1 /* $NetBSD: apprentice.c,v 1.16 2017/02/10 18:06:59 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 1.257 2017/02/04 16:46:16 christos Exp $") 39 #else 40 __RCSID("$NetBSD: apprentice.c,v 1.16 2017/02/10 18:06:59 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 if (ms->mlist[0] != NULL) 622 file_reset(ms); 623 624 init_file_tables(); 625 626 for (i = 0; i < MAGIC_SETS; i++) { 627 mlist_free(ms->mlist[i]); 628 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 629 file_oomem(ms, sizeof(*ms->mlist[i])); 630 goto fail; 631 } 632 } 633 634 for (i = 0; i < nbufs; i++) { 635 map = apprentice_buf(ms, bufs[i], sizes[i]); 636 if (map == NULL) 637 goto fail; 638 639 for (j = 0; j < MAGIC_SETS; j++) { 640 if (add_mlist(ms->mlist[j], map, j) == -1) { 641 file_oomem(ms, sizeof(*ml)); 642 goto fail; 643 } 644 } 645 } 646 647 return 0; 648 fail: 649 for (i = 0; i < MAGIC_SETS; i++) { 650 mlist_free(ms->mlist[i]); 651 ms->mlist[i] = NULL; 652 } 653 return -1; 654 } 655 #endif 656 657 /* const char *fn: list of magic files and directories */ 658 protected int 659 file_apprentice(struct magic_set *ms, const char *fn, int action) 660 { 661 char *p, *mfn; 662 int file_err, errs = -1; 663 size_t i; 664 665 if (ms->mlist[0] != NULL) 666 file_reset(ms); 667 668 if ((fn = magic_getpath(fn, action)) == NULL) 669 return -1; 670 671 init_file_tables(); 672 673 if ((mfn = strdup(fn)) == NULL) { 674 file_oomem(ms, strlen(fn)); 675 return -1; 676 } 677 678 for (i = 0; i < MAGIC_SETS; i++) { 679 mlist_free(ms->mlist[i]); 680 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 681 file_oomem(ms, sizeof(*ms->mlist[i])); 682 while (i-- > 0) { 683 mlist_free(ms->mlist[i]); 684 ms->mlist[i] = NULL; 685 } 686 free(mfn); 687 return -1; 688 } 689 } 690 fn = mfn; 691 692 while (fn) { 693 p = strchr(fn, PATHSEP); 694 if (p) 695 *p++ = '\0'; 696 if (*fn == '\0') 697 break; 698 file_err = apprentice_1(ms, fn, action); 699 errs = MAX(errs, file_err); 700 fn = p; 701 } 702 703 free(mfn); 704 705 if (errs == -1) { 706 for (i = 0; i < MAGIC_SETS; i++) { 707 mlist_free(ms->mlist[i]); 708 ms->mlist[i] = NULL; 709 } 710 file_error(ms, 0, "could not find any valid magic files!"); 711 return -1; 712 } 713 714 #if 0 715 /* 716 * Always leave the database loaded 717 */ 718 if (action == FILE_LOAD) 719 return 0; 720 721 for (i = 0; i < MAGIC_SETS; i++) { 722 mlist_free(ms->mlist[i]); 723 ms->mlist[i] = NULL; 724 } 725 #endif 726 727 switch (action) { 728 case FILE_LOAD: 729 case FILE_COMPILE: 730 case FILE_CHECK: 731 case FILE_LIST: 732 return 0; 733 default: 734 file_error(ms, 0, "Invalid action %d", action); 735 return -1; 736 } 737 } 738 739 /* 740 * Compute the real length of a magic expression, for the purposes 741 * of determining how "strong" a magic expression is (approximating 742 * how specific its matches are): 743 * - magic characters count 0 unless escaped. 744 * - [] expressions count 1 745 * - {} expressions count 0 746 * - regular characters or escaped magic characters count 1 747 * - 0 length expressions count as one 748 */ 749 private size_t 750 nonmagic(const char *str) 751 { 752 const char *p; 753 size_t rv = 0; 754 755 for (p = str; *p; p++) 756 switch (*p) { 757 case '\\': /* Escaped anything counts 1 */ 758 if (!*++p) 759 p--; 760 rv++; 761 continue; 762 case '?': /* Magic characters count 0 */ 763 case '*': 764 case '.': 765 case '+': 766 case '^': 767 case '$': 768 continue; 769 case '[': /* Bracketed expressions count 1 the ']' */ 770 while (*p && *p != ']') 771 p++; 772 p--; 773 continue; 774 case '{': /* Braced expressions count 0 */ 775 while (*p && *p != '}') 776 p++; 777 if (!*p) 778 p--; 779 continue; 780 default: /* Anything else counts 1 */ 781 rv++; 782 continue; 783 } 784 785 return rv == 0 ? 1 : rv; /* Return at least 1 */ 786 } 787 788 /* 789 * Get weight of this magic entry, for sorting purposes. 790 */ 791 private size_t 792 apprentice_magic_strength(const struct magic *m) 793 { 794 #define MULT 10 795 size_t v, val = 2 * MULT; /* baseline strength */ 796 797 switch (m->type) { 798 case FILE_DEFAULT: /* make sure this sorts last */ 799 if (m->factor_op != FILE_FACTOR_OP_NONE) 800 abort(); 801 return 0; 802 803 case FILE_BYTE: 804 val += 1 * MULT; 805 break; 806 807 case FILE_SHORT: 808 case FILE_LESHORT: 809 case FILE_BESHORT: 810 val += 2 * MULT; 811 break; 812 813 case FILE_LONG: 814 case FILE_LELONG: 815 case FILE_BELONG: 816 case FILE_MELONG: 817 val += 4 * MULT; 818 break; 819 820 case FILE_PSTRING: 821 case FILE_STRING: 822 val += m->vallen * MULT; 823 break; 824 825 case FILE_BESTRING16: 826 case FILE_LESTRING16: 827 val += m->vallen * MULT / 2; 828 break; 829 830 case FILE_SEARCH: 831 val += m->vallen * MAX(MULT / m->vallen, 1); 832 break; 833 834 case FILE_REGEX: 835 v = nonmagic(m->value.s); 836 val += v * MAX(MULT / v, 1); 837 break; 838 839 case FILE_DATE: 840 case FILE_LEDATE: 841 case FILE_BEDATE: 842 case FILE_MEDATE: 843 case FILE_LDATE: 844 case FILE_LELDATE: 845 case FILE_BELDATE: 846 case FILE_MELDATE: 847 case FILE_FLOAT: 848 case FILE_BEFLOAT: 849 case FILE_LEFLOAT: 850 val += 4 * MULT; 851 break; 852 853 case FILE_QUAD: 854 case FILE_BEQUAD: 855 case FILE_LEQUAD: 856 case FILE_QDATE: 857 case FILE_LEQDATE: 858 case FILE_BEQDATE: 859 case FILE_QLDATE: 860 case FILE_LEQLDATE: 861 case FILE_BEQLDATE: 862 case FILE_QWDATE: 863 case FILE_LEQWDATE: 864 case FILE_BEQWDATE: 865 case FILE_DOUBLE: 866 case FILE_BEDOUBLE: 867 case FILE_LEDOUBLE: 868 val += 8 * MULT; 869 break; 870 871 case FILE_INDIRECT: 872 case FILE_NAME: 873 case FILE_USE: 874 break; 875 876 case FILE_DER: 877 val += MULT; 878 break; 879 880 default: 881 (void)fprintf(stderr, "Bad type %d\n", m->type); 882 abort(); 883 } 884 885 switch (m->reln) { 886 case 'x': /* matches anything penalize */ 887 case '!': /* matches almost anything penalize */ 888 val = 0; 889 break; 890 891 case '=': /* Exact match, prefer */ 892 val += MULT; 893 break; 894 895 case '>': 896 case '<': /* comparison match reduce strength */ 897 val -= 2 * MULT; 898 break; 899 900 case '^': 901 case '&': /* masking bits, we could count them too */ 902 val -= MULT; 903 break; 904 905 default: 906 (void)fprintf(stderr, "Bad relation %c\n", m->reln); 907 abort(); 908 } 909 910 if (val == 0) /* ensure we only return 0 for FILE_DEFAULT */ 911 val = 1; 912 913 switch (m->factor_op) { 914 case FILE_FACTOR_OP_NONE: 915 break; 916 case FILE_FACTOR_OP_PLUS: 917 val += m->factor; 918 break; 919 case FILE_FACTOR_OP_MINUS: 920 val -= m->factor; 921 break; 922 case FILE_FACTOR_OP_TIMES: 923 val *= m->factor; 924 break; 925 case FILE_FACTOR_OP_DIV: 926 val /= m->factor; 927 break; 928 default: 929 abort(); 930 } 931 932 /* 933 * Magic entries with no description get a bonus because they depend 934 * on subsequent magic entries to print something. 935 */ 936 if (m->desc[0] == '\0') 937 val++; 938 return val; 939 } 940 941 /* 942 * Sort callback for sorting entries by "strength" (basically length) 943 */ 944 private int 945 apprentice_sort(const void *a, const void *b) 946 { 947 const struct magic_entry *ma = CAST(const struct magic_entry *, a); 948 const struct magic_entry *mb = CAST(const struct magic_entry *, b); 949 size_t sa = apprentice_magic_strength(ma->mp); 950 size_t sb = apprentice_magic_strength(mb->mp); 951 if (sa == sb) 952 return 0; 953 else if (sa > sb) 954 return -1; 955 else 956 return 1; 957 } 958 959 /* 960 * Shows sorted patterns list in the order which is used for the matching 961 */ 962 private void 963 apprentice_list(struct mlist *mlist, int mode) 964 { 965 uint32_t magindex = 0; 966 struct mlist *ml; 967 for (ml = mlist->next; ml != mlist; ml = ml->next) { 968 for (magindex = 0; magindex < ml->nmagic; magindex++) { 969 struct magic *m = &ml->magic[magindex]; 970 if ((m->flag & mode) != mode) { 971 /* Skip sub-tests */ 972 while (magindex + 1 < ml->nmagic && 973 ml->magic[magindex + 1].cont_level != 0) 974 ++magindex; 975 continue; /* Skip to next top-level test*/ 976 } 977 978 /* 979 * Try to iterate over the tree until we find item with 980 * description/mimetype. 981 */ 982 while (magindex + 1 < ml->nmagic && 983 ml->magic[magindex + 1].cont_level != 0 && 984 *ml->magic[magindex].desc == '\0' && 985 *ml->magic[magindex].mimetype == '\0') 986 magindex++; 987 988 printf("Strength = %3" SIZE_T_FORMAT "u@%u: %s [%s]\n", 989 apprentice_magic_strength(m), 990 ml->magic[magindex].lineno, 991 ml->magic[magindex].desc, 992 ml->magic[magindex].mimetype); 993 } 994 } 995 } 996 997 private void 998 set_test_type(struct magic *mstart, struct magic *m) 999 { 1000 switch (m->type) { 1001 case FILE_BYTE: 1002 case FILE_SHORT: 1003 case FILE_LONG: 1004 case FILE_DATE: 1005 case FILE_BESHORT: 1006 case FILE_BELONG: 1007 case FILE_BEDATE: 1008 case FILE_LESHORT: 1009 case FILE_LELONG: 1010 case FILE_LEDATE: 1011 case FILE_LDATE: 1012 case FILE_BELDATE: 1013 case FILE_LELDATE: 1014 case FILE_MEDATE: 1015 case FILE_MELDATE: 1016 case FILE_MELONG: 1017 case FILE_QUAD: 1018 case FILE_LEQUAD: 1019 case FILE_BEQUAD: 1020 case FILE_QDATE: 1021 case FILE_LEQDATE: 1022 case FILE_BEQDATE: 1023 case FILE_QLDATE: 1024 case FILE_LEQLDATE: 1025 case FILE_BEQLDATE: 1026 case FILE_QWDATE: 1027 case FILE_LEQWDATE: 1028 case FILE_BEQWDATE: 1029 case FILE_FLOAT: 1030 case FILE_BEFLOAT: 1031 case FILE_LEFLOAT: 1032 case FILE_DOUBLE: 1033 case FILE_BEDOUBLE: 1034 case FILE_LEDOUBLE: 1035 case FILE_DER: 1036 mstart->flag |= BINTEST; 1037 break; 1038 case FILE_STRING: 1039 case FILE_PSTRING: 1040 case FILE_BESTRING16: 1041 case FILE_LESTRING16: 1042 /* Allow text overrides */ 1043 if (mstart->str_flags & STRING_TEXTTEST) 1044 mstart->flag |= TEXTTEST; 1045 else 1046 mstart->flag |= BINTEST; 1047 break; 1048 case FILE_REGEX: 1049 case FILE_SEARCH: 1050 /* Check for override */ 1051 if (mstart->str_flags & STRING_BINTEST) 1052 mstart->flag |= BINTEST; 1053 if (mstart->str_flags & STRING_TEXTTEST) 1054 mstart->flag |= TEXTTEST; 1055 1056 if (mstart->flag & (TEXTTEST|BINTEST)) 1057 break; 1058 1059 /* binary test if pattern is not text */ 1060 if (file_looks_utf8(m->value.us, (size_t)m->vallen, NULL, 1061 NULL) <= 0) 1062 mstart->flag |= BINTEST; 1063 else 1064 mstart->flag |= TEXTTEST; 1065 break; 1066 case FILE_DEFAULT: 1067 /* can't deduce anything; we shouldn't see this at the 1068 top level anyway */ 1069 break; 1070 case FILE_INVALID: 1071 default: 1072 /* invalid search type, but no need to complain here */ 1073 break; 1074 } 1075 } 1076 1077 private int 1078 addentry(struct magic_set *ms, struct magic_entry *me, 1079 struct magic_entry_set *mset) 1080 { 1081 size_t i = me->mp->type == FILE_NAME ? 1 : 0; 1082 if (mset[i].count == mset[i].max) { 1083 struct magic_entry *mp; 1084 1085 mset[i].max += ALLOC_INCR; 1086 if ((mp = CAST(struct magic_entry *, 1087 realloc(mset[i].me, sizeof(*mp) * mset[i].max))) == 1088 NULL) { 1089 file_oomem(ms, sizeof(*mp) * mset[i].max); 1090 return -1; 1091 } 1092 (void)memset(&mp[mset[i].count], 0, sizeof(*mp) * 1093 ALLOC_INCR); 1094 mset[i].me = mp; 1095 } 1096 mset[i].me[mset[i].count++] = *me; 1097 memset(me, 0, sizeof(*me)); 1098 return 0; 1099 } 1100 1101 /* 1102 * Load and parse one file. 1103 */ 1104 private void 1105 load_1(struct magic_set *ms, int action, const char *fn, int *errs, 1106 struct magic_entry_set *mset) 1107 { 1108 size_t lineno = 0, llen = 0; 1109 char *line = NULL; 1110 ssize_t len; 1111 struct magic_entry me; 1112 1113 FILE *f = fopen(ms->file = fn, "r"); 1114 if (f == NULL) { 1115 if (errno != ENOENT) 1116 file_error(ms, errno, "cannot read magic file `%s'", 1117 fn); 1118 (*errs)++; 1119 return; 1120 } 1121 1122 memset(&me, 0, sizeof(me)); 1123 /* read and parse this file */ 1124 for (ms->line = 1; (len = getline(&line, &llen, f)) != -1; 1125 ms->line++) { 1126 if (len == 0) /* null line, garbage, etc */ 1127 continue; 1128 if (line[len - 1] == '\n') { 1129 lineno++; 1130 line[len - 1] = '\0'; /* delete newline */ 1131 } 1132 switch (line[0]) { 1133 case '\0': /* empty, do not parse */ 1134 case '#': /* comment, do not parse */ 1135 continue; 1136 case '!': 1137 if (line[1] == ':') { 1138 size_t i; 1139 1140 for (i = 0; bang[i].name != NULL; i++) { 1141 if ((size_t)(len - 2) > bang[i].len && 1142 memcmp(bang[i].name, line + 2, 1143 bang[i].len) == 0) 1144 break; 1145 } 1146 if (bang[i].name == NULL) { 1147 file_error(ms, 0, 1148 "Unknown !: entry `%s'", line); 1149 (*errs)++; 1150 continue; 1151 } 1152 if (me.mp == NULL) { 1153 file_error(ms, 0, 1154 "No current entry for :!%s type", 1155 bang[i].name); 1156 (*errs)++; 1157 continue; 1158 } 1159 if ((*bang[i].fun)(ms, &me, 1160 line + bang[i].len + 2) != 0) { 1161 (*errs)++; 1162 continue; 1163 } 1164 continue; 1165 } 1166 /*FALLTHROUGH*/ 1167 default: 1168 again: 1169 switch (parse(ms, &me, line, lineno, action)) { 1170 case 0: 1171 continue; 1172 case 1: 1173 (void)addentry(ms, &me, mset); 1174 goto again; 1175 default: 1176 (*errs)++; 1177 break; 1178 } 1179 } 1180 } 1181 if (me.mp) 1182 (void)addentry(ms, &me, mset); 1183 free(line); 1184 (void)fclose(f); 1185 } 1186 1187 /* 1188 * parse a file or directory of files 1189 * const char *fn: name of magic file or directory 1190 */ 1191 private int 1192 cmpstrp(const void *p1, const void *p2) 1193 { 1194 return strcmp(*(char *const *)p1, *(char *const *)p2); 1195 } 1196 1197 1198 private uint32_t 1199 set_text_binary(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1200 uint32_t starttest) 1201 { 1202 static const char text[] = "text"; 1203 static const char binary[] = "binary"; 1204 static const size_t len = sizeof(text); 1205 1206 uint32_t i = starttest; 1207 1208 do { 1209 set_test_type(me[starttest].mp, me[i].mp); 1210 if ((ms->flags & MAGIC_DEBUG) == 0) 1211 continue; 1212 (void)fprintf(stderr, "%s%s%s: %s\n", 1213 me[i].mp->mimetype, 1214 me[i].mp->mimetype[0] == '\0' ? "" : "; ", 1215 me[i].mp->desc[0] ? me[i].mp->desc : "(no description)", 1216 me[i].mp->flag & BINTEST ? binary : text); 1217 if (me[i].mp->flag & BINTEST) { 1218 char *p = strstr(me[i].mp->desc, text); 1219 if (p && (p == me[i].mp->desc || 1220 isspace((unsigned char)p[-1])) && 1221 (p + len - me[i].mp->desc == MAXstring 1222 || (p[len] == '\0' || 1223 isspace((unsigned char)p[len])))) 1224 (void)fprintf(stderr, "*** Possible " 1225 "binary test for text type\n"); 1226 } 1227 } while (++i < nme && me[i].mp->cont_level != 0); 1228 return i; 1229 } 1230 1231 private void 1232 set_last_default(struct magic_set *ms, struct magic_entry *me, uint32_t nme) 1233 { 1234 uint32_t i; 1235 for (i = 0; i < nme; i++) { 1236 if (me[i].mp->cont_level == 0 && 1237 me[i].mp->type == FILE_DEFAULT) { 1238 while (++i < nme) 1239 if (me[i].mp->cont_level == 0) 1240 break; 1241 if (i != nme) { 1242 /* XXX - Ugh! */ 1243 ms->line = me[i].mp->lineno; 1244 file_magwarn(ms, 1245 "level 0 \"default\" did not sort last"); 1246 } 1247 return; 1248 } 1249 } 1250 } 1251 1252 private int 1253 coalesce_entries(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1254 struct magic **ma, uint32_t *nma) 1255 { 1256 uint32_t i, mentrycount = 0; 1257 size_t slen; 1258 1259 for (i = 0; i < nme; i++) 1260 mentrycount += me[i].cont_count; 1261 1262 slen = sizeof(**ma) * mentrycount; 1263 if ((*ma = CAST(struct magic *, malloc(slen))) == NULL) { 1264 file_oomem(ms, slen); 1265 return -1; 1266 } 1267 1268 mentrycount = 0; 1269 for (i = 0; i < nme; i++) { 1270 (void)memcpy(*ma + mentrycount, me[i].mp, 1271 me[i].cont_count * sizeof(**ma)); 1272 mentrycount += me[i].cont_count; 1273 } 1274 *nma = mentrycount; 1275 return 0; 1276 } 1277 1278 private void 1279 magic_entry_free(struct magic_entry *me, uint32_t nme) 1280 { 1281 uint32_t i; 1282 if (me == NULL) 1283 return; 1284 for (i = 0; i < nme; i++) 1285 free(me[i].mp); 1286 free(me); 1287 } 1288 1289 private struct magic_map * 1290 apprentice_load(struct magic_set *ms, const char *fn, int action) 1291 { 1292 int errs = 0; 1293 uint32_t i, j; 1294 size_t files = 0, maxfiles = 0; 1295 char **filearr = NULL, *mfn; 1296 struct stat st; 1297 struct magic_map *map; 1298 struct magic_entry_set mset[MAGIC_SETS]; 1299 DIR *dir; 1300 struct dirent *d; 1301 1302 memset(mset, 0, sizeof(mset)); 1303 ms->flags |= MAGIC_CHECK; /* Enable checks for parsed files */ 1304 1305 1306 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) 1307 { 1308 file_oomem(ms, sizeof(*map)); 1309 return NULL; 1310 } 1311 map->type = MAP_TYPE_MALLOC; 1312 1313 /* print silly verbose header for USG compat. */ 1314 if (action == FILE_CHECK) 1315 (void)fprintf(stderr, "%s\n", usg_hdr); 1316 1317 /* load directory or file */ 1318 if (stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) { 1319 dir = opendir(fn); 1320 if (!dir) { 1321 errs++; 1322 goto out; 1323 } 1324 while ((d = readdir(dir)) != NULL) { 1325 if (asprintf(&mfn, "%s/%s", fn, d->d_name) < 0) { 1326 file_oomem(ms, 1327 strlen(fn) + strlen(d->d_name) + 2); 1328 errs++; 1329 closedir(dir); 1330 goto out; 1331 } 1332 if (stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) { 1333 free(mfn); 1334 continue; 1335 } 1336 if (files >= maxfiles) { 1337 size_t mlen; 1338 maxfiles = (maxfiles + 1) * 2; 1339 mlen = maxfiles * sizeof(*filearr); 1340 if ((filearr = CAST(char **, 1341 realloc(filearr, mlen))) == NULL) { 1342 file_oomem(ms, mlen); 1343 free(mfn); 1344 closedir(dir); 1345 errs++; 1346 goto out; 1347 } 1348 } 1349 filearr[files++] = mfn; 1350 } 1351 closedir(dir); 1352 qsort(filearr, files, sizeof(*filearr), cmpstrp); 1353 for (i = 0; i < files; i++) { 1354 load_1(ms, action, filearr[i], &errs, mset); 1355 free(filearr[i]); 1356 } 1357 free(filearr); 1358 } else 1359 load_1(ms, action, fn, &errs, mset); 1360 if (errs) 1361 goto out; 1362 1363 for (j = 0; j < MAGIC_SETS; j++) { 1364 /* Set types of tests */ 1365 for (i = 0; i < mset[j].count; ) { 1366 if (mset[j].me[i].mp->cont_level != 0) { 1367 i++; 1368 continue; 1369 } 1370 i = set_text_binary(ms, mset[j].me, mset[j].count, i); 1371 } 1372 if (mset[j].me) 1373 qsort(mset[j].me, mset[j].count, sizeof(*mset[j].me), 1374 apprentice_sort); 1375 1376 /* 1377 * Make sure that any level 0 "default" line is last 1378 * (if one exists). 1379 */ 1380 set_last_default(ms, mset[j].me, mset[j].count); 1381 1382 /* coalesce per file arrays into a single one */ 1383 if (coalesce_entries(ms, mset[j].me, mset[j].count, 1384 &map->magic[j], &map->nmagic[j]) == -1) { 1385 errs++; 1386 goto out; 1387 } 1388 } 1389 1390 out: 1391 for (j = 0; j < MAGIC_SETS; j++) 1392 magic_entry_free(mset[j].me, mset[j].count); 1393 1394 if (errs) { 1395 apprentice_unmap(map); 1396 return NULL; 1397 } 1398 return map; 1399 } 1400 1401 /* 1402 * extend the sign bit if the comparison is to be signed 1403 */ 1404 protected uint64_t 1405 file_signextend(struct magic_set *ms, struct magic *m, uint64_t v) 1406 { 1407 if (!(m->flag & UNSIGNED)) { 1408 switch(m->type) { 1409 /* 1410 * Do not remove the casts below. They are 1411 * vital. When later compared with the data, 1412 * the sign extension must have happened. 1413 */ 1414 case FILE_BYTE: 1415 v = (signed char) v; 1416 break; 1417 case FILE_SHORT: 1418 case FILE_BESHORT: 1419 case FILE_LESHORT: 1420 v = (short) v; 1421 break; 1422 case FILE_DATE: 1423 case FILE_BEDATE: 1424 case FILE_LEDATE: 1425 case FILE_MEDATE: 1426 case FILE_LDATE: 1427 case FILE_BELDATE: 1428 case FILE_LELDATE: 1429 case FILE_MELDATE: 1430 case FILE_LONG: 1431 case FILE_BELONG: 1432 case FILE_LELONG: 1433 case FILE_MELONG: 1434 case FILE_FLOAT: 1435 case FILE_BEFLOAT: 1436 case FILE_LEFLOAT: 1437 v = (int32_t) v; 1438 break; 1439 case FILE_QUAD: 1440 case FILE_BEQUAD: 1441 case FILE_LEQUAD: 1442 case FILE_QDATE: 1443 case FILE_QLDATE: 1444 case FILE_QWDATE: 1445 case FILE_BEQDATE: 1446 case FILE_BEQLDATE: 1447 case FILE_BEQWDATE: 1448 case FILE_LEQDATE: 1449 case FILE_LEQLDATE: 1450 case FILE_LEQWDATE: 1451 case FILE_DOUBLE: 1452 case FILE_BEDOUBLE: 1453 case FILE_LEDOUBLE: 1454 v = (int64_t) v; 1455 break; 1456 case FILE_STRING: 1457 case FILE_PSTRING: 1458 case FILE_BESTRING16: 1459 case FILE_LESTRING16: 1460 case FILE_REGEX: 1461 case FILE_SEARCH: 1462 case FILE_DEFAULT: 1463 case FILE_INDIRECT: 1464 case FILE_NAME: 1465 case FILE_USE: 1466 case FILE_CLEAR: 1467 case FILE_DER: 1468 break; 1469 default: 1470 if (ms->flags & MAGIC_CHECK) 1471 file_magwarn(ms, "cannot happen: m->type=%d\n", 1472 m->type); 1473 return ~0U; 1474 } 1475 } 1476 return v; 1477 } 1478 1479 private int 1480 string_modifier_check(struct magic_set *ms, struct magic *m) 1481 { 1482 if ((ms->flags & MAGIC_CHECK) == 0) 1483 return 0; 1484 1485 if ((m->type != FILE_REGEX || (m->str_flags & REGEX_LINE_COUNT) == 0) && 1486 (m->type != FILE_PSTRING && (m->str_flags & PSTRING_LEN) != 0)) { 1487 file_magwarn(ms, 1488 "'/BHhLl' modifiers are only allowed for pascal strings\n"); 1489 return -1; 1490 } 1491 switch (m->type) { 1492 case FILE_BESTRING16: 1493 case FILE_LESTRING16: 1494 if (m->str_flags != 0) { 1495 file_magwarn(ms, 1496 "no modifiers allowed for 16-bit strings\n"); 1497 return -1; 1498 } 1499 break; 1500 case FILE_STRING: 1501 case FILE_PSTRING: 1502 if ((m->str_flags & REGEX_OFFSET_START) != 0) { 1503 file_magwarn(ms, 1504 "'/%c' only allowed on regex and search\n", 1505 CHAR_REGEX_OFFSET_START); 1506 return -1; 1507 } 1508 break; 1509 case FILE_SEARCH: 1510 if (m->str_range == 0) { 1511 file_magwarn(ms, 1512 "missing range; defaulting to %d\n", 1513 STRING_DEFAULT_RANGE); 1514 m->str_range = STRING_DEFAULT_RANGE; 1515 return -1; 1516 } 1517 break; 1518 case FILE_REGEX: 1519 if ((m->str_flags & STRING_COMPACT_WHITESPACE) != 0) { 1520 file_magwarn(ms, "'/%c' not allowed on regex\n", 1521 CHAR_COMPACT_WHITESPACE); 1522 return -1; 1523 } 1524 if ((m->str_flags & STRING_COMPACT_OPTIONAL_WHITESPACE) != 0) { 1525 file_magwarn(ms, "'/%c' not allowed on regex\n", 1526 CHAR_COMPACT_OPTIONAL_WHITESPACE); 1527 return -1; 1528 } 1529 break; 1530 default: 1531 file_magwarn(ms, "coding error: m->type=%d\n", 1532 m->type); 1533 return -1; 1534 } 1535 return 0; 1536 } 1537 1538 private int 1539 get_op(char c) 1540 { 1541 switch (c) { 1542 case '&': 1543 return FILE_OPAND; 1544 case '|': 1545 return FILE_OPOR; 1546 case '^': 1547 return FILE_OPXOR; 1548 case '+': 1549 return FILE_OPADD; 1550 case '-': 1551 return FILE_OPMINUS; 1552 case '*': 1553 return FILE_OPMULTIPLY; 1554 case '/': 1555 return FILE_OPDIVIDE; 1556 case '%': 1557 return FILE_OPMODULO; 1558 default: 1559 return -1; 1560 } 1561 } 1562 1563 #ifdef ENABLE_CONDITIONALS 1564 private int 1565 get_cond(const char *l, const char **t) 1566 { 1567 static const struct cond_tbl_s { 1568 char name[8]; 1569 size_t len; 1570 int cond; 1571 } cond_tbl[] = { 1572 { "if", 2, COND_IF }, 1573 { "elif", 4, COND_ELIF }, 1574 { "else", 4, COND_ELSE }, 1575 { "", 0, COND_NONE }, 1576 }; 1577 const struct cond_tbl_s *p; 1578 1579 for (p = cond_tbl; p->len; p++) { 1580 if (strncmp(l, p->name, p->len) == 0 && 1581 isspace((unsigned char)l[p->len])) { 1582 if (t) 1583 *t = l + p->len; 1584 break; 1585 } 1586 } 1587 return p->cond; 1588 } 1589 1590 private int 1591 check_cond(struct magic_set *ms, int cond, uint32_t cont_level) 1592 { 1593 int last_cond; 1594 last_cond = ms->c.li[cont_level].last_cond; 1595 1596 switch (cond) { 1597 case COND_IF: 1598 if (last_cond != COND_NONE && last_cond != COND_ELIF) { 1599 if (ms->flags & MAGIC_CHECK) 1600 file_magwarn(ms, "syntax error: `if'"); 1601 return -1; 1602 } 1603 last_cond = COND_IF; 1604 break; 1605 1606 case COND_ELIF: 1607 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1608 if (ms->flags & MAGIC_CHECK) 1609 file_magwarn(ms, "syntax error: `elif'"); 1610 return -1; 1611 } 1612 last_cond = COND_ELIF; 1613 break; 1614 1615 case COND_ELSE: 1616 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1617 if (ms->flags & MAGIC_CHECK) 1618 file_magwarn(ms, "syntax error: `else'"); 1619 return -1; 1620 } 1621 last_cond = COND_NONE; 1622 break; 1623 1624 case COND_NONE: 1625 last_cond = COND_NONE; 1626 break; 1627 } 1628 1629 ms->c.li[cont_level].last_cond = last_cond; 1630 return 0; 1631 } 1632 #endif /* ENABLE_CONDITIONALS */ 1633 1634 private int 1635 parse_indirect_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1636 { 1637 const char *l = *lp; 1638 1639 while (!isspace((unsigned char)*++l)) 1640 switch (*l) { 1641 case CHAR_INDIRECT_RELATIVE: 1642 m->str_flags |= INDIRECT_RELATIVE; 1643 break; 1644 default: 1645 if (ms->flags & MAGIC_CHECK) 1646 file_magwarn(ms, "indirect modifier `%c' " 1647 "invalid", *l); 1648 *lp = l; 1649 return -1; 1650 } 1651 *lp = l; 1652 return 0; 1653 } 1654 1655 private void 1656 parse_op_modifier(struct magic_set *ms, struct magic *m, const char **lp, 1657 int op) 1658 { 1659 const char *l = *lp; 1660 char *t; 1661 uint64_t val; 1662 1663 ++l; 1664 m->mask_op |= op; 1665 val = (uint64_t)strtoull(l, &t, 0); 1666 l = t; 1667 m->num_mask = file_signextend(ms, m, val); 1668 eatsize(&l); 1669 *lp = l; 1670 } 1671 1672 private int 1673 parse_string_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1674 { 1675 const char *l = *lp; 1676 char *t; 1677 int have_range = 0; 1678 1679 while (!isspace((unsigned char)*++l)) { 1680 switch (*l) { 1681 case '0': case '1': case '2': 1682 case '3': case '4': case '5': 1683 case '6': case '7': case '8': 1684 case '9': 1685 if (have_range && (ms->flags & MAGIC_CHECK)) 1686 file_magwarn(ms, "multiple ranges"); 1687 have_range = 1; 1688 m->str_range = CAST(uint32_t, strtoul(l, &t, 0)); 1689 if (m->str_range == 0) 1690 file_magwarn(ms, "zero range"); 1691 l = t - 1; 1692 break; 1693 case CHAR_COMPACT_WHITESPACE: 1694 m->str_flags |= STRING_COMPACT_WHITESPACE; 1695 break; 1696 case CHAR_COMPACT_OPTIONAL_WHITESPACE: 1697 m->str_flags |= STRING_COMPACT_OPTIONAL_WHITESPACE; 1698 break; 1699 case CHAR_IGNORE_LOWERCASE: 1700 m->str_flags |= STRING_IGNORE_LOWERCASE; 1701 break; 1702 case CHAR_IGNORE_UPPERCASE: 1703 m->str_flags |= STRING_IGNORE_UPPERCASE; 1704 break; 1705 case CHAR_REGEX_OFFSET_START: 1706 m->str_flags |= REGEX_OFFSET_START; 1707 break; 1708 case CHAR_BINTEST: 1709 m->str_flags |= STRING_BINTEST; 1710 break; 1711 case CHAR_TEXTTEST: 1712 m->str_flags |= STRING_TEXTTEST; 1713 break; 1714 case CHAR_TRIM: 1715 m->str_flags |= STRING_TRIM; 1716 break; 1717 case CHAR_PSTRING_1_LE: 1718 #define SET_LENGTH(a) m->str_flags = (m->str_flags & ~PSTRING_LEN) | (a) 1719 if (m->type != FILE_PSTRING) 1720 goto bad; 1721 SET_LENGTH(PSTRING_1_LE); 1722 break; 1723 case CHAR_PSTRING_2_BE: 1724 if (m->type != FILE_PSTRING) 1725 goto bad; 1726 SET_LENGTH(PSTRING_2_BE); 1727 break; 1728 case CHAR_PSTRING_2_LE: 1729 if (m->type != FILE_PSTRING) 1730 goto bad; 1731 SET_LENGTH(PSTRING_2_LE); 1732 break; 1733 case CHAR_PSTRING_4_BE: 1734 if (m->type != FILE_PSTRING) 1735 goto bad; 1736 SET_LENGTH(PSTRING_4_BE); 1737 break; 1738 case CHAR_PSTRING_4_LE: 1739 switch (m->type) { 1740 case FILE_PSTRING: 1741 case FILE_REGEX: 1742 break; 1743 default: 1744 goto bad; 1745 } 1746 SET_LENGTH(PSTRING_4_LE); 1747 break; 1748 case CHAR_PSTRING_LENGTH_INCLUDES_ITSELF: 1749 if (m->type != FILE_PSTRING) 1750 goto bad; 1751 m->str_flags |= PSTRING_LENGTH_INCLUDES_ITSELF; 1752 break; 1753 default: 1754 bad: 1755 if (ms->flags & MAGIC_CHECK) 1756 file_magwarn(ms, "string modifier `%c' " 1757 "invalid", *l); 1758 goto out; 1759 } 1760 /* allow multiple '/' for readability */ 1761 if (l[1] == '/' && !isspace((unsigned char)l[2])) 1762 l++; 1763 } 1764 if (string_modifier_check(ms, m) == -1) 1765 goto out; 1766 *lp = l; 1767 return 0; 1768 out: 1769 *lp = l; 1770 return -1; 1771 } 1772 1773 /* 1774 * parse one line from magic file, put into magic[index++] if valid 1775 */ 1776 private int 1777 parse(struct magic_set *ms, struct magic_entry *me, const char *line, 1778 size_t lineno, int action) 1779 { 1780 #ifdef ENABLE_CONDITIONALS 1781 static uint32_t last_cont_level = 0; 1782 #endif 1783 size_t i; 1784 struct magic *m; 1785 const char *l = line; 1786 char *t; 1787 int op; 1788 uint32_t cont_level; 1789 int32_t diff; 1790 1791 cont_level = 0; 1792 1793 /* 1794 * Parse the offset. 1795 */ 1796 while (*l == '>') { 1797 ++l; /* step over */ 1798 cont_level++; 1799 } 1800 #ifdef ENABLE_CONDITIONALS 1801 if (cont_level == 0 || cont_level > last_cont_level) 1802 if (file_check_mem(ms, cont_level) == -1) 1803 return -1; 1804 last_cont_level = cont_level; 1805 #endif 1806 if (cont_level != 0) { 1807 if (me->mp == NULL) { 1808 file_magerror(ms, "No current entry for continuation"); 1809 return -1; 1810 } 1811 if (me->cont_count == 0) { 1812 file_magerror(ms, "Continuations present with 0 count"); 1813 return -1; 1814 } 1815 m = &me->mp[me->cont_count - 1]; 1816 diff = (int32_t)cont_level - (int32_t)m->cont_level; 1817 if (diff > 1) 1818 file_magwarn(ms, "New continuation level %u is more " 1819 "than one larger than current level %u", cont_level, 1820 m->cont_level); 1821 if (me->cont_count == me->max_count) { 1822 struct magic *nm; 1823 size_t cnt = me->max_count + ALLOC_CHUNK; 1824 if ((nm = CAST(struct magic *, realloc(me->mp, 1825 sizeof(*nm) * cnt))) == NULL) { 1826 file_oomem(ms, sizeof(*nm) * cnt); 1827 return -1; 1828 } 1829 me->mp = m = nm; 1830 me->max_count = CAST(uint32_t, cnt); 1831 } 1832 m = &me->mp[me->cont_count++]; 1833 (void)memset(m, 0, sizeof(*m)); 1834 m->cont_level = cont_level; 1835 } else { 1836 static const size_t len = sizeof(*m) * ALLOC_CHUNK; 1837 if (me->mp != NULL) 1838 return 1; 1839 if ((m = CAST(struct magic *, malloc(len))) == NULL) { 1840 file_oomem(ms, len); 1841 return -1; 1842 } 1843 me->mp = m; 1844 me->max_count = ALLOC_CHUNK; 1845 (void)memset(m, 0, sizeof(*m)); 1846 m->factor_op = FILE_FACTOR_OP_NONE; 1847 m->cont_level = 0; 1848 me->cont_count = 1; 1849 } 1850 m->lineno = CAST(uint32_t, lineno); 1851 1852 if (*l == '&') { /* m->cont_level == 0 checked below. */ 1853 ++l; /* step over */ 1854 m->flag |= OFFADD; 1855 } 1856 if (*l == '(') { 1857 ++l; /* step over */ 1858 m->flag |= INDIR; 1859 if (m->flag & OFFADD) 1860 m->flag = (m->flag & ~OFFADD) | INDIROFFADD; 1861 1862 if (*l == '&') { /* m->cont_level == 0 checked below */ 1863 ++l; /* step over */ 1864 m->flag |= OFFADD; 1865 } 1866 } 1867 /* Indirect offsets are not valid at level 0. */ 1868 if (m->cont_level == 0 && (m->flag & (OFFADD | INDIROFFADD))) { 1869 if (ms->flags & MAGIC_CHECK) 1870 file_magwarn(ms, "relative offset at level 0"); 1871 return -1; 1872 } 1873 1874 /* get offset, then skip over it */ 1875 m->offset = (uint32_t)strtoul(l, &t, 0); 1876 if (l == t) { 1877 if (ms->flags & MAGIC_CHECK) 1878 file_magwarn(ms, "offset `%s' invalid", l); 1879 return -1; 1880 } 1881 l = t; 1882 1883 if (m->flag & INDIR) { 1884 m->in_type = FILE_LONG; 1885 m->in_offset = 0; 1886 m->in_op = 0; 1887 /* 1888 * read [.,lbs][+-]nnnnn) 1889 */ 1890 if (*l == '.' || *l == ',') { 1891 if (*l == ',') 1892 m->in_op |= FILE_OPSIGNED; 1893 l++; 1894 switch (*l) { 1895 case 'l': 1896 m->in_type = FILE_LELONG; 1897 break; 1898 case 'L': 1899 m->in_type = FILE_BELONG; 1900 break; 1901 case 'm': 1902 m->in_type = FILE_MELONG; 1903 break; 1904 case 'h': 1905 case 's': 1906 m->in_type = FILE_LESHORT; 1907 break; 1908 case 'H': 1909 case 'S': 1910 m->in_type = FILE_BESHORT; 1911 break; 1912 case 'c': 1913 case 'b': 1914 case 'C': 1915 case 'B': 1916 m->in_type = FILE_BYTE; 1917 break; 1918 case 'e': 1919 case 'f': 1920 case 'g': 1921 m->in_type = FILE_LEDOUBLE; 1922 break; 1923 case 'E': 1924 case 'F': 1925 case 'G': 1926 m->in_type = FILE_BEDOUBLE; 1927 break; 1928 case 'i': 1929 m->in_type = FILE_LEID3; 1930 break; 1931 case 'I': 1932 m->in_type = FILE_BEID3; 1933 break; 1934 default: 1935 if (ms->flags & MAGIC_CHECK) 1936 file_magwarn(ms, 1937 "indirect offset type `%c' invalid", 1938 *l); 1939 return -1; 1940 } 1941 l++; 1942 } 1943 1944 if (*l == '~') { 1945 m->in_op |= FILE_OPINVERSE; 1946 l++; 1947 } 1948 if ((op = get_op(*l)) != -1) { 1949 m->in_op |= op; 1950 l++; 1951 } 1952 if (*l == '(') { 1953 m->in_op |= FILE_OPINDIRECT; 1954 l++; 1955 } 1956 if (isdigit((unsigned char)*l) || *l == '-') { 1957 m->in_offset = (int32_t)strtol(l, &t, 0); 1958 if (l == t) { 1959 if (ms->flags & MAGIC_CHECK) 1960 file_magwarn(ms, 1961 "in_offset `%s' invalid", l); 1962 return -1; 1963 } 1964 l = t; 1965 } 1966 if (*l++ != ')' || 1967 ((m->in_op & FILE_OPINDIRECT) && *l++ != ')')) { 1968 if (ms->flags & MAGIC_CHECK) 1969 file_magwarn(ms, 1970 "missing ')' in indirect offset"); 1971 return -1; 1972 } 1973 } 1974 EATAB; 1975 1976 #ifdef ENABLE_CONDITIONALS 1977 m->cond = get_cond(l, &l); 1978 if (check_cond(ms, m->cond, cont_level) == -1) 1979 return -1; 1980 1981 EATAB; 1982 #endif 1983 1984 /* 1985 * Parse the type. 1986 */ 1987 if (*l == 'u') { 1988 /* 1989 * Try it as a keyword type prefixed by "u"; match what 1990 * follows the "u". If that fails, try it as an SUS 1991 * integer type. 1992 */ 1993 m->type = get_type(type_tbl, l + 1, &l); 1994 if (m->type == FILE_INVALID) { 1995 /* 1996 * Not a keyword type; parse it as an SUS type, 1997 * 'u' possibly followed by a number or C/S/L. 1998 */ 1999 m->type = get_standard_integer_type(l, &l); 2000 } 2001 /* It's unsigned. */ 2002 if (m->type != FILE_INVALID) 2003 m->flag |= UNSIGNED; 2004 } else { 2005 /* 2006 * Try it as a keyword type. If that fails, try it as 2007 * an SUS integer type if it begins with "d" or as an 2008 * SUS string type if it begins with "s". In any case, 2009 * it's not unsigned. 2010 */ 2011 m->type = get_type(type_tbl, l, &l); 2012 if (m->type == FILE_INVALID) { 2013 /* 2014 * Not a keyword type; parse it as an SUS type, 2015 * either 'd' possibly followed by a number or 2016 * C/S/L, or just 's'. 2017 */ 2018 if (*l == 'd') 2019 m->type = get_standard_integer_type(l, &l); 2020 else if (*l == 's' && !isalpha((unsigned char)l[1])) { 2021 m->type = FILE_STRING; 2022 ++l; 2023 } 2024 } 2025 } 2026 2027 if (m->type == FILE_INVALID) { 2028 /* Not found - try it as a special keyword. */ 2029 m->type = get_type(special_tbl, l, &l); 2030 } 2031 2032 if (m->type == FILE_INVALID) { 2033 if (ms->flags & MAGIC_CHECK) 2034 file_magwarn(ms, "type `%s' invalid", l); 2035 return -1; 2036 } 2037 2038 /* New-style anding: "0 byte&0x80 =0x80 dynamically linked" */ 2039 /* New and improved: ~ & | ^ + - * / % -- exciting, isn't it? */ 2040 2041 m->mask_op = 0; 2042 if (*l == '~') { 2043 if (!IS_STRING(m->type)) 2044 m->mask_op |= FILE_OPINVERSE; 2045 else if (ms->flags & MAGIC_CHECK) 2046 file_magwarn(ms, "'~' invalid for string types"); 2047 ++l; 2048 } 2049 m->str_range = 0; 2050 m->str_flags = m->type == FILE_PSTRING ? PSTRING_1_LE : 0; 2051 if ((op = get_op(*l)) != -1) { 2052 if (IS_STRING(m->type)) { 2053 int r; 2054 2055 if (op != FILE_OPDIVIDE) { 2056 if (ms->flags & MAGIC_CHECK) 2057 file_magwarn(ms, 2058 "invalid string/indirect op: " 2059 "`%c'", *t); 2060 return -1; 2061 } 2062 2063 if (m->type == FILE_INDIRECT) 2064 r = parse_indirect_modifier(ms, m, &l); 2065 else 2066 r = parse_string_modifier(ms, m, &l); 2067 if (r == -1) 2068 return -1; 2069 } else 2070 parse_op_modifier(ms, m, &l, op); 2071 } 2072 2073 /* 2074 * We used to set mask to all 1's here, instead let's just not do 2075 * anything if mask = 0 (unless you have a better idea) 2076 */ 2077 EATAB; 2078 2079 switch (*l) { 2080 case '>': 2081 case '<': 2082 m->reln = *l; 2083 ++l; 2084 if (*l == '=') { 2085 if (ms->flags & MAGIC_CHECK) { 2086 file_magwarn(ms, "%c= not supported", 2087 m->reln); 2088 return -1; 2089 } 2090 ++l; 2091 } 2092 break; 2093 /* Old-style anding: "0 byte &0x80 dynamically linked" */ 2094 case '&': 2095 case '^': 2096 case '=': 2097 m->reln = *l; 2098 ++l; 2099 if (*l == '=') { 2100 /* HP compat: ignore &= etc. */ 2101 ++l; 2102 } 2103 break; 2104 case '!': 2105 m->reln = *l; 2106 ++l; 2107 break; 2108 default: 2109 m->reln = '='; /* the default relation */ 2110 if (*l == 'x' && ((isascii((unsigned char)l[1]) && 2111 isspace((unsigned char)l[1])) || !l[1])) { 2112 m->reln = *l; 2113 ++l; 2114 } 2115 break; 2116 } 2117 /* 2118 * Grab the value part, except for an 'x' reln. 2119 */ 2120 if (m->reln != 'x' && getvalue(ms, m, &l, action)) 2121 return -1; 2122 2123 /* 2124 * TODO finish this macro and start using it! 2125 * #define offsetcheck {if (offset > ms->bytes_max -1) 2126 * magwarn("offset too big"); } 2127 */ 2128 2129 /* 2130 * Now get last part - the description 2131 */ 2132 EATAB; 2133 if (l[0] == '\b') { 2134 ++l; 2135 m->flag |= NOSPACE; 2136 } else if ((l[0] == '\\') && (l[1] == 'b')) { 2137 ++l; 2138 ++l; 2139 m->flag |= NOSPACE; 2140 } 2141 for (i = 0; (m->desc[i++] = *l++) != '\0' && i < sizeof(m->desc); ) 2142 continue; 2143 if (i == sizeof(m->desc)) { 2144 m->desc[sizeof(m->desc) - 1] = '\0'; 2145 if (ms->flags & MAGIC_CHECK) 2146 file_magwarn(ms, "description `%s' truncated", m->desc); 2147 } 2148 2149 /* 2150 * We only do this check while compiling, or if any of the magic 2151 * files were not compiled. 2152 */ 2153 if (ms->flags & MAGIC_CHECK) { 2154 if (check_format(ms, m) == -1) 2155 return -1; 2156 } 2157 #ifndef COMPILE_ONLY 2158 if (action == FILE_CHECK) { 2159 file_mdump(m); 2160 } 2161 #endif 2162 m->mimetype[0] = '\0'; /* initialise MIME type to none */ 2163 return 0; 2164 } 2165 2166 /* 2167 * parse a STRENGTH annotation line from magic file, put into magic[index - 1] 2168 * if valid 2169 */ 2170 private int 2171 parse_strength(struct magic_set *ms, struct magic_entry *me, const char *line) 2172 { 2173 const char *l = line; 2174 char *el; 2175 unsigned long factor; 2176 struct magic *m = &me->mp[0]; 2177 2178 if (m->factor_op != FILE_FACTOR_OP_NONE) { 2179 file_magwarn(ms, 2180 "Current entry already has a strength type: %c %d", 2181 m->factor_op, m->factor); 2182 return -1; 2183 } 2184 if (m->type == FILE_NAME) { 2185 file_magwarn(ms, "%s: Strength setting is not supported in " 2186 "\"name\" magic entries", m->value.s); 2187 return -1; 2188 } 2189 EATAB; 2190 switch (*l) { 2191 case FILE_FACTOR_OP_NONE: 2192 case FILE_FACTOR_OP_PLUS: 2193 case FILE_FACTOR_OP_MINUS: 2194 case FILE_FACTOR_OP_TIMES: 2195 case FILE_FACTOR_OP_DIV: 2196 m->factor_op = *l++; 2197 break; 2198 default: 2199 file_magwarn(ms, "Unknown factor op `%c'", *l); 2200 return -1; 2201 } 2202 EATAB; 2203 factor = strtoul(l, &el, 0); 2204 if (factor > 255) { 2205 file_magwarn(ms, "Too large factor `%lu'", factor); 2206 goto out; 2207 } 2208 if (*el && !isspace((unsigned char)*el)) { 2209 file_magwarn(ms, "Bad factor `%s'", l); 2210 goto out; 2211 } 2212 m->factor = (uint8_t)factor; 2213 if (m->factor == 0 && m->factor_op == FILE_FACTOR_OP_DIV) { 2214 file_magwarn(ms, "Cannot have factor op `%c' and factor %u", 2215 m->factor_op, m->factor); 2216 goto out; 2217 } 2218 return 0; 2219 out: 2220 m->factor_op = FILE_FACTOR_OP_NONE; 2221 m->factor = 0; 2222 return -1; 2223 } 2224 2225 private int 2226 goodchar(unsigned char x, const char *extra) 2227 { 2228 return (isascii(x) && isalnum(x)) || strchr(extra, x); 2229 } 2230 2231 private int 2232 parse_extra(struct magic_set *ms, struct magic_entry *me, const char *line, 2233 off_t off, size_t len, const char *name, const char *extra, int nt) 2234 { 2235 size_t i; 2236 const char *l = line; 2237 struct magic *m = &me->mp[me->cont_count == 0 ? 0 : me->cont_count - 1]; 2238 char *buf = CAST(char *, CAST(void *, m)) + off; 2239 2240 if (buf[0] != '\0') { 2241 len = nt ? strlen(buf) : len; 2242 file_magwarn(ms, "Current entry already has a %s type " 2243 "`%.*s', new type `%s'", name, (int)len, buf, l); 2244 return -1; 2245 } 2246 2247 if (*m->desc == '\0') { 2248 file_magwarn(ms, "Current entry does not yet have a " 2249 "description for adding a %s type", name); 2250 return -1; 2251 } 2252 2253 EATAB; 2254 for (i = 0; *l && i < len && goodchar(*l, extra); buf[i++] = *l++) 2255 continue; 2256 2257 if (i == len && *l) { 2258 if (nt) 2259 buf[len - 1] = '\0'; 2260 if (ms->flags & MAGIC_CHECK) 2261 file_magwarn(ms, "%s type `%s' truncated %" 2262 SIZE_T_FORMAT "u", name, line, i); 2263 } else { 2264 if (!isspace((unsigned char)*l) && !goodchar(*l, extra)) 2265 file_magwarn(ms, "%s type `%s' has bad char '%c'", 2266 name, line, *l); 2267 if (nt) 2268 buf[i] = '\0'; 2269 } 2270 2271 if (i > 0) 2272 return 0; 2273 2274 file_magerror(ms, "Bad magic entry '%s'", line); 2275 return -1; 2276 } 2277 2278 /* 2279 * Parse an Apple CREATOR/TYPE annotation from magic file and put it into 2280 * magic[index - 1] 2281 */ 2282 private int 2283 parse_apple(struct magic_set *ms, struct magic_entry *me, const char *line) 2284 { 2285 struct magic *m = &me->mp[0]; 2286 2287 return parse_extra(ms, me, line, 2288 CAST(off_t, offsetof(struct magic, apple)), 2289 sizeof(m->apple), "APPLE", "!+-./?", 0); 2290 } 2291 2292 /* 2293 * Parse a comma-separated list of extensions 2294 */ 2295 private int 2296 parse_ext(struct magic_set *ms, struct magic_entry *me, const char *line) 2297 { 2298 struct magic *m = &me->mp[0]; 2299 2300 return parse_extra(ms, me, line, 2301 CAST(off_t, offsetof(struct magic, ext)), 2302 sizeof(m->ext), "EXTENSION", ",!+-/@", 0); 2303 } 2304 2305 /* 2306 * parse a MIME annotation line from magic file, put into magic[index - 1] 2307 * if valid 2308 */ 2309 private int 2310 parse_mime(struct magic_set *ms, struct magic_entry *me, const char *line) 2311 { 2312 struct magic *m = &me->mp[0]; 2313 2314 return parse_extra(ms, me, line, 2315 CAST(off_t, offsetof(struct magic, mimetype)), 2316 sizeof(m->mimetype), "MIME", "+-/.", 1); 2317 } 2318 2319 private int 2320 check_format_type(const char *ptr, int type, const char **estr) 2321 { 2322 int quad = 0, h; 2323 size_t len, cnt; 2324 if (*ptr == '\0') { 2325 /* Missing format string; bad */ 2326 *estr = "missing format spec"; 2327 return -1; 2328 } 2329 2330 switch (file_formats[type]) { 2331 case FILE_FMT_QUAD: 2332 quad = 1; 2333 /*FALLTHROUGH*/ 2334 case FILE_FMT_NUM: 2335 if (quad == 0) { 2336 switch (type) { 2337 case FILE_BYTE: 2338 h = 2; 2339 break; 2340 case FILE_SHORT: 2341 case FILE_BESHORT: 2342 case FILE_LESHORT: 2343 h = 1; 2344 break; 2345 case FILE_LONG: 2346 case FILE_BELONG: 2347 case FILE_LELONG: 2348 case FILE_MELONG: 2349 case FILE_LEID3: 2350 case FILE_BEID3: 2351 case FILE_INDIRECT: 2352 h = 0; 2353 break; 2354 default: 2355 abort(); 2356 } 2357 } else 2358 h = 0; 2359 if (*ptr == '-') 2360 ptr++; 2361 if (*ptr == '.') 2362 ptr++; 2363 #define CHECKLEN() do { \ 2364 for (len = cnt = 0; isdigit((unsigned char)*ptr); ptr++, cnt++) \ 2365 len = len * 10 + (*ptr - '0'); \ 2366 if (cnt > 5 || len > 1024) \ 2367 goto toolong; \ 2368 } while (/*CONSTCOND*/0) 2369 2370 CHECKLEN(); 2371 if (*ptr == '.') 2372 ptr++; 2373 CHECKLEN(); 2374 if (quad) { 2375 if (*ptr++ != 'l') 2376 goto invalid; 2377 if (*ptr++ != 'l') 2378 goto invalid; 2379 } 2380 2381 switch (*ptr++) { 2382 #ifdef STRICT_FORMAT /* "long" formats are int formats for us */ 2383 /* so don't accept the 'l' modifier */ 2384 case 'l': 2385 switch (*ptr++) { 2386 case 'i': 2387 case 'd': 2388 case 'u': 2389 case 'o': 2390 case 'x': 2391 case 'X': 2392 if (h == 0) 2393 return 0; 2394 /*FALLTHROUGH*/ 2395 default: 2396 goto invalid; 2397 } 2398 2399 /* 2400 * Don't accept h and hh modifiers. They make writing 2401 * magic entries more complicated, for very little benefit 2402 */ 2403 case 'h': 2404 if (h-- <= 0) 2405 goto invalid; 2406 switch (*ptr++) { 2407 case 'h': 2408 if (h-- <= 0) 2409 goto invalid; 2410 switch (*ptr++) { 2411 case 'i': 2412 case 'd': 2413 case 'u': 2414 case 'o': 2415 case 'x': 2416 case 'X': 2417 return 0; 2418 default: 2419 goto invalid; 2420 } 2421 case 'i': 2422 case 'd': 2423 case 'u': 2424 case 'o': 2425 case 'x': 2426 case 'X': 2427 if (h == 0) 2428 return 0; 2429 /*FALLTHROUGH*/ 2430 default: 2431 goto invalid; 2432 } 2433 #endif 2434 case 'c': 2435 if (h == 2) 2436 return 0; 2437 goto invalid; 2438 case 'i': 2439 case 'd': 2440 case 'u': 2441 case 'o': 2442 case 'x': 2443 case 'X': 2444 #ifdef STRICT_FORMAT 2445 if (h == 0) 2446 return 0; 2447 /*FALLTHROUGH*/ 2448 #else 2449 return 0; 2450 #endif 2451 default: 2452 goto invalid; 2453 } 2454 2455 case FILE_FMT_FLOAT: 2456 case FILE_FMT_DOUBLE: 2457 if (*ptr == '-') 2458 ptr++; 2459 if (*ptr == '.') 2460 ptr++; 2461 CHECKLEN(); 2462 if (*ptr == '.') 2463 ptr++; 2464 CHECKLEN(); 2465 switch (*ptr++) { 2466 case 'e': 2467 case 'E': 2468 case 'f': 2469 case 'F': 2470 case 'g': 2471 case 'G': 2472 return 0; 2473 2474 default: 2475 goto invalid; 2476 } 2477 2478 2479 case FILE_FMT_STR: 2480 if (*ptr == '-') 2481 ptr++; 2482 while (isdigit((unsigned char )*ptr)) 2483 ptr++; 2484 if (*ptr == '.') { 2485 ptr++; 2486 while (isdigit((unsigned char )*ptr)) 2487 ptr++; 2488 } 2489 2490 switch (*ptr++) { 2491 case 's': 2492 return 0; 2493 default: 2494 goto invalid; 2495 } 2496 2497 default: 2498 /* internal error */ 2499 abort(); 2500 } 2501 invalid: 2502 *estr = "not valid"; 2503 toolong: 2504 *estr = "too long"; 2505 return -1; 2506 } 2507 2508 /* 2509 * Check that the optional printf format in description matches 2510 * the type of the magic. 2511 */ 2512 private int 2513 check_format(struct magic_set *ms, struct magic *m) 2514 { 2515 char *ptr; 2516 const char *estr; 2517 2518 for (ptr = m->desc; *ptr; ptr++) 2519 if (*ptr == '%') 2520 break; 2521 if (*ptr == '\0') { 2522 /* No format string; ok */ 2523 return 1; 2524 } 2525 2526 assert(file_nformats == file_nnames); 2527 2528 if (m->type >= file_nformats) { 2529 file_magwarn(ms, "Internal error inconsistency between " 2530 "m->type and format strings"); 2531 return -1; 2532 } 2533 if (file_formats[m->type] == FILE_FMT_NONE) { 2534 file_magwarn(ms, "No format string for `%s' with description " 2535 "`%s'", m->desc, file_names[m->type]); 2536 return -1; 2537 } 2538 2539 ptr++; 2540 if (check_format_type(ptr, m->type, &estr) == -1) { 2541 /* 2542 * TODO: this error message is unhelpful if the format 2543 * string is not one character long 2544 */ 2545 file_magwarn(ms, "Printf format is %s for type " 2546 "`%s' in description `%s'", estr, 2547 file_names[m->type], m->desc); 2548 return -1; 2549 } 2550 2551 for (; *ptr; ptr++) { 2552 if (*ptr == '%') { 2553 file_magwarn(ms, 2554 "Too many format strings (should have at most one) " 2555 "for `%s' with description `%s'", 2556 file_names[m->type], m->desc); 2557 return -1; 2558 } 2559 } 2560 return 0; 2561 } 2562 2563 /* 2564 * Read a numeric value from a pointer, into the value union of a magic 2565 * pointer, according to the magic type. Update the string pointer to point 2566 * just after the number read. Return 0 for success, non-zero for failure. 2567 */ 2568 private int 2569 getvalue(struct magic_set *ms, struct magic *m, const char **p, int action) 2570 { 2571 switch (m->type) { 2572 case FILE_BESTRING16: 2573 case FILE_LESTRING16: 2574 case FILE_STRING: 2575 case FILE_PSTRING: 2576 case FILE_REGEX: 2577 case FILE_SEARCH: 2578 case FILE_NAME: 2579 case FILE_USE: 2580 case FILE_DER: 2581 *p = getstr(ms, m, *p, action == FILE_COMPILE); 2582 if (*p == NULL) { 2583 if (ms->flags & MAGIC_CHECK) 2584 file_magwarn(ms, "cannot get string from `%s'", 2585 m->value.s); 2586 return -1; 2587 } 2588 if (m->type == FILE_REGEX) { 2589 file_regex_t rx; 2590 int rc = file_regcomp(&rx, m->value.s, REG_EXTENDED); 2591 if (rc) { 2592 if (ms->flags & MAGIC_CHECK) 2593 file_regerror(&rx, rc, ms); 2594 } 2595 file_regfree(&rx); 2596 return rc ? -1 : 0; 2597 } 2598 return 0; 2599 case FILE_FLOAT: 2600 case FILE_BEFLOAT: 2601 case FILE_LEFLOAT: 2602 if (m->reln != 'x') { 2603 char *ep; 2604 errno = 0; 2605 #ifdef HAVE_STRTOF 2606 m->value.f = strtof(*p, &ep); 2607 #else 2608 m->value.f = (float)strtod(*p, &ep); 2609 #endif 2610 if (errno == 0) 2611 *p = ep; 2612 } 2613 return 0; 2614 case FILE_DOUBLE: 2615 case FILE_BEDOUBLE: 2616 case FILE_LEDOUBLE: 2617 if (m->reln != 'x') { 2618 char *ep; 2619 errno = 0; 2620 m->value.d = strtod(*p, &ep); 2621 if (errno == 0) 2622 *p = ep; 2623 } 2624 return 0; 2625 default: 2626 if (m->reln != 'x') { 2627 char *ep; 2628 errno = 0; 2629 m->value.q = file_signextend(ms, m, 2630 (uint64_t)strtoull(*p, &ep, 0)); 2631 if (errno == 0) { 2632 *p = ep; 2633 eatsize(p); 2634 } 2635 } 2636 return 0; 2637 } 2638 } 2639 2640 /* 2641 * Convert a string containing C character escapes. Stop at an unescaped 2642 * space or tab. 2643 * Copy the converted version to "m->value.s", and the length in m->vallen. 2644 * Return updated scan pointer as function result. Warn if set. 2645 */ 2646 private const char * 2647 getstr(struct magic_set *ms, struct magic *m, const char *s, int warn) 2648 { 2649 const char *origs = s; 2650 char *p = m->value.s; 2651 size_t plen = sizeof(m->value.s); 2652 char *origp = p; 2653 char *pmax = p + plen - 1; 2654 int c; 2655 int val; 2656 2657 while ((c = *s++) != '\0') { 2658 if (isspace((unsigned char) c)) 2659 break; 2660 if (p >= pmax) { 2661 file_error(ms, 0, "string too long: `%s'", origs); 2662 return NULL; 2663 } 2664 if (c == '\\') { 2665 switch(c = *s++) { 2666 2667 case '\0': 2668 if (warn) 2669 file_magwarn(ms, "incomplete escape"); 2670 s--; 2671 goto out; 2672 2673 case '\t': 2674 if (warn) { 2675 file_magwarn(ms, 2676 "escaped tab found, use \\t instead"); 2677 warn = 0; /* already did */ 2678 } 2679 /*FALLTHROUGH*/ 2680 default: 2681 if (warn) { 2682 if (isprint((unsigned char)c)) { 2683 /* Allow escaping of 2684 * ``relations'' */ 2685 if (strchr("<>&^=!", c) == NULL 2686 && (m->type != FILE_REGEX || 2687 strchr("[]().*?^$|{}", c) 2688 == NULL)) { 2689 file_magwarn(ms, "no " 2690 "need to escape " 2691 "`%c'", c); 2692 } 2693 } else { 2694 file_magwarn(ms, 2695 "unknown escape sequence: " 2696 "\\%03o", c); 2697 } 2698 } 2699 /*FALLTHROUGH*/ 2700 /* space, perhaps force people to use \040? */ 2701 case ' ': 2702 #if 0 2703 /* 2704 * Other things people escape, but shouldn't need to, 2705 * so we disallow them 2706 */ 2707 case '\'': 2708 case '"': 2709 case '?': 2710 #endif 2711 /* Relations */ 2712 case '>': 2713 case '<': 2714 case '&': 2715 case '^': 2716 case '=': 2717 case '!': 2718 /* and baskslash itself */ 2719 case '\\': 2720 *p++ = (char) c; 2721 break; 2722 2723 case 'a': 2724 *p++ = '\a'; 2725 break; 2726 2727 case 'b': 2728 *p++ = '\b'; 2729 break; 2730 2731 case 'f': 2732 *p++ = '\f'; 2733 break; 2734 2735 case 'n': 2736 *p++ = '\n'; 2737 break; 2738 2739 case 'r': 2740 *p++ = '\r'; 2741 break; 2742 2743 case 't': 2744 *p++ = '\t'; 2745 break; 2746 2747 case 'v': 2748 *p++ = '\v'; 2749 break; 2750 2751 /* \ and up to 3 octal digits */ 2752 case '0': 2753 case '1': 2754 case '2': 2755 case '3': 2756 case '4': 2757 case '5': 2758 case '6': 2759 case '7': 2760 val = c - '0'; 2761 c = *s++; /* try for 2 */ 2762 if (c >= '0' && c <= '7') { 2763 val = (val << 3) | (c - '0'); 2764 c = *s++; /* try for 3 */ 2765 if (c >= '0' && c <= '7') 2766 val = (val << 3) | (c-'0'); 2767 else 2768 --s; 2769 } 2770 else 2771 --s; 2772 *p++ = (char)val; 2773 break; 2774 2775 /* \x and up to 2 hex digits */ 2776 case 'x': 2777 val = 'x'; /* Default if no digits */ 2778 c = hextoint(*s++); /* Get next char */ 2779 if (c >= 0) { 2780 val = c; 2781 c = hextoint(*s++); 2782 if (c >= 0) 2783 val = (val << 4) + c; 2784 else 2785 --s; 2786 } else 2787 --s; 2788 *p++ = (char)val; 2789 break; 2790 } 2791 } else 2792 *p++ = (char)c; 2793 } 2794 --s; 2795 out: 2796 *p = '\0'; 2797 m->vallen = CAST(unsigned char, (p - origp)); 2798 if (m->type == FILE_PSTRING) 2799 m->vallen += (unsigned char)file_pstring_length_size(m); 2800 return s; 2801 } 2802 2803 2804 /* Single hex char to int; -1 if not a hex char. */ 2805 private int 2806 hextoint(int c) 2807 { 2808 if (!isascii((unsigned char) c)) 2809 return -1; 2810 if (isdigit((unsigned char) c)) 2811 return c - '0'; 2812 if ((c >= 'a') && (c <= 'f')) 2813 return c + 10 - 'a'; 2814 if (( c>= 'A') && (c <= 'F')) 2815 return c + 10 - 'A'; 2816 return -1; 2817 } 2818 2819 2820 /* 2821 * Print a string containing C character escapes. 2822 */ 2823 protected void 2824 file_showstr(FILE *fp, const char *s, size_t len) 2825 { 2826 char c; 2827 2828 for (;;) { 2829 if (len == ~0U) { 2830 c = *s++; 2831 if (c == '\0') 2832 break; 2833 } 2834 else { 2835 if (len-- == 0) 2836 break; 2837 c = *s++; 2838 } 2839 if (c >= 040 && c <= 0176) /* TODO isprint && !iscntrl */ 2840 (void) fputc(c, fp); 2841 else { 2842 (void) fputc('\\', fp); 2843 switch (c) { 2844 case '\a': 2845 (void) fputc('a', fp); 2846 break; 2847 2848 case '\b': 2849 (void) fputc('b', fp); 2850 break; 2851 2852 case '\f': 2853 (void) fputc('f', fp); 2854 break; 2855 2856 case '\n': 2857 (void) fputc('n', fp); 2858 break; 2859 2860 case '\r': 2861 (void) fputc('r', fp); 2862 break; 2863 2864 case '\t': 2865 (void) fputc('t', fp); 2866 break; 2867 2868 case '\v': 2869 (void) fputc('v', fp); 2870 break; 2871 2872 default: 2873 (void) fprintf(fp, "%.3o", c & 0377); 2874 break; 2875 } 2876 } 2877 } 2878 } 2879 2880 /* 2881 * eatsize(): Eat the size spec from a number [eg. 10UL] 2882 */ 2883 private void 2884 eatsize(const char **p) 2885 { 2886 const char *l = *p; 2887 2888 if (LOWCASE(*l) == 'u') 2889 l++; 2890 2891 switch (LOWCASE(*l)) { 2892 case 'l': /* long */ 2893 case 's': /* short */ 2894 case 'h': /* short */ 2895 case 'b': /* char/byte */ 2896 case 'c': /* char/byte */ 2897 l++; 2898 /*FALLTHROUGH*/ 2899 default: 2900 break; 2901 } 2902 2903 *p = l; 2904 } 2905 2906 /* 2907 * handle a buffer containing a compiled file. 2908 */ 2909 private struct magic_map * 2910 apprentice_buf(struct magic_set *ms, struct magic *buf, size_t len) 2911 { 2912 struct magic_map *map; 2913 2914 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 2915 file_oomem(ms, sizeof(*map)); 2916 return NULL; 2917 } 2918 map->len = len; 2919 map->p = buf; 2920 map->type = MAP_TYPE_USER; 2921 if (check_buffer(ms, map, "buffer") != 0) { 2922 apprentice_unmap(map); 2923 return NULL; 2924 } 2925 return map; 2926 } 2927 2928 /* 2929 * handle a compiled file. 2930 */ 2931 2932 private struct magic_map * 2933 apprentice_map(struct magic_set *ms, const char *fn) 2934 { 2935 int fd; 2936 struct stat st; 2937 char *dbname = NULL; 2938 struct magic_map *map; 2939 struct magic_map *rv = NULL; 2940 2941 fd = -1; 2942 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 2943 file_oomem(ms, sizeof(*map)); 2944 goto error; 2945 } 2946 map->type = MAP_TYPE_USER; /* unspecified */ 2947 2948 dbname = mkdbname(ms, fn, 0); 2949 if (dbname == NULL) 2950 goto error; 2951 2952 if ((fd = open(dbname, O_RDONLY|O_BINARY)) == -1) 2953 goto error; 2954 2955 if (fstat(fd, &st) == -1) { 2956 file_error(ms, errno, "cannot stat `%s'", dbname); 2957 goto error; 2958 } 2959 if (st.st_size < 8 || st.st_size > MAXMAGIC_SIZE) { 2960 file_error(ms, 0, "file `%s' is too %s", dbname, 2961 st.st_size < 8 ? "small" : "large"); 2962 goto error; 2963 } 2964 2965 map->len = (size_t)st.st_size; 2966 #ifdef QUICK 2967 map->type = MAP_TYPE_MMAP; 2968 if ((map->p = mmap(0, (size_t)st.st_size, PROT_READ|PROT_WRITE, 2969 MAP_PRIVATE|MAP_FILE, fd, (off_t)0)) == MAP_FAILED) { 2970 file_error(ms, errno, "cannot map `%s'", dbname); 2971 goto error; 2972 } 2973 #else 2974 map->type = MAP_TYPE_MALLOC; 2975 if ((map->p = CAST(void *, malloc(map->len))) == NULL) { 2976 file_oomem(ms, map->len); 2977 goto error; 2978 } 2979 if (read(fd, map->p, map->len) != (ssize_t)map->len) { 2980 file_badread(ms); 2981 goto error; 2982 } 2983 #define RET 1 2984 #endif 2985 (void)close(fd); 2986 fd = -1; 2987 2988 if (check_buffer(ms, map, dbname) != 0) { 2989 rv = (struct magic_map *)-1; 2990 goto error; 2991 } 2992 #ifdef QUICK 2993 if (mprotect(map->p, (size_t)st.st_size, PROT_READ) == -1) { 2994 file_error(ms, errno, "cannot mprotect `%s'", dbname); 2995 goto error; 2996 } 2997 #endif 2998 2999 free(dbname); 3000 return map; 3001 3002 error: 3003 if (fd != -1) 3004 (void)close(fd); 3005 apprentice_unmap(map); 3006 free(dbname); 3007 return rv; 3008 } 3009 3010 private int 3011 check_buffer(struct magic_set *ms, struct magic_map *map, const char *dbname) 3012 { 3013 uint32_t *ptr; 3014 uint32_t entries, nentries; 3015 uint32_t version; 3016 int i, needsbyteswap; 3017 3018 ptr = CAST(uint32_t *, map->p); 3019 if (*ptr != MAGICNO) { 3020 if (swap4(*ptr) != MAGICNO) { 3021 file_error(ms, 0, "bad magic in `%s'", dbname); 3022 return -1; 3023 } 3024 needsbyteswap = 1; 3025 } else 3026 needsbyteswap = 0; 3027 if (needsbyteswap) 3028 version = swap4(ptr[1]); 3029 else 3030 version = ptr[1]; 3031 if (version != VERSIONNO) { 3032 file_error(ms, 0, "File %s supports only version %d magic " 3033 "files. `%s' is version %d", VERSION, 3034 VERSIONNO, dbname, version); 3035 return -1; 3036 } 3037 entries = (uint32_t)(map->len / sizeof(struct magic)); 3038 if ((entries * sizeof(struct magic)) != map->len) { 3039 file_error(ms, 0, "Size of `%s' %" SIZE_T_FORMAT "u is not " 3040 "a multiple of %" SIZE_T_FORMAT "u", 3041 dbname, map->len, sizeof(struct magic)); 3042 return -1; 3043 } 3044 map->magic[0] = CAST(struct magic *, map->p) + 1; 3045 nentries = 0; 3046 for (i = 0; i < MAGIC_SETS; i++) { 3047 if (needsbyteswap) 3048 map->nmagic[i] = swap4(ptr[i + 2]); 3049 else 3050 map->nmagic[i] = ptr[i + 2]; 3051 if (i != MAGIC_SETS - 1) 3052 map->magic[i + 1] = map->magic[i] + map->nmagic[i]; 3053 nentries += map->nmagic[i]; 3054 } 3055 if (entries != nentries + 1) { 3056 file_error(ms, 0, "Inconsistent entries in `%s' %u != %u", 3057 dbname, entries, nentries + 1); 3058 return -1; 3059 } 3060 if (needsbyteswap) 3061 for (i = 0; i < MAGIC_SETS; i++) 3062 byteswap(map->magic[i], map->nmagic[i]); 3063 return 0; 3064 } 3065 3066 /* 3067 * handle an mmaped file. 3068 */ 3069 private int 3070 apprentice_compile(struct magic_set *ms, struct magic_map *map, const char *fn) 3071 { 3072 static const size_t nm = sizeof(*map->nmagic) * MAGIC_SETS; 3073 static const size_t m = sizeof(**map->magic); 3074 int fd = -1; 3075 size_t len; 3076 char *dbname; 3077 int rv = -1; 3078 uint32_t i; 3079 union { 3080 struct magic m; 3081 uint32_t h[2 + MAGIC_SETS]; 3082 } hdr; 3083 3084 dbname = mkdbname(ms, fn, 1); 3085 3086 if (dbname == NULL) 3087 goto out; 3088 3089 if ((fd = open(dbname, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY, 0644)) == -1) 3090 { 3091 file_error(ms, errno, "cannot open `%s'", dbname); 3092 goto out; 3093 } 3094 memset(&hdr, 0, sizeof(hdr)); 3095 hdr.h[0] = MAGICNO; 3096 hdr.h[1] = VERSIONNO; 3097 memcpy(hdr.h + 2, map->nmagic, nm); 3098 3099 if (write(fd, &hdr, sizeof(hdr)) != (ssize_t)sizeof(hdr)) { 3100 file_error(ms, errno, "error writing `%s'", dbname); 3101 goto out; 3102 } 3103 3104 for (i = 0; i < MAGIC_SETS; i++) { 3105 len = m * map->nmagic[i]; 3106 if (write(fd, map->magic[i], len) != (ssize_t)len) { 3107 file_error(ms, errno, "error writing `%s'", dbname); 3108 goto out; 3109 } 3110 } 3111 3112 if (fd != -1) 3113 (void)close(fd); 3114 rv = 0; 3115 out: 3116 apprentice_unmap(map); 3117 free(dbname); 3118 return rv; 3119 } 3120 3121 private const char ext[] = ".mgc"; 3122 /* 3123 * make a dbname 3124 */ 3125 private char * 3126 mkdbname(struct magic_set *ms, const char *fn, int strip) 3127 { 3128 const char *p, *q; 3129 char *buf; 3130 3131 if (strip) { 3132 if ((p = strrchr(fn, '/')) != NULL) 3133 fn = ++p; 3134 } 3135 3136 for (q = fn; *q; q++) 3137 continue; 3138 /* Look for .mgc */ 3139 for (p = ext + sizeof(ext) - 1; p >= ext && q >= fn; p--, q--) 3140 if (*p != *q) 3141 break; 3142 3143 /* Did not find .mgc, restore q */ 3144 if (p >= ext) 3145 while (*q) 3146 q++; 3147 3148 q++; 3149 /* Compatibility with old code that looked in .mime */ 3150 if (ms->flags & MAGIC_MIME) { 3151 if (asprintf(&buf, "%.*s.mime%s", (int)(q - fn), fn, ext) < 0) 3152 return NULL; 3153 if (access(buf, R_OK) != -1) { 3154 ms->flags &= MAGIC_MIME_TYPE; 3155 return buf; 3156 } 3157 free(buf); 3158 } 3159 if (asprintf(&buf, "%.*s%s", (int)(q - fn), fn, ext) < 0) 3160 return NULL; 3161 3162 /* Compatibility with old code that looked in .mime */ 3163 if (strstr(fn, ".mime") != NULL) 3164 ms->flags &= MAGIC_MIME_TYPE; 3165 return buf; 3166 } 3167 3168 /* 3169 * Byteswap an mmap'ed file if needed 3170 */ 3171 private void 3172 byteswap(struct magic *magic, uint32_t nmagic) 3173 { 3174 uint32_t i; 3175 for (i = 0; i < nmagic; i++) 3176 bs1(&magic[i]); 3177 } 3178 3179 /* 3180 * swap a short 3181 */ 3182 private uint16_t 3183 swap2(uint16_t sv) 3184 { 3185 uint16_t rv; 3186 uint8_t *s = (uint8_t *)(void *)&sv; 3187 uint8_t *d = (uint8_t *)(void *)&rv; 3188 d[0] = s[1]; 3189 d[1] = s[0]; 3190 return rv; 3191 } 3192 3193 /* 3194 * swap an int 3195 */ 3196 private uint32_t 3197 swap4(uint32_t sv) 3198 { 3199 uint32_t rv; 3200 uint8_t *s = (uint8_t *)(void *)&sv; 3201 uint8_t *d = (uint8_t *)(void *)&rv; 3202 d[0] = s[3]; 3203 d[1] = s[2]; 3204 d[2] = s[1]; 3205 d[3] = s[0]; 3206 return rv; 3207 } 3208 3209 /* 3210 * swap a quad 3211 */ 3212 private uint64_t 3213 swap8(uint64_t sv) 3214 { 3215 uint64_t rv; 3216 uint8_t *s = (uint8_t *)(void *)&sv; 3217 uint8_t *d = (uint8_t *)(void *)&rv; 3218 #if 0 3219 d[0] = s[3]; 3220 d[1] = s[2]; 3221 d[2] = s[1]; 3222 d[3] = s[0]; 3223 d[4] = s[7]; 3224 d[5] = s[6]; 3225 d[6] = s[5]; 3226 d[7] = s[4]; 3227 #else 3228 d[0] = s[7]; 3229 d[1] = s[6]; 3230 d[2] = s[5]; 3231 d[3] = s[4]; 3232 d[4] = s[3]; 3233 d[5] = s[2]; 3234 d[6] = s[1]; 3235 d[7] = s[0]; 3236 #endif 3237 return rv; 3238 } 3239 3240 /* 3241 * byteswap a single magic entry 3242 */ 3243 private void 3244 bs1(struct magic *m) 3245 { 3246 m->cont_level = swap2(m->cont_level); 3247 m->offset = swap4((uint32_t)m->offset); 3248 m->in_offset = swap4((uint32_t)m->in_offset); 3249 m->lineno = swap4((uint32_t)m->lineno); 3250 if (IS_STRING(m->type)) { 3251 m->str_range = swap4(m->str_range); 3252 m->str_flags = swap4(m->str_flags); 3253 } 3254 else { 3255 m->value.q = swap8(m->value.q); 3256 m->num_mask = swap8(m->num_mask); 3257 } 3258 } 3259 3260 protected size_t 3261 file_pstring_length_size(const struct magic *m) 3262 { 3263 switch (m->str_flags & PSTRING_LEN) { 3264 case PSTRING_1_LE: 3265 return 1; 3266 case PSTRING_2_LE: 3267 case PSTRING_2_BE: 3268 return 2; 3269 case PSTRING_4_LE: 3270 case PSTRING_4_BE: 3271 return 4; 3272 default: 3273 abort(); /* Impossible */ 3274 return 1; 3275 } 3276 } 3277 protected size_t 3278 file_pstring_get_length(const struct magic *m, const char *ss) 3279 { 3280 size_t len = 0; 3281 const unsigned char *s = (const unsigned char *)ss; 3282 unsigned int s3, s2, s1, s0; 3283 3284 switch (m->str_flags & PSTRING_LEN) { 3285 case PSTRING_1_LE: 3286 len = *s; 3287 break; 3288 case PSTRING_2_LE: 3289 s0 = s[0]; 3290 s1 = s[1]; 3291 len = (s1 << 8) | s0; 3292 break; 3293 case PSTRING_2_BE: 3294 s0 = s[0]; 3295 s1 = s[1]; 3296 len = (s0 << 8) | s1; 3297 break; 3298 case PSTRING_4_LE: 3299 s0 = s[0]; 3300 s1 = s[1]; 3301 s2 = s[2]; 3302 s3 = s[3]; 3303 len = (s3 << 24) | (s2 << 16) | (s1 << 8) | s0; 3304 break; 3305 case PSTRING_4_BE: 3306 s0 = s[0]; 3307 s1 = s[1]; 3308 s2 = s[2]; 3309 s3 = s[3]; 3310 len = (s0 << 24) | (s1 << 16) | (s2 << 8) | s3; 3311 break; 3312 default: 3313 abort(); /* Impossible */ 3314 } 3315 3316 if (m->str_flags & PSTRING_LENGTH_INCLUDES_ITSELF) 3317 len -= file_pstring_length_size(m); 3318 3319 return len; 3320 } 3321 3322 protected int 3323 file_magicfind(struct magic_set *ms, const char *name, struct mlist *v) 3324 { 3325 uint32_t i, j; 3326 struct mlist *mlist, *ml; 3327 3328 mlist = ms->mlist[1]; 3329 3330 for (ml = mlist->next; ml != mlist; ml = ml->next) { 3331 struct magic *ma = ml->magic; 3332 uint32_t nma = ml->nmagic; 3333 for (i = 0; i < nma; i++) { 3334 if (ma[i].type != FILE_NAME) 3335 continue; 3336 if (strcmp(ma[i].value.s, name) == 0) { 3337 v->magic = &ma[i]; 3338 for (j = i + 1; j < nma; j++) 3339 if (ma[j].cont_level == 0) 3340 break; 3341 v->nmagic = j - i; 3342 return 0; 3343 } 3344 } 3345 } 3346 return -1; 3347 } 3348