Commit | Line | Data |
---|---|---|
12a0b67d | 1 | /* Symbol, variable and name lookup. |
250d07de | 2 | Copyright (C) 2019-2021 Free Software Foundation, Inc. |
47d546f4 NA |
3 | |
4 | This file is part of libctf. | |
5 | ||
6 | libctf is free software; you can redistribute it and/or modify it under | |
7 | the terms of the GNU General Public License as published by the Free | |
8 | Software Foundation; either version 3, or (at your option) any later | |
9 | version. | |
10 | ||
11 | This program is distributed in the hope that it will be useful, but | |
12 | WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. | |
14 | See the GNU General Public License for more details. | |
15 | ||
16 | You should have received a copy of the GNU General Public License | |
17 | along with this program; see the file COPYING. If not see | |
18 | <http://www.gnu.org/licenses/>. */ | |
19 | ||
20 | #include <ctf-impl.h> | |
21 | #include <elf.h> | |
22 | #include <string.h> | |
1136c379 | 23 | #include <assert.h> |
47d546f4 | 24 | |
abe4ca69 NA |
25 | /* Grow the pptrtab so that it is at least NEW_LEN long. */ |
26 | static int | |
27 | grow_pptrtab (ctf_dict_t *fp, size_t new_len) | |
28 | { | |
29 | uint32_t *new_pptrtab; | |
30 | ||
31 | if ((new_pptrtab = realloc (fp->ctf_pptrtab, sizeof (uint32_t) | |
32 | * new_len)) == NULL) | |
33 | return (ctf_set_errno (fp, ENOMEM)); | |
34 | ||
35 | fp->ctf_pptrtab = new_pptrtab; | |
36 | ||
37 | memset (fp->ctf_pptrtab + fp->ctf_pptrtab_len, 0, | |
38 | sizeof (uint32_t) * (new_len - fp->ctf_pptrtab_len)); | |
39 | ||
40 | fp->ctf_pptrtab_len = new_len; | |
41 | return 0; | |
42 | } | |
43 | ||
44 | /* Update entries in the pptrtab that relate to types newly added in the | |
45 | child. */ | |
46 | static int | |
47 | refresh_pptrtab (ctf_dict_t *fp, ctf_dict_t *pfp) | |
48 | { | |
49 | uint32_t i; | |
50 | for (i = fp->ctf_pptrtab_typemax; i <= fp->ctf_typemax; i++) | |
51 | { | |
52 | ctf_id_t type = LCTF_INDEX_TO_TYPE (fp, i, 1); | |
53 | ctf_id_t reffed_type; | |
abe4ca69 NA |
54 | |
55 | if (ctf_type_kind (fp, type) != CTF_K_POINTER) | |
56 | continue; | |
57 | ||
58 | reffed_type = ctf_type_reference (fp, type); | |
59 | ||
60 | if (LCTF_TYPE_ISPARENT (fp, reffed_type)) | |
61 | { | |
62 | uint32_t idx = LCTF_TYPE_TO_INDEX (fp, reffed_type); | |
63 | ||
64 | /* Guard against references to invalid types. No need to consider | |
65 | the CTF dict corrupt in this case: this pointer just can't be a | |
66 | pointer to any type we know about. */ | |
abe4ca69 NA |
67 | if (idx <= pfp->ctf_typemax) |
68 | { | |
69 | if (idx >= fp->ctf_pptrtab_len | |
70 | && grow_pptrtab (fp, pfp->ctf_ptrtab_len) < 0) | |
71 | return -1; /* errno is set for us. */ | |
72 | ||
73 | fp->ctf_pptrtab[idx] = i; | |
74 | } | |
75 | } | |
76 | } | |
77 | ||
78 | fp->ctf_pptrtab_typemax = fp->ctf_typemax; | |
79 | ||
80 | return 0; | |
81 | } | |
82 | ||
b437bfe0 NA |
83 | /* Compare the given input string and length against a table of known C storage |
84 | qualifier keywords. We just ignore these in ctf_lookup_by_name, below. To | |
85 | do this quickly, we use a pre-computed Perfect Hash Function similar to the | |
86 | technique originally described in the classic paper: | |
87 | ||
88 | R.J. Cichelli, "Minimal Perfect Hash Functions Made Simple", | |
89 | Communications of the ACM, Volume 23, Issue 1, January 1980, pp. 17-19. | |
90 | ||
91 | For an input string S of length N, we use hash H = S[N - 1] + N - 105, which | |
92 | for the current set of qualifiers yields a unique H in the range [0 .. 20]. | |
93 | The hash can be modified when the keyword set changes as necessary. We also | |
94 | store the length of each keyword and check it prior to the final strcmp(). | |
95 | ||
96 | TODO: just use gperf. */ | |
97 | ||
98 | static int | |
99 | isqualifier (const char *s, size_t len) | |
100 | { | |
101 | static const struct qual | |
102 | { | |
103 | const char *q_name; | |
104 | size_t q_len; | |
105 | } qhash[] = { | |
106 | {"static", 6}, {"", 0}, {"", 0}, {"", 0}, | |
107 | {"volatile", 8}, {"", 0}, {"", 0}, {"", 0}, {"", 0}, | |
108 | {"", 0}, {"auto", 4}, {"extern", 6}, {"", 0}, {"", 0}, | |
109 | {"", 0}, {"", 0}, {"const", 5}, {"register", 8}, | |
110 | {"", 0}, {"restrict", 8}, {"_Restrict", 9} | |
111 | }; | |
112 | ||
113 | int h = s[len - 1] + (int) len - 105; | |
114 | const struct qual *qp = &qhash[h]; | |
115 | ||
116 | return (h >= 0 && (size_t) h < sizeof (qhash) / sizeof (qhash[0]) | |
117 | && (size_t) len == qp->q_len && | |
118 | strncmp (qp->q_name, s, qp->q_len) == 0); | |
119 | } | |
120 | ||
121 | /* Attempt to convert the given C type name into the corresponding CTF type ID. | |
122 | It is not possible to do complete and proper conversion of type names | |
123 | without implementing a more full-fledged parser, which is necessary to | |
124 | handle things like types that are function pointers to functions that | |
125 | have arguments that are function pointers, and fun stuff like that. | |
126 | Instead, this function implements a very simple conversion algorithm that | |
127 | finds the things that we actually care about: structs, unions, enums, | |
128 | integers, floats, typedefs, and pointers to any of these named types. */ | |
129 | ||
abe4ca69 NA |
130 | static ctf_id_t |
131 | ctf_lookup_by_name_internal (ctf_dict_t *fp, ctf_dict_t *child, | |
132 | const char *name) | |
b437bfe0 NA |
133 | { |
134 | static const char delimiters[] = " \t\n\r\v\f*"; | |
135 | ||
136 | const ctf_lookup_t *lp; | |
137 | const char *p, *q, *end; | |
138 | ctf_id_t type = 0; | |
139 | ctf_id_t ntype, ptype; | |
140 | ||
141 | if (name == NULL) | |
142 | return (ctf_set_errno (fp, EINVAL)); | |
143 | ||
144 | for (p = name, end = name + strlen (name); *p != '\0'; p = q) | |
145 | { | |
734c8942 | 146 | while (isspace ((int) *p)) |
b437bfe0 NA |
147 | p++; /* Skip leading whitespace. */ |
148 | ||
149 | if (p == end) | |
150 | break; | |
151 | ||
152 | if ((q = strpbrk (p + 1, delimiters)) == NULL) | |
153 | q = end; /* Compare until end. */ | |
154 | ||
155 | if (*p == '*') | |
156 | { | |
abe4ca69 NA |
157 | /* Find a pointer to type by looking in child->ctf_pptrtab (if child |
158 | is set) and fp->ctf_ptrtab. If we can't find a pointer to the | |
159 | given type, see if we can compute a pointer to the type resulting | |
160 | from resolving the type down to its base type and use that instead. | |
161 | This helps with cases where the CTF data includes "struct foo *" | |
162 | but not "foo_t *" and the user tries to access "foo_t *" in the | |
e05a3e5a NA |
163 | debugger. |
164 | ||
165 | There is extra complexity here because uninitialized elements in | |
166 | the pptrtab and ptrtab are set to zero, but zero (as the type ID | |
167 | meaning the unimplemented type) is a valid return type from | |
168 | ctf_lookup_by_name. (Pointers to types are never of type 0, so | |
169 | this is unambiguous, just fiddly to deal with.) */ | |
abe4ca69 NA |
170 | |
171 | uint32_t idx = LCTF_TYPE_TO_INDEX (fp, type); | |
172 | int in_child = 0; | |
173 | ||
e05a3e5a | 174 | ntype = CTF_ERR; |
abe4ca69 NA |
175 | if (child && idx <= child->ctf_pptrtab_len) |
176 | { | |
177 | ntype = child->ctf_pptrtab[idx]; | |
178 | if (ntype) | |
179 | in_child = 1; | |
e05a3e5a NA |
180 | else |
181 | ntype = CTF_ERR; | |
abe4ca69 | 182 | } |
b437bfe0 | 183 | |
e05a3e5a NA |
184 | if (ntype == CTF_ERR) |
185 | { | |
186 | ntype = fp->ctf_ptrtab[idx]; | |
187 | if (ntype == 0) | |
188 | ntype = CTF_ERR; | |
189 | } | |
b437bfe0 | 190 | |
abe4ca69 | 191 | /* Try resolving to its base type and check again. */ |
e05a3e5a | 192 | if (ntype == CTF_ERR) |
b437bfe0 | 193 | { |
abe4ca69 NA |
194 | if (child) |
195 | ntype = ctf_type_resolve_unsliced (child, type); | |
196 | else | |
197 | ntype = ctf_type_resolve_unsliced (fp, type); | |
198 | ||
199 | if (ntype == CTF_ERR) | |
200 | goto notype; | |
201 | ||
202 | idx = LCTF_TYPE_TO_INDEX (fp, ntype); | |
203 | ||
e05a3e5a | 204 | ntype = CTF_ERR; |
abe4ca69 | 205 | if (child && idx <= child->ctf_pptrtab_len) |
b437bfe0 | 206 | { |
abe4ca69 NA |
207 | ntype = child->ctf_pptrtab[idx]; |
208 | if (ntype) | |
209 | in_child = 1; | |
e05a3e5a NA |
210 | else |
211 | ntype = CTF_ERR; | |
b437bfe0 | 212 | } |
abe4ca69 | 213 | |
e05a3e5a NA |
214 | if (ntype == CTF_ERR) |
215 | { | |
216 | ntype = fp->ctf_ptrtab[idx]; | |
217 | if (ntype == 0) | |
218 | ntype = CTF_ERR; | |
219 | } | |
abe4ca69 NA |
220 | if (ntype == CTF_ERR) |
221 | goto notype; | |
b437bfe0 NA |
222 | } |
223 | ||
abe4ca69 NA |
224 | type = LCTF_INDEX_TO_TYPE (fp, ntype, (fp->ctf_flags & LCTF_CHILD) |
225 | || in_child); | |
226 | ||
227 | /* We are looking up a type in the parent, but the pointed-to type is | |
228 | in the child. Switch to looking in the child: if we need to go | |
229 | back into the parent, we can recurse again. */ | |
230 | if (in_child) | |
231 | { | |
232 | fp = child; | |
233 | child = NULL; | |
234 | } | |
b437bfe0 NA |
235 | |
236 | q = p + 1; | |
237 | continue; | |
238 | } | |
239 | ||
240 | if (isqualifier (p, (size_t) (q - p))) | |
241 | continue; /* Skip qualifier keyword. */ | |
242 | ||
243 | for (lp = fp->ctf_lookups; lp->ctl_prefix != NULL; lp++) | |
244 | { | |
245 | /* TODO: This is not MT-safe. */ | |
246 | if ((lp->ctl_prefix[0] == '\0' || | |
247 | strncmp (p, lp->ctl_prefix, (size_t) (q - p)) == 0) && | |
248 | (size_t) (q - p) >= lp->ctl_len) | |
249 | { | |
734c8942 | 250 | for (p += lp->ctl_len; isspace ((int) *p); p++) |
b437bfe0 NA |
251 | continue; /* Skip prefix and next whitespace. */ |
252 | ||
253 | if ((q = strchr (p, '*')) == NULL) | |
254 | q = end; /* Compare until end. */ | |
255 | ||
734c8942 | 256 | while (isspace ((int) q[-1])) |
b437bfe0 NA |
257 | q--; /* Exclude trailing whitespace. */ |
258 | ||
259 | /* Expand and/or allocate storage for a slice of the name, then | |
260 | copy it in. */ | |
261 | ||
262 | if (fp->ctf_tmp_typeslicelen >= (size_t) (q - p) + 1) | |
263 | { | |
264 | memcpy (fp->ctf_tmp_typeslice, p, (size_t) (q - p)); | |
265 | fp->ctf_tmp_typeslice[(size_t) (q - p)] = '\0'; | |
266 | } | |
267 | else | |
268 | { | |
269 | free (fp->ctf_tmp_typeslice); | |
942d35f7 | 270 | fp->ctf_tmp_typeslice = xstrndup (p, (size_t) (q - p)); |
b437bfe0 NA |
271 | if (fp->ctf_tmp_typeslice == NULL) |
272 | { | |
abe4ca69 | 273 | ctf_set_errno (fp, ENOMEM); |
b437bfe0 NA |
274 | return CTF_ERR; |
275 | } | |
276 | } | |
277 | ||
676c3ecb NA |
278 | if ((type = ctf_lookup_by_rawhash (fp, lp->ctl_hash, |
279 | fp->ctf_tmp_typeslice)) == 0) | |
abe4ca69 | 280 | goto notype; |
b437bfe0 NA |
281 | |
282 | break; | |
283 | } | |
284 | } | |
285 | ||
286 | if (lp->ctl_prefix == NULL) | |
abe4ca69 | 287 | goto notype; |
b437bfe0 NA |
288 | } |
289 | ||
290 | if (*p != '\0' || type == 0) | |
291 | return (ctf_set_errno (fp, ECTF_SYNTAX)); | |
292 | ||
293 | return type; | |
294 | ||
abe4ca69 NA |
295 | notype: |
296 | ctf_set_errno (fp, ECTF_NOTYPE); | |
297 | if (fp->ctf_parent != NULL) | |
298 | { | |
299 | /* Need to look up in the parent, from the child's perspective. | |
300 | Make sure the pptrtab is up to date. */ | |
301 | ||
302 | if (fp->ctf_pptrtab_typemax < fp->ctf_typemax) | |
303 | { | |
304 | if (refresh_pptrtab (fp, fp->ctf_parent) < 0) | |
305 | return -1; /* errno is set for us. */ | |
306 | } | |
307 | ||
308 | if ((ptype = ctf_lookup_by_name_internal (fp->ctf_parent, fp, | |
309 | name)) != CTF_ERR) | |
310 | return ptype; | |
311 | return (ctf_set_errno (fp, ctf_errno (fp->ctf_parent))); | |
312 | } | |
b437bfe0 NA |
313 | |
314 | return CTF_ERR; | |
315 | } | |
316 | ||
abe4ca69 NA |
317 | ctf_id_t |
318 | ctf_lookup_by_name (ctf_dict_t *fp, const char *name) | |
319 | { | |
320 | return ctf_lookup_by_name_internal (fp, NULL, name); | |
321 | } | |
322 | ||
1136c379 NA |
323 | /* Return the pointer to the internal CTF type data corresponding to the |
324 | given type ID. If the ID is invalid, the function returns NULL. | |
325 | This function is not exported outside of the library. */ | |
326 | ||
327 | const ctf_type_t * | |
328 | ctf_lookup_by_id (ctf_dict_t **fpp, ctf_id_t type) | |
b437bfe0 | 329 | { |
1136c379 NA |
330 | ctf_dict_t *fp = *fpp; /* Caller passes in starting CTF dict. */ |
331 | ctf_id_t idx; | |
332 | ||
333 | if ((fp = ctf_get_dict (fp, type)) == NULL) | |
334 | { | |
335 | (void) ctf_set_errno (*fpp, ECTF_NOPARENT); | |
336 | return NULL; | |
337 | } | |
338 | ||
339 | /* If this dict is writable, check for a dynamic type. */ | |
340 | ||
341 | if (fp->ctf_flags & LCTF_RDWR) | |
342 | { | |
343 | ctf_dtdef_t *dtd; | |
344 | ||
345 | if ((dtd = ctf_dynamic_type (fp, type)) != NULL) | |
346 | { | |
347 | *fpp = fp; | |
348 | return &dtd->dtd_data; | |
349 | } | |
350 | (void) ctf_set_errno (*fpp, ECTF_BADID); | |
351 | return NULL; | |
352 | } | |
353 | ||
354 | /* Check for a type in the static portion. */ | |
355 | ||
356 | idx = LCTF_TYPE_TO_INDEX (fp, type); | |
357 | if (idx > 0 && (unsigned long) idx <= fp->ctf_typemax) | |
358 | { | |
359 | *fpp = fp; /* Function returns ending CTF dict. */ | |
360 | return (LCTF_INDEX_TO_TYPEPTR (fp, idx)); | |
361 | } | |
362 | ||
363 | (void) ctf_set_errno (*fpp, ECTF_BADID); | |
364 | return NULL; | |
365 | } | |
366 | ||
367 | typedef struct ctf_lookup_idx_key | |
368 | { | |
369 | ctf_dict_t *clik_fp; | |
370 | const char *clik_name; | |
371 | uint32_t *clik_names; | |
372 | } ctf_lookup_idx_key_t; | |
b437bfe0 NA |
373 | |
374 | /* A bsearch function for variable names. */ | |
375 | ||
376 | static int | |
1136c379 | 377 | ctf_lookup_var (const void *key_, const void *lookup_) |
b437bfe0 | 378 | { |
1136c379 NA |
379 | const ctf_lookup_idx_key_t *key = key_; |
380 | const ctf_varent_t *lookup = lookup_; | |
b437bfe0 | 381 | |
1136c379 | 382 | return (strcmp (key->clik_name, ctf_strptr (key->clik_fp, lookup->ctv_name))); |
b437bfe0 NA |
383 | } |
384 | ||
385 | /* Given a variable name, return the type of the variable with that name. */ | |
386 | ||
387 | ctf_id_t | |
139633c3 | 388 | ctf_lookup_variable (ctf_dict_t *fp, const char *name) |
b437bfe0 NA |
389 | { |
390 | ctf_varent_t *ent; | |
1136c379 | 391 | ctf_lookup_idx_key_t key = { fp, name, NULL }; |
b437bfe0 NA |
392 | |
393 | /* This array is sorted, so we can bsearch for it. */ | |
394 | ||
395 | ent = bsearch (&key, fp->ctf_vars, fp->ctf_nvars, sizeof (ctf_varent_t), | |
396 | ctf_lookup_var); | |
397 | ||
398 | if (ent == NULL) | |
399 | { | |
400 | if (fp->ctf_parent != NULL) | |
401 | return ctf_lookup_variable (fp->ctf_parent, name); | |
402 | ||
403 | return (ctf_set_errno (fp, ECTF_NOTYPEDAT)); | |
404 | } | |
405 | ||
406 | return ent->ctv_type; | |
407 | } | |
408 | ||
1136c379 NA |
409 | typedef struct ctf_symidx_sort_arg_cb |
410 | { | |
411 | ctf_dict_t *fp; | |
412 | uint32_t *names; | |
413 | } ctf_symidx_sort_arg_cb_t; | |
414 | ||
415 | static int | |
416 | sort_symidx_by_name (const void *one_, const void *two_, void *arg_) | |
417 | { | |
418 | const uint32_t *one = one_; | |
419 | const uint32_t *two = two_; | |
420 | ctf_symidx_sort_arg_cb_t *arg = arg_; | |
421 | ||
422 | return (strcmp (ctf_strptr (arg->fp, arg->names[*one]), | |
423 | ctf_strptr (arg->fp, arg->names[*two]))); | |
424 | } | |
425 | ||
426 | /* Sort a symbol index section by name. Takes a 1:1 mapping of names to the | |
427 | corresponding symbol table. Returns a lexicographically sorted array of idx | |
428 | indexes (and thus, of indexes into the corresponding func info / data object | |
429 | section). */ | |
430 | ||
431 | static uint32_t * | |
432 | ctf_symidx_sort (ctf_dict_t *fp, uint32_t *idx, size_t *nidx, | |
433 | size_t len) | |
434 | { | |
435 | uint32_t *sorted; | |
436 | size_t i; | |
437 | ||
438 | if ((sorted = malloc (len)) == NULL) | |
439 | { | |
440 | ctf_set_errno (fp, ENOMEM); | |
441 | return NULL; | |
442 | } | |
443 | ||
444 | *nidx = len / sizeof (uint32_t); | |
445 | for (i = 0; i < *nidx; i++) | |
446 | sorted[i] = i; | |
447 | ||
448 | if (!(fp->ctf_header->cth_flags & CTF_F_IDXSORTED)) | |
449 | { | |
450 | ctf_symidx_sort_arg_cb_t arg = { fp, idx }; | |
451 | ctf_dprintf ("Index section unsorted: sorting."); | |
452 | ctf_qsort_r (sorted, *nidx, sizeof (uint32_t), sort_symidx_by_name, &arg); | |
453 | fp->ctf_header->cth_flags |= CTF_F_IDXSORTED; | |
454 | } | |
455 | ||
456 | return sorted; | |
457 | } | |
458 | ||
459 | /* Given a symbol index, return the name of that symbol from the table provided | |
460 | by ctf_link_shuffle_syms, or failing that from the secondary string table, or | |
461 | the null string. */ | |
b437bfe0 | 462 | const char * |
139633c3 | 463 | ctf_lookup_symbol_name (ctf_dict_t *fp, unsigned long symidx) |
b437bfe0 NA |
464 | { |
465 | const ctf_sect_t *sp = &fp->ctf_symtab; | |
1136c379 NA |
466 | ctf_link_sym_t sym; |
467 | int err; | |
b437bfe0 | 468 | |
1136c379 | 469 | if (fp->ctf_dynsymidx) |
b437bfe0 | 470 | { |
1136c379 NA |
471 | err = EINVAL; |
472 | if (symidx > fp->ctf_dynsymmax) | |
473 | goto try_parent; | |
474 | ||
475 | ctf_link_sym_t *symp = fp->ctf_dynsymidx[symidx]; | |
476 | ||
477 | if (!symp) | |
478 | goto try_parent; | |
479 | ||
480 | return symp->st_name; | |
b437bfe0 NA |
481 | } |
482 | ||
1136c379 NA |
483 | err = ECTF_NOSYMTAB; |
484 | if (sp->cts_data == NULL) | |
485 | goto try_parent; | |
486 | ||
b437bfe0 | 487 | if (symidx >= fp->ctf_nsyms) |
1136c379 NA |
488 | goto try_parent; |
489 | ||
490 | switch (sp->cts_entsize) | |
b437bfe0 | 491 | { |
1136c379 NA |
492 | case sizeof (Elf64_Sym): |
493 | { | |
494 | const Elf64_Sym *symp = (Elf64_Sym *) sp->cts_data + symidx; | |
495 | ctf_elf64_to_link_sym (fp, &sym, symp, symidx); | |
496 | } | |
497 | break; | |
498 | case sizeof (Elf32_Sym): | |
499 | { | |
500 | const Elf32_Sym *symp = (Elf32_Sym *) sp->cts_data + symidx; | |
501 | ctf_elf32_to_link_sym (fp, &sym, symp, symidx); | |
502 | } | |
503 | break; | |
504 | default: | |
505 | ctf_set_errno (fp, ECTF_SYMTAB); | |
b437bfe0 NA |
506 | return _CTF_NULLSTR; |
507 | } | |
508 | ||
1136c379 | 509 | assert (!sym.st_nameidx_set); |
b437bfe0 | 510 | |
1136c379 | 511 | return sym.st_name; |
b437bfe0 | 512 | |
1136c379 NA |
513 | try_parent: |
514 | if (fp->ctf_parent) | |
515 | return ctf_lookup_symbol_name (fp->ctf_parent, symidx); | |
516 | else | |
517 | { | |
518 | ctf_set_errno (fp, err); | |
519 | return _CTF_NULLSTR; | |
520 | } | |
b437bfe0 NA |
521 | } |
522 | ||
1136c379 NA |
523 | /* Iterate over all symbols with types: if FUNC, function symbols, otherwise, |
524 | data symbols. The name argument is not optional. The return order is | |
525 | arbitrary, though is likely to be in symbol index or name order. You can | |
526 | change the value of 'functions' in the middle of iteration over non-dynamic | |
527 | dicts, but doing so on dynamic dicts will fail. (This is probably not very | |
528 | useful, but there is no reason to prohibit it.) */ | |
b437bfe0 NA |
529 | |
530 | ctf_id_t | |
1136c379 NA |
531 | ctf_symbol_next (ctf_dict_t *fp, ctf_next_t **it, const char **name, |
532 | int functions) | |
b437bfe0 | 533 | { |
1136c379 NA |
534 | ctf_id_t sym; |
535 | ctf_next_t *i = *it; | |
536 | int err; | |
b437bfe0 | 537 | |
1136c379 NA |
538 | if (!i) |
539 | { | |
540 | if ((i = ctf_next_create ()) == NULL) | |
541 | return ctf_set_errno (fp, ENOMEM); | |
b437bfe0 | 542 | |
1136c379 NA |
543 | i->cu.ctn_fp = fp; |
544 | i->ctn_iter_fun = (void (*) (void)) ctf_symbol_next; | |
545 | i->ctn_n = 0; | |
546 | *it = i; | |
547 | } | |
548 | ||
549 | if ((void (*) (void)) ctf_symbol_next != i->ctn_iter_fun) | |
550 | return (ctf_set_errno (fp, ECTF_NEXT_WRONGFUN)); | |
b437bfe0 | 551 | |
1136c379 NA |
552 | if (fp != i->cu.ctn_fp) |
553 | return (ctf_set_errno (fp, ECTF_NEXT_WRONGFP)); | |
554 | ||
555 | /* We intentionally use raw access, not ctf_lookup_by_symbol, to avoid | |
556 | incurring additional sorting cost for unsorted symtypetabs coming from the | |
557 | compiler, to allow ctf_symbol_next to work in the absence of a symtab, and | |
558 | finally because it's easier to work out what the name of each symbol is if | |
559 | we do that. */ | |
560 | ||
561 | if (fp->ctf_flags & LCTF_RDWR) | |
b437bfe0 | 562 | { |
1136c379 NA |
563 | ctf_dynhash_t *dynh = functions ? fp->ctf_funchash : fp->ctf_objthash; |
564 | void *dyn_name = NULL, *dyn_value = NULL; | |
565 | ||
566 | if (!dynh) | |
567 | { | |
568 | ctf_next_destroy (i); | |
569 | return (ctf_set_errno (fp, ECTF_NEXT_END)); | |
570 | } | |
571 | ||
6c3a3877 | 572 | err = ctf_dynhash_next (dynh, &i->ctn_next, &dyn_name, &dyn_value); |
1136c379 NA |
573 | /* This covers errors and also end-of-iteration. */ |
574 | if (err != 0) | |
575 | { | |
576 | ctf_next_destroy (i); | |
577 | *it = NULL; | |
578 | return ctf_set_errno (fp, err); | |
579 | } | |
580 | ||
581 | *name = dyn_name; | |
582 | sym = (ctf_id_t) (uintptr_t) dyn_value; | |
b437bfe0 | 583 | } |
1136c379 NA |
584 | else if ((!functions && fp->ctf_objtidx_names) || |
585 | (functions && fp->ctf_funcidx_names)) | |
b437bfe0 | 586 | { |
1136c379 NA |
587 | ctf_header_t *hp = fp->ctf_header; |
588 | uint32_t *idx = functions ? fp->ctf_funcidx_names : fp->ctf_objtidx_names; | |
589 | uint32_t *tab; | |
590 | size_t len; | |
591 | ||
592 | if (functions) | |
593 | { | |
594 | len = (hp->cth_varoff - hp->cth_funcidxoff) / sizeof (uint32_t); | |
595 | tab = (uint32_t *) (fp->ctf_buf + hp->cth_funcoff); | |
596 | } | |
597 | else | |
598 | { | |
599 | len = (hp->cth_funcidxoff - hp->cth_objtidxoff) / sizeof (uint32_t); | |
600 | tab = (uint32_t *) (fp->ctf_buf + hp->cth_objtoff); | |
601 | } | |
602 | ||
603 | do | |
604 | { | |
605 | if (i->ctn_n >= len) | |
606 | goto end; | |
607 | ||
608 | *name = ctf_strptr (fp, idx[i->ctn_n]); | |
609 | sym = tab[i->ctn_n++]; | |
610 | } while (sym == -1u || sym == 0); | |
b437bfe0 | 611 | } |
1136c379 NA |
612 | else |
613 | { | |
614 | /* Skip over pads in ctf_xslate, padding for typeless symbols in the | |
615 | symtypetab itself, and symbols in the wrong table. */ | |
616 | for (; i->ctn_n < fp->ctf_nsyms; i->ctn_n++) | |
617 | { | |
618 | ctf_header_t *hp = fp->ctf_header; | |
b437bfe0 | 619 | |
1136c379 NA |
620 | if (fp->ctf_sxlate[i->ctn_n] == -1u) |
621 | continue; | |
b437bfe0 | 622 | |
1136c379 | 623 | sym = *(uint32_t *) ((uintptr_t) fp->ctf_buf + fp->ctf_sxlate[i->ctn_n]); |
b437bfe0 | 624 | |
1136c379 NA |
625 | if (sym == 0) |
626 | continue; | |
627 | ||
628 | if (functions) | |
629 | { | |
630 | if (fp->ctf_sxlate[i->ctn_n] >= hp->cth_funcoff | |
631 | && fp->ctf_sxlate[i->ctn_n] < hp->cth_objtidxoff) | |
632 | break; | |
633 | } | |
634 | else | |
635 | { | |
636 | if (fp->ctf_sxlate[i->ctn_n] >= hp->cth_objtoff | |
637 | && fp->ctf_sxlate[i->ctn_n] < hp->cth_funcoff) | |
638 | break; | |
639 | } | |
640 | } | |
641 | ||
642 | if (i->ctn_n >= fp->ctf_nsyms) | |
643 | goto end; | |
644 | ||
645 | *name = ctf_lookup_symbol_name (fp, i->ctn_n++); | |
646 | } | |
647 | ||
648 | return sym; | |
649 | ||
650 | end: | |
651 | ctf_next_destroy (i); | |
652 | *it = NULL; | |
653 | return (ctf_set_errno (fp, ECTF_NEXT_END)); | |
b437bfe0 NA |
654 | } |
655 | ||
1136c379 NA |
656 | /* A bsearch function for function and object index names. */ |
657 | ||
658 | static int | |
659 | ctf_lookup_idx_name (const void *key_, const void *idx_) | |
688d28f6 | 660 | { |
1136c379 NA |
661 | const ctf_lookup_idx_key_t *key = key_; |
662 | const uint32_t *idx = idx_; | |
688d28f6 | 663 | |
1136c379 | 664 | return (strcmp (key->clik_name, ctf_strptr (key->clik_fp, key->clik_names[*idx]))); |
688d28f6 NA |
665 | } |
666 | ||
1136c379 NA |
667 | /* Given a symbol number, look up that symbol in the function or object |
668 | index table (which must exist). Return 0 if not found there (or pad). */ | |
47d546f4 | 669 | |
1136c379 NA |
670 | static ctf_id_t |
671 | ctf_try_lookup_indexed (ctf_dict_t *fp, unsigned long symidx, int is_function) | |
47d546f4 | 672 | { |
1136c379 NA |
673 | const char *symname = ctf_lookup_symbol_name (fp, symidx); |
674 | struct ctf_header *hp = fp->ctf_header; | |
675 | uint32_t *symtypetab; | |
676 | uint32_t *names; | |
677 | uint32_t *sxlate; | |
678 | size_t nidx; | |
47d546f4 | 679 | |
1136c379 NA |
680 | ctf_dprintf ("Looking up type of object with symtab idx %lx (%s) in " |
681 | "indexed symtypetab\n", symidx, symname); | |
47d546f4 | 682 | |
1136c379 NA |
683 | if (symname[0] == '\0') |
684 | return -1; /* errno is set for us. */ | |
47d546f4 | 685 | |
1136c379 | 686 | if (is_function) |
47d546f4 | 687 | { |
1136c379 | 688 | if (!fp->ctf_funcidx_sxlate) |
47d546f4 | 689 | { |
1136c379 NA |
690 | if ((fp->ctf_funcidx_sxlate |
691 | = ctf_symidx_sort (fp, (uint32_t *) | |
692 | (fp->ctf_buf + hp->cth_funcidxoff), | |
693 | &fp->ctf_nfuncidx, | |
694 | hp->cth_varoff - hp->cth_funcidxoff)) | |
695 | == NULL) | |
696 | { | |
697 | ctf_err_warn (fp, 0, 0, _("cannot sort function symidx")); | |
698 | return -1; /* errno is set for us. */ | |
699 | } | |
47d546f4 | 700 | } |
1136c379 NA |
701 | symtypetab = (uint32_t *) (fp->ctf_buf + hp->cth_funcoff); |
702 | sxlate = fp->ctf_funcidx_sxlate; | |
703 | names = fp->ctf_funcidx_names; | |
704 | nidx = fp->ctf_nfuncidx; | |
47d546f4 | 705 | } |
1136c379 NA |
706 | else |
707 | { | |
708 | if (!fp->ctf_objtidx_sxlate) | |
709 | { | |
710 | if ((fp->ctf_objtidx_sxlate | |
711 | = ctf_symidx_sort (fp, (uint32_t *) | |
712 | (fp->ctf_buf + hp->cth_objtidxoff), | |
713 | &fp->ctf_nobjtidx, | |
714 | hp->cth_funcidxoff - hp->cth_objtidxoff)) | |
715 | == NULL) | |
716 | { | |
717 | ctf_err_warn (fp, 0, 0, _("cannot sort object symidx")); | |
718 | return -1; /* errno is set for us. */ | |
719 | } | |
720 | } | |
676c3ecb | 721 | |
1136c379 NA |
722 | symtypetab = (uint32_t *) (fp->ctf_buf + hp->cth_objtoff); |
723 | sxlate = fp->ctf_objtidx_sxlate; | |
724 | names = fp->ctf_objtidx_names; | |
725 | nidx = fp->ctf_nobjtidx; | |
726 | } | |
676c3ecb | 727 | |
1136c379 NA |
728 | ctf_lookup_idx_key_t key = { fp, symname, names }; |
729 | uint32_t *idx; | |
730 | ||
731 | idx = bsearch (&key, sxlate, nidx, sizeof (uint32_t), ctf_lookup_idx_name); | |
732 | ||
733 | if (!idx) | |
676c3ecb | 734 | { |
1136c379 NA |
735 | ctf_dprintf ("%s not found in idx\n", symname); |
736 | return 0; | |
676c3ecb NA |
737 | } |
738 | ||
1136c379 NA |
739 | /* Should be impossible, but be paranoid. */ |
740 | if ((idx - sxlate) > (ptrdiff_t) nidx) | |
741 | return (ctf_set_errno (fp, ECTF_CORRUPT)); | |
742 | ||
743 | ctf_dprintf ("Symbol %lx (%s) is of type %x\n", symidx, symname, | |
744 | symtypetab[*idx]); | |
745 | return symtypetab[*idx]; | |
47d546f4 NA |
746 | } |
747 | ||
1136c379 NA |
748 | /* Given a symbol table index, return the type of the function or data object |
749 | described by the corresponding entry in the symbol table. We can only return | |
750 | symbols in read-only dicts and in dicts for which ctf_link_shuffle_syms has | |
751 | been called to assign symbol indexes to symbol names. */ | |
b437bfe0 | 752 | |
1136c379 NA |
753 | ctf_id_t |
754 | ctf_lookup_by_symbol (ctf_dict_t *fp, unsigned long symidx) | |
b437bfe0 NA |
755 | { |
756 | const ctf_sect_t *sp = &fp->ctf_symtab; | |
1136c379 NA |
757 | ctf_id_t type = 0; |
758 | int err = 0; | |
759 | ||
760 | /* Shuffled dynsymidx present? Use that. */ | |
761 | if (fp->ctf_dynsymidx) | |
762 | { | |
763 | const ctf_link_sym_t *sym; | |
764 | ||
765 | ctf_dprintf ("Looking up type of object with symtab idx %lx in " | |
766 | "writable dict symtypetab\n", symidx); | |
767 | ||
768 | /* The dict must be dynamic. */ | |
769 | if (!ctf_assert (fp, fp->ctf_flags & LCTF_RDWR)) | |
770 | return CTF_ERR; | |
771 | ||
772 | err = EINVAL; | |
773 | if (symidx > fp->ctf_dynsymmax) | |
774 | goto try_parent; | |
775 | ||
776 | sym = fp->ctf_dynsymidx[symidx]; | |
777 | err = ECTF_NOTYPEDAT; | |
778 | if (!sym || (sym->st_shndx != STT_OBJECT && sym->st_shndx != STT_FUNC)) | |
779 | goto try_parent; | |
780 | ||
781 | if (!ctf_assert (fp, !sym->st_nameidx_set)) | |
782 | return CTF_ERR; | |
783 | ||
784 | if (fp->ctf_objthash == NULL | |
785 | || ((type = (ctf_id_t) (uintptr_t) | |
786 | ctf_dynhash_lookup (fp->ctf_objthash, sym->st_name)) == 0)) | |
787 | { | |
788 | if (fp->ctf_funchash == NULL | |
789 | || ((type = (ctf_id_t) (uintptr_t) | |
790 | ctf_dynhash_lookup (fp->ctf_funchash, sym->st_name)) == 0)) | |
791 | goto try_parent; | |
792 | } | |
793 | ||
794 | return type; | |
795 | } | |
b437bfe0 | 796 | |
1136c379 | 797 | err = ECTF_NOSYMTAB; |
b437bfe0 | 798 | if (sp->cts_data == NULL) |
1136c379 | 799 | goto try_parent; |
b437bfe0 | 800 | |
1136c379 NA |
801 | /* This covers both out-of-range lookups and a dynamic dict which hasn't been |
802 | shuffled yet. */ | |
803 | err = EINVAL; | |
b437bfe0 | 804 | if (symidx >= fp->ctf_nsyms) |
1136c379 | 805 | goto try_parent; |
b437bfe0 | 806 | |
1136c379 | 807 | if (fp->ctf_objtidx_names) |
b437bfe0 | 808 | { |
1136c379 NA |
809 | if ((type = ctf_try_lookup_indexed (fp, symidx, 0)) == CTF_ERR) |
810 | return CTF_ERR; /* errno is set for us. */ | |
b437bfe0 | 811 | } |
1136c379 | 812 | if (type == 0 && fp->ctf_funcidx_names) |
b437bfe0 | 813 | { |
1136c379 NA |
814 | if ((type = ctf_try_lookup_indexed (fp, symidx, 1)) == CTF_ERR) |
815 | return CTF_ERR; /* errno is set for us. */ | |
b437bfe0 | 816 | } |
1136c379 NA |
817 | if (type != 0) |
818 | return type; | |
819 | ||
820 | err = ECTF_NOTYPEDAT; | |
821 | if (fp->ctf_objtidx_names && fp->ctf_funcidx_names) | |
822 | goto try_parent; | |
823 | ||
824 | /* Table must be nonindexed. */ | |
825 | ||
826 | ctf_dprintf ("Looking up object type %lx in 1:1 dict symtypetab\n", symidx); | |
b437bfe0 NA |
827 | |
828 | if (fp->ctf_sxlate[symidx] == -1u) | |
1136c379 NA |
829 | goto try_parent; |
830 | ||
831 | type = *(uint32_t *) ((uintptr_t) fp->ctf_buf + fp->ctf_sxlate[symidx]); | |
b437bfe0 | 832 | |
1136c379 NA |
833 | if (type == 0) |
834 | goto try_parent; | |
b437bfe0 | 835 | |
1136c379 NA |
836 | return type; |
837 | try_parent: | |
838 | if (fp->ctf_parent) | |
839 | return ctf_lookup_by_symbol (fp->ctf_parent, symidx); | |
840 | else | |
841 | return (ctf_set_errno (fp, err)); | |
842 | } | |
b437bfe0 | 843 | |
1136c379 NA |
844 | /* Given a symbol table index, return the info for the function described |
845 | by the corresponding entry in the symbol table, which may be a function | |
846 | symbol or may be a data symbol that happens to be a function pointer. */ | |
b437bfe0 | 847 | |
1136c379 NA |
848 | int |
849 | ctf_func_info (ctf_dict_t *fp, unsigned long symidx, ctf_funcinfo_t *fip) | |
850 | { | |
851 | ctf_id_t type; | |
b437bfe0 | 852 | |
1136c379 NA |
853 | if ((type = ctf_lookup_by_symbol (fp, symidx)) == CTF_ERR) |
854 | return -1; /* errno is set for us. */ | |
b437bfe0 | 855 | |
1136c379 NA |
856 | if (ctf_type_kind (fp, type) != CTF_K_FUNCTION) |
857 | return (ctf_set_errno (fp, ECTF_NOTFUNC)); | |
b437bfe0 | 858 | |
1136c379 | 859 | return ctf_func_type_info (fp, type, fip); |
b437bfe0 NA |
860 | } |
861 | ||
862 | /* Given a symbol table index, return the arguments for the function described | |
863 | by the corresponding entry in the symbol table. */ | |
864 | ||
865 | int | |
139633c3 | 866 | ctf_func_args (ctf_dict_t *fp, unsigned long symidx, uint32_t argc, |
1136c379 | 867 | ctf_id_t *argv) |
b437bfe0 | 868 | { |
1136c379 | 869 | ctf_id_t type; |
b437bfe0 | 870 | |
1136c379 NA |
871 | if ((type = ctf_lookup_by_symbol (fp, symidx)) == CTF_ERR) |
872 | return -1; /* errno is set for us. */ | |
b437bfe0 | 873 | |
1136c379 NA |
874 | if (ctf_type_kind (fp, type) != CTF_K_FUNCTION) |
875 | return (ctf_set_errno (fp, ECTF_NOTFUNC)); | |
b437bfe0 | 876 | |
1136c379 | 877 | return ctf_func_type_args (fp, type, argc, argv); |
b437bfe0 | 878 | } |