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 1.233 2015/06/10 00:57:41 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 return ms; 535 free: 536 free(ms); 537 return NULL; 538 } 539 540 private void 541 apprentice_unmap(struct magic_map *map) 542 { 543 if (map == NULL) 544 return; 545 546 switch (map->type) { 547 #ifdef QUICK 548 case MAP_TYPE_MMAP: 549 if (map->p) 550 (void)munmap(map->p, map->len); 551 break; 552 #endif 553 case MAP_TYPE_MALLOC: 554 free(map->p); 555 break; 556 case MAP_TYPE_USER: 557 break; 558 default: 559 abort(); 560 } 561 free(map); 562 } 563 564 private struct mlist * 565 mlist_alloc(void) 566 { 567 struct mlist *mlist; 568 if ((mlist = CAST(struct mlist *, calloc(1, sizeof(*mlist)))) == NULL) { 569 return NULL; 570 } 571 mlist->next = mlist->prev = mlist; 572 return mlist; 573 } 574 575 private void 576 mlist_free(struct mlist *mlist) 577 { 578 struct mlist *ml, *next; 579 580 if (mlist == NULL) 581 return; 582 583 ml = mlist->next; 584 for (ml = mlist->next; (next = ml->next) != NULL; ml = next) { 585 if (ml->map) 586 apprentice_unmap(ml->map); 587 free(ml); 588 if (ml == mlist) 589 break; 590 } 591 } 592 593 #ifndef COMPILE_ONLY 594 /* void **bufs: an array of compiled magic files */ 595 protected int 596 buffer_apprentice(struct magic_set *ms, struct magic **bufs, 597 size_t *sizes, size_t nbufs) 598 { 599 size_t i, j; 600 struct mlist *ml; 601 struct magic_map *map; 602 603 if (nbufs == 0) 604 return -1; 605 606 if (ms->mlist[0] != NULL) 607 file_reset(ms); 608 609 init_file_tables(); 610 611 for (i = 0; i < MAGIC_SETS; i++) { 612 mlist_free(ms->mlist[i]); 613 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 614 file_oomem(ms, sizeof(*ms->mlist[i])); 615 goto fail; 616 } 617 } 618 619 for (i = 0; i < nbufs; i++) { 620 map = apprentice_buf(ms, bufs[i], sizes[i]); 621 if (map == NULL) 622 goto fail; 623 624 for (j = 0; j < MAGIC_SETS; j++) { 625 if (add_mlist(ms->mlist[j], map, j) == -1) { 626 file_oomem(ms, sizeof(*ml)); 627 goto fail; 628 } 629 } 630 } 631 632 return 0; 633 fail: 634 for (i = 0; i < MAGIC_SETS; i++) { 635 mlist_free(ms->mlist[i]); 636 ms->mlist[i] = NULL; 637 } 638 return -1; 639 } 640 #endif 641 642 /* const char *fn: list of magic files and directories */ 643 protected int 644 file_apprentice(struct magic_set *ms, const char *fn, int action) 645 { 646 char *p, *mfn; 647 int file_err, errs = -1; 648 size_t i; 649 650 if (ms->mlist[0] != NULL) 651 file_reset(ms); 652 653 if ((fn = magic_getpath(fn, action)) == NULL) 654 return -1; 655 656 init_file_tables(); 657 658 if ((mfn = strdup(fn)) == NULL) { 659 file_oomem(ms, strlen(fn)); 660 return -1; 661 } 662 663 for (i = 0; i < MAGIC_SETS; i++) { 664 mlist_free(ms->mlist[i]); 665 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 666 file_oomem(ms, sizeof(*ms->mlist[i])); 667 while (i-- > 0) { 668 mlist_free(ms->mlist[i]); 669 ms->mlist[i] = NULL; 670 } 671 free(mfn); 672 return -1; 673 } 674 } 675 fn = mfn; 676 677 while (fn) { 678 p = strchr(fn, PATHSEP); 679 if (p) 680 *p++ = '\0'; 681 if (*fn == '\0') 682 break; 683 file_err = apprentice_1(ms, fn, action); 684 errs = MAX(errs, file_err); 685 fn = p; 686 } 687 688 free(mfn); 689 690 if (errs == -1) { 691 for (i = 0; i < MAGIC_SETS; i++) { 692 mlist_free(ms->mlist[i]); 693 ms->mlist[i] = NULL; 694 } 695 file_error(ms, 0, "could not find any valid magic files!"); 696 return -1; 697 } 698 699 #if 0 700 /* 701 * Always leave the database loaded 702 */ 703 if (action == FILE_LOAD) 704 return 0; 705 706 for (i = 0; i < MAGIC_SETS; i++) { 707 mlist_free(ms->mlist[i]); 708 ms->mlist[i] = NULL; 709 } 710 #endif 711 712 switch (action) { 713 case FILE_LOAD: 714 case FILE_COMPILE: 715 case FILE_CHECK: 716 case FILE_LIST: 717 return 0; 718 default: 719 file_error(ms, 0, "Invalid action %d", action); 720 return -1; 721 } 722 } 723 724 /* 725 * Compute the real length of a magic expression, for the purposes 726 * of determining how "strong" a magic expression is (approximating 727 * how specific its matches are): 728 * - magic characters count 0 unless escaped. 729 * - [] expressions count 1 730 * - {} expressions count 0 731 * - regular characters or escaped magic characters count 1 732 * - 0 length expressions count as one 733 */ 734 private size_t 735 nonmagic(const char *str) 736 { 737 const char *p; 738 size_t rv = 0; 739 740 for (p = str; *p; p++) 741 switch (*p) { 742 case '\\': /* Escaped anything counts 1 */ 743 if (!*++p) 744 p--; 745 rv++; 746 continue; 747 case '?': /* Magic characters count 0 */ 748 case '*': 749 case '.': 750 case '+': 751 case '^': 752 case '$': 753 continue; 754 case '[': /* Bracketed expressions count 1 the ']' */ 755 while (*p && *p != ']') 756 p++; 757 p--; 758 continue; 759 case '{': /* Braced expressions count 0 */ 760 while (*p && *p != '}') 761 p++; 762 if (!*p) 763 p--; 764 continue; 765 default: /* Anything else counts 1 */ 766 rv++; 767 continue; 768 } 769 770 return rv == 0 ? 1 : rv; /* Return at least 1 */ 771 } 772 773 /* 774 * Get weight of this magic entry, for sorting purposes. 775 */ 776 private size_t 777 apprentice_magic_strength(const struct magic *m) 778 { 779 #define MULT 10 780 size_t v, val = 2 * MULT; /* baseline strength */ 781 782 switch (m->type) { 783 case FILE_DEFAULT: /* make sure this sorts last */ 784 if (m->factor_op != FILE_FACTOR_OP_NONE) 785 abort(); 786 return 0; 787 788 case FILE_BYTE: 789 val += 1 * MULT; 790 break; 791 792 case FILE_SHORT: 793 case FILE_LESHORT: 794 case FILE_BESHORT: 795 val += 2 * MULT; 796 break; 797 798 case FILE_LONG: 799 case FILE_LELONG: 800 case FILE_BELONG: 801 case FILE_MELONG: 802 val += 4 * MULT; 803 break; 804 805 case FILE_PSTRING: 806 case FILE_STRING: 807 val += m->vallen * MULT; 808 break; 809 810 case FILE_BESTRING16: 811 case FILE_LESTRING16: 812 val += m->vallen * MULT / 2; 813 break; 814 815 case FILE_SEARCH: 816 val += m->vallen * MAX(MULT / m->vallen, 1); 817 break; 818 819 case FILE_REGEX: 820 v = nonmagic(m->value.s); 821 val += v * MAX(MULT / v, 1); 822 break; 823 824 case FILE_DATE: 825 case FILE_LEDATE: 826 case FILE_BEDATE: 827 case FILE_MEDATE: 828 case FILE_LDATE: 829 case FILE_LELDATE: 830 case FILE_BELDATE: 831 case FILE_MELDATE: 832 case FILE_FLOAT: 833 case FILE_BEFLOAT: 834 case FILE_LEFLOAT: 835 val += 4 * MULT; 836 break; 837 838 case FILE_QUAD: 839 case FILE_BEQUAD: 840 case FILE_LEQUAD: 841 case FILE_QDATE: 842 case FILE_LEQDATE: 843 case FILE_BEQDATE: 844 case FILE_QLDATE: 845 case FILE_LEQLDATE: 846 case FILE_BEQLDATE: 847 case FILE_QWDATE: 848 case FILE_LEQWDATE: 849 case FILE_BEQWDATE: 850 case FILE_DOUBLE: 851 case FILE_BEDOUBLE: 852 case FILE_LEDOUBLE: 853 val += 8 * MULT; 854 break; 855 856 case FILE_INDIRECT: 857 case FILE_NAME: 858 case FILE_USE: 859 break; 860 861 default: 862 (void)fprintf(stderr, "Bad type %d\n", m->type); 863 abort(); 864 } 865 866 switch (m->reln) { 867 case 'x': /* matches anything penalize */ 868 case '!': /* matches almost anything penalize */ 869 val = 0; 870 break; 871 872 case '=': /* Exact match, prefer */ 873 val += MULT; 874 break; 875 876 case '>': 877 case '<': /* comparison match reduce strength */ 878 val -= 2 * MULT; 879 break; 880 881 case '^': 882 case '&': /* masking bits, we could count them too */ 883 val -= MULT; 884 break; 885 886 default: 887 (void)fprintf(stderr, "Bad relation %c\n", m->reln); 888 abort(); 889 } 890 891 if (val == 0) /* ensure we only return 0 for FILE_DEFAULT */ 892 val = 1; 893 894 switch (m->factor_op) { 895 case FILE_FACTOR_OP_NONE: 896 break; 897 case FILE_FACTOR_OP_PLUS: 898 val += m->factor; 899 break; 900 case FILE_FACTOR_OP_MINUS: 901 val -= m->factor; 902 break; 903 case FILE_FACTOR_OP_TIMES: 904 val *= m->factor; 905 break; 906 case FILE_FACTOR_OP_DIV: 907 val /= m->factor; 908 break; 909 default: 910 abort(); 911 } 912 913 /* 914 * Magic entries with no description get a bonus because they depend 915 * on subsequent magic entries to print something. 916 */ 917 if (m->desc[0] == '\0') 918 val++; 919 return val; 920 } 921 922 /* 923 * Sort callback for sorting entries by "strength" (basically length) 924 */ 925 private int 926 apprentice_sort(const void *a, const void *b) 927 { 928 const struct magic_entry *ma = CAST(const struct magic_entry *, a); 929 const struct magic_entry *mb = CAST(const struct magic_entry *, b); 930 size_t sa = apprentice_magic_strength(ma->mp); 931 size_t sb = apprentice_magic_strength(mb->mp); 932 if (sa == sb) 933 return 0; 934 else if (sa > sb) 935 return -1; 936 else 937 return 1; 938 } 939 940 /* 941 * Shows sorted patterns list in the order which is used for the matching 942 */ 943 private void 944 apprentice_list(struct mlist *mlist, int mode) 945 { 946 uint32_t magindex = 0; 947 struct mlist *ml; 948 for (ml = mlist->next; ml != mlist; ml = ml->next) { 949 for (magindex = 0; magindex < ml->nmagic; magindex++) { 950 struct magic *m = &ml->magic[magindex]; 951 if ((m->flag & mode) != mode) { 952 /* Skip sub-tests */ 953 while (magindex + 1 < ml->nmagic && 954 ml->magic[magindex + 1].cont_level != 0) 955 ++magindex; 956 continue; /* Skip to next top-level test*/ 957 } 958 959 /* 960 * Try to iterate over the tree until we find item with 961 * description/mimetype. 962 */ 963 while (magindex + 1 < ml->nmagic && 964 ml->magic[magindex + 1].cont_level != 0 && 965 *ml->magic[magindex].desc == '\0' && 966 *ml->magic[magindex].mimetype == '\0') 967 magindex++; 968 969 printf("Strength = %3" SIZE_T_FORMAT "u@%u: %s [%s]\n", 970 apprentice_magic_strength(m), 971 ml->magic[magindex].lineno, 972 ml->magic[magindex].desc, 973 ml->magic[magindex].mimetype); 974 } 975 } 976 } 977 978 private void 979 set_test_type(struct magic *mstart, struct magic *m) 980 { 981 switch (m->type) { 982 case FILE_BYTE: 983 case FILE_SHORT: 984 case FILE_LONG: 985 case FILE_DATE: 986 case FILE_BESHORT: 987 case FILE_BELONG: 988 case FILE_BEDATE: 989 case FILE_LESHORT: 990 case FILE_LELONG: 991 case FILE_LEDATE: 992 case FILE_LDATE: 993 case FILE_BELDATE: 994 case FILE_LELDATE: 995 case FILE_MEDATE: 996 case FILE_MELDATE: 997 case FILE_MELONG: 998 case FILE_QUAD: 999 case FILE_LEQUAD: 1000 case FILE_BEQUAD: 1001 case FILE_QDATE: 1002 case FILE_LEQDATE: 1003 case FILE_BEQDATE: 1004 case FILE_QLDATE: 1005 case FILE_LEQLDATE: 1006 case FILE_BEQLDATE: 1007 case FILE_QWDATE: 1008 case FILE_LEQWDATE: 1009 case FILE_BEQWDATE: 1010 case FILE_FLOAT: 1011 case FILE_BEFLOAT: 1012 case FILE_LEFLOAT: 1013 case FILE_DOUBLE: 1014 case FILE_BEDOUBLE: 1015 case FILE_LEDOUBLE: 1016 mstart->flag |= BINTEST; 1017 break; 1018 case FILE_STRING: 1019 case FILE_PSTRING: 1020 case FILE_BESTRING16: 1021 case FILE_LESTRING16: 1022 /* Allow text overrides */ 1023 if (mstart->str_flags & STRING_TEXTTEST) 1024 mstart->flag |= TEXTTEST; 1025 else 1026 mstart->flag |= BINTEST; 1027 break; 1028 case FILE_REGEX: 1029 case FILE_SEARCH: 1030 /* Check for override */ 1031 if (mstart->str_flags & STRING_BINTEST) 1032 mstart->flag |= BINTEST; 1033 if (mstart->str_flags & STRING_TEXTTEST) 1034 mstart->flag |= TEXTTEST; 1035 1036 if (mstart->flag & (TEXTTEST|BINTEST)) 1037 break; 1038 1039 /* binary test if pattern is not text */ 1040 if (file_looks_utf8(m->value.us, (size_t)m->vallen, NULL, 1041 NULL) <= 0) 1042 mstart->flag |= BINTEST; 1043 else 1044 mstart->flag |= TEXTTEST; 1045 break; 1046 case FILE_DEFAULT: 1047 /* can't deduce anything; we shouldn't see this at the 1048 top level anyway */ 1049 break; 1050 case FILE_INVALID: 1051 default: 1052 /* invalid search type, but no need to complain here */ 1053 break; 1054 } 1055 } 1056 1057 private int 1058 addentry(struct magic_set *ms, struct magic_entry *me, 1059 struct magic_entry_set *mset) 1060 { 1061 size_t i = me->mp->type == FILE_NAME ? 1 : 0; 1062 if (mset[i].count == mset[i].max) { 1063 struct magic_entry *mp; 1064 1065 mset[i].max += ALLOC_INCR; 1066 if ((mp = CAST(struct magic_entry *, 1067 realloc(mset[i].me, sizeof(*mp) * mset[i].max))) == 1068 NULL) { 1069 file_oomem(ms, sizeof(*mp) * mset[i].max); 1070 return -1; 1071 } 1072 (void)memset(&mp[mset[i].count], 0, sizeof(*mp) * 1073 ALLOC_INCR); 1074 mset[i].me = mp; 1075 } 1076 mset[i].me[mset[i].count++] = *me; 1077 memset(me, 0, sizeof(*me)); 1078 return 0; 1079 } 1080 1081 /* 1082 * Load and parse one file. 1083 */ 1084 private void 1085 load_1(struct magic_set *ms, int action, const char *fn, int *errs, 1086 struct magic_entry_set *mset) 1087 { 1088 size_t lineno = 0, llen = 0; 1089 char *line = NULL; 1090 ssize_t len; 1091 struct magic_entry me; 1092 1093 FILE *f = fopen(ms->file = fn, "r"); 1094 if (f == NULL) { 1095 if (errno != ENOENT) 1096 file_error(ms, errno, "cannot read magic file `%s'", 1097 fn); 1098 (*errs)++; 1099 return; 1100 } 1101 1102 memset(&me, 0, sizeof(me)); 1103 /* read and parse this file */ 1104 for (ms->line = 1; (len = getline(&line, &llen, f)) != -1; 1105 ms->line++) { 1106 if (len == 0) /* null line, garbage, etc */ 1107 continue; 1108 if (line[len - 1] == '\n') { 1109 lineno++; 1110 line[len - 1] = '\0'; /* delete newline */ 1111 } 1112 switch (line[0]) { 1113 case '\0': /* empty, do not parse */ 1114 case '#': /* comment, do not parse */ 1115 continue; 1116 case '!': 1117 if (line[1] == ':') { 1118 size_t i; 1119 1120 for (i = 0; bang[i].name != NULL; i++) { 1121 if ((size_t)(len - 2) > bang[i].len && 1122 memcmp(bang[i].name, line + 2, 1123 bang[i].len) == 0) 1124 break; 1125 } 1126 if (bang[i].name == NULL) { 1127 file_error(ms, 0, 1128 "Unknown !: entry `%s'", line); 1129 (*errs)++; 1130 continue; 1131 } 1132 if (me.mp == NULL) { 1133 file_error(ms, 0, 1134 "No current entry for :!%s type", 1135 bang[i].name); 1136 (*errs)++; 1137 continue; 1138 } 1139 if ((*bang[i].fun)(ms, &me, 1140 line + bang[i].len + 2) != 0) { 1141 (*errs)++; 1142 continue; 1143 } 1144 continue; 1145 } 1146 /*FALLTHROUGH*/ 1147 default: 1148 again: 1149 switch (parse(ms, &me, line, lineno, action)) { 1150 case 0: 1151 continue; 1152 case 1: 1153 (void)addentry(ms, &me, mset); 1154 goto again; 1155 default: 1156 (*errs)++; 1157 break; 1158 } 1159 } 1160 } 1161 if (me.mp) 1162 (void)addentry(ms, &me, mset); 1163 free(line); 1164 (void)fclose(f); 1165 } 1166 1167 /* 1168 * parse a file or directory of files 1169 * const char *fn: name of magic file or directory 1170 */ 1171 private int 1172 cmpstrp(const void *p1, const void *p2) 1173 { 1174 return strcmp(*(char *const *)p1, *(char *const *)p2); 1175 } 1176 1177 1178 private uint32_t 1179 set_text_binary(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1180 uint32_t starttest) 1181 { 1182 static const char text[] = "text"; 1183 static const char binary[] = "binary"; 1184 static const size_t len = sizeof(text); 1185 1186 uint32_t i = starttest; 1187 1188 do { 1189 set_test_type(me[starttest].mp, me[i].mp); 1190 if ((ms->flags & MAGIC_DEBUG) == 0) 1191 continue; 1192 (void)fprintf(stderr, "%s%s%s: %s\n", 1193 me[i].mp->mimetype, 1194 me[i].mp->mimetype[0] == '\0' ? "" : "; ", 1195 me[i].mp->desc[0] ? me[i].mp->desc : "(no description)", 1196 me[i].mp->flag & BINTEST ? binary : text); 1197 if (me[i].mp->flag & BINTEST) { 1198 char *p = strstr(me[i].mp->desc, text); 1199 if (p && (p == me[i].mp->desc || 1200 isspace((unsigned char)p[-1])) && 1201 (p + len - me[i].mp->desc == MAXstring 1202 || (p[len] == '\0' || 1203 isspace((unsigned char)p[len])))) 1204 (void)fprintf(stderr, "*** Possible " 1205 "binary test for text type\n"); 1206 } 1207 } while (++i < nme && me[i].mp->cont_level != 0); 1208 return i; 1209 } 1210 1211 private void 1212 set_last_default(struct magic_set *ms, struct magic_entry *me, uint32_t nme) 1213 { 1214 uint32_t i; 1215 for (i = 0; i < nme; i++) { 1216 if (me[i].mp->cont_level == 0 && 1217 me[i].mp->type == FILE_DEFAULT) { 1218 while (++i < nme) 1219 if (me[i].mp->cont_level == 0) 1220 break; 1221 if (i != nme) { 1222 /* XXX - Ugh! */ 1223 ms->line = me[i].mp->lineno; 1224 file_magwarn(ms, 1225 "level 0 \"default\" did not sort last"); 1226 } 1227 return; 1228 } 1229 } 1230 } 1231 1232 private int 1233 coalesce_entries(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1234 struct magic **ma, uint32_t *nma) 1235 { 1236 uint32_t i, mentrycount = 0; 1237 size_t slen; 1238 1239 for (i = 0; i < nme; i++) 1240 mentrycount += me[i].cont_count; 1241 1242 slen = sizeof(**ma) * mentrycount; 1243 if ((*ma = CAST(struct magic *, malloc(slen))) == NULL) { 1244 file_oomem(ms, slen); 1245 return -1; 1246 } 1247 1248 mentrycount = 0; 1249 for (i = 0; i < nme; i++) { 1250 (void)memcpy(*ma + mentrycount, me[i].mp, 1251 me[i].cont_count * sizeof(**ma)); 1252 mentrycount += me[i].cont_count; 1253 } 1254 *nma = mentrycount; 1255 return 0; 1256 } 1257 1258 private void 1259 magic_entry_free(struct magic_entry *me, uint32_t nme) 1260 { 1261 uint32_t i; 1262 if (me == NULL) 1263 return; 1264 for (i = 0; i < nme; i++) 1265 free(me[i].mp); 1266 free(me); 1267 } 1268 1269 private struct magic_map * 1270 apprentice_load(struct magic_set *ms, const char *fn, int action) 1271 { 1272 int errs = 0; 1273 uint32_t i, j; 1274 size_t files = 0, maxfiles = 0; 1275 char **filearr = NULL, *mfn; 1276 struct stat st; 1277 struct magic_map *map; 1278 struct magic_entry_set mset[MAGIC_SETS]; 1279 DIR *dir; 1280 struct dirent *d; 1281 1282 memset(mset, 0, sizeof(mset)); 1283 ms->flags |= MAGIC_CHECK; /* Enable checks for parsed files */ 1284 1285 1286 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) 1287 { 1288 file_oomem(ms, sizeof(*map)); 1289 return NULL; 1290 } 1291 1292 /* print silly verbose header for USG compat. */ 1293 if (action == FILE_CHECK) 1294 (void)fprintf(stderr, "%s\n", usg_hdr); 1295 1296 /* load directory or file */ 1297 if (stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) { 1298 dir = opendir(fn); 1299 if (!dir) { 1300 errs++; 1301 goto out; 1302 } 1303 while ((d = readdir(dir)) != NULL) { 1304 if (asprintf(&mfn, "%s/%s", fn, d->d_name) < 0) { 1305 file_oomem(ms, 1306 strlen(fn) + strlen(d->d_name) + 2); 1307 errs++; 1308 closedir(dir); 1309 goto out; 1310 } 1311 if (stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) { 1312 free(mfn); 1313 continue; 1314 } 1315 if (files >= maxfiles) { 1316 size_t mlen; 1317 maxfiles = (maxfiles + 1) * 2; 1318 mlen = maxfiles * sizeof(*filearr); 1319 if ((filearr = CAST(char **, 1320 realloc(filearr, mlen))) == NULL) { 1321 file_oomem(ms, mlen); 1322 free(mfn); 1323 closedir(dir); 1324 errs++; 1325 goto out; 1326 } 1327 } 1328 filearr[files++] = mfn; 1329 } 1330 closedir(dir); 1331 qsort(filearr, files, sizeof(*filearr), cmpstrp); 1332 for (i = 0; i < files; i++) { 1333 load_1(ms, action, filearr[i], &errs, mset); 1334 free(filearr[i]); 1335 } 1336 free(filearr); 1337 } else 1338 load_1(ms, action, fn, &errs, mset); 1339 if (errs) 1340 goto out; 1341 1342 for (j = 0; j < MAGIC_SETS; j++) { 1343 /* Set types of tests */ 1344 for (i = 0; i < mset[j].count; ) { 1345 if (mset[j].me[i].mp->cont_level != 0) { 1346 i++; 1347 continue; 1348 } 1349 i = set_text_binary(ms, mset[j].me, mset[j].count, i); 1350 } 1351 qsort(mset[j].me, mset[j].count, sizeof(*mset[j].me), 1352 apprentice_sort); 1353 1354 /* 1355 * Make sure that any level 0 "default" line is last 1356 * (if one exists). 1357 */ 1358 set_last_default(ms, mset[j].me, mset[j].count); 1359 1360 /* coalesce per file arrays into a single one */ 1361 if (coalesce_entries(ms, mset[j].me, mset[j].count, 1362 &map->magic[j], &map->nmagic[j]) == -1) { 1363 errs++; 1364 goto out; 1365 } 1366 } 1367 1368 out: 1369 for (j = 0; j < MAGIC_SETS; j++) 1370 magic_entry_free(mset[j].me, mset[j].count); 1371 1372 if (errs) { 1373 apprentice_unmap(map); 1374 return NULL; 1375 } 1376 return map; 1377 } 1378 1379 /* 1380 * extend the sign bit if the comparison is to be signed 1381 */ 1382 protected uint64_t 1383 file_signextend(struct magic_set *ms, struct magic *m, uint64_t v) 1384 { 1385 if (!(m->flag & UNSIGNED)) { 1386 switch(m->type) { 1387 /* 1388 * Do not remove the casts below. They are 1389 * vital. When later compared with the data, 1390 * the sign extension must have happened. 1391 */ 1392 case FILE_BYTE: 1393 v = (signed char) v; 1394 break; 1395 case FILE_SHORT: 1396 case FILE_BESHORT: 1397 case FILE_LESHORT: 1398 v = (short) v; 1399 break; 1400 case FILE_DATE: 1401 case FILE_BEDATE: 1402 case FILE_LEDATE: 1403 case FILE_MEDATE: 1404 case FILE_LDATE: 1405 case FILE_BELDATE: 1406 case FILE_LELDATE: 1407 case FILE_MELDATE: 1408 case FILE_LONG: 1409 case FILE_BELONG: 1410 case FILE_LELONG: 1411 case FILE_MELONG: 1412 case FILE_FLOAT: 1413 case FILE_BEFLOAT: 1414 case FILE_LEFLOAT: 1415 v = (int32_t) v; 1416 break; 1417 case FILE_QUAD: 1418 case FILE_BEQUAD: 1419 case FILE_LEQUAD: 1420 case FILE_QDATE: 1421 case FILE_QLDATE: 1422 case FILE_QWDATE: 1423 case FILE_BEQDATE: 1424 case FILE_BEQLDATE: 1425 case FILE_BEQWDATE: 1426 case FILE_LEQDATE: 1427 case FILE_LEQLDATE: 1428 case FILE_LEQWDATE: 1429 case FILE_DOUBLE: 1430 case FILE_BEDOUBLE: 1431 case FILE_LEDOUBLE: 1432 v = (int64_t) v; 1433 break; 1434 case FILE_STRING: 1435 case FILE_PSTRING: 1436 case FILE_BESTRING16: 1437 case FILE_LESTRING16: 1438 case FILE_REGEX: 1439 case FILE_SEARCH: 1440 case FILE_DEFAULT: 1441 case FILE_INDIRECT: 1442 case FILE_NAME: 1443 case FILE_USE: 1444 case FILE_CLEAR: 1445 break; 1446 default: 1447 if (ms->flags & MAGIC_CHECK) 1448 file_magwarn(ms, "cannot happen: m->type=%d\n", 1449 m->type); 1450 return ~0U; 1451 } 1452 } 1453 return v; 1454 } 1455 1456 private int 1457 string_modifier_check(struct magic_set *ms, struct magic *m) 1458 { 1459 if ((ms->flags & MAGIC_CHECK) == 0) 1460 return 0; 1461 1462 if ((m->type != FILE_REGEX || (m->str_flags & REGEX_LINE_COUNT) == 0) && 1463 (m->type != FILE_PSTRING && (m->str_flags & PSTRING_LEN) != 0)) { 1464 file_magwarn(ms, 1465 "'/BHhLl' modifiers are only allowed for pascal strings\n"); 1466 return -1; 1467 } 1468 switch (m->type) { 1469 case FILE_BESTRING16: 1470 case FILE_LESTRING16: 1471 if (m->str_flags != 0) { 1472 file_magwarn(ms, 1473 "no modifiers allowed for 16-bit strings\n"); 1474 return -1; 1475 } 1476 break; 1477 case FILE_STRING: 1478 case FILE_PSTRING: 1479 if ((m->str_flags & REGEX_OFFSET_START) != 0) { 1480 file_magwarn(ms, 1481 "'/%c' only allowed on regex and search\n", 1482 CHAR_REGEX_OFFSET_START); 1483 return -1; 1484 } 1485 break; 1486 case FILE_SEARCH: 1487 if (m->str_range == 0) { 1488 file_magwarn(ms, 1489 "missing range; defaulting to %d\n", 1490 STRING_DEFAULT_RANGE); 1491 m->str_range = STRING_DEFAULT_RANGE; 1492 return -1; 1493 } 1494 break; 1495 case FILE_REGEX: 1496 if ((m->str_flags & STRING_COMPACT_WHITESPACE) != 0) { 1497 file_magwarn(ms, "'/%c' not allowed on regex\n", 1498 CHAR_COMPACT_WHITESPACE); 1499 return -1; 1500 } 1501 if ((m->str_flags & STRING_COMPACT_OPTIONAL_WHITESPACE) != 0) { 1502 file_magwarn(ms, "'/%c' not allowed on regex\n", 1503 CHAR_COMPACT_OPTIONAL_WHITESPACE); 1504 return -1; 1505 } 1506 break; 1507 default: 1508 file_magwarn(ms, "coding error: m->type=%d\n", 1509 m->type); 1510 return -1; 1511 } 1512 return 0; 1513 } 1514 1515 private int 1516 get_op(char c) 1517 { 1518 switch (c) { 1519 case '&': 1520 return FILE_OPAND; 1521 case '|': 1522 return FILE_OPOR; 1523 case '^': 1524 return FILE_OPXOR; 1525 case '+': 1526 return FILE_OPADD; 1527 case '-': 1528 return FILE_OPMINUS; 1529 case '*': 1530 return FILE_OPMULTIPLY; 1531 case '/': 1532 return FILE_OPDIVIDE; 1533 case '%': 1534 return FILE_OPMODULO; 1535 default: 1536 return -1; 1537 } 1538 } 1539 1540 #ifdef ENABLE_CONDITIONALS 1541 private int 1542 get_cond(const char *l, const char **t) 1543 { 1544 static const struct cond_tbl_s { 1545 char name[8]; 1546 size_t len; 1547 int cond; 1548 } cond_tbl[] = { 1549 { "if", 2, COND_IF }, 1550 { "elif", 4, COND_ELIF }, 1551 { "else", 4, COND_ELSE }, 1552 { "", 0, COND_NONE }, 1553 }; 1554 const struct cond_tbl_s *p; 1555 1556 for (p = cond_tbl; p->len; p++) { 1557 if (strncmp(l, p->name, p->len) == 0 && 1558 isspace((unsigned char)l[p->len])) { 1559 if (t) 1560 *t = l + p->len; 1561 break; 1562 } 1563 } 1564 return p->cond; 1565 } 1566 1567 private int 1568 check_cond(struct magic_set *ms, int cond, uint32_t cont_level) 1569 { 1570 int last_cond; 1571 last_cond = ms->c.li[cont_level].last_cond; 1572 1573 switch (cond) { 1574 case COND_IF: 1575 if (last_cond != COND_NONE && last_cond != COND_ELIF) { 1576 if (ms->flags & MAGIC_CHECK) 1577 file_magwarn(ms, "syntax error: `if'"); 1578 return -1; 1579 } 1580 last_cond = COND_IF; 1581 break; 1582 1583 case COND_ELIF: 1584 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1585 if (ms->flags & MAGIC_CHECK) 1586 file_magwarn(ms, "syntax error: `elif'"); 1587 return -1; 1588 } 1589 last_cond = COND_ELIF; 1590 break; 1591 1592 case COND_ELSE: 1593 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1594 if (ms->flags & MAGIC_CHECK) 1595 file_magwarn(ms, "syntax error: `else'"); 1596 return -1; 1597 } 1598 last_cond = COND_NONE; 1599 break; 1600 1601 case COND_NONE: 1602 last_cond = COND_NONE; 1603 break; 1604 } 1605 1606 ms->c.li[cont_level].last_cond = last_cond; 1607 return 0; 1608 } 1609 #endif /* ENABLE_CONDITIONALS */ 1610 1611 private int 1612 parse_indirect_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1613 { 1614 const char *l = *lp; 1615 1616 while (!isspace((unsigned char)*++l)) 1617 switch (*l) { 1618 case CHAR_INDIRECT_RELATIVE: 1619 m->str_flags |= INDIRECT_RELATIVE; 1620 break; 1621 default: 1622 if (ms->flags & MAGIC_CHECK) 1623 file_magwarn(ms, "indirect modifier `%c' " 1624 "invalid", *l); 1625 *lp = l; 1626 return -1; 1627 } 1628 *lp = l; 1629 return 0; 1630 } 1631 1632 private void 1633 parse_op_modifier(struct magic_set *ms, struct magic *m, const char **lp, 1634 int op) 1635 { 1636 const char *l = *lp; 1637 char *t; 1638 uint64_t val; 1639 1640 ++l; 1641 m->mask_op |= op; 1642 val = (uint64_t)strtoull(l, &t, 0); 1643 l = t; 1644 m->num_mask = file_signextend(ms, m, val); 1645 eatsize(&l); 1646 *lp = l; 1647 } 1648 1649 private int 1650 parse_string_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1651 { 1652 const char *l = *lp; 1653 char *t; 1654 int have_range = 0; 1655 1656 while (!isspace((unsigned char)*++l)) { 1657 switch (*l) { 1658 case '0': case '1': case '2': 1659 case '3': case '4': case '5': 1660 case '6': case '7': case '8': 1661 case '9': 1662 if (have_range && (ms->flags & MAGIC_CHECK)) 1663 file_magwarn(ms, "multiple ranges"); 1664 have_range = 1; 1665 m->str_range = CAST(uint32_t, strtoul(l, &t, 0)); 1666 if (m->str_range == 0) 1667 file_magwarn(ms, "zero range"); 1668 l = t - 1; 1669 break; 1670 case CHAR_COMPACT_WHITESPACE: 1671 m->str_flags |= STRING_COMPACT_WHITESPACE; 1672 break; 1673 case CHAR_COMPACT_OPTIONAL_WHITESPACE: 1674 m->str_flags |= STRING_COMPACT_OPTIONAL_WHITESPACE; 1675 break; 1676 case CHAR_IGNORE_LOWERCASE: 1677 m->str_flags |= STRING_IGNORE_LOWERCASE; 1678 break; 1679 case CHAR_IGNORE_UPPERCASE: 1680 m->str_flags |= STRING_IGNORE_UPPERCASE; 1681 break; 1682 case CHAR_REGEX_OFFSET_START: 1683 m->str_flags |= REGEX_OFFSET_START; 1684 break; 1685 case CHAR_BINTEST: 1686 m->str_flags |= STRING_BINTEST; 1687 break; 1688 case CHAR_TEXTTEST: 1689 m->str_flags |= STRING_TEXTTEST; 1690 break; 1691 case CHAR_TRIM: 1692 m->str_flags |= STRING_TRIM; 1693 break; 1694 case CHAR_PSTRING_1_LE: 1695 #define SET_LENGTH(a) m->str_flags = (m->str_flags & ~PSTRING_LEN) | (a) 1696 if (m->type != FILE_PSTRING) 1697 goto bad; 1698 SET_LENGTH(PSTRING_1_LE); 1699 break; 1700 case CHAR_PSTRING_2_BE: 1701 if (m->type != FILE_PSTRING) 1702 goto bad; 1703 SET_LENGTH(PSTRING_2_BE); 1704 break; 1705 case CHAR_PSTRING_2_LE: 1706 if (m->type != FILE_PSTRING) 1707 goto bad; 1708 SET_LENGTH(PSTRING_2_LE); 1709 break; 1710 case CHAR_PSTRING_4_BE: 1711 if (m->type != FILE_PSTRING) 1712 goto bad; 1713 SET_LENGTH(PSTRING_4_BE); 1714 break; 1715 case CHAR_PSTRING_4_LE: 1716 switch (m->type) { 1717 case FILE_PSTRING: 1718 case FILE_REGEX: 1719 break; 1720 default: 1721 goto bad; 1722 } 1723 SET_LENGTH(PSTRING_4_LE); 1724 break; 1725 case CHAR_PSTRING_LENGTH_INCLUDES_ITSELF: 1726 if (m->type != FILE_PSTRING) 1727 goto bad; 1728 m->str_flags |= PSTRING_LENGTH_INCLUDES_ITSELF; 1729 break; 1730 default: 1731 bad: 1732 if (ms->flags & MAGIC_CHECK) 1733 file_magwarn(ms, "string modifier `%c' " 1734 "invalid", *l); 1735 goto out; 1736 } 1737 /* allow multiple '/' for readability */ 1738 if (l[1] == '/' && !isspace((unsigned char)l[2])) 1739 l++; 1740 } 1741 if (string_modifier_check(ms, m) == -1) 1742 goto out; 1743 *lp = l; 1744 return 0; 1745 out: 1746 *lp = l; 1747 return -1; 1748 } 1749 1750 /* 1751 * parse one line from magic file, put into magic[index++] if valid 1752 */ 1753 private int 1754 parse(struct magic_set *ms, struct magic_entry *me, const char *line, 1755 size_t lineno, int action) 1756 { 1757 #ifdef ENABLE_CONDITIONALS 1758 static uint32_t last_cont_level = 0; 1759 #endif 1760 size_t i; 1761 struct magic *m; 1762 const char *l = line; 1763 char *t; 1764 int op; 1765 uint32_t cont_level; 1766 int32_t diff; 1767 1768 cont_level = 0; 1769 1770 /* 1771 * Parse the offset. 1772 */ 1773 while (*l == '>') { 1774 ++l; /* step over */ 1775 cont_level++; 1776 } 1777 #ifdef ENABLE_CONDITIONALS 1778 if (cont_level == 0 || cont_level > last_cont_level) 1779 if (file_check_mem(ms, cont_level) == -1) 1780 return -1; 1781 last_cont_level = cont_level; 1782 #endif 1783 if (cont_level != 0) { 1784 if (me->mp == NULL) { 1785 file_magerror(ms, "No current entry for continuation"); 1786 return -1; 1787 } 1788 if (me->cont_count == 0) { 1789 file_magerror(ms, "Continuations present with 0 count"); 1790 return -1; 1791 } 1792 m = &me->mp[me->cont_count - 1]; 1793 diff = (int32_t)cont_level - (int32_t)m->cont_level; 1794 if (diff > 1) 1795 file_magwarn(ms, "New continuation level %u is more " 1796 "than one larger than current level %u", cont_level, 1797 m->cont_level); 1798 if (me->cont_count == me->max_count) { 1799 struct magic *nm; 1800 size_t cnt = me->max_count + ALLOC_CHUNK; 1801 if ((nm = CAST(struct magic *, realloc(me->mp, 1802 sizeof(*nm) * cnt))) == NULL) { 1803 file_oomem(ms, sizeof(*nm) * cnt); 1804 return -1; 1805 } 1806 me->mp = m = nm; 1807 me->max_count = CAST(uint32_t, cnt); 1808 } 1809 m = &me->mp[me->cont_count++]; 1810 (void)memset(m, 0, sizeof(*m)); 1811 m->cont_level = cont_level; 1812 } else { 1813 static const size_t len = sizeof(*m) * ALLOC_CHUNK; 1814 if (me->mp != NULL) 1815 return 1; 1816 if ((m = CAST(struct magic *, malloc(len))) == NULL) { 1817 file_oomem(ms, len); 1818 return -1; 1819 } 1820 me->mp = m; 1821 me->max_count = ALLOC_CHUNK; 1822 (void)memset(m, 0, sizeof(*m)); 1823 m->factor_op = FILE_FACTOR_OP_NONE; 1824 m->cont_level = 0; 1825 me->cont_count = 1; 1826 } 1827 m->lineno = CAST(uint32_t, lineno); 1828 1829 if (*l == '&') { /* m->cont_level == 0 checked below. */ 1830 ++l; /* step over */ 1831 m->flag |= OFFADD; 1832 } 1833 if (*l == '(') { 1834 ++l; /* step over */ 1835 m->flag |= INDIR; 1836 if (m->flag & OFFADD) 1837 m->flag = (m->flag & ~OFFADD) | INDIROFFADD; 1838 1839 if (*l == '&') { /* m->cont_level == 0 checked below */ 1840 ++l; /* step over */ 1841 m->flag |= OFFADD; 1842 } 1843 } 1844 /* Indirect offsets are not valid at level 0. */ 1845 if (m->cont_level == 0 && (m->flag & (OFFADD | INDIROFFADD))) { 1846 if (ms->flags & MAGIC_CHECK) 1847 file_magwarn(ms, "relative offset at level 0"); 1848 return -1; 1849 } 1850 1851 /* get offset, then skip over it */ 1852 m->offset = (uint32_t)strtoul(l, &t, 0); 1853 if (l == t) { 1854 if (ms->flags & MAGIC_CHECK) 1855 file_magwarn(ms, "offset `%s' invalid", l); 1856 return -1; 1857 } 1858 l = t; 1859 1860 if (m->flag & INDIR) { 1861 m->in_type = FILE_LONG; 1862 m->in_offset = 0; 1863 /* 1864 * read [.lbs][+-]nnnnn) 1865 */ 1866 if (*l == '.') { 1867 l++; 1868 switch (*l) { 1869 case 'l': 1870 m->in_type = FILE_LELONG; 1871 break; 1872 case 'L': 1873 m->in_type = FILE_BELONG; 1874 break; 1875 case 'm': 1876 m->in_type = FILE_MELONG; 1877 break; 1878 case 'h': 1879 case 's': 1880 m->in_type = FILE_LESHORT; 1881 break; 1882 case 'H': 1883 case 'S': 1884 m->in_type = FILE_BESHORT; 1885 break; 1886 case 'c': 1887 case 'b': 1888 case 'C': 1889 case 'B': 1890 m->in_type = FILE_BYTE; 1891 break; 1892 case 'e': 1893 case 'f': 1894 case 'g': 1895 m->in_type = FILE_LEDOUBLE; 1896 break; 1897 case 'E': 1898 case 'F': 1899 case 'G': 1900 m->in_type = FILE_BEDOUBLE; 1901 break; 1902 case 'i': 1903 m->in_type = FILE_LEID3; 1904 break; 1905 case 'I': 1906 m->in_type = FILE_BEID3; 1907 break; 1908 default: 1909 if (ms->flags & MAGIC_CHECK) 1910 file_magwarn(ms, 1911 "indirect offset type `%c' invalid", 1912 *l); 1913 return -1; 1914 } 1915 l++; 1916 } 1917 1918 m->in_op = 0; 1919 if (*l == '~') { 1920 m->in_op |= FILE_OPINVERSE; 1921 l++; 1922 } 1923 if ((op = get_op(*l)) != -1) { 1924 m->in_op |= op; 1925 l++; 1926 } 1927 if (*l == '(') { 1928 m->in_op |= FILE_OPINDIRECT; 1929 l++; 1930 } 1931 if (isdigit((unsigned char)*l) || *l == '-') { 1932 m->in_offset = (int32_t)strtol(l, &t, 0); 1933 if (l == t) { 1934 if (ms->flags & MAGIC_CHECK) 1935 file_magwarn(ms, 1936 "in_offset `%s' invalid", l); 1937 return -1; 1938 } 1939 l = t; 1940 } 1941 if (*l++ != ')' || 1942 ((m->in_op & FILE_OPINDIRECT) && *l++ != ')')) { 1943 if (ms->flags & MAGIC_CHECK) 1944 file_magwarn(ms, 1945 "missing ')' in indirect offset"); 1946 return -1; 1947 } 1948 } 1949 EATAB; 1950 1951 #ifdef ENABLE_CONDITIONALS 1952 m->cond = get_cond(l, &l); 1953 if (check_cond(ms, m->cond, cont_level) == -1) 1954 return -1; 1955 1956 EATAB; 1957 #endif 1958 1959 /* 1960 * Parse the type. 1961 */ 1962 if (*l == 'u') { 1963 /* 1964 * Try it as a keyword type prefixed by "u"; match what 1965 * follows the "u". If that fails, try it as an SUS 1966 * integer type. 1967 */ 1968 m->type = get_type(type_tbl, l + 1, &l); 1969 if (m->type == FILE_INVALID) { 1970 /* 1971 * Not a keyword type; parse it as an SUS type, 1972 * 'u' possibly followed by a number or C/S/L. 1973 */ 1974 m->type = get_standard_integer_type(l, &l); 1975 } 1976 /* It's unsigned. */ 1977 if (m->type != FILE_INVALID) 1978 m->flag |= UNSIGNED; 1979 } else { 1980 /* 1981 * Try it as a keyword type. If that fails, try it as 1982 * an SUS integer type if it begins with "d" or as an 1983 * SUS string type if it begins with "s". In any case, 1984 * it's not unsigned. 1985 */ 1986 m->type = get_type(type_tbl, l, &l); 1987 if (m->type == FILE_INVALID) { 1988 /* 1989 * Not a keyword type; parse it as an SUS type, 1990 * either 'd' possibly followed by a number or 1991 * C/S/L, or just 's'. 1992 */ 1993 if (*l == 'd') 1994 m->type = get_standard_integer_type(l, &l); 1995 else if (*l == 's' && !isalpha((unsigned char)l[1])) { 1996 m->type = FILE_STRING; 1997 ++l; 1998 } 1999 } 2000 } 2001 2002 if (m->type == FILE_INVALID) { 2003 /* Not found - try it as a special keyword. */ 2004 m->type = get_type(special_tbl, l, &l); 2005 } 2006 2007 if (m->type == FILE_INVALID) { 2008 if (ms->flags & MAGIC_CHECK) 2009 file_magwarn(ms, "type `%s' invalid", l); 2010 return -1; 2011 } 2012 2013 /* New-style anding: "0 byte&0x80 =0x80 dynamically linked" */ 2014 /* New and improved: ~ & | ^ + - * / % -- exciting, isn't it? */ 2015 2016 m->mask_op = 0; 2017 if (*l == '~') { 2018 if (!IS_STRING(m->type)) 2019 m->mask_op |= FILE_OPINVERSE; 2020 else if (ms->flags & MAGIC_CHECK) 2021 file_magwarn(ms, "'~' invalid for string types"); 2022 ++l; 2023 } 2024 m->str_range = 0; 2025 m->str_flags = m->type == FILE_PSTRING ? PSTRING_1_LE : 0; 2026 if ((op = get_op(*l)) != -1) { 2027 if (IS_STRING(m->type)) { 2028 int r; 2029 2030 if (op != FILE_OPDIVIDE) { 2031 if (ms->flags & MAGIC_CHECK) 2032 file_magwarn(ms, 2033 "invalid string/indirect op: " 2034 "`%c'", *t); 2035 return -1; 2036 } 2037 2038 if (m->type == FILE_INDIRECT) 2039 r = parse_indirect_modifier(ms, m, &l); 2040 else 2041 r = parse_string_modifier(ms, m, &l); 2042 if (r == -1) 2043 return -1; 2044 } else 2045 parse_op_modifier(ms, m, &l, op); 2046 } 2047 2048 /* 2049 * We used to set mask to all 1's here, instead let's just not do 2050 * anything if mask = 0 (unless you have a better idea) 2051 */ 2052 EATAB; 2053 2054 switch (*l) { 2055 case '>': 2056 case '<': 2057 m->reln = *l; 2058 ++l; 2059 if (*l == '=') { 2060 if (ms->flags & MAGIC_CHECK) { 2061 file_magwarn(ms, "%c= not supported", 2062 m->reln); 2063 return -1; 2064 } 2065 ++l; 2066 } 2067 break; 2068 /* Old-style anding: "0 byte &0x80 dynamically linked" */ 2069 case '&': 2070 case '^': 2071 case '=': 2072 m->reln = *l; 2073 ++l; 2074 if (*l == '=') { 2075 /* HP compat: ignore &= etc. */ 2076 ++l; 2077 } 2078 break; 2079 case '!': 2080 m->reln = *l; 2081 ++l; 2082 break; 2083 default: 2084 m->reln = '='; /* the default relation */ 2085 if (*l == 'x' && ((isascii((unsigned char)l[1]) && 2086 isspace((unsigned char)l[1])) || !l[1])) { 2087 m->reln = *l; 2088 ++l; 2089 } 2090 break; 2091 } 2092 /* 2093 * Grab the value part, except for an 'x' reln. 2094 */ 2095 if (m->reln != 'x' && getvalue(ms, m, &l, action)) 2096 return -1; 2097 2098 /* 2099 * TODO finish this macro and start using it! 2100 * #define offsetcheck {if (offset > HOWMANY-1) 2101 * magwarn("offset too big"); } 2102 */ 2103 2104 /* 2105 * Now get last part - the description 2106 */ 2107 EATAB; 2108 if (l[0] == '\b') { 2109 ++l; 2110 m->flag |= NOSPACE; 2111 } else if ((l[0] == '\\') && (l[1] == 'b')) { 2112 ++l; 2113 ++l; 2114 m->flag |= NOSPACE; 2115 } 2116 for (i = 0; (m->desc[i++] = *l++) != '\0' && i < sizeof(m->desc); ) 2117 continue; 2118 if (i == sizeof(m->desc)) { 2119 m->desc[sizeof(m->desc) - 1] = '\0'; 2120 if (ms->flags & MAGIC_CHECK) 2121 file_magwarn(ms, "description `%s' truncated", m->desc); 2122 } 2123 2124 /* 2125 * We only do this check while compiling, or if any of the magic 2126 * files were not compiled. 2127 */ 2128 if (ms->flags & MAGIC_CHECK) { 2129 if (check_format(ms, m) == -1) 2130 return -1; 2131 } 2132 #ifndef COMPILE_ONLY 2133 if (action == FILE_CHECK) { 2134 file_mdump(m); 2135 } 2136 #endif 2137 m->mimetype[0] = '\0'; /* initialise MIME type to none */ 2138 return 0; 2139 } 2140 2141 /* 2142 * parse a STRENGTH annotation line from magic file, put into magic[index - 1] 2143 * if valid 2144 */ 2145 private int 2146 parse_strength(struct magic_set *ms, struct magic_entry *me, const char *line) 2147 { 2148 const char *l = line; 2149 char *el; 2150 unsigned long factor; 2151 struct magic *m = &me->mp[0]; 2152 2153 if (m->factor_op != FILE_FACTOR_OP_NONE) { 2154 file_magwarn(ms, 2155 "Current entry already has a strength type: %c %d", 2156 m->factor_op, m->factor); 2157 return -1; 2158 } 2159 if (m->type == FILE_NAME) { 2160 file_magwarn(ms, "%s: Strength setting is not supported in " 2161 "\"name\" magic entries", m->value.s); 2162 return -1; 2163 } 2164 EATAB; 2165 switch (*l) { 2166 case FILE_FACTOR_OP_NONE: 2167 case FILE_FACTOR_OP_PLUS: 2168 case FILE_FACTOR_OP_MINUS: 2169 case FILE_FACTOR_OP_TIMES: 2170 case FILE_FACTOR_OP_DIV: 2171 m->factor_op = *l++; 2172 break; 2173 default: 2174 file_magwarn(ms, "Unknown factor op `%c'", *l); 2175 return -1; 2176 } 2177 EATAB; 2178 factor = strtoul(l, &el, 0); 2179 if (factor > 255) { 2180 file_magwarn(ms, "Too large factor `%lu'", factor); 2181 goto out; 2182 } 2183 if (*el && !isspace((unsigned char)*el)) { 2184 file_magwarn(ms, "Bad factor `%s'", l); 2185 goto out; 2186 } 2187 m->factor = (uint8_t)factor; 2188 if (m->factor == 0 && m->factor_op == FILE_FACTOR_OP_DIV) { 2189 file_magwarn(ms, "Cannot have factor op `%c' and factor %u", 2190 m->factor_op, m->factor); 2191 goto out; 2192 } 2193 return 0; 2194 out: 2195 m->factor_op = FILE_FACTOR_OP_NONE; 2196 m->factor = 0; 2197 return -1; 2198 } 2199 2200 private int 2201 goodchar(unsigned char x, const char *extra) 2202 { 2203 return (isascii(x) && isalnum(x)) || strchr(extra, x); 2204 } 2205 2206 private int 2207 parse_extra(struct magic_set *ms, struct magic_entry *me, const char *line, 2208 off_t off, size_t len, const char *name, const char *extra, int nt) 2209 { 2210 size_t i; 2211 const char *l = line; 2212 struct magic *m = &me->mp[me->cont_count == 0 ? 0 : me->cont_count - 1]; 2213 char *buf = CAST(char *, CAST(void *, m)) + off; 2214 2215 if (buf[0] != '\0') { 2216 len = nt ? strlen(buf) : len; 2217 file_magwarn(ms, "Current entry already has a %s type " 2218 "`%.*s', new type `%s'", name, (int)len, buf, l); 2219 return -1; 2220 } 2221 2222 if (*m->desc == '\0') { 2223 file_magwarn(ms, "Current entry does not yet have a " 2224 "description for adding a %s type", name); 2225 return -1; 2226 } 2227 2228 EATAB; 2229 for (i = 0; *l && i < len && goodchar(*l, extra); buf[i++] = *l++) 2230 continue; 2231 2232 if (i == len && *l) { 2233 if (nt) 2234 buf[len - 1] = '\0'; 2235 if (ms->flags & MAGIC_CHECK) 2236 file_magwarn(ms, "%s type `%s' truncated %" 2237 SIZE_T_FORMAT "u", name, line, i); 2238 } else { 2239 if (!isspace((unsigned char)*l) && !goodchar(*l, extra)) 2240 file_magwarn(ms, "%s type `%s' has bad char '%c'", 2241 name, line, *l); 2242 if (nt) 2243 buf[i] = '\0'; 2244 } 2245 2246 if (i > 0) 2247 return 0; 2248 2249 file_magerror(ms, "Bad magic entry '%s'", line); 2250 return -1; 2251 } 2252 2253 /* 2254 * Parse an Apple CREATOR/TYPE annotation from magic file and put it into 2255 * magic[index - 1] 2256 */ 2257 private int 2258 parse_apple(struct magic_set *ms, struct magic_entry *me, const char *line) 2259 { 2260 struct magic *m = &me->mp[0]; 2261 2262 return parse_extra(ms, me, line, 2263 CAST(off_t, offsetof(struct magic, apple)), 2264 sizeof(m->apple), "APPLE", "!+-./", 0); 2265 } 2266 2267 /* 2268 * Parse a comma-separated list of extensions 2269 */ 2270 private int 2271 parse_ext(struct magic_set *ms, struct magic_entry *me, const char *line) 2272 { 2273 struct magic *m = &me->mp[0]; 2274 2275 return parse_extra(ms, me, line, 2276 CAST(off_t, offsetof(struct magic, ext)), 2277 sizeof(m->ext), "EXTENSION", ",!+-/", 0); 2278 } 2279 2280 /* 2281 * parse a MIME annotation line from magic file, put into magic[index - 1] 2282 * if valid 2283 */ 2284 private int 2285 parse_mime(struct magic_set *ms, struct magic_entry *me, const char *line) 2286 { 2287 struct magic *m = &me->mp[0]; 2288 2289 return parse_extra(ms, me, line, 2290 CAST(off_t, offsetof(struct magic, mimetype)), 2291 sizeof(m->mimetype), "MIME", "+-/.", 1); 2292 } 2293 2294 private int 2295 check_format_type(const char *ptr, int type) 2296 { 2297 int quad = 0, h; 2298 if (*ptr == '\0') { 2299 /* Missing format string; bad */ 2300 return -1; 2301 } 2302 2303 switch (file_formats[type]) { 2304 case FILE_FMT_QUAD: 2305 quad = 1; 2306 /*FALLTHROUGH*/ 2307 case FILE_FMT_NUM: 2308 if (quad == 0) { 2309 switch (type) { 2310 case FILE_BYTE: 2311 h = 2; 2312 break; 2313 case FILE_SHORT: 2314 case FILE_BESHORT: 2315 case FILE_LESHORT: 2316 h = 1; 2317 break; 2318 case FILE_LONG: 2319 case FILE_BELONG: 2320 case FILE_LELONG: 2321 case FILE_MELONG: 2322 case FILE_LEID3: 2323 case FILE_BEID3: 2324 case FILE_INDIRECT: 2325 h = 0; 2326 break; 2327 default: 2328 abort(); 2329 } 2330 } else 2331 h = 0; 2332 if (*ptr == '-') 2333 ptr++; 2334 if (*ptr == '.') 2335 ptr++; 2336 while (isdigit((unsigned char)*ptr)) ptr++; 2337 if (*ptr == '.') 2338 ptr++; 2339 while (isdigit((unsigned char)*ptr)) ptr++; 2340 if (quad) { 2341 if (*ptr++ != 'l') 2342 return -1; 2343 if (*ptr++ != 'l') 2344 return -1; 2345 } 2346 2347 switch (*ptr++) { 2348 #ifdef STRICT_FORMAT /* "long" formats are int formats for us */ 2349 /* so don't accept the 'l' modifier */ 2350 case 'l': 2351 switch (*ptr++) { 2352 case 'i': 2353 case 'd': 2354 case 'u': 2355 case 'o': 2356 case 'x': 2357 case 'X': 2358 return h != 0 ? -1 : 0; 2359 default: 2360 return -1; 2361 } 2362 2363 /* 2364 * Don't accept h and hh modifiers. They make writing 2365 * magic entries more complicated, for very little benefit 2366 */ 2367 case 'h': 2368 if (h-- <= 0) 2369 return -1; 2370 switch (*ptr++) { 2371 case 'h': 2372 if (h-- <= 0) 2373 return -1; 2374 switch (*ptr++) { 2375 case 'i': 2376 case 'd': 2377 case 'u': 2378 case 'o': 2379 case 'x': 2380 case 'X': 2381 return 0; 2382 default: 2383 return -1; 2384 } 2385 case 'i': 2386 case 'd': 2387 case 'u': 2388 case 'o': 2389 case 'x': 2390 case 'X': 2391 return h != 0 ? -1 : 0; 2392 default: 2393 return -1; 2394 } 2395 #endif 2396 case 'c': 2397 return h != 2 ? -1 : 0; 2398 case 'i': 2399 case 'd': 2400 case 'u': 2401 case 'o': 2402 case 'x': 2403 case 'X': 2404 #ifdef STRICT_FORMAT 2405 return h != 0 ? -1 : 0; 2406 #else 2407 return 0; 2408 #endif 2409 default: 2410 return -1; 2411 } 2412 2413 case FILE_FMT_FLOAT: 2414 case FILE_FMT_DOUBLE: 2415 if (*ptr == '-') 2416 ptr++; 2417 if (*ptr == '.') 2418 ptr++; 2419 while (isdigit((unsigned char)*ptr)) ptr++; 2420 if (*ptr == '.') 2421 ptr++; 2422 while (isdigit((unsigned char)*ptr)) ptr++; 2423 2424 switch (*ptr++) { 2425 case 'e': 2426 case 'E': 2427 case 'f': 2428 case 'F': 2429 case 'g': 2430 case 'G': 2431 return 0; 2432 2433 default: 2434 return -1; 2435 } 2436 2437 2438 case FILE_FMT_STR: 2439 if (*ptr == '-') 2440 ptr++; 2441 while (isdigit((unsigned char )*ptr)) 2442 ptr++; 2443 if (*ptr == '.') { 2444 ptr++; 2445 while (isdigit((unsigned char )*ptr)) 2446 ptr++; 2447 } 2448 2449 switch (*ptr++) { 2450 case 's': 2451 return 0; 2452 default: 2453 return -1; 2454 } 2455 2456 default: 2457 /* internal error */ 2458 abort(); 2459 } 2460 /*NOTREACHED*/ 2461 return -1; 2462 } 2463 2464 /* 2465 * Check that the optional printf format in description matches 2466 * the type of the magic. 2467 */ 2468 private int 2469 check_format(struct magic_set *ms, struct magic *m) 2470 { 2471 char *ptr; 2472 2473 for (ptr = m->desc; *ptr; ptr++) 2474 if (*ptr == '%') 2475 break; 2476 if (*ptr == '\0') { 2477 /* No format string; ok */ 2478 return 1; 2479 } 2480 2481 assert(file_nformats == file_nnames); 2482 2483 if (m->type >= file_nformats) { 2484 file_magwarn(ms, "Internal error inconsistency between " 2485 "m->type and format strings"); 2486 return -1; 2487 } 2488 if (file_formats[m->type] == FILE_FMT_NONE) { 2489 file_magwarn(ms, "No format string for `%s' with description " 2490 "`%s'", m->desc, file_names[m->type]); 2491 return -1; 2492 } 2493 2494 ptr++; 2495 if (check_format_type(ptr, m->type) == -1) { 2496 /* 2497 * TODO: this error message is unhelpful if the format 2498 * string is not one character long 2499 */ 2500 file_magwarn(ms, "Printf format `%c' is not valid for type " 2501 "`%s' in description `%s'", *ptr ? *ptr : '?', 2502 file_names[m->type], m->desc); 2503 return -1; 2504 } 2505 2506 for (; *ptr; ptr++) { 2507 if (*ptr == '%') { 2508 file_magwarn(ms, 2509 "Too many format strings (should have at most one) " 2510 "for `%s' with description `%s'", 2511 file_names[m->type], m->desc); 2512 return -1; 2513 } 2514 } 2515 return 0; 2516 } 2517 2518 /* 2519 * Read a numeric value from a pointer, into the value union of a magic 2520 * pointer, according to the magic type. Update the string pointer to point 2521 * just after the number read. Return 0 for success, non-zero for failure. 2522 */ 2523 private int 2524 getvalue(struct magic_set *ms, struct magic *m, const char **p, int action) 2525 { 2526 switch (m->type) { 2527 case FILE_BESTRING16: 2528 case FILE_LESTRING16: 2529 case FILE_STRING: 2530 case FILE_PSTRING: 2531 case FILE_REGEX: 2532 case FILE_SEARCH: 2533 case FILE_NAME: 2534 case FILE_USE: 2535 *p = getstr(ms, m, *p, action == FILE_COMPILE); 2536 if (*p == NULL) { 2537 if (ms->flags & MAGIC_CHECK) 2538 file_magwarn(ms, "cannot get string from `%s'", 2539 m->value.s); 2540 return -1; 2541 } 2542 if (m->type == FILE_REGEX) { 2543 file_regex_t rx; 2544 int rc = file_regcomp(&rx, m->value.s, REG_EXTENDED); 2545 if (rc) { 2546 if (ms->flags & MAGIC_CHECK) 2547 file_regerror(&rx, rc, ms); 2548 } 2549 file_regfree(&rx); 2550 return rc ? -1 : 0; 2551 } 2552 return 0; 2553 case FILE_FLOAT: 2554 case FILE_BEFLOAT: 2555 case FILE_LEFLOAT: 2556 if (m->reln != 'x') { 2557 char *ep; 2558 #ifdef HAVE_STRTOF 2559 m->value.f = strtof(*p, &ep); 2560 #else 2561 m->value.f = (float)strtod(*p, &ep); 2562 #endif 2563 *p = ep; 2564 } 2565 return 0; 2566 case FILE_DOUBLE: 2567 case FILE_BEDOUBLE: 2568 case FILE_LEDOUBLE: 2569 if (m->reln != 'x') { 2570 char *ep; 2571 m->value.d = strtod(*p, &ep); 2572 *p = ep; 2573 } 2574 return 0; 2575 default: 2576 if (m->reln != 'x') { 2577 char *ep; 2578 m->value.q = file_signextend(ms, m, 2579 (uint64_t)strtoull(*p, &ep, 0)); 2580 *p = ep; 2581 eatsize(p); 2582 } 2583 return 0; 2584 } 2585 } 2586 2587 /* 2588 * Convert a string containing C character escapes. Stop at an unescaped 2589 * space or tab. 2590 * Copy the converted version to "m->value.s", and the length in m->vallen. 2591 * Return updated scan pointer as function result. Warn if set. 2592 */ 2593 private const char * 2594 getstr(struct magic_set *ms, struct magic *m, const char *s, int warn) 2595 { 2596 const char *origs = s; 2597 char *p = m->value.s; 2598 size_t plen = sizeof(m->value.s); 2599 char *origp = p; 2600 char *pmax = p + plen - 1; 2601 int c; 2602 int val; 2603 2604 while ((c = *s++) != '\0') { 2605 if (isspace((unsigned char) c)) 2606 break; 2607 if (p >= pmax) { 2608 file_error(ms, 0, "string too long: `%s'", origs); 2609 return NULL; 2610 } 2611 if (c == '\\') { 2612 switch(c = *s++) { 2613 2614 case '\0': 2615 if (warn) 2616 file_magwarn(ms, "incomplete escape"); 2617 goto out; 2618 2619 case '\t': 2620 if (warn) { 2621 file_magwarn(ms, 2622 "escaped tab found, use \\t instead"); 2623 warn = 0; /* already did */ 2624 } 2625 /*FALLTHROUGH*/ 2626 default: 2627 if (warn) { 2628 if (isprint((unsigned char)c)) { 2629 /* Allow escaping of 2630 * ``relations'' */ 2631 if (strchr("<>&^=!", c) == NULL 2632 && (m->type != FILE_REGEX || 2633 strchr("[]().*?^$|{}", c) 2634 == NULL)) { 2635 file_magwarn(ms, "no " 2636 "need to escape " 2637 "`%c'", c); 2638 } 2639 } else { 2640 file_magwarn(ms, 2641 "unknown escape sequence: " 2642 "\\%03o", c); 2643 } 2644 } 2645 /*FALLTHROUGH*/ 2646 /* space, perhaps force people to use \040? */ 2647 case ' ': 2648 #if 0 2649 /* 2650 * Other things people escape, but shouldn't need to, 2651 * so we disallow them 2652 */ 2653 case '\'': 2654 case '"': 2655 case '?': 2656 #endif 2657 /* Relations */ 2658 case '>': 2659 case '<': 2660 case '&': 2661 case '^': 2662 case '=': 2663 case '!': 2664 /* and baskslash itself */ 2665 case '\\': 2666 *p++ = (char) c; 2667 break; 2668 2669 case 'a': 2670 *p++ = '\a'; 2671 break; 2672 2673 case 'b': 2674 *p++ = '\b'; 2675 break; 2676 2677 case 'f': 2678 *p++ = '\f'; 2679 break; 2680 2681 case 'n': 2682 *p++ = '\n'; 2683 break; 2684 2685 case 'r': 2686 *p++ = '\r'; 2687 break; 2688 2689 case 't': 2690 *p++ = '\t'; 2691 break; 2692 2693 case 'v': 2694 *p++ = '\v'; 2695 break; 2696 2697 /* \ and up to 3 octal digits */ 2698 case '0': 2699 case '1': 2700 case '2': 2701 case '3': 2702 case '4': 2703 case '5': 2704 case '6': 2705 case '7': 2706 val = c - '0'; 2707 c = *s++; /* try for 2 */ 2708 if (c >= '0' && c <= '7') { 2709 val = (val << 3) | (c - '0'); 2710 c = *s++; /* try for 3 */ 2711 if (c >= '0' && c <= '7') 2712 val = (val << 3) | (c-'0'); 2713 else 2714 --s; 2715 } 2716 else 2717 --s; 2718 *p++ = (char)val; 2719 break; 2720 2721 /* \x and up to 2 hex digits */ 2722 case 'x': 2723 val = 'x'; /* Default if no digits */ 2724 c = hextoint(*s++); /* Get next char */ 2725 if (c >= 0) { 2726 val = c; 2727 c = hextoint(*s++); 2728 if (c >= 0) 2729 val = (val << 4) + c; 2730 else 2731 --s; 2732 } else 2733 --s; 2734 *p++ = (char)val; 2735 break; 2736 } 2737 } else 2738 *p++ = (char)c; 2739 } 2740 out: 2741 *p = '\0'; 2742 m->vallen = CAST(unsigned char, (p - origp)); 2743 if (m->type == FILE_PSTRING) 2744 m->vallen += (unsigned char)file_pstring_length_size(m); 2745 return s; 2746 } 2747 2748 2749 /* Single hex char to int; -1 if not a hex char. */ 2750 private int 2751 hextoint(int c) 2752 { 2753 if (!isascii((unsigned char) c)) 2754 return -1; 2755 if (isdigit((unsigned char) c)) 2756 return c - '0'; 2757 if ((c >= 'a') && (c <= 'f')) 2758 return c + 10 - 'a'; 2759 if (( c>= 'A') && (c <= 'F')) 2760 return c + 10 - 'A'; 2761 return -1; 2762 } 2763 2764 2765 /* 2766 * Print a string containing C character escapes. 2767 */ 2768 protected void 2769 file_showstr(FILE *fp, const char *s, size_t len) 2770 { 2771 char c; 2772 2773 for (;;) { 2774 if (len == ~0U) { 2775 c = *s++; 2776 if (c == '\0') 2777 break; 2778 } 2779 else { 2780 if (len-- == 0) 2781 break; 2782 c = *s++; 2783 } 2784 if (c >= 040 && c <= 0176) /* TODO isprint && !iscntrl */ 2785 (void) fputc(c, fp); 2786 else { 2787 (void) fputc('\\', fp); 2788 switch (c) { 2789 case '\a': 2790 (void) fputc('a', fp); 2791 break; 2792 2793 case '\b': 2794 (void) fputc('b', fp); 2795 break; 2796 2797 case '\f': 2798 (void) fputc('f', fp); 2799 break; 2800 2801 case '\n': 2802 (void) fputc('n', fp); 2803 break; 2804 2805 case '\r': 2806 (void) fputc('r', fp); 2807 break; 2808 2809 case '\t': 2810 (void) fputc('t', fp); 2811 break; 2812 2813 case '\v': 2814 (void) fputc('v', fp); 2815 break; 2816 2817 default: 2818 (void) fprintf(fp, "%.3o", c & 0377); 2819 break; 2820 } 2821 } 2822 } 2823 } 2824 2825 /* 2826 * eatsize(): Eat the size spec from a number [eg. 10UL] 2827 */ 2828 private void 2829 eatsize(const char **p) 2830 { 2831 const char *l = *p; 2832 2833 if (LOWCASE(*l) == 'u') 2834 l++; 2835 2836 switch (LOWCASE(*l)) { 2837 case 'l': /* long */ 2838 case 's': /* short */ 2839 case 'h': /* short */ 2840 case 'b': /* char/byte */ 2841 case 'c': /* char/byte */ 2842 l++; 2843 /*FALLTHROUGH*/ 2844 default: 2845 break; 2846 } 2847 2848 *p = l; 2849 } 2850 2851 /* 2852 * handle a buffer containing a compiled file. 2853 */ 2854 private struct magic_map * 2855 apprentice_buf(struct magic_set *ms, struct magic *buf, size_t len) 2856 { 2857 struct magic_map *map; 2858 2859 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 2860 file_oomem(ms, sizeof(*map)); 2861 return NULL; 2862 } 2863 map->len = len; 2864 map->p = buf; 2865 map->type = MAP_TYPE_USER; 2866 if (check_buffer(ms, map, "buffer") != 0) { 2867 apprentice_unmap(map); 2868 return NULL; 2869 } 2870 return map; 2871 } 2872 2873 /* 2874 * handle a compiled file. 2875 */ 2876 2877 private struct magic_map * 2878 apprentice_map(struct magic_set *ms, const char *fn) 2879 { 2880 int fd; 2881 struct stat st; 2882 char *dbname = NULL; 2883 struct magic_map *map; 2884 2885 fd = -1; 2886 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 2887 file_oomem(ms, sizeof(*map)); 2888 goto error; 2889 } 2890 2891 dbname = mkdbname(ms, fn, 0); 2892 if (dbname == NULL) 2893 goto error; 2894 2895 if ((fd = open(dbname, O_RDONLY|O_BINARY)) == -1) 2896 goto error; 2897 2898 if (fstat(fd, &st) == -1) { 2899 file_error(ms, errno, "cannot stat `%s'", dbname); 2900 goto error; 2901 } 2902 if (st.st_size < 8 || st.st_size > MAXMAGIC_SIZE) { 2903 file_error(ms, 0, "file `%s' is too %s", dbname, 2904 st.st_size < 8 ? "small" : "large"); 2905 goto error; 2906 } 2907 2908 map->len = (size_t)st.st_size; 2909 #ifdef QUICK 2910 if ((map->p = mmap(0, (size_t)st.st_size, PROT_READ|PROT_WRITE, 2911 MAP_PRIVATE|MAP_FILE, fd, (off_t)0)) == MAP_FAILED) { 2912 file_error(ms, errno, "cannot map `%s'", dbname); 2913 goto error; 2914 } 2915 map->type = MAP_TYPE_MMAP; 2916 #else 2917 if ((map->p = CAST(void *, malloc(map->len))) == NULL) { 2918 file_oomem(ms, map->len); 2919 goto error; 2920 } 2921 if (read(fd, map->p, map->len) != (ssize_t)map->len) { 2922 file_badread(ms); 2923 goto error; 2924 } 2925 map->type = MAP_TYPE_MALLOC; 2926 #define RET 1 2927 #endif 2928 (void)close(fd); 2929 fd = -1; 2930 2931 if (check_buffer(ms, map, dbname) != 0) 2932 goto error; 2933 2934 free(dbname); 2935 return map; 2936 2937 error: 2938 if (fd != -1) 2939 (void)close(fd); 2940 apprentice_unmap(map); 2941 free(dbname); 2942 return NULL; 2943 } 2944 2945 private int 2946 check_buffer(struct magic_set *ms, struct magic_map *map, const char *dbname) 2947 { 2948 uint32_t *ptr; 2949 uint32_t entries, nentries; 2950 uint32_t version; 2951 int i, needsbyteswap; 2952 2953 ptr = CAST(uint32_t *, map->p); 2954 if (*ptr != MAGICNO) { 2955 if (swap4(*ptr) != MAGICNO) { 2956 file_error(ms, 0, "bad magic in `%s'", dbname); 2957 return -1; 2958 } 2959 needsbyteswap = 1; 2960 } else 2961 needsbyteswap = 0; 2962 if (needsbyteswap) 2963 version = swap4(ptr[1]); 2964 else 2965 version = ptr[1]; 2966 if (version != VERSIONNO) { 2967 file_error(ms, 0, "File %s supports only version %d magic " 2968 "files. `%s' is version %d", VERSION, 2969 VERSIONNO, dbname, version); 2970 return -1; 2971 } 2972 entries = (uint32_t)(map->len / sizeof(struct magic)); 2973 if ((entries * sizeof(struct magic)) != map->len) { 2974 file_error(ms, 0, "Size of `%s' %" SIZE_T_FORMAT "u is not " 2975 "a multiple of %" SIZE_T_FORMAT "u", 2976 dbname, map->len, sizeof(struct magic)); 2977 return -1; 2978 } 2979 map->magic[0] = CAST(struct magic *, map->p) + 1; 2980 nentries = 0; 2981 for (i = 0; i < MAGIC_SETS; i++) { 2982 if (needsbyteswap) 2983 map->nmagic[i] = swap4(ptr[i + 2]); 2984 else 2985 map->nmagic[i] = ptr[i + 2]; 2986 if (i != MAGIC_SETS - 1) 2987 map->magic[i + 1] = map->magic[i] + map->nmagic[i]; 2988 nentries += map->nmagic[i]; 2989 } 2990 if (entries != nentries + 1) { 2991 file_error(ms, 0, "Inconsistent entries in `%s' %u != %u", 2992 dbname, entries, nentries + 1); 2993 return -1; 2994 } 2995 if (needsbyteswap) 2996 for (i = 0; i < MAGIC_SETS; i++) 2997 byteswap(map->magic[i], map->nmagic[i]); 2998 return 0; 2999 } 3000 3001 /* 3002 * handle an mmaped file. 3003 */ 3004 private int 3005 apprentice_compile(struct magic_set *ms, struct magic_map *map, const char *fn) 3006 { 3007 static const size_t nm = sizeof(*map->nmagic) * MAGIC_SETS; 3008 static const size_t m = sizeof(**map->magic); 3009 int fd = -1; 3010 size_t len; 3011 char *dbname; 3012 int rv = -1; 3013 uint32_t i; 3014 union { 3015 struct magic m; 3016 uint32_t h[2 + MAGIC_SETS]; 3017 } hdr; 3018 3019 dbname = mkdbname(ms, fn, 1); 3020 3021 if (dbname == NULL) 3022 goto out; 3023 3024 if ((fd = open(dbname, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY, 0644)) == -1) 3025 { 3026 file_error(ms, errno, "cannot open `%s'", dbname); 3027 goto out; 3028 } 3029 memset(&hdr, 0, sizeof(hdr)); 3030 hdr.h[0] = MAGICNO; 3031 hdr.h[1] = VERSIONNO; 3032 memcpy(hdr.h + 2, map->nmagic, nm); 3033 3034 if (write(fd, &hdr, sizeof(hdr)) != (ssize_t)sizeof(hdr)) { 3035 file_error(ms, errno, "error writing `%s'", dbname); 3036 goto out; 3037 } 3038 3039 for (i = 0; i < MAGIC_SETS; i++) { 3040 len = m * map->nmagic[i]; 3041 if (write(fd, map->magic[i], len) != (ssize_t)len) { 3042 file_error(ms, errno, "error writing `%s'", dbname); 3043 goto out; 3044 } 3045 } 3046 3047 if (fd != -1) 3048 (void)close(fd); 3049 rv = 0; 3050 out: 3051 free(dbname); 3052 return rv; 3053 } 3054 3055 private const char ext[] = ".mgc"; 3056 /* 3057 * make a dbname 3058 */ 3059 private char * 3060 mkdbname(struct magic_set *ms, const char *fn, int strip) 3061 { 3062 const char *p, *q; 3063 char *buf; 3064 3065 if (strip) { 3066 if ((p = strrchr(fn, '/')) != NULL) 3067 fn = ++p; 3068 } 3069 3070 for (q = fn; *q; q++) 3071 continue; 3072 /* Look for .mgc */ 3073 for (p = ext + sizeof(ext) - 1; p >= ext && q >= fn; p--, q--) 3074 if (*p != *q) 3075 break; 3076 3077 /* Did not find .mgc, restore q */ 3078 if (p >= ext) 3079 while (*q) 3080 q++; 3081 3082 q++; 3083 /* Compatibility with old code that looked in .mime */ 3084 if (ms->flags & MAGIC_MIME) { 3085 if (asprintf(&buf, "%.*s.mime%s", (int)(q - fn), fn, ext) < 0) 3086 return NULL; 3087 if (access(buf, R_OK) != -1) { 3088 ms->flags &= MAGIC_MIME_TYPE; 3089 return buf; 3090 } 3091 free(buf); 3092 } 3093 if (asprintf(&buf, "%.*s%s", (int)(q - fn), fn, ext) < 0) 3094 return NULL; 3095 3096 /* Compatibility with old code that looked in .mime */ 3097 if (strstr(p, ".mime") != NULL) 3098 ms->flags &= MAGIC_MIME_TYPE; 3099 return buf; 3100 } 3101 3102 /* 3103 * Byteswap an mmap'ed file if needed 3104 */ 3105 private void 3106 byteswap(struct magic *magic, uint32_t nmagic) 3107 { 3108 uint32_t i; 3109 for (i = 0; i < nmagic; i++) 3110 bs1(&magic[i]); 3111 } 3112 3113 /* 3114 * swap a short 3115 */ 3116 private uint16_t 3117 swap2(uint16_t sv) 3118 { 3119 uint16_t rv; 3120 uint8_t *s = (uint8_t *)(void *)&sv; 3121 uint8_t *d = (uint8_t *)(void *)&rv; 3122 d[0] = s[1]; 3123 d[1] = s[0]; 3124 return rv; 3125 } 3126 3127 /* 3128 * swap an int 3129 */ 3130 private uint32_t 3131 swap4(uint32_t sv) 3132 { 3133 uint32_t rv; 3134 uint8_t *s = (uint8_t *)(void *)&sv; 3135 uint8_t *d = (uint8_t *)(void *)&rv; 3136 d[0] = s[3]; 3137 d[1] = s[2]; 3138 d[2] = s[1]; 3139 d[3] = s[0]; 3140 return rv; 3141 } 3142 3143 /* 3144 * swap a quad 3145 */ 3146 private uint64_t 3147 swap8(uint64_t sv) 3148 { 3149 uint64_t rv; 3150 uint8_t *s = (uint8_t *)(void *)&sv; 3151 uint8_t *d = (uint8_t *)(void *)&rv; 3152 #if 0 3153 d[0] = s[3]; 3154 d[1] = s[2]; 3155 d[2] = s[1]; 3156 d[3] = s[0]; 3157 d[4] = s[7]; 3158 d[5] = s[6]; 3159 d[6] = s[5]; 3160 d[7] = s[4]; 3161 #else 3162 d[0] = s[7]; 3163 d[1] = s[6]; 3164 d[2] = s[5]; 3165 d[3] = s[4]; 3166 d[4] = s[3]; 3167 d[5] = s[2]; 3168 d[6] = s[1]; 3169 d[7] = s[0]; 3170 #endif 3171 return rv; 3172 } 3173 3174 /* 3175 * byteswap a single magic entry 3176 */ 3177 private void 3178 bs1(struct magic *m) 3179 { 3180 m->cont_level = swap2(m->cont_level); 3181 m->offset = swap4((uint32_t)m->offset); 3182 m->in_offset = swap4((uint32_t)m->in_offset); 3183 m->lineno = swap4((uint32_t)m->lineno); 3184 if (IS_STRING(m->type)) { 3185 m->str_range = swap4(m->str_range); 3186 m->str_flags = swap4(m->str_flags); 3187 } 3188 else { 3189 m->value.q = swap8(m->value.q); 3190 m->num_mask = swap8(m->num_mask); 3191 } 3192 } 3193 3194 protected size_t 3195 file_pstring_length_size(const struct magic *m) 3196 { 3197 switch (m->str_flags & PSTRING_LEN) { 3198 case PSTRING_1_LE: 3199 return 1; 3200 case PSTRING_2_LE: 3201 case PSTRING_2_BE: 3202 return 2; 3203 case PSTRING_4_LE: 3204 case PSTRING_4_BE: 3205 return 4; 3206 default: 3207 abort(); /* Impossible */ 3208 return 1; 3209 } 3210 } 3211 protected size_t 3212 file_pstring_get_length(const struct magic *m, const char *s) 3213 { 3214 size_t len = 0; 3215 3216 switch (m->str_flags & PSTRING_LEN) { 3217 case PSTRING_1_LE: 3218 len = *s; 3219 break; 3220 case PSTRING_2_LE: 3221 len = (s[1] << 8) | s[0]; 3222 break; 3223 case PSTRING_2_BE: 3224 len = (s[0] << 8) | s[1]; 3225 break; 3226 case PSTRING_4_LE: 3227 len = (s[3] << 24) | (s[2] << 16) | (s[1] << 8) | s[0]; 3228 break; 3229 case PSTRING_4_BE: 3230 len = (s[0] << 24) | (s[1] << 16) | (s[2] << 8) | s[3]; 3231 break; 3232 default: 3233 abort(); /* Impossible */ 3234 } 3235 3236 if (m->str_flags & PSTRING_LENGTH_INCLUDES_ITSELF) 3237 len -= file_pstring_length_size(m); 3238 3239 return len; 3240 } 3241 3242 protected int 3243 file_magicfind(struct magic_set *ms, const char *name, struct mlist *v) 3244 { 3245 uint32_t i, j; 3246 struct mlist *mlist, *ml; 3247 3248 mlist = ms->mlist[1]; 3249 3250 for (ml = mlist->next; ml != mlist; ml = ml->next) { 3251 struct magic *ma = ml->magic; 3252 uint32_t nma = ml->nmagic; 3253 for (i = 0; i < nma; i++) { 3254 if (ma[i].type != FILE_NAME) 3255 continue; 3256 if (strcmp(ma[i].value.s, name) == 0) { 3257 v->magic = &ma[i]; 3258 for (j = i + 1; j < nma; j++) 3259 if (ma[j].cont_level == 0) 3260 break; 3261 v->nmagic = j - i; 3262 return 0; 3263 } 3264 } 3265 } 3266 return -1; 3267 } 3268