1 /* 2 * Copyright (c) Ian F. Darwin 1986-1995. 3 * Software written by Ian F. Darwin and others; 4 * maintained 1995-present by Christos Zoulas and others. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice immediately at the beginning of the file, without modification, 11 * this list of conditions, and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR 20 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 */ 28 /* 29 * apprentice - make one pass through /etc/magic, learning its secrets. 30 */ 31 32 #include "file.h" 33 34 #ifndef lint 35 FILE_RCSID("@(#)$File: apprentice.c,v 1.283 2019/02/20 02:35:27 christos Exp $") 36 #endif /* lint */ 37 38 #include "magic.h" 39 #include <stdlib.h> 40 #ifdef HAVE_UNISTD_H 41 #include <unistd.h> 42 #endif 43 #include <stddef.h> 44 #include <string.h> 45 #include <assert.h> 46 #include <ctype.h> 47 #include <fcntl.h> 48 #ifdef QUICK 49 #include <sys/mman.h> 50 #endif 51 #include <dirent.h> 52 #include <limits.h> 53 54 55 #define EATAB {while (isascii(CAST(unsigned char, *l)) && \ 56 isspace(CAST(unsigned char, *l))) ++l;} 57 #define LOWCASE(l) (isupper(CAST(unsigned char, l)) ? \ 58 tolower(CAST(unsigned char, l)) : (l)) 59 /* 60 * Work around a bug in headers on Digital Unix. 61 * At least confirmed for: OSF1 V4.0 878 62 */ 63 #if defined(__osf__) && defined(__DECC) 64 #ifdef MAP_FAILED 65 #undef MAP_FAILED 66 #endif 67 #endif 68 69 #ifndef MAP_FAILED 70 #define MAP_FAILED (void *) -1 71 #endif 72 73 #ifndef MAP_FILE 74 #define MAP_FILE 0 75 #endif 76 77 #define ALLOC_CHUNK CAST(size_t, 10) 78 #define ALLOC_INCR CAST(size_t, 200) 79 80 #define MAP_TYPE_USER 0 81 #define MAP_TYPE_MALLOC 1 82 #define MAP_TYPE_MMAP 2 83 84 struct magic_entry { 85 struct magic *mp; 86 uint32_t cont_count; 87 uint32_t max_count; 88 }; 89 90 struct magic_entry_set { 91 struct magic_entry *me; 92 uint32_t count; 93 uint32_t max; 94 }; 95 96 struct magic_map { 97 void *p; 98 size_t len; 99 int type; 100 struct magic *magic[MAGIC_SETS]; 101 uint32_t nmagic[MAGIC_SETS]; 102 }; 103 104 int file_formats[FILE_NAMES_SIZE]; 105 const size_t file_nformats = FILE_NAMES_SIZE; 106 const char *file_names[FILE_NAMES_SIZE]; 107 const size_t file_nnames = FILE_NAMES_SIZE; 108 109 private int getvalue(struct magic_set *ms, struct magic *, const char **, int); 110 private int hextoint(int); 111 private const char *getstr(struct magic_set *, struct magic *, const char *, 112 int); 113 private int parse(struct magic_set *, struct magic_entry *, const char *, 114 size_t, int); 115 private void eatsize(const char **); 116 private int apprentice_1(struct magic_set *, const char *, int); 117 private size_t apprentice_magic_strength(const struct magic *); 118 private int apprentice_sort(const void *, const void *); 119 private void apprentice_list(struct mlist *, int ); 120 private struct magic_map *apprentice_load(struct magic_set *, 121 const char *, int); 122 private struct mlist *mlist_alloc(void); 123 private void mlist_free(struct mlist *); 124 private void byteswap(struct magic *, uint32_t); 125 private void bs1(struct magic *); 126 private uint16_t swap2(uint16_t); 127 private uint32_t swap4(uint32_t); 128 private uint64_t swap8(uint64_t); 129 private char *mkdbname(struct magic_set *, const char *, int); 130 private struct magic_map *apprentice_buf(struct magic_set *, struct magic *, 131 size_t); 132 private struct magic_map *apprentice_map(struct magic_set *, const char *); 133 private int check_buffer(struct magic_set *, struct magic_map *, const char *); 134 private void apprentice_unmap(struct magic_map *); 135 private int apprentice_compile(struct magic_set *, struct magic_map *, 136 const char *); 137 private int check_format_type(const char *, int, const char **); 138 private int check_format(struct magic_set *, struct magic *); 139 private int get_op(char); 140 private int parse_mime(struct magic_set *, struct magic_entry *, const char *); 141 private int parse_strength(struct magic_set *, struct magic_entry *, const char *); 142 private int parse_apple(struct magic_set *, struct magic_entry *, const char *); 143 private int parse_ext(struct magic_set *, struct magic_entry *, const char *); 144 145 146 private size_t magicsize = sizeof(struct magic); 147 148 private const char usg_hdr[] = "cont\toffset\ttype\topcode\tmask\tvalue\tdesc"; 149 150 private struct { 151 const char *name; 152 size_t len; 153 int (*fun)(struct magic_set *, struct magic_entry *, const char *); 154 } bang[] = { 155 #define DECLARE_FIELD(name) { # name, sizeof(# name) - 1, parse_ ## name } 156 DECLARE_FIELD(mime), 157 DECLARE_FIELD(apple), 158 DECLARE_FIELD(ext), 159 DECLARE_FIELD(strength), 160 #undef DECLARE_FIELD 161 { NULL, 0, NULL } 162 }; 163 164 #ifdef COMPILE_ONLY 165 166 int main(int, char *[]); 167 168 int 169 main(int argc, char *argv[]) 170 { 171 int ret; 172 struct magic_set *ms; 173 char *progname; 174 175 if ((progname = strrchr(argv[0], '/')) != NULL) 176 progname++; 177 else 178 progname = argv[0]; 179 180 if (argc != 2) { 181 (void)fprintf(stderr, "Usage: %s file\n", progname); 182 return 1; 183 } 184 185 if ((ms = magic_open(MAGIC_CHECK)) == NULL) { 186 (void)fprintf(stderr, "%s: %s\n", progname, strerror(errno)); 187 return 1; 188 } 189 ret = magic_compile(ms, argv[1]) == -1 ? 1 : 0; 190 if (ret == 1) 191 (void)fprintf(stderr, "%s: %s\n", progname, magic_error(ms)); 192 magic_close(ms); 193 return ret; 194 } 195 #endif /* COMPILE_ONLY */ 196 197 struct type_tbl_s { 198 const char name[16]; 199 const size_t len; 200 const int type; 201 const int format; 202 }; 203 204 /* 205 * XXX - the actual Single UNIX Specification says that "long" means "long", 206 * as in the C data type, but we treat it as meaning "4-byte integer". 207 * Given that the OS X version of file 5.04 did the same, I guess that passes 208 * the actual test; having "long" be dependent on how big a "long" is on 209 * the machine running "file" is silly. 210 */ 211 static const struct type_tbl_s type_tbl[] = { 212 # define XX(s) s, (sizeof(s) - 1) 213 # define XX_NULL "", 0 214 { XX("invalid"), FILE_INVALID, FILE_FMT_NONE }, 215 { XX("byte"), FILE_BYTE, FILE_FMT_NUM }, 216 { XX("short"), FILE_SHORT, FILE_FMT_NUM }, 217 { XX("default"), FILE_DEFAULT, FILE_FMT_NONE }, 218 { XX("long"), FILE_LONG, FILE_FMT_NUM }, 219 { XX("string"), FILE_STRING, FILE_FMT_STR }, 220 { XX("date"), FILE_DATE, FILE_FMT_STR }, 221 { XX("beshort"), FILE_BESHORT, FILE_FMT_NUM }, 222 { XX("belong"), FILE_BELONG, FILE_FMT_NUM }, 223 { XX("bedate"), FILE_BEDATE, FILE_FMT_STR }, 224 { XX("leshort"), FILE_LESHORT, FILE_FMT_NUM }, 225 { XX("lelong"), FILE_LELONG, FILE_FMT_NUM }, 226 { XX("ledate"), FILE_LEDATE, FILE_FMT_STR }, 227 { XX("pstring"), FILE_PSTRING, FILE_FMT_STR }, 228 { XX("ldate"), FILE_LDATE, FILE_FMT_STR }, 229 { XX("beldate"), FILE_BELDATE, FILE_FMT_STR }, 230 { XX("leldate"), FILE_LELDATE, FILE_FMT_STR }, 231 { XX("regex"), FILE_REGEX, FILE_FMT_STR }, 232 { XX("bestring16"), FILE_BESTRING16, FILE_FMT_STR }, 233 { XX("lestring16"), FILE_LESTRING16, FILE_FMT_STR }, 234 { XX("search"), FILE_SEARCH, FILE_FMT_STR }, 235 { XX("medate"), FILE_MEDATE, FILE_FMT_STR }, 236 { XX("meldate"), FILE_MELDATE, FILE_FMT_STR }, 237 { XX("melong"), FILE_MELONG, FILE_FMT_NUM }, 238 { XX("quad"), FILE_QUAD, FILE_FMT_QUAD }, 239 { XX("lequad"), FILE_LEQUAD, FILE_FMT_QUAD }, 240 { XX("bequad"), FILE_BEQUAD, FILE_FMT_QUAD }, 241 { XX("qdate"), FILE_QDATE, FILE_FMT_STR }, 242 { XX("leqdate"), FILE_LEQDATE, FILE_FMT_STR }, 243 { XX("beqdate"), FILE_BEQDATE, FILE_FMT_STR }, 244 { XX("qldate"), FILE_QLDATE, FILE_FMT_STR }, 245 { XX("leqldate"), FILE_LEQLDATE, FILE_FMT_STR }, 246 { XX("beqldate"), FILE_BEQLDATE, FILE_FMT_STR }, 247 { XX("float"), FILE_FLOAT, FILE_FMT_FLOAT }, 248 { XX("befloat"), FILE_BEFLOAT, FILE_FMT_FLOAT }, 249 { XX("lefloat"), FILE_LEFLOAT, FILE_FMT_FLOAT }, 250 { XX("double"), FILE_DOUBLE, FILE_FMT_DOUBLE }, 251 { XX("bedouble"), FILE_BEDOUBLE, FILE_FMT_DOUBLE }, 252 { XX("ledouble"), FILE_LEDOUBLE, FILE_FMT_DOUBLE }, 253 { XX("leid3"), FILE_LEID3, FILE_FMT_NUM }, 254 { XX("beid3"), FILE_BEID3, FILE_FMT_NUM }, 255 { XX("indirect"), FILE_INDIRECT, FILE_FMT_NUM }, 256 { XX("qwdate"), FILE_QWDATE, FILE_FMT_STR }, 257 { XX("leqwdate"), FILE_LEQWDATE, FILE_FMT_STR }, 258 { XX("beqwdate"), FILE_BEQWDATE, FILE_FMT_STR }, 259 { XX("name"), FILE_NAME, FILE_FMT_NONE }, 260 { XX("use"), FILE_USE, FILE_FMT_NONE }, 261 { XX("clear"), FILE_CLEAR, FILE_FMT_NONE }, 262 { XX("der"), FILE_DER, FILE_FMT_STR }, 263 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 264 }; 265 266 /* 267 * These are not types, and cannot be preceded by "u" to make them 268 * unsigned. 269 */ 270 static const struct type_tbl_s special_tbl[] = { 271 { XX("der"), FILE_DER, FILE_FMT_STR }, 272 { XX("name"), FILE_NAME, FILE_FMT_STR }, 273 { XX("use"), FILE_USE, FILE_FMT_STR }, 274 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 275 }; 276 # undef XX 277 # undef XX_NULL 278 279 private int 280 get_type(const struct type_tbl_s *tbl, const char *l, const char **t) 281 { 282 const struct type_tbl_s *p; 283 284 for (p = tbl; p->len; p++) { 285 if (strncmp(l, p->name, p->len) == 0) { 286 if (t) 287 *t = l + p->len; 288 break; 289 } 290 } 291 return p->type; 292 } 293 294 private off_t 295 maxoff_t(void) { 296 if (/*CONSTCOND*/sizeof(off_t) == sizeof(int)) 297 return CAST(off_t, INT_MAX); 298 if (/*CONSTCOND*/sizeof(off_t) == sizeof(long)) 299 return CAST(off_t, LONG_MAX); 300 return 0x7fffffff; 301 } 302 303 private int 304 get_standard_integer_type(const char *l, const char **t) 305 { 306 int type; 307 308 if (isalpha(CAST(unsigned char, l[1]))) { 309 switch (l[1]) { 310 case 'C': 311 /* "dC" and "uC" */ 312 type = FILE_BYTE; 313 break; 314 case 'S': 315 /* "dS" and "uS" */ 316 type = FILE_SHORT; 317 break; 318 case 'I': 319 case 'L': 320 /* 321 * "dI", "dL", "uI", and "uL". 322 * 323 * XXX - the actual Single UNIX Specification says 324 * that "L" means "long", as in the C data type, 325 * but we treat it as meaning "4-byte integer". 326 * Given that the OS X version of file 5.04 did 327 * the same, I guess that passes the actual SUS 328 * validation suite; having "dL" be dependent on 329 * how big a "long" is on the machine running 330 * "file" is silly. 331 */ 332 type = FILE_LONG; 333 break; 334 case 'Q': 335 /* "dQ" and "uQ" */ 336 type = FILE_QUAD; 337 break; 338 default: 339 /* "d{anything else}", "u{anything else}" */ 340 return FILE_INVALID; 341 } 342 l += 2; 343 } else if (isdigit(CAST(unsigned char, l[1]))) { 344 /* 345 * "d{num}" and "u{num}"; we only support {num} values 346 * of 1, 2, 4, and 8 - the Single UNIX Specification 347 * doesn't say anything about whether arbitrary 348 * values should be supported, but both the Solaris 10 349 * and OS X Mountain Lion versions of file passed the 350 * Single UNIX Specification validation suite, and 351 * neither of them support values bigger than 8 or 352 * non-power-of-2 values. 353 */ 354 if (isdigit(CAST(unsigned char, l[2]))) { 355 /* Multi-digit, so > 9 */ 356 return FILE_INVALID; 357 } 358 switch (l[1]) { 359 case '1': 360 type = FILE_BYTE; 361 break; 362 case '2': 363 type = FILE_SHORT; 364 break; 365 case '4': 366 type = FILE_LONG; 367 break; 368 case '8': 369 type = FILE_QUAD; 370 break; 371 default: 372 /* XXX - what about 3, 5, 6, or 7? */ 373 return FILE_INVALID; 374 } 375 l += 2; 376 } else { 377 /* 378 * "d" or "u" by itself. 379 */ 380 type = FILE_LONG; 381 ++l; 382 } 383 if (t) 384 *t = l; 385 return type; 386 } 387 388 private void 389 init_file_tables(void) 390 { 391 static int done = 0; 392 const struct type_tbl_s *p; 393 394 if (done) 395 return; 396 done++; 397 398 for (p = type_tbl; p->len; p++) { 399 assert(p->type < FILE_NAMES_SIZE); 400 file_names[p->type] = p->name; 401 file_formats[p->type] = p->format; 402 } 403 assert(p - type_tbl == FILE_NAMES_SIZE); 404 } 405 406 private int 407 add_mlist(struct mlist *mlp, struct magic_map *map, size_t idx) 408 { 409 struct mlist *ml; 410 411 mlp->map = NULL; 412 if ((ml = CAST(struct mlist *, malloc(sizeof(*ml)))) == NULL) 413 return -1; 414 415 ml->map = idx == 0 ? map : NULL; 416 ml->magic = map->magic[idx]; 417 ml->nmagic = map->nmagic[idx]; 418 419 mlp->prev->next = ml; 420 ml->prev = mlp->prev; 421 ml->next = mlp; 422 mlp->prev = ml; 423 return 0; 424 } 425 426 /* 427 * Handle one file or directory. 428 */ 429 private int 430 apprentice_1(struct magic_set *ms, const char *fn, int action) 431 { 432 struct magic_map *map; 433 #ifndef COMPILE_ONLY 434 struct mlist *ml; 435 size_t i; 436 #endif 437 438 if (magicsize != FILE_MAGICSIZE) { 439 file_error(ms, 0, "magic element size %lu != %lu", 440 CAST(unsigned long, sizeof(*map->magic[0])), 441 CAST(unsigned long, FILE_MAGICSIZE)); 442 return -1; 443 } 444 445 if (action == FILE_COMPILE) { 446 map = apprentice_load(ms, fn, action); 447 if (map == NULL) 448 return -1; 449 return apprentice_compile(ms, map, fn); 450 } 451 452 #ifndef COMPILE_ONLY 453 map = apprentice_map(ms, fn); 454 if (map == RCAST(struct magic_map *, -1)) 455 return -1; 456 if (map == NULL) { 457 if (ms->flags & MAGIC_CHECK) 458 file_magwarn(ms, "using regular magic file `%s'", fn); 459 map = apprentice_load(ms, fn, action); 460 if (map == NULL) 461 return -1; 462 } 463 464 for (i = 0; i < MAGIC_SETS; i++) { 465 if (add_mlist(ms->mlist[i], map, i) == -1) { 466 file_oomem(ms, sizeof(*ml)); 467 return -1; 468 } 469 } 470 471 if (action == FILE_LIST) { 472 for (i = 0; i < MAGIC_SETS; i++) { 473 printf("Set %" SIZE_T_FORMAT "u:\nBinary patterns:\n", 474 i); 475 apprentice_list(ms->mlist[i], BINTEST); 476 printf("Text patterns:\n"); 477 apprentice_list(ms->mlist[i], TEXTTEST); 478 } 479 } 480 return 0; 481 #else 482 return 0; 483 #endif /* COMPILE_ONLY */ 484 } 485 486 protected void 487 file_ms_free(struct magic_set *ms) 488 { 489 size_t i; 490 if (ms == NULL) 491 return; 492 for (i = 0; i < MAGIC_SETS; i++) 493 mlist_free(ms->mlist[i]); 494 free(ms->o.pbuf); 495 free(ms->o.buf); 496 free(ms->c.li); 497 free(ms); 498 } 499 500 protected struct magic_set * 501 file_ms_alloc(int flags) 502 { 503 struct magic_set *ms; 504 size_t i, len; 505 506 if ((ms = CAST(struct magic_set *, calloc(CAST(size_t, 1u), 507 sizeof(struct magic_set)))) == NULL) 508 return NULL; 509 510 if (magic_setflags(ms, flags) == -1) { 511 errno = EINVAL; 512 goto free; 513 } 514 515 ms->o.buf = ms->o.pbuf = NULL; 516 len = (ms->c.len = 10) * sizeof(*ms->c.li); 517 518 if ((ms->c.li = CAST(struct level_info *, malloc(len))) == NULL) 519 goto free; 520 521 ms->event_flags = 0; 522 ms->error = -1; 523 for (i = 0; i < MAGIC_SETS; i++) 524 ms->mlist[i] = NULL; 525 ms->file = "unknown"; 526 ms->line = 0; 527 ms->indir_max = FILE_INDIR_MAX; 528 ms->name_max = FILE_NAME_MAX; 529 ms->elf_shnum_max = FILE_ELF_SHNUM_MAX; 530 ms->elf_phnum_max = FILE_ELF_PHNUM_MAX; 531 ms->elf_notes_max = FILE_ELF_NOTES_MAX; 532 ms->regex_max = FILE_REGEX_MAX; 533 ms->bytes_max = FILE_BYTES_MAX; 534 return ms; 535 free: 536 free(ms); 537 return NULL; 538 } 539 540 private void 541 apprentice_unmap(struct magic_map *map) 542 { 543 size_t i; 544 if (map == NULL) 545 return; 546 547 switch (map->type) { 548 case MAP_TYPE_USER: 549 break; 550 case MAP_TYPE_MALLOC: 551 for (i = 0; i < MAGIC_SETS; i++) { 552 void *b = map->magic[i]; 553 void *p = map->p; 554 if (CAST(char *, b) >= CAST(char *, p) && 555 CAST(char *, b) <= CAST(char *, p) + map->len) 556 continue; 557 free(map->magic[i]); 558 } 559 free(map->p); 560 break; 561 #ifdef QUICK 562 case MAP_TYPE_MMAP: 563 if (map->p && map->p != MAP_FAILED) 564 (void)munmap(map->p, map->len); 565 break; 566 #endif 567 default: 568 abort(); 569 } 570 free(map); 571 } 572 573 private struct mlist * 574 mlist_alloc(void) 575 { 576 struct mlist *mlist; 577 if ((mlist = CAST(struct mlist *, calloc(1, sizeof(*mlist)))) == NULL) { 578 return NULL; 579 } 580 mlist->next = mlist->prev = mlist; 581 return mlist; 582 } 583 584 private void 585 mlist_free_one(struct mlist *ml) 586 { 587 if (ml->map) 588 apprentice_unmap(CAST(struct magic_map *, ml->map)); 589 free(ml); 590 } 591 592 private void 593 mlist_free(struct mlist *mlist) 594 { 595 struct mlist *ml, *next; 596 597 if (mlist == NULL) 598 return; 599 600 for (ml = mlist->next; ml != mlist; ml = next) { 601 next = ml->next; 602 mlist_free_one(ml); 603 } 604 mlist_free_one(mlist); 605 } 606 607 #ifndef COMPILE_ONLY 608 /* void **bufs: an array of compiled magic files */ 609 protected int 610 buffer_apprentice(struct magic_set *ms, struct magic **bufs, 611 size_t *sizes, size_t nbufs) 612 { 613 size_t i, j; 614 struct mlist *ml; 615 struct magic_map *map; 616 617 if (nbufs == 0) 618 return -1; 619 620 (void)file_reset(ms, 0); 621 622 init_file_tables(); 623 624 for (i = 0; i < MAGIC_SETS; i++) { 625 mlist_free(ms->mlist[i]); 626 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 627 file_oomem(ms, sizeof(*ms->mlist[i])); 628 goto fail; 629 } 630 } 631 632 for (i = 0; i < nbufs; i++) { 633 map = apprentice_buf(ms, bufs[i], sizes[i]); 634 if (map == NULL) 635 goto fail; 636 637 for (j = 0; j < MAGIC_SETS; j++) { 638 if (add_mlist(ms->mlist[j], map, j) == -1) { 639 file_oomem(ms, sizeof(*ml)); 640 goto fail; 641 } 642 } 643 } 644 645 return 0; 646 fail: 647 for (i = 0; i < MAGIC_SETS; i++) { 648 mlist_free(ms->mlist[i]); 649 ms->mlist[i] = NULL; 650 } 651 return -1; 652 } 653 #endif 654 655 /* const char *fn: list of magic files and directories */ 656 protected int 657 file_apprentice(struct magic_set *ms, const char *fn, int action) 658 { 659 char *p, *mfn; 660 int fileerr, errs = -1; 661 size_t i; 662 663 (void)file_reset(ms, 0); 664 665 if ((fn = magic_getpath(fn, action)) == NULL) 666 return -1; 667 668 init_file_tables(); 669 670 if ((mfn = strdup(fn)) == NULL) { 671 file_oomem(ms, strlen(fn)); 672 return -1; 673 } 674 675 for (i = 0; i < MAGIC_SETS; i++) { 676 mlist_free(ms->mlist[i]); 677 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 678 file_oomem(ms, sizeof(*ms->mlist[i])); 679 while (i-- > 0) { 680 mlist_free(ms->mlist[i]); 681 ms->mlist[i] = NULL; 682 } 683 free(mfn); 684 return -1; 685 } 686 } 687 fn = mfn; 688 689 while (fn) { 690 p = strchr(fn, PATHSEP); 691 if (p) 692 *p++ = '\0'; 693 if (*fn == '\0') 694 break; 695 fileerr = apprentice_1(ms, fn, action); 696 errs = MAX(errs, fileerr); 697 fn = p; 698 } 699 700 free(mfn); 701 702 if (errs == -1) { 703 for (i = 0; i < MAGIC_SETS; i++) { 704 mlist_free(ms->mlist[i]); 705 ms->mlist[i] = NULL; 706 } 707 file_error(ms, 0, "could not find any valid magic files!"); 708 return -1; 709 } 710 711 #if 0 712 /* 713 * Always leave the database loaded 714 */ 715 if (action == FILE_LOAD) 716 return 0; 717 718 for (i = 0; i < MAGIC_SETS; i++) { 719 mlist_free(ms->mlist[i]); 720 ms->mlist[i] = NULL; 721 } 722 #endif 723 724 switch (action) { 725 case FILE_LOAD: 726 case FILE_COMPILE: 727 case FILE_CHECK: 728 case FILE_LIST: 729 return 0; 730 default: 731 file_error(ms, 0, "Invalid action %d", action); 732 return -1; 733 } 734 } 735 736 /* 737 * Compute the real length of a magic expression, for the purposes 738 * of determining how "strong" a magic expression is (approximating 739 * how specific its matches are): 740 * - magic characters count 0 unless escaped. 741 * - [] expressions count 1 742 * - {} expressions count 0 743 * - regular characters or escaped magic characters count 1 744 * - 0 length expressions count as one 745 */ 746 private size_t 747 nonmagic(const char *str) 748 { 749 const char *p; 750 size_t rv = 0; 751 752 for (p = str; *p; p++) 753 switch (*p) { 754 case '\\': /* Escaped anything counts 1 */ 755 if (!*++p) 756 p--; 757 rv++; 758 continue; 759 case '?': /* Magic characters count 0 */ 760 case '*': 761 case '.': 762 case '+': 763 case '^': 764 case '$': 765 continue; 766 case '[': /* Bracketed expressions count 1 the ']' */ 767 while (*p && *p != ']') 768 p++; 769 p--; 770 continue; 771 case '{': /* Braced expressions count 0 */ 772 while (*p && *p != '}') 773 p++; 774 if (!*p) 775 p--; 776 continue; 777 default: /* Anything else counts 1 */ 778 rv++; 779 continue; 780 } 781 782 return rv == 0 ? 1 : rv; /* Return at least 1 */ 783 } 784 785 786 private size_t 787 typesize(int type) 788 { 789 switch (type) { 790 case FILE_BYTE: 791 return 1; 792 793 case FILE_SHORT: 794 case FILE_LESHORT: 795 case FILE_BESHORT: 796 return 2; 797 798 case FILE_LONG: 799 case FILE_LELONG: 800 case FILE_BELONG: 801 case FILE_MELONG: 802 return 4; 803 804 case FILE_DATE: 805 case FILE_LEDATE: 806 case FILE_BEDATE: 807 case FILE_MEDATE: 808 case FILE_LDATE: 809 case FILE_LELDATE: 810 case FILE_BELDATE: 811 case FILE_MELDATE: 812 case FILE_FLOAT: 813 case FILE_BEFLOAT: 814 case FILE_LEFLOAT: 815 return 4; 816 817 case FILE_QUAD: 818 case FILE_BEQUAD: 819 case FILE_LEQUAD: 820 case FILE_QDATE: 821 case FILE_LEQDATE: 822 case FILE_BEQDATE: 823 case FILE_QLDATE: 824 case FILE_LEQLDATE: 825 case FILE_BEQLDATE: 826 case FILE_QWDATE: 827 case FILE_LEQWDATE: 828 case FILE_BEQWDATE: 829 case FILE_DOUBLE: 830 case FILE_BEDOUBLE: 831 case FILE_LEDOUBLE: 832 return 8; 833 default: 834 return CAST(size_t, ~0); 835 } 836 } 837 838 /* 839 * Get weight of this magic entry, for sorting purposes. 840 */ 841 private size_t 842 apprentice_magic_strength(const struct magic *m) 843 { 844 #define MULT 10U 845 size_t ts, v; 846 ssize_t val = 2 * MULT; /* baseline strength */ 847 848 switch (m->type) { 849 case FILE_DEFAULT: /* make sure this sorts last */ 850 if (m->factor_op != FILE_FACTOR_OP_NONE) 851 abort(); 852 return 0; 853 854 case FILE_BYTE: 855 case FILE_SHORT: 856 case FILE_LESHORT: 857 case FILE_BESHORT: 858 case FILE_LONG: 859 case FILE_LELONG: 860 case FILE_BELONG: 861 case FILE_MELONG: 862 case FILE_DATE: 863 case FILE_LEDATE: 864 case FILE_BEDATE: 865 case FILE_MEDATE: 866 case FILE_LDATE: 867 case FILE_LELDATE: 868 case FILE_BELDATE: 869 case FILE_MELDATE: 870 case FILE_FLOAT: 871 case FILE_BEFLOAT: 872 case FILE_LEFLOAT: 873 case FILE_QUAD: 874 case FILE_BEQUAD: 875 case FILE_LEQUAD: 876 case FILE_QDATE: 877 case FILE_LEQDATE: 878 case FILE_BEQDATE: 879 case FILE_QLDATE: 880 case FILE_LEQLDATE: 881 case FILE_BEQLDATE: 882 case FILE_QWDATE: 883 case FILE_LEQWDATE: 884 case FILE_BEQWDATE: 885 case FILE_DOUBLE: 886 case FILE_BEDOUBLE: 887 case FILE_LEDOUBLE: 888 ts = typesize(m->type); 889 if (ts == CAST(size_t, ~0)) 890 abort(); 891 val += ts * MULT; 892 break; 893 894 case FILE_PSTRING: 895 case FILE_STRING: 896 val += m->vallen * MULT; 897 break; 898 899 case FILE_BESTRING16: 900 case FILE_LESTRING16: 901 val += m->vallen * MULT / 2; 902 break; 903 904 case FILE_SEARCH: 905 if (m->vallen == 0) 906 break; 907 val += m->vallen * MAX(MULT / m->vallen, 1); 908 break; 909 910 case FILE_REGEX: 911 v = nonmagic(m->value.s); 912 val += v * MAX(MULT / v, 1); 913 break; 914 915 case FILE_INDIRECT: 916 case FILE_NAME: 917 case FILE_USE: 918 break; 919 920 case FILE_DER: 921 val += MULT; 922 break; 923 924 default: 925 (void)fprintf(stderr, "Bad type %d\n", m->type); 926 abort(); 927 } 928 929 switch (m->reln) { 930 case 'x': /* matches anything penalize */ 931 case '!': /* matches almost anything penalize */ 932 val = 0; 933 break; 934 935 case '=': /* Exact match, prefer */ 936 val += MULT; 937 break; 938 939 case '>': 940 case '<': /* comparison match reduce strength */ 941 val -= 2 * MULT; 942 break; 943 944 case '^': 945 case '&': /* masking bits, we could count them too */ 946 val -= MULT; 947 break; 948 949 default: 950 (void)fprintf(stderr, "Bad relation %c\n", m->reln); 951 abort(); 952 } 953 954 switch (m->factor_op) { 955 case FILE_FACTOR_OP_NONE: 956 break; 957 case FILE_FACTOR_OP_PLUS: 958 val += m->factor; 959 break; 960 case FILE_FACTOR_OP_MINUS: 961 val -= m->factor; 962 break; 963 case FILE_FACTOR_OP_TIMES: 964 val *= m->factor; 965 break; 966 case FILE_FACTOR_OP_DIV: 967 val /= m->factor; 968 break; 969 default: 970 abort(); 971 } 972 973 if (val <= 0) /* ensure we only return 0 for FILE_DEFAULT */ 974 val = 1; 975 976 /* 977 * Magic entries with no description get a bonus because they depend 978 * on subsequent magic entries to print something. 979 */ 980 if (m->desc[0] == '\0') 981 val++; 982 return val; 983 } 984 985 /* 986 * Sort callback for sorting entries by "strength" (basically length) 987 */ 988 private int 989 apprentice_sort(const void *a, const void *b) 990 { 991 const struct magic_entry *ma = CAST(const struct magic_entry *, a); 992 const struct magic_entry *mb = CAST(const struct magic_entry *, b); 993 size_t sa = apprentice_magic_strength(ma->mp); 994 size_t sb = apprentice_magic_strength(mb->mp); 995 if (sa == sb) 996 return 0; 997 else if (sa > sb) 998 return -1; 999 else 1000 return 1; 1001 } 1002 1003 /* 1004 * Shows sorted patterns list in the order which is used for the matching 1005 */ 1006 private void 1007 apprentice_list(struct mlist *mlist, int mode) 1008 { 1009 uint32_t magindex = 0; 1010 struct mlist *ml; 1011 for (ml = mlist->next; ml != mlist; ml = ml->next) { 1012 for (magindex = 0; magindex < ml->nmagic; magindex++) { 1013 struct magic *m = &ml->magic[magindex]; 1014 if ((m->flag & mode) != mode) { 1015 /* Skip sub-tests */ 1016 while (magindex + 1 < ml->nmagic && 1017 ml->magic[magindex + 1].cont_level != 0) 1018 ++magindex; 1019 continue; /* Skip to next top-level test*/ 1020 } 1021 1022 /* 1023 * Try to iterate over the tree until we find item with 1024 * description/mimetype. 1025 */ 1026 while (magindex + 1 < ml->nmagic && 1027 ml->magic[magindex + 1].cont_level != 0 && 1028 *ml->magic[magindex].desc == '\0' && 1029 *ml->magic[magindex].mimetype == '\0') 1030 magindex++; 1031 1032 printf("Strength = %3" SIZE_T_FORMAT "u@%u: %s [%s]\n", 1033 apprentice_magic_strength(m), 1034 ml->magic[magindex].lineno, 1035 ml->magic[magindex].desc, 1036 ml->magic[magindex].mimetype); 1037 } 1038 } 1039 } 1040 1041 private void 1042 set_test_type(struct magic *mstart, struct magic *m) 1043 { 1044 switch (m->type) { 1045 case FILE_BYTE: 1046 case FILE_SHORT: 1047 case FILE_LONG: 1048 case FILE_DATE: 1049 case FILE_BESHORT: 1050 case FILE_BELONG: 1051 case FILE_BEDATE: 1052 case FILE_LESHORT: 1053 case FILE_LELONG: 1054 case FILE_LEDATE: 1055 case FILE_LDATE: 1056 case FILE_BELDATE: 1057 case FILE_LELDATE: 1058 case FILE_MEDATE: 1059 case FILE_MELDATE: 1060 case FILE_MELONG: 1061 case FILE_QUAD: 1062 case FILE_LEQUAD: 1063 case FILE_BEQUAD: 1064 case FILE_QDATE: 1065 case FILE_LEQDATE: 1066 case FILE_BEQDATE: 1067 case FILE_QLDATE: 1068 case FILE_LEQLDATE: 1069 case FILE_BEQLDATE: 1070 case FILE_QWDATE: 1071 case FILE_LEQWDATE: 1072 case FILE_BEQWDATE: 1073 case FILE_FLOAT: 1074 case FILE_BEFLOAT: 1075 case FILE_LEFLOAT: 1076 case FILE_DOUBLE: 1077 case FILE_BEDOUBLE: 1078 case FILE_LEDOUBLE: 1079 case FILE_DER: 1080 mstart->flag |= BINTEST; 1081 break; 1082 case FILE_STRING: 1083 case FILE_PSTRING: 1084 case FILE_BESTRING16: 1085 case FILE_LESTRING16: 1086 /* Allow text overrides */ 1087 if (mstart->str_flags & STRING_TEXTTEST) 1088 mstart->flag |= TEXTTEST; 1089 else 1090 mstart->flag |= BINTEST; 1091 break; 1092 case FILE_REGEX: 1093 case FILE_SEARCH: 1094 /* Check for override */ 1095 if (mstart->str_flags & STRING_BINTEST) 1096 mstart->flag |= BINTEST; 1097 if (mstart->str_flags & STRING_TEXTTEST) 1098 mstart->flag |= TEXTTEST; 1099 1100 if (mstart->flag & (TEXTTEST|BINTEST)) 1101 break; 1102 1103 /* binary test if pattern is not text */ 1104 if (file_looks_utf8(m->value.us, CAST(size_t, m->vallen), NULL, 1105 NULL) <= 0) 1106 mstart->flag |= BINTEST; 1107 else 1108 mstart->flag |= TEXTTEST; 1109 break; 1110 case FILE_DEFAULT: 1111 /* can't deduce anything; we shouldn't see this at the 1112 top level anyway */ 1113 break; 1114 case FILE_INVALID: 1115 default: 1116 /* invalid search type, but no need to complain here */ 1117 break; 1118 } 1119 } 1120 1121 private int 1122 addentry(struct magic_set *ms, struct magic_entry *me, 1123 struct magic_entry_set *mset) 1124 { 1125 size_t i = me->mp->type == FILE_NAME ? 1 : 0; 1126 if (mset[i].count == mset[i].max) { 1127 struct magic_entry *mp; 1128 1129 mset[i].max += ALLOC_INCR; 1130 if ((mp = CAST(struct magic_entry *, 1131 realloc(mset[i].me, sizeof(*mp) * mset[i].max))) == 1132 NULL) { 1133 file_oomem(ms, sizeof(*mp) * mset[i].max); 1134 return -1; 1135 } 1136 (void)memset(&mp[mset[i].count], 0, sizeof(*mp) * 1137 ALLOC_INCR); 1138 mset[i].me = mp; 1139 } 1140 mset[i].me[mset[i].count++] = *me; 1141 memset(me, 0, sizeof(*me)); 1142 return 0; 1143 } 1144 1145 /* 1146 * Load and parse one file. 1147 */ 1148 private void 1149 load_1(struct magic_set *ms, int action, const char *fn, int *errs, 1150 struct magic_entry_set *mset) 1151 { 1152 size_t lineno = 0, llen = 0; 1153 char *line = NULL; 1154 ssize_t len; 1155 struct magic_entry me; 1156 1157 FILE *f = fopen(ms->file = fn, "r"); 1158 if (f == NULL) { 1159 if (errno != ENOENT) 1160 file_error(ms, errno, "cannot read magic file `%s'", 1161 fn); 1162 (*errs)++; 1163 return; 1164 } 1165 1166 memset(&me, 0, sizeof(me)); 1167 /* read and parse this file */ 1168 for (ms->line = 1; (len = getline(&line, &llen, f)) != -1; 1169 ms->line++) { 1170 if (len == 0) /* null line, garbage, etc */ 1171 continue; 1172 if (line[len - 1] == '\n') { 1173 lineno++; 1174 line[len - 1] = '\0'; /* delete newline */ 1175 } 1176 switch (line[0]) { 1177 case '\0': /* empty, do not parse */ 1178 case '#': /* comment, do not parse */ 1179 continue; 1180 case '!': 1181 if (line[1] == ':') { 1182 size_t i; 1183 1184 for (i = 0; bang[i].name != NULL; i++) { 1185 if (CAST(size_t, len - 2) > bang[i].len && 1186 memcmp(bang[i].name, line + 2, 1187 bang[i].len) == 0) 1188 break; 1189 } 1190 if (bang[i].name == NULL) { 1191 file_error(ms, 0, 1192 "Unknown !: entry `%s'", line); 1193 (*errs)++; 1194 continue; 1195 } 1196 if (me.mp == NULL) { 1197 file_error(ms, 0, 1198 "No current entry for :!%s type", 1199 bang[i].name); 1200 (*errs)++; 1201 continue; 1202 } 1203 if ((*bang[i].fun)(ms, &me, 1204 line + bang[i].len + 2) != 0) { 1205 (*errs)++; 1206 continue; 1207 } 1208 continue; 1209 } 1210 /*FALLTHROUGH*/ 1211 default: 1212 again: 1213 switch (parse(ms, &me, line, lineno, action)) { 1214 case 0: 1215 continue; 1216 case 1: 1217 (void)addentry(ms, &me, mset); 1218 goto again; 1219 default: 1220 (*errs)++; 1221 break; 1222 } 1223 } 1224 } 1225 if (me.mp) 1226 (void)addentry(ms, &me, mset); 1227 free(line); 1228 (void)fclose(f); 1229 } 1230 1231 /* 1232 * parse a file or directory of files 1233 * const char *fn: name of magic file or directory 1234 */ 1235 private int 1236 cmpstrp(const void *p1, const void *p2) 1237 { 1238 return strcmp(*RCAST(char *const *, p1), *RCAST(char *const *, p2)); 1239 } 1240 1241 1242 private uint32_t 1243 set_text_binary(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1244 uint32_t starttest) 1245 { 1246 static const char text[] = "text"; 1247 static const char binary[] = "binary"; 1248 static const size_t len = sizeof(text); 1249 1250 uint32_t i = starttest; 1251 1252 do { 1253 set_test_type(me[starttest].mp, me[i].mp); 1254 if ((ms->flags & MAGIC_DEBUG) == 0) 1255 continue; 1256 (void)fprintf(stderr, "%s%s%s: %s\n", 1257 me[i].mp->mimetype, 1258 me[i].mp->mimetype[0] == '\0' ? "" : "; ", 1259 me[i].mp->desc[0] ? me[i].mp->desc : "(no description)", 1260 me[i].mp->flag & BINTEST ? binary : text); 1261 if (me[i].mp->flag & BINTEST) { 1262 char *p = strstr(me[i].mp->desc, text); 1263 if (p && (p == me[i].mp->desc || 1264 isspace(CAST(unsigned char, p[-1]))) && 1265 (p + len - me[i].mp->desc == MAXstring 1266 || (p[len] == '\0' || 1267 isspace(CAST(unsigned char, p[len]))))) 1268 (void)fprintf(stderr, "*** Possible " 1269 "binary test for text type\n"); 1270 } 1271 } while (++i < nme && me[i].mp->cont_level != 0); 1272 return i; 1273 } 1274 1275 private void 1276 set_last_default(struct magic_set *ms, struct magic_entry *me, uint32_t nme) 1277 { 1278 uint32_t i; 1279 for (i = 0; i < nme; i++) { 1280 if (me[i].mp->cont_level == 0 && 1281 me[i].mp->type == FILE_DEFAULT) { 1282 while (++i < nme) 1283 if (me[i].mp->cont_level == 0) 1284 break; 1285 if (i != nme) { 1286 /* XXX - Ugh! */ 1287 ms->line = me[i].mp->lineno; 1288 file_magwarn(ms, 1289 "level 0 \"default\" did not sort last"); 1290 } 1291 return; 1292 } 1293 } 1294 } 1295 1296 private int 1297 coalesce_entries(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1298 struct magic **ma, uint32_t *nma) 1299 { 1300 uint32_t i, mentrycount = 0; 1301 size_t slen; 1302 1303 for (i = 0; i < nme; i++) 1304 mentrycount += me[i].cont_count; 1305 1306 slen = sizeof(**ma) * mentrycount; 1307 if ((*ma = CAST(struct magic *, malloc(slen))) == NULL) { 1308 file_oomem(ms, slen); 1309 return -1; 1310 } 1311 1312 mentrycount = 0; 1313 for (i = 0; i < nme; i++) { 1314 (void)memcpy(*ma + mentrycount, me[i].mp, 1315 me[i].cont_count * sizeof(**ma)); 1316 mentrycount += me[i].cont_count; 1317 } 1318 *nma = mentrycount; 1319 return 0; 1320 } 1321 1322 private void 1323 magic_entry_free(struct magic_entry *me, uint32_t nme) 1324 { 1325 uint32_t i; 1326 if (me == NULL) 1327 return; 1328 for (i = 0; i < nme; i++) 1329 free(me[i].mp); 1330 free(me); 1331 } 1332 1333 private struct magic_map * 1334 apprentice_load(struct magic_set *ms, const char *fn, int action) 1335 { 1336 int errs = 0; 1337 uint32_t i, j; 1338 size_t files = 0, maxfiles = 0; 1339 char **filearr = NULL, *mfn; 1340 struct stat st; 1341 struct magic_map *map; 1342 struct magic_entry_set mset[MAGIC_SETS]; 1343 DIR *dir; 1344 struct dirent *d; 1345 1346 memset(mset, 0, sizeof(mset)); 1347 ms->flags |= MAGIC_CHECK; /* Enable checks for parsed files */ 1348 1349 1350 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) 1351 { 1352 file_oomem(ms, sizeof(*map)); 1353 return NULL; 1354 } 1355 map->type = MAP_TYPE_MALLOC; 1356 1357 /* print silly verbose header for USG compat. */ 1358 if (action == FILE_CHECK) 1359 (void)fprintf(stderr, "%s\n", usg_hdr); 1360 1361 /* load directory or file */ 1362 if (stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) { 1363 dir = opendir(fn); 1364 if (!dir) { 1365 errs++; 1366 goto out; 1367 } 1368 while ((d = readdir(dir)) != NULL) { 1369 if (d->d_name[0] == '.') 1370 continue; 1371 if (asprintf(&mfn, "%s/%s", fn, d->d_name) < 0) { 1372 file_oomem(ms, 1373 strlen(fn) + strlen(d->d_name) + 2); 1374 errs++; 1375 closedir(dir); 1376 goto out; 1377 } 1378 if (stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) { 1379 free(mfn); 1380 continue; 1381 } 1382 if (files >= maxfiles) { 1383 size_t mlen; 1384 maxfiles = (maxfiles + 1) * 2; 1385 mlen = maxfiles * sizeof(*filearr); 1386 if ((filearr = CAST(char **, 1387 realloc(filearr, mlen))) == NULL) { 1388 file_oomem(ms, mlen); 1389 free(mfn); 1390 closedir(dir); 1391 errs++; 1392 goto out; 1393 } 1394 } 1395 filearr[files++] = mfn; 1396 } 1397 closedir(dir); 1398 if (filearr) { 1399 qsort(filearr, files, sizeof(*filearr), cmpstrp); 1400 for (i = 0; i < files; i++) { 1401 load_1(ms, action, filearr[i], &errs, mset); 1402 free(filearr[i]); 1403 } 1404 free(filearr); 1405 } 1406 } else 1407 load_1(ms, action, fn, &errs, mset); 1408 if (errs) 1409 goto out; 1410 1411 for (j = 0; j < MAGIC_SETS; j++) { 1412 /* Set types of tests */ 1413 for (i = 0; i < mset[j].count; ) { 1414 if (mset[j].me[i].mp->cont_level != 0) { 1415 i++; 1416 continue; 1417 } 1418 i = set_text_binary(ms, mset[j].me, mset[j].count, i); 1419 } 1420 if (mset[j].me) 1421 qsort(mset[j].me, mset[j].count, sizeof(*mset[j].me), 1422 apprentice_sort); 1423 1424 /* 1425 * Make sure that any level 0 "default" line is last 1426 * (if one exists). 1427 */ 1428 set_last_default(ms, mset[j].me, mset[j].count); 1429 1430 /* coalesce per file arrays into a single one */ 1431 if (coalesce_entries(ms, mset[j].me, mset[j].count, 1432 &map->magic[j], &map->nmagic[j]) == -1) { 1433 errs++; 1434 goto out; 1435 } 1436 } 1437 1438 out: 1439 for (j = 0; j < MAGIC_SETS; j++) 1440 magic_entry_free(mset[j].me, mset[j].count); 1441 1442 if (errs) { 1443 apprentice_unmap(map); 1444 return NULL; 1445 } 1446 return map; 1447 } 1448 1449 /* 1450 * extend the sign bit if the comparison is to be signed 1451 */ 1452 protected uint64_t 1453 file_signextend(struct magic_set *ms, struct magic *m, uint64_t v) 1454 { 1455 if (!(m->flag & UNSIGNED)) { 1456 switch(m->type) { 1457 /* 1458 * Do not remove the casts below. They are 1459 * vital. When later compared with the data, 1460 * the sign extension must have happened. 1461 */ 1462 case FILE_BYTE: 1463 v = CAST(signed char, v); 1464 break; 1465 case FILE_SHORT: 1466 case FILE_BESHORT: 1467 case FILE_LESHORT: 1468 v = CAST(short, v); 1469 break; 1470 case FILE_DATE: 1471 case FILE_BEDATE: 1472 case FILE_LEDATE: 1473 case FILE_MEDATE: 1474 case FILE_LDATE: 1475 case FILE_BELDATE: 1476 case FILE_LELDATE: 1477 case FILE_MELDATE: 1478 case FILE_LONG: 1479 case FILE_BELONG: 1480 case FILE_LELONG: 1481 case FILE_MELONG: 1482 case FILE_FLOAT: 1483 case FILE_BEFLOAT: 1484 case FILE_LEFLOAT: 1485 v = CAST(int32_t, v); 1486 break; 1487 case FILE_QUAD: 1488 case FILE_BEQUAD: 1489 case FILE_LEQUAD: 1490 case FILE_QDATE: 1491 case FILE_QLDATE: 1492 case FILE_QWDATE: 1493 case FILE_BEQDATE: 1494 case FILE_BEQLDATE: 1495 case FILE_BEQWDATE: 1496 case FILE_LEQDATE: 1497 case FILE_LEQLDATE: 1498 case FILE_LEQWDATE: 1499 case FILE_DOUBLE: 1500 case FILE_BEDOUBLE: 1501 case FILE_LEDOUBLE: 1502 v = CAST(int64_t, v); 1503 break; 1504 case FILE_STRING: 1505 case FILE_PSTRING: 1506 case FILE_BESTRING16: 1507 case FILE_LESTRING16: 1508 case FILE_REGEX: 1509 case FILE_SEARCH: 1510 case FILE_DEFAULT: 1511 case FILE_INDIRECT: 1512 case FILE_NAME: 1513 case FILE_USE: 1514 case FILE_CLEAR: 1515 case FILE_DER: 1516 break; 1517 default: 1518 if (ms->flags & MAGIC_CHECK) 1519 file_magwarn(ms, "cannot happen: m->type=%d\n", 1520 m->type); 1521 return ~0U; 1522 } 1523 } 1524 return v; 1525 } 1526 1527 private int 1528 string_modifier_check(struct magic_set *ms, struct magic *m) 1529 { 1530 if ((ms->flags & MAGIC_CHECK) == 0) 1531 return 0; 1532 1533 if ((m->type != FILE_REGEX || (m->str_flags & REGEX_LINE_COUNT) == 0) && 1534 (m->type != FILE_PSTRING && (m->str_flags & PSTRING_LEN) != 0)) { 1535 file_magwarn(ms, 1536 "'/BHhLl' modifiers are only allowed for pascal strings\n"); 1537 return -1; 1538 } 1539 switch (m->type) { 1540 case FILE_BESTRING16: 1541 case FILE_LESTRING16: 1542 if (m->str_flags != 0) { 1543 file_magwarn(ms, 1544 "no modifiers allowed for 16-bit strings\n"); 1545 return -1; 1546 } 1547 break; 1548 case FILE_STRING: 1549 case FILE_PSTRING: 1550 if ((m->str_flags & REGEX_OFFSET_START) != 0) { 1551 file_magwarn(ms, 1552 "'/%c' only allowed on regex and search\n", 1553 CHAR_REGEX_OFFSET_START); 1554 return -1; 1555 } 1556 break; 1557 case FILE_SEARCH: 1558 if (m->str_range == 0) { 1559 file_magwarn(ms, 1560 "missing range; defaulting to %d\n", 1561 STRING_DEFAULT_RANGE); 1562 m->str_range = STRING_DEFAULT_RANGE; 1563 return -1; 1564 } 1565 break; 1566 case FILE_REGEX: 1567 if ((m->str_flags & STRING_COMPACT_WHITESPACE) != 0) { 1568 file_magwarn(ms, "'/%c' not allowed on regex\n", 1569 CHAR_COMPACT_WHITESPACE); 1570 return -1; 1571 } 1572 if ((m->str_flags & STRING_COMPACT_OPTIONAL_WHITESPACE) != 0) { 1573 file_magwarn(ms, "'/%c' not allowed on regex\n", 1574 CHAR_COMPACT_OPTIONAL_WHITESPACE); 1575 return -1; 1576 } 1577 break; 1578 default: 1579 file_magwarn(ms, "coding error: m->type=%d\n", 1580 m->type); 1581 return -1; 1582 } 1583 return 0; 1584 } 1585 1586 private int 1587 get_op(char c) 1588 { 1589 switch (c) { 1590 case '&': 1591 return FILE_OPAND; 1592 case '|': 1593 return FILE_OPOR; 1594 case '^': 1595 return FILE_OPXOR; 1596 case '+': 1597 return FILE_OPADD; 1598 case '-': 1599 return FILE_OPMINUS; 1600 case '*': 1601 return FILE_OPMULTIPLY; 1602 case '/': 1603 return FILE_OPDIVIDE; 1604 case '%': 1605 return FILE_OPMODULO; 1606 default: 1607 return -1; 1608 } 1609 } 1610 1611 #ifdef ENABLE_CONDITIONALS 1612 private int 1613 get_cond(const char *l, const char **t) 1614 { 1615 static const struct cond_tbl_s { 1616 char name[8]; 1617 size_t len; 1618 int cond; 1619 } cond_tbl[] = { 1620 { "if", 2, COND_IF }, 1621 { "elif", 4, COND_ELIF }, 1622 { "else", 4, COND_ELSE }, 1623 { "", 0, COND_NONE }, 1624 }; 1625 const struct cond_tbl_s *p; 1626 1627 for (p = cond_tbl; p->len; p++) { 1628 if (strncmp(l, p->name, p->len) == 0 && 1629 isspace(CAST(unsigned char, l[p->len]))) { 1630 if (t) 1631 *t = l + p->len; 1632 break; 1633 } 1634 } 1635 return p->cond; 1636 } 1637 1638 private int 1639 check_cond(struct magic_set *ms, int cond, uint32_t cont_level) 1640 { 1641 int last_cond; 1642 last_cond = ms->c.li[cont_level].last_cond; 1643 1644 switch (cond) { 1645 case COND_IF: 1646 if (last_cond != COND_NONE && last_cond != COND_ELIF) { 1647 if (ms->flags & MAGIC_CHECK) 1648 file_magwarn(ms, "syntax error: `if'"); 1649 return -1; 1650 } 1651 last_cond = COND_IF; 1652 break; 1653 1654 case COND_ELIF: 1655 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1656 if (ms->flags & MAGIC_CHECK) 1657 file_magwarn(ms, "syntax error: `elif'"); 1658 return -1; 1659 } 1660 last_cond = COND_ELIF; 1661 break; 1662 1663 case COND_ELSE: 1664 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1665 if (ms->flags & MAGIC_CHECK) 1666 file_magwarn(ms, "syntax error: `else'"); 1667 return -1; 1668 } 1669 last_cond = COND_NONE; 1670 break; 1671 1672 case COND_NONE: 1673 last_cond = COND_NONE; 1674 break; 1675 } 1676 1677 ms->c.li[cont_level].last_cond = last_cond; 1678 return 0; 1679 } 1680 #endif /* ENABLE_CONDITIONALS */ 1681 1682 private int 1683 parse_indirect_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1684 { 1685 const char *l = *lp; 1686 1687 while (!isspace(CAST(unsigned char, *++l))) 1688 switch (*l) { 1689 case CHAR_INDIRECT_RELATIVE: 1690 m->str_flags |= INDIRECT_RELATIVE; 1691 break; 1692 default: 1693 if (ms->flags & MAGIC_CHECK) 1694 file_magwarn(ms, "indirect modifier `%c' " 1695 "invalid", *l); 1696 *lp = l; 1697 return -1; 1698 } 1699 *lp = l; 1700 return 0; 1701 } 1702 1703 private void 1704 parse_op_modifier(struct magic_set *ms, struct magic *m, const char **lp, 1705 int op) 1706 { 1707 const char *l = *lp; 1708 char *t; 1709 uint64_t val; 1710 1711 ++l; 1712 m->mask_op |= op; 1713 val = CAST(uint64_t, strtoull(l, &t, 0)); 1714 l = t; 1715 m->num_mask = file_signextend(ms, m, val); 1716 eatsize(&l); 1717 *lp = l; 1718 } 1719 1720 private int 1721 parse_string_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1722 { 1723 const char *l = *lp; 1724 char *t; 1725 int have_range = 0; 1726 1727 while (!isspace(CAST(unsigned char, *++l))) { 1728 switch (*l) { 1729 case '0': case '1': case '2': 1730 case '3': case '4': case '5': 1731 case '6': case '7': case '8': 1732 case '9': 1733 if (have_range && (ms->flags & MAGIC_CHECK)) 1734 file_magwarn(ms, "multiple ranges"); 1735 have_range = 1; 1736 m->str_range = CAST(uint32_t, strtoul(l, &t, 0)); 1737 if (m->str_range == 0) 1738 file_magwarn(ms, "zero range"); 1739 l = t - 1; 1740 break; 1741 case CHAR_COMPACT_WHITESPACE: 1742 m->str_flags |= STRING_COMPACT_WHITESPACE; 1743 break; 1744 case CHAR_COMPACT_OPTIONAL_WHITESPACE: 1745 m->str_flags |= STRING_COMPACT_OPTIONAL_WHITESPACE; 1746 break; 1747 case CHAR_IGNORE_LOWERCASE: 1748 m->str_flags |= STRING_IGNORE_LOWERCASE; 1749 break; 1750 case CHAR_IGNORE_UPPERCASE: 1751 m->str_flags |= STRING_IGNORE_UPPERCASE; 1752 break; 1753 case CHAR_REGEX_OFFSET_START: 1754 m->str_flags |= REGEX_OFFSET_START; 1755 break; 1756 case CHAR_BINTEST: 1757 m->str_flags |= STRING_BINTEST; 1758 break; 1759 case CHAR_TEXTTEST: 1760 m->str_flags |= STRING_TEXTTEST; 1761 break; 1762 case CHAR_TRIM: 1763 m->str_flags |= STRING_TRIM; 1764 break; 1765 case CHAR_PSTRING_1_LE: 1766 #define SET_LENGTH(a) m->str_flags = (m->str_flags & ~PSTRING_LEN) | (a) 1767 if (m->type != FILE_PSTRING) 1768 goto bad; 1769 SET_LENGTH(PSTRING_1_LE); 1770 break; 1771 case CHAR_PSTRING_2_BE: 1772 if (m->type != FILE_PSTRING) 1773 goto bad; 1774 SET_LENGTH(PSTRING_2_BE); 1775 break; 1776 case CHAR_PSTRING_2_LE: 1777 if (m->type != FILE_PSTRING) 1778 goto bad; 1779 SET_LENGTH(PSTRING_2_LE); 1780 break; 1781 case CHAR_PSTRING_4_BE: 1782 if (m->type != FILE_PSTRING) 1783 goto bad; 1784 SET_LENGTH(PSTRING_4_BE); 1785 break; 1786 case CHAR_PSTRING_4_LE: 1787 switch (m->type) { 1788 case FILE_PSTRING: 1789 case FILE_REGEX: 1790 break; 1791 default: 1792 goto bad; 1793 } 1794 SET_LENGTH(PSTRING_4_LE); 1795 break; 1796 case CHAR_PSTRING_LENGTH_INCLUDES_ITSELF: 1797 if (m->type != FILE_PSTRING) 1798 goto bad; 1799 m->str_flags |= PSTRING_LENGTH_INCLUDES_ITSELF; 1800 break; 1801 default: 1802 bad: 1803 if (ms->flags & MAGIC_CHECK) 1804 file_magwarn(ms, "string modifier `%c' " 1805 "invalid", *l); 1806 goto out; 1807 } 1808 /* allow multiple '/' for readability */ 1809 if (l[1] == '/' && !isspace(CAST(unsigned char, l[2]))) 1810 l++; 1811 } 1812 if (string_modifier_check(ms, m) == -1) 1813 goto out; 1814 *lp = l; 1815 return 0; 1816 out: 1817 *lp = l; 1818 return -1; 1819 } 1820 1821 /* 1822 * parse one line from magic file, put into magic[index++] if valid 1823 */ 1824 private int 1825 parse(struct magic_set *ms, struct magic_entry *me, const char *line, 1826 size_t lineno, int action) 1827 { 1828 #ifdef ENABLE_CONDITIONALS 1829 static uint32_t last_cont_level = 0; 1830 #endif 1831 size_t i; 1832 struct magic *m; 1833 const char *l = line; 1834 char *t; 1835 int op; 1836 uint32_t cont_level; 1837 int32_t diff; 1838 1839 cont_level = 0; 1840 1841 /* 1842 * Parse the offset. 1843 */ 1844 while (*l == '>') { 1845 ++l; /* step over */ 1846 cont_level++; 1847 } 1848 #ifdef ENABLE_CONDITIONALS 1849 if (cont_level == 0 || cont_level > last_cont_level) 1850 if (file_check_mem(ms, cont_level) == -1) 1851 return -1; 1852 last_cont_level = cont_level; 1853 #endif 1854 if (cont_level != 0) { 1855 if (me->mp == NULL) { 1856 file_magerror(ms, "No current entry for continuation"); 1857 return -1; 1858 } 1859 if (me->cont_count == 0) { 1860 file_magerror(ms, "Continuations present with 0 count"); 1861 return -1; 1862 } 1863 m = &me->mp[me->cont_count - 1]; 1864 diff = CAST(int32_t, cont_level) - CAST(int32_t, m->cont_level); 1865 if (diff > 1) 1866 file_magwarn(ms, "New continuation level %u is more " 1867 "than one larger than current level %u", cont_level, 1868 m->cont_level); 1869 if (me->cont_count == me->max_count) { 1870 struct magic *nm; 1871 size_t cnt = me->max_count + ALLOC_CHUNK; 1872 if ((nm = CAST(struct magic *, realloc(me->mp, 1873 sizeof(*nm) * cnt))) == NULL) { 1874 file_oomem(ms, sizeof(*nm) * cnt); 1875 return -1; 1876 } 1877 me->mp = nm; 1878 me->max_count = CAST(uint32_t, cnt); 1879 } 1880 m = &me->mp[me->cont_count++]; 1881 (void)memset(m, 0, sizeof(*m)); 1882 m->cont_level = cont_level; 1883 } else { 1884 static const size_t len = sizeof(*m) * ALLOC_CHUNK; 1885 if (me->mp != NULL) 1886 return 1; 1887 if ((m = CAST(struct magic *, malloc(len))) == NULL) { 1888 file_oomem(ms, len); 1889 return -1; 1890 } 1891 me->mp = m; 1892 me->max_count = ALLOC_CHUNK; 1893 (void)memset(m, 0, sizeof(*m)); 1894 m->factor_op = FILE_FACTOR_OP_NONE; 1895 m->cont_level = 0; 1896 me->cont_count = 1; 1897 } 1898 m->lineno = CAST(uint32_t, lineno); 1899 1900 if (*l == '&') { /* m->cont_level == 0 checked below. */ 1901 ++l; /* step over */ 1902 m->flag |= OFFADD; 1903 } 1904 if (*l == '(') { 1905 ++l; /* step over */ 1906 m->flag |= INDIR; 1907 if (m->flag & OFFADD) 1908 m->flag = (m->flag & ~OFFADD) | INDIROFFADD; 1909 1910 if (*l == '&') { /* m->cont_level == 0 checked below */ 1911 ++l; /* step over */ 1912 m->flag |= OFFADD; 1913 } 1914 } 1915 /* Indirect offsets are not valid at level 0. */ 1916 if (m->cont_level == 0 && (m->flag & (OFFADD | INDIROFFADD))) { 1917 if (ms->flags & MAGIC_CHECK) 1918 file_magwarn(ms, "relative offset at level 0"); 1919 return -1; 1920 } 1921 1922 /* get offset, then skip over it */ 1923 m->offset = CAST(int32_t, strtol(l, &t, 0)); 1924 if (l == t) { 1925 if (ms->flags & MAGIC_CHECK) 1926 file_magwarn(ms, "offset `%s' invalid", l); 1927 return -1; 1928 } 1929 #if 0 1930 if (m->offset < 0 && cont_level != 0 && 1931 (m->flag & (OFFADD | INDIROFFADD)) == 0) { 1932 if (ms->flags & MAGIC_CHECK) { 1933 file_magwarn(ms, 1934 "negative direct offset `%s' at level %u", 1935 l, cont_level); 1936 } 1937 return -1; 1938 } 1939 #endif 1940 l = t; 1941 1942 if (m->flag & INDIR) { 1943 m->in_type = FILE_LONG; 1944 m->in_offset = 0; 1945 m->in_op = 0; 1946 /* 1947 * read [.,lbs][+-]nnnnn) 1948 */ 1949 if (*l == '.' || *l == ',') { 1950 if (*l == ',') 1951 m->in_op |= FILE_OPSIGNED; 1952 l++; 1953 switch (*l) { 1954 case 'l': 1955 m->in_type = FILE_LELONG; 1956 break; 1957 case 'L': 1958 m->in_type = FILE_BELONG; 1959 break; 1960 case 'm': 1961 m->in_type = FILE_MELONG; 1962 break; 1963 case 'h': 1964 case 's': 1965 m->in_type = FILE_LESHORT; 1966 break; 1967 case 'H': 1968 case 'S': 1969 m->in_type = FILE_BESHORT; 1970 break; 1971 case 'c': 1972 case 'b': 1973 case 'C': 1974 case 'B': 1975 m->in_type = FILE_BYTE; 1976 break; 1977 case 'e': 1978 case 'f': 1979 case 'g': 1980 m->in_type = FILE_LEDOUBLE; 1981 break; 1982 case 'E': 1983 case 'F': 1984 case 'G': 1985 m->in_type = FILE_BEDOUBLE; 1986 break; 1987 case 'i': 1988 m->in_type = FILE_LEID3; 1989 break; 1990 case 'I': 1991 m->in_type = FILE_BEID3; 1992 break; 1993 case 'q': 1994 m->in_type = FILE_LEQUAD; 1995 break; 1996 case 'Q': 1997 m->in_type = FILE_BEQUAD; 1998 break; 1999 default: 2000 if (ms->flags & MAGIC_CHECK) 2001 file_magwarn(ms, 2002 "indirect offset type `%c' invalid", 2003 *l); 2004 return -1; 2005 } 2006 l++; 2007 } 2008 2009 if (*l == '~') { 2010 m->in_op |= FILE_OPINVERSE; 2011 l++; 2012 } 2013 if ((op = get_op(*l)) != -1) { 2014 m->in_op |= op; 2015 l++; 2016 } 2017 if (*l == '(') { 2018 m->in_op |= FILE_OPINDIRECT; 2019 l++; 2020 } 2021 if (isdigit(CAST(unsigned char, *l)) || *l == '-') { 2022 m->in_offset = CAST(int32_t, strtol(l, &t, 0)); 2023 if (l == t) { 2024 if (ms->flags & MAGIC_CHECK) 2025 file_magwarn(ms, 2026 "in_offset `%s' invalid", l); 2027 return -1; 2028 } 2029 l = t; 2030 } 2031 if (*l++ != ')' || 2032 ((m->in_op & FILE_OPINDIRECT) && *l++ != ')')) { 2033 if (ms->flags & MAGIC_CHECK) 2034 file_magwarn(ms, 2035 "missing ')' in indirect offset"); 2036 return -1; 2037 } 2038 } 2039 EATAB; 2040 2041 #ifdef ENABLE_CONDITIONALS 2042 m->cond = get_cond(l, &l); 2043 if (check_cond(ms, m->cond, cont_level) == -1) 2044 return -1; 2045 2046 EATAB; 2047 #endif 2048 2049 /* 2050 * Parse the type. 2051 */ 2052 if (*l == 'u') { 2053 /* 2054 * Try it as a keyword type prefixed by "u"; match what 2055 * follows the "u". If that fails, try it as an SUS 2056 * integer type. 2057 */ 2058 m->type = get_type(type_tbl, l + 1, &l); 2059 if (m->type == FILE_INVALID) { 2060 /* 2061 * Not a keyword type; parse it as an SUS type, 2062 * 'u' possibly followed by a number or C/S/L. 2063 */ 2064 m->type = get_standard_integer_type(l, &l); 2065 } 2066 /* It's unsigned. */ 2067 if (m->type != FILE_INVALID) 2068 m->flag |= UNSIGNED; 2069 } else { 2070 /* 2071 * Try it as a keyword type. If that fails, try it as 2072 * an SUS integer type if it begins with "d" or as an 2073 * SUS string type if it begins with "s". In any case, 2074 * it's not unsigned. 2075 */ 2076 m->type = get_type(type_tbl, l, &l); 2077 if (m->type == FILE_INVALID) { 2078 /* 2079 * Not a keyword type; parse it as an SUS type, 2080 * either 'd' possibly followed by a number or 2081 * C/S/L, or just 's'. 2082 */ 2083 if (*l == 'd') 2084 m->type = get_standard_integer_type(l, &l); 2085 else if (*l == 's' 2086 && !isalpha(CAST(unsigned char, l[1]))) { 2087 m->type = FILE_STRING; 2088 ++l; 2089 } 2090 } 2091 } 2092 2093 if (m->type == FILE_INVALID) { 2094 /* Not found - try it as a special keyword. */ 2095 m->type = get_type(special_tbl, l, &l); 2096 } 2097 2098 if (m->type == FILE_INVALID) { 2099 if (ms->flags & MAGIC_CHECK) 2100 file_magwarn(ms, "type `%s' invalid", l); 2101 return -1; 2102 } 2103 2104 /* New-style anding: "0 byte&0x80 =0x80 dynamically linked" */ 2105 /* New and improved: ~ & | ^ + - * / % -- exciting, isn't it? */ 2106 2107 m->mask_op = 0; 2108 if (*l == '~') { 2109 if (!IS_STRING(m->type)) 2110 m->mask_op |= FILE_OPINVERSE; 2111 else if (ms->flags & MAGIC_CHECK) 2112 file_magwarn(ms, "'~' invalid for string types"); 2113 ++l; 2114 } 2115 m->str_range = 0; 2116 m->str_flags = m->type == FILE_PSTRING ? PSTRING_1_LE : 0; 2117 if ((op = get_op(*l)) != -1) { 2118 if (IS_STRING(m->type)) { 2119 int r; 2120 2121 if (op != FILE_OPDIVIDE) { 2122 if (ms->flags & MAGIC_CHECK) 2123 file_magwarn(ms, 2124 "invalid string/indirect op: " 2125 "`%c'", *t); 2126 return -1; 2127 } 2128 2129 if (m->type == FILE_INDIRECT) 2130 r = parse_indirect_modifier(ms, m, &l); 2131 else 2132 r = parse_string_modifier(ms, m, &l); 2133 if (r == -1) 2134 return -1; 2135 } else 2136 parse_op_modifier(ms, m, &l, op); 2137 } 2138 2139 /* 2140 * We used to set mask to all 1's here, instead let's just not do 2141 * anything if mask = 0 (unless you have a better idea) 2142 */ 2143 EATAB; 2144 2145 switch (*l) { 2146 case '>': 2147 case '<': 2148 m->reln = *l; 2149 ++l; 2150 if (*l == '=') { 2151 if (ms->flags & MAGIC_CHECK) { 2152 file_magwarn(ms, "%c= not supported", 2153 m->reln); 2154 return -1; 2155 } 2156 ++l; 2157 } 2158 break; 2159 /* Old-style anding: "0 byte &0x80 dynamically linked" */ 2160 case '&': 2161 case '^': 2162 case '=': 2163 m->reln = *l; 2164 ++l; 2165 if (*l == '=') { 2166 /* HP compat: ignore &= etc. */ 2167 ++l; 2168 } 2169 break; 2170 case '!': 2171 m->reln = *l; 2172 ++l; 2173 break; 2174 default: 2175 m->reln = '='; /* the default relation */ 2176 if (*l == 'x' && ((isascii(CAST(unsigned char, l[1])) && 2177 isspace(CAST(unsigned char, l[1]))) || !l[1])) { 2178 m->reln = *l; 2179 ++l; 2180 } 2181 break; 2182 } 2183 /* 2184 * Grab the value part, except for an 'x' reln. 2185 */ 2186 if (m->reln != 'x' && getvalue(ms, m, &l, action)) 2187 return -1; 2188 2189 /* 2190 * TODO finish this macro and start using it! 2191 * #define offsetcheck {if (offset > ms->bytes_max -1) 2192 * magwarn("offset too big"); } 2193 */ 2194 2195 /* 2196 * Now get last part - the description 2197 */ 2198 EATAB; 2199 if (l[0] == '\b') { 2200 ++l; 2201 m->flag |= NOSPACE; 2202 } else if ((l[0] == '\\') && (l[1] == 'b')) { 2203 ++l; 2204 ++l; 2205 m->flag |= NOSPACE; 2206 } 2207 for (i = 0; (m->desc[i++] = *l++) != '\0' && i < sizeof(m->desc); ) 2208 continue; 2209 if (i == sizeof(m->desc)) { 2210 m->desc[sizeof(m->desc) - 1] = '\0'; 2211 if (ms->flags & MAGIC_CHECK) 2212 file_magwarn(ms, "description `%s' truncated", m->desc); 2213 } 2214 2215 /* 2216 * We only do this check while compiling, or if any of the magic 2217 * files were not compiled. 2218 */ 2219 if (ms->flags & MAGIC_CHECK) { 2220 if (check_format(ms, m) == -1) 2221 return -1; 2222 } 2223 #ifndef COMPILE_ONLY 2224 if (action == FILE_CHECK) { 2225 file_mdump(m); 2226 } 2227 #endif 2228 m->mimetype[0] = '\0'; /* initialise MIME type to none */ 2229 return 0; 2230 } 2231 2232 /* 2233 * parse a STRENGTH annotation line from magic file, put into magic[index - 1] 2234 * if valid 2235 */ 2236 private int 2237 parse_strength(struct magic_set *ms, struct magic_entry *me, const char *line) 2238 { 2239 const char *l = line; 2240 char *el; 2241 unsigned long factor; 2242 struct magic *m = &me->mp[0]; 2243 2244 if (m->factor_op != FILE_FACTOR_OP_NONE) { 2245 file_magwarn(ms, 2246 "Current entry already has a strength type: %c %d", 2247 m->factor_op, m->factor); 2248 return -1; 2249 } 2250 if (m->type == FILE_NAME) { 2251 file_magwarn(ms, "%s: Strength setting is not supported in " 2252 "\"name\" magic entries", m->value.s); 2253 return -1; 2254 } 2255 EATAB; 2256 switch (*l) { 2257 case FILE_FACTOR_OP_NONE: 2258 case FILE_FACTOR_OP_PLUS: 2259 case FILE_FACTOR_OP_MINUS: 2260 case FILE_FACTOR_OP_TIMES: 2261 case FILE_FACTOR_OP_DIV: 2262 m->factor_op = *l++; 2263 break; 2264 default: 2265 file_magwarn(ms, "Unknown factor op `%c'", *l); 2266 return -1; 2267 } 2268 EATAB; 2269 factor = strtoul(l, &el, 0); 2270 if (factor > 255) { 2271 file_magwarn(ms, "Too large factor `%lu'", factor); 2272 goto out; 2273 } 2274 if (*el && !isspace(CAST(unsigned char, *el))) { 2275 file_magwarn(ms, "Bad factor `%s'", l); 2276 goto out; 2277 } 2278 m->factor = CAST(uint8_t, factor); 2279 if (m->factor == 0 && m->factor_op == FILE_FACTOR_OP_DIV) { 2280 file_magwarn(ms, "Cannot have factor op `%c' and factor %u", 2281 m->factor_op, m->factor); 2282 goto out; 2283 } 2284 return 0; 2285 out: 2286 m->factor_op = FILE_FACTOR_OP_NONE; 2287 m->factor = 0; 2288 return -1; 2289 } 2290 2291 private int 2292 goodchar(unsigned char x, const char *extra) 2293 { 2294 return (isascii(x) && isalnum(x)) || strchr(extra, x); 2295 } 2296 2297 private int 2298 parse_extra(struct magic_set *ms, struct magic_entry *me, const char *line, 2299 off_t off, size_t len, const char *name, const char *extra, int nt) 2300 { 2301 size_t i; 2302 const char *l = line; 2303 struct magic *m = &me->mp[me->cont_count == 0 ? 0 : me->cont_count - 1]; 2304 char *buf = CAST(char *, CAST(void *, m)) + off; 2305 2306 if (buf[0] != '\0') { 2307 len = nt ? strlen(buf) : len; 2308 file_magwarn(ms, "Current entry already has a %s type " 2309 "`%.*s', new type `%s'", name, CAST(int, len), buf, l); 2310 return -1; 2311 } 2312 2313 if (*m->desc == '\0') { 2314 file_magwarn(ms, "Current entry does not yet have a " 2315 "description for adding a %s type", name); 2316 return -1; 2317 } 2318 2319 EATAB; 2320 for (i = 0; *l && i < len && goodchar(*l, extra); buf[i++] = *l++) 2321 continue; 2322 2323 if (i == len && *l) { 2324 if (nt) 2325 buf[len - 1] = '\0'; 2326 if (ms->flags & MAGIC_CHECK) 2327 file_magwarn(ms, "%s type `%s' truncated %" 2328 SIZE_T_FORMAT "u", name, line, i); 2329 } else { 2330 if (!isspace(CAST(unsigned char, *l)) && !goodchar(*l, extra)) 2331 file_magwarn(ms, "%s type `%s' has bad char '%c'", 2332 name, line, *l); 2333 if (nt) 2334 buf[i] = '\0'; 2335 } 2336 2337 if (i > 0) 2338 return 0; 2339 2340 file_magerror(ms, "Bad magic entry '%s'", line); 2341 return -1; 2342 } 2343 2344 /* 2345 * Parse an Apple CREATOR/TYPE annotation from magic file and put it into 2346 * magic[index - 1] 2347 */ 2348 private int 2349 parse_apple(struct magic_set *ms, struct magic_entry *me, const char *line) 2350 { 2351 struct magic *m = &me->mp[0]; 2352 2353 return parse_extra(ms, me, line, 2354 CAST(off_t, offsetof(struct magic, apple)), 2355 sizeof(m->apple), "APPLE", "!+-./?", 0); 2356 } 2357 2358 /* 2359 * Parse a comma-separated list of extensions 2360 */ 2361 private int 2362 parse_ext(struct magic_set *ms, struct magic_entry *me, const char *line) 2363 { 2364 struct magic *m = &me->mp[0]; 2365 2366 return parse_extra(ms, me, line, 2367 CAST(off_t, offsetof(struct magic, ext)), 2368 sizeof(m->ext), "EXTENSION", ",!+-/@?_$", 0); 2369 } 2370 2371 /* 2372 * parse a MIME annotation line from magic file, put into magic[index - 1] 2373 * if valid 2374 */ 2375 private int 2376 parse_mime(struct magic_set *ms, struct magic_entry *me, const char *line) 2377 { 2378 struct magic *m = &me->mp[0]; 2379 2380 return parse_extra(ms, me, line, 2381 CAST(off_t, offsetof(struct magic, mimetype)), 2382 sizeof(m->mimetype), "MIME", "+-/.$?:{}", 1); 2383 } 2384 2385 private int 2386 check_format_type(const char *ptr, int type, const char **estr) 2387 { 2388 int quad = 0, h; 2389 size_t len, cnt; 2390 if (*ptr == '\0') { 2391 /* Missing format string; bad */ 2392 *estr = "missing format spec"; 2393 return -1; 2394 } 2395 2396 switch (file_formats[type]) { 2397 case FILE_FMT_QUAD: 2398 quad = 1; 2399 /*FALLTHROUGH*/ 2400 case FILE_FMT_NUM: 2401 if (quad == 0) { 2402 switch (type) { 2403 case FILE_BYTE: 2404 h = 2; 2405 break; 2406 case FILE_SHORT: 2407 case FILE_BESHORT: 2408 case FILE_LESHORT: 2409 h = 1; 2410 break; 2411 case FILE_LONG: 2412 case FILE_BELONG: 2413 case FILE_LELONG: 2414 case FILE_MELONG: 2415 case FILE_LEID3: 2416 case FILE_BEID3: 2417 case FILE_INDIRECT: 2418 h = 0; 2419 break; 2420 default: 2421 abort(); 2422 } 2423 } else 2424 h = 0; 2425 if (*ptr == '-') 2426 ptr++; 2427 if (*ptr == '.') 2428 ptr++; 2429 if (*ptr == '#') 2430 ptr++; 2431 #define CHECKLEN() do { \ 2432 for (len = cnt = 0; isdigit(CAST(unsigned char, *ptr)); ptr++, cnt++) \ 2433 len = len * 10 + (*ptr - '0'); \ 2434 if (cnt > 5 || len > 1024) \ 2435 goto toolong; \ 2436 } while (/*CONSTCOND*/0) 2437 2438 CHECKLEN(); 2439 if (*ptr == '.') 2440 ptr++; 2441 CHECKLEN(); 2442 if (quad) { 2443 if (*ptr++ != 'l') 2444 goto invalid; 2445 if (*ptr++ != 'l') 2446 goto invalid; 2447 } 2448 2449 switch (*ptr++) { 2450 #ifdef STRICT_FORMAT /* "long" formats are int formats for us */ 2451 /* so don't accept the 'l' modifier */ 2452 case 'l': 2453 switch (*ptr++) { 2454 case 'i': 2455 case 'd': 2456 case 'u': 2457 case 'o': 2458 case 'x': 2459 case 'X': 2460 if (h == 0) 2461 return 0; 2462 /*FALLTHROUGH*/ 2463 default: 2464 goto invalid; 2465 } 2466 2467 /* 2468 * Don't accept h and hh modifiers. They make writing 2469 * magic entries more complicated, for very little benefit 2470 */ 2471 case 'h': 2472 if (h-- <= 0) 2473 goto invalid; 2474 switch (*ptr++) { 2475 case 'h': 2476 if (h-- <= 0) 2477 goto invalid; 2478 switch (*ptr++) { 2479 case 'i': 2480 case 'd': 2481 case 'u': 2482 case 'o': 2483 case 'x': 2484 case 'X': 2485 return 0; 2486 default: 2487 goto invalid; 2488 } 2489 case 'i': 2490 case 'd': 2491 case 'u': 2492 case 'o': 2493 case 'x': 2494 case 'X': 2495 if (h == 0) 2496 return 0; 2497 /*FALLTHROUGH*/ 2498 default: 2499 goto invalid; 2500 } 2501 #endif 2502 case 'c': 2503 if (h == 2) 2504 return 0; 2505 goto invalid; 2506 case 'i': 2507 case 'd': 2508 case 'u': 2509 case 'o': 2510 case 'x': 2511 case 'X': 2512 #ifdef STRICT_FORMAT 2513 if (h == 0) 2514 return 0; 2515 /*FALLTHROUGH*/ 2516 #else 2517 return 0; 2518 #endif 2519 default: 2520 goto invalid; 2521 } 2522 2523 case FILE_FMT_FLOAT: 2524 case FILE_FMT_DOUBLE: 2525 if (*ptr == '-') 2526 ptr++; 2527 if (*ptr == '.') 2528 ptr++; 2529 CHECKLEN(); 2530 if (*ptr == '.') 2531 ptr++; 2532 CHECKLEN(); 2533 switch (*ptr++) { 2534 case 'e': 2535 case 'E': 2536 case 'f': 2537 case 'F': 2538 case 'g': 2539 case 'G': 2540 return 0; 2541 2542 default: 2543 goto invalid; 2544 } 2545 2546 2547 case FILE_FMT_STR: 2548 if (*ptr == '-') 2549 ptr++; 2550 while (isdigit(CAST(unsigned char, *ptr))) 2551 ptr++; 2552 if (*ptr == '.') { 2553 ptr++; 2554 while (isdigit(CAST(unsigned char , *ptr))) 2555 ptr++; 2556 } 2557 2558 switch (*ptr++) { 2559 case 's': 2560 return 0; 2561 default: 2562 goto invalid; 2563 } 2564 2565 default: 2566 /* internal error */ 2567 abort(); 2568 } 2569 invalid: 2570 *estr = "not valid"; 2571 toolong: 2572 *estr = "too long"; 2573 return -1; 2574 } 2575 2576 /* 2577 * Check that the optional printf format in description matches 2578 * the type of the magic. 2579 */ 2580 private int 2581 check_format(struct magic_set *ms, struct magic *m) 2582 { 2583 char *ptr; 2584 const char *estr; 2585 2586 for (ptr = m->desc; *ptr; ptr++) 2587 if (*ptr == '%') 2588 break; 2589 if (*ptr == '\0') { 2590 /* No format string; ok */ 2591 return 1; 2592 } 2593 2594 assert(file_nformats == file_nnames); 2595 2596 if (m->type >= file_nformats) { 2597 file_magwarn(ms, "Internal error inconsistency between " 2598 "m->type and format strings"); 2599 return -1; 2600 } 2601 if (file_formats[m->type] == FILE_FMT_NONE) { 2602 file_magwarn(ms, "No format string for `%s' with description " 2603 "`%s'", m->desc, file_names[m->type]); 2604 return -1; 2605 } 2606 2607 ptr++; 2608 if (check_format_type(ptr, m->type, &estr) == -1) { 2609 /* 2610 * TODO: this error message is unhelpful if the format 2611 * string is not one character long 2612 */ 2613 file_magwarn(ms, "Printf format is %s for type " 2614 "`%s' in description `%s'", estr, 2615 file_names[m->type], m->desc); 2616 return -1; 2617 } 2618 2619 for (; *ptr; ptr++) { 2620 if (*ptr == '%') { 2621 file_magwarn(ms, 2622 "Too many format strings (should have at most one) " 2623 "for `%s' with description `%s'", 2624 file_names[m->type], m->desc); 2625 return -1; 2626 } 2627 } 2628 return 0; 2629 } 2630 2631 /* 2632 * Read a numeric value from a pointer, into the value union of a magic 2633 * pointer, according to the magic type. Update the string pointer to point 2634 * just after the number read. Return 0 for success, non-zero for failure. 2635 */ 2636 private int 2637 getvalue(struct magic_set *ms, struct magic *m, const char **p, int action) 2638 { 2639 char *ep; 2640 uint64_t ull; 2641 2642 switch (m->type) { 2643 case FILE_BESTRING16: 2644 case FILE_LESTRING16: 2645 case FILE_STRING: 2646 case FILE_PSTRING: 2647 case FILE_REGEX: 2648 case FILE_SEARCH: 2649 case FILE_NAME: 2650 case FILE_USE: 2651 case FILE_DER: 2652 *p = getstr(ms, m, *p, action == FILE_COMPILE); 2653 if (*p == NULL) { 2654 if (ms->flags & MAGIC_CHECK) 2655 file_magwarn(ms, "cannot get string from `%s'", 2656 m->value.s); 2657 return -1; 2658 } 2659 if (m->type == FILE_REGEX) { 2660 file_regex_t rx; 2661 int rc = file_regcomp(&rx, m->value.s, REG_EXTENDED); 2662 if (rc) { 2663 if (ms->flags & MAGIC_CHECK) 2664 file_regerror(&rx, rc, ms); 2665 } 2666 file_regfree(&rx); 2667 return rc ? -1 : 0; 2668 } 2669 return 0; 2670 default: 2671 if (m->reln == 'x') 2672 return 0; 2673 break; 2674 } 2675 2676 switch (m->type) { 2677 case FILE_FLOAT: 2678 case FILE_BEFLOAT: 2679 case FILE_LEFLOAT: 2680 errno = 0; 2681 #ifdef HAVE_STRTOF 2682 m->value.f = strtof(*p, &ep); 2683 #else 2684 m->value.f = (float)strtod(*p, &ep); 2685 #endif 2686 if (errno == 0) 2687 *p = ep; 2688 return 0; 2689 case FILE_DOUBLE: 2690 case FILE_BEDOUBLE: 2691 case FILE_LEDOUBLE: 2692 errno = 0; 2693 m->value.d = strtod(*p, &ep); 2694 if (errno == 0) 2695 *p = ep; 2696 return 0; 2697 default: 2698 errno = 0; 2699 ull = CAST(uint64_t, strtoull(*p, &ep, 0)); 2700 m->value.q = file_signextend(ms, m, ull); 2701 if (*p == ep) { 2702 file_magwarn(ms, "Unparseable number `%s'", *p); 2703 } else { 2704 size_t ts = typesize(m->type); 2705 uint64_t x; 2706 const char *q; 2707 2708 if (ts == CAST(size_t, ~0)) { 2709 file_magwarn(ms, 2710 "Expected numeric type got `%s'", 2711 type_tbl[m->type].name); 2712 } 2713 for (q = *p; isspace(CAST(unsigned char, *q)); q++) 2714 continue; 2715 if (*q == '-') 2716 ull = -CAST(int64_t, ull); 2717 switch (ts) { 2718 case 1: 2719 x = CAST(uint64_t, ull & ~0xffULL); 2720 break; 2721 case 2: 2722 x = CAST(uint64_t, ull & ~0xffffULL); 2723 break; 2724 case 4: 2725 x = CAST(uint64_t, ull & ~0xffffffffULL); 2726 break; 2727 case 8: 2728 x = 0; 2729 break; 2730 default: 2731 abort(); 2732 } 2733 if (x) { 2734 file_magwarn(ms, "Overflow for numeric" 2735 " type `%s' value %#" PRIx64, 2736 type_tbl[m->type].name, ull); 2737 } 2738 } 2739 if (errno == 0) { 2740 *p = ep; 2741 eatsize(p); 2742 } 2743 return 0; 2744 } 2745 } 2746 2747 /* 2748 * Convert a string containing C character escapes. Stop at an unescaped 2749 * space or tab. 2750 * Copy the converted version to "m->value.s", and the length in m->vallen. 2751 * Return updated scan pointer as function result. Warn if set. 2752 */ 2753 private const char * 2754 getstr(struct magic_set *ms, struct magic *m, const char *s, int warn) 2755 { 2756 const char *origs = s; 2757 char *p = m->value.s; 2758 size_t plen = sizeof(m->value.s); 2759 char *origp = p; 2760 char *pmax = p + plen - 1; 2761 int c; 2762 int val; 2763 2764 while ((c = *s++) != '\0') { 2765 if (isspace(CAST(unsigned char, c))) 2766 break; 2767 if (p >= pmax) { 2768 file_error(ms, 0, "string too long: `%s'", origs); 2769 return NULL; 2770 } 2771 if (c == '\\') { 2772 switch(c = *s++) { 2773 2774 case '\0': 2775 if (warn) 2776 file_magwarn(ms, "incomplete escape"); 2777 s--; 2778 goto out; 2779 2780 case '\t': 2781 if (warn) { 2782 file_magwarn(ms, 2783 "escaped tab found, use \\t instead"); 2784 warn = 0; /* already did */ 2785 } 2786 /*FALLTHROUGH*/ 2787 default: 2788 if (warn) { 2789 if (isprint(CAST(unsigned char, c))) { 2790 /* Allow escaping of 2791 * ``relations'' */ 2792 if (strchr("<>&^=!", c) == NULL 2793 && (m->type != FILE_REGEX || 2794 strchr("[]().*?^$|{}", c) 2795 == NULL)) { 2796 file_magwarn(ms, "no " 2797 "need to escape " 2798 "`%c'", c); 2799 } 2800 } else { 2801 file_magwarn(ms, 2802 "unknown escape sequence: " 2803 "\\%03o", c); 2804 } 2805 } 2806 /*FALLTHROUGH*/ 2807 /* space, perhaps force people to use \040? */ 2808 case ' ': 2809 #if 0 2810 /* 2811 * Other things people escape, but shouldn't need to, 2812 * so we disallow them 2813 */ 2814 case '\'': 2815 case '"': 2816 case '?': 2817 #endif 2818 /* Relations */ 2819 case '>': 2820 case '<': 2821 case '&': 2822 case '^': 2823 case '=': 2824 case '!': 2825 /* and baskslash itself */ 2826 case '\\': 2827 *p++ = CAST(char, c); 2828 break; 2829 2830 case 'a': 2831 *p++ = '\a'; 2832 break; 2833 2834 case 'b': 2835 *p++ = '\b'; 2836 break; 2837 2838 case 'f': 2839 *p++ = '\f'; 2840 break; 2841 2842 case 'n': 2843 *p++ = '\n'; 2844 break; 2845 2846 case 'r': 2847 *p++ = '\r'; 2848 break; 2849 2850 case 't': 2851 *p++ = '\t'; 2852 break; 2853 2854 case 'v': 2855 *p++ = '\v'; 2856 break; 2857 2858 /* \ and up to 3 octal digits */ 2859 case '0': 2860 case '1': 2861 case '2': 2862 case '3': 2863 case '4': 2864 case '5': 2865 case '6': 2866 case '7': 2867 val = c - '0'; 2868 c = *s++; /* try for 2 */ 2869 if (c >= '0' && c <= '7') { 2870 val = (val << 3) | (c - '0'); 2871 c = *s++; /* try for 3 */ 2872 if (c >= '0' && c <= '7') 2873 val = (val << 3) | (c-'0'); 2874 else 2875 --s; 2876 } 2877 else 2878 --s; 2879 *p++ = CAST(char, val); 2880 break; 2881 2882 /* \x and up to 2 hex digits */ 2883 case 'x': 2884 val = 'x'; /* Default if no digits */ 2885 c = hextoint(*s++); /* Get next char */ 2886 if (c >= 0) { 2887 val = c; 2888 c = hextoint(*s++); 2889 if (c >= 0) 2890 val = (val << 4) + c; 2891 else 2892 --s; 2893 } else 2894 --s; 2895 *p++ = CAST(char, val); 2896 break; 2897 } 2898 } else 2899 *p++ = CAST(char, c); 2900 } 2901 --s; 2902 out: 2903 *p = '\0'; 2904 m->vallen = CAST(unsigned char, (p - origp)); 2905 if (m->type == FILE_PSTRING) 2906 m->vallen += CAST(unsigned char, file_pstring_length_size(m)); 2907 return s; 2908 } 2909 2910 2911 /* Single hex char to int; -1 if not a hex char. */ 2912 private int 2913 hextoint(int c) 2914 { 2915 if (!isascii(CAST(unsigned char, c))) 2916 return -1; 2917 if (isdigit(CAST(unsigned char, c))) 2918 return c - '0'; 2919 if ((c >= 'a') && (c <= 'f')) 2920 return c + 10 - 'a'; 2921 if (( c>= 'A') && (c <= 'F')) 2922 return c + 10 - 'A'; 2923 return -1; 2924 } 2925 2926 2927 /* 2928 * Print a string containing C character escapes. 2929 */ 2930 protected void 2931 file_showstr(FILE *fp, const char *s, size_t len) 2932 { 2933 char c; 2934 2935 for (;;) { 2936 if (len == ~0U) { 2937 c = *s++; 2938 if (c == '\0') 2939 break; 2940 } 2941 else { 2942 if (len-- == 0) 2943 break; 2944 c = *s++; 2945 } 2946 if (c >= 040 && c <= 0176) /* TODO isprint && !iscntrl */ 2947 (void) fputc(c, fp); 2948 else { 2949 (void) fputc('\\', fp); 2950 switch (c) { 2951 case '\a': 2952 (void) fputc('a', fp); 2953 break; 2954 2955 case '\b': 2956 (void) fputc('b', fp); 2957 break; 2958 2959 case '\f': 2960 (void) fputc('f', fp); 2961 break; 2962 2963 case '\n': 2964 (void) fputc('n', fp); 2965 break; 2966 2967 case '\r': 2968 (void) fputc('r', fp); 2969 break; 2970 2971 case '\t': 2972 (void) fputc('t', fp); 2973 break; 2974 2975 case '\v': 2976 (void) fputc('v', fp); 2977 break; 2978 2979 default: 2980 (void) fprintf(fp, "%.3o", c & 0377); 2981 break; 2982 } 2983 } 2984 } 2985 } 2986 2987 /* 2988 * eatsize(): Eat the size spec from a number [eg. 10UL] 2989 */ 2990 private void 2991 eatsize(const char **p) 2992 { 2993 const char *l = *p; 2994 2995 if (LOWCASE(*l) == 'u') 2996 l++; 2997 2998 switch (LOWCASE(*l)) { 2999 case 'l': /* long */ 3000 case 's': /* short */ 3001 case 'h': /* short */ 3002 case 'b': /* char/byte */ 3003 case 'c': /* char/byte */ 3004 l++; 3005 /*FALLTHROUGH*/ 3006 default: 3007 break; 3008 } 3009 3010 *p = l; 3011 } 3012 3013 /* 3014 * handle a buffer containing a compiled file. 3015 */ 3016 private struct magic_map * 3017 apprentice_buf(struct magic_set *ms, struct magic *buf, size_t len) 3018 { 3019 struct magic_map *map; 3020 3021 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 3022 file_oomem(ms, sizeof(*map)); 3023 return NULL; 3024 } 3025 map->len = len; 3026 map->p = buf; 3027 map->type = MAP_TYPE_USER; 3028 if (check_buffer(ms, map, "buffer") != 0) { 3029 apprentice_unmap(map); 3030 return NULL; 3031 } 3032 return map; 3033 } 3034 3035 /* 3036 * handle a compiled file. 3037 */ 3038 3039 private struct magic_map * 3040 apprentice_map(struct magic_set *ms, const char *fn) 3041 { 3042 int fd; 3043 struct stat st; 3044 char *dbname = NULL; 3045 struct magic_map *map; 3046 struct magic_map *rv = NULL; 3047 3048 fd = -1; 3049 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 3050 file_oomem(ms, sizeof(*map)); 3051 goto error; 3052 } 3053 map->type = MAP_TYPE_USER; /* unspecified */ 3054 3055 dbname = mkdbname(ms, fn, 0); 3056 if (dbname == NULL) 3057 goto error; 3058 3059 if ((fd = open(dbname, O_RDONLY|O_BINARY)) == -1) 3060 goto error; 3061 3062 if (fstat(fd, &st) == -1) { 3063 file_error(ms, errno, "cannot stat `%s'", dbname); 3064 goto error; 3065 } 3066 if (st.st_size < 8 || st.st_size > maxoff_t()) { 3067 file_error(ms, 0, "file `%s' is too %s", dbname, 3068 st.st_size < 8 ? "small" : "large"); 3069 goto error; 3070 } 3071 3072 map->len = CAST(size_t, st.st_size); 3073 #ifdef QUICK 3074 map->type = MAP_TYPE_MMAP; 3075 if ((map->p = mmap(0, CAST(size_t, st.st_size), PROT_READ|PROT_WRITE, 3076 MAP_PRIVATE|MAP_FILE, fd, CAST(off_t, 0))) == MAP_FAILED) { 3077 file_error(ms, errno, "cannot map `%s'", dbname); 3078 goto error; 3079 } 3080 #else 3081 map->type = MAP_TYPE_MALLOC; 3082 if ((map->p = CAST(void *, malloc(map->len))) == NULL) { 3083 file_oomem(ms, map->len); 3084 goto error; 3085 } 3086 if (read(fd, map->p, map->len) != (ssize_t)map->len) { 3087 file_badread(ms); 3088 goto error; 3089 } 3090 #define RET 1 3091 #endif 3092 (void)close(fd); 3093 fd = -1; 3094 3095 if (check_buffer(ms, map, dbname) != 0) { 3096 rv = RCAST(struct magic_map *, -1); 3097 goto error; 3098 } 3099 #ifdef QUICK 3100 if (mprotect(map->p, CAST(size_t, st.st_size), PROT_READ) == -1) { 3101 file_error(ms, errno, "cannot mprotect `%s'", dbname); 3102 goto error; 3103 } 3104 #endif 3105 3106 free(dbname); 3107 return map; 3108 3109 error: 3110 if (fd != -1) 3111 (void)close(fd); 3112 apprentice_unmap(map); 3113 free(dbname); 3114 return rv; 3115 } 3116 3117 private int 3118 check_buffer(struct magic_set *ms, struct magic_map *map, const char *dbname) 3119 { 3120 uint32_t *ptr; 3121 uint32_t entries, nentries; 3122 uint32_t version; 3123 int i, needsbyteswap; 3124 3125 ptr = CAST(uint32_t *, map->p); 3126 if (*ptr != MAGICNO) { 3127 if (swap4(*ptr) != MAGICNO) { 3128 file_error(ms, 0, "bad magic in `%s'", dbname); 3129 return -1; 3130 } 3131 needsbyteswap = 1; 3132 } else 3133 needsbyteswap = 0; 3134 if (needsbyteswap) 3135 version = swap4(ptr[1]); 3136 else 3137 version = ptr[1]; 3138 if (version != VERSIONNO) { 3139 file_error(ms, 0, "File %s supports only version %d magic " 3140 "files. `%s' is version %d", VERSION, 3141 VERSIONNO, dbname, version); 3142 return -1; 3143 } 3144 entries = CAST(uint32_t, map->len / sizeof(struct magic)); 3145 if ((entries * sizeof(struct magic)) != map->len) { 3146 file_error(ms, 0, "Size of `%s' %" SIZE_T_FORMAT "u is not " 3147 "a multiple of %" SIZE_T_FORMAT "u", 3148 dbname, map->len, sizeof(struct magic)); 3149 return -1; 3150 } 3151 map->magic[0] = CAST(struct magic *, map->p) + 1; 3152 nentries = 0; 3153 for (i = 0; i < MAGIC_SETS; i++) { 3154 if (needsbyteswap) 3155 map->nmagic[i] = swap4(ptr[i + 2]); 3156 else 3157 map->nmagic[i] = ptr[i + 2]; 3158 if (i != MAGIC_SETS - 1) 3159 map->magic[i + 1] = map->magic[i] + map->nmagic[i]; 3160 nentries += map->nmagic[i]; 3161 } 3162 if (entries != nentries + 1) { 3163 file_error(ms, 0, "Inconsistent entries in `%s' %u != %u", 3164 dbname, entries, nentries + 1); 3165 return -1; 3166 } 3167 if (needsbyteswap) 3168 for (i = 0; i < MAGIC_SETS; i++) 3169 byteswap(map->magic[i], map->nmagic[i]); 3170 return 0; 3171 } 3172 3173 /* 3174 * handle an mmaped file. 3175 */ 3176 private int 3177 apprentice_compile(struct magic_set *ms, struct magic_map *map, const char *fn) 3178 { 3179 static const size_t nm = sizeof(*map->nmagic) * MAGIC_SETS; 3180 static const size_t m = sizeof(**map->magic); 3181 int fd = -1; 3182 size_t len; 3183 char *dbname; 3184 int rv = -1; 3185 uint32_t i; 3186 union { 3187 struct magic m; 3188 uint32_t h[2 + MAGIC_SETS]; 3189 } hdr; 3190 3191 dbname = mkdbname(ms, fn, 1); 3192 3193 if (dbname == NULL) 3194 goto out; 3195 3196 if ((fd = open(dbname, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY, 0644)) == -1) 3197 { 3198 file_error(ms, errno, "cannot open `%s'", dbname); 3199 goto out; 3200 } 3201 memset(&hdr, 0, sizeof(hdr)); 3202 hdr.h[0] = MAGICNO; 3203 hdr.h[1] = VERSIONNO; 3204 memcpy(hdr.h + 2, map->nmagic, nm); 3205 3206 if (write(fd, &hdr, sizeof(hdr)) != CAST(ssize_t, sizeof(hdr))) { 3207 file_error(ms, errno, "error writing `%s'", dbname); 3208 goto out2; 3209 } 3210 3211 for (i = 0; i < MAGIC_SETS; i++) { 3212 len = m * map->nmagic[i]; 3213 if (write(fd, map->magic[i], len) != CAST(ssize_t, len)) { 3214 file_error(ms, errno, "error writing `%s'", dbname); 3215 goto out2; 3216 } 3217 } 3218 3219 rv = 0; 3220 out2: 3221 if (fd != -1) 3222 (void)close(fd); 3223 out: 3224 apprentice_unmap(map); 3225 free(dbname); 3226 return rv; 3227 } 3228 3229 private const char ext[] = ".mgc"; 3230 /* 3231 * make a dbname 3232 */ 3233 private char * 3234 mkdbname(struct magic_set *ms, const char *fn, int strip) 3235 { 3236 const char *p, *q; 3237 char *buf; 3238 3239 if (strip) { 3240 if ((p = strrchr(fn, '/')) != NULL) 3241 fn = ++p; 3242 } 3243 3244 for (q = fn; *q; q++) 3245 continue; 3246 /* Look for .mgc */ 3247 for (p = ext + sizeof(ext) - 1; p >= ext && q >= fn; p--, q--) 3248 if (*p != *q) 3249 break; 3250 3251 /* Did not find .mgc, restore q */ 3252 if (p >= ext) 3253 while (*q) 3254 q++; 3255 3256 q++; 3257 /* Compatibility with old code that looked in .mime */ 3258 if (ms->flags & MAGIC_MIME) { 3259 if (asprintf(&buf, "%.*s.mime%s", CAST(int, q - fn), fn, ext) 3260 < 0) 3261 return NULL; 3262 if (access(buf, R_OK) != -1) { 3263 ms->flags &= MAGIC_MIME_TYPE; 3264 return buf; 3265 } 3266 free(buf); 3267 } 3268 if (asprintf(&buf, "%.*s%s", CAST(int, q - fn), fn, ext) < 0) 3269 return NULL; 3270 3271 /* Compatibility with old code that looked in .mime */ 3272 if (strstr(fn, ".mime") != NULL) 3273 ms->flags &= MAGIC_MIME_TYPE; 3274 return buf; 3275 } 3276 3277 /* 3278 * Byteswap an mmap'ed file if needed 3279 */ 3280 private void 3281 byteswap(struct magic *magic, uint32_t nmagic) 3282 { 3283 uint32_t i; 3284 for (i = 0; i < nmagic; i++) 3285 bs1(&magic[i]); 3286 } 3287 3288 /* 3289 * swap a short 3290 */ 3291 private uint16_t 3292 swap2(uint16_t sv) 3293 { 3294 uint16_t rv; 3295 uint8_t *s = RCAST(uint8_t *, RCAST(void *, &sv)); 3296 uint8_t *d = RCAST(uint8_t *, RCAST(void *, &rv)); 3297 d[0] = s[1]; 3298 d[1] = s[0]; 3299 return rv; 3300 } 3301 3302 /* 3303 * swap an int 3304 */ 3305 private uint32_t 3306 swap4(uint32_t sv) 3307 { 3308 uint32_t rv; 3309 uint8_t *s = RCAST(uint8_t *, RCAST(void *, &sv)); 3310 uint8_t *d = RCAST(uint8_t *, RCAST(void *, &rv)); 3311 d[0] = s[3]; 3312 d[1] = s[2]; 3313 d[2] = s[1]; 3314 d[3] = s[0]; 3315 return rv; 3316 } 3317 3318 /* 3319 * swap a quad 3320 */ 3321 private uint64_t 3322 swap8(uint64_t sv) 3323 { 3324 uint64_t rv; 3325 uint8_t *s = RCAST(uint8_t *, RCAST(void *, &sv)); 3326 uint8_t *d = RCAST(uint8_t *, RCAST(void *, &rv)); 3327 #if 0 3328 d[0] = s[3]; 3329 d[1] = s[2]; 3330 d[2] = s[1]; 3331 d[3] = s[0]; 3332 d[4] = s[7]; 3333 d[5] = s[6]; 3334 d[6] = s[5]; 3335 d[7] = s[4]; 3336 #else 3337 d[0] = s[7]; 3338 d[1] = s[6]; 3339 d[2] = s[5]; 3340 d[3] = s[4]; 3341 d[4] = s[3]; 3342 d[5] = s[2]; 3343 d[6] = s[1]; 3344 d[7] = s[0]; 3345 #endif 3346 return rv; 3347 } 3348 3349 /* 3350 * byteswap a single magic entry 3351 */ 3352 private void 3353 bs1(struct magic *m) 3354 { 3355 m->cont_level = swap2(m->cont_level); 3356 m->offset = swap4(CAST(uint32_t, m->offset)); 3357 m->in_offset = swap4(CAST(uint32_t, m->in_offset)); 3358 m->lineno = swap4(CAST(uint32_t, m->lineno)); 3359 if (IS_STRING(m->type)) { 3360 m->str_range = swap4(m->str_range); 3361 m->str_flags = swap4(m->str_flags); 3362 } 3363 else { 3364 m->value.q = swap8(m->value.q); 3365 m->num_mask = swap8(m->num_mask); 3366 } 3367 } 3368 3369 protected size_t 3370 file_pstring_length_size(const struct magic *m) 3371 { 3372 switch (m->str_flags & PSTRING_LEN) { 3373 case PSTRING_1_LE: 3374 return 1; 3375 case PSTRING_2_LE: 3376 case PSTRING_2_BE: 3377 return 2; 3378 case PSTRING_4_LE: 3379 case PSTRING_4_BE: 3380 return 4; 3381 default: 3382 abort(); /* Impossible */ 3383 return 1; 3384 } 3385 } 3386 protected size_t 3387 file_pstring_get_length(const struct magic *m, const char *ss) 3388 { 3389 size_t len = 0; 3390 const unsigned char *s = RCAST(const unsigned char *, ss); 3391 unsigned int s3, s2, s1, s0; 3392 3393 switch (m->str_flags & PSTRING_LEN) { 3394 case PSTRING_1_LE: 3395 len = *s; 3396 break; 3397 case PSTRING_2_LE: 3398 s0 = s[0]; 3399 s1 = s[1]; 3400 len = (s1 << 8) | s0; 3401 break; 3402 case PSTRING_2_BE: 3403 s0 = s[0]; 3404 s1 = s[1]; 3405 len = (s0 << 8) | s1; 3406 break; 3407 case PSTRING_4_LE: 3408 s0 = s[0]; 3409 s1 = s[1]; 3410 s2 = s[2]; 3411 s3 = s[3]; 3412 len = (s3 << 24) | (s2 << 16) | (s1 << 8) | s0; 3413 break; 3414 case PSTRING_4_BE: 3415 s0 = s[0]; 3416 s1 = s[1]; 3417 s2 = s[2]; 3418 s3 = s[3]; 3419 len = (s0 << 24) | (s1 << 16) | (s2 << 8) | s3; 3420 break; 3421 default: 3422 abort(); /* Impossible */ 3423 } 3424 3425 if (m->str_flags & PSTRING_LENGTH_INCLUDES_ITSELF) 3426 len -= file_pstring_length_size(m); 3427 3428 return len; 3429 } 3430 3431 protected int 3432 file_magicfind(struct magic_set *ms, const char *name, struct mlist *v) 3433 { 3434 uint32_t i, j; 3435 struct mlist *mlist, *ml; 3436 3437 mlist = ms->mlist[1]; 3438 3439 for (ml = mlist->next; ml != mlist; ml = ml->next) { 3440 struct magic *ma = ml->magic; 3441 uint32_t nma = ml->nmagic; 3442 for (i = 0; i < nma; i++) { 3443 if (ma[i].type != FILE_NAME) 3444 continue; 3445 if (strcmp(ma[i].value.s, name) == 0) { 3446 v->magic = &ma[i]; 3447 for (j = i + 1; j < nma; j++) 3448 if (ma[j].cont_level == 0) 3449 break; 3450 v->nmagic = j - i; 3451 return 0; 3452 } 3453 } 3454 } 3455 return -1; 3456 } 3457