will contain the base string | |
550 | * followed by the offset string of the form "+0x1234". On failure, | |
551 | * `result` remains unchanged. | |
552 | * | |
553 | * @param base_str The string to which to append an offset string | |
554 | * @param low_addr The lower virtual memory address, the base from | |
555 | * which the offset is computed | |
556 | * @param high_addr The higher virtual memory address | |
557 | * @param result Out parameter, the base string followed by the | |
558 | * offset string | |
559 | * @returns 0 on success, -1 on failure | |
560 | */ | |
561 | static | |
562 | int so_info_append_offset_str(const char *base_str, uint64_t low_addr, | |
563 | uint64_t high_addr, char **result) | |
564 | { | |
565 | int ret; | |
566 | uint64_t offset; | |
567 | char *_result = NULL; | |
568 | char offset_str[ADDR_STR_LEN]; | |
569 | ||
570 | if (!base_str || !result) { | |
571 | goto error; | |
572 | } | |
573 | ||
574 | offset = high_addr - low_addr; | |
575 | ||
576 | _result = malloc(strlen(base_str) + ADDR_STR_LEN); | |
577 | if (!_result) { | |
578 | goto error; | |
579 | } | |
580 | ||
581 | ret = snprintf(offset_str, ADDR_STR_LEN, "+%#0" PRIx64, offset); | |
582 | if (ret < 0) { | |
583 | goto error; | |
584 | } | |
585 | strcpy(_result, base_str); | |
586 | strcat(_result, offset_str); | |
587 | *result = _result; | |
588 | ||
589 | return 0; | |
590 | ||
591 | error: | |
592 | free(_result); | |
593 | return -1; | |
594 | } | |
c40a57e5 AB |
595 | |
596 | /** | |
597 | * Try to find the symbol closest to an address within a given ELF | |
598 | * section. | |
599 | * | |
600 | * Only function symbols are taken into account. The symbol's address | |
601 | * must precede `addr`. A symbol with a closer address might exist | |
602 | * after `addr` but is irrelevant because it cannot encompass `addr`. | |
603 | * | |
604 | * On success, if found, the out parameters `sym` and `shdr` are | |
605 | * set. On failure or if none are found, they remain unchanged. | |
606 | * | |
607 | * @param scn ELF section in which to look for the address | |
608 | * @param addr Virtual memory address for which to find the | |
609 | * nearest function symbol | |
610 | * @param sym Out parameter, the nearest function symbol | |
611 | * @param shdr Out parameter, the section header for scn | |
612 | * @returns 0 on success, -1 on failure | |
613 | */ | |
614 | static | |
615 | int so_info_get_nearest_symbol_from_section(Elf_Scn *scn, uint64_t addr, | |
616 | GElf_Sym **sym, GElf_Shdr **shdr) | |
617 | { | |
618 | int i; | |
619 | size_t symbol_count; | |
620 | Elf_Data *data = NULL; | |
621 | GElf_Shdr *_shdr = NULL; | |
622 | GElf_Sym *nearest_sym = NULL; | |
623 | ||
624 | if (!scn || !sym || !shdr) { | |
625 | goto error; | |
626 | } | |
627 | ||
628 | _shdr = g_new0(GElf_Shdr, 1); | |
629 | if (!_shdr) { | |
630 | goto error; | |
631 | } | |
632 | ||
633 | _shdr = gelf_getshdr(scn, _shdr); | |
634 | if (!_shdr) { | |
635 | goto error; | |
636 | } | |
637 | ||
638 | if (_shdr->sh_type != SHT_SYMTAB) { | |
639 | /* | |
640 | * We are only interested in symbol table (symtab) | |
641 | * sections, skip this one. | |
642 | */ | |
643 | goto end; | |
644 | } | |
645 | ||
646 | data = elf_getdata(scn, NULL); | |
647 | if (!data) { | |
648 | goto error; | |
649 | } | |
650 | ||
651 | symbol_count = _shdr->sh_size / _shdr->sh_entsize; | |
652 | ||
653 | for (i = 0; i < symbol_count; ++i) { | |
654 | GElf_Sym *cur_sym = NULL; | |
655 | ||
656 | cur_sym = g_new0(GElf_Sym, 1); | |
657 | if (!cur_sym) { | |
658 | goto error; | |
659 | } | |
660 | cur_sym = gelf_getsym(data, i, cur_sym); | |
661 | if (!cur_sym) { | |
662 | goto error; | |
663 | } | |
664 | if (GELF_ST_TYPE(cur_sym->st_info) != STT_FUNC) { | |
665 | /* We're only interested in the functions. */ | |
666 | g_free(cur_sym); | |
667 | continue; | |
668 | } | |
669 | ||
670 | if (cur_sym->st_value <= addr && | |
671 | (!nearest_sym || | |
672 | cur_sym->st_value > nearest_sym->st_value)) { | |
673 | g_free(nearest_sym); | |
674 | nearest_sym = cur_sym; | |
675 | } else { | |
676 | g_free(cur_sym); | |
677 | } | |
678 | } | |
679 | ||
680 | end: | |
681 | if (nearest_sym) { | |
682 | *sym = nearest_sym; | |
683 | *shdr = _shdr; | |
684 | } else { | |
685 | g_free(_shdr); | |
686 | } | |
687 | ||
688 | return 0; | |
689 | ||
690 | error: | |
691 | g_free(nearest_sym); | |
692 | g_free(_shdr); | |
693 | return -1; | |
694 | } | |
695 | ||
696 | /** | |
697 | * Get the name of the function containing a given address within an | |
698 | * executable using ELF symbols. | |
699 | * | |
700 | * The function name is in fact the name of the nearest ELF symbol, | |
701 | * followed by the offset in bytes between the address and the symbol | |
702 | * (in hex), separated by a '+' character. | |
703 | * | |
704 | * If found, the out parameter `func_name` is set on success. On failure, | |
705 | * it remains unchanged. | |
706 | * | |
707 | * @param so so_info instance for the executable containing | |
708 | * the address | |
709 | * @param addr Virtual memory address for which to find the | |
710 | * function name | |
711 | * @param func_name Out parameter, the function name | |
712 | * @returns 0 on success, -1 on failure | |
713 | */ | |
714 | static | |
715 | int so_info_lookup_elf_function_name(struct so_info *so, uint64_t addr, | |
716 | char **func_name) | |
717 | { | |
718 | /* | |
719 | * TODO (possible optimisation): if an ELF has no symtab | |
720 | * section, it has been stripped. Therefore, it would be wise | |
721 | * to store a flag indicating the stripped status after the | |
722 | * first iteration to prevent subsequent ones. | |
723 | */ | |
724 | int ret = 0; | |
725 | Elf_Scn *scn = NULL; | |
726 | GElf_Sym *sym = NULL; | |
727 | GElf_Shdr *shdr = NULL; | |
728 | char *sym_name = NULL; | |
c40a57e5 | 729 | |
49824faa AB |
730 | /* Set ELF file if it hasn't been accessed yet. */ |
731 | if (!so->elf_file) { | |
732 | ret = so_info_set_elf_file(so); | |
733 | if (ret) { | |
734 | /* Failed to set ELF file. */ | |
735 | goto error; | |
736 | } | |
737 | } | |
738 | ||
c40a57e5 AB |
739 | scn = elf_nextscn(so->elf_file, scn); |
740 | if (!scn) { | |
741 | goto error; | |
742 | } | |
743 | ||
744 | while (scn && !sym) { | |
745 | ret = so_info_get_nearest_symbol_from_section( | |
746 | scn, addr, &sym, &shdr); | |
747 | if (ret) { | |
748 | goto error; | |
749 | } | |
750 | ||
751 | scn = elf_nextscn(so->elf_file, scn); | |
752 | } | |
753 | ||
754 | if (sym) { | |
755 | sym_name = elf_strptr(so->elf_file, shdr->sh_link, | |
756 | sym->st_name); | |
757 | if (!sym_name) { | |
758 | goto error; | |
759 | } | |
760 | ||
d2ac1099 AB |
761 | ret = so_info_append_offset_str(sym_name, sym->st_value, addr, |
762 | func_name); | |
763 | if (ret) { | |
c40a57e5 AB |
764 | goto error; |
765 | } | |
c40a57e5 AB |
766 | } |
767 | ||
768 | g_free(shdr); | |
769 | g_free(sym); | |
770 | return 0; | |
771 | ||
772 | error: | |
773 | g_free(shdr); | |
774 | g_free(sym); | |
c40a57e5 AB |
775 | return -1; |
776 | } | |
777 | ||
778 | /** | |
779 | * Get the name of the function containing a given address within a | |
780 | * given compile unit (CU). | |
781 | * | |
782 | * If found, the out parameter `func_name` is set on success. On | |
783 | * failure, it remains unchanged. | |
784 | * | |
785 | * @param cu bt_dwarf_cu instance which may contain the address | |
786 | * @param addr Virtual memory address for which to find the | |
787 | * function name | |
788 | * @param func_name Out parameter, the function name | |
789 | * @returns 0 on success, -1 on failure | |
790 | */ | |
791 | static | |
792 | int so_info_lookup_cu_function_name(struct bt_dwarf_cu *cu, uint64_t addr, | |
793 | char **func_name) | |
794 | { | |
795 | int ret = 0, found = 0; | |
c40a57e5 AB |
796 | struct bt_dwarf_die *die = NULL; |
797 | ||
798 | if (!cu || !func_name) { | |
799 | goto error; | |
800 | } | |
801 | ||
802 | die = bt_dwarf_die_create(cu); | |
803 | if (!die) { | |
804 | goto error; | |
805 | } | |
806 | ||
807 | while (bt_dwarf_die_next(die) == 0) { | |
808 | int tag; | |
809 | ||
810 | ret = bt_dwarf_die_get_tag(die, &tag); | |
811 | if (ret) { | |
812 | goto error; | |
813 | } | |
814 | ||
815 | if (tag == DW_TAG_subprogram) { | |
816 | ret = bt_dwarf_die_contains_addr(die, addr, &found); | |
817 | if (ret) { | |
818 | goto error; | |
819 | } | |
820 | ||
821 | if (found) { | |
822 | break; | |
823 | } | |
824 | } | |
825 | } | |
826 | ||
827 | if (found) { | |
d2ac1099 AB |
828 | uint64_t low_addr = 0; |
829 | char *die_name = NULL; | |
830 | ||
7935ee7a AB |
831 | ret = bt_dwarf_die_get_name(die, &die_name); |
832 | if (ret) { | |
833 | goto error; | |
834 | } | |
835 | ||
836 | ret = dwarf_lowpc(die->dwarf_die, &low_addr); | |
c40a57e5 AB |
837 | if (ret) { |
838 | goto error; | |
839 | } | |
840 | ||
d2ac1099 AB |
841 | ret = so_info_append_offset_str(die_name, low_addr, addr, |
842 | func_name); | |
843 | if (ret) { | |
7935ee7a AB |
844 | goto error; |
845 | } | |
c40a57e5 AB |
846 | } |
847 | ||
848 | bt_dwarf_die_destroy(die); | |
849 | return 0; | |
850 | ||
851 | error: | |
852 | bt_dwarf_die_destroy(die); | |
853 | return -1; | |
854 | } | |
855 | ||
856 | /** | |
857 | * Get the name of the function containing a given address within an | |
858 | * executable using DWARF debug info. | |
859 | * | |
860 | * If found, the out parameter `func_name` is set on success. On | |
861 | * failure, it remains unchanged. | |
862 | * | |
863 | * @param so so_info instance for the executable containing | |
864 | * the address | |
865 | * @param addr Virtual memory address for which to find the | |
866 | * function name | |
867 | * @param func_name Out parameter, the function name | |
868 | * @returns 0 on success, -1 on failure | |
869 | */ | |
870 | static | |
871 | int so_info_lookup_dwarf_function_name(struct so_info *so, uint64_t addr, | |
872 | char **func_name) | |
873 | { | |
874 | int ret = 0; | |
875 | char *_func_name = NULL; | |
876 | struct bt_dwarf_cu *cu = NULL; | |
877 | ||
878 | if (!so || !func_name) { | |
879 | goto error; | |
880 | } | |
881 | ||
882 | cu = bt_dwarf_cu_create(so->dwarf_info); | |
883 | if (!cu) { | |
884 | goto error; | |
885 | } | |
886 | ||
887 | while (bt_dwarf_cu_next(cu) == 0) { | |
888 | ret = so_info_lookup_cu_function_name(cu, addr, &_func_name); | |
889 | if (ret) { | |
890 | goto error; | |
891 | } | |
892 | ||
893 | if (_func_name) { | |
894 | break; | |
895 | } | |
896 | } | |
897 | ||
898 | if (_func_name) { | |
899 | *func_name = _func_name; | |
900 | } | |
901 | ||
902 | bt_dwarf_cu_destroy(cu); | |
903 | return 0; | |
904 | ||
905 | error: | |
906 | bt_dwarf_cu_destroy(cu); | |
907 | return -1; | |
908 | } | |
909 | ||
910 | BT_HIDDEN | |
36ae9941 | 911 | int so_info_lookup_function_name(struct so_info *so, uint64_t addr, |
c40a57e5 AB |
912 | char **func_name) |
913 | { | |
914 | int ret = 0; | |
915 | char *_func_name = NULL; | |
916 | ||
917 | if (!so || !func_name) { | |
918 | goto error; | |
919 | } | |
920 | ||
921 | /* Set DWARF info if it hasn't been accessed yet. */ | |
922 | if (!so->dwarf_info && !so->is_elf_only) { | |
923 | ret = so_info_set_dwarf_info(so); | |
924 | if (ret) { | |
925 | /* Failed to set DWARF info, fallback to ELF. */ | |
926 | so->is_elf_only = true; | |
927 | } | |
928 | } | |
929 | ||
36ae9941 | 930 | if (!so_info_has_address(so, addr)) { |
c40a57e5 AB |
931 | goto error; |
932 | } | |
933 | ||
934 | /* | |
935 | * Addresses in ELF and DWARF are relative to base address for | |
936 | * PIC, so make the address argument relative too if needed. | |
937 | */ | |
36ae9941 AB |
938 | if (so->is_pic) { |
939 | addr -= so->low_addr; | |
940 | } | |
941 | ||
c40a57e5 | 942 | if (so->is_elf_only) { |
36ae9941 | 943 | ret = so_info_lookup_elf_function_name(so, addr, &_func_name); |
c40a57e5 | 944 | } else { |
36ae9941 | 945 | ret = so_info_lookup_dwarf_function_name(so, addr, &_func_name); |
c40a57e5 AB |
946 | } |
947 | ||
36ae9941 | 948 | if (ret || !_func_name) { |
c40a57e5 AB |
949 | goto error; |
950 | } | |
951 | ||
36ae9941 AB |
952 | *func_name = _func_name; |
953 | return 0; | |
55cd033d | 954 | |
36ae9941 AB |
955 | error: |
956 | return -1; | |
957 | } | |
958 | ||
959 | BT_HIDDEN | |
960 | int so_info_get_bin_loc(struct so_info *so, uint64_t addr, char **bin_loc) | |
961 | { | |
962 | int ret = 0; | |
963 | char *_bin_loc = NULL; | |
964 | ||
965 | if (!so || !bin_loc) { | |
966 | goto error; | |
c40a57e5 AB |
967 | } |
968 | ||
36ae9941 AB |
969 | if (so->is_pic) { |
970 | addr -= so->low_addr; | |
971 | ret = asprintf(&_bin_loc, "+%#0" PRIx64, addr); | |
972 | } else { | |
973 | ret = asprintf(&_bin_loc, "@%#0" PRIx64, addr); | |
974 | } | |
975 | ||
976 | if (ret == -1 || !_bin_loc) { | |
977 | goto error; | |
978 | } | |
979 | ||
980 | *bin_loc = _bin_loc; | |
c40a57e5 AB |
981 | return 0; |
982 | ||
983 | error: | |
984 | return -1; | |
985 | } | |
986 | ||
987 | /** | |
988 | * Predicate used to determine whether the children of a given DIE | |
989 | * contain a specific address. | |
990 | * | |
991 | * More specifically, the parameter `die` is expected to be a | |
992 | * subprogram (function) DIE, and this predicate tells whether any | |
993 | * subroutines are inlined within this function and would contain | |
994 | * `addr`. | |
995 | * | |
996 | * Do note that this function advances the position of `die`. If the | |
997 | * address is found within one of its children, `die` will be pointing | |
998 | * to that child upon returning from the function, allowing to extract | |
999 | * the information deemed necessary. | |
1000 | * | |
1001 | * @param die The parent DIE in whose children the address will be | |
1002 | * looked for | |
1003 | * @param addr The address for which to look for in the DIEs | |
1004 | * @returns Returns 1 if the address was found, 0 if not | |
1005 | */ | |
1006 | static | |
1007 | int so_info_child_die_has_address(struct bt_dwarf_die *die, uint64_t addr) | |
1008 | { | |
1009 | int ret = 0, contains = 0; | |
1010 | ||
1011 | if (!die) { | |
1012 | goto error; | |
1013 | } | |
1014 | ||
1015 | ret = bt_dwarf_die_child(die); | |
1016 | if (ret) { | |
1017 | goto error; | |
1018 | } | |
1019 | ||
1020 | do { | |
1021 | int tag; | |
1022 | ||
1023 | ret = bt_dwarf_die_get_tag(die, &tag); | |
1024 | if (ret) { | |
1025 | goto error; | |
1026 | } | |
1027 | ||
1028 | if (tag == DW_TAG_inlined_subroutine) { | |
1029 | ret = bt_dwarf_die_contains_addr(die, addr, &contains); | |
1030 | if (ret) { | |
1031 | goto error; | |
1032 | } | |
1033 | ||
1034 | if (contains) { | |
1035 | ret = 1; | |
1036 | goto end; | |
1037 | } | |
1038 | } | |
1039 | } while (bt_dwarf_die_next(die) == 0); | |
1040 | ||
1041 | end: | |
1042 | return ret; | |
1043 | ||
1044 | error: | |
1045 | ret = 0; | |
1046 | goto end; | |
1047 | } | |
1048 | ||
1049 | /** | |
1050 | * Lookup the source location for a given address within a CU, making | |
1051 | * the assumption that it is contained within an inline routine in a | |
1052 | * function. | |
1053 | * | |
1054 | * @param cu bt_dwarf_cu instance in which to look for the address | |
1055 | * @param addr The address for which to look for | |
1056 | * @param src_loc Out parameter, the source location (filename and | |
1057 | * line number) for the address | |
1058 | * @returns 0 on success, -1 on failure | |
1059 | */ | |
1060 | static | |
1061 | int so_info_lookup_cu_src_loc_inl(struct bt_dwarf_cu *cu, uint64_t addr, | |
1062 | struct source_location **src_loc) | |
1063 | { | |
1064 | int ret = 0, found = 0; | |
1065 | struct bt_dwarf_die *die = NULL; | |
1066 | struct source_location *_src_loc = NULL; | |
1067 | ||
1068 | if (!cu || !src_loc) { | |
1069 | goto error; | |
1070 | } | |
1071 | ||
1072 | die = bt_dwarf_die_create(cu); | |
1073 | if (!die) { | |
1074 | goto error; | |
1075 | } | |
1076 | ||
1077 | while (bt_dwarf_die_next(die) == 0) { | |
1078 | int tag; | |
1079 | ||
1080 | ret = bt_dwarf_die_get_tag(die, &tag); | |
1081 | if (ret) { | |
1082 | goto error; | |
1083 | } | |
1084 | ||
1085 | if (tag == DW_TAG_subprogram) { | |
1086 | int contains = 0; | |
1087 | ||
1088 | ret = bt_dwarf_die_contains_addr(die, addr, &contains); | |
1089 | if (ret) { | |
1090 | goto error; | |
1091 | } | |
1092 | ||
1093 | if (contains) { | |
1094 | /* | |
1095 | * Try to find an inlined subroutine | |
1096 | * child of this DIE containing addr. | |
1097 | */ | |
1098 | found = so_info_child_die_has_address( | |
1099 | die, addr); | |
1100 | goto end; | |
1101 | } | |
1102 | } | |
1103 | } | |
1104 | ||
1105 | end: | |
1106 | if (found) { | |
1107 | char *filename = NULL; | |
1108 | uint64_t line_no; | |
1109 | ||
1110 | _src_loc = g_new0(struct source_location, 1); | |
1111 | if (!_src_loc) { | |
1112 | goto error; | |
1113 | } | |
1114 | ||
1115 | ret = bt_dwarf_die_get_call_file(die, &filename); | |
1116 | if (ret) { | |
1117 | goto error; | |
1118 | } | |
1119 | ret = bt_dwarf_die_get_call_line(die, &line_no); | |
1120 | if (ret) { | |
1121 | free(filename); | |
1122 | goto error; | |
1123 | } | |
1124 | ||
1125 | _src_loc->filename = filename; | |
1126 | _src_loc->line_no = line_no; | |
1127 | *src_loc = _src_loc; | |
1128 | } | |
1129 | ||
1130 | bt_dwarf_die_destroy(die); | |
1131 | return 0; | |
1132 | ||
1133 | error: | |
1134 | source_location_destroy(_src_loc); | |
1135 | bt_dwarf_die_destroy(die); | |
1136 | return -1; | |
1137 | } | |
1138 | ||
1139 | /** | |
1140 | * Lookup the source location for a given address within a CU, | |
1141 | * assuming that it is contained within an inlined function. | |
1142 | * | |
1143 | * A source location can be found regardless of inlining status for | |
1144 | * this method, but in the case of an inlined function, the returned | |
1145 | * source location will point not to the callsite but rather to the | |
1146 | * definition site of the inline function. | |
1147 | * | |
1148 | * @param cu bt_dwarf_cu instance in which to look for the address | |
1149 | * @param addr The address for which to look for | |
1150 | * @param src_loc Out parameter, the source location (filename and | |
1151 | * line number) for the address | |
1152 | * @returns 0 on success, -1 on failure | |
1153 | */ | |
1154 | static | |
1155 | int so_info_lookup_cu_src_loc_no_inl(struct bt_dwarf_cu *cu, uint64_t addr, | |
1156 | struct source_location **src_loc) | |
1157 | { | |
1158 | struct source_location *_src_loc = NULL; | |
1159 | struct bt_dwarf_die *die = NULL; | |
1160 | const char *filename = NULL; | |
1161 | Dwarf_Line *line = NULL; | |
1162 | Dwarf_Addr line_addr; | |
1163 | int ret, line_no; | |
1164 | ||
1165 | if (!cu || !src_loc) { | |
1166 | goto error; | |
1167 | } | |
1168 | ||
1169 | die = bt_dwarf_die_create(cu); | |
1170 | if (!die) { | |
1171 | goto error; | |
1172 | } | |
1173 | ||
1174 | line = dwarf_getsrc_die(die->dwarf_die, addr); | |
1175 | if (!line) { | |
1176 | goto error; | |
1177 | } | |
1178 | ||
1179 | ret = dwarf_lineaddr(line, &line_addr); | |
1180 | if (ret) { | |
1181 | goto error; | |
1182 | } | |
1183 | ||
1184 | filename = dwarf_linesrc(line, NULL, NULL); | |
1185 | if (!filename) { | |
1186 | goto error; | |
1187 | } | |
1188 | ||
1189 | if (addr == line_addr) { | |
1190 | _src_loc = g_new0(struct source_location, 1); | |
1191 | if (!_src_loc) { | |
1192 | goto error; | |
1193 | } | |
1194 | ||
1195 | ret = dwarf_lineno(line, &line_no); | |
1196 | if (ret) { | |
1197 | goto error; | |
1198 | } | |
1199 | ||
1200 | _src_loc->line_no = line_no; | |
1201 | _src_loc->filename = strdup(filename); | |
1202 | } | |
1203 | ||
1204 | bt_dwarf_die_destroy(die); | |
1205 | ||
1206 | if (_src_loc) { | |
1207 | *src_loc = _src_loc; | |
1208 | } | |
1209 | ||
1210 | return 0; | |
1211 | ||
1212 | error: | |
1213 | source_location_destroy(_src_loc); | |
1214 | bt_dwarf_die_destroy(die); | |
1215 | return -1; | |
1216 | } | |
1217 | ||
1218 | /** | |
1219 | * Get the source location (file name and line number) for a given | |
1220 | * address within a compile unit (CU). | |
1221 | * | |
1222 | * On success, the out parameter `src_loc` is set if found. On | |
1223 | * failure, it remains unchanged. | |
1224 | * | |
1225 | * @param so bt_dwarf_cu instance for the compile unit which | |
1226 | * may contain the address | |
1227 | * @param addr Virtual memory address for which to find the | |
1228 | * source location | |
1229 | * @param src_loc Out parameter, the source location | |
1230 | * @returns 0 on success, -1 on failure | |
1231 | */ | |
1232 | static | |
1233 | int so_info_lookup_cu_src_loc(struct bt_dwarf_cu *cu, uint64_t addr, | |
1234 | struct source_location **src_loc) | |
1235 | { | |
1236 | int ret = 0; | |
1237 | struct source_location *_src_loc = NULL; | |
1238 | ||
1239 | if (!cu || !src_loc) { | |
1240 | goto error; | |
1241 | } | |
1242 | ||
1243 | ret = so_info_lookup_cu_src_loc_inl(cu, addr, &_src_loc); | |
1244 | if (ret) { | |
1245 | goto error; | |
1246 | } | |
1247 | ||
1248 | if (_src_loc) { | |
1249 | goto end; | |
1250 | } | |
1251 | ||
1252 | ret = so_info_lookup_cu_src_loc_no_inl(cu, addr, &_src_loc); | |
1253 | if (ret) { | |
1254 | goto error; | |
1255 | } | |
1256 | ||
1257 | if (_src_loc) { | |
1258 | goto end; | |
1259 | } | |
1260 | ||
1261 | end: | |
1262 | if (_src_loc) { | |
1263 | *src_loc = _src_loc; | |
1264 | } | |
1265 | ||
1266 | return 0; | |
1267 | ||
1268 | error: | |
1269 | source_location_destroy(_src_loc); | |
1270 | return -1; | |
1271 | } | |
1272 | ||
1273 | BT_HIDDEN | |
1274 | int so_info_lookup_source_location(struct so_info *so, uint64_t addr, | |
1275 | struct source_location **src_loc) | |
1276 | { | |
1277 | struct bt_dwarf_cu *cu = NULL; | |
1278 | struct source_location *_src_loc = NULL; | |
1279 | ||
1280 | if (!so || !src_loc) { | |
1281 | goto error; | |
1282 | } | |
1283 | ||
1284 | /* Set DWARF info if it hasn't been accessed yet. */ | |
1285 | if (!so->dwarf_info && !so->is_elf_only) { | |
1286 | if (so_info_set_dwarf_info(so)) { | |
1287 | /* Failed to set DWARF info. */ | |
1288 | so->is_elf_only = true; | |
1289 | } | |
1290 | } | |
1291 | ||
1292 | if (so->is_elf_only) { | |
1293 | /* We cannot lookup source location without DWARF info. */ | |
1294 | goto error; | |
1295 | } | |
1296 | ||
1297 | if (!so_info_has_address(so, addr)) { | |
1298 | goto error; | |
1299 | } | |
1300 | ||
1301 | /* | |
1302 | * Addresses in ELF and DWARF are relative to base address for | |
1303 | * PIC, so make the address argument relative too if needed. | |
1304 | */ | |
1305 | if (so->is_pic) { | |
1306 | addr -= so->low_addr; | |
1307 | } | |
1308 | ||
1309 | cu = bt_dwarf_cu_create(so->dwarf_info); | |
1310 | if (!cu) { | |
1311 | goto error; | |
1312 | } | |
1313 | ||
1314 | while (bt_dwarf_cu_next(cu) == 0) { | |
1315 | int ret; | |
1316 | ||
1317 | ret = so_info_lookup_cu_src_loc(cu, addr, &_src_loc); | |
1318 | if (ret) { | |
1319 | goto error; | |
1320 | } | |
1321 | ||
1322 | if (_src_loc) { | |
1323 | break; | |
1324 | } | |
1325 | } | |
1326 | ||
1327 | bt_dwarf_cu_destroy(cu); | |
1328 | if (_src_loc) { | |
1329 | *src_loc = _src_loc; | |
1330 | } | |
1331 | ||
1332 | return 0; | |
1333 | ||
1334 | error: | |
1335 | source_location_destroy(_src_loc); | |
1336 | bt_dwarf_cu_destroy(cu); | |
1337 | return -1; | |
1338 | } |