| 1 | /* GDB routines for manipulating the minimal symbol tables. |
| 2 | Copyright 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001 |
| 3 | Free Software Foundation, Inc. |
| 4 | Contributed by Cygnus Support, using pieces from other GDB modules. |
| 5 | |
| 6 | This file is part of GDB. |
| 7 | |
| 8 | This program is free software; you can redistribute it and/or modify |
| 9 | it under the terms of the GNU General Public License as published by |
| 10 | the Free Software Foundation; either version 2 of the License, or |
| 11 | (at your option) any later version. |
| 12 | |
| 13 | This program is distributed in the hope that it will be useful, |
| 14 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 16 | GNU General Public License for more details. |
| 17 | |
| 18 | You should have received a copy of the GNU General Public License |
| 19 | along with this program; if not, write to the Free Software |
| 20 | Foundation, Inc., 59 Temple Place - Suite 330, |
| 21 | Boston, MA 02111-1307, USA. */ |
| 22 | |
| 23 | |
| 24 | /* This file contains support routines for creating, manipulating, and |
| 25 | destroying minimal symbol tables. |
| 26 | |
| 27 | Minimal symbol tables are used to hold some very basic information about |
| 28 | all defined global symbols (text, data, bss, abs, etc). The only two |
| 29 | required pieces of information are the symbol's name and the address |
| 30 | associated with that symbol. |
| 31 | |
| 32 | In many cases, even if a file was compiled with no special options for |
| 33 | debugging at all, as long as was not stripped it will contain sufficient |
| 34 | information to build useful minimal symbol tables using this structure. |
| 35 | |
| 36 | Even when a file contains enough debugging information to build a full |
| 37 | symbol table, these minimal symbols are still useful for quickly mapping |
| 38 | between names and addresses, and vice versa. They are also sometimes used |
| 39 | to figure out what full symbol table entries need to be read in. */ |
| 40 | |
| 41 | |
| 42 | #include "defs.h" |
| 43 | #include <ctype.h> |
| 44 | #include "gdb_string.h" |
| 45 | #include "symtab.h" |
| 46 | #include "bfd.h" |
| 47 | #include "symfile.h" |
| 48 | #include "objfiles.h" |
| 49 | #include "demangle.h" |
| 50 | #include "value.h" |
| 51 | #include "cp-abi.h" |
| 52 | |
| 53 | /* Accumulate the minimal symbols for each objfile in bunches of BUNCH_SIZE. |
| 54 | At the end, copy them all into one newly allocated location on an objfile's |
| 55 | symbol obstack. */ |
| 56 | |
| 57 | #define BUNCH_SIZE 127 |
| 58 | |
| 59 | struct msym_bunch |
| 60 | { |
| 61 | struct msym_bunch *next; |
| 62 | struct minimal_symbol contents[BUNCH_SIZE]; |
| 63 | }; |
| 64 | |
| 65 | /* Bunch currently being filled up. |
| 66 | The next field points to chain of filled bunches. */ |
| 67 | |
| 68 | static struct msym_bunch *msym_bunch; |
| 69 | |
| 70 | /* Number of slots filled in current bunch. */ |
| 71 | |
| 72 | static int msym_bunch_index; |
| 73 | |
| 74 | /* Total number of minimal symbols recorded so far for the objfile. */ |
| 75 | |
| 76 | static int msym_count; |
| 77 | |
| 78 | /* Prototypes for local functions. */ |
| 79 | |
| 80 | static int compare_minimal_symbols (const PTR, const PTR); |
| 81 | |
| 82 | static int |
| 83 | compact_minimal_symbols (struct minimal_symbol *, int, struct objfile *); |
| 84 | |
| 85 | static void add_minsym_to_demangled_hash_table (struct minimal_symbol *sym, |
| 86 | struct minimal_symbol **table); |
| 87 | |
| 88 | /* Compute a hash code based using the same criteria as `strcmp_iw'. */ |
| 89 | |
| 90 | unsigned int |
| 91 | msymbol_hash_iw (const char *string) |
| 92 | { |
| 93 | unsigned int hash = 0; |
| 94 | while (*string && *string != '(') |
| 95 | { |
| 96 | while (isspace (*string)) |
| 97 | ++string; |
| 98 | if (*string && *string != '(') |
| 99 | hash = (31 * hash) + *string; |
| 100 | ++string; |
| 101 | } |
| 102 | return hash % MINIMAL_SYMBOL_HASH_SIZE; |
| 103 | } |
| 104 | |
| 105 | /* Compute a hash code for a string. */ |
| 106 | |
| 107 | unsigned int |
| 108 | msymbol_hash (const char *string) |
| 109 | { |
| 110 | unsigned int hash = 0; |
| 111 | for (; *string; ++string) |
| 112 | hash = (31 * hash) + *string; |
| 113 | return hash % MINIMAL_SYMBOL_HASH_SIZE; |
| 114 | } |
| 115 | |
| 116 | /* Add the minimal symbol SYM to an objfile's minsym hash table, TABLE. */ |
| 117 | void |
| 118 | add_minsym_to_hash_table (struct minimal_symbol *sym, |
| 119 | struct minimal_symbol **table) |
| 120 | { |
| 121 | if (sym->hash_next == NULL) |
| 122 | { |
| 123 | unsigned int hash = msymbol_hash (SYMBOL_NAME (sym)); |
| 124 | sym->hash_next = table[hash]; |
| 125 | table[hash] = sym; |
| 126 | } |
| 127 | } |
| 128 | |
| 129 | /* Add the minimal symbol SYM to an objfile's minsym demangled hash table, |
| 130 | TABLE. */ |
| 131 | static void |
| 132 | add_minsym_to_demangled_hash_table (struct minimal_symbol *sym, |
| 133 | struct minimal_symbol **table) |
| 134 | { |
| 135 | if (sym->demangled_hash_next == NULL) |
| 136 | { |
| 137 | unsigned int hash = msymbol_hash_iw (SYMBOL_DEMANGLED_NAME (sym)); |
| 138 | sym->demangled_hash_next = table[hash]; |
| 139 | table[hash] = sym; |
| 140 | } |
| 141 | } |
| 142 | |
| 143 | |
| 144 | /* Look through all the current minimal symbol tables and find the |
| 145 | first minimal symbol that matches NAME. If OBJF is non-NULL, limit |
| 146 | the search to that objfile. If SFILE is non-NULL, limit the search |
| 147 | to that source file. Returns a pointer to the minimal symbol that |
| 148 | matches, or NULL if no match is found. |
| 149 | |
| 150 | Note: One instance where there may be duplicate minimal symbols with |
| 151 | the same name is when the symbol tables for a shared library and the |
| 152 | symbol tables for an executable contain global symbols with the same |
| 153 | names (the dynamic linker deals with the duplication). */ |
| 154 | |
| 155 | struct minimal_symbol * |
| 156 | lookup_minimal_symbol (register const char *name, const char *sfile, |
| 157 | struct objfile *objf) |
| 158 | { |
| 159 | struct objfile *objfile; |
| 160 | struct minimal_symbol *msymbol; |
| 161 | struct minimal_symbol *found_symbol = NULL; |
| 162 | struct minimal_symbol *found_file_symbol = NULL; |
| 163 | struct minimal_symbol *trampoline_symbol = NULL; |
| 164 | |
| 165 | unsigned int hash = msymbol_hash (name); |
| 166 | unsigned int dem_hash = msymbol_hash_iw (name); |
| 167 | |
| 168 | #ifdef SOFUN_ADDRESS_MAYBE_MISSING |
| 169 | if (sfile != NULL) |
| 170 | { |
| 171 | char *p = strrchr (sfile, '/'); |
| 172 | if (p != NULL) |
| 173 | sfile = p + 1; |
| 174 | } |
| 175 | #endif |
| 176 | |
| 177 | for (objfile = object_files; |
| 178 | objfile != NULL && found_symbol == NULL; |
| 179 | objfile = objfile->next) |
| 180 | { |
| 181 | if (objf == NULL || objf == objfile) |
| 182 | { |
| 183 | /* Do two passes: the first over the ordinary hash table, |
| 184 | and the second over the demangled hash table. */ |
| 185 | int pass; |
| 186 | |
| 187 | for (pass = 1; pass <= 2 && found_symbol == NULL; pass++) |
| 188 | { |
| 189 | /* Select hash list according to pass. */ |
| 190 | if (pass == 1) |
| 191 | msymbol = objfile->msymbol_hash[hash]; |
| 192 | else |
| 193 | msymbol = objfile->msymbol_demangled_hash[dem_hash]; |
| 194 | |
| 195 | while (msymbol != NULL && found_symbol == NULL) |
| 196 | { |
| 197 | if (SYMBOL_MATCHES_NAME (msymbol, name)) |
| 198 | { |
| 199 | switch (MSYMBOL_TYPE (msymbol)) |
| 200 | { |
| 201 | case mst_file_text: |
| 202 | case mst_file_data: |
| 203 | case mst_file_bss: |
| 204 | #ifdef SOFUN_ADDRESS_MAYBE_MISSING |
| 205 | if (sfile == NULL || STREQ (msymbol->filename, sfile)) |
| 206 | found_file_symbol = msymbol; |
| 207 | #else |
| 208 | /* We have neither the ability nor the need to |
| 209 | deal with the SFILE parameter. If we find |
| 210 | more than one symbol, just return the latest |
| 211 | one (the user can't expect useful behavior in |
| 212 | that case). */ |
| 213 | found_file_symbol = msymbol; |
| 214 | #endif |
| 215 | break; |
| 216 | |
| 217 | case mst_solib_trampoline: |
| 218 | |
| 219 | /* If a trampoline symbol is found, we prefer to |
| 220 | keep looking for the *real* symbol. If the |
| 221 | actual symbol is not found, then we'll use the |
| 222 | trampoline entry. */ |
| 223 | if (trampoline_symbol == NULL) |
| 224 | trampoline_symbol = msymbol; |
| 225 | break; |
| 226 | |
| 227 | case mst_unknown: |
| 228 | default: |
| 229 | found_symbol = msymbol; |
| 230 | break; |
| 231 | } |
| 232 | } |
| 233 | |
| 234 | /* Find the next symbol on the hash chain. */ |
| 235 | if (pass == 1) |
| 236 | msymbol = msymbol->hash_next; |
| 237 | else |
| 238 | msymbol = msymbol->demangled_hash_next; |
| 239 | } |
| 240 | } |
| 241 | } |
| 242 | } |
| 243 | /* External symbols are best. */ |
| 244 | if (found_symbol) |
| 245 | return found_symbol; |
| 246 | |
| 247 | /* File-local symbols are next best. */ |
| 248 | if (found_file_symbol) |
| 249 | return found_file_symbol; |
| 250 | |
| 251 | /* Symbols for shared library trampolines are next best. */ |
| 252 | if (trampoline_symbol) |
| 253 | return trampoline_symbol; |
| 254 | |
| 255 | return NULL; |
| 256 | } |
| 257 | |
| 258 | /* Look through all the current minimal symbol tables and find the |
| 259 | first minimal symbol that matches NAME and of text type. |
| 260 | If OBJF is non-NULL, limit |
| 261 | the search to that objfile. If SFILE is non-NULL, limit the search |
| 262 | to that source file. Returns a pointer to the minimal symbol that |
| 263 | matches, or NULL if no match is found. |
| 264 | */ |
| 265 | |
| 266 | struct minimal_symbol * |
| 267 | lookup_minimal_symbol_text (register const char *name, const char *sfile, |
| 268 | struct objfile *objf) |
| 269 | { |
| 270 | struct objfile *objfile; |
| 271 | struct minimal_symbol *msymbol; |
| 272 | struct minimal_symbol *found_symbol = NULL; |
| 273 | struct minimal_symbol *found_file_symbol = NULL; |
| 274 | |
| 275 | #ifdef SOFUN_ADDRESS_MAYBE_MISSING |
| 276 | if (sfile != NULL) |
| 277 | { |
| 278 | char *p = strrchr (sfile, '/'); |
| 279 | if (p != NULL) |
| 280 | sfile = p + 1; |
| 281 | } |
| 282 | #endif |
| 283 | |
| 284 | for (objfile = object_files; |
| 285 | objfile != NULL && found_symbol == NULL; |
| 286 | objfile = objfile->next) |
| 287 | { |
| 288 | if (objf == NULL || objf == objfile) |
| 289 | { |
| 290 | for (msymbol = objfile->msymbols; |
| 291 | msymbol != NULL && SYMBOL_NAME (msymbol) != NULL && |
| 292 | found_symbol == NULL; |
| 293 | msymbol++) |
| 294 | { |
| 295 | if (SYMBOL_MATCHES_NAME (msymbol, name) && |
| 296 | (MSYMBOL_TYPE (msymbol) == mst_text || |
| 297 | MSYMBOL_TYPE (msymbol) == mst_file_text)) |
| 298 | { |
| 299 | switch (MSYMBOL_TYPE (msymbol)) |
| 300 | { |
| 301 | case mst_file_text: |
| 302 | #ifdef SOFUN_ADDRESS_MAYBE_MISSING |
| 303 | if (sfile == NULL || STREQ (msymbol->filename, sfile)) |
| 304 | found_file_symbol = msymbol; |
| 305 | #else |
| 306 | /* We have neither the ability nor the need to |
| 307 | deal with the SFILE parameter. If we find |
| 308 | more than one symbol, just return the latest |
| 309 | one (the user can't expect useful behavior in |
| 310 | that case). */ |
| 311 | found_file_symbol = msymbol; |
| 312 | #endif |
| 313 | break; |
| 314 | default: |
| 315 | found_symbol = msymbol; |
| 316 | break; |
| 317 | } |
| 318 | } |
| 319 | } |
| 320 | } |
| 321 | } |
| 322 | /* External symbols are best. */ |
| 323 | if (found_symbol) |
| 324 | return found_symbol; |
| 325 | |
| 326 | /* File-local symbols are next best. */ |
| 327 | if (found_file_symbol) |
| 328 | return found_file_symbol; |
| 329 | |
| 330 | return NULL; |
| 331 | } |
| 332 | |
| 333 | /* Look through all the current minimal symbol tables and find the |
| 334 | first minimal symbol that matches NAME and of solib trampoline type. |
| 335 | If OBJF is non-NULL, limit |
| 336 | the search to that objfile. If SFILE is non-NULL, limit the search |
| 337 | to that source file. Returns a pointer to the minimal symbol that |
| 338 | matches, or NULL if no match is found. |
| 339 | */ |
| 340 | |
| 341 | struct minimal_symbol * |
| 342 | lookup_minimal_symbol_solib_trampoline (register const char *name, |
| 343 | const char *sfile, struct objfile *objf) |
| 344 | { |
| 345 | struct objfile *objfile; |
| 346 | struct minimal_symbol *msymbol; |
| 347 | struct minimal_symbol *found_symbol = NULL; |
| 348 | |
| 349 | #ifdef SOFUN_ADDRESS_MAYBE_MISSING |
| 350 | if (sfile != NULL) |
| 351 | { |
| 352 | char *p = strrchr (sfile, '/'); |
| 353 | if (p != NULL) |
| 354 | sfile = p + 1; |
| 355 | } |
| 356 | #endif |
| 357 | |
| 358 | for (objfile = object_files; |
| 359 | objfile != NULL && found_symbol == NULL; |
| 360 | objfile = objfile->next) |
| 361 | { |
| 362 | if (objf == NULL || objf == objfile) |
| 363 | { |
| 364 | for (msymbol = objfile->msymbols; |
| 365 | msymbol != NULL && SYMBOL_NAME (msymbol) != NULL && |
| 366 | found_symbol == NULL; |
| 367 | msymbol++) |
| 368 | { |
| 369 | if (SYMBOL_MATCHES_NAME (msymbol, name) && |
| 370 | MSYMBOL_TYPE (msymbol) == mst_solib_trampoline) |
| 371 | return msymbol; |
| 372 | } |
| 373 | } |
| 374 | } |
| 375 | |
| 376 | return NULL; |
| 377 | } |
| 378 | |
| 379 | |
| 380 | /* Search through the minimal symbol table for each objfile and find |
| 381 | the symbol whose address is the largest address that is still less |
| 382 | than or equal to PC, and matches SECTION (if non-null). Returns a |
| 383 | pointer to the minimal symbol if such a symbol is found, or NULL if |
| 384 | PC is not in a suitable range. Note that we need to look through |
| 385 | ALL the minimal symbol tables before deciding on the symbol that |
| 386 | comes closest to the specified PC. This is because objfiles can |
| 387 | overlap, for example objfile A has .text at 0x100 and .data at |
| 388 | 0x40000 and objfile B has .text at 0x234 and .data at 0x40048. */ |
| 389 | |
| 390 | struct minimal_symbol * |
| 391 | lookup_minimal_symbol_by_pc_section (CORE_ADDR pc, asection *section) |
| 392 | { |
| 393 | int lo; |
| 394 | int hi; |
| 395 | int new; |
| 396 | struct objfile *objfile; |
| 397 | struct minimal_symbol *msymbol; |
| 398 | struct minimal_symbol *best_symbol = NULL; |
| 399 | |
| 400 | /* pc has to be in a known section. This ensures that anything beyond |
| 401 | the end of the last segment doesn't appear to be part of the last |
| 402 | function in the last segment. */ |
| 403 | if (find_pc_section (pc) == NULL) |
| 404 | return NULL; |
| 405 | |
| 406 | for (objfile = object_files; |
| 407 | objfile != NULL; |
| 408 | objfile = objfile->next) |
| 409 | { |
| 410 | /* If this objfile has a minimal symbol table, go search it using |
| 411 | a binary search. Note that a minimal symbol table always consists |
| 412 | of at least two symbols, a "real" symbol and the terminating |
| 413 | "null symbol". If there are no real symbols, then there is no |
| 414 | minimal symbol table at all. */ |
| 415 | |
| 416 | if ((msymbol = objfile->msymbols) != NULL) |
| 417 | { |
| 418 | lo = 0; |
| 419 | hi = objfile->minimal_symbol_count - 1; |
| 420 | |
| 421 | /* This code assumes that the minimal symbols are sorted by |
| 422 | ascending address values. If the pc value is greater than or |
| 423 | equal to the first symbol's address, then some symbol in this |
| 424 | minimal symbol table is a suitable candidate for being the |
| 425 | "best" symbol. This includes the last real symbol, for cases |
| 426 | where the pc value is larger than any address in this vector. |
| 427 | |
| 428 | By iterating until the address associated with the current |
| 429 | hi index (the endpoint of the test interval) is less than |
| 430 | or equal to the desired pc value, we accomplish two things: |
| 431 | (1) the case where the pc value is larger than any minimal |
| 432 | symbol address is trivially solved, (2) the address associated |
| 433 | with the hi index is always the one we want when the interation |
| 434 | terminates. In essence, we are iterating the test interval |
| 435 | down until the pc value is pushed out of it from the high end. |
| 436 | |
| 437 | Warning: this code is trickier than it would appear at first. */ |
| 438 | |
| 439 | /* Should also require that pc is <= end of objfile. FIXME! */ |
| 440 | if (pc >= SYMBOL_VALUE_ADDRESS (&msymbol[lo])) |
| 441 | { |
| 442 | while (SYMBOL_VALUE_ADDRESS (&msymbol[hi]) > pc) |
| 443 | { |
| 444 | /* pc is still strictly less than highest address */ |
| 445 | /* Note "new" will always be >= lo */ |
| 446 | new = (lo + hi) / 2; |
| 447 | if ((SYMBOL_VALUE_ADDRESS (&msymbol[new]) >= pc) || |
| 448 | (lo == new)) |
| 449 | { |
| 450 | hi = new; |
| 451 | } |
| 452 | else |
| 453 | { |
| 454 | lo = new; |
| 455 | } |
| 456 | } |
| 457 | |
| 458 | /* If we have multiple symbols at the same address, we want |
| 459 | hi to point to the last one. That way we can find the |
| 460 | right symbol if it has an index greater than hi. */ |
| 461 | while (hi < objfile->minimal_symbol_count - 1 |
| 462 | && (SYMBOL_VALUE_ADDRESS (&msymbol[hi]) |
| 463 | == SYMBOL_VALUE_ADDRESS (&msymbol[hi + 1]))) |
| 464 | hi++; |
| 465 | |
| 466 | /* The minimal symbol indexed by hi now is the best one in this |
| 467 | objfile's minimal symbol table. See if it is the best one |
| 468 | overall. */ |
| 469 | |
| 470 | /* Skip any absolute symbols. This is apparently what adb |
| 471 | and dbx do, and is needed for the CM-5. There are two |
| 472 | known possible problems: (1) on ELF, apparently end, edata, |
| 473 | etc. are absolute. Not sure ignoring them here is a big |
| 474 | deal, but if we want to use them, the fix would go in |
| 475 | elfread.c. (2) I think shared library entry points on the |
| 476 | NeXT are absolute. If we want special handling for this |
| 477 | it probably should be triggered by a special |
| 478 | mst_abs_or_lib or some such. */ |
| 479 | while (hi >= 0 |
| 480 | && msymbol[hi].type == mst_abs) |
| 481 | --hi; |
| 482 | |
| 483 | /* If "section" specified, skip any symbol from wrong section */ |
| 484 | /* This is the new code that distinguishes it from the old function */ |
| 485 | if (section) |
| 486 | while (hi >= 0 |
| 487 | /* Some types of debug info, such as COFF, |
| 488 | don't fill the bfd_section member, so don't |
| 489 | throw away symbols on those platforms. */ |
| 490 | && SYMBOL_BFD_SECTION (&msymbol[hi]) != NULL |
| 491 | && SYMBOL_BFD_SECTION (&msymbol[hi]) != section) |
| 492 | --hi; |
| 493 | |
| 494 | if (hi >= 0 |
| 495 | && ((best_symbol == NULL) || |
| 496 | (SYMBOL_VALUE_ADDRESS (best_symbol) < |
| 497 | SYMBOL_VALUE_ADDRESS (&msymbol[hi])))) |
| 498 | { |
| 499 | best_symbol = &msymbol[hi]; |
| 500 | } |
| 501 | } |
| 502 | } |
| 503 | } |
| 504 | return (best_symbol); |
| 505 | } |
| 506 | |
| 507 | /* Backward compatibility: search through the minimal symbol table |
| 508 | for a matching PC (no section given) */ |
| 509 | |
| 510 | struct minimal_symbol * |
| 511 | lookup_minimal_symbol_by_pc (CORE_ADDR pc) |
| 512 | { |
| 513 | return lookup_minimal_symbol_by_pc_section (pc, find_pc_mapped_section (pc)); |
| 514 | } |
| 515 | |
| 516 | #ifdef SOFUN_ADDRESS_MAYBE_MISSING |
| 517 | CORE_ADDR |
| 518 | find_stab_function_addr (char *namestring, char *filename, |
| 519 | struct objfile *objfile) |
| 520 | { |
| 521 | struct minimal_symbol *msym; |
| 522 | char *p; |
| 523 | int n; |
| 524 | |
| 525 | p = strchr (namestring, ':'); |
| 526 | if (p == NULL) |
| 527 | p = namestring; |
| 528 | n = p - namestring; |
| 529 | p = alloca (n + 2); |
| 530 | strncpy (p, namestring, n); |
| 531 | p[n] = 0; |
| 532 | |
| 533 | msym = lookup_minimal_symbol (p, filename, objfile); |
| 534 | if (msym == NULL) |
| 535 | { |
| 536 | /* Sun Fortran appends an underscore to the minimal symbol name, |
| 537 | try again with an appended underscore if the minimal symbol |
| 538 | was not found. */ |
| 539 | p[n] = '_'; |
| 540 | p[n + 1] = 0; |
| 541 | msym = lookup_minimal_symbol (p, filename, objfile); |
| 542 | } |
| 543 | |
| 544 | if (msym == NULL && filename != NULL) |
| 545 | { |
| 546 | /* Try again without the filename. */ |
| 547 | p[n] = 0; |
| 548 | msym = lookup_minimal_symbol (p, 0, objfile); |
| 549 | } |
| 550 | if (msym == NULL && filename != NULL) |
| 551 | { |
| 552 | /* And try again for Sun Fortran, but without the filename. */ |
| 553 | p[n] = '_'; |
| 554 | p[n + 1] = 0; |
| 555 | msym = lookup_minimal_symbol (p, 0, objfile); |
| 556 | } |
| 557 | |
| 558 | return msym == NULL ? 0 : SYMBOL_VALUE_ADDRESS (msym); |
| 559 | } |
| 560 | #endif /* SOFUN_ADDRESS_MAYBE_MISSING */ |
| 561 | \f |
| 562 | |
| 563 | /* Return leading symbol character for a BFD. If BFD is NULL, |
| 564 | return the leading symbol character from the main objfile. */ |
| 565 | |
| 566 | static int get_symbol_leading_char (bfd *); |
| 567 | |
| 568 | static int |
| 569 | get_symbol_leading_char (bfd *abfd) |
| 570 | { |
| 571 | if (abfd != NULL) |
| 572 | return bfd_get_symbol_leading_char (abfd); |
| 573 | if (symfile_objfile != NULL && symfile_objfile->obfd != NULL) |
| 574 | return bfd_get_symbol_leading_char (symfile_objfile->obfd); |
| 575 | return 0; |
| 576 | } |
| 577 | |
| 578 | /* Prepare to start collecting minimal symbols. Note that presetting |
| 579 | msym_bunch_index to BUNCH_SIZE causes the first call to save a minimal |
| 580 | symbol to allocate the memory for the first bunch. */ |
| 581 | |
| 582 | void |
| 583 | init_minimal_symbol_collection (void) |
| 584 | { |
| 585 | msym_count = 0; |
| 586 | msym_bunch = NULL; |
| 587 | msym_bunch_index = BUNCH_SIZE; |
| 588 | } |
| 589 | |
| 590 | void |
| 591 | prim_record_minimal_symbol (const char *name, CORE_ADDR address, |
| 592 | enum minimal_symbol_type ms_type, |
| 593 | struct objfile *objfile) |
| 594 | { |
| 595 | int section; |
| 596 | |
| 597 | switch (ms_type) |
| 598 | { |
| 599 | case mst_text: |
| 600 | case mst_file_text: |
| 601 | case mst_solib_trampoline: |
| 602 | section = SECT_OFF_TEXT (objfile); |
| 603 | break; |
| 604 | case mst_data: |
| 605 | case mst_file_data: |
| 606 | section = SECT_OFF_DATA (objfile); |
| 607 | break; |
| 608 | case mst_bss: |
| 609 | case mst_file_bss: |
| 610 | section = SECT_OFF_BSS (objfile); |
| 611 | break; |
| 612 | default: |
| 613 | section = -1; |
| 614 | } |
| 615 | |
| 616 | prim_record_minimal_symbol_and_info (name, address, ms_type, |
| 617 | NULL, section, NULL, objfile); |
| 618 | } |
| 619 | |
| 620 | /* Record a minimal symbol in the msym bunches. Returns the symbol |
| 621 | newly created. */ |
| 622 | |
| 623 | struct minimal_symbol * |
| 624 | prim_record_minimal_symbol_and_info (const char *name, CORE_ADDR address, |
| 625 | enum minimal_symbol_type ms_type, |
| 626 | char *info, int section, |
| 627 | asection *bfd_section, |
| 628 | struct objfile *objfile) |
| 629 | { |
| 630 | register struct msym_bunch *new; |
| 631 | register struct minimal_symbol *msymbol; |
| 632 | |
| 633 | if (ms_type == mst_file_text) |
| 634 | { |
| 635 | /* Don't put gcc_compiled, __gnu_compiled_cplus, and friends into |
| 636 | the minimal symbols, because if there is also another symbol |
| 637 | at the same address (e.g. the first function of the file), |
| 638 | lookup_minimal_symbol_by_pc would have no way of getting the |
| 639 | right one. */ |
| 640 | if (name[0] == 'g' |
| 641 | && (strcmp (name, GCC_COMPILED_FLAG_SYMBOL) == 0 |
| 642 | || strcmp (name, GCC2_COMPILED_FLAG_SYMBOL) == 0)) |
| 643 | return (NULL); |
| 644 | |
| 645 | { |
| 646 | const char *tempstring = name; |
| 647 | if (tempstring[0] == get_symbol_leading_char (objfile->obfd)) |
| 648 | ++tempstring; |
| 649 | if (STREQN (tempstring, "__gnu_compiled", 14)) |
| 650 | return (NULL); |
| 651 | } |
| 652 | } |
| 653 | |
| 654 | if (msym_bunch_index == BUNCH_SIZE) |
| 655 | { |
| 656 | new = (struct msym_bunch *) xmalloc (sizeof (struct msym_bunch)); |
| 657 | msym_bunch_index = 0; |
| 658 | new->next = msym_bunch; |
| 659 | msym_bunch = new; |
| 660 | } |
| 661 | msymbol = &msym_bunch->contents[msym_bunch_index]; |
| 662 | SYMBOL_NAME (msymbol) = obsavestring ((char *) name, strlen (name), |
| 663 | &objfile->symbol_obstack); |
| 664 | SYMBOL_INIT_LANGUAGE_SPECIFIC (msymbol, language_unknown); |
| 665 | SYMBOL_VALUE_ADDRESS (msymbol) = address; |
| 666 | SYMBOL_SECTION (msymbol) = section; |
| 667 | SYMBOL_BFD_SECTION (msymbol) = bfd_section; |
| 668 | |
| 669 | MSYMBOL_TYPE (msymbol) = ms_type; |
| 670 | /* FIXME: This info, if it remains, needs its own field. */ |
| 671 | MSYMBOL_INFO (msymbol) = info; /* FIXME! */ |
| 672 | |
| 673 | /* The hash pointers must be cleared! If they're not, |
| 674 | add_minsym_to_hash_table will NOT add this msymbol to the hash table. */ |
| 675 | msymbol->hash_next = NULL; |
| 676 | msymbol->demangled_hash_next = NULL; |
| 677 | |
| 678 | msym_bunch_index++; |
| 679 | msym_count++; |
| 680 | OBJSTAT (objfile, n_minsyms++); |
| 681 | return msymbol; |
| 682 | } |
| 683 | |
| 684 | /* Compare two minimal symbols by address and return a signed result based |
| 685 | on unsigned comparisons, so that we sort into unsigned numeric order. |
| 686 | Within groups with the same address, sort by name. */ |
| 687 | |
| 688 | static int |
| 689 | compare_minimal_symbols (const PTR fn1p, const PTR fn2p) |
| 690 | { |
| 691 | register const struct minimal_symbol *fn1; |
| 692 | register const struct minimal_symbol *fn2; |
| 693 | |
| 694 | fn1 = (const struct minimal_symbol *) fn1p; |
| 695 | fn2 = (const struct minimal_symbol *) fn2p; |
| 696 | |
| 697 | if (SYMBOL_VALUE_ADDRESS (fn1) < SYMBOL_VALUE_ADDRESS (fn2)) |
| 698 | { |
| 699 | return (-1); /* addr 1 is less than addr 2 */ |
| 700 | } |
| 701 | else if (SYMBOL_VALUE_ADDRESS (fn1) > SYMBOL_VALUE_ADDRESS (fn2)) |
| 702 | { |
| 703 | return (1); /* addr 1 is greater than addr 2 */ |
| 704 | } |
| 705 | else |
| 706 | /* addrs are equal: sort by name */ |
| 707 | { |
| 708 | char *name1 = SYMBOL_NAME (fn1); |
| 709 | char *name2 = SYMBOL_NAME (fn2); |
| 710 | |
| 711 | if (name1 && name2) /* both have names */ |
| 712 | return strcmp (name1, name2); |
| 713 | else if (name2) |
| 714 | return 1; /* fn1 has no name, so it is "less" */ |
| 715 | else if (name1) /* fn2 has no name, so it is "less" */ |
| 716 | return -1; |
| 717 | else |
| 718 | return (0); /* neither has a name, so they're equal. */ |
| 719 | } |
| 720 | } |
| 721 | |
| 722 | /* Discard the currently collected minimal symbols, if any. If we wish |
| 723 | to save them for later use, we must have already copied them somewhere |
| 724 | else before calling this function. |
| 725 | |
| 726 | FIXME: We could allocate the minimal symbol bunches on their own |
| 727 | obstack and then simply blow the obstack away when we are done with |
| 728 | it. Is it worth the extra trouble though? */ |
| 729 | |
| 730 | static void |
| 731 | do_discard_minimal_symbols_cleanup (void *arg) |
| 732 | { |
| 733 | register struct msym_bunch *next; |
| 734 | |
| 735 | while (msym_bunch != NULL) |
| 736 | { |
| 737 | next = msym_bunch->next; |
| 738 | xfree (msym_bunch); |
| 739 | msym_bunch = next; |
| 740 | } |
| 741 | } |
| 742 | |
| 743 | struct cleanup * |
| 744 | make_cleanup_discard_minimal_symbols (void) |
| 745 | { |
| 746 | return make_cleanup (do_discard_minimal_symbols_cleanup, 0); |
| 747 | } |
| 748 | |
| 749 | |
| 750 | |
| 751 | /* Compact duplicate entries out of a minimal symbol table by walking |
| 752 | through the table and compacting out entries with duplicate addresses |
| 753 | and matching names. Return the number of entries remaining. |
| 754 | |
| 755 | On entry, the table resides between msymbol[0] and msymbol[mcount]. |
| 756 | On exit, it resides between msymbol[0] and msymbol[result_count]. |
| 757 | |
| 758 | When files contain multiple sources of symbol information, it is |
| 759 | possible for the minimal symbol table to contain many duplicate entries. |
| 760 | As an example, SVR4 systems use ELF formatted object files, which |
| 761 | usually contain at least two different types of symbol tables (a |
| 762 | standard ELF one and a smaller dynamic linking table), as well as |
| 763 | DWARF debugging information for files compiled with -g. |
| 764 | |
| 765 | Without compacting, the minimal symbol table for gdb itself contains |
| 766 | over a 1000 duplicates, about a third of the total table size. Aside |
| 767 | from the potential trap of not noticing that two successive entries |
| 768 | identify the same location, this duplication impacts the time required |
| 769 | to linearly scan the table, which is done in a number of places. So we |
| 770 | just do one linear scan here and toss out the duplicates. |
| 771 | |
| 772 | Note that we are not concerned here about recovering the space that |
| 773 | is potentially freed up, because the strings themselves are allocated |
| 774 | on the symbol_obstack, and will get automatically freed when the symbol |
| 775 | table is freed. The caller can free up the unused minimal symbols at |
| 776 | the end of the compacted region if their allocation strategy allows it. |
| 777 | |
| 778 | Also note we only go up to the next to last entry within the loop |
| 779 | and then copy the last entry explicitly after the loop terminates. |
| 780 | |
| 781 | Since the different sources of information for each symbol may |
| 782 | have different levels of "completeness", we may have duplicates |
| 783 | that have one entry with type "mst_unknown" and the other with a |
| 784 | known type. So if the one we are leaving alone has type mst_unknown, |
| 785 | overwrite its type with the type from the one we are compacting out. */ |
| 786 | |
| 787 | static int |
| 788 | compact_minimal_symbols (struct minimal_symbol *msymbol, int mcount, |
| 789 | struct objfile *objfile) |
| 790 | { |
| 791 | struct minimal_symbol *copyfrom; |
| 792 | struct minimal_symbol *copyto; |
| 793 | |
| 794 | if (mcount > 0) |
| 795 | { |
| 796 | copyfrom = copyto = msymbol; |
| 797 | while (copyfrom < msymbol + mcount - 1) |
| 798 | { |
| 799 | if (SYMBOL_VALUE_ADDRESS (copyfrom) == |
| 800 | SYMBOL_VALUE_ADDRESS ((copyfrom + 1)) && |
| 801 | (STREQ (SYMBOL_NAME (copyfrom), SYMBOL_NAME ((copyfrom + 1))))) |
| 802 | { |
| 803 | if (MSYMBOL_TYPE ((copyfrom + 1)) == mst_unknown) |
| 804 | { |
| 805 | MSYMBOL_TYPE ((copyfrom + 1)) = MSYMBOL_TYPE (copyfrom); |
| 806 | } |
| 807 | copyfrom++; |
| 808 | } |
| 809 | else |
| 810 | *copyto++ = *copyfrom++; |
| 811 | } |
| 812 | *copyto++ = *copyfrom++; |
| 813 | mcount = copyto - msymbol; |
| 814 | } |
| 815 | return (mcount); |
| 816 | } |
| 817 | |
| 818 | /* Build (or rebuild) the minimal symbol hash tables. This is necessary |
| 819 | after compacting or sorting the table since the entries move around |
| 820 | thus causing the internal minimal_symbol pointers to become jumbled. */ |
| 821 | |
| 822 | static void |
| 823 | build_minimal_symbol_hash_tables (struct objfile *objfile) |
| 824 | { |
| 825 | int i; |
| 826 | struct minimal_symbol *msym; |
| 827 | |
| 828 | /* Clear the hash tables. */ |
| 829 | for (i = 0; i < MINIMAL_SYMBOL_HASH_SIZE; i++) |
| 830 | { |
| 831 | objfile->msymbol_hash[i] = 0; |
| 832 | objfile->msymbol_demangled_hash[i] = 0; |
| 833 | } |
| 834 | |
| 835 | /* Now, (re)insert the actual entries. */ |
| 836 | for (i = objfile->minimal_symbol_count, msym = objfile->msymbols; |
| 837 | i > 0; |
| 838 | i--, msym++) |
| 839 | { |
| 840 | msym->hash_next = 0; |
| 841 | add_minsym_to_hash_table (msym, objfile->msymbol_hash); |
| 842 | |
| 843 | msym->demangled_hash_next = 0; |
| 844 | if (SYMBOL_DEMANGLED_NAME (msym) != NULL) |
| 845 | add_minsym_to_demangled_hash_table (msym, |
| 846 | objfile->msymbol_demangled_hash); |
| 847 | } |
| 848 | } |
| 849 | |
| 850 | /* Add the minimal symbols in the existing bunches to the objfile's official |
| 851 | minimal symbol table. In most cases there is no minimal symbol table yet |
| 852 | for this objfile, and the existing bunches are used to create one. Once |
| 853 | in a while (for shared libraries for example), we add symbols (e.g. common |
| 854 | symbols) to an existing objfile. |
| 855 | |
| 856 | Because of the way minimal symbols are collected, we generally have no way |
| 857 | of knowing what source language applies to any particular minimal symbol. |
| 858 | Specifically, we have no way of knowing if the minimal symbol comes from a |
| 859 | C++ compilation unit or not. So for the sake of supporting cached |
| 860 | demangled C++ names, we have no choice but to try and demangle each new one |
| 861 | that comes in. If the demangling succeeds, then we assume it is a C++ |
| 862 | symbol and set the symbol's language and demangled name fields |
| 863 | appropriately. Note that in order to avoid unnecessary demanglings, and |
| 864 | allocating obstack space that subsequently can't be freed for the demangled |
| 865 | names, we mark all newly added symbols with language_auto. After |
| 866 | compaction of the minimal symbols, we go back and scan the entire minimal |
| 867 | symbol table looking for these new symbols. For each new symbol we attempt |
| 868 | to demangle it, and if successful, record it as a language_cplus symbol |
| 869 | and cache the demangled form on the symbol obstack. Symbols which don't |
| 870 | demangle are marked as language_unknown symbols, which inhibits future |
| 871 | attempts to demangle them if we later add more minimal symbols. */ |
| 872 | |
| 873 | void |
| 874 | install_minimal_symbols (struct objfile *objfile) |
| 875 | { |
| 876 | register int bindex; |
| 877 | register int mcount; |
| 878 | register struct msym_bunch *bunch; |
| 879 | register struct minimal_symbol *msymbols; |
| 880 | int alloc_count; |
| 881 | register char leading_char; |
| 882 | |
| 883 | if (msym_count > 0) |
| 884 | { |
| 885 | /* Allocate enough space in the obstack, into which we will gather the |
| 886 | bunches of new and existing minimal symbols, sort them, and then |
| 887 | compact out the duplicate entries. Once we have a final table, |
| 888 | we will give back the excess space. */ |
| 889 | |
| 890 | alloc_count = msym_count + objfile->minimal_symbol_count + 1; |
| 891 | obstack_blank (&objfile->symbol_obstack, |
| 892 | alloc_count * sizeof (struct minimal_symbol)); |
| 893 | msymbols = (struct minimal_symbol *) |
| 894 | obstack_base (&objfile->symbol_obstack); |
| 895 | |
| 896 | /* Copy in the existing minimal symbols, if there are any. */ |
| 897 | |
| 898 | if (objfile->minimal_symbol_count) |
| 899 | memcpy ((char *) msymbols, (char *) objfile->msymbols, |
| 900 | objfile->minimal_symbol_count * sizeof (struct minimal_symbol)); |
| 901 | |
| 902 | /* Walk through the list of minimal symbol bunches, adding each symbol |
| 903 | to the new contiguous array of symbols. Note that we start with the |
| 904 | current, possibly partially filled bunch (thus we use the current |
| 905 | msym_bunch_index for the first bunch we copy over), and thereafter |
| 906 | each bunch is full. */ |
| 907 | |
| 908 | mcount = objfile->minimal_symbol_count; |
| 909 | leading_char = get_symbol_leading_char (objfile->obfd); |
| 910 | |
| 911 | for (bunch = msym_bunch; bunch != NULL; bunch = bunch->next) |
| 912 | { |
| 913 | for (bindex = 0; bindex < msym_bunch_index; bindex++, mcount++) |
| 914 | { |
| 915 | msymbols[mcount] = bunch->contents[bindex]; |
| 916 | SYMBOL_LANGUAGE (&msymbols[mcount]) = language_auto; |
| 917 | if (SYMBOL_NAME (&msymbols[mcount])[0] == leading_char) |
| 918 | { |
| 919 | SYMBOL_NAME (&msymbols[mcount])++; |
| 920 | } |
| 921 | } |
| 922 | msym_bunch_index = BUNCH_SIZE; |
| 923 | } |
| 924 | |
| 925 | /* Sort the minimal symbols by address. */ |
| 926 | |
| 927 | qsort (msymbols, mcount, sizeof (struct minimal_symbol), |
| 928 | compare_minimal_symbols); |
| 929 | |
| 930 | /* Compact out any duplicates, and free up whatever space we are |
| 931 | no longer using. */ |
| 932 | |
| 933 | mcount = compact_minimal_symbols (msymbols, mcount, objfile); |
| 934 | |
| 935 | obstack_blank (&objfile->symbol_obstack, |
| 936 | (mcount + 1 - alloc_count) * sizeof (struct minimal_symbol)); |
| 937 | msymbols = (struct minimal_symbol *) |
| 938 | obstack_finish (&objfile->symbol_obstack); |
| 939 | |
| 940 | /* We also terminate the minimal symbol table with a "null symbol", |
| 941 | which is *not* included in the size of the table. This makes it |
| 942 | easier to find the end of the table when we are handed a pointer |
| 943 | to some symbol in the middle of it. Zero out the fields in the |
| 944 | "null symbol" allocated at the end of the array. Note that the |
| 945 | symbol count does *not* include this null symbol, which is why it |
| 946 | is indexed by mcount and not mcount-1. */ |
| 947 | |
| 948 | SYMBOL_NAME (&msymbols[mcount]) = NULL; |
| 949 | SYMBOL_VALUE_ADDRESS (&msymbols[mcount]) = 0; |
| 950 | MSYMBOL_INFO (&msymbols[mcount]) = NULL; |
| 951 | MSYMBOL_TYPE (&msymbols[mcount]) = mst_unknown; |
| 952 | SYMBOL_INIT_LANGUAGE_SPECIFIC (&msymbols[mcount], language_unknown); |
| 953 | |
| 954 | /* Attach the minimal symbol table to the specified objfile. |
| 955 | The strings themselves are also located in the symbol_obstack |
| 956 | of this objfile. */ |
| 957 | |
| 958 | objfile->minimal_symbol_count = mcount; |
| 959 | objfile->msymbols = msymbols; |
| 960 | |
| 961 | /* Try to guess the appropriate C++ ABI by looking at the names |
| 962 | of the minimal symbols in the table. */ |
| 963 | { |
| 964 | int i; |
| 965 | |
| 966 | for (i = 0; i < mcount; i++) |
| 967 | { |
| 968 | const char *name = SYMBOL_NAME (&objfile->msymbols[i]); |
| 969 | if (name[0] == '_' && name[1] == 'Z') |
| 970 | { |
| 971 | switch_to_cp_abi ("gnu-v3"); |
| 972 | break; |
| 973 | } |
| 974 | } |
| 975 | } |
| 976 | |
| 977 | /* Now walk through all the minimal symbols, selecting the newly added |
| 978 | ones and attempting to cache their C++ demangled names. */ |
| 979 | for (; mcount-- > 0; msymbols++) |
| 980 | SYMBOL_INIT_DEMANGLED_NAME (msymbols, &objfile->symbol_obstack); |
| 981 | |
| 982 | /* Now build the hash tables; we can't do this incrementally |
| 983 | at an earlier point since we weren't finished with the obstack |
| 984 | yet. (And if the msymbol obstack gets moved, all the internal |
| 985 | pointers to other msymbols need to be adjusted.) */ |
| 986 | build_minimal_symbol_hash_tables (objfile); |
| 987 | } |
| 988 | } |
| 989 | |
| 990 | /* Sort all the minimal symbols in OBJFILE. */ |
| 991 | |
| 992 | void |
| 993 | msymbols_sort (struct objfile *objfile) |
| 994 | { |
| 995 | qsort (objfile->msymbols, objfile->minimal_symbol_count, |
| 996 | sizeof (struct minimal_symbol), compare_minimal_symbols); |
| 997 | build_minimal_symbol_hash_tables (objfile); |
| 998 | } |
| 999 | |
| 1000 | /* Check if PC is in a shared library trampoline code stub. |
| 1001 | Return minimal symbol for the trampoline entry or NULL if PC is not |
| 1002 | in a trampoline code stub. */ |
| 1003 | |
| 1004 | struct minimal_symbol * |
| 1005 | lookup_solib_trampoline_symbol_by_pc (CORE_ADDR pc) |
| 1006 | { |
| 1007 | struct minimal_symbol *msymbol = lookup_minimal_symbol_by_pc (pc); |
| 1008 | |
| 1009 | if (msymbol != NULL && MSYMBOL_TYPE (msymbol) == mst_solib_trampoline) |
| 1010 | return msymbol; |
| 1011 | return NULL; |
| 1012 | } |
| 1013 | |
| 1014 | /* If PC is in a shared library trampoline code stub, return the |
| 1015 | address of the `real' function belonging to the stub. |
| 1016 | Return 0 if PC is not in a trampoline code stub or if the real |
| 1017 | function is not found in the minimal symbol table. |
| 1018 | |
| 1019 | We may fail to find the right function if a function with the |
| 1020 | same name is defined in more than one shared library, but this |
| 1021 | is considered bad programming style. We could return 0 if we find |
| 1022 | a duplicate function in case this matters someday. */ |
| 1023 | |
| 1024 | CORE_ADDR |
| 1025 | find_solib_trampoline_target (CORE_ADDR pc) |
| 1026 | { |
| 1027 | struct objfile *objfile; |
| 1028 | struct minimal_symbol *msymbol; |
| 1029 | struct minimal_symbol *tsymbol = lookup_solib_trampoline_symbol_by_pc (pc); |
| 1030 | |
| 1031 | if (tsymbol != NULL) |
| 1032 | { |
| 1033 | ALL_MSYMBOLS (objfile, msymbol) |
| 1034 | { |
| 1035 | if (MSYMBOL_TYPE (msymbol) == mst_text |
| 1036 | && STREQ (SYMBOL_NAME (msymbol), SYMBOL_NAME (tsymbol))) |
| 1037 | return SYMBOL_VALUE_ADDRESS (msymbol); |
| 1038 | } |
| 1039 | } |
| 1040 | return 0; |
| 1041 | } |