1 /* 2 * Copyright (c) Ian F. Darwin 1986-1995. 3 * Software written by Ian F. Darwin and others; 4 * maintained 1995-present by Christos Zoulas and others. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice immediately at the beginning of the file, without modification, 11 * this list of conditions, and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR 20 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 */ 28 /* 29 * apprentice - make one pass through /etc/magic, learning its secrets. 30 */ 31 32 #include "file.h" 33 34 #ifndef lint 35 FILE_RCSID("@(#)$File: apprentice.c,v 1.301 2021/02/23 00:51:11 christos Exp $") 36 #endif /* lint */ 37 38 #include "magic.h" 39 #include <stdlib.h> 40 #ifdef HAVE_UNISTD_H 41 #include <unistd.h> 42 #endif 43 #include <stddef.h> 44 #include <string.h> 45 #include <assert.h> 46 #include <ctype.h> 47 #include <fcntl.h> 48 #ifdef QUICK 49 #include <sys/mman.h> 50 #endif 51 #include <dirent.h> 52 #include <limits.h> 53 54 55 #define EATAB {while (isascii(CAST(unsigned char, *l)) && \ 56 isspace(CAST(unsigned char, *l))) ++l;} 57 #define LOWCASE(l) (isupper(CAST(unsigned char, l)) ? \ 58 tolower(CAST(unsigned char, l)) : (l)) 59 /* 60 * Work around a bug in headers on Digital Unix. 61 * At least confirmed for: OSF1 V4.0 878 62 */ 63 #if defined(__osf__) && defined(__DECC) 64 #ifdef MAP_FAILED 65 #undef MAP_FAILED 66 #endif 67 #endif 68 69 #ifndef MAP_FAILED 70 #define MAP_FAILED (void *) -1 71 #endif 72 73 #ifndef MAP_FILE 74 #define MAP_FILE 0 75 #endif 76 77 #define ALLOC_CHUNK CAST(size_t, 10) 78 #define ALLOC_INCR CAST(size_t, 200) 79 80 #define MAP_TYPE_USER 0 81 #define MAP_TYPE_MALLOC 1 82 #define MAP_TYPE_MMAP 2 83 84 struct magic_entry { 85 struct magic *mp; 86 uint32_t cont_count; 87 uint32_t max_count; 88 }; 89 90 struct magic_entry_set { 91 struct magic_entry *me; 92 uint32_t count; 93 uint32_t max; 94 }; 95 96 struct magic_map { 97 void *p; 98 size_t len; 99 int type; 100 struct magic *magic[MAGIC_SETS]; 101 uint32_t nmagic[MAGIC_SETS]; 102 }; 103 104 int file_formats[FILE_NAMES_SIZE]; 105 const size_t file_nformats = FILE_NAMES_SIZE; 106 const char *file_names[FILE_NAMES_SIZE]; 107 const size_t file_nnames = FILE_NAMES_SIZE; 108 109 private int getvalue(struct magic_set *ms, struct magic *, const char **, int); 110 private int hextoint(int); 111 private const char *getstr(struct magic_set *, struct magic *, const char *, 112 int); 113 private int parse(struct magic_set *, struct magic_entry *, const char *, 114 size_t, int); 115 private void eatsize(const char **); 116 private int apprentice_1(struct magic_set *, const char *, int); 117 private size_t apprentice_magic_strength(const struct magic *); 118 private int apprentice_sort(const void *, const void *); 119 private void apprentice_list(struct mlist *, int ); 120 private struct magic_map *apprentice_load(struct magic_set *, 121 const char *, int); 122 private struct mlist *mlist_alloc(void); 123 private void mlist_free_all(struct magic_set *); 124 private void mlist_free(struct mlist *); 125 private void byteswap(struct magic *, uint32_t); 126 private void bs1(struct magic *); 127 private uint16_t swap2(uint16_t); 128 private uint32_t swap4(uint32_t); 129 private uint64_t swap8(uint64_t); 130 private char *mkdbname(struct magic_set *, const char *, int); 131 private struct magic_map *apprentice_buf(struct magic_set *, struct magic *, 132 size_t); 133 private struct magic_map *apprentice_map(struct magic_set *, const char *); 134 private int check_buffer(struct magic_set *, struct magic_map *, const char *); 135 private void apprentice_unmap(struct magic_map *); 136 private int apprentice_compile(struct magic_set *, struct magic_map *, 137 const char *); 138 private int check_format_type(const char *, int, const char **); 139 private int check_format(struct magic_set *, struct magic *); 140 private int get_op(char); 141 private int parse_mime(struct magic_set *, struct magic_entry *, const char *, 142 size_t); 143 private int parse_strength(struct magic_set *, struct magic_entry *, 144 const char *, size_t); 145 private int parse_apple(struct magic_set *, struct magic_entry *, const char *, 146 size_t); 147 private int parse_ext(struct magic_set *, struct magic_entry *, const char *, 148 size_t); 149 150 151 private size_t magicsize = sizeof(struct magic); 152 153 private const char usg_hdr[] = "cont\toffset\ttype\topcode\tmask\tvalue\tdesc"; 154 155 private struct { 156 const char *name; 157 size_t len; 158 int (*fun)(struct magic_set *, struct magic_entry *, const char *, 159 size_t); 160 } bang[] = { 161 #define DECLARE_FIELD(name) { # name, sizeof(# name) - 1, parse_ ## name } 162 DECLARE_FIELD(mime), 163 DECLARE_FIELD(apple), 164 DECLARE_FIELD(ext), 165 DECLARE_FIELD(strength), 166 #undef DECLARE_FIELD 167 { NULL, 0, NULL } 168 }; 169 170 #ifdef COMPILE_ONLY 171 172 int main(int, char *[]); 173 174 int 175 main(int argc, char *argv[]) 176 { 177 int ret; 178 struct magic_set *ms; 179 char *progname; 180 181 if ((progname = strrchr(argv[0], '/')) != NULL) 182 progname++; 183 else 184 progname = argv[0]; 185 186 if (argc != 2) { 187 (void)fprintf(stderr, "Usage: %s file\n", progname); 188 return 1; 189 } 190 191 if ((ms = magic_open(MAGIC_CHECK)) == NULL) { 192 (void)fprintf(stderr, "%s: %s\n", progname, strerror(errno)); 193 return 1; 194 } 195 ret = magic_compile(ms, argv[1]) == -1 ? 1 : 0; 196 if (ret == 1) 197 (void)fprintf(stderr, "%s: %s\n", progname, magic_error(ms)); 198 magic_close(ms); 199 return ret; 200 } 201 #endif /* COMPILE_ONLY */ 202 203 struct type_tbl_s { 204 const char name[16]; 205 const size_t len; 206 const int type; 207 const int format; 208 }; 209 210 /* 211 * XXX - the actual Single UNIX Specification says that "long" means "long", 212 * as in the C data type, but we treat it as meaning "4-byte integer". 213 * Given that the OS X version of file 5.04 did the same, I guess that passes 214 * the actual test; having "long" be dependent on how big a "long" is on 215 * the machine running "file" is silly. 216 */ 217 static const struct type_tbl_s type_tbl[] = { 218 # define XX(s) s, (sizeof(s) - 1) 219 # define XX_NULL "", 0 220 { XX("invalid"), FILE_INVALID, FILE_FMT_NONE }, 221 { XX("byte"), FILE_BYTE, FILE_FMT_NUM }, 222 { XX("short"), FILE_SHORT, FILE_FMT_NUM }, 223 { XX("default"), FILE_DEFAULT, FILE_FMT_NONE }, 224 { XX("long"), FILE_LONG, FILE_FMT_NUM }, 225 { XX("string"), FILE_STRING, FILE_FMT_STR }, 226 { XX("date"), FILE_DATE, FILE_FMT_STR }, 227 { XX("beshort"), FILE_BESHORT, FILE_FMT_NUM }, 228 { XX("belong"), FILE_BELONG, FILE_FMT_NUM }, 229 { XX("bedate"), FILE_BEDATE, FILE_FMT_STR }, 230 { XX("leshort"), FILE_LESHORT, FILE_FMT_NUM }, 231 { XX("lelong"), FILE_LELONG, FILE_FMT_NUM }, 232 { XX("ledate"), FILE_LEDATE, FILE_FMT_STR }, 233 { XX("pstring"), FILE_PSTRING, FILE_FMT_STR }, 234 { XX("ldate"), FILE_LDATE, FILE_FMT_STR }, 235 { XX("beldate"), FILE_BELDATE, FILE_FMT_STR }, 236 { XX("leldate"), FILE_LELDATE, FILE_FMT_STR }, 237 { XX("regex"), FILE_REGEX, FILE_FMT_STR }, 238 { XX("bestring16"), FILE_BESTRING16, FILE_FMT_STR }, 239 { XX("lestring16"), FILE_LESTRING16, FILE_FMT_STR }, 240 { XX("search"), FILE_SEARCH, FILE_FMT_STR }, 241 { XX("medate"), FILE_MEDATE, FILE_FMT_STR }, 242 { XX("meldate"), FILE_MELDATE, FILE_FMT_STR }, 243 { XX("melong"), FILE_MELONG, FILE_FMT_NUM }, 244 { XX("quad"), FILE_QUAD, FILE_FMT_QUAD }, 245 { XX("lequad"), FILE_LEQUAD, FILE_FMT_QUAD }, 246 { XX("bequad"), FILE_BEQUAD, FILE_FMT_QUAD }, 247 { XX("qdate"), FILE_QDATE, FILE_FMT_STR }, 248 { XX("leqdate"), FILE_LEQDATE, FILE_FMT_STR }, 249 { XX("beqdate"), FILE_BEQDATE, FILE_FMT_STR }, 250 { XX("qldate"), FILE_QLDATE, FILE_FMT_STR }, 251 { XX("leqldate"), FILE_LEQLDATE, FILE_FMT_STR }, 252 { XX("beqldate"), FILE_BEQLDATE, FILE_FMT_STR }, 253 { XX("float"), FILE_FLOAT, FILE_FMT_FLOAT }, 254 { XX("befloat"), FILE_BEFLOAT, FILE_FMT_FLOAT }, 255 { XX("lefloat"), FILE_LEFLOAT, FILE_FMT_FLOAT }, 256 { XX("double"), FILE_DOUBLE, FILE_FMT_DOUBLE }, 257 { XX("bedouble"), FILE_BEDOUBLE, FILE_FMT_DOUBLE }, 258 { XX("ledouble"), FILE_LEDOUBLE, FILE_FMT_DOUBLE }, 259 { XX("leid3"), FILE_LEID3, FILE_FMT_NUM }, 260 { XX("beid3"), FILE_BEID3, FILE_FMT_NUM }, 261 { XX("indirect"), FILE_INDIRECT, FILE_FMT_NUM }, 262 { XX("qwdate"), FILE_QWDATE, FILE_FMT_STR }, 263 { XX("leqwdate"), FILE_LEQWDATE, FILE_FMT_STR }, 264 { XX("beqwdate"), FILE_BEQWDATE, FILE_FMT_STR }, 265 { XX("name"), FILE_NAME, FILE_FMT_NONE }, 266 { XX("use"), FILE_USE, FILE_FMT_NONE }, 267 { XX("clear"), FILE_CLEAR, FILE_FMT_NONE }, 268 { XX("der"), FILE_DER, FILE_FMT_STR }, 269 { XX("guid"), FILE_GUID, FILE_FMT_STR }, 270 { XX("offset"), FILE_OFFSET, FILE_FMT_QUAD }, 271 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 272 }; 273 274 /* 275 * These are not types, and cannot be preceded by "u" to make them 276 * unsigned. 277 */ 278 static const struct type_tbl_s special_tbl[] = { 279 { XX("der"), FILE_DER, FILE_FMT_STR }, 280 { XX("name"), FILE_NAME, FILE_FMT_STR }, 281 { XX("use"), FILE_USE, FILE_FMT_STR }, 282 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 283 }; 284 # undef XX 285 # undef XX_NULL 286 287 private int 288 get_type(const struct type_tbl_s *tbl, const char *l, const char **t) 289 { 290 const struct type_tbl_s *p; 291 292 for (p = tbl; p->len; p++) { 293 if (strncmp(l, p->name, p->len) == 0) { 294 if (t) 295 *t = l + p->len; 296 break; 297 } 298 } 299 return p->type; 300 } 301 302 private off_t 303 maxoff_t(void) { 304 if (/*CONSTCOND*/sizeof(off_t) == sizeof(int)) 305 return CAST(off_t, INT_MAX); 306 if (/*CONSTCOND*/sizeof(off_t) == sizeof(long)) 307 return CAST(off_t, LONG_MAX); 308 return 0x7fffffff; 309 } 310 311 private int 312 get_standard_integer_type(const char *l, const char **t) 313 { 314 int type; 315 316 if (isalpha(CAST(unsigned char, l[1]))) { 317 switch (l[1]) { 318 case 'C': 319 /* "dC" and "uC" */ 320 type = FILE_BYTE; 321 break; 322 case 'S': 323 /* "dS" and "uS" */ 324 type = FILE_SHORT; 325 break; 326 case 'I': 327 case 'L': 328 /* 329 * "dI", "dL", "uI", and "uL". 330 * 331 * XXX - the actual Single UNIX Specification says 332 * that "L" means "long", as in the C data type, 333 * but we treat it as meaning "4-byte integer". 334 * Given that the OS X version of file 5.04 did 335 * the same, I guess that passes the actual SUS 336 * validation suite; having "dL" be dependent on 337 * how big a "long" is on the machine running 338 * "file" is silly. 339 */ 340 type = FILE_LONG; 341 break; 342 case 'Q': 343 /* "dQ" and "uQ" */ 344 type = FILE_QUAD; 345 break; 346 default: 347 /* "d{anything else}", "u{anything else}" */ 348 return FILE_INVALID; 349 } 350 l += 2; 351 } else if (isdigit(CAST(unsigned char, l[1]))) { 352 /* 353 * "d{num}" and "u{num}"; we only support {num} values 354 * of 1, 2, 4, and 8 - the Single UNIX Specification 355 * doesn't say anything about whether arbitrary 356 * values should be supported, but both the Solaris 10 357 * and OS X Mountain Lion versions of file passed the 358 * Single UNIX Specification validation suite, and 359 * neither of them support values bigger than 8 or 360 * non-power-of-2 values. 361 */ 362 if (isdigit(CAST(unsigned char, l[2]))) { 363 /* Multi-digit, so > 9 */ 364 return FILE_INVALID; 365 } 366 switch (l[1]) { 367 case '1': 368 type = FILE_BYTE; 369 break; 370 case '2': 371 type = FILE_SHORT; 372 break; 373 case '4': 374 type = FILE_LONG; 375 break; 376 case '8': 377 type = FILE_QUAD; 378 break; 379 default: 380 /* XXX - what about 3, 5, 6, or 7? */ 381 return FILE_INVALID; 382 } 383 l += 2; 384 } else { 385 /* 386 * "d" or "u" by itself. 387 */ 388 type = FILE_LONG; 389 ++l; 390 } 391 if (t) 392 *t = l; 393 return type; 394 } 395 396 private void 397 init_file_tables(void) 398 { 399 static int done = 0; 400 const struct type_tbl_s *p; 401 402 if (done) 403 return; 404 done++; 405 406 for (p = type_tbl; p->len; p++) { 407 assert(p->type < FILE_NAMES_SIZE); 408 file_names[p->type] = p->name; 409 file_formats[p->type] = p->format; 410 } 411 assert(p - type_tbl == FILE_NAMES_SIZE); 412 } 413 414 private int 415 add_mlist(struct mlist *mlp, struct magic_map *map, size_t idx) 416 { 417 struct mlist *ml; 418 419 mlp->map = NULL; 420 if ((ml = CAST(struct mlist *, malloc(sizeof(*ml)))) == NULL) 421 return -1; 422 423 ml->map = idx == 0 ? map : NULL; 424 ml->magic = map->magic[idx]; 425 ml->nmagic = map->nmagic[idx]; 426 427 mlp->prev->next = ml; 428 ml->prev = mlp->prev; 429 ml->next = mlp; 430 mlp->prev = ml; 431 return 0; 432 } 433 434 /* 435 * Handle one file or directory. 436 */ 437 private int 438 apprentice_1(struct magic_set *ms, const char *fn, int action) 439 { 440 struct magic_map *map; 441 #ifndef COMPILE_ONLY 442 struct mlist *ml; 443 size_t i; 444 #endif 445 446 if (magicsize != FILE_MAGICSIZE) { 447 file_error(ms, 0, "magic element size %lu != %lu", 448 CAST(unsigned long, sizeof(*map->magic[0])), 449 CAST(unsigned long, FILE_MAGICSIZE)); 450 return -1; 451 } 452 453 if (action == FILE_COMPILE) { 454 map = apprentice_load(ms, fn, action); 455 if (map == NULL) 456 return -1; 457 return apprentice_compile(ms, map, fn); 458 } 459 460 #ifndef COMPILE_ONLY 461 map = apprentice_map(ms, fn); 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 /* failed to add to any list, free explicitly */ 473 if (i == 0) 474 apprentice_unmap(map); 475 else 476 mlist_free_all(ms); 477 file_oomem(ms, sizeof(*ml)); 478 return -1; 479 } 480 } 481 482 if (action == FILE_LIST) { 483 for (i = 0; i < MAGIC_SETS; i++) { 484 printf("Set %" SIZE_T_FORMAT "u:\nBinary patterns:\n", 485 i); 486 apprentice_list(ms->mlist[i], BINTEST); 487 printf("Text patterns:\n"); 488 apprentice_list(ms->mlist[i], TEXTTEST); 489 } 490 } 491 return 0; 492 #else 493 return 0; 494 #endif /* COMPILE_ONLY */ 495 } 496 497 protected void 498 file_ms_free(struct magic_set *ms) 499 { 500 size_t i; 501 if (ms == NULL) 502 return; 503 for (i = 0; i < MAGIC_SETS; i++) 504 mlist_free(ms->mlist[i]); 505 free(ms->o.pbuf); 506 free(ms->o.buf); 507 free(ms->c.li); 508 free(ms); 509 } 510 511 protected struct magic_set * 512 file_ms_alloc(int flags) 513 { 514 struct magic_set *ms; 515 size_t i, len; 516 517 if ((ms = CAST(struct magic_set *, calloc(CAST(size_t, 1u), 518 sizeof(struct magic_set)))) == NULL) 519 return NULL; 520 521 if (magic_setflags(ms, flags) == -1) { 522 errno = EINVAL; 523 goto free; 524 } 525 526 ms->o.buf = ms->o.pbuf = NULL; 527 ms->o.blen = 0; 528 len = (ms->c.len = 10) * sizeof(*ms->c.li); 529 530 if ((ms->c.li = CAST(struct level_info *, malloc(len))) == NULL) 531 goto free; 532 533 ms->event_flags = 0; 534 ms->error = -1; 535 for (i = 0; i < MAGIC_SETS; i++) 536 ms->mlist[i] = NULL; 537 ms->file = "unknown"; 538 ms->line = 0; 539 ms->indir_max = FILE_INDIR_MAX; 540 ms->name_max = FILE_NAME_MAX; 541 ms->elf_shnum_max = FILE_ELF_SHNUM_MAX; 542 ms->elf_phnum_max = FILE_ELF_PHNUM_MAX; 543 ms->elf_notes_max = FILE_ELF_NOTES_MAX; 544 ms->regex_max = FILE_REGEX_MAX; 545 ms->bytes_max = FILE_BYTES_MAX; 546 ms->encoding_max = FILE_ENCODING_MAX; 547 return ms; 548 free: 549 free(ms); 550 return NULL; 551 } 552 553 private void 554 apprentice_unmap(struct magic_map *map) 555 { 556 size_t i; 557 if (map == NULL) 558 return; 559 560 switch (map->type) { 561 case MAP_TYPE_USER: 562 break; 563 case MAP_TYPE_MALLOC: 564 for (i = 0; i < MAGIC_SETS; i++) { 565 void *b = map->magic[i]; 566 void *p = map->p; 567 if (CAST(char *, b) >= CAST(char *, p) && 568 CAST(char *, b) <= CAST(char *, p) + map->len) 569 continue; 570 free(map->magic[i]); 571 } 572 free(map->p); 573 break; 574 #ifdef QUICK 575 case MAP_TYPE_MMAP: 576 if (map->p && map->p != MAP_FAILED) 577 (void)munmap(map->p, map->len); 578 break; 579 #endif 580 default: 581 abort(); 582 } 583 free(map); 584 } 585 586 private struct mlist * 587 mlist_alloc(void) 588 { 589 struct mlist *mlist; 590 if ((mlist = CAST(struct mlist *, calloc(1, sizeof(*mlist)))) == NULL) { 591 return NULL; 592 } 593 mlist->next = mlist->prev = mlist; 594 return mlist; 595 } 596 597 private void 598 mlist_free_all(struct magic_set *ms) 599 { 600 size_t i; 601 602 for (i = 0; i < MAGIC_SETS; i++) { 603 mlist_free(ms->mlist[i]); 604 ms->mlist[i] = NULL; 605 } 606 } 607 608 private void 609 mlist_free_one(struct mlist *ml) 610 { 611 if (ml->map) 612 apprentice_unmap(CAST(struct magic_map *, ml->map)); 613 free(ml); 614 } 615 616 private void 617 mlist_free(struct mlist *mlist) 618 { 619 struct mlist *ml, *next; 620 621 if (mlist == NULL) 622 return; 623 624 for (ml = mlist->next; ml != mlist;) { 625 next = ml->next; 626 mlist_free_one(ml); 627 ml = next; 628 } 629 mlist_free_one(mlist); 630 } 631 632 #ifndef COMPILE_ONLY 633 /* void **bufs: an array of compiled magic files */ 634 protected int 635 buffer_apprentice(struct magic_set *ms, struct magic **bufs, 636 size_t *sizes, size_t nbufs) 637 { 638 size_t i, j; 639 struct mlist *ml; 640 struct magic_map *map; 641 642 if (nbufs == 0) 643 return -1; 644 645 (void)file_reset(ms, 0); 646 647 init_file_tables(); 648 649 for (i = 0; i < MAGIC_SETS; i++) { 650 mlist_free(ms->mlist[i]); 651 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 652 file_oomem(ms, sizeof(*ms->mlist[i])); 653 goto fail; 654 } 655 } 656 657 for (i = 0; i < nbufs; i++) { 658 map = apprentice_buf(ms, bufs[i], sizes[i]); 659 if (map == NULL) 660 goto fail; 661 662 for (j = 0; j < MAGIC_SETS; j++) { 663 if (add_mlist(ms->mlist[j], map, j) == -1) { 664 file_oomem(ms, sizeof(*ml)); 665 goto fail; 666 } 667 } 668 } 669 670 return 0; 671 fail: 672 mlist_free_all(ms); 673 return -1; 674 } 675 #endif 676 677 /* const char *fn: list of magic files and directories */ 678 protected int 679 file_apprentice(struct magic_set *ms, const char *fn, int action) 680 { 681 char *p, *mfn; 682 int fileerr, errs = -1; 683 size_t i, j; 684 685 (void)file_reset(ms, 0); 686 687 if ((fn = magic_getpath(fn, action)) == NULL) 688 return -1; 689 690 init_file_tables(); 691 692 if ((mfn = strdup(fn)) == NULL) { 693 file_oomem(ms, strlen(fn)); 694 return -1; 695 } 696 697 for (i = 0; i < MAGIC_SETS; i++) { 698 mlist_free(ms->mlist[i]); 699 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 700 file_oomem(ms, sizeof(*ms->mlist[i])); 701 for (j = 0; j < i; j++) { 702 mlist_free(ms->mlist[j]); 703 ms->mlist[j] = NULL; 704 } 705 free(mfn); 706 return -1; 707 } 708 } 709 fn = mfn; 710 711 while (fn) { 712 p = strchr(fn, PATHSEP); 713 if (p) 714 *p++ = '\0'; 715 if (*fn == '\0') 716 break; 717 fileerr = apprentice_1(ms, fn, action); 718 errs = MAX(errs, fileerr); 719 fn = p; 720 } 721 722 free(mfn); 723 724 if (errs == -1) { 725 for (i = 0; i < MAGIC_SETS; i++) { 726 mlist_free(ms->mlist[i]); 727 ms->mlist[i] = NULL; 728 } 729 file_error(ms, 0, "could not find any valid magic files!"); 730 return -1; 731 } 732 733 #if 0 734 /* 735 * Always leave the database loaded 736 */ 737 if (action == FILE_LOAD) 738 return 0; 739 740 for (i = 0; i < MAGIC_SETS; i++) { 741 mlist_free(ms->mlist[i]); 742 ms->mlist[i] = NULL; 743 } 744 #endif 745 746 switch (action) { 747 case FILE_LOAD: 748 case FILE_COMPILE: 749 case FILE_CHECK: 750 case FILE_LIST: 751 return 0; 752 default: 753 file_error(ms, 0, "Invalid action %d", action); 754 return -1; 755 } 756 } 757 758 /* 759 * Compute the real length of a magic expression, for the purposes 760 * of determining how "strong" a magic expression is (approximating 761 * how specific its matches are): 762 * - magic characters count 0 unless escaped. 763 * - [] expressions count 1 764 * - {} expressions count 0 765 * - regular characters or escaped magic characters count 1 766 * - 0 length expressions count as one 767 */ 768 private size_t 769 nonmagic(const char *str) 770 { 771 const char *p; 772 size_t rv = 0; 773 774 for (p = str; *p; p++) 775 switch (*p) { 776 case '\\': /* Escaped anything counts 1 */ 777 if (!*++p) 778 p--; 779 rv++; 780 continue; 781 case '?': /* Magic characters count 0 */ 782 case '*': 783 case '.': 784 case '+': 785 case '^': 786 case '$': 787 continue; 788 case '[': /* Bracketed expressions count 1 the ']' */ 789 while (*p && *p != ']') 790 p++; 791 p--; 792 continue; 793 case '{': /* Braced expressions count 0 */ 794 while (*p && *p != '}') 795 p++; 796 if (!*p) 797 p--; 798 continue; 799 default: /* Anything else counts 1 */ 800 rv++; 801 continue; 802 } 803 804 return rv == 0 ? 1 : rv; /* Return at least 1 */ 805 } 806 807 808 private size_t 809 typesize(int type) 810 { 811 switch (type) { 812 case FILE_BYTE: 813 return 1; 814 815 case FILE_SHORT: 816 case FILE_LESHORT: 817 case FILE_BESHORT: 818 return 2; 819 820 case FILE_LONG: 821 case FILE_LELONG: 822 case FILE_BELONG: 823 case FILE_MELONG: 824 return 4; 825 826 case FILE_DATE: 827 case FILE_LEDATE: 828 case FILE_BEDATE: 829 case FILE_MEDATE: 830 case FILE_LDATE: 831 case FILE_LELDATE: 832 case FILE_BELDATE: 833 case FILE_MELDATE: 834 case FILE_FLOAT: 835 case FILE_BEFLOAT: 836 case FILE_LEFLOAT: 837 return 4; 838 839 case FILE_QUAD: 840 case FILE_BEQUAD: 841 case FILE_LEQUAD: 842 case FILE_QDATE: 843 case FILE_LEQDATE: 844 case FILE_BEQDATE: 845 case FILE_QLDATE: 846 case FILE_LEQLDATE: 847 case FILE_BEQLDATE: 848 case FILE_QWDATE: 849 case FILE_LEQWDATE: 850 case FILE_BEQWDATE: 851 case FILE_DOUBLE: 852 case FILE_BEDOUBLE: 853 case FILE_LEDOUBLE: 854 case FILE_OFFSET: 855 return 8; 856 857 case FILE_GUID: 858 return 16; 859 860 default: 861 return FILE_BADSIZE; 862 } 863 } 864 865 /* 866 * Get weight of this magic entry, for sorting purposes. 867 */ 868 private size_t 869 apprentice_magic_strength(const struct magic *m) 870 { 871 #define MULT 10U 872 size_t ts, v; 873 ssize_t val = 2 * MULT; /* baseline strength */ 874 875 switch (m->type) { 876 case FILE_DEFAULT: /* make sure this sorts last */ 877 if (m->factor_op != FILE_FACTOR_OP_NONE) 878 abort(); 879 return 0; 880 881 case FILE_BYTE: 882 case FILE_SHORT: 883 case FILE_LESHORT: 884 case FILE_BESHORT: 885 case FILE_LONG: 886 case FILE_LELONG: 887 case FILE_BELONG: 888 case FILE_MELONG: 889 case FILE_DATE: 890 case FILE_LEDATE: 891 case FILE_BEDATE: 892 case FILE_MEDATE: 893 case FILE_LDATE: 894 case FILE_LELDATE: 895 case FILE_BELDATE: 896 case FILE_MELDATE: 897 case FILE_FLOAT: 898 case FILE_BEFLOAT: 899 case FILE_LEFLOAT: 900 case FILE_QUAD: 901 case FILE_BEQUAD: 902 case FILE_LEQUAD: 903 case FILE_QDATE: 904 case FILE_LEQDATE: 905 case FILE_BEQDATE: 906 case FILE_QLDATE: 907 case FILE_LEQLDATE: 908 case FILE_BEQLDATE: 909 case FILE_QWDATE: 910 case FILE_LEQWDATE: 911 case FILE_BEQWDATE: 912 case FILE_DOUBLE: 913 case FILE_BEDOUBLE: 914 case FILE_LEDOUBLE: 915 case FILE_GUID: 916 case FILE_OFFSET: 917 ts = typesize(m->type); 918 if (ts == FILE_BADSIZE) 919 abort(); 920 val += ts * MULT; 921 break; 922 923 case FILE_PSTRING: 924 case FILE_STRING: 925 val += m->vallen * MULT; 926 break; 927 928 case FILE_BESTRING16: 929 case FILE_LESTRING16: 930 val += m->vallen * MULT / 2; 931 break; 932 933 case FILE_SEARCH: 934 if (m->vallen == 0) 935 break; 936 val += m->vallen * MAX(MULT / m->vallen, 1); 937 break; 938 939 case FILE_REGEX: 940 v = nonmagic(m->value.s); 941 val += v * MAX(MULT / v, 1); 942 break; 943 944 case FILE_INDIRECT: 945 case FILE_NAME: 946 case FILE_USE: 947 break; 948 949 case FILE_DER: 950 val += MULT; 951 break; 952 953 default: 954 (void)fprintf(stderr, "Bad type %d\n", m->type); 955 abort(); 956 } 957 958 switch (m->reln) { 959 case 'x': /* matches anything penalize */ 960 case '!': /* matches almost anything penalize */ 961 val = 0; 962 break; 963 964 case '=': /* Exact match, prefer */ 965 val += MULT; 966 break; 967 968 case '>': 969 case '<': /* comparison match reduce strength */ 970 val -= 2 * MULT; 971 break; 972 973 case '^': 974 case '&': /* masking bits, we could count them too */ 975 val -= MULT; 976 break; 977 978 default: 979 (void)fprintf(stderr, "Bad relation %c\n", m->reln); 980 abort(); 981 } 982 983 switch (m->factor_op) { 984 case FILE_FACTOR_OP_NONE: 985 break; 986 case FILE_FACTOR_OP_PLUS: 987 val += m->factor; 988 break; 989 case FILE_FACTOR_OP_MINUS: 990 val -= m->factor; 991 break; 992 case FILE_FACTOR_OP_TIMES: 993 val *= m->factor; 994 break; 995 case FILE_FACTOR_OP_DIV: 996 val /= m->factor; 997 break; 998 default: 999 abort(); 1000 } 1001 1002 if (val <= 0) /* ensure we only return 0 for FILE_DEFAULT */ 1003 val = 1; 1004 1005 /* 1006 * Magic entries with no description get a bonus because they depend 1007 * on subsequent magic entries to print something. 1008 */ 1009 if (m->desc[0] == '\0') 1010 val++; 1011 return val; 1012 } 1013 1014 /* 1015 * Sort callback for sorting entries by "strength" (basically length) 1016 */ 1017 private int 1018 apprentice_sort(const void *a, const void *b) 1019 { 1020 const struct magic_entry *ma = CAST(const struct magic_entry *, a); 1021 const struct magic_entry *mb = CAST(const struct magic_entry *, b); 1022 size_t sa = apprentice_magic_strength(ma->mp); 1023 size_t sb = apprentice_magic_strength(mb->mp); 1024 if (sa == sb) 1025 return 0; 1026 else if (sa > sb) 1027 return -1; 1028 else 1029 return 1; 1030 } 1031 1032 /* 1033 * Shows sorted patterns list in the order which is used for the matching 1034 */ 1035 private void 1036 apprentice_list(struct mlist *mlist, int mode) 1037 { 1038 uint32_t magindex = 0; 1039 struct mlist *ml; 1040 for (ml = mlist->next; ml != mlist; ml = ml->next) { 1041 for (magindex = 0; magindex < ml->nmagic; magindex++) { 1042 struct magic *m = &ml->magic[magindex]; 1043 if ((m->flag & mode) != mode) { 1044 /* Skip sub-tests */ 1045 while (magindex + 1 < ml->nmagic && 1046 ml->magic[magindex + 1].cont_level != 0) 1047 ++magindex; 1048 continue; /* Skip to next top-level test*/ 1049 } 1050 1051 /* 1052 * Try to iterate over the tree until we find item with 1053 * description/mimetype. 1054 */ 1055 while (magindex + 1 < ml->nmagic && 1056 ml->magic[magindex + 1].cont_level != 0 && 1057 *ml->magic[magindex].desc == '\0' && 1058 *ml->magic[magindex].mimetype == '\0') 1059 magindex++; 1060 1061 printf("Strength = %3" SIZE_T_FORMAT "u@%u: %s [%s]\n", 1062 apprentice_magic_strength(m), 1063 ml->magic[magindex].lineno, 1064 ml->magic[magindex].desc, 1065 ml->magic[magindex].mimetype); 1066 } 1067 } 1068 } 1069 1070 private void 1071 set_test_type(struct magic *mstart, struct magic *m) 1072 { 1073 switch (m->type) { 1074 case FILE_BYTE: 1075 case FILE_SHORT: 1076 case FILE_LONG: 1077 case FILE_DATE: 1078 case FILE_BESHORT: 1079 case FILE_BELONG: 1080 case FILE_BEDATE: 1081 case FILE_LESHORT: 1082 case FILE_LELONG: 1083 case FILE_LEDATE: 1084 case FILE_LDATE: 1085 case FILE_BELDATE: 1086 case FILE_LELDATE: 1087 case FILE_MEDATE: 1088 case FILE_MELDATE: 1089 case FILE_MELONG: 1090 case FILE_QUAD: 1091 case FILE_LEQUAD: 1092 case FILE_BEQUAD: 1093 case FILE_QDATE: 1094 case FILE_LEQDATE: 1095 case FILE_BEQDATE: 1096 case FILE_QLDATE: 1097 case FILE_LEQLDATE: 1098 case FILE_BEQLDATE: 1099 case FILE_QWDATE: 1100 case FILE_LEQWDATE: 1101 case FILE_BEQWDATE: 1102 case FILE_FLOAT: 1103 case FILE_BEFLOAT: 1104 case FILE_LEFLOAT: 1105 case FILE_DOUBLE: 1106 case FILE_BEDOUBLE: 1107 case FILE_LEDOUBLE: 1108 case FILE_DER: 1109 case FILE_GUID: 1110 case FILE_OFFSET: 1111 mstart->flag |= BINTEST; 1112 break; 1113 case FILE_STRING: 1114 case FILE_PSTRING: 1115 case FILE_BESTRING16: 1116 case FILE_LESTRING16: 1117 /* Allow text overrides */ 1118 if (mstart->str_flags & STRING_TEXTTEST) 1119 mstart->flag |= TEXTTEST; 1120 else 1121 mstart->flag |= BINTEST; 1122 break; 1123 case FILE_REGEX: 1124 case FILE_SEARCH: 1125 /* Check for override */ 1126 if (mstart->str_flags & STRING_BINTEST) 1127 mstart->flag |= BINTEST; 1128 if (mstart->str_flags & STRING_TEXTTEST) 1129 mstart->flag |= TEXTTEST; 1130 1131 if (mstart->flag & (TEXTTEST|BINTEST)) 1132 break; 1133 1134 /* binary test if pattern is not text */ 1135 if (file_looks_utf8(m->value.us, CAST(size_t, m->vallen), NULL, 1136 NULL) <= 0) 1137 mstart->flag |= BINTEST; 1138 else 1139 mstart->flag |= TEXTTEST; 1140 break; 1141 case FILE_DEFAULT: 1142 /* can't deduce anything; we shouldn't see this at the 1143 top level anyway */ 1144 break; 1145 case FILE_INVALID: 1146 default: 1147 /* invalid search type, but no need to complain here */ 1148 break; 1149 } 1150 } 1151 1152 private int 1153 addentry(struct magic_set *ms, struct magic_entry *me, 1154 struct magic_entry_set *mset) 1155 { 1156 size_t i = me->mp->type == FILE_NAME ? 1 : 0; 1157 if (mset[i].count == mset[i].max) { 1158 struct magic_entry *mp; 1159 1160 mset[i].max += ALLOC_INCR; 1161 if ((mp = CAST(struct magic_entry *, 1162 realloc(mset[i].me, sizeof(*mp) * mset[i].max))) == 1163 NULL) { 1164 file_oomem(ms, sizeof(*mp) * mset[i].max); 1165 return -1; 1166 } 1167 (void)memset(&mp[mset[i].count], 0, sizeof(*mp) * 1168 ALLOC_INCR); 1169 mset[i].me = mp; 1170 } 1171 mset[i].me[mset[i].count++] = *me; 1172 memset(me, 0, sizeof(*me)); 1173 return 0; 1174 } 1175 1176 /* 1177 * Load and parse one file. 1178 */ 1179 private void 1180 load_1(struct magic_set *ms, int action, const char *fn, int *errs, 1181 struct magic_entry_set *mset) 1182 { 1183 size_t lineno = 0, llen = 0; 1184 char *line = NULL; 1185 ssize_t len; 1186 struct magic_entry me; 1187 1188 FILE *f = fopen(ms->file = fn, "r"); 1189 if (f == NULL) { 1190 if (errno != ENOENT) 1191 file_error(ms, errno, "cannot read magic file `%s'", 1192 fn); 1193 (*errs)++; 1194 return; 1195 } 1196 1197 memset(&me, 0, sizeof(me)); 1198 /* read and parse this file */ 1199 for (ms->line = 1; (len = getline(&line, &llen, f)) != -1; 1200 ms->line++) { 1201 if (len == 0) /* null line, garbage, etc */ 1202 continue; 1203 if (line[len - 1] == '\n') { 1204 lineno++; 1205 line[len - 1] = '\0'; /* delete newline */ 1206 } 1207 switch (line[0]) { 1208 case '\0': /* empty, do not parse */ 1209 case '#': /* comment, do not parse */ 1210 continue; 1211 case '!': 1212 if (line[1] == ':') { 1213 size_t i; 1214 1215 for (i = 0; bang[i].name != NULL; i++) { 1216 if (CAST(size_t, len - 2) > bang[i].len && 1217 memcmp(bang[i].name, line + 2, 1218 bang[i].len) == 0) 1219 break; 1220 } 1221 if (bang[i].name == NULL) { 1222 file_error(ms, 0, 1223 "Unknown !: entry `%s'", line); 1224 (*errs)++; 1225 continue; 1226 } 1227 if (me.mp == NULL) { 1228 file_error(ms, 0, 1229 "No current entry for :!%s type", 1230 bang[i].name); 1231 (*errs)++; 1232 continue; 1233 } 1234 if ((*bang[i].fun)(ms, &me, 1235 line + bang[i].len + 2, 1236 len - bang[i].len - 2) != 0) { 1237 (*errs)++; 1238 continue; 1239 } 1240 continue; 1241 } 1242 /*FALLTHROUGH*/ 1243 default: 1244 again: 1245 switch (parse(ms, &me, line, lineno, action)) { 1246 case 0: 1247 continue; 1248 case 1: 1249 (void)addentry(ms, &me, mset); 1250 goto again; 1251 default: 1252 (*errs)++; 1253 break; 1254 } 1255 } 1256 } 1257 if (me.mp) 1258 (void)addentry(ms, &me, mset); 1259 free(line); 1260 (void)fclose(f); 1261 } 1262 1263 /* 1264 * parse a file or directory of files 1265 * const char *fn: name of magic file or directory 1266 */ 1267 private int 1268 cmpstrp(const void *p1, const void *p2) 1269 { 1270 return strcmp(*RCAST(char *const *, p1), *RCAST(char *const *, p2)); 1271 } 1272 1273 1274 private uint32_t 1275 set_text_binary(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1276 uint32_t starttest) 1277 { 1278 static const char text[] = "text"; 1279 static const char binary[] = "binary"; 1280 static const size_t len = sizeof(text); 1281 1282 uint32_t i = starttest; 1283 1284 do { 1285 set_test_type(me[starttest].mp, me[i].mp); 1286 if ((ms->flags & MAGIC_DEBUG) == 0) 1287 continue; 1288 (void)fprintf(stderr, "%s%s%s: %s\n", 1289 me[i].mp->mimetype, 1290 me[i].mp->mimetype[0] == '\0' ? "" : "; ", 1291 me[i].mp->desc[0] ? me[i].mp->desc : "(no description)", 1292 me[i].mp->flag & BINTEST ? binary : text); 1293 if (me[i].mp->flag & BINTEST) { 1294 char *p = strstr(me[i].mp->desc, text); 1295 if (p && (p == me[i].mp->desc || 1296 isspace(CAST(unsigned char, p[-1]))) && 1297 (p + len - me[i].mp->desc == MAXstring 1298 || (p[len] == '\0' || 1299 isspace(CAST(unsigned char, p[len]))))) 1300 (void)fprintf(stderr, "*** Possible " 1301 "binary test for text type\n"); 1302 } 1303 } while (++i < nme && me[i].mp->cont_level != 0); 1304 return i; 1305 } 1306 1307 private void 1308 set_last_default(struct magic_set *ms, struct magic_entry *me, uint32_t nme) 1309 { 1310 uint32_t i; 1311 for (i = 0; i < nme; i++) { 1312 if (me[i].mp->cont_level == 0 && 1313 me[i].mp->type == FILE_DEFAULT) { 1314 while (++i < nme) 1315 if (me[i].mp->cont_level == 0) 1316 break; 1317 if (i != nme) { 1318 /* XXX - Ugh! */ 1319 ms->line = me[i].mp->lineno; 1320 file_magwarn(ms, 1321 "level 0 \"default\" did not sort last"); 1322 } 1323 return; 1324 } 1325 } 1326 } 1327 1328 private int 1329 coalesce_entries(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1330 struct magic **ma, uint32_t *nma) 1331 { 1332 uint32_t i, mentrycount = 0; 1333 size_t slen; 1334 1335 for (i = 0; i < nme; i++) 1336 mentrycount += me[i].cont_count; 1337 1338 slen = sizeof(**ma) * mentrycount; 1339 if ((*ma = CAST(struct magic *, malloc(slen))) == NULL) { 1340 file_oomem(ms, slen); 1341 return -1; 1342 } 1343 1344 mentrycount = 0; 1345 for (i = 0; i < nme; i++) { 1346 (void)memcpy(*ma + mentrycount, me[i].mp, 1347 me[i].cont_count * sizeof(**ma)); 1348 mentrycount += me[i].cont_count; 1349 } 1350 *nma = mentrycount; 1351 return 0; 1352 } 1353 1354 private void 1355 magic_entry_free(struct magic_entry *me, uint32_t nme) 1356 { 1357 uint32_t i; 1358 if (me == NULL) 1359 return; 1360 for (i = 0; i < nme; i++) 1361 free(me[i].mp); 1362 free(me); 1363 } 1364 1365 private struct magic_map * 1366 apprentice_load(struct magic_set *ms, const char *fn, int action) 1367 { 1368 int errs = 0; 1369 uint32_t i, j; 1370 size_t files = 0, maxfiles = 0; 1371 char **filearr = NULL, *mfn; 1372 struct stat st; 1373 struct magic_map *map; 1374 struct magic_entry_set mset[MAGIC_SETS]; 1375 DIR *dir; 1376 struct dirent *d; 1377 1378 memset(mset, 0, sizeof(mset)); 1379 ms->flags |= MAGIC_CHECK; /* Enable checks for parsed files */ 1380 1381 1382 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) 1383 { 1384 file_oomem(ms, sizeof(*map)); 1385 return NULL; 1386 } 1387 map->type = MAP_TYPE_MALLOC; 1388 1389 /* print silly verbose header for USG compat. */ 1390 if (action == FILE_CHECK) 1391 (void)fprintf(stderr, "%s\n", usg_hdr); 1392 1393 /* load directory or file */ 1394 if (stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) { 1395 dir = opendir(fn); 1396 if (!dir) { 1397 errs++; 1398 goto out; 1399 } 1400 while ((d = readdir(dir)) != NULL) { 1401 if (d->d_name[0] == '.') 1402 continue; 1403 if (asprintf(&mfn, "%s/%s", fn, d->d_name) < 0) { 1404 file_oomem(ms, 1405 strlen(fn) + strlen(d->d_name) + 2); 1406 errs++; 1407 closedir(dir); 1408 goto out; 1409 } 1410 if (stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) { 1411 free(mfn); 1412 continue; 1413 } 1414 if (files >= maxfiles) { 1415 size_t mlen; 1416 char **nfilearr; 1417 maxfiles = (maxfiles + 1) * 2; 1418 mlen = maxfiles * sizeof(*filearr); 1419 if ((nfilearr = CAST(char **, 1420 realloc(filearr, mlen))) == NULL) { 1421 file_oomem(ms, mlen); 1422 free(mfn); 1423 closedir(dir); 1424 errs++; 1425 goto out; 1426 } 1427 filearr = nfilearr; 1428 } 1429 filearr[files++] = mfn; 1430 } 1431 closedir(dir); 1432 if (filearr) { 1433 qsort(filearr, files, sizeof(*filearr), cmpstrp); 1434 for (i = 0; i < files; i++) { 1435 load_1(ms, action, filearr[i], &errs, mset); 1436 free(filearr[i]); 1437 } 1438 free(filearr); 1439 filearr = NULL; 1440 } 1441 } else 1442 load_1(ms, action, fn, &errs, mset); 1443 if (errs) 1444 goto out; 1445 1446 for (j = 0; j < MAGIC_SETS; j++) { 1447 /* Set types of tests */ 1448 for (i = 0; i < mset[j].count; ) { 1449 if (mset[j].me[i].mp->cont_level != 0) { 1450 i++; 1451 continue; 1452 } 1453 i = set_text_binary(ms, mset[j].me, mset[j].count, i); 1454 } 1455 if (mset[j].me) 1456 qsort(mset[j].me, mset[j].count, sizeof(*mset[j].me), 1457 apprentice_sort); 1458 1459 /* 1460 * Make sure that any level 0 "default" line is last 1461 * (if one exists). 1462 */ 1463 set_last_default(ms, mset[j].me, mset[j].count); 1464 1465 /* coalesce per file arrays into a single one, if needed */ 1466 if (mset[j].count == 0) 1467 continue; 1468 1469 if (coalesce_entries(ms, mset[j].me, mset[j].count, 1470 &map->magic[j], &map->nmagic[j]) == -1) { 1471 errs++; 1472 goto out; 1473 } 1474 } 1475 1476 out: 1477 free(filearr); 1478 for (j = 0; j < MAGIC_SETS; j++) 1479 magic_entry_free(mset[j].me, mset[j].count); 1480 1481 if (errs) { 1482 apprentice_unmap(map); 1483 return NULL; 1484 } 1485 return map; 1486 } 1487 1488 /* 1489 * extend the sign bit if the comparison is to be signed 1490 */ 1491 protected uint64_t 1492 file_signextend(struct magic_set *ms, struct magic *m, uint64_t v) 1493 { 1494 if (!(m->flag & UNSIGNED)) { 1495 switch(m->type) { 1496 /* 1497 * Do not remove the casts below. They are 1498 * vital. When later compared with the data, 1499 * the sign extension must have happened. 1500 */ 1501 case FILE_BYTE: 1502 v = CAST(signed char, v); 1503 break; 1504 case FILE_SHORT: 1505 case FILE_BESHORT: 1506 case FILE_LESHORT: 1507 v = CAST(short, v); 1508 break; 1509 case FILE_DATE: 1510 case FILE_BEDATE: 1511 case FILE_LEDATE: 1512 case FILE_MEDATE: 1513 case FILE_LDATE: 1514 case FILE_BELDATE: 1515 case FILE_LELDATE: 1516 case FILE_MELDATE: 1517 case FILE_LONG: 1518 case FILE_BELONG: 1519 case FILE_LELONG: 1520 case FILE_MELONG: 1521 case FILE_FLOAT: 1522 case FILE_BEFLOAT: 1523 case FILE_LEFLOAT: 1524 v = CAST(int32_t, v); 1525 break; 1526 case FILE_QUAD: 1527 case FILE_BEQUAD: 1528 case FILE_LEQUAD: 1529 case FILE_QDATE: 1530 case FILE_QLDATE: 1531 case FILE_QWDATE: 1532 case FILE_BEQDATE: 1533 case FILE_BEQLDATE: 1534 case FILE_BEQWDATE: 1535 case FILE_LEQDATE: 1536 case FILE_LEQLDATE: 1537 case FILE_LEQWDATE: 1538 case FILE_DOUBLE: 1539 case FILE_BEDOUBLE: 1540 case FILE_LEDOUBLE: 1541 case FILE_OFFSET: 1542 v = CAST(int64_t, v); 1543 break; 1544 case FILE_STRING: 1545 case FILE_PSTRING: 1546 case FILE_BESTRING16: 1547 case FILE_LESTRING16: 1548 case FILE_REGEX: 1549 case FILE_SEARCH: 1550 case FILE_DEFAULT: 1551 case FILE_INDIRECT: 1552 case FILE_NAME: 1553 case FILE_USE: 1554 case FILE_CLEAR: 1555 case FILE_DER: 1556 case FILE_GUID: 1557 break; 1558 default: 1559 if (ms->flags & MAGIC_CHECK) 1560 file_magwarn(ms, "cannot happen: m->type=%d\n", 1561 m->type); 1562 return FILE_BADSIZE; 1563 } 1564 } 1565 return v; 1566 } 1567 1568 private int 1569 string_modifier_check(struct magic_set *ms, struct magic *m) 1570 { 1571 if ((ms->flags & MAGIC_CHECK) == 0) 1572 return 0; 1573 1574 if ((m->type != FILE_REGEX || (m->str_flags & REGEX_LINE_COUNT) == 0) && 1575 (m->type != FILE_PSTRING && (m->str_flags & PSTRING_LEN) != 0)) { 1576 file_magwarn(ms, 1577 "'/BHhLl' modifiers are only allowed for pascal strings\n"); 1578 return -1; 1579 } 1580 switch (m->type) { 1581 case FILE_BESTRING16: 1582 case FILE_LESTRING16: 1583 if (m->str_flags != 0) { 1584 file_magwarn(ms, 1585 "no modifiers allowed for 16-bit strings\n"); 1586 return -1; 1587 } 1588 break; 1589 case FILE_STRING: 1590 case FILE_PSTRING: 1591 if ((m->str_flags & REGEX_OFFSET_START) != 0) { 1592 file_magwarn(ms, 1593 "'/%c' only allowed on regex and search\n", 1594 CHAR_REGEX_OFFSET_START); 1595 return -1; 1596 } 1597 break; 1598 case FILE_SEARCH: 1599 if (m->str_range == 0) { 1600 file_magwarn(ms, 1601 "missing range; defaulting to %d\n", 1602 STRING_DEFAULT_RANGE); 1603 m->str_range = STRING_DEFAULT_RANGE; 1604 return -1; 1605 } 1606 break; 1607 case FILE_REGEX: 1608 if ((m->str_flags & STRING_COMPACT_WHITESPACE) != 0) { 1609 file_magwarn(ms, "'/%c' not allowed on regex\n", 1610 CHAR_COMPACT_WHITESPACE); 1611 return -1; 1612 } 1613 if ((m->str_flags & STRING_COMPACT_OPTIONAL_WHITESPACE) != 0) { 1614 file_magwarn(ms, "'/%c' not allowed on regex\n", 1615 CHAR_COMPACT_OPTIONAL_WHITESPACE); 1616 return -1; 1617 } 1618 break; 1619 default: 1620 file_magwarn(ms, "coding error: m->type=%d\n", 1621 m->type); 1622 return -1; 1623 } 1624 return 0; 1625 } 1626 1627 private int 1628 get_op(char c) 1629 { 1630 switch (c) { 1631 case '&': 1632 return FILE_OPAND; 1633 case '|': 1634 return FILE_OPOR; 1635 case '^': 1636 return FILE_OPXOR; 1637 case '+': 1638 return FILE_OPADD; 1639 case '-': 1640 return FILE_OPMINUS; 1641 case '*': 1642 return FILE_OPMULTIPLY; 1643 case '/': 1644 return FILE_OPDIVIDE; 1645 case '%': 1646 return FILE_OPMODULO; 1647 default: 1648 return -1; 1649 } 1650 } 1651 1652 #ifdef ENABLE_CONDITIONALS 1653 private int 1654 get_cond(const char *l, const char **t) 1655 { 1656 static const struct cond_tbl_s { 1657 char name[8]; 1658 size_t len; 1659 int cond; 1660 } cond_tbl[] = { 1661 { "if", 2, COND_IF }, 1662 { "elif", 4, COND_ELIF }, 1663 { "else", 4, COND_ELSE }, 1664 { "", 0, COND_NONE }, 1665 }; 1666 const struct cond_tbl_s *p; 1667 1668 for (p = cond_tbl; p->len; p++) { 1669 if (strncmp(l, p->name, p->len) == 0 && 1670 isspace(CAST(unsigned char, l[p->len]))) { 1671 if (t) 1672 *t = l + p->len; 1673 break; 1674 } 1675 } 1676 return p->cond; 1677 } 1678 1679 private int 1680 check_cond(struct magic_set *ms, int cond, uint32_t cont_level) 1681 { 1682 int last_cond; 1683 last_cond = ms->c.li[cont_level].last_cond; 1684 1685 switch (cond) { 1686 case COND_IF: 1687 if (last_cond != COND_NONE && last_cond != COND_ELIF) { 1688 if (ms->flags & MAGIC_CHECK) 1689 file_magwarn(ms, "syntax error: `if'"); 1690 return -1; 1691 } 1692 last_cond = COND_IF; 1693 break; 1694 1695 case COND_ELIF: 1696 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1697 if (ms->flags & MAGIC_CHECK) 1698 file_magwarn(ms, "syntax error: `elif'"); 1699 return -1; 1700 } 1701 last_cond = COND_ELIF; 1702 break; 1703 1704 case COND_ELSE: 1705 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1706 if (ms->flags & MAGIC_CHECK) 1707 file_magwarn(ms, "syntax error: `else'"); 1708 return -1; 1709 } 1710 last_cond = COND_NONE; 1711 break; 1712 1713 case COND_NONE: 1714 last_cond = COND_NONE; 1715 break; 1716 } 1717 1718 ms->c.li[cont_level].last_cond = last_cond; 1719 return 0; 1720 } 1721 #endif /* ENABLE_CONDITIONALS */ 1722 1723 private int 1724 parse_indirect_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1725 { 1726 const char *l = *lp; 1727 1728 while (!isspace(CAST(unsigned char, *++l))) 1729 switch (*l) { 1730 case CHAR_INDIRECT_RELATIVE: 1731 m->str_flags |= INDIRECT_RELATIVE; 1732 break; 1733 default: 1734 if (ms->flags & MAGIC_CHECK) 1735 file_magwarn(ms, "indirect modifier `%c' " 1736 "invalid", *l); 1737 *lp = l; 1738 return -1; 1739 } 1740 *lp = l; 1741 return 0; 1742 } 1743 1744 private void 1745 parse_op_modifier(struct magic_set *ms, struct magic *m, const char **lp, 1746 int op) 1747 { 1748 const char *l = *lp; 1749 char *t; 1750 uint64_t val; 1751 1752 ++l; 1753 m->mask_op |= op; 1754 val = CAST(uint64_t, strtoull(l, &t, 0)); 1755 l = t; 1756 m->num_mask = file_signextend(ms, m, val); 1757 eatsize(&l); 1758 *lp = l; 1759 } 1760 1761 private int 1762 parse_string_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1763 { 1764 const char *l = *lp; 1765 char *t; 1766 int have_range = 0; 1767 1768 while (!isspace(CAST(unsigned char, *++l))) { 1769 switch (*l) { 1770 case '0': case '1': case '2': 1771 case '3': case '4': case '5': 1772 case '6': case '7': case '8': 1773 case '9': 1774 if (have_range && (ms->flags & MAGIC_CHECK)) 1775 file_magwarn(ms, "multiple ranges"); 1776 have_range = 1; 1777 m->str_range = CAST(uint32_t, strtoul(l, &t, 0)); 1778 if (m->str_range == 0) 1779 file_magwarn(ms, "zero range"); 1780 l = t - 1; 1781 break; 1782 case CHAR_COMPACT_WHITESPACE: 1783 m->str_flags |= STRING_COMPACT_WHITESPACE; 1784 break; 1785 case CHAR_COMPACT_OPTIONAL_WHITESPACE: 1786 m->str_flags |= STRING_COMPACT_OPTIONAL_WHITESPACE; 1787 break; 1788 case CHAR_IGNORE_LOWERCASE: 1789 m->str_flags |= STRING_IGNORE_LOWERCASE; 1790 break; 1791 case CHAR_IGNORE_UPPERCASE: 1792 m->str_flags |= STRING_IGNORE_UPPERCASE; 1793 break; 1794 case CHAR_REGEX_OFFSET_START: 1795 m->str_flags |= REGEX_OFFSET_START; 1796 break; 1797 case CHAR_BINTEST: 1798 m->str_flags |= STRING_BINTEST; 1799 break; 1800 case CHAR_TEXTTEST: 1801 m->str_flags |= STRING_TEXTTEST; 1802 break; 1803 case CHAR_TRIM: 1804 m->str_flags |= STRING_TRIM; 1805 break; 1806 case CHAR_PSTRING_1_LE: 1807 #define SET_LENGTH(a) m->str_flags = (m->str_flags & ~PSTRING_LEN) | (a) 1808 if (m->type != FILE_PSTRING) 1809 goto bad; 1810 SET_LENGTH(PSTRING_1_LE); 1811 break; 1812 case CHAR_PSTRING_2_BE: 1813 if (m->type != FILE_PSTRING) 1814 goto bad; 1815 SET_LENGTH(PSTRING_2_BE); 1816 break; 1817 case CHAR_PSTRING_2_LE: 1818 if (m->type != FILE_PSTRING) 1819 goto bad; 1820 SET_LENGTH(PSTRING_2_LE); 1821 break; 1822 case CHAR_PSTRING_4_BE: 1823 if (m->type != FILE_PSTRING) 1824 goto bad; 1825 SET_LENGTH(PSTRING_4_BE); 1826 break; 1827 case CHAR_PSTRING_4_LE: 1828 switch (m->type) { 1829 case FILE_PSTRING: 1830 case FILE_REGEX: 1831 break; 1832 default: 1833 goto bad; 1834 } 1835 SET_LENGTH(PSTRING_4_LE); 1836 break; 1837 case CHAR_PSTRING_LENGTH_INCLUDES_ITSELF: 1838 if (m->type != FILE_PSTRING) 1839 goto bad; 1840 m->str_flags |= PSTRING_LENGTH_INCLUDES_ITSELF; 1841 break; 1842 default: 1843 bad: 1844 if (ms->flags & MAGIC_CHECK) 1845 file_magwarn(ms, "string modifier `%c' " 1846 "invalid", *l); 1847 goto out; 1848 } 1849 /* allow multiple '/' for readability */ 1850 if (l[1] == '/' && !isspace(CAST(unsigned char, l[2]))) 1851 l++; 1852 } 1853 if (string_modifier_check(ms, m) == -1) 1854 goto out; 1855 *lp = l; 1856 return 0; 1857 out: 1858 *lp = l; 1859 return -1; 1860 } 1861 1862 /* 1863 * parse one line from magic file, put into magic[index++] if valid 1864 */ 1865 private int 1866 parse(struct magic_set *ms, struct magic_entry *me, const char *line, 1867 size_t lineno, int action) 1868 { 1869 #ifdef ENABLE_CONDITIONALS 1870 static uint32_t last_cont_level = 0; 1871 #endif 1872 size_t i; 1873 struct magic *m; 1874 const char *l = line; 1875 char *t; 1876 int op; 1877 uint32_t cont_level; 1878 int32_t diff; 1879 1880 cont_level = 0; 1881 1882 /* 1883 * Parse the offset. 1884 */ 1885 while (*l == '>') { 1886 ++l; /* step over */ 1887 cont_level++; 1888 } 1889 #ifdef ENABLE_CONDITIONALS 1890 if (cont_level == 0 || cont_level > last_cont_level) 1891 if (file_check_mem(ms, cont_level) == -1) 1892 return -1; 1893 last_cont_level = cont_level; 1894 #endif 1895 if (cont_level != 0) { 1896 if (me->mp == NULL) { 1897 file_magerror(ms, "No current entry for continuation"); 1898 return -1; 1899 } 1900 if (me->cont_count == 0) { 1901 file_magerror(ms, "Continuations present with 0 count"); 1902 return -1; 1903 } 1904 m = &me->mp[me->cont_count - 1]; 1905 diff = CAST(int32_t, cont_level) - CAST(int32_t, m->cont_level); 1906 if (diff > 1) 1907 file_magwarn(ms, "New continuation level %u is more " 1908 "than one larger than current level %u", cont_level, 1909 m->cont_level); 1910 if (me->cont_count == me->max_count) { 1911 struct magic *nm; 1912 size_t cnt = me->max_count + ALLOC_CHUNK; 1913 if ((nm = CAST(struct magic *, realloc(me->mp, 1914 sizeof(*nm) * cnt))) == NULL) { 1915 file_oomem(ms, sizeof(*nm) * cnt); 1916 return -1; 1917 } 1918 me->mp = nm; 1919 me->max_count = CAST(uint32_t, cnt); 1920 } 1921 m = &me->mp[me->cont_count++]; 1922 (void)memset(m, 0, sizeof(*m)); 1923 m->cont_level = cont_level; 1924 } else { 1925 static const size_t len = sizeof(*m) * ALLOC_CHUNK; 1926 if (me->mp != NULL) 1927 return 1; 1928 if ((m = CAST(struct magic *, malloc(len))) == NULL) { 1929 file_oomem(ms, len); 1930 return -1; 1931 } 1932 me->mp = m; 1933 me->max_count = ALLOC_CHUNK; 1934 (void)memset(m, 0, sizeof(*m)); 1935 m->factor_op = FILE_FACTOR_OP_NONE; 1936 m->cont_level = 0; 1937 me->cont_count = 1; 1938 } 1939 m->lineno = CAST(uint32_t, lineno); 1940 1941 if (*l == '&') { /* m->cont_level == 0 checked below. */ 1942 ++l; /* step over */ 1943 m->flag |= OFFADD; 1944 } 1945 if (*l == '(') { 1946 ++l; /* step over */ 1947 m->flag |= INDIR; 1948 if (m->flag & OFFADD) 1949 m->flag = (m->flag & ~OFFADD) | INDIROFFADD; 1950 1951 if (*l == '&') { /* m->cont_level == 0 checked below */ 1952 ++l; /* step over */ 1953 m->flag |= OFFADD; 1954 } 1955 } 1956 /* Indirect offsets are not valid at level 0. */ 1957 if (m->cont_level == 0 && (m->flag & (OFFADD | INDIROFFADD))) { 1958 if (ms->flags & MAGIC_CHECK) 1959 file_magwarn(ms, "relative offset at level 0"); 1960 return -1; 1961 } 1962 1963 /* get offset, then skip over it */ 1964 if (*l == '-') { 1965 ++l; /* step over */ 1966 m->flag |= OFFNEGATIVE; 1967 } 1968 m->offset = CAST(int32_t, strtol(l, &t, 0)); 1969 if (l == t) { 1970 if (ms->flags & MAGIC_CHECK) 1971 file_magwarn(ms, "offset `%s' invalid", l); 1972 return -1; 1973 } 1974 1975 l = t; 1976 1977 if (m->flag & INDIR) { 1978 m->in_type = FILE_LONG; 1979 m->in_offset = 0; 1980 m->in_op = 0; 1981 /* 1982 * read [.,lbs][+-]nnnnn) 1983 */ 1984 if (*l == '.' || *l == ',') { 1985 if (*l == ',') 1986 m->in_op |= FILE_OPSIGNED; 1987 l++; 1988 switch (*l) { 1989 case 'l': 1990 m->in_type = FILE_LELONG; 1991 break; 1992 case 'L': 1993 m->in_type = FILE_BELONG; 1994 break; 1995 case 'm': 1996 m->in_type = FILE_MELONG; 1997 break; 1998 case 'h': 1999 case 's': 2000 m->in_type = FILE_LESHORT; 2001 break; 2002 case 'H': 2003 case 'S': 2004 m->in_type = FILE_BESHORT; 2005 break; 2006 case 'c': 2007 case 'b': 2008 case 'C': 2009 case 'B': 2010 m->in_type = FILE_BYTE; 2011 break; 2012 case 'e': 2013 case 'f': 2014 case 'g': 2015 m->in_type = FILE_LEDOUBLE; 2016 break; 2017 case 'E': 2018 case 'F': 2019 case 'G': 2020 m->in_type = FILE_BEDOUBLE; 2021 break; 2022 case 'i': 2023 m->in_type = FILE_LEID3; 2024 break; 2025 case 'I': 2026 m->in_type = FILE_BEID3; 2027 break; 2028 case 'q': 2029 m->in_type = FILE_LEQUAD; 2030 break; 2031 case 'Q': 2032 m->in_type = FILE_BEQUAD; 2033 break; 2034 default: 2035 if (ms->flags & MAGIC_CHECK) 2036 file_magwarn(ms, 2037 "indirect offset type `%c' invalid", 2038 *l); 2039 return -1; 2040 } 2041 l++; 2042 } 2043 2044 if (*l == '~') { 2045 m->in_op |= FILE_OPINVERSE; 2046 l++; 2047 } 2048 if ((op = get_op(*l)) != -1) { 2049 m->in_op |= op; 2050 l++; 2051 } 2052 if (*l == '(') { 2053 m->in_op |= FILE_OPINDIRECT; 2054 l++; 2055 } 2056 if (isdigit(CAST(unsigned char, *l)) || *l == '-') { 2057 m->in_offset = CAST(int32_t, strtol(l, &t, 0)); 2058 if (l == t) { 2059 if (ms->flags & MAGIC_CHECK) 2060 file_magwarn(ms, 2061 "in_offset `%s' invalid", l); 2062 return -1; 2063 } 2064 l = t; 2065 } 2066 if (*l++ != ')' || 2067 ((m->in_op & FILE_OPINDIRECT) && *l++ != ')')) { 2068 if (ms->flags & MAGIC_CHECK) 2069 file_magwarn(ms, 2070 "missing ')' in indirect offset"); 2071 return -1; 2072 } 2073 } 2074 EATAB; 2075 2076 #ifdef ENABLE_CONDITIONALS 2077 m->cond = get_cond(l, &l); 2078 if (check_cond(ms, m->cond, cont_level) == -1) 2079 return -1; 2080 2081 EATAB; 2082 #endif 2083 2084 /* 2085 * Parse the type. 2086 */ 2087 if (*l == 'u') { 2088 /* 2089 * Try it as a keyword type prefixed by "u"; match what 2090 * follows the "u". If that fails, try it as an SUS 2091 * integer type. 2092 */ 2093 m->type = get_type(type_tbl, l + 1, &l); 2094 if (m->type == FILE_INVALID) { 2095 /* 2096 * Not a keyword type; parse it as an SUS type, 2097 * 'u' possibly followed by a number or C/S/L. 2098 */ 2099 m->type = get_standard_integer_type(l, &l); 2100 } 2101 /* It's unsigned. */ 2102 if (m->type != FILE_INVALID) 2103 m->flag |= UNSIGNED; 2104 } else { 2105 /* 2106 * Try it as a keyword type. If that fails, try it as 2107 * an SUS integer type if it begins with "d" or as an 2108 * SUS string type if it begins with "s". In any case, 2109 * it's not unsigned. 2110 */ 2111 m->type = get_type(type_tbl, l, &l); 2112 if (m->type == FILE_INVALID) { 2113 /* 2114 * Not a keyword type; parse it as an SUS type, 2115 * either 'd' possibly followed by a number or 2116 * C/S/L, or just 's'. 2117 */ 2118 if (*l == 'd') 2119 m->type = get_standard_integer_type(l, &l); 2120 else if (*l == 's' 2121 && !isalpha(CAST(unsigned char, l[1]))) { 2122 m->type = FILE_STRING; 2123 ++l; 2124 } 2125 } 2126 } 2127 2128 if (m->type == FILE_INVALID) { 2129 /* Not found - try it as a special keyword. */ 2130 m->type = get_type(special_tbl, l, &l); 2131 } 2132 2133 if (m->type == FILE_INVALID) { 2134 if (ms->flags & MAGIC_CHECK) 2135 file_magwarn(ms, "type `%s' invalid", l); 2136 return -1; 2137 } 2138 2139 if (m->type == FILE_NAME && cont_level != 0) { 2140 if (ms->flags & MAGIC_CHECK) 2141 file_magwarn(ms, "`name%s' entries can only be " 2142 "declared at top level", l); 2143 return -1; 2144 } 2145 2146 /* New-style anding: "0 byte&0x80 =0x80 dynamically linked" */ 2147 /* New and improved: ~ & | ^ + - * / % -- exciting, isn't it? */ 2148 2149 m->mask_op = 0; 2150 if (*l == '~') { 2151 if (!IS_STRING(m->type)) 2152 m->mask_op |= FILE_OPINVERSE; 2153 else if (ms->flags & MAGIC_CHECK) 2154 file_magwarn(ms, "'~' invalid for string types"); 2155 ++l; 2156 } 2157 m->str_range = 0; 2158 m->str_flags = m->type == FILE_PSTRING ? PSTRING_1_LE : 0; 2159 if ((op = get_op(*l)) != -1) { 2160 if (IS_STRING(m->type)) { 2161 int r; 2162 2163 if (op != FILE_OPDIVIDE) { 2164 if (ms->flags & MAGIC_CHECK) 2165 file_magwarn(ms, 2166 "invalid string/indirect op: " 2167 "`%c'", *t); 2168 return -1; 2169 } 2170 2171 if (m->type == FILE_INDIRECT) 2172 r = parse_indirect_modifier(ms, m, &l); 2173 else 2174 r = parse_string_modifier(ms, m, &l); 2175 if (r == -1) 2176 return -1; 2177 } else 2178 parse_op_modifier(ms, m, &l, op); 2179 } 2180 2181 /* 2182 * We used to set mask to all 1's here, instead let's just not do 2183 * anything if mask = 0 (unless you have a better idea) 2184 */ 2185 EATAB; 2186 2187 switch (*l) { 2188 case '>': 2189 case '<': 2190 m->reln = *l; 2191 ++l; 2192 if (*l == '=') { 2193 if (ms->flags & MAGIC_CHECK) { 2194 file_magwarn(ms, "%c= not supported", 2195 m->reln); 2196 return -1; 2197 } 2198 ++l; 2199 } 2200 break; 2201 /* Old-style anding: "0 byte &0x80 dynamically linked" */ 2202 case '&': 2203 case '^': 2204 case '=': 2205 m->reln = *l; 2206 ++l; 2207 if (*l == '=') { 2208 /* HP compat: ignore &= etc. */ 2209 ++l; 2210 } 2211 break; 2212 case '!': 2213 m->reln = *l; 2214 ++l; 2215 break; 2216 default: 2217 m->reln = '='; /* the default relation */ 2218 if (*l == 'x' && ((isascii(CAST(unsigned char, l[1])) && 2219 isspace(CAST(unsigned char, l[1]))) || !l[1])) { 2220 m->reln = *l; 2221 ++l; 2222 } 2223 break; 2224 } 2225 /* 2226 * Grab the value part, except for an 'x' reln. 2227 */ 2228 if (m->reln != 'x' && getvalue(ms, m, &l, action)) 2229 return -1; 2230 2231 /* 2232 * TODO finish this macro and start using it! 2233 * #define offsetcheck {if (offset > ms->bytes_max -1) 2234 * magwarn("offset too big"); } 2235 */ 2236 2237 /* 2238 * Now get last part - the description 2239 */ 2240 EATAB; 2241 if (l[0] == '\b') { 2242 ++l; 2243 m->flag |= NOSPACE; 2244 } else if ((l[0] == '\\') && (l[1] == 'b')) { 2245 ++l; 2246 ++l; 2247 m->flag |= NOSPACE; 2248 } 2249 for (i = 0; (m->desc[i++] = *l++) != '\0' && i < sizeof(m->desc); ) 2250 continue; 2251 if (i == sizeof(m->desc)) { 2252 m->desc[sizeof(m->desc) - 1] = '\0'; 2253 if (ms->flags & MAGIC_CHECK) 2254 file_magwarn(ms, "description `%s' truncated", m->desc); 2255 } 2256 2257 /* 2258 * We only do this check while compiling, or if any of the magic 2259 * files were not compiled. 2260 */ 2261 if (ms->flags & MAGIC_CHECK) { 2262 if (check_format(ms, m) == -1) 2263 return -1; 2264 } 2265 #ifndef COMPILE_ONLY 2266 if (action == FILE_CHECK) { 2267 file_mdump(m); 2268 } 2269 #endif 2270 m->mimetype[0] = '\0'; /* initialise MIME type to none */ 2271 return 0; 2272 } 2273 2274 /* 2275 * parse a STRENGTH annotation line from magic file, put into magic[index - 1] 2276 * if valid 2277 */ 2278 private int 2279 parse_strength(struct magic_set *ms, struct magic_entry *me, const char *line, 2280 size_t len __attribute__((__unused__))) 2281 { 2282 const char *l = line; 2283 char *el; 2284 unsigned long factor; 2285 struct magic *m = &me->mp[0]; 2286 2287 if (m->factor_op != FILE_FACTOR_OP_NONE) { 2288 file_magwarn(ms, 2289 "Current entry already has a strength type: %c %d", 2290 m->factor_op, m->factor); 2291 return -1; 2292 } 2293 if (m->type == FILE_NAME) { 2294 file_magwarn(ms, "%s: Strength setting is not supported in " 2295 "\"name\" magic entries", m->value.s); 2296 return -1; 2297 } 2298 EATAB; 2299 switch (*l) { 2300 case FILE_FACTOR_OP_NONE: 2301 case FILE_FACTOR_OP_PLUS: 2302 case FILE_FACTOR_OP_MINUS: 2303 case FILE_FACTOR_OP_TIMES: 2304 case FILE_FACTOR_OP_DIV: 2305 m->factor_op = *l++; 2306 break; 2307 default: 2308 file_magwarn(ms, "Unknown factor op `%c'", *l); 2309 return -1; 2310 } 2311 EATAB; 2312 factor = strtoul(l, &el, 0); 2313 if (factor > 255) { 2314 file_magwarn(ms, "Too large factor `%lu'", factor); 2315 goto out; 2316 } 2317 if (*el && !isspace(CAST(unsigned char, *el))) { 2318 file_magwarn(ms, "Bad factor `%s'", l); 2319 goto out; 2320 } 2321 m->factor = CAST(uint8_t, factor); 2322 if (m->factor == 0 && m->factor_op == FILE_FACTOR_OP_DIV) { 2323 file_magwarn(ms, "Cannot have factor op `%c' and factor %u", 2324 m->factor_op, m->factor); 2325 goto out; 2326 } 2327 return 0; 2328 out: 2329 m->factor_op = FILE_FACTOR_OP_NONE; 2330 m->factor = 0; 2331 return -1; 2332 } 2333 2334 private int 2335 goodchar(unsigned char x, const char *extra) 2336 { 2337 return (isascii(x) && isalnum(x)) || strchr(extra, x); 2338 } 2339 2340 private int 2341 parse_extra(struct magic_set *ms, struct magic_entry *me, const char *line, 2342 size_t llen, off_t off, size_t len, const char *name, const char *extra, 2343 int nt) 2344 { 2345 size_t i; 2346 const char *l = line; 2347 struct magic *m = &me->mp[me->cont_count == 0 ? 0 : me->cont_count - 1]; 2348 char *buf = CAST(char *, CAST(void *, m)) + off; 2349 2350 if (buf[0] != '\0') { 2351 len = nt ? strlen(buf) : len; 2352 file_magwarn(ms, "Current entry already has a %s type " 2353 "`%.*s', new type `%s'", name, CAST(int, len), buf, l); 2354 return -1; 2355 } 2356 2357 if (*m->desc == '\0') { 2358 file_magwarn(ms, "Current entry does not yet have a " 2359 "description for adding a %s type", name); 2360 return -1; 2361 } 2362 2363 EATAB; 2364 for (i = 0; *l && i < llen && i < len && goodchar(*l, extra); 2365 buf[i++] = *l++) 2366 continue; 2367 2368 if (i == len && *l) { 2369 if (nt) 2370 buf[len - 1] = '\0'; 2371 if (ms->flags & MAGIC_CHECK) 2372 file_magwarn(ms, "%s type `%s' truncated %" 2373 SIZE_T_FORMAT "u", name, line, i); 2374 } else { 2375 if (!isspace(CAST(unsigned char, *l)) && !goodchar(*l, extra)) 2376 file_magwarn(ms, "%s type `%s' has bad char '%c'", 2377 name, line, *l); 2378 if (nt) 2379 buf[i] = '\0'; 2380 } 2381 2382 if (i > 0) 2383 return 0; 2384 2385 file_magerror(ms, "Bad magic entry '%s'", line); 2386 return -1; 2387 } 2388 2389 /* 2390 * Parse an Apple CREATOR/TYPE annotation from magic file and put it into 2391 * magic[index - 1] 2392 */ 2393 private int 2394 parse_apple(struct magic_set *ms, struct magic_entry *me, const char *line, 2395 size_t len) 2396 { 2397 struct magic *m = &me->mp[0]; 2398 2399 return parse_extra(ms, me, line, len, 2400 CAST(off_t, offsetof(struct magic, apple)), 2401 sizeof(m->apple), "APPLE", "!+-./?", 0); 2402 } 2403 2404 /* 2405 * Parse a comma-separated list of extensions 2406 */ 2407 private int 2408 parse_ext(struct magic_set *ms, struct magic_entry *me, const char *line, 2409 size_t len) 2410 { 2411 struct magic *m = &me->mp[0]; 2412 2413 return parse_extra(ms, me, line, len, 2414 CAST(off_t, offsetof(struct magic, ext)), 2415 sizeof(m->ext), "EXTENSION", ",!+-/@?_$", 0); 2416 } 2417 2418 /* 2419 * parse a MIME annotation line from magic file, put into magic[index - 1] 2420 * if valid 2421 */ 2422 private int 2423 parse_mime(struct magic_set *ms, struct magic_entry *me, const char *line, 2424 size_t len) 2425 { 2426 struct magic *m = &me->mp[0]; 2427 2428 return parse_extra(ms, me, line, len, 2429 CAST(off_t, offsetof(struct magic, mimetype)), 2430 sizeof(m->mimetype), "MIME", "+-/.$?:{}", 1); 2431 } 2432 2433 private int 2434 check_format_type(const char *ptr, int type, const char **estr) 2435 { 2436 int quad = 0, h; 2437 size_t len, cnt; 2438 if (*ptr == '\0') { 2439 /* Missing format string; bad */ 2440 *estr = "missing format spec"; 2441 return -1; 2442 } 2443 2444 switch (file_formats[type]) { 2445 case FILE_FMT_QUAD: 2446 quad = 1; 2447 /*FALLTHROUGH*/ 2448 case FILE_FMT_NUM: 2449 if (quad == 0) { 2450 switch (type) { 2451 case FILE_BYTE: 2452 h = 2; 2453 break; 2454 case FILE_SHORT: 2455 case FILE_BESHORT: 2456 case FILE_LESHORT: 2457 h = 1; 2458 break; 2459 case FILE_LONG: 2460 case FILE_BELONG: 2461 case FILE_LELONG: 2462 case FILE_MELONG: 2463 case FILE_LEID3: 2464 case FILE_BEID3: 2465 case FILE_INDIRECT: 2466 h = 0; 2467 break; 2468 default: 2469 abort(); 2470 } 2471 } else 2472 h = 0; 2473 if (*ptr == '-') 2474 ptr++; 2475 if (*ptr == '.') 2476 ptr++; 2477 if (*ptr == '#') 2478 ptr++; 2479 #define CHECKLEN() do { \ 2480 for (len = cnt = 0; isdigit(CAST(unsigned char, *ptr)); ptr++, cnt++) \ 2481 len = len * 10 + (*ptr - '0'); \ 2482 if (cnt > 5 || len > 1024) \ 2483 goto toolong; \ 2484 } while (/*CONSTCOND*/0) 2485 2486 CHECKLEN(); 2487 if (*ptr == '.') 2488 ptr++; 2489 CHECKLEN(); 2490 if (quad) { 2491 if (*ptr++ != 'l') 2492 goto invalid; 2493 if (*ptr++ != 'l') 2494 goto invalid; 2495 } 2496 2497 switch (*ptr++) { 2498 #ifdef STRICT_FORMAT /* "long" formats are int formats for us */ 2499 /* so don't accept the 'l' modifier */ 2500 case 'l': 2501 switch (*ptr++) { 2502 case 'i': 2503 case 'd': 2504 case 'u': 2505 case 'o': 2506 case 'x': 2507 case 'X': 2508 if (h == 0) 2509 return 0; 2510 /*FALLTHROUGH*/ 2511 default: 2512 goto invalid; 2513 } 2514 2515 /* 2516 * Don't accept h and hh modifiers. They make writing 2517 * magic entries more complicated, for very little benefit 2518 */ 2519 case 'h': 2520 if (h-- <= 0) 2521 goto invalid; 2522 switch (*ptr++) { 2523 case 'h': 2524 if (h-- <= 0) 2525 goto invalid; 2526 switch (*ptr++) { 2527 case 'i': 2528 case 'd': 2529 case 'u': 2530 case 'o': 2531 case 'x': 2532 case 'X': 2533 return 0; 2534 default: 2535 goto invalid; 2536 } 2537 case 'i': 2538 case 'd': 2539 case 'u': 2540 case 'o': 2541 case 'x': 2542 case 'X': 2543 if (h == 0) 2544 return 0; 2545 /*FALLTHROUGH*/ 2546 default: 2547 goto invalid; 2548 } 2549 #endif 2550 case 'c': 2551 if (h == 2) 2552 return 0; 2553 goto invalid; 2554 case 'i': 2555 case 'd': 2556 case 'u': 2557 case 'o': 2558 case 'x': 2559 case 'X': 2560 #ifdef STRICT_FORMAT 2561 if (h == 0) 2562 return 0; 2563 /*FALLTHROUGH*/ 2564 #else 2565 return 0; 2566 #endif 2567 default: 2568 goto invalid; 2569 } 2570 2571 case FILE_FMT_FLOAT: 2572 case FILE_FMT_DOUBLE: 2573 if (*ptr == '-') 2574 ptr++; 2575 if (*ptr == '.') 2576 ptr++; 2577 CHECKLEN(); 2578 if (*ptr == '.') 2579 ptr++; 2580 CHECKLEN(); 2581 switch (*ptr++) { 2582 case 'e': 2583 case 'E': 2584 case 'f': 2585 case 'F': 2586 case 'g': 2587 case 'G': 2588 return 0; 2589 2590 default: 2591 goto invalid; 2592 } 2593 2594 2595 case FILE_FMT_STR: 2596 if (*ptr == '-') 2597 ptr++; 2598 while (isdigit(CAST(unsigned char, *ptr))) 2599 ptr++; 2600 if (*ptr == '.') { 2601 ptr++; 2602 while (isdigit(CAST(unsigned char , *ptr))) 2603 ptr++; 2604 } 2605 2606 switch (*ptr++) { 2607 case 's': 2608 return 0; 2609 default: 2610 goto invalid; 2611 } 2612 2613 default: 2614 /* internal error */ 2615 abort(); 2616 } 2617 invalid: 2618 *estr = "not valid"; 2619 toolong: 2620 *estr = "too long"; 2621 return -1; 2622 } 2623 2624 /* 2625 * Check that the optional printf format in description matches 2626 * the type of the magic. 2627 */ 2628 private int 2629 check_format(struct magic_set *ms, struct magic *m) 2630 { 2631 char *ptr; 2632 const char *estr; 2633 2634 for (ptr = m->desc; *ptr; ptr++) 2635 if (*ptr == '%') 2636 break; 2637 if (*ptr == '\0') { 2638 /* No format string; ok */ 2639 return 1; 2640 } 2641 2642 assert(file_nformats == file_nnames); 2643 2644 if (m->type >= file_nformats) { 2645 file_magwarn(ms, "Internal error inconsistency between " 2646 "m->type and format strings"); 2647 return -1; 2648 } 2649 if (file_formats[m->type] == FILE_FMT_NONE) { 2650 file_magwarn(ms, "No format string for `%s' with description " 2651 "`%s'", m->desc, file_names[m->type]); 2652 return -1; 2653 } 2654 2655 ptr++; 2656 if (check_format_type(ptr, m->type, &estr) == -1) { 2657 /* 2658 * TODO: this error message is unhelpful if the format 2659 * string is not one character long 2660 */ 2661 file_magwarn(ms, "Printf format is %s for type " 2662 "`%s' in description `%s'", estr, 2663 file_names[m->type], m->desc); 2664 return -1; 2665 } 2666 2667 for (; *ptr; ptr++) { 2668 if (*ptr == '%') { 2669 file_magwarn(ms, 2670 "Too many format strings (should have at most one) " 2671 "for `%s' with description `%s'", 2672 file_names[m->type], m->desc); 2673 return -1; 2674 } 2675 } 2676 return 0; 2677 } 2678 2679 /* 2680 * Read a numeric value from a pointer, into the value union of a magic 2681 * pointer, according to the magic type. Update the string pointer to point 2682 * just after the number read. Return 0 for success, non-zero for failure. 2683 */ 2684 private int 2685 getvalue(struct magic_set *ms, struct magic *m, const char **p, int action) 2686 { 2687 char *ep; 2688 uint64_t ull; 2689 2690 switch (m->type) { 2691 case FILE_BESTRING16: 2692 case FILE_LESTRING16: 2693 case FILE_STRING: 2694 case FILE_PSTRING: 2695 case FILE_REGEX: 2696 case FILE_SEARCH: 2697 case FILE_NAME: 2698 case FILE_USE: 2699 case FILE_DER: 2700 *p = getstr(ms, m, *p, action == FILE_COMPILE); 2701 if (*p == NULL) { 2702 if (ms->flags & MAGIC_CHECK) 2703 file_magwarn(ms, "cannot get string from `%s'", 2704 m->value.s); 2705 return -1; 2706 } 2707 if (m->type == FILE_REGEX) { 2708 file_regex_t rx; 2709 int rc = file_regcomp(&rx, m->value.s, REG_EXTENDED); 2710 if (rc) { 2711 if (ms->flags & MAGIC_CHECK) 2712 file_regerror(&rx, rc, ms); 2713 } 2714 file_regfree(&rx); 2715 return rc ? -1 : 0; 2716 } 2717 return 0; 2718 default: 2719 if (m->reln == 'x') 2720 return 0; 2721 break; 2722 } 2723 2724 switch (m->type) { 2725 case FILE_FLOAT: 2726 case FILE_BEFLOAT: 2727 case FILE_LEFLOAT: 2728 errno = 0; 2729 #ifdef HAVE_STRTOF 2730 m->value.f = strtof(*p, &ep); 2731 #else 2732 m->value.f = (float)strtod(*p, &ep); 2733 #endif 2734 if (errno == 0) 2735 *p = ep; 2736 return 0; 2737 case FILE_DOUBLE: 2738 case FILE_BEDOUBLE: 2739 case FILE_LEDOUBLE: 2740 errno = 0; 2741 m->value.d = strtod(*p, &ep); 2742 if (errno == 0) 2743 *p = ep; 2744 return 0; 2745 case FILE_GUID: 2746 if (file_parse_guid(*p, m->value.guid) == -1) 2747 return -1; 2748 *p += FILE_GUID_SIZE - 1; 2749 return 0; 2750 default: 2751 errno = 0; 2752 ull = CAST(uint64_t, strtoull(*p, &ep, 0)); 2753 m->value.q = file_signextend(ms, m, ull); 2754 if (*p == ep) { 2755 file_magwarn(ms, "Unparsable number `%s'", *p); 2756 } else { 2757 size_t ts = typesize(m->type); 2758 uint64_t x; 2759 const char *q; 2760 2761 if (ts == FILE_BADSIZE) { 2762 file_magwarn(ms, 2763 "Expected numeric type got `%s'", 2764 type_tbl[m->type].name); 2765 } 2766 for (q = *p; isspace(CAST(unsigned char, *q)); q++) 2767 continue; 2768 if (*q == '-') 2769 ull = -CAST(int64_t, ull); 2770 switch (ts) { 2771 case 1: 2772 x = CAST(uint64_t, ull & ~0xffULL); 2773 break; 2774 case 2: 2775 x = CAST(uint64_t, ull & ~0xffffULL); 2776 break; 2777 case 4: 2778 x = CAST(uint64_t, ull & ~0xffffffffULL); 2779 break; 2780 case 8: 2781 x = 0; 2782 break; 2783 default: 2784 abort(); 2785 } 2786 if (x) { 2787 file_magwarn(ms, "Overflow for numeric" 2788 " type `%s' value %#" PRIx64, 2789 type_tbl[m->type].name, ull); 2790 } 2791 } 2792 if (errno == 0) { 2793 *p = ep; 2794 eatsize(p); 2795 } 2796 return 0; 2797 } 2798 } 2799 2800 /* 2801 * Convert a string containing C character escapes. Stop at an unescaped 2802 * space or tab. 2803 * Copy the converted version to "m->value.s", and the length in m->vallen. 2804 * Return updated scan pointer as function result. Warn if set. 2805 */ 2806 private const char * 2807 getstr(struct magic_set *ms, struct magic *m, const char *s, int warn) 2808 { 2809 const char *origs = s; 2810 char *p = m->value.s; 2811 size_t plen = sizeof(m->value.s); 2812 char *origp = p; 2813 char *pmax = p + plen - 1; 2814 int c; 2815 int val; 2816 2817 while ((c = *s++) != '\0') { 2818 if (isspace(CAST(unsigned char, c))) 2819 break; 2820 if (p >= pmax) { 2821 file_error(ms, 0, "string too long: `%s'", origs); 2822 return NULL; 2823 } 2824 if (c == '\\') { 2825 switch(c = *s++) { 2826 2827 case '\0': 2828 if (warn) 2829 file_magwarn(ms, "incomplete escape"); 2830 s--; 2831 goto out; 2832 2833 case '\t': 2834 if (warn) { 2835 file_magwarn(ms, 2836 "escaped tab found, use \\t instead"); 2837 warn = 0; /* already did */ 2838 } 2839 /*FALLTHROUGH*/ 2840 default: 2841 if (warn) { 2842 if (isprint(CAST(unsigned char, c))) { 2843 /* Allow escaping of 2844 * ``relations'' */ 2845 if (strchr("<>&^=!", c) == NULL 2846 && (m->type != FILE_REGEX || 2847 strchr("[]().*?^$|{}", c) 2848 == NULL)) { 2849 file_magwarn(ms, "no " 2850 "need to escape " 2851 "`%c'", c); 2852 } 2853 } else { 2854 file_magwarn(ms, 2855 "unknown escape sequence: " 2856 "\\%03o", c); 2857 } 2858 } 2859 /*FALLTHROUGH*/ 2860 /* space, perhaps force people to use \040? */ 2861 case ' ': 2862 #if 0 2863 /* 2864 * Other things people escape, but shouldn't need to, 2865 * so we disallow them 2866 */ 2867 case '\'': 2868 case '"': 2869 case '?': 2870 #endif 2871 /* Relations */ 2872 case '>': 2873 case '<': 2874 case '&': 2875 case '^': 2876 case '=': 2877 case '!': 2878 /* and baskslash itself */ 2879 case '\\': 2880 *p++ = CAST(char, c); 2881 break; 2882 2883 case 'a': 2884 *p++ = '\a'; 2885 break; 2886 2887 case 'b': 2888 *p++ = '\b'; 2889 break; 2890 2891 case 'f': 2892 *p++ = '\f'; 2893 break; 2894 2895 case 'n': 2896 *p++ = '\n'; 2897 break; 2898 2899 case 'r': 2900 *p++ = '\r'; 2901 break; 2902 2903 case 't': 2904 *p++ = '\t'; 2905 break; 2906 2907 case 'v': 2908 *p++ = '\v'; 2909 break; 2910 2911 /* \ and up to 3 octal digits */ 2912 case '0': 2913 case '1': 2914 case '2': 2915 case '3': 2916 case '4': 2917 case '5': 2918 case '6': 2919 case '7': 2920 val = c - '0'; 2921 c = *s++; /* try for 2 */ 2922 if (c >= '0' && c <= '7') { 2923 val = (val << 3) | (c - '0'); 2924 c = *s++; /* try for 3 */ 2925 if (c >= '0' && c <= '7') 2926 val = (val << 3) | (c-'0'); 2927 else 2928 --s; 2929 } 2930 else 2931 --s; 2932 *p++ = CAST(char, val); 2933 break; 2934 2935 /* \x and up to 2 hex digits */ 2936 case 'x': 2937 val = 'x'; /* Default if no digits */ 2938 c = hextoint(*s++); /* Get next char */ 2939 if (c >= 0) { 2940 val = c; 2941 c = hextoint(*s++); 2942 if (c >= 0) 2943 val = (val << 4) + c; 2944 else 2945 --s; 2946 } else 2947 --s; 2948 *p++ = CAST(char, val); 2949 break; 2950 } 2951 } else 2952 *p++ = CAST(char, c); 2953 } 2954 --s; 2955 out: 2956 *p = '\0'; 2957 m->vallen = CAST(unsigned char, (p - origp)); 2958 if (m->type == FILE_PSTRING) { 2959 size_t l = file_pstring_length_size(ms, m); 2960 if (l == FILE_BADSIZE) 2961 return NULL; 2962 m->vallen += CAST(unsigned char, l); 2963 } 2964 return s; 2965 } 2966 2967 2968 /* Single hex char to int; -1 if not a hex char. */ 2969 private int 2970 hextoint(int c) 2971 { 2972 if (!isascii(CAST(unsigned char, c))) 2973 return -1; 2974 if (isdigit(CAST(unsigned char, c))) 2975 return c - '0'; 2976 if ((c >= 'a') && (c <= 'f')) 2977 return c + 10 - 'a'; 2978 if (( c>= 'A') && (c <= 'F')) 2979 return c + 10 - 'A'; 2980 return -1; 2981 } 2982 2983 2984 /* 2985 * Print a string containing C character escapes. 2986 */ 2987 protected void 2988 file_showstr(FILE *fp, const char *s, size_t len) 2989 { 2990 char c; 2991 2992 for (;;) { 2993 if (len == FILE_BADSIZE) { 2994 c = *s++; 2995 if (c == '\0') 2996 break; 2997 } 2998 else { 2999 if (len-- == 0) 3000 break; 3001 c = *s++; 3002 } 3003 if (c >= 040 && c <= 0176) /* TODO isprint && !iscntrl */ 3004 (void) fputc(c, fp); 3005 else { 3006 (void) fputc('\\', fp); 3007 switch (c) { 3008 case '\a': 3009 (void) fputc('a', fp); 3010 break; 3011 3012 case '\b': 3013 (void) fputc('b', fp); 3014 break; 3015 3016 case '\f': 3017 (void) fputc('f', fp); 3018 break; 3019 3020 case '\n': 3021 (void) fputc('n', fp); 3022 break; 3023 3024 case '\r': 3025 (void) fputc('r', fp); 3026 break; 3027 3028 case '\t': 3029 (void) fputc('t', fp); 3030 break; 3031 3032 case '\v': 3033 (void) fputc('v', fp); 3034 break; 3035 3036 default: 3037 (void) fprintf(fp, "%.3o", c & 0377); 3038 break; 3039 } 3040 } 3041 } 3042 } 3043 3044 /* 3045 * eatsize(): Eat the size spec from a number [eg. 10UL] 3046 */ 3047 private void 3048 eatsize(const char **p) 3049 { 3050 const char *l = *p; 3051 3052 if (LOWCASE(*l) == 'u') 3053 l++; 3054 3055 switch (LOWCASE(*l)) { 3056 case 'l': /* long */ 3057 case 's': /* short */ 3058 case 'h': /* short */ 3059 case 'b': /* char/byte */ 3060 case 'c': /* char/byte */ 3061 l++; 3062 /*FALLTHROUGH*/ 3063 default: 3064 break; 3065 } 3066 3067 *p = l; 3068 } 3069 3070 /* 3071 * handle a buffer containing a compiled file. 3072 */ 3073 private struct magic_map * 3074 apprentice_buf(struct magic_set *ms, struct magic *buf, size_t len) 3075 { 3076 struct magic_map *map; 3077 3078 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 3079 file_oomem(ms, sizeof(*map)); 3080 return NULL; 3081 } 3082 map->len = len; 3083 map->p = buf; 3084 map->type = MAP_TYPE_USER; 3085 if (check_buffer(ms, map, "buffer") != 0) { 3086 apprentice_unmap(map); 3087 return NULL; 3088 } 3089 return map; 3090 } 3091 3092 /* 3093 * handle a compiled file. 3094 */ 3095 3096 private struct magic_map * 3097 apprentice_map(struct magic_set *ms, const char *fn) 3098 { 3099 int fd; 3100 struct stat st; 3101 char *dbname = NULL; 3102 struct magic_map *map; 3103 struct magic_map *rv = NULL; 3104 3105 fd = -1; 3106 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 3107 file_oomem(ms, sizeof(*map)); 3108 goto error; 3109 } 3110 map->type = MAP_TYPE_USER; /* unspecified */ 3111 3112 dbname = mkdbname(ms, fn, 0); 3113 if (dbname == NULL) 3114 goto error; 3115 3116 if ((fd = open(dbname, O_RDONLY|O_BINARY)) == -1) 3117 goto error; 3118 3119 if (fstat(fd, &st) == -1) { 3120 file_error(ms, errno, "cannot stat `%s'", dbname); 3121 goto error; 3122 } 3123 if (st.st_size < 8 || st.st_size > maxoff_t()) { 3124 file_error(ms, 0, "file `%s' is too %s", dbname, 3125 st.st_size < 8 ? "small" : "large"); 3126 goto error; 3127 } 3128 3129 map->len = CAST(size_t, st.st_size); 3130 #ifdef QUICK 3131 map->type = MAP_TYPE_MMAP; 3132 if ((map->p = mmap(0, CAST(size_t, st.st_size), PROT_READ|PROT_WRITE, 3133 MAP_PRIVATE|MAP_FILE, fd, CAST(off_t, 0))) == MAP_FAILED) { 3134 file_error(ms, errno, "cannot map `%s'", dbname); 3135 goto error; 3136 } 3137 #else 3138 map->type = MAP_TYPE_MALLOC; 3139 if ((map->p = CAST(void *, malloc(map->len))) == NULL) { 3140 file_oomem(ms, map->len); 3141 goto error; 3142 } 3143 if (read(fd, map->p, map->len) != (ssize_t)map->len) { 3144 file_badread(ms); 3145 goto error; 3146 } 3147 #endif 3148 (void)close(fd); 3149 fd = -1; 3150 3151 if (check_buffer(ms, map, dbname) != 0) { 3152 goto error; 3153 } 3154 #ifdef QUICK 3155 if (mprotect(map->p, CAST(size_t, st.st_size), PROT_READ) == -1) { 3156 file_error(ms, errno, "cannot mprotect `%s'", dbname); 3157 goto error; 3158 } 3159 #endif 3160 3161 free(dbname); 3162 return map; 3163 3164 error: 3165 if (fd != -1) 3166 (void)close(fd); 3167 apprentice_unmap(map); 3168 free(dbname); 3169 return rv; 3170 } 3171 3172 private int 3173 check_buffer(struct magic_set *ms, struct magic_map *map, const char *dbname) 3174 { 3175 uint32_t *ptr; 3176 uint32_t entries, nentries; 3177 uint32_t version; 3178 int i, needsbyteswap; 3179 3180 ptr = CAST(uint32_t *, map->p); 3181 if (*ptr != MAGICNO) { 3182 if (swap4(*ptr) != MAGICNO) { 3183 file_error(ms, 0, "bad magic in `%s'", dbname); 3184 return -1; 3185 } 3186 needsbyteswap = 1; 3187 } else 3188 needsbyteswap = 0; 3189 if (needsbyteswap) 3190 version = swap4(ptr[1]); 3191 else 3192 version = ptr[1]; 3193 if (version != VERSIONNO) { 3194 file_error(ms, 0, "File %s supports only version %d magic " 3195 "files. `%s' is version %d", VERSION, 3196 VERSIONNO, dbname, version); 3197 return -1; 3198 } 3199 entries = CAST(uint32_t, map->len / sizeof(struct magic)); 3200 if ((entries * sizeof(struct magic)) != map->len) { 3201 file_error(ms, 0, "Size of `%s' %" SIZE_T_FORMAT "u is not " 3202 "a multiple of %" SIZE_T_FORMAT "u", 3203 dbname, map->len, sizeof(struct magic)); 3204 return -1; 3205 } 3206 map->magic[0] = CAST(struct magic *, map->p) + 1; 3207 nentries = 0; 3208 for (i = 0; i < MAGIC_SETS; i++) { 3209 if (needsbyteswap) 3210 map->nmagic[i] = swap4(ptr[i + 2]); 3211 else 3212 map->nmagic[i] = ptr[i + 2]; 3213 if (i != MAGIC_SETS - 1) 3214 map->magic[i + 1] = map->magic[i] + map->nmagic[i]; 3215 nentries += map->nmagic[i]; 3216 } 3217 if (entries != nentries + 1) { 3218 file_error(ms, 0, "Inconsistent entries in `%s' %u != %u", 3219 dbname, entries, nentries + 1); 3220 return -1; 3221 } 3222 if (needsbyteswap) 3223 for (i = 0; i < MAGIC_SETS; i++) 3224 byteswap(map->magic[i], map->nmagic[i]); 3225 return 0; 3226 } 3227 3228 /* 3229 * handle an mmaped file. 3230 */ 3231 private int 3232 apprentice_compile(struct magic_set *ms, struct magic_map *map, const char *fn) 3233 { 3234 static const size_t nm = sizeof(*map->nmagic) * MAGIC_SETS; 3235 static const size_t m = sizeof(**map->magic); 3236 int fd = -1; 3237 size_t len; 3238 char *dbname; 3239 int rv = -1; 3240 uint32_t i; 3241 union { 3242 struct magic m; 3243 uint32_t h[2 + MAGIC_SETS]; 3244 } hdr; 3245 3246 dbname = mkdbname(ms, fn, 1); 3247 3248 if (dbname == NULL) 3249 goto out; 3250 3251 if ((fd = open(dbname, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY, 0644)) == -1) 3252 { 3253 file_error(ms, errno, "cannot open `%s'", dbname); 3254 goto out; 3255 } 3256 memset(&hdr, 0, sizeof(hdr)); 3257 hdr.h[0] = MAGICNO; 3258 hdr.h[1] = VERSIONNO; 3259 memcpy(hdr.h + 2, map->nmagic, nm); 3260 3261 if (write(fd, &hdr, sizeof(hdr)) != CAST(ssize_t, sizeof(hdr))) { 3262 file_error(ms, errno, "error writing `%s'", dbname); 3263 goto out2; 3264 } 3265 3266 for (i = 0; i < MAGIC_SETS; i++) { 3267 len = m * map->nmagic[i]; 3268 if (write(fd, map->magic[i], len) != CAST(ssize_t, len)) { 3269 file_error(ms, errno, "error writing `%s'", dbname); 3270 goto out2; 3271 } 3272 } 3273 3274 rv = 0; 3275 out2: 3276 if (fd != -1) 3277 (void)close(fd); 3278 out: 3279 apprentice_unmap(map); 3280 free(dbname); 3281 return rv; 3282 } 3283 3284 private const char ext[] = ".mgc"; 3285 /* 3286 * make a dbname 3287 */ 3288 private char * 3289 mkdbname(struct magic_set *ms, const char *fn, int strip) 3290 { 3291 const char *p, *q; 3292 char *buf; 3293 3294 if (strip) { 3295 if ((p = strrchr(fn, '/')) != NULL) 3296 fn = ++p; 3297 } 3298 3299 for (q = fn; *q; q++) 3300 continue; 3301 /* Look for .mgc */ 3302 for (p = ext + sizeof(ext) - 1; p >= ext && q >= fn; p--, q--) 3303 if (*p != *q) 3304 break; 3305 3306 /* Did not find .mgc, restore q */ 3307 if (p >= ext) 3308 while (*q) 3309 q++; 3310 3311 q++; 3312 /* Compatibility with old code that looked in .mime */ 3313 if (ms->flags & MAGIC_MIME) { 3314 if (asprintf(&buf, "%.*s.mime%s", CAST(int, q - fn), fn, ext) 3315 < 0) 3316 return NULL; 3317 if (access(buf, R_OK) != -1) { 3318 ms->flags &= MAGIC_MIME_TYPE; 3319 return buf; 3320 } 3321 free(buf); 3322 } 3323 if (asprintf(&buf, "%.*s%s", CAST(int, q - fn), fn, ext) < 0) 3324 return NULL; 3325 3326 /* Compatibility with old code that looked in .mime */ 3327 if (strstr(fn, ".mime") != NULL) 3328 ms->flags &= MAGIC_MIME_TYPE; 3329 return buf; 3330 } 3331 3332 /* 3333 * Byteswap an mmap'ed file if needed 3334 */ 3335 private void 3336 byteswap(struct magic *magic, uint32_t nmagic) 3337 { 3338 uint32_t i; 3339 for (i = 0; i < nmagic; i++) 3340 bs1(&magic[i]); 3341 } 3342 3343 /* 3344 * swap a short 3345 */ 3346 private uint16_t 3347 swap2(uint16_t sv) 3348 { 3349 uint16_t rv; 3350 uint8_t *s = RCAST(uint8_t *, RCAST(void *, &sv)); 3351 uint8_t *d = RCAST(uint8_t *, RCAST(void *, &rv)); 3352 d[0] = s[1]; 3353 d[1] = s[0]; 3354 return rv; 3355 } 3356 3357 /* 3358 * swap an int 3359 */ 3360 private uint32_t 3361 swap4(uint32_t sv) 3362 { 3363 uint32_t rv; 3364 uint8_t *s = RCAST(uint8_t *, RCAST(void *, &sv)); 3365 uint8_t *d = RCAST(uint8_t *, RCAST(void *, &rv)); 3366 d[0] = s[3]; 3367 d[1] = s[2]; 3368 d[2] = s[1]; 3369 d[3] = s[0]; 3370 return rv; 3371 } 3372 3373 /* 3374 * swap a quad 3375 */ 3376 private uint64_t 3377 swap8(uint64_t sv) 3378 { 3379 uint64_t rv; 3380 uint8_t *s = RCAST(uint8_t *, RCAST(void *, &sv)); 3381 uint8_t *d = RCAST(uint8_t *, RCAST(void *, &rv)); 3382 #if 0 3383 d[0] = s[3]; 3384 d[1] = s[2]; 3385 d[2] = s[1]; 3386 d[3] = s[0]; 3387 d[4] = s[7]; 3388 d[5] = s[6]; 3389 d[6] = s[5]; 3390 d[7] = s[4]; 3391 #else 3392 d[0] = s[7]; 3393 d[1] = s[6]; 3394 d[2] = s[5]; 3395 d[3] = s[4]; 3396 d[4] = s[3]; 3397 d[5] = s[2]; 3398 d[6] = s[1]; 3399 d[7] = s[0]; 3400 #endif 3401 return rv; 3402 } 3403 3404 /* 3405 * byteswap a single magic entry 3406 */ 3407 private void 3408 bs1(struct magic *m) 3409 { 3410 m->cont_level = swap2(m->cont_level); 3411 m->offset = swap4(CAST(uint32_t, m->offset)); 3412 m->in_offset = swap4(CAST(uint32_t, m->in_offset)); 3413 m->lineno = swap4(CAST(uint32_t, m->lineno)); 3414 if (IS_STRING(m->type)) { 3415 m->str_range = swap4(m->str_range); 3416 m->str_flags = swap4(m->str_flags); 3417 } 3418 else { 3419 m->value.q = swap8(m->value.q); 3420 m->num_mask = swap8(m->num_mask); 3421 } 3422 } 3423 3424 protected size_t 3425 file_pstring_length_size(struct magic_set *ms, const struct magic *m) 3426 { 3427 switch (m->str_flags & PSTRING_LEN) { 3428 case PSTRING_1_LE: 3429 return 1; 3430 case PSTRING_2_LE: 3431 case PSTRING_2_BE: 3432 return 2; 3433 case PSTRING_4_LE: 3434 case PSTRING_4_BE: 3435 return 4; 3436 default: 3437 file_error(ms, 0, "corrupt magic file " 3438 "(bad pascal string length %d)", 3439 m->str_flags & PSTRING_LEN); 3440 return FILE_BADSIZE; 3441 } 3442 } 3443 protected size_t 3444 file_pstring_get_length(struct magic_set *ms, const struct magic *m, 3445 const char *ss) 3446 { 3447 size_t len = 0; 3448 const unsigned char *s = RCAST(const unsigned char *, ss); 3449 unsigned int s3, s2, s1, s0; 3450 3451 switch (m->str_flags & PSTRING_LEN) { 3452 case PSTRING_1_LE: 3453 len = *s; 3454 break; 3455 case PSTRING_2_LE: 3456 s0 = s[0]; 3457 s1 = s[1]; 3458 len = (s1 << 8) | s0; 3459 break; 3460 case PSTRING_2_BE: 3461 s0 = s[0]; 3462 s1 = s[1]; 3463 len = (s0 << 8) | s1; 3464 break; 3465 case PSTRING_4_LE: 3466 s0 = s[0]; 3467 s1 = s[1]; 3468 s2 = s[2]; 3469 s3 = s[3]; 3470 len = (s3 << 24) | (s2 << 16) | (s1 << 8) | s0; 3471 break; 3472 case PSTRING_4_BE: 3473 s0 = s[0]; 3474 s1 = s[1]; 3475 s2 = s[2]; 3476 s3 = s[3]; 3477 len = (s0 << 24) | (s1 << 16) | (s2 << 8) | s3; 3478 break; 3479 default: 3480 file_error(ms, 0, "corrupt magic file " 3481 "(bad pascal string length %d)", 3482 m->str_flags & PSTRING_LEN); 3483 return FILE_BADSIZE; 3484 } 3485 3486 if (m->str_flags & PSTRING_LENGTH_INCLUDES_ITSELF) { 3487 size_t l = file_pstring_length_size(ms, m); 3488 if (l == FILE_BADSIZE) 3489 return l; 3490 len -= l; 3491 } 3492 3493 return len; 3494 } 3495 3496 protected int 3497 file_magicfind(struct magic_set *ms, const char *name, struct mlist *v) 3498 { 3499 uint32_t i, j; 3500 struct mlist *mlist, *ml; 3501 3502 mlist = ms->mlist[1]; 3503 3504 for (ml = mlist->next; ml != mlist; ml = ml->next) { 3505 struct magic *ma = ml->magic; 3506 uint32_t nma = ml->nmagic; 3507 for (i = 0; i < nma; i++) { 3508 if (ma[i].type != FILE_NAME) 3509 continue; 3510 if (strcmp(ma[i].value.s, name) == 0) { 3511 v->magic = &ma[i]; 3512 for (j = i + 1; j < nma; j++) 3513 if (ma[j].cont_level == 0) 3514 break; 3515 v->nmagic = j - i; 3516 return 0; 3517 } 3518 } 3519 } 3520 return -1; 3521 } 3522