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 1.249 2016/05/17 21:43:07 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 #define MAP_TYPE_USER 0 90 #define MAP_TYPE_MALLOC 1 91 #define MAP_TYPE_MMAP 2 92 93 struct magic_entry { 94 struct magic *mp; 95 uint32_t cont_count; 96 uint32_t max_count; 97 }; 98 99 struct magic_entry_set { 100 struct magic_entry *me; 101 uint32_t count; 102 uint32_t max; 103 }; 104 105 struct magic_map { 106 void *p; 107 size_t len; 108 int type; 109 struct magic *magic[MAGIC_SETS]; 110 uint32_t nmagic[MAGIC_SETS]; 111 }; 112 113 int file_formats[FILE_NAMES_SIZE]; 114 const size_t file_nformats = FILE_NAMES_SIZE; 115 const char *file_names[FILE_NAMES_SIZE]; 116 const size_t file_nnames = FILE_NAMES_SIZE; 117 118 private int getvalue(struct magic_set *ms, struct magic *, const char **, int); 119 private int hextoint(int); 120 private const char *getstr(struct magic_set *, struct magic *, const char *, 121 int); 122 private int parse(struct magic_set *, struct magic_entry *, const char *, 123 size_t, int); 124 private void eatsize(const char **); 125 private int apprentice_1(struct magic_set *, const char *, int); 126 private size_t apprentice_magic_strength(const struct magic *); 127 private int apprentice_sort(const void *, const void *); 128 private void apprentice_list(struct mlist *, int ); 129 private struct magic_map *apprentice_load(struct magic_set *, 130 const char *, int); 131 private struct mlist *mlist_alloc(void); 132 private void mlist_free(struct mlist *); 133 private void byteswap(struct magic *, uint32_t); 134 private void bs1(struct magic *); 135 private uint16_t swap2(uint16_t); 136 private uint32_t swap4(uint32_t); 137 private uint64_t swap8(uint64_t); 138 private char *mkdbname(struct magic_set *, const char *, int); 139 private struct magic_map *apprentice_buf(struct magic_set *, struct magic *, 140 size_t); 141 private struct magic_map *apprentice_map(struct magic_set *, const char *); 142 private int check_buffer(struct magic_set *, struct magic_map *, const char *); 143 private void apprentice_unmap(struct magic_map *); 144 private int apprentice_compile(struct magic_set *, struct magic_map *, 145 const char *); 146 private int check_format_type(const char *, int, const char **); 147 private int check_format(struct magic_set *, struct magic *); 148 private int get_op(char); 149 private int parse_mime(struct magic_set *, struct magic_entry *, const char *); 150 private int parse_strength(struct magic_set *, struct magic_entry *, const char *); 151 private int parse_apple(struct magic_set *, struct magic_entry *, const char *); 152 private int parse_ext(struct magic_set *, struct magic_entry *, const char *); 153 154 155 private size_t magicsize = sizeof(struct magic); 156 157 private const char usg_hdr[] = "cont\toffset\ttype\topcode\tmask\tvalue\tdesc"; 158 159 private struct { 160 const char *name; 161 size_t len; 162 int (*fun)(struct magic_set *, struct magic_entry *, const char *); 163 } bang[] = { 164 #define DECLARE_FIELD(name) { # name, sizeof(# name) - 1, parse_ ## name } 165 DECLARE_FIELD(mime), 166 DECLARE_FIELD(apple), 167 DECLARE_FIELD(ext), 168 DECLARE_FIELD(strength), 169 #undef DECLARE_FIELD 170 { NULL, 0, NULL } 171 }; 172 173 #ifdef COMPILE_ONLY 174 175 int main(int, char *[]); 176 177 int 178 main(int argc, char *argv[]) 179 { 180 int ret; 181 struct magic_set *ms; 182 char *progname; 183 184 if ((progname = strrchr(argv[0], '/')) != NULL) 185 progname++; 186 else 187 progname = argv[0]; 188 189 if (argc != 2) { 190 (void)fprintf(stderr, "Usage: %s file\n", progname); 191 return 1; 192 } 193 194 if ((ms = magic_open(MAGIC_CHECK)) == NULL) { 195 (void)fprintf(stderr, "%s: %s\n", progname, strerror(errno)); 196 return 1; 197 } 198 ret = magic_compile(ms, argv[1]) == -1 ? 1 : 0; 199 if (ret == 1) 200 (void)fprintf(stderr, "%s: %s\n", progname, magic_error(ms)); 201 magic_close(ms); 202 return ret; 203 } 204 #endif /* COMPILE_ONLY */ 205 206 struct type_tbl_s { 207 const char name[16]; 208 const size_t len; 209 const int type; 210 const int format; 211 }; 212 213 /* 214 * XXX - the actual Single UNIX Specification says that "long" means "long", 215 * as in the C data type, but we treat it as meaning "4-byte integer". 216 * Given that the OS X version of file 5.04 did the same, I guess that passes 217 * the actual test; having "long" be dependent on how big a "long" is on 218 * the machine running "file" is silly. 219 */ 220 static const struct type_tbl_s type_tbl[] = { 221 # define XX(s) s, (sizeof(s) - 1) 222 # define XX_NULL "", 0 223 { XX("invalid"), FILE_INVALID, FILE_FMT_NONE }, 224 { XX("byte"), FILE_BYTE, FILE_FMT_NUM }, 225 { XX("short"), FILE_SHORT, FILE_FMT_NUM }, 226 { XX("default"), FILE_DEFAULT, FILE_FMT_NONE }, 227 { XX("long"), FILE_LONG, FILE_FMT_NUM }, 228 { XX("string"), FILE_STRING, FILE_FMT_STR }, 229 { XX("date"), FILE_DATE, FILE_FMT_STR }, 230 { XX("beshort"), FILE_BESHORT, FILE_FMT_NUM }, 231 { XX("belong"), FILE_BELONG, FILE_FMT_NUM }, 232 { XX("bedate"), FILE_BEDATE, FILE_FMT_STR }, 233 { XX("leshort"), FILE_LESHORT, FILE_FMT_NUM }, 234 { XX("lelong"), FILE_LELONG, FILE_FMT_NUM }, 235 { XX("ledate"), FILE_LEDATE, FILE_FMT_STR }, 236 { XX("pstring"), FILE_PSTRING, FILE_FMT_STR }, 237 { XX("ldate"), FILE_LDATE, FILE_FMT_STR }, 238 { XX("beldate"), FILE_BELDATE, FILE_FMT_STR }, 239 { XX("leldate"), FILE_LELDATE, FILE_FMT_STR }, 240 { XX("regex"), FILE_REGEX, FILE_FMT_STR }, 241 { XX("bestring16"), FILE_BESTRING16, FILE_FMT_STR }, 242 { XX("lestring16"), FILE_LESTRING16, FILE_FMT_STR }, 243 { XX("search"), FILE_SEARCH, FILE_FMT_STR }, 244 { XX("medate"), FILE_MEDATE, FILE_FMT_STR }, 245 { XX("meldate"), FILE_MELDATE, FILE_FMT_STR }, 246 { XX("melong"), FILE_MELONG, FILE_FMT_NUM }, 247 { XX("quad"), FILE_QUAD, FILE_FMT_QUAD }, 248 { XX("lequad"), FILE_LEQUAD, FILE_FMT_QUAD }, 249 { XX("bequad"), FILE_BEQUAD, FILE_FMT_QUAD }, 250 { XX("qdate"), FILE_QDATE, FILE_FMT_STR }, 251 { XX("leqdate"), FILE_LEQDATE, FILE_FMT_STR }, 252 { XX("beqdate"), FILE_BEQDATE, FILE_FMT_STR }, 253 { XX("qldate"), FILE_QLDATE, FILE_FMT_STR }, 254 { XX("leqldate"), FILE_LEQLDATE, FILE_FMT_STR }, 255 { XX("beqldate"), FILE_BEQLDATE, FILE_FMT_STR }, 256 { XX("float"), FILE_FLOAT, FILE_FMT_FLOAT }, 257 { XX("befloat"), FILE_BEFLOAT, FILE_FMT_FLOAT }, 258 { XX("lefloat"), FILE_LEFLOAT, FILE_FMT_FLOAT }, 259 { XX("double"), FILE_DOUBLE, FILE_FMT_DOUBLE }, 260 { XX("bedouble"), FILE_BEDOUBLE, FILE_FMT_DOUBLE }, 261 { XX("ledouble"), FILE_LEDOUBLE, FILE_FMT_DOUBLE }, 262 { XX("leid3"), FILE_LEID3, FILE_FMT_NUM }, 263 { XX("beid3"), FILE_BEID3, FILE_FMT_NUM }, 264 { XX("indirect"), FILE_INDIRECT, FILE_FMT_NUM }, 265 { XX("qwdate"), FILE_QWDATE, FILE_FMT_STR }, 266 { XX("leqwdate"), FILE_LEQWDATE, FILE_FMT_STR }, 267 { XX("beqwdate"), FILE_BEQWDATE, FILE_FMT_STR }, 268 { XX("name"), FILE_NAME, FILE_FMT_NONE }, 269 { XX("use"), FILE_USE, FILE_FMT_NONE }, 270 { XX("clear"), FILE_CLEAR, FILE_FMT_NONE }, 271 { XX("der"), FILE_DER, FILE_FMT_STR }, 272 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 273 }; 274 275 /* 276 * These are not types, and cannot be preceded by "u" to make them 277 * unsigned. 278 */ 279 static const struct type_tbl_s special_tbl[] = { 280 { XX("der"), FILE_DER, FILE_FMT_STR }, 281 { XX("name"), FILE_NAME, FILE_FMT_STR }, 282 { XX("use"), FILE_USE, FILE_FMT_STR }, 283 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 284 }; 285 # undef XX 286 # undef XX_NULL 287 288 private int 289 get_type(const struct type_tbl_s *tbl, const char *l, const char **t) 290 { 291 const struct type_tbl_s *p; 292 293 for (p = tbl; p->len; p++) { 294 if (strncmp(l, p->name, p->len) == 0) { 295 if (t) 296 *t = l + p->len; 297 break; 298 } 299 } 300 return p->type; 301 } 302 303 private int 304 get_standard_integer_type(const char *l, const char **t) 305 { 306 int type; 307 308 if (isalpha((unsigned char)l[1])) { 309 switch (l[1]) { 310 case 'C': 311 /* "dC" and "uC" */ 312 type = FILE_BYTE; 313 break; 314 case 'S': 315 /* "dS" and "uS" */ 316 type = FILE_SHORT; 317 break; 318 case 'I': 319 case 'L': 320 /* 321 * "dI", "dL", "uI", and "uL". 322 * 323 * XXX - the actual Single UNIX Specification says 324 * that "L" means "long", as in the C data type, 325 * but we treat it as meaning "4-byte integer". 326 * Given that the OS X version of file 5.04 did 327 * the same, I guess that passes the actual SUS 328 * validation suite; having "dL" be dependent on 329 * how big a "long" is on the machine running 330 * "file" is silly. 331 */ 332 type = FILE_LONG; 333 break; 334 case 'Q': 335 /* "dQ" and "uQ" */ 336 type = FILE_QUAD; 337 break; 338 default: 339 /* "d{anything else}", "u{anything else}" */ 340 return FILE_INVALID; 341 } 342 l += 2; 343 } else if (isdigit((unsigned char)l[1])) { 344 /* 345 * "d{num}" and "u{num}"; we only support {num} values 346 * of 1, 2, 4, and 8 - the Single UNIX Specification 347 * doesn't say anything about whether arbitrary 348 * values should be supported, but both the Solaris 10 349 * and OS X Mountain Lion versions of file passed the 350 * Single UNIX Specification validation suite, and 351 * neither of them support values bigger than 8 or 352 * non-power-of-2 values. 353 */ 354 if (isdigit((unsigned char)l[2])) { 355 /* Multi-digit, so > 9 */ 356 return FILE_INVALID; 357 } 358 switch (l[1]) { 359 case '1': 360 type = FILE_BYTE; 361 break; 362 case '2': 363 type = FILE_SHORT; 364 break; 365 case '4': 366 type = FILE_LONG; 367 break; 368 case '8': 369 type = FILE_QUAD; 370 break; 371 default: 372 /* XXX - what about 3, 5, 6, or 7? */ 373 return FILE_INVALID; 374 } 375 l += 2; 376 } else { 377 /* 378 * "d" or "u" by itself. 379 */ 380 type = FILE_LONG; 381 ++l; 382 } 383 if (t) 384 *t = l; 385 return type; 386 } 387 388 private void 389 init_file_tables(void) 390 { 391 static int done = 0; 392 const struct type_tbl_s *p; 393 394 if (done) 395 return; 396 done++; 397 398 for (p = type_tbl; p->len; p++) { 399 assert(p->type < FILE_NAMES_SIZE); 400 file_names[p->type] = p->name; 401 file_formats[p->type] = p->format; 402 } 403 assert(p - type_tbl == FILE_NAMES_SIZE); 404 } 405 406 private int 407 add_mlist(struct mlist *mlp, struct magic_map *map, size_t idx) 408 { 409 struct mlist *ml; 410 411 mlp->map = idx == 0 ? map : NULL; 412 if ((ml = CAST(struct mlist *, malloc(sizeof(*ml)))) == NULL) 413 return -1; 414 415 ml->map = NULL; 416 ml->magic = map->magic[idx]; 417 ml->nmagic = map->nmagic[idx]; 418 419 mlp->prev->next = ml; 420 ml->prev = mlp->prev; 421 ml->next = mlp; 422 mlp->prev = ml; 423 return 0; 424 } 425 426 /* 427 * Handle one file or directory. 428 */ 429 private int 430 apprentice_1(struct magic_set *ms, const char *fn, int action) 431 { 432 struct magic_map *map; 433 #ifndef COMPILE_ONLY 434 struct mlist *ml; 435 size_t i; 436 #endif 437 438 if (magicsize != FILE_MAGICSIZE) { 439 file_error(ms, 0, "magic element size %lu != %lu", 440 (unsigned long)sizeof(*map->magic[0]), 441 (unsigned long)FILE_MAGICSIZE); 442 return -1; 443 } 444 445 if (action == FILE_COMPILE) { 446 map = apprentice_load(ms, fn, action); 447 if (map == NULL) 448 return -1; 449 return apprentice_compile(ms, map, fn); 450 } 451 452 #ifndef COMPILE_ONLY 453 map = apprentice_map(ms, fn); 454 if (map == NULL) { 455 if (ms->flags & MAGIC_CHECK) 456 file_magwarn(ms, "using regular magic file `%s'", fn); 457 map = apprentice_load(ms, fn, action); 458 if (map == NULL) 459 return -1; 460 } 461 462 for (i = 0; i < MAGIC_SETS; i++) { 463 if (add_mlist(ms->mlist[i], map, i) == -1) { 464 file_oomem(ms, sizeof(*ml)); 465 goto fail; 466 } 467 } 468 469 if (action == FILE_LIST) { 470 for (i = 0; i < MAGIC_SETS; i++) { 471 printf("Set %" SIZE_T_FORMAT "u:\nBinary patterns:\n", 472 i); 473 apprentice_list(ms->mlist[i], BINTEST); 474 printf("Text patterns:\n"); 475 apprentice_list(ms->mlist[i], TEXTTEST); 476 } 477 } 478 return 0; 479 fail: 480 for (i = 0; i < MAGIC_SETS; i++) { 481 mlist_free(ms->mlist[i]); 482 ms->mlist[i] = NULL; 483 } 484 return -1; 485 #else 486 return 0; 487 #endif /* COMPILE_ONLY */ 488 } 489 490 protected void 491 file_ms_free(struct magic_set *ms) 492 { 493 size_t i; 494 if (ms == NULL) 495 return; 496 for (i = 0; i < MAGIC_SETS; i++) 497 mlist_free(ms->mlist[i]); 498 free(ms->o.pbuf); 499 free(ms->o.buf); 500 free(ms->c.li); 501 free(ms); 502 } 503 504 protected struct magic_set * 505 file_ms_alloc(int flags) 506 { 507 struct magic_set *ms; 508 size_t i, len; 509 510 if ((ms = CAST(struct magic_set *, calloc((size_t)1, 511 sizeof(struct magic_set)))) == NULL) 512 return NULL; 513 514 if (magic_setflags(ms, flags) == -1) { 515 errno = EINVAL; 516 goto free; 517 } 518 519 ms->o.buf = ms->o.pbuf = NULL; 520 len = (ms->c.len = 10) * sizeof(*ms->c.li); 521 522 if ((ms->c.li = CAST(struct level_info *, malloc(len))) == NULL) 523 goto free; 524 525 ms->event_flags = 0; 526 ms->error = -1; 527 for (i = 0; i < MAGIC_SETS; i++) 528 ms->mlist[i] = NULL; 529 ms->file = "unknown"; 530 ms->line = 0; 531 ms->indir_max = FILE_INDIR_MAX; 532 ms->name_max = FILE_NAME_MAX; 533 ms->elf_shnum_max = FILE_ELF_SHNUM_MAX; 534 ms->elf_phnum_max = FILE_ELF_PHNUM_MAX; 535 ms->elf_notes_max = FILE_ELF_NOTES_MAX; 536 ms->regex_max = FILE_REGEX_MAX; 537 ms->bytes_max = FILE_BYTES_MAX; 538 return ms; 539 free: 540 free(ms); 541 return NULL; 542 } 543 544 private void 545 apprentice_unmap(struct magic_map *map) 546 { 547 size_t i; 548 if (map == NULL) 549 return; 550 551 switch (map->type) { 552 case MAP_TYPE_USER: 553 break; 554 case MAP_TYPE_MALLOC: 555 for (i = 0; i < MAGIC_SETS; i++) { 556 if ((char *)map->magic[i] >= (char *)map->p && 557 (char *)map->magic[i] < (char *)map->p + map->len) 558 continue; 559 free(map->magic[i]); 560 } 561 free(map->p); 562 break; 563 #ifdef QUICK 564 case MAP_TYPE_MMAP: 565 if (map->p && map->p != MAP_FAILED) 566 (void)munmap(map->p, map->len); 567 break; 568 #endif 569 default: 570 abort(); 571 } 572 free(map); 573 } 574 575 private struct mlist * 576 mlist_alloc(void) 577 { 578 struct mlist *mlist; 579 if ((mlist = CAST(struct mlist *, calloc(1, sizeof(*mlist)))) == NULL) { 580 return NULL; 581 } 582 mlist->next = mlist->prev = mlist; 583 return mlist; 584 } 585 586 private void 587 mlist_free(struct mlist *mlist) 588 { 589 struct mlist *ml, *next; 590 591 if (mlist == NULL) 592 return; 593 594 ml = mlist->next; 595 for (ml = mlist->next; (next = ml->next) != NULL; ml = next) { 596 if (ml->map) 597 apprentice_unmap(ml->map); 598 free(ml); 599 if (ml == mlist) 600 break; 601 } 602 } 603 604 #ifndef COMPILE_ONLY 605 /* void **bufs: an array of compiled magic files */ 606 protected int 607 buffer_apprentice(struct magic_set *ms, struct magic **bufs, 608 size_t *sizes, size_t nbufs) 609 { 610 size_t i, j; 611 struct mlist *ml; 612 struct magic_map *map; 613 614 if (nbufs == 0) 615 return -1; 616 617 if (ms->mlist[0] != NULL) 618 file_reset(ms); 619 620 init_file_tables(); 621 622 for (i = 0; i < MAGIC_SETS; i++) { 623 mlist_free(ms->mlist[i]); 624 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 625 file_oomem(ms, sizeof(*ms->mlist[i])); 626 goto fail; 627 } 628 } 629 630 for (i = 0; i < nbufs; i++) { 631 map = apprentice_buf(ms, bufs[i], sizes[i]); 632 if (map == NULL) 633 goto fail; 634 635 for (j = 0; j < MAGIC_SETS; j++) { 636 if (add_mlist(ms->mlist[j], map, j) == -1) { 637 file_oomem(ms, sizeof(*ml)); 638 goto fail; 639 } 640 } 641 } 642 643 return 0; 644 fail: 645 for (i = 0; i < MAGIC_SETS; i++) { 646 mlist_free(ms->mlist[i]); 647 ms->mlist[i] = NULL; 648 } 649 return -1; 650 } 651 #endif 652 653 /* const char *fn: list of magic files and directories */ 654 protected int 655 file_apprentice(struct magic_set *ms, const char *fn, int action) 656 { 657 char *p, *mfn; 658 int file_err, errs = -1; 659 size_t i; 660 661 if (ms->mlist[0] != NULL) 662 file_reset(ms); 663 664 if ((fn = magic_getpath(fn, action)) == NULL) 665 return -1; 666 667 init_file_tables(); 668 669 if ((mfn = strdup(fn)) == NULL) { 670 file_oomem(ms, strlen(fn)); 671 return -1; 672 } 673 674 for (i = 0; i < MAGIC_SETS; i++) { 675 mlist_free(ms->mlist[i]); 676 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 677 file_oomem(ms, sizeof(*ms->mlist[i])); 678 while (i-- > 0) { 679 mlist_free(ms->mlist[i]); 680 ms->mlist[i] = NULL; 681 } 682 free(mfn); 683 return -1; 684 } 685 } 686 fn = mfn; 687 688 while (fn) { 689 p = strchr(fn, PATHSEP); 690 if (p) 691 *p++ = '\0'; 692 if (*fn == '\0') 693 break; 694 file_err = apprentice_1(ms, fn, action); 695 errs = MAX(errs, file_err); 696 fn = p; 697 } 698 699 free(mfn); 700 701 if (errs == -1) { 702 for (i = 0; i < MAGIC_SETS; i++) { 703 mlist_free(ms->mlist[i]); 704 ms->mlist[i] = NULL; 705 } 706 file_error(ms, 0, "could not find any valid magic files!"); 707 return -1; 708 } 709 710 #if 0 711 /* 712 * Always leave the database loaded 713 */ 714 if (action == FILE_LOAD) 715 return 0; 716 717 for (i = 0; i < MAGIC_SETS; i++) { 718 mlist_free(ms->mlist[i]); 719 ms->mlist[i] = NULL; 720 } 721 #endif 722 723 switch (action) { 724 case FILE_LOAD: 725 case FILE_COMPILE: 726 case FILE_CHECK: 727 case FILE_LIST: 728 return 0; 729 default: 730 file_error(ms, 0, "Invalid action %d", action); 731 return -1; 732 } 733 } 734 735 /* 736 * Compute the real length of a magic expression, for the purposes 737 * of determining how "strong" a magic expression is (approximating 738 * how specific its matches are): 739 * - magic characters count 0 unless escaped. 740 * - [] expressions count 1 741 * - {} expressions count 0 742 * - regular characters or escaped magic characters count 1 743 * - 0 length expressions count as one 744 */ 745 private size_t 746 nonmagic(const char *str) 747 { 748 const char *p; 749 size_t rv = 0; 750 751 for (p = str; *p; p++) 752 switch (*p) { 753 case '\\': /* Escaped anything counts 1 */ 754 if (!*++p) 755 p--; 756 rv++; 757 continue; 758 case '?': /* Magic characters count 0 */ 759 case '*': 760 case '.': 761 case '+': 762 case '^': 763 case '$': 764 continue; 765 case '[': /* Bracketed expressions count 1 the ']' */ 766 while (*p && *p != ']') 767 p++; 768 p--; 769 continue; 770 case '{': /* Braced expressions count 0 */ 771 while (*p && *p != '}') 772 p++; 773 if (!*p) 774 p--; 775 continue; 776 default: /* Anything else counts 1 */ 777 rv++; 778 continue; 779 } 780 781 return rv == 0 ? 1 : rv; /* Return at least 1 */ 782 } 783 784 /* 785 * Get weight of this magic entry, for sorting purposes. 786 */ 787 private size_t 788 apprentice_magic_strength(const struct magic *m) 789 { 790 #define MULT 10 791 size_t v, val = 2 * MULT; /* baseline strength */ 792 793 switch (m->type) { 794 case FILE_DEFAULT: /* make sure this sorts last */ 795 if (m->factor_op != FILE_FACTOR_OP_NONE) 796 abort(); 797 return 0; 798 799 case FILE_BYTE: 800 val += 1 * MULT; 801 break; 802 803 case FILE_SHORT: 804 case FILE_LESHORT: 805 case FILE_BESHORT: 806 val += 2 * MULT; 807 break; 808 809 case FILE_LONG: 810 case FILE_LELONG: 811 case FILE_BELONG: 812 case FILE_MELONG: 813 val += 4 * MULT; 814 break; 815 816 case FILE_PSTRING: 817 case FILE_STRING: 818 val += m->vallen * MULT; 819 break; 820 821 case FILE_BESTRING16: 822 case FILE_LESTRING16: 823 val += m->vallen * MULT / 2; 824 break; 825 826 case FILE_SEARCH: 827 val += m->vallen * MAX(MULT / m->vallen, 1); 828 break; 829 830 case FILE_REGEX: 831 v = nonmagic(m->value.s); 832 val += v * MAX(MULT / v, 1); 833 break; 834 835 case FILE_DATE: 836 case FILE_LEDATE: 837 case FILE_BEDATE: 838 case FILE_MEDATE: 839 case FILE_LDATE: 840 case FILE_LELDATE: 841 case FILE_BELDATE: 842 case FILE_MELDATE: 843 case FILE_FLOAT: 844 case FILE_BEFLOAT: 845 case FILE_LEFLOAT: 846 val += 4 * MULT; 847 break; 848 849 case FILE_QUAD: 850 case FILE_BEQUAD: 851 case FILE_LEQUAD: 852 case FILE_QDATE: 853 case FILE_LEQDATE: 854 case FILE_BEQDATE: 855 case FILE_QLDATE: 856 case FILE_LEQLDATE: 857 case FILE_BEQLDATE: 858 case FILE_QWDATE: 859 case FILE_LEQWDATE: 860 case FILE_BEQWDATE: 861 case FILE_DOUBLE: 862 case FILE_BEDOUBLE: 863 case FILE_LEDOUBLE: 864 val += 8 * MULT; 865 break; 866 867 case FILE_INDIRECT: 868 case FILE_NAME: 869 case FILE_USE: 870 break; 871 872 case FILE_DER: 873 val += MULT; 874 break; 875 876 default: 877 (void)fprintf(stderr, "Bad type %d\n", m->type); 878 abort(); 879 } 880 881 switch (m->reln) { 882 case 'x': /* matches anything penalize */ 883 case '!': /* matches almost anything penalize */ 884 val = 0; 885 break; 886 887 case '=': /* Exact match, prefer */ 888 val += MULT; 889 break; 890 891 case '>': 892 case '<': /* comparison match reduce strength */ 893 val -= 2 * MULT; 894 break; 895 896 case '^': 897 case '&': /* masking bits, we could count them too */ 898 val -= MULT; 899 break; 900 901 default: 902 (void)fprintf(stderr, "Bad relation %c\n", m->reln); 903 abort(); 904 } 905 906 if (val == 0) /* ensure we only return 0 for FILE_DEFAULT */ 907 val = 1; 908 909 switch (m->factor_op) { 910 case FILE_FACTOR_OP_NONE: 911 break; 912 case FILE_FACTOR_OP_PLUS: 913 val += m->factor; 914 break; 915 case FILE_FACTOR_OP_MINUS: 916 val -= m->factor; 917 break; 918 case FILE_FACTOR_OP_TIMES: 919 val *= m->factor; 920 break; 921 case FILE_FACTOR_OP_DIV: 922 val /= m->factor; 923 break; 924 default: 925 abort(); 926 } 927 928 /* 929 * Magic entries with no description get a bonus because they depend 930 * on subsequent magic entries to print something. 931 */ 932 if (m->desc[0] == '\0') 933 val++; 934 return val; 935 } 936 937 /* 938 * Sort callback for sorting entries by "strength" (basically length) 939 */ 940 private int 941 apprentice_sort(const void *a, const void *b) 942 { 943 const struct magic_entry *ma = CAST(const struct magic_entry *, a); 944 const struct magic_entry *mb = CAST(const struct magic_entry *, b); 945 size_t sa = apprentice_magic_strength(ma->mp); 946 size_t sb = apprentice_magic_strength(mb->mp); 947 if (sa == sb) 948 return 0; 949 else if (sa > sb) 950 return -1; 951 else 952 return 1; 953 } 954 955 /* 956 * Shows sorted patterns list in the order which is used for the matching 957 */ 958 private void 959 apprentice_list(struct mlist *mlist, int mode) 960 { 961 uint32_t magindex = 0; 962 struct mlist *ml; 963 for (ml = mlist->next; ml != mlist; ml = ml->next) { 964 for (magindex = 0; magindex < ml->nmagic; magindex++) { 965 struct magic *m = &ml->magic[magindex]; 966 if ((m->flag & mode) != mode) { 967 /* Skip sub-tests */ 968 while (magindex + 1 < ml->nmagic && 969 ml->magic[magindex + 1].cont_level != 0) 970 ++magindex; 971 continue; /* Skip to next top-level test*/ 972 } 973 974 /* 975 * Try to iterate over the tree until we find item with 976 * description/mimetype. 977 */ 978 while (magindex + 1 < ml->nmagic && 979 ml->magic[magindex + 1].cont_level != 0 && 980 *ml->magic[magindex].desc == '\0' && 981 *ml->magic[magindex].mimetype == '\0') 982 magindex++; 983 984 printf("Strength = %3" SIZE_T_FORMAT "u@%u: %s [%s]\n", 985 apprentice_magic_strength(m), 986 ml->magic[magindex].lineno, 987 ml->magic[magindex].desc, 988 ml->magic[magindex].mimetype); 989 } 990 } 991 } 992 993 private void 994 set_test_type(struct magic *mstart, struct magic *m) 995 { 996 switch (m->type) { 997 case FILE_BYTE: 998 case FILE_SHORT: 999 case FILE_LONG: 1000 case FILE_DATE: 1001 case FILE_BESHORT: 1002 case FILE_BELONG: 1003 case FILE_BEDATE: 1004 case FILE_LESHORT: 1005 case FILE_LELONG: 1006 case FILE_LEDATE: 1007 case FILE_LDATE: 1008 case FILE_BELDATE: 1009 case FILE_LELDATE: 1010 case FILE_MEDATE: 1011 case FILE_MELDATE: 1012 case FILE_MELONG: 1013 case FILE_QUAD: 1014 case FILE_LEQUAD: 1015 case FILE_BEQUAD: 1016 case FILE_QDATE: 1017 case FILE_LEQDATE: 1018 case FILE_BEQDATE: 1019 case FILE_QLDATE: 1020 case FILE_LEQLDATE: 1021 case FILE_BEQLDATE: 1022 case FILE_QWDATE: 1023 case FILE_LEQWDATE: 1024 case FILE_BEQWDATE: 1025 case FILE_FLOAT: 1026 case FILE_BEFLOAT: 1027 case FILE_LEFLOAT: 1028 case FILE_DOUBLE: 1029 case FILE_BEDOUBLE: 1030 case FILE_LEDOUBLE: 1031 case FILE_DER: 1032 mstart->flag |= BINTEST; 1033 break; 1034 case FILE_STRING: 1035 case FILE_PSTRING: 1036 case FILE_BESTRING16: 1037 case FILE_LESTRING16: 1038 /* Allow text overrides */ 1039 if (mstart->str_flags & STRING_TEXTTEST) 1040 mstart->flag |= TEXTTEST; 1041 else 1042 mstart->flag |= BINTEST; 1043 break; 1044 case FILE_REGEX: 1045 case FILE_SEARCH: 1046 /* Check for override */ 1047 if (mstart->str_flags & STRING_BINTEST) 1048 mstart->flag |= BINTEST; 1049 if (mstart->str_flags & STRING_TEXTTEST) 1050 mstart->flag |= TEXTTEST; 1051 1052 if (mstart->flag & (TEXTTEST|BINTEST)) 1053 break; 1054 1055 /* binary test if pattern is not text */ 1056 if (file_looks_utf8(m->value.us, (size_t)m->vallen, NULL, 1057 NULL) <= 0) 1058 mstart->flag |= BINTEST; 1059 else 1060 mstart->flag |= TEXTTEST; 1061 break; 1062 case FILE_DEFAULT: 1063 /* can't deduce anything; we shouldn't see this at the 1064 top level anyway */ 1065 break; 1066 case FILE_INVALID: 1067 default: 1068 /* invalid search type, but no need to complain here */ 1069 break; 1070 } 1071 } 1072 1073 private int 1074 addentry(struct magic_set *ms, struct magic_entry *me, 1075 struct magic_entry_set *mset) 1076 { 1077 size_t i = me->mp->type == FILE_NAME ? 1 : 0; 1078 if (mset[i].count == mset[i].max) { 1079 struct magic_entry *mp; 1080 1081 mset[i].max += ALLOC_INCR; 1082 if ((mp = CAST(struct magic_entry *, 1083 realloc(mset[i].me, sizeof(*mp) * mset[i].max))) == 1084 NULL) { 1085 file_oomem(ms, sizeof(*mp) * mset[i].max); 1086 return -1; 1087 } 1088 (void)memset(&mp[mset[i].count], 0, sizeof(*mp) * 1089 ALLOC_INCR); 1090 mset[i].me = mp; 1091 } 1092 mset[i].me[mset[i].count++] = *me; 1093 memset(me, 0, sizeof(*me)); 1094 return 0; 1095 } 1096 1097 /* 1098 * Load and parse one file. 1099 */ 1100 private void 1101 load_1(struct magic_set *ms, int action, const char *fn, int *errs, 1102 struct magic_entry_set *mset) 1103 { 1104 size_t lineno = 0, llen = 0; 1105 char *line = NULL; 1106 ssize_t len; 1107 struct magic_entry me; 1108 1109 FILE *f = fopen(ms->file = fn, "r"); 1110 if (f == NULL) { 1111 if (errno != ENOENT) 1112 file_error(ms, errno, "cannot read magic file `%s'", 1113 fn); 1114 (*errs)++; 1115 return; 1116 } 1117 1118 memset(&me, 0, sizeof(me)); 1119 /* read and parse this file */ 1120 for (ms->line = 1; (len = getline(&line, &llen, f)) != -1; 1121 ms->line++) { 1122 if (len == 0) /* null line, garbage, etc */ 1123 continue; 1124 if (line[len - 1] == '\n') { 1125 lineno++; 1126 line[len - 1] = '\0'; /* delete newline */ 1127 } 1128 switch (line[0]) { 1129 case '\0': /* empty, do not parse */ 1130 case '#': /* comment, do not parse */ 1131 continue; 1132 case '!': 1133 if (line[1] == ':') { 1134 size_t i; 1135 1136 for (i = 0; bang[i].name != NULL; i++) { 1137 if ((size_t)(len - 2) > bang[i].len && 1138 memcmp(bang[i].name, line + 2, 1139 bang[i].len) == 0) 1140 break; 1141 } 1142 if (bang[i].name == NULL) { 1143 file_error(ms, 0, 1144 "Unknown !: entry `%s'", line); 1145 (*errs)++; 1146 continue; 1147 } 1148 if (me.mp == NULL) { 1149 file_error(ms, 0, 1150 "No current entry for :!%s type", 1151 bang[i].name); 1152 (*errs)++; 1153 continue; 1154 } 1155 if ((*bang[i].fun)(ms, &me, 1156 line + bang[i].len + 2) != 0) { 1157 (*errs)++; 1158 continue; 1159 } 1160 continue; 1161 } 1162 /*FALLTHROUGH*/ 1163 default: 1164 again: 1165 switch (parse(ms, &me, line, lineno, action)) { 1166 case 0: 1167 continue; 1168 case 1: 1169 (void)addentry(ms, &me, mset); 1170 goto again; 1171 default: 1172 (*errs)++; 1173 break; 1174 } 1175 } 1176 } 1177 if (me.mp) 1178 (void)addentry(ms, &me, mset); 1179 free(line); 1180 (void)fclose(f); 1181 } 1182 1183 /* 1184 * parse a file or directory of files 1185 * const char *fn: name of magic file or directory 1186 */ 1187 private int 1188 cmpstrp(const void *p1, const void *p2) 1189 { 1190 return strcmp(*(char *const *)p1, *(char *const *)p2); 1191 } 1192 1193 1194 private uint32_t 1195 set_text_binary(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1196 uint32_t starttest) 1197 { 1198 static const char text[] = "text"; 1199 static const char binary[] = "binary"; 1200 static const size_t len = sizeof(text); 1201 1202 uint32_t i = starttest; 1203 1204 do { 1205 set_test_type(me[starttest].mp, me[i].mp); 1206 if ((ms->flags & MAGIC_DEBUG) == 0) 1207 continue; 1208 (void)fprintf(stderr, "%s%s%s: %s\n", 1209 me[i].mp->mimetype, 1210 me[i].mp->mimetype[0] == '\0' ? "" : "; ", 1211 me[i].mp->desc[0] ? me[i].mp->desc : "(no description)", 1212 me[i].mp->flag & BINTEST ? binary : text); 1213 if (me[i].mp->flag & BINTEST) { 1214 char *p = strstr(me[i].mp->desc, text); 1215 if (p && (p == me[i].mp->desc || 1216 isspace((unsigned char)p[-1])) && 1217 (p + len - me[i].mp->desc == MAXstring 1218 || (p[len] == '\0' || 1219 isspace((unsigned char)p[len])))) 1220 (void)fprintf(stderr, "*** Possible " 1221 "binary test for text type\n"); 1222 } 1223 } while (++i < nme && me[i].mp->cont_level != 0); 1224 return i; 1225 } 1226 1227 private void 1228 set_last_default(struct magic_set *ms, struct magic_entry *me, uint32_t nme) 1229 { 1230 uint32_t i; 1231 for (i = 0; i < nme; i++) { 1232 if (me[i].mp->cont_level == 0 && 1233 me[i].mp->type == FILE_DEFAULT) { 1234 while (++i < nme) 1235 if (me[i].mp->cont_level == 0) 1236 break; 1237 if (i != nme) { 1238 /* XXX - Ugh! */ 1239 ms->line = me[i].mp->lineno; 1240 file_magwarn(ms, 1241 "level 0 \"default\" did not sort last"); 1242 } 1243 return; 1244 } 1245 } 1246 } 1247 1248 private int 1249 coalesce_entries(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1250 struct magic **ma, uint32_t *nma) 1251 { 1252 uint32_t i, mentrycount = 0; 1253 size_t slen; 1254 1255 for (i = 0; i < nme; i++) 1256 mentrycount += me[i].cont_count; 1257 1258 slen = sizeof(**ma) * mentrycount; 1259 if ((*ma = CAST(struct magic *, malloc(slen))) == NULL) { 1260 file_oomem(ms, slen); 1261 return -1; 1262 } 1263 1264 mentrycount = 0; 1265 for (i = 0; i < nme; i++) { 1266 (void)memcpy(*ma + mentrycount, me[i].mp, 1267 me[i].cont_count * sizeof(**ma)); 1268 mentrycount += me[i].cont_count; 1269 } 1270 *nma = mentrycount; 1271 return 0; 1272 } 1273 1274 private void 1275 magic_entry_free(struct magic_entry *me, uint32_t nme) 1276 { 1277 uint32_t i; 1278 if (me == NULL) 1279 return; 1280 for (i = 0; i < nme; i++) 1281 free(me[i].mp); 1282 free(me); 1283 } 1284 1285 private struct magic_map * 1286 apprentice_load(struct magic_set *ms, const char *fn, int action) 1287 { 1288 int errs = 0; 1289 uint32_t i, j; 1290 size_t files = 0, maxfiles = 0; 1291 char **filearr = NULL, *mfn; 1292 struct stat st; 1293 struct magic_map *map; 1294 struct magic_entry_set mset[MAGIC_SETS]; 1295 DIR *dir; 1296 struct dirent *d; 1297 1298 memset(mset, 0, sizeof(mset)); 1299 ms->flags |= MAGIC_CHECK; /* Enable checks for parsed files */ 1300 1301 1302 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) 1303 { 1304 file_oomem(ms, sizeof(*map)); 1305 return NULL; 1306 } 1307 map->type = MAP_TYPE_MALLOC; 1308 1309 /* print silly verbose header for USG compat. */ 1310 if (action == FILE_CHECK) 1311 (void)fprintf(stderr, "%s\n", usg_hdr); 1312 1313 /* load directory or file */ 1314 if (stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) { 1315 dir = opendir(fn); 1316 if (!dir) { 1317 errs++; 1318 goto out; 1319 } 1320 while ((d = readdir(dir)) != NULL) { 1321 if (asprintf(&mfn, "%s/%s", fn, d->d_name) < 0) { 1322 file_oomem(ms, 1323 strlen(fn) + strlen(d->d_name) + 2); 1324 errs++; 1325 closedir(dir); 1326 goto out; 1327 } 1328 if (stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) { 1329 free(mfn); 1330 continue; 1331 } 1332 if (files >= maxfiles) { 1333 size_t mlen; 1334 maxfiles = (maxfiles + 1) * 2; 1335 mlen = maxfiles * sizeof(*filearr); 1336 if ((filearr = CAST(char **, 1337 realloc(filearr, mlen))) == NULL) { 1338 file_oomem(ms, mlen); 1339 free(mfn); 1340 closedir(dir); 1341 errs++; 1342 goto out; 1343 } 1344 } 1345 filearr[files++] = mfn; 1346 } 1347 closedir(dir); 1348 qsort(filearr, files, sizeof(*filearr), cmpstrp); 1349 for (i = 0; i < files; i++) { 1350 load_1(ms, action, filearr[i], &errs, mset); 1351 free(filearr[i]); 1352 } 1353 free(filearr); 1354 } else 1355 load_1(ms, action, fn, &errs, mset); 1356 if (errs) 1357 goto out; 1358 1359 for (j = 0; j < MAGIC_SETS; j++) { 1360 /* Set types of tests */ 1361 for (i = 0; i < mset[j].count; ) { 1362 if (mset[j].me[i].mp->cont_level != 0) { 1363 i++; 1364 continue; 1365 } 1366 i = set_text_binary(ms, mset[j].me, mset[j].count, i); 1367 } 1368 if (mset[j].me) 1369 qsort(mset[j].me, mset[j].count, sizeof(*mset[j].me), 1370 apprentice_sort); 1371 1372 /* 1373 * Make sure that any level 0 "default" line is last 1374 * (if one exists). 1375 */ 1376 set_last_default(ms, mset[j].me, mset[j].count); 1377 1378 /* coalesce per file arrays into a single one */ 1379 if (coalesce_entries(ms, mset[j].me, mset[j].count, 1380 &map->magic[j], &map->nmagic[j]) == -1) { 1381 errs++; 1382 goto out; 1383 } 1384 } 1385 1386 out: 1387 for (j = 0; j < MAGIC_SETS; j++) 1388 magic_entry_free(mset[j].me, mset[j].count); 1389 1390 if (errs) { 1391 apprentice_unmap(map); 1392 return NULL; 1393 } 1394 return map; 1395 } 1396 1397 /* 1398 * extend the sign bit if the comparison is to be signed 1399 */ 1400 protected uint64_t 1401 file_signextend(struct magic_set *ms, struct magic *m, uint64_t v) 1402 { 1403 if (!(m->flag & UNSIGNED)) { 1404 switch(m->type) { 1405 /* 1406 * Do not remove the casts below. They are 1407 * vital. When later compared with the data, 1408 * the sign extension must have happened. 1409 */ 1410 case FILE_BYTE: 1411 v = (signed char) v; 1412 break; 1413 case FILE_SHORT: 1414 case FILE_BESHORT: 1415 case FILE_LESHORT: 1416 v = (short) v; 1417 break; 1418 case FILE_DATE: 1419 case FILE_BEDATE: 1420 case FILE_LEDATE: 1421 case FILE_MEDATE: 1422 case FILE_LDATE: 1423 case FILE_BELDATE: 1424 case FILE_LELDATE: 1425 case FILE_MELDATE: 1426 case FILE_LONG: 1427 case FILE_BELONG: 1428 case FILE_LELONG: 1429 case FILE_MELONG: 1430 case FILE_FLOAT: 1431 case FILE_BEFLOAT: 1432 case FILE_LEFLOAT: 1433 v = (int32_t) v; 1434 break; 1435 case FILE_QUAD: 1436 case FILE_BEQUAD: 1437 case FILE_LEQUAD: 1438 case FILE_QDATE: 1439 case FILE_QLDATE: 1440 case FILE_QWDATE: 1441 case FILE_BEQDATE: 1442 case FILE_BEQLDATE: 1443 case FILE_BEQWDATE: 1444 case FILE_LEQDATE: 1445 case FILE_LEQLDATE: 1446 case FILE_LEQWDATE: 1447 case FILE_DOUBLE: 1448 case FILE_BEDOUBLE: 1449 case FILE_LEDOUBLE: 1450 v = (int64_t) v; 1451 break; 1452 case FILE_STRING: 1453 case FILE_PSTRING: 1454 case FILE_BESTRING16: 1455 case FILE_LESTRING16: 1456 case FILE_REGEX: 1457 case FILE_SEARCH: 1458 case FILE_DEFAULT: 1459 case FILE_INDIRECT: 1460 case FILE_NAME: 1461 case FILE_USE: 1462 case FILE_CLEAR: 1463 case FILE_DER: 1464 break; 1465 default: 1466 if (ms->flags & MAGIC_CHECK) 1467 file_magwarn(ms, "cannot happen: m->type=%d\n", 1468 m->type); 1469 return ~0U; 1470 } 1471 } 1472 return v; 1473 } 1474 1475 private int 1476 string_modifier_check(struct magic_set *ms, struct magic *m) 1477 { 1478 if ((ms->flags & MAGIC_CHECK) == 0) 1479 return 0; 1480 1481 if ((m->type != FILE_REGEX || (m->str_flags & REGEX_LINE_COUNT) == 0) && 1482 (m->type != FILE_PSTRING && (m->str_flags & PSTRING_LEN) != 0)) { 1483 file_magwarn(ms, 1484 "'/BHhLl' modifiers are only allowed for pascal strings\n"); 1485 return -1; 1486 } 1487 switch (m->type) { 1488 case FILE_BESTRING16: 1489 case FILE_LESTRING16: 1490 if (m->str_flags != 0) { 1491 file_magwarn(ms, 1492 "no modifiers allowed for 16-bit strings\n"); 1493 return -1; 1494 } 1495 break; 1496 case FILE_STRING: 1497 case FILE_PSTRING: 1498 if ((m->str_flags & REGEX_OFFSET_START) != 0) { 1499 file_magwarn(ms, 1500 "'/%c' only allowed on regex and search\n", 1501 CHAR_REGEX_OFFSET_START); 1502 return -1; 1503 } 1504 break; 1505 case FILE_SEARCH: 1506 if (m->str_range == 0) { 1507 file_magwarn(ms, 1508 "missing range; defaulting to %d\n", 1509 STRING_DEFAULT_RANGE); 1510 m->str_range = STRING_DEFAULT_RANGE; 1511 return -1; 1512 } 1513 break; 1514 case FILE_REGEX: 1515 if ((m->str_flags & STRING_COMPACT_WHITESPACE) != 0) { 1516 file_magwarn(ms, "'/%c' not allowed on regex\n", 1517 CHAR_COMPACT_WHITESPACE); 1518 return -1; 1519 } 1520 if ((m->str_flags & STRING_COMPACT_OPTIONAL_WHITESPACE) != 0) { 1521 file_magwarn(ms, "'/%c' not allowed on regex\n", 1522 CHAR_COMPACT_OPTIONAL_WHITESPACE); 1523 return -1; 1524 } 1525 break; 1526 default: 1527 file_magwarn(ms, "coding error: m->type=%d\n", 1528 m->type); 1529 return -1; 1530 } 1531 return 0; 1532 } 1533 1534 private int 1535 get_op(char c) 1536 { 1537 switch (c) { 1538 case '&': 1539 return FILE_OPAND; 1540 case '|': 1541 return FILE_OPOR; 1542 case '^': 1543 return FILE_OPXOR; 1544 case '+': 1545 return FILE_OPADD; 1546 case '-': 1547 return FILE_OPMINUS; 1548 case '*': 1549 return FILE_OPMULTIPLY; 1550 case '/': 1551 return FILE_OPDIVIDE; 1552 case '%': 1553 return FILE_OPMODULO; 1554 default: 1555 return -1; 1556 } 1557 } 1558 1559 #ifdef ENABLE_CONDITIONALS 1560 private int 1561 get_cond(const char *l, const char **t) 1562 { 1563 static const struct cond_tbl_s { 1564 char name[8]; 1565 size_t len; 1566 int cond; 1567 } cond_tbl[] = { 1568 { "if", 2, COND_IF }, 1569 { "elif", 4, COND_ELIF }, 1570 { "else", 4, COND_ELSE }, 1571 { "", 0, COND_NONE }, 1572 }; 1573 const struct cond_tbl_s *p; 1574 1575 for (p = cond_tbl; p->len; p++) { 1576 if (strncmp(l, p->name, p->len) == 0 && 1577 isspace((unsigned char)l[p->len])) { 1578 if (t) 1579 *t = l + p->len; 1580 break; 1581 } 1582 } 1583 return p->cond; 1584 } 1585 1586 private int 1587 check_cond(struct magic_set *ms, int cond, uint32_t cont_level) 1588 { 1589 int last_cond; 1590 last_cond = ms->c.li[cont_level].last_cond; 1591 1592 switch (cond) { 1593 case COND_IF: 1594 if (last_cond != COND_NONE && last_cond != COND_ELIF) { 1595 if (ms->flags & MAGIC_CHECK) 1596 file_magwarn(ms, "syntax error: `if'"); 1597 return -1; 1598 } 1599 last_cond = COND_IF; 1600 break; 1601 1602 case COND_ELIF: 1603 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1604 if (ms->flags & MAGIC_CHECK) 1605 file_magwarn(ms, "syntax error: `elif'"); 1606 return -1; 1607 } 1608 last_cond = COND_ELIF; 1609 break; 1610 1611 case COND_ELSE: 1612 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1613 if (ms->flags & MAGIC_CHECK) 1614 file_magwarn(ms, "syntax error: `else'"); 1615 return -1; 1616 } 1617 last_cond = COND_NONE; 1618 break; 1619 1620 case COND_NONE: 1621 last_cond = COND_NONE; 1622 break; 1623 } 1624 1625 ms->c.li[cont_level].last_cond = last_cond; 1626 return 0; 1627 } 1628 #endif /* ENABLE_CONDITIONALS */ 1629 1630 private int 1631 parse_indirect_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1632 { 1633 const char *l = *lp; 1634 1635 while (!isspace((unsigned char)*++l)) 1636 switch (*l) { 1637 case CHAR_INDIRECT_RELATIVE: 1638 m->str_flags |= INDIRECT_RELATIVE; 1639 break; 1640 default: 1641 if (ms->flags & MAGIC_CHECK) 1642 file_magwarn(ms, "indirect modifier `%c' " 1643 "invalid", *l); 1644 *lp = l; 1645 return -1; 1646 } 1647 *lp = l; 1648 return 0; 1649 } 1650 1651 private void 1652 parse_op_modifier(struct magic_set *ms, struct magic *m, const char **lp, 1653 int op) 1654 { 1655 const char *l = *lp; 1656 char *t; 1657 uint64_t val; 1658 1659 ++l; 1660 m->mask_op |= op; 1661 val = (uint64_t)strtoull(l, &t, 0); 1662 l = t; 1663 m->num_mask = file_signextend(ms, m, val); 1664 eatsize(&l); 1665 *lp = l; 1666 } 1667 1668 private int 1669 parse_string_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1670 { 1671 const char *l = *lp; 1672 char *t; 1673 int have_range = 0; 1674 1675 while (!isspace((unsigned char)*++l)) { 1676 switch (*l) { 1677 case '0': case '1': case '2': 1678 case '3': case '4': case '5': 1679 case '6': case '7': case '8': 1680 case '9': 1681 if (have_range && (ms->flags & MAGIC_CHECK)) 1682 file_magwarn(ms, "multiple ranges"); 1683 have_range = 1; 1684 m->str_range = CAST(uint32_t, strtoul(l, &t, 0)); 1685 if (m->str_range == 0) 1686 file_magwarn(ms, "zero range"); 1687 l = t - 1; 1688 break; 1689 case CHAR_COMPACT_WHITESPACE: 1690 m->str_flags |= STRING_COMPACT_WHITESPACE; 1691 break; 1692 case CHAR_COMPACT_OPTIONAL_WHITESPACE: 1693 m->str_flags |= STRING_COMPACT_OPTIONAL_WHITESPACE; 1694 break; 1695 case CHAR_IGNORE_LOWERCASE: 1696 m->str_flags |= STRING_IGNORE_LOWERCASE; 1697 break; 1698 case CHAR_IGNORE_UPPERCASE: 1699 m->str_flags |= STRING_IGNORE_UPPERCASE; 1700 break; 1701 case CHAR_REGEX_OFFSET_START: 1702 m->str_flags |= REGEX_OFFSET_START; 1703 break; 1704 case CHAR_BINTEST: 1705 m->str_flags |= STRING_BINTEST; 1706 break; 1707 case CHAR_TEXTTEST: 1708 m->str_flags |= STRING_TEXTTEST; 1709 break; 1710 case CHAR_TRIM: 1711 m->str_flags |= STRING_TRIM; 1712 break; 1713 case CHAR_PSTRING_1_LE: 1714 #define SET_LENGTH(a) m->str_flags = (m->str_flags & ~PSTRING_LEN) | (a) 1715 if (m->type != FILE_PSTRING) 1716 goto bad; 1717 SET_LENGTH(PSTRING_1_LE); 1718 break; 1719 case CHAR_PSTRING_2_BE: 1720 if (m->type != FILE_PSTRING) 1721 goto bad; 1722 SET_LENGTH(PSTRING_2_BE); 1723 break; 1724 case CHAR_PSTRING_2_LE: 1725 if (m->type != FILE_PSTRING) 1726 goto bad; 1727 SET_LENGTH(PSTRING_2_LE); 1728 break; 1729 case CHAR_PSTRING_4_BE: 1730 if (m->type != FILE_PSTRING) 1731 goto bad; 1732 SET_LENGTH(PSTRING_4_BE); 1733 break; 1734 case CHAR_PSTRING_4_LE: 1735 switch (m->type) { 1736 case FILE_PSTRING: 1737 case FILE_REGEX: 1738 break; 1739 default: 1740 goto bad; 1741 } 1742 SET_LENGTH(PSTRING_4_LE); 1743 break; 1744 case CHAR_PSTRING_LENGTH_INCLUDES_ITSELF: 1745 if (m->type != FILE_PSTRING) 1746 goto bad; 1747 m->str_flags |= PSTRING_LENGTH_INCLUDES_ITSELF; 1748 break; 1749 default: 1750 bad: 1751 if (ms->flags & MAGIC_CHECK) 1752 file_magwarn(ms, "string modifier `%c' " 1753 "invalid", *l); 1754 goto out; 1755 } 1756 /* allow multiple '/' for readability */ 1757 if (l[1] == '/' && !isspace((unsigned char)l[2])) 1758 l++; 1759 } 1760 if (string_modifier_check(ms, m) == -1) 1761 goto out; 1762 *lp = l; 1763 return 0; 1764 out: 1765 *lp = l; 1766 return -1; 1767 } 1768 1769 /* 1770 * parse one line from magic file, put into magic[index++] if valid 1771 */ 1772 private int 1773 parse(struct magic_set *ms, struct magic_entry *me, const char *line, 1774 size_t lineno, int action) 1775 { 1776 #ifdef ENABLE_CONDITIONALS 1777 static uint32_t last_cont_level = 0; 1778 #endif 1779 size_t i; 1780 struct magic *m; 1781 const char *l = line; 1782 char *t; 1783 int op; 1784 uint32_t cont_level; 1785 int32_t diff; 1786 1787 cont_level = 0; 1788 1789 /* 1790 * Parse the offset. 1791 */ 1792 while (*l == '>') { 1793 ++l; /* step over */ 1794 cont_level++; 1795 } 1796 #ifdef ENABLE_CONDITIONALS 1797 if (cont_level == 0 || cont_level > last_cont_level) 1798 if (file_check_mem(ms, cont_level) == -1) 1799 return -1; 1800 last_cont_level = cont_level; 1801 #endif 1802 if (cont_level != 0) { 1803 if (me->mp == NULL) { 1804 file_magerror(ms, "No current entry for continuation"); 1805 return -1; 1806 } 1807 if (me->cont_count == 0) { 1808 file_magerror(ms, "Continuations present with 0 count"); 1809 return -1; 1810 } 1811 m = &me->mp[me->cont_count - 1]; 1812 diff = (int32_t)cont_level - (int32_t)m->cont_level; 1813 if (diff > 1) 1814 file_magwarn(ms, "New continuation level %u is more " 1815 "than one larger than current level %u", cont_level, 1816 m->cont_level); 1817 if (me->cont_count == me->max_count) { 1818 struct magic *nm; 1819 size_t cnt = me->max_count + ALLOC_CHUNK; 1820 if ((nm = CAST(struct magic *, realloc(me->mp, 1821 sizeof(*nm) * cnt))) == NULL) { 1822 file_oomem(ms, sizeof(*nm) * cnt); 1823 return -1; 1824 } 1825 me->mp = m = nm; 1826 me->max_count = CAST(uint32_t, cnt); 1827 } 1828 m = &me->mp[me->cont_count++]; 1829 (void)memset(m, 0, sizeof(*m)); 1830 m->cont_level = cont_level; 1831 } else { 1832 static const size_t len = sizeof(*m) * ALLOC_CHUNK; 1833 if (me->mp != NULL) 1834 return 1; 1835 if ((m = CAST(struct magic *, malloc(len))) == NULL) { 1836 file_oomem(ms, len); 1837 return -1; 1838 } 1839 me->mp = m; 1840 me->max_count = ALLOC_CHUNK; 1841 (void)memset(m, 0, sizeof(*m)); 1842 m->factor_op = FILE_FACTOR_OP_NONE; 1843 m->cont_level = 0; 1844 me->cont_count = 1; 1845 } 1846 m->lineno = CAST(uint32_t, lineno); 1847 1848 if (*l == '&') { /* m->cont_level == 0 checked below. */ 1849 ++l; /* step over */ 1850 m->flag |= OFFADD; 1851 } 1852 if (*l == '(') { 1853 ++l; /* step over */ 1854 m->flag |= INDIR; 1855 if (m->flag & OFFADD) 1856 m->flag = (m->flag & ~OFFADD) | INDIROFFADD; 1857 1858 if (*l == '&') { /* m->cont_level == 0 checked below */ 1859 ++l; /* step over */ 1860 m->flag |= OFFADD; 1861 } 1862 } 1863 /* Indirect offsets are not valid at level 0. */ 1864 if (m->cont_level == 0 && (m->flag & (OFFADD | INDIROFFADD))) { 1865 if (ms->flags & MAGIC_CHECK) 1866 file_magwarn(ms, "relative offset at level 0"); 1867 return -1; 1868 } 1869 1870 /* get offset, then skip over it */ 1871 m->offset = (uint32_t)strtoul(l, &t, 0); 1872 if (l == t) { 1873 if (ms->flags & MAGIC_CHECK) 1874 file_magwarn(ms, "offset `%s' invalid", l); 1875 return -1; 1876 } 1877 l = t; 1878 1879 if (m->flag & INDIR) { 1880 m->in_type = FILE_LONG; 1881 m->in_offset = 0; 1882 /* 1883 * read [.lbs][+-]nnnnn) 1884 */ 1885 if (*l == '.') { 1886 l++; 1887 switch (*l) { 1888 case 'l': 1889 m->in_type = FILE_LELONG; 1890 break; 1891 case 'L': 1892 m->in_type = FILE_BELONG; 1893 break; 1894 case 'm': 1895 m->in_type = FILE_MELONG; 1896 break; 1897 case 'h': 1898 case 's': 1899 m->in_type = FILE_LESHORT; 1900 break; 1901 case 'H': 1902 case 'S': 1903 m->in_type = FILE_BESHORT; 1904 break; 1905 case 'c': 1906 case 'b': 1907 case 'C': 1908 case 'B': 1909 m->in_type = FILE_BYTE; 1910 break; 1911 case 'e': 1912 case 'f': 1913 case 'g': 1914 m->in_type = FILE_LEDOUBLE; 1915 break; 1916 case 'E': 1917 case 'F': 1918 case 'G': 1919 m->in_type = FILE_BEDOUBLE; 1920 break; 1921 case 'i': 1922 m->in_type = FILE_LEID3; 1923 break; 1924 case 'I': 1925 m->in_type = FILE_BEID3; 1926 break; 1927 default: 1928 if (ms->flags & MAGIC_CHECK) 1929 file_magwarn(ms, 1930 "indirect offset type `%c' invalid", 1931 *l); 1932 return -1; 1933 } 1934 l++; 1935 } 1936 1937 m->in_op = 0; 1938 if (*l == '~') { 1939 m->in_op |= FILE_OPINVERSE; 1940 l++; 1941 } 1942 if ((op = get_op(*l)) != -1) { 1943 m->in_op |= op; 1944 l++; 1945 } 1946 if (*l == '(') { 1947 m->in_op |= FILE_OPINDIRECT; 1948 l++; 1949 } 1950 if (isdigit((unsigned char)*l) || *l == '-') { 1951 m->in_offset = (int32_t)strtol(l, &t, 0); 1952 if (l == t) { 1953 if (ms->flags & MAGIC_CHECK) 1954 file_magwarn(ms, 1955 "in_offset `%s' invalid", l); 1956 return -1; 1957 } 1958 l = t; 1959 } 1960 if (*l++ != ')' || 1961 ((m->in_op & FILE_OPINDIRECT) && *l++ != ')')) { 1962 if (ms->flags & MAGIC_CHECK) 1963 file_magwarn(ms, 1964 "missing ')' in indirect offset"); 1965 return -1; 1966 } 1967 } 1968 EATAB; 1969 1970 #ifdef ENABLE_CONDITIONALS 1971 m->cond = get_cond(l, &l); 1972 if (check_cond(ms, m->cond, cont_level) == -1) 1973 return -1; 1974 1975 EATAB; 1976 #endif 1977 1978 /* 1979 * Parse the type. 1980 */ 1981 if (*l == 'u') { 1982 /* 1983 * Try it as a keyword type prefixed by "u"; match what 1984 * follows the "u". If that fails, try it as an SUS 1985 * integer type. 1986 */ 1987 m->type = get_type(type_tbl, l + 1, &l); 1988 if (m->type == FILE_INVALID) { 1989 /* 1990 * Not a keyword type; parse it as an SUS type, 1991 * 'u' possibly followed by a number or C/S/L. 1992 */ 1993 m->type = get_standard_integer_type(l, &l); 1994 } 1995 /* It's unsigned. */ 1996 if (m->type != FILE_INVALID) 1997 m->flag |= UNSIGNED; 1998 } else { 1999 /* 2000 * Try it as a keyword type. If that fails, try it as 2001 * an SUS integer type if it begins with "d" or as an 2002 * SUS string type if it begins with "s". In any case, 2003 * it's not unsigned. 2004 */ 2005 m->type = get_type(type_tbl, l, &l); 2006 if (m->type == FILE_INVALID) { 2007 /* 2008 * Not a keyword type; parse it as an SUS type, 2009 * either 'd' possibly followed by a number or 2010 * C/S/L, or just 's'. 2011 */ 2012 if (*l == 'd') 2013 m->type = get_standard_integer_type(l, &l); 2014 else if (*l == 's' && !isalpha((unsigned char)l[1])) { 2015 m->type = FILE_STRING; 2016 ++l; 2017 } 2018 } 2019 } 2020 2021 if (m->type == FILE_INVALID) { 2022 /* Not found - try it as a special keyword. */ 2023 m->type = get_type(special_tbl, l, &l); 2024 } 2025 2026 if (m->type == FILE_INVALID) { 2027 if (ms->flags & MAGIC_CHECK) 2028 file_magwarn(ms, "type `%s' invalid", l); 2029 return -1; 2030 } 2031 2032 /* New-style anding: "0 byte&0x80 =0x80 dynamically linked" */ 2033 /* New and improved: ~ & | ^ + - * / % -- exciting, isn't it? */ 2034 2035 m->mask_op = 0; 2036 if (*l == '~') { 2037 if (!IS_STRING(m->type)) 2038 m->mask_op |= FILE_OPINVERSE; 2039 else if (ms->flags & MAGIC_CHECK) 2040 file_magwarn(ms, "'~' invalid for string types"); 2041 ++l; 2042 } 2043 m->str_range = 0; 2044 m->str_flags = m->type == FILE_PSTRING ? PSTRING_1_LE : 0; 2045 if ((op = get_op(*l)) != -1) { 2046 if (IS_STRING(m->type)) { 2047 int r; 2048 2049 if (op != FILE_OPDIVIDE) { 2050 if (ms->flags & MAGIC_CHECK) 2051 file_magwarn(ms, 2052 "invalid string/indirect op: " 2053 "`%c'", *t); 2054 return -1; 2055 } 2056 2057 if (m->type == FILE_INDIRECT) 2058 r = parse_indirect_modifier(ms, m, &l); 2059 else 2060 r = parse_string_modifier(ms, m, &l); 2061 if (r == -1) 2062 return -1; 2063 } else 2064 parse_op_modifier(ms, m, &l, op); 2065 } 2066 2067 /* 2068 * We used to set mask to all 1's here, instead let's just not do 2069 * anything if mask = 0 (unless you have a better idea) 2070 */ 2071 EATAB; 2072 2073 switch (*l) { 2074 case '>': 2075 case '<': 2076 m->reln = *l; 2077 ++l; 2078 if (*l == '=') { 2079 if (ms->flags & MAGIC_CHECK) { 2080 file_magwarn(ms, "%c= not supported", 2081 m->reln); 2082 return -1; 2083 } 2084 ++l; 2085 } 2086 break; 2087 /* Old-style anding: "0 byte &0x80 dynamically linked" */ 2088 case '&': 2089 case '^': 2090 case '=': 2091 m->reln = *l; 2092 ++l; 2093 if (*l == '=') { 2094 /* HP compat: ignore &= etc. */ 2095 ++l; 2096 } 2097 break; 2098 case '!': 2099 m->reln = *l; 2100 ++l; 2101 break; 2102 default: 2103 m->reln = '='; /* the default relation */ 2104 if (*l == 'x' && ((isascii((unsigned char)l[1]) && 2105 isspace((unsigned char)l[1])) || !l[1])) { 2106 m->reln = *l; 2107 ++l; 2108 } 2109 break; 2110 } 2111 /* 2112 * Grab the value part, except for an 'x' reln. 2113 */ 2114 if (m->reln != 'x' && getvalue(ms, m, &l, action)) 2115 return -1; 2116 2117 /* 2118 * TODO finish this macro and start using it! 2119 * #define offsetcheck {if (offset > ms->bytes_max -1) 2120 * magwarn("offset too big"); } 2121 */ 2122 2123 /* 2124 * Now get last part - the description 2125 */ 2126 EATAB; 2127 if (l[0] == '\b') { 2128 ++l; 2129 m->flag |= NOSPACE; 2130 } else if ((l[0] == '\\') && (l[1] == 'b')) { 2131 ++l; 2132 ++l; 2133 m->flag |= NOSPACE; 2134 } 2135 for (i = 0; (m->desc[i++] = *l++) != '\0' && i < sizeof(m->desc); ) 2136 continue; 2137 if (i == sizeof(m->desc)) { 2138 m->desc[sizeof(m->desc) - 1] = '\0'; 2139 if (ms->flags & MAGIC_CHECK) 2140 file_magwarn(ms, "description `%s' truncated", m->desc); 2141 } 2142 2143 /* 2144 * We only do this check while compiling, or if any of the magic 2145 * files were not compiled. 2146 */ 2147 if (ms->flags & MAGIC_CHECK) { 2148 if (check_format(ms, m) == -1) 2149 return -1; 2150 } 2151 #ifndef COMPILE_ONLY 2152 if (action == FILE_CHECK) { 2153 file_mdump(m); 2154 } 2155 #endif 2156 m->mimetype[0] = '\0'; /* initialise MIME type to none */ 2157 return 0; 2158 } 2159 2160 /* 2161 * parse a STRENGTH annotation line from magic file, put into magic[index - 1] 2162 * if valid 2163 */ 2164 private int 2165 parse_strength(struct magic_set *ms, struct magic_entry *me, const char *line) 2166 { 2167 const char *l = line; 2168 char *el; 2169 unsigned long factor; 2170 struct magic *m = &me->mp[0]; 2171 2172 if (m->factor_op != FILE_FACTOR_OP_NONE) { 2173 file_magwarn(ms, 2174 "Current entry already has a strength type: %c %d", 2175 m->factor_op, m->factor); 2176 return -1; 2177 } 2178 if (m->type == FILE_NAME) { 2179 file_magwarn(ms, "%s: Strength setting is not supported in " 2180 "\"name\" magic entries", m->value.s); 2181 return -1; 2182 } 2183 EATAB; 2184 switch (*l) { 2185 case FILE_FACTOR_OP_NONE: 2186 case FILE_FACTOR_OP_PLUS: 2187 case FILE_FACTOR_OP_MINUS: 2188 case FILE_FACTOR_OP_TIMES: 2189 case FILE_FACTOR_OP_DIV: 2190 m->factor_op = *l++; 2191 break; 2192 default: 2193 file_magwarn(ms, "Unknown factor op `%c'", *l); 2194 return -1; 2195 } 2196 EATAB; 2197 factor = strtoul(l, &el, 0); 2198 if (factor > 255) { 2199 file_magwarn(ms, "Too large factor `%lu'", factor); 2200 goto out; 2201 } 2202 if (*el && !isspace((unsigned char)*el)) { 2203 file_magwarn(ms, "Bad factor `%s'", l); 2204 goto out; 2205 } 2206 m->factor = (uint8_t)factor; 2207 if (m->factor == 0 && m->factor_op == FILE_FACTOR_OP_DIV) { 2208 file_magwarn(ms, "Cannot have factor op `%c' and factor %u", 2209 m->factor_op, m->factor); 2210 goto out; 2211 } 2212 return 0; 2213 out: 2214 m->factor_op = FILE_FACTOR_OP_NONE; 2215 m->factor = 0; 2216 return -1; 2217 } 2218 2219 private int 2220 goodchar(unsigned char x, const char *extra) 2221 { 2222 return (isascii(x) && isalnum(x)) || strchr(extra, x); 2223 } 2224 2225 private int 2226 parse_extra(struct magic_set *ms, struct magic_entry *me, const char *line, 2227 off_t off, size_t len, const char *name, const char *extra, int nt) 2228 { 2229 size_t i; 2230 const char *l = line; 2231 struct magic *m = &me->mp[me->cont_count == 0 ? 0 : me->cont_count - 1]; 2232 char *buf = CAST(char *, CAST(void *, m)) + off; 2233 2234 if (buf[0] != '\0') { 2235 len = nt ? strlen(buf) : len; 2236 file_magwarn(ms, "Current entry already has a %s type " 2237 "`%.*s', new type `%s'", name, (int)len, buf, l); 2238 return -1; 2239 } 2240 2241 if (*m->desc == '\0') { 2242 file_magwarn(ms, "Current entry does not yet have a " 2243 "description for adding a %s type", name); 2244 return -1; 2245 } 2246 2247 EATAB; 2248 for (i = 0; *l && i < len && goodchar(*l, extra); buf[i++] = *l++) 2249 continue; 2250 2251 if (i == len && *l) { 2252 if (nt) 2253 buf[len - 1] = '\0'; 2254 if (ms->flags & MAGIC_CHECK) 2255 file_magwarn(ms, "%s type `%s' truncated %" 2256 SIZE_T_FORMAT "u", name, line, i); 2257 } else { 2258 if (!isspace((unsigned char)*l) && !goodchar(*l, extra)) 2259 file_magwarn(ms, "%s type `%s' has bad char '%c'", 2260 name, line, *l); 2261 if (nt) 2262 buf[i] = '\0'; 2263 } 2264 2265 if (i > 0) 2266 return 0; 2267 2268 file_magerror(ms, "Bad magic entry '%s'", line); 2269 return -1; 2270 } 2271 2272 /* 2273 * Parse an Apple CREATOR/TYPE annotation from magic file and put it into 2274 * magic[index - 1] 2275 */ 2276 private int 2277 parse_apple(struct magic_set *ms, struct magic_entry *me, const char *line) 2278 { 2279 struct magic *m = &me->mp[0]; 2280 2281 return parse_extra(ms, me, line, 2282 CAST(off_t, offsetof(struct magic, apple)), 2283 sizeof(m->apple), "APPLE", "!+-./?", 0); 2284 } 2285 2286 /* 2287 * Parse a comma-separated list of extensions 2288 */ 2289 private int 2290 parse_ext(struct magic_set *ms, struct magic_entry *me, const char *line) 2291 { 2292 struct magic *m = &me->mp[0]; 2293 2294 return parse_extra(ms, me, line, 2295 CAST(off_t, offsetof(struct magic, ext)), 2296 sizeof(m->ext), "EXTENSION", ",!+-/", 0); 2297 } 2298 2299 /* 2300 * parse a MIME annotation line from magic file, put into magic[index - 1] 2301 * if valid 2302 */ 2303 private int 2304 parse_mime(struct magic_set *ms, struct magic_entry *me, const char *line) 2305 { 2306 struct magic *m = &me->mp[0]; 2307 2308 return parse_extra(ms, me, line, 2309 CAST(off_t, offsetof(struct magic, mimetype)), 2310 sizeof(m->mimetype), "MIME", "+-/.", 1); 2311 } 2312 2313 private int 2314 check_format_type(const char *ptr, int type, const char **estr) 2315 { 2316 int quad = 0, h; 2317 size_t len, cnt; 2318 if (*ptr == '\0') { 2319 /* Missing format string; bad */ 2320 *estr = "missing format spec"; 2321 return -1; 2322 } 2323 2324 switch (file_formats[type]) { 2325 case FILE_FMT_QUAD: 2326 quad = 1; 2327 /*FALLTHROUGH*/ 2328 case FILE_FMT_NUM: 2329 if (quad == 0) { 2330 switch (type) { 2331 case FILE_BYTE: 2332 h = 2; 2333 break; 2334 case FILE_SHORT: 2335 case FILE_BESHORT: 2336 case FILE_LESHORT: 2337 h = 1; 2338 break; 2339 case FILE_LONG: 2340 case FILE_BELONG: 2341 case FILE_LELONG: 2342 case FILE_MELONG: 2343 case FILE_LEID3: 2344 case FILE_BEID3: 2345 case FILE_INDIRECT: 2346 h = 0; 2347 break; 2348 default: 2349 abort(); 2350 } 2351 } else 2352 h = 0; 2353 if (*ptr == '-') 2354 ptr++; 2355 if (*ptr == '.') 2356 ptr++; 2357 #define CHECKLEN() do { \ 2358 for (len = cnt = 0; isdigit((unsigned char)*ptr); ptr++, cnt++) \ 2359 len = len * 10 + (*ptr - '0'); \ 2360 if (cnt > 5 || len > 1024) \ 2361 goto toolong; \ 2362 } while (/*CONSTCOND*/0) 2363 2364 CHECKLEN(); 2365 if (*ptr == '.') 2366 ptr++; 2367 CHECKLEN(); 2368 if (quad) { 2369 if (*ptr++ != 'l') 2370 goto invalid; 2371 if (*ptr++ != 'l') 2372 goto invalid; 2373 } 2374 2375 switch (*ptr++) { 2376 #ifdef STRICT_FORMAT /* "long" formats are int formats for us */ 2377 /* so don't accept the 'l' modifier */ 2378 case 'l': 2379 switch (*ptr++) { 2380 case 'i': 2381 case 'd': 2382 case 'u': 2383 case 'o': 2384 case 'x': 2385 case 'X': 2386 if (h == 0) 2387 return 0; 2388 /*FALLTHROUGH*/ 2389 default: 2390 goto invalid; 2391 } 2392 2393 /* 2394 * Don't accept h and hh modifiers. They make writing 2395 * magic entries more complicated, for very little benefit 2396 */ 2397 case 'h': 2398 if (h-- <= 0) 2399 goto invalid; 2400 switch (*ptr++) { 2401 case 'h': 2402 if (h-- <= 0) 2403 goto invalid; 2404 switch (*ptr++) { 2405 case 'i': 2406 case 'd': 2407 case 'u': 2408 case 'o': 2409 case 'x': 2410 case 'X': 2411 return 0; 2412 default: 2413 goto invalid; 2414 } 2415 case 'i': 2416 case 'd': 2417 case 'u': 2418 case 'o': 2419 case 'x': 2420 case 'X': 2421 if (h == 0) 2422 return 0; 2423 /*FALLTHROUGH*/ 2424 default: 2425 goto invalid; 2426 } 2427 #endif 2428 case 'c': 2429 if (h == 2) 2430 return 0; 2431 goto invalid; 2432 case 'i': 2433 case 'd': 2434 case 'u': 2435 case 'o': 2436 case 'x': 2437 case 'X': 2438 #ifdef STRICT_FORMAT 2439 if (h == 0) 2440 return 0; 2441 /*FALLTHROUGH*/ 2442 #else 2443 return 0; 2444 #endif 2445 default: 2446 goto invalid; 2447 } 2448 2449 case FILE_FMT_FLOAT: 2450 case FILE_FMT_DOUBLE: 2451 if (*ptr == '-') 2452 ptr++; 2453 if (*ptr == '.') 2454 ptr++; 2455 CHECKLEN(); 2456 if (*ptr == '.') 2457 ptr++; 2458 CHECKLEN(); 2459 switch (*ptr++) { 2460 case 'e': 2461 case 'E': 2462 case 'f': 2463 case 'F': 2464 case 'g': 2465 case 'G': 2466 return 0; 2467 2468 default: 2469 goto invalid; 2470 } 2471 2472 2473 case FILE_FMT_STR: 2474 if (*ptr == '-') 2475 ptr++; 2476 while (isdigit((unsigned char )*ptr)) 2477 ptr++; 2478 if (*ptr == '.') { 2479 ptr++; 2480 while (isdigit((unsigned char )*ptr)) 2481 ptr++; 2482 } 2483 2484 switch (*ptr++) { 2485 case 's': 2486 return 0; 2487 default: 2488 goto invalid; 2489 } 2490 2491 default: 2492 /* internal error */ 2493 abort(); 2494 } 2495 invalid: 2496 *estr = "not valid"; 2497 toolong: 2498 *estr = "too long"; 2499 return -1; 2500 } 2501 2502 /* 2503 * Check that the optional printf format in description matches 2504 * the type of the magic. 2505 */ 2506 private int 2507 check_format(struct magic_set *ms, struct magic *m) 2508 { 2509 char *ptr; 2510 const char *estr; 2511 2512 for (ptr = m->desc; *ptr; ptr++) 2513 if (*ptr == '%') 2514 break; 2515 if (*ptr == '\0') { 2516 /* No format string; ok */ 2517 return 1; 2518 } 2519 2520 assert(file_nformats == file_nnames); 2521 2522 if (m->type >= file_nformats) { 2523 file_magwarn(ms, "Internal error inconsistency between " 2524 "m->type and format strings"); 2525 return -1; 2526 } 2527 if (file_formats[m->type] == FILE_FMT_NONE) { 2528 file_magwarn(ms, "No format string for `%s' with description " 2529 "`%s'", m->desc, file_names[m->type]); 2530 return -1; 2531 } 2532 2533 ptr++; 2534 if (check_format_type(ptr, m->type, &estr) == -1) { 2535 /* 2536 * TODO: this error message is unhelpful if the format 2537 * string is not one character long 2538 */ 2539 file_magwarn(ms, "Printf format is %s for type " 2540 "`%s' in description `%s'", estr, 2541 file_names[m->type], m->desc); 2542 return -1; 2543 } 2544 2545 for (; *ptr; ptr++) { 2546 if (*ptr == '%') { 2547 file_magwarn(ms, 2548 "Too many format strings (should have at most one) " 2549 "for `%s' with description `%s'", 2550 file_names[m->type], m->desc); 2551 return -1; 2552 } 2553 } 2554 return 0; 2555 } 2556 2557 /* 2558 * Read a numeric value from a pointer, into the value union of a magic 2559 * pointer, according to the magic type. Update the string pointer to point 2560 * just after the number read. Return 0 for success, non-zero for failure. 2561 */ 2562 private int 2563 getvalue(struct magic_set *ms, struct magic *m, const char **p, int action) 2564 { 2565 switch (m->type) { 2566 case FILE_BESTRING16: 2567 case FILE_LESTRING16: 2568 case FILE_STRING: 2569 case FILE_PSTRING: 2570 case FILE_REGEX: 2571 case FILE_SEARCH: 2572 case FILE_NAME: 2573 case FILE_USE: 2574 case FILE_DER: 2575 *p = getstr(ms, m, *p, action == FILE_COMPILE); 2576 if (*p == NULL) { 2577 if (ms->flags & MAGIC_CHECK) 2578 file_magwarn(ms, "cannot get string from `%s'", 2579 m->value.s); 2580 return -1; 2581 } 2582 if (m->type == FILE_REGEX) { 2583 file_regex_t rx; 2584 int rc = file_regcomp(&rx, m->value.s, REG_EXTENDED); 2585 if (rc) { 2586 if (ms->flags & MAGIC_CHECK) 2587 file_regerror(&rx, rc, ms); 2588 } 2589 file_regfree(&rx); 2590 return rc ? -1 : 0; 2591 } 2592 return 0; 2593 case FILE_FLOAT: 2594 case FILE_BEFLOAT: 2595 case FILE_LEFLOAT: 2596 if (m->reln != 'x') { 2597 char *ep; 2598 errno = 0; 2599 #ifdef HAVE_STRTOF 2600 m->value.f = strtof(*p, &ep); 2601 #else 2602 m->value.f = (float)strtod(*p, &ep); 2603 #endif 2604 if (errno == 0) 2605 *p = ep; 2606 } 2607 return 0; 2608 case FILE_DOUBLE: 2609 case FILE_BEDOUBLE: 2610 case FILE_LEDOUBLE: 2611 if (m->reln != 'x') { 2612 char *ep; 2613 errno = 0; 2614 m->value.d = strtod(*p, &ep); 2615 if (errno == 0) 2616 *p = ep; 2617 } 2618 return 0; 2619 default: 2620 if (m->reln != 'x') { 2621 char *ep; 2622 errno = 0; 2623 m->value.q = file_signextend(ms, m, 2624 (uint64_t)strtoull(*p, &ep, 0)); 2625 if (errno == 0) { 2626 *p = ep; 2627 eatsize(p); 2628 } 2629 } 2630 return 0; 2631 } 2632 } 2633 2634 /* 2635 * Convert a string containing C character escapes. Stop at an unescaped 2636 * space or tab. 2637 * Copy the converted version to "m->value.s", and the length in m->vallen. 2638 * Return updated scan pointer as function result. Warn if set. 2639 */ 2640 private const char * 2641 getstr(struct magic_set *ms, struct magic *m, const char *s, int warn) 2642 { 2643 const char *origs = s; 2644 char *p = m->value.s; 2645 size_t plen = sizeof(m->value.s); 2646 char *origp = p; 2647 char *pmax = p + plen - 1; 2648 int c; 2649 int val; 2650 2651 while ((c = *s++) != '\0') { 2652 if (isspace((unsigned char) c)) 2653 break; 2654 if (p >= pmax) { 2655 file_error(ms, 0, "string too long: `%s'", origs); 2656 return NULL; 2657 } 2658 if (c == '\\') { 2659 switch(c = *s++) { 2660 2661 case '\0': 2662 if (warn) 2663 file_magwarn(ms, "incomplete escape"); 2664 s--; 2665 goto out; 2666 2667 case '\t': 2668 if (warn) { 2669 file_magwarn(ms, 2670 "escaped tab found, use \\t instead"); 2671 warn = 0; /* already did */ 2672 } 2673 /*FALLTHROUGH*/ 2674 default: 2675 if (warn) { 2676 if (isprint((unsigned char)c)) { 2677 /* Allow escaping of 2678 * ``relations'' */ 2679 if (strchr("<>&^=!", c) == NULL 2680 && (m->type != FILE_REGEX || 2681 strchr("[]().*?^$|{}", c) 2682 == NULL)) { 2683 file_magwarn(ms, "no " 2684 "need to escape " 2685 "`%c'", c); 2686 } 2687 } else { 2688 file_magwarn(ms, 2689 "unknown escape sequence: " 2690 "\\%03o", c); 2691 } 2692 } 2693 /*FALLTHROUGH*/ 2694 /* space, perhaps force people to use \040? */ 2695 case ' ': 2696 #if 0 2697 /* 2698 * Other things people escape, but shouldn't need to, 2699 * so we disallow them 2700 */ 2701 case '\'': 2702 case '"': 2703 case '?': 2704 #endif 2705 /* Relations */ 2706 case '>': 2707 case '<': 2708 case '&': 2709 case '^': 2710 case '=': 2711 case '!': 2712 /* and baskslash itself */ 2713 case '\\': 2714 *p++ = (char) c; 2715 break; 2716 2717 case 'a': 2718 *p++ = '\a'; 2719 break; 2720 2721 case 'b': 2722 *p++ = '\b'; 2723 break; 2724 2725 case 'f': 2726 *p++ = '\f'; 2727 break; 2728 2729 case 'n': 2730 *p++ = '\n'; 2731 break; 2732 2733 case 'r': 2734 *p++ = '\r'; 2735 break; 2736 2737 case 't': 2738 *p++ = '\t'; 2739 break; 2740 2741 case 'v': 2742 *p++ = '\v'; 2743 break; 2744 2745 /* \ and up to 3 octal digits */ 2746 case '0': 2747 case '1': 2748 case '2': 2749 case '3': 2750 case '4': 2751 case '5': 2752 case '6': 2753 case '7': 2754 val = c - '0'; 2755 c = *s++; /* try for 2 */ 2756 if (c >= '0' && c <= '7') { 2757 val = (val << 3) | (c - '0'); 2758 c = *s++; /* try for 3 */ 2759 if (c >= '0' && c <= '7') 2760 val = (val << 3) | (c-'0'); 2761 else 2762 --s; 2763 } 2764 else 2765 --s; 2766 *p++ = (char)val; 2767 break; 2768 2769 /* \x and up to 2 hex digits */ 2770 case 'x': 2771 val = 'x'; /* Default if no digits */ 2772 c = hextoint(*s++); /* Get next char */ 2773 if (c >= 0) { 2774 val = c; 2775 c = hextoint(*s++); 2776 if (c >= 0) 2777 val = (val << 4) + c; 2778 else 2779 --s; 2780 } else 2781 --s; 2782 *p++ = (char)val; 2783 break; 2784 } 2785 } else 2786 *p++ = (char)c; 2787 } 2788 --s; 2789 out: 2790 *p = '\0'; 2791 m->vallen = CAST(unsigned char, (p - origp)); 2792 if (m->type == FILE_PSTRING) 2793 m->vallen += (unsigned char)file_pstring_length_size(m); 2794 return s; 2795 } 2796 2797 2798 /* Single hex char to int; -1 if not a hex char. */ 2799 private int 2800 hextoint(int c) 2801 { 2802 if (!isascii((unsigned char) c)) 2803 return -1; 2804 if (isdigit((unsigned char) c)) 2805 return c - '0'; 2806 if ((c >= 'a') && (c <= 'f')) 2807 return c + 10 - 'a'; 2808 if (( c>= 'A') && (c <= 'F')) 2809 return c + 10 - 'A'; 2810 return -1; 2811 } 2812 2813 2814 /* 2815 * Print a string containing C character escapes. 2816 */ 2817 protected void 2818 file_showstr(FILE *fp, const char *s, size_t len) 2819 { 2820 char c; 2821 2822 for (;;) { 2823 if (len == ~0U) { 2824 c = *s++; 2825 if (c == '\0') 2826 break; 2827 } 2828 else { 2829 if (len-- == 0) 2830 break; 2831 c = *s++; 2832 } 2833 if (c >= 040 && c <= 0176) /* TODO isprint && !iscntrl */ 2834 (void) fputc(c, fp); 2835 else { 2836 (void) fputc('\\', fp); 2837 switch (c) { 2838 case '\a': 2839 (void) fputc('a', fp); 2840 break; 2841 2842 case '\b': 2843 (void) fputc('b', fp); 2844 break; 2845 2846 case '\f': 2847 (void) fputc('f', fp); 2848 break; 2849 2850 case '\n': 2851 (void) fputc('n', fp); 2852 break; 2853 2854 case '\r': 2855 (void) fputc('r', fp); 2856 break; 2857 2858 case '\t': 2859 (void) fputc('t', fp); 2860 break; 2861 2862 case '\v': 2863 (void) fputc('v', fp); 2864 break; 2865 2866 default: 2867 (void) fprintf(fp, "%.3o", c & 0377); 2868 break; 2869 } 2870 } 2871 } 2872 } 2873 2874 /* 2875 * eatsize(): Eat the size spec from a number [eg. 10UL] 2876 */ 2877 private void 2878 eatsize(const char **p) 2879 { 2880 const char *l = *p; 2881 2882 if (LOWCASE(*l) == 'u') 2883 l++; 2884 2885 switch (LOWCASE(*l)) { 2886 case 'l': /* long */ 2887 case 's': /* short */ 2888 case 'h': /* short */ 2889 case 'b': /* char/byte */ 2890 case 'c': /* char/byte */ 2891 l++; 2892 /*FALLTHROUGH*/ 2893 default: 2894 break; 2895 } 2896 2897 *p = l; 2898 } 2899 2900 /* 2901 * handle a buffer containing a compiled file. 2902 */ 2903 private struct magic_map * 2904 apprentice_buf(struct magic_set *ms, struct magic *buf, size_t len) 2905 { 2906 struct magic_map *map; 2907 2908 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 2909 file_oomem(ms, sizeof(*map)); 2910 return NULL; 2911 } 2912 map->len = len; 2913 map->p = buf; 2914 map->type = MAP_TYPE_USER; 2915 if (check_buffer(ms, map, "buffer") != 0) { 2916 apprentice_unmap(map); 2917 return NULL; 2918 } 2919 return map; 2920 } 2921 2922 /* 2923 * handle a compiled file. 2924 */ 2925 2926 private struct magic_map * 2927 apprentice_map(struct magic_set *ms, const char *fn) 2928 { 2929 int fd; 2930 struct stat st; 2931 char *dbname = NULL; 2932 struct magic_map *map; 2933 2934 fd = -1; 2935 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 2936 file_oomem(ms, sizeof(*map)); 2937 goto error; 2938 } 2939 map->type = MAP_TYPE_USER; /* unspecified */ 2940 2941 dbname = mkdbname(ms, fn, 0); 2942 if (dbname == NULL) 2943 goto error; 2944 2945 if ((fd = open(dbname, O_RDONLY|O_BINARY)) == -1) 2946 goto error; 2947 2948 if (fstat(fd, &st) == -1) { 2949 file_error(ms, errno, "cannot stat `%s'", dbname); 2950 goto error; 2951 } 2952 if (st.st_size < 8 || st.st_size > MAXMAGIC_SIZE) { 2953 file_error(ms, 0, "file `%s' is too %s", dbname, 2954 st.st_size < 8 ? "small" : "large"); 2955 goto error; 2956 } 2957 2958 map->len = (size_t)st.st_size; 2959 #ifdef QUICK 2960 map->type = MAP_TYPE_MMAP; 2961 if ((map->p = mmap(0, (size_t)st.st_size, PROT_READ|PROT_WRITE, 2962 MAP_PRIVATE|MAP_FILE, fd, (off_t)0)) == MAP_FAILED) { 2963 file_error(ms, errno, "cannot map `%s'", dbname); 2964 goto error; 2965 } 2966 #else 2967 map->type = MAP_TYPE_MALLOC; 2968 if ((map->p = CAST(void *, malloc(map->len))) == NULL) { 2969 file_oomem(ms, map->len); 2970 goto error; 2971 } 2972 if (read(fd, map->p, map->len) != (ssize_t)map->len) { 2973 file_badread(ms); 2974 goto error; 2975 } 2976 #define RET 1 2977 #endif 2978 (void)close(fd); 2979 fd = -1; 2980 2981 if (check_buffer(ms, map, dbname) != 0) 2982 goto error; 2983 #ifdef QUICK 2984 if (mprotect(map->p, (size_t)st.st_size, PROT_READ) == -1) { 2985 file_error(ms, errno, "cannot mprotect `%s'", dbname); 2986 goto error; 2987 } 2988 #endif 2989 2990 free(dbname); 2991 return map; 2992 2993 error: 2994 if (fd != -1) 2995 (void)close(fd); 2996 apprentice_unmap(map); 2997 free(dbname); 2998 return NULL; 2999 } 3000 3001 private int 3002 check_buffer(struct magic_set *ms, struct magic_map *map, const char *dbname) 3003 { 3004 uint32_t *ptr; 3005 uint32_t entries, nentries; 3006 uint32_t version; 3007 int i, needsbyteswap; 3008 3009 ptr = CAST(uint32_t *, map->p); 3010 if (*ptr != MAGICNO) { 3011 if (swap4(*ptr) != MAGICNO) { 3012 file_error(ms, 0, "bad magic in `%s'", dbname); 3013 return -1; 3014 } 3015 needsbyteswap = 1; 3016 } else 3017 needsbyteswap = 0; 3018 if (needsbyteswap) 3019 version = swap4(ptr[1]); 3020 else 3021 version = ptr[1]; 3022 if (version != VERSIONNO) { 3023 file_error(ms, 0, "File %s supports only version %d magic " 3024 "files. `%s' is version %d", VERSION, 3025 VERSIONNO, dbname, version); 3026 return -1; 3027 } 3028 entries = (uint32_t)(map->len / sizeof(struct magic)); 3029 if ((entries * sizeof(struct magic)) != map->len) { 3030 file_error(ms, 0, "Size of `%s' %" SIZE_T_FORMAT "u is not " 3031 "a multiple of %" SIZE_T_FORMAT "u", 3032 dbname, map->len, sizeof(struct magic)); 3033 return -1; 3034 } 3035 map->magic[0] = CAST(struct magic *, map->p) + 1; 3036 nentries = 0; 3037 for (i = 0; i < MAGIC_SETS; i++) { 3038 if (needsbyteswap) 3039 map->nmagic[i] = swap4(ptr[i + 2]); 3040 else 3041 map->nmagic[i] = ptr[i + 2]; 3042 if (i != MAGIC_SETS - 1) 3043 map->magic[i + 1] = map->magic[i] + map->nmagic[i]; 3044 nentries += map->nmagic[i]; 3045 } 3046 if (entries != nentries + 1) { 3047 file_error(ms, 0, "Inconsistent entries in `%s' %u != %u", 3048 dbname, entries, nentries + 1); 3049 return -1; 3050 } 3051 if (needsbyteswap) 3052 for (i = 0; i < MAGIC_SETS; i++) 3053 byteswap(map->magic[i], map->nmagic[i]); 3054 return 0; 3055 } 3056 3057 /* 3058 * handle an mmaped file. 3059 */ 3060 private int 3061 apprentice_compile(struct magic_set *ms, struct magic_map *map, const char *fn) 3062 { 3063 static const size_t nm = sizeof(*map->nmagic) * MAGIC_SETS; 3064 static const size_t m = sizeof(**map->magic); 3065 int fd = -1; 3066 size_t len; 3067 char *dbname; 3068 int rv = -1; 3069 uint32_t i; 3070 union { 3071 struct magic m; 3072 uint32_t h[2 + MAGIC_SETS]; 3073 } hdr; 3074 3075 dbname = mkdbname(ms, fn, 1); 3076 3077 if (dbname == NULL) 3078 goto out; 3079 3080 if ((fd = open(dbname, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY, 0644)) == -1) 3081 { 3082 file_error(ms, errno, "cannot open `%s'", dbname); 3083 goto out; 3084 } 3085 memset(&hdr, 0, sizeof(hdr)); 3086 hdr.h[0] = MAGICNO; 3087 hdr.h[1] = VERSIONNO; 3088 memcpy(hdr.h + 2, map->nmagic, nm); 3089 3090 if (write(fd, &hdr, sizeof(hdr)) != (ssize_t)sizeof(hdr)) { 3091 file_error(ms, errno, "error writing `%s'", dbname); 3092 goto out; 3093 } 3094 3095 for (i = 0; i < MAGIC_SETS; i++) { 3096 len = m * map->nmagic[i]; 3097 if (write(fd, map->magic[i], len) != (ssize_t)len) { 3098 file_error(ms, errno, "error writing `%s'", dbname); 3099 goto out; 3100 } 3101 } 3102 3103 if (fd != -1) 3104 (void)close(fd); 3105 rv = 0; 3106 out: 3107 apprentice_unmap(map); 3108 free(dbname); 3109 return rv; 3110 } 3111 3112 private const char ext[] = ".mgc"; 3113 /* 3114 * make a dbname 3115 */ 3116 private char * 3117 mkdbname(struct magic_set *ms, const char *fn, int strip) 3118 { 3119 const char *p, *q; 3120 char *buf; 3121 3122 if (strip) { 3123 if ((p = strrchr(fn, '/')) != NULL) 3124 fn = ++p; 3125 } 3126 3127 for (q = fn; *q; q++) 3128 continue; 3129 /* Look for .mgc */ 3130 for (p = ext + sizeof(ext) - 1; p >= ext && q >= fn; p--, q--) 3131 if (*p != *q) 3132 break; 3133 3134 /* Did not find .mgc, restore q */ 3135 if (p >= ext) 3136 while (*q) 3137 q++; 3138 3139 q++; 3140 /* Compatibility with old code that looked in .mime */ 3141 if (ms->flags & MAGIC_MIME) { 3142 if (asprintf(&buf, "%.*s.mime%s", (int)(q - fn), fn, ext) < 0) 3143 return NULL; 3144 if (access(buf, R_OK) != -1) { 3145 ms->flags &= MAGIC_MIME_TYPE; 3146 return buf; 3147 } 3148 free(buf); 3149 } 3150 if (asprintf(&buf, "%.*s%s", (int)(q - fn), fn, ext) < 0) 3151 return NULL; 3152 3153 /* Compatibility with old code that looked in .mime */ 3154 if (strstr(p, ".mime") != NULL) 3155 ms->flags &= MAGIC_MIME_TYPE; 3156 return buf; 3157 } 3158 3159 /* 3160 * Byteswap an mmap'ed file if needed 3161 */ 3162 private void 3163 byteswap(struct magic *magic, uint32_t nmagic) 3164 { 3165 uint32_t i; 3166 for (i = 0; i < nmagic; i++) 3167 bs1(&magic[i]); 3168 } 3169 3170 /* 3171 * swap a short 3172 */ 3173 private uint16_t 3174 swap2(uint16_t sv) 3175 { 3176 uint16_t rv; 3177 uint8_t *s = (uint8_t *)(void *)&sv; 3178 uint8_t *d = (uint8_t *)(void *)&rv; 3179 d[0] = s[1]; 3180 d[1] = s[0]; 3181 return rv; 3182 } 3183 3184 /* 3185 * swap an int 3186 */ 3187 private uint32_t 3188 swap4(uint32_t sv) 3189 { 3190 uint32_t rv; 3191 uint8_t *s = (uint8_t *)(void *)&sv; 3192 uint8_t *d = (uint8_t *)(void *)&rv; 3193 d[0] = s[3]; 3194 d[1] = s[2]; 3195 d[2] = s[1]; 3196 d[3] = s[0]; 3197 return rv; 3198 } 3199 3200 /* 3201 * swap a quad 3202 */ 3203 private uint64_t 3204 swap8(uint64_t sv) 3205 { 3206 uint64_t rv; 3207 uint8_t *s = (uint8_t *)(void *)&sv; 3208 uint8_t *d = (uint8_t *)(void *)&rv; 3209 #if 0 3210 d[0] = s[3]; 3211 d[1] = s[2]; 3212 d[2] = s[1]; 3213 d[3] = s[0]; 3214 d[4] = s[7]; 3215 d[5] = s[6]; 3216 d[6] = s[5]; 3217 d[7] = s[4]; 3218 #else 3219 d[0] = s[7]; 3220 d[1] = s[6]; 3221 d[2] = s[5]; 3222 d[3] = s[4]; 3223 d[4] = s[3]; 3224 d[5] = s[2]; 3225 d[6] = s[1]; 3226 d[7] = s[0]; 3227 #endif 3228 return rv; 3229 } 3230 3231 /* 3232 * byteswap a single magic entry 3233 */ 3234 private void 3235 bs1(struct magic *m) 3236 { 3237 m->cont_level = swap2(m->cont_level); 3238 m->offset = swap4((uint32_t)m->offset); 3239 m->in_offset = swap4((uint32_t)m->in_offset); 3240 m->lineno = swap4((uint32_t)m->lineno); 3241 if (IS_STRING(m->type)) { 3242 m->str_range = swap4(m->str_range); 3243 m->str_flags = swap4(m->str_flags); 3244 } 3245 else { 3246 m->value.q = swap8(m->value.q); 3247 m->num_mask = swap8(m->num_mask); 3248 } 3249 } 3250 3251 protected size_t 3252 file_pstring_length_size(const struct magic *m) 3253 { 3254 switch (m->str_flags & PSTRING_LEN) { 3255 case PSTRING_1_LE: 3256 return 1; 3257 case PSTRING_2_LE: 3258 case PSTRING_2_BE: 3259 return 2; 3260 case PSTRING_4_LE: 3261 case PSTRING_4_BE: 3262 return 4; 3263 default: 3264 abort(); /* Impossible */ 3265 return 1; 3266 } 3267 } 3268 protected size_t 3269 file_pstring_get_length(const struct magic *m, const char *ss) 3270 { 3271 size_t len = 0; 3272 const unsigned char *s = (const unsigned char *)ss; 3273 3274 switch (m->str_flags & PSTRING_LEN) { 3275 case PSTRING_1_LE: 3276 len = *s; 3277 break; 3278 case PSTRING_2_LE: 3279 len = (s[1] << 8) | s[0]; 3280 break; 3281 case PSTRING_2_BE: 3282 len = (s[0] << 8) | s[1]; 3283 break; 3284 case PSTRING_4_LE: 3285 len = (s[3] << 24) | (s[2] << 16) | (s[1] << 8) | s[0]; 3286 break; 3287 case PSTRING_4_BE: 3288 len = (s[0] << 24) | (s[1] << 16) | (s[2] << 8) | s[3]; 3289 break; 3290 default: 3291 abort(); /* Impossible */ 3292 } 3293 3294 if (m->str_flags & PSTRING_LENGTH_INCLUDES_ITSELF) 3295 len -= file_pstring_length_size(m); 3296 3297 return len; 3298 } 3299 3300 protected int 3301 file_magicfind(struct magic_set *ms, const char *name, struct mlist *v) 3302 { 3303 uint32_t i, j; 3304 struct mlist *mlist, *ml; 3305 3306 mlist = ms->mlist[1]; 3307 3308 for (ml = mlist->next; ml != mlist; ml = ml->next) { 3309 struct magic *ma = ml->magic; 3310 uint32_t nma = ml->nmagic; 3311 for (i = 0; i < nma; i++) { 3312 if (ma[i].type != FILE_NAME) 3313 continue; 3314 if (strcmp(ma[i].value.s, name) == 0) { 3315 v->magic = &ma[i]; 3316 for (j = i + 1; j < nma; j++) 3317 if (ma[j].cont_level == 0) 3318 break; 3319 v->nmagic = j - i; 3320 return 0; 3321 } 3322 } 3323 } 3324 return -1; 3325 } 3326