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 1.202 2014/03/14 18:48:11 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 #ifdef HAVE_STDDEF_H 44 #include <stddef.h> 45 #endif 46 #include <string.h> 47 #include <assert.h> 48 #include <ctype.h> 49 #include <fcntl.h> 50 #ifdef QUICK 51 #include <sys/mman.h> 52 #endif 53 #include <dirent.h> 54 #if defined(HAVE_LIMITS_H) 55 #include <limits.h> 56 #endif 57 58 #ifndef SSIZE_MAX 59 #define MAXMAGIC_SIZE ((ssize_t)0x7fffffff) 60 #else 61 #define MAXMAGIC_SIZE SSIZE_MAX 62 #endif 63 64 #define EATAB {while (isascii((unsigned char) *l) && \ 65 isspace((unsigned char) *l)) ++l;} 66 #define LOWCASE(l) (isupper((unsigned char) (l)) ? \ 67 tolower((unsigned char) (l)) : (l)) 68 /* 69 * Work around a bug in headers on Digital Unix. 70 * At least confirmed for: OSF1 V4.0 878 71 */ 72 #if defined(__osf__) && defined(__DECC) 73 #ifdef MAP_FAILED 74 #undef MAP_FAILED 75 #endif 76 #endif 77 78 #ifndef MAP_FAILED 79 #define MAP_FAILED (void *) -1 80 #endif 81 82 #ifndef MAP_FILE 83 #define MAP_FILE 0 84 #endif 85 86 #define ALLOC_CHUNK (size_t)10 87 #define ALLOC_INCR (size_t)200 88 89 struct magic_entry { 90 struct magic *mp; 91 uint32_t cont_count; 92 uint32_t max_count; 93 }; 94 95 struct magic_entry_set { 96 struct magic_entry *me; 97 uint32_t count; 98 uint32_t max; 99 }; 100 101 struct magic_map { 102 void *p; 103 size_t len; 104 struct magic *magic[MAGIC_SETS]; 105 uint32_t nmagic[MAGIC_SETS]; 106 }; 107 108 int file_formats[FILE_NAMES_SIZE]; 109 const size_t file_nformats = FILE_NAMES_SIZE; 110 const char *file_names[FILE_NAMES_SIZE]; 111 const size_t file_nnames = FILE_NAMES_SIZE; 112 113 private int getvalue(struct magic_set *ms, struct magic *, const char **, int); 114 private int hextoint(int); 115 private const char *getstr(struct magic_set *, struct magic *, const char *, 116 int); 117 private int parse(struct magic_set *, struct magic_entry *, const char *, 118 size_t, int); 119 private void eatsize(const char **); 120 private int apprentice_1(struct magic_set *, const char *, int); 121 private size_t apprentice_magic_strength(const struct magic *); 122 private int apprentice_sort(const void *, const void *); 123 private void apprentice_list(struct mlist *, int ); 124 private struct magic_map *apprentice_load(struct magic_set *, 125 const char *, int); 126 private struct mlist *mlist_alloc(void); 127 private void mlist_free(struct mlist *); 128 private void byteswap(struct magic *, uint32_t); 129 private void bs1(struct magic *); 130 private uint16_t swap2(uint16_t); 131 private uint32_t swap4(uint32_t); 132 private uint64_t swap8(uint64_t); 133 private char *mkdbname(struct magic_set *, const char *, int); 134 private struct magic_map *apprentice_map(struct magic_set *, const char *); 135 private void apprentice_unmap(struct magic_map *); 136 private int apprentice_compile(struct magic_set *, struct magic_map *, 137 const char *); 138 private int check_format_type(const char *, int); 139 private int check_format(struct magic_set *, struct magic *); 140 private int get_op(char); 141 private int parse_mime(struct magic_set *, struct magic_entry *, const char *); 142 private int parse_strength(struct magic_set *, struct magic_entry *, const char *); 143 private int parse_apple(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(strength), 159 #undef DECLARE_FIELD 160 { NULL, 0, NULL } 161 }; 162 163 #ifdef COMPILE_ONLY 164 165 int main(int, char *[]); 166 167 int 168 main(int argc, char *argv[]) 169 { 170 int ret; 171 struct magic_set *ms; 172 char *progname; 173 174 if ((progname = strrchr(argv[0], '/')) != NULL) 175 progname++; 176 else 177 progname = argv[0]; 178 179 if (argc != 2) { 180 (void)fprintf(stderr, "Usage: %s file\n", progname); 181 return 1; 182 } 183 184 if ((ms = magic_open(MAGIC_CHECK)) == NULL) { 185 (void)fprintf(stderr, "%s: %s\n", progname, strerror(errno)); 186 return 1; 187 } 188 ret = magic_compile(ms, argv[1]) == -1 ? 1 : 0; 189 if (ret == 1) 190 (void)fprintf(stderr, "%s: %s\n", progname, magic_error(ms)); 191 magic_close(ms); 192 return ret; 193 } 194 #endif /* COMPILE_ONLY */ 195 196 struct type_tbl_s { 197 const char name[16]; 198 const size_t len; 199 const int type; 200 const int format; 201 }; 202 203 /* 204 * XXX - the actual Single UNIX Specification says that "long" means "long", 205 * as in the C data type, but we treat it as meaning "4-byte integer". 206 * Given that the OS X version of file 5.04 did the same, I guess that passes 207 * the actual test; having "long" be dependent on how big a "long" is on 208 * the machine running "file" is silly. 209 */ 210 static const struct type_tbl_s type_tbl[] = { 211 # define XX(s) s, (sizeof(s) - 1) 212 # define XX_NULL "", 0 213 { XX("invalid"), FILE_INVALID, FILE_FMT_NONE }, 214 { XX("byte"), FILE_BYTE, FILE_FMT_NUM }, 215 { XX("short"), FILE_SHORT, FILE_FMT_NUM }, 216 { XX("default"), FILE_DEFAULT, FILE_FMT_NONE }, 217 { XX("long"), FILE_LONG, FILE_FMT_NUM }, 218 { XX("string"), FILE_STRING, FILE_FMT_STR }, 219 { XX("date"), FILE_DATE, FILE_FMT_STR }, 220 { XX("beshort"), FILE_BESHORT, FILE_FMT_NUM }, 221 { XX("belong"), FILE_BELONG, FILE_FMT_NUM }, 222 { XX("bedate"), FILE_BEDATE, FILE_FMT_STR }, 223 { XX("leshort"), FILE_LESHORT, FILE_FMT_NUM }, 224 { XX("lelong"), FILE_LELONG, FILE_FMT_NUM }, 225 { XX("ledate"), FILE_LEDATE, FILE_FMT_STR }, 226 { XX("pstring"), FILE_PSTRING, FILE_FMT_STR }, 227 { XX("ldate"), FILE_LDATE, FILE_FMT_STR }, 228 { XX("beldate"), FILE_BELDATE, FILE_FMT_STR }, 229 { XX("leldate"), FILE_LELDATE, FILE_FMT_STR }, 230 { XX("regex"), FILE_REGEX, FILE_FMT_STR }, 231 { XX("bestring16"), FILE_BESTRING16, FILE_FMT_STR }, 232 { XX("lestring16"), FILE_LESTRING16, FILE_FMT_STR }, 233 { XX("search"), FILE_SEARCH, FILE_FMT_STR }, 234 { XX("medate"), FILE_MEDATE, FILE_FMT_STR }, 235 { XX("meldate"), FILE_MELDATE, FILE_FMT_STR }, 236 { XX("melong"), FILE_MELONG, FILE_FMT_NUM }, 237 { XX("quad"), FILE_QUAD, FILE_FMT_QUAD }, 238 { XX("lequad"), FILE_LEQUAD, FILE_FMT_QUAD }, 239 { XX("bequad"), FILE_BEQUAD, FILE_FMT_QUAD }, 240 { XX("qdate"), FILE_QDATE, FILE_FMT_STR }, 241 { XX("leqdate"), FILE_LEQDATE, FILE_FMT_STR }, 242 { XX("beqdate"), FILE_BEQDATE, FILE_FMT_STR }, 243 { XX("qldate"), FILE_QLDATE, FILE_FMT_STR }, 244 { XX("leqldate"), FILE_LEQLDATE, FILE_FMT_STR }, 245 { XX("beqldate"), FILE_BEQLDATE, FILE_FMT_STR }, 246 { XX("float"), FILE_FLOAT, FILE_FMT_FLOAT }, 247 { XX("befloat"), FILE_BEFLOAT, FILE_FMT_FLOAT }, 248 { XX("lefloat"), FILE_LEFLOAT, FILE_FMT_FLOAT }, 249 { XX("double"), FILE_DOUBLE, FILE_FMT_DOUBLE }, 250 { XX("bedouble"), FILE_BEDOUBLE, FILE_FMT_DOUBLE }, 251 { XX("ledouble"), FILE_LEDOUBLE, FILE_FMT_DOUBLE }, 252 { XX("leid3"), FILE_LEID3, FILE_FMT_NUM }, 253 { XX("beid3"), FILE_BEID3, FILE_FMT_NUM }, 254 { XX("indirect"), FILE_INDIRECT, FILE_FMT_NUM }, 255 { XX("qwdate"), FILE_QWDATE, FILE_FMT_STR }, 256 { XX("leqwdate"), FILE_LEQWDATE, FILE_FMT_STR }, 257 { XX("beqwdate"), FILE_BEQWDATE, FILE_FMT_STR }, 258 { XX("name"), FILE_NAME, FILE_FMT_NONE }, 259 { XX("use"), FILE_USE, FILE_FMT_NONE }, 260 { XX("clear"), FILE_CLEAR, FILE_FMT_NONE }, 261 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 262 }; 263 264 /* 265 * These are not types, and cannot be preceded by "u" to make them 266 * unsigned. 267 */ 268 static const struct type_tbl_s special_tbl[] = { 269 { XX("name"), FILE_NAME, FILE_FMT_STR }, 270 { XX("use"), FILE_USE, FILE_FMT_STR }, 271 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 272 }; 273 # undef XX 274 # undef XX_NULL 275 276 private int 277 get_type(const struct type_tbl_s *tbl, const char *l, const char **t) 278 { 279 const struct type_tbl_s *p; 280 281 for (p = tbl; p->len; p++) { 282 if (strncmp(l, p->name, p->len) == 0) { 283 if (t) 284 *t = l + p->len; 285 break; 286 } 287 } 288 return p->type; 289 } 290 291 private int 292 get_standard_integer_type(const char *l, const char **t) 293 { 294 int type; 295 296 if (isalpha((unsigned char)l[1])) { 297 switch (l[1]) { 298 case 'C': 299 /* "dC" and "uC" */ 300 type = FILE_BYTE; 301 break; 302 case 'S': 303 /* "dS" and "uS" */ 304 type = FILE_SHORT; 305 break; 306 case 'I': 307 case 'L': 308 /* 309 * "dI", "dL", "uI", and "uL". 310 * 311 * XXX - the actual Single UNIX Specification says 312 * that "L" means "long", as in the C data type, 313 * but we treat it as meaning "4-byte integer". 314 * Given that the OS X version of file 5.04 did 315 * the same, I guess that passes the actual SUS 316 * validation suite; having "dL" be dependent on 317 * how big a "long" is on the machine running 318 * "file" is silly. 319 */ 320 type = FILE_LONG; 321 break; 322 case 'Q': 323 /* "dQ" and "uQ" */ 324 type = FILE_QUAD; 325 break; 326 default: 327 /* "d{anything else}", "u{anything else}" */ 328 return FILE_INVALID; 329 } 330 l += 2; 331 } else if (isdigit((unsigned char)l[1])) { 332 /* 333 * "d{num}" and "u{num}"; we only support {num} values 334 * of 1, 2, 4, and 8 - the Single UNIX Specification 335 * doesn't say anything about whether arbitrary 336 * values should be supported, but both the Solaris 10 337 * and OS X Mountain Lion versions of file passed the 338 * Single UNIX Specification validation suite, and 339 * neither of them support values bigger than 8 or 340 * non-power-of-2 values. 341 */ 342 if (isdigit((unsigned char)l[2])) { 343 /* Multi-digit, so > 9 */ 344 return FILE_INVALID; 345 } 346 switch (l[1]) { 347 case '1': 348 type = FILE_BYTE; 349 break; 350 case '2': 351 type = FILE_SHORT; 352 break; 353 case '4': 354 type = FILE_LONG; 355 break; 356 case '8': 357 type = FILE_QUAD; 358 break; 359 default: 360 /* XXX - what about 3, 5, 6, or 7? */ 361 return FILE_INVALID; 362 } 363 l += 2; 364 } else { 365 /* 366 * "d" or "u" by itself. 367 */ 368 type = FILE_LONG; 369 ++l; 370 } 371 if (t) 372 *t = l; 373 return type; 374 } 375 376 private void 377 init_file_tables(void) 378 { 379 static int done = 0; 380 const struct type_tbl_s *p; 381 382 if (done) 383 return; 384 done++; 385 386 for (p = type_tbl; p->len; p++) { 387 assert(p->type < FILE_NAMES_SIZE); 388 file_names[p->type] = p->name; 389 file_formats[p->type] = p->format; 390 } 391 assert(p - type_tbl == FILE_NAMES_SIZE); 392 } 393 394 private int 395 add_mlist(struct mlist *mlp, struct magic_map *map, size_t idx) 396 { 397 struct mlist *ml; 398 399 if ((ml = CAST(struct mlist *, malloc(sizeof(*ml)))) == NULL) 400 return -1; 401 402 ml->map = idx == 0 ? map : NULL; 403 ml->magic = map->magic[idx]; 404 ml->nmagic = map->nmagic[idx]; 405 406 mlp->prev->next = ml; 407 ml->prev = mlp->prev; 408 ml->next = mlp; 409 mlp->prev = ml; 410 return 0; 411 } 412 413 /* 414 * Handle one file or directory. 415 */ 416 private int 417 apprentice_1(struct magic_set *ms, const char *fn, int action) 418 { 419 struct mlist *ml; 420 struct magic_map *map; 421 size_t i; 422 423 if (magicsize != FILE_MAGICSIZE) { 424 file_error(ms, 0, "magic element size %lu != %lu", 425 (unsigned long)sizeof(*map->magic[0]), 426 (unsigned long)FILE_MAGICSIZE); 427 return -1; 428 } 429 430 if (action == FILE_COMPILE) { 431 map = apprentice_load(ms, fn, action); 432 if (map == NULL) 433 return -1; 434 return apprentice_compile(ms, map, fn); 435 } 436 437 #ifndef COMPILE_ONLY 438 map = apprentice_map(ms, fn); 439 if (map == NULL) { 440 if (ms->flags & MAGIC_CHECK) 441 file_magwarn(ms, "using regular magic file `%s'", fn); 442 map = apprentice_load(ms, fn, action); 443 if (map == NULL) 444 return -1; 445 } 446 447 for (i = 0; i < MAGIC_SETS; i++) { 448 if (add_mlist(ms->mlist[i], map, i) == -1) { 449 file_oomem(ms, sizeof(*ml)); 450 apprentice_unmap(map); 451 return -1; 452 } 453 } 454 455 if (action == FILE_LIST) { 456 for (i = 0; i < MAGIC_SETS; i++) { 457 printf("Set %zu:\nBinary patterns:\n", i); 458 apprentice_list(ms->mlist[i], BINTEST); 459 printf("Text patterns:\n"); 460 apprentice_list(ms->mlist[i], TEXTTEST); 461 } 462 } 463 464 return 0; 465 #endif /* COMPILE_ONLY */ 466 } 467 468 protected void 469 file_ms_free(struct magic_set *ms) 470 { 471 size_t i; 472 if (ms == NULL) 473 return; 474 for (i = 0; i < MAGIC_SETS; i++) 475 mlist_free(ms->mlist[i]); 476 free(ms->o.pbuf); 477 free(ms->o.buf); 478 free(ms->c.li); 479 free(ms); 480 } 481 482 protected struct magic_set * 483 file_ms_alloc(int flags) 484 { 485 struct magic_set *ms; 486 size_t i, len; 487 488 if ((ms = CAST(struct magic_set *, calloc((size_t)1, 489 sizeof(struct magic_set)))) == NULL) 490 return NULL; 491 492 if (magic_setflags(ms, flags) == -1) { 493 errno = EINVAL; 494 goto free; 495 } 496 497 ms->o.buf = ms->o.pbuf = NULL; 498 len = (ms->c.len = 10) * sizeof(*ms->c.li); 499 500 if ((ms->c.li = CAST(struct level_info *, malloc(len))) == NULL) 501 goto free; 502 503 ms->event_flags = 0; 504 ms->error = -1; 505 for (i = 0; i < MAGIC_SETS; i++) 506 ms->mlist[i] = NULL; 507 ms->file = "unknown"; 508 ms->line = 0; 509 return ms; 510 free: 511 free(ms); 512 return NULL; 513 } 514 515 private void 516 apprentice_unmap(struct magic_map *map) 517 { 518 if (map == NULL) 519 return; 520 if (map->p == NULL) 521 return; 522 #ifdef QUICK 523 if (map->len) 524 (void)munmap(map->p, map->len); 525 else 526 #endif 527 free(map->p); 528 free(map); 529 } 530 531 private struct mlist * 532 mlist_alloc(void) 533 { 534 struct mlist *mlist; 535 if ((mlist = CAST(struct mlist *, calloc(1, sizeof(*mlist)))) == NULL) { 536 return NULL; 537 } 538 mlist->next = mlist->prev = mlist; 539 return mlist; 540 } 541 542 private void 543 mlist_free(struct mlist *mlist) 544 { 545 struct mlist *ml; 546 547 if (mlist == NULL) 548 return; 549 550 for (ml = mlist->next; ml != mlist;) { 551 struct mlist *next = ml->next; 552 if (ml->map) 553 apprentice_unmap(ml->map); 554 free(ml); 555 ml = next; 556 } 557 free(ml); 558 } 559 560 /* const char *fn: list of magic files and directories */ 561 protected int 562 file_apprentice(struct magic_set *ms, const char *fn, int action) 563 { 564 char *p, *mfn; 565 int file_err, errs = -1; 566 size_t i; 567 568 if (ms->mlist[0] != NULL) 569 file_reset(ms); 570 571 if ((fn = magic_getpath(fn, action)) == NULL) 572 return -1; 573 574 init_file_tables(); 575 576 if ((mfn = strdup(fn)) == NULL) { 577 file_oomem(ms, strlen(fn)); 578 return -1; 579 } 580 581 for (i = 0; i < MAGIC_SETS; i++) { 582 mlist_free(ms->mlist[i]); 583 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 584 file_oomem(ms, sizeof(*ms->mlist[i])); 585 if (i != 0) { 586 --i; 587 do 588 mlist_free(ms->mlist[i]); 589 while (i != 0); 590 } 591 free(mfn); 592 return -1; 593 } 594 } 595 fn = mfn; 596 597 while (fn) { 598 p = strchr(fn, PATHSEP); 599 if (p) 600 *p++ = '\0'; 601 if (*fn == '\0') 602 break; 603 file_err = apprentice_1(ms, fn, action); 604 errs = MAX(errs, file_err); 605 fn = p; 606 } 607 608 free(mfn); 609 610 if (errs == -1) { 611 for (i = 0; i < MAGIC_SETS; i++) { 612 mlist_free(ms->mlist[i]); 613 ms->mlist[i] = NULL; 614 } 615 file_error(ms, 0, "could not find any valid magic files!"); 616 return -1; 617 } 618 619 #if 0 620 /* 621 * Always leave the database loaded 622 */ 623 if (action == FILE_LOAD) 624 return 0; 625 626 for (i = 0; i < MAGIC_SETS; i++) { 627 mlist_free(ms->mlist[i]); 628 ms->mlist[i] = NULL; 629 } 630 #endif 631 632 switch (action) { 633 case FILE_LOAD: 634 case FILE_COMPILE: 635 case FILE_CHECK: 636 case FILE_LIST: 637 return 0; 638 default: 639 file_error(ms, 0, "Invalid action %d", action); 640 return -1; 641 } 642 } 643 644 /* 645 * Compute the real length of a magic expression, for the purposes 646 * of determining how "strong" a magic expression is (approximating 647 * how specific its matches are): 648 * - magic characters count 0 unless escaped. 649 * - [] expressions count 1 650 * - {} expressions count 0 651 * - regular characters or escaped magic characters count 1 652 * - 0 length expressions count as one 653 */ 654 private size_t 655 nonmagic(const char *str) 656 { 657 const char *p; 658 size_t rv = 0; 659 660 for (p = str; *p; p++) 661 switch (*p) { 662 case '\\': /* Escaped anything counts 1 */ 663 if (!*++p) 664 p--; 665 rv++; 666 continue; 667 case '?': /* Magic characters count 0 */ 668 case '*': 669 case '.': 670 case '+': 671 case '^': 672 case '$': 673 continue; 674 case '[': /* Bracketed expressions count 1 the ']' */ 675 while (*p && *p != ']') 676 p++; 677 p--; 678 continue; 679 case '{': /* Braced expressions count 0 */ 680 while (*p && *p != '}') 681 p++; 682 if (!*p) 683 p--; 684 continue; 685 default: /* Anything else counts 1 */ 686 rv++; 687 continue; 688 } 689 690 return rv == 0 ? 1 : rv; /* Return at least 1 */ 691 } 692 693 /* 694 * Get weight of this magic entry, for sorting purposes. 695 */ 696 private size_t 697 apprentice_magic_strength(const struct magic *m) 698 { 699 #define MULT 10 700 size_t v, val = 2 * MULT; /* baseline strength */ 701 702 switch (m->type) { 703 case FILE_DEFAULT: /* make sure this sorts last */ 704 if (m->factor_op != FILE_FACTOR_OP_NONE) 705 abort(); 706 return 0; 707 708 case FILE_BYTE: 709 val += 1 * MULT; 710 break; 711 712 case FILE_SHORT: 713 case FILE_LESHORT: 714 case FILE_BESHORT: 715 val += 2 * MULT; 716 break; 717 718 case FILE_LONG: 719 case FILE_LELONG: 720 case FILE_BELONG: 721 case FILE_MELONG: 722 val += 4 * MULT; 723 break; 724 725 case FILE_PSTRING: 726 case FILE_STRING: 727 val += m->vallen * MULT; 728 break; 729 730 case FILE_BESTRING16: 731 case FILE_LESTRING16: 732 val += m->vallen * MULT / 2; 733 break; 734 735 case FILE_SEARCH: 736 val += m->vallen * MAX(MULT / m->vallen, 1); 737 break; 738 739 case FILE_REGEX: 740 v = nonmagic(m->value.s); 741 val += v * MAX(MULT / v, 1); 742 break; 743 744 case FILE_DATE: 745 case FILE_LEDATE: 746 case FILE_BEDATE: 747 case FILE_MEDATE: 748 case FILE_LDATE: 749 case FILE_LELDATE: 750 case FILE_BELDATE: 751 case FILE_MELDATE: 752 case FILE_FLOAT: 753 case FILE_BEFLOAT: 754 case FILE_LEFLOAT: 755 val += 4 * MULT; 756 break; 757 758 case FILE_QUAD: 759 case FILE_BEQUAD: 760 case FILE_LEQUAD: 761 case FILE_QDATE: 762 case FILE_LEQDATE: 763 case FILE_BEQDATE: 764 case FILE_QLDATE: 765 case FILE_LEQLDATE: 766 case FILE_BEQLDATE: 767 case FILE_QWDATE: 768 case FILE_LEQWDATE: 769 case FILE_BEQWDATE: 770 case FILE_DOUBLE: 771 case FILE_BEDOUBLE: 772 case FILE_LEDOUBLE: 773 val += 8 * MULT; 774 break; 775 776 case FILE_INDIRECT: 777 case FILE_NAME: 778 case FILE_USE: 779 break; 780 781 default: 782 val = 0; 783 (void)fprintf(stderr, "Bad type %d\n", m->type); 784 abort(); 785 } 786 787 switch (m->reln) { 788 case 'x': /* matches anything penalize */ 789 case '!': /* matches almost anything penalize */ 790 val = 0; 791 break; 792 793 case '=': /* Exact match, prefer */ 794 val += MULT; 795 break; 796 797 case '>': 798 case '<': /* comparison match reduce strength */ 799 val -= 2 * MULT; 800 break; 801 802 case '^': 803 case '&': /* masking bits, we could count them too */ 804 val -= MULT; 805 break; 806 807 default: 808 (void)fprintf(stderr, "Bad relation %c\n", m->reln); 809 abort(); 810 } 811 812 if (val == 0) /* ensure we only return 0 for FILE_DEFAULT */ 813 val = 1; 814 815 switch (m->factor_op) { 816 case FILE_FACTOR_OP_NONE: 817 break; 818 case FILE_FACTOR_OP_PLUS: 819 val += m->factor; 820 break; 821 case FILE_FACTOR_OP_MINUS: 822 val -= m->factor; 823 break; 824 case FILE_FACTOR_OP_TIMES: 825 val *= m->factor; 826 break; 827 case FILE_FACTOR_OP_DIV: 828 val /= m->factor; 829 break; 830 default: 831 abort(); 832 } 833 834 /* 835 * Magic entries with no description get a bonus because they depend 836 * on subsequent magic entries to print something. 837 */ 838 if (m->desc[0] == '\0') 839 val++; 840 return val; 841 } 842 843 /* 844 * Sort callback for sorting entries by "strength" (basically length) 845 */ 846 private int 847 apprentice_sort(const void *a, const void *b) 848 { 849 const struct magic_entry *ma = CAST(const struct magic_entry *, a); 850 const struct magic_entry *mb = CAST(const struct magic_entry *, b); 851 size_t sa = apprentice_magic_strength(ma->mp); 852 size_t sb = apprentice_magic_strength(mb->mp); 853 if (sa == sb) 854 return 0; 855 else if (sa > sb) 856 return -1; 857 else 858 return 1; 859 } 860 861 /* 862 * Shows sorted patterns list in the order which is used for the matching 863 */ 864 private void 865 apprentice_list(struct mlist *mlist, int mode) 866 { 867 uint32_t magindex = 0; 868 struct mlist *ml; 869 for (ml = mlist->next; ml != mlist; ml = ml->next) { 870 for (magindex = 0; magindex < ml->nmagic; magindex++) { 871 struct magic *m = &ml->magic[magindex]; 872 if ((m->flag & mode) != mode) { 873 /* Skip sub-tests */ 874 while (magindex + 1 < ml->nmagic && 875 ml->magic[magindex + 1].cont_level != 0) 876 ++magindex; 877 continue; /* Skip to next top-level test*/ 878 } 879 880 /* 881 * Try to iterate over the tree until we find item with 882 * description/mimetype. 883 */ 884 while (magindex + 1 < ml->nmagic && 885 ml->magic[magindex + 1].cont_level != 0 && 886 *ml->magic[magindex].desc == '\0' && 887 *ml->magic[magindex].mimetype == '\0') 888 magindex++; 889 890 printf("Strength = %3" SIZE_T_FORMAT "u : %s [%s]\n", 891 apprentice_magic_strength(m), 892 ml->magic[magindex].desc, 893 ml->magic[magindex].mimetype); 894 } 895 } 896 } 897 898 private void 899 set_test_type(struct magic *mstart, struct magic *m) 900 { 901 switch (m->type) { 902 case FILE_BYTE: 903 case FILE_SHORT: 904 case FILE_LONG: 905 case FILE_DATE: 906 case FILE_BESHORT: 907 case FILE_BELONG: 908 case FILE_BEDATE: 909 case FILE_LESHORT: 910 case FILE_LELONG: 911 case FILE_LEDATE: 912 case FILE_LDATE: 913 case FILE_BELDATE: 914 case FILE_LELDATE: 915 case FILE_MEDATE: 916 case FILE_MELDATE: 917 case FILE_MELONG: 918 case FILE_QUAD: 919 case FILE_LEQUAD: 920 case FILE_BEQUAD: 921 case FILE_QDATE: 922 case FILE_LEQDATE: 923 case FILE_BEQDATE: 924 case FILE_QLDATE: 925 case FILE_LEQLDATE: 926 case FILE_BEQLDATE: 927 case FILE_QWDATE: 928 case FILE_LEQWDATE: 929 case FILE_BEQWDATE: 930 case FILE_FLOAT: 931 case FILE_BEFLOAT: 932 case FILE_LEFLOAT: 933 case FILE_DOUBLE: 934 case FILE_BEDOUBLE: 935 case FILE_LEDOUBLE: 936 mstart->flag |= BINTEST; 937 break; 938 case FILE_STRING: 939 case FILE_PSTRING: 940 case FILE_BESTRING16: 941 case FILE_LESTRING16: 942 /* Allow text overrides */ 943 if (mstart->str_flags & STRING_TEXTTEST) 944 mstart->flag |= TEXTTEST; 945 else 946 mstart->flag |= BINTEST; 947 break; 948 case FILE_REGEX: 949 case FILE_SEARCH: 950 /* Check for override */ 951 if (mstart->str_flags & STRING_BINTEST) 952 mstart->flag |= BINTEST; 953 if (mstart->str_flags & STRING_TEXTTEST) 954 mstart->flag |= TEXTTEST; 955 956 if (mstart->flag & (TEXTTEST|BINTEST)) 957 break; 958 959 /* binary test if pattern is not text */ 960 if (file_looks_utf8(m->value.us, (size_t)m->vallen, NULL, 961 NULL) <= 0) 962 mstart->flag |= BINTEST; 963 else 964 mstart->flag |= TEXTTEST; 965 break; 966 case FILE_DEFAULT: 967 /* can't deduce anything; we shouldn't see this at the 968 top level anyway */ 969 break; 970 case FILE_INVALID: 971 default: 972 /* invalid search type, but no need to complain here */ 973 break; 974 } 975 } 976 977 private int 978 addentry(struct magic_set *ms, struct magic_entry *me, 979 struct magic_entry_set *mset) 980 { 981 size_t i = me->mp->type == FILE_NAME ? 1 : 0; 982 if (mset[i].count == mset[i].max) { 983 struct magic_entry *mp; 984 985 mset[i].max += ALLOC_INCR; 986 if ((mp = CAST(struct magic_entry *, 987 realloc(mset[i].me, sizeof(*mp) * mset[i].max))) == 988 NULL) { 989 file_oomem(ms, sizeof(*mp) * mset[i].max); 990 return -1; 991 } 992 (void)memset(&mp[mset[i].count], 0, sizeof(*mp) * 993 ALLOC_INCR); 994 mset[i].me = mp; 995 } 996 mset[i].me[mset[i].count++] = *me; 997 memset(me, 0, sizeof(*me)); 998 return 0; 999 } 1000 1001 /* 1002 * Load and parse one file. 1003 */ 1004 private void 1005 load_1(struct magic_set *ms, int action, const char *fn, int *errs, 1006 struct magic_entry_set *mset) 1007 { 1008 size_t lineno = 0, llen = 0; 1009 char *line = NULL; 1010 ssize_t len; 1011 struct magic_entry me; 1012 1013 FILE *f = fopen(ms->file = fn, "r"); 1014 if (f == NULL) { 1015 if (errno != ENOENT) 1016 file_error(ms, errno, "cannot read magic file `%s'", 1017 fn); 1018 (*errs)++; 1019 return; 1020 } 1021 1022 memset(&me, 0, sizeof(me)); 1023 /* read and parse this file */ 1024 for (ms->line = 1; (len = getline(&line, &llen, f)) != -1; 1025 ms->line++) { 1026 if (len == 0) /* null line, garbage, etc */ 1027 continue; 1028 if (line[len - 1] == '\n') { 1029 lineno++; 1030 line[len - 1] = '\0'; /* delete newline */ 1031 } 1032 switch (line[0]) { 1033 case '\0': /* empty, do not parse */ 1034 case '#': /* comment, do not parse */ 1035 continue; 1036 case '!': 1037 if (line[1] == ':') { 1038 size_t i; 1039 1040 for (i = 0; bang[i].name != NULL; i++) { 1041 if ((size_t)(len - 2) > bang[i].len && 1042 memcmp(bang[i].name, line + 2, 1043 bang[i].len) == 0) 1044 break; 1045 } 1046 if (bang[i].name == NULL) { 1047 file_error(ms, 0, 1048 "Unknown !: entry `%s'", line); 1049 (*errs)++; 1050 continue; 1051 } 1052 if (me.mp == NULL) { 1053 file_error(ms, 0, 1054 "No current entry for :!%s type", 1055 bang[i].name); 1056 (*errs)++; 1057 continue; 1058 } 1059 if ((*bang[i].fun)(ms, &me, 1060 line + bang[i].len + 2) != 0) { 1061 (*errs)++; 1062 continue; 1063 } 1064 continue; 1065 } 1066 /*FALLTHROUGH*/ 1067 default: 1068 again: 1069 switch (parse(ms, &me, line, lineno, action)) { 1070 case 0: 1071 continue; 1072 case 1: 1073 (void)addentry(ms, &me, mset); 1074 goto again; 1075 default: 1076 (*errs)++; 1077 break; 1078 } 1079 } 1080 } 1081 if (me.mp) 1082 (void)addentry(ms, &me, mset); 1083 free(line); 1084 (void)fclose(f); 1085 } 1086 1087 /* 1088 * parse a file or directory of files 1089 * const char *fn: name of magic file or directory 1090 */ 1091 private int 1092 cmpstrp(const void *p1, const void *p2) 1093 { 1094 return strcmp(*(char *const *)p1, *(char *const *)p2); 1095 } 1096 1097 1098 private uint32_t 1099 set_text_binary(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1100 uint32_t starttest) 1101 { 1102 static const char text[] = "text"; 1103 static const char binary[] = "binary"; 1104 static const size_t len = sizeof(text); 1105 1106 uint32_t i = starttest; 1107 1108 do { 1109 set_test_type(me[starttest].mp, me[i].mp); 1110 if ((ms->flags & MAGIC_DEBUG) == 0) 1111 continue; 1112 (void)fprintf(stderr, "%s%s%s: %s\n", 1113 me[i].mp->mimetype, 1114 me[i].mp->mimetype[0] == '\0' ? "" : "; ", 1115 me[i].mp->desc[0] ? me[i].mp->desc : "(no description)", 1116 me[i].mp->flag & BINTEST ? binary : text); 1117 if (me[i].mp->flag & BINTEST) { 1118 char *p = strstr(me[i].mp->desc, text); 1119 if (p && (p == me[i].mp->desc || 1120 isspace((unsigned char)p[-1])) && 1121 (p + len - me[i].mp->desc == MAXstring 1122 || (p[len] == '\0' || 1123 isspace((unsigned char)p[len])))) 1124 (void)fprintf(stderr, "*** Possible " 1125 "binary test for text type\n"); 1126 } 1127 } while (++i < nme && me[i].mp->cont_level != 0); 1128 return i; 1129 } 1130 1131 private void 1132 set_last_default(struct magic_set *ms, struct magic_entry *me, uint32_t nme) 1133 { 1134 uint32_t i; 1135 for (i = 0; i < nme; i++) { 1136 if (me[i].mp->cont_level == 0 && 1137 me[i].mp->type == FILE_DEFAULT) { 1138 while (++i < nme) 1139 if (me[i].mp->cont_level == 0) 1140 break; 1141 if (i != nme) { 1142 /* XXX - Ugh! */ 1143 ms->line = me[i].mp->lineno; 1144 file_magwarn(ms, 1145 "level 0 \"default\" did not sort last"); 1146 } 1147 return; 1148 } 1149 } 1150 } 1151 1152 private int 1153 coalesce_entries(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1154 struct magic **ma, uint32_t *nma) 1155 { 1156 uint32_t i, mentrycount = 0; 1157 size_t slen; 1158 1159 for (i = 0; i < nme; i++) 1160 mentrycount += me[i].cont_count; 1161 1162 slen = sizeof(**ma) * mentrycount; 1163 if ((*ma = CAST(struct magic *, malloc(slen))) == NULL) { 1164 file_oomem(ms, slen); 1165 return -1; 1166 } 1167 1168 mentrycount = 0; 1169 for (i = 0; i < nme; i++) { 1170 (void)memcpy(*ma + mentrycount, me[i].mp, 1171 me[i].cont_count * sizeof(**ma)); 1172 mentrycount += me[i].cont_count; 1173 } 1174 *nma = mentrycount; 1175 return 0; 1176 } 1177 1178 private void 1179 magic_entry_free(struct magic_entry *me, uint32_t nme) 1180 { 1181 uint32_t i; 1182 if (me == NULL) 1183 return; 1184 for (i = 0; i < nme; i++) 1185 free(me[i].mp); 1186 free(me); 1187 } 1188 1189 private struct magic_map * 1190 apprentice_load(struct magic_set *ms, const char *fn, int action) 1191 { 1192 int errs = 0; 1193 uint32_t i, j; 1194 size_t files = 0, maxfiles = 0; 1195 char **filearr = NULL, *mfn; 1196 struct stat st; 1197 struct magic_map *map; 1198 struct magic_entry_set mset[MAGIC_SETS]; 1199 DIR *dir; 1200 struct dirent *d; 1201 1202 memset(mset, 0, sizeof(mset)); 1203 ms->flags |= MAGIC_CHECK; /* Enable checks for parsed files */ 1204 1205 1206 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) 1207 { 1208 file_oomem(ms, sizeof(*map)); 1209 return NULL; 1210 } 1211 1212 /* print silly verbose header for USG compat. */ 1213 if (action == FILE_CHECK) 1214 (void)fprintf(stderr, "%s\n", usg_hdr); 1215 1216 /* load directory or file */ 1217 if (stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) { 1218 dir = opendir(fn); 1219 if (!dir) { 1220 errs++; 1221 goto out; 1222 } 1223 while ((d = readdir(dir)) != NULL) { 1224 if (asprintf(&mfn, "%s/%s", fn, d->d_name) < 0) { 1225 file_oomem(ms, 1226 strlen(fn) + strlen(d->d_name) + 2); 1227 errs++; 1228 closedir(dir); 1229 goto out; 1230 } 1231 if (stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) { 1232 free(mfn); 1233 continue; 1234 } 1235 if (files >= maxfiles) { 1236 size_t mlen; 1237 maxfiles = (maxfiles + 1) * 2; 1238 mlen = maxfiles * sizeof(*filearr); 1239 if ((filearr = CAST(char **, 1240 realloc(filearr, mlen))) == NULL) { 1241 file_oomem(ms, mlen); 1242 free(mfn); 1243 closedir(dir); 1244 errs++; 1245 goto out; 1246 } 1247 } 1248 filearr[files++] = mfn; 1249 } 1250 closedir(dir); 1251 qsort(filearr, files, sizeof(*filearr), cmpstrp); 1252 for (i = 0; i < files; i++) { 1253 load_1(ms, action, filearr[i], &errs, mset); 1254 free(filearr[i]); 1255 } 1256 free(filearr); 1257 } else 1258 load_1(ms, action, fn, &errs, mset); 1259 if (errs) 1260 goto out; 1261 1262 for (j = 0; j < MAGIC_SETS; j++) { 1263 /* Set types of tests */ 1264 for (i = 0; i < mset[j].count; ) { 1265 if (mset[j].me[i].mp->cont_level != 0) { 1266 i++; 1267 continue; 1268 } 1269 i = set_text_binary(ms, mset[j].me, mset[j].count, i); 1270 } 1271 qsort(mset[j].me, mset[j].count, sizeof(*mset[j].me), 1272 apprentice_sort); 1273 1274 /* 1275 * Make sure that any level 0 "default" line is last 1276 * (if one exists). 1277 */ 1278 set_last_default(ms, mset[j].me, mset[j].count); 1279 1280 /* coalesce per file arrays into a single one */ 1281 if (coalesce_entries(ms, mset[j].me, mset[j].count, 1282 &map->magic[j], &map->nmagic[j]) == -1) { 1283 errs++; 1284 goto out; 1285 } 1286 } 1287 1288 out: 1289 for (j = 0; j < MAGIC_SETS; j++) 1290 magic_entry_free(mset[j].me, mset[j].count); 1291 1292 if (errs) { 1293 for (j = 0; j < MAGIC_SETS; j++) { 1294 if (map->magic[j]) 1295 free(map->magic[j]); 1296 } 1297 free(map); 1298 return NULL; 1299 } 1300 return map; 1301 } 1302 1303 /* 1304 * extend the sign bit if the comparison is to be signed 1305 */ 1306 protected uint64_t 1307 file_signextend(struct magic_set *ms, struct magic *m, uint64_t v) 1308 { 1309 if (!(m->flag & UNSIGNED)) { 1310 switch(m->type) { 1311 /* 1312 * Do not remove the casts below. They are 1313 * vital. When later compared with the data, 1314 * the sign extension must have happened. 1315 */ 1316 case FILE_BYTE: 1317 v = (char) v; 1318 break; 1319 case FILE_SHORT: 1320 case FILE_BESHORT: 1321 case FILE_LESHORT: 1322 v = (short) v; 1323 break; 1324 case FILE_DATE: 1325 case FILE_BEDATE: 1326 case FILE_LEDATE: 1327 case FILE_MEDATE: 1328 case FILE_LDATE: 1329 case FILE_BELDATE: 1330 case FILE_LELDATE: 1331 case FILE_MELDATE: 1332 case FILE_LONG: 1333 case FILE_BELONG: 1334 case FILE_LELONG: 1335 case FILE_MELONG: 1336 case FILE_FLOAT: 1337 case FILE_BEFLOAT: 1338 case FILE_LEFLOAT: 1339 v = (int32_t) v; 1340 break; 1341 case FILE_QUAD: 1342 case FILE_BEQUAD: 1343 case FILE_LEQUAD: 1344 case FILE_QDATE: 1345 case FILE_QLDATE: 1346 case FILE_QWDATE: 1347 case FILE_BEQDATE: 1348 case FILE_BEQLDATE: 1349 case FILE_BEQWDATE: 1350 case FILE_LEQDATE: 1351 case FILE_LEQLDATE: 1352 case FILE_LEQWDATE: 1353 case FILE_DOUBLE: 1354 case FILE_BEDOUBLE: 1355 case FILE_LEDOUBLE: 1356 v = (int64_t) v; 1357 break; 1358 case FILE_STRING: 1359 case FILE_PSTRING: 1360 case FILE_BESTRING16: 1361 case FILE_LESTRING16: 1362 case FILE_REGEX: 1363 case FILE_SEARCH: 1364 case FILE_DEFAULT: 1365 case FILE_INDIRECT: 1366 case FILE_NAME: 1367 case FILE_USE: 1368 case FILE_CLEAR: 1369 break; 1370 default: 1371 if (ms->flags & MAGIC_CHECK) 1372 file_magwarn(ms, "cannot happen: m->type=%d\n", 1373 m->type); 1374 return ~0U; 1375 } 1376 } 1377 return v; 1378 } 1379 1380 private int 1381 string_modifier_check(struct magic_set *ms, struct magic *m) 1382 { 1383 if ((ms->flags & MAGIC_CHECK) == 0) 1384 return 0; 1385 1386 if (m->type != FILE_PSTRING && (m->str_flags & PSTRING_LEN) != 0) { 1387 file_magwarn(ms, 1388 "'/BHhLl' modifiers are only allowed for pascal strings\n"); 1389 return -1; 1390 } 1391 switch (m->type) { 1392 case FILE_BESTRING16: 1393 case FILE_LESTRING16: 1394 if (m->str_flags != 0) { 1395 file_magwarn(ms, 1396 "no modifiers allowed for 16-bit strings\n"); 1397 return -1; 1398 } 1399 break; 1400 case FILE_STRING: 1401 case FILE_PSTRING: 1402 if ((m->str_flags & REGEX_OFFSET_START) != 0) { 1403 file_magwarn(ms, 1404 "'/%c' only allowed on regex and search\n", 1405 CHAR_REGEX_OFFSET_START); 1406 return -1; 1407 } 1408 break; 1409 case FILE_SEARCH: 1410 if (m->str_range == 0) { 1411 file_magwarn(ms, 1412 "missing range; defaulting to %d\n", 1413 STRING_DEFAULT_RANGE); 1414 m->str_range = STRING_DEFAULT_RANGE; 1415 return -1; 1416 } 1417 break; 1418 case FILE_REGEX: 1419 if ((m->str_flags & STRING_COMPACT_WHITESPACE) != 0) { 1420 file_magwarn(ms, "'/%c' not allowed on regex\n", 1421 CHAR_COMPACT_WHITESPACE); 1422 return -1; 1423 } 1424 if ((m->str_flags & STRING_COMPACT_OPTIONAL_WHITESPACE) != 0) { 1425 file_magwarn(ms, "'/%c' not allowed on regex\n", 1426 CHAR_COMPACT_OPTIONAL_WHITESPACE); 1427 return -1; 1428 } 1429 break; 1430 default: 1431 file_magwarn(ms, "coding error: m->type=%d\n", 1432 m->type); 1433 return -1; 1434 } 1435 return 0; 1436 } 1437 1438 private int 1439 get_op(char c) 1440 { 1441 switch (c) { 1442 case '&': 1443 return FILE_OPAND; 1444 case '|': 1445 return FILE_OPOR; 1446 case '^': 1447 return FILE_OPXOR; 1448 case '+': 1449 return FILE_OPADD; 1450 case '-': 1451 return FILE_OPMINUS; 1452 case '*': 1453 return FILE_OPMULTIPLY; 1454 case '/': 1455 return FILE_OPDIVIDE; 1456 case '%': 1457 return FILE_OPMODULO; 1458 default: 1459 return -1; 1460 } 1461 } 1462 1463 #ifdef ENABLE_CONDITIONALS 1464 private int 1465 get_cond(const char *l, const char **t) 1466 { 1467 static const struct cond_tbl_s { 1468 char name[8]; 1469 size_t len; 1470 int cond; 1471 } cond_tbl[] = { 1472 { "if", 2, COND_IF }, 1473 { "elif", 4, COND_ELIF }, 1474 { "else", 4, COND_ELSE }, 1475 { "", 0, COND_NONE }, 1476 }; 1477 const struct cond_tbl_s *p; 1478 1479 for (p = cond_tbl; p->len; p++) { 1480 if (strncmp(l, p->name, p->len) == 0 && 1481 isspace((unsigned char)l[p->len])) { 1482 if (t) 1483 *t = l + p->len; 1484 break; 1485 } 1486 } 1487 return p->cond; 1488 } 1489 1490 private int 1491 check_cond(struct magic_set *ms, int cond, uint32_t cont_level) 1492 { 1493 int last_cond; 1494 last_cond = ms->c.li[cont_level].last_cond; 1495 1496 switch (cond) { 1497 case COND_IF: 1498 if (last_cond != COND_NONE && last_cond != COND_ELIF) { 1499 if (ms->flags & MAGIC_CHECK) 1500 file_magwarn(ms, "syntax error: `if'"); 1501 return -1; 1502 } 1503 last_cond = COND_IF; 1504 break; 1505 1506 case COND_ELIF: 1507 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1508 if (ms->flags & MAGIC_CHECK) 1509 file_magwarn(ms, "syntax error: `elif'"); 1510 return -1; 1511 } 1512 last_cond = COND_ELIF; 1513 break; 1514 1515 case COND_ELSE: 1516 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1517 if (ms->flags & MAGIC_CHECK) 1518 file_magwarn(ms, "syntax error: `else'"); 1519 return -1; 1520 } 1521 last_cond = COND_NONE; 1522 break; 1523 1524 case COND_NONE: 1525 last_cond = COND_NONE; 1526 break; 1527 } 1528 1529 ms->c.li[cont_level].last_cond = last_cond; 1530 return 0; 1531 } 1532 #endif /* ENABLE_CONDITIONALS */ 1533 1534 /* 1535 * parse one line from magic file, put into magic[index++] if valid 1536 */ 1537 private int 1538 parse(struct magic_set *ms, struct magic_entry *me, const char *line, 1539 size_t lineno, int action) 1540 { 1541 #ifdef ENABLE_CONDITIONALS 1542 static uint32_t last_cont_level = 0; 1543 #endif 1544 size_t i; 1545 struct magic *m; 1546 const char *l = line; 1547 char *t; 1548 int op; 1549 uint32_t cont_level; 1550 int32_t diff; 1551 1552 cont_level = 0; 1553 1554 /* 1555 * Parse the offset. 1556 */ 1557 while (*l == '>') { 1558 ++l; /* step over */ 1559 cont_level++; 1560 } 1561 #ifdef ENABLE_CONDITIONALS 1562 if (cont_level == 0 || cont_level > last_cont_level) 1563 if (file_check_mem(ms, cont_level) == -1) 1564 return -1; 1565 last_cont_level = cont_level; 1566 #endif 1567 if (cont_level != 0) { 1568 if (me->mp == NULL) { 1569 file_magerror(ms, "No current entry for continuation"); 1570 return -1; 1571 } 1572 if (me->cont_count == 0) { 1573 file_magerror(ms, "Continuations present with 0 count"); 1574 return -1; 1575 } 1576 m = &me->mp[me->cont_count - 1]; 1577 diff = (int32_t)cont_level - (int32_t)m->cont_level; 1578 if (diff > 1) 1579 file_magwarn(ms, "New continuation level %u is more " 1580 "than one larger than current level %u", cont_level, 1581 m->cont_level); 1582 if (me->cont_count == me->max_count) { 1583 struct magic *nm; 1584 size_t cnt = me->max_count + ALLOC_CHUNK; 1585 if ((nm = CAST(struct magic *, realloc(me->mp, 1586 sizeof(*nm) * cnt))) == NULL) { 1587 file_oomem(ms, sizeof(*nm) * cnt); 1588 return -1; 1589 } 1590 me->mp = m = nm; 1591 me->max_count = CAST(uint32_t, cnt); 1592 } 1593 m = &me->mp[me->cont_count++]; 1594 (void)memset(m, 0, sizeof(*m)); 1595 m->cont_level = cont_level; 1596 } else { 1597 static const size_t len = sizeof(*m) * ALLOC_CHUNK; 1598 if (me->mp != NULL) 1599 return 1; 1600 if ((m = CAST(struct magic *, malloc(len))) == NULL) { 1601 file_oomem(ms, len); 1602 return -1; 1603 } 1604 me->mp = m; 1605 me->max_count = ALLOC_CHUNK; 1606 (void)memset(m, 0, sizeof(*m)); 1607 m->factor_op = FILE_FACTOR_OP_NONE; 1608 m->cont_level = 0; 1609 me->cont_count = 1; 1610 } 1611 m->lineno = CAST(uint32_t, lineno); 1612 1613 if (*l == '&') { /* m->cont_level == 0 checked below. */ 1614 ++l; /* step over */ 1615 m->flag |= OFFADD; 1616 } 1617 if (*l == '(') { 1618 ++l; /* step over */ 1619 m->flag |= INDIR; 1620 if (m->flag & OFFADD) 1621 m->flag = (m->flag & ~OFFADD) | INDIROFFADD; 1622 1623 if (*l == '&') { /* m->cont_level == 0 checked below */ 1624 ++l; /* step over */ 1625 m->flag |= OFFADD; 1626 } 1627 } 1628 /* Indirect offsets are not valid at level 0. */ 1629 if (m->cont_level == 0 && (m->flag & (OFFADD | INDIROFFADD))) 1630 if (ms->flags & MAGIC_CHECK) 1631 file_magwarn(ms, "relative offset at level 0"); 1632 1633 /* get offset, then skip over it */ 1634 m->offset = (uint32_t)strtoul(l, &t, 0); 1635 if (l == t) 1636 if (ms->flags & MAGIC_CHECK) 1637 file_magwarn(ms, "offset `%s' invalid", l); 1638 l = t; 1639 1640 if (m->flag & INDIR) { 1641 m->in_type = FILE_LONG; 1642 m->in_offset = 0; 1643 /* 1644 * read [.lbs][+-]nnnnn) 1645 */ 1646 if (*l == '.') { 1647 l++; 1648 switch (*l) { 1649 case 'l': 1650 m->in_type = FILE_LELONG; 1651 break; 1652 case 'L': 1653 m->in_type = FILE_BELONG; 1654 break; 1655 case 'm': 1656 m->in_type = FILE_MELONG; 1657 break; 1658 case 'h': 1659 case 's': 1660 m->in_type = FILE_LESHORT; 1661 break; 1662 case 'H': 1663 case 'S': 1664 m->in_type = FILE_BESHORT; 1665 break; 1666 case 'c': 1667 case 'b': 1668 case 'C': 1669 case 'B': 1670 m->in_type = FILE_BYTE; 1671 break; 1672 case 'e': 1673 case 'f': 1674 case 'g': 1675 m->in_type = FILE_LEDOUBLE; 1676 break; 1677 case 'E': 1678 case 'F': 1679 case 'G': 1680 m->in_type = FILE_BEDOUBLE; 1681 break; 1682 case 'i': 1683 m->in_type = FILE_LEID3; 1684 break; 1685 case 'I': 1686 m->in_type = FILE_BEID3; 1687 break; 1688 default: 1689 if (ms->flags & MAGIC_CHECK) 1690 file_magwarn(ms, 1691 "indirect offset type `%c' invalid", 1692 *l); 1693 break; 1694 } 1695 l++; 1696 } 1697 1698 m->in_op = 0; 1699 if (*l == '~') { 1700 m->in_op |= FILE_OPINVERSE; 1701 l++; 1702 } 1703 if ((op = get_op(*l)) != -1) { 1704 m->in_op |= op; 1705 l++; 1706 } 1707 if (*l == '(') { 1708 m->in_op |= FILE_OPINDIRECT; 1709 l++; 1710 } 1711 if (isdigit((unsigned char)*l) || *l == '-') { 1712 m->in_offset = (int32_t)strtol(l, &t, 0); 1713 if (l == t) 1714 if (ms->flags & MAGIC_CHECK) 1715 file_magwarn(ms, 1716 "in_offset `%s' invalid", l); 1717 l = t; 1718 } 1719 if (*l++ != ')' || 1720 ((m->in_op & FILE_OPINDIRECT) && *l++ != ')')) 1721 if (ms->flags & MAGIC_CHECK) 1722 file_magwarn(ms, 1723 "missing ')' in indirect offset"); 1724 } 1725 EATAB; 1726 1727 #ifdef ENABLE_CONDITIONALS 1728 m->cond = get_cond(l, &l); 1729 if (check_cond(ms, m->cond, cont_level) == -1) 1730 return -1; 1731 1732 EATAB; 1733 #endif 1734 1735 /* 1736 * Parse the type. 1737 */ 1738 if (*l == 'u') { 1739 /* 1740 * Try it as a keyword type prefixed by "u"; match what 1741 * follows the "u". If that fails, try it as an SUS 1742 * integer type. 1743 */ 1744 m->type = get_type(type_tbl, l + 1, &l); 1745 if (m->type == FILE_INVALID) { 1746 /* 1747 * Not a keyword type; parse it as an SUS type, 1748 * 'u' possibly followed by a number or C/S/L. 1749 */ 1750 m->type = get_standard_integer_type(l, &l); 1751 } 1752 // It's unsigned. 1753 if (m->type != FILE_INVALID) 1754 m->flag |= UNSIGNED; 1755 } else { 1756 /* 1757 * Try it as a keyword type. If that fails, try it as 1758 * an SUS integer type if it begins with "d" or as an 1759 * SUS string type if it begins with "s". In any case, 1760 * it's not unsigned. 1761 */ 1762 m->type = get_type(type_tbl, l, &l); 1763 if (m->type == FILE_INVALID) { 1764 /* 1765 * Not a keyword type; parse it as an SUS type, 1766 * either 'd' possibly followed by a number or 1767 * C/S/L, or just 's'. 1768 */ 1769 if (*l == 'd') 1770 m->type = get_standard_integer_type(l, &l); 1771 else if (*l == 's' && !isalpha((unsigned char)l[1])) { 1772 m->type = FILE_STRING; 1773 ++l; 1774 } 1775 } 1776 } 1777 1778 if (m->type == FILE_INVALID) { 1779 /* Not found - try it as a special keyword. */ 1780 m->type = get_type(special_tbl, l, &l); 1781 } 1782 1783 if (m->type == FILE_INVALID) { 1784 if (ms->flags & MAGIC_CHECK) 1785 file_magwarn(ms, "type `%s' invalid", l); 1786 return -1; 1787 } 1788 1789 /* New-style anding: "0 byte&0x80 =0x80 dynamically linked" */ 1790 /* New and improved: ~ & | ^ + - * / % -- exciting, isn't it? */ 1791 1792 m->mask_op = 0; 1793 if (*l == '~') { 1794 if (!IS_STRING(m->type)) 1795 m->mask_op |= FILE_OPINVERSE; 1796 else if (ms->flags & MAGIC_CHECK) 1797 file_magwarn(ms, "'~' invalid for string types"); 1798 ++l; 1799 } 1800 m->str_range = 0; 1801 m->str_flags = m->type == FILE_PSTRING ? PSTRING_1_LE : 0; 1802 if ((op = get_op(*l)) != -1) { 1803 if (!IS_STRING(m->type)) { 1804 uint64_t val; 1805 ++l; 1806 m->mask_op |= op; 1807 val = (uint64_t)strtoull(l, &t, 0); 1808 l = t; 1809 m->num_mask = file_signextend(ms, m, val); 1810 eatsize(&l); 1811 } 1812 else if (op == FILE_OPDIVIDE) { 1813 int have_range = 0; 1814 while (!isspace((unsigned char)*++l)) { 1815 switch (*l) { 1816 case '0': case '1': case '2': 1817 case '3': case '4': case '5': 1818 case '6': case '7': case '8': 1819 case '9': 1820 if (have_range && 1821 (ms->flags & MAGIC_CHECK)) 1822 file_magwarn(ms, 1823 "multiple ranges"); 1824 have_range = 1; 1825 m->str_range = CAST(uint32_t, 1826 strtoul(l, &t, 0)); 1827 if (m->str_range == 0) 1828 file_magwarn(ms, 1829 "zero range"); 1830 l = t - 1; 1831 break; 1832 case CHAR_COMPACT_WHITESPACE: 1833 m->str_flags |= 1834 STRING_COMPACT_WHITESPACE; 1835 break; 1836 case CHAR_COMPACT_OPTIONAL_WHITESPACE: 1837 m->str_flags |= 1838 STRING_COMPACT_OPTIONAL_WHITESPACE; 1839 break; 1840 case CHAR_IGNORE_LOWERCASE: 1841 m->str_flags |= STRING_IGNORE_LOWERCASE; 1842 break; 1843 case CHAR_IGNORE_UPPERCASE: 1844 m->str_flags |= STRING_IGNORE_UPPERCASE; 1845 break; 1846 case CHAR_REGEX_OFFSET_START: 1847 m->str_flags |= REGEX_OFFSET_START; 1848 break; 1849 case CHAR_BINTEST: 1850 m->str_flags |= STRING_BINTEST; 1851 break; 1852 case CHAR_TEXTTEST: 1853 m->str_flags |= STRING_TEXTTEST; 1854 break; 1855 case CHAR_TRIM: 1856 m->str_flags |= STRING_TRIM; 1857 break; 1858 case CHAR_PSTRING_1_LE: 1859 if (m->type != FILE_PSTRING) 1860 goto bad; 1861 m->str_flags = (m->str_flags & ~PSTRING_LEN) | PSTRING_1_LE; 1862 break; 1863 case CHAR_PSTRING_2_BE: 1864 if (m->type != FILE_PSTRING) 1865 goto bad; 1866 m->str_flags = (m->str_flags & ~PSTRING_LEN) | PSTRING_2_BE; 1867 break; 1868 case CHAR_PSTRING_2_LE: 1869 if (m->type != FILE_PSTRING) 1870 goto bad; 1871 m->str_flags = (m->str_flags & ~PSTRING_LEN) | PSTRING_2_LE; 1872 break; 1873 case CHAR_PSTRING_4_BE: 1874 if (m->type != FILE_PSTRING) 1875 goto bad; 1876 m->str_flags = (m->str_flags & ~PSTRING_LEN) | PSTRING_4_BE; 1877 break; 1878 case CHAR_PSTRING_4_LE: 1879 if (m->type != FILE_PSTRING) 1880 goto bad; 1881 m->str_flags = (m->str_flags & ~PSTRING_LEN) | PSTRING_4_LE; 1882 break; 1883 case CHAR_PSTRING_LENGTH_INCLUDES_ITSELF: 1884 if (m->type != FILE_PSTRING) 1885 goto bad; 1886 m->str_flags |= PSTRING_LENGTH_INCLUDES_ITSELF; 1887 break; 1888 default: 1889 bad: 1890 if (ms->flags & MAGIC_CHECK) 1891 file_magwarn(ms, 1892 "string extension `%c' " 1893 "invalid", *l); 1894 return -1; 1895 } 1896 /* allow multiple '/' for readability */ 1897 if (l[1] == '/' && 1898 !isspace((unsigned char)l[2])) 1899 l++; 1900 } 1901 if (string_modifier_check(ms, m) == -1) 1902 return -1; 1903 } 1904 else { 1905 if (ms->flags & MAGIC_CHECK) 1906 file_magwarn(ms, "invalid string op: %c", *t); 1907 return -1; 1908 } 1909 } 1910 /* 1911 * We used to set mask to all 1's here, instead let's just not do 1912 * anything if mask = 0 (unless you have a better idea) 1913 */ 1914 EATAB; 1915 1916 switch (*l) { 1917 case '>': 1918 case '<': 1919 m->reln = *l; 1920 ++l; 1921 if (*l == '=') { 1922 if (ms->flags & MAGIC_CHECK) { 1923 file_magwarn(ms, "%c= not supported", 1924 m->reln); 1925 return -1; 1926 } 1927 ++l; 1928 } 1929 break; 1930 /* Old-style anding: "0 byte &0x80 dynamically linked" */ 1931 case '&': 1932 case '^': 1933 case '=': 1934 m->reln = *l; 1935 ++l; 1936 if (*l == '=') { 1937 /* HP compat: ignore &= etc. */ 1938 ++l; 1939 } 1940 break; 1941 case '!': 1942 m->reln = *l; 1943 ++l; 1944 break; 1945 default: 1946 m->reln = '='; /* the default relation */ 1947 if (*l == 'x' && ((isascii((unsigned char)l[1]) && 1948 isspace((unsigned char)l[1])) || !l[1])) { 1949 m->reln = *l; 1950 ++l; 1951 } 1952 break; 1953 } 1954 /* 1955 * Grab the value part, except for an 'x' reln. 1956 */ 1957 if (m->reln != 'x' && getvalue(ms, m, &l, action)) 1958 return -1; 1959 1960 /* 1961 * TODO finish this macro and start using it! 1962 * #define offsetcheck {if (offset > HOWMANY-1) 1963 * magwarn("offset too big"); } 1964 */ 1965 1966 /* 1967 * Now get last part - the description 1968 */ 1969 EATAB; 1970 if (l[0] == '\b') { 1971 ++l; 1972 m->flag |= NOSPACE; 1973 } else if ((l[0] == '\\') && (l[1] == 'b')) { 1974 ++l; 1975 ++l; 1976 m->flag |= NOSPACE; 1977 } 1978 for (i = 0; (m->desc[i++] = *l++) != '\0' && i < sizeof(m->desc); ) 1979 continue; 1980 if (i == sizeof(m->desc)) { 1981 m->desc[sizeof(m->desc) - 1] = '\0'; 1982 if (ms->flags & MAGIC_CHECK) 1983 file_magwarn(ms, "description `%s' truncated", m->desc); 1984 } 1985 1986 /* 1987 * We only do this check while compiling, or if any of the magic 1988 * files were not compiled. 1989 */ 1990 if (ms->flags & MAGIC_CHECK) { 1991 if (check_format(ms, m) == -1) 1992 return -1; 1993 } 1994 #ifndef COMPILE_ONLY 1995 if (action == FILE_CHECK) { 1996 file_mdump(m); 1997 } 1998 #endif 1999 m->mimetype[0] = '\0'; /* initialise MIME type to none */ 2000 return 0; 2001 } 2002 2003 /* 2004 * parse a STRENGTH annotation line from magic file, put into magic[index - 1] 2005 * if valid 2006 */ 2007 private int 2008 parse_strength(struct magic_set *ms, struct magic_entry *me, const char *line) 2009 { 2010 const char *l = line; 2011 char *el; 2012 unsigned long factor; 2013 struct magic *m = &me->mp[0]; 2014 2015 if (m->factor_op != FILE_FACTOR_OP_NONE) { 2016 file_magwarn(ms, 2017 "Current entry already has a strength type: %c %d", 2018 m->factor_op, m->factor); 2019 return -1; 2020 } 2021 if (m->type == FILE_NAME) { 2022 file_magwarn(ms, "%s: Strength setting is not supported in " 2023 "\"name\" magic entries", m->value.s); 2024 return -1; 2025 } 2026 EATAB; 2027 switch (*l) { 2028 case FILE_FACTOR_OP_NONE: 2029 case FILE_FACTOR_OP_PLUS: 2030 case FILE_FACTOR_OP_MINUS: 2031 case FILE_FACTOR_OP_TIMES: 2032 case FILE_FACTOR_OP_DIV: 2033 m->factor_op = *l++; 2034 break; 2035 default: 2036 file_magwarn(ms, "Unknown factor op `%c'", *l); 2037 return -1; 2038 } 2039 EATAB; 2040 factor = strtoul(l, &el, 0); 2041 if (factor > 255) { 2042 file_magwarn(ms, "Too large factor `%lu'", factor); 2043 goto out; 2044 } 2045 if (*el && !isspace((unsigned char)*el)) { 2046 file_magwarn(ms, "Bad factor `%s'", l); 2047 goto out; 2048 } 2049 m->factor = (uint8_t)factor; 2050 if (m->factor == 0 && m->factor_op == FILE_FACTOR_OP_DIV) { 2051 file_magwarn(ms, "Cannot have factor op `%c' and factor %u", 2052 m->factor_op, m->factor); 2053 goto out; 2054 } 2055 return 0; 2056 out: 2057 m->factor_op = FILE_FACTOR_OP_NONE; 2058 m->factor = 0; 2059 return -1; 2060 } 2061 2062 private int 2063 parse_extra(struct magic_set *ms, struct magic_entry *me, const char *line, 2064 off_t off, size_t len, const char *name, int nt) 2065 { 2066 size_t i; 2067 const char *l = line; 2068 struct magic *m = &me->mp[me->cont_count == 0 ? 0 : me->cont_count - 1]; 2069 char *buf = (char *)m + off; 2070 2071 if (buf[0] != '\0') { 2072 len = nt ? strlen(buf) : len; 2073 file_magwarn(ms, "Current entry already has a %s type " 2074 "`%.*s', new type `%s'", name, (int)len, buf, l); 2075 return -1; 2076 } 2077 2078 if (*m->desc == '\0') { 2079 file_magwarn(ms, "Current entry does not yet have a " 2080 "description for adding a %s type", name); 2081 return -1; 2082 } 2083 2084 EATAB; 2085 for (i = 0; *l && ((isascii((unsigned char)*l) && 2086 isalnum((unsigned char)*l)) || strchr("-+/.", *l)) && 2087 i < len; buf[i++] = *l++) 2088 continue; 2089 2090 if (i == len && *l) { 2091 if (nt) 2092 buf[len - 1] = '\0'; 2093 if (ms->flags & MAGIC_CHECK) 2094 file_magwarn(ms, "%s type `%s' truncated %" 2095 SIZE_T_FORMAT "u", name, line, i); 2096 } else { 2097 if (nt) 2098 buf[i] = '\0'; 2099 } 2100 2101 if (i > 0) 2102 return 0; 2103 else 2104 return -1; 2105 } 2106 2107 /* 2108 * Parse an Apple CREATOR/TYPE annotation from magic file and put it into 2109 * magic[index - 1] 2110 */ 2111 private int 2112 parse_apple(struct magic_set *ms, struct magic_entry *me, const char *line) 2113 { 2114 struct magic *m = &me->mp[0]; 2115 2116 return parse_extra(ms, me, line, offsetof(struct magic, apple), 2117 sizeof(m->apple), "APPLE", 0); 2118 } 2119 2120 /* 2121 * parse a MIME annotation line from magic file, put into magic[index - 1] 2122 * if valid 2123 */ 2124 private int 2125 parse_mime(struct magic_set *ms, struct magic_entry *me, const char *line) 2126 { 2127 struct magic *m = &me->mp[0]; 2128 2129 return parse_extra(ms, me, line, offsetof(struct magic, mimetype), 2130 sizeof(m->mimetype), "MIME", 1); 2131 } 2132 2133 private int 2134 check_format_type(const char *ptr, int type) 2135 { 2136 int quad = 0; 2137 if (*ptr == '\0') { 2138 /* Missing format string; bad */ 2139 return -1; 2140 } 2141 2142 switch (type) { 2143 case FILE_FMT_QUAD: 2144 quad = 1; 2145 /*FALLTHROUGH*/ 2146 case FILE_FMT_NUM: 2147 if (*ptr == '-') 2148 ptr++; 2149 if (*ptr == '.') 2150 ptr++; 2151 while (isdigit((unsigned char)*ptr)) ptr++; 2152 if (*ptr == '.') 2153 ptr++; 2154 while (isdigit((unsigned char)*ptr)) ptr++; 2155 if (quad) { 2156 if (*ptr++ != 'l') 2157 return -1; 2158 if (*ptr++ != 'l') 2159 return -1; 2160 } 2161 2162 switch (*ptr++) { 2163 case 'l': 2164 switch (*ptr++) { 2165 case 'i': 2166 case 'd': 2167 case 'u': 2168 case 'o': 2169 case 'x': 2170 case 'X': 2171 return 0; 2172 default: 2173 return -1; 2174 } 2175 2176 case 'h': 2177 switch (*ptr++) { 2178 case 'h': 2179 switch (*ptr++) { 2180 case 'i': 2181 case 'd': 2182 case 'u': 2183 case 'o': 2184 case 'x': 2185 case 'X': 2186 return 0; 2187 default: 2188 return -1; 2189 } 2190 case 'd': 2191 return 0; 2192 default: 2193 return -1; 2194 } 2195 2196 case 'i': 2197 case 'c': 2198 case 'd': 2199 case 'u': 2200 case 'o': 2201 case 'x': 2202 case 'X': 2203 return 0; 2204 2205 default: 2206 return -1; 2207 } 2208 2209 case FILE_FMT_FLOAT: 2210 case FILE_FMT_DOUBLE: 2211 if (*ptr == '-') 2212 ptr++; 2213 if (*ptr == '.') 2214 ptr++; 2215 while (isdigit((unsigned char)*ptr)) ptr++; 2216 if (*ptr == '.') 2217 ptr++; 2218 while (isdigit((unsigned char)*ptr)) ptr++; 2219 2220 switch (*ptr++) { 2221 case 'e': 2222 case 'E': 2223 case 'f': 2224 case 'F': 2225 case 'g': 2226 case 'G': 2227 return 0; 2228 2229 default: 2230 return -1; 2231 } 2232 2233 2234 case FILE_FMT_STR: 2235 if (*ptr == '-') 2236 ptr++; 2237 while (isdigit((unsigned char )*ptr)) 2238 ptr++; 2239 if (*ptr == '.') { 2240 ptr++; 2241 while (isdigit((unsigned char )*ptr)) 2242 ptr++; 2243 } 2244 2245 switch (*ptr++) { 2246 case 's': 2247 return 0; 2248 default: 2249 return -1; 2250 } 2251 2252 default: 2253 /* internal error */ 2254 abort(); 2255 } 2256 /*NOTREACHED*/ 2257 return -1; 2258 } 2259 2260 /* 2261 * Check that the optional printf format in description matches 2262 * the type of the magic. 2263 */ 2264 private int 2265 check_format(struct magic_set *ms, struct magic *m) 2266 { 2267 char *ptr; 2268 2269 for (ptr = m->desc; *ptr; ptr++) 2270 if (*ptr == '%') 2271 break; 2272 if (*ptr == '\0') { 2273 /* No format string; ok */ 2274 return 1; 2275 } 2276 2277 assert(file_nformats == file_nnames); 2278 2279 if (m->type >= file_nformats) { 2280 file_magwarn(ms, "Internal error inconsistency between " 2281 "m->type and format strings"); 2282 return -1; 2283 } 2284 if (file_formats[m->type] == FILE_FMT_NONE) { 2285 file_magwarn(ms, "No format string for `%s' with description " 2286 "`%s'", m->desc, file_names[m->type]); 2287 return -1; 2288 } 2289 2290 ptr++; 2291 if (check_format_type(ptr, file_formats[m->type]) == -1) { 2292 /* 2293 * TODO: this error message is unhelpful if the format 2294 * string is not one character long 2295 */ 2296 file_magwarn(ms, "Printf format `%c' is not valid for type " 2297 "`%s' in description `%s'", *ptr ? *ptr : '?', 2298 file_names[m->type], m->desc); 2299 return -1; 2300 } 2301 2302 for (; *ptr; ptr++) { 2303 if (*ptr == '%') { 2304 file_magwarn(ms, 2305 "Too many format strings (should have at most one) " 2306 "for `%s' with description `%s'", 2307 file_names[m->type], m->desc); 2308 return -1; 2309 } 2310 } 2311 return 0; 2312 } 2313 2314 /* 2315 * Read a numeric value from a pointer, into the value union of a magic 2316 * pointer, according to the magic type. Update the string pointer to point 2317 * just after the number read. Return 0 for success, non-zero for failure. 2318 */ 2319 private int 2320 getvalue(struct magic_set *ms, struct magic *m, const char **p, int action) 2321 { 2322 switch (m->type) { 2323 case FILE_BESTRING16: 2324 case FILE_LESTRING16: 2325 case FILE_STRING: 2326 case FILE_PSTRING: 2327 case FILE_REGEX: 2328 case FILE_SEARCH: 2329 case FILE_NAME: 2330 case FILE_USE: 2331 *p = getstr(ms, m, *p, action == FILE_COMPILE); 2332 if (*p == NULL) { 2333 if (ms->flags & MAGIC_CHECK) 2334 file_magwarn(ms, "cannot get string from `%s'", 2335 m->value.s); 2336 return -1; 2337 } 2338 return 0; 2339 case FILE_FLOAT: 2340 case FILE_BEFLOAT: 2341 case FILE_LEFLOAT: 2342 if (m->reln != 'x') { 2343 char *ep; 2344 #ifdef HAVE_STRTOF 2345 m->value.f = strtof(*p, &ep); 2346 #else 2347 m->value.f = (float)strtod(*p, &ep); 2348 #endif 2349 *p = ep; 2350 } 2351 return 0; 2352 case FILE_DOUBLE: 2353 case FILE_BEDOUBLE: 2354 case FILE_LEDOUBLE: 2355 if (m->reln != 'x') { 2356 char *ep; 2357 m->value.d = strtod(*p, &ep); 2358 *p = ep; 2359 } 2360 return 0; 2361 default: 2362 if (m->reln != 'x') { 2363 char *ep; 2364 m->value.q = file_signextend(ms, m, 2365 (uint64_t)strtoull(*p, &ep, 0)); 2366 *p = ep; 2367 eatsize(p); 2368 } 2369 return 0; 2370 } 2371 } 2372 2373 /* 2374 * Convert a string containing C character escapes. Stop at an unescaped 2375 * space or tab. 2376 * Copy the converted version to "m->value.s", and the length in m->vallen. 2377 * Return updated scan pointer as function result. Warn if set. 2378 */ 2379 private const char * 2380 getstr(struct magic_set *ms, struct magic *m, const char *s, int warn) 2381 { 2382 const char *origs = s; 2383 char *p = m->value.s; 2384 size_t plen = sizeof(m->value.s); 2385 char *origp = p; 2386 char *pmax = p + plen - 1; 2387 int c; 2388 int val; 2389 2390 while ((c = *s++) != '\0') { 2391 if (isspace((unsigned char) c)) 2392 break; 2393 if (p >= pmax) { 2394 file_error(ms, 0, "string too long: `%s'", origs); 2395 return NULL; 2396 } 2397 if (c == '\\') { 2398 switch(c = *s++) { 2399 2400 case '\0': 2401 if (warn) 2402 file_magwarn(ms, "incomplete escape"); 2403 goto out; 2404 2405 case '\t': 2406 if (warn) { 2407 file_magwarn(ms, 2408 "escaped tab found, use \\t instead"); 2409 warn = 0; /* already did */ 2410 } 2411 /*FALLTHROUGH*/ 2412 default: 2413 if (warn) { 2414 if (isprint((unsigned char)c)) { 2415 /* Allow escaping of 2416 * ``relations'' */ 2417 if (strchr("<>&^=!", c) == NULL 2418 && (m->type != FILE_REGEX || 2419 strchr("[]().*?^$|{}", c) 2420 == NULL)) { 2421 file_magwarn(ms, "no " 2422 "need to escape " 2423 "`%c'", c); 2424 } 2425 } else { 2426 file_magwarn(ms, 2427 "unknown escape sequence: " 2428 "\\%03o", c); 2429 } 2430 } 2431 /*FALLTHROUGH*/ 2432 /* space, perhaps force people to use \040? */ 2433 case ' ': 2434 #if 0 2435 /* 2436 * Other things people escape, but shouldn't need to, 2437 * so we disallow them 2438 */ 2439 case '\'': 2440 case '"': 2441 case '?': 2442 #endif 2443 /* Relations */ 2444 case '>': 2445 case '<': 2446 case '&': 2447 case '^': 2448 case '=': 2449 case '!': 2450 /* and baskslash itself */ 2451 case '\\': 2452 *p++ = (char) c; 2453 break; 2454 2455 case 'a': 2456 *p++ = '\a'; 2457 break; 2458 2459 case 'b': 2460 *p++ = '\b'; 2461 break; 2462 2463 case 'f': 2464 *p++ = '\f'; 2465 break; 2466 2467 case 'n': 2468 *p++ = '\n'; 2469 break; 2470 2471 case 'r': 2472 *p++ = '\r'; 2473 break; 2474 2475 case 't': 2476 *p++ = '\t'; 2477 break; 2478 2479 case 'v': 2480 *p++ = '\v'; 2481 break; 2482 2483 /* \ and up to 3 octal digits */ 2484 case '0': 2485 case '1': 2486 case '2': 2487 case '3': 2488 case '4': 2489 case '5': 2490 case '6': 2491 case '7': 2492 val = c - '0'; 2493 c = *s++; /* try for 2 */ 2494 if (c >= '0' && c <= '7') { 2495 val = (val << 3) | (c - '0'); 2496 c = *s++; /* try for 3 */ 2497 if (c >= '0' && c <= '7') 2498 val = (val << 3) | (c-'0'); 2499 else 2500 --s; 2501 } 2502 else 2503 --s; 2504 *p++ = (char)val; 2505 break; 2506 2507 /* \x and up to 2 hex digits */ 2508 case 'x': 2509 val = 'x'; /* Default if no digits */ 2510 c = hextoint(*s++); /* Get next char */ 2511 if (c >= 0) { 2512 val = c; 2513 c = hextoint(*s++); 2514 if (c >= 0) 2515 val = (val << 4) + c; 2516 else 2517 --s; 2518 } else 2519 --s; 2520 *p++ = (char)val; 2521 break; 2522 } 2523 } else 2524 *p++ = (char)c; 2525 } 2526 out: 2527 *p = '\0'; 2528 m->vallen = CAST(unsigned char, (p - origp)); 2529 if (m->type == FILE_PSTRING) 2530 m->vallen += (unsigned char)file_pstring_length_size(m); 2531 return s; 2532 } 2533 2534 2535 /* Single hex char to int; -1 if not a hex char. */ 2536 private int 2537 hextoint(int c) 2538 { 2539 if (!isascii((unsigned char) c)) 2540 return -1; 2541 if (isdigit((unsigned char) c)) 2542 return c - '0'; 2543 if ((c >= 'a') && (c <= 'f')) 2544 return c + 10 - 'a'; 2545 if (( c>= 'A') && (c <= 'F')) 2546 return c + 10 - 'A'; 2547 return -1; 2548 } 2549 2550 2551 /* 2552 * Print a string containing C character escapes. 2553 */ 2554 protected void 2555 file_showstr(FILE *fp, const char *s, size_t len) 2556 { 2557 char c; 2558 2559 for (;;) { 2560 if (len == ~0U) { 2561 c = *s++; 2562 if (c == '\0') 2563 break; 2564 } 2565 else { 2566 if (len-- == 0) 2567 break; 2568 c = *s++; 2569 } 2570 if (c >= 040 && c <= 0176) /* TODO isprint && !iscntrl */ 2571 (void) fputc(c, fp); 2572 else { 2573 (void) fputc('\\', fp); 2574 switch (c) { 2575 case '\a': 2576 (void) fputc('a', fp); 2577 break; 2578 2579 case '\b': 2580 (void) fputc('b', fp); 2581 break; 2582 2583 case '\f': 2584 (void) fputc('f', fp); 2585 break; 2586 2587 case '\n': 2588 (void) fputc('n', fp); 2589 break; 2590 2591 case '\r': 2592 (void) fputc('r', fp); 2593 break; 2594 2595 case '\t': 2596 (void) fputc('t', fp); 2597 break; 2598 2599 case '\v': 2600 (void) fputc('v', fp); 2601 break; 2602 2603 default: 2604 (void) fprintf(fp, "%.3o", c & 0377); 2605 break; 2606 } 2607 } 2608 } 2609 } 2610 2611 /* 2612 * eatsize(): Eat the size spec from a number [eg. 10UL] 2613 */ 2614 private void 2615 eatsize(const char **p) 2616 { 2617 const char *l = *p; 2618 2619 if (LOWCASE(*l) == 'u') 2620 l++; 2621 2622 switch (LOWCASE(*l)) { 2623 case 'l': /* long */ 2624 case 's': /* short */ 2625 case 'h': /* short */ 2626 case 'b': /* char/byte */ 2627 case 'c': /* char/byte */ 2628 l++; 2629 /*FALLTHROUGH*/ 2630 default: 2631 break; 2632 } 2633 2634 *p = l; 2635 } 2636 2637 /* 2638 * handle a compiled file. 2639 */ 2640 2641 private struct magic_map * 2642 apprentice_map(struct magic_set *ms, const char *fn) 2643 { 2644 int fd; 2645 struct stat st; 2646 uint32_t *ptr; 2647 uint32_t version, entries, nentries; 2648 int needsbyteswap; 2649 char *dbname = NULL; 2650 struct magic_map *map; 2651 size_t i; 2652 2653 fd = -1; 2654 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 2655 file_oomem(ms, sizeof(*map)); 2656 goto error; 2657 } 2658 2659 dbname = mkdbname(ms, fn, 0); 2660 if (dbname == NULL) 2661 goto error; 2662 2663 if ((fd = open(dbname, O_RDONLY|O_BINARY)) == -1) 2664 goto error; 2665 2666 if (fstat(fd, &st) == -1) { 2667 file_error(ms, errno, "cannot stat `%s'", dbname); 2668 goto error; 2669 } 2670 if (st.st_size < 8 || st.st_size > MAXMAGIC_SIZE) { 2671 file_error(ms, 0, "file `%s' is too %s", dbname, 2672 st.st_size < 8 ? "small" : "large"); 2673 goto error; 2674 } 2675 2676 map->len = (size_t)st.st_size; 2677 #ifdef QUICK 2678 if ((map->p = mmap(0, (size_t)st.st_size, PROT_READ|PROT_WRITE, 2679 MAP_PRIVATE|MAP_FILE, fd, (off_t)0)) == MAP_FAILED) { 2680 file_error(ms, errno, "cannot map `%s'", dbname); 2681 goto error; 2682 } 2683 #else 2684 if ((map->p = CAST(void *, malloc(map->len))) == NULL) { 2685 file_oomem(ms, map->len); 2686 goto error; 2687 } 2688 if (read(fd, map->p, map->len) != (ssize_t)map->len) { 2689 file_badread(ms); 2690 goto error; 2691 } 2692 map->len = 0; 2693 #define RET 1 2694 #endif 2695 (void)close(fd); 2696 fd = -1; 2697 ptr = CAST(uint32_t *, map->p); 2698 if (*ptr != MAGICNO) { 2699 if (swap4(*ptr) != MAGICNO) { 2700 file_error(ms, 0, "bad magic in `%s'", dbname); 2701 goto error; 2702 } 2703 needsbyteswap = 1; 2704 } else 2705 needsbyteswap = 0; 2706 if (needsbyteswap) 2707 version = swap4(ptr[1]); 2708 else 2709 version = ptr[1]; 2710 if (version != VERSIONNO) { 2711 file_error(ms, 0, "File %s supports only version %d magic " 2712 "files. `%s' is version %d", VERSION, 2713 VERSIONNO, dbname, version); 2714 goto error; 2715 } 2716 entries = (uint32_t)(st.st_size / sizeof(struct magic)); 2717 if ((off_t)(entries * sizeof(struct magic)) != st.st_size) { 2718 file_error(ms, 0, "Size of `%s' %llu is not a multiple of %zu", 2719 dbname, (unsigned long long)st.st_size, 2720 sizeof(struct magic)); 2721 goto error; 2722 } 2723 map->magic[0] = CAST(struct magic *, map->p) + 1; 2724 nentries = 0; 2725 for (i = 0; i < MAGIC_SETS; i++) { 2726 if (needsbyteswap) 2727 map->nmagic[i] = swap4(ptr[i + 2]); 2728 else 2729 map->nmagic[i] = ptr[i + 2]; 2730 if (i != MAGIC_SETS - 1) 2731 map->magic[i + 1] = map->magic[i] + map->nmagic[i]; 2732 nentries += map->nmagic[i]; 2733 } 2734 if (entries != nentries + 1) { 2735 file_error(ms, 0, "Inconsistent entries in `%s' %u != %u", 2736 dbname, entries, nentries + 1); 2737 goto error; 2738 } 2739 if (needsbyteswap) 2740 for (i = 0; i < MAGIC_SETS; i++) 2741 byteswap(map->magic[i], map->nmagic[i]); 2742 free(dbname); 2743 return map; 2744 2745 error: 2746 if (fd != -1) 2747 (void)close(fd); 2748 apprentice_unmap(map); 2749 free(dbname); 2750 return NULL; 2751 } 2752 2753 private const uint32_t ar[] = { 2754 MAGICNO, VERSIONNO 2755 }; 2756 2757 /* 2758 * handle an mmaped file. 2759 */ 2760 private int 2761 apprentice_compile(struct magic_set *ms, struct magic_map *map, const char *fn) 2762 { 2763 static const size_t nm = sizeof(*map->nmagic) * MAGIC_SETS; 2764 static const size_t m = sizeof(**map->magic); 2765 int fd = -1; 2766 size_t len; 2767 char *dbname; 2768 int rv = -1; 2769 uint32_t i; 2770 2771 dbname = mkdbname(ms, fn, 1); 2772 2773 if (dbname == NULL) 2774 goto out; 2775 2776 if ((fd = open(dbname, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY, 0644)) == -1) 2777 { 2778 file_error(ms, errno, "cannot open `%s'", dbname); 2779 goto out; 2780 } 2781 2782 if (write(fd, ar, sizeof(ar)) != (ssize_t)sizeof(ar)) { 2783 file_error(ms, errno, "error writing `%s'", dbname); 2784 goto out; 2785 } 2786 2787 if (write(fd, map->nmagic, nm) != (ssize_t)nm) { 2788 file_error(ms, errno, "error writing `%s'", dbname); 2789 goto out; 2790 } 2791 2792 assert(nm + sizeof(ar) < m); 2793 2794 if (lseek(fd, (off_t)m, SEEK_SET) != (off_t)m) { 2795 file_error(ms, errno, "error seeking `%s'", dbname); 2796 goto out; 2797 } 2798 2799 for (i = 0; i < MAGIC_SETS; i++) { 2800 len = m * map->nmagic[i]; 2801 if (write(fd, map->magic[i], len) != (ssize_t)len) { 2802 file_error(ms, errno, "error writing `%s'", dbname); 2803 goto out; 2804 } 2805 } 2806 2807 if (fd != -1) 2808 (void)close(fd); 2809 rv = 0; 2810 out: 2811 free(dbname); 2812 return rv; 2813 } 2814 2815 private const char ext[] = ".mgc"; 2816 /* 2817 * make a dbname 2818 */ 2819 private char * 2820 mkdbname(struct magic_set *ms, const char *fn, int strip) 2821 { 2822 const char *p, *q; 2823 char *buf; 2824 2825 if (strip) { 2826 if ((p = strrchr(fn, '/')) != NULL) 2827 fn = ++p; 2828 } 2829 2830 for (q = fn; *q; q++) 2831 continue; 2832 /* Look for .mgc */ 2833 for (p = ext + sizeof(ext) - 1; p >= ext && q >= fn; p--, q--) 2834 if (*p != *q) 2835 break; 2836 2837 /* Did not find .mgc, restore q */ 2838 if (p >= ext) 2839 while (*q) 2840 q++; 2841 2842 q++; 2843 /* Compatibility with old code that looked in .mime */ 2844 if (ms->flags & MAGIC_MIME) { 2845 if (asprintf(&buf, "%.*s.mime%s", (int)(q - fn), fn, ext) < 0) 2846 return NULL; 2847 if (access(buf, R_OK) != -1) { 2848 ms->flags &= MAGIC_MIME_TYPE; 2849 return buf; 2850 } 2851 free(buf); 2852 } 2853 if (asprintf(&buf, "%.*s%s", (int)(q - fn), fn, ext) < 0) 2854 return NULL; 2855 2856 /* Compatibility with old code that looked in .mime */ 2857 if (strstr(p, ".mime") != NULL) 2858 ms->flags &= MAGIC_MIME_TYPE; 2859 return buf; 2860 } 2861 2862 /* 2863 * Byteswap an mmap'ed file if needed 2864 */ 2865 private void 2866 byteswap(struct magic *magic, uint32_t nmagic) 2867 { 2868 uint32_t i; 2869 for (i = 0; i < nmagic; i++) 2870 bs1(&magic[i]); 2871 } 2872 2873 /* 2874 * swap a short 2875 */ 2876 private uint16_t 2877 swap2(uint16_t sv) 2878 { 2879 uint16_t rv; 2880 uint8_t *s = (uint8_t *)(void *)&sv; 2881 uint8_t *d = (uint8_t *)(void *)&rv; 2882 d[0] = s[1]; 2883 d[1] = s[0]; 2884 return rv; 2885 } 2886 2887 /* 2888 * swap an int 2889 */ 2890 private uint32_t 2891 swap4(uint32_t sv) 2892 { 2893 uint32_t rv; 2894 uint8_t *s = (uint8_t *)(void *)&sv; 2895 uint8_t *d = (uint8_t *)(void *)&rv; 2896 d[0] = s[3]; 2897 d[1] = s[2]; 2898 d[2] = s[1]; 2899 d[3] = s[0]; 2900 return rv; 2901 } 2902 2903 /* 2904 * swap a quad 2905 */ 2906 private uint64_t 2907 swap8(uint64_t sv) 2908 { 2909 uint64_t rv; 2910 uint8_t *s = (uint8_t *)(void *)&sv; 2911 uint8_t *d = (uint8_t *)(void *)&rv; 2912 #if 0 2913 d[0] = s[3]; 2914 d[1] = s[2]; 2915 d[2] = s[1]; 2916 d[3] = s[0]; 2917 d[4] = s[7]; 2918 d[5] = s[6]; 2919 d[6] = s[5]; 2920 d[7] = s[4]; 2921 #else 2922 d[0] = s[7]; 2923 d[1] = s[6]; 2924 d[2] = s[5]; 2925 d[3] = s[4]; 2926 d[4] = s[3]; 2927 d[5] = s[2]; 2928 d[6] = s[1]; 2929 d[7] = s[0]; 2930 #endif 2931 return rv; 2932 } 2933 2934 /* 2935 * byteswap a single magic entry 2936 */ 2937 private void 2938 bs1(struct magic *m) 2939 { 2940 m->cont_level = swap2(m->cont_level); 2941 m->offset = swap4((uint32_t)m->offset); 2942 m->in_offset = swap4((uint32_t)m->in_offset); 2943 m->lineno = swap4((uint32_t)m->lineno); 2944 if (IS_STRING(m->type)) { 2945 m->str_range = swap4(m->str_range); 2946 m->str_flags = swap4(m->str_flags); 2947 } 2948 else { 2949 m->value.q = swap8(m->value.q); 2950 m->num_mask = swap8(m->num_mask); 2951 } 2952 } 2953 2954 protected size_t 2955 file_pstring_length_size(const struct magic *m) 2956 { 2957 switch (m->str_flags & PSTRING_LEN) { 2958 case PSTRING_1_LE: 2959 return 1; 2960 case PSTRING_2_LE: 2961 case PSTRING_2_BE: 2962 return 2; 2963 case PSTRING_4_LE: 2964 case PSTRING_4_BE: 2965 return 4; 2966 default: 2967 abort(); /* Impossible */ 2968 return 1; 2969 } 2970 } 2971 protected size_t 2972 file_pstring_get_length(const struct magic *m, const char *s) 2973 { 2974 size_t len = 0; 2975 2976 switch (m->str_flags & PSTRING_LEN) { 2977 case PSTRING_1_LE: 2978 len = *s; 2979 break; 2980 case PSTRING_2_LE: 2981 len = (s[1] << 8) | s[0]; 2982 break; 2983 case PSTRING_2_BE: 2984 len = (s[0] << 8) | s[1]; 2985 break; 2986 case PSTRING_4_LE: 2987 len = (s[3] << 24) | (s[2] << 16) | (s[1] << 8) | s[0]; 2988 break; 2989 case PSTRING_4_BE: 2990 len = (s[0] << 24) | (s[1] << 16) | (s[2] << 8) | s[3]; 2991 break; 2992 default: 2993 abort(); /* Impossible */ 2994 } 2995 2996 if (m->str_flags & PSTRING_LENGTH_INCLUDES_ITSELF) 2997 len -= file_pstring_length_size(m); 2998 2999 return len; 3000 } 3001 3002 protected int 3003 file_magicfind(struct magic_set *ms, const char *name, struct mlist *v) 3004 { 3005 uint32_t i, j; 3006 struct mlist *mlist, *ml; 3007 3008 mlist = ms->mlist[1]; 3009 3010 for (ml = mlist->next; ml != mlist; ml = ml->next) { 3011 struct magic *ma = ml->magic; 3012 uint32_t nma = ml->nmagic; 3013 for (i = 0; i < nma; i++) { 3014 if (ma[i].type != FILE_NAME) 3015 continue; 3016 if (strcmp(ma[i].value.s, name) == 0) { 3017 v->magic = &ma[i]; 3018 for (j = i + 1; j < nma; j++) 3019 if (ma[j].cont_level == 0) 3020 break; 3021 v->nmagic = j - i; 3022 return 0; 3023 } 3024 } 3025 } 3026 return -1; 3027 } 3028