1 /* $NetBSD: apprentice.c,v 1.20 2018/04/15 19:45:32 christos Exp $ */ 2 3 /* 4 * Copyright (c) Ian F. Darwin 1986-1995. 5 * Software written by Ian F. Darwin and others; 6 * maintained 1995-present by Christos Zoulas and others. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice immediately at the beginning of the file, without modification, 13 * this list of conditions, and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 21 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR 22 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 28 * SUCH DAMAGE. 29 */ 30 /* 31 * apprentice - make one pass through /etc/magic, learning its secrets. 32 */ 33 34 #include "file.h" 35 36 #ifndef lint 37 #if 0 38 FILE_RCSID("@(#)$File: apprentice.c,v 1.270 2018/02/21 21:26:48 christos Exp $") 39 #else 40 __RCSID("$NetBSD: apprentice.c,v 1.20 2018/04/15 19:45:32 christos Exp $"); 41 #endif 42 #endif /* lint */ 43 44 #include "magic.h" 45 #include <stdlib.h> 46 #ifdef HAVE_UNISTD_H 47 #include <unistd.h> 48 #endif 49 #ifdef HAVE_STDDEF_H 50 #include <stddef.h> 51 #endif 52 #include <string.h> 53 #include <assert.h> 54 #include <ctype.h> 55 #include <fcntl.h> 56 #ifdef QUICK 57 #include <sys/mman.h> 58 #endif 59 #include <dirent.h> 60 #if defined(HAVE_LIMITS_H) 61 #include <limits.h> 62 #endif 63 64 #ifndef SSIZE_MAX 65 #define MAXMAGIC_SIZE ((ssize_t)0x7fffffff) 66 #else 67 #define MAXMAGIC_SIZE SSIZE_MAX 68 #endif 69 70 #define EATAB {while (isascii((unsigned char) *l) && \ 71 isspace((unsigned char) *l)) ++l;} 72 #define LOWCASE(l) (isupper((unsigned char) (l)) ? \ 73 tolower((unsigned char) (l)) : (l)) 74 /* 75 * Work around a bug in headers on Digital Unix. 76 * At least confirmed for: OSF1 V4.0 878 77 */ 78 #if defined(__osf__) && defined(__DECC) 79 #ifdef MAP_FAILED 80 #undef MAP_FAILED 81 #endif 82 #endif 83 84 #ifndef MAP_FAILED 85 #define MAP_FAILED (void *) -1 86 #endif 87 88 #ifndef MAP_FILE 89 #define MAP_FILE 0 90 #endif 91 92 #define ALLOC_CHUNK (size_t)10 93 #define ALLOC_INCR (size_t)200 94 95 #define MAP_TYPE_USER 0 96 #define MAP_TYPE_MALLOC 1 97 #define MAP_TYPE_MMAP 2 98 99 struct magic_entry { 100 struct magic *mp; 101 uint32_t cont_count; 102 uint32_t max_count; 103 }; 104 105 struct magic_entry_set { 106 struct magic_entry *me; 107 uint32_t count; 108 uint32_t max; 109 }; 110 111 struct magic_map { 112 void *p; 113 size_t len; 114 int type; 115 struct magic *magic[MAGIC_SETS]; 116 uint32_t nmagic[MAGIC_SETS]; 117 }; 118 119 int file_formats[FILE_NAMES_SIZE]; 120 const size_t file_nformats = FILE_NAMES_SIZE; 121 const char *file_names[FILE_NAMES_SIZE]; 122 const size_t file_nnames = FILE_NAMES_SIZE; 123 124 private int getvalue(struct magic_set *ms, struct magic *, const char **, int); 125 private int hextoint(int); 126 private const char *getstr(struct magic_set *, struct magic *, const char *, 127 int); 128 private int parse(struct magic_set *, struct magic_entry *, const char *, 129 size_t, int); 130 private void eatsize(const char **); 131 private int apprentice_1(struct magic_set *, const char *, int); 132 private size_t apprentice_magic_strength(const struct magic *); 133 private int apprentice_sort(const void *, const void *); 134 private void apprentice_list(struct mlist *, int ); 135 private struct magic_map *apprentice_load(struct magic_set *, 136 const char *, int); 137 private struct mlist *mlist_alloc(void); 138 private void mlist_free(struct mlist *); 139 private void byteswap(struct magic *, uint32_t); 140 private void bs1(struct magic *); 141 private uint16_t swap2(uint16_t); 142 private uint32_t swap4(uint32_t); 143 private uint64_t swap8(uint64_t); 144 private char *mkdbname(struct magic_set *, const char *, int); 145 private struct magic_map *apprentice_buf(struct magic_set *, struct magic *, 146 size_t); 147 private struct magic_map *apprentice_map(struct magic_set *, const char *); 148 private int check_buffer(struct magic_set *, struct magic_map *, const char *); 149 private void apprentice_unmap(struct magic_map *); 150 private int apprentice_compile(struct magic_set *, struct magic_map *, 151 const char *); 152 private int check_format_type(const char *, int, const char **); 153 private int check_format(struct magic_set *, struct magic *); 154 private int get_op(char); 155 private int parse_mime(struct magic_set *, struct magic_entry *, const char *); 156 private int parse_strength(struct magic_set *, struct magic_entry *, const char *); 157 private int parse_apple(struct magic_set *, struct magic_entry *, const char *); 158 private int parse_ext(struct magic_set *, struct magic_entry *, const char *); 159 160 161 private size_t magicsize = sizeof(struct magic); 162 163 private const char usg_hdr[] = "cont\toffset\ttype\topcode\tmask\tvalue\tdesc"; 164 165 private struct { 166 const char *name; 167 size_t len; 168 int (*fun)(struct magic_set *, struct magic_entry *, const char *); 169 } bang[] = { 170 #define DECLARE_FIELD(name) { # name, sizeof(# name) - 1, parse_ ## name } 171 DECLARE_FIELD(mime), 172 DECLARE_FIELD(apple), 173 DECLARE_FIELD(ext), 174 DECLARE_FIELD(strength), 175 #undef DECLARE_FIELD 176 { NULL, 0, NULL } 177 }; 178 179 #ifdef COMPILE_ONLY 180 181 int main(int, char *[]); 182 183 int 184 main(int argc, char *argv[]) 185 { 186 int ret; 187 struct magic_set *ms; 188 char *progname; 189 190 if ((progname = strrchr(argv[0], '/')) != NULL) 191 progname++; 192 else 193 progname = argv[0]; 194 195 if (argc != 2) { 196 (void)fprintf(stderr, "Usage: %s file\n", progname); 197 return 1; 198 } 199 200 if ((ms = magic_open(MAGIC_CHECK)) == NULL) { 201 (void)fprintf(stderr, "%s: %s\n", progname, strerror(errno)); 202 return 1; 203 } 204 ret = magic_compile(ms, argv[1]) == -1 ? 1 : 0; 205 if (ret == 1) 206 (void)fprintf(stderr, "%s: %s\n", progname, magic_error(ms)); 207 magic_close(ms); 208 return ret; 209 } 210 #endif /* COMPILE_ONLY */ 211 212 struct type_tbl_s { 213 const char name[16]; 214 const size_t len; 215 const int type; 216 const int format; 217 }; 218 219 /* 220 * XXX - the actual Single UNIX Specification says that "long" means "long", 221 * as in the C data type, but we treat it as meaning "4-byte integer". 222 * Given that the OS X version of file 5.04 did the same, I guess that passes 223 * the actual test; having "long" be dependent on how big a "long" is on 224 * the machine running "file" is silly. 225 */ 226 static const struct type_tbl_s type_tbl[] = { 227 # define XX(s) s, (sizeof(s) - 1) 228 # define XX_NULL "", 0 229 { XX("invalid"), FILE_INVALID, FILE_FMT_NONE }, 230 { XX("byte"), FILE_BYTE, FILE_FMT_NUM }, 231 { XX("short"), FILE_SHORT, FILE_FMT_NUM }, 232 { XX("default"), FILE_DEFAULT, FILE_FMT_NONE }, 233 { XX("long"), FILE_LONG, FILE_FMT_NUM }, 234 { XX("string"), FILE_STRING, FILE_FMT_STR }, 235 { XX("date"), FILE_DATE, FILE_FMT_STR }, 236 { XX("beshort"), FILE_BESHORT, FILE_FMT_NUM }, 237 { XX("belong"), FILE_BELONG, FILE_FMT_NUM }, 238 { XX("bedate"), FILE_BEDATE, FILE_FMT_STR }, 239 { XX("leshort"), FILE_LESHORT, FILE_FMT_NUM }, 240 { XX("lelong"), FILE_LELONG, FILE_FMT_NUM }, 241 { XX("ledate"), FILE_LEDATE, FILE_FMT_STR }, 242 { XX("pstring"), FILE_PSTRING, FILE_FMT_STR }, 243 { XX("ldate"), FILE_LDATE, FILE_FMT_STR }, 244 { XX("beldate"), FILE_BELDATE, FILE_FMT_STR }, 245 { XX("leldate"), FILE_LELDATE, FILE_FMT_STR }, 246 { XX("regex"), FILE_REGEX, FILE_FMT_STR }, 247 { XX("bestring16"), FILE_BESTRING16, FILE_FMT_STR }, 248 { XX("lestring16"), FILE_LESTRING16, FILE_FMT_STR }, 249 { XX("search"), FILE_SEARCH, FILE_FMT_STR }, 250 { XX("medate"), FILE_MEDATE, FILE_FMT_STR }, 251 { XX("meldate"), FILE_MELDATE, FILE_FMT_STR }, 252 { XX("melong"), FILE_MELONG, FILE_FMT_NUM }, 253 { XX("quad"), FILE_QUAD, FILE_FMT_QUAD }, 254 { XX("lequad"), FILE_LEQUAD, FILE_FMT_QUAD }, 255 { XX("bequad"), FILE_BEQUAD, FILE_FMT_QUAD }, 256 { XX("qdate"), FILE_QDATE, FILE_FMT_STR }, 257 { XX("leqdate"), FILE_LEQDATE, FILE_FMT_STR }, 258 { XX("beqdate"), FILE_BEQDATE, FILE_FMT_STR }, 259 { XX("qldate"), FILE_QLDATE, FILE_FMT_STR }, 260 { XX("leqldate"), FILE_LEQLDATE, FILE_FMT_STR }, 261 { XX("beqldate"), FILE_BEQLDATE, FILE_FMT_STR }, 262 { XX("float"), FILE_FLOAT, FILE_FMT_FLOAT }, 263 { XX("befloat"), FILE_BEFLOAT, FILE_FMT_FLOAT }, 264 { XX("lefloat"), FILE_LEFLOAT, FILE_FMT_FLOAT }, 265 { XX("double"), FILE_DOUBLE, FILE_FMT_DOUBLE }, 266 { XX("bedouble"), FILE_BEDOUBLE, FILE_FMT_DOUBLE }, 267 { XX("ledouble"), FILE_LEDOUBLE, FILE_FMT_DOUBLE }, 268 { XX("leid3"), FILE_LEID3, FILE_FMT_NUM }, 269 { XX("beid3"), FILE_BEID3, FILE_FMT_NUM }, 270 { XX("indirect"), FILE_INDIRECT, FILE_FMT_NUM }, 271 { XX("qwdate"), FILE_QWDATE, FILE_FMT_STR }, 272 { XX("leqwdate"), FILE_LEQWDATE, FILE_FMT_STR }, 273 { XX("beqwdate"), FILE_BEQWDATE, FILE_FMT_STR }, 274 { XX("name"), FILE_NAME, FILE_FMT_NONE }, 275 { XX("use"), FILE_USE, FILE_FMT_NONE }, 276 { XX("clear"), FILE_CLEAR, FILE_FMT_NONE }, 277 { XX("der"), FILE_DER, FILE_FMT_STR }, 278 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 279 }; 280 281 /* 282 * These are not types, and cannot be preceded by "u" to make them 283 * unsigned. 284 */ 285 static const struct type_tbl_s special_tbl[] = { 286 { XX("der"), FILE_DER, FILE_FMT_STR }, 287 { XX("name"), FILE_NAME, FILE_FMT_STR }, 288 { XX("use"), FILE_USE, FILE_FMT_STR }, 289 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 290 }; 291 # undef XX 292 # undef XX_NULL 293 294 private int 295 get_type(const struct type_tbl_s *tbl, const char *l, const char **t) 296 { 297 const struct type_tbl_s *p; 298 299 for (p = tbl; p->len; p++) { 300 if (strncmp(l, p->name, p->len) == 0) { 301 if (t) 302 *t = l + p->len; 303 break; 304 } 305 } 306 return p->type; 307 } 308 309 private int 310 get_standard_integer_type(const char *l, const char **t) 311 { 312 int type; 313 314 if (isalpha((unsigned char)l[1])) { 315 switch (l[1]) { 316 case 'C': 317 /* "dC" and "uC" */ 318 type = FILE_BYTE; 319 break; 320 case 'S': 321 /* "dS" and "uS" */ 322 type = FILE_SHORT; 323 break; 324 case 'I': 325 case 'L': 326 /* 327 * "dI", "dL", "uI", and "uL". 328 * 329 * XXX - the actual Single UNIX Specification says 330 * that "L" means "long", as in the C data type, 331 * but we treat it as meaning "4-byte integer". 332 * Given that the OS X version of file 5.04 did 333 * the same, I guess that passes the actual SUS 334 * validation suite; having "dL" be dependent on 335 * how big a "long" is on the machine running 336 * "file" is silly. 337 */ 338 type = FILE_LONG; 339 break; 340 case 'Q': 341 /* "dQ" and "uQ" */ 342 type = FILE_QUAD; 343 break; 344 default: 345 /* "d{anything else}", "u{anything else}" */ 346 return FILE_INVALID; 347 } 348 l += 2; 349 } else if (isdigit((unsigned char)l[1])) { 350 /* 351 * "d{num}" and "u{num}"; we only support {num} values 352 * of 1, 2, 4, and 8 - the Single UNIX Specification 353 * doesn't say anything about whether arbitrary 354 * values should be supported, but both the Solaris 10 355 * and OS X Mountain Lion versions of file passed the 356 * Single UNIX Specification validation suite, and 357 * neither of them support values bigger than 8 or 358 * non-power-of-2 values. 359 */ 360 if (isdigit((unsigned char)l[2])) { 361 /* Multi-digit, so > 9 */ 362 return FILE_INVALID; 363 } 364 switch (l[1]) { 365 case '1': 366 type = FILE_BYTE; 367 break; 368 case '2': 369 type = FILE_SHORT; 370 break; 371 case '4': 372 type = FILE_LONG; 373 break; 374 case '8': 375 type = FILE_QUAD; 376 break; 377 default: 378 /* XXX - what about 3, 5, 6, or 7? */ 379 return FILE_INVALID; 380 } 381 l += 2; 382 } else { 383 /* 384 * "d" or "u" by itself. 385 */ 386 type = FILE_LONG; 387 ++l; 388 } 389 if (t) 390 *t = l; 391 return type; 392 } 393 394 private void 395 init_file_tables(void) 396 { 397 static int done = 0; 398 const struct type_tbl_s *p; 399 400 if (done) 401 return; 402 done++; 403 404 for (p = type_tbl; p->len; p++) { 405 assert(p->type < FILE_NAMES_SIZE); 406 file_names[p->type] = p->name; 407 file_formats[p->type] = p->format; 408 } 409 assert(p - type_tbl == FILE_NAMES_SIZE); 410 } 411 412 private int 413 add_mlist(struct mlist *mlp, struct magic_map *map, size_t idx) 414 { 415 struct mlist *ml; 416 417 mlp->map = NULL; 418 if ((ml = CAST(struct mlist *, malloc(sizeof(*ml)))) == NULL) 419 return -1; 420 421 ml->map = idx == 0 ? map : NULL; 422 ml->magic = map->magic[idx]; 423 ml->nmagic = map->nmagic[idx]; 424 425 mlp->prev->next = ml; 426 ml->prev = mlp->prev; 427 ml->next = mlp; 428 mlp->prev = ml; 429 return 0; 430 } 431 432 /* 433 * Handle one file or directory. 434 */ 435 private int 436 apprentice_1(struct magic_set *ms, const char *fn, int action) 437 { 438 struct magic_map *map; 439 #ifndef COMPILE_ONLY 440 struct mlist *ml; 441 size_t i; 442 #endif 443 444 if (magicsize != FILE_MAGICSIZE) { 445 file_error(ms, 0, "magic element size %lu != %lu", 446 (unsigned long)sizeof(*map->magic[0]), 447 (unsigned long)FILE_MAGICSIZE); 448 return -1; 449 } 450 451 if (action == FILE_COMPILE) { 452 map = apprentice_load(ms, fn, action); 453 if (map == NULL) 454 return -1; 455 return apprentice_compile(ms, map, fn); 456 } 457 458 #ifndef COMPILE_ONLY 459 map = apprentice_map(ms, fn); 460 if (map == (struct magic_map *)-1) 461 return -1; 462 if (map == NULL) { 463 if (ms->flags & MAGIC_CHECK) 464 file_magwarn(ms, "using regular magic file `%s'", fn); 465 map = apprentice_load(ms, fn, action); 466 if (map == NULL) 467 return -1; 468 } 469 470 for (i = 0; i < MAGIC_SETS; i++) { 471 if (add_mlist(ms->mlist[i], map, i) == -1) { 472 file_oomem(ms, sizeof(*ml)); 473 return -1; 474 } 475 } 476 477 if (action == FILE_LIST) { 478 for (i = 0; i < MAGIC_SETS; i++) { 479 printf("Set %" SIZE_T_FORMAT "u:\nBinary patterns:\n", 480 i); 481 apprentice_list(ms->mlist[i], BINTEST); 482 printf("Text patterns:\n"); 483 apprentice_list(ms->mlist[i], TEXTTEST); 484 } 485 } 486 return 0; 487 #else 488 return 0; 489 #endif /* COMPILE_ONLY */ 490 } 491 492 protected void 493 file_ms_free(struct magic_set *ms) 494 { 495 size_t i; 496 if (ms == NULL) 497 return; 498 for (i = 0; i < MAGIC_SETS; i++) 499 mlist_free(ms->mlist[i]); 500 free(ms->o.pbuf); 501 free(ms->o.buf); 502 free(ms->c.li); 503 free(ms); 504 } 505 506 protected struct magic_set * 507 file_ms_alloc(int flags) 508 { 509 struct magic_set *ms; 510 size_t i, len; 511 512 if ((ms = CAST(struct magic_set *, calloc((size_t)1, 513 sizeof(struct magic_set)))) == NULL) 514 return NULL; 515 516 if (magic_setflags(ms, flags) == -1) { 517 errno = EINVAL; 518 goto free; 519 } 520 521 ms->o.buf = ms->o.pbuf = NULL; 522 len = (ms->c.len = 10) * sizeof(*ms->c.li); 523 524 if ((ms->c.li = CAST(struct level_info *, malloc(len))) == NULL) 525 goto free; 526 527 ms->event_flags = 0; 528 ms->error = -1; 529 for (i = 0; i < MAGIC_SETS; i++) 530 ms->mlist[i] = NULL; 531 ms->file = "unknown"; 532 ms->line = 0; 533 ms->indir_max = FILE_INDIR_MAX; 534 ms->name_max = FILE_NAME_MAX; 535 ms->elf_shnum_max = FILE_ELF_SHNUM_MAX; 536 ms->elf_phnum_max = FILE_ELF_PHNUM_MAX; 537 ms->elf_notes_max = FILE_ELF_NOTES_MAX; 538 ms->regex_max = FILE_REGEX_MAX; 539 ms->bytes_max = FILE_BYTES_MAX; 540 return ms; 541 free: 542 free(ms); 543 return NULL; 544 } 545 546 private void 547 apprentice_unmap(struct magic_map *map) 548 { 549 size_t i; 550 if (map == NULL) 551 return; 552 553 switch (map->type) { 554 case MAP_TYPE_USER: 555 break; 556 case MAP_TYPE_MALLOC: 557 for (i = 0; i < MAGIC_SETS; i++) { 558 void *b = map->magic[i]; 559 void *p = map->p; 560 if (CAST(char *, b) >= CAST(char *, p) && 561 CAST(char *, b) <= CAST(char *, p) + map->len) 562 continue; 563 free(map->magic[i]); 564 } 565 free(map->p); 566 break; 567 #ifdef QUICK 568 case MAP_TYPE_MMAP: 569 if (map->p && map->p != MAP_FAILED) 570 (void)munmap(map->p, map->len); 571 break; 572 #endif 573 default: 574 abort(); 575 } 576 free(map); 577 } 578 579 private struct mlist * 580 mlist_alloc(void) 581 { 582 struct mlist *mlist; 583 if ((mlist = CAST(struct mlist *, calloc(1, sizeof(*mlist)))) == NULL) { 584 return NULL; 585 } 586 mlist->next = mlist->prev = mlist; 587 return mlist; 588 } 589 590 private void 591 mlist_free(struct mlist *mlist) 592 { 593 struct mlist *ml, *next; 594 595 if (mlist == NULL) 596 return; 597 598 ml = mlist->next; 599 for (ml = mlist->next; (next = ml->next) != NULL; ml = next) { 600 if (ml->map) 601 apprentice_unmap(CAST(struct magic_map *, ml->map)); 602 free(ml); 603 if (ml == mlist) 604 break; 605 } 606 } 607 608 #ifndef COMPILE_ONLY 609 /* void **bufs: an array of compiled magic files */ 610 protected int 611 buffer_apprentice(struct magic_set *ms, struct magic **bufs, 612 size_t *sizes, size_t nbufs) 613 { 614 size_t i, j; 615 struct mlist *ml; 616 struct magic_map *map; 617 618 if (nbufs == 0) 619 return -1; 620 621 (void)file_reset(ms, 0); 622 623 init_file_tables(); 624 625 for (i = 0; i < MAGIC_SETS; i++) { 626 mlist_free(ms->mlist[i]); 627 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 628 file_oomem(ms, sizeof(*ms->mlist[i])); 629 goto fail; 630 } 631 } 632 633 for (i = 0; i < nbufs; i++) { 634 map = apprentice_buf(ms, bufs[i], sizes[i]); 635 if (map == NULL) 636 goto fail; 637 638 for (j = 0; j < MAGIC_SETS; j++) { 639 if (add_mlist(ms->mlist[j], map, j) == -1) { 640 file_oomem(ms, sizeof(*ml)); 641 goto fail; 642 } 643 } 644 } 645 646 return 0; 647 fail: 648 for (i = 0; i < MAGIC_SETS; i++) { 649 mlist_free(ms->mlist[i]); 650 ms->mlist[i] = NULL; 651 } 652 return -1; 653 } 654 #endif 655 656 /* const char *fn: list of magic files and directories */ 657 protected int 658 file_apprentice(struct magic_set *ms, const char *fn, int action) 659 { 660 char *p, *mfn; 661 int fileerr, errs = -1; 662 size_t i; 663 664 (void)file_reset(ms, 0); 665 666 if ((fn = magic_getpath(fn, action)) == NULL) 667 return -1; 668 669 init_file_tables(); 670 671 if ((mfn = strdup(fn)) == NULL) { 672 file_oomem(ms, strlen(fn)); 673 return -1; 674 } 675 676 for (i = 0; i < MAGIC_SETS; i++) { 677 mlist_free(ms->mlist[i]); 678 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 679 file_oomem(ms, sizeof(*ms->mlist[i])); 680 while (i-- > 0) { 681 mlist_free(ms->mlist[i]); 682 ms->mlist[i] = NULL; 683 } 684 free(mfn); 685 return -1; 686 } 687 } 688 fn = mfn; 689 690 while (fn) { 691 p = strchr(fn, PATHSEP); 692 if (p) 693 *p++ = '\0'; 694 if (*fn == '\0') 695 break; 696 fileerr = apprentice_1(ms, fn, action); 697 errs = MAX(errs, fileerr); 698 fn = p; 699 } 700 701 free(mfn); 702 703 if (errs == -1) { 704 for (i = 0; i < MAGIC_SETS; i++) { 705 mlist_free(ms->mlist[i]); 706 ms->mlist[i] = NULL; 707 } 708 file_error(ms, 0, "could not find any valid magic files!"); 709 return -1; 710 } 711 712 #if 0 713 /* 714 * Always leave the database loaded 715 */ 716 if (action == FILE_LOAD) 717 return 0; 718 719 for (i = 0; i < MAGIC_SETS; i++) { 720 mlist_free(ms->mlist[i]); 721 ms->mlist[i] = NULL; 722 } 723 #endif 724 725 switch (action) { 726 case FILE_LOAD: 727 case FILE_COMPILE: 728 case FILE_CHECK: 729 case FILE_LIST: 730 return 0; 731 default: 732 file_error(ms, 0, "Invalid action %d", action); 733 return -1; 734 } 735 } 736 737 /* 738 * Compute the real length of a magic expression, for the purposes 739 * of determining how "strong" a magic expression is (approximating 740 * how specific its matches are): 741 * - magic characters count 0 unless escaped. 742 * - [] expressions count 1 743 * - {} expressions count 0 744 * - regular characters or escaped magic characters count 1 745 * - 0 length expressions count as one 746 */ 747 private size_t 748 nonmagic(const char *str) 749 { 750 const char *p; 751 size_t rv = 0; 752 753 for (p = str; *p; p++) 754 switch (*p) { 755 case '\\': /* Escaped anything counts 1 */ 756 if (!*++p) 757 p--; 758 rv++; 759 continue; 760 case '?': /* Magic characters count 0 */ 761 case '*': 762 case '.': 763 case '+': 764 case '^': 765 case '$': 766 continue; 767 case '[': /* Bracketed expressions count 1 the ']' */ 768 while (*p && *p != ']') 769 p++; 770 p--; 771 continue; 772 case '{': /* Braced expressions count 0 */ 773 while (*p && *p != '}') 774 p++; 775 if (!*p) 776 p--; 777 continue; 778 default: /* Anything else counts 1 */ 779 rv++; 780 continue; 781 } 782 783 return rv == 0 ? 1 : rv; /* Return at least 1 */ 784 } 785 786 787 private size_t 788 typesize(int type) 789 { 790 switch (type) { 791 case FILE_BYTE: 792 return 1; 793 794 case FILE_SHORT: 795 case FILE_LESHORT: 796 case FILE_BESHORT: 797 return 2; 798 799 case FILE_LONG: 800 case FILE_LELONG: 801 case FILE_BELONG: 802 case FILE_MELONG: 803 return 4; 804 805 case FILE_DATE: 806 case FILE_LEDATE: 807 case FILE_BEDATE: 808 case FILE_MEDATE: 809 case FILE_LDATE: 810 case FILE_LELDATE: 811 case FILE_BELDATE: 812 case FILE_MELDATE: 813 case FILE_FLOAT: 814 case FILE_BEFLOAT: 815 case FILE_LEFLOAT: 816 return 4; 817 818 case FILE_QUAD: 819 case FILE_BEQUAD: 820 case FILE_LEQUAD: 821 case FILE_QDATE: 822 case FILE_LEQDATE: 823 case FILE_BEQDATE: 824 case FILE_QLDATE: 825 case FILE_LEQLDATE: 826 case FILE_BEQLDATE: 827 case FILE_QWDATE: 828 case FILE_LEQWDATE: 829 case FILE_BEQWDATE: 830 case FILE_DOUBLE: 831 case FILE_BEDOUBLE: 832 case FILE_LEDOUBLE: 833 return 8; 834 default: 835 return (size_t)~0; 836 } 837 } 838 839 /* 840 * Get weight of this magic entry, for sorting purposes. 841 */ 842 private size_t 843 apprentice_magic_strength(const struct magic *m) 844 { 845 #define MULT 10 846 size_t ts, v, val = 2 * MULT; /* baseline strength */ 847 848 switch (m->type) { 849 case FILE_DEFAULT: /* make sure this sorts last */ 850 if (m->factor_op != FILE_FACTOR_OP_NONE) 851 abort(); 852 return 0; 853 854 case FILE_BYTE: 855 case FILE_SHORT: 856 case FILE_LESHORT: 857 case FILE_BESHORT: 858 case FILE_LONG: 859 case FILE_LELONG: 860 case FILE_BELONG: 861 case FILE_MELONG: 862 case FILE_DATE: 863 case FILE_LEDATE: 864 case FILE_BEDATE: 865 case FILE_MEDATE: 866 case FILE_LDATE: 867 case FILE_LELDATE: 868 case FILE_BELDATE: 869 case FILE_MELDATE: 870 case FILE_FLOAT: 871 case FILE_BEFLOAT: 872 case FILE_LEFLOAT: 873 case FILE_QUAD: 874 case FILE_BEQUAD: 875 case FILE_LEQUAD: 876 case FILE_QDATE: 877 case FILE_LEQDATE: 878 case FILE_BEQDATE: 879 case FILE_QLDATE: 880 case FILE_LEQLDATE: 881 case FILE_BEQLDATE: 882 case FILE_QWDATE: 883 case FILE_LEQWDATE: 884 case FILE_BEQWDATE: 885 case FILE_DOUBLE: 886 case FILE_BEDOUBLE: 887 case FILE_LEDOUBLE: 888 ts = typesize(m->type); 889 if (ts == (size_t)~0) 890 abort(); 891 val += ts * MULT; 892 break; 893 894 case FILE_PSTRING: 895 case FILE_STRING: 896 val += m->vallen * MULT; 897 break; 898 899 case FILE_BESTRING16: 900 case FILE_LESTRING16: 901 val += m->vallen * MULT / 2; 902 break; 903 904 case FILE_SEARCH: 905 val += m->vallen * MAX(MULT / m->vallen, 1); 906 break; 907 908 case FILE_REGEX: 909 v = nonmagic(m->value.s); 910 val += v * MAX(MULT / v, 1); 911 break; 912 913 case FILE_INDIRECT: 914 case FILE_NAME: 915 case FILE_USE: 916 break; 917 918 case FILE_DER: 919 val += MULT; 920 break; 921 922 default: 923 (void)fprintf(stderr, "Bad type %d\n", m->type); 924 abort(); 925 } 926 927 switch (m->reln) { 928 case 'x': /* matches anything penalize */ 929 case '!': /* matches almost anything penalize */ 930 val = 0; 931 break; 932 933 case '=': /* Exact match, prefer */ 934 val += MULT; 935 break; 936 937 case '>': 938 case '<': /* comparison match reduce strength */ 939 val -= 2 * MULT; 940 break; 941 942 case '^': 943 case '&': /* masking bits, we could count them too */ 944 val -= MULT; 945 break; 946 947 default: 948 (void)fprintf(stderr, "Bad relation %c\n", m->reln); 949 abort(); 950 } 951 952 if (val == 0) /* ensure we only return 0 for FILE_DEFAULT */ 953 val = 1; 954 955 switch (m->factor_op) { 956 case FILE_FACTOR_OP_NONE: 957 break; 958 case FILE_FACTOR_OP_PLUS: 959 val += m->factor; 960 break; 961 case FILE_FACTOR_OP_MINUS: 962 val -= m->factor; 963 break; 964 case FILE_FACTOR_OP_TIMES: 965 val *= m->factor; 966 break; 967 case FILE_FACTOR_OP_DIV: 968 val /= m->factor; 969 break; 970 default: 971 abort(); 972 } 973 974 /* 975 * Magic entries with no description get a bonus because they depend 976 * on subsequent magic entries to print something. 977 */ 978 if (m->desc[0] == '\0') 979 val++; 980 return val; 981 } 982 983 /* 984 * Sort callback for sorting entries by "strength" (basically length) 985 */ 986 private int 987 apprentice_sort(const void *a, const void *b) 988 { 989 const struct magic_entry *ma = CAST(const struct magic_entry *, a); 990 const struct magic_entry *mb = CAST(const struct magic_entry *, b); 991 size_t sa = apprentice_magic_strength(ma->mp); 992 size_t sb = apprentice_magic_strength(mb->mp); 993 if (sa == sb) 994 return 0; 995 else if (sa > sb) 996 return -1; 997 else 998 return 1; 999 } 1000 1001 /* 1002 * Shows sorted patterns list in the order which is used for the matching 1003 */ 1004 private void 1005 apprentice_list(struct mlist *mlist, int mode) 1006 { 1007 uint32_t magindex = 0; 1008 struct mlist *ml; 1009 for (ml = mlist->next; ml != mlist; ml = ml->next) { 1010 for (magindex = 0; magindex < ml->nmagic; magindex++) { 1011 struct magic *m = &ml->magic[magindex]; 1012 if ((m->flag & mode) != mode) { 1013 /* Skip sub-tests */ 1014 while (magindex + 1 < ml->nmagic && 1015 ml->magic[magindex + 1].cont_level != 0) 1016 ++magindex; 1017 continue; /* Skip to next top-level test*/ 1018 } 1019 1020 /* 1021 * Try to iterate over the tree until we find item with 1022 * description/mimetype. 1023 */ 1024 while (magindex + 1 < ml->nmagic && 1025 ml->magic[magindex + 1].cont_level != 0 && 1026 *ml->magic[magindex].desc == '\0' && 1027 *ml->magic[magindex].mimetype == '\0') 1028 magindex++; 1029 1030 printf("Strength = %3" SIZE_T_FORMAT "u@%u: %s [%s]\n", 1031 apprentice_magic_strength(m), 1032 ml->magic[magindex].lineno, 1033 ml->magic[magindex].desc, 1034 ml->magic[magindex].mimetype); 1035 } 1036 } 1037 } 1038 1039 private void 1040 set_test_type(struct magic *mstart, struct magic *m) 1041 { 1042 switch (m->type) { 1043 case FILE_BYTE: 1044 case FILE_SHORT: 1045 case FILE_LONG: 1046 case FILE_DATE: 1047 case FILE_BESHORT: 1048 case FILE_BELONG: 1049 case FILE_BEDATE: 1050 case FILE_LESHORT: 1051 case FILE_LELONG: 1052 case FILE_LEDATE: 1053 case FILE_LDATE: 1054 case FILE_BELDATE: 1055 case FILE_LELDATE: 1056 case FILE_MEDATE: 1057 case FILE_MELDATE: 1058 case FILE_MELONG: 1059 case FILE_QUAD: 1060 case FILE_LEQUAD: 1061 case FILE_BEQUAD: 1062 case FILE_QDATE: 1063 case FILE_LEQDATE: 1064 case FILE_BEQDATE: 1065 case FILE_QLDATE: 1066 case FILE_LEQLDATE: 1067 case FILE_BEQLDATE: 1068 case FILE_QWDATE: 1069 case FILE_LEQWDATE: 1070 case FILE_BEQWDATE: 1071 case FILE_FLOAT: 1072 case FILE_BEFLOAT: 1073 case FILE_LEFLOAT: 1074 case FILE_DOUBLE: 1075 case FILE_BEDOUBLE: 1076 case FILE_LEDOUBLE: 1077 case FILE_DER: 1078 mstart->flag |= BINTEST; 1079 break; 1080 case FILE_STRING: 1081 case FILE_PSTRING: 1082 case FILE_BESTRING16: 1083 case FILE_LESTRING16: 1084 /* Allow text overrides */ 1085 if (mstart->str_flags & STRING_TEXTTEST) 1086 mstart->flag |= TEXTTEST; 1087 else 1088 mstart->flag |= BINTEST; 1089 break; 1090 case FILE_REGEX: 1091 case FILE_SEARCH: 1092 /* Check for override */ 1093 if (mstart->str_flags & STRING_BINTEST) 1094 mstart->flag |= BINTEST; 1095 if (mstart->str_flags & STRING_TEXTTEST) 1096 mstart->flag |= TEXTTEST; 1097 1098 if (mstart->flag & (TEXTTEST|BINTEST)) 1099 break; 1100 1101 /* binary test if pattern is not text */ 1102 if (file_looks_utf8(m->value.us, (size_t)m->vallen, NULL, 1103 NULL) <= 0) 1104 mstart->flag |= BINTEST; 1105 else 1106 mstart->flag |= TEXTTEST; 1107 break; 1108 case FILE_DEFAULT: 1109 /* can't deduce anything; we shouldn't see this at the 1110 top level anyway */ 1111 break; 1112 case FILE_INVALID: 1113 default: 1114 /* invalid search type, but no need to complain here */ 1115 break; 1116 } 1117 } 1118 1119 private int 1120 addentry(struct magic_set *ms, struct magic_entry *me, 1121 struct magic_entry_set *mset) 1122 { 1123 size_t i = me->mp->type == FILE_NAME ? 1 : 0; 1124 if (mset[i].count == mset[i].max) { 1125 struct magic_entry *mp; 1126 1127 mset[i].max += ALLOC_INCR; 1128 if ((mp = CAST(struct magic_entry *, 1129 realloc(mset[i].me, sizeof(*mp) * mset[i].max))) == 1130 NULL) { 1131 file_oomem(ms, sizeof(*mp) * mset[i].max); 1132 return -1; 1133 } 1134 (void)memset(&mp[mset[i].count], 0, sizeof(*mp) * 1135 ALLOC_INCR); 1136 mset[i].me = mp; 1137 } 1138 mset[i].me[mset[i].count++] = *me; 1139 memset(me, 0, sizeof(*me)); 1140 return 0; 1141 } 1142 1143 /* 1144 * Load and parse one file. 1145 */ 1146 private void 1147 load_1(struct magic_set *ms, int action, const char *fn, int *errs, 1148 struct magic_entry_set *mset) 1149 { 1150 size_t lineno = 0, llen = 0; 1151 char *line = NULL; 1152 ssize_t len; 1153 struct magic_entry me; 1154 1155 FILE *f = fopen(ms->file = fn, "r"); 1156 if (f == NULL) { 1157 if (errno != ENOENT) 1158 file_error(ms, errno, "cannot read magic file `%s'", 1159 fn); 1160 (*errs)++; 1161 return; 1162 } 1163 1164 memset(&me, 0, sizeof(me)); 1165 /* read and parse this file */ 1166 for (ms->line = 1; (len = getline(&line, &llen, f)) != -1; 1167 ms->line++) { 1168 if (len == 0) /* null line, garbage, etc */ 1169 continue; 1170 if (line[len - 1] == '\n') { 1171 lineno++; 1172 line[len - 1] = '\0'; /* delete newline */ 1173 } 1174 switch (line[0]) { 1175 case '\0': /* empty, do not parse */ 1176 case '#': /* comment, do not parse */ 1177 continue; 1178 case '!': 1179 if (line[1] == ':') { 1180 size_t i; 1181 1182 for (i = 0; bang[i].name != NULL; i++) { 1183 if ((size_t)(len - 2) > bang[i].len && 1184 memcmp(bang[i].name, line + 2, 1185 bang[i].len) == 0) 1186 break; 1187 } 1188 if (bang[i].name == NULL) { 1189 file_error(ms, 0, 1190 "Unknown !: entry `%s'", line); 1191 (*errs)++; 1192 continue; 1193 } 1194 if (me.mp == NULL) { 1195 file_error(ms, 0, 1196 "No current entry for :!%s type", 1197 bang[i].name); 1198 (*errs)++; 1199 continue; 1200 } 1201 if ((*bang[i].fun)(ms, &me, 1202 line + bang[i].len + 2) != 0) { 1203 (*errs)++; 1204 continue; 1205 } 1206 continue; 1207 } 1208 /*FALLTHROUGH*/ 1209 default: 1210 again: 1211 switch (parse(ms, &me, line, lineno, action)) { 1212 case 0: 1213 continue; 1214 case 1: 1215 (void)addentry(ms, &me, mset); 1216 goto again; 1217 default: 1218 (*errs)++; 1219 break; 1220 } 1221 } 1222 } 1223 if (me.mp) 1224 (void)addentry(ms, &me, mset); 1225 free(line); 1226 (void)fclose(f); 1227 } 1228 1229 /* 1230 * parse a file or directory of files 1231 * const char *fn: name of magic file or directory 1232 */ 1233 private int 1234 cmpstrp(const void *p1, const void *p2) 1235 { 1236 return strcmp(*(char *const *)p1, *(char *const *)p2); 1237 } 1238 1239 1240 private uint32_t 1241 set_text_binary(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1242 uint32_t starttest) 1243 { 1244 static const char text[] = "text"; 1245 static const char binary[] = "binary"; 1246 static const size_t len = sizeof(text); 1247 1248 uint32_t i = starttest; 1249 1250 do { 1251 set_test_type(me[starttest].mp, me[i].mp); 1252 if ((ms->flags & MAGIC_DEBUG) == 0) 1253 continue; 1254 (void)fprintf(stderr, "%s%s%s: %s\n", 1255 me[i].mp->mimetype, 1256 me[i].mp->mimetype[0] == '\0' ? "" : "; ", 1257 me[i].mp->desc[0] ? me[i].mp->desc : "(no description)", 1258 me[i].mp->flag & BINTEST ? binary : text); 1259 if (me[i].mp->flag & BINTEST) { 1260 char *p = strstr(me[i].mp->desc, text); 1261 if (p && (p == me[i].mp->desc || 1262 isspace((unsigned char)p[-1])) && 1263 (p + len - me[i].mp->desc == MAXstring 1264 || (p[len] == '\0' || 1265 isspace((unsigned char)p[len])))) 1266 (void)fprintf(stderr, "*** Possible " 1267 "binary test for text type\n"); 1268 } 1269 } while (++i < nme && me[i].mp->cont_level != 0); 1270 return i; 1271 } 1272 1273 private void 1274 set_last_default(struct magic_set *ms, struct magic_entry *me, uint32_t nme) 1275 { 1276 uint32_t i; 1277 for (i = 0; i < nme; i++) { 1278 if (me[i].mp->cont_level == 0 && 1279 me[i].mp->type == FILE_DEFAULT) { 1280 while (++i < nme) 1281 if (me[i].mp->cont_level == 0) 1282 break; 1283 if (i != nme) { 1284 /* XXX - Ugh! */ 1285 ms->line = me[i].mp->lineno; 1286 file_magwarn(ms, 1287 "level 0 \"default\" did not sort last"); 1288 } 1289 return; 1290 } 1291 } 1292 } 1293 1294 private int 1295 coalesce_entries(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1296 struct magic **ma, uint32_t *nma) 1297 { 1298 uint32_t i, mentrycount = 0; 1299 size_t slen; 1300 1301 for (i = 0; i < nme; i++) 1302 mentrycount += me[i].cont_count; 1303 1304 slen = sizeof(**ma) * mentrycount; 1305 if ((*ma = CAST(struct magic *, malloc(slen))) == NULL) { 1306 file_oomem(ms, slen); 1307 return -1; 1308 } 1309 1310 mentrycount = 0; 1311 for (i = 0; i < nme; i++) { 1312 (void)memcpy(*ma + mentrycount, me[i].mp, 1313 me[i].cont_count * sizeof(**ma)); 1314 mentrycount += me[i].cont_count; 1315 } 1316 *nma = mentrycount; 1317 return 0; 1318 } 1319 1320 private void 1321 magic_entry_free(struct magic_entry *me, uint32_t nme) 1322 { 1323 uint32_t i; 1324 if (me == NULL) 1325 return; 1326 for (i = 0; i < nme; i++) 1327 free(me[i].mp); 1328 free(me); 1329 } 1330 1331 private struct magic_map * 1332 apprentice_load(struct magic_set *ms, const char *fn, int action) 1333 { 1334 int errs = 0; 1335 uint32_t i, j; 1336 size_t files = 0, maxfiles = 0; 1337 char **filearr = NULL, *mfn; 1338 struct stat st; 1339 struct magic_map *map; 1340 struct magic_entry_set mset[MAGIC_SETS]; 1341 DIR *dir; 1342 struct dirent *d; 1343 1344 memset(mset, 0, sizeof(mset)); 1345 ms->flags |= MAGIC_CHECK; /* Enable checks for parsed files */ 1346 1347 1348 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) 1349 { 1350 file_oomem(ms, sizeof(*map)); 1351 return NULL; 1352 } 1353 map->type = MAP_TYPE_MALLOC; 1354 1355 /* print silly verbose header for USG compat. */ 1356 if (action == FILE_CHECK) 1357 (void)fprintf(stderr, "%s\n", usg_hdr); 1358 1359 /* load directory or file */ 1360 if (stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) { 1361 dir = opendir(fn); 1362 if (!dir) { 1363 errs++; 1364 goto out; 1365 } 1366 while ((d = readdir(dir)) != NULL) { 1367 if (d->d_name[0] == '.') 1368 continue; 1369 if (asprintf(&mfn, "%s/%s", fn, d->d_name) < 0) { 1370 file_oomem(ms, 1371 strlen(fn) + strlen(d->d_name) + 2); 1372 errs++; 1373 closedir(dir); 1374 goto out; 1375 } 1376 if (stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) { 1377 free(mfn); 1378 continue; 1379 } 1380 if (files >= maxfiles) { 1381 size_t mlen; 1382 maxfiles = (maxfiles + 1) * 2; 1383 mlen = maxfiles * sizeof(*filearr); 1384 if ((filearr = CAST(char **, 1385 realloc(filearr, mlen))) == NULL) { 1386 file_oomem(ms, mlen); 1387 free(mfn); 1388 closedir(dir); 1389 errs++; 1390 goto out; 1391 } 1392 } 1393 filearr[files++] = mfn; 1394 } 1395 closedir(dir); 1396 qsort(filearr, files, sizeof(*filearr), cmpstrp); 1397 for (i = 0; i < files; i++) { 1398 load_1(ms, action, filearr[i], &errs, mset); 1399 free(filearr[i]); 1400 } 1401 free(filearr); 1402 } else 1403 load_1(ms, action, fn, &errs, mset); 1404 if (errs) 1405 goto out; 1406 1407 for (j = 0; j < MAGIC_SETS; j++) { 1408 /* Set types of tests */ 1409 for (i = 0; i < mset[j].count; ) { 1410 if (mset[j].me[i].mp->cont_level != 0) { 1411 i++; 1412 continue; 1413 } 1414 i = set_text_binary(ms, mset[j].me, mset[j].count, i); 1415 } 1416 if (mset[j].me) 1417 qsort(mset[j].me, mset[j].count, sizeof(*mset[j].me), 1418 apprentice_sort); 1419 1420 /* 1421 * Make sure that any level 0 "default" line is last 1422 * (if one exists). 1423 */ 1424 set_last_default(ms, mset[j].me, mset[j].count); 1425 1426 /* coalesce per file arrays into a single one */ 1427 if (coalesce_entries(ms, mset[j].me, mset[j].count, 1428 &map->magic[j], &map->nmagic[j]) == -1) { 1429 errs++; 1430 goto out; 1431 } 1432 } 1433 1434 out: 1435 for (j = 0; j < MAGIC_SETS; j++) 1436 magic_entry_free(mset[j].me, mset[j].count); 1437 1438 if (errs) { 1439 apprentice_unmap(map); 1440 return NULL; 1441 } 1442 return map; 1443 } 1444 1445 /* 1446 * extend the sign bit if the comparison is to be signed 1447 */ 1448 protected uint64_t 1449 file_signextend(struct magic_set *ms, struct magic *m, uint64_t v) 1450 { 1451 if (!(m->flag & UNSIGNED)) { 1452 switch(m->type) { 1453 /* 1454 * Do not remove the casts below. They are 1455 * vital. When later compared with the data, 1456 * the sign extension must have happened. 1457 */ 1458 case FILE_BYTE: 1459 v = (signed char) v; 1460 break; 1461 case FILE_SHORT: 1462 case FILE_BESHORT: 1463 case FILE_LESHORT: 1464 v = (short) v; 1465 break; 1466 case FILE_DATE: 1467 case FILE_BEDATE: 1468 case FILE_LEDATE: 1469 case FILE_MEDATE: 1470 case FILE_LDATE: 1471 case FILE_BELDATE: 1472 case FILE_LELDATE: 1473 case FILE_MELDATE: 1474 case FILE_LONG: 1475 case FILE_BELONG: 1476 case FILE_LELONG: 1477 case FILE_MELONG: 1478 case FILE_FLOAT: 1479 case FILE_BEFLOAT: 1480 case FILE_LEFLOAT: 1481 v = (int32_t) v; 1482 break; 1483 case FILE_QUAD: 1484 case FILE_BEQUAD: 1485 case FILE_LEQUAD: 1486 case FILE_QDATE: 1487 case FILE_QLDATE: 1488 case FILE_QWDATE: 1489 case FILE_BEQDATE: 1490 case FILE_BEQLDATE: 1491 case FILE_BEQWDATE: 1492 case FILE_LEQDATE: 1493 case FILE_LEQLDATE: 1494 case FILE_LEQWDATE: 1495 case FILE_DOUBLE: 1496 case FILE_BEDOUBLE: 1497 case FILE_LEDOUBLE: 1498 v = (int64_t) v; 1499 break; 1500 case FILE_STRING: 1501 case FILE_PSTRING: 1502 case FILE_BESTRING16: 1503 case FILE_LESTRING16: 1504 case FILE_REGEX: 1505 case FILE_SEARCH: 1506 case FILE_DEFAULT: 1507 case FILE_INDIRECT: 1508 case FILE_NAME: 1509 case FILE_USE: 1510 case FILE_CLEAR: 1511 case FILE_DER: 1512 break; 1513 default: 1514 if (ms->flags & MAGIC_CHECK) 1515 file_magwarn(ms, "cannot happen: m->type=%d\n", 1516 m->type); 1517 return ~0U; 1518 } 1519 } 1520 return v; 1521 } 1522 1523 private int 1524 string_modifier_check(struct magic_set *ms, struct magic *m) 1525 { 1526 if ((ms->flags & MAGIC_CHECK) == 0) 1527 return 0; 1528 1529 if ((m->type != FILE_REGEX || (m->str_flags & REGEX_LINE_COUNT) == 0) && 1530 (m->type != FILE_PSTRING && (m->str_flags & PSTRING_LEN) != 0)) { 1531 file_magwarn(ms, 1532 "'/BHhLl' modifiers are only allowed for pascal strings\n"); 1533 return -1; 1534 } 1535 switch (m->type) { 1536 case FILE_BESTRING16: 1537 case FILE_LESTRING16: 1538 if (m->str_flags != 0) { 1539 file_magwarn(ms, 1540 "no modifiers allowed for 16-bit strings\n"); 1541 return -1; 1542 } 1543 break; 1544 case FILE_STRING: 1545 case FILE_PSTRING: 1546 if ((m->str_flags & REGEX_OFFSET_START) != 0) { 1547 file_magwarn(ms, 1548 "'/%c' only allowed on regex and search\n", 1549 CHAR_REGEX_OFFSET_START); 1550 return -1; 1551 } 1552 break; 1553 case FILE_SEARCH: 1554 if (m->str_range == 0) { 1555 file_magwarn(ms, 1556 "missing range; defaulting to %d\n", 1557 STRING_DEFAULT_RANGE); 1558 m->str_range = STRING_DEFAULT_RANGE; 1559 return -1; 1560 } 1561 break; 1562 case FILE_REGEX: 1563 if ((m->str_flags & STRING_COMPACT_WHITESPACE) != 0) { 1564 file_magwarn(ms, "'/%c' not allowed on regex\n", 1565 CHAR_COMPACT_WHITESPACE); 1566 return -1; 1567 } 1568 if ((m->str_flags & STRING_COMPACT_OPTIONAL_WHITESPACE) != 0) { 1569 file_magwarn(ms, "'/%c' not allowed on regex\n", 1570 CHAR_COMPACT_OPTIONAL_WHITESPACE); 1571 return -1; 1572 } 1573 break; 1574 default: 1575 file_magwarn(ms, "coding error: m->type=%d\n", 1576 m->type); 1577 return -1; 1578 } 1579 return 0; 1580 } 1581 1582 private int 1583 get_op(char c) 1584 { 1585 switch (c) { 1586 case '&': 1587 return FILE_OPAND; 1588 case '|': 1589 return FILE_OPOR; 1590 case '^': 1591 return FILE_OPXOR; 1592 case '+': 1593 return FILE_OPADD; 1594 case '-': 1595 return FILE_OPMINUS; 1596 case '*': 1597 return FILE_OPMULTIPLY; 1598 case '/': 1599 return FILE_OPDIVIDE; 1600 case '%': 1601 return FILE_OPMODULO; 1602 default: 1603 return -1; 1604 } 1605 } 1606 1607 #ifdef ENABLE_CONDITIONALS 1608 private int 1609 get_cond(const char *l, const char **t) 1610 { 1611 static const struct cond_tbl_s { 1612 char name[8]; 1613 size_t len; 1614 int cond; 1615 } cond_tbl[] = { 1616 { "if", 2, COND_IF }, 1617 { "elif", 4, COND_ELIF }, 1618 { "else", 4, COND_ELSE }, 1619 { "", 0, COND_NONE }, 1620 }; 1621 const struct cond_tbl_s *p; 1622 1623 for (p = cond_tbl; p->len; p++) { 1624 if (strncmp(l, p->name, p->len) == 0 && 1625 isspace((unsigned char)l[p->len])) { 1626 if (t) 1627 *t = l + p->len; 1628 break; 1629 } 1630 } 1631 return p->cond; 1632 } 1633 1634 private int 1635 check_cond(struct magic_set *ms, int cond, uint32_t cont_level) 1636 { 1637 int last_cond; 1638 last_cond = ms->c.li[cont_level].last_cond; 1639 1640 switch (cond) { 1641 case COND_IF: 1642 if (last_cond != COND_NONE && last_cond != COND_ELIF) { 1643 if (ms->flags & MAGIC_CHECK) 1644 file_magwarn(ms, "syntax error: `if'"); 1645 return -1; 1646 } 1647 last_cond = COND_IF; 1648 break; 1649 1650 case COND_ELIF: 1651 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1652 if (ms->flags & MAGIC_CHECK) 1653 file_magwarn(ms, "syntax error: `elif'"); 1654 return -1; 1655 } 1656 last_cond = COND_ELIF; 1657 break; 1658 1659 case COND_ELSE: 1660 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1661 if (ms->flags & MAGIC_CHECK) 1662 file_magwarn(ms, "syntax error: `else'"); 1663 return -1; 1664 } 1665 last_cond = COND_NONE; 1666 break; 1667 1668 case COND_NONE: 1669 last_cond = COND_NONE; 1670 break; 1671 } 1672 1673 ms->c.li[cont_level].last_cond = last_cond; 1674 return 0; 1675 } 1676 #endif /* ENABLE_CONDITIONALS */ 1677 1678 private int 1679 parse_indirect_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1680 { 1681 const char *l = *lp; 1682 1683 while (!isspace((unsigned char)*++l)) 1684 switch (*l) { 1685 case CHAR_INDIRECT_RELATIVE: 1686 m->str_flags |= INDIRECT_RELATIVE; 1687 break; 1688 default: 1689 if (ms->flags & MAGIC_CHECK) 1690 file_magwarn(ms, "indirect modifier `%c' " 1691 "invalid", *l); 1692 *lp = l; 1693 return -1; 1694 } 1695 *lp = l; 1696 return 0; 1697 } 1698 1699 private void 1700 parse_op_modifier(struct magic_set *ms, struct magic *m, const char **lp, 1701 int op) 1702 { 1703 const char *l = *lp; 1704 char *t; 1705 uint64_t val; 1706 1707 ++l; 1708 m->mask_op |= op; 1709 val = (uint64_t)strtoull(l, &t, 0); 1710 l = t; 1711 m->num_mask = file_signextend(ms, m, val); 1712 eatsize(&l); 1713 *lp = l; 1714 } 1715 1716 private int 1717 parse_string_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1718 { 1719 const char *l = *lp; 1720 char *t; 1721 int have_range = 0; 1722 1723 while (!isspace((unsigned char)*++l)) { 1724 switch (*l) { 1725 case '0': case '1': case '2': 1726 case '3': case '4': case '5': 1727 case '6': case '7': case '8': 1728 case '9': 1729 if (have_range && (ms->flags & MAGIC_CHECK)) 1730 file_magwarn(ms, "multiple ranges"); 1731 have_range = 1; 1732 m->str_range = CAST(uint32_t, strtoul(l, &t, 0)); 1733 if (m->str_range == 0) 1734 file_magwarn(ms, "zero range"); 1735 l = t - 1; 1736 break; 1737 case CHAR_COMPACT_WHITESPACE: 1738 m->str_flags |= STRING_COMPACT_WHITESPACE; 1739 break; 1740 case CHAR_COMPACT_OPTIONAL_WHITESPACE: 1741 m->str_flags |= STRING_COMPACT_OPTIONAL_WHITESPACE; 1742 break; 1743 case CHAR_IGNORE_LOWERCASE: 1744 m->str_flags |= STRING_IGNORE_LOWERCASE; 1745 break; 1746 case CHAR_IGNORE_UPPERCASE: 1747 m->str_flags |= STRING_IGNORE_UPPERCASE; 1748 break; 1749 case CHAR_REGEX_OFFSET_START: 1750 m->str_flags |= REGEX_OFFSET_START; 1751 break; 1752 case CHAR_BINTEST: 1753 m->str_flags |= STRING_BINTEST; 1754 break; 1755 case CHAR_TEXTTEST: 1756 m->str_flags |= STRING_TEXTTEST; 1757 break; 1758 case CHAR_TRIM: 1759 m->str_flags |= STRING_TRIM; 1760 break; 1761 case CHAR_PSTRING_1_LE: 1762 #define SET_LENGTH(a) m->str_flags = (m->str_flags & ~PSTRING_LEN) | (a) 1763 if (m->type != FILE_PSTRING) 1764 goto bad; 1765 SET_LENGTH(PSTRING_1_LE); 1766 break; 1767 case CHAR_PSTRING_2_BE: 1768 if (m->type != FILE_PSTRING) 1769 goto bad; 1770 SET_LENGTH(PSTRING_2_BE); 1771 break; 1772 case CHAR_PSTRING_2_LE: 1773 if (m->type != FILE_PSTRING) 1774 goto bad; 1775 SET_LENGTH(PSTRING_2_LE); 1776 break; 1777 case CHAR_PSTRING_4_BE: 1778 if (m->type != FILE_PSTRING) 1779 goto bad; 1780 SET_LENGTH(PSTRING_4_BE); 1781 break; 1782 case CHAR_PSTRING_4_LE: 1783 switch (m->type) { 1784 case FILE_PSTRING: 1785 case FILE_REGEX: 1786 break; 1787 default: 1788 goto bad; 1789 } 1790 SET_LENGTH(PSTRING_4_LE); 1791 break; 1792 case CHAR_PSTRING_LENGTH_INCLUDES_ITSELF: 1793 if (m->type != FILE_PSTRING) 1794 goto bad; 1795 m->str_flags |= PSTRING_LENGTH_INCLUDES_ITSELF; 1796 break; 1797 default: 1798 bad: 1799 if (ms->flags & MAGIC_CHECK) 1800 file_magwarn(ms, "string modifier `%c' " 1801 "invalid", *l); 1802 goto out; 1803 } 1804 /* allow multiple '/' for readability */ 1805 if (l[1] == '/' && !isspace((unsigned char)l[2])) 1806 l++; 1807 } 1808 if (string_modifier_check(ms, m) == -1) 1809 goto out; 1810 *lp = l; 1811 return 0; 1812 out: 1813 *lp = l; 1814 return -1; 1815 } 1816 1817 /* 1818 * parse one line from magic file, put into magic[index++] if valid 1819 */ 1820 private int 1821 parse(struct magic_set *ms, struct magic_entry *me, const char *line, 1822 size_t lineno, int action) 1823 { 1824 #ifdef ENABLE_CONDITIONALS 1825 static uint32_t last_cont_level = 0; 1826 #endif 1827 size_t i; 1828 struct magic *m; 1829 const char *l = line; 1830 char *t; 1831 int op; 1832 uint32_t cont_level; 1833 int32_t diff; 1834 1835 cont_level = 0; 1836 1837 /* 1838 * Parse the offset. 1839 */ 1840 while (*l == '>') { 1841 ++l; /* step over */ 1842 cont_level++; 1843 } 1844 #ifdef ENABLE_CONDITIONALS 1845 if (cont_level == 0 || cont_level > last_cont_level) 1846 if (file_check_mem(ms, cont_level) == -1) 1847 return -1; 1848 last_cont_level = cont_level; 1849 #endif 1850 if (cont_level != 0) { 1851 if (me->mp == NULL) { 1852 file_magerror(ms, "No current entry for continuation"); 1853 return -1; 1854 } 1855 if (me->cont_count == 0) { 1856 file_magerror(ms, "Continuations present with 0 count"); 1857 return -1; 1858 } 1859 m = &me->mp[me->cont_count - 1]; 1860 diff = (int32_t)cont_level - (int32_t)m->cont_level; 1861 if (diff > 1) 1862 file_magwarn(ms, "New continuation level %u is more " 1863 "than one larger than current level %u", cont_level, 1864 m->cont_level); 1865 if (me->cont_count == me->max_count) { 1866 struct magic *nm; 1867 size_t cnt = me->max_count + ALLOC_CHUNK; 1868 if ((nm = CAST(struct magic *, realloc(me->mp, 1869 sizeof(*nm) * cnt))) == NULL) { 1870 file_oomem(ms, sizeof(*nm) * cnt); 1871 return -1; 1872 } 1873 me->mp = m = nm; 1874 me->max_count = CAST(uint32_t, cnt); 1875 } 1876 m = &me->mp[me->cont_count++]; 1877 (void)memset(m, 0, sizeof(*m)); 1878 m->cont_level = cont_level; 1879 } else { 1880 static const size_t len = sizeof(*m) * ALLOC_CHUNK; 1881 if (me->mp != NULL) 1882 return 1; 1883 if ((m = CAST(struct magic *, malloc(len))) == NULL) { 1884 file_oomem(ms, len); 1885 return -1; 1886 } 1887 me->mp = m; 1888 me->max_count = ALLOC_CHUNK; 1889 (void)memset(m, 0, sizeof(*m)); 1890 m->factor_op = FILE_FACTOR_OP_NONE; 1891 m->cont_level = 0; 1892 me->cont_count = 1; 1893 } 1894 m->lineno = CAST(uint32_t, lineno); 1895 1896 if (*l == '&') { /* m->cont_level == 0 checked below. */ 1897 ++l; /* step over */ 1898 m->flag |= OFFADD; 1899 } 1900 if (*l == '(') { 1901 ++l; /* step over */ 1902 m->flag |= INDIR; 1903 if (m->flag & OFFADD) 1904 m->flag = (m->flag & ~OFFADD) | INDIROFFADD; 1905 1906 if (*l == '&') { /* m->cont_level == 0 checked below */ 1907 ++l; /* step over */ 1908 m->flag |= OFFADD; 1909 } 1910 } 1911 /* Indirect offsets are not valid at level 0. */ 1912 if (m->cont_level == 0 && (m->flag & (OFFADD | INDIROFFADD))) { 1913 if (ms->flags & MAGIC_CHECK) 1914 file_magwarn(ms, "relative offset at level 0"); 1915 return -1; 1916 } 1917 1918 /* get offset, then skip over it */ 1919 m->offset = (int32_t)strtol(l, &t, 0); 1920 if (l == t) { 1921 if (ms->flags & MAGIC_CHECK) 1922 file_magwarn(ms, "offset `%s' invalid", l); 1923 return -1; 1924 } 1925 #if 0 1926 if (m->offset < 0 && cont_level != 0 && 1927 (m->flag & (OFFADD | INDIROFFADD)) == 0) { 1928 if (ms->flags & MAGIC_CHECK) { 1929 file_magwarn(ms, 1930 "negative direct offset `%s' at level %u", 1931 l, cont_level); 1932 } 1933 return -1; 1934 } 1935 #endif 1936 l = t; 1937 1938 if (m->flag & INDIR) { 1939 m->in_type = FILE_LONG; 1940 m->in_offset = 0; 1941 m->in_op = 0; 1942 /* 1943 * read [.,lbs][+-]nnnnn) 1944 */ 1945 if (*l == '.' || *l == ',') { 1946 if (*l == ',') 1947 m->in_op |= FILE_OPSIGNED; 1948 l++; 1949 switch (*l) { 1950 case 'l': 1951 m->in_type = FILE_LELONG; 1952 break; 1953 case 'L': 1954 m->in_type = FILE_BELONG; 1955 break; 1956 case 'm': 1957 m->in_type = FILE_MELONG; 1958 break; 1959 case 'h': 1960 case 's': 1961 m->in_type = FILE_LESHORT; 1962 break; 1963 case 'H': 1964 case 'S': 1965 m->in_type = FILE_BESHORT; 1966 break; 1967 case 'c': 1968 case 'b': 1969 case 'C': 1970 case 'B': 1971 m->in_type = FILE_BYTE; 1972 break; 1973 case 'e': 1974 case 'f': 1975 case 'g': 1976 m->in_type = FILE_LEDOUBLE; 1977 break; 1978 case 'E': 1979 case 'F': 1980 case 'G': 1981 m->in_type = FILE_BEDOUBLE; 1982 break; 1983 case 'i': 1984 m->in_type = FILE_LEID3; 1985 break; 1986 case 'I': 1987 m->in_type = FILE_BEID3; 1988 break; 1989 default: 1990 if (ms->flags & MAGIC_CHECK) 1991 file_magwarn(ms, 1992 "indirect offset type `%c' invalid", 1993 *l); 1994 return -1; 1995 } 1996 l++; 1997 } 1998 1999 if (*l == '~') { 2000 m->in_op |= FILE_OPINVERSE; 2001 l++; 2002 } 2003 if ((op = get_op(*l)) != -1) { 2004 m->in_op |= op; 2005 l++; 2006 } 2007 if (*l == '(') { 2008 m->in_op |= FILE_OPINDIRECT; 2009 l++; 2010 } 2011 if (isdigit((unsigned char)*l) || *l == '-') { 2012 m->in_offset = (int32_t)strtol(l, &t, 0); 2013 if (l == t) { 2014 if (ms->flags & MAGIC_CHECK) 2015 file_magwarn(ms, 2016 "in_offset `%s' invalid", l); 2017 return -1; 2018 } 2019 l = t; 2020 } 2021 if (*l++ != ')' || 2022 ((m->in_op & FILE_OPINDIRECT) && *l++ != ')')) { 2023 if (ms->flags & MAGIC_CHECK) 2024 file_magwarn(ms, 2025 "missing ')' in indirect offset"); 2026 return -1; 2027 } 2028 } 2029 EATAB; 2030 2031 #ifdef ENABLE_CONDITIONALS 2032 m->cond = get_cond(l, &l); 2033 if (check_cond(ms, m->cond, cont_level) == -1) 2034 return -1; 2035 2036 EATAB; 2037 #endif 2038 2039 /* 2040 * Parse the type. 2041 */ 2042 if (*l == 'u') { 2043 /* 2044 * Try it as a keyword type prefixed by "u"; match what 2045 * follows the "u". If that fails, try it as an SUS 2046 * integer type. 2047 */ 2048 m->type = get_type(type_tbl, l + 1, &l); 2049 if (m->type == FILE_INVALID) { 2050 /* 2051 * Not a keyword type; parse it as an SUS type, 2052 * 'u' possibly followed by a number or C/S/L. 2053 */ 2054 m->type = get_standard_integer_type(l, &l); 2055 } 2056 /* It's unsigned. */ 2057 if (m->type != FILE_INVALID) 2058 m->flag |= UNSIGNED; 2059 } else { 2060 /* 2061 * Try it as a keyword type. If that fails, try it as 2062 * an SUS integer type if it begins with "d" or as an 2063 * SUS string type if it begins with "s". In any case, 2064 * it's not unsigned. 2065 */ 2066 m->type = get_type(type_tbl, l, &l); 2067 if (m->type == FILE_INVALID) { 2068 /* 2069 * Not a keyword type; parse it as an SUS type, 2070 * either 'd' possibly followed by a number or 2071 * C/S/L, or just 's'. 2072 */ 2073 if (*l == 'd') 2074 m->type = get_standard_integer_type(l, &l); 2075 else if (*l == 's' && !isalpha((unsigned char)l[1])) { 2076 m->type = FILE_STRING; 2077 ++l; 2078 } 2079 } 2080 } 2081 2082 if (m->type == FILE_INVALID) { 2083 /* Not found - try it as a special keyword. */ 2084 m->type = get_type(special_tbl, l, &l); 2085 } 2086 2087 if (m->type == FILE_INVALID) { 2088 if (ms->flags & MAGIC_CHECK) 2089 file_magwarn(ms, "type `%s' invalid", l); 2090 return -1; 2091 } 2092 2093 /* New-style anding: "0 byte&0x80 =0x80 dynamically linked" */ 2094 /* New and improved: ~ & | ^ + - * / % -- exciting, isn't it? */ 2095 2096 m->mask_op = 0; 2097 if (*l == '~') { 2098 if (!IS_STRING(m->type)) 2099 m->mask_op |= FILE_OPINVERSE; 2100 else if (ms->flags & MAGIC_CHECK) 2101 file_magwarn(ms, "'~' invalid for string types"); 2102 ++l; 2103 } 2104 m->str_range = 0; 2105 m->str_flags = m->type == FILE_PSTRING ? PSTRING_1_LE : 0; 2106 if ((op = get_op(*l)) != -1) { 2107 if (IS_STRING(m->type)) { 2108 int r; 2109 2110 if (op != FILE_OPDIVIDE) { 2111 if (ms->flags & MAGIC_CHECK) 2112 file_magwarn(ms, 2113 "invalid string/indirect op: " 2114 "`%c'", *t); 2115 return -1; 2116 } 2117 2118 if (m->type == FILE_INDIRECT) 2119 r = parse_indirect_modifier(ms, m, &l); 2120 else 2121 r = parse_string_modifier(ms, m, &l); 2122 if (r == -1) 2123 return -1; 2124 } else 2125 parse_op_modifier(ms, m, &l, op); 2126 } 2127 2128 /* 2129 * We used to set mask to all 1's here, instead let's just not do 2130 * anything if mask = 0 (unless you have a better idea) 2131 */ 2132 EATAB; 2133 2134 switch (*l) { 2135 case '>': 2136 case '<': 2137 m->reln = *l; 2138 ++l; 2139 if (*l == '=') { 2140 if (ms->flags & MAGIC_CHECK) { 2141 file_magwarn(ms, "%c= not supported", 2142 m->reln); 2143 return -1; 2144 } 2145 ++l; 2146 } 2147 break; 2148 /* Old-style anding: "0 byte &0x80 dynamically linked" */ 2149 case '&': 2150 case '^': 2151 case '=': 2152 m->reln = *l; 2153 ++l; 2154 if (*l == '=') { 2155 /* HP compat: ignore &= etc. */ 2156 ++l; 2157 } 2158 break; 2159 case '!': 2160 m->reln = *l; 2161 ++l; 2162 break; 2163 default: 2164 m->reln = '='; /* the default relation */ 2165 if (*l == 'x' && ((isascii((unsigned char)l[1]) && 2166 isspace((unsigned char)l[1])) || !l[1])) { 2167 m->reln = *l; 2168 ++l; 2169 } 2170 break; 2171 } 2172 /* 2173 * Grab the value part, except for an 'x' reln. 2174 */ 2175 if (m->reln != 'x' && getvalue(ms, m, &l, action)) 2176 return -1; 2177 2178 /* 2179 * TODO finish this macro and start using it! 2180 * #define offsetcheck {if (offset > ms->bytes_max -1) 2181 * magwarn("offset too big"); } 2182 */ 2183 2184 /* 2185 * Now get last part - the description 2186 */ 2187 EATAB; 2188 if (l[0] == '\b') { 2189 ++l; 2190 m->flag |= NOSPACE; 2191 } else if ((l[0] == '\\') && (l[1] == 'b')) { 2192 ++l; 2193 ++l; 2194 m->flag |= NOSPACE; 2195 } 2196 for (i = 0; (m->desc[i++] = *l++) != '\0' && i < sizeof(m->desc); ) 2197 continue; 2198 if (i == sizeof(m->desc)) { 2199 m->desc[sizeof(m->desc) - 1] = '\0'; 2200 if (ms->flags & MAGIC_CHECK) 2201 file_magwarn(ms, "description `%s' truncated", m->desc); 2202 } 2203 2204 /* 2205 * We only do this check while compiling, or if any of the magic 2206 * files were not compiled. 2207 */ 2208 if (ms->flags & MAGIC_CHECK) { 2209 if (check_format(ms, m) == -1) 2210 return -1; 2211 } 2212 #ifndef COMPILE_ONLY 2213 if (action == FILE_CHECK) { 2214 file_mdump(m); 2215 } 2216 #endif 2217 m->mimetype[0] = '\0'; /* initialise MIME type to none */ 2218 return 0; 2219 } 2220 2221 /* 2222 * parse a STRENGTH annotation line from magic file, put into magic[index - 1] 2223 * if valid 2224 */ 2225 private int 2226 parse_strength(struct magic_set *ms, struct magic_entry *me, const char *line) 2227 { 2228 const char *l = line; 2229 char *el; 2230 unsigned long factor; 2231 struct magic *m = &me->mp[0]; 2232 2233 if (m->factor_op != FILE_FACTOR_OP_NONE) { 2234 file_magwarn(ms, 2235 "Current entry already has a strength type: %c %d", 2236 m->factor_op, m->factor); 2237 return -1; 2238 } 2239 if (m->type == FILE_NAME) { 2240 file_magwarn(ms, "%s: Strength setting is not supported in " 2241 "\"name\" magic entries", m->value.s); 2242 return -1; 2243 } 2244 EATAB; 2245 switch (*l) { 2246 case FILE_FACTOR_OP_NONE: 2247 case FILE_FACTOR_OP_PLUS: 2248 case FILE_FACTOR_OP_MINUS: 2249 case FILE_FACTOR_OP_TIMES: 2250 case FILE_FACTOR_OP_DIV: 2251 m->factor_op = *l++; 2252 break; 2253 default: 2254 file_magwarn(ms, "Unknown factor op `%c'", *l); 2255 return -1; 2256 } 2257 EATAB; 2258 factor = strtoul(l, &el, 0); 2259 if (factor > 255) { 2260 file_magwarn(ms, "Too large factor `%lu'", factor); 2261 goto out; 2262 } 2263 if (*el && !isspace((unsigned char)*el)) { 2264 file_magwarn(ms, "Bad factor `%s'", l); 2265 goto out; 2266 } 2267 m->factor = (uint8_t)factor; 2268 if (m->factor == 0 && m->factor_op == FILE_FACTOR_OP_DIV) { 2269 file_magwarn(ms, "Cannot have factor op `%c' and factor %u", 2270 m->factor_op, m->factor); 2271 goto out; 2272 } 2273 return 0; 2274 out: 2275 m->factor_op = FILE_FACTOR_OP_NONE; 2276 m->factor = 0; 2277 return -1; 2278 } 2279 2280 private int 2281 goodchar(unsigned char x, const char *extra) 2282 { 2283 return (isascii(x) && isalnum(x)) || strchr(extra, x); 2284 } 2285 2286 private int 2287 parse_extra(struct magic_set *ms, struct magic_entry *me, const char *line, 2288 off_t off, size_t len, const char *name, const char *extra, int nt) 2289 { 2290 size_t i; 2291 const char *l = line; 2292 struct magic *m = &me->mp[me->cont_count == 0 ? 0 : me->cont_count - 1]; 2293 char *buf = CAST(char *, CAST(void *, m)) + off; 2294 2295 if (buf[0] != '\0') { 2296 len = nt ? strlen(buf) : len; 2297 file_magwarn(ms, "Current entry already has a %s type " 2298 "`%.*s', new type `%s'", name, (int)len, buf, l); 2299 return -1; 2300 } 2301 2302 if (*m->desc == '\0') { 2303 file_magwarn(ms, "Current entry does not yet have a " 2304 "description for adding a %s type", name); 2305 return -1; 2306 } 2307 2308 EATAB; 2309 for (i = 0; *l && i < len && goodchar(*l, extra); buf[i++] = *l++) 2310 continue; 2311 2312 if (i == len && *l) { 2313 if (nt) 2314 buf[len - 1] = '\0'; 2315 if (ms->flags & MAGIC_CHECK) 2316 file_magwarn(ms, "%s type `%s' truncated %" 2317 SIZE_T_FORMAT "u", name, line, i); 2318 } else { 2319 if (!isspace((unsigned char)*l) && !goodchar(*l, extra)) 2320 file_magwarn(ms, "%s type `%s' has bad char '%c'", 2321 name, line, *l); 2322 if (nt) 2323 buf[i] = '\0'; 2324 } 2325 2326 if (i > 0) 2327 return 0; 2328 2329 file_magerror(ms, "Bad magic entry '%s'", line); 2330 return -1; 2331 } 2332 2333 /* 2334 * Parse an Apple CREATOR/TYPE annotation from magic file and put it into 2335 * magic[index - 1] 2336 */ 2337 private int 2338 parse_apple(struct magic_set *ms, struct magic_entry *me, const char *line) 2339 { 2340 struct magic *m = &me->mp[0]; 2341 2342 return parse_extra(ms, me, line, 2343 CAST(off_t, offsetof(struct magic, apple)), 2344 sizeof(m->apple), "APPLE", "!+-./?", 0); 2345 } 2346 2347 /* 2348 * Parse a comma-separated list of extensions 2349 */ 2350 private int 2351 parse_ext(struct magic_set *ms, struct magic_entry *me, const char *line) 2352 { 2353 struct magic *m = &me->mp[0]; 2354 2355 return parse_extra(ms, me, line, 2356 CAST(off_t, offsetof(struct magic, ext)), 2357 sizeof(m->ext), "EXTENSION", ",!+-/@?_$", 0); 2358 } 2359 2360 /* 2361 * parse a MIME annotation line from magic file, put into magic[index - 1] 2362 * if valid 2363 */ 2364 private int 2365 parse_mime(struct magic_set *ms, struct magic_entry *me, const char *line) 2366 { 2367 struct magic *m = &me->mp[0]; 2368 2369 return parse_extra(ms, me, line, 2370 CAST(off_t, offsetof(struct magic, mimetype)), 2371 sizeof(m->mimetype), "MIME", "+-/.$?:{}", 1); 2372 } 2373 2374 private int 2375 check_format_type(const char *ptr, int type, const char **estr) 2376 { 2377 int quad = 0, h; 2378 size_t len, cnt; 2379 if (*ptr == '\0') { 2380 /* Missing format string; bad */ 2381 *estr = "missing format spec"; 2382 return -1; 2383 } 2384 2385 switch (file_formats[type]) { 2386 case FILE_FMT_QUAD: 2387 quad = 1; 2388 /*FALLTHROUGH*/ 2389 case FILE_FMT_NUM: 2390 if (quad == 0) { 2391 switch (type) { 2392 case FILE_BYTE: 2393 h = 2; 2394 break; 2395 case FILE_SHORT: 2396 case FILE_BESHORT: 2397 case FILE_LESHORT: 2398 h = 1; 2399 break; 2400 case FILE_LONG: 2401 case FILE_BELONG: 2402 case FILE_LELONG: 2403 case FILE_MELONG: 2404 case FILE_LEID3: 2405 case FILE_BEID3: 2406 case FILE_INDIRECT: 2407 h = 0; 2408 break; 2409 default: 2410 abort(); 2411 } 2412 } else 2413 h = 0; 2414 if (*ptr == '-') 2415 ptr++; 2416 if (*ptr == '.') 2417 ptr++; 2418 if (*ptr == '#') 2419 ptr++; 2420 #define CHECKLEN() do { \ 2421 for (len = cnt = 0; isdigit((unsigned char)*ptr); ptr++, cnt++) \ 2422 len = len * 10 + (*ptr - '0'); \ 2423 if (cnt > 5 || len > 1024) \ 2424 goto toolong; \ 2425 } while (/*CONSTCOND*/0) 2426 2427 CHECKLEN(); 2428 if (*ptr == '.') 2429 ptr++; 2430 CHECKLEN(); 2431 if (quad) { 2432 if (*ptr++ != 'l') 2433 goto invalid; 2434 if (*ptr++ != 'l') 2435 goto invalid; 2436 } 2437 2438 switch (*ptr++) { 2439 #ifdef STRICT_FORMAT /* "long" formats are int formats for us */ 2440 /* so don't accept the 'l' modifier */ 2441 case 'l': 2442 switch (*ptr++) { 2443 case 'i': 2444 case 'd': 2445 case 'u': 2446 case 'o': 2447 case 'x': 2448 case 'X': 2449 if (h == 0) 2450 return 0; 2451 /*FALLTHROUGH*/ 2452 default: 2453 goto invalid; 2454 } 2455 2456 /* 2457 * Don't accept h and hh modifiers. They make writing 2458 * magic entries more complicated, for very little benefit 2459 */ 2460 case 'h': 2461 if (h-- <= 0) 2462 goto invalid; 2463 switch (*ptr++) { 2464 case 'h': 2465 if (h-- <= 0) 2466 goto invalid; 2467 switch (*ptr++) { 2468 case 'i': 2469 case 'd': 2470 case 'u': 2471 case 'o': 2472 case 'x': 2473 case 'X': 2474 return 0; 2475 default: 2476 goto invalid; 2477 } 2478 case 'i': 2479 case 'd': 2480 case 'u': 2481 case 'o': 2482 case 'x': 2483 case 'X': 2484 if (h == 0) 2485 return 0; 2486 /*FALLTHROUGH*/ 2487 default: 2488 goto invalid; 2489 } 2490 #endif 2491 case 'c': 2492 if (h == 2) 2493 return 0; 2494 goto invalid; 2495 case 'i': 2496 case 'd': 2497 case 'u': 2498 case 'o': 2499 case 'x': 2500 case 'X': 2501 #ifdef STRICT_FORMAT 2502 if (h == 0) 2503 return 0; 2504 /*FALLTHROUGH*/ 2505 #else 2506 return 0; 2507 #endif 2508 default: 2509 goto invalid; 2510 } 2511 2512 case FILE_FMT_FLOAT: 2513 case FILE_FMT_DOUBLE: 2514 if (*ptr == '-') 2515 ptr++; 2516 if (*ptr == '.') 2517 ptr++; 2518 CHECKLEN(); 2519 if (*ptr == '.') 2520 ptr++; 2521 CHECKLEN(); 2522 switch (*ptr++) { 2523 case 'e': 2524 case 'E': 2525 case 'f': 2526 case 'F': 2527 case 'g': 2528 case 'G': 2529 return 0; 2530 2531 default: 2532 goto invalid; 2533 } 2534 2535 2536 case FILE_FMT_STR: 2537 if (*ptr == '-') 2538 ptr++; 2539 while (isdigit((unsigned char )*ptr)) 2540 ptr++; 2541 if (*ptr == '.') { 2542 ptr++; 2543 while (isdigit((unsigned char )*ptr)) 2544 ptr++; 2545 } 2546 2547 switch (*ptr++) { 2548 case 's': 2549 return 0; 2550 default: 2551 goto invalid; 2552 } 2553 2554 default: 2555 /* internal error */ 2556 abort(); 2557 } 2558 invalid: 2559 *estr = "not valid"; 2560 toolong: 2561 *estr = "too long"; 2562 return -1; 2563 } 2564 2565 /* 2566 * Check that the optional printf format in description matches 2567 * the type of the magic. 2568 */ 2569 private int 2570 check_format(struct magic_set *ms, struct magic *m) 2571 { 2572 char *ptr; 2573 const char *estr; 2574 2575 for (ptr = m->desc; *ptr; ptr++) 2576 if (*ptr == '%') 2577 break; 2578 if (*ptr == '\0') { 2579 /* No format string; ok */ 2580 return 1; 2581 } 2582 2583 assert(file_nformats == file_nnames); 2584 2585 if (m->type >= file_nformats) { 2586 file_magwarn(ms, "Internal error inconsistency between " 2587 "m->type and format strings"); 2588 return -1; 2589 } 2590 if (file_formats[m->type] == FILE_FMT_NONE) { 2591 file_magwarn(ms, "No format string for `%s' with description " 2592 "`%s'", m->desc, file_names[m->type]); 2593 return -1; 2594 } 2595 2596 ptr++; 2597 if (check_format_type(ptr, m->type, &estr) == -1) { 2598 /* 2599 * TODO: this error message is unhelpful if the format 2600 * string is not one character long 2601 */ 2602 file_magwarn(ms, "Printf format is %s for type " 2603 "`%s' in description `%s'", estr, 2604 file_names[m->type], m->desc); 2605 return -1; 2606 } 2607 2608 for (; *ptr; ptr++) { 2609 if (*ptr == '%') { 2610 file_magwarn(ms, 2611 "Too many format strings (should have at most one) " 2612 "for `%s' with description `%s'", 2613 file_names[m->type], m->desc); 2614 return -1; 2615 } 2616 } 2617 return 0; 2618 } 2619 2620 /* 2621 * Read a numeric value from a pointer, into the value union of a magic 2622 * pointer, according to the magic type. Update the string pointer to point 2623 * just after the number read. Return 0 for success, non-zero for failure. 2624 */ 2625 private int 2626 getvalue(struct magic_set *ms, struct magic *m, const char **p, int action) 2627 { 2628 char *ep; 2629 uint64_t ull; 2630 2631 switch (m->type) { 2632 case FILE_BESTRING16: 2633 case FILE_LESTRING16: 2634 case FILE_STRING: 2635 case FILE_PSTRING: 2636 case FILE_REGEX: 2637 case FILE_SEARCH: 2638 case FILE_NAME: 2639 case FILE_USE: 2640 case FILE_DER: 2641 *p = getstr(ms, m, *p, action == FILE_COMPILE); 2642 if (*p == NULL) { 2643 if (ms->flags & MAGIC_CHECK) 2644 file_magwarn(ms, "cannot get string from `%s'", 2645 m->value.s); 2646 return -1; 2647 } 2648 if (m->type == FILE_REGEX) { 2649 file_regex_t rx; 2650 int rc = file_regcomp(&rx, m->value.s, REG_EXTENDED); 2651 if (rc) { 2652 if (ms->flags & MAGIC_CHECK) 2653 file_regerror(&rx, rc, ms); 2654 } 2655 file_regfree(&rx); 2656 return rc ? -1 : 0; 2657 } 2658 return 0; 2659 default: 2660 if (m->reln == 'x') 2661 return 0; 2662 break; 2663 } 2664 2665 switch (m->type) { 2666 case FILE_FLOAT: 2667 case FILE_BEFLOAT: 2668 case FILE_LEFLOAT: 2669 errno = 0; 2670 #ifdef HAVE_STRTOF 2671 m->value.f = strtof(*p, &ep); 2672 #else 2673 m->value.f = (float)strtod(*p, &ep); 2674 #endif 2675 if (errno == 0) 2676 *p = ep; 2677 return 0; 2678 case FILE_DOUBLE: 2679 case FILE_BEDOUBLE: 2680 case FILE_LEDOUBLE: 2681 errno = 0; 2682 m->value.d = strtod(*p, &ep); 2683 if (errno == 0) 2684 *p = ep; 2685 return 0; 2686 default: 2687 errno = 0; 2688 ull = (uint64_t)strtoull(*p, &ep, 0); 2689 m->value.q = file_signextend(ms, m, ull); 2690 if (*p == ep) { 2691 file_magwarn(ms, "Unparseable number `%s'", *p); 2692 } else { 2693 size_t ts = typesize(m->type); 2694 uint64_t x; 2695 const char *q; 2696 2697 if (ts == (size_t)~0) { 2698 file_magwarn(ms, 2699 "Expected numeric type got `%s'", 2700 type_tbl[m->type].name); 2701 } 2702 for (q = *p; isspace((unsigned char)*q); q++) 2703 continue; 2704 if (*q == '-') 2705 ull = -(int64_t)ull; 2706 switch (ts) { 2707 case 1: 2708 x = (uint64_t)(ull & ~0xffULL); 2709 break; 2710 case 2: 2711 x = (uint64_t)(ull & ~0xffffULL); 2712 break; 2713 case 4: 2714 x = (uint64_t)(ull & ~0xffffffffULL); 2715 break; 2716 case 8: 2717 x = 0; 2718 break; 2719 default: 2720 abort(); 2721 } 2722 if (x) { 2723 file_magwarn(ms, "Overflow for numeric" 2724 " type `%s' value %#" PRIx64, 2725 type_tbl[m->type].name, ull); 2726 } 2727 } 2728 if (errno == 0) { 2729 *p = ep; 2730 eatsize(p); 2731 } 2732 return 0; 2733 } 2734 } 2735 2736 /* 2737 * Convert a string containing C character escapes. Stop at an unescaped 2738 * space or tab. 2739 * Copy the converted version to "m->value.s", and the length in m->vallen. 2740 * Return updated scan pointer as function result. Warn if set. 2741 */ 2742 private const char * 2743 getstr(struct magic_set *ms, struct magic *m, const char *s, int warn) 2744 { 2745 const char *origs = s; 2746 char *p = m->value.s; 2747 size_t plen = sizeof(m->value.s); 2748 char *origp = p; 2749 char *pmax = p + plen - 1; 2750 int c; 2751 int val; 2752 2753 while ((c = *s++) != '\0') { 2754 if (isspace((unsigned char) c)) 2755 break; 2756 if (p >= pmax) { 2757 file_error(ms, 0, "string too long: `%s'", origs); 2758 return NULL; 2759 } 2760 if (c == '\\') { 2761 switch(c = *s++) { 2762 2763 case '\0': 2764 if (warn) 2765 file_magwarn(ms, "incomplete escape"); 2766 s--; 2767 goto out; 2768 2769 case '\t': 2770 if (warn) { 2771 file_magwarn(ms, 2772 "escaped tab found, use \\t instead"); 2773 warn = 0; /* already did */ 2774 } 2775 /*FALLTHROUGH*/ 2776 default: 2777 if (warn) { 2778 if (isprint((unsigned char)c)) { 2779 /* Allow escaping of 2780 * ``relations'' */ 2781 if (strchr("<>&^=!", c) == NULL 2782 && (m->type != FILE_REGEX || 2783 strchr("[]().*?^$|{}", c) 2784 == NULL)) { 2785 file_magwarn(ms, "no " 2786 "need to escape " 2787 "`%c'", c); 2788 } 2789 } else { 2790 file_magwarn(ms, 2791 "unknown escape sequence: " 2792 "\\%03o", c); 2793 } 2794 } 2795 /*FALLTHROUGH*/ 2796 /* space, perhaps force people to use \040? */ 2797 case ' ': 2798 #if 0 2799 /* 2800 * Other things people escape, but shouldn't need to, 2801 * so we disallow them 2802 */ 2803 case '\'': 2804 case '"': 2805 case '?': 2806 #endif 2807 /* Relations */ 2808 case '>': 2809 case '<': 2810 case '&': 2811 case '^': 2812 case '=': 2813 case '!': 2814 /* and baskslash itself */ 2815 case '\\': 2816 *p++ = (char) c; 2817 break; 2818 2819 case 'a': 2820 *p++ = '\a'; 2821 break; 2822 2823 case 'b': 2824 *p++ = '\b'; 2825 break; 2826 2827 case 'f': 2828 *p++ = '\f'; 2829 break; 2830 2831 case 'n': 2832 *p++ = '\n'; 2833 break; 2834 2835 case 'r': 2836 *p++ = '\r'; 2837 break; 2838 2839 case 't': 2840 *p++ = '\t'; 2841 break; 2842 2843 case 'v': 2844 *p++ = '\v'; 2845 break; 2846 2847 /* \ and up to 3 octal digits */ 2848 case '0': 2849 case '1': 2850 case '2': 2851 case '3': 2852 case '4': 2853 case '5': 2854 case '6': 2855 case '7': 2856 val = c - '0'; 2857 c = *s++; /* try for 2 */ 2858 if (c >= '0' && c <= '7') { 2859 val = (val << 3) | (c - '0'); 2860 c = *s++; /* try for 3 */ 2861 if (c >= '0' && c <= '7') 2862 val = (val << 3) | (c-'0'); 2863 else 2864 --s; 2865 } 2866 else 2867 --s; 2868 *p++ = (char)val; 2869 break; 2870 2871 /* \x and up to 2 hex digits */ 2872 case 'x': 2873 val = 'x'; /* Default if no digits */ 2874 c = hextoint(*s++); /* Get next char */ 2875 if (c >= 0) { 2876 val = c; 2877 c = hextoint(*s++); 2878 if (c >= 0) 2879 val = (val << 4) + c; 2880 else 2881 --s; 2882 } else 2883 --s; 2884 *p++ = (char)val; 2885 break; 2886 } 2887 } else 2888 *p++ = (char)c; 2889 } 2890 --s; 2891 out: 2892 *p = '\0'; 2893 m->vallen = CAST(unsigned char, (p - origp)); 2894 if (m->type == FILE_PSTRING) 2895 m->vallen += (unsigned char)file_pstring_length_size(m); 2896 return s; 2897 } 2898 2899 2900 /* Single hex char to int; -1 if not a hex char. */ 2901 private int 2902 hextoint(int c) 2903 { 2904 if (!isascii((unsigned char) c)) 2905 return -1; 2906 if (isdigit((unsigned char) c)) 2907 return c - '0'; 2908 if ((c >= 'a') && (c <= 'f')) 2909 return c + 10 - 'a'; 2910 if (( c>= 'A') && (c <= 'F')) 2911 return c + 10 - 'A'; 2912 return -1; 2913 } 2914 2915 2916 /* 2917 * Print a string containing C character escapes. 2918 */ 2919 protected void 2920 file_showstr(FILE *fp, const char *s, size_t len) 2921 { 2922 char c; 2923 2924 for (;;) { 2925 if (len == ~0U) { 2926 c = *s++; 2927 if (c == '\0') 2928 break; 2929 } 2930 else { 2931 if (len-- == 0) 2932 break; 2933 c = *s++; 2934 } 2935 if (c >= 040 && c <= 0176) /* TODO isprint && !iscntrl */ 2936 (void) fputc(c, fp); 2937 else { 2938 (void) fputc('\\', fp); 2939 switch (c) { 2940 case '\a': 2941 (void) fputc('a', fp); 2942 break; 2943 2944 case '\b': 2945 (void) fputc('b', fp); 2946 break; 2947 2948 case '\f': 2949 (void) fputc('f', fp); 2950 break; 2951 2952 case '\n': 2953 (void) fputc('n', fp); 2954 break; 2955 2956 case '\r': 2957 (void) fputc('r', fp); 2958 break; 2959 2960 case '\t': 2961 (void) fputc('t', fp); 2962 break; 2963 2964 case '\v': 2965 (void) fputc('v', fp); 2966 break; 2967 2968 default: 2969 (void) fprintf(fp, "%.3o", c & 0377); 2970 break; 2971 } 2972 } 2973 } 2974 } 2975 2976 /* 2977 * eatsize(): Eat the size spec from a number [eg. 10UL] 2978 */ 2979 private void 2980 eatsize(const char **p) 2981 { 2982 const char *l = *p; 2983 2984 if (LOWCASE(*l) == 'u') 2985 l++; 2986 2987 switch (LOWCASE(*l)) { 2988 case 'l': /* long */ 2989 case 's': /* short */ 2990 case 'h': /* short */ 2991 case 'b': /* char/byte */ 2992 case 'c': /* char/byte */ 2993 l++; 2994 /*FALLTHROUGH*/ 2995 default: 2996 break; 2997 } 2998 2999 *p = l; 3000 } 3001 3002 /* 3003 * handle a buffer containing a compiled file. 3004 */ 3005 private struct magic_map * 3006 apprentice_buf(struct magic_set *ms, struct magic *buf, size_t len) 3007 { 3008 struct magic_map *map; 3009 3010 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 3011 file_oomem(ms, sizeof(*map)); 3012 return NULL; 3013 } 3014 map->len = len; 3015 map->p = buf; 3016 map->type = MAP_TYPE_USER; 3017 if (check_buffer(ms, map, "buffer") != 0) { 3018 apprentice_unmap(map); 3019 return NULL; 3020 } 3021 return map; 3022 } 3023 3024 /* 3025 * handle a compiled file. 3026 */ 3027 3028 private struct magic_map * 3029 apprentice_map(struct magic_set *ms, const char *fn) 3030 { 3031 int fd; 3032 struct stat st; 3033 char *dbname = NULL; 3034 struct magic_map *map; 3035 struct magic_map *rv = NULL; 3036 3037 fd = -1; 3038 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 3039 file_oomem(ms, sizeof(*map)); 3040 goto error; 3041 } 3042 map->type = MAP_TYPE_USER; /* unspecified */ 3043 3044 dbname = mkdbname(ms, fn, 0); 3045 if (dbname == NULL) 3046 goto error; 3047 3048 if ((fd = open(dbname, O_RDONLY|O_BINARY)) == -1) 3049 goto error; 3050 3051 if (fstat(fd, &st) == -1) { 3052 file_error(ms, errno, "cannot stat `%s'", dbname); 3053 goto error; 3054 } 3055 if (st.st_size < 8 || st.st_size > MAXMAGIC_SIZE) { 3056 file_error(ms, 0, "file `%s' is too %s", dbname, 3057 st.st_size < 8 ? "small" : "large"); 3058 goto error; 3059 } 3060 3061 map->len = (size_t)st.st_size; 3062 #ifdef QUICK 3063 map->type = MAP_TYPE_MMAP; 3064 if ((map->p = mmap(0, (size_t)st.st_size, PROT_READ|PROT_WRITE, 3065 MAP_PRIVATE|MAP_FILE, fd, (off_t)0)) == MAP_FAILED) { 3066 file_error(ms, errno, "cannot map `%s'", dbname); 3067 goto error; 3068 } 3069 #else 3070 map->type = MAP_TYPE_MALLOC; 3071 if ((map->p = CAST(void *, malloc(map->len))) == NULL) { 3072 file_oomem(ms, map->len); 3073 goto error; 3074 } 3075 if (read(fd, map->p, map->len) != (ssize_t)map->len) { 3076 file_badread(ms); 3077 goto error; 3078 } 3079 #define RET 1 3080 #endif 3081 (void)close(fd); 3082 fd = -1; 3083 3084 if (check_buffer(ms, map, dbname) != 0) { 3085 rv = (struct magic_map *)-1; 3086 goto error; 3087 } 3088 #ifdef QUICK 3089 if (mprotect(map->p, (size_t)st.st_size, PROT_READ) == -1) { 3090 file_error(ms, errno, "cannot mprotect `%s'", dbname); 3091 goto error; 3092 } 3093 #endif 3094 3095 free(dbname); 3096 return map; 3097 3098 error: 3099 if (fd != -1) 3100 (void)close(fd); 3101 apprentice_unmap(map); 3102 free(dbname); 3103 return rv; 3104 } 3105 3106 private int 3107 check_buffer(struct magic_set *ms, struct magic_map *map, const char *dbname) 3108 { 3109 uint32_t *ptr; 3110 uint32_t entries, nentries; 3111 uint32_t version; 3112 int i, needsbyteswap; 3113 3114 ptr = CAST(uint32_t *, map->p); 3115 if (*ptr != MAGICNO) { 3116 if (swap4(*ptr) != MAGICNO) { 3117 file_error(ms, 0, "bad magic in `%s'", dbname); 3118 return -1; 3119 } 3120 needsbyteswap = 1; 3121 } else 3122 needsbyteswap = 0; 3123 if (needsbyteswap) 3124 version = swap4(ptr[1]); 3125 else 3126 version = ptr[1]; 3127 if (version != VERSIONNO) { 3128 file_error(ms, 0, "File %s supports only version %d magic " 3129 "files. `%s' is version %d", VERSION, 3130 VERSIONNO, dbname, version); 3131 return -1; 3132 } 3133 entries = (uint32_t)(map->len / sizeof(struct magic)); 3134 if ((entries * sizeof(struct magic)) != map->len) { 3135 file_error(ms, 0, "Size of `%s' %" SIZE_T_FORMAT "u is not " 3136 "a multiple of %" SIZE_T_FORMAT "u", 3137 dbname, map->len, sizeof(struct magic)); 3138 return -1; 3139 } 3140 map->magic[0] = CAST(struct magic *, map->p) + 1; 3141 nentries = 0; 3142 for (i = 0; i < MAGIC_SETS; i++) { 3143 if (needsbyteswap) 3144 map->nmagic[i] = swap4(ptr[i + 2]); 3145 else 3146 map->nmagic[i] = ptr[i + 2]; 3147 if (i != MAGIC_SETS - 1) 3148 map->magic[i + 1] = map->magic[i] + map->nmagic[i]; 3149 nentries += map->nmagic[i]; 3150 } 3151 if (entries != nentries + 1) { 3152 file_error(ms, 0, "Inconsistent entries in `%s' %u != %u", 3153 dbname, entries, nentries + 1); 3154 return -1; 3155 } 3156 if (needsbyteswap) 3157 for (i = 0; i < MAGIC_SETS; i++) 3158 byteswap(map->magic[i], map->nmagic[i]); 3159 return 0; 3160 } 3161 3162 /* 3163 * handle an mmaped file. 3164 */ 3165 private int 3166 apprentice_compile(struct magic_set *ms, struct magic_map *map, const char *fn) 3167 { 3168 static const size_t nm = sizeof(*map->nmagic) * MAGIC_SETS; 3169 static const size_t m = sizeof(**map->magic); 3170 int fd = -1; 3171 size_t len; 3172 char *dbname; 3173 int rv = -1; 3174 uint32_t i; 3175 union { 3176 struct magic m; 3177 uint32_t h[2 + MAGIC_SETS]; 3178 } hdr; 3179 3180 dbname = mkdbname(ms, fn, 1); 3181 3182 if (dbname == NULL) 3183 goto out; 3184 3185 if ((fd = open(dbname, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY, 0644)) == -1) 3186 { 3187 file_error(ms, errno, "cannot open `%s'", dbname); 3188 goto out; 3189 } 3190 memset(&hdr, 0, sizeof(hdr)); 3191 hdr.h[0] = MAGICNO; 3192 hdr.h[1] = VERSIONNO; 3193 memcpy(hdr.h + 2, map->nmagic, nm); 3194 3195 if (write(fd, &hdr, sizeof(hdr)) != (ssize_t)sizeof(hdr)) { 3196 file_error(ms, errno, "error writing `%s'", dbname); 3197 goto out2; 3198 } 3199 3200 for (i = 0; i < MAGIC_SETS; i++) { 3201 len = m * map->nmagic[i]; 3202 if (write(fd, map->magic[i], len) != (ssize_t)len) { 3203 file_error(ms, errno, "error writing `%s'", dbname); 3204 goto out2; 3205 } 3206 } 3207 3208 rv = 0; 3209 out2: 3210 if (fd != -1) 3211 (void)close(fd); 3212 out: 3213 apprentice_unmap(map); 3214 free(dbname); 3215 return rv; 3216 } 3217 3218 private const char ext[] = ".mgc"; 3219 /* 3220 * make a dbname 3221 */ 3222 private char * 3223 mkdbname(struct magic_set *ms, const char *fn, int strip) 3224 { 3225 const char *p, *q; 3226 char *buf; 3227 3228 if (strip) { 3229 if ((p = strrchr(fn, '/')) != NULL) 3230 fn = ++p; 3231 } 3232 3233 for (q = fn; *q; q++) 3234 continue; 3235 /* Look for .mgc */ 3236 for (p = ext + sizeof(ext) - 1; p >= ext && q >= fn; p--, q--) 3237 if (*p != *q) 3238 break; 3239 3240 /* Did not find .mgc, restore q */ 3241 if (p >= ext) 3242 while (*q) 3243 q++; 3244 3245 q++; 3246 /* Compatibility with old code that looked in .mime */ 3247 if (ms->flags & MAGIC_MIME) { 3248 if (asprintf(&buf, "%.*s.mime%s", (int)(q - fn), fn, ext) < 0) 3249 return NULL; 3250 if (access(buf, R_OK) != -1) { 3251 ms->flags &= MAGIC_MIME_TYPE; 3252 return buf; 3253 } 3254 free(buf); 3255 } 3256 if (asprintf(&buf, "%.*s%s", (int)(q - fn), fn, ext) < 0) 3257 return NULL; 3258 3259 /* Compatibility with old code that looked in .mime */ 3260 if (strstr(fn, ".mime") != NULL) 3261 ms->flags &= MAGIC_MIME_TYPE; 3262 return buf; 3263 } 3264 3265 /* 3266 * Byteswap an mmap'ed file if needed 3267 */ 3268 private void 3269 byteswap(struct magic *magic, uint32_t nmagic) 3270 { 3271 uint32_t i; 3272 for (i = 0; i < nmagic; i++) 3273 bs1(&magic[i]); 3274 } 3275 3276 /* 3277 * swap a short 3278 */ 3279 private uint16_t 3280 swap2(uint16_t sv) 3281 { 3282 uint16_t rv; 3283 uint8_t *s = (uint8_t *)(void *)&sv; 3284 uint8_t *d = (uint8_t *)(void *)&rv; 3285 d[0] = s[1]; 3286 d[1] = s[0]; 3287 return rv; 3288 } 3289 3290 /* 3291 * swap an int 3292 */ 3293 private uint32_t 3294 swap4(uint32_t sv) 3295 { 3296 uint32_t rv; 3297 uint8_t *s = (uint8_t *)(void *)&sv; 3298 uint8_t *d = (uint8_t *)(void *)&rv; 3299 d[0] = s[3]; 3300 d[1] = s[2]; 3301 d[2] = s[1]; 3302 d[3] = s[0]; 3303 return rv; 3304 } 3305 3306 /* 3307 * swap a quad 3308 */ 3309 private uint64_t 3310 swap8(uint64_t sv) 3311 { 3312 uint64_t rv; 3313 uint8_t *s = (uint8_t *)(void *)&sv; 3314 uint8_t *d = (uint8_t *)(void *)&rv; 3315 #if 0 3316 d[0] = s[3]; 3317 d[1] = s[2]; 3318 d[2] = s[1]; 3319 d[3] = s[0]; 3320 d[4] = s[7]; 3321 d[5] = s[6]; 3322 d[6] = s[5]; 3323 d[7] = s[4]; 3324 #else 3325 d[0] = s[7]; 3326 d[1] = s[6]; 3327 d[2] = s[5]; 3328 d[3] = s[4]; 3329 d[4] = s[3]; 3330 d[5] = s[2]; 3331 d[6] = s[1]; 3332 d[7] = s[0]; 3333 #endif 3334 return rv; 3335 } 3336 3337 /* 3338 * byteswap a single magic entry 3339 */ 3340 private void 3341 bs1(struct magic *m) 3342 { 3343 m->cont_level = swap2(m->cont_level); 3344 m->offset = swap4((uint32_t)m->offset); 3345 m->in_offset = swap4((uint32_t)m->in_offset); 3346 m->lineno = swap4((uint32_t)m->lineno); 3347 if (IS_STRING(m->type)) { 3348 m->str_range = swap4(m->str_range); 3349 m->str_flags = swap4(m->str_flags); 3350 } 3351 else { 3352 m->value.q = swap8(m->value.q); 3353 m->num_mask = swap8(m->num_mask); 3354 } 3355 } 3356 3357 protected size_t 3358 file_pstring_length_size(const struct magic *m) 3359 { 3360 switch (m->str_flags & PSTRING_LEN) { 3361 case PSTRING_1_LE: 3362 return 1; 3363 case PSTRING_2_LE: 3364 case PSTRING_2_BE: 3365 return 2; 3366 case PSTRING_4_LE: 3367 case PSTRING_4_BE: 3368 return 4; 3369 default: 3370 abort(); /* Impossible */ 3371 return 1; 3372 } 3373 } 3374 protected size_t 3375 file_pstring_get_length(const struct magic *m, const char *ss) 3376 { 3377 size_t len = 0; 3378 const unsigned char *s = (const unsigned char *)ss; 3379 unsigned int s3, s2, s1, s0; 3380 3381 switch (m->str_flags & PSTRING_LEN) { 3382 case PSTRING_1_LE: 3383 len = *s; 3384 break; 3385 case PSTRING_2_LE: 3386 s0 = s[0]; 3387 s1 = s[1]; 3388 len = (s1 << 8) | s0; 3389 break; 3390 case PSTRING_2_BE: 3391 s0 = s[0]; 3392 s1 = s[1]; 3393 len = (s0 << 8) | s1; 3394 break; 3395 case PSTRING_4_LE: 3396 s0 = s[0]; 3397 s1 = s[1]; 3398 s2 = s[2]; 3399 s3 = s[3]; 3400 len = (s3 << 24) | (s2 << 16) | (s1 << 8) | s0; 3401 break; 3402 case PSTRING_4_BE: 3403 s0 = s[0]; 3404 s1 = s[1]; 3405 s2 = s[2]; 3406 s3 = s[3]; 3407 len = (s0 << 24) | (s1 << 16) | (s2 << 8) | s3; 3408 break; 3409 default: 3410 abort(); /* Impossible */ 3411 } 3412 3413 if (m->str_flags & PSTRING_LENGTH_INCLUDES_ITSELF) 3414 len -= file_pstring_length_size(m); 3415 3416 return len; 3417 } 3418 3419 protected int 3420 file_magicfind(struct magic_set *ms, const char *name, struct mlist *v) 3421 { 3422 uint32_t i, j; 3423 struct mlist *mlist, *ml; 3424 3425 mlist = ms->mlist[1]; 3426 3427 for (ml = mlist->next; ml != mlist; ml = ml->next) { 3428 struct magic *ma = ml->magic; 3429 uint32_t nma = ml->nmagic; 3430 for (i = 0; i < nma; i++) { 3431 if (ma[i].type != FILE_NAME) 3432 continue; 3433 if (strcmp(ma[i].value.s, name) == 0) { 3434 v->magic = &ma[i]; 3435 for (j = i + 1; j < nma; j++) 3436 if (ma[j].cont_level == 0) 3437 break; 3438 v->nmagic = j - i; 3439 return 0; 3440 } 3441 } 3442 } 3443 return -1; 3444 } 3445