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 1.324 2022/05/31 18:54:25 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 #ifdef HAVE_BYTESWAP_H 54 #include <byteswap.h> 55 #endif 56 #ifdef HAVE_SYS_BSWAP_H 57 #include <sys/bswap.h> 58 #endif 59 60 61 #define EATAB {while (isascii(CAST(unsigned char, *l)) && \ 62 isspace(CAST(unsigned char, *l))) ++l;} 63 #define LOWCASE(l) (isupper(CAST(unsigned char, l)) ? \ 64 tolower(CAST(unsigned char, l)) : (l)) 65 /* 66 * Work around a bug in headers on Digital Unix. 67 * At least confirmed for: OSF1 V4.0 878 68 */ 69 #if defined(__osf__) && defined(__DECC) 70 #ifdef MAP_FAILED 71 #undef MAP_FAILED 72 #endif 73 #endif 74 75 #ifndef MAP_FAILED 76 #define MAP_FAILED (void *) -1 77 #endif 78 79 #ifndef MAP_FILE 80 #define MAP_FILE 0 81 #endif 82 83 #define ALLOC_CHUNK CAST(size_t, 10) 84 #define ALLOC_INCR CAST(size_t, 200) 85 86 #define MAP_TYPE_USER 0 87 #define MAP_TYPE_MALLOC 1 88 #define MAP_TYPE_MMAP 2 89 90 struct magic_entry { 91 struct magic *mp; 92 uint32_t cont_count; 93 uint32_t max_count; 94 }; 95 96 struct magic_entry_set { 97 struct magic_entry *me; 98 uint32_t count; 99 uint32_t max; 100 }; 101 102 struct magic_map { 103 void *p; 104 size_t len; 105 int type; 106 struct magic *magic[MAGIC_SETS]; 107 uint32_t nmagic[MAGIC_SETS]; 108 }; 109 110 int file_formats[FILE_NAMES_SIZE]; 111 const size_t file_nformats = FILE_NAMES_SIZE; 112 const char *file_names[FILE_NAMES_SIZE]; 113 const size_t file_nnames = FILE_NAMES_SIZE; 114 115 private int getvalue(struct magic_set *ms, struct magic *, const char **, int); 116 private int hextoint(int); 117 private const char *getstr(struct magic_set *, struct magic *, const char *, 118 int); 119 private int parse(struct magic_set *, struct magic_entry *, const char *, 120 size_t, int); 121 private void eatsize(const char **); 122 private int apprentice_1(struct magic_set *, const char *, int); 123 private ssize_t apprentice_magic_strength_1(const struct magic *); 124 private size_t apprentice_magic_strength(const struct magic *, size_t); 125 private int apprentice_sort(const void *, const void *); 126 private void apprentice_list(struct mlist *, int ); 127 private struct magic_map *apprentice_load(struct magic_set *, 128 const char *, int); 129 private struct mlist *mlist_alloc(void); 130 private void mlist_free_all(struct magic_set *); 131 private void mlist_free(struct mlist *); 132 private void byteswap(struct magic *, uint32_t); 133 private void bs1(struct magic *); 134 135 #if defined(HAVE_BYTESWAP_H) 136 #define swap2(x) bswap_16(x) 137 #define swap4(x) bswap_32(x) 138 #define swap8(x) bswap_64(x) 139 #elif defined(HAVE_SYS_BSWAP_H) 140 #define swap2(x) bswap16(x) 141 #define swap4(x) bswap32(x) 142 #define swap8(x) bswap64(x) 143 #else 144 private uint16_t swap2(uint16_t); 145 private uint32_t swap4(uint32_t); 146 private uint64_t swap8(uint64_t); 147 #endif 148 149 private char *mkdbname(struct magic_set *, const char *, int); 150 private struct magic_map *apprentice_buf(struct magic_set *, struct magic *, 151 size_t); 152 private struct magic_map *apprentice_map(struct magic_set *, const char *); 153 private int check_buffer(struct magic_set *, struct magic_map *, const char *); 154 private void apprentice_unmap(struct magic_map *); 155 private int apprentice_compile(struct magic_set *, struct magic_map *, 156 const char *); 157 private int check_format_type(const char *, int, const char **); 158 private int check_format(struct magic_set *, struct magic *); 159 private int get_op(char); 160 private int parse_mime(struct magic_set *, struct magic_entry *, const char *, 161 size_t); 162 private int parse_strength(struct magic_set *, struct magic_entry *, 163 const char *, size_t); 164 private int parse_apple(struct magic_set *, struct magic_entry *, const char *, 165 size_t); 166 private int parse_ext(struct magic_set *, struct magic_entry *, const char *, 167 size_t); 168 169 170 private size_t magicsize = sizeof(struct magic); 171 172 private const char usg_hdr[] = "cont\toffset\ttype\topcode\tmask\tvalue\tdesc"; 173 174 private struct { 175 const char *name; 176 size_t len; 177 int (*fun)(struct magic_set *, struct magic_entry *, const char *, 178 size_t); 179 } bang[] = { 180 #define DECLARE_FIELD(name) { # name, sizeof(# name) - 1, parse_ ## name } 181 DECLARE_FIELD(mime), 182 DECLARE_FIELD(apple), 183 DECLARE_FIELD(ext), 184 DECLARE_FIELD(strength), 185 #undef DECLARE_FIELD 186 { NULL, 0, NULL } 187 }; 188 189 #ifdef COMPILE_ONLY 190 191 int main(int, char *[]); 192 193 int 194 main(int argc, char *argv[]) 195 { 196 int ret; 197 struct magic_set *ms; 198 char *progname; 199 200 if ((progname = strrchr(argv[0], '/')) != NULL) 201 progname++; 202 else 203 progname = argv[0]; 204 205 if (argc != 2) { 206 (void)fprintf(stderr, "Usage: %s file\n", progname); 207 return 1; 208 } 209 210 if ((ms = magic_open(MAGIC_CHECK)) == NULL) { 211 (void)fprintf(stderr, "%s: %s\n", progname, strerror(errno)); 212 return 1; 213 } 214 ret = magic_compile(ms, argv[1]) == -1 ? 1 : 0; 215 if (ret == 1) 216 (void)fprintf(stderr, "%s: %s\n", progname, magic_error(ms)); 217 magic_close(ms); 218 return ret; 219 } 220 #endif /* COMPILE_ONLY */ 221 222 struct type_tbl_s { 223 const char name[16]; 224 const size_t len; 225 const int type; 226 const int format; 227 }; 228 229 /* 230 * XXX - the actual Single UNIX Specification says that "long" means "long", 231 * as in the C data type, but we treat it as meaning "4-byte integer". 232 * Given that the OS X version of file 5.04 did the same, I guess that passes 233 * the actual test; having "long" be dependent on how big a "long" is on 234 * the machine running "file" is silly. 235 */ 236 static const struct type_tbl_s type_tbl[] = { 237 # define XX(s) s, (sizeof(s) - 1) 238 # define XX_NULL "", 0 239 { XX("invalid"), FILE_INVALID, FILE_FMT_NONE }, 240 { XX("byte"), FILE_BYTE, FILE_FMT_NUM }, 241 { XX("short"), FILE_SHORT, FILE_FMT_NUM }, 242 { XX("default"), FILE_DEFAULT, FILE_FMT_NONE }, 243 { XX("long"), FILE_LONG, FILE_FMT_NUM }, 244 { XX("string"), FILE_STRING, FILE_FMT_STR }, 245 { XX("date"), FILE_DATE, FILE_FMT_STR }, 246 { XX("beshort"), FILE_BESHORT, FILE_FMT_NUM }, 247 { XX("belong"), FILE_BELONG, FILE_FMT_NUM }, 248 { XX("bedate"), FILE_BEDATE, FILE_FMT_STR }, 249 { XX("leshort"), FILE_LESHORT, FILE_FMT_NUM }, 250 { XX("lelong"), FILE_LELONG, FILE_FMT_NUM }, 251 { XX("ledate"), FILE_LEDATE, FILE_FMT_STR }, 252 { XX("pstring"), FILE_PSTRING, FILE_FMT_STR }, 253 { XX("ldate"), FILE_LDATE, FILE_FMT_STR }, 254 { XX("beldate"), FILE_BELDATE, FILE_FMT_STR }, 255 { XX("leldate"), FILE_LELDATE, FILE_FMT_STR }, 256 { XX("regex"), FILE_REGEX, FILE_FMT_STR }, 257 { XX("bestring16"), FILE_BESTRING16, FILE_FMT_STR }, 258 { XX("lestring16"), FILE_LESTRING16, FILE_FMT_STR }, 259 { XX("search"), FILE_SEARCH, FILE_FMT_STR }, 260 { XX("medate"), FILE_MEDATE, FILE_FMT_STR }, 261 { XX("meldate"), FILE_MELDATE, FILE_FMT_STR }, 262 { XX("melong"), FILE_MELONG, FILE_FMT_NUM }, 263 { XX("quad"), FILE_QUAD, FILE_FMT_QUAD }, 264 { XX("lequad"), FILE_LEQUAD, FILE_FMT_QUAD }, 265 { XX("bequad"), FILE_BEQUAD, FILE_FMT_QUAD }, 266 { XX("qdate"), FILE_QDATE, FILE_FMT_STR }, 267 { XX("leqdate"), FILE_LEQDATE, FILE_FMT_STR }, 268 { XX("beqdate"), FILE_BEQDATE, FILE_FMT_STR }, 269 { XX("qldate"), FILE_QLDATE, FILE_FMT_STR }, 270 { XX("leqldate"), FILE_LEQLDATE, FILE_FMT_STR }, 271 { XX("beqldate"), FILE_BEQLDATE, FILE_FMT_STR }, 272 { XX("float"), FILE_FLOAT, FILE_FMT_FLOAT }, 273 { XX("befloat"), FILE_BEFLOAT, FILE_FMT_FLOAT }, 274 { XX("lefloat"), FILE_LEFLOAT, FILE_FMT_FLOAT }, 275 { XX("double"), FILE_DOUBLE, FILE_FMT_DOUBLE }, 276 { XX("bedouble"), FILE_BEDOUBLE, FILE_FMT_DOUBLE }, 277 { XX("ledouble"), FILE_LEDOUBLE, FILE_FMT_DOUBLE }, 278 { XX("leid3"), FILE_LEID3, FILE_FMT_NUM }, 279 { XX("beid3"), FILE_BEID3, FILE_FMT_NUM }, 280 { XX("indirect"), FILE_INDIRECT, FILE_FMT_NUM }, 281 { XX("qwdate"), FILE_QWDATE, FILE_FMT_STR }, 282 { XX("leqwdate"), FILE_LEQWDATE, FILE_FMT_STR }, 283 { XX("beqwdate"), FILE_BEQWDATE, FILE_FMT_STR }, 284 { XX("name"), FILE_NAME, FILE_FMT_NONE }, 285 { XX("use"), FILE_USE, FILE_FMT_NONE }, 286 { XX("clear"), FILE_CLEAR, FILE_FMT_NONE }, 287 { XX("der"), FILE_DER, FILE_FMT_STR }, 288 { XX("guid"), FILE_GUID, FILE_FMT_STR }, 289 { XX("offset"), FILE_OFFSET, FILE_FMT_QUAD }, 290 { XX("bevarint"), FILE_BEVARINT, FILE_FMT_STR }, 291 { XX("levarint"), FILE_LEVARINT, FILE_FMT_STR }, 292 { XX("msdosdate"), FILE_MSDOSDATE, FILE_FMT_STR }, 293 { XX("lemsdosdate"), FILE_LEMSDOSDATE, FILE_FMT_STR }, 294 { XX("bemsdosdate"), FILE_BEMSDOSDATE, FILE_FMT_STR }, 295 { XX("msdostime"), FILE_MSDOSTIME, FILE_FMT_STR }, 296 { XX("lemsdostime"), FILE_LEMSDOSTIME, FILE_FMT_STR }, 297 { XX("bemsdostime"), FILE_BEMSDOSTIME, FILE_FMT_STR }, 298 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 299 }; 300 301 /* 302 * These are not types, and cannot be preceded by "u" to make them 303 * unsigned. 304 */ 305 static const struct type_tbl_s special_tbl[] = { 306 { XX("der"), FILE_DER, FILE_FMT_STR }, 307 { XX("name"), FILE_NAME, FILE_FMT_STR }, 308 { XX("use"), FILE_USE, FILE_FMT_STR }, 309 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 310 }; 311 # undef XX 312 # undef XX_NULL 313 314 private int 315 get_type(const struct type_tbl_s *tbl, const char *l, const char **t) 316 { 317 const struct type_tbl_s *p; 318 319 for (p = tbl; p->len; p++) { 320 if (strncmp(l, p->name, p->len) == 0) { 321 if (t) 322 *t = l + p->len; 323 break; 324 } 325 } 326 return p->type; 327 } 328 329 private off_t 330 maxoff_t(void) { 331 if (/*CONSTCOND*/sizeof(off_t) == sizeof(int)) 332 return CAST(off_t, INT_MAX); 333 if (/*CONSTCOND*/sizeof(off_t) == sizeof(long)) 334 return CAST(off_t, LONG_MAX); 335 return 0x7fffffff; 336 } 337 338 private int 339 get_standard_integer_type(const char *l, const char **t) 340 { 341 int type; 342 343 if (isalpha(CAST(unsigned char, l[1]))) { 344 switch (l[1]) { 345 case 'C': 346 /* "dC" and "uC" */ 347 type = FILE_BYTE; 348 break; 349 case 'S': 350 /* "dS" and "uS" */ 351 type = FILE_SHORT; 352 break; 353 case 'I': 354 case 'L': 355 /* 356 * "dI", "dL", "uI", and "uL". 357 * 358 * XXX - the actual Single UNIX Specification says 359 * that "L" means "long", as in the C data type, 360 * but we treat it as meaning "4-byte integer". 361 * Given that the OS X version of file 5.04 did 362 * the same, I guess that passes the actual SUS 363 * validation suite; having "dL" be dependent on 364 * how big a "long" is on the machine running 365 * "file" is silly. 366 */ 367 type = FILE_LONG; 368 break; 369 case 'Q': 370 /* "dQ" and "uQ" */ 371 type = FILE_QUAD; 372 break; 373 default: 374 /* "d{anything else}", "u{anything else}" */ 375 return FILE_INVALID; 376 } 377 l += 2; 378 } else if (isdigit(CAST(unsigned char, l[1]))) { 379 /* 380 * "d{num}" and "u{num}"; we only support {num} values 381 * of 1, 2, 4, and 8 - the Single UNIX Specification 382 * doesn't say anything about whether arbitrary 383 * values should be supported, but both the Solaris 10 384 * and OS X Mountain Lion versions of file passed the 385 * Single UNIX Specification validation suite, and 386 * neither of them support values bigger than 8 or 387 * non-power-of-2 values. 388 */ 389 if (isdigit(CAST(unsigned char, l[2]))) { 390 /* Multi-digit, so > 9 */ 391 return FILE_INVALID; 392 } 393 switch (l[1]) { 394 case '1': 395 type = FILE_BYTE; 396 break; 397 case '2': 398 type = FILE_SHORT; 399 break; 400 case '4': 401 type = FILE_LONG; 402 break; 403 case '8': 404 type = FILE_QUAD; 405 break; 406 default: 407 /* XXX - what about 3, 5, 6, or 7? */ 408 return FILE_INVALID; 409 } 410 l += 2; 411 } else { 412 /* 413 * "d" or "u" by itself. 414 */ 415 type = FILE_LONG; 416 ++l; 417 } 418 if (t) 419 *t = l; 420 return type; 421 } 422 423 private void 424 init_file_tables(void) 425 { 426 static int done = 0; 427 const struct type_tbl_s *p; 428 429 if (done) 430 return; 431 done++; 432 433 for (p = type_tbl; p->len; p++) { 434 assert(p->type < FILE_NAMES_SIZE); 435 file_names[p->type] = p->name; 436 file_formats[p->type] = p->format; 437 } 438 assert(p - type_tbl == FILE_NAMES_SIZE); 439 } 440 441 private int 442 add_mlist(struct mlist *mlp, struct magic_map *map, size_t idx) 443 { 444 struct mlist *ml; 445 446 mlp->map = NULL; 447 if ((ml = CAST(struct mlist *, malloc(sizeof(*ml)))) == NULL) 448 return -1; 449 450 ml->map = idx == 0 ? map : NULL; 451 ml->magic = map->magic[idx]; 452 ml->nmagic = map->nmagic[idx]; 453 if (ml->nmagic) { 454 ml->magic_rxcomp = CAST(file_regex_t **, 455 calloc(ml->nmagic, sizeof(*ml->magic_rxcomp))); 456 if (ml->magic_rxcomp == NULL) { 457 free(ml); 458 return -1; 459 } 460 } else 461 ml->magic_rxcomp = NULL; 462 mlp->prev->next = ml; 463 ml->prev = mlp->prev; 464 ml->next = mlp; 465 mlp->prev = ml; 466 return 0; 467 } 468 469 /* 470 * Handle one file or directory. 471 */ 472 private int 473 apprentice_1(struct magic_set *ms, const char *fn, int action) 474 { 475 struct magic_map *map; 476 #ifndef COMPILE_ONLY 477 struct mlist *ml; 478 size_t i; 479 #endif 480 481 if (magicsize != FILE_MAGICSIZE) { 482 file_error(ms, 0, "magic element size %lu != %lu", 483 CAST(unsigned long, sizeof(*map->magic[0])), 484 CAST(unsigned long, FILE_MAGICSIZE)); 485 return -1; 486 } 487 488 if (action == FILE_COMPILE) { 489 map = apprentice_load(ms, fn, action); 490 if (map == NULL) 491 return -1; 492 return apprentice_compile(ms, map, fn); 493 } 494 495 #ifndef COMPILE_ONLY 496 map = apprentice_map(ms, fn); 497 if (map == NULL) { 498 if (ms->flags & MAGIC_CHECK) 499 file_magwarn(ms, "using regular magic file `%s'", fn); 500 map = apprentice_load(ms, fn, action); 501 if (map == NULL) 502 return -1; 503 } 504 505 for (i = 0; i < MAGIC_SETS; i++) { 506 if (add_mlist(ms->mlist[i], map, i) == -1) { 507 /* failed to add to any list, free explicitly */ 508 if (i == 0) 509 apprentice_unmap(map); 510 else 511 mlist_free_all(ms); 512 file_oomem(ms, sizeof(*ml)); 513 return -1; 514 } 515 } 516 517 if (action == FILE_LIST) { 518 for (i = 0; i < MAGIC_SETS; i++) { 519 printf("Set %" SIZE_T_FORMAT "u:\nBinary patterns:\n", 520 i); 521 apprentice_list(ms->mlist[i], BINTEST); 522 printf("Text patterns:\n"); 523 apprentice_list(ms->mlist[i], TEXTTEST); 524 } 525 } 526 return 0; 527 #else 528 return 0; 529 #endif /* COMPILE_ONLY */ 530 } 531 532 protected void 533 file_ms_free(struct magic_set *ms) 534 { 535 size_t i; 536 if (ms == NULL) 537 return; 538 for (i = 0; i < MAGIC_SETS; i++) 539 mlist_free(ms->mlist[i]); 540 free(ms->o.pbuf); 541 free(ms->o.buf); 542 free(ms->c.li); 543 #ifdef USE_C_LOCALE 544 freelocale(ms->c_lc_ctype); 545 #endif 546 free(ms); 547 } 548 549 protected struct magic_set * 550 file_ms_alloc(int flags) 551 { 552 struct magic_set *ms; 553 size_t i, len; 554 555 if ((ms = CAST(struct magic_set *, calloc(CAST(size_t, 1u), 556 sizeof(struct magic_set)))) == NULL) 557 return NULL; 558 559 if (magic_setflags(ms, flags) == -1) { 560 errno = EINVAL; 561 goto free; 562 } 563 564 ms->o.buf = ms->o.pbuf = NULL; 565 ms->o.blen = 0; 566 len = (ms->c.len = 10) * sizeof(*ms->c.li); 567 568 if ((ms->c.li = CAST(struct level_info *, malloc(len))) == NULL) 569 goto free; 570 571 ms->event_flags = 0; 572 ms->error = -1; 573 for (i = 0; i < MAGIC_SETS; i++) 574 ms->mlist[i] = NULL; 575 ms->file = "unknown"; 576 ms->line = 0; 577 ms->indir_max = FILE_INDIR_MAX; 578 ms->name_max = FILE_NAME_MAX; 579 ms->elf_shnum_max = FILE_ELF_SHNUM_MAX; 580 ms->elf_phnum_max = FILE_ELF_PHNUM_MAX; 581 ms->elf_notes_max = FILE_ELF_NOTES_MAX; 582 ms->regex_max = FILE_REGEX_MAX; 583 ms->bytes_max = FILE_BYTES_MAX; 584 ms->encoding_max = FILE_ENCODING_MAX; 585 #ifdef USE_C_LOCALE 586 ms->c_lc_ctype = newlocale(LC_CTYPE_MASK, "C", 0); 587 assert(ms->c_lc_ctype != NULL); 588 #endif 589 return ms; 590 free: 591 free(ms); 592 return NULL; 593 } 594 595 private void 596 apprentice_unmap(struct magic_map *map) 597 { 598 size_t i; 599 char *p; 600 if (map == NULL) 601 return; 602 603 switch (map->type) { 604 case MAP_TYPE_USER: 605 break; 606 case MAP_TYPE_MALLOC: 607 p = CAST(char *, map->p); 608 for (i = 0; i < MAGIC_SETS; i++) { 609 char *b = RCAST(char *, map->magic[i]); 610 if (p != NULL && b >= p && b <= p + map->len) 611 continue; 612 free(b); 613 } 614 free(p); 615 break; 616 #ifdef QUICK 617 case MAP_TYPE_MMAP: 618 if (map->p && map->p != MAP_FAILED) 619 (void)munmap(map->p, map->len); 620 break; 621 #endif 622 default: 623 fprintf(stderr, "Bad map type %d", map->type); 624 abort(); 625 } 626 free(map); 627 } 628 629 private struct mlist * 630 mlist_alloc(void) 631 { 632 struct mlist *mlist; 633 if ((mlist = CAST(struct mlist *, calloc(1, sizeof(*mlist)))) == NULL) { 634 return NULL; 635 } 636 mlist->next = mlist->prev = mlist; 637 return mlist; 638 } 639 640 private void 641 mlist_free_all(struct magic_set *ms) 642 { 643 size_t i; 644 645 for (i = 0; i < MAGIC_SETS; i++) { 646 mlist_free(ms->mlist[i]); 647 ms->mlist[i] = NULL; 648 } 649 } 650 651 private void 652 mlist_free_one(struct mlist *ml) 653 { 654 size_t i; 655 656 if (ml->map) 657 apprentice_unmap(CAST(struct magic_map *, ml->map)); 658 659 for (i = 0; i < ml->nmagic; ++i) { 660 if (ml->magic_rxcomp[i]) { 661 file_regfree(ml->magic_rxcomp[i]); 662 free(ml->magic_rxcomp[i]); 663 ml->magic_rxcomp[i] = NULL; 664 } 665 } 666 free(ml->magic_rxcomp); 667 ml->magic_rxcomp = NULL; 668 free(ml); 669 } 670 671 private void 672 mlist_free(struct mlist *mlist) 673 { 674 struct mlist *ml, *next; 675 676 if (mlist == NULL) 677 return; 678 679 for (ml = mlist->next; ml != mlist;) { 680 next = ml->next; 681 mlist_free_one(ml); 682 ml = next; 683 } 684 mlist_free_one(mlist); 685 } 686 687 #ifndef COMPILE_ONLY 688 /* void **bufs: an array of compiled magic files */ 689 protected int 690 buffer_apprentice(struct magic_set *ms, struct magic **bufs, 691 size_t *sizes, size_t nbufs) 692 { 693 size_t i, j; 694 struct mlist *ml; 695 struct magic_map *map; 696 697 if (nbufs == 0) 698 return -1; 699 700 (void)file_reset(ms, 0); 701 702 init_file_tables(); 703 704 for (i = 0; i < MAGIC_SETS; i++) { 705 mlist_free(ms->mlist[i]); 706 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 707 file_oomem(ms, sizeof(*ms->mlist[i])); 708 goto fail; 709 } 710 } 711 712 for (i = 0; i < nbufs; i++) { 713 map = apprentice_buf(ms, bufs[i], sizes[i]); 714 if (map == NULL) 715 goto fail; 716 717 for (j = 0; j < MAGIC_SETS; j++) { 718 if (add_mlist(ms->mlist[j], map, j) == -1) { 719 file_oomem(ms, sizeof(*ml)); 720 goto fail; 721 } 722 } 723 } 724 725 return 0; 726 fail: 727 mlist_free_all(ms); 728 return -1; 729 } 730 #endif 731 732 /* const char *fn: list of magic files and directories */ 733 protected int 734 file_apprentice(struct magic_set *ms, const char *fn, int action) 735 { 736 char *p, *mfn; 737 int fileerr, errs = -1; 738 size_t i, j; 739 740 (void)file_reset(ms, 0); 741 742 if ((fn = magic_getpath(fn, action)) == NULL) 743 return -1; 744 745 init_file_tables(); 746 747 if ((mfn = strdup(fn)) == NULL) { 748 file_oomem(ms, strlen(fn)); 749 return -1; 750 } 751 752 for (i = 0; i < MAGIC_SETS; i++) { 753 mlist_free(ms->mlist[i]); 754 if ((ms->mlist[i] = mlist_alloc()) == NULL) { 755 file_oomem(ms, sizeof(*ms->mlist[i])); 756 for (j = 0; j < i; j++) { 757 mlist_free(ms->mlist[j]); 758 ms->mlist[j] = NULL; 759 } 760 free(mfn); 761 return -1; 762 } 763 } 764 fn = mfn; 765 766 while (fn) { 767 p = CCAST(char *, strchr(fn, PATHSEP)); 768 if (p) 769 *p++ = '\0'; 770 if (*fn == '\0') 771 break; 772 fileerr = apprentice_1(ms, fn, action); 773 errs = MAX(errs, fileerr); 774 fn = p; 775 } 776 777 free(mfn); 778 779 if (errs == -1) { 780 for (i = 0; i < MAGIC_SETS; i++) { 781 mlist_free(ms->mlist[i]); 782 ms->mlist[i] = NULL; 783 } 784 file_error(ms, 0, "could not find any valid magic files!"); 785 return -1; 786 } 787 788 #if 0 789 /* 790 * Always leave the database loaded 791 */ 792 if (action == FILE_LOAD) 793 return 0; 794 795 for (i = 0; i < MAGIC_SETS; i++) { 796 mlist_free(ms->mlist[i]); 797 ms->mlist[i] = NULL; 798 } 799 #endif 800 801 switch (action) { 802 case FILE_LOAD: 803 case FILE_COMPILE: 804 case FILE_CHECK: 805 case FILE_LIST: 806 return 0; 807 default: 808 file_error(ms, 0, "Invalid action %d", action); 809 return -1; 810 } 811 } 812 813 /* 814 * Compute the real length of a magic expression, for the purposes 815 * of determining how "strong" a magic expression is (approximating 816 * how specific its matches are): 817 * - magic characters count 0 unless escaped. 818 * - [] expressions count 1 819 * - {} expressions count 0 820 * - regular characters or escaped magic characters count 1 821 * - 0 length expressions count as one 822 */ 823 private size_t 824 nonmagic(const char *str) 825 { 826 const char *p; 827 size_t rv = 0; 828 829 for (p = str; *p; p++) 830 switch (*p) { 831 case '\\': /* Escaped anything counts 1 */ 832 if (!*++p) 833 p--; 834 rv++; 835 continue; 836 case '?': /* Magic characters count 0 */ 837 case '*': 838 case '.': 839 case '+': 840 case '^': 841 case '$': 842 continue; 843 case '[': /* Bracketed expressions count 1 the ']' */ 844 while (*p && *p != ']') 845 p++; 846 p--; 847 continue; 848 case '{': /* Braced expressions count 0 */ 849 while (*p && *p != '}') 850 p++; 851 if (!*p) 852 p--; 853 continue; 854 default: /* Anything else counts 1 */ 855 rv++; 856 continue; 857 } 858 859 return rv == 0 ? 1 : rv; /* Return at least 1 */ 860 } 861 862 863 private size_t 864 typesize(int type) 865 { 866 switch (type) { 867 case FILE_BYTE: 868 return 1; 869 870 case FILE_SHORT: 871 case FILE_LESHORT: 872 case FILE_BESHORT: 873 case FILE_MSDOSDATE: 874 case FILE_BEMSDOSDATE: 875 case FILE_LEMSDOSDATE: 876 case FILE_MSDOSTIME: 877 case FILE_BEMSDOSTIME: 878 case FILE_LEMSDOSTIME: 879 return 2; 880 881 case FILE_LONG: 882 case FILE_LELONG: 883 case FILE_BELONG: 884 case FILE_MELONG: 885 return 4; 886 887 case FILE_DATE: 888 case FILE_LEDATE: 889 case FILE_BEDATE: 890 case FILE_MEDATE: 891 case FILE_LDATE: 892 case FILE_LELDATE: 893 case FILE_BELDATE: 894 case FILE_MELDATE: 895 case FILE_FLOAT: 896 case FILE_BEFLOAT: 897 case FILE_LEFLOAT: 898 case FILE_BEID3: 899 case FILE_LEID3: 900 return 4; 901 902 case FILE_QUAD: 903 case FILE_BEQUAD: 904 case FILE_LEQUAD: 905 case FILE_QDATE: 906 case FILE_LEQDATE: 907 case FILE_BEQDATE: 908 case FILE_QLDATE: 909 case FILE_LEQLDATE: 910 case FILE_BEQLDATE: 911 case FILE_QWDATE: 912 case FILE_LEQWDATE: 913 case FILE_BEQWDATE: 914 case FILE_DOUBLE: 915 case FILE_BEDOUBLE: 916 case FILE_LEDOUBLE: 917 case FILE_OFFSET: 918 case FILE_BEVARINT: 919 case FILE_LEVARINT: 920 return 8; 921 922 case FILE_GUID: 923 return 16; 924 925 default: 926 return FILE_BADSIZE; 927 } 928 } 929 930 /* 931 * Get weight of this magic entry, for sorting purposes. 932 */ 933 private ssize_t 934 apprentice_magic_strength_1(const struct magic *m) 935 { 936 #define MULT 10U 937 size_t ts, v; 938 ssize_t val = 2 * MULT; /* baseline strength */ 939 940 switch (m->type) { 941 case FILE_DEFAULT: /* make sure this sorts last */ 942 if (m->factor_op != FILE_FACTOR_OP_NONE) { 943 fprintf(stderr, "Bad factor_op %d", m->factor_op); 944 abort(); 945 } 946 return 0; 947 948 case FILE_BYTE: 949 case FILE_SHORT: 950 case FILE_LESHORT: 951 case FILE_BESHORT: 952 case FILE_LONG: 953 case FILE_LELONG: 954 case FILE_BELONG: 955 case FILE_MELONG: 956 case FILE_DATE: 957 case FILE_LEDATE: 958 case FILE_BEDATE: 959 case FILE_MEDATE: 960 case FILE_LDATE: 961 case FILE_LELDATE: 962 case FILE_BELDATE: 963 case FILE_MELDATE: 964 case FILE_FLOAT: 965 case FILE_BEFLOAT: 966 case FILE_LEFLOAT: 967 case FILE_QUAD: 968 case FILE_BEQUAD: 969 case FILE_LEQUAD: 970 case FILE_QDATE: 971 case FILE_LEQDATE: 972 case FILE_BEQDATE: 973 case FILE_QLDATE: 974 case FILE_LEQLDATE: 975 case FILE_BEQLDATE: 976 case FILE_QWDATE: 977 case FILE_LEQWDATE: 978 case FILE_BEQWDATE: 979 case FILE_DOUBLE: 980 case FILE_BEDOUBLE: 981 case FILE_LEDOUBLE: 982 case FILE_BEVARINT: 983 case FILE_LEVARINT: 984 case FILE_GUID: 985 case FILE_BEID3: 986 case FILE_LEID3: 987 case FILE_OFFSET: 988 case FILE_MSDOSDATE: 989 case FILE_BEMSDOSDATE: 990 case FILE_LEMSDOSDATE: 991 case FILE_MSDOSTIME: 992 case FILE_BEMSDOSTIME: 993 case FILE_LEMSDOSTIME: 994 ts = typesize(m->type); 995 if (ts == FILE_BADSIZE) { 996 (void)fprintf(stderr, "Bad size for type %d\n", 997 m->type); 998 abort(); 999 } 1000 val += ts * MULT; 1001 break; 1002 1003 case FILE_PSTRING: 1004 case FILE_STRING: 1005 val += m->vallen * MULT; 1006 break; 1007 1008 case FILE_BESTRING16: 1009 case FILE_LESTRING16: 1010 val += m->vallen * MULT / 2; 1011 break; 1012 1013 case FILE_SEARCH: 1014 if (m->vallen == 0) 1015 break; 1016 val += m->vallen * MAX(MULT / m->vallen, 1); 1017 break; 1018 1019 case FILE_REGEX: 1020 v = nonmagic(m->value.s); 1021 val += v * MAX(MULT / v, 1); 1022 break; 1023 1024 case FILE_INDIRECT: 1025 case FILE_NAME: 1026 case FILE_USE: 1027 case FILE_CLEAR: 1028 break; 1029 1030 case FILE_DER: 1031 val += MULT; 1032 break; 1033 1034 default: 1035 (void)fprintf(stderr, "Bad type %d\n", m->type); 1036 abort(); 1037 } 1038 1039 switch (m->reln) { 1040 case 'x': /* matches anything penalize */ 1041 case '!': /* matches almost anything penalize */ 1042 val = 0; 1043 break; 1044 1045 case '=': /* Exact match, prefer */ 1046 val += MULT; 1047 break; 1048 1049 case '>': 1050 case '<': /* comparison match reduce strength */ 1051 val -= 2 * MULT; 1052 break; 1053 1054 case '^': 1055 case '&': /* masking bits, we could count them too */ 1056 val -= MULT; 1057 break; 1058 1059 default: 1060 (void)fprintf(stderr, "Bad relation %c\n", m->reln); 1061 abort(); 1062 } 1063 1064 return val; 1065 } 1066 1067 1068 private size_t 1069 apprentice_magic_strength(const struct magic *m, 1070 size_t nmagic __attribute__((__unused__))) 1071 { 1072 ssize_t val = apprentice_magic_strength_1(m); 1073 1074 #ifdef notyet 1075 if (m->desc[0] == '\0') { 1076 size_t i; 1077 /* 1078 * Magic entries with no description get their continuations 1079 * added 1080 */ 1081 for (i = 1; m[i].cont_level != 0 && i < MIN(nmagic, 3); i++) { 1082 ssize_t v = apprentice_magic_strength_1(&m[i]) >> 1083 (i + 1); 1084 val += v; 1085 if (m[i].desc[0] != '\0') 1086 break; 1087 } 1088 } 1089 #endif 1090 1091 switch (m->factor_op) { 1092 case FILE_FACTOR_OP_NONE: 1093 break; 1094 case FILE_FACTOR_OP_PLUS: 1095 val += m->factor; 1096 break; 1097 case FILE_FACTOR_OP_MINUS: 1098 val -= m->factor; 1099 break; 1100 case FILE_FACTOR_OP_TIMES: 1101 val *= m->factor; 1102 break; 1103 case FILE_FACTOR_OP_DIV: 1104 val /= m->factor; 1105 break; 1106 default: 1107 (void)fprintf(stderr, "Bad factor_op %u\n", m->factor_op); 1108 abort(); 1109 } 1110 1111 if (val <= 0) /* ensure we only return 0 for FILE_DEFAULT */ 1112 val = 1; 1113 1114 #ifndef notyet 1115 /* 1116 * Magic entries with no description get a bonus because they depend 1117 * on subsequent magic entries to print something. 1118 */ 1119 if (m->desc[0] == '\0') 1120 val++; 1121 #endif 1122 1123 return val; 1124 } 1125 1126 /* 1127 * Sort callback for sorting entries by "strength" (basically length) 1128 */ 1129 private int 1130 apprentice_sort(const void *a, const void *b) 1131 { 1132 const struct magic_entry *ma = CAST(const struct magic_entry *, a); 1133 const struct magic_entry *mb = CAST(const struct magic_entry *, b); 1134 size_t sa = apprentice_magic_strength(ma->mp, ma->cont_count); 1135 size_t sb = apprentice_magic_strength(mb->mp, mb->cont_count); 1136 if (sa == sb) 1137 return 0; 1138 else if (sa > sb) 1139 return -1; 1140 else 1141 return 1; 1142 } 1143 1144 /* 1145 * Shows sorted patterns list in the order which is used for the matching 1146 */ 1147 private void 1148 apprentice_list(struct mlist *mlist, int mode) 1149 { 1150 uint32_t magindex, descindex, mimeindex, lineindex; 1151 struct mlist *ml; 1152 for (ml = mlist->next; ml != mlist; ml = ml->next) { 1153 for (magindex = 0; magindex < ml->nmagic; magindex++) { 1154 struct magic *m = &ml->magic[magindex]; 1155 if ((m->flag & mode) != mode) { 1156 /* Skip sub-tests */ 1157 while (magindex + 1 < ml->nmagic && 1158 ml->magic[magindex + 1].cont_level != 0) 1159 ++magindex; 1160 continue; /* Skip to next top-level test*/ 1161 } 1162 1163 /* 1164 * Try to iterate over the tree until we find item with 1165 * description/mimetype. 1166 */ 1167 lineindex = descindex = mimeindex = magindex; 1168 for (magindex++; magindex < ml->nmagic && 1169 ml->magic[magindex].cont_level != 0; magindex++) { 1170 if (*ml->magic[descindex].desc == '\0' 1171 && *ml->magic[magindex].desc) 1172 descindex = magindex; 1173 if (*ml->magic[mimeindex].mimetype == '\0' 1174 && *ml->magic[magindex].mimetype) 1175 mimeindex = magindex; 1176 } 1177 1178 printf("Strength = %3" SIZE_T_FORMAT "u@%u: %s [%s]\n", 1179 apprentice_magic_strength(m, ml->nmagic - magindex), 1180 ml->magic[lineindex].lineno, 1181 ml->magic[descindex].desc, 1182 ml->magic[mimeindex].mimetype); 1183 } 1184 } 1185 } 1186 1187 private void 1188 set_test_type(struct magic *mstart, struct magic *m) 1189 { 1190 switch (m->type) { 1191 case FILE_BYTE: 1192 case FILE_SHORT: 1193 case FILE_LONG: 1194 case FILE_DATE: 1195 case FILE_BESHORT: 1196 case FILE_BELONG: 1197 case FILE_BEDATE: 1198 case FILE_LESHORT: 1199 case FILE_LELONG: 1200 case FILE_LEDATE: 1201 case FILE_LDATE: 1202 case FILE_BELDATE: 1203 case FILE_LELDATE: 1204 case FILE_MEDATE: 1205 case FILE_MELDATE: 1206 case FILE_MELONG: 1207 case FILE_QUAD: 1208 case FILE_LEQUAD: 1209 case FILE_BEQUAD: 1210 case FILE_QDATE: 1211 case FILE_LEQDATE: 1212 case FILE_BEQDATE: 1213 case FILE_QLDATE: 1214 case FILE_LEQLDATE: 1215 case FILE_BEQLDATE: 1216 case FILE_QWDATE: 1217 case FILE_LEQWDATE: 1218 case FILE_BEQWDATE: 1219 case FILE_FLOAT: 1220 case FILE_BEFLOAT: 1221 case FILE_LEFLOAT: 1222 case FILE_DOUBLE: 1223 case FILE_BEDOUBLE: 1224 case FILE_LEDOUBLE: 1225 case FILE_BEVARINT: 1226 case FILE_LEVARINT: 1227 case FILE_DER: 1228 case FILE_GUID: 1229 case FILE_OFFSET: 1230 case FILE_MSDOSDATE: 1231 case FILE_BEMSDOSDATE: 1232 case FILE_LEMSDOSDATE: 1233 case FILE_MSDOSTIME: 1234 case FILE_BEMSDOSTIME: 1235 case FILE_LEMSDOSTIME: 1236 mstart->flag |= BINTEST; 1237 break; 1238 case FILE_STRING: 1239 case FILE_PSTRING: 1240 case FILE_BESTRING16: 1241 case FILE_LESTRING16: 1242 /* Allow text overrides */ 1243 if (mstart->str_flags & STRING_TEXTTEST) 1244 mstart->flag |= TEXTTEST; 1245 else 1246 mstart->flag |= BINTEST; 1247 break; 1248 case FILE_REGEX: 1249 case FILE_SEARCH: 1250 /* Check for override */ 1251 if (mstart->str_flags & STRING_BINTEST) 1252 mstart->flag |= BINTEST; 1253 if (mstart->str_flags & STRING_TEXTTEST) 1254 mstart->flag |= TEXTTEST; 1255 1256 if (mstart->flag & (TEXTTEST|BINTEST)) 1257 break; 1258 1259 /* binary test if pattern is not text */ 1260 if (file_looks_utf8(m->value.us, CAST(size_t, m->vallen), NULL, 1261 NULL) <= 0) 1262 mstart->flag |= BINTEST; 1263 else 1264 mstart->flag |= TEXTTEST; 1265 break; 1266 case FILE_DEFAULT: 1267 /* can't deduce anything; we shouldn't see this at the 1268 top level anyway */ 1269 break; 1270 case FILE_INVALID: 1271 default: 1272 /* invalid search type, but no need to complain here */ 1273 break; 1274 } 1275 } 1276 1277 private int 1278 addentry(struct magic_set *ms, struct magic_entry *me, 1279 struct magic_entry_set *mset) 1280 { 1281 size_t i = me->mp->type == FILE_NAME ? 1 : 0; 1282 if (mset[i].me == NULL || mset[i].count == mset[i].max) { 1283 struct magic_entry *mp; 1284 1285 size_t incr = mset[i].max + ALLOC_INCR; 1286 if ((mp = CAST(struct magic_entry *, 1287 realloc(mset[i].me, sizeof(*mp) * incr))) == 1288 NULL) { 1289 file_oomem(ms, sizeof(*mp) * incr); 1290 return -1; 1291 } 1292 (void)memset(&mp[mset[i].count], 0, sizeof(*mp) * 1293 ALLOC_INCR); 1294 mset[i].me = mp; 1295 mset[i].max = incr; 1296 } 1297 mset[i].me[mset[i].count++] = *me; 1298 memset(me, 0, sizeof(*me)); 1299 return 0; 1300 } 1301 1302 /* 1303 * Load and parse one file. 1304 */ 1305 private void 1306 load_1(struct magic_set *ms, int action, const char *fn, int *errs, 1307 struct magic_entry_set *mset) 1308 { 1309 size_t lineno = 0, llen = 0; 1310 char *line = NULL; 1311 ssize_t len; 1312 struct magic_entry me; 1313 1314 FILE *f = fopen(ms->file = fn, "r"); 1315 if (f == NULL) { 1316 if (errno != ENOENT) 1317 file_error(ms, errno, "cannot read magic file `%s'", 1318 fn); 1319 (*errs)++; 1320 return; 1321 } 1322 1323 memset(&me, 0, sizeof(me)); 1324 /* read and parse this file */ 1325 for (ms->line = 1; (len = getline(&line, &llen, f)) != -1; 1326 ms->line++) { 1327 if (len == 0) /* null line, garbage, etc */ 1328 continue; 1329 if (line[len - 1] == '\n') { 1330 lineno++; 1331 line[len - 1] = '\0'; /* delete newline */ 1332 } 1333 switch (line[0]) { 1334 case '\0': /* empty, do not parse */ 1335 case '#': /* comment, do not parse */ 1336 continue; 1337 case '!': 1338 if (line[1] == ':') { 1339 size_t i; 1340 1341 for (i = 0; bang[i].name != NULL; i++) { 1342 if (CAST(size_t, len - 2) > bang[i].len && 1343 memcmp(bang[i].name, line + 2, 1344 bang[i].len) == 0) 1345 break; 1346 } 1347 if (bang[i].name == NULL) { 1348 file_error(ms, 0, 1349 "Unknown !: entry `%s'", line); 1350 (*errs)++; 1351 continue; 1352 } 1353 if (me.mp == NULL) { 1354 file_error(ms, 0, 1355 "No current entry for :!%s type", 1356 bang[i].name); 1357 (*errs)++; 1358 continue; 1359 } 1360 if ((*bang[i].fun)(ms, &me, 1361 line + bang[i].len + 2, 1362 len - bang[i].len - 2) != 0) { 1363 (*errs)++; 1364 continue; 1365 } 1366 continue; 1367 } 1368 /*FALLTHROUGH*/ 1369 default: 1370 again: 1371 switch (parse(ms, &me, line, lineno, action)) { 1372 case 0: 1373 continue; 1374 case 1: 1375 (void)addentry(ms, &me, mset); 1376 goto again; 1377 default: 1378 (*errs)++; 1379 break; 1380 } 1381 } 1382 } 1383 if (me.mp) 1384 (void)addentry(ms, &me, mset); 1385 free(line); 1386 (void)fclose(f); 1387 } 1388 1389 /* 1390 * parse a file or directory of files 1391 * const char *fn: name of magic file or directory 1392 */ 1393 private int 1394 cmpstrp(const void *p1, const void *p2) 1395 { 1396 return strcmp(*RCAST(char *const *, p1), *RCAST(char *const *, p2)); 1397 } 1398 1399 1400 private uint32_t 1401 set_text_binary(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1402 uint32_t starttest) 1403 { 1404 static const char text[] = "text"; 1405 static const char binary[] = "binary"; 1406 static const size_t len = sizeof(text); 1407 1408 uint32_t i = starttest; 1409 1410 do { 1411 set_test_type(me[starttest].mp, me[i].mp); 1412 if ((ms->flags & MAGIC_DEBUG) == 0) 1413 continue; 1414 (void)fprintf(stderr, "%s%s%s: %s\n", 1415 me[i].mp->mimetype, 1416 me[i].mp->mimetype[0] == '\0' ? "" : "; ", 1417 me[i].mp->desc[0] ? me[i].mp->desc : "(no description)", 1418 me[i].mp->flag & BINTEST ? binary : text); 1419 if (me[i].mp->flag & BINTEST) { 1420 char *p = strstr(me[i].mp->desc, text); 1421 if (p && (p == me[i].mp->desc || 1422 isspace(CAST(unsigned char, p[-1]))) && 1423 (p + len - me[i].mp->desc == MAXstring 1424 || (p[len] == '\0' || 1425 isspace(CAST(unsigned char, p[len]))))) 1426 (void)fprintf(stderr, "*** Possible " 1427 "binary test for text type\n"); 1428 } 1429 } while (++i < nme && me[i].mp->cont_level != 0); 1430 return i; 1431 } 1432 1433 private void 1434 set_last_default(struct magic_set *ms, struct magic_entry *me, uint32_t nme) 1435 { 1436 uint32_t i; 1437 for (i = 0; i < nme; i++) { 1438 if (me[i].mp->cont_level == 0 && 1439 me[i].mp->type == FILE_DEFAULT) { 1440 while (++i < nme) 1441 if (me[i].mp->cont_level == 0) 1442 break; 1443 if (i != nme) { 1444 /* XXX - Ugh! */ 1445 ms->line = me[i].mp->lineno; 1446 file_magwarn(ms, 1447 "level 0 \"default\" did not sort last"); 1448 } 1449 return; 1450 } 1451 } 1452 } 1453 1454 private int 1455 coalesce_entries(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1456 struct magic **ma, uint32_t *nma) 1457 { 1458 uint32_t i, mentrycount = 0; 1459 size_t slen; 1460 1461 for (i = 0; i < nme; i++) 1462 mentrycount += me[i].cont_count; 1463 1464 if (mentrycount == 0) { 1465 *ma = NULL; 1466 *nma = 0; 1467 return 0; 1468 } 1469 1470 slen = sizeof(**ma) * mentrycount; 1471 if ((*ma = CAST(struct magic *, malloc(slen))) == NULL) { 1472 file_oomem(ms, slen); 1473 return -1; 1474 } 1475 1476 mentrycount = 0; 1477 for (i = 0; i < nme; i++) { 1478 (void)memcpy(*ma + mentrycount, me[i].mp, 1479 me[i].cont_count * sizeof(**ma)); 1480 mentrycount += me[i].cont_count; 1481 } 1482 *nma = mentrycount; 1483 return 0; 1484 } 1485 1486 private void 1487 magic_entry_free(struct magic_entry *me, uint32_t nme) 1488 { 1489 uint32_t i; 1490 if (me == NULL) 1491 return; 1492 for (i = 0; i < nme; i++) 1493 free(me[i].mp); 1494 free(me); 1495 } 1496 1497 private struct magic_map * 1498 apprentice_load(struct magic_set *ms, const char *fn, int action) 1499 { 1500 int errs = 0; 1501 uint32_t i, j; 1502 size_t files = 0, maxfiles = 0; 1503 char **filearr = NULL, *mfn; 1504 struct stat st; 1505 struct magic_map *map; 1506 struct magic_entry_set mset[MAGIC_SETS]; 1507 DIR *dir; 1508 struct dirent *d; 1509 1510 memset(mset, 0, sizeof(mset)); 1511 ms->flags |= MAGIC_CHECK; /* Enable checks for parsed files */ 1512 1513 1514 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) 1515 { 1516 file_oomem(ms, sizeof(*map)); 1517 return NULL; 1518 } 1519 map->type = MAP_TYPE_MALLOC; 1520 1521 /* print silly verbose header for USG compat. */ 1522 if (action == FILE_CHECK) 1523 (void)fprintf(stderr, "%s\n", usg_hdr); 1524 1525 /* load directory or file */ 1526 if (stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) { 1527 dir = opendir(fn); 1528 if (!dir) { 1529 errs++; 1530 goto out; 1531 } 1532 while ((d = readdir(dir)) != NULL) { 1533 if (d->d_name[0] == '.') 1534 continue; 1535 if (asprintf(&mfn, "%s/%s", fn, d->d_name) < 0) { 1536 file_oomem(ms, 1537 strlen(fn) + strlen(d->d_name) + 2); 1538 errs++; 1539 closedir(dir); 1540 goto out; 1541 } 1542 if (stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) { 1543 free(mfn); 1544 continue; 1545 } 1546 if (files >= maxfiles) { 1547 size_t mlen; 1548 char **nfilearr; 1549 maxfiles = (maxfiles + 1) * 2; 1550 mlen = maxfiles * sizeof(*filearr); 1551 if ((nfilearr = CAST(char **, 1552 realloc(filearr, mlen))) == NULL) { 1553 file_oomem(ms, mlen); 1554 free(mfn); 1555 closedir(dir); 1556 errs++; 1557 goto out; 1558 } 1559 filearr = nfilearr; 1560 } 1561 filearr[files++] = mfn; 1562 } 1563 closedir(dir); 1564 if (filearr) { 1565 qsort(filearr, files, sizeof(*filearr), cmpstrp); 1566 for (i = 0; i < files; i++) { 1567 load_1(ms, action, filearr[i], &errs, mset); 1568 free(filearr[i]); 1569 } 1570 free(filearr); 1571 filearr = NULL; 1572 } 1573 } else 1574 load_1(ms, action, fn, &errs, mset); 1575 if (errs) 1576 goto out; 1577 1578 for (j = 0; j < MAGIC_SETS; j++) { 1579 /* Set types of tests */ 1580 for (i = 0; i < mset[j].count; ) { 1581 if (mset[j].me[i].mp->cont_level != 0) { 1582 i++; 1583 continue; 1584 } 1585 i = set_text_binary(ms, mset[j].me, mset[j].count, i); 1586 } 1587 if (mset[j].me) 1588 qsort(mset[j].me, mset[j].count, sizeof(*mset[j].me), 1589 apprentice_sort); 1590 1591 /* 1592 * Make sure that any level 0 "default" line is last 1593 * (if one exists). 1594 */ 1595 set_last_default(ms, mset[j].me, mset[j].count); 1596 1597 /* coalesce per file arrays into a single one, if needed */ 1598 if (mset[j].count == 0) 1599 continue; 1600 1601 if (coalesce_entries(ms, mset[j].me, mset[j].count, 1602 &map->magic[j], &map->nmagic[j]) == -1) { 1603 errs++; 1604 goto out; 1605 } 1606 } 1607 1608 out: 1609 free(filearr); 1610 for (j = 0; j < MAGIC_SETS; j++) 1611 magic_entry_free(mset[j].me, mset[j].count); 1612 1613 if (errs) { 1614 apprentice_unmap(map); 1615 return NULL; 1616 } 1617 return map; 1618 } 1619 1620 /* 1621 * extend the sign bit if the comparison is to be signed 1622 */ 1623 protected uint64_t 1624 file_signextend(struct magic_set *ms, struct magic *m, uint64_t v) 1625 { 1626 if (!(m->flag & UNSIGNED)) { 1627 switch(m->type) { 1628 /* 1629 * Do not remove the casts below. They are 1630 * vital. When later compared with the data, 1631 * the sign extension must have happened. 1632 */ 1633 case FILE_BYTE: 1634 v = CAST(signed char, v); 1635 break; 1636 case FILE_SHORT: 1637 case FILE_BESHORT: 1638 case FILE_LESHORT: 1639 v = CAST(short, v); 1640 break; 1641 case FILE_DATE: 1642 case FILE_BEDATE: 1643 case FILE_LEDATE: 1644 case FILE_MEDATE: 1645 case FILE_LDATE: 1646 case FILE_BELDATE: 1647 case FILE_LELDATE: 1648 case FILE_MELDATE: 1649 case FILE_LONG: 1650 case FILE_BELONG: 1651 case FILE_LELONG: 1652 case FILE_MELONG: 1653 case FILE_FLOAT: 1654 case FILE_BEFLOAT: 1655 case FILE_LEFLOAT: 1656 case FILE_MSDOSDATE: 1657 case FILE_BEMSDOSDATE: 1658 case FILE_LEMSDOSDATE: 1659 case FILE_MSDOSTIME: 1660 case FILE_BEMSDOSTIME: 1661 case FILE_LEMSDOSTIME: 1662 v = CAST(int32_t, v); 1663 break; 1664 case FILE_QUAD: 1665 case FILE_BEQUAD: 1666 case FILE_LEQUAD: 1667 case FILE_QDATE: 1668 case FILE_QLDATE: 1669 case FILE_QWDATE: 1670 case FILE_BEQDATE: 1671 case FILE_BEQLDATE: 1672 case FILE_BEQWDATE: 1673 case FILE_LEQDATE: 1674 case FILE_LEQLDATE: 1675 case FILE_LEQWDATE: 1676 case FILE_DOUBLE: 1677 case FILE_BEDOUBLE: 1678 case FILE_LEDOUBLE: 1679 case FILE_OFFSET: 1680 case FILE_BEVARINT: 1681 case FILE_LEVARINT: 1682 v = CAST(int64_t, v); 1683 break; 1684 case FILE_STRING: 1685 case FILE_PSTRING: 1686 case FILE_BESTRING16: 1687 case FILE_LESTRING16: 1688 case FILE_REGEX: 1689 case FILE_SEARCH: 1690 case FILE_DEFAULT: 1691 case FILE_INDIRECT: 1692 case FILE_NAME: 1693 case FILE_USE: 1694 case FILE_CLEAR: 1695 case FILE_DER: 1696 case FILE_GUID: 1697 break; 1698 default: 1699 if (ms->flags & MAGIC_CHECK) 1700 file_magwarn(ms, "cannot happen: m->type=%d\n", 1701 m->type); 1702 return FILE_BADSIZE; 1703 } 1704 } 1705 return v; 1706 } 1707 1708 private int 1709 string_modifier_check(struct magic_set *ms, struct magic *m) 1710 { 1711 if ((ms->flags & MAGIC_CHECK) == 0) 1712 return 0; 1713 1714 if ((m->type != FILE_REGEX || (m->str_flags & REGEX_LINE_COUNT) == 0) && 1715 (m->type != FILE_PSTRING && (m->str_flags & PSTRING_LEN) != 0)) { 1716 file_magwarn(ms, 1717 "'/BHhLl' modifiers are only allowed for pascal strings\n"); 1718 return -1; 1719 } 1720 switch (m->type) { 1721 case FILE_BESTRING16: 1722 case FILE_LESTRING16: 1723 if (m->str_flags != 0) { 1724 file_magwarn(ms, 1725 "no modifiers allowed for 16-bit strings\n"); 1726 return -1; 1727 } 1728 break; 1729 case FILE_STRING: 1730 case FILE_PSTRING: 1731 if ((m->str_flags & REGEX_OFFSET_START) != 0) { 1732 file_magwarn(ms, 1733 "'/%c' only allowed on regex and search\n", 1734 CHAR_REGEX_OFFSET_START); 1735 return -1; 1736 } 1737 break; 1738 case FILE_SEARCH: 1739 if (m->str_range == 0) { 1740 file_magwarn(ms, 1741 "missing range; defaulting to %d\n", 1742 STRING_DEFAULT_RANGE); 1743 m->str_range = STRING_DEFAULT_RANGE; 1744 return -1; 1745 } 1746 break; 1747 case FILE_REGEX: 1748 if ((m->str_flags & STRING_COMPACT_WHITESPACE) != 0) { 1749 file_magwarn(ms, "'/%c' not allowed on regex\n", 1750 CHAR_COMPACT_WHITESPACE); 1751 return -1; 1752 } 1753 if ((m->str_flags & STRING_COMPACT_OPTIONAL_WHITESPACE) != 0) { 1754 file_magwarn(ms, "'/%c' not allowed on regex\n", 1755 CHAR_COMPACT_OPTIONAL_WHITESPACE); 1756 return -1; 1757 } 1758 break; 1759 default: 1760 file_magwarn(ms, "coding error: m->type=%d\n", 1761 m->type); 1762 return -1; 1763 } 1764 return 0; 1765 } 1766 1767 private int 1768 get_op(char c) 1769 { 1770 switch (c) { 1771 case '&': 1772 return FILE_OPAND; 1773 case '|': 1774 return FILE_OPOR; 1775 case '^': 1776 return FILE_OPXOR; 1777 case '+': 1778 return FILE_OPADD; 1779 case '-': 1780 return FILE_OPMINUS; 1781 case '*': 1782 return FILE_OPMULTIPLY; 1783 case '/': 1784 return FILE_OPDIVIDE; 1785 case '%': 1786 return FILE_OPMODULO; 1787 default: 1788 return -1; 1789 } 1790 } 1791 1792 #ifdef ENABLE_CONDITIONALS 1793 private int 1794 get_cond(const char *l, const char **t) 1795 { 1796 static const struct cond_tbl_s { 1797 char name[8]; 1798 size_t len; 1799 int cond; 1800 } cond_tbl[] = { 1801 { "if", 2, COND_IF }, 1802 { "elif", 4, COND_ELIF }, 1803 { "else", 4, COND_ELSE }, 1804 { "", 0, COND_NONE }, 1805 }; 1806 const struct cond_tbl_s *p; 1807 1808 for (p = cond_tbl; p->len; p++) { 1809 if (strncmp(l, p->name, p->len) == 0 && 1810 isspace(CAST(unsigned char, l[p->len]))) { 1811 if (t) 1812 *t = l + p->len; 1813 break; 1814 } 1815 } 1816 return p->cond; 1817 } 1818 1819 private int 1820 check_cond(struct magic_set *ms, int cond, uint32_t cont_level) 1821 { 1822 int last_cond; 1823 last_cond = ms->c.li[cont_level].last_cond; 1824 1825 switch (cond) { 1826 case COND_IF: 1827 if (last_cond != COND_NONE && last_cond != COND_ELIF) { 1828 if (ms->flags & MAGIC_CHECK) 1829 file_magwarn(ms, "syntax error: `if'"); 1830 return -1; 1831 } 1832 last_cond = COND_IF; 1833 break; 1834 1835 case COND_ELIF: 1836 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1837 if (ms->flags & MAGIC_CHECK) 1838 file_magwarn(ms, "syntax error: `elif'"); 1839 return -1; 1840 } 1841 last_cond = COND_ELIF; 1842 break; 1843 1844 case COND_ELSE: 1845 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1846 if (ms->flags & MAGIC_CHECK) 1847 file_magwarn(ms, "syntax error: `else'"); 1848 return -1; 1849 } 1850 last_cond = COND_NONE; 1851 break; 1852 1853 case COND_NONE: 1854 last_cond = COND_NONE; 1855 break; 1856 } 1857 1858 ms->c.li[cont_level].last_cond = last_cond; 1859 return 0; 1860 } 1861 #endif /* ENABLE_CONDITIONALS */ 1862 1863 private int 1864 parse_indirect_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1865 { 1866 const char *l = *lp; 1867 1868 while (!isspace(CAST(unsigned char, *++l))) 1869 switch (*l) { 1870 case CHAR_INDIRECT_RELATIVE: 1871 m->str_flags |= INDIRECT_RELATIVE; 1872 break; 1873 default: 1874 if (ms->flags & MAGIC_CHECK) 1875 file_magwarn(ms, "indirect modifier `%c' " 1876 "invalid", *l); 1877 *lp = l; 1878 return -1; 1879 } 1880 *lp = l; 1881 return 0; 1882 } 1883 1884 private void 1885 parse_op_modifier(struct magic_set *ms, struct magic *m, const char **lp, 1886 int op) 1887 { 1888 const char *l = *lp; 1889 char *t; 1890 uint64_t val; 1891 1892 ++l; 1893 m->mask_op |= op; 1894 val = CAST(uint64_t, strtoull(l, &t, 0)); 1895 l = t; 1896 m->num_mask = file_signextend(ms, m, val); 1897 eatsize(&l); 1898 *lp = l; 1899 } 1900 1901 private int 1902 parse_string_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1903 { 1904 const char *l = *lp; 1905 char *t; 1906 int have_range = 0; 1907 1908 while (!isspace(CAST(unsigned char, *++l))) { 1909 switch (*l) { 1910 case '0': case '1': case '2': 1911 case '3': case '4': case '5': 1912 case '6': case '7': case '8': 1913 case '9': 1914 if (have_range && (ms->flags & MAGIC_CHECK)) 1915 file_magwarn(ms, "multiple ranges"); 1916 have_range = 1; 1917 m->str_range = CAST(uint32_t, strtoul(l, &t, 0)); 1918 if (m->str_range == 0) 1919 file_magwarn(ms, "zero range"); 1920 l = t - 1; 1921 break; 1922 case CHAR_COMPACT_WHITESPACE: 1923 m->str_flags |= STRING_COMPACT_WHITESPACE; 1924 break; 1925 case CHAR_COMPACT_OPTIONAL_WHITESPACE: 1926 m->str_flags |= STRING_COMPACT_OPTIONAL_WHITESPACE; 1927 break; 1928 case CHAR_IGNORE_LOWERCASE: 1929 m->str_flags |= STRING_IGNORE_LOWERCASE; 1930 break; 1931 case CHAR_IGNORE_UPPERCASE: 1932 m->str_flags |= STRING_IGNORE_UPPERCASE; 1933 break; 1934 case CHAR_REGEX_OFFSET_START: 1935 m->str_flags |= REGEX_OFFSET_START; 1936 break; 1937 case CHAR_BINTEST: 1938 m->str_flags |= STRING_BINTEST; 1939 break; 1940 case CHAR_TEXTTEST: 1941 m->str_flags |= STRING_TEXTTEST; 1942 break; 1943 case CHAR_TRIM: 1944 m->str_flags |= STRING_TRIM; 1945 break; 1946 case CHAR_FULL_WORD: 1947 m->str_flags |= STRING_FULL_WORD; 1948 break; 1949 case CHAR_PSTRING_1_LE: 1950 #define SET_LENGTH(a) m->str_flags = (m->str_flags & ~PSTRING_LEN) | (a) 1951 if (m->type != FILE_PSTRING) 1952 goto bad; 1953 SET_LENGTH(PSTRING_1_LE); 1954 break; 1955 case CHAR_PSTRING_2_BE: 1956 if (m->type != FILE_PSTRING) 1957 goto bad; 1958 SET_LENGTH(PSTRING_2_BE); 1959 break; 1960 case CHAR_PSTRING_2_LE: 1961 if (m->type != FILE_PSTRING) 1962 goto bad; 1963 SET_LENGTH(PSTRING_2_LE); 1964 break; 1965 case CHAR_PSTRING_4_BE: 1966 if (m->type != FILE_PSTRING) 1967 goto bad; 1968 SET_LENGTH(PSTRING_4_BE); 1969 break; 1970 case CHAR_PSTRING_4_LE: 1971 switch (m->type) { 1972 case FILE_PSTRING: 1973 case FILE_REGEX: 1974 break; 1975 default: 1976 goto bad; 1977 } 1978 SET_LENGTH(PSTRING_4_LE); 1979 break; 1980 case CHAR_PSTRING_LENGTH_INCLUDES_ITSELF: 1981 if (m->type != FILE_PSTRING) 1982 goto bad; 1983 m->str_flags |= PSTRING_LENGTH_INCLUDES_ITSELF; 1984 break; 1985 default: 1986 bad: 1987 if (ms->flags & MAGIC_CHECK) 1988 file_magwarn(ms, "string modifier `%c' " 1989 "invalid", *l); 1990 goto out; 1991 } 1992 /* allow multiple '/' for readability */ 1993 if (l[1] == '/' && !isspace(CAST(unsigned char, l[2]))) 1994 l++; 1995 } 1996 if (string_modifier_check(ms, m) == -1) 1997 goto out; 1998 *lp = l; 1999 return 0; 2000 out: 2001 *lp = l; 2002 return -1; 2003 } 2004 2005 /* 2006 * parse one line from magic file, put into magic[index++] if valid 2007 */ 2008 private int 2009 parse(struct magic_set *ms, struct magic_entry *me, const char *line, 2010 size_t lineno, int action) 2011 { 2012 #ifdef ENABLE_CONDITIONALS 2013 static uint32_t last_cont_level = 0; 2014 #endif 2015 size_t i; 2016 struct magic *m; 2017 const char *l = line; 2018 char *t; 2019 int op; 2020 uint32_t cont_level; 2021 int32_t diff; 2022 2023 cont_level = 0; 2024 2025 /* 2026 * Parse the offset. 2027 */ 2028 while (*l == '>') { 2029 ++l; /* step over */ 2030 cont_level++; 2031 } 2032 #ifdef ENABLE_CONDITIONALS 2033 if (cont_level == 0 || cont_level > last_cont_level) 2034 if (file_check_mem(ms, cont_level) == -1) 2035 return -1; 2036 last_cont_level = cont_level; 2037 #endif 2038 if (cont_level != 0) { 2039 if (me->mp == NULL) { 2040 file_magerror(ms, "No current entry for continuation"); 2041 return -1; 2042 } 2043 if (me->cont_count == 0) { 2044 file_magerror(ms, "Continuations present with 0 count"); 2045 return -1; 2046 } 2047 m = &me->mp[me->cont_count - 1]; 2048 diff = CAST(int32_t, cont_level) - CAST(int32_t, m->cont_level); 2049 if (diff > 1) 2050 file_magwarn(ms, "New continuation level %u is more " 2051 "than one larger than current level %u", cont_level, 2052 m->cont_level); 2053 if (me->cont_count == me->max_count) { 2054 struct magic *nm; 2055 size_t cnt = me->max_count + ALLOC_CHUNK; 2056 if ((nm = CAST(struct magic *, realloc(me->mp, 2057 sizeof(*nm) * cnt))) == NULL) { 2058 file_oomem(ms, sizeof(*nm) * cnt); 2059 return -1; 2060 } 2061 me->mp = nm; 2062 me->max_count = CAST(uint32_t, cnt); 2063 } 2064 m = &me->mp[me->cont_count++]; 2065 (void)memset(m, 0, sizeof(*m)); 2066 m->cont_level = cont_level; 2067 } else { 2068 static const size_t len = sizeof(*m) * ALLOC_CHUNK; 2069 if (me->mp != NULL) 2070 return 1; 2071 if ((m = CAST(struct magic *, malloc(len))) == NULL) { 2072 file_oomem(ms, len); 2073 return -1; 2074 } 2075 me->mp = m; 2076 me->max_count = ALLOC_CHUNK; 2077 (void)memset(m, 0, sizeof(*m)); 2078 m->factor_op = FILE_FACTOR_OP_NONE; 2079 m->cont_level = 0; 2080 me->cont_count = 1; 2081 } 2082 m->lineno = CAST(uint32_t, lineno); 2083 2084 if (*l == '&') { /* m->cont_level == 0 checked below. */ 2085 ++l; /* step over */ 2086 m->flag |= OFFADD; 2087 } 2088 if (*l == '(') { 2089 ++l; /* step over */ 2090 m->flag |= INDIR; 2091 if (m->flag & OFFADD) 2092 m->flag = (m->flag & ~OFFADD) | INDIROFFADD; 2093 2094 if (*l == '&') { /* m->cont_level == 0 checked below */ 2095 ++l; /* step over */ 2096 m->flag |= OFFADD; 2097 } 2098 } 2099 /* Indirect offsets are not valid at level 0. */ 2100 if (m->cont_level == 0 && (m->flag & (OFFADD | INDIROFFADD))) { 2101 if (ms->flags & MAGIC_CHECK) 2102 file_magwarn(ms, "relative offset at level 0"); 2103 return -1; 2104 } 2105 2106 /* get offset, then skip over it */ 2107 if (*l == '-') { 2108 ++l; /* step over */ 2109 m->flag |= OFFNEGATIVE; 2110 } 2111 m->offset = CAST(int32_t, strtol(l, &t, 0)); 2112 if (l == t) { 2113 if (ms->flags & MAGIC_CHECK) 2114 file_magwarn(ms, "offset `%s' invalid", l); 2115 return -1; 2116 } 2117 2118 l = t; 2119 2120 if (m->flag & INDIR) { 2121 m->in_type = FILE_LONG; 2122 m->in_offset = 0; 2123 m->in_op = 0; 2124 /* 2125 * read [.,lbs][+-]nnnnn) 2126 */ 2127 if (*l == '.' || *l == ',') { 2128 if (*l == ',') 2129 m->in_op |= FILE_OPSIGNED; 2130 l++; 2131 switch (*l) { 2132 case 'l': 2133 m->in_type = FILE_LELONG; 2134 break; 2135 case 'L': 2136 m->in_type = FILE_BELONG; 2137 break; 2138 case 'm': 2139 m->in_type = FILE_MELONG; 2140 break; 2141 case 'h': 2142 case 's': 2143 m->in_type = FILE_LESHORT; 2144 break; 2145 case 'H': 2146 case 'S': 2147 m->in_type = FILE_BESHORT; 2148 break; 2149 case 'c': 2150 case 'b': 2151 case 'C': 2152 case 'B': 2153 m->in_type = FILE_BYTE; 2154 break; 2155 case 'e': 2156 case 'f': 2157 case 'g': 2158 m->in_type = FILE_LEDOUBLE; 2159 break; 2160 case 'E': 2161 case 'F': 2162 case 'G': 2163 m->in_type = FILE_BEDOUBLE; 2164 break; 2165 case 'i': 2166 m->in_type = FILE_LEID3; 2167 break; 2168 case 'I': 2169 m->in_type = FILE_BEID3; 2170 break; 2171 case 'q': 2172 m->in_type = FILE_LEQUAD; 2173 break; 2174 case 'Q': 2175 m->in_type = FILE_BEQUAD; 2176 break; 2177 default: 2178 if (ms->flags & MAGIC_CHECK) 2179 file_magwarn(ms, 2180 "indirect offset type `%c' invalid", 2181 *l); 2182 return -1; 2183 } 2184 l++; 2185 } 2186 2187 if (*l == '~') { 2188 m->in_op |= FILE_OPINVERSE; 2189 l++; 2190 } 2191 if ((op = get_op(*l)) != -1) { 2192 m->in_op |= op; 2193 l++; 2194 } 2195 if (*l == '(') { 2196 m->in_op |= FILE_OPINDIRECT; 2197 l++; 2198 } 2199 if (isdigit(CAST(unsigned char, *l)) || *l == '-') { 2200 m->in_offset = CAST(int32_t, strtol(l, &t, 0)); 2201 if (l == t) { 2202 if (ms->flags & MAGIC_CHECK) 2203 file_magwarn(ms, 2204 "in_offset `%s' invalid", l); 2205 return -1; 2206 } 2207 l = t; 2208 } 2209 if (*l++ != ')' || 2210 ((m->in_op & FILE_OPINDIRECT) && *l++ != ')')) { 2211 if (ms->flags & MAGIC_CHECK) 2212 file_magwarn(ms, 2213 "missing ')' in indirect offset"); 2214 return -1; 2215 } 2216 } 2217 EATAB; 2218 2219 #ifdef ENABLE_CONDITIONALS 2220 m->cond = get_cond(l, &l); 2221 if (check_cond(ms, m->cond, cont_level) == -1) 2222 return -1; 2223 2224 EATAB; 2225 #endif 2226 2227 /* 2228 * Parse the type. 2229 */ 2230 if (*l == 'u') { 2231 /* 2232 * Try it as a keyword type prefixed by "u"; match what 2233 * follows the "u". If that fails, try it as an SUS 2234 * integer type. 2235 */ 2236 m->type = get_type(type_tbl, l + 1, &l); 2237 if (m->type == FILE_INVALID) { 2238 /* 2239 * Not a keyword type; parse it as an SUS type, 2240 * 'u' possibly followed by a number or C/S/L. 2241 */ 2242 m->type = get_standard_integer_type(l, &l); 2243 } 2244 /* It's unsigned. */ 2245 if (m->type != FILE_INVALID) 2246 m->flag |= UNSIGNED; 2247 } else { 2248 /* 2249 * Try it as a keyword type. If that fails, try it as 2250 * an SUS integer type if it begins with "d" or as an 2251 * SUS string type if it begins with "s". In any case, 2252 * it's not unsigned. 2253 */ 2254 m->type = get_type(type_tbl, l, &l); 2255 if (m->type == FILE_INVALID) { 2256 /* 2257 * Not a keyword type; parse it as an SUS type, 2258 * either 'd' possibly followed by a number or 2259 * C/S/L, or just 's'. 2260 */ 2261 if (*l == 'd') 2262 m->type = get_standard_integer_type(l, &l); 2263 else if (*l == 's' 2264 && !isalpha(CAST(unsigned char, l[1]))) { 2265 m->type = FILE_STRING; 2266 ++l; 2267 } 2268 } 2269 } 2270 2271 if (m->type == FILE_INVALID) { 2272 /* Not found - try it as a special keyword. */ 2273 m->type = get_type(special_tbl, l, &l); 2274 } 2275 2276 if (m->type == FILE_INVALID) { 2277 if (ms->flags & MAGIC_CHECK) 2278 file_magwarn(ms, "type `%s' invalid", l); 2279 return -1; 2280 } 2281 2282 if (m->type == FILE_NAME && cont_level != 0) { 2283 if (ms->flags & MAGIC_CHECK) 2284 file_magwarn(ms, "`name%s' entries can only be " 2285 "declared at top level", l); 2286 return -1; 2287 } 2288 2289 /* New-style anding: "0 byte&0x80 =0x80 dynamically linked" */ 2290 /* New and improved: ~ & | ^ + - * / % -- exciting, isn't it? */ 2291 2292 m->mask_op = 0; 2293 if (*l == '~') { 2294 if (!IS_STRING(m->type)) 2295 m->mask_op |= FILE_OPINVERSE; 2296 else if (ms->flags & MAGIC_CHECK) 2297 file_magwarn(ms, "'~' invalid for string types"); 2298 ++l; 2299 } 2300 m->str_range = 0; 2301 m->str_flags = m->type == FILE_PSTRING ? PSTRING_1_LE : 0; 2302 if ((op = get_op(*l)) != -1) { 2303 if (IS_STRING(m->type)) { 2304 int r; 2305 2306 if (op != FILE_OPDIVIDE) { 2307 if (ms->flags & MAGIC_CHECK) 2308 file_magwarn(ms, 2309 "invalid string/indirect op: " 2310 "`%c'", *t); 2311 return -1; 2312 } 2313 2314 if (m->type == FILE_INDIRECT) 2315 r = parse_indirect_modifier(ms, m, &l); 2316 else 2317 r = parse_string_modifier(ms, m, &l); 2318 if (r == -1) 2319 return -1; 2320 } else 2321 parse_op_modifier(ms, m, &l, op); 2322 } 2323 2324 /* 2325 * We used to set mask to all 1's here, instead let's just not do 2326 * anything if mask = 0 (unless you have a better idea) 2327 */ 2328 EATAB; 2329 2330 switch (*l) { 2331 case '>': 2332 case '<': 2333 m->reln = *l; 2334 ++l; 2335 if (*l == '=') { 2336 if (ms->flags & MAGIC_CHECK) { 2337 file_magwarn(ms, "%c= not supported", 2338 m->reln); 2339 return -1; 2340 } 2341 ++l; 2342 } 2343 break; 2344 /* Old-style anding: "0 byte &0x80 dynamically linked" */ 2345 case '&': 2346 case '^': 2347 case '=': 2348 m->reln = *l; 2349 ++l; 2350 if (*l == '=') { 2351 /* HP compat: ignore &= etc. */ 2352 ++l; 2353 } 2354 break; 2355 case '!': 2356 m->reln = *l; 2357 ++l; 2358 break; 2359 default: 2360 m->reln = '='; /* the default relation */ 2361 if (*l == 'x' && ((isascii(CAST(unsigned char, l[1])) && 2362 isspace(CAST(unsigned char, l[1]))) || !l[1])) { 2363 m->reln = *l; 2364 ++l; 2365 } 2366 break; 2367 } 2368 /* 2369 * Grab the value part, except for an 'x' reln. 2370 */ 2371 if (m->reln != 'x' && getvalue(ms, m, &l, action)) 2372 return -1; 2373 2374 /* 2375 * TODO finish this macro and start using it! 2376 * #define offsetcheck {if (offset > ms->bytes_max -1) 2377 * magwarn("offset too big"); } 2378 */ 2379 2380 /* 2381 * Now get last part - the description 2382 */ 2383 EATAB; 2384 if (l[0] == '\b') { 2385 ++l; 2386 m->flag |= NOSPACE; 2387 } else if ((l[0] == '\\') && (l[1] == 'b')) { 2388 ++l; 2389 ++l; 2390 m->flag |= NOSPACE; 2391 } 2392 for (i = 0; (m->desc[i++] = *l++) != '\0' && i < sizeof(m->desc); ) 2393 continue; 2394 if (i == sizeof(m->desc)) { 2395 m->desc[sizeof(m->desc) - 1] = '\0'; 2396 if (ms->flags & MAGIC_CHECK) 2397 file_magwarn(ms, "description `%s' truncated", m->desc); 2398 } 2399 2400 /* 2401 * We only do this check while compiling, or if any of the magic 2402 * files were not compiled. 2403 */ 2404 if (ms->flags & MAGIC_CHECK) { 2405 if (check_format(ms, m) == -1) 2406 return -1; 2407 } 2408 #ifndef COMPILE_ONLY 2409 if (action == FILE_CHECK) { 2410 file_mdump(m); 2411 } 2412 #endif 2413 m->mimetype[0] = '\0'; /* initialise MIME type to none */ 2414 return 0; 2415 } 2416 2417 /* 2418 * parse a STRENGTH annotation line from magic file, put into magic[index - 1] 2419 * if valid 2420 */ 2421 private int 2422 parse_strength(struct magic_set *ms, struct magic_entry *me, const char *line, 2423 size_t len __attribute__((__unused__))) 2424 { 2425 const char *l = line; 2426 char *el; 2427 unsigned long factor; 2428 struct magic *m = &me->mp[0]; 2429 2430 if (m->factor_op != FILE_FACTOR_OP_NONE) { 2431 file_magwarn(ms, 2432 "Current entry already has a strength type: %c %d", 2433 m->factor_op, m->factor); 2434 return -1; 2435 } 2436 if (m->type == FILE_NAME) { 2437 file_magwarn(ms, "%s: Strength setting is not supported in " 2438 "\"name\" magic entries", m->value.s); 2439 return -1; 2440 } 2441 EATAB; 2442 switch (*l) { 2443 case FILE_FACTOR_OP_NONE: 2444 case FILE_FACTOR_OP_PLUS: 2445 case FILE_FACTOR_OP_MINUS: 2446 case FILE_FACTOR_OP_TIMES: 2447 case FILE_FACTOR_OP_DIV: 2448 m->factor_op = *l++; 2449 break; 2450 default: 2451 file_magwarn(ms, "Unknown factor op `%c'", *l); 2452 return -1; 2453 } 2454 EATAB; 2455 factor = strtoul(l, &el, 0); 2456 if (factor > 255) { 2457 file_magwarn(ms, "Too large factor `%lu'", factor); 2458 goto out; 2459 } 2460 if (*el && !isspace(CAST(unsigned char, *el))) { 2461 file_magwarn(ms, "Bad factor `%s'", l); 2462 goto out; 2463 } 2464 m->factor = CAST(uint8_t, factor); 2465 if (m->factor == 0 && m->factor_op == FILE_FACTOR_OP_DIV) { 2466 file_magwarn(ms, "Cannot have factor op `%c' and factor %u", 2467 m->factor_op, m->factor); 2468 goto out; 2469 } 2470 return 0; 2471 out: 2472 m->factor_op = FILE_FACTOR_OP_NONE; 2473 m->factor = 0; 2474 return -1; 2475 } 2476 2477 private int 2478 goodchar(unsigned char x, const char *extra) 2479 { 2480 return (isascii(x) && isalnum(x)) || strchr(extra, x); 2481 } 2482 2483 private int 2484 parse_extra(struct magic_set *ms, struct magic_entry *me, const char *line, 2485 size_t llen, off_t off, size_t len, const char *name, const char *extra, 2486 int nt) 2487 { 2488 size_t i; 2489 const char *l = line; 2490 struct magic *m = &me->mp[me->cont_count == 0 ? 0 : me->cont_count - 1]; 2491 char *buf = CAST(char *, CAST(void *, m)) + off; 2492 2493 if (buf[0] != '\0') { 2494 len = nt ? strlen(buf) : len; 2495 file_magwarn(ms, "Current entry already has a %s type " 2496 "`%.*s', new type `%s'", name, CAST(int, len), buf, l); 2497 return -1; 2498 } 2499 2500 if (*m->desc == '\0') { 2501 file_magwarn(ms, "Current entry does not yet have a " 2502 "description for adding a %s type", name); 2503 return -1; 2504 } 2505 2506 EATAB; 2507 for (i = 0; *l && i < llen && i < len && goodchar(*l, extra); 2508 buf[i++] = *l++) 2509 continue; 2510 2511 if (i == len && *l) { 2512 if (nt) 2513 buf[len - 1] = '\0'; 2514 if (ms->flags & MAGIC_CHECK) 2515 file_magwarn(ms, "%s type `%s' truncated %" 2516 SIZE_T_FORMAT "u", name, line, i); 2517 } else { 2518 if (!isspace(CAST(unsigned char, *l)) && !goodchar(*l, extra)) 2519 file_magwarn(ms, "%s type `%s' has bad char '%c'", 2520 name, line, *l); 2521 if (nt) 2522 buf[i] = '\0'; 2523 } 2524 2525 if (i > 0) 2526 return 0; 2527 2528 file_magerror(ms, "Bad magic entry '%s'", line); 2529 return -1; 2530 } 2531 2532 /* 2533 * Parse an Apple CREATOR/TYPE annotation from magic file and put it into 2534 * magic[index - 1] 2535 */ 2536 private int 2537 parse_apple(struct magic_set *ms, struct magic_entry *me, const char *line, 2538 size_t len) 2539 { 2540 return parse_extra(ms, me, line, len, 2541 CAST(off_t, offsetof(struct magic, apple)), 2542 sizeof(me->mp[0].apple), "APPLE", "!+-./?", 0); 2543 } 2544 2545 /* 2546 * Parse a comma-separated list of extensions 2547 */ 2548 private int 2549 parse_ext(struct magic_set *ms, struct magic_entry *me, const char *line, 2550 size_t len) 2551 { 2552 return parse_extra(ms, me, line, len, 2553 CAST(off_t, offsetof(struct magic, ext)), 2554 sizeof(me->mp[0].ext), "EXTENSION", ",!+-/@?_$&", 0); /* & for b&w */ 2555 } 2556 2557 /* 2558 * parse a MIME annotation line from magic file, put into magic[index - 1] 2559 * if valid 2560 */ 2561 private int 2562 parse_mime(struct magic_set *ms, struct magic_entry *me, const char *line, 2563 size_t len) 2564 { 2565 return parse_extra(ms, me, line, len, 2566 CAST(off_t, offsetof(struct magic, mimetype)), 2567 sizeof(me->mp[0].mimetype), "MIME", "+-/.$?:{}", 1); 2568 } 2569 2570 private int 2571 check_format_type(const char *ptr, int type, const char **estr) 2572 { 2573 int quad = 0, h; 2574 size_t len, cnt; 2575 if (*ptr == '\0') { 2576 /* Missing format string; bad */ 2577 *estr = "missing format spec"; 2578 return -1; 2579 } 2580 2581 switch (file_formats[type]) { 2582 case FILE_FMT_QUAD: 2583 quad = 1; 2584 /*FALLTHROUGH*/ 2585 case FILE_FMT_NUM: 2586 if (quad == 0) { 2587 switch (type) { 2588 case FILE_BYTE: 2589 h = 2; 2590 break; 2591 case FILE_SHORT: 2592 case FILE_BESHORT: 2593 case FILE_LESHORT: 2594 h = 1; 2595 break; 2596 case FILE_LONG: 2597 case FILE_BELONG: 2598 case FILE_LELONG: 2599 case FILE_MELONG: 2600 case FILE_LEID3: 2601 case FILE_BEID3: 2602 case FILE_INDIRECT: 2603 h = 0; 2604 break; 2605 default: 2606 fprintf(stderr, "Bad number format %d", type); 2607 abort(); 2608 } 2609 } else 2610 h = 0; 2611 while (*ptr && strchr("-.#", *ptr) != NULL) 2612 ptr++; 2613 #define CHECKLEN() do { \ 2614 for (len = cnt = 0; isdigit(CAST(unsigned char, *ptr)); ptr++, cnt++) \ 2615 len = len * 10 + (*ptr - '0'); \ 2616 if (cnt > 5 || len > 1024) \ 2617 goto toolong; \ 2618 } while (/*CONSTCOND*/0) 2619 2620 CHECKLEN(); 2621 if (*ptr == '.') 2622 ptr++; 2623 CHECKLEN(); 2624 if (quad) { 2625 if (*ptr++ != 'l') 2626 goto invalid; 2627 if (*ptr++ != 'l') 2628 goto invalid; 2629 } 2630 2631 switch (*ptr++) { 2632 #ifdef STRICT_FORMAT /* "long" formats are int formats for us */ 2633 /* so don't accept the 'l' modifier */ 2634 case 'l': 2635 switch (*ptr++) { 2636 case 'i': 2637 case 'd': 2638 case 'u': 2639 case 'o': 2640 case 'x': 2641 case 'X': 2642 if (h == 0) 2643 return 0; 2644 /*FALLTHROUGH*/ 2645 default: 2646 goto invalid; 2647 } 2648 2649 /* 2650 * Don't accept h and hh modifiers. They make writing 2651 * magic entries more complicated, for very little benefit 2652 */ 2653 case 'h': 2654 if (h-- <= 0) 2655 goto invalid; 2656 switch (*ptr++) { 2657 case 'h': 2658 if (h-- <= 0) 2659 goto invalid; 2660 switch (*ptr++) { 2661 case 'i': 2662 case 'd': 2663 case 'u': 2664 case 'o': 2665 case 'x': 2666 case 'X': 2667 return 0; 2668 default: 2669 goto invalid; 2670 } 2671 case 'i': 2672 case 'd': 2673 case 'u': 2674 case 'o': 2675 case 'x': 2676 case 'X': 2677 if (h == 0) 2678 return 0; 2679 /*FALLTHROUGH*/ 2680 default: 2681 goto invalid; 2682 } 2683 #endif 2684 case 'c': 2685 if (h == 2) 2686 return 0; 2687 goto invalid; 2688 case 'i': 2689 case 'd': 2690 case 'u': 2691 case 'o': 2692 case 'x': 2693 case 'X': 2694 #ifdef STRICT_FORMAT 2695 if (h == 0) 2696 return 0; 2697 /*FALLTHROUGH*/ 2698 #else 2699 return 0; 2700 #endif 2701 default: 2702 goto invalid; 2703 } 2704 2705 case FILE_FMT_FLOAT: 2706 case FILE_FMT_DOUBLE: 2707 if (*ptr == '-') 2708 ptr++; 2709 if (*ptr == '.') 2710 ptr++; 2711 CHECKLEN(); 2712 if (*ptr == '.') 2713 ptr++; 2714 CHECKLEN(); 2715 switch (*ptr++) { 2716 case 'e': 2717 case 'E': 2718 case 'f': 2719 case 'F': 2720 case 'g': 2721 case 'G': 2722 return 0; 2723 2724 default: 2725 goto invalid; 2726 } 2727 2728 2729 case FILE_FMT_STR: 2730 if (*ptr == '-') 2731 ptr++; 2732 while (isdigit(CAST(unsigned char, *ptr))) 2733 ptr++; 2734 if (*ptr == '.') { 2735 ptr++; 2736 while (isdigit(CAST(unsigned char , *ptr))) 2737 ptr++; 2738 } 2739 2740 switch (*ptr++) { 2741 case 's': 2742 return 0; 2743 default: 2744 goto invalid; 2745 } 2746 2747 default: 2748 /* internal error */ 2749 fprintf(stderr, "Bad file format %d", type); 2750 abort(); 2751 } 2752 invalid: 2753 *estr = "not valid"; 2754 toolong: 2755 *estr = "too long"; 2756 return -1; 2757 } 2758 2759 /* 2760 * Check that the optional printf format in description matches 2761 * the type of the magic. 2762 */ 2763 private int 2764 check_format(struct magic_set *ms, struct magic *m) 2765 { 2766 char *ptr; 2767 const char *estr; 2768 2769 for (ptr = m->desc; *ptr; ptr++) 2770 if (*ptr == '%') 2771 break; 2772 if (*ptr == '\0') { 2773 /* No format string; ok */ 2774 return 1; 2775 } 2776 2777 assert(file_nformats == file_nnames); 2778 2779 if (m->type >= file_nformats) { 2780 file_magwarn(ms, "Internal error inconsistency between " 2781 "m->type and format strings"); 2782 return -1; 2783 } 2784 if (file_formats[m->type] == FILE_FMT_NONE) { 2785 file_magwarn(ms, "No format string for `%s' with description " 2786 "`%s'", m->desc, file_names[m->type]); 2787 return -1; 2788 } 2789 2790 ptr++; 2791 if (check_format_type(ptr, m->type, &estr) == -1) { 2792 /* 2793 * TODO: this error message is unhelpful if the format 2794 * string is not one character long 2795 */ 2796 file_magwarn(ms, "Printf format is %s for type " 2797 "`%s' in description `%s'", estr, 2798 file_names[m->type], m->desc); 2799 return -1; 2800 } 2801 2802 for (; *ptr; ptr++) { 2803 if (*ptr == '%') { 2804 file_magwarn(ms, 2805 "Too many format strings (should have at most one) " 2806 "for `%s' with description `%s'", 2807 file_names[m->type], m->desc); 2808 return -1; 2809 } 2810 } 2811 return 0; 2812 } 2813 2814 /* 2815 * Read a numeric value from a pointer, into the value union of a magic 2816 * pointer, according to the magic type. Update the string pointer to point 2817 * just after the number read. Return 0 for success, non-zero for failure. 2818 */ 2819 private int 2820 getvalue(struct magic_set *ms, struct magic *m, const char **p, int action) 2821 { 2822 char *ep; 2823 uint64_t ull; 2824 2825 switch (m->type) { 2826 case FILE_BESTRING16: 2827 case FILE_LESTRING16: 2828 case FILE_STRING: 2829 case FILE_PSTRING: 2830 case FILE_REGEX: 2831 case FILE_SEARCH: 2832 case FILE_NAME: 2833 case FILE_USE: 2834 case FILE_DER: 2835 *p = getstr(ms, m, *p, action == FILE_COMPILE); 2836 if (*p == NULL) { 2837 if (ms->flags & MAGIC_CHECK) 2838 file_magwarn(ms, "cannot get string from `%s'", 2839 m->value.s); 2840 return -1; 2841 } 2842 if (m->type == FILE_REGEX) { 2843 file_regex_t rx; 2844 int rc = file_regcomp(ms, &rx, m->value.s, 2845 REG_EXTENDED); 2846 if (rc == 0) { 2847 file_regfree(&rx); 2848 } 2849 return rc ? -1 : 0; 2850 } 2851 return 0; 2852 default: 2853 if (m->reln == 'x') 2854 return 0; 2855 break; 2856 } 2857 2858 switch (m->type) { 2859 case FILE_FLOAT: 2860 case FILE_BEFLOAT: 2861 case FILE_LEFLOAT: 2862 errno = 0; 2863 #ifdef HAVE_STRTOF 2864 m->value.f = strtof(*p, &ep); 2865 #else 2866 m->value.f = (float)strtod(*p, &ep); 2867 #endif 2868 if (errno == 0) 2869 *p = ep; 2870 return 0; 2871 case FILE_DOUBLE: 2872 case FILE_BEDOUBLE: 2873 case FILE_LEDOUBLE: 2874 errno = 0; 2875 m->value.d = strtod(*p, &ep); 2876 if (errno == 0) 2877 *p = ep; 2878 return 0; 2879 case FILE_GUID: 2880 if (file_parse_guid(*p, m->value.guid) == -1) 2881 return -1; 2882 *p += FILE_GUID_SIZE - 1; 2883 return 0; 2884 default: 2885 errno = 0; 2886 ull = CAST(uint64_t, strtoull(*p, &ep, 0)); 2887 m->value.q = file_signextend(ms, m, ull); 2888 if (*p == ep) { 2889 file_magwarn(ms, "Unparsable number `%s'", *p); 2890 } else { 2891 size_t ts = typesize(m->type); 2892 uint64_t x; 2893 const char *q; 2894 2895 if (ts == FILE_BADSIZE) { 2896 file_magwarn(ms, 2897 "Expected numeric type got `%s'", 2898 type_tbl[m->type].name); 2899 } 2900 for (q = *p; isspace(CAST(unsigned char, *q)); q++) 2901 continue; 2902 if (*q == '-') 2903 ull = -CAST(int64_t, ull); 2904 switch (ts) { 2905 case 1: 2906 x = CAST(uint64_t, ull & ~0xffULL); 2907 break; 2908 case 2: 2909 x = CAST(uint64_t, ull & ~0xffffULL); 2910 break; 2911 case 4: 2912 x = CAST(uint64_t, ull & ~0xffffffffULL); 2913 break; 2914 case 8: 2915 x = 0; 2916 break; 2917 default: 2918 fprintf(stderr, "Bad width %zu", ts); 2919 abort(); 2920 } 2921 if (x) { 2922 file_magwarn(ms, "Overflow for numeric" 2923 " type `%s' value %#" PRIx64, 2924 type_tbl[m->type].name, ull); 2925 } 2926 } 2927 if (errno == 0) { 2928 *p = ep; 2929 eatsize(p); 2930 } 2931 return 0; 2932 } 2933 } 2934 2935 /* 2936 * Convert a string containing C character escapes. Stop at an unescaped 2937 * space or tab. 2938 * Copy the converted version to "m->value.s", and the length in m->vallen. 2939 * Return updated scan pointer as function result. Warn if set. 2940 */ 2941 private const char * 2942 getstr(struct magic_set *ms, struct magic *m, const char *s, int warn) 2943 { 2944 const char *origs = s; 2945 char *p = m->value.s; 2946 size_t plen = sizeof(m->value.s); 2947 char *origp = p; 2948 char *pmax = p + plen - 1; 2949 int c; 2950 int val; 2951 size_t bracket_nesting = 0; 2952 2953 while ((c = *s++) != '\0') { 2954 if (isspace(CAST(unsigned char, c))) 2955 break; 2956 if (p >= pmax) { 2957 file_error(ms, 0, "string too long: `%s'", origs); 2958 return NULL; 2959 } 2960 if (c != '\\') { 2961 if (c == '[') { 2962 bracket_nesting++; 2963 } 2964 if (c == ']' && bracket_nesting > 0) { 2965 bracket_nesting--; 2966 } 2967 *p++ = CAST(char, c); 2968 continue; 2969 } 2970 switch(c = *s++) { 2971 2972 case '\0': 2973 if (warn) 2974 file_magwarn(ms, "incomplete escape"); 2975 s--; 2976 goto out; 2977 case '.': 2978 if (m->type == FILE_REGEX && 2979 bracket_nesting == 0 && warn) { 2980 file_magwarn(ms, "escaped dot ('.') found, " 2981 "use \\\\. instead"); 2982 } 2983 warn = 0; /* already did */ 2984 /*FALLTHROUGH*/ 2985 case '\t': 2986 if (warn) { 2987 file_magwarn(ms, 2988 "escaped tab found, use \\\\t instead"); 2989 warn = 0; /* already did */ 2990 } 2991 /*FALLTHROUGH*/ 2992 default: 2993 if (warn) { 2994 if (isprint(CAST(unsigned char, c))) { 2995 /* Allow escaping of 2996 * ``relations'' */ 2997 if (strchr("<>&^=!", c) == NULL 2998 && (m->type != FILE_REGEX || 2999 strchr("[]().*?^$|{}", c) 3000 == NULL)) { 3001 file_magwarn(ms, "no " 3002 "need to escape " 3003 "`%c'", c); 3004 } 3005 } else { 3006 file_magwarn(ms, 3007 "unknown escape sequence: " 3008 "\\%03o", c); 3009 } 3010 } 3011 /*FALLTHROUGH*/ 3012 /* space, perhaps force people to use \040? */ 3013 case ' ': 3014 #if 0 3015 /* 3016 * Other things people escape, but shouldn't need to, 3017 * so we disallow them 3018 */ 3019 case '\'': 3020 case '"': 3021 case '?': 3022 #endif 3023 /* Relations */ 3024 case '>': 3025 case '<': 3026 case '&': 3027 case '^': 3028 case '=': 3029 case '!': 3030 /* and backslash itself */ 3031 case '\\': 3032 *p++ = CAST(char, c); 3033 break; 3034 3035 case 'a': 3036 *p++ = '\a'; 3037 break; 3038 3039 case 'b': 3040 *p++ = '\b'; 3041 break; 3042 3043 case 'f': 3044 *p++ = '\f'; 3045 break; 3046 3047 case 'n': 3048 *p++ = '\n'; 3049 break; 3050 3051 case 'r': 3052 *p++ = '\r'; 3053 break; 3054 3055 case 't': 3056 *p++ = '\t'; 3057 break; 3058 3059 case 'v': 3060 *p++ = '\v'; 3061 break; 3062 3063 /* \ and up to 3 octal digits */ 3064 case '0': 3065 case '1': 3066 case '2': 3067 case '3': 3068 case '4': 3069 case '5': 3070 case '6': 3071 case '7': 3072 val = c - '0'; 3073 c = *s++; /* try for 2 */ 3074 if (c >= '0' && c <= '7') { 3075 val = (val << 3) | (c - '0'); 3076 c = *s++; /* try for 3 */ 3077 if (c >= '0' && c <= '7') 3078 val = (val << 3) | (c-'0'); 3079 else 3080 --s; 3081 } 3082 else 3083 --s; 3084 *p++ = CAST(char, val); 3085 break; 3086 3087 /* \x and up to 2 hex digits */ 3088 case 'x': 3089 val = 'x'; /* Default if no digits */ 3090 c = hextoint(*s++); /* Get next char */ 3091 if (c >= 0) { 3092 val = c; 3093 c = hextoint(*s++); 3094 if (c >= 0) 3095 val = (val << 4) + c; 3096 else 3097 --s; 3098 } else 3099 --s; 3100 *p++ = CAST(char, val); 3101 break; 3102 } 3103 } 3104 --s; 3105 out: 3106 *p = '\0'; 3107 m->vallen = CAST(unsigned char, (p - origp)); 3108 if (m->type == FILE_PSTRING) { 3109 size_t l = file_pstring_length_size(ms, m); 3110 if (l == FILE_BADSIZE) 3111 return NULL; 3112 m->vallen += CAST(unsigned char, l); 3113 } 3114 return s; 3115 } 3116 3117 3118 /* Single hex char to int; -1 if not a hex char. */ 3119 private int 3120 hextoint(int c) 3121 { 3122 if (!isascii(CAST(unsigned char, c))) 3123 return -1; 3124 if (isdigit(CAST(unsigned char, c))) 3125 return c - '0'; 3126 if ((c >= 'a') && (c <= 'f')) 3127 return c + 10 - 'a'; 3128 if (( c>= 'A') && (c <= 'F')) 3129 return c + 10 - 'A'; 3130 return -1; 3131 } 3132 3133 3134 /* 3135 * Print a string containing C character escapes. 3136 */ 3137 protected void 3138 file_showstr(FILE *fp, const char *s, size_t len) 3139 { 3140 char c; 3141 3142 for (;;) { 3143 if (len == FILE_BADSIZE) { 3144 c = *s++; 3145 if (c == '\0') 3146 break; 3147 } 3148 else { 3149 if (len-- == 0) 3150 break; 3151 c = *s++; 3152 } 3153 if (c >= 040 && c <= 0176) /* TODO isprint && !iscntrl */ 3154 (void) fputc(c, fp); 3155 else { 3156 (void) fputc('\\', fp); 3157 switch (c) { 3158 case '\a': 3159 (void) fputc('a', fp); 3160 break; 3161 3162 case '\b': 3163 (void) fputc('b', fp); 3164 break; 3165 3166 case '\f': 3167 (void) fputc('f', fp); 3168 break; 3169 3170 case '\n': 3171 (void) fputc('n', fp); 3172 break; 3173 3174 case '\r': 3175 (void) fputc('r', fp); 3176 break; 3177 3178 case '\t': 3179 (void) fputc('t', fp); 3180 break; 3181 3182 case '\v': 3183 (void) fputc('v', fp); 3184 break; 3185 3186 default: 3187 (void) fprintf(fp, "%.3o", c & 0377); 3188 break; 3189 } 3190 } 3191 } 3192 } 3193 3194 /* 3195 * eatsize(): Eat the size spec from a number [eg. 10UL] 3196 */ 3197 private void 3198 eatsize(const char **p) 3199 { 3200 const char *l = *p; 3201 3202 if (LOWCASE(*l) == 'u') 3203 l++; 3204 3205 switch (LOWCASE(*l)) { 3206 case 'l': /* long */ 3207 case 's': /* short */ 3208 case 'h': /* short */ 3209 case 'b': /* char/byte */ 3210 case 'c': /* char/byte */ 3211 l++; 3212 /*FALLTHROUGH*/ 3213 default: 3214 break; 3215 } 3216 3217 *p = l; 3218 } 3219 3220 /* 3221 * handle a buffer containing a compiled file. 3222 */ 3223 private struct magic_map * 3224 apprentice_buf(struct magic_set *ms, struct magic *buf, size_t len) 3225 { 3226 struct magic_map *map; 3227 3228 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 3229 file_oomem(ms, sizeof(*map)); 3230 return NULL; 3231 } 3232 map->len = len; 3233 map->p = buf; 3234 map->type = MAP_TYPE_USER; 3235 if (check_buffer(ms, map, "buffer") != 0) { 3236 apprentice_unmap(map); 3237 return NULL; 3238 } 3239 return map; 3240 } 3241 3242 /* 3243 * handle a compiled file. 3244 */ 3245 3246 private struct magic_map * 3247 apprentice_map(struct magic_set *ms, const char *fn) 3248 { 3249 int fd; 3250 struct stat st; 3251 char *dbname = NULL; 3252 struct magic_map *map; 3253 struct magic_map *rv = NULL; 3254 3255 fd = -1; 3256 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 3257 file_oomem(ms, sizeof(*map)); 3258 goto error; 3259 } 3260 map->type = MAP_TYPE_USER; /* unspecified */ 3261 3262 dbname = mkdbname(ms, fn, 0); 3263 if (dbname == NULL) 3264 goto error; 3265 3266 if ((fd = open(dbname, O_RDONLY|O_BINARY)) == -1) 3267 goto error; 3268 3269 if (fstat(fd, &st) == -1) { 3270 file_error(ms, errno, "cannot stat `%s'", dbname); 3271 goto error; 3272 } 3273 if (st.st_size < 8 || st.st_size > maxoff_t()) { 3274 file_error(ms, 0, "file `%s' is too %s", dbname, 3275 st.st_size < 8 ? "small" : "large"); 3276 goto error; 3277 } 3278 3279 map->len = CAST(size_t, st.st_size); 3280 #ifdef QUICK 3281 map->type = MAP_TYPE_MMAP; 3282 if ((map->p = mmap(0, CAST(size_t, st.st_size), PROT_READ|PROT_WRITE, 3283 MAP_PRIVATE|MAP_FILE, fd, CAST(off_t, 0))) == MAP_FAILED) { 3284 file_error(ms, errno, "cannot map `%s'", dbname); 3285 goto error; 3286 } 3287 #else 3288 map->type = MAP_TYPE_MALLOC; 3289 if ((map->p = CAST(void *, malloc(map->len))) == NULL) { 3290 file_oomem(ms, map->len); 3291 goto error; 3292 } 3293 if (read(fd, map->p, map->len) != (ssize_t)map->len) { 3294 file_badread(ms); 3295 goto error; 3296 } 3297 #endif 3298 (void)close(fd); 3299 fd = -1; 3300 3301 if (check_buffer(ms, map, dbname) != 0) { 3302 goto error; 3303 } 3304 #ifdef QUICK 3305 if (mprotect(map->p, CAST(size_t, st.st_size), PROT_READ) == -1) { 3306 file_error(ms, errno, "cannot mprotect `%s'", dbname); 3307 goto error; 3308 } 3309 #endif 3310 3311 free(dbname); 3312 return map; 3313 3314 error: 3315 if (fd != -1) 3316 (void)close(fd); 3317 apprentice_unmap(map); 3318 free(dbname); 3319 return rv; 3320 } 3321 3322 private int 3323 check_buffer(struct magic_set *ms, struct magic_map *map, const char *dbname) 3324 { 3325 uint32_t *ptr; 3326 uint32_t entries, nentries; 3327 uint32_t version; 3328 int i, needsbyteswap; 3329 3330 ptr = CAST(uint32_t *, map->p); 3331 if (*ptr != MAGICNO) { 3332 if (swap4(*ptr) != MAGICNO) { 3333 file_error(ms, 0, "bad magic in `%s'", dbname); 3334 return -1; 3335 } 3336 needsbyteswap = 1; 3337 } else 3338 needsbyteswap = 0; 3339 if (needsbyteswap) 3340 version = swap4(ptr[1]); 3341 else 3342 version = ptr[1]; 3343 if (version != VERSIONNO) { 3344 file_error(ms, 0, "File %s supports only version %d magic " 3345 "files. `%s' is version %d", VERSION, 3346 VERSIONNO, dbname, version); 3347 return -1; 3348 } 3349 entries = CAST(uint32_t, map->len / sizeof(struct magic)); 3350 if ((entries * sizeof(struct magic)) != map->len) { 3351 file_error(ms, 0, "Size of `%s' %" SIZE_T_FORMAT "u is not " 3352 "a multiple of %" SIZE_T_FORMAT "u", 3353 dbname, map->len, sizeof(struct magic)); 3354 return -1; 3355 } 3356 map->magic[0] = CAST(struct magic *, map->p) + 1; 3357 nentries = 0; 3358 for (i = 0; i < MAGIC_SETS; i++) { 3359 if (needsbyteswap) 3360 map->nmagic[i] = swap4(ptr[i + 2]); 3361 else 3362 map->nmagic[i] = ptr[i + 2]; 3363 if (i != MAGIC_SETS - 1) 3364 map->magic[i + 1] = map->magic[i] + map->nmagic[i]; 3365 nentries += map->nmagic[i]; 3366 } 3367 if (entries != nentries + 1) { 3368 file_error(ms, 0, "Inconsistent entries in `%s' %u != %u", 3369 dbname, entries, nentries + 1); 3370 return -1; 3371 } 3372 if (needsbyteswap) 3373 for (i = 0; i < MAGIC_SETS; i++) 3374 byteswap(map->magic[i], map->nmagic[i]); 3375 return 0; 3376 } 3377 3378 /* 3379 * handle an mmaped file. 3380 */ 3381 private int 3382 apprentice_compile(struct magic_set *ms, struct magic_map *map, const char *fn) 3383 { 3384 static const size_t nm = sizeof(*map->nmagic) * MAGIC_SETS; 3385 static const size_t m = sizeof(**map->magic); 3386 int fd = -1; 3387 size_t len; 3388 char *dbname; 3389 int rv = -1; 3390 uint32_t i; 3391 union { 3392 struct magic m; 3393 uint32_t h[2 + MAGIC_SETS]; 3394 } hdr; 3395 3396 dbname = mkdbname(ms, fn, 1); 3397 3398 if (dbname == NULL) 3399 goto out; 3400 3401 if ((fd = open(dbname, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY, 0644)) == -1) 3402 { 3403 file_error(ms, errno, "cannot open `%s'", dbname); 3404 goto out; 3405 } 3406 memset(&hdr, 0, sizeof(hdr)); 3407 hdr.h[0] = MAGICNO; 3408 hdr.h[1] = VERSIONNO; 3409 memcpy(hdr.h + 2, map->nmagic, nm); 3410 3411 if (write(fd, &hdr, sizeof(hdr)) != CAST(ssize_t, sizeof(hdr))) { 3412 file_error(ms, errno, "error writing `%s'", dbname); 3413 goto out2; 3414 } 3415 3416 for (i = 0; i < MAGIC_SETS; i++) { 3417 len = m * map->nmagic[i]; 3418 if (write(fd, map->magic[i], len) != CAST(ssize_t, len)) { 3419 file_error(ms, errno, "error writing `%s'", dbname); 3420 goto out2; 3421 } 3422 } 3423 3424 rv = 0; 3425 out2: 3426 if (fd != -1) 3427 (void)close(fd); 3428 out: 3429 apprentice_unmap(map); 3430 free(dbname); 3431 return rv; 3432 } 3433 3434 private const char ext[] = ".mgc"; 3435 /* 3436 * make a dbname 3437 */ 3438 private char * 3439 mkdbname(struct magic_set *ms, const char *fn, int strip) 3440 { 3441 const char *p, *q; 3442 char *buf; 3443 3444 if (strip) { 3445 if ((p = strrchr(fn, '/')) != NULL) 3446 fn = ++p; 3447 } 3448 3449 for (q = fn; *q; q++) 3450 continue; 3451 /* Look for .mgc */ 3452 for (p = ext + sizeof(ext) - 1; p >= ext && q >= fn; p--, q--) 3453 if (*p != *q) 3454 break; 3455 3456 /* Did not find .mgc, restore q */ 3457 if (p >= ext) 3458 while (*q) 3459 q++; 3460 3461 q++; 3462 /* Compatibility with old code that looked in .mime */ 3463 if (ms->flags & MAGIC_MIME) { 3464 if (asprintf(&buf, "%.*s.mime%s", CAST(int, q - fn), fn, ext) 3465 < 0) 3466 return NULL; 3467 if (access(buf, R_OK) != -1) { 3468 ms->flags &= MAGIC_MIME_TYPE; 3469 return buf; 3470 } 3471 free(buf); 3472 } 3473 if (asprintf(&buf, "%.*s%s", CAST(int, q - fn), fn, ext) < 0) 3474 return NULL; 3475 3476 /* Compatibility with old code that looked in .mime */ 3477 if (strstr(fn, ".mime") != NULL) 3478 ms->flags &= MAGIC_MIME_TYPE; 3479 return buf; 3480 } 3481 3482 /* 3483 * Byteswap an mmap'ed file if needed 3484 */ 3485 private void 3486 byteswap(struct magic *magic, uint32_t nmagic) 3487 { 3488 uint32_t i; 3489 for (i = 0; i < nmagic; i++) 3490 bs1(&magic[i]); 3491 } 3492 3493 #if !defined(HAVE_BYTESWAP_H) && !defined(HAVE_SYS_BSWAP_H) 3494 /* 3495 * swap a short 3496 */ 3497 private uint16_t 3498 swap2(uint16_t sv) 3499 { 3500 uint16_t rv; 3501 uint8_t *s = RCAST(uint8_t *, RCAST(void *, &sv)); 3502 uint8_t *d = RCAST(uint8_t *, RCAST(void *, &rv)); 3503 d[0] = s[1]; 3504 d[1] = s[0]; 3505 return rv; 3506 } 3507 3508 /* 3509 * swap an int 3510 */ 3511 private uint32_t 3512 swap4(uint32_t sv) 3513 { 3514 uint32_t rv; 3515 uint8_t *s = RCAST(uint8_t *, RCAST(void *, &sv)); 3516 uint8_t *d = RCAST(uint8_t *, RCAST(void *, &rv)); 3517 d[0] = s[3]; 3518 d[1] = s[2]; 3519 d[2] = s[1]; 3520 d[3] = s[0]; 3521 return rv; 3522 } 3523 3524 /* 3525 * swap a quad 3526 */ 3527 private uint64_t 3528 swap8(uint64_t sv) 3529 { 3530 uint64_t rv; 3531 uint8_t *s = RCAST(uint8_t *, RCAST(void *, &sv)); 3532 uint8_t *d = RCAST(uint8_t *, RCAST(void *, &rv)); 3533 # if 0 3534 d[0] = s[3]; 3535 d[1] = s[2]; 3536 d[2] = s[1]; 3537 d[3] = s[0]; 3538 d[4] = s[7]; 3539 d[5] = s[6]; 3540 d[6] = s[5]; 3541 d[7] = s[4]; 3542 # else 3543 d[0] = s[7]; 3544 d[1] = s[6]; 3545 d[2] = s[5]; 3546 d[3] = s[4]; 3547 d[4] = s[3]; 3548 d[5] = s[2]; 3549 d[6] = s[1]; 3550 d[7] = s[0]; 3551 # endif 3552 return rv; 3553 } 3554 #endif 3555 3556 protected uintmax_t 3557 file_varint2uintmax_t(const unsigned char *us, int t, size_t *l) 3558 { 3559 uintmax_t x = 0; 3560 const unsigned char *c; 3561 if (t == FILE_LEVARINT) { 3562 for (c = us; *c; c++) { 3563 if ((*c & 0x80) == 0) 3564 break; 3565 } 3566 if (l) 3567 *l = c - us + 1; 3568 for (; c >= us; c--) { 3569 x |= *c & 0x7f; 3570 x <<= 7; 3571 } 3572 } else { 3573 for (c = us; *c; c++) { 3574 x |= *c & 0x7f; 3575 if ((*c & 0x80) == 0) 3576 break; 3577 x <<= 7; 3578 } 3579 if (l) 3580 *l = c - us + 1; 3581 } 3582 return x; 3583 } 3584 3585 3586 /* 3587 * byteswap a single magic entry 3588 */ 3589 private void 3590 bs1(struct magic *m) 3591 { 3592 m->cont_level = swap2(m->cont_level); 3593 m->offset = swap4(CAST(uint32_t, m->offset)); 3594 m->in_offset = swap4(CAST(uint32_t, m->in_offset)); 3595 m->lineno = swap4(CAST(uint32_t, m->lineno)); 3596 if (IS_STRING(m->type)) { 3597 m->str_range = swap4(m->str_range); 3598 m->str_flags = swap4(m->str_flags); 3599 } 3600 else { 3601 m->value.q = swap8(m->value.q); 3602 m->num_mask = swap8(m->num_mask); 3603 } 3604 } 3605 3606 protected size_t 3607 file_pstring_length_size(struct magic_set *ms, const struct magic *m) 3608 { 3609 switch (m->str_flags & PSTRING_LEN) { 3610 case PSTRING_1_LE: 3611 return 1; 3612 case PSTRING_2_LE: 3613 case PSTRING_2_BE: 3614 return 2; 3615 case PSTRING_4_LE: 3616 case PSTRING_4_BE: 3617 return 4; 3618 default: 3619 file_error(ms, 0, "corrupt magic file " 3620 "(bad pascal string length %d)", 3621 m->str_flags & PSTRING_LEN); 3622 return FILE_BADSIZE; 3623 } 3624 } 3625 protected size_t 3626 file_pstring_get_length(struct magic_set *ms, const struct magic *m, 3627 const char *ss) 3628 { 3629 size_t len = 0; 3630 const unsigned char *s = RCAST(const unsigned char *, ss); 3631 unsigned int s3, s2, s1, s0; 3632 3633 switch (m->str_flags & PSTRING_LEN) { 3634 case PSTRING_1_LE: 3635 len = *s; 3636 break; 3637 case PSTRING_2_LE: 3638 s0 = s[0]; 3639 s1 = s[1]; 3640 len = (s1 << 8) | s0; 3641 break; 3642 case PSTRING_2_BE: 3643 s0 = s[0]; 3644 s1 = s[1]; 3645 len = (s0 << 8) | s1; 3646 break; 3647 case PSTRING_4_LE: 3648 s0 = s[0]; 3649 s1 = s[1]; 3650 s2 = s[2]; 3651 s3 = s[3]; 3652 len = (s3 << 24) | (s2 << 16) | (s1 << 8) | s0; 3653 break; 3654 case PSTRING_4_BE: 3655 s0 = s[0]; 3656 s1 = s[1]; 3657 s2 = s[2]; 3658 s3 = s[3]; 3659 len = (s0 << 24) | (s1 << 16) | (s2 << 8) | s3; 3660 break; 3661 default: 3662 file_error(ms, 0, "corrupt magic file " 3663 "(bad pascal string length %d)", 3664 m->str_flags & PSTRING_LEN); 3665 return FILE_BADSIZE; 3666 } 3667 3668 if (m->str_flags & PSTRING_LENGTH_INCLUDES_ITSELF) { 3669 size_t l = file_pstring_length_size(ms, m); 3670 if (l == FILE_BADSIZE) 3671 return l; 3672 len -= l; 3673 } 3674 3675 return len; 3676 } 3677 3678 protected int 3679 file_magicfind(struct magic_set *ms, const char *name, struct mlist *v) 3680 { 3681 uint32_t i, j; 3682 struct mlist *mlist, *ml; 3683 3684 mlist = ms->mlist[1]; 3685 3686 for (ml = mlist->next; ml != mlist; ml = ml->next) { 3687 struct magic *ma = ml->magic; 3688 for (i = 0; i < ml->nmagic; i++) { 3689 if (ma[i].type != FILE_NAME) 3690 continue; 3691 if (strcmp(ma[i].value.s, name) == 0) { 3692 v->magic = &ma[i]; 3693 v->magic_rxcomp = &(ml->magic_rxcomp[i]); 3694 for (j = i + 1; j < ml->nmagic; j++) 3695 if (ma[j].cont_level == 0) 3696 break; 3697 v->nmagic = j - i; 3698 return 0; 3699 } 3700 } 3701 } 3702 return -1; 3703 } 3704