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 1.262 2017/08/28 13:39:18 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 = NULL; 412 if ((ml = CAST(struct mlist *, malloc(sizeof(*ml)))) == NULL) 413 return -1; 414 415 ml->map = idx == 0 ? map : NULL; 416 ml->magic = map->magic[idx]; 417 ml->nmagic = map->nmagic[idx]; 418 419 mlp->prev->next = ml; 420 ml->prev = mlp->prev; 421 ml->next = mlp; 422 mlp->prev = ml; 423 return 0; 424 } 425 426 /* 427 * Handle one file or directory. 428 */ 429 private int 430 apprentice_1(struct magic_set *ms, const char *fn, int action) 431 { 432 struct magic_map *map; 433 #ifndef COMPILE_ONLY 434 struct mlist *ml; 435 size_t i; 436 #endif 437 438 if (magicsize != FILE_MAGICSIZE) { 439 file_error(ms, 0, "magic element size %lu != %lu", 440 (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 == (struct magic_map *)-1) 455 return -1; 456 if (map == NULL) { 457 if (ms->flags & MAGIC_CHECK) 458 file_magwarn(ms, "using regular magic file `%s'", fn); 459 map = apprentice_load(ms, fn, action); 460 if (map == NULL) 461 return -1; 462 } 463 464 for (i = 0; i < MAGIC_SETS; i++) { 465 if (add_mlist(ms->mlist[i], map, i) == -1) { 466 file_oomem(ms, sizeof(*ml)); 467 return -1; 468 } 469 } 470 471 if (action == FILE_LIST) { 472 for (i = 0; i < MAGIC_SETS; i++) { 473 printf("Set %" SIZE_T_FORMAT "u:\nBinary patterns:\n", 474 i); 475 apprentice_list(ms->mlist[i], BINTEST); 476 printf("Text patterns:\n"); 477 apprentice_list(ms->mlist[i], TEXTTEST); 478 } 479 } 480 return 0; 481 #else 482 return 0; 483 #endif /* COMPILE_ONLY */ 484 } 485 486 protected void 487 file_ms_free(struct magic_set *ms) 488 { 489 size_t i; 490 if (ms == NULL) 491 return; 492 for (i = 0; i < MAGIC_SETS; i++) 493 mlist_free(ms->mlist[i]); 494 free(ms->o.pbuf); 495 free(ms->o.buf); 496 free(ms->c.li); 497 free(ms); 498 } 499 500 protected struct magic_set * 501 file_ms_alloc(int flags) 502 { 503 struct magic_set *ms; 504 size_t i, len; 505 506 if ((ms = CAST(struct magic_set *, calloc((size_t)1, 507 sizeof(struct magic_set)))) == NULL) 508 return NULL; 509 510 if (magic_setflags(ms, flags) == -1) { 511 errno = EINVAL; 512 goto free; 513 } 514 515 ms->o.buf = ms->o.pbuf = NULL; 516 len = (ms->c.len = 10) * sizeof(*ms->c.li); 517 518 if ((ms->c.li = CAST(struct level_info *, malloc(len))) == NULL) 519 goto free; 520 521 ms->event_flags = 0; 522 ms->error = -1; 523 for (i = 0; i < MAGIC_SETS; i++) 524 ms->mlist[i] = NULL; 525 ms->file = "unknown"; 526 ms->line = 0; 527 ms->indir_max = FILE_INDIR_MAX; 528 ms->name_max = FILE_NAME_MAX; 529 ms->elf_shnum_max = FILE_ELF_SHNUM_MAX; 530 ms->elf_phnum_max = FILE_ELF_PHNUM_MAX; 531 ms->elf_notes_max = FILE_ELF_NOTES_MAX; 532 ms->regex_max = FILE_REGEX_MAX; 533 ms->bytes_max = FILE_BYTES_MAX; 534 return ms; 535 free: 536 free(ms); 537 return NULL; 538 } 539 540 private void 541 apprentice_unmap(struct magic_map *map) 542 { 543 size_t i; 544 if (map == NULL) 545 return; 546 547 switch (map->type) { 548 case MAP_TYPE_USER: 549 break; 550 case MAP_TYPE_MALLOC: 551 for (i = 0; i < MAGIC_SETS; i++) { 552 void *b = map->magic[i]; 553 void *p = map->p; 554 if (CAST(char *, b) >= CAST(char *, p) && 555 CAST(char *, b) <= CAST(char *, p) + map->len) 556 continue; 557 free(map->magic[i]); 558 } 559 free(map->p); 560 break; 561 #ifdef QUICK 562 case MAP_TYPE_MMAP: 563 if (map->p && map->p != MAP_FAILED) 564 (void)munmap(map->p, map->len); 565 break; 566 #endif 567 default: 568 abort(); 569 } 570 free(map); 571 } 572 573 private struct mlist * 574 mlist_alloc(void) 575 { 576 struct mlist *mlist; 577 if ((mlist = CAST(struct mlist *, calloc(1, sizeof(*mlist)))) == NULL) { 578 return NULL; 579 } 580 mlist->next = mlist->prev = mlist; 581 return mlist; 582 } 583 584 private void 585 mlist_free(struct mlist *mlist) 586 { 587 struct mlist *ml, *next; 588 589 if (mlist == NULL) 590 return; 591 592 ml = mlist->next; 593 for (ml = mlist->next; (next = ml->next) != NULL; ml = next) { 594 if (ml->map) 595 apprentice_unmap(CAST(struct magic_map *, ml->map)); 596 free(ml); 597 if (ml == mlist) 598 break; 599 } 600 } 601 602 #ifndef COMPILE_ONLY 603 /* void **bufs: an array of compiled magic files */ 604 protected int 605 buffer_apprentice(struct magic_set *ms, struct magic **bufs, 606 size_t *sizes, size_t nbufs) 607 { 608 size_t i, j; 609 struct mlist *ml; 610 struct magic_map *map; 611 612 if (nbufs == 0) 613 return -1; 614 615 (void)file_reset(ms, 0); 616 617 init_file_tables(); 618 619 for (i = 0; i < MAGIC_SETS; i++) { 620 mlist_free(ms->mlist[i]); 621 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 622 file_oomem(ms, sizeof(*ms->mlist[i])); 623 goto fail; 624 } 625 } 626 627 for (i = 0; i < nbufs; i++) { 628 map = apprentice_buf(ms, bufs[i], sizes[i]); 629 if (map == NULL) 630 goto fail; 631 632 for (j = 0; j < MAGIC_SETS; j++) { 633 if (add_mlist(ms->mlist[j], map, j) == -1) { 634 file_oomem(ms, sizeof(*ml)); 635 goto fail; 636 } 637 } 638 } 639 640 return 0; 641 fail: 642 for (i = 0; i < MAGIC_SETS; i++) { 643 mlist_free(ms->mlist[i]); 644 ms->mlist[i] = NULL; 645 } 646 return -1; 647 } 648 #endif 649 650 /* const char *fn: list of magic files and directories */ 651 protected int 652 file_apprentice(struct magic_set *ms, const char *fn, int action) 653 { 654 char *p, *mfn; 655 int file_err, errs = -1; 656 size_t i; 657 658 (void)file_reset(ms, 0); 659 660 if ((fn = magic_getpath(fn, action)) == NULL) 661 return -1; 662 663 init_file_tables(); 664 665 if ((mfn = strdup(fn)) == NULL) { 666 file_oomem(ms, strlen(fn)); 667 return -1; 668 } 669 670 for (i = 0; i < MAGIC_SETS; i++) { 671 mlist_free(ms->mlist[i]); 672 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 673 file_oomem(ms, sizeof(*ms->mlist[i])); 674 while (i-- > 0) { 675 mlist_free(ms->mlist[i]); 676 ms->mlist[i] = NULL; 677 } 678 free(mfn); 679 return -1; 680 } 681 } 682 fn = mfn; 683 684 while (fn) { 685 p = strchr(fn, PATHSEP); 686 if (p) 687 *p++ = '\0'; 688 if (*fn == '\0') 689 break; 690 file_err = apprentice_1(ms, fn, action); 691 errs = MAX(errs, file_err); 692 fn = p; 693 } 694 695 free(mfn); 696 697 if (errs == -1) { 698 for (i = 0; i < MAGIC_SETS; i++) { 699 mlist_free(ms->mlist[i]); 700 ms->mlist[i] = NULL; 701 } 702 file_error(ms, 0, "could not find any valid magic files!"); 703 return -1; 704 } 705 706 #if 0 707 /* 708 * Always leave the database loaded 709 */ 710 if (action == FILE_LOAD) 711 return 0; 712 713 for (i = 0; i < MAGIC_SETS; i++) { 714 mlist_free(ms->mlist[i]); 715 ms->mlist[i] = NULL; 716 } 717 #endif 718 719 switch (action) { 720 case FILE_LOAD: 721 case FILE_COMPILE: 722 case FILE_CHECK: 723 case FILE_LIST: 724 return 0; 725 default: 726 file_error(ms, 0, "Invalid action %d", action); 727 return -1; 728 } 729 } 730 731 /* 732 * Compute the real length of a magic expression, for the purposes 733 * of determining how "strong" a magic expression is (approximating 734 * how specific its matches are): 735 * - magic characters count 0 unless escaped. 736 * - [] expressions count 1 737 * - {} expressions count 0 738 * - regular characters or escaped magic characters count 1 739 * - 0 length expressions count as one 740 */ 741 private size_t 742 nonmagic(const char *str) 743 { 744 const char *p; 745 size_t rv = 0; 746 747 for (p = str; *p; p++) 748 switch (*p) { 749 case '\\': /* Escaped anything counts 1 */ 750 if (!*++p) 751 p--; 752 rv++; 753 continue; 754 case '?': /* Magic characters count 0 */ 755 case '*': 756 case '.': 757 case '+': 758 case '^': 759 case '$': 760 continue; 761 case '[': /* Bracketed expressions count 1 the ']' */ 762 while (*p && *p != ']') 763 p++; 764 p--; 765 continue; 766 case '{': /* Braced expressions count 0 */ 767 while (*p && *p != '}') 768 p++; 769 if (!*p) 770 p--; 771 continue; 772 default: /* Anything else counts 1 */ 773 rv++; 774 continue; 775 } 776 777 return rv == 0 ? 1 : rv; /* Return at least 1 */ 778 } 779 780 781 private size_t 782 typesize(int type) 783 { 784 switch (type) { 785 case FILE_BYTE: 786 return 1; 787 788 case FILE_SHORT: 789 case FILE_LESHORT: 790 case FILE_BESHORT: 791 return 2; 792 793 case FILE_LONG: 794 case FILE_LELONG: 795 case FILE_BELONG: 796 case FILE_MELONG: 797 return 4; 798 799 case FILE_DATE: 800 case FILE_LEDATE: 801 case FILE_BEDATE: 802 case FILE_MEDATE: 803 case FILE_LDATE: 804 case FILE_LELDATE: 805 case FILE_BELDATE: 806 case FILE_MELDATE: 807 case FILE_FLOAT: 808 case FILE_BEFLOAT: 809 case FILE_LEFLOAT: 810 return 4; 811 812 case FILE_QUAD: 813 case FILE_BEQUAD: 814 case FILE_LEQUAD: 815 case FILE_QDATE: 816 case FILE_LEQDATE: 817 case FILE_BEQDATE: 818 case FILE_QLDATE: 819 case FILE_LEQLDATE: 820 case FILE_BEQLDATE: 821 case FILE_QWDATE: 822 case FILE_LEQWDATE: 823 case FILE_BEQWDATE: 824 case FILE_DOUBLE: 825 case FILE_BEDOUBLE: 826 case FILE_LEDOUBLE: 827 return 8; 828 default: 829 return (size_t)~0; 830 } 831 } 832 833 /* 834 * Get weight of this magic entry, for sorting purposes. 835 */ 836 private size_t 837 apprentice_magic_strength(const struct magic *m) 838 { 839 #define MULT 10 840 size_t ts, v, val = 2 * MULT; /* baseline strength */ 841 842 switch (m->type) { 843 case FILE_DEFAULT: /* make sure this sorts last */ 844 if (m->factor_op != FILE_FACTOR_OP_NONE) 845 abort(); 846 return 0; 847 848 case FILE_BYTE: 849 case FILE_SHORT: 850 case FILE_LESHORT: 851 case FILE_BESHORT: 852 case FILE_LONG: 853 case FILE_LELONG: 854 case FILE_BELONG: 855 case FILE_MELONG: 856 case FILE_DATE: 857 case FILE_LEDATE: 858 case FILE_BEDATE: 859 case FILE_MEDATE: 860 case FILE_LDATE: 861 case FILE_LELDATE: 862 case FILE_BELDATE: 863 case FILE_MELDATE: 864 case FILE_FLOAT: 865 case FILE_BEFLOAT: 866 case FILE_LEFLOAT: 867 case FILE_QUAD: 868 case FILE_BEQUAD: 869 case FILE_LEQUAD: 870 case FILE_QDATE: 871 case FILE_LEQDATE: 872 case FILE_BEQDATE: 873 case FILE_QLDATE: 874 case FILE_LEQLDATE: 875 case FILE_BEQLDATE: 876 case FILE_QWDATE: 877 case FILE_LEQWDATE: 878 case FILE_BEQWDATE: 879 case FILE_DOUBLE: 880 case FILE_BEDOUBLE: 881 case FILE_LEDOUBLE: 882 ts = typesize(m->type); 883 if (ts == (size_t)~0) 884 abort(); 885 val += ts * MULT; 886 break; 887 888 case FILE_PSTRING: 889 case FILE_STRING: 890 val += m->vallen * MULT; 891 break; 892 893 case FILE_BESTRING16: 894 case FILE_LESTRING16: 895 val += m->vallen * MULT / 2; 896 break; 897 898 case FILE_SEARCH: 899 val += m->vallen * MAX(MULT / m->vallen, 1); 900 break; 901 902 case FILE_REGEX: 903 v = nonmagic(m->value.s); 904 val += v * MAX(MULT / v, 1); 905 break; 906 907 case FILE_INDIRECT: 908 case FILE_NAME: 909 case FILE_USE: 910 break; 911 912 case FILE_DER: 913 val += MULT; 914 break; 915 916 default: 917 (void)fprintf(stderr, "Bad type %d\n", m->type); 918 abort(); 919 } 920 921 switch (m->reln) { 922 case 'x': /* matches anything penalize */ 923 case '!': /* matches almost anything penalize */ 924 val = 0; 925 break; 926 927 case '=': /* Exact match, prefer */ 928 val += MULT; 929 break; 930 931 case '>': 932 case '<': /* comparison match reduce strength */ 933 val -= 2 * MULT; 934 break; 935 936 case '^': 937 case '&': /* masking bits, we could count them too */ 938 val -= MULT; 939 break; 940 941 default: 942 (void)fprintf(stderr, "Bad relation %c\n", m->reln); 943 abort(); 944 } 945 946 if (val == 0) /* ensure we only return 0 for FILE_DEFAULT */ 947 val = 1; 948 949 switch (m->factor_op) { 950 case FILE_FACTOR_OP_NONE: 951 break; 952 case FILE_FACTOR_OP_PLUS: 953 val += m->factor; 954 break; 955 case FILE_FACTOR_OP_MINUS: 956 val -= m->factor; 957 break; 958 case FILE_FACTOR_OP_TIMES: 959 val *= m->factor; 960 break; 961 case FILE_FACTOR_OP_DIV: 962 val /= m->factor; 963 break; 964 default: 965 abort(); 966 } 967 968 /* 969 * Magic entries with no description get a bonus because they depend 970 * on subsequent magic entries to print something. 971 */ 972 if (m->desc[0] == '\0') 973 val++; 974 return val; 975 } 976 977 /* 978 * Sort callback for sorting entries by "strength" (basically length) 979 */ 980 private int 981 apprentice_sort(const void *a, const void *b) 982 { 983 const struct magic_entry *ma = CAST(const struct magic_entry *, a); 984 const struct magic_entry *mb = CAST(const struct magic_entry *, b); 985 size_t sa = apprentice_magic_strength(ma->mp); 986 size_t sb = apprentice_magic_strength(mb->mp); 987 if (sa == sb) 988 return 0; 989 else if (sa > sb) 990 return -1; 991 else 992 return 1; 993 } 994 995 /* 996 * Shows sorted patterns list in the order which is used for the matching 997 */ 998 private void 999 apprentice_list(struct mlist *mlist, int mode) 1000 { 1001 uint32_t magindex = 0; 1002 struct mlist *ml; 1003 for (ml = mlist->next; ml != mlist; ml = ml->next) { 1004 for (magindex = 0; magindex < ml->nmagic; magindex++) { 1005 struct magic *m = &ml->magic[magindex]; 1006 if ((m->flag & mode) != mode) { 1007 /* Skip sub-tests */ 1008 while (magindex + 1 < ml->nmagic && 1009 ml->magic[magindex + 1].cont_level != 0) 1010 ++magindex; 1011 continue; /* Skip to next top-level test*/ 1012 } 1013 1014 /* 1015 * Try to iterate over the tree until we find item with 1016 * description/mimetype. 1017 */ 1018 while (magindex + 1 < ml->nmagic && 1019 ml->magic[magindex + 1].cont_level != 0 && 1020 *ml->magic[magindex].desc == '\0' && 1021 *ml->magic[magindex].mimetype == '\0') 1022 magindex++; 1023 1024 printf("Strength = %3" SIZE_T_FORMAT "u@%u: %s [%s]\n", 1025 apprentice_magic_strength(m), 1026 ml->magic[magindex].lineno, 1027 ml->magic[magindex].desc, 1028 ml->magic[magindex].mimetype); 1029 } 1030 } 1031 } 1032 1033 private void 1034 set_test_type(struct magic *mstart, struct magic *m) 1035 { 1036 switch (m->type) { 1037 case FILE_BYTE: 1038 case FILE_SHORT: 1039 case FILE_LONG: 1040 case FILE_DATE: 1041 case FILE_BESHORT: 1042 case FILE_BELONG: 1043 case FILE_BEDATE: 1044 case FILE_LESHORT: 1045 case FILE_LELONG: 1046 case FILE_LEDATE: 1047 case FILE_LDATE: 1048 case FILE_BELDATE: 1049 case FILE_LELDATE: 1050 case FILE_MEDATE: 1051 case FILE_MELDATE: 1052 case FILE_MELONG: 1053 case FILE_QUAD: 1054 case FILE_LEQUAD: 1055 case FILE_BEQUAD: 1056 case FILE_QDATE: 1057 case FILE_LEQDATE: 1058 case FILE_BEQDATE: 1059 case FILE_QLDATE: 1060 case FILE_LEQLDATE: 1061 case FILE_BEQLDATE: 1062 case FILE_QWDATE: 1063 case FILE_LEQWDATE: 1064 case FILE_BEQWDATE: 1065 case FILE_FLOAT: 1066 case FILE_BEFLOAT: 1067 case FILE_LEFLOAT: 1068 case FILE_DOUBLE: 1069 case FILE_BEDOUBLE: 1070 case FILE_LEDOUBLE: 1071 case FILE_DER: 1072 mstart->flag |= BINTEST; 1073 break; 1074 case FILE_STRING: 1075 case FILE_PSTRING: 1076 case FILE_BESTRING16: 1077 case FILE_LESTRING16: 1078 /* Allow text overrides */ 1079 if (mstart->str_flags & STRING_TEXTTEST) 1080 mstart->flag |= TEXTTEST; 1081 else 1082 mstart->flag |= BINTEST; 1083 break; 1084 case FILE_REGEX: 1085 case FILE_SEARCH: 1086 /* Check for override */ 1087 if (mstart->str_flags & STRING_BINTEST) 1088 mstart->flag |= BINTEST; 1089 if (mstart->str_flags & STRING_TEXTTEST) 1090 mstart->flag |= TEXTTEST; 1091 1092 if (mstart->flag & (TEXTTEST|BINTEST)) 1093 break; 1094 1095 /* binary test if pattern is not text */ 1096 if (file_looks_utf8(m->value.us, (size_t)m->vallen, NULL, 1097 NULL) <= 0) 1098 mstart->flag |= BINTEST; 1099 else 1100 mstart->flag |= TEXTTEST; 1101 break; 1102 case FILE_DEFAULT: 1103 /* can't deduce anything; we shouldn't see this at the 1104 top level anyway */ 1105 break; 1106 case FILE_INVALID: 1107 default: 1108 /* invalid search type, but no need to complain here */ 1109 break; 1110 } 1111 } 1112 1113 private int 1114 addentry(struct magic_set *ms, struct magic_entry *me, 1115 struct magic_entry_set *mset) 1116 { 1117 size_t i = me->mp->type == FILE_NAME ? 1 : 0; 1118 if (mset[i].count == mset[i].max) { 1119 struct magic_entry *mp; 1120 1121 mset[i].max += ALLOC_INCR; 1122 if ((mp = CAST(struct magic_entry *, 1123 realloc(mset[i].me, sizeof(*mp) * mset[i].max))) == 1124 NULL) { 1125 file_oomem(ms, sizeof(*mp) * mset[i].max); 1126 return -1; 1127 } 1128 (void)memset(&mp[mset[i].count], 0, sizeof(*mp) * 1129 ALLOC_INCR); 1130 mset[i].me = mp; 1131 } 1132 mset[i].me[mset[i].count++] = *me; 1133 memset(me, 0, sizeof(*me)); 1134 return 0; 1135 } 1136 1137 /* 1138 * Load and parse one file. 1139 */ 1140 private void 1141 load_1(struct magic_set *ms, int action, const char *fn, int *errs, 1142 struct magic_entry_set *mset) 1143 { 1144 size_t lineno = 0, llen = 0; 1145 char *line = NULL; 1146 ssize_t len; 1147 struct magic_entry me; 1148 1149 FILE *f = fopen(ms->file = fn, "r"); 1150 if (f == NULL) { 1151 if (errno != ENOENT) 1152 file_error(ms, errno, "cannot read magic file `%s'", 1153 fn); 1154 (*errs)++; 1155 return; 1156 } 1157 1158 memset(&me, 0, sizeof(me)); 1159 /* read and parse this file */ 1160 for (ms->line = 1; (len = getline(&line, &llen, f)) != -1; 1161 ms->line++) { 1162 if (len == 0) /* null line, garbage, etc */ 1163 continue; 1164 if (line[len - 1] == '\n') { 1165 lineno++; 1166 line[len - 1] = '\0'; /* delete newline */ 1167 } 1168 switch (line[0]) { 1169 case '\0': /* empty, do not parse */ 1170 case '#': /* comment, do not parse */ 1171 continue; 1172 case '!': 1173 if (line[1] == ':') { 1174 size_t i; 1175 1176 for (i = 0; bang[i].name != NULL; i++) { 1177 if ((size_t)(len - 2) > bang[i].len && 1178 memcmp(bang[i].name, line + 2, 1179 bang[i].len) == 0) 1180 break; 1181 } 1182 if (bang[i].name == NULL) { 1183 file_error(ms, 0, 1184 "Unknown !: entry `%s'", line); 1185 (*errs)++; 1186 continue; 1187 } 1188 if (me.mp == NULL) { 1189 file_error(ms, 0, 1190 "No current entry for :!%s type", 1191 bang[i].name); 1192 (*errs)++; 1193 continue; 1194 } 1195 if ((*bang[i].fun)(ms, &me, 1196 line + bang[i].len + 2) != 0) { 1197 (*errs)++; 1198 continue; 1199 } 1200 continue; 1201 } 1202 /*FALLTHROUGH*/ 1203 default: 1204 again: 1205 switch (parse(ms, &me, line, lineno, action)) { 1206 case 0: 1207 continue; 1208 case 1: 1209 (void)addentry(ms, &me, mset); 1210 goto again; 1211 default: 1212 (*errs)++; 1213 break; 1214 } 1215 } 1216 } 1217 if (me.mp) 1218 (void)addentry(ms, &me, mset); 1219 free(line); 1220 (void)fclose(f); 1221 } 1222 1223 /* 1224 * parse a file or directory of files 1225 * const char *fn: name of magic file or directory 1226 */ 1227 private int 1228 cmpstrp(const void *p1, const void *p2) 1229 { 1230 return strcmp(*(char *const *)p1, *(char *const *)p2); 1231 } 1232 1233 1234 private uint32_t 1235 set_text_binary(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1236 uint32_t starttest) 1237 { 1238 static const char text[] = "text"; 1239 static const char binary[] = "binary"; 1240 static const size_t len = sizeof(text); 1241 1242 uint32_t i = starttest; 1243 1244 do { 1245 set_test_type(me[starttest].mp, me[i].mp); 1246 if ((ms->flags & MAGIC_DEBUG) == 0) 1247 continue; 1248 (void)fprintf(stderr, "%s%s%s: %s\n", 1249 me[i].mp->mimetype, 1250 me[i].mp->mimetype[0] == '\0' ? "" : "; ", 1251 me[i].mp->desc[0] ? me[i].mp->desc : "(no description)", 1252 me[i].mp->flag & BINTEST ? binary : text); 1253 if (me[i].mp->flag & BINTEST) { 1254 char *p = strstr(me[i].mp->desc, text); 1255 if (p && (p == me[i].mp->desc || 1256 isspace((unsigned char)p[-1])) && 1257 (p + len - me[i].mp->desc == MAXstring 1258 || (p[len] == '\0' || 1259 isspace((unsigned char)p[len])))) 1260 (void)fprintf(stderr, "*** Possible " 1261 "binary test for text type\n"); 1262 } 1263 } while (++i < nme && me[i].mp->cont_level != 0); 1264 return i; 1265 } 1266 1267 private void 1268 set_last_default(struct magic_set *ms, struct magic_entry *me, uint32_t nme) 1269 { 1270 uint32_t i; 1271 for (i = 0; i < nme; i++) { 1272 if (me[i].mp->cont_level == 0 && 1273 me[i].mp->type == FILE_DEFAULT) { 1274 while (++i < nme) 1275 if (me[i].mp->cont_level == 0) 1276 break; 1277 if (i != nme) { 1278 /* XXX - Ugh! */ 1279 ms->line = me[i].mp->lineno; 1280 file_magwarn(ms, 1281 "level 0 \"default\" did not sort last"); 1282 } 1283 return; 1284 } 1285 } 1286 } 1287 1288 private int 1289 coalesce_entries(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1290 struct magic **ma, uint32_t *nma) 1291 { 1292 uint32_t i, mentrycount = 0; 1293 size_t slen; 1294 1295 for (i = 0; i < nme; i++) 1296 mentrycount += me[i].cont_count; 1297 1298 slen = sizeof(**ma) * mentrycount; 1299 if ((*ma = CAST(struct magic *, malloc(slen))) == NULL) { 1300 file_oomem(ms, slen); 1301 return -1; 1302 } 1303 1304 mentrycount = 0; 1305 for (i = 0; i < nme; i++) { 1306 (void)memcpy(*ma + mentrycount, me[i].mp, 1307 me[i].cont_count * sizeof(**ma)); 1308 mentrycount += me[i].cont_count; 1309 } 1310 *nma = mentrycount; 1311 return 0; 1312 } 1313 1314 private void 1315 magic_entry_free(struct magic_entry *me, uint32_t nme) 1316 { 1317 uint32_t i; 1318 if (me == NULL) 1319 return; 1320 for (i = 0; i < nme; i++) 1321 free(me[i].mp); 1322 free(me); 1323 } 1324 1325 private struct magic_map * 1326 apprentice_load(struct magic_set *ms, const char *fn, int action) 1327 { 1328 int errs = 0; 1329 uint32_t i, j; 1330 size_t files = 0, maxfiles = 0; 1331 char **filearr = NULL, *mfn; 1332 struct stat st; 1333 struct magic_map *map; 1334 struct magic_entry_set mset[MAGIC_SETS]; 1335 DIR *dir; 1336 struct dirent *d; 1337 1338 memset(mset, 0, sizeof(mset)); 1339 ms->flags |= MAGIC_CHECK; /* Enable checks for parsed files */ 1340 1341 1342 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) 1343 { 1344 file_oomem(ms, sizeof(*map)); 1345 return NULL; 1346 } 1347 map->type = MAP_TYPE_MALLOC; 1348 1349 /* print silly verbose header for USG compat. */ 1350 if (action == FILE_CHECK) 1351 (void)fprintf(stderr, "%s\n", usg_hdr); 1352 1353 /* load directory or file */ 1354 if (stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) { 1355 dir = opendir(fn); 1356 if (!dir) { 1357 errs++; 1358 goto out; 1359 } 1360 while ((d = readdir(dir)) != NULL) { 1361 if (d->d_name[0] == '.') 1362 continue; 1363 if (asprintf(&mfn, "%s/%s", fn, d->d_name) < 0) { 1364 file_oomem(ms, 1365 strlen(fn) + strlen(d->d_name) + 2); 1366 errs++; 1367 closedir(dir); 1368 goto out; 1369 } 1370 if (stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) { 1371 free(mfn); 1372 continue; 1373 } 1374 if (files >= maxfiles) { 1375 size_t mlen; 1376 maxfiles = (maxfiles + 1) * 2; 1377 mlen = maxfiles * sizeof(*filearr); 1378 if ((filearr = CAST(char **, 1379 realloc(filearr, mlen))) == NULL) { 1380 file_oomem(ms, mlen); 1381 free(mfn); 1382 closedir(dir); 1383 errs++; 1384 goto out; 1385 } 1386 } 1387 filearr[files++] = mfn; 1388 } 1389 closedir(dir); 1390 qsort(filearr, files, sizeof(*filearr), cmpstrp); 1391 for (i = 0; i < files; i++) { 1392 load_1(ms, action, filearr[i], &errs, mset); 1393 free(filearr[i]); 1394 } 1395 free(filearr); 1396 } else 1397 load_1(ms, action, fn, &errs, mset); 1398 if (errs) 1399 goto out; 1400 1401 for (j = 0; j < MAGIC_SETS; j++) { 1402 /* Set types of tests */ 1403 for (i = 0; i < mset[j].count; ) { 1404 if (mset[j].me[i].mp->cont_level != 0) { 1405 i++; 1406 continue; 1407 } 1408 i = set_text_binary(ms, mset[j].me, mset[j].count, i); 1409 } 1410 if (mset[j].me) 1411 qsort(mset[j].me, mset[j].count, sizeof(*mset[j].me), 1412 apprentice_sort); 1413 1414 /* 1415 * Make sure that any level 0 "default" line is last 1416 * (if one exists). 1417 */ 1418 set_last_default(ms, mset[j].me, mset[j].count); 1419 1420 /* coalesce per file arrays into a single one */ 1421 if (coalesce_entries(ms, mset[j].me, mset[j].count, 1422 &map->magic[j], &map->nmagic[j]) == -1) { 1423 errs++; 1424 goto out; 1425 } 1426 } 1427 1428 out: 1429 for (j = 0; j < MAGIC_SETS; j++) 1430 magic_entry_free(mset[j].me, mset[j].count); 1431 1432 if (errs) { 1433 apprentice_unmap(map); 1434 return NULL; 1435 } 1436 return map; 1437 } 1438 1439 /* 1440 * extend the sign bit if the comparison is to be signed 1441 */ 1442 protected uint64_t 1443 file_signextend(struct magic_set *ms, struct magic *m, uint64_t v) 1444 { 1445 if (!(m->flag & UNSIGNED)) { 1446 switch(m->type) { 1447 /* 1448 * Do not remove the casts below. They are 1449 * vital. When later compared with the data, 1450 * the sign extension must have happened. 1451 */ 1452 case FILE_BYTE: 1453 v = (signed char) v; 1454 break; 1455 case FILE_SHORT: 1456 case FILE_BESHORT: 1457 case FILE_LESHORT: 1458 v = (short) v; 1459 break; 1460 case FILE_DATE: 1461 case FILE_BEDATE: 1462 case FILE_LEDATE: 1463 case FILE_MEDATE: 1464 case FILE_LDATE: 1465 case FILE_BELDATE: 1466 case FILE_LELDATE: 1467 case FILE_MELDATE: 1468 case FILE_LONG: 1469 case FILE_BELONG: 1470 case FILE_LELONG: 1471 case FILE_MELONG: 1472 case FILE_FLOAT: 1473 case FILE_BEFLOAT: 1474 case FILE_LEFLOAT: 1475 v = (int32_t) v; 1476 break; 1477 case FILE_QUAD: 1478 case FILE_BEQUAD: 1479 case FILE_LEQUAD: 1480 case FILE_QDATE: 1481 case FILE_QLDATE: 1482 case FILE_QWDATE: 1483 case FILE_BEQDATE: 1484 case FILE_BEQLDATE: 1485 case FILE_BEQWDATE: 1486 case FILE_LEQDATE: 1487 case FILE_LEQLDATE: 1488 case FILE_LEQWDATE: 1489 case FILE_DOUBLE: 1490 case FILE_BEDOUBLE: 1491 case FILE_LEDOUBLE: 1492 v = (int64_t) v; 1493 break; 1494 case FILE_STRING: 1495 case FILE_PSTRING: 1496 case FILE_BESTRING16: 1497 case FILE_LESTRING16: 1498 case FILE_REGEX: 1499 case FILE_SEARCH: 1500 case FILE_DEFAULT: 1501 case FILE_INDIRECT: 1502 case FILE_NAME: 1503 case FILE_USE: 1504 case FILE_CLEAR: 1505 case FILE_DER: 1506 break; 1507 default: 1508 if (ms->flags & MAGIC_CHECK) 1509 file_magwarn(ms, "cannot happen: m->type=%d\n", 1510 m->type); 1511 return ~0U; 1512 } 1513 } 1514 return v; 1515 } 1516 1517 private int 1518 string_modifier_check(struct magic_set *ms, struct magic *m) 1519 { 1520 if ((ms->flags & MAGIC_CHECK) == 0) 1521 return 0; 1522 1523 if ((m->type != FILE_REGEX || (m->str_flags & REGEX_LINE_COUNT) == 0) && 1524 (m->type != FILE_PSTRING && (m->str_flags & PSTRING_LEN) != 0)) { 1525 file_magwarn(ms, 1526 "'/BHhLl' modifiers are only allowed for pascal strings\n"); 1527 return -1; 1528 } 1529 switch (m->type) { 1530 case FILE_BESTRING16: 1531 case FILE_LESTRING16: 1532 if (m->str_flags != 0) { 1533 file_magwarn(ms, 1534 "no modifiers allowed for 16-bit strings\n"); 1535 return -1; 1536 } 1537 break; 1538 case FILE_STRING: 1539 case FILE_PSTRING: 1540 if ((m->str_flags & REGEX_OFFSET_START) != 0) { 1541 file_magwarn(ms, 1542 "'/%c' only allowed on regex and search\n", 1543 CHAR_REGEX_OFFSET_START); 1544 return -1; 1545 } 1546 break; 1547 case FILE_SEARCH: 1548 if (m->str_range == 0) { 1549 file_magwarn(ms, 1550 "missing range; defaulting to %d\n", 1551 STRING_DEFAULT_RANGE); 1552 m->str_range = STRING_DEFAULT_RANGE; 1553 return -1; 1554 } 1555 break; 1556 case FILE_REGEX: 1557 if ((m->str_flags & STRING_COMPACT_WHITESPACE) != 0) { 1558 file_magwarn(ms, "'/%c' not allowed on regex\n", 1559 CHAR_COMPACT_WHITESPACE); 1560 return -1; 1561 } 1562 if ((m->str_flags & STRING_COMPACT_OPTIONAL_WHITESPACE) != 0) { 1563 file_magwarn(ms, "'/%c' not allowed on regex\n", 1564 CHAR_COMPACT_OPTIONAL_WHITESPACE); 1565 return -1; 1566 } 1567 break; 1568 default: 1569 file_magwarn(ms, "coding error: m->type=%d\n", 1570 m->type); 1571 return -1; 1572 } 1573 return 0; 1574 } 1575 1576 private int 1577 get_op(char c) 1578 { 1579 switch (c) { 1580 case '&': 1581 return FILE_OPAND; 1582 case '|': 1583 return FILE_OPOR; 1584 case '^': 1585 return FILE_OPXOR; 1586 case '+': 1587 return FILE_OPADD; 1588 case '-': 1589 return FILE_OPMINUS; 1590 case '*': 1591 return FILE_OPMULTIPLY; 1592 case '/': 1593 return FILE_OPDIVIDE; 1594 case '%': 1595 return FILE_OPMODULO; 1596 default: 1597 return -1; 1598 } 1599 } 1600 1601 #ifdef ENABLE_CONDITIONALS 1602 private int 1603 get_cond(const char *l, const char **t) 1604 { 1605 static const struct cond_tbl_s { 1606 char name[8]; 1607 size_t len; 1608 int cond; 1609 } cond_tbl[] = { 1610 { "if", 2, COND_IF }, 1611 { "elif", 4, COND_ELIF }, 1612 { "else", 4, COND_ELSE }, 1613 { "", 0, COND_NONE }, 1614 }; 1615 const struct cond_tbl_s *p; 1616 1617 for (p = cond_tbl; p->len; p++) { 1618 if (strncmp(l, p->name, p->len) == 0 && 1619 isspace((unsigned char)l[p->len])) { 1620 if (t) 1621 *t = l + p->len; 1622 break; 1623 } 1624 } 1625 return p->cond; 1626 } 1627 1628 private int 1629 check_cond(struct magic_set *ms, int cond, uint32_t cont_level) 1630 { 1631 int last_cond; 1632 last_cond = ms->c.li[cont_level].last_cond; 1633 1634 switch (cond) { 1635 case COND_IF: 1636 if (last_cond != COND_NONE && last_cond != COND_ELIF) { 1637 if (ms->flags & MAGIC_CHECK) 1638 file_magwarn(ms, "syntax error: `if'"); 1639 return -1; 1640 } 1641 last_cond = COND_IF; 1642 break; 1643 1644 case COND_ELIF: 1645 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1646 if (ms->flags & MAGIC_CHECK) 1647 file_magwarn(ms, "syntax error: `elif'"); 1648 return -1; 1649 } 1650 last_cond = COND_ELIF; 1651 break; 1652 1653 case COND_ELSE: 1654 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1655 if (ms->flags & MAGIC_CHECK) 1656 file_magwarn(ms, "syntax error: `else'"); 1657 return -1; 1658 } 1659 last_cond = COND_NONE; 1660 break; 1661 1662 case COND_NONE: 1663 last_cond = COND_NONE; 1664 break; 1665 } 1666 1667 ms->c.li[cont_level].last_cond = last_cond; 1668 return 0; 1669 } 1670 #endif /* ENABLE_CONDITIONALS */ 1671 1672 private int 1673 parse_indirect_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1674 { 1675 const char *l = *lp; 1676 1677 while (!isspace((unsigned char)*++l)) 1678 switch (*l) { 1679 case CHAR_INDIRECT_RELATIVE: 1680 m->str_flags |= INDIRECT_RELATIVE; 1681 break; 1682 default: 1683 if (ms->flags & MAGIC_CHECK) 1684 file_magwarn(ms, "indirect modifier `%c' " 1685 "invalid", *l); 1686 *lp = l; 1687 return -1; 1688 } 1689 *lp = l; 1690 return 0; 1691 } 1692 1693 private void 1694 parse_op_modifier(struct magic_set *ms, struct magic *m, const char **lp, 1695 int op) 1696 { 1697 const char *l = *lp; 1698 char *t; 1699 uint64_t val; 1700 1701 ++l; 1702 m->mask_op |= op; 1703 val = (uint64_t)strtoull(l, &t, 0); 1704 l = t; 1705 m->num_mask = file_signextend(ms, m, val); 1706 eatsize(&l); 1707 *lp = l; 1708 } 1709 1710 private int 1711 parse_string_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1712 { 1713 const char *l = *lp; 1714 char *t; 1715 int have_range = 0; 1716 1717 while (!isspace((unsigned char)*++l)) { 1718 switch (*l) { 1719 case '0': case '1': case '2': 1720 case '3': case '4': case '5': 1721 case '6': case '7': case '8': 1722 case '9': 1723 if (have_range && (ms->flags & MAGIC_CHECK)) 1724 file_magwarn(ms, "multiple ranges"); 1725 have_range = 1; 1726 m->str_range = CAST(uint32_t, strtoul(l, &t, 0)); 1727 if (m->str_range == 0) 1728 file_magwarn(ms, "zero range"); 1729 l = t - 1; 1730 break; 1731 case CHAR_COMPACT_WHITESPACE: 1732 m->str_flags |= STRING_COMPACT_WHITESPACE; 1733 break; 1734 case CHAR_COMPACT_OPTIONAL_WHITESPACE: 1735 m->str_flags |= STRING_COMPACT_OPTIONAL_WHITESPACE; 1736 break; 1737 case CHAR_IGNORE_LOWERCASE: 1738 m->str_flags |= STRING_IGNORE_LOWERCASE; 1739 break; 1740 case CHAR_IGNORE_UPPERCASE: 1741 m->str_flags |= STRING_IGNORE_UPPERCASE; 1742 break; 1743 case CHAR_REGEX_OFFSET_START: 1744 m->str_flags |= REGEX_OFFSET_START; 1745 break; 1746 case CHAR_BINTEST: 1747 m->str_flags |= STRING_BINTEST; 1748 break; 1749 case CHAR_TEXTTEST: 1750 m->str_flags |= STRING_TEXTTEST; 1751 break; 1752 case CHAR_TRIM: 1753 m->str_flags |= STRING_TRIM; 1754 break; 1755 case CHAR_PSTRING_1_LE: 1756 #define SET_LENGTH(a) m->str_flags = (m->str_flags & ~PSTRING_LEN) | (a) 1757 if (m->type != FILE_PSTRING) 1758 goto bad; 1759 SET_LENGTH(PSTRING_1_LE); 1760 break; 1761 case CHAR_PSTRING_2_BE: 1762 if (m->type != FILE_PSTRING) 1763 goto bad; 1764 SET_LENGTH(PSTRING_2_BE); 1765 break; 1766 case CHAR_PSTRING_2_LE: 1767 if (m->type != FILE_PSTRING) 1768 goto bad; 1769 SET_LENGTH(PSTRING_2_LE); 1770 break; 1771 case CHAR_PSTRING_4_BE: 1772 if (m->type != FILE_PSTRING) 1773 goto bad; 1774 SET_LENGTH(PSTRING_4_BE); 1775 break; 1776 case CHAR_PSTRING_4_LE: 1777 switch (m->type) { 1778 case FILE_PSTRING: 1779 case FILE_REGEX: 1780 break; 1781 default: 1782 goto bad; 1783 } 1784 SET_LENGTH(PSTRING_4_LE); 1785 break; 1786 case CHAR_PSTRING_LENGTH_INCLUDES_ITSELF: 1787 if (m->type != FILE_PSTRING) 1788 goto bad; 1789 m->str_flags |= PSTRING_LENGTH_INCLUDES_ITSELF; 1790 break; 1791 default: 1792 bad: 1793 if (ms->flags & MAGIC_CHECK) 1794 file_magwarn(ms, "string modifier `%c' " 1795 "invalid", *l); 1796 goto out; 1797 } 1798 /* allow multiple '/' for readability */ 1799 if (l[1] == '/' && !isspace((unsigned char)l[2])) 1800 l++; 1801 } 1802 if (string_modifier_check(ms, m) == -1) 1803 goto out; 1804 *lp = l; 1805 return 0; 1806 out: 1807 *lp = l; 1808 return -1; 1809 } 1810 1811 /* 1812 * parse one line from magic file, put into magic[index++] if valid 1813 */ 1814 private int 1815 parse(struct magic_set *ms, struct magic_entry *me, const char *line, 1816 size_t lineno, int action) 1817 { 1818 #ifdef ENABLE_CONDITIONALS 1819 static uint32_t last_cont_level = 0; 1820 #endif 1821 size_t i; 1822 struct magic *m; 1823 const char *l = line; 1824 char *t; 1825 int op; 1826 uint32_t cont_level; 1827 int32_t diff; 1828 1829 cont_level = 0; 1830 1831 /* 1832 * Parse the offset. 1833 */ 1834 while (*l == '>') { 1835 ++l; /* step over */ 1836 cont_level++; 1837 } 1838 #ifdef ENABLE_CONDITIONALS 1839 if (cont_level == 0 || cont_level > last_cont_level) 1840 if (file_check_mem(ms, cont_level) == -1) 1841 return -1; 1842 last_cont_level = cont_level; 1843 #endif 1844 if (cont_level != 0) { 1845 if (me->mp == NULL) { 1846 file_magerror(ms, "No current entry for continuation"); 1847 return -1; 1848 } 1849 if (me->cont_count == 0) { 1850 file_magerror(ms, "Continuations present with 0 count"); 1851 return -1; 1852 } 1853 m = &me->mp[me->cont_count - 1]; 1854 diff = (int32_t)cont_level - (int32_t)m->cont_level; 1855 if (diff > 1) 1856 file_magwarn(ms, "New continuation level %u is more " 1857 "than one larger than current level %u", cont_level, 1858 m->cont_level); 1859 if (me->cont_count == me->max_count) { 1860 struct magic *nm; 1861 size_t cnt = me->max_count + ALLOC_CHUNK; 1862 if ((nm = CAST(struct magic *, realloc(me->mp, 1863 sizeof(*nm) * cnt))) == NULL) { 1864 file_oomem(ms, sizeof(*nm) * cnt); 1865 return -1; 1866 } 1867 me->mp = m = nm; 1868 me->max_count = CAST(uint32_t, cnt); 1869 } 1870 m = &me->mp[me->cont_count++]; 1871 (void)memset(m, 0, sizeof(*m)); 1872 m->cont_level = cont_level; 1873 } else { 1874 static const size_t len = sizeof(*m) * ALLOC_CHUNK; 1875 if (me->mp != NULL) 1876 return 1; 1877 if ((m = CAST(struct magic *, malloc(len))) == NULL) { 1878 file_oomem(ms, len); 1879 return -1; 1880 } 1881 me->mp = m; 1882 me->max_count = ALLOC_CHUNK; 1883 (void)memset(m, 0, sizeof(*m)); 1884 m->factor_op = FILE_FACTOR_OP_NONE; 1885 m->cont_level = 0; 1886 me->cont_count = 1; 1887 } 1888 m->lineno = CAST(uint32_t, lineno); 1889 1890 if (*l == '&') { /* m->cont_level == 0 checked below. */ 1891 ++l; /* step over */ 1892 m->flag |= OFFADD; 1893 } 1894 if (*l == '(') { 1895 ++l; /* step over */ 1896 m->flag |= INDIR; 1897 if (m->flag & OFFADD) 1898 m->flag = (m->flag & ~OFFADD) | INDIROFFADD; 1899 1900 if (*l == '&') { /* m->cont_level == 0 checked below */ 1901 ++l; /* step over */ 1902 m->flag |= OFFADD; 1903 } 1904 } 1905 /* Indirect offsets are not valid at level 0. */ 1906 if (m->cont_level == 0 && (m->flag & (OFFADD | INDIROFFADD))) { 1907 if (ms->flags & MAGIC_CHECK) 1908 file_magwarn(ms, "relative offset at level 0"); 1909 return -1; 1910 } 1911 1912 /* get offset, then skip over it */ 1913 m->offset = (uint32_t)strtoul(l, &t, 0); 1914 if (l == t) { 1915 if (ms->flags & MAGIC_CHECK) 1916 file_magwarn(ms, "offset `%s' invalid", l); 1917 return -1; 1918 } 1919 l = t; 1920 1921 if (m->flag & INDIR) { 1922 m->in_type = FILE_LONG; 1923 m->in_offset = 0; 1924 m->in_op = 0; 1925 /* 1926 * read [.,lbs][+-]nnnnn) 1927 */ 1928 if (*l == '.' || *l == ',') { 1929 if (*l == ',') 1930 m->in_op |= FILE_OPSIGNED; 1931 l++; 1932 switch (*l) { 1933 case 'l': 1934 m->in_type = FILE_LELONG; 1935 break; 1936 case 'L': 1937 m->in_type = FILE_BELONG; 1938 break; 1939 case 'm': 1940 m->in_type = FILE_MELONG; 1941 break; 1942 case 'h': 1943 case 's': 1944 m->in_type = FILE_LESHORT; 1945 break; 1946 case 'H': 1947 case 'S': 1948 m->in_type = FILE_BESHORT; 1949 break; 1950 case 'c': 1951 case 'b': 1952 case 'C': 1953 case 'B': 1954 m->in_type = FILE_BYTE; 1955 break; 1956 case 'e': 1957 case 'f': 1958 case 'g': 1959 m->in_type = FILE_LEDOUBLE; 1960 break; 1961 case 'E': 1962 case 'F': 1963 case 'G': 1964 m->in_type = FILE_BEDOUBLE; 1965 break; 1966 case 'i': 1967 m->in_type = FILE_LEID3; 1968 break; 1969 case 'I': 1970 m->in_type = FILE_BEID3; 1971 break; 1972 default: 1973 if (ms->flags & MAGIC_CHECK) 1974 file_magwarn(ms, 1975 "indirect offset type `%c' invalid", 1976 *l); 1977 return -1; 1978 } 1979 l++; 1980 } 1981 1982 if (*l == '~') { 1983 m->in_op |= FILE_OPINVERSE; 1984 l++; 1985 } 1986 if ((op = get_op(*l)) != -1) { 1987 m->in_op |= op; 1988 l++; 1989 } 1990 if (*l == '(') { 1991 m->in_op |= FILE_OPINDIRECT; 1992 l++; 1993 } 1994 if (isdigit((unsigned char)*l) || *l == '-') { 1995 m->in_offset = (int32_t)strtol(l, &t, 0); 1996 if (l == t) { 1997 if (ms->flags & MAGIC_CHECK) 1998 file_magwarn(ms, 1999 "in_offset `%s' invalid", l); 2000 return -1; 2001 } 2002 l = t; 2003 } 2004 if (*l++ != ')' || 2005 ((m->in_op & FILE_OPINDIRECT) && *l++ != ')')) { 2006 if (ms->flags & MAGIC_CHECK) 2007 file_magwarn(ms, 2008 "missing ')' in indirect offset"); 2009 return -1; 2010 } 2011 } 2012 EATAB; 2013 2014 #ifdef ENABLE_CONDITIONALS 2015 m->cond = get_cond(l, &l); 2016 if (check_cond(ms, m->cond, cont_level) == -1) 2017 return -1; 2018 2019 EATAB; 2020 #endif 2021 2022 /* 2023 * Parse the type. 2024 */ 2025 if (*l == 'u') { 2026 /* 2027 * Try it as a keyword type prefixed by "u"; match what 2028 * follows the "u". If that fails, try it as an SUS 2029 * integer type. 2030 */ 2031 m->type = get_type(type_tbl, l + 1, &l); 2032 if (m->type == FILE_INVALID) { 2033 /* 2034 * Not a keyword type; parse it as an SUS type, 2035 * 'u' possibly followed by a number or C/S/L. 2036 */ 2037 m->type = get_standard_integer_type(l, &l); 2038 } 2039 /* It's unsigned. */ 2040 if (m->type != FILE_INVALID) 2041 m->flag |= UNSIGNED; 2042 } else { 2043 /* 2044 * Try it as a keyword type. If that fails, try it as 2045 * an SUS integer type if it begins with "d" or as an 2046 * SUS string type if it begins with "s". In any case, 2047 * it's not unsigned. 2048 */ 2049 m->type = get_type(type_tbl, l, &l); 2050 if (m->type == FILE_INVALID) { 2051 /* 2052 * Not a keyword type; parse it as an SUS type, 2053 * either 'd' possibly followed by a number or 2054 * C/S/L, or just 's'. 2055 */ 2056 if (*l == 'd') 2057 m->type = get_standard_integer_type(l, &l); 2058 else if (*l == 's' && !isalpha((unsigned char)l[1])) { 2059 m->type = FILE_STRING; 2060 ++l; 2061 } 2062 } 2063 } 2064 2065 if (m->type == FILE_INVALID) { 2066 /* Not found - try it as a special keyword. */ 2067 m->type = get_type(special_tbl, l, &l); 2068 } 2069 2070 if (m->type == FILE_INVALID) { 2071 if (ms->flags & MAGIC_CHECK) 2072 file_magwarn(ms, "type `%s' invalid", l); 2073 return -1; 2074 } 2075 2076 /* New-style anding: "0 byte&0x80 =0x80 dynamically linked" */ 2077 /* New and improved: ~ & | ^ + - * / % -- exciting, isn't it? */ 2078 2079 m->mask_op = 0; 2080 if (*l == '~') { 2081 if (!IS_STRING(m->type)) 2082 m->mask_op |= FILE_OPINVERSE; 2083 else if (ms->flags & MAGIC_CHECK) 2084 file_magwarn(ms, "'~' invalid for string types"); 2085 ++l; 2086 } 2087 m->str_range = 0; 2088 m->str_flags = m->type == FILE_PSTRING ? PSTRING_1_LE : 0; 2089 if ((op = get_op(*l)) != -1) { 2090 if (IS_STRING(m->type)) { 2091 int r; 2092 2093 if (op != FILE_OPDIVIDE) { 2094 if (ms->flags & MAGIC_CHECK) 2095 file_magwarn(ms, 2096 "invalid string/indirect op: " 2097 "`%c'", *t); 2098 return -1; 2099 } 2100 2101 if (m->type == FILE_INDIRECT) 2102 r = parse_indirect_modifier(ms, m, &l); 2103 else 2104 r = parse_string_modifier(ms, m, &l); 2105 if (r == -1) 2106 return -1; 2107 } else 2108 parse_op_modifier(ms, m, &l, op); 2109 } 2110 2111 /* 2112 * We used to set mask to all 1's here, instead let's just not do 2113 * anything if mask = 0 (unless you have a better idea) 2114 */ 2115 EATAB; 2116 2117 switch (*l) { 2118 case '>': 2119 case '<': 2120 m->reln = *l; 2121 ++l; 2122 if (*l == '=') { 2123 if (ms->flags & MAGIC_CHECK) { 2124 file_magwarn(ms, "%c= not supported", 2125 m->reln); 2126 return -1; 2127 } 2128 ++l; 2129 } 2130 break; 2131 /* Old-style anding: "0 byte &0x80 dynamically linked" */ 2132 case '&': 2133 case '^': 2134 case '=': 2135 m->reln = *l; 2136 ++l; 2137 if (*l == '=') { 2138 /* HP compat: ignore &= etc. */ 2139 ++l; 2140 } 2141 break; 2142 case '!': 2143 m->reln = *l; 2144 ++l; 2145 break; 2146 default: 2147 m->reln = '='; /* the default relation */ 2148 if (*l == 'x' && ((isascii((unsigned char)l[1]) && 2149 isspace((unsigned char)l[1])) || !l[1])) { 2150 m->reln = *l; 2151 ++l; 2152 } 2153 break; 2154 } 2155 /* 2156 * Grab the value part, except for an 'x' reln. 2157 */ 2158 if (m->reln != 'x' && getvalue(ms, m, &l, action)) 2159 return -1; 2160 2161 /* 2162 * TODO finish this macro and start using it! 2163 * #define offsetcheck {if (offset > ms->bytes_max -1) 2164 * magwarn("offset too big"); } 2165 */ 2166 2167 /* 2168 * Now get last part - the description 2169 */ 2170 EATAB; 2171 if (l[0] == '\b') { 2172 ++l; 2173 m->flag |= NOSPACE; 2174 } else if ((l[0] == '\\') && (l[1] == 'b')) { 2175 ++l; 2176 ++l; 2177 m->flag |= NOSPACE; 2178 } 2179 for (i = 0; (m->desc[i++] = *l++) != '\0' && i < sizeof(m->desc); ) 2180 continue; 2181 if (i == sizeof(m->desc)) { 2182 m->desc[sizeof(m->desc) - 1] = '\0'; 2183 if (ms->flags & MAGIC_CHECK) 2184 file_magwarn(ms, "description `%s' truncated", m->desc); 2185 } 2186 2187 /* 2188 * We only do this check while compiling, or if any of the magic 2189 * files were not compiled. 2190 */ 2191 if (ms->flags & MAGIC_CHECK) { 2192 if (check_format(ms, m) == -1) 2193 return -1; 2194 } 2195 #ifndef COMPILE_ONLY 2196 if (action == FILE_CHECK) { 2197 file_mdump(m); 2198 } 2199 #endif 2200 m->mimetype[0] = '\0'; /* initialise MIME type to none */ 2201 return 0; 2202 } 2203 2204 /* 2205 * parse a STRENGTH annotation line from magic file, put into magic[index - 1] 2206 * if valid 2207 */ 2208 private int 2209 parse_strength(struct magic_set *ms, struct magic_entry *me, const char *line) 2210 { 2211 const char *l = line; 2212 char *el; 2213 unsigned long factor; 2214 struct magic *m = &me->mp[0]; 2215 2216 if (m->factor_op != FILE_FACTOR_OP_NONE) { 2217 file_magwarn(ms, 2218 "Current entry already has a strength type: %c %d", 2219 m->factor_op, m->factor); 2220 return -1; 2221 } 2222 if (m->type == FILE_NAME) { 2223 file_magwarn(ms, "%s: Strength setting is not supported in " 2224 "\"name\" magic entries", m->value.s); 2225 return -1; 2226 } 2227 EATAB; 2228 switch (*l) { 2229 case FILE_FACTOR_OP_NONE: 2230 case FILE_FACTOR_OP_PLUS: 2231 case FILE_FACTOR_OP_MINUS: 2232 case FILE_FACTOR_OP_TIMES: 2233 case FILE_FACTOR_OP_DIV: 2234 m->factor_op = *l++; 2235 break; 2236 default: 2237 file_magwarn(ms, "Unknown factor op `%c'", *l); 2238 return -1; 2239 } 2240 EATAB; 2241 factor = strtoul(l, &el, 0); 2242 if (factor > 255) { 2243 file_magwarn(ms, "Too large factor `%lu'", factor); 2244 goto out; 2245 } 2246 if (*el && !isspace((unsigned char)*el)) { 2247 file_magwarn(ms, "Bad factor `%s'", l); 2248 goto out; 2249 } 2250 m->factor = (uint8_t)factor; 2251 if (m->factor == 0 && m->factor_op == FILE_FACTOR_OP_DIV) { 2252 file_magwarn(ms, "Cannot have factor op `%c' and factor %u", 2253 m->factor_op, m->factor); 2254 goto out; 2255 } 2256 return 0; 2257 out: 2258 m->factor_op = FILE_FACTOR_OP_NONE; 2259 m->factor = 0; 2260 return -1; 2261 } 2262 2263 private int 2264 goodchar(unsigned char x, const char *extra) 2265 { 2266 return (isascii(x) && isalnum(x)) || strchr(extra, x); 2267 } 2268 2269 private int 2270 parse_extra(struct magic_set *ms, struct magic_entry *me, const char *line, 2271 off_t off, size_t len, const char *name, const char *extra, int nt) 2272 { 2273 size_t i; 2274 const char *l = line; 2275 struct magic *m = &me->mp[me->cont_count == 0 ? 0 : me->cont_count - 1]; 2276 char *buf = CAST(char *, CAST(void *, m)) + off; 2277 2278 if (buf[0] != '\0') { 2279 len = nt ? strlen(buf) : len; 2280 file_magwarn(ms, "Current entry already has a %s type " 2281 "`%.*s', new type `%s'", name, (int)len, buf, l); 2282 return -1; 2283 } 2284 2285 if (*m->desc == '\0') { 2286 file_magwarn(ms, "Current entry does not yet have a " 2287 "description for adding a %s type", name); 2288 return -1; 2289 } 2290 2291 EATAB; 2292 for (i = 0; *l && i < len && goodchar(*l, extra); buf[i++] = *l++) 2293 continue; 2294 2295 if (i == len && *l) { 2296 if (nt) 2297 buf[len - 1] = '\0'; 2298 if (ms->flags & MAGIC_CHECK) 2299 file_magwarn(ms, "%s type `%s' truncated %" 2300 SIZE_T_FORMAT "u", name, line, i); 2301 } else { 2302 if (!isspace((unsigned char)*l) && !goodchar(*l, extra)) 2303 file_magwarn(ms, "%s type `%s' has bad char '%c'", 2304 name, line, *l); 2305 if (nt) 2306 buf[i] = '\0'; 2307 } 2308 2309 if (i > 0) 2310 return 0; 2311 2312 file_magerror(ms, "Bad magic entry '%s'", line); 2313 return -1; 2314 } 2315 2316 /* 2317 * Parse an Apple CREATOR/TYPE annotation from magic file and put it into 2318 * magic[index - 1] 2319 */ 2320 private int 2321 parse_apple(struct magic_set *ms, struct magic_entry *me, const char *line) 2322 { 2323 struct magic *m = &me->mp[0]; 2324 2325 return parse_extra(ms, me, line, 2326 CAST(off_t, offsetof(struct magic, apple)), 2327 sizeof(m->apple), "APPLE", "!+-./?", 0); 2328 } 2329 2330 /* 2331 * Parse a comma-separated list of extensions 2332 */ 2333 private int 2334 parse_ext(struct magic_set *ms, struct magic_entry *me, const char *line) 2335 { 2336 struct magic *m = &me->mp[0]; 2337 2338 return parse_extra(ms, me, line, 2339 CAST(off_t, offsetof(struct magic, ext)), 2340 sizeof(m->ext), "EXTENSION", ",!+-/@", 0); 2341 } 2342 2343 /* 2344 * parse a MIME annotation line from magic file, put into magic[index - 1] 2345 * if valid 2346 */ 2347 private int 2348 parse_mime(struct magic_set *ms, struct magic_entry *me, const char *line) 2349 { 2350 struct magic *m = &me->mp[0]; 2351 2352 return parse_extra(ms, me, line, 2353 CAST(off_t, offsetof(struct magic, mimetype)), 2354 sizeof(m->mimetype), "MIME", "+-/.", 1); 2355 } 2356 2357 private int 2358 check_format_type(const char *ptr, int type, const char **estr) 2359 { 2360 int quad = 0, h; 2361 size_t len, cnt; 2362 if (*ptr == '\0') { 2363 /* Missing format string; bad */ 2364 *estr = "missing format spec"; 2365 return -1; 2366 } 2367 2368 switch (file_formats[type]) { 2369 case FILE_FMT_QUAD: 2370 quad = 1; 2371 /*FALLTHROUGH*/ 2372 case FILE_FMT_NUM: 2373 if (quad == 0) { 2374 switch (type) { 2375 case FILE_BYTE: 2376 h = 2; 2377 break; 2378 case FILE_SHORT: 2379 case FILE_BESHORT: 2380 case FILE_LESHORT: 2381 h = 1; 2382 break; 2383 case FILE_LONG: 2384 case FILE_BELONG: 2385 case FILE_LELONG: 2386 case FILE_MELONG: 2387 case FILE_LEID3: 2388 case FILE_BEID3: 2389 case FILE_INDIRECT: 2390 h = 0; 2391 break; 2392 default: 2393 abort(); 2394 } 2395 } else 2396 h = 0; 2397 if (*ptr == '-') 2398 ptr++; 2399 if (*ptr == '.') 2400 ptr++; 2401 if (*ptr == '#') 2402 ptr++; 2403 #define CHECKLEN() do { \ 2404 for (len = cnt = 0; isdigit((unsigned char)*ptr); ptr++, cnt++) \ 2405 len = len * 10 + (*ptr - '0'); \ 2406 if (cnt > 5 || len > 1024) \ 2407 goto toolong; \ 2408 } while (/*CONSTCOND*/0) 2409 2410 CHECKLEN(); 2411 if (*ptr == '.') 2412 ptr++; 2413 CHECKLEN(); 2414 if (quad) { 2415 if (*ptr++ != 'l') 2416 goto invalid; 2417 if (*ptr++ != 'l') 2418 goto invalid; 2419 } 2420 2421 switch (*ptr++) { 2422 #ifdef STRICT_FORMAT /* "long" formats are int formats for us */ 2423 /* so don't accept the 'l' modifier */ 2424 case 'l': 2425 switch (*ptr++) { 2426 case 'i': 2427 case 'd': 2428 case 'u': 2429 case 'o': 2430 case 'x': 2431 case 'X': 2432 if (h == 0) 2433 return 0; 2434 /*FALLTHROUGH*/ 2435 default: 2436 goto invalid; 2437 } 2438 2439 /* 2440 * Don't accept h and hh modifiers. They make writing 2441 * magic entries more complicated, for very little benefit 2442 */ 2443 case 'h': 2444 if (h-- <= 0) 2445 goto invalid; 2446 switch (*ptr++) { 2447 case 'h': 2448 if (h-- <= 0) 2449 goto invalid; 2450 switch (*ptr++) { 2451 case 'i': 2452 case 'd': 2453 case 'u': 2454 case 'o': 2455 case 'x': 2456 case 'X': 2457 return 0; 2458 default: 2459 goto invalid; 2460 } 2461 case 'i': 2462 case 'd': 2463 case 'u': 2464 case 'o': 2465 case 'x': 2466 case 'X': 2467 if (h == 0) 2468 return 0; 2469 /*FALLTHROUGH*/ 2470 default: 2471 goto invalid; 2472 } 2473 #endif 2474 case 'c': 2475 if (h == 2) 2476 return 0; 2477 goto invalid; 2478 case 'i': 2479 case 'd': 2480 case 'u': 2481 case 'o': 2482 case 'x': 2483 case 'X': 2484 #ifdef STRICT_FORMAT 2485 if (h == 0) 2486 return 0; 2487 /*FALLTHROUGH*/ 2488 #else 2489 return 0; 2490 #endif 2491 default: 2492 goto invalid; 2493 } 2494 2495 case FILE_FMT_FLOAT: 2496 case FILE_FMT_DOUBLE: 2497 if (*ptr == '-') 2498 ptr++; 2499 if (*ptr == '.') 2500 ptr++; 2501 CHECKLEN(); 2502 if (*ptr == '.') 2503 ptr++; 2504 CHECKLEN(); 2505 switch (*ptr++) { 2506 case 'e': 2507 case 'E': 2508 case 'f': 2509 case 'F': 2510 case 'g': 2511 case 'G': 2512 return 0; 2513 2514 default: 2515 goto invalid; 2516 } 2517 2518 2519 case FILE_FMT_STR: 2520 if (*ptr == '-') 2521 ptr++; 2522 while (isdigit((unsigned char )*ptr)) 2523 ptr++; 2524 if (*ptr == '.') { 2525 ptr++; 2526 while (isdigit((unsigned char )*ptr)) 2527 ptr++; 2528 } 2529 2530 switch (*ptr++) { 2531 case 's': 2532 return 0; 2533 default: 2534 goto invalid; 2535 } 2536 2537 default: 2538 /* internal error */ 2539 abort(); 2540 } 2541 invalid: 2542 *estr = "not valid"; 2543 toolong: 2544 *estr = "too long"; 2545 return -1; 2546 } 2547 2548 /* 2549 * Check that the optional printf format in description matches 2550 * the type of the magic. 2551 */ 2552 private int 2553 check_format(struct magic_set *ms, struct magic *m) 2554 { 2555 char *ptr; 2556 const char *estr; 2557 2558 for (ptr = m->desc; *ptr; ptr++) 2559 if (*ptr == '%') 2560 break; 2561 if (*ptr == '\0') { 2562 /* No format string; ok */ 2563 return 1; 2564 } 2565 2566 assert(file_nformats == file_nnames); 2567 2568 if (m->type >= file_nformats) { 2569 file_magwarn(ms, "Internal error inconsistency between " 2570 "m->type and format strings"); 2571 return -1; 2572 } 2573 if (file_formats[m->type] == FILE_FMT_NONE) { 2574 file_magwarn(ms, "No format string for `%s' with description " 2575 "`%s'", m->desc, file_names[m->type]); 2576 return -1; 2577 } 2578 2579 ptr++; 2580 if (check_format_type(ptr, m->type, &estr) == -1) { 2581 /* 2582 * TODO: this error message is unhelpful if the format 2583 * string is not one character long 2584 */ 2585 file_magwarn(ms, "Printf format is %s for type " 2586 "`%s' in description `%s'", estr, 2587 file_names[m->type], m->desc); 2588 return -1; 2589 } 2590 2591 for (; *ptr; ptr++) { 2592 if (*ptr == '%') { 2593 file_magwarn(ms, 2594 "Too many format strings (should have at most one) " 2595 "for `%s' with description `%s'", 2596 file_names[m->type], m->desc); 2597 return -1; 2598 } 2599 } 2600 return 0; 2601 } 2602 2603 /* 2604 * Read a numeric value from a pointer, into the value union of a magic 2605 * pointer, according to the magic type. Update the string pointer to point 2606 * just after the number read. Return 0 for success, non-zero for failure. 2607 */ 2608 private int 2609 getvalue(struct magic_set *ms, struct magic *m, const char **p, int action) 2610 { 2611 switch (m->type) { 2612 case FILE_BESTRING16: 2613 case FILE_LESTRING16: 2614 case FILE_STRING: 2615 case FILE_PSTRING: 2616 case FILE_REGEX: 2617 case FILE_SEARCH: 2618 case FILE_NAME: 2619 case FILE_USE: 2620 case FILE_DER: 2621 *p = getstr(ms, m, *p, action == FILE_COMPILE); 2622 if (*p == NULL) { 2623 if (ms->flags & MAGIC_CHECK) 2624 file_magwarn(ms, "cannot get string from `%s'", 2625 m->value.s); 2626 return -1; 2627 } 2628 if (m->type == FILE_REGEX) { 2629 file_regex_t rx; 2630 int rc = file_regcomp(&rx, m->value.s, REG_EXTENDED); 2631 if (rc) { 2632 if (ms->flags & MAGIC_CHECK) 2633 file_regerror(&rx, rc, ms); 2634 } 2635 file_regfree(&rx); 2636 return rc ? -1 : 0; 2637 } 2638 return 0; 2639 case FILE_FLOAT: 2640 case FILE_BEFLOAT: 2641 case FILE_LEFLOAT: 2642 if (m->reln != 'x') { 2643 char *ep; 2644 errno = 0; 2645 #ifdef HAVE_STRTOF 2646 m->value.f = strtof(*p, &ep); 2647 #else 2648 m->value.f = (float)strtod(*p, &ep); 2649 #endif 2650 if (errno == 0) 2651 *p = ep; 2652 } 2653 return 0; 2654 case FILE_DOUBLE: 2655 case FILE_BEDOUBLE: 2656 case FILE_LEDOUBLE: 2657 if (m->reln != 'x') { 2658 char *ep; 2659 errno = 0; 2660 m->value.d = strtod(*p, &ep); 2661 if (errno == 0) 2662 *p = ep; 2663 } 2664 return 0; 2665 default: 2666 if (m->reln != 'x') { 2667 char *ep; 2668 uint64_t ull; 2669 errno = 0; 2670 ull = (uint64_t)strtoull(*p, &ep, 0); 2671 m->value.q = file_signextend(ms, m, ull); 2672 if (*p == ep) { 2673 file_magwarn(ms, "Unparseable number `%s'", *p); 2674 } else { 2675 size_t ts = typesize(m->type); 2676 uint64_t x; 2677 const char *q; 2678 2679 if (ts == (size_t)~0) { 2680 file_magwarn(ms, "Expected numeric type got `%s'", 2681 type_tbl[m->type].name); 2682 } 2683 for (q = *p; isspace((unsigned char)*q); q++) 2684 continue; 2685 if (*q == '-') 2686 ull = -(int64_t)ull; 2687 switch (ts) { 2688 case 1: 2689 x = ull & ~0xffULL; 2690 break; 2691 case 2: 2692 x = ull & ~0xffffULL; 2693 break; 2694 case 4: 2695 x = ull & ~0xffffffffULL; 2696 break; 2697 case 8: 2698 x = 0; 2699 break; 2700 default: 2701 abort(); 2702 } 2703 if (x) { 2704 file_magwarn(ms, "Overflow for numeric type `%s' value %#" PRIx64, 2705 type_tbl[m->type].name, ull); 2706 } 2707 } 2708 if (errno == 0) { 2709 *p = ep; 2710 eatsize(p); 2711 } 2712 } 2713 return 0; 2714 } 2715 } 2716 2717 /* 2718 * Convert a string containing C character escapes. Stop at an unescaped 2719 * space or tab. 2720 * Copy the converted version to "m->value.s", and the length in m->vallen. 2721 * Return updated scan pointer as function result. Warn if set. 2722 */ 2723 private const char * 2724 getstr(struct magic_set *ms, struct magic *m, const char *s, int warn) 2725 { 2726 const char *origs = s; 2727 char *p = m->value.s; 2728 size_t plen = sizeof(m->value.s); 2729 char *origp = p; 2730 char *pmax = p + plen - 1; 2731 int c; 2732 int val; 2733 2734 while ((c = *s++) != '\0') { 2735 if (isspace((unsigned char) c)) 2736 break; 2737 if (p >= pmax) { 2738 file_error(ms, 0, "string too long: `%s'", origs); 2739 return NULL; 2740 } 2741 if (c == '\\') { 2742 switch(c = *s++) { 2743 2744 case '\0': 2745 if (warn) 2746 file_magwarn(ms, "incomplete escape"); 2747 s--; 2748 goto out; 2749 2750 case '\t': 2751 if (warn) { 2752 file_magwarn(ms, 2753 "escaped tab found, use \\t instead"); 2754 warn = 0; /* already did */ 2755 } 2756 /*FALLTHROUGH*/ 2757 default: 2758 if (warn) { 2759 if (isprint((unsigned char)c)) { 2760 /* Allow escaping of 2761 * ``relations'' */ 2762 if (strchr("<>&^=!", c) == NULL 2763 && (m->type != FILE_REGEX || 2764 strchr("[]().*?^$|{}", c) 2765 == NULL)) { 2766 file_magwarn(ms, "no " 2767 "need to escape " 2768 "`%c'", c); 2769 } 2770 } else { 2771 file_magwarn(ms, 2772 "unknown escape sequence: " 2773 "\\%03o", c); 2774 } 2775 } 2776 /*FALLTHROUGH*/ 2777 /* space, perhaps force people to use \040? */ 2778 case ' ': 2779 #if 0 2780 /* 2781 * Other things people escape, but shouldn't need to, 2782 * so we disallow them 2783 */ 2784 case '\'': 2785 case '"': 2786 case '?': 2787 #endif 2788 /* Relations */ 2789 case '>': 2790 case '<': 2791 case '&': 2792 case '^': 2793 case '=': 2794 case '!': 2795 /* and baskslash itself */ 2796 case '\\': 2797 *p++ = (char) c; 2798 break; 2799 2800 case 'a': 2801 *p++ = '\a'; 2802 break; 2803 2804 case 'b': 2805 *p++ = '\b'; 2806 break; 2807 2808 case 'f': 2809 *p++ = '\f'; 2810 break; 2811 2812 case 'n': 2813 *p++ = '\n'; 2814 break; 2815 2816 case 'r': 2817 *p++ = '\r'; 2818 break; 2819 2820 case 't': 2821 *p++ = '\t'; 2822 break; 2823 2824 case 'v': 2825 *p++ = '\v'; 2826 break; 2827 2828 /* \ and up to 3 octal digits */ 2829 case '0': 2830 case '1': 2831 case '2': 2832 case '3': 2833 case '4': 2834 case '5': 2835 case '6': 2836 case '7': 2837 val = c - '0'; 2838 c = *s++; /* try for 2 */ 2839 if (c >= '0' && c <= '7') { 2840 val = (val << 3) | (c - '0'); 2841 c = *s++; /* try for 3 */ 2842 if (c >= '0' && c <= '7') 2843 val = (val << 3) | (c-'0'); 2844 else 2845 --s; 2846 } 2847 else 2848 --s; 2849 *p++ = (char)val; 2850 break; 2851 2852 /* \x and up to 2 hex digits */ 2853 case 'x': 2854 val = 'x'; /* Default if no digits */ 2855 c = hextoint(*s++); /* Get next char */ 2856 if (c >= 0) { 2857 val = c; 2858 c = hextoint(*s++); 2859 if (c >= 0) 2860 val = (val << 4) + c; 2861 else 2862 --s; 2863 } else 2864 --s; 2865 *p++ = (char)val; 2866 break; 2867 } 2868 } else 2869 *p++ = (char)c; 2870 } 2871 --s; 2872 out: 2873 *p = '\0'; 2874 m->vallen = CAST(unsigned char, (p - origp)); 2875 if (m->type == FILE_PSTRING) 2876 m->vallen += (unsigned char)file_pstring_length_size(m); 2877 return s; 2878 } 2879 2880 2881 /* Single hex char to int; -1 if not a hex char. */ 2882 private int 2883 hextoint(int c) 2884 { 2885 if (!isascii((unsigned char) c)) 2886 return -1; 2887 if (isdigit((unsigned char) c)) 2888 return c - '0'; 2889 if ((c >= 'a') && (c <= 'f')) 2890 return c + 10 - 'a'; 2891 if (( c>= 'A') && (c <= 'F')) 2892 return c + 10 - 'A'; 2893 return -1; 2894 } 2895 2896 2897 /* 2898 * Print a string containing C character escapes. 2899 */ 2900 protected void 2901 file_showstr(FILE *fp, const char *s, size_t len) 2902 { 2903 char c; 2904 2905 for (;;) { 2906 if (len == ~0U) { 2907 c = *s++; 2908 if (c == '\0') 2909 break; 2910 } 2911 else { 2912 if (len-- == 0) 2913 break; 2914 c = *s++; 2915 } 2916 if (c >= 040 && c <= 0176) /* TODO isprint && !iscntrl */ 2917 (void) fputc(c, fp); 2918 else { 2919 (void) fputc('\\', fp); 2920 switch (c) { 2921 case '\a': 2922 (void) fputc('a', fp); 2923 break; 2924 2925 case '\b': 2926 (void) fputc('b', fp); 2927 break; 2928 2929 case '\f': 2930 (void) fputc('f', fp); 2931 break; 2932 2933 case '\n': 2934 (void) fputc('n', fp); 2935 break; 2936 2937 case '\r': 2938 (void) fputc('r', fp); 2939 break; 2940 2941 case '\t': 2942 (void) fputc('t', fp); 2943 break; 2944 2945 case '\v': 2946 (void) fputc('v', fp); 2947 break; 2948 2949 default: 2950 (void) fprintf(fp, "%.3o", c & 0377); 2951 break; 2952 } 2953 } 2954 } 2955 } 2956 2957 /* 2958 * eatsize(): Eat the size spec from a number [eg. 10UL] 2959 */ 2960 private void 2961 eatsize(const char **p) 2962 { 2963 const char *l = *p; 2964 2965 if (LOWCASE(*l) == 'u') 2966 l++; 2967 2968 switch (LOWCASE(*l)) { 2969 case 'l': /* long */ 2970 case 's': /* short */ 2971 case 'h': /* short */ 2972 case 'b': /* char/byte */ 2973 case 'c': /* char/byte */ 2974 l++; 2975 /*FALLTHROUGH*/ 2976 default: 2977 break; 2978 } 2979 2980 *p = l; 2981 } 2982 2983 /* 2984 * handle a buffer containing a compiled file. 2985 */ 2986 private struct magic_map * 2987 apprentice_buf(struct magic_set *ms, struct magic *buf, size_t len) 2988 { 2989 struct magic_map *map; 2990 2991 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 2992 file_oomem(ms, sizeof(*map)); 2993 return NULL; 2994 } 2995 map->len = len; 2996 map->p = buf; 2997 map->type = MAP_TYPE_USER; 2998 if (check_buffer(ms, map, "buffer") != 0) { 2999 apprentice_unmap(map); 3000 return NULL; 3001 } 3002 return map; 3003 } 3004 3005 /* 3006 * handle a compiled file. 3007 */ 3008 3009 private struct magic_map * 3010 apprentice_map(struct magic_set *ms, const char *fn) 3011 { 3012 int fd; 3013 struct stat st; 3014 char *dbname = NULL; 3015 struct magic_map *map; 3016 struct magic_map *rv = NULL; 3017 3018 fd = -1; 3019 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 3020 file_oomem(ms, sizeof(*map)); 3021 goto error; 3022 } 3023 map->type = MAP_TYPE_USER; /* unspecified */ 3024 3025 dbname = mkdbname(ms, fn, 0); 3026 if (dbname == NULL) 3027 goto error; 3028 3029 if ((fd = open(dbname, O_RDONLY|O_BINARY)) == -1) 3030 goto error; 3031 3032 if (fstat(fd, &st) == -1) { 3033 file_error(ms, errno, "cannot stat `%s'", dbname); 3034 goto error; 3035 } 3036 if (st.st_size < 8 || st.st_size > MAXMAGIC_SIZE) { 3037 file_error(ms, 0, "file `%s' is too %s", dbname, 3038 st.st_size < 8 ? "small" : "large"); 3039 goto error; 3040 } 3041 3042 map->len = (size_t)st.st_size; 3043 #ifdef QUICK 3044 map->type = MAP_TYPE_MMAP; 3045 if ((map->p = mmap(0, (size_t)st.st_size, PROT_READ|PROT_WRITE, 3046 MAP_PRIVATE|MAP_FILE, fd, (off_t)0)) == MAP_FAILED) { 3047 file_error(ms, errno, "cannot map `%s'", dbname); 3048 goto error; 3049 } 3050 #else 3051 map->type = MAP_TYPE_MALLOC; 3052 if ((map->p = CAST(void *, malloc(map->len))) == NULL) { 3053 file_oomem(ms, map->len); 3054 goto error; 3055 } 3056 if (read(fd, map->p, map->len) != (ssize_t)map->len) { 3057 file_badread(ms); 3058 goto error; 3059 } 3060 #define RET 1 3061 #endif 3062 (void)close(fd); 3063 fd = -1; 3064 3065 if (check_buffer(ms, map, dbname) != 0) { 3066 rv = (struct magic_map *)-1; 3067 goto error; 3068 } 3069 #ifdef QUICK 3070 if (mprotect(map->p, (size_t)st.st_size, PROT_READ) == -1) { 3071 file_error(ms, errno, "cannot mprotect `%s'", dbname); 3072 goto error; 3073 } 3074 #endif 3075 3076 free(dbname); 3077 return map; 3078 3079 error: 3080 if (fd != -1) 3081 (void)close(fd); 3082 apprentice_unmap(map); 3083 free(dbname); 3084 return rv; 3085 } 3086 3087 private int 3088 check_buffer(struct magic_set *ms, struct magic_map *map, const char *dbname) 3089 { 3090 uint32_t *ptr; 3091 uint32_t entries, nentries; 3092 uint32_t version; 3093 int i, needsbyteswap; 3094 3095 ptr = CAST(uint32_t *, map->p); 3096 if (*ptr != MAGICNO) { 3097 if (swap4(*ptr) != MAGICNO) { 3098 file_error(ms, 0, "bad magic in `%s'", dbname); 3099 return -1; 3100 } 3101 needsbyteswap = 1; 3102 } else 3103 needsbyteswap = 0; 3104 if (needsbyteswap) 3105 version = swap4(ptr[1]); 3106 else 3107 version = ptr[1]; 3108 if (version != VERSIONNO) { 3109 file_error(ms, 0, "File %s supports only version %d magic " 3110 "files. `%s' is version %d", VERSION, 3111 VERSIONNO, dbname, version); 3112 return -1; 3113 } 3114 entries = (uint32_t)(map->len / sizeof(struct magic)); 3115 if ((entries * sizeof(struct magic)) != map->len) { 3116 file_error(ms, 0, "Size of `%s' %" SIZE_T_FORMAT "u is not " 3117 "a multiple of %" SIZE_T_FORMAT "u", 3118 dbname, map->len, sizeof(struct magic)); 3119 return -1; 3120 } 3121 map->magic[0] = CAST(struct magic *, map->p) + 1; 3122 nentries = 0; 3123 for (i = 0; i < MAGIC_SETS; i++) { 3124 if (needsbyteswap) 3125 map->nmagic[i] = swap4(ptr[i + 2]); 3126 else 3127 map->nmagic[i] = ptr[i + 2]; 3128 if (i != MAGIC_SETS - 1) 3129 map->magic[i + 1] = map->magic[i] + map->nmagic[i]; 3130 nentries += map->nmagic[i]; 3131 } 3132 if (entries != nentries + 1) { 3133 file_error(ms, 0, "Inconsistent entries in `%s' %u != %u", 3134 dbname, entries, nentries + 1); 3135 return -1; 3136 } 3137 if (needsbyteswap) 3138 for (i = 0; i < MAGIC_SETS; i++) 3139 byteswap(map->magic[i], map->nmagic[i]); 3140 return 0; 3141 } 3142 3143 /* 3144 * handle an mmaped file. 3145 */ 3146 private int 3147 apprentice_compile(struct magic_set *ms, struct magic_map *map, const char *fn) 3148 { 3149 static const size_t nm = sizeof(*map->nmagic) * MAGIC_SETS; 3150 static const size_t m = sizeof(**map->magic); 3151 int fd = -1; 3152 size_t len; 3153 char *dbname; 3154 int rv = -1; 3155 uint32_t i; 3156 union { 3157 struct magic m; 3158 uint32_t h[2 + MAGIC_SETS]; 3159 } hdr; 3160 3161 dbname = mkdbname(ms, fn, 1); 3162 3163 if (dbname == NULL) 3164 goto out; 3165 3166 if ((fd = open(dbname, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY, 0644)) == -1) 3167 { 3168 file_error(ms, errno, "cannot open `%s'", dbname); 3169 goto out; 3170 } 3171 memset(&hdr, 0, sizeof(hdr)); 3172 hdr.h[0] = MAGICNO; 3173 hdr.h[1] = VERSIONNO; 3174 memcpy(hdr.h + 2, map->nmagic, nm); 3175 3176 if (write(fd, &hdr, sizeof(hdr)) != (ssize_t)sizeof(hdr)) { 3177 file_error(ms, errno, "error writing `%s'", dbname); 3178 goto out; 3179 } 3180 3181 for (i = 0; i < MAGIC_SETS; i++) { 3182 len = m * map->nmagic[i]; 3183 if (write(fd, map->magic[i], len) != (ssize_t)len) { 3184 file_error(ms, errno, "error writing `%s'", dbname); 3185 goto out; 3186 } 3187 } 3188 3189 if (fd != -1) 3190 (void)close(fd); 3191 rv = 0; 3192 out: 3193 apprentice_unmap(map); 3194 free(dbname); 3195 return rv; 3196 } 3197 3198 private const char ext[] = ".mgc"; 3199 /* 3200 * make a dbname 3201 */ 3202 private char * 3203 mkdbname(struct magic_set *ms, const char *fn, int strip) 3204 { 3205 const char *p, *q; 3206 char *buf; 3207 3208 if (strip) { 3209 if ((p = strrchr(fn, '/')) != NULL) 3210 fn = ++p; 3211 } 3212 3213 for (q = fn; *q; q++) 3214 continue; 3215 /* Look for .mgc */ 3216 for (p = ext + sizeof(ext) - 1; p >= ext && q >= fn; p--, q--) 3217 if (*p != *q) 3218 break; 3219 3220 /* Did not find .mgc, restore q */ 3221 if (p >= ext) 3222 while (*q) 3223 q++; 3224 3225 q++; 3226 /* Compatibility with old code that looked in .mime */ 3227 if (ms->flags & MAGIC_MIME) { 3228 if (asprintf(&buf, "%.*s.mime%s", (int)(q - fn), fn, ext) < 0) 3229 return NULL; 3230 if (access(buf, R_OK) != -1) { 3231 ms->flags &= MAGIC_MIME_TYPE; 3232 return buf; 3233 } 3234 free(buf); 3235 } 3236 if (asprintf(&buf, "%.*s%s", (int)(q - fn), fn, ext) < 0) 3237 return NULL; 3238 3239 /* Compatibility with old code that looked in .mime */ 3240 if (strstr(fn, ".mime") != NULL) 3241 ms->flags &= MAGIC_MIME_TYPE; 3242 return buf; 3243 } 3244 3245 /* 3246 * Byteswap an mmap'ed file if needed 3247 */ 3248 private void 3249 byteswap(struct magic *magic, uint32_t nmagic) 3250 { 3251 uint32_t i; 3252 for (i = 0; i < nmagic; i++) 3253 bs1(&magic[i]); 3254 } 3255 3256 /* 3257 * swap a short 3258 */ 3259 private uint16_t 3260 swap2(uint16_t sv) 3261 { 3262 uint16_t rv; 3263 uint8_t *s = (uint8_t *)(void *)&sv; 3264 uint8_t *d = (uint8_t *)(void *)&rv; 3265 d[0] = s[1]; 3266 d[1] = s[0]; 3267 return rv; 3268 } 3269 3270 /* 3271 * swap an int 3272 */ 3273 private uint32_t 3274 swap4(uint32_t sv) 3275 { 3276 uint32_t rv; 3277 uint8_t *s = (uint8_t *)(void *)&sv; 3278 uint8_t *d = (uint8_t *)(void *)&rv; 3279 d[0] = s[3]; 3280 d[1] = s[2]; 3281 d[2] = s[1]; 3282 d[3] = s[0]; 3283 return rv; 3284 } 3285 3286 /* 3287 * swap a quad 3288 */ 3289 private uint64_t 3290 swap8(uint64_t sv) 3291 { 3292 uint64_t rv; 3293 uint8_t *s = (uint8_t *)(void *)&sv; 3294 uint8_t *d = (uint8_t *)(void *)&rv; 3295 #if 0 3296 d[0] = s[3]; 3297 d[1] = s[2]; 3298 d[2] = s[1]; 3299 d[3] = s[0]; 3300 d[4] = s[7]; 3301 d[5] = s[6]; 3302 d[6] = s[5]; 3303 d[7] = s[4]; 3304 #else 3305 d[0] = s[7]; 3306 d[1] = s[6]; 3307 d[2] = s[5]; 3308 d[3] = s[4]; 3309 d[4] = s[3]; 3310 d[5] = s[2]; 3311 d[6] = s[1]; 3312 d[7] = s[0]; 3313 #endif 3314 return rv; 3315 } 3316 3317 /* 3318 * byteswap a single magic entry 3319 */ 3320 private void 3321 bs1(struct magic *m) 3322 { 3323 m->cont_level = swap2(m->cont_level); 3324 m->offset = swap4((uint32_t)m->offset); 3325 m->in_offset = swap4((uint32_t)m->in_offset); 3326 m->lineno = swap4((uint32_t)m->lineno); 3327 if (IS_STRING(m->type)) { 3328 m->str_range = swap4(m->str_range); 3329 m->str_flags = swap4(m->str_flags); 3330 } 3331 else { 3332 m->value.q = swap8(m->value.q); 3333 m->num_mask = swap8(m->num_mask); 3334 } 3335 } 3336 3337 protected size_t 3338 file_pstring_length_size(const struct magic *m) 3339 { 3340 switch (m->str_flags & PSTRING_LEN) { 3341 case PSTRING_1_LE: 3342 return 1; 3343 case PSTRING_2_LE: 3344 case PSTRING_2_BE: 3345 return 2; 3346 case PSTRING_4_LE: 3347 case PSTRING_4_BE: 3348 return 4; 3349 default: 3350 abort(); /* Impossible */ 3351 return 1; 3352 } 3353 } 3354 protected size_t 3355 file_pstring_get_length(const struct magic *m, const char *ss) 3356 { 3357 size_t len = 0; 3358 const unsigned char *s = (const unsigned char *)ss; 3359 unsigned int s3, s2, s1, s0; 3360 3361 switch (m->str_flags & PSTRING_LEN) { 3362 case PSTRING_1_LE: 3363 len = *s; 3364 break; 3365 case PSTRING_2_LE: 3366 s0 = s[0]; 3367 s1 = s[1]; 3368 len = (s1 << 8) | s0; 3369 break; 3370 case PSTRING_2_BE: 3371 s0 = s[0]; 3372 s1 = s[1]; 3373 len = (s0 << 8) | s1; 3374 break; 3375 case PSTRING_4_LE: 3376 s0 = s[0]; 3377 s1 = s[1]; 3378 s2 = s[2]; 3379 s3 = s[3]; 3380 len = (s3 << 24) | (s2 << 16) | (s1 << 8) | s0; 3381 break; 3382 case PSTRING_4_BE: 3383 s0 = s[0]; 3384 s1 = s[1]; 3385 s2 = s[2]; 3386 s3 = s[3]; 3387 len = (s0 << 24) | (s1 << 16) | (s2 << 8) | s3; 3388 break; 3389 default: 3390 abort(); /* Impossible */ 3391 } 3392 3393 if (m->str_flags & PSTRING_LENGTH_INCLUDES_ITSELF) 3394 len -= file_pstring_length_size(m); 3395 3396 return len; 3397 } 3398 3399 protected int 3400 file_magicfind(struct magic_set *ms, const char *name, struct mlist *v) 3401 { 3402 uint32_t i, j; 3403 struct mlist *mlist, *ml; 3404 3405 mlist = ms->mlist[1]; 3406 3407 for (ml = mlist->next; ml != mlist; ml = ml->next) { 3408 struct magic *ma = ml->magic; 3409 uint32_t nma = ml->nmagic; 3410 for (i = 0; i < nma; i++) { 3411 if (ma[i].type != FILE_NAME) 3412 continue; 3413 if (strcmp(ma[i].value.s, name) == 0) { 3414 v->magic = &ma[i]; 3415 for (j = i + 1; j < nma; j++) 3416 if (ma[j].cont_level == 0) 3417 break; 3418 v->nmagic = j - i; 3419 return 0; 3420 } 3421 } 3422 } 3423 return -1; 3424 } 3425