1 /* $NetBSD: apprentice.c,v 1.10 2013/12/01 19:32:15 christos Exp $ */ 2 3 /* 4 * Copyright (c) Ian F. Darwin 1986-1995. 5 * Software written by Ian F. Darwin and others; 6 * maintained 1995-present by Christos Zoulas and others. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice immediately at the beginning of the file, without modification, 13 * this list of conditions, and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 21 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR 22 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 28 * SUCH DAMAGE. 29 */ 30 /* 31 * apprentice - make one pass through /etc/magic, learning its secrets. 32 */ 33 34 #include "file.h" 35 36 #ifndef lint 37 #if 0 38 FILE_RCSID("@(#)$File: apprentice.c,v 1.196 2013/11/19 21:01:12 christos Exp $") 39 #else 40 __RCSID("$NetBSD: apprentice.c,v 1.10 2013/12/01 19:32:15 christos Exp $"); 41 #endif 42 #endif /* lint */ 43 44 #include "magic.h" 45 #include <stdlib.h> 46 #ifdef HAVE_UNISTD_H 47 #include <unistd.h> 48 #endif 49 #include <string.h> 50 #include <assert.h> 51 #include <ctype.h> 52 #include <fcntl.h> 53 #ifdef QUICK 54 #include <sys/mman.h> 55 #endif 56 #include <dirent.h> 57 58 #define EATAB {while (isascii((unsigned char) *l) && \ 59 isspace((unsigned char) *l)) ++l;} 60 #define LOWCASE(l) (isupper((unsigned char) (l)) ? \ 61 tolower((unsigned char) (l)) : (l)) 62 /* 63 * Work around a bug in headers on Digital Unix. 64 * At least confirmed for: OSF1 V4.0 878 65 */ 66 #if defined(__osf__) && defined(__DECC) 67 #ifdef MAP_FAILED 68 #undef MAP_FAILED 69 #endif 70 #endif 71 72 #ifndef MAP_FAILED 73 #define MAP_FAILED (void *) -1 74 #endif 75 76 #ifndef MAP_FILE 77 #define MAP_FILE 0 78 #endif 79 80 #define ALLOC_CHUNK (size_t)10 81 #define ALLOC_INCR (size_t)200 82 83 struct magic_entry { 84 struct magic *mp; 85 uint32_t cont_count; 86 uint32_t max_count; 87 }; 88 89 struct magic_entry_set { 90 struct magic_entry *me; 91 uint32_t count; 92 uint32_t max; 93 }; 94 95 struct magic_map { 96 void *p; 97 size_t len; 98 struct magic *magic[MAGIC_SETS]; 99 uint32_t nmagic[MAGIC_SETS]; 100 }; 101 102 int file_formats[FILE_NAMES_SIZE]; 103 const size_t file_nformats = FILE_NAMES_SIZE; 104 const char *file_names[FILE_NAMES_SIZE]; 105 const size_t file_nnames = FILE_NAMES_SIZE; 106 107 private int getvalue(struct magic_set *ms, struct magic *, const char **, int); 108 private int hextoint(int); 109 private const char *getstr(struct magic_set *, struct magic *, const char *, 110 int); 111 private int parse(struct magic_set *, struct magic_entry *, const char *, 112 size_t, int); 113 private void eatsize(const char **); 114 private int apprentice_1(struct magic_set *, const char *, int); 115 private size_t apprentice_magic_strength(const struct magic *); 116 private int apprentice_sort(const void *, const void *); 117 private void apprentice_list(struct mlist *, int ); 118 private struct magic_map *apprentice_load(struct magic_set *, 119 const char *, int); 120 private struct mlist *mlist_alloc(void); 121 private void mlist_free(struct mlist *); 122 private void byteswap(struct magic *, uint32_t); 123 private void bs1(struct magic *); 124 private uint16_t swap2(uint16_t); 125 private uint32_t swap4(uint32_t); 126 private uint64_t swap8(uint64_t); 127 private char *mkdbname(struct magic_set *, const char *, int); 128 private struct magic_map *apprentice_map(struct magic_set *, const char *); 129 private void apprentice_unmap(struct magic_map *); 130 private int apprentice_compile(struct magic_set *, struct magic_map *, 131 const char *); 132 private int check_format_type(const char *, int); 133 private int check_format(struct magic_set *, struct magic *); 134 private int get_op(char); 135 private int parse_mime(struct magic_set *, struct magic_entry *, const char *); 136 private int parse_strength(struct magic_set *, struct magic_entry *, const char *); 137 private int parse_apple(struct magic_set *, struct magic_entry *, const char *); 138 139 140 private size_t magicsize = sizeof(struct magic); 141 142 private const char usg_hdr[] = "cont\toffset\ttype\topcode\tmask\tvalue\tdesc"; 143 144 private struct { 145 const char *name; 146 size_t len; 147 int (*fun)(struct magic_set *, struct magic_entry *, const char *); 148 } bang[] = { 149 #define DECLARE_FIELD(name) { # name, sizeof(# name) - 1, parse_ ## name } 150 DECLARE_FIELD(mime), 151 DECLARE_FIELD(apple), 152 DECLARE_FIELD(strength), 153 #undef DECLARE_FIELD 154 { NULL, 0, NULL } 155 }; 156 157 #ifdef COMPILE_ONLY 158 159 int main(int, char *[]); 160 161 int 162 main(int argc, char *argv[]) 163 { 164 int ret; 165 struct magic_set *ms; 166 char *progname; 167 168 if ((progname = strrchr(argv[0], '/')) != NULL) 169 progname++; 170 else 171 progname = argv[0]; 172 173 if (argc != 2) { 174 (void)fprintf(stderr, "Usage: %s file\n", progname); 175 return 1; 176 } 177 178 if ((ms = magic_open(MAGIC_CHECK)) == NULL) { 179 (void)fprintf(stderr, "%s: %s\n", progname, strerror(errno)); 180 return 1; 181 } 182 ret = magic_compile(ms, argv[1]) == -1 ? 1 : 0; 183 if (ret == 1) 184 (void)fprintf(stderr, "%s: %s\n", progname, magic_error(ms)); 185 magic_close(ms); 186 return ret; 187 } 188 #endif /* COMPILE_ONLY */ 189 190 struct type_tbl_s { 191 const char name[16]; 192 const size_t len; 193 const int type; 194 const int format; 195 }; 196 197 /* 198 * XXX - the actual Single UNIX Specification says that "long" means "long", 199 * as in the C data type, but we treat it as meaning "4-byte integer". 200 * Given that the OS X version of file 5.04 did the same, I guess that passes 201 * the actual test; having "long" be dependent on how big a "long" is on 202 * the machine running "file" is silly. 203 */ 204 static const struct type_tbl_s type_tbl[] = { 205 # define XX(s) s, (sizeof(s) - 1) 206 # define XX_NULL "", 0 207 { XX("invalid"), FILE_INVALID, FILE_FMT_NONE }, 208 { XX("byte"), FILE_BYTE, FILE_FMT_NUM }, 209 { XX("short"), FILE_SHORT, FILE_FMT_NUM }, 210 { XX("default"), FILE_DEFAULT, FILE_FMT_NONE }, 211 { XX("long"), FILE_LONG, FILE_FMT_NUM }, 212 { XX("string"), FILE_STRING, FILE_FMT_STR }, 213 { XX("date"), FILE_DATE, FILE_FMT_STR }, 214 { XX("beshort"), FILE_BESHORT, FILE_FMT_NUM }, 215 { XX("belong"), FILE_BELONG, FILE_FMT_NUM }, 216 { XX("bedate"), FILE_BEDATE, FILE_FMT_STR }, 217 { XX("leshort"), FILE_LESHORT, FILE_FMT_NUM }, 218 { XX("lelong"), FILE_LELONG, FILE_FMT_NUM }, 219 { XX("ledate"), FILE_LEDATE, FILE_FMT_STR }, 220 { XX("pstring"), FILE_PSTRING, FILE_FMT_STR }, 221 { XX("ldate"), FILE_LDATE, FILE_FMT_STR }, 222 { XX("beldate"), FILE_BELDATE, FILE_FMT_STR }, 223 { XX("leldate"), FILE_LELDATE, FILE_FMT_STR }, 224 { XX("regex"), FILE_REGEX, FILE_FMT_STR }, 225 { XX("bestring16"), FILE_BESTRING16, FILE_FMT_STR }, 226 { XX("lestring16"), FILE_LESTRING16, FILE_FMT_STR }, 227 { XX("search"), FILE_SEARCH, FILE_FMT_STR }, 228 { XX("medate"), FILE_MEDATE, FILE_FMT_STR }, 229 { XX("meldate"), FILE_MELDATE, FILE_FMT_STR }, 230 { XX("melong"), FILE_MELONG, FILE_FMT_NUM }, 231 { XX("quad"), FILE_QUAD, FILE_FMT_QUAD }, 232 { XX("lequad"), FILE_LEQUAD, FILE_FMT_QUAD }, 233 { XX("bequad"), FILE_BEQUAD, FILE_FMT_QUAD }, 234 { XX("qdate"), FILE_QDATE, FILE_FMT_STR }, 235 { XX("leqdate"), FILE_LEQDATE, FILE_FMT_STR }, 236 { XX("beqdate"), FILE_BEQDATE, FILE_FMT_STR }, 237 { XX("qldate"), FILE_QLDATE, FILE_FMT_STR }, 238 { XX("leqldate"), FILE_LEQLDATE, FILE_FMT_STR }, 239 { XX("beqldate"), FILE_BEQLDATE, FILE_FMT_STR }, 240 { XX("float"), FILE_FLOAT, FILE_FMT_FLOAT }, 241 { XX("befloat"), FILE_BEFLOAT, FILE_FMT_FLOAT }, 242 { XX("lefloat"), FILE_LEFLOAT, FILE_FMT_FLOAT }, 243 { XX("double"), FILE_DOUBLE, FILE_FMT_DOUBLE }, 244 { XX("bedouble"), FILE_BEDOUBLE, FILE_FMT_DOUBLE }, 245 { XX("ledouble"), FILE_LEDOUBLE, FILE_FMT_DOUBLE }, 246 { XX("leid3"), FILE_LEID3, FILE_FMT_NUM }, 247 { XX("beid3"), FILE_BEID3, FILE_FMT_NUM }, 248 { XX("indirect"), FILE_INDIRECT, FILE_FMT_NUM }, 249 { XX("qwdate"), FILE_QWDATE, FILE_FMT_STR }, 250 { XX("leqwdate"), FILE_LEQWDATE, FILE_FMT_STR }, 251 { XX("beqwdate"), FILE_BEQWDATE, FILE_FMT_STR }, 252 { XX("name"), FILE_NAME, FILE_FMT_NONE }, 253 { XX("use"), FILE_USE, FILE_FMT_NONE }, 254 { XX("clear"), FILE_CLEAR, FILE_FMT_NONE }, 255 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 256 }; 257 258 /* 259 * These are not types, and cannot be preceded by "u" to make them 260 * unsigned. 261 */ 262 static const struct type_tbl_s special_tbl[] = { 263 { XX("name"), FILE_NAME, FILE_FMT_STR }, 264 { XX("use"), FILE_USE, FILE_FMT_STR }, 265 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 266 }; 267 # undef XX 268 # undef XX_NULL 269 270 private int 271 get_type(const struct type_tbl_s *tbl, const char *l, const char **t) 272 { 273 const struct type_tbl_s *p; 274 275 for (p = tbl; p->len; p++) { 276 if (strncmp(l, p->name, p->len) == 0) { 277 if (t) 278 *t = l + p->len; 279 break; 280 } 281 } 282 return p->type; 283 } 284 285 private int 286 get_standard_integer_type(const char *l, const char **t) 287 { 288 int type; 289 290 if (isalpha((unsigned char)l[1])) { 291 switch (l[1]) { 292 case 'C': 293 /* "dC" and "uC" */ 294 type = FILE_BYTE; 295 break; 296 case 'S': 297 /* "dS" and "uS" */ 298 type = FILE_SHORT; 299 break; 300 case 'I': 301 case 'L': 302 /* 303 * "dI", "dL", "uI", and "uL". 304 * 305 * XXX - the actual Single UNIX Specification says 306 * that "L" means "long", as in the C data type, 307 * but we treat it as meaning "4-byte integer". 308 * Given that the OS X version of file 5.04 did 309 * the same, I guess that passes the actual SUS 310 * validation suite; having "dL" be dependent on 311 * how big a "long" is on the machine running 312 * "file" is silly. 313 */ 314 type = FILE_LONG; 315 break; 316 case 'Q': 317 /* "dQ" and "uQ" */ 318 type = FILE_QUAD; 319 break; 320 default: 321 /* "d{anything else}", "u{anything else}" */ 322 return FILE_INVALID; 323 } 324 l += 2; 325 } else if (isdigit((unsigned char)l[1])) { 326 /* 327 * "d{num}" and "u{num}"; we only support {num} values 328 * of 1, 2, 4, and 8 - the Single UNIX Specification 329 * doesn't say anything about whether arbitrary 330 * values should be supported, but both the Solaris 10 331 * and OS X Mountain Lion versions of file passed the 332 * Single UNIX Specification validation suite, and 333 * neither of them support values bigger than 8 or 334 * non-power-of-2 values. 335 */ 336 if (isdigit((unsigned char)l[2])) { 337 /* Multi-digit, so > 9 */ 338 return FILE_INVALID; 339 } 340 switch (l[1]) { 341 case '1': 342 type = FILE_BYTE; 343 break; 344 case '2': 345 type = FILE_SHORT; 346 break; 347 case '4': 348 type = FILE_LONG; 349 break; 350 case '8': 351 type = FILE_QUAD; 352 break; 353 default: 354 /* XXX - what about 3, 5, 6, or 7? */ 355 return FILE_INVALID; 356 } 357 l += 2; 358 } else { 359 /* 360 * "d" or "u" by itself. 361 */ 362 type = FILE_LONG; 363 ++l; 364 } 365 if (t) 366 *t = l; 367 return type; 368 } 369 370 private void 371 init_file_tables(void) 372 { 373 static int done = 0; 374 const struct type_tbl_s *p; 375 376 if (done) 377 return; 378 done++; 379 380 for (p = type_tbl; p->len; p++) { 381 assert(p->type < FILE_NAMES_SIZE); 382 file_names[p->type] = p->name; 383 file_formats[p->type] = p->format; 384 } 385 assert(p - type_tbl == FILE_NAMES_SIZE); 386 } 387 388 private int 389 add_mlist(struct mlist *mlp, struct magic_map *map, size_t idx) 390 { 391 struct mlist *ml; 392 393 if ((ml = CAST(struct mlist *, malloc(sizeof(*ml)))) == NULL) 394 return -1; 395 396 ml->map = idx == 0 ? map : NULL; 397 ml->magic = map->magic[idx]; 398 ml->nmagic = map->nmagic[idx]; 399 400 mlp->prev->next = ml; 401 ml->prev = mlp->prev; 402 ml->next = mlp; 403 mlp->prev = ml; 404 return 0; 405 } 406 407 /* 408 * Handle one file or directory. 409 */ 410 private int 411 apprentice_1(struct magic_set *ms, const char *fn, int action) 412 { 413 struct mlist *ml; 414 struct magic_map *map; 415 size_t i; 416 417 if (magicsize != FILE_MAGICSIZE) { 418 file_error(ms, 0, "magic element size %lu != %lu", 419 (unsigned long)sizeof(*map->magic[0]), 420 (unsigned long)FILE_MAGICSIZE); 421 return -1; 422 } 423 424 if (action == FILE_COMPILE) { 425 map = apprentice_load(ms, fn, action); 426 if (map == NULL) 427 return -1; 428 return apprentice_compile(ms, map, fn); 429 } 430 431 #ifndef COMPILE_ONLY 432 map = apprentice_map(ms, fn); 433 if (map == NULL) { 434 if (ms->flags & MAGIC_CHECK) 435 file_magwarn(ms, "using regular magic file `%s'", fn); 436 map = apprentice_load(ms, fn, action); 437 if (map == NULL) 438 return -1; 439 } 440 441 for (i = 0; i < MAGIC_SETS; i++) { 442 if (add_mlist(ms->mlist[i], map, i) == -1) { 443 file_oomem(ms, sizeof(*ml)); 444 apprentice_unmap(map); 445 return -1; 446 } 447 } 448 449 if (action == FILE_LIST) { 450 for (i = 0; i < MAGIC_SETS; i++) { 451 printf("Set %zu:\nBinary patterns:\n", i); 452 apprentice_list(ms->mlist[i], BINTEST); 453 printf("Text patterns:\n"); 454 apprentice_list(ms->mlist[i], TEXTTEST); 455 } 456 } 457 458 return 0; 459 #endif /* COMPILE_ONLY */ 460 } 461 462 protected void 463 file_ms_free(struct magic_set *ms) 464 { 465 size_t i; 466 if (ms == NULL) 467 return; 468 for (i = 0; i < MAGIC_SETS; i++) 469 mlist_free(ms->mlist[i]); 470 free(ms->o.pbuf); 471 free(ms->o.buf); 472 free(ms->c.li); 473 free(ms); 474 } 475 476 protected struct magic_set * 477 file_ms_alloc(int flags) 478 { 479 struct magic_set *ms; 480 size_t i, len; 481 482 if ((ms = CAST(struct magic_set *, calloc((size_t)1, 483 sizeof(struct magic_set)))) == NULL) 484 return NULL; 485 486 if (magic_setflags(ms, flags) == -1) { 487 errno = EINVAL; 488 goto free; 489 } 490 491 ms->o.buf = ms->o.pbuf = NULL; 492 len = (ms->c.len = 10) * sizeof(*ms->c.li); 493 494 if ((ms->c.li = CAST(struct level_info *, malloc(len))) == NULL) 495 goto free; 496 497 ms->event_flags = 0; 498 ms->error = -1; 499 for (i = 0; i < MAGIC_SETS; i++) 500 ms->mlist[i] = NULL; 501 ms->file = "unknown"; 502 ms->line = 0; 503 return ms; 504 free: 505 free(ms); 506 return NULL; 507 } 508 509 private void 510 apprentice_unmap(struct magic_map *map) 511 { 512 if (map == NULL) 513 return; 514 if (map->p == NULL) 515 return; 516 #ifdef QUICK 517 if (map->len) 518 (void)munmap(map->p, map->len); 519 else 520 #endif 521 free(map->p); 522 free(map); 523 } 524 525 private struct mlist * 526 mlist_alloc(void) 527 { 528 struct mlist *mlist; 529 if ((mlist = CAST(struct mlist *, calloc(1, sizeof(*mlist)))) == NULL) { 530 return NULL; 531 } 532 mlist->next = mlist->prev = mlist; 533 return mlist; 534 } 535 536 private void 537 mlist_free(struct mlist *mlist) 538 { 539 struct mlist *ml; 540 541 if (mlist == NULL) 542 return; 543 544 for (ml = mlist->next; ml != mlist;) { 545 struct mlist *next = ml->next; 546 if (ml->map) 547 apprentice_unmap(ml->map); 548 free(ml); 549 ml = next; 550 } 551 free(ml); 552 } 553 554 /* const char *fn: list of magic files and directories */ 555 protected int 556 file_apprentice(struct magic_set *ms, const char *fn, int action) 557 { 558 char *p, *mfn; 559 int file_err, errs = -1; 560 size_t i; 561 562 file_reset(ms); 563 564 if ((fn = magic_getpath(fn, action)) == NULL) 565 return -1; 566 567 init_file_tables(); 568 569 if ((mfn = strdup(fn)) == NULL) { 570 file_oomem(ms, strlen(fn)); 571 return -1; 572 } 573 574 for (i = 0; i < MAGIC_SETS; i++) { 575 mlist_free(ms->mlist[i]); 576 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 577 file_oomem(ms, sizeof(*ms->mlist[i])); 578 if (i != 0) { 579 --i; 580 do 581 mlist_free(ms->mlist[i]); 582 while (i != 0); 583 } 584 free(mfn); 585 return -1; 586 } 587 } 588 fn = mfn; 589 590 while (fn) { 591 p = strchr(fn, PATHSEP); 592 if (p) 593 *p++ = '\0'; 594 if (*fn == '\0') 595 break; 596 file_err = apprentice_1(ms, fn, action); 597 errs = MAX(errs, file_err); 598 fn = p; 599 } 600 601 free(mfn); 602 603 if (errs == -1) { 604 for (i = 0; i < MAGIC_SETS; i++) { 605 mlist_free(ms->mlist[i]); 606 ms->mlist[i] = NULL; 607 } 608 file_error(ms, 0, "could not find any valid magic files!"); 609 return -1; 610 } 611 612 #if 0 613 /* 614 * Always leave the database loaded 615 */ 616 if (action == FILE_LOAD) 617 return 0; 618 619 for (i = 0; i < MAGIC_SETS; i++) { 620 mlist_free(ms->mlist[i]); 621 ms->mlist[i] = NULL; 622 } 623 #endif 624 625 switch (action) { 626 case FILE_LOAD: 627 case FILE_COMPILE: 628 case FILE_CHECK: 629 case FILE_LIST: 630 return 0; 631 default: 632 file_error(ms, 0, "Invalid action %d", action); 633 return -1; 634 } 635 } 636 637 /* 638 * Get weight of this magic entry, for sorting purposes. 639 */ 640 private size_t 641 apprentice_magic_strength(const struct magic *m) 642 { 643 #define MULT 10 644 size_t val = 2 * MULT; /* baseline strength */ 645 646 switch (m->type) { 647 case FILE_DEFAULT: /* make sure this sorts last */ 648 if (m->factor_op != FILE_FACTOR_OP_NONE) 649 abort(); 650 return 0; 651 652 case FILE_BYTE: 653 val += 1 * MULT; 654 break; 655 656 case FILE_SHORT: 657 case FILE_LESHORT: 658 case FILE_BESHORT: 659 val += 2 * MULT; 660 break; 661 662 case FILE_LONG: 663 case FILE_LELONG: 664 case FILE_BELONG: 665 case FILE_MELONG: 666 val += 4 * MULT; 667 break; 668 669 case FILE_PSTRING: 670 case FILE_STRING: 671 val += m->vallen * MULT; 672 break; 673 674 case FILE_BESTRING16: 675 case FILE_LESTRING16: 676 val += m->vallen * MULT / 2; 677 break; 678 679 case FILE_SEARCH: 680 case FILE_REGEX: 681 val += m->vallen * MAX(MULT / m->vallen, 1); 682 break; 683 684 case FILE_DATE: 685 case FILE_LEDATE: 686 case FILE_BEDATE: 687 case FILE_MEDATE: 688 case FILE_LDATE: 689 case FILE_LELDATE: 690 case FILE_BELDATE: 691 case FILE_MELDATE: 692 case FILE_FLOAT: 693 case FILE_BEFLOAT: 694 case FILE_LEFLOAT: 695 val += 4 * MULT; 696 break; 697 698 case FILE_QUAD: 699 case FILE_BEQUAD: 700 case FILE_LEQUAD: 701 case FILE_QDATE: 702 case FILE_LEQDATE: 703 case FILE_BEQDATE: 704 case FILE_QLDATE: 705 case FILE_LEQLDATE: 706 case FILE_BEQLDATE: 707 case FILE_QWDATE: 708 case FILE_LEQWDATE: 709 case FILE_BEQWDATE: 710 case FILE_DOUBLE: 711 case FILE_BEDOUBLE: 712 case FILE_LEDOUBLE: 713 val += 8 * MULT; 714 break; 715 716 case FILE_INDIRECT: 717 case FILE_NAME: 718 case FILE_USE: 719 break; 720 721 default: 722 val = 0; 723 (void)fprintf(stderr, "Bad type %d\n", m->type); 724 abort(); 725 } 726 727 switch (m->reln) { 728 case 'x': /* matches anything penalize */ 729 case '!': /* matches almost anything penalize */ 730 val = 0; 731 break; 732 733 case '=': /* Exact match, prefer */ 734 val += MULT; 735 break; 736 737 case '>': 738 case '<': /* comparison match reduce strength */ 739 val -= 2 * MULT; 740 break; 741 742 case '^': 743 case '&': /* masking bits, we could count them too */ 744 val -= MULT; 745 break; 746 747 default: 748 (void)fprintf(stderr, "Bad relation %c\n", m->reln); 749 abort(); 750 } 751 752 if (val == 0) /* ensure we only return 0 for FILE_DEFAULT */ 753 val = 1; 754 755 switch (m->factor_op) { 756 case FILE_FACTOR_OP_NONE: 757 break; 758 case FILE_FACTOR_OP_PLUS: 759 val += m->factor; 760 break; 761 case FILE_FACTOR_OP_MINUS: 762 val -= m->factor; 763 break; 764 case FILE_FACTOR_OP_TIMES: 765 val *= m->factor; 766 break; 767 case FILE_FACTOR_OP_DIV: 768 val /= m->factor; 769 break; 770 default: 771 abort(); 772 } 773 774 /* 775 * Magic entries with no description get a bonus because they depend 776 * on subsequent magic entries to print something. 777 */ 778 if (m->desc[0] == '\0') 779 val++; 780 return val; 781 } 782 783 /* 784 * Sort callback for sorting entries by "strength" (basically length) 785 */ 786 private int 787 apprentice_sort(const void *a, const void *b) 788 { 789 const struct magic_entry *ma = CAST(const struct magic_entry *, a); 790 const struct magic_entry *mb = CAST(const struct magic_entry *, b); 791 size_t sa = apprentice_magic_strength(ma->mp); 792 size_t sb = apprentice_magic_strength(mb->mp); 793 if (sa == sb) 794 return 0; 795 else if (sa > sb) 796 return -1; 797 else 798 return 1; 799 } 800 801 /* 802 * Shows sorted patterns list in the order which is used for the matching 803 */ 804 private void 805 apprentice_list(struct mlist *mlist, int mode) 806 { 807 uint32_t magindex = 0; 808 struct mlist *ml; 809 for (ml = mlist->next; ml != mlist; ml = ml->next) { 810 for (magindex = 0; magindex < ml->nmagic; magindex++) { 811 struct magic *m = &ml->magic[magindex]; 812 if ((m->flag & mode) != mode) { 813 /* Skip sub-tests */ 814 while (magindex + 1 < ml->nmagic && 815 ml->magic[magindex + 1].cont_level != 0) 816 ++magindex; 817 continue; /* Skip to next top-level test*/ 818 } 819 820 /* 821 * Try to iterate over the tree until we find item with 822 * description/mimetype. 823 */ 824 while (magindex + 1 < ml->nmagic && 825 ml->magic[magindex + 1].cont_level != 0 && 826 *ml->magic[magindex].desc == '\0' && 827 *ml->magic[magindex].mimetype == '\0') 828 magindex++; 829 830 printf("Strength = %3" SIZE_T_FORMAT "u : %s [%s]\n", 831 apprentice_magic_strength(m), 832 ml->magic[magindex].desc, 833 ml->magic[magindex].mimetype); 834 } 835 } 836 } 837 838 private void 839 set_test_type(struct magic *mstart, struct magic *m) 840 { 841 switch (m->type) { 842 case FILE_BYTE: 843 case FILE_SHORT: 844 case FILE_LONG: 845 case FILE_DATE: 846 case FILE_BESHORT: 847 case FILE_BELONG: 848 case FILE_BEDATE: 849 case FILE_LESHORT: 850 case FILE_LELONG: 851 case FILE_LEDATE: 852 case FILE_LDATE: 853 case FILE_BELDATE: 854 case FILE_LELDATE: 855 case FILE_MEDATE: 856 case FILE_MELDATE: 857 case FILE_MELONG: 858 case FILE_QUAD: 859 case FILE_LEQUAD: 860 case FILE_BEQUAD: 861 case FILE_QDATE: 862 case FILE_LEQDATE: 863 case FILE_BEQDATE: 864 case FILE_QLDATE: 865 case FILE_LEQLDATE: 866 case FILE_BEQLDATE: 867 case FILE_QWDATE: 868 case FILE_LEQWDATE: 869 case FILE_BEQWDATE: 870 case FILE_FLOAT: 871 case FILE_BEFLOAT: 872 case FILE_LEFLOAT: 873 case FILE_DOUBLE: 874 case FILE_BEDOUBLE: 875 case FILE_LEDOUBLE: 876 mstart->flag |= BINTEST; 877 break; 878 case FILE_STRING: 879 case FILE_PSTRING: 880 case FILE_BESTRING16: 881 case FILE_LESTRING16: 882 /* Allow text overrides */ 883 if (mstart->str_flags & STRING_TEXTTEST) 884 mstart->flag |= TEXTTEST; 885 else 886 mstart->flag |= BINTEST; 887 break; 888 case FILE_REGEX: 889 case FILE_SEARCH: 890 /* Check for override */ 891 if (mstart->str_flags & STRING_BINTEST) 892 mstart->flag |= BINTEST; 893 if (mstart->str_flags & STRING_TEXTTEST) 894 mstart->flag |= TEXTTEST; 895 896 if (mstart->flag & (TEXTTEST|BINTEST)) 897 break; 898 899 /* binary test if pattern is not text */ 900 if (file_looks_utf8(m->value.us, (size_t)m->vallen, NULL, 901 NULL) <= 0) 902 mstart->flag |= BINTEST; 903 else 904 mstart->flag |= TEXTTEST; 905 break; 906 case FILE_DEFAULT: 907 /* can't deduce anything; we shouldn't see this at the 908 top level anyway */ 909 break; 910 case FILE_INVALID: 911 default: 912 /* invalid search type, but no need to complain here */ 913 break; 914 } 915 } 916 917 private int 918 addentry(struct magic_set *ms, struct magic_entry *me, 919 struct magic_entry_set *mset) 920 { 921 size_t i = me->mp->type == FILE_NAME ? 1 : 0; 922 if (mset[i].count == mset[i].max) { 923 struct magic_entry *mp; 924 925 mset[i].max += ALLOC_INCR; 926 if ((mp = CAST(struct magic_entry *, 927 realloc(mset[i].me, sizeof(*mp) * mset[i].max))) == 928 NULL) { 929 file_oomem(ms, sizeof(*mp) * mset[i].max); 930 return -1; 931 } 932 (void)memset(&mp[mset[i].count], 0, sizeof(*mp) * 933 ALLOC_INCR); 934 mset[i].me = mp; 935 } 936 mset[i].me[mset[i].count++] = *me; 937 memset(me, 0, sizeof(*me)); 938 return 0; 939 } 940 941 /* 942 * Load and parse one file. 943 */ 944 private void 945 load_1(struct magic_set *ms, int action, const char *fn, int *errs, 946 struct magic_entry_set *mset) 947 { 948 size_t lineno = 0, llen = 0; 949 char *line = NULL; 950 ssize_t len; 951 struct magic_entry me; 952 953 FILE *f = fopen(ms->file = fn, "r"); 954 if (f == NULL) { 955 if (errno != ENOENT) 956 file_error(ms, errno, "cannot read magic file `%s'", 957 fn); 958 (*errs)++; 959 return; 960 } 961 962 memset(&me, 0, sizeof(me)); 963 /* read and parse this file */ 964 for (ms->line = 1; (len = getline(&line, &llen, f)) != -1; 965 ms->line++) { 966 if (len == 0) /* null line, garbage, etc */ 967 continue; 968 if (line[len - 1] == '\n') { 969 lineno++; 970 line[len - 1] = '\0'; /* delete newline */ 971 } 972 switch (line[0]) { 973 case '\0': /* empty, do not parse */ 974 case '#': /* comment, do not parse */ 975 continue; 976 case '!': 977 if (line[1] == ':') { 978 size_t i; 979 980 for (i = 0; bang[i].name != NULL; i++) { 981 if ((size_t)(len - 2) > bang[i].len && 982 memcmp(bang[i].name, line + 2, 983 bang[i].len) == 0) 984 break; 985 } 986 if (bang[i].name == NULL) { 987 file_error(ms, 0, 988 "Unknown !: entry `%s'", line); 989 (*errs)++; 990 continue; 991 } 992 if (me.mp == NULL) { 993 file_error(ms, 0, 994 "No current entry for :!%s type", 995 bang[i].name); 996 (*errs)++; 997 continue; 998 } 999 if ((*bang[i].fun)(ms, &me, 1000 line + bang[i].len + 2) != 0) { 1001 (*errs)++; 1002 continue; 1003 } 1004 continue; 1005 } 1006 /*FALLTHROUGH*/ 1007 default: 1008 again: 1009 switch (parse(ms, &me, line, lineno, action)) { 1010 case 0: 1011 continue; 1012 case 1: 1013 (void)addentry(ms, &me, mset); 1014 goto again; 1015 default: 1016 (*errs)++; 1017 break; 1018 } 1019 } 1020 } 1021 if (me.mp) 1022 (void)addentry(ms, &me, mset); 1023 free(line); 1024 (void)fclose(f); 1025 } 1026 1027 /* 1028 * parse a file or directory of files 1029 * const char *fn: name of magic file or directory 1030 */ 1031 private int 1032 cmpstrp(const void *p1, const void *p2) 1033 { 1034 return strcmp(*(char *const *)p1, *(char *const *)p2); 1035 } 1036 1037 1038 private uint32_t 1039 set_text_binary(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1040 uint32_t starttest) 1041 { 1042 static const char text[] = "text"; 1043 static const char binary[] = "binary"; 1044 static const size_t len = sizeof(text); 1045 1046 uint32_t i = starttest; 1047 1048 do { 1049 set_test_type(me[starttest].mp, me[i].mp); 1050 if ((ms->flags & MAGIC_DEBUG) == 0) 1051 continue; 1052 (void)fprintf(stderr, "%s%s%s: %s\n", 1053 me[i].mp->mimetype, 1054 me[i].mp->mimetype[0] == '\0' ? "" : "; ", 1055 me[i].mp->desc[0] ? me[i].mp->desc : "(no description)", 1056 me[i].mp->flag & BINTEST ? binary : text); 1057 if (me[i].mp->flag & BINTEST) { 1058 char *p = strstr(me[i].mp->desc, text); 1059 if (p && (p == me[i].mp->desc || 1060 isspace((unsigned char)p[-1])) && 1061 (p + len - me[i].mp->desc == MAXstring 1062 || (p[len] == '\0' || 1063 isspace((unsigned char)p[len])))) 1064 (void)fprintf(stderr, "*** Possible " 1065 "binary test for text type\n"); 1066 } 1067 } while (++i < nme && me[i].mp->cont_level != 0); 1068 return i; 1069 } 1070 1071 private void 1072 set_last_default(struct magic_set *ms, struct magic_entry *me, uint32_t nme) 1073 { 1074 uint32_t i; 1075 for (i = 0; i < nme; i++) { 1076 if (me[i].mp->cont_level == 0 && 1077 me[i].mp->type == FILE_DEFAULT) { 1078 while (++i < nme) 1079 if (me[i].mp->cont_level == 0) 1080 break; 1081 if (i != nme) { 1082 /* XXX - Ugh! */ 1083 ms->line = me[i].mp->lineno; 1084 file_magwarn(ms, 1085 "level 0 \"default\" did not sort last"); 1086 } 1087 return; 1088 } 1089 } 1090 } 1091 1092 private int 1093 coalesce_entries(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1094 struct magic **ma, uint32_t *nma) 1095 { 1096 uint32_t i, mentrycount = 0; 1097 size_t slen; 1098 1099 for (i = 0; i < nme; i++) 1100 mentrycount += me[i].cont_count; 1101 1102 slen = sizeof(**ma) * mentrycount; 1103 if ((*ma = CAST(struct magic *, malloc(slen))) == NULL) { 1104 file_oomem(ms, slen); 1105 return -1; 1106 } 1107 1108 mentrycount = 0; 1109 for (i = 0; i < nme; i++) { 1110 (void)memcpy(*ma + mentrycount, me[i].mp, 1111 me[i].cont_count * sizeof(**ma)); 1112 mentrycount += me[i].cont_count; 1113 } 1114 *nma = mentrycount; 1115 return 0; 1116 } 1117 1118 private void 1119 magic_entry_free(struct magic_entry *me, uint32_t nme) 1120 { 1121 uint32_t i; 1122 if (me == NULL) 1123 return; 1124 for (i = 0; i < nme; i++) 1125 free(me[i].mp); 1126 free(me); 1127 } 1128 1129 private struct magic_map * 1130 apprentice_load(struct magic_set *ms, const char *fn, int action) 1131 { 1132 int errs = 0; 1133 uint32_t i, j; 1134 size_t files = 0, maxfiles = 0; 1135 char **filearr = NULL, *mfn; 1136 struct stat st; 1137 struct magic_map *map; 1138 struct magic_entry_set mset[MAGIC_SETS]; 1139 DIR *dir; 1140 struct dirent *d; 1141 1142 memset(mset, 0, sizeof(mset)); 1143 ms->flags |= MAGIC_CHECK; /* Enable checks for parsed files */ 1144 1145 1146 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) 1147 { 1148 file_oomem(ms, sizeof(*map)); 1149 return NULL; 1150 } 1151 1152 /* print silly verbose header for USG compat. */ 1153 if (action == FILE_CHECK) 1154 (void)fprintf(stderr, "%s\n", usg_hdr); 1155 1156 /* load directory or file */ 1157 if (stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) { 1158 dir = opendir(fn); 1159 if (!dir) { 1160 errs++; 1161 goto out; 1162 } 1163 while ((d = readdir(dir)) != NULL) { 1164 if (asprintf(&mfn, "%s/%s", fn, d->d_name) < 0) { 1165 file_oomem(ms, 1166 strlen(fn) + strlen(d->d_name) + 2); 1167 errs++; 1168 closedir(dir); 1169 goto out; 1170 } 1171 if (stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) { 1172 free(mfn); 1173 continue; 1174 } 1175 if (files >= maxfiles) { 1176 size_t mlen; 1177 maxfiles = (maxfiles + 1) * 2; 1178 mlen = maxfiles * sizeof(*filearr); 1179 if ((filearr = CAST(char **, 1180 realloc(filearr, mlen))) == NULL) { 1181 file_oomem(ms, mlen); 1182 free(mfn); 1183 closedir(dir); 1184 errs++; 1185 goto out; 1186 } 1187 } 1188 filearr[files++] = mfn; 1189 } 1190 closedir(dir); 1191 qsort(filearr, files, sizeof(*filearr), cmpstrp); 1192 for (i = 0; i < files; i++) { 1193 load_1(ms, action, filearr[i], &errs, mset); 1194 free(filearr[i]); 1195 } 1196 free(filearr); 1197 } else 1198 load_1(ms, action, fn, &errs, mset); 1199 if (errs) 1200 goto out; 1201 1202 for (j = 0; j < MAGIC_SETS; j++) { 1203 /* Set types of tests */ 1204 for (i = 0; i < mset[j].count; ) { 1205 if (mset[j].me[i].mp->cont_level != 0) { 1206 i++; 1207 continue; 1208 } 1209 i = set_text_binary(ms, mset[j].me, mset[j].count, i); 1210 } 1211 qsort(mset[j].me, mset[j].count, sizeof(*mset[j].me), 1212 apprentice_sort); 1213 1214 /* 1215 * Make sure that any level 0 "default" line is last 1216 * (if one exists). 1217 */ 1218 set_last_default(ms, mset[j].me, mset[j].count); 1219 1220 /* coalesce per file arrays into a single one */ 1221 if (coalesce_entries(ms, mset[j].me, mset[j].count, 1222 &map->magic[j], &map->nmagic[j]) == -1) { 1223 errs++; 1224 goto out; 1225 } 1226 } 1227 1228 out: 1229 for (j = 0; j < MAGIC_SETS; j++) 1230 magic_entry_free(mset[j].me, mset[j].count); 1231 1232 if (errs) { 1233 for (j = 0; j < MAGIC_SETS; j++) { 1234 if (map->magic[j]) 1235 free(map->magic[j]); 1236 } 1237 free(map); 1238 return NULL; 1239 } 1240 return map; 1241 } 1242 1243 /* 1244 * extend the sign bit if the comparison is to be signed 1245 */ 1246 protected uint64_t 1247 file_signextend(struct magic_set *ms, struct magic *m, uint64_t v) 1248 { 1249 if (!(m->flag & UNSIGNED)) { 1250 switch(m->type) { 1251 /* 1252 * Do not remove the casts below. They are 1253 * vital. When later compared with the data, 1254 * the sign extension must have happened. 1255 */ 1256 case FILE_BYTE: 1257 v = (char) v; 1258 break; 1259 case FILE_SHORT: 1260 case FILE_BESHORT: 1261 case FILE_LESHORT: 1262 v = (short) v; 1263 break; 1264 case FILE_DATE: 1265 case FILE_BEDATE: 1266 case FILE_LEDATE: 1267 case FILE_MEDATE: 1268 case FILE_LDATE: 1269 case FILE_BELDATE: 1270 case FILE_LELDATE: 1271 case FILE_MELDATE: 1272 case FILE_LONG: 1273 case FILE_BELONG: 1274 case FILE_LELONG: 1275 case FILE_MELONG: 1276 case FILE_FLOAT: 1277 case FILE_BEFLOAT: 1278 case FILE_LEFLOAT: 1279 v = (int32_t) v; 1280 break; 1281 case FILE_QUAD: 1282 case FILE_BEQUAD: 1283 case FILE_LEQUAD: 1284 case FILE_QDATE: 1285 case FILE_QLDATE: 1286 case FILE_QWDATE: 1287 case FILE_BEQDATE: 1288 case FILE_BEQLDATE: 1289 case FILE_BEQWDATE: 1290 case FILE_LEQDATE: 1291 case FILE_LEQLDATE: 1292 case FILE_LEQWDATE: 1293 case FILE_DOUBLE: 1294 case FILE_BEDOUBLE: 1295 case FILE_LEDOUBLE: 1296 v = (int64_t) v; 1297 break; 1298 case FILE_STRING: 1299 case FILE_PSTRING: 1300 case FILE_BESTRING16: 1301 case FILE_LESTRING16: 1302 case FILE_REGEX: 1303 case FILE_SEARCH: 1304 case FILE_DEFAULT: 1305 case FILE_INDIRECT: 1306 case FILE_NAME: 1307 case FILE_USE: 1308 case FILE_CLEAR: 1309 break; 1310 default: 1311 if (ms->flags & MAGIC_CHECK) 1312 file_magwarn(ms, "cannot happen: m->type=%d\n", 1313 m->type); 1314 return ~0U; 1315 } 1316 } 1317 return v; 1318 } 1319 1320 private int 1321 string_modifier_check(struct magic_set *ms, struct magic *m) 1322 { 1323 if ((ms->flags & MAGIC_CHECK) == 0) 1324 return 0; 1325 1326 if (m->type != FILE_PSTRING && (m->str_flags & PSTRING_LEN) != 0) { 1327 file_magwarn(ms, 1328 "'/BHhLl' modifiers are only allowed for pascal strings\n"); 1329 return -1; 1330 } 1331 switch (m->type) { 1332 case FILE_BESTRING16: 1333 case FILE_LESTRING16: 1334 if (m->str_flags != 0) { 1335 file_magwarn(ms, 1336 "no modifiers allowed for 16-bit strings\n"); 1337 return -1; 1338 } 1339 break; 1340 case FILE_STRING: 1341 case FILE_PSTRING: 1342 if ((m->str_flags & REGEX_OFFSET_START) != 0) { 1343 file_magwarn(ms, 1344 "'/%c' only allowed on regex and search\n", 1345 CHAR_REGEX_OFFSET_START); 1346 return -1; 1347 } 1348 break; 1349 case FILE_SEARCH: 1350 if (m->str_range == 0) { 1351 file_magwarn(ms, 1352 "missing range; defaulting to %d\n", 1353 STRING_DEFAULT_RANGE); 1354 m->str_range = STRING_DEFAULT_RANGE; 1355 return -1; 1356 } 1357 break; 1358 case FILE_REGEX: 1359 if ((m->str_flags & STRING_COMPACT_WHITESPACE) != 0) { 1360 file_magwarn(ms, "'/%c' not allowed on regex\n", 1361 CHAR_COMPACT_WHITESPACE); 1362 return -1; 1363 } 1364 if ((m->str_flags & STRING_COMPACT_OPTIONAL_WHITESPACE) != 0) { 1365 file_magwarn(ms, "'/%c' not allowed on regex\n", 1366 CHAR_COMPACT_OPTIONAL_WHITESPACE); 1367 return -1; 1368 } 1369 break; 1370 default: 1371 file_magwarn(ms, "coding error: m->type=%d\n", 1372 m->type); 1373 return -1; 1374 } 1375 return 0; 1376 } 1377 1378 private int 1379 get_op(char c) 1380 { 1381 switch (c) { 1382 case '&': 1383 return FILE_OPAND; 1384 case '|': 1385 return FILE_OPOR; 1386 case '^': 1387 return FILE_OPXOR; 1388 case '+': 1389 return FILE_OPADD; 1390 case '-': 1391 return FILE_OPMINUS; 1392 case '*': 1393 return FILE_OPMULTIPLY; 1394 case '/': 1395 return FILE_OPDIVIDE; 1396 case '%': 1397 return FILE_OPMODULO; 1398 default: 1399 return -1; 1400 } 1401 } 1402 1403 #ifdef ENABLE_CONDITIONALS 1404 private int 1405 get_cond(const char *l, const char **t) 1406 { 1407 static const struct cond_tbl_s { 1408 char name[8]; 1409 size_t len; 1410 int cond; 1411 } cond_tbl[] = { 1412 { "if", 2, COND_IF }, 1413 { "elif", 4, COND_ELIF }, 1414 { "else", 4, COND_ELSE }, 1415 { "", 0, COND_NONE }, 1416 }; 1417 const struct cond_tbl_s *p; 1418 1419 for (p = cond_tbl; p->len; p++) { 1420 if (strncmp(l, p->name, p->len) == 0 && 1421 isspace((unsigned char)l[p->len])) { 1422 if (t) 1423 *t = l + p->len; 1424 break; 1425 } 1426 } 1427 return p->cond; 1428 } 1429 1430 private int 1431 check_cond(struct magic_set *ms, int cond, uint32_t cont_level) 1432 { 1433 int last_cond; 1434 last_cond = ms->c.li[cont_level].last_cond; 1435 1436 switch (cond) { 1437 case COND_IF: 1438 if (last_cond != COND_NONE && last_cond != COND_ELIF) { 1439 if (ms->flags & MAGIC_CHECK) 1440 file_magwarn(ms, "syntax error: `if'"); 1441 return -1; 1442 } 1443 last_cond = COND_IF; 1444 break; 1445 1446 case COND_ELIF: 1447 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1448 if (ms->flags & MAGIC_CHECK) 1449 file_magwarn(ms, "syntax error: `elif'"); 1450 return -1; 1451 } 1452 last_cond = COND_ELIF; 1453 break; 1454 1455 case COND_ELSE: 1456 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1457 if (ms->flags & MAGIC_CHECK) 1458 file_magwarn(ms, "syntax error: `else'"); 1459 return -1; 1460 } 1461 last_cond = COND_NONE; 1462 break; 1463 1464 case COND_NONE: 1465 last_cond = COND_NONE; 1466 break; 1467 } 1468 1469 ms->c.li[cont_level].last_cond = last_cond; 1470 return 0; 1471 } 1472 #endif /* ENABLE_CONDITIONALS */ 1473 1474 /* 1475 * parse one line from magic file, put into magic[index++] if valid 1476 */ 1477 private int 1478 parse(struct magic_set *ms, struct magic_entry *me, const char *line, 1479 size_t lineno, int action) 1480 { 1481 #ifdef ENABLE_CONDITIONALS 1482 static uint32_t last_cont_level = 0; 1483 #endif 1484 size_t i; 1485 struct magic *m; 1486 const char *l = line; 1487 char *t; 1488 int op; 1489 uint32_t cont_level; 1490 int32_t diff; 1491 1492 cont_level = 0; 1493 1494 /* 1495 * Parse the offset. 1496 */ 1497 while (*l == '>') { 1498 ++l; /* step over */ 1499 cont_level++; 1500 } 1501 #ifdef ENABLE_CONDITIONALS 1502 if (cont_level == 0 || cont_level > last_cont_level) 1503 if (file_check_mem(ms, cont_level) == -1) 1504 return -1; 1505 last_cont_level = cont_level; 1506 #endif 1507 if (cont_level != 0) { 1508 if (me->mp == NULL) { 1509 file_magerror(ms, "No current entry for continuation"); 1510 return -1; 1511 } 1512 if (me->cont_count == 0) { 1513 file_magerror(ms, "Continuations present with 0 count"); 1514 return -1; 1515 } 1516 m = &me->mp[me->cont_count - 1]; 1517 diff = (int32_t)cont_level - (int32_t)m->cont_level; 1518 if (diff > 1) 1519 file_magwarn(ms, "New continuation level %u is more " 1520 "than one larger than current level %u", cont_level, 1521 m->cont_level); 1522 if (me->cont_count == me->max_count) { 1523 struct magic *nm; 1524 size_t cnt = me->max_count + ALLOC_CHUNK; 1525 if ((nm = CAST(struct magic *, realloc(me->mp, 1526 sizeof(*nm) * cnt))) == NULL) { 1527 file_oomem(ms, sizeof(*nm) * cnt); 1528 return -1; 1529 } 1530 me->mp = m = nm; 1531 me->max_count = CAST(uint32_t, cnt); 1532 } 1533 m = &me->mp[me->cont_count++]; 1534 (void)memset(m, 0, sizeof(*m)); 1535 m->cont_level = cont_level; 1536 } else { 1537 static const size_t len = sizeof(*m) * ALLOC_CHUNK; 1538 if (me->mp != NULL) 1539 return 1; 1540 if ((m = CAST(struct magic *, malloc(len))) == NULL) { 1541 file_oomem(ms, len); 1542 return -1; 1543 } 1544 me->mp = m; 1545 me->max_count = ALLOC_CHUNK; 1546 (void)memset(m, 0, sizeof(*m)); 1547 m->factor_op = FILE_FACTOR_OP_NONE; 1548 m->cont_level = 0; 1549 me->cont_count = 1; 1550 } 1551 m->lineno = CAST(uint32_t, lineno); 1552 1553 if (*l == '&') { /* m->cont_level == 0 checked below. */ 1554 ++l; /* step over */ 1555 m->flag |= OFFADD; 1556 } 1557 if (*l == '(') { 1558 ++l; /* step over */ 1559 m->flag |= INDIR; 1560 if (m->flag & OFFADD) 1561 m->flag = (m->flag & ~OFFADD) | INDIROFFADD; 1562 1563 if (*l == '&') { /* m->cont_level == 0 checked below */ 1564 ++l; /* step over */ 1565 m->flag |= OFFADD; 1566 } 1567 } 1568 /* Indirect offsets are not valid at level 0. */ 1569 if (m->cont_level == 0 && (m->flag & (OFFADD | INDIROFFADD))) 1570 if (ms->flags & MAGIC_CHECK) 1571 file_magwarn(ms, "relative offset at level 0"); 1572 1573 /* get offset, then skip over it */ 1574 m->offset = (uint32_t)strtoul(l, &t, 0); 1575 if (l == t) 1576 if (ms->flags & MAGIC_CHECK) 1577 file_magwarn(ms, "offset `%s' invalid", l); 1578 l = t; 1579 1580 if (m->flag & INDIR) { 1581 m->in_type = FILE_LONG; 1582 m->in_offset = 0; 1583 /* 1584 * read [.lbs][+-]nnnnn) 1585 */ 1586 if (*l == '.') { 1587 l++; 1588 switch (*l) { 1589 case 'l': 1590 m->in_type = FILE_LELONG; 1591 break; 1592 case 'L': 1593 m->in_type = FILE_BELONG; 1594 break; 1595 case 'm': 1596 m->in_type = FILE_MELONG; 1597 break; 1598 case 'h': 1599 case 's': 1600 m->in_type = FILE_LESHORT; 1601 break; 1602 case 'H': 1603 case 'S': 1604 m->in_type = FILE_BESHORT; 1605 break; 1606 case 'c': 1607 case 'b': 1608 case 'C': 1609 case 'B': 1610 m->in_type = FILE_BYTE; 1611 break; 1612 case 'e': 1613 case 'f': 1614 case 'g': 1615 m->in_type = FILE_LEDOUBLE; 1616 break; 1617 case 'E': 1618 case 'F': 1619 case 'G': 1620 m->in_type = FILE_BEDOUBLE; 1621 break; 1622 case 'i': 1623 m->in_type = FILE_LEID3; 1624 break; 1625 case 'I': 1626 m->in_type = FILE_BEID3; 1627 break; 1628 default: 1629 if (ms->flags & MAGIC_CHECK) 1630 file_magwarn(ms, 1631 "indirect offset type `%c' invalid", 1632 *l); 1633 break; 1634 } 1635 l++; 1636 } 1637 1638 m->in_op = 0; 1639 if (*l == '~') { 1640 m->in_op |= FILE_OPINVERSE; 1641 l++; 1642 } 1643 if ((op = get_op(*l)) != -1) { 1644 m->in_op |= op; 1645 l++; 1646 } 1647 if (*l == '(') { 1648 m->in_op |= FILE_OPINDIRECT; 1649 l++; 1650 } 1651 if (isdigit((unsigned char)*l) || *l == '-') { 1652 m->in_offset = (int32_t)strtol(l, &t, 0); 1653 if (l == t) 1654 if (ms->flags & MAGIC_CHECK) 1655 file_magwarn(ms, 1656 "in_offset `%s' invalid", l); 1657 l = t; 1658 } 1659 if (*l++ != ')' || 1660 ((m->in_op & FILE_OPINDIRECT) && *l++ != ')')) 1661 if (ms->flags & MAGIC_CHECK) 1662 file_magwarn(ms, 1663 "missing ')' in indirect offset"); 1664 } 1665 EATAB; 1666 1667 #ifdef ENABLE_CONDITIONALS 1668 m->cond = get_cond(l, &l); 1669 if (check_cond(ms, m->cond, cont_level) == -1) 1670 return -1; 1671 1672 EATAB; 1673 #endif 1674 1675 /* 1676 * Parse the type. 1677 */ 1678 if (*l == 'u') { 1679 /* 1680 * Try it as a keyword type prefixed by "u"; match what 1681 * follows the "u". If that fails, try it as an SUS 1682 * integer type. 1683 */ 1684 m->type = get_type(type_tbl, l + 1, &l); 1685 if (m->type == FILE_INVALID) { 1686 /* 1687 * Not a keyword type; parse it as an SUS type, 1688 * 'u' possibly followed by a number or C/S/L. 1689 */ 1690 m->type = get_standard_integer_type(l, &l); 1691 } 1692 // It's unsigned. 1693 if (m->type != FILE_INVALID) 1694 m->flag |= UNSIGNED; 1695 } else { 1696 /* 1697 * Try it as a keyword type. If that fails, try it as 1698 * an SUS integer type if it begins with "d" or as an 1699 * SUS string type if it begins with "s". In any case, 1700 * it's not unsigned. 1701 */ 1702 m->type = get_type(type_tbl, l, &l); 1703 if (m->type == FILE_INVALID) { 1704 /* 1705 * Not a keyword type; parse it as an SUS type, 1706 * either 'd' possibly followed by a number or 1707 * C/S/L, or just 's'. 1708 */ 1709 if (*l == 'd') 1710 m->type = get_standard_integer_type(l, &l); 1711 else if (*l == 's' && !isalpha((unsigned char)l[1])) { 1712 m->type = FILE_STRING; 1713 ++l; 1714 } 1715 } 1716 } 1717 1718 if (m->type == FILE_INVALID) { 1719 /* Not found - try it as a special keyword. */ 1720 m->type = get_type(special_tbl, l, &l); 1721 } 1722 1723 if (m->type == FILE_INVALID) { 1724 if (ms->flags & MAGIC_CHECK) 1725 file_magwarn(ms, "type `%s' invalid", l); 1726 return -1; 1727 } 1728 1729 /* New-style anding: "0 byte&0x80 =0x80 dynamically linked" */ 1730 /* New and improved: ~ & | ^ + - * / % -- exciting, isn't it? */ 1731 1732 m->mask_op = 0; 1733 if (*l == '~') { 1734 if (!IS_STRING(m->type)) 1735 m->mask_op |= FILE_OPINVERSE; 1736 else if (ms->flags & MAGIC_CHECK) 1737 file_magwarn(ms, "'~' invalid for string types"); 1738 ++l; 1739 } 1740 m->str_range = 0; 1741 m->str_flags = m->type == FILE_PSTRING ? PSTRING_1_LE : 0; 1742 if ((op = get_op(*l)) != -1) { 1743 if (!IS_STRING(m->type)) { 1744 uint64_t val; 1745 ++l; 1746 m->mask_op |= op; 1747 val = (uint64_t)strtoull(l, &t, 0); 1748 l = t; 1749 m->num_mask = file_signextend(ms, m, val); 1750 eatsize(&l); 1751 } 1752 else if (op == FILE_OPDIVIDE) { 1753 int have_range = 0; 1754 while (!isspace((unsigned char)*++l)) { 1755 switch (*l) { 1756 case '0': case '1': case '2': 1757 case '3': case '4': case '5': 1758 case '6': case '7': case '8': 1759 case '9': 1760 if (have_range && 1761 (ms->flags & MAGIC_CHECK)) 1762 file_magwarn(ms, 1763 "multiple ranges"); 1764 have_range = 1; 1765 m->str_range = CAST(uint32_t, 1766 strtoul(l, &t, 0)); 1767 if (m->str_range == 0) 1768 file_magwarn(ms, 1769 "zero range"); 1770 l = t - 1; 1771 break; 1772 case CHAR_COMPACT_WHITESPACE: 1773 m->str_flags |= 1774 STRING_COMPACT_WHITESPACE; 1775 break; 1776 case CHAR_COMPACT_OPTIONAL_WHITESPACE: 1777 m->str_flags |= 1778 STRING_COMPACT_OPTIONAL_WHITESPACE; 1779 break; 1780 case CHAR_IGNORE_LOWERCASE: 1781 m->str_flags |= STRING_IGNORE_LOWERCASE; 1782 break; 1783 case CHAR_IGNORE_UPPERCASE: 1784 m->str_flags |= STRING_IGNORE_UPPERCASE; 1785 break; 1786 case CHAR_REGEX_OFFSET_START: 1787 m->str_flags |= REGEX_OFFSET_START; 1788 break; 1789 case CHAR_BINTEST: 1790 m->str_flags |= STRING_BINTEST; 1791 break; 1792 case CHAR_TEXTTEST: 1793 m->str_flags |= STRING_TEXTTEST; 1794 break; 1795 case CHAR_TRIM: 1796 m->str_flags |= STRING_TRIM; 1797 break; 1798 case CHAR_PSTRING_1_LE: 1799 if (m->type != FILE_PSTRING) 1800 goto bad; 1801 m->str_flags = (m->str_flags & ~PSTRING_LEN) | PSTRING_1_LE; 1802 break; 1803 case CHAR_PSTRING_2_BE: 1804 if (m->type != FILE_PSTRING) 1805 goto bad; 1806 m->str_flags = (m->str_flags & ~PSTRING_LEN) | PSTRING_2_BE; 1807 break; 1808 case CHAR_PSTRING_2_LE: 1809 if (m->type != FILE_PSTRING) 1810 goto bad; 1811 m->str_flags = (m->str_flags & ~PSTRING_LEN) | PSTRING_2_LE; 1812 break; 1813 case CHAR_PSTRING_4_BE: 1814 if (m->type != FILE_PSTRING) 1815 goto bad; 1816 m->str_flags = (m->str_flags & ~PSTRING_LEN) | PSTRING_4_BE; 1817 break; 1818 case CHAR_PSTRING_4_LE: 1819 if (m->type != FILE_PSTRING) 1820 goto bad; 1821 m->str_flags = (m->str_flags & ~PSTRING_LEN) | PSTRING_4_LE; 1822 break; 1823 case CHAR_PSTRING_LENGTH_INCLUDES_ITSELF: 1824 if (m->type != FILE_PSTRING) 1825 goto bad; 1826 m->str_flags |= PSTRING_LENGTH_INCLUDES_ITSELF; 1827 break; 1828 default: 1829 bad: 1830 if (ms->flags & MAGIC_CHECK) 1831 file_magwarn(ms, 1832 "string extension `%c' " 1833 "invalid", *l); 1834 return -1; 1835 } 1836 /* allow multiple '/' for readability */ 1837 if (l[1] == '/' && 1838 !isspace((unsigned char)l[2])) 1839 l++; 1840 } 1841 if (string_modifier_check(ms, m) == -1) 1842 return -1; 1843 } 1844 else { 1845 if (ms->flags & MAGIC_CHECK) 1846 file_magwarn(ms, "invalid string op: %c", *t); 1847 return -1; 1848 } 1849 } 1850 /* 1851 * We used to set mask to all 1's here, instead let's just not do 1852 * anything if mask = 0 (unless you have a better idea) 1853 */ 1854 EATAB; 1855 1856 switch (*l) { 1857 case '>': 1858 case '<': 1859 m->reln = *l; 1860 ++l; 1861 if (*l == '=') { 1862 if (ms->flags & MAGIC_CHECK) { 1863 file_magwarn(ms, "%c= not supported", 1864 m->reln); 1865 return -1; 1866 } 1867 ++l; 1868 } 1869 break; 1870 /* Old-style anding: "0 byte &0x80 dynamically linked" */ 1871 case '&': 1872 case '^': 1873 case '=': 1874 m->reln = *l; 1875 ++l; 1876 if (*l == '=') { 1877 /* HP compat: ignore &= etc. */ 1878 ++l; 1879 } 1880 break; 1881 case '!': 1882 m->reln = *l; 1883 ++l; 1884 break; 1885 default: 1886 m->reln = '='; /* the default relation */ 1887 if (*l == 'x' && ((isascii((unsigned char)l[1]) && 1888 isspace((unsigned char)l[1])) || !l[1])) { 1889 m->reln = *l; 1890 ++l; 1891 } 1892 break; 1893 } 1894 /* 1895 * Grab the value part, except for an 'x' reln. 1896 */ 1897 if (m->reln != 'x' && getvalue(ms, m, &l, action)) 1898 return -1; 1899 1900 /* 1901 * TODO finish this macro and start using it! 1902 * #define offsetcheck {if (offset > HOWMANY-1) 1903 * magwarn("offset too big"); } 1904 */ 1905 1906 /* 1907 * Now get last part - the description 1908 */ 1909 EATAB; 1910 if (l[0] == '\b') { 1911 ++l; 1912 m->flag |= NOSPACE; 1913 } else if ((l[0] == '\\') && (l[1] == 'b')) { 1914 ++l; 1915 ++l; 1916 m->flag |= NOSPACE; 1917 } 1918 for (i = 0; (m->desc[i++] = *l++) != '\0' && i < sizeof(m->desc); ) 1919 continue; 1920 if (i == sizeof(m->desc)) { 1921 m->desc[sizeof(m->desc) - 1] = '\0'; 1922 if (ms->flags & MAGIC_CHECK) 1923 file_magwarn(ms, "description `%s' truncated", m->desc); 1924 } 1925 1926 /* 1927 * We only do this check while compiling, or if any of the magic 1928 * files were not compiled. 1929 */ 1930 if (ms->flags & MAGIC_CHECK) { 1931 if (check_format(ms, m) == -1) 1932 return -1; 1933 } 1934 #ifndef COMPILE_ONLY 1935 if (action == FILE_CHECK) { 1936 file_mdump(m); 1937 } 1938 #endif 1939 m->mimetype[0] = '\0'; /* initialise MIME type to none */ 1940 return 0; 1941 } 1942 1943 /* 1944 * parse a STRENGTH annotation line from magic file, put into magic[index - 1] 1945 * if valid 1946 */ 1947 private int 1948 parse_strength(struct magic_set *ms, struct magic_entry *me, const char *line) 1949 { 1950 const char *l = line; 1951 char *el; 1952 unsigned long factor; 1953 struct magic *m = &me->mp[0]; 1954 1955 if (m->factor_op != FILE_FACTOR_OP_NONE) { 1956 file_magwarn(ms, 1957 "Current entry already has a strength type: %c %d", 1958 m->factor_op, m->factor); 1959 return -1; 1960 } 1961 if (m->type == FILE_NAME) { 1962 file_magwarn(ms, "%s: Strength setting is not supported in " 1963 "\"name\" magic entries", m->value.s); 1964 return -1; 1965 } 1966 EATAB; 1967 switch (*l) { 1968 case FILE_FACTOR_OP_NONE: 1969 case FILE_FACTOR_OP_PLUS: 1970 case FILE_FACTOR_OP_MINUS: 1971 case FILE_FACTOR_OP_TIMES: 1972 case FILE_FACTOR_OP_DIV: 1973 m->factor_op = *l++; 1974 break; 1975 default: 1976 file_magwarn(ms, "Unknown factor op `%c'", *l); 1977 return -1; 1978 } 1979 EATAB; 1980 factor = strtoul(l, &el, 0); 1981 if (factor > 255) { 1982 file_magwarn(ms, "Too large factor `%lu'", factor); 1983 goto out; 1984 } 1985 if (*el && !isspace((unsigned char)*el)) { 1986 file_magwarn(ms, "Bad factor `%s'", l); 1987 goto out; 1988 } 1989 m->factor = (uint8_t)factor; 1990 if (m->factor == 0 && m->factor_op == FILE_FACTOR_OP_DIV) { 1991 file_magwarn(ms, "Cannot have factor op `%c' and factor %u", 1992 m->factor_op, m->factor); 1993 goto out; 1994 } 1995 return 0; 1996 out: 1997 m->factor_op = FILE_FACTOR_OP_NONE; 1998 m->factor = 0; 1999 return -1; 2000 } 2001 2002 /* 2003 * Parse an Apple CREATOR/TYPE annotation from magic file and put it into 2004 * magic[index - 1] 2005 */ 2006 private int 2007 parse_apple(struct magic_set *ms, struct magic_entry *me, const char *line) 2008 { 2009 size_t i; 2010 const char *l = line; 2011 struct magic *m = &me->mp[me->cont_count == 0 ? 0 : me->cont_count - 1]; 2012 2013 if (m->apple[0] != '\0') { 2014 file_magwarn(ms, "Current entry already has a APPLE type " 2015 "`%.8s', new type `%s'", m->mimetype, l); 2016 return -1; 2017 } 2018 2019 EATAB; 2020 for (i = 0; *l && ((isascii((unsigned char)*l) && 2021 isalnum((unsigned char)*l)) || strchr("-+/.", *l)) && 2022 i < sizeof(m->apple); m->apple[i++] = *l++) 2023 continue; 2024 if (i == sizeof(m->apple) && *l) { 2025 /* We don't need to NUL terminate here, printing handles it */ 2026 if (ms->flags & MAGIC_CHECK) 2027 file_magwarn(ms, "APPLE type `%s' truncated %" 2028 SIZE_T_FORMAT "u", line, i); 2029 } 2030 2031 if (i > 0) 2032 return 0; 2033 else 2034 return -1; 2035 } 2036 2037 /* 2038 * parse a MIME annotation line from magic file, put into magic[index - 1] 2039 * if valid 2040 */ 2041 private int 2042 parse_mime(struct magic_set *ms, struct magic_entry *me, const char *line) 2043 { 2044 size_t i; 2045 const char *l = line; 2046 struct magic *m = &me->mp[me->cont_count == 0 ? 0 : me->cont_count - 1]; 2047 2048 if (m->mimetype[0] != '\0') { 2049 file_magwarn(ms, "Current entry already has a MIME type `%s'," 2050 " new type `%s'", m->mimetype, l); 2051 return -1; 2052 } 2053 2054 EATAB; 2055 for (i = 0; *l && ((isascii((unsigned char)*l) && 2056 isalnum((unsigned char)*l)) || strchr("-+/.", *l)) && 2057 i < sizeof(m->mimetype); m->mimetype[i++] = *l++) 2058 continue; 2059 if (i == sizeof(m->mimetype)) { 2060 m->mimetype[sizeof(m->mimetype) - 1] = '\0'; 2061 if (ms->flags & MAGIC_CHECK) 2062 file_magwarn(ms, "MIME type `%s' truncated %" 2063 SIZE_T_FORMAT "u", m->mimetype, i); 2064 } else 2065 m->mimetype[i] = '\0'; 2066 2067 if (i > 0) 2068 return 0; 2069 else 2070 return -1; 2071 } 2072 2073 private int 2074 check_format_type(const char *ptr, int type) 2075 { 2076 int quad = 0; 2077 if (*ptr == '\0') { 2078 /* Missing format string; bad */ 2079 return -1; 2080 } 2081 2082 switch (type) { 2083 case FILE_FMT_QUAD: 2084 quad = 1; 2085 /*FALLTHROUGH*/ 2086 case FILE_FMT_NUM: 2087 if (*ptr == '-') 2088 ptr++; 2089 if (*ptr == '.') 2090 ptr++; 2091 while (isdigit((unsigned char)*ptr)) ptr++; 2092 if (*ptr == '.') 2093 ptr++; 2094 while (isdigit((unsigned char)*ptr)) ptr++; 2095 if (quad) { 2096 if (*ptr++ != 'l') 2097 return -1; 2098 if (*ptr++ != 'l') 2099 return -1; 2100 } 2101 2102 switch (*ptr++) { 2103 case 'l': 2104 switch (*ptr++) { 2105 case 'i': 2106 case 'd': 2107 case 'u': 2108 case 'o': 2109 case 'x': 2110 case 'X': 2111 return 0; 2112 default: 2113 return -1; 2114 } 2115 2116 case 'h': 2117 switch (*ptr++) { 2118 case 'h': 2119 switch (*ptr++) { 2120 case 'i': 2121 case 'd': 2122 case 'u': 2123 case 'o': 2124 case 'x': 2125 case 'X': 2126 return 0; 2127 default: 2128 return -1; 2129 } 2130 case 'd': 2131 return 0; 2132 default: 2133 return -1; 2134 } 2135 2136 case 'i': 2137 case 'c': 2138 case 'd': 2139 case 'u': 2140 case 'o': 2141 case 'x': 2142 case 'X': 2143 return 0; 2144 2145 default: 2146 return -1; 2147 } 2148 2149 case FILE_FMT_FLOAT: 2150 case FILE_FMT_DOUBLE: 2151 if (*ptr == '-') 2152 ptr++; 2153 if (*ptr == '.') 2154 ptr++; 2155 while (isdigit((unsigned char)*ptr)) ptr++; 2156 if (*ptr == '.') 2157 ptr++; 2158 while (isdigit((unsigned char)*ptr)) ptr++; 2159 2160 switch (*ptr++) { 2161 case 'e': 2162 case 'E': 2163 case 'f': 2164 case 'F': 2165 case 'g': 2166 case 'G': 2167 return 0; 2168 2169 default: 2170 return -1; 2171 } 2172 2173 2174 case FILE_FMT_STR: 2175 if (*ptr == '-') 2176 ptr++; 2177 while (isdigit((unsigned char )*ptr)) 2178 ptr++; 2179 if (*ptr == '.') { 2180 ptr++; 2181 while (isdigit((unsigned char )*ptr)) 2182 ptr++; 2183 } 2184 2185 switch (*ptr++) { 2186 case 's': 2187 return 0; 2188 default: 2189 return -1; 2190 } 2191 2192 default: 2193 /* internal error */ 2194 abort(); 2195 } 2196 /*NOTREACHED*/ 2197 return -1; 2198 } 2199 2200 /* 2201 * Check that the optional printf format in description matches 2202 * the type of the magic. 2203 */ 2204 private int 2205 check_format(struct magic_set *ms, struct magic *m) 2206 { 2207 char *ptr; 2208 2209 for (ptr = m->desc; *ptr; ptr++) 2210 if (*ptr == '%') 2211 break; 2212 if (*ptr == '\0') { 2213 /* No format string; ok */ 2214 return 1; 2215 } 2216 2217 assert(file_nformats == file_nnames); 2218 2219 if (m->type >= file_nformats) { 2220 file_magwarn(ms, "Internal error inconsistency between " 2221 "m->type and format strings"); 2222 return -1; 2223 } 2224 if (file_formats[m->type] == FILE_FMT_NONE) { 2225 file_magwarn(ms, "No format string for `%s' with description " 2226 "`%s'", m->desc, file_names[m->type]); 2227 return -1; 2228 } 2229 2230 ptr++; 2231 if (check_format_type(ptr, file_formats[m->type]) == -1) { 2232 /* 2233 * TODO: this error message is unhelpful if the format 2234 * string is not one character long 2235 */ 2236 file_magwarn(ms, "Printf format `%c' is not valid for type " 2237 "`%s' in description `%s'", *ptr ? *ptr : '?', 2238 file_names[m->type], m->desc); 2239 return -1; 2240 } 2241 2242 for (; *ptr; ptr++) { 2243 if (*ptr == '%') { 2244 file_magwarn(ms, 2245 "Too many format strings (should have at most one) " 2246 "for `%s' with description `%s'", 2247 file_names[m->type], m->desc); 2248 return -1; 2249 } 2250 } 2251 return 0; 2252 } 2253 2254 /* 2255 * Read a numeric value from a pointer, into the value union of a magic 2256 * pointer, according to the magic type. Update the string pointer to point 2257 * just after the number read. Return 0 for success, non-zero for failure. 2258 */ 2259 private int 2260 getvalue(struct magic_set *ms, struct magic *m, const char **p, int action) 2261 { 2262 switch (m->type) { 2263 case FILE_BESTRING16: 2264 case FILE_LESTRING16: 2265 case FILE_STRING: 2266 case FILE_PSTRING: 2267 case FILE_REGEX: 2268 case FILE_SEARCH: 2269 case FILE_NAME: 2270 case FILE_USE: 2271 *p = getstr(ms, m, *p, action == FILE_COMPILE); 2272 if (*p == NULL) { 2273 if (ms->flags & MAGIC_CHECK) 2274 file_magwarn(ms, "cannot get string from `%s'", 2275 m->value.s); 2276 return -1; 2277 } 2278 return 0; 2279 case FILE_FLOAT: 2280 case FILE_BEFLOAT: 2281 case FILE_LEFLOAT: 2282 if (m->reln != 'x') { 2283 char *ep; 2284 #ifdef HAVE_STRTOF 2285 m->value.f = strtof(*p, &ep); 2286 #else 2287 m->value.f = (float)strtod(*p, &ep); 2288 #endif 2289 *p = ep; 2290 } 2291 return 0; 2292 case FILE_DOUBLE: 2293 case FILE_BEDOUBLE: 2294 case FILE_LEDOUBLE: 2295 if (m->reln != 'x') { 2296 char *ep; 2297 m->value.d = strtod(*p, &ep); 2298 *p = ep; 2299 } 2300 return 0; 2301 default: 2302 if (m->reln != 'x') { 2303 char *ep; 2304 m->value.q = file_signextend(ms, m, 2305 (uint64_t)strtoull(*p, &ep, 0)); 2306 *p = ep; 2307 eatsize(p); 2308 } 2309 return 0; 2310 } 2311 } 2312 2313 /* 2314 * Convert a string containing C character escapes. Stop at an unescaped 2315 * space or tab. 2316 * Copy the converted version to "m->value.s", and the length in m->vallen. 2317 * Return updated scan pointer as function result. Warn if set. 2318 */ 2319 private const char * 2320 getstr(struct magic_set *ms, struct magic *m, const char *s, int warn) 2321 { 2322 const char *origs = s; 2323 char *p = m->value.s; 2324 size_t plen = sizeof(m->value.s); 2325 char *origp = p; 2326 char *pmax = p + plen - 1; 2327 int c; 2328 int val; 2329 2330 while ((c = *s++) != '\0') { 2331 if (isspace((unsigned char) c)) 2332 break; 2333 if (p >= pmax) { 2334 file_error(ms, 0, "string too long: `%s'", origs); 2335 return NULL; 2336 } 2337 if (c == '\\') { 2338 switch(c = *s++) { 2339 2340 case '\0': 2341 if (warn) 2342 file_magwarn(ms, "incomplete escape"); 2343 goto out; 2344 2345 case '\t': 2346 if (warn) { 2347 file_magwarn(ms, 2348 "escaped tab found, use \\t instead"); 2349 warn = 0; /* already did */ 2350 } 2351 /*FALLTHROUGH*/ 2352 default: 2353 if (warn) { 2354 if (isprint((unsigned char)c)) { 2355 /* Allow escaping of 2356 * ``relations'' */ 2357 if (strchr("<>&^=!", c) == NULL 2358 && (m->type != FILE_REGEX || 2359 strchr("[]().*?^$|{}", c) 2360 == NULL)) { 2361 file_magwarn(ms, "no " 2362 "need to escape " 2363 "`%c'", c); 2364 } 2365 } else { 2366 file_magwarn(ms, 2367 "unknown escape sequence: " 2368 "\\%03o", c); 2369 } 2370 } 2371 /*FALLTHROUGH*/ 2372 /* space, perhaps force people to use \040? */ 2373 case ' ': 2374 #if 0 2375 /* 2376 * Other things people escape, but shouldn't need to, 2377 * so we disallow them 2378 */ 2379 case '\'': 2380 case '"': 2381 case '?': 2382 #endif 2383 /* Relations */ 2384 case '>': 2385 case '<': 2386 case '&': 2387 case '^': 2388 case '=': 2389 case '!': 2390 /* and baskslash itself */ 2391 case '\\': 2392 *p++ = (char) c; 2393 break; 2394 2395 case 'a': 2396 *p++ = '\a'; 2397 break; 2398 2399 case 'b': 2400 *p++ = '\b'; 2401 break; 2402 2403 case 'f': 2404 *p++ = '\f'; 2405 break; 2406 2407 case 'n': 2408 *p++ = '\n'; 2409 break; 2410 2411 case 'r': 2412 *p++ = '\r'; 2413 break; 2414 2415 case 't': 2416 *p++ = '\t'; 2417 break; 2418 2419 case 'v': 2420 *p++ = '\v'; 2421 break; 2422 2423 /* \ and up to 3 octal digits */ 2424 case '0': 2425 case '1': 2426 case '2': 2427 case '3': 2428 case '4': 2429 case '5': 2430 case '6': 2431 case '7': 2432 val = c - '0'; 2433 c = *s++; /* try for 2 */ 2434 if (c >= '0' && c <= '7') { 2435 val = (val << 3) | (c - '0'); 2436 c = *s++; /* try for 3 */ 2437 if (c >= '0' && c <= '7') 2438 val = (val << 3) | (c-'0'); 2439 else 2440 --s; 2441 } 2442 else 2443 --s; 2444 *p++ = (char)val; 2445 break; 2446 2447 /* \x and up to 2 hex digits */ 2448 case 'x': 2449 val = 'x'; /* Default if no digits */ 2450 c = hextoint(*s++); /* Get next char */ 2451 if (c >= 0) { 2452 val = c; 2453 c = hextoint(*s++); 2454 if (c >= 0) 2455 val = (val << 4) + c; 2456 else 2457 --s; 2458 } else 2459 --s; 2460 *p++ = (char)val; 2461 break; 2462 } 2463 } else 2464 *p++ = (char)c; 2465 } 2466 out: 2467 *p = '\0'; 2468 m->vallen = CAST(unsigned char, (p - origp)); 2469 if (m->type == FILE_PSTRING) 2470 m->vallen += (unsigned char)file_pstring_length_size(m); 2471 return s; 2472 } 2473 2474 2475 /* Single hex char to int; -1 if not a hex char. */ 2476 private int 2477 hextoint(int c) 2478 { 2479 if (!isascii((unsigned char) c)) 2480 return -1; 2481 if (isdigit((unsigned char) c)) 2482 return c - '0'; 2483 if ((c >= 'a') && (c <= 'f')) 2484 return c + 10 - 'a'; 2485 if (( c>= 'A') && (c <= 'F')) 2486 return c + 10 - 'A'; 2487 return -1; 2488 } 2489 2490 2491 /* 2492 * Print a string containing C character escapes. 2493 */ 2494 protected void 2495 file_showstr(FILE *fp, const char *s, size_t len) 2496 { 2497 char c; 2498 2499 for (;;) { 2500 if (len == ~0U) { 2501 c = *s++; 2502 if (c == '\0') 2503 break; 2504 } 2505 else { 2506 if (len-- == 0) 2507 break; 2508 c = *s++; 2509 } 2510 if (c >= 040 && c <= 0176) /* TODO isprint && !iscntrl */ 2511 (void) fputc(c, fp); 2512 else { 2513 (void) fputc('\\', fp); 2514 switch (c) { 2515 case '\a': 2516 (void) fputc('a', fp); 2517 break; 2518 2519 case '\b': 2520 (void) fputc('b', fp); 2521 break; 2522 2523 case '\f': 2524 (void) fputc('f', fp); 2525 break; 2526 2527 case '\n': 2528 (void) fputc('n', fp); 2529 break; 2530 2531 case '\r': 2532 (void) fputc('r', fp); 2533 break; 2534 2535 case '\t': 2536 (void) fputc('t', fp); 2537 break; 2538 2539 case '\v': 2540 (void) fputc('v', fp); 2541 break; 2542 2543 default: 2544 (void) fprintf(fp, "%.3o", c & 0377); 2545 break; 2546 } 2547 } 2548 } 2549 } 2550 2551 /* 2552 * eatsize(): Eat the size spec from a number [eg. 10UL] 2553 */ 2554 private void 2555 eatsize(const char **p) 2556 { 2557 const char *l = *p; 2558 2559 if (LOWCASE(*l) == 'u') 2560 l++; 2561 2562 switch (LOWCASE(*l)) { 2563 case 'l': /* long */ 2564 case 's': /* short */ 2565 case 'h': /* short */ 2566 case 'b': /* char/byte */ 2567 case 'c': /* char/byte */ 2568 l++; 2569 /*FALLTHROUGH*/ 2570 default: 2571 break; 2572 } 2573 2574 *p = l; 2575 } 2576 2577 /* 2578 * handle a compiled file. 2579 */ 2580 2581 private struct magic_map * 2582 apprentice_map(struct magic_set *ms, const char *fn) 2583 { 2584 int fd; 2585 struct stat st; 2586 uint32_t *ptr; 2587 uint32_t version, entries, nentries; 2588 int needsbyteswap; 2589 char *dbname = NULL; 2590 struct magic_map *map; 2591 size_t i; 2592 2593 fd = -1; 2594 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 2595 file_oomem(ms, sizeof(*map)); 2596 goto error; 2597 } 2598 2599 dbname = mkdbname(ms, fn, 0); 2600 if (dbname == NULL) 2601 goto error; 2602 2603 if ((fd = open(dbname, O_RDONLY|O_BINARY)) == -1) 2604 goto error; 2605 2606 if (fstat(fd, &st) == -1) { 2607 file_error(ms, errno, "cannot stat `%s'", dbname); 2608 goto error; 2609 } 2610 if (st.st_size < 8) { 2611 file_error(ms, 0, "file `%s' is too small", dbname); 2612 goto error; 2613 } 2614 2615 map->len = (size_t)st.st_size; 2616 #ifdef QUICK 2617 if ((map->p = mmap(0, (size_t)st.st_size, PROT_READ|PROT_WRITE, 2618 MAP_PRIVATE|MAP_FILE, fd, (off_t)0)) == MAP_FAILED) { 2619 file_error(ms, errno, "cannot map `%s'", dbname); 2620 goto error; 2621 } 2622 #else 2623 if ((map->p = CAST(void *, malloc(map->len))) == NULL) { 2624 file_oomem(ms, map->len); 2625 goto error; 2626 } 2627 if (read(fd, map->p, map->len) != (ssize_t)map->len) { 2628 file_badread(ms); 2629 goto error; 2630 } 2631 map->len = 0; 2632 #define RET 1 2633 #endif 2634 (void)close(fd); 2635 fd = -1; 2636 ptr = CAST(uint32_t *, map->p); 2637 if (*ptr != MAGICNO) { 2638 if (swap4(*ptr) != MAGICNO) { 2639 file_error(ms, 0, "bad magic in `%s'", dbname); 2640 goto error; 2641 } 2642 needsbyteswap = 1; 2643 } else 2644 needsbyteswap = 0; 2645 if (needsbyteswap) 2646 version = swap4(ptr[1]); 2647 else 2648 version = ptr[1]; 2649 if (version != VERSIONNO) { 2650 file_error(ms, 0, "File %s supports only version %d magic " 2651 "files. `%s' is version %d", VERSION, 2652 VERSIONNO, dbname, version); 2653 goto error; 2654 } 2655 entries = (uint32_t)(st.st_size / sizeof(struct magic)); 2656 if ((off_t)(entries * sizeof(struct magic)) != st.st_size) { 2657 file_error(ms, 0, "Size of `%s' %llu is not a multiple of %zu", 2658 dbname, (unsigned long long)st.st_size, 2659 sizeof(struct magic)); 2660 goto error; 2661 } 2662 map->magic[0] = CAST(struct magic *, map->p) + 1; 2663 nentries = 0; 2664 for (i = 0; i < MAGIC_SETS; i++) { 2665 if (needsbyteswap) 2666 map->nmagic[i] = swap4(ptr[i + 2]); 2667 else 2668 map->nmagic[i] = ptr[i + 2]; 2669 if (i != MAGIC_SETS - 1) 2670 map->magic[i + 1] = map->magic[i] + map->nmagic[i]; 2671 nentries += map->nmagic[i]; 2672 } 2673 if (entries != nentries + 1) { 2674 file_error(ms, 0, "Inconsistent entries in `%s' %u != %u", 2675 dbname, entries, nentries + 1); 2676 goto error; 2677 } 2678 if (needsbyteswap) 2679 for (i = 0; i < MAGIC_SETS; i++) 2680 byteswap(map->magic[i], map->nmagic[i]); 2681 free(dbname); 2682 return map; 2683 2684 error: 2685 if (fd != -1) 2686 (void)close(fd); 2687 apprentice_unmap(map); 2688 free(dbname); 2689 return NULL; 2690 } 2691 2692 private const uint32_t ar[] = { 2693 MAGICNO, VERSIONNO 2694 }; 2695 2696 /* 2697 * handle an mmaped file. 2698 */ 2699 private int 2700 apprentice_compile(struct magic_set *ms, struct magic_map *map, const char *fn) 2701 { 2702 static const size_t nm = sizeof(*map->nmagic) * MAGIC_SETS; 2703 static const size_t m = sizeof(**map->magic); 2704 int fd = -1; 2705 size_t len; 2706 char *dbname; 2707 int rv = -1; 2708 uint32_t i; 2709 2710 dbname = mkdbname(ms, fn, 1); 2711 2712 if (dbname == NULL) 2713 goto out; 2714 2715 if ((fd = open(dbname, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY, 0644)) == -1) 2716 { 2717 file_error(ms, errno, "cannot open `%s'", dbname); 2718 goto out; 2719 } 2720 2721 if (write(fd, ar, sizeof(ar)) != (ssize_t)sizeof(ar)) { 2722 file_error(ms, errno, "error writing `%s'", dbname); 2723 goto out; 2724 } 2725 2726 if (write(fd, map->nmagic, nm) != (ssize_t)nm) { 2727 file_error(ms, errno, "error writing `%s'", dbname); 2728 goto out; 2729 } 2730 2731 assert(nm + sizeof(ar) < m); 2732 2733 if (lseek(fd, (off_t)m, SEEK_SET) != (off_t)m) { 2734 file_error(ms, errno, "error seeking `%s'", dbname); 2735 goto out; 2736 } 2737 2738 for (i = 0; i < MAGIC_SETS; i++) { 2739 len = m * map->nmagic[i]; 2740 if (write(fd, map->magic[i], len) != (ssize_t)len) { 2741 file_error(ms, errno, "error writing `%s'", dbname); 2742 goto out; 2743 } 2744 } 2745 2746 if (fd != -1) 2747 (void)close(fd); 2748 rv = 0; 2749 out: 2750 free(dbname); 2751 return rv; 2752 } 2753 2754 private const char ext[] = ".mgc"; 2755 /* 2756 * make a dbname 2757 */ 2758 private char * 2759 mkdbname(struct magic_set *ms, const char *fn, int strip) 2760 { 2761 const char *p, *q; 2762 char *buf; 2763 2764 if (strip) { 2765 if ((p = strrchr(fn, '/')) != NULL) 2766 fn = ++p; 2767 } 2768 2769 for (q = fn; *q; q++) 2770 continue; 2771 /* Look for .mgc */ 2772 for (p = ext + sizeof(ext) - 1; p >= ext && q >= fn; p--, q--) 2773 if (*p != *q) 2774 break; 2775 2776 /* Did not find .mgc, restore q */ 2777 if (p >= ext) 2778 while (*q) 2779 q++; 2780 2781 q++; 2782 /* Compatibility with old code that looked in .mime */ 2783 if (ms->flags & MAGIC_MIME) { 2784 if (asprintf(&buf, "%.*s.mime%s", (int)(q - fn), fn, ext) < 0) 2785 return NULL; 2786 if (access(buf, R_OK) != -1) { 2787 ms->flags &= MAGIC_MIME_TYPE; 2788 return buf; 2789 } 2790 free(buf); 2791 } 2792 if (asprintf(&buf, "%.*s%s", (int)(q - fn), fn, ext) < 0) 2793 return NULL; 2794 2795 /* Compatibility with old code that looked in .mime */ 2796 if (strstr(p, ".mime") != NULL) 2797 ms->flags &= MAGIC_MIME_TYPE; 2798 return buf; 2799 } 2800 2801 /* 2802 * Byteswap an mmap'ed file if needed 2803 */ 2804 private void 2805 byteswap(struct magic *magic, uint32_t nmagic) 2806 { 2807 uint32_t i; 2808 for (i = 0; i < nmagic; i++) 2809 bs1(&magic[i]); 2810 } 2811 2812 /* 2813 * swap a short 2814 */ 2815 private uint16_t 2816 swap2(uint16_t sv) 2817 { 2818 uint16_t rv; 2819 uint8_t *s = (uint8_t *)(void *)&sv; 2820 uint8_t *d = (uint8_t *)(void *)&rv; 2821 d[0] = s[1]; 2822 d[1] = s[0]; 2823 return rv; 2824 } 2825 2826 /* 2827 * swap an int 2828 */ 2829 private uint32_t 2830 swap4(uint32_t sv) 2831 { 2832 uint32_t rv; 2833 uint8_t *s = (uint8_t *)(void *)&sv; 2834 uint8_t *d = (uint8_t *)(void *)&rv; 2835 d[0] = s[3]; 2836 d[1] = s[2]; 2837 d[2] = s[1]; 2838 d[3] = s[0]; 2839 return rv; 2840 } 2841 2842 /* 2843 * swap a quad 2844 */ 2845 private uint64_t 2846 swap8(uint64_t sv) 2847 { 2848 uint64_t rv; 2849 uint8_t *s = (uint8_t *)(void *)&sv; 2850 uint8_t *d = (uint8_t *)(void *)&rv; 2851 #if 0 2852 d[0] = s[3]; 2853 d[1] = s[2]; 2854 d[2] = s[1]; 2855 d[3] = s[0]; 2856 d[4] = s[7]; 2857 d[5] = s[6]; 2858 d[6] = s[5]; 2859 d[7] = s[4]; 2860 #else 2861 d[0] = s[7]; 2862 d[1] = s[6]; 2863 d[2] = s[5]; 2864 d[3] = s[4]; 2865 d[4] = s[3]; 2866 d[5] = s[2]; 2867 d[6] = s[1]; 2868 d[7] = s[0]; 2869 #endif 2870 return rv; 2871 } 2872 2873 /* 2874 * byteswap a single magic entry 2875 */ 2876 private void 2877 bs1(struct magic *m) 2878 { 2879 m->cont_level = swap2(m->cont_level); 2880 m->offset = swap4((uint32_t)m->offset); 2881 m->in_offset = swap4((uint32_t)m->in_offset); 2882 m->lineno = swap4((uint32_t)m->lineno); 2883 if (IS_STRING(m->type)) { 2884 m->str_range = swap4(m->str_range); 2885 m->str_flags = swap4(m->str_flags); 2886 } 2887 else { 2888 m->value.q = swap8(m->value.q); 2889 m->num_mask = swap8(m->num_mask); 2890 } 2891 } 2892 2893 protected size_t 2894 file_pstring_length_size(const struct magic *m) 2895 { 2896 switch (m->str_flags & PSTRING_LEN) { 2897 case PSTRING_1_LE: 2898 return 1; 2899 case PSTRING_2_LE: 2900 case PSTRING_2_BE: 2901 return 2; 2902 case PSTRING_4_LE: 2903 case PSTRING_4_BE: 2904 return 4; 2905 default: 2906 abort(); /* Impossible */ 2907 return 1; 2908 } 2909 } 2910 protected size_t 2911 file_pstring_get_length(const struct magic *m, const char *s) 2912 { 2913 size_t len = 0; 2914 2915 switch (m->str_flags & PSTRING_LEN) { 2916 case PSTRING_1_LE: 2917 len = *s; 2918 break; 2919 case PSTRING_2_LE: 2920 len = (s[1] << 8) | s[0]; 2921 break; 2922 case PSTRING_2_BE: 2923 len = (s[0] << 8) | s[1]; 2924 break; 2925 case PSTRING_4_LE: 2926 len = (s[3] << 24) | (s[2] << 16) | (s[1] << 8) | s[0]; 2927 break; 2928 case PSTRING_4_BE: 2929 len = (s[0] << 24) | (s[1] << 16) | (s[2] << 8) | s[3]; 2930 break; 2931 default: 2932 abort(); /* Impossible */ 2933 } 2934 2935 if (m->str_flags & PSTRING_LENGTH_INCLUDES_ITSELF) 2936 len -= file_pstring_length_size(m); 2937 2938 return len; 2939 } 2940 2941 protected int 2942 file_magicfind(struct magic_set *ms, const char *name, struct mlist *v) 2943 { 2944 uint32_t i, j; 2945 struct mlist *mlist, *ml; 2946 2947 mlist = ms->mlist[1]; 2948 2949 for (ml = mlist->next; ml != mlist; ml = ml->next) { 2950 struct magic *ma = ml->magic; 2951 uint32_t nma = ml->nmagic; 2952 for (i = 0; i < nma; i++) { 2953 if (ma[i].type != FILE_NAME) 2954 continue; 2955 if (strcmp(ma[i].value.s, name) == 0) { 2956 v->magic = &ma[i]; 2957 for (j = i + 1; j < nma; j++) 2958 if (ma[j].cont_level == 0) 2959 break; 2960 v->nmagic = j - i; 2961 return 0; 2962 } 2963 } 2964 } 2965 return -1; 2966 } 2967