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 1.238 2015/09/12 18:10:42 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_MMAP 0 90 #define MAP_TYPE_MALLOC 1 91 #define MAP_TYPE_USER 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); 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_NULL, FILE_INVALID, FILE_FMT_NONE }, 272 }; 273 274 /* 275 * These are not types, and cannot be preceded by "u" to make them 276 * unsigned. 277 */ 278 static const struct type_tbl_s special_tbl[] = { 279 { XX("name"), FILE_NAME, FILE_FMT_STR }, 280 { XX("use"), FILE_USE, FILE_FMT_STR }, 281 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 282 }; 283 # undef XX 284 # undef XX_NULL 285 286 private int 287 get_type(const struct type_tbl_s *tbl, const char *l, const char **t) 288 { 289 const struct type_tbl_s *p; 290 291 for (p = tbl; p->len; p++) { 292 if (strncmp(l, p->name, p->len) == 0) { 293 if (t) 294 *t = l + p->len; 295 break; 296 } 297 } 298 return p->type; 299 } 300 301 private int 302 get_standard_integer_type(const char *l, const char **t) 303 { 304 int type; 305 306 if (isalpha((unsigned char)l[1])) { 307 switch (l[1]) { 308 case 'C': 309 /* "dC" and "uC" */ 310 type = FILE_BYTE; 311 break; 312 case 'S': 313 /* "dS" and "uS" */ 314 type = FILE_SHORT; 315 break; 316 case 'I': 317 case 'L': 318 /* 319 * "dI", "dL", "uI", and "uL". 320 * 321 * XXX - the actual Single UNIX Specification says 322 * that "L" means "long", as in the C data type, 323 * but we treat it as meaning "4-byte integer". 324 * Given that the OS X version of file 5.04 did 325 * the same, I guess that passes the actual SUS 326 * validation suite; having "dL" be dependent on 327 * how big a "long" is on the machine running 328 * "file" is silly. 329 */ 330 type = FILE_LONG; 331 break; 332 case 'Q': 333 /* "dQ" and "uQ" */ 334 type = FILE_QUAD; 335 break; 336 default: 337 /* "d{anything else}", "u{anything else}" */ 338 return FILE_INVALID; 339 } 340 l += 2; 341 } else if (isdigit((unsigned char)l[1])) { 342 /* 343 * "d{num}" and "u{num}"; we only support {num} values 344 * of 1, 2, 4, and 8 - the Single UNIX Specification 345 * doesn't say anything about whether arbitrary 346 * values should be supported, but both the Solaris 10 347 * and OS X Mountain Lion versions of file passed the 348 * Single UNIX Specification validation suite, and 349 * neither of them support values bigger than 8 or 350 * non-power-of-2 values. 351 */ 352 if (isdigit((unsigned char)l[2])) { 353 /* Multi-digit, so > 9 */ 354 return FILE_INVALID; 355 } 356 switch (l[1]) { 357 case '1': 358 type = FILE_BYTE; 359 break; 360 case '2': 361 type = FILE_SHORT; 362 break; 363 case '4': 364 type = FILE_LONG; 365 break; 366 case '8': 367 type = FILE_QUAD; 368 break; 369 default: 370 /* XXX - what about 3, 5, 6, or 7? */ 371 return FILE_INVALID; 372 } 373 l += 2; 374 } else { 375 /* 376 * "d" or "u" by itself. 377 */ 378 type = FILE_LONG; 379 ++l; 380 } 381 if (t) 382 *t = l; 383 return type; 384 } 385 386 private void 387 init_file_tables(void) 388 { 389 static int done = 0; 390 const struct type_tbl_s *p; 391 392 if (done) 393 return; 394 done++; 395 396 for (p = type_tbl; p->len; p++) { 397 assert(p->type < FILE_NAMES_SIZE); 398 file_names[p->type] = p->name; 399 file_formats[p->type] = p->format; 400 } 401 assert(p - type_tbl == FILE_NAMES_SIZE); 402 } 403 404 private int 405 add_mlist(struct mlist *mlp, struct magic_map *map, size_t idx) 406 { 407 struct mlist *ml; 408 409 mlp->map = idx == 0 ? map : NULL; 410 if ((ml = CAST(struct mlist *, malloc(sizeof(*ml)))) == NULL) 411 return -1; 412 413 ml->map = NULL; 414 ml->magic = map->magic[idx]; 415 ml->nmagic = map->nmagic[idx]; 416 417 mlp->prev->next = ml; 418 ml->prev = mlp->prev; 419 ml->next = mlp; 420 mlp->prev = ml; 421 return 0; 422 } 423 424 /* 425 * Handle one file or directory. 426 */ 427 private int 428 apprentice_1(struct magic_set *ms, const char *fn, int action) 429 { 430 struct magic_map *map; 431 #ifndef COMPILE_ONLY 432 struct mlist *ml; 433 size_t i; 434 #endif 435 436 if (magicsize != FILE_MAGICSIZE) { 437 file_error(ms, 0, "magic element size %lu != %lu", 438 (unsigned long)sizeof(*map->magic[0]), 439 (unsigned long)FILE_MAGICSIZE); 440 return -1; 441 } 442 443 if (action == FILE_COMPILE) { 444 map = apprentice_load(ms, fn, action); 445 if (map == NULL) 446 return -1; 447 return apprentice_compile(ms, map, fn); 448 } 449 450 #ifndef COMPILE_ONLY 451 map = apprentice_map(ms, fn); 452 if (map == NULL) { 453 if (ms->flags & MAGIC_CHECK) 454 file_magwarn(ms, "using regular magic file `%s'", fn); 455 map = apprentice_load(ms, fn, action); 456 if (map == NULL) 457 return -1; 458 } 459 460 for (i = 0; i < MAGIC_SETS; i++) { 461 if (add_mlist(ms->mlist[i], map, i) == -1) { 462 file_oomem(ms, sizeof(*ml)); 463 goto fail; 464 } 465 } 466 467 if (action == FILE_LIST) { 468 for (i = 0; i < MAGIC_SETS; i++) { 469 printf("Set %" SIZE_T_FORMAT "u:\nBinary patterns:\n", 470 i); 471 apprentice_list(ms->mlist[i], BINTEST); 472 printf("Text patterns:\n"); 473 apprentice_list(ms->mlist[i], TEXTTEST); 474 } 475 } 476 return 0; 477 fail: 478 for (i = 0; i < MAGIC_SETS; i++) { 479 mlist_free(ms->mlist[i]); 480 ms->mlist[i] = NULL; 481 } 482 return -1; 483 #else 484 return 0; 485 #endif /* COMPILE_ONLY */ 486 } 487 488 protected void 489 file_ms_free(struct magic_set *ms) 490 { 491 size_t i; 492 if (ms == NULL) 493 return; 494 for (i = 0; i < MAGIC_SETS; i++) 495 mlist_free(ms->mlist[i]); 496 free(ms->o.pbuf); 497 free(ms->o.buf); 498 free(ms->c.li); 499 free(ms); 500 } 501 502 protected struct magic_set * 503 file_ms_alloc(int flags) 504 { 505 struct magic_set *ms; 506 size_t i, len; 507 508 if ((ms = CAST(struct magic_set *, calloc((size_t)1, 509 sizeof(struct magic_set)))) == NULL) 510 return NULL; 511 512 if (magic_setflags(ms, flags) == -1) { 513 errno = EINVAL; 514 goto free; 515 } 516 517 ms->o.buf = ms->o.pbuf = NULL; 518 len = (ms->c.len = 10) * sizeof(*ms->c.li); 519 520 if ((ms->c.li = CAST(struct level_info *, malloc(len))) == NULL) 521 goto free; 522 523 ms->event_flags = 0; 524 ms->error = -1; 525 for (i = 0; i < MAGIC_SETS; i++) 526 ms->mlist[i] = NULL; 527 ms->file = "unknown"; 528 ms->line = 0; 529 ms->indir_max = FILE_INDIR_MAX; 530 ms->name_max = FILE_NAME_MAX; 531 ms->elf_shnum_max = FILE_ELF_SHNUM_MAX; 532 ms->elf_phnum_max = FILE_ELF_PHNUM_MAX; 533 ms->elf_notes_max = FILE_ELF_NOTES_MAX; 534 ms->regex_max = FILE_REGEX_MAX; 535 return ms; 536 free: 537 free(ms); 538 return NULL; 539 } 540 541 private void 542 apprentice_unmap(struct magic_map *map) 543 { 544 size_t i; 545 if (map == NULL) 546 return; 547 548 switch (map->type) { 549 #ifdef QUICK 550 case MAP_TYPE_MMAP: 551 if (map->p) 552 (void)munmap(map->p, map->len); 553 break; 554 #endif 555 case MAP_TYPE_MALLOC: 556 free(map->p); 557 for (i = 0; i < MAGIC_SETS; i++) 558 free(map->magic[i]); 559 break; 560 case MAP_TYPE_USER: 561 break; 562 default: 563 abort(); 564 } 565 free(map); 566 } 567 568 private struct mlist * 569 mlist_alloc(void) 570 { 571 struct mlist *mlist; 572 if ((mlist = CAST(struct mlist *, calloc(1, sizeof(*mlist)))) == NULL) { 573 return NULL; 574 } 575 mlist->next = mlist->prev = mlist; 576 return mlist; 577 } 578 579 private void 580 mlist_free(struct mlist *mlist) 581 { 582 struct mlist *ml, *next; 583 584 if (mlist == NULL) 585 return; 586 587 ml = mlist->next; 588 for (ml = mlist->next; (next = ml->next) != NULL; ml = next) { 589 if (ml->map) 590 apprentice_unmap(ml->map); 591 free(ml); 592 if (ml == mlist) 593 break; 594 } 595 } 596 597 #ifndef COMPILE_ONLY 598 /* void **bufs: an array of compiled magic files */ 599 protected int 600 buffer_apprentice(struct magic_set *ms, struct magic **bufs, 601 size_t *sizes, size_t nbufs) 602 { 603 size_t i, j; 604 struct mlist *ml; 605 struct magic_map *map; 606 607 if (nbufs == 0) 608 return -1; 609 610 if (ms->mlist[0] != NULL) 611 file_reset(ms); 612 613 init_file_tables(); 614 615 for (i = 0; i < MAGIC_SETS; i++) { 616 mlist_free(ms->mlist[i]); 617 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 618 file_oomem(ms, sizeof(*ms->mlist[i])); 619 goto fail; 620 } 621 } 622 623 for (i = 0; i < nbufs; i++) { 624 map = apprentice_buf(ms, bufs[i], sizes[i]); 625 if (map == NULL) 626 goto fail; 627 628 for (j = 0; j < MAGIC_SETS; j++) { 629 if (add_mlist(ms->mlist[j], map, j) == -1) { 630 file_oomem(ms, sizeof(*ml)); 631 goto fail; 632 } 633 } 634 } 635 636 return 0; 637 fail: 638 for (i = 0; i < MAGIC_SETS; i++) { 639 mlist_free(ms->mlist[i]); 640 ms->mlist[i] = NULL; 641 } 642 return -1; 643 } 644 #endif 645 646 /* const char *fn: list of magic files and directories */ 647 protected int 648 file_apprentice(struct magic_set *ms, const char *fn, int action) 649 { 650 char *p, *mfn; 651 int file_err, errs = -1; 652 size_t i; 653 654 if (ms->mlist[0] != NULL) 655 file_reset(ms); 656 657 if ((fn = magic_getpath(fn, action)) == NULL) 658 return -1; 659 660 init_file_tables(); 661 662 if ((mfn = strdup(fn)) == NULL) { 663 file_oomem(ms, strlen(fn)); 664 return -1; 665 } 666 667 for (i = 0; i < MAGIC_SETS; i++) { 668 mlist_free(ms->mlist[i]); 669 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 670 file_oomem(ms, sizeof(*ms->mlist[i])); 671 while (i-- > 0) { 672 mlist_free(ms->mlist[i]); 673 ms->mlist[i] = NULL; 674 } 675 free(mfn); 676 return -1; 677 } 678 } 679 fn = mfn; 680 681 while (fn) { 682 p = strchr(fn, PATHSEP); 683 if (p) 684 *p++ = '\0'; 685 if (*fn == '\0') 686 break; 687 file_err = apprentice_1(ms, fn, action); 688 errs = MAX(errs, file_err); 689 fn = p; 690 } 691 692 free(mfn); 693 694 if (errs == -1) { 695 for (i = 0; i < MAGIC_SETS; i++) { 696 mlist_free(ms->mlist[i]); 697 ms->mlist[i] = NULL; 698 } 699 file_error(ms, 0, "could not find any valid magic files!"); 700 return -1; 701 } 702 703 #if 0 704 /* 705 * Always leave the database loaded 706 */ 707 if (action == FILE_LOAD) 708 return 0; 709 710 for (i = 0; i < MAGIC_SETS; i++) { 711 mlist_free(ms->mlist[i]); 712 ms->mlist[i] = NULL; 713 } 714 #endif 715 716 switch (action) { 717 case FILE_LOAD: 718 case FILE_COMPILE: 719 case FILE_CHECK: 720 case FILE_LIST: 721 return 0; 722 default: 723 file_error(ms, 0, "Invalid action %d", action); 724 return -1; 725 } 726 } 727 728 /* 729 * Compute the real length of a magic expression, for the purposes 730 * of determining how "strong" a magic expression is (approximating 731 * how specific its matches are): 732 * - magic characters count 0 unless escaped. 733 * - [] expressions count 1 734 * - {} expressions count 0 735 * - regular characters or escaped magic characters count 1 736 * - 0 length expressions count as one 737 */ 738 private size_t 739 nonmagic(const char *str) 740 { 741 const char *p; 742 size_t rv = 0; 743 744 for (p = str; *p; p++) 745 switch (*p) { 746 case '\\': /* Escaped anything counts 1 */ 747 if (!*++p) 748 p--; 749 rv++; 750 continue; 751 case '?': /* Magic characters count 0 */ 752 case '*': 753 case '.': 754 case '+': 755 case '^': 756 case '$': 757 continue; 758 case '[': /* Bracketed expressions count 1 the ']' */ 759 while (*p && *p != ']') 760 p++; 761 p--; 762 continue; 763 case '{': /* Braced expressions count 0 */ 764 while (*p && *p != '}') 765 p++; 766 if (!*p) 767 p--; 768 continue; 769 default: /* Anything else counts 1 */ 770 rv++; 771 continue; 772 } 773 774 return rv == 0 ? 1 : rv; /* Return at least 1 */ 775 } 776 777 /* 778 * Get weight of this magic entry, for sorting purposes. 779 */ 780 private size_t 781 apprentice_magic_strength(const struct magic *m) 782 { 783 #define MULT 10 784 size_t v, val = 2 * MULT; /* baseline strength */ 785 786 switch (m->type) { 787 case FILE_DEFAULT: /* make sure this sorts last */ 788 if (m->factor_op != FILE_FACTOR_OP_NONE) 789 abort(); 790 return 0; 791 792 case FILE_BYTE: 793 val += 1 * MULT; 794 break; 795 796 case FILE_SHORT: 797 case FILE_LESHORT: 798 case FILE_BESHORT: 799 val += 2 * MULT; 800 break; 801 802 case FILE_LONG: 803 case FILE_LELONG: 804 case FILE_BELONG: 805 case FILE_MELONG: 806 val += 4 * MULT; 807 break; 808 809 case FILE_PSTRING: 810 case FILE_STRING: 811 val += m->vallen * MULT; 812 break; 813 814 case FILE_BESTRING16: 815 case FILE_LESTRING16: 816 val += m->vallen * MULT / 2; 817 break; 818 819 case FILE_SEARCH: 820 val += m->vallen * MAX(MULT / m->vallen, 1); 821 break; 822 823 case FILE_REGEX: 824 v = nonmagic(m->value.s); 825 val += v * MAX(MULT / v, 1); 826 break; 827 828 case FILE_DATE: 829 case FILE_LEDATE: 830 case FILE_BEDATE: 831 case FILE_MEDATE: 832 case FILE_LDATE: 833 case FILE_LELDATE: 834 case FILE_BELDATE: 835 case FILE_MELDATE: 836 case FILE_FLOAT: 837 case FILE_BEFLOAT: 838 case FILE_LEFLOAT: 839 val += 4 * MULT; 840 break; 841 842 case FILE_QUAD: 843 case FILE_BEQUAD: 844 case FILE_LEQUAD: 845 case FILE_QDATE: 846 case FILE_LEQDATE: 847 case FILE_BEQDATE: 848 case FILE_QLDATE: 849 case FILE_LEQLDATE: 850 case FILE_BEQLDATE: 851 case FILE_QWDATE: 852 case FILE_LEQWDATE: 853 case FILE_BEQWDATE: 854 case FILE_DOUBLE: 855 case FILE_BEDOUBLE: 856 case FILE_LEDOUBLE: 857 val += 8 * MULT; 858 break; 859 860 case FILE_INDIRECT: 861 case FILE_NAME: 862 case FILE_USE: 863 break; 864 865 default: 866 (void)fprintf(stderr, "Bad type %d\n", m->type); 867 abort(); 868 } 869 870 switch (m->reln) { 871 case 'x': /* matches anything penalize */ 872 case '!': /* matches almost anything penalize */ 873 val = 0; 874 break; 875 876 case '=': /* Exact match, prefer */ 877 val += MULT; 878 break; 879 880 case '>': 881 case '<': /* comparison match reduce strength */ 882 val -= 2 * MULT; 883 break; 884 885 case '^': 886 case '&': /* masking bits, we could count them too */ 887 val -= MULT; 888 break; 889 890 default: 891 (void)fprintf(stderr, "Bad relation %c\n", m->reln); 892 abort(); 893 } 894 895 if (val == 0) /* ensure we only return 0 for FILE_DEFAULT */ 896 val = 1; 897 898 switch (m->factor_op) { 899 case FILE_FACTOR_OP_NONE: 900 break; 901 case FILE_FACTOR_OP_PLUS: 902 val += m->factor; 903 break; 904 case FILE_FACTOR_OP_MINUS: 905 val -= m->factor; 906 break; 907 case FILE_FACTOR_OP_TIMES: 908 val *= m->factor; 909 break; 910 case FILE_FACTOR_OP_DIV: 911 val /= m->factor; 912 break; 913 default: 914 abort(); 915 } 916 917 /* 918 * Magic entries with no description get a bonus because they depend 919 * on subsequent magic entries to print something. 920 */ 921 if (m->desc[0] == '\0') 922 val++; 923 return val; 924 } 925 926 /* 927 * Sort callback for sorting entries by "strength" (basically length) 928 */ 929 private int 930 apprentice_sort(const void *a, const void *b) 931 { 932 const struct magic_entry *ma = CAST(const struct magic_entry *, a); 933 const struct magic_entry *mb = CAST(const struct magic_entry *, b); 934 size_t sa = apprentice_magic_strength(ma->mp); 935 size_t sb = apprentice_magic_strength(mb->mp); 936 if (sa == sb) 937 return 0; 938 else if (sa > sb) 939 return -1; 940 else 941 return 1; 942 } 943 944 /* 945 * Shows sorted patterns list in the order which is used for the matching 946 */ 947 private void 948 apprentice_list(struct mlist *mlist, int mode) 949 { 950 uint32_t magindex = 0; 951 struct mlist *ml; 952 for (ml = mlist->next; ml != mlist; ml = ml->next) { 953 for (magindex = 0; magindex < ml->nmagic; magindex++) { 954 struct magic *m = &ml->magic[magindex]; 955 if ((m->flag & mode) != mode) { 956 /* Skip sub-tests */ 957 while (magindex + 1 < ml->nmagic && 958 ml->magic[magindex + 1].cont_level != 0) 959 ++magindex; 960 continue; /* Skip to next top-level test*/ 961 } 962 963 /* 964 * Try to iterate over the tree until we find item with 965 * description/mimetype. 966 */ 967 while (magindex + 1 < ml->nmagic && 968 ml->magic[magindex + 1].cont_level != 0 && 969 *ml->magic[magindex].desc == '\0' && 970 *ml->magic[magindex].mimetype == '\0') 971 magindex++; 972 973 printf("Strength = %3" SIZE_T_FORMAT "u@%u: %s [%s]\n", 974 apprentice_magic_strength(m), 975 ml->magic[magindex].lineno, 976 ml->magic[magindex].desc, 977 ml->magic[magindex].mimetype); 978 } 979 } 980 } 981 982 private void 983 set_test_type(struct magic *mstart, struct magic *m) 984 { 985 switch (m->type) { 986 case FILE_BYTE: 987 case FILE_SHORT: 988 case FILE_LONG: 989 case FILE_DATE: 990 case FILE_BESHORT: 991 case FILE_BELONG: 992 case FILE_BEDATE: 993 case FILE_LESHORT: 994 case FILE_LELONG: 995 case FILE_LEDATE: 996 case FILE_LDATE: 997 case FILE_BELDATE: 998 case FILE_LELDATE: 999 case FILE_MEDATE: 1000 case FILE_MELDATE: 1001 case FILE_MELONG: 1002 case FILE_QUAD: 1003 case FILE_LEQUAD: 1004 case FILE_BEQUAD: 1005 case FILE_QDATE: 1006 case FILE_LEQDATE: 1007 case FILE_BEQDATE: 1008 case FILE_QLDATE: 1009 case FILE_LEQLDATE: 1010 case FILE_BEQLDATE: 1011 case FILE_QWDATE: 1012 case FILE_LEQWDATE: 1013 case FILE_BEQWDATE: 1014 case FILE_FLOAT: 1015 case FILE_BEFLOAT: 1016 case FILE_LEFLOAT: 1017 case FILE_DOUBLE: 1018 case FILE_BEDOUBLE: 1019 case FILE_LEDOUBLE: 1020 mstart->flag |= BINTEST; 1021 break; 1022 case FILE_STRING: 1023 case FILE_PSTRING: 1024 case FILE_BESTRING16: 1025 case FILE_LESTRING16: 1026 /* Allow text overrides */ 1027 if (mstart->str_flags & STRING_TEXTTEST) 1028 mstart->flag |= TEXTTEST; 1029 else 1030 mstart->flag |= BINTEST; 1031 break; 1032 case FILE_REGEX: 1033 case FILE_SEARCH: 1034 /* Check for override */ 1035 if (mstart->str_flags & STRING_BINTEST) 1036 mstart->flag |= BINTEST; 1037 if (mstart->str_flags & STRING_TEXTTEST) 1038 mstart->flag |= TEXTTEST; 1039 1040 if (mstart->flag & (TEXTTEST|BINTEST)) 1041 break; 1042 1043 /* binary test if pattern is not text */ 1044 if (file_looks_utf8(m->value.us, (size_t)m->vallen, NULL, 1045 NULL) <= 0) 1046 mstart->flag |= BINTEST; 1047 else 1048 mstart->flag |= TEXTTEST; 1049 break; 1050 case FILE_DEFAULT: 1051 /* can't deduce anything; we shouldn't see this at the 1052 top level anyway */ 1053 break; 1054 case FILE_INVALID: 1055 default: 1056 /* invalid search type, but no need to complain here */ 1057 break; 1058 } 1059 } 1060 1061 private int 1062 addentry(struct magic_set *ms, struct magic_entry *me, 1063 struct magic_entry_set *mset) 1064 { 1065 size_t i = me->mp->type == FILE_NAME ? 1 : 0; 1066 if (mset[i].count == mset[i].max) { 1067 struct magic_entry *mp; 1068 1069 mset[i].max += ALLOC_INCR; 1070 if ((mp = CAST(struct magic_entry *, 1071 realloc(mset[i].me, sizeof(*mp) * mset[i].max))) == 1072 NULL) { 1073 file_oomem(ms, sizeof(*mp) * mset[i].max); 1074 return -1; 1075 } 1076 (void)memset(&mp[mset[i].count], 0, sizeof(*mp) * 1077 ALLOC_INCR); 1078 mset[i].me = mp; 1079 } 1080 mset[i].me[mset[i].count++] = *me; 1081 memset(me, 0, sizeof(*me)); 1082 return 0; 1083 } 1084 1085 /* 1086 * Load and parse one file. 1087 */ 1088 private void 1089 load_1(struct magic_set *ms, int action, const char *fn, int *errs, 1090 struct magic_entry_set *mset) 1091 { 1092 size_t lineno = 0, llen = 0; 1093 char *line = NULL; 1094 ssize_t len; 1095 struct magic_entry me; 1096 1097 FILE *f = fopen(ms->file = fn, "r"); 1098 if (f == NULL) { 1099 if (errno != ENOENT) 1100 file_error(ms, errno, "cannot read magic file `%s'", 1101 fn); 1102 (*errs)++; 1103 return; 1104 } 1105 1106 memset(&me, 0, sizeof(me)); 1107 /* read and parse this file */ 1108 for (ms->line = 1; (len = getline(&line, &llen, f)) != -1; 1109 ms->line++) { 1110 if (len == 0) /* null line, garbage, etc */ 1111 continue; 1112 if (line[len - 1] == '\n') { 1113 lineno++; 1114 line[len - 1] = '\0'; /* delete newline */ 1115 } 1116 switch (line[0]) { 1117 case '\0': /* empty, do not parse */ 1118 case '#': /* comment, do not parse */ 1119 continue; 1120 case '!': 1121 if (line[1] == ':') { 1122 size_t i; 1123 1124 for (i = 0; bang[i].name != NULL; i++) { 1125 if ((size_t)(len - 2) > bang[i].len && 1126 memcmp(bang[i].name, line + 2, 1127 bang[i].len) == 0) 1128 break; 1129 } 1130 if (bang[i].name == NULL) { 1131 file_error(ms, 0, 1132 "Unknown !: entry `%s'", line); 1133 (*errs)++; 1134 continue; 1135 } 1136 if (me.mp == NULL) { 1137 file_error(ms, 0, 1138 "No current entry for :!%s type", 1139 bang[i].name); 1140 (*errs)++; 1141 continue; 1142 } 1143 if ((*bang[i].fun)(ms, &me, 1144 line + bang[i].len + 2) != 0) { 1145 (*errs)++; 1146 continue; 1147 } 1148 continue; 1149 } 1150 /*FALLTHROUGH*/ 1151 default: 1152 again: 1153 switch (parse(ms, &me, line, lineno, action)) { 1154 case 0: 1155 continue; 1156 case 1: 1157 (void)addentry(ms, &me, mset); 1158 goto again; 1159 default: 1160 (*errs)++; 1161 break; 1162 } 1163 } 1164 } 1165 if (me.mp) 1166 (void)addentry(ms, &me, mset); 1167 free(line); 1168 (void)fclose(f); 1169 } 1170 1171 /* 1172 * parse a file or directory of files 1173 * const char *fn: name of magic file or directory 1174 */ 1175 private int 1176 cmpstrp(const void *p1, const void *p2) 1177 { 1178 return strcmp(*(char *const *)p1, *(char *const *)p2); 1179 } 1180 1181 1182 private uint32_t 1183 set_text_binary(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1184 uint32_t starttest) 1185 { 1186 static const char text[] = "text"; 1187 static const char binary[] = "binary"; 1188 static const size_t len = sizeof(text); 1189 1190 uint32_t i = starttest; 1191 1192 do { 1193 set_test_type(me[starttest].mp, me[i].mp); 1194 if ((ms->flags & MAGIC_DEBUG) == 0) 1195 continue; 1196 (void)fprintf(stderr, "%s%s%s: %s\n", 1197 me[i].mp->mimetype, 1198 me[i].mp->mimetype[0] == '\0' ? "" : "; ", 1199 me[i].mp->desc[0] ? me[i].mp->desc : "(no description)", 1200 me[i].mp->flag & BINTEST ? binary : text); 1201 if (me[i].mp->flag & BINTEST) { 1202 char *p = strstr(me[i].mp->desc, text); 1203 if (p && (p == me[i].mp->desc || 1204 isspace((unsigned char)p[-1])) && 1205 (p + len - me[i].mp->desc == MAXstring 1206 || (p[len] == '\0' || 1207 isspace((unsigned char)p[len])))) 1208 (void)fprintf(stderr, "*** Possible " 1209 "binary test for text type\n"); 1210 } 1211 } while (++i < nme && me[i].mp->cont_level != 0); 1212 return i; 1213 } 1214 1215 private void 1216 set_last_default(struct magic_set *ms, struct magic_entry *me, uint32_t nme) 1217 { 1218 uint32_t i; 1219 for (i = 0; i < nme; i++) { 1220 if (me[i].mp->cont_level == 0 && 1221 me[i].mp->type == FILE_DEFAULT) { 1222 while (++i < nme) 1223 if (me[i].mp->cont_level == 0) 1224 break; 1225 if (i != nme) { 1226 /* XXX - Ugh! */ 1227 ms->line = me[i].mp->lineno; 1228 file_magwarn(ms, 1229 "level 0 \"default\" did not sort last"); 1230 } 1231 return; 1232 } 1233 } 1234 } 1235 1236 private int 1237 coalesce_entries(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1238 struct magic **ma, uint32_t *nma) 1239 { 1240 uint32_t i, mentrycount = 0; 1241 size_t slen; 1242 1243 for (i = 0; i < nme; i++) 1244 mentrycount += me[i].cont_count; 1245 1246 slen = sizeof(**ma) * mentrycount; 1247 if ((*ma = CAST(struct magic *, malloc(slen))) == NULL) { 1248 file_oomem(ms, slen); 1249 return -1; 1250 } 1251 1252 mentrycount = 0; 1253 for (i = 0; i < nme; i++) { 1254 (void)memcpy(*ma + mentrycount, me[i].mp, 1255 me[i].cont_count * sizeof(**ma)); 1256 mentrycount += me[i].cont_count; 1257 } 1258 *nma = mentrycount; 1259 return 0; 1260 } 1261 1262 private void 1263 magic_entry_free(struct magic_entry *me, uint32_t nme) 1264 { 1265 uint32_t i; 1266 if (me == NULL) 1267 return; 1268 for (i = 0; i < nme; i++) 1269 free(me[i].mp); 1270 free(me); 1271 } 1272 1273 private struct magic_map * 1274 apprentice_load(struct magic_set *ms, const char *fn, int action) 1275 { 1276 int errs = 0; 1277 uint32_t i, j; 1278 size_t files = 0, maxfiles = 0; 1279 char **filearr = NULL, *mfn; 1280 struct stat st; 1281 struct magic_map *map; 1282 struct magic_entry_set mset[MAGIC_SETS]; 1283 DIR *dir; 1284 struct dirent *d; 1285 1286 memset(mset, 0, sizeof(mset)); 1287 ms->flags |= MAGIC_CHECK; /* Enable checks for parsed files */ 1288 1289 1290 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) 1291 { 1292 file_oomem(ms, sizeof(*map)); 1293 return NULL; 1294 } 1295 map->type = MAP_TYPE_MALLOC; 1296 1297 /* print silly verbose header for USG compat. */ 1298 if (action == FILE_CHECK) 1299 (void)fprintf(stderr, "%s\n", usg_hdr); 1300 1301 /* load directory or file */ 1302 if (stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) { 1303 dir = opendir(fn); 1304 if (!dir) { 1305 errs++; 1306 goto out; 1307 } 1308 while ((d = readdir(dir)) != NULL) { 1309 if (asprintf(&mfn, "%s/%s", fn, d->d_name) < 0) { 1310 file_oomem(ms, 1311 strlen(fn) + strlen(d->d_name) + 2); 1312 errs++; 1313 closedir(dir); 1314 goto out; 1315 } 1316 if (stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) { 1317 free(mfn); 1318 continue; 1319 } 1320 if (files >= maxfiles) { 1321 size_t mlen; 1322 maxfiles = (maxfiles + 1) * 2; 1323 mlen = maxfiles * sizeof(*filearr); 1324 if ((filearr = CAST(char **, 1325 realloc(filearr, mlen))) == NULL) { 1326 file_oomem(ms, mlen); 1327 free(mfn); 1328 closedir(dir); 1329 errs++; 1330 goto out; 1331 } 1332 } 1333 filearr[files++] = mfn; 1334 } 1335 closedir(dir); 1336 qsort(filearr, files, sizeof(*filearr), cmpstrp); 1337 for (i = 0; i < files; i++) { 1338 load_1(ms, action, filearr[i], &errs, mset); 1339 free(filearr[i]); 1340 } 1341 free(filearr); 1342 } else 1343 load_1(ms, action, fn, &errs, mset); 1344 if (errs) 1345 goto out; 1346 1347 for (j = 0; j < MAGIC_SETS; j++) { 1348 /* Set types of tests */ 1349 for (i = 0; i < mset[j].count; ) { 1350 if (mset[j].me[i].mp->cont_level != 0) { 1351 i++; 1352 continue; 1353 } 1354 i = set_text_binary(ms, mset[j].me, mset[j].count, i); 1355 } 1356 if (mset[j].me) 1357 qsort(mset[j].me, mset[j].count, sizeof(*mset[j].me), 1358 apprentice_sort); 1359 1360 /* 1361 * Make sure that any level 0 "default" line is last 1362 * (if one exists). 1363 */ 1364 set_last_default(ms, mset[j].me, mset[j].count); 1365 1366 /* coalesce per file arrays into a single one */ 1367 if (coalesce_entries(ms, mset[j].me, mset[j].count, 1368 &map->magic[j], &map->nmagic[j]) == -1) { 1369 errs++; 1370 goto out; 1371 } 1372 } 1373 1374 out: 1375 for (j = 0; j < MAGIC_SETS; j++) 1376 magic_entry_free(mset[j].me, mset[j].count); 1377 1378 if (errs) { 1379 apprentice_unmap(map); 1380 return NULL; 1381 } 1382 return map; 1383 } 1384 1385 /* 1386 * extend the sign bit if the comparison is to be signed 1387 */ 1388 protected uint64_t 1389 file_signextend(struct magic_set *ms, struct magic *m, uint64_t v) 1390 { 1391 if (!(m->flag & UNSIGNED)) { 1392 switch(m->type) { 1393 /* 1394 * Do not remove the casts below. They are 1395 * vital. When later compared with the data, 1396 * the sign extension must have happened. 1397 */ 1398 case FILE_BYTE: 1399 v = (signed char) v; 1400 break; 1401 case FILE_SHORT: 1402 case FILE_BESHORT: 1403 case FILE_LESHORT: 1404 v = (short) v; 1405 break; 1406 case FILE_DATE: 1407 case FILE_BEDATE: 1408 case FILE_LEDATE: 1409 case FILE_MEDATE: 1410 case FILE_LDATE: 1411 case FILE_BELDATE: 1412 case FILE_LELDATE: 1413 case FILE_MELDATE: 1414 case FILE_LONG: 1415 case FILE_BELONG: 1416 case FILE_LELONG: 1417 case FILE_MELONG: 1418 case FILE_FLOAT: 1419 case FILE_BEFLOAT: 1420 case FILE_LEFLOAT: 1421 v = (int32_t) v; 1422 break; 1423 case FILE_QUAD: 1424 case FILE_BEQUAD: 1425 case FILE_LEQUAD: 1426 case FILE_QDATE: 1427 case FILE_QLDATE: 1428 case FILE_QWDATE: 1429 case FILE_BEQDATE: 1430 case FILE_BEQLDATE: 1431 case FILE_BEQWDATE: 1432 case FILE_LEQDATE: 1433 case FILE_LEQLDATE: 1434 case FILE_LEQWDATE: 1435 case FILE_DOUBLE: 1436 case FILE_BEDOUBLE: 1437 case FILE_LEDOUBLE: 1438 v = (int64_t) v; 1439 break; 1440 case FILE_STRING: 1441 case FILE_PSTRING: 1442 case FILE_BESTRING16: 1443 case FILE_LESTRING16: 1444 case FILE_REGEX: 1445 case FILE_SEARCH: 1446 case FILE_DEFAULT: 1447 case FILE_INDIRECT: 1448 case FILE_NAME: 1449 case FILE_USE: 1450 case FILE_CLEAR: 1451 break; 1452 default: 1453 if (ms->flags & MAGIC_CHECK) 1454 file_magwarn(ms, "cannot happen: m->type=%d\n", 1455 m->type); 1456 return ~0U; 1457 } 1458 } 1459 return v; 1460 } 1461 1462 private int 1463 string_modifier_check(struct magic_set *ms, struct magic *m) 1464 { 1465 if ((ms->flags & MAGIC_CHECK) == 0) 1466 return 0; 1467 1468 if ((m->type != FILE_REGEX || (m->str_flags & REGEX_LINE_COUNT) == 0) && 1469 (m->type != FILE_PSTRING && (m->str_flags & PSTRING_LEN) != 0)) { 1470 file_magwarn(ms, 1471 "'/BHhLl' modifiers are only allowed for pascal strings\n"); 1472 return -1; 1473 } 1474 switch (m->type) { 1475 case FILE_BESTRING16: 1476 case FILE_LESTRING16: 1477 if (m->str_flags != 0) { 1478 file_magwarn(ms, 1479 "no modifiers allowed for 16-bit strings\n"); 1480 return -1; 1481 } 1482 break; 1483 case FILE_STRING: 1484 case FILE_PSTRING: 1485 if ((m->str_flags & REGEX_OFFSET_START) != 0) { 1486 file_magwarn(ms, 1487 "'/%c' only allowed on regex and search\n", 1488 CHAR_REGEX_OFFSET_START); 1489 return -1; 1490 } 1491 break; 1492 case FILE_SEARCH: 1493 if (m->str_range == 0) { 1494 file_magwarn(ms, 1495 "missing range; defaulting to %d\n", 1496 STRING_DEFAULT_RANGE); 1497 m->str_range = STRING_DEFAULT_RANGE; 1498 return -1; 1499 } 1500 break; 1501 case FILE_REGEX: 1502 if ((m->str_flags & STRING_COMPACT_WHITESPACE) != 0) { 1503 file_magwarn(ms, "'/%c' not allowed on regex\n", 1504 CHAR_COMPACT_WHITESPACE); 1505 return -1; 1506 } 1507 if ((m->str_flags & STRING_COMPACT_OPTIONAL_WHITESPACE) != 0) { 1508 file_magwarn(ms, "'/%c' not allowed on regex\n", 1509 CHAR_COMPACT_OPTIONAL_WHITESPACE); 1510 return -1; 1511 } 1512 break; 1513 default: 1514 file_magwarn(ms, "coding error: m->type=%d\n", 1515 m->type); 1516 return -1; 1517 } 1518 return 0; 1519 } 1520 1521 private int 1522 get_op(char c) 1523 { 1524 switch (c) { 1525 case '&': 1526 return FILE_OPAND; 1527 case '|': 1528 return FILE_OPOR; 1529 case '^': 1530 return FILE_OPXOR; 1531 case '+': 1532 return FILE_OPADD; 1533 case '-': 1534 return FILE_OPMINUS; 1535 case '*': 1536 return FILE_OPMULTIPLY; 1537 case '/': 1538 return FILE_OPDIVIDE; 1539 case '%': 1540 return FILE_OPMODULO; 1541 default: 1542 return -1; 1543 } 1544 } 1545 1546 #ifdef ENABLE_CONDITIONALS 1547 private int 1548 get_cond(const char *l, const char **t) 1549 { 1550 static const struct cond_tbl_s { 1551 char name[8]; 1552 size_t len; 1553 int cond; 1554 } cond_tbl[] = { 1555 { "if", 2, COND_IF }, 1556 { "elif", 4, COND_ELIF }, 1557 { "else", 4, COND_ELSE }, 1558 { "", 0, COND_NONE }, 1559 }; 1560 const struct cond_tbl_s *p; 1561 1562 for (p = cond_tbl; p->len; p++) { 1563 if (strncmp(l, p->name, p->len) == 0 && 1564 isspace((unsigned char)l[p->len])) { 1565 if (t) 1566 *t = l + p->len; 1567 break; 1568 } 1569 } 1570 return p->cond; 1571 } 1572 1573 private int 1574 check_cond(struct magic_set *ms, int cond, uint32_t cont_level) 1575 { 1576 int last_cond; 1577 last_cond = ms->c.li[cont_level].last_cond; 1578 1579 switch (cond) { 1580 case COND_IF: 1581 if (last_cond != COND_NONE && last_cond != COND_ELIF) { 1582 if (ms->flags & MAGIC_CHECK) 1583 file_magwarn(ms, "syntax error: `if'"); 1584 return -1; 1585 } 1586 last_cond = COND_IF; 1587 break; 1588 1589 case COND_ELIF: 1590 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1591 if (ms->flags & MAGIC_CHECK) 1592 file_magwarn(ms, "syntax error: `elif'"); 1593 return -1; 1594 } 1595 last_cond = COND_ELIF; 1596 break; 1597 1598 case COND_ELSE: 1599 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1600 if (ms->flags & MAGIC_CHECK) 1601 file_magwarn(ms, "syntax error: `else'"); 1602 return -1; 1603 } 1604 last_cond = COND_NONE; 1605 break; 1606 1607 case COND_NONE: 1608 last_cond = COND_NONE; 1609 break; 1610 } 1611 1612 ms->c.li[cont_level].last_cond = last_cond; 1613 return 0; 1614 } 1615 #endif /* ENABLE_CONDITIONALS */ 1616 1617 private int 1618 parse_indirect_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1619 { 1620 const char *l = *lp; 1621 1622 while (!isspace((unsigned char)*++l)) 1623 switch (*l) { 1624 case CHAR_INDIRECT_RELATIVE: 1625 m->str_flags |= INDIRECT_RELATIVE; 1626 break; 1627 default: 1628 if (ms->flags & MAGIC_CHECK) 1629 file_magwarn(ms, "indirect modifier `%c' " 1630 "invalid", *l); 1631 *lp = l; 1632 return -1; 1633 } 1634 *lp = l; 1635 return 0; 1636 } 1637 1638 private void 1639 parse_op_modifier(struct magic_set *ms, struct magic *m, const char **lp, 1640 int op) 1641 { 1642 const char *l = *lp; 1643 char *t; 1644 uint64_t val; 1645 1646 ++l; 1647 m->mask_op |= op; 1648 val = (uint64_t)strtoull(l, &t, 0); 1649 l = t; 1650 m->num_mask = file_signextend(ms, m, val); 1651 eatsize(&l); 1652 *lp = l; 1653 } 1654 1655 private int 1656 parse_string_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1657 { 1658 const char *l = *lp; 1659 char *t; 1660 int have_range = 0; 1661 1662 while (!isspace((unsigned char)*++l)) { 1663 switch (*l) { 1664 case '0': case '1': case '2': 1665 case '3': case '4': case '5': 1666 case '6': case '7': case '8': 1667 case '9': 1668 if (have_range && (ms->flags & MAGIC_CHECK)) 1669 file_magwarn(ms, "multiple ranges"); 1670 have_range = 1; 1671 m->str_range = CAST(uint32_t, strtoul(l, &t, 0)); 1672 if (m->str_range == 0) 1673 file_magwarn(ms, "zero range"); 1674 l = t - 1; 1675 break; 1676 case CHAR_COMPACT_WHITESPACE: 1677 m->str_flags |= STRING_COMPACT_WHITESPACE; 1678 break; 1679 case CHAR_COMPACT_OPTIONAL_WHITESPACE: 1680 m->str_flags |= STRING_COMPACT_OPTIONAL_WHITESPACE; 1681 break; 1682 case CHAR_IGNORE_LOWERCASE: 1683 m->str_flags |= STRING_IGNORE_LOWERCASE; 1684 break; 1685 case CHAR_IGNORE_UPPERCASE: 1686 m->str_flags |= STRING_IGNORE_UPPERCASE; 1687 break; 1688 case CHAR_REGEX_OFFSET_START: 1689 m->str_flags |= REGEX_OFFSET_START; 1690 break; 1691 case CHAR_BINTEST: 1692 m->str_flags |= STRING_BINTEST; 1693 break; 1694 case CHAR_TEXTTEST: 1695 m->str_flags |= STRING_TEXTTEST; 1696 break; 1697 case CHAR_TRIM: 1698 m->str_flags |= STRING_TRIM; 1699 break; 1700 case CHAR_PSTRING_1_LE: 1701 #define SET_LENGTH(a) m->str_flags = (m->str_flags & ~PSTRING_LEN) | (a) 1702 if (m->type != FILE_PSTRING) 1703 goto bad; 1704 SET_LENGTH(PSTRING_1_LE); 1705 break; 1706 case CHAR_PSTRING_2_BE: 1707 if (m->type != FILE_PSTRING) 1708 goto bad; 1709 SET_LENGTH(PSTRING_2_BE); 1710 break; 1711 case CHAR_PSTRING_2_LE: 1712 if (m->type != FILE_PSTRING) 1713 goto bad; 1714 SET_LENGTH(PSTRING_2_LE); 1715 break; 1716 case CHAR_PSTRING_4_BE: 1717 if (m->type != FILE_PSTRING) 1718 goto bad; 1719 SET_LENGTH(PSTRING_4_BE); 1720 break; 1721 case CHAR_PSTRING_4_LE: 1722 switch (m->type) { 1723 case FILE_PSTRING: 1724 case FILE_REGEX: 1725 break; 1726 default: 1727 goto bad; 1728 } 1729 SET_LENGTH(PSTRING_4_LE); 1730 break; 1731 case CHAR_PSTRING_LENGTH_INCLUDES_ITSELF: 1732 if (m->type != FILE_PSTRING) 1733 goto bad; 1734 m->str_flags |= PSTRING_LENGTH_INCLUDES_ITSELF; 1735 break; 1736 default: 1737 bad: 1738 if (ms->flags & MAGIC_CHECK) 1739 file_magwarn(ms, "string modifier `%c' " 1740 "invalid", *l); 1741 goto out; 1742 } 1743 /* allow multiple '/' for readability */ 1744 if (l[1] == '/' && !isspace((unsigned char)l[2])) 1745 l++; 1746 } 1747 if (string_modifier_check(ms, m) == -1) 1748 goto out; 1749 *lp = l; 1750 return 0; 1751 out: 1752 *lp = l; 1753 return -1; 1754 } 1755 1756 /* 1757 * parse one line from magic file, put into magic[index++] if valid 1758 */ 1759 private int 1760 parse(struct magic_set *ms, struct magic_entry *me, const char *line, 1761 size_t lineno, int action) 1762 { 1763 #ifdef ENABLE_CONDITIONALS 1764 static uint32_t last_cont_level = 0; 1765 #endif 1766 size_t i; 1767 struct magic *m; 1768 const char *l = line; 1769 char *t; 1770 int op; 1771 uint32_t cont_level; 1772 int32_t diff; 1773 1774 cont_level = 0; 1775 1776 /* 1777 * Parse the offset. 1778 */ 1779 while (*l == '>') { 1780 ++l; /* step over */ 1781 cont_level++; 1782 } 1783 #ifdef ENABLE_CONDITIONALS 1784 if (cont_level == 0 || cont_level > last_cont_level) 1785 if (file_check_mem(ms, cont_level) == -1) 1786 return -1; 1787 last_cont_level = cont_level; 1788 #endif 1789 if (cont_level != 0) { 1790 if (me->mp == NULL) { 1791 file_magerror(ms, "No current entry for continuation"); 1792 return -1; 1793 } 1794 if (me->cont_count == 0) { 1795 file_magerror(ms, "Continuations present with 0 count"); 1796 return -1; 1797 } 1798 m = &me->mp[me->cont_count - 1]; 1799 diff = (int32_t)cont_level - (int32_t)m->cont_level; 1800 if (diff > 1) 1801 file_magwarn(ms, "New continuation level %u is more " 1802 "than one larger than current level %u", cont_level, 1803 m->cont_level); 1804 if (me->cont_count == me->max_count) { 1805 struct magic *nm; 1806 size_t cnt = me->max_count + ALLOC_CHUNK; 1807 if ((nm = CAST(struct magic *, realloc(me->mp, 1808 sizeof(*nm) * cnt))) == NULL) { 1809 file_oomem(ms, sizeof(*nm) * cnt); 1810 return -1; 1811 } 1812 me->mp = m = nm; 1813 me->max_count = CAST(uint32_t, cnt); 1814 } 1815 m = &me->mp[me->cont_count++]; 1816 (void)memset(m, 0, sizeof(*m)); 1817 m->cont_level = cont_level; 1818 } else { 1819 static const size_t len = sizeof(*m) * ALLOC_CHUNK; 1820 if (me->mp != NULL) 1821 return 1; 1822 if ((m = CAST(struct magic *, malloc(len))) == NULL) { 1823 file_oomem(ms, len); 1824 return -1; 1825 } 1826 me->mp = m; 1827 me->max_count = ALLOC_CHUNK; 1828 (void)memset(m, 0, sizeof(*m)); 1829 m->factor_op = FILE_FACTOR_OP_NONE; 1830 m->cont_level = 0; 1831 me->cont_count = 1; 1832 } 1833 m->lineno = CAST(uint32_t, lineno); 1834 1835 if (*l == '&') { /* m->cont_level == 0 checked below. */ 1836 ++l; /* step over */ 1837 m->flag |= OFFADD; 1838 } 1839 if (*l == '(') { 1840 ++l; /* step over */ 1841 m->flag |= INDIR; 1842 if (m->flag & OFFADD) 1843 m->flag = (m->flag & ~OFFADD) | INDIROFFADD; 1844 1845 if (*l == '&') { /* m->cont_level == 0 checked below */ 1846 ++l; /* step over */ 1847 m->flag |= OFFADD; 1848 } 1849 } 1850 /* Indirect offsets are not valid at level 0. */ 1851 if (m->cont_level == 0 && (m->flag & (OFFADD | INDIROFFADD))) { 1852 if (ms->flags & MAGIC_CHECK) 1853 file_magwarn(ms, "relative offset at level 0"); 1854 return -1; 1855 } 1856 1857 /* get offset, then skip over it */ 1858 m->offset = (uint32_t)strtoul(l, &t, 0); 1859 if (l == t) { 1860 if (ms->flags & MAGIC_CHECK) 1861 file_magwarn(ms, "offset `%s' invalid", l); 1862 return -1; 1863 } 1864 l = t; 1865 1866 if (m->flag & INDIR) { 1867 m->in_type = FILE_LONG; 1868 m->in_offset = 0; 1869 /* 1870 * read [.lbs][+-]nnnnn) 1871 */ 1872 if (*l == '.') { 1873 l++; 1874 switch (*l) { 1875 case 'l': 1876 m->in_type = FILE_LELONG; 1877 break; 1878 case 'L': 1879 m->in_type = FILE_BELONG; 1880 break; 1881 case 'm': 1882 m->in_type = FILE_MELONG; 1883 break; 1884 case 'h': 1885 case 's': 1886 m->in_type = FILE_LESHORT; 1887 break; 1888 case 'H': 1889 case 'S': 1890 m->in_type = FILE_BESHORT; 1891 break; 1892 case 'c': 1893 case 'b': 1894 case 'C': 1895 case 'B': 1896 m->in_type = FILE_BYTE; 1897 break; 1898 case 'e': 1899 case 'f': 1900 case 'g': 1901 m->in_type = FILE_LEDOUBLE; 1902 break; 1903 case 'E': 1904 case 'F': 1905 case 'G': 1906 m->in_type = FILE_BEDOUBLE; 1907 break; 1908 case 'i': 1909 m->in_type = FILE_LEID3; 1910 break; 1911 case 'I': 1912 m->in_type = FILE_BEID3; 1913 break; 1914 default: 1915 if (ms->flags & MAGIC_CHECK) 1916 file_magwarn(ms, 1917 "indirect offset type `%c' invalid", 1918 *l); 1919 return -1; 1920 } 1921 l++; 1922 } 1923 1924 m->in_op = 0; 1925 if (*l == '~') { 1926 m->in_op |= FILE_OPINVERSE; 1927 l++; 1928 } 1929 if ((op = get_op(*l)) != -1) { 1930 m->in_op |= op; 1931 l++; 1932 } 1933 if (*l == '(') { 1934 m->in_op |= FILE_OPINDIRECT; 1935 l++; 1936 } 1937 if (isdigit((unsigned char)*l) || *l == '-') { 1938 m->in_offset = (int32_t)strtol(l, &t, 0); 1939 if (l == t) { 1940 if (ms->flags & MAGIC_CHECK) 1941 file_magwarn(ms, 1942 "in_offset `%s' invalid", l); 1943 return -1; 1944 } 1945 l = t; 1946 } 1947 if (*l++ != ')' || 1948 ((m->in_op & FILE_OPINDIRECT) && *l++ != ')')) { 1949 if (ms->flags & MAGIC_CHECK) 1950 file_magwarn(ms, 1951 "missing ')' in indirect offset"); 1952 return -1; 1953 } 1954 } 1955 EATAB; 1956 1957 #ifdef ENABLE_CONDITIONALS 1958 m->cond = get_cond(l, &l); 1959 if (check_cond(ms, m->cond, cont_level) == -1) 1960 return -1; 1961 1962 EATAB; 1963 #endif 1964 1965 /* 1966 * Parse the type. 1967 */ 1968 if (*l == 'u') { 1969 /* 1970 * Try it as a keyword type prefixed by "u"; match what 1971 * follows the "u". If that fails, try it as an SUS 1972 * integer type. 1973 */ 1974 m->type = get_type(type_tbl, l + 1, &l); 1975 if (m->type == FILE_INVALID) { 1976 /* 1977 * Not a keyword type; parse it as an SUS type, 1978 * 'u' possibly followed by a number or C/S/L. 1979 */ 1980 m->type = get_standard_integer_type(l, &l); 1981 } 1982 /* It's unsigned. */ 1983 if (m->type != FILE_INVALID) 1984 m->flag |= UNSIGNED; 1985 } else { 1986 /* 1987 * Try it as a keyword type. If that fails, try it as 1988 * an SUS integer type if it begins with "d" or as an 1989 * SUS string type if it begins with "s". In any case, 1990 * it's not unsigned. 1991 */ 1992 m->type = get_type(type_tbl, l, &l); 1993 if (m->type == FILE_INVALID) { 1994 /* 1995 * Not a keyword type; parse it as an SUS type, 1996 * either 'd' possibly followed by a number or 1997 * C/S/L, or just 's'. 1998 */ 1999 if (*l == 'd') 2000 m->type = get_standard_integer_type(l, &l); 2001 else if (*l == 's' && !isalpha((unsigned char)l[1])) { 2002 m->type = FILE_STRING; 2003 ++l; 2004 } 2005 } 2006 } 2007 2008 if (m->type == FILE_INVALID) { 2009 /* Not found - try it as a special keyword. */ 2010 m->type = get_type(special_tbl, l, &l); 2011 } 2012 2013 if (m->type == FILE_INVALID) { 2014 if (ms->flags & MAGIC_CHECK) 2015 file_magwarn(ms, "type `%s' invalid", l); 2016 return -1; 2017 } 2018 2019 /* New-style anding: "0 byte&0x80 =0x80 dynamically linked" */ 2020 /* New and improved: ~ & | ^ + - * / % -- exciting, isn't it? */ 2021 2022 m->mask_op = 0; 2023 if (*l == '~') { 2024 if (!IS_STRING(m->type)) 2025 m->mask_op |= FILE_OPINVERSE; 2026 else if (ms->flags & MAGIC_CHECK) 2027 file_magwarn(ms, "'~' invalid for string types"); 2028 ++l; 2029 } 2030 m->str_range = 0; 2031 m->str_flags = m->type == FILE_PSTRING ? PSTRING_1_LE : 0; 2032 if ((op = get_op(*l)) != -1) { 2033 if (IS_STRING(m->type)) { 2034 int r; 2035 2036 if (op != FILE_OPDIVIDE) { 2037 if (ms->flags & MAGIC_CHECK) 2038 file_magwarn(ms, 2039 "invalid string/indirect op: " 2040 "`%c'", *t); 2041 return -1; 2042 } 2043 2044 if (m->type == FILE_INDIRECT) 2045 r = parse_indirect_modifier(ms, m, &l); 2046 else 2047 r = parse_string_modifier(ms, m, &l); 2048 if (r == -1) 2049 return -1; 2050 } else 2051 parse_op_modifier(ms, m, &l, op); 2052 } 2053 2054 /* 2055 * We used to set mask to all 1's here, instead let's just not do 2056 * anything if mask = 0 (unless you have a better idea) 2057 */ 2058 EATAB; 2059 2060 switch (*l) { 2061 case '>': 2062 case '<': 2063 m->reln = *l; 2064 ++l; 2065 if (*l == '=') { 2066 if (ms->flags & MAGIC_CHECK) { 2067 file_magwarn(ms, "%c= not supported", 2068 m->reln); 2069 return -1; 2070 } 2071 ++l; 2072 } 2073 break; 2074 /* Old-style anding: "0 byte &0x80 dynamically linked" */ 2075 case '&': 2076 case '^': 2077 case '=': 2078 m->reln = *l; 2079 ++l; 2080 if (*l == '=') { 2081 /* HP compat: ignore &= etc. */ 2082 ++l; 2083 } 2084 break; 2085 case '!': 2086 m->reln = *l; 2087 ++l; 2088 break; 2089 default: 2090 m->reln = '='; /* the default relation */ 2091 if (*l == 'x' && ((isascii((unsigned char)l[1]) && 2092 isspace((unsigned char)l[1])) || !l[1])) { 2093 m->reln = *l; 2094 ++l; 2095 } 2096 break; 2097 } 2098 /* 2099 * Grab the value part, except for an 'x' reln. 2100 */ 2101 if (m->reln != 'x' && getvalue(ms, m, &l, action)) 2102 return -1; 2103 2104 /* 2105 * TODO finish this macro and start using it! 2106 * #define offsetcheck {if (offset > HOWMANY-1) 2107 * magwarn("offset too big"); } 2108 */ 2109 2110 /* 2111 * Now get last part - the description 2112 */ 2113 EATAB; 2114 if (l[0] == '\b') { 2115 ++l; 2116 m->flag |= NOSPACE; 2117 } else if ((l[0] == '\\') && (l[1] == 'b')) { 2118 ++l; 2119 ++l; 2120 m->flag |= NOSPACE; 2121 } 2122 for (i = 0; (m->desc[i++] = *l++) != '\0' && i < sizeof(m->desc); ) 2123 continue; 2124 if (i == sizeof(m->desc)) { 2125 m->desc[sizeof(m->desc) - 1] = '\0'; 2126 if (ms->flags & MAGIC_CHECK) 2127 file_magwarn(ms, "description `%s' truncated", m->desc); 2128 } 2129 2130 /* 2131 * We only do this check while compiling, or if any of the magic 2132 * files were not compiled. 2133 */ 2134 if (ms->flags & MAGIC_CHECK) { 2135 if (check_format(ms, m) == -1) 2136 return -1; 2137 } 2138 #ifndef COMPILE_ONLY 2139 if (action == FILE_CHECK) { 2140 file_mdump(m); 2141 } 2142 #endif 2143 m->mimetype[0] = '\0'; /* initialise MIME type to none */ 2144 return 0; 2145 } 2146 2147 /* 2148 * parse a STRENGTH annotation line from magic file, put into magic[index - 1] 2149 * if valid 2150 */ 2151 private int 2152 parse_strength(struct magic_set *ms, struct magic_entry *me, const char *line) 2153 { 2154 const char *l = line; 2155 char *el; 2156 unsigned long factor; 2157 struct magic *m = &me->mp[0]; 2158 2159 if (m->factor_op != FILE_FACTOR_OP_NONE) { 2160 file_magwarn(ms, 2161 "Current entry already has a strength type: %c %d", 2162 m->factor_op, m->factor); 2163 return -1; 2164 } 2165 if (m->type == FILE_NAME) { 2166 file_magwarn(ms, "%s: Strength setting is not supported in " 2167 "\"name\" magic entries", m->value.s); 2168 return -1; 2169 } 2170 EATAB; 2171 switch (*l) { 2172 case FILE_FACTOR_OP_NONE: 2173 case FILE_FACTOR_OP_PLUS: 2174 case FILE_FACTOR_OP_MINUS: 2175 case FILE_FACTOR_OP_TIMES: 2176 case FILE_FACTOR_OP_DIV: 2177 m->factor_op = *l++; 2178 break; 2179 default: 2180 file_magwarn(ms, "Unknown factor op `%c'", *l); 2181 return -1; 2182 } 2183 EATAB; 2184 factor = strtoul(l, &el, 0); 2185 if (factor > 255) { 2186 file_magwarn(ms, "Too large factor `%lu'", factor); 2187 goto out; 2188 } 2189 if (*el && !isspace((unsigned char)*el)) { 2190 file_magwarn(ms, "Bad factor `%s'", l); 2191 goto out; 2192 } 2193 m->factor = (uint8_t)factor; 2194 if (m->factor == 0 && m->factor_op == FILE_FACTOR_OP_DIV) { 2195 file_magwarn(ms, "Cannot have factor op `%c' and factor %u", 2196 m->factor_op, m->factor); 2197 goto out; 2198 } 2199 return 0; 2200 out: 2201 m->factor_op = FILE_FACTOR_OP_NONE; 2202 m->factor = 0; 2203 return -1; 2204 } 2205 2206 private int 2207 goodchar(unsigned char x, const char *extra) 2208 { 2209 return (isascii(x) && isalnum(x)) || strchr(extra, x); 2210 } 2211 2212 private int 2213 parse_extra(struct magic_set *ms, struct magic_entry *me, const char *line, 2214 off_t off, size_t len, const char *name, const char *extra, int nt) 2215 { 2216 size_t i; 2217 const char *l = line; 2218 struct magic *m = &me->mp[me->cont_count == 0 ? 0 : me->cont_count - 1]; 2219 char *buf = CAST(char *, CAST(void *, m)) + off; 2220 2221 if (buf[0] != '\0') { 2222 len = nt ? strlen(buf) : len; 2223 file_magwarn(ms, "Current entry already has a %s type " 2224 "`%.*s', new type `%s'", name, (int)len, buf, l); 2225 return -1; 2226 } 2227 2228 if (*m->desc == '\0') { 2229 file_magwarn(ms, "Current entry does not yet have a " 2230 "description for adding a %s type", name); 2231 return -1; 2232 } 2233 2234 EATAB; 2235 for (i = 0; *l && i < len && goodchar(*l, extra); buf[i++] = *l++) 2236 continue; 2237 2238 if (i == len && *l) { 2239 if (nt) 2240 buf[len - 1] = '\0'; 2241 if (ms->flags & MAGIC_CHECK) 2242 file_magwarn(ms, "%s type `%s' truncated %" 2243 SIZE_T_FORMAT "u", name, line, i); 2244 } else { 2245 if (!isspace((unsigned char)*l) && !goodchar(*l, extra)) 2246 file_magwarn(ms, "%s type `%s' has bad char '%c'", 2247 name, line, *l); 2248 if (nt) 2249 buf[i] = '\0'; 2250 } 2251 2252 if (i > 0) 2253 return 0; 2254 2255 file_magerror(ms, "Bad magic entry '%s'", line); 2256 return -1; 2257 } 2258 2259 /* 2260 * Parse an Apple CREATOR/TYPE annotation from magic file and put it into 2261 * magic[index - 1] 2262 */ 2263 private int 2264 parse_apple(struct magic_set *ms, struct magic_entry *me, const char *line) 2265 { 2266 struct magic *m = &me->mp[0]; 2267 2268 return parse_extra(ms, me, line, 2269 CAST(off_t, offsetof(struct magic, apple)), 2270 sizeof(m->apple), "APPLE", "!+-./", 0); 2271 } 2272 2273 /* 2274 * Parse a comma-separated list of extensions 2275 */ 2276 private int 2277 parse_ext(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, ext)), 2283 sizeof(m->ext), "EXTENSION", ",!+-/", 0); 2284 } 2285 2286 /* 2287 * parse a MIME annotation line from magic file, put into magic[index - 1] 2288 * if valid 2289 */ 2290 private int 2291 parse_mime(struct magic_set *ms, struct magic_entry *me, const char *line) 2292 { 2293 struct magic *m = &me->mp[0]; 2294 2295 return parse_extra(ms, me, line, 2296 CAST(off_t, offsetof(struct magic, mimetype)), 2297 sizeof(m->mimetype), "MIME", "+-/.", 1); 2298 } 2299 2300 private int 2301 check_format_type(const char *ptr, int type) 2302 { 2303 int quad = 0, h; 2304 if (*ptr == '\0') { 2305 /* Missing format string; bad */ 2306 return -1; 2307 } 2308 2309 switch (file_formats[type]) { 2310 case FILE_FMT_QUAD: 2311 quad = 1; 2312 /*FALLTHROUGH*/ 2313 case FILE_FMT_NUM: 2314 if (quad == 0) { 2315 switch (type) { 2316 case FILE_BYTE: 2317 h = 2; 2318 break; 2319 case FILE_SHORT: 2320 case FILE_BESHORT: 2321 case FILE_LESHORT: 2322 h = 1; 2323 break; 2324 case FILE_LONG: 2325 case FILE_BELONG: 2326 case FILE_LELONG: 2327 case FILE_MELONG: 2328 case FILE_LEID3: 2329 case FILE_BEID3: 2330 case FILE_INDIRECT: 2331 h = 0; 2332 break; 2333 default: 2334 abort(); 2335 } 2336 } else 2337 h = 0; 2338 if (*ptr == '-') 2339 ptr++; 2340 if (*ptr == '.') 2341 ptr++; 2342 while (isdigit((unsigned char)*ptr)) ptr++; 2343 if (*ptr == '.') 2344 ptr++; 2345 while (isdigit((unsigned char)*ptr)) ptr++; 2346 if (quad) { 2347 if (*ptr++ != 'l') 2348 return -1; 2349 if (*ptr++ != 'l') 2350 return -1; 2351 } 2352 2353 switch (*ptr++) { 2354 #ifdef STRICT_FORMAT /* "long" formats are int formats for us */ 2355 /* so don't accept the 'l' modifier */ 2356 case 'l': 2357 switch (*ptr++) { 2358 case 'i': 2359 case 'd': 2360 case 'u': 2361 case 'o': 2362 case 'x': 2363 case 'X': 2364 return h != 0 ? -1 : 0; 2365 default: 2366 return -1; 2367 } 2368 2369 /* 2370 * Don't accept h and hh modifiers. They make writing 2371 * magic entries more complicated, for very little benefit 2372 */ 2373 case 'h': 2374 if (h-- <= 0) 2375 return -1; 2376 switch (*ptr++) { 2377 case 'h': 2378 if (h-- <= 0) 2379 return -1; 2380 switch (*ptr++) { 2381 case 'i': 2382 case 'd': 2383 case 'u': 2384 case 'o': 2385 case 'x': 2386 case 'X': 2387 return 0; 2388 default: 2389 return -1; 2390 } 2391 case 'i': 2392 case 'd': 2393 case 'u': 2394 case 'o': 2395 case 'x': 2396 case 'X': 2397 return h != 0 ? -1 : 0; 2398 default: 2399 return -1; 2400 } 2401 #endif 2402 case 'c': 2403 return h != 2 ? -1 : 0; 2404 case 'i': 2405 case 'd': 2406 case 'u': 2407 case 'o': 2408 case 'x': 2409 case 'X': 2410 #ifdef STRICT_FORMAT 2411 return h != 0 ? -1 : 0; 2412 #else 2413 return 0; 2414 #endif 2415 default: 2416 return -1; 2417 } 2418 2419 case FILE_FMT_FLOAT: 2420 case FILE_FMT_DOUBLE: 2421 if (*ptr == '-') 2422 ptr++; 2423 if (*ptr == '.') 2424 ptr++; 2425 while (isdigit((unsigned char)*ptr)) ptr++; 2426 if (*ptr == '.') 2427 ptr++; 2428 while (isdigit((unsigned char)*ptr)) ptr++; 2429 2430 switch (*ptr++) { 2431 case 'e': 2432 case 'E': 2433 case 'f': 2434 case 'F': 2435 case 'g': 2436 case 'G': 2437 return 0; 2438 2439 default: 2440 return -1; 2441 } 2442 2443 2444 case FILE_FMT_STR: 2445 if (*ptr == '-') 2446 ptr++; 2447 while (isdigit((unsigned char )*ptr)) 2448 ptr++; 2449 if (*ptr == '.') { 2450 ptr++; 2451 while (isdigit((unsigned char )*ptr)) 2452 ptr++; 2453 } 2454 2455 switch (*ptr++) { 2456 case 's': 2457 return 0; 2458 default: 2459 return -1; 2460 } 2461 2462 default: 2463 /* internal error */ 2464 abort(); 2465 } 2466 /*NOTREACHED*/ 2467 return -1; 2468 } 2469 2470 /* 2471 * Check that the optional printf format in description matches 2472 * the type of the magic. 2473 */ 2474 private int 2475 check_format(struct magic_set *ms, struct magic *m) 2476 { 2477 char *ptr; 2478 2479 for (ptr = m->desc; *ptr; ptr++) 2480 if (*ptr == '%') 2481 break; 2482 if (*ptr == '\0') { 2483 /* No format string; ok */ 2484 return 1; 2485 } 2486 2487 assert(file_nformats == file_nnames); 2488 2489 if (m->type >= file_nformats) { 2490 file_magwarn(ms, "Internal error inconsistency between " 2491 "m->type and format strings"); 2492 return -1; 2493 } 2494 if (file_formats[m->type] == FILE_FMT_NONE) { 2495 file_magwarn(ms, "No format string for `%s' with description " 2496 "`%s'", m->desc, file_names[m->type]); 2497 return -1; 2498 } 2499 2500 ptr++; 2501 if (check_format_type(ptr, m->type) == -1) { 2502 /* 2503 * TODO: this error message is unhelpful if the format 2504 * string is not one character long 2505 */ 2506 file_magwarn(ms, "Printf format `%c' is not valid for type " 2507 "`%s' in description `%s'", *ptr ? *ptr : '?', 2508 file_names[m->type], m->desc); 2509 return -1; 2510 } 2511 2512 for (; *ptr; ptr++) { 2513 if (*ptr == '%') { 2514 file_magwarn(ms, 2515 "Too many format strings (should have at most one) " 2516 "for `%s' with description `%s'", 2517 file_names[m->type], m->desc); 2518 return -1; 2519 } 2520 } 2521 return 0; 2522 } 2523 2524 /* 2525 * Read a numeric value from a pointer, into the value union of a magic 2526 * pointer, according to the magic type. Update the string pointer to point 2527 * just after the number read. Return 0 for success, non-zero for failure. 2528 */ 2529 private int 2530 getvalue(struct magic_set *ms, struct magic *m, const char **p, int action) 2531 { 2532 switch (m->type) { 2533 case FILE_BESTRING16: 2534 case FILE_LESTRING16: 2535 case FILE_STRING: 2536 case FILE_PSTRING: 2537 case FILE_REGEX: 2538 case FILE_SEARCH: 2539 case FILE_NAME: 2540 case FILE_USE: 2541 *p = getstr(ms, m, *p, action == FILE_COMPILE); 2542 if (*p == NULL) { 2543 if (ms->flags & MAGIC_CHECK) 2544 file_magwarn(ms, "cannot get string from `%s'", 2545 m->value.s); 2546 return -1; 2547 } 2548 if (m->type == FILE_REGEX) { 2549 file_regex_t rx; 2550 int rc = file_regcomp(&rx, m->value.s, REG_EXTENDED); 2551 if (rc) { 2552 if (ms->flags & MAGIC_CHECK) 2553 file_regerror(&rx, rc, ms); 2554 } 2555 file_regfree(&rx); 2556 return rc ? -1 : 0; 2557 } 2558 return 0; 2559 case FILE_FLOAT: 2560 case FILE_BEFLOAT: 2561 case FILE_LEFLOAT: 2562 if (m->reln != 'x') { 2563 char *ep; 2564 errno = 0; 2565 #ifdef HAVE_STRTOF 2566 m->value.f = strtof(*p, &ep); 2567 #else 2568 m->value.f = (float)strtod(*p, &ep); 2569 #endif 2570 if (errno == 0) 2571 *p = ep; 2572 } 2573 return 0; 2574 case FILE_DOUBLE: 2575 case FILE_BEDOUBLE: 2576 case FILE_LEDOUBLE: 2577 if (m->reln != 'x') { 2578 char *ep; 2579 errno = 0; 2580 m->value.d = strtod(*p, &ep); 2581 if (errno == 0) 2582 *p = ep; 2583 } 2584 return 0; 2585 default: 2586 if (m->reln != 'x') { 2587 char *ep; 2588 errno = 0; 2589 m->value.q = file_signextend(ms, m, 2590 (uint64_t)strtoull(*p, &ep, 0)); 2591 if (errno == 0) { 2592 *p = ep; 2593 eatsize(p); 2594 } 2595 } 2596 return 0; 2597 } 2598 } 2599 2600 /* 2601 * Convert a string containing C character escapes. Stop at an unescaped 2602 * space or tab. 2603 * Copy the converted version to "m->value.s", and the length in m->vallen. 2604 * Return updated scan pointer as function result. Warn if set. 2605 */ 2606 private const char * 2607 getstr(struct magic_set *ms, struct magic *m, const char *s, int warn) 2608 { 2609 const char *origs = s; 2610 char *p = m->value.s; 2611 size_t plen = sizeof(m->value.s); 2612 char *origp = p; 2613 char *pmax = p + plen - 1; 2614 int c; 2615 int val; 2616 2617 while ((c = *s++) != '\0') { 2618 if (isspace((unsigned char) c)) 2619 break; 2620 if (p >= pmax) { 2621 file_error(ms, 0, "string too long: `%s'", origs); 2622 return NULL; 2623 } 2624 if (c == '\\') { 2625 switch(c = *s++) { 2626 2627 case '\0': 2628 if (warn) 2629 file_magwarn(ms, "incomplete escape"); 2630 s--; 2631 goto out; 2632 2633 case '\t': 2634 if (warn) { 2635 file_magwarn(ms, 2636 "escaped tab found, use \\t instead"); 2637 warn = 0; /* already did */ 2638 } 2639 /*FALLTHROUGH*/ 2640 default: 2641 if (warn) { 2642 if (isprint((unsigned char)c)) { 2643 /* Allow escaping of 2644 * ``relations'' */ 2645 if (strchr("<>&^=!", c) == NULL 2646 && (m->type != FILE_REGEX || 2647 strchr("[]().*?^$|{}", c) 2648 == NULL)) { 2649 file_magwarn(ms, "no " 2650 "need to escape " 2651 "`%c'", c); 2652 } 2653 } else { 2654 file_magwarn(ms, 2655 "unknown escape sequence: " 2656 "\\%03o", c); 2657 } 2658 } 2659 /*FALLTHROUGH*/ 2660 /* space, perhaps force people to use \040? */ 2661 case ' ': 2662 #if 0 2663 /* 2664 * Other things people escape, but shouldn't need to, 2665 * so we disallow them 2666 */ 2667 case '\'': 2668 case '"': 2669 case '?': 2670 #endif 2671 /* Relations */ 2672 case '>': 2673 case '<': 2674 case '&': 2675 case '^': 2676 case '=': 2677 case '!': 2678 /* and baskslash itself */ 2679 case '\\': 2680 *p++ = (char) c; 2681 break; 2682 2683 case 'a': 2684 *p++ = '\a'; 2685 break; 2686 2687 case 'b': 2688 *p++ = '\b'; 2689 break; 2690 2691 case 'f': 2692 *p++ = '\f'; 2693 break; 2694 2695 case 'n': 2696 *p++ = '\n'; 2697 break; 2698 2699 case 'r': 2700 *p++ = '\r'; 2701 break; 2702 2703 case 't': 2704 *p++ = '\t'; 2705 break; 2706 2707 case 'v': 2708 *p++ = '\v'; 2709 break; 2710 2711 /* \ and up to 3 octal digits */ 2712 case '0': 2713 case '1': 2714 case '2': 2715 case '3': 2716 case '4': 2717 case '5': 2718 case '6': 2719 case '7': 2720 val = c - '0'; 2721 c = *s++; /* try for 2 */ 2722 if (c >= '0' && c <= '7') { 2723 val = (val << 3) | (c - '0'); 2724 c = *s++; /* try for 3 */ 2725 if (c >= '0' && c <= '7') 2726 val = (val << 3) | (c-'0'); 2727 else 2728 --s; 2729 } 2730 else 2731 --s; 2732 *p++ = (char)val; 2733 break; 2734 2735 /* \x and up to 2 hex digits */ 2736 case 'x': 2737 val = 'x'; /* Default if no digits */ 2738 c = hextoint(*s++); /* Get next char */ 2739 if (c >= 0) { 2740 val = c; 2741 c = hextoint(*s++); 2742 if (c >= 0) 2743 val = (val << 4) + c; 2744 else 2745 --s; 2746 } else 2747 --s; 2748 *p++ = (char)val; 2749 break; 2750 } 2751 } else 2752 *p++ = (char)c; 2753 } 2754 --s; 2755 out: 2756 *p = '\0'; 2757 m->vallen = CAST(unsigned char, (p - origp)); 2758 if (m->type == FILE_PSTRING) 2759 m->vallen += (unsigned char)file_pstring_length_size(m); 2760 return s; 2761 } 2762 2763 2764 /* Single hex char to int; -1 if not a hex char. */ 2765 private int 2766 hextoint(int c) 2767 { 2768 if (!isascii((unsigned char) c)) 2769 return -1; 2770 if (isdigit((unsigned char) c)) 2771 return c - '0'; 2772 if ((c >= 'a') && (c <= 'f')) 2773 return c + 10 - 'a'; 2774 if (( c>= 'A') && (c <= 'F')) 2775 return c + 10 - 'A'; 2776 return -1; 2777 } 2778 2779 2780 /* 2781 * Print a string containing C character escapes. 2782 */ 2783 protected void 2784 file_showstr(FILE *fp, const char *s, size_t len) 2785 { 2786 char c; 2787 2788 for (;;) { 2789 if (len == ~0U) { 2790 c = *s++; 2791 if (c == '\0') 2792 break; 2793 } 2794 else { 2795 if (len-- == 0) 2796 break; 2797 c = *s++; 2798 } 2799 if (c >= 040 && c <= 0176) /* TODO isprint && !iscntrl */ 2800 (void) fputc(c, fp); 2801 else { 2802 (void) fputc('\\', fp); 2803 switch (c) { 2804 case '\a': 2805 (void) fputc('a', fp); 2806 break; 2807 2808 case '\b': 2809 (void) fputc('b', fp); 2810 break; 2811 2812 case '\f': 2813 (void) fputc('f', fp); 2814 break; 2815 2816 case '\n': 2817 (void) fputc('n', fp); 2818 break; 2819 2820 case '\r': 2821 (void) fputc('r', fp); 2822 break; 2823 2824 case '\t': 2825 (void) fputc('t', fp); 2826 break; 2827 2828 case '\v': 2829 (void) fputc('v', fp); 2830 break; 2831 2832 default: 2833 (void) fprintf(fp, "%.3o", c & 0377); 2834 break; 2835 } 2836 } 2837 } 2838 } 2839 2840 /* 2841 * eatsize(): Eat the size spec from a number [eg. 10UL] 2842 */ 2843 private void 2844 eatsize(const char **p) 2845 { 2846 const char *l = *p; 2847 2848 if (LOWCASE(*l) == 'u') 2849 l++; 2850 2851 switch (LOWCASE(*l)) { 2852 case 'l': /* long */ 2853 case 's': /* short */ 2854 case 'h': /* short */ 2855 case 'b': /* char/byte */ 2856 case 'c': /* char/byte */ 2857 l++; 2858 /*FALLTHROUGH*/ 2859 default: 2860 break; 2861 } 2862 2863 *p = l; 2864 } 2865 2866 /* 2867 * handle a buffer containing a compiled file. 2868 */ 2869 private struct magic_map * 2870 apprentice_buf(struct magic_set *ms, struct magic *buf, size_t len) 2871 { 2872 struct magic_map *map; 2873 2874 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 2875 file_oomem(ms, sizeof(*map)); 2876 return NULL; 2877 } 2878 map->len = len; 2879 map->p = buf; 2880 map->type = MAP_TYPE_USER; 2881 if (check_buffer(ms, map, "buffer") != 0) { 2882 apprentice_unmap(map); 2883 return NULL; 2884 } 2885 return map; 2886 } 2887 2888 /* 2889 * handle a compiled file. 2890 */ 2891 2892 private struct magic_map * 2893 apprentice_map(struct magic_set *ms, const char *fn) 2894 { 2895 int fd; 2896 struct stat st; 2897 char *dbname = NULL; 2898 struct magic_map *map; 2899 2900 fd = -1; 2901 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 2902 file_oomem(ms, sizeof(*map)); 2903 goto error; 2904 } 2905 2906 dbname = mkdbname(ms, fn, 0); 2907 if (dbname == NULL) 2908 goto error; 2909 2910 if ((fd = open(dbname, O_RDONLY|O_BINARY)) == -1) 2911 goto error; 2912 2913 if (fstat(fd, &st) == -1) { 2914 file_error(ms, errno, "cannot stat `%s'", dbname); 2915 goto error; 2916 } 2917 if (st.st_size < 8 || st.st_size > MAXMAGIC_SIZE) { 2918 file_error(ms, 0, "file `%s' is too %s", dbname, 2919 st.st_size < 8 ? "small" : "large"); 2920 goto error; 2921 } 2922 2923 map->len = (size_t)st.st_size; 2924 #ifdef QUICK 2925 if ((map->p = mmap(0, (size_t)st.st_size, PROT_READ|PROT_WRITE, 2926 MAP_PRIVATE|MAP_FILE, fd, (off_t)0)) == MAP_FAILED) { 2927 file_error(ms, errno, "cannot map `%s'", dbname); 2928 goto error; 2929 } 2930 map->type = MAP_TYPE_MMAP; 2931 #else 2932 if ((map->p = CAST(void *, malloc(map->len))) == NULL) { 2933 file_oomem(ms, map->len); 2934 goto error; 2935 } 2936 if (read(fd, map->p, map->len) != (ssize_t)map->len) { 2937 file_badread(ms); 2938 goto error; 2939 } 2940 map->type = MAP_TYPE_MALLOC; 2941 #define RET 1 2942 #endif 2943 (void)close(fd); 2944 fd = -1; 2945 2946 if (check_buffer(ms, map, dbname) != 0) 2947 goto error; 2948 2949 free(dbname); 2950 return map; 2951 2952 error: 2953 if (fd != -1) 2954 (void)close(fd); 2955 apprentice_unmap(map); 2956 free(dbname); 2957 return NULL; 2958 } 2959 2960 private int 2961 check_buffer(struct magic_set *ms, struct magic_map *map, const char *dbname) 2962 { 2963 uint32_t *ptr; 2964 uint32_t entries, nentries; 2965 uint32_t version; 2966 int i, needsbyteswap; 2967 2968 ptr = CAST(uint32_t *, map->p); 2969 if (*ptr != MAGICNO) { 2970 if (swap4(*ptr) != MAGICNO) { 2971 file_error(ms, 0, "bad magic in `%s'", dbname); 2972 return -1; 2973 } 2974 needsbyteswap = 1; 2975 } else 2976 needsbyteswap = 0; 2977 if (needsbyteswap) 2978 version = swap4(ptr[1]); 2979 else 2980 version = ptr[1]; 2981 if (version != VERSIONNO) { 2982 file_error(ms, 0, "File %s supports only version %d magic " 2983 "files. `%s' is version %d", VERSION, 2984 VERSIONNO, dbname, version); 2985 return -1; 2986 } 2987 entries = (uint32_t)(map->len / sizeof(struct magic)); 2988 if ((entries * sizeof(struct magic)) != map->len) { 2989 file_error(ms, 0, "Size of `%s' %" SIZE_T_FORMAT "u is not " 2990 "a multiple of %" SIZE_T_FORMAT "u", 2991 dbname, map->len, sizeof(struct magic)); 2992 return -1; 2993 } 2994 map->magic[0] = CAST(struct magic *, map->p) + 1; 2995 nentries = 0; 2996 for (i = 0; i < MAGIC_SETS; i++) { 2997 if (needsbyteswap) 2998 map->nmagic[i] = swap4(ptr[i + 2]); 2999 else 3000 map->nmagic[i] = ptr[i + 2]; 3001 if (i != MAGIC_SETS - 1) 3002 map->magic[i + 1] = map->magic[i] + map->nmagic[i]; 3003 nentries += map->nmagic[i]; 3004 } 3005 if (entries != nentries + 1) { 3006 file_error(ms, 0, "Inconsistent entries in `%s' %u != %u", 3007 dbname, entries, nentries + 1); 3008 return -1; 3009 } 3010 if (needsbyteswap) 3011 for (i = 0; i < MAGIC_SETS; i++) 3012 byteswap(map->magic[i], map->nmagic[i]); 3013 return 0; 3014 } 3015 3016 /* 3017 * handle an mmaped file. 3018 */ 3019 private int 3020 apprentice_compile(struct magic_set *ms, struct magic_map *map, const char *fn) 3021 { 3022 static const size_t nm = sizeof(*map->nmagic) * MAGIC_SETS; 3023 static const size_t m = sizeof(**map->magic); 3024 int fd = -1; 3025 size_t len; 3026 char *dbname; 3027 int rv = -1; 3028 uint32_t i; 3029 union { 3030 struct magic m; 3031 uint32_t h[2 + MAGIC_SETS]; 3032 } hdr; 3033 3034 dbname = mkdbname(ms, fn, 1); 3035 3036 if (dbname == NULL) 3037 goto out; 3038 3039 if ((fd = open(dbname, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY, 0644)) == -1) 3040 { 3041 file_error(ms, errno, "cannot open `%s'", dbname); 3042 goto out; 3043 } 3044 memset(&hdr, 0, sizeof(hdr)); 3045 hdr.h[0] = MAGICNO; 3046 hdr.h[1] = VERSIONNO; 3047 memcpy(hdr.h + 2, map->nmagic, nm); 3048 3049 if (write(fd, &hdr, sizeof(hdr)) != (ssize_t)sizeof(hdr)) { 3050 file_error(ms, errno, "error writing `%s'", dbname); 3051 goto out; 3052 } 3053 3054 for (i = 0; i < MAGIC_SETS; i++) { 3055 len = m * map->nmagic[i]; 3056 if (write(fd, map->magic[i], len) != (ssize_t)len) { 3057 file_error(ms, errno, "error writing `%s'", dbname); 3058 goto out; 3059 } 3060 } 3061 3062 if (fd != -1) 3063 (void)close(fd); 3064 rv = 0; 3065 out: 3066 free(dbname); 3067 return rv; 3068 } 3069 3070 private const char ext[] = ".mgc"; 3071 /* 3072 * make a dbname 3073 */ 3074 private char * 3075 mkdbname(struct magic_set *ms, const char *fn, int strip) 3076 { 3077 const char *p, *q; 3078 char *buf; 3079 3080 if (strip) { 3081 if ((p = strrchr(fn, '/')) != NULL) 3082 fn = ++p; 3083 } 3084 3085 for (q = fn; *q; q++) 3086 continue; 3087 /* Look for .mgc */ 3088 for (p = ext + sizeof(ext) - 1; p >= ext && q >= fn; p--, q--) 3089 if (*p != *q) 3090 break; 3091 3092 /* Did not find .mgc, restore q */ 3093 if (p >= ext) 3094 while (*q) 3095 q++; 3096 3097 q++; 3098 /* Compatibility with old code that looked in .mime */ 3099 if (ms->flags & MAGIC_MIME) { 3100 if (asprintf(&buf, "%.*s.mime%s", (int)(q - fn), fn, ext) < 0) 3101 return NULL; 3102 if (access(buf, R_OK) != -1) { 3103 ms->flags &= MAGIC_MIME_TYPE; 3104 return buf; 3105 } 3106 free(buf); 3107 } 3108 if (asprintf(&buf, "%.*s%s", (int)(q - fn), fn, ext) < 0) 3109 return NULL; 3110 3111 /* Compatibility with old code that looked in .mime */ 3112 if (strstr(p, ".mime") != NULL) 3113 ms->flags &= MAGIC_MIME_TYPE; 3114 return buf; 3115 } 3116 3117 /* 3118 * Byteswap an mmap'ed file if needed 3119 */ 3120 private void 3121 byteswap(struct magic *magic, uint32_t nmagic) 3122 { 3123 uint32_t i; 3124 for (i = 0; i < nmagic; i++) 3125 bs1(&magic[i]); 3126 } 3127 3128 /* 3129 * swap a short 3130 */ 3131 private uint16_t 3132 swap2(uint16_t sv) 3133 { 3134 uint16_t rv; 3135 uint8_t *s = (uint8_t *)(void *)&sv; 3136 uint8_t *d = (uint8_t *)(void *)&rv; 3137 d[0] = s[1]; 3138 d[1] = s[0]; 3139 return rv; 3140 } 3141 3142 /* 3143 * swap an int 3144 */ 3145 private uint32_t 3146 swap4(uint32_t sv) 3147 { 3148 uint32_t rv; 3149 uint8_t *s = (uint8_t *)(void *)&sv; 3150 uint8_t *d = (uint8_t *)(void *)&rv; 3151 d[0] = s[3]; 3152 d[1] = s[2]; 3153 d[2] = s[1]; 3154 d[3] = s[0]; 3155 return rv; 3156 } 3157 3158 /* 3159 * swap a quad 3160 */ 3161 private uint64_t 3162 swap8(uint64_t sv) 3163 { 3164 uint64_t rv; 3165 uint8_t *s = (uint8_t *)(void *)&sv; 3166 uint8_t *d = (uint8_t *)(void *)&rv; 3167 #if 0 3168 d[0] = s[3]; 3169 d[1] = s[2]; 3170 d[2] = s[1]; 3171 d[3] = s[0]; 3172 d[4] = s[7]; 3173 d[5] = s[6]; 3174 d[6] = s[5]; 3175 d[7] = s[4]; 3176 #else 3177 d[0] = s[7]; 3178 d[1] = s[6]; 3179 d[2] = s[5]; 3180 d[3] = s[4]; 3181 d[4] = s[3]; 3182 d[5] = s[2]; 3183 d[6] = s[1]; 3184 d[7] = s[0]; 3185 #endif 3186 return rv; 3187 } 3188 3189 /* 3190 * byteswap a single magic entry 3191 */ 3192 private void 3193 bs1(struct magic *m) 3194 { 3195 m->cont_level = swap2(m->cont_level); 3196 m->offset = swap4((uint32_t)m->offset); 3197 m->in_offset = swap4((uint32_t)m->in_offset); 3198 m->lineno = swap4((uint32_t)m->lineno); 3199 if (IS_STRING(m->type)) { 3200 m->str_range = swap4(m->str_range); 3201 m->str_flags = swap4(m->str_flags); 3202 } 3203 else { 3204 m->value.q = swap8(m->value.q); 3205 m->num_mask = swap8(m->num_mask); 3206 } 3207 } 3208 3209 protected size_t 3210 file_pstring_length_size(const struct magic *m) 3211 { 3212 switch (m->str_flags & PSTRING_LEN) { 3213 case PSTRING_1_LE: 3214 return 1; 3215 case PSTRING_2_LE: 3216 case PSTRING_2_BE: 3217 return 2; 3218 case PSTRING_4_LE: 3219 case PSTRING_4_BE: 3220 return 4; 3221 default: 3222 abort(); /* Impossible */ 3223 return 1; 3224 } 3225 } 3226 protected size_t 3227 file_pstring_get_length(const struct magic *m, const char *ss) 3228 { 3229 size_t len = 0; 3230 const unsigned char *s = (const unsigned char *)ss; 3231 3232 switch (m->str_flags & PSTRING_LEN) { 3233 case PSTRING_1_LE: 3234 len = *s; 3235 break; 3236 case PSTRING_2_LE: 3237 len = (s[1] << 8) | s[0]; 3238 break; 3239 case PSTRING_2_BE: 3240 len = (s[0] << 8) | s[1]; 3241 break; 3242 case PSTRING_4_LE: 3243 len = (s[3] << 24) | (s[2] << 16) | (s[1] << 8) | s[0]; 3244 break; 3245 case PSTRING_4_BE: 3246 len = (s[0] << 24) | (s[1] << 16) | (s[2] << 8) | s[3]; 3247 break; 3248 default: 3249 abort(); /* Impossible */ 3250 } 3251 3252 if (m->str_flags & PSTRING_LENGTH_INCLUDES_ITSELF) 3253 len -= file_pstring_length_size(m); 3254 3255 return len; 3256 } 3257 3258 protected int 3259 file_magicfind(struct magic_set *ms, const char *name, struct mlist *v) 3260 { 3261 uint32_t i, j; 3262 struct mlist *mlist, *ml; 3263 3264 mlist = ms->mlist[1]; 3265 3266 for (ml = mlist->next; ml != mlist; ml = ml->next) { 3267 struct magic *ma = ml->magic; 3268 uint32_t nma = ml->nmagic; 3269 for (i = 0; i < nma; i++) { 3270 if (ma[i].type != FILE_NAME) 3271 continue; 3272 if (strcmp(ma[i].value.s, name) == 0) { 3273 v->magic = &ma[i]; 3274 for (j = i + 1; j < nma; j++) 3275 if (ma[j].cont_level == 0) 3276 break; 3277 v->nmagic = j - i; 3278 return 0; 3279 } 3280 } 3281 } 3282 return -1; 3283 } 3284