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 35FILE_RCSID("@(#)$File: apprentice.c,v 1.342 2023/07/17 14:38:35 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 90struct magic_entry { 91 struct magic *mp; 92 uint32_t cont_count; 93 uint32_t max_count; 94}; 95 96struct magic_entry_set { 97 struct magic_entry *me; 98 uint32_t count; 99 uint32_t max; 100}; 101 102struct 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 110int file_formats[FILE_NAMES_SIZE]; 111const size_t file_nformats = FILE_NAMES_SIZE; 112const char *file_names[FILE_NAMES_SIZE]; 113const size_t file_nnames = FILE_NAMES_SIZE; 114 115file_private int getvalue(struct magic_set *ms, struct magic *, const char **, int); 116file_private int hextoint(int); 117file_private const char *getstr(struct magic_set *, struct magic *, const char *, 118 int); 119file_private int parse(struct magic_set *, struct magic_entry *, const char *, 120 size_t, int); 121file_private void eatsize(const char **); 122file_private int apprentice_1(struct magic_set *, const char *, int); 123file_private ssize_t apprentice_magic_strength_1(const struct magic *); 124file_private int apprentice_sort(const void *, const void *); 125file_private void apprentice_list(struct mlist *, int ); 126file_private struct magic_map *apprentice_load(struct magic_set *, 127 const char *, int); 128file_private struct mlist *mlist_alloc(void); 129file_private void mlist_free_all(struct magic_set *); 130file_private void mlist_free(struct mlist *); 131file_private void byteswap(struct magic *, uint32_t); 132file_private void bs1(struct magic *); 133 134#if defined(HAVE_BYTESWAP_H) 135#define swap2(x) bswap_16(x) 136#define swap4(x) bswap_32(x) 137#define swap8(x) bswap_64(x) 138#elif defined(HAVE_SYS_BSWAP_H) 139#define swap2(x) bswap16(x) 140#define swap4(x) bswap32(x) 141#define swap8(x) bswap64(x) 142#else 143file_private uint16_t swap2(uint16_t); 144file_private uint32_t swap4(uint32_t); 145file_private uint64_t swap8(uint64_t); 146#endif 147 148file_private char *mkdbname(struct magic_set *, const char *, int); 149file_private struct magic_map *apprentice_buf(struct magic_set *, struct magic *, 150 size_t); 151file_private struct magic_map *apprentice_map(struct magic_set *, const char *); 152file_private int check_buffer(struct magic_set *, struct magic_map *, const char *); 153file_private void apprentice_unmap(struct magic_map *); 154file_private int apprentice_compile(struct magic_set *, struct magic_map *, 155 const char *); 156file_private int check_format_type(const char *, int, const char **); 157file_private int check_format(struct magic_set *, struct magic *); 158file_private int get_op(char); 159file_private int parse_mime(struct magic_set *, struct magic_entry *, const char *, 160 size_t); 161file_private int parse_strength(struct magic_set *, struct magic_entry *, 162 const char *, size_t); 163file_private int parse_apple(struct magic_set *, struct magic_entry *, const char *, 164 size_t); 165file_private int parse_ext(struct magic_set *, struct magic_entry *, const char *, 166 size_t); 167 168 169file_private size_t magicsize = sizeof(struct magic); 170 171file_private const char usg_hdr[] = "cont\toffset\ttype\topcode\tmask\tvalue\tdesc"; 172 173file_private struct { 174 const char *name; 175 size_t len; 176 int (*fun)(struct magic_set *, struct magic_entry *, const char *, 177 size_t); 178} bang[] = { 179#define DECLARE_FIELD(name) { # name, sizeof(# name) - 1, parse_ ## name } 180 DECLARE_FIELD(mime), 181 DECLARE_FIELD(apple), 182 DECLARE_FIELD(ext), 183 DECLARE_FIELD(strength), 184#undef DECLARE_FIELD 185 { NULL, 0, NULL } 186}; 187 188#ifdef COMPILE_ONLY 189 190int main(int, char *[]); 191 192int 193main(int argc, char *argv[]) 194{ 195 int ret; 196 struct magic_set *ms; 197 char *progname; 198 199 if ((progname = strrchr(argv[0], '/')) != NULL) 200 progname++; 201 else 202 progname = argv[0]; 203 204 if (argc != 2) { 205 (void)fprintf(stderr, "Usage: %s file\n", progname); 206 return 1; 207 } 208 209 if ((ms = magic_open(MAGIC_CHECK)) == NULL) { 210 (void)fprintf(stderr, "%s: %s\n", progname, strerror(errno)); 211 return 1; 212 } 213 ret = magic_compile(ms, argv[1]) == -1 ? 1 : 0; 214 if (ret == 1) 215 (void)fprintf(stderr, "%s: %s\n", progname, magic_error(ms)); 216 magic_close(ms); 217 return ret; 218} 219#endif /* COMPILE_ONLY */ 220 221struct type_tbl_s { 222 const char name[16]; 223 const size_t len; 224 const int type; 225 const int format; 226}; 227 228/* 229 * XXX - the actual Single UNIX Specification says that "long" means "long", 230 * as in the C data type, but we treat it as meaning "4-byte integer". 231 * Given that the OS X version of file 5.04 did the same, I guess that passes 232 * the actual test; having "long" be dependent on how big a "long" is on 233 * the machine running "file" is silly. 234 */ 235static const struct type_tbl_s type_tbl[] = { 236# define XX(s) s, (sizeof(s) - 1) 237# define XX_NULL "", 0 238 { XX("invalid"), FILE_INVALID, FILE_FMT_NONE }, 239 { XX("byte"), FILE_BYTE, FILE_FMT_NUM }, 240 { XX("short"), FILE_SHORT, FILE_FMT_NUM }, 241 { XX("default"), FILE_DEFAULT, FILE_FMT_NONE }, 242 { XX("long"), FILE_LONG, FILE_FMT_NUM }, 243 { XX("string"), FILE_STRING, FILE_FMT_STR }, 244 { XX("date"), FILE_DATE, FILE_FMT_STR }, 245 { XX("beshort"), FILE_BESHORT, FILE_FMT_NUM }, 246 { XX("belong"), FILE_BELONG, FILE_FMT_NUM }, 247 { XX("bedate"), FILE_BEDATE, FILE_FMT_STR }, 248 { XX("leshort"), FILE_LESHORT, FILE_FMT_NUM }, 249 { XX("lelong"), FILE_LELONG, FILE_FMT_NUM }, 250 { XX("ledate"), FILE_LEDATE, FILE_FMT_STR }, 251 { XX("pstring"), FILE_PSTRING, FILE_FMT_STR }, 252 { XX("ldate"), FILE_LDATE, FILE_FMT_STR }, 253 { XX("beldate"), FILE_BELDATE, FILE_FMT_STR }, 254 { XX("leldate"), FILE_LELDATE, FILE_FMT_STR }, 255 { XX("regex"), FILE_REGEX, FILE_FMT_STR }, 256 { XX("bestring16"), FILE_BESTRING16, FILE_FMT_STR }, 257 { XX("lestring16"), FILE_LESTRING16, FILE_FMT_STR }, 258 { XX("search"), FILE_SEARCH, FILE_FMT_STR }, 259 { XX("medate"), FILE_MEDATE, FILE_FMT_STR }, 260 { XX("meldate"), FILE_MELDATE, FILE_FMT_STR }, 261 { XX("melong"), FILE_MELONG, FILE_FMT_NUM }, 262 { XX("quad"), FILE_QUAD, FILE_FMT_QUAD }, 263 { XX("lequad"), FILE_LEQUAD, FILE_FMT_QUAD }, 264 { XX("bequad"), FILE_BEQUAD, FILE_FMT_QUAD }, 265 { XX("qdate"), FILE_QDATE, FILE_FMT_STR }, 266 { XX("leqdate"), FILE_LEQDATE, FILE_FMT_STR }, 267 { XX("beqdate"), FILE_BEQDATE, FILE_FMT_STR }, 268 { XX("qldate"), FILE_QLDATE, FILE_FMT_STR }, 269 { XX("leqldate"), FILE_LEQLDATE, FILE_FMT_STR }, 270 { XX("beqldate"), FILE_BEQLDATE, FILE_FMT_STR }, 271 { XX("float"), FILE_FLOAT, FILE_FMT_FLOAT }, 272 { XX("befloat"), FILE_BEFLOAT, FILE_FMT_FLOAT }, 273 { XX("lefloat"), FILE_LEFLOAT, FILE_FMT_FLOAT }, 274 { XX("double"), FILE_DOUBLE, FILE_FMT_DOUBLE }, 275 { XX("bedouble"), FILE_BEDOUBLE, FILE_FMT_DOUBLE }, 276 { XX("ledouble"), FILE_LEDOUBLE, FILE_FMT_DOUBLE }, 277 { XX("leid3"), FILE_LEID3, FILE_FMT_NUM }, 278 { XX("beid3"), FILE_BEID3, FILE_FMT_NUM }, 279 { XX("indirect"), FILE_INDIRECT, FILE_FMT_NUM }, 280 { XX("qwdate"), FILE_QWDATE, FILE_FMT_STR }, 281 { XX("leqwdate"), FILE_LEQWDATE, FILE_FMT_STR }, 282 { XX("beqwdate"), FILE_BEQWDATE, FILE_FMT_STR }, 283 { XX("name"), FILE_NAME, FILE_FMT_NONE }, 284 { XX("use"), FILE_USE, FILE_FMT_NONE }, 285 { XX("clear"), FILE_CLEAR, FILE_FMT_NONE }, 286 { XX("der"), FILE_DER, FILE_FMT_STR }, 287 { XX("guid"), FILE_GUID, FILE_FMT_STR }, 288 { XX("offset"), FILE_OFFSET, FILE_FMT_QUAD }, 289 { XX("bevarint"), FILE_BEVARINT, FILE_FMT_STR }, 290 { XX("levarint"), FILE_LEVARINT, FILE_FMT_STR }, 291 { XX("msdosdate"), FILE_MSDOSDATE, FILE_FMT_STR }, 292 { XX("lemsdosdate"), FILE_LEMSDOSDATE, FILE_FMT_STR }, 293 { XX("bemsdosdate"), FILE_BEMSDOSDATE, FILE_FMT_STR }, 294 { XX("msdostime"), FILE_MSDOSTIME, FILE_FMT_STR }, 295 { XX("lemsdostime"), FILE_LEMSDOSTIME, FILE_FMT_STR }, 296 { XX("bemsdostime"), FILE_BEMSDOSTIME, FILE_FMT_STR }, 297 { XX("octal"), FILE_OCTAL, 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 */ 305static 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("octal"), FILE_OCTAL, FILE_FMT_STR }, 310 { XX_NULL, FILE_INVALID, FILE_FMT_NONE }, 311}; 312# undef XX 313# undef XX_NULL 314 315file_private int 316get_type(const struct type_tbl_s *tbl, const char *l, const char **t) 317{ 318 const struct type_tbl_s *p; 319 320 for (p = tbl; p->len; p++) { 321 if (strncmp(l, p->name, p->len) == 0) { 322 if (t) 323 *t = l + p->len; 324 break; 325 } 326 } 327 return p->type; 328} 329 330file_private off_t 331maxoff_t(void) { 332 if (/*CONSTCOND*/sizeof(off_t) == sizeof(int)) 333 return CAST(off_t, INT_MAX); 334 if (/*CONSTCOND*/sizeof(off_t) == sizeof(long)) 335 return CAST(off_t, LONG_MAX); 336 return 0x7fffffff; 337} 338 339file_private int 340get_standard_integer_type(const char *l, const char **t) 341{ 342 int type; 343 344 if (isalpha(CAST(unsigned char, l[1]))) { 345 switch (l[1]) { 346 case 'C': 347 /* "dC" and "uC" */ 348 type = FILE_BYTE; 349 break; 350 case 'S': 351 /* "dS" and "uS" */ 352 type = FILE_SHORT; 353 break; 354 case 'I': 355 case 'L': 356 /* 357 * "dI", "dL", "uI", and "uL". 358 * 359 * XXX - the actual Single UNIX Specification says 360 * that "L" means "long", as in the C data type, 361 * but we treat it as meaning "4-byte integer". 362 * Given that the OS X version of file 5.04 did 363 * the same, I guess that passes the actual SUS 364 * validation suite; having "dL" be dependent on 365 * how big a "long" is on the machine running 366 * "file" is silly. 367 */ 368 type = FILE_LONG; 369 break; 370 case 'Q': 371 /* "dQ" and "uQ" */ 372 type = FILE_QUAD; 373 break; 374 default: 375 /* "d{anything else}", "u{anything else}" */ 376 return FILE_INVALID; 377 } 378 l += 2; 379 } else if (isdigit(CAST(unsigned char, l[1]))) { 380 /* 381 * "d{num}" and "u{num}"; we only support {num} values 382 * of 1, 2, 4, and 8 - the Single UNIX Specification 383 * doesn't say anything about whether arbitrary 384 * values should be supported, but both the Solaris 10 385 * and OS X Mountain Lion versions of file passed the 386 * Single UNIX Specification validation suite, and 387 * neither of them support values bigger than 8 or 388 * non-power-of-2 values. 389 */ 390 if (isdigit(CAST(unsigned char, l[2]))) { 391 /* Multi-digit, so > 9 */ 392 return FILE_INVALID; 393 } 394 switch (l[1]) { 395 case '1': 396 type = FILE_BYTE; 397 break; 398 case '2': 399 type = FILE_SHORT; 400 break; 401 case '4': 402 type = FILE_LONG; 403 break; 404 case '8': 405 type = FILE_QUAD; 406 break; 407 default: 408 /* XXX - what about 3, 5, 6, or 7? */ 409 return FILE_INVALID; 410 } 411 l += 2; 412 } else { 413 /* 414 * "d" or "u" by itself. 415 */ 416 type = FILE_LONG; 417 ++l; 418 } 419 if (t) 420 *t = l; 421 return type; 422} 423 424file_private void 425init_file_tables(void) 426{ 427 static int done = 0; 428 const struct type_tbl_s *p; 429 430 if (done) 431 return; 432 done++; 433 434 for (p = type_tbl; p->len; p++) { 435 assert(p->type < FILE_NAMES_SIZE); 436 file_names[p->type] = p->name; 437 file_formats[p->type] = p->format; 438 } 439 assert(p - type_tbl == FILE_NAMES_SIZE); 440} 441 442file_private int 443add_mlist(struct mlist *mlp, struct magic_map *map, size_t idx) 444{ 445 struct mlist *ml; 446 447 mlp->map = NULL; 448 if ((ml = CAST(struct mlist *, malloc(sizeof(*ml)))) == NULL) 449 return -1; 450 451 ml->map = idx == 0 ? map : NULL; 452 ml->magic = map->magic[idx]; 453 ml->nmagic = map->nmagic[idx]; 454 if (ml->nmagic) { 455 ml->magic_rxcomp = CAST(file_regex_t **, 456 calloc(ml->nmagic, sizeof(*ml->magic_rxcomp))); 457 if (ml->magic_rxcomp == NULL) { 458 free(ml); 459 return -1; 460 } 461 } else 462 ml->magic_rxcomp = NULL; 463 mlp->prev->next = ml; 464 ml->prev = mlp->prev; 465 ml->next = mlp; 466 mlp->prev = ml; 467 return 0; 468} 469 470/* 471 * Handle one file or directory. 472 */ 473file_private int 474apprentice_1(struct magic_set *ms, const char *fn, int action) 475{ 476 struct magic_map *map; 477#ifndef COMPILE_ONLY 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(NULL, "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(*ms->mlist[0])); 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 532file_protected void 533file_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 549file_protected struct magic_set * 550file_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(*ms)))) == 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_shsize_max = FILE_ELF_SHSIZE_MAX; 581 ms->elf_phnum_max = FILE_ELF_PHNUM_MAX; 582 ms->elf_notes_max = FILE_ELF_NOTES_MAX; 583 ms->regex_max = FILE_REGEX_MAX; 584 ms->bytes_max = FILE_BYTES_MAX; 585 ms->encoding_max = FILE_ENCODING_MAX; 586#ifdef USE_C_LOCALE 587 ms->c_lc_ctype = newlocale(LC_CTYPE_MASK, "C", 0); 588 assert(ms->c_lc_ctype != NULL); 589#endif 590 return ms; 591free: 592 free(ms); 593 return NULL; 594} 595 596file_private void 597apprentice_unmap(struct magic_map *map) 598{ 599 size_t i; 600 char *p; 601 if (map == NULL) 602 return; 603 604 switch (map->type) { 605 case MAP_TYPE_USER: 606 break; 607 case MAP_TYPE_MALLOC: 608 p = CAST(char *, map->p); 609 for (i = 0; i < MAGIC_SETS; i++) { 610 char *b = RCAST(char *, map->magic[i]); 611 if (p != NULL && b >= p && b <= p + map->len) 612 continue; 613 free(b); 614 } 615 free(p); 616 break; 617#ifdef QUICK 618 case MAP_TYPE_MMAP: 619 if (map->p && map->p != MAP_FAILED) 620 (void)munmap(map->p, map->len); 621 break; 622#endif 623 default: 624 fprintf(stderr, "Bad map type %d", map->type); 625 abort(); 626 } 627 free(map); 628} 629 630file_private struct mlist * 631mlist_alloc(void) 632{ 633 struct mlist *mlist; 634 if ((mlist = CAST(struct mlist *, calloc(1, sizeof(*mlist)))) == NULL) { 635 return NULL; 636 } 637 mlist->next = mlist->prev = mlist; 638 return mlist; 639} 640 641file_private void 642mlist_free_all(struct magic_set *ms) 643{ 644 size_t i; 645 646 for (i = 0; i < MAGIC_SETS; i++) { 647 mlist_free(ms->mlist[i]); 648 ms->mlist[i] = NULL; 649 } 650} 651 652file_private void 653mlist_free_one(struct mlist *ml) 654{ 655 size_t i; 656 657 if (ml->map) 658 apprentice_unmap(CAST(struct magic_map *, ml->map)); 659 660 for (i = 0; i < ml->nmagic; ++i) { 661 if (ml->magic_rxcomp[i]) { 662 file_regfree(ml->magic_rxcomp[i]); 663 free(ml->magic_rxcomp[i]); 664 ml->magic_rxcomp[i] = NULL; 665 } 666 } 667 free(ml->magic_rxcomp); 668 ml->magic_rxcomp = NULL; 669 free(ml); 670} 671 672file_private void 673mlist_free(struct mlist *mlist) 674{ 675 struct mlist *ml, *next; 676 677 if (mlist == NULL) 678 return; 679 680 for (ml = mlist->next; ml != mlist;) { 681 next = ml->next; 682 mlist_free_one(ml); 683 ml = next; 684 } 685 mlist_free_one(mlist); 686} 687 688#ifndef COMPILE_ONLY 689/* void **bufs: an array of compiled magic files */ 690file_protected int 691buffer_apprentice(struct magic_set *ms, struct magic **bufs, 692 size_t *sizes, size_t nbufs) 693{ 694 size_t i, j; 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[0])); 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(*ms->mlist[0])); 720 goto fail; 721 } 722 } 723 } 724 725 return 0; 726fail: 727 mlist_free_all(ms); 728 return -1; 729} 730#endif 731 732/* const char *fn: list of magic files and directories */ 733file_protected int 734file_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[0])); 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 */ 823file_private size_t 824nonmagic(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 863file_private size_t 864typesize(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 */ 933file_private ssize_t 934apprentice_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 file_magwarn(NULL, "Usupported factor_op in default %d", 944 m->factor_op); 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 case FILE_OCTAL: 1006 val += m->vallen * MULT; 1007 break; 1008 1009 case FILE_BESTRING16: 1010 case FILE_LESTRING16: 1011 val += m->vallen * MULT / 2; 1012 break; 1013 1014 case FILE_SEARCH: 1015 if (m->vallen == 0) 1016 break; 1017 val += m->vallen * MAX(MULT / m->vallen, 1); 1018 break; 1019 1020 case FILE_REGEX: 1021 v = nonmagic(m->value.s); 1022 val += v * MAX(MULT / v, 1); 1023 break; 1024 1025 case FILE_INDIRECT: 1026 case FILE_NAME: 1027 case FILE_USE: 1028 case FILE_CLEAR: 1029 break; 1030 1031 case FILE_DER: 1032 val += MULT; 1033 break; 1034 1035 default: 1036 (void)fprintf(stderr, "Bad type %d\n", m->type); 1037 abort(); 1038 } 1039 1040 switch (m->reln) { 1041 case 'x': /* matches anything penalize */ 1042 case '!': /* matches almost anything penalize */ 1043 val = 0; 1044 break; 1045 1046 case '=': /* Exact match, prefer */ 1047 val += MULT; 1048 break; 1049 1050 case '>': 1051 case '<': /* comparison match reduce strength */ 1052 val -= 2 * MULT; 1053 break; 1054 1055 case '^': 1056 case '&': /* masking bits, we could count them too */ 1057 val -= MULT; 1058 break; 1059 1060 default: 1061 (void)fprintf(stderr, "Bad relation %c\n", m->reln); 1062 abort(); 1063 } 1064 1065 return val; 1066} 1067 1068 1069/*ARGSUSED*/ 1070file_protected size_t 1071file_magic_strength(const struct magic *m, 1072 size_t nmagic __attribute__((__unused__))) 1073{ 1074 ssize_t val = apprentice_magic_strength_1(m); 1075 1076#ifdef notyet 1077 if (m->desc[0] == '\0') { 1078 size_t i; 1079 /* 1080 * Magic entries with no description get their continuations 1081 * added 1082 */ 1083 for (i = 1; m[i].cont_level != 0 && i < MIN(nmagic, 3); i++) { 1084 ssize_t v = apprentice_magic_strength_1(&m[i]) >> 1085 (i + 1); 1086 val += v; 1087 if (m[i].desc[0] != '\0') 1088 break; 1089 } 1090 } 1091#endif 1092 1093 switch (m->factor_op) { 1094 case FILE_FACTOR_OP_NONE: 1095 break; 1096 case FILE_FACTOR_OP_PLUS: 1097 val += m->factor; 1098 break; 1099 case FILE_FACTOR_OP_MINUS: 1100 val -= m->factor; 1101 break; 1102 case FILE_FACTOR_OP_TIMES: 1103 val *= m->factor; 1104 break; 1105 case FILE_FACTOR_OP_DIV: 1106 val /= m->factor; 1107 break; 1108 default: 1109 (void)fprintf(stderr, "Bad factor_op %u\n", m->factor_op); 1110 abort(); 1111 } 1112 1113 if (val <= 0) /* ensure we only return 0 for FILE_DEFAULT */ 1114 val = 1; 1115 1116#ifndef notyet 1117 /* 1118 * Magic entries with no description get a bonus because they depend 1119 * on subsequent magic entries to print something. 1120 */ 1121 if (m->desc[0] == '\0') 1122 val++; 1123#endif 1124 1125 return val; 1126} 1127 1128/* 1129 * Sort callback for sorting entries by "strength" (basically length) 1130 */ 1131file_private int 1132apprentice_sort(const void *a, const void *b) 1133{ 1134 const struct magic_entry *ma = CAST(const struct magic_entry *, a); 1135 const struct magic_entry *mb = CAST(const struct magic_entry *, b); 1136 size_t sa = file_magic_strength(ma->mp, ma->cont_count); 1137 size_t sb = file_magic_strength(mb->mp, mb->cont_count); 1138 if (sa == sb) 1139 return 0; 1140 else if (sa > sb) 1141 return -1; 1142 else 1143 return 1; 1144} 1145 1146/* 1147 * Shows sorted patterns list in the order which is used for the matching 1148 */ 1149file_private void 1150apprentice_list(struct mlist *mlist, int mode) 1151{ 1152 uint32_t magindex, descindex, mimeindex, lineindex; 1153 struct mlist *ml; 1154 for (ml = mlist->next; ml != mlist; ml = ml->next) { 1155 for (magindex = 0; magindex < ml->nmagic; magindex++) { 1156 struct magic *m = &ml->magic[magindex]; 1157 if ((m->flag & mode) != mode) { 1158 /* Skip sub-tests */ 1159 while (magindex + 1 < ml->nmagic && 1160 ml->magic[magindex + 1].cont_level != 0) 1161 ++magindex; 1162 continue; /* Skip to next top-level test*/ 1163 } 1164 1165 /* 1166 * Try to iterate over the tree until we find item with 1167 * description/mimetype. 1168 */ 1169 lineindex = descindex = mimeindex = magindex; 1170 for (; magindex + 1 < ml->nmagic && 1171 ml->magic[magindex + 1].cont_level != 0; 1172 magindex++) { 1173 uint32_t mi = magindex + 1; 1174 if (*ml->magic[descindex].desc == '\0' 1175 && *ml->magic[mi].desc) 1176 descindex = mi; 1177 if (*ml->magic[mimeindex].mimetype == '\0' 1178 && *ml->magic[mi].mimetype) 1179 mimeindex = mi; 1180 } 1181 1182 printf("Strength = %3" SIZE_T_FORMAT "u@%u: %s [%s]\n", 1183 file_magic_strength(m, ml->nmagic - magindex), 1184 ml->magic[lineindex].lineno, 1185 ml->magic[descindex].desc, 1186 ml->magic[mimeindex].mimetype); 1187 } 1188 } 1189} 1190 1191file_private void 1192set_test_type(struct magic *mstart, struct magic *m) 1193{ 1194 switch (m->type) { 1195 case FILE_BYTE: 1196 case FILE_SHORT: 1197 case FILE_LONG: 1198 case FILE_DATE: 1199 case FILE_BESHORT: 1200 case FILE_BELONG: 1201 case FILE_BEDATE: 1202 case FILE_LESHORT: 1203 case FILE_LELONG: 1204 case FILE_LEDATE: 1205 case FILE_LDATE: 1206 case FILE_BELDATE: 1207 case FILE_LELDATE: 1208 case FILE_MEDATE: 1209 case FILE_MELDATE: 1210 case FILE_MELONG: 1211 case FILE_QUAD: 1212 case FILE_LEQUAD: 1213 case FILE_BEQUAD: 1214 case FILE_QDATE: 1215 case FILE_LEQDATE: 1216 case FILE_BEQDATE: 1217 case FILE_QLDATE: 1218 case FILE_LEQLDATE: 1219 case FILE_BEQLDATE: 1220 case FILE_QWDATE: 1221 case FILE_LEQWDATE: 1222 case FILE_BEQWDATE: 1223 case FILE_FLOAT: 1224 case FILE_BEFLOAT: 1225 case FILE_LEFLOAT: 1226 case FILE_DOUBLE: 1227 case FILE_BEDOUBLE: 1228 case FILE_LEDOUBLE: 1229 case FILE_BEVARINT: 1230 case FILE_LEVARINT: 1231 case FILE_DER: 1232 case FILE_GUID: 1233 case FILE_OFFSET: 1234 case FILE_MSDOSDATE: 1235 case FILE_BEMSDOSDATE: 1236 case FILE_LEMSDOSDATE: 1237 case FILE_MSDOSTIME: 1238 case FILE_BEMSDOSTIME: 1239 case FILE_LEMSDOSTIME: 1240 case FILE_OCTAL: 1241 mstart->flag |= BINTEST; 1242 break; 1243 case FILE_STRING: 1244 case FILE_PSTRING: 1245 case FILE_BESTRING16: 1246 case FILE_LESTRING16: 1247 /* Allow text overrides */ 1248 if (mstart->str_flags & STRING_TEXTTEST) 1249 mstart->flag |= TEXTTEST; 1250 else 1251 mstart->flag |= BINTEST; 1252 break; 1253 case FILE_REGEX: 1254 case FILE_SEARCH: 1255 /* Check for override */ 1256 if (mstart->str_flags & STRING_BINTEST) 1257 mstart->flag |= BINTEST; 1258 if (mstart->str_flags & STRING_TEXTTEST) 1259 mstart->flag |= TEXTTEST; 1260 1261 if (mstart->flag & (TEXTTEST|BINTEST)) 1262 break; 1263 1264 /* binary test if pattern is not text */ 1265 if (file_looks_utf8(m->value.us, CAST(size_t, m->vallen), NULL, 1266 NULL) <= 0) 1267 mstart->flag |= BINTEST; 1268 else 1269 mstart->flag |= TEXTTEST; 1270 break; 1271 case FILE_DEFAULT: 1272 /* can't deduce anything; we shouldn't see this at the 1273 top level anyway */ 1274 break; 1275 case FILE_INVALID: 1276 default: 1277 /* invalid search type, but no need to complain here */ 1278 break; 1279 } 1280} 1281 1282file_private int 1283addentry(struct magic_set *ms, struct magic_entry *me, 1284 struct magic_entry_set *mset) 1285{ 1286 size_t i = me->mp->type == FILE_NAME ? 1 : 0; 1287 if (mset[i].me == NULL || mset[i].count == mset[i].max) { 1288 struct magic_entry *mp; 1289 1290 size_t incr = mset[i].max + ALLOC_INCR; 1291 if ((mp = CAST(struct magic_entry *, 1292 realloc(mset[i].me, sizeof(*mp) * incr))) == 1293 NULL) { 1294 file_oomem(ms, sizeof(*mp) * incr); 1295 return -1; 1296 } 1297 (void)memset(&mp[mset[i].count], 0, sizeof(*mp) * 1298 ALLOC_INCR); 1299 mset[i].me = mp; 1300 mset[i].max = CAST(uint32_t, incr); 1301 assert(mset[i].max == incr); 1302 } 1303 mset[i].me[mset[i].count++] = *me; 1304 memset(me, 0, sizeof(*me)); 1305 return 0; 1306} 1307 1308/* 1309 * Load and parse one file. 1310 */ 1311file_private void 1312load_1(struct magic_set *ms, int action, const char *fn, int *errs, 1313 struct magic_entry_set *mset) 1314{ 1315 size_t lineno = 0, llen = 0; 1316 char *line = NULL; 1317 ssize_t len; 1318 struct magic_entry me; 1319 1320 FILE *f = fopen(ms->file = fn, "r"); 1321 if (f == NULL) { 1322 if (errno != ENOENT) 1323 file_error(ms, errno, "cannot read magic file `%s'", 1324 fn); 1325 (*errs)++; 1326 return; 1327 } 1328 1329 memset(&me, 0, sizeof(me)); 1330 /* read and parse this file */ 1331 for (ms->line = 1; (len = getline(&line, &llen, f)) != -1; 1332 ms->line++) { 1333 if (len == 0) /* null line, garbage, etc */ 1334 continue; 1335 if (line[len - 1] == '\n') { 1336 lineno++; 1337 line[len - 1] = '\0'; /* delete newline */ 1338 } 1339 switch (line[0]) { 1340 case '\0': /* empty, do not parse */ 1341 case '#': /* comment, do not parse */ 1342 continue; 1343 case '!': 1344 if (line[1] == ':') { 1345 size_t i; 1346 1347 for (i = 0; bang[i].name != NULL; i++) { 1348 if (CAST(size_t, len - 2) > bang[i].len && 1349 memcmp(bang[i].name, line + 2, 1350 bang[i].len) == 0) 1351 break; 1352 } 1353 if (bang[i].name == NULL) { 1354 file_error(ms, 0, 1355 "Unknown !: entry `%s'", line); 1356 (*errs)++; 1357 continue; 1358 } 1359 if (me.mp == NULL) { 1360 file_error(ms, 0, 1361 "No current entry for :!%s type", 1362 bang[i].name); 1363 (*errs)++; 1364 continue; 1365 } 1366 if ((*bang[i].fun)(ms, &me, 1367 line + bang[i].len + 2, 1368 len - bang[i].len - 2) != 0) { 1369 (*errs)++; 1370 continue; 1371 } 1372 continue; 1373 } 1374 /*FALLTHROUGH*/ 1375 default: 1376 again: 1377 switch (parse(ms, &me, line, lineno, action)) { 1378 case 0: 1379 continue; 1380 case 1: 1381 (void)addentry(ms, &me, mset); 1382 goto again; 1383 default: 1384 (*errs)++; 1385 break; 1386 } 1387 } 1388 } 1389 if (me.mp) 1390 (void)addentry(ms, &me, mset); 1391 free(line); 1392 (void)fclose(f); 1393} 1394 1395/* 1396 * parse a file or directory of files 1397 * const char *fn: name of magic file or directory 1398 */ 1399file_private int 1400cmpstrp(const void *p1, const void *p2) 1401{ 1402 return strcmp(*RCAST(char *const *, p1), *RCAST(char *const *, p2)); 1403} 1404 1405 1406file_private uint32_t 1407set_text_binary(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1408 uint32_t starttest) 1409{ 1410 static const char text[] = "text"; 1411 static const char binary[] = "binary"; 1412 static const size_t len = sizeof(text); 1413 1414 uint32_t i = starttest; 1415 1416 do { 1417 set_test_type(me[starttest].mp, me[i].mp); 1418 if ((ms->flags & MAGIC_DEBUG) == 0) 1419 continue; 1420 (void)fprintf(stderr, "%s%s%s: %s\n", 1421 me[i].mp->mimetype, 1422 me[i].mp->mimetype[0] == '\0' ? "" : "; ", 1423 me[i].mp->desc[0] ? me[i].mp->desc : "(no description)", 1424 me[i].mp->flag & BINTEST ? binary : text); 1425 if (me[i].mp->flag & BINTEST) { 1426 char *p = strstr(me[i].mp->desc, text); 1427 if (p && (p == me[i].mp->desc || 1428 isspace(CAST(unsigned char, p[-1]))) && 1429 (p + len - me[i].mp->desc == MAXstring 1430 || (p[len] == '\0' || 1431 isspace(CAST(unsigned char, p[len]))))) 1432 (void)fprintf(stderr, "*** Possible " 1433 "binary test for text type\n"); 1434 } 1435 } while (++i < nme && me[i].mp->cont_level != 0); 1436 return i; 1437} 1438 1439file_private void 1440set_last_default(struct magic_set *ms, struct magic_entry *me, uint32_t nme) 1441{ 1442 uint32_t i; 1443 for (i = 0; i < nme; i++) { 1444 if (me[i].mp->cont_level == 0 && 1445 me[i].mp->type == FILE_DEFAULT) { 1446 while (++i < nme) 1447 if (me[i].mp->cont_level == 0) 1448 break; 1449 if (i != nme) { 1450 /* XXX - Ugh! */ 1451 ms->line = me[i].mp->lineno; 1452 file_magwarn(ms, 1453 "level 0 \"default\" did not sort last"); 1454 } 1455 return; 1456 } 1457 } 1458} 1459 1460file_private int 1461coalesce_entries(struct magic_set *ms, struct magic_entry *me, uint32_t nme, 1462 struct magic **ma, uint32_t *nma) 1463{ 1464 uint32_t i, mentrycount = 0; 1465 size_t slen; 1466 1467 for (i = 0; i < nme; i++) 1468 mentrycount += me[i].cont_count; 1469 1470 if (mentrycount == 0) { 1471 *ma = NULL; 1472 *nma = 0; 1473 return 0; 1474 } 1475 1476 slen = sizeof(**ma) * mentrycount; 1477 if ((*ma = CAST(struct magic *, malloc(slen))) == NULL) { 1478 file_oomem(ms, slen); 1479 return -1; 1480 } 1481 1482 mentrycount = 0; 1483 for (i = 0; i < nme; i++) { 1484 (void)memcpy(*ma + mentrycount, me[i].mp, 1485 me[i].cont_count * sizeof(**ma)); 1486 mentrycount += me[i].cont_count; 1487 } 1488 *nma = mentrycount; 1489 return 0; 1490} 1491 1492file_private void 1493magic_entry_free(struct magic_entry *me, uint32_t nme) 1494{ 1495 uint32_t i; 1496 if (me == NULL) 1497 return; 1498 for (i = 0; i < nme; i++) 1499 free(me[i].mp); 1500 free(me); 1501} 1502 1503file_private struct magic_map * 1504apprentice_load(struct magic_set *ms, const char *fn, int action) 1505{ 1506 int errs = 0; 1507 uint32_t i, j; 1508 size_t files = 0, maxfiles = 0; 1509 char **filearr = NULL, *mfn; 1510 struct stat st; 1511 struct magic_map *map; 1512 struct magic_entry_set mset[MAGIC_SETS]; 1513 DIR *dir; 1514 struct dirent *d; 1515 1516 memset(mset, 0, sizeof(mset)); 1517 ms->flags |= MAGIC_CHECK; /* Enable checks for parsed files */ 1518 1519 1520 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) 1521 { 1522 file_oomem(ms, sizeof(*map)); 1523 return NULL; 1524 } 1525 map->type = MAP_TYPE_MALLOC; 1526 1527 /* print silly verbose header for USG compat. */ 1528 if (action == FILE_CHECK) 1529 (void)fprintf(stderr, "%s\n", usg_hdr); 1530 1531 /* load directory or file */ 1532 if (stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) { 1533 dir = opendir(fn); 1534 if (!dir) { 1535 errs++; 1536 goto out; 1537 } 1538 while ((d = readdir(dir)) != NULL) { 1539 if (d->d_name[0] == '.') 1540 continue; 1541 if (asprintf(&mfn, "%s/%s", fn, d->d_name) < 0) { 1542 file_oomem(ms, 1543 strlen(fn) + strlen(d->d_name) + 2); 1544 errs++; 1545 closedir(dir); 1546 goto out; 1547 } 1548 if (stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) { 1549 free(mfn); 1550 continue; 1551 } 1552 if (files >= maxfiles) { 1553 size_t mlen; 1554 char **nfilearr; 1555 maxfiles = (maxfiles + 1) * 2; 1556 mlen = maxfiles * sizeof(*filearr); 1557 if ((nfilearr = CAST(char **, 1558 realloc(filearr, mlen))) == NULL) { 1559 file_oomem(ms, mlen); 1560 free(mfn); 1561 closedir(dir); 1562 errs++; 1563 goto out; 1564 } 1565 filearr = nfilearr; 1566 } 1567 filearr[files++] = mfn; 1568 } 1569 closedir(dir); 1570 if (filearr) { 1571 qsort(filearr, files, sizeof(*filearr), cmpstrp); 1572 for (i = 0; i < files; i++) { 1573 load_1(ms, action, filearr[i], &errs, mset); 1574 free(filearr[i]); 1575 } 1576 free(filearr); 1577 filearr = NULL; 1578 } 1579 } else 1580 load_1(ms, action, fn, &errs, mset); 1581 if (errs) 1582 goto out; 1583 1584 for (j = 0; j < MAGIC_SETS; j++) { 1585 /* Set types of tests */ 1586 for (i = 0; i < mset[j].count; ) { 1587 if (mset[j].me[i].mp->cont_level != 0) { 1588 i++; 1589 continue; 1590 } 1591 i = set_text_binary(ms, mset[j].me, mset[j].count, i); 1592 } 1593 if (mset[j].me) 1594 qsort(mset[j].me, mset[j].count, sizeof(*mset[0].me), 1595 apprentice_sort); 1596 1597 /* 1598 * Make sure that any level 0 "default" line is last 1599 * (if one exists). 1600 */ 1601 set_last_default(ms, mset[j].me, mset[j].count); 1602 1603 /* coalesce per file arrays into a single one, if needed */ 1604 if (mset[j].count == 0) 1605 continue; 1606 1607 if (coalesce_entries(ms, mset[j].me, mset[j].count, 1608 &map->magic[j], &map->nmagic[j]) == -1) { 1609 errs++; 1610 goto out; 1611 } 1612 } 1613 1614out: 1615 free(filearr); 1616 for (j = 0; j < MAGIC_SETS; j++) 1617 magic_entry_free(mset[j].me, mset[j].count); 1618 1619 if (errs) { 1620 apprentice_unmap(map); 1621 return NULL; 1622 } 1623 return map; 1624} 1625 1626/* 1627 * extend the sign bit if the comparison is to be signed 1628 */ 1629file_protected uint64_t 1630file_signextend(struct magic_set *ms, struct magic *m, uint64_t v) 1631{ 1632 if (!(m->flag & UNSIGNED)) { 1633 switch(m->type) { 1634 /* 1635 * Do not remove the casts below. They are 1636 * vital. When later compared with the data, 1637 * the sign extension must have happened. 1638 */ 1639 case FILE_BYTE: 1640 v = CAST(signed char, v); 1641 break; 1642 case FILE_SHORT: 1643 case FILE_BESHORT: 1644 case FILE_LESHORT: 1645 v = CAST(short, v); 1646 break; 1647 case FILE_DATE: 1648 case FILE_BEDATE: 1649 case FILE_LEDATE: 1650 case FILE_MEDATE: 1651 case FILE_LDATE: 1652 case FILE_BELDATE: 1653 case FILE_LELDATE: 1654 case FILE_MELDATE: 1655 case FILE_LONG: 1656 case FILE_BELONG: 1657 case FILE_LELONG: 1658 case FILE_MELONG: 1659 case FILE_FLOAT: 1660 case FILE_BEFLOAT: 1661 case FILE_LEFLOAT: 1662 case FILE_MSDOSDATE: 1663 case FILE_BEMSDOSDATE: 1664 case FILE_LEMSDOSDATE: 1665 case FILE_MSDOSTIME: 1666 case FILE_BEMSDOSTIME: 1667 case FILE_LEMSDOSTIME: 1668 v = CAST(int32_t, v); 1669 break; 1670 case FILE_QUAD: 1671 case FILE_BEQUAD: 1672 case FILE_LEQUAD: 1673 case FILE_QDATE: 1674 case FILE_QLDATE: 1675 case FILE_QWDATE: 1676 case FILE_BEQDATE: 1677 case FILE_BEQLDATE: 1678 case FILE_BEQWDATE: 1679 case FILE_LEQDATE: 1680 case FILE_LEQLDATE: 1681 case FILE_LEQWDATE: 1682 case FILE_DOUBLE: 1683 case FILE_BEDOUBLE: 1684 case FILE_LEDOUBLE: 1685 case FILE_OFFSET: 1686 case FILE_BEVARINT: 1687 case FILE_LEVARINT: 1688 v = CAST(int64_t, v); 1689 break; 1690 case FILE_STRING: 1691 case FILE_PSTRING: 1692 case FILE_BESTRING16: 1693 case FILE_LESTRING16: 1694 case FILE_REGEX: 1695 case FILE_SEARCH: 1696 case FILE_DEFAULT: 1697 case FILE_INDIRECT: 1698 case FILE_NAME: 1699 case FILE_USE: 1700 case FILE_CLEAR: 1701 case FILE_DER: 1702 case FILE_GUID: 1703 case FILE_OCTAL: 1704 break; 1705 default: 1706 if (ms->flags & MAGIC_CHECK) 1707 file_magwarn(ms, "cannot happen: m->type=%d\n", 1708 m->type); 1709 return FILE_BADSIZE; 1710 } 1711 } 1712 return v; 1713} 1714 1715file_private int 1716string_modifier_check(struct magic_set *ms, struct magic *m) 1717{ 1718 if ((ms->flags & MAGIC_CHECK) == 0) 1719 return 0; 1720 1721 if ((m->type != FILE_REGEX || (m->str_flags & REGEX_LINE_COUNT) == 0) && 1722 (m->type != FILE_PSTRING && (m->str_flags & PSTRING_LEN) != 0)) { 1723 file_magwarn(ms, 1724 "'/BHhLl' modifiers are only allowed for pascal strings\n"); 1725 return -1; 1726 } 1727 switch (m->type) { 1728 case FILE_BESTRING16: 1729 case FILE_LESTRING16: 1730 if (m->str_flags != 0) { 1731 file_magwarn(ms, 1732 "no modifiers allowed for 16-bit strings\n"); 1733 return -1; 1734 } 1735 break; 1736 case FILE_STRING: 1737 case FILE_PSTRING: 1738 if ((m->str_flags & REGEX_OFFSET_START) != 0) { 1739 file_magwarn(ms, 1740 "'/%c' only allowed on regex and search\n", 1741 CHAR_REGEX_OFFSET_START); 1742 return -1; 1743 } 1744 break; 1745 case FILE_SEARCH: 1746 if (m->str_range == 0) { 1747 file_magwarn(ms, 1748 "missing range; defaulting to %d\n", 1749 STRING_DEFAULT_RANGE); 1750 m->str_range = STRING_DEFAULT_RANGE; 1751 return -1; 1752 } 1753 break; 1754 case FILE_REGEX: 1755 if ((m->str_flags & STRING_COMPACT_WHITESPACE) != 0) { 1756 file_magwarn(ms, "'/%c' not allowed on regex\n", 1757 CHAR_COMPACT_WHITESPACE); 1758 return -1; 1759 } 1760 if ((m->str_flags & STRING_COMPACT_OPTIONAL_WHITESPACE) != 0) { 1761 file_magwarn(ms, "'/%c' not allowed on regex\n", 1762 CHAR_COMPACT_OPTIONAL_WHITESPACE); 1763 return -1; 1764 } 1765 break; 1766 default: 1767 file_magwarn(ms, "coding error: m->type=%d\n", 1768 m->type); 1769 return -1; 1770 } 1771 return 0; 1772} 1773 1774file_private int 1775get_op(char c) 1776{ 1777 switch (c) { 1778 case '&': 1779 return FILE_OPAND; 1780 case '|': 1781 return FILE_OPOR; 1782 case '^': 1783 return FILE_OPXOR; 1784 case '+': 1785 return FILE_OPADD; 1786 case '-': 1787 return FILE_OPMINUS; 1788 case '*': 1789 return FILE_OPMULTIPLY; 1790 case '/': 1791 return FILE_OPDIVIDE; 1792 case '%': 1793 return FILE_OPMODULO; 1794 default: 1795 return -1; 1796 } 1797} 1798 1799#ifdef ENABLE_CONDITIONALS 1800file_private int 1801get_cond(const char *l, const char **t) 1802{ 1803 static const struct cond_tbl_s { 1804 char name[8]; 1805 size_t len; 1806 int cond; 1807 } cond_tbl[] = { 1808 { "if", 2, COND_IF }, 1809 { "elif", 4, COND_ELIF }, 1810 { "else", 4, COND_ELSE }, 1811 { "", 0, COND_NONE }, 1812 }; 1813 const struct cond_tbl_s *p; 1814 1815 for (p = cond_tbl; p->len; p++) { 1816 if (strncmp(l, p->name, p->len) == 0 && 1817 isspace(CAST(unsigned char, l[p->len]))) { 1818 if (t) 1819 *t = l + p->len; 1820 break; 1821 } 1822 } 1823 return p->cond; 1824} 1825 1826file_private int 1827check_cond(struct magic_set *ms, int cond, uint32_t cont_level) 1828{ 1829 int last_cond; 1830 last_cond = ms->c.li[cont_level].last_cond; 1831 1832 switch (cond) { 1833 case COND_IF: 1834 if (last_cond != COND_NONE && last_cond != COND_ELIF) { 1835 if (ms->flags & MAGIC_CHECK) 1836 file_magwarn(ms, "syntax error: `if'"); 1837 return -1; 1838 } 1839 last_cond = COND_IF; 1840 break; 1841 1842 case COND_ELIF: 1843 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1844 if (ms->flags & MAGIC_CHECK) 1845 file_magwarn(ms, "syntax error: `elif'"); 1846 return -1; 1847 } 1848 last_cond = COND_ELIF; 1849 break; 1850 1851 case COND_ELSE: 1852 if (last_cond != COND_IF && last_cond != COND_ELIF) { 1853 if (ms->flags & MAGIC_CHECK) 1854 file_magwarn(ms, "syntax error: `else'"); 1855 return -1; 1856 } 1857 last_cond = COND_NONE; 1858 break; 1859 1860 case COND_NONE: 1861 last_cond = COND_NONE; 1862 break; 1863 } 1864 1865 ms->c.li[cont_level].last_cond = last_cond; 1866 return 0; 1867} 1868#endif /* ENABLE_CONDITIONALS */ 1869 1870file_private int 1871parse_indirect_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1872{ 1873 const char *l = *lp; 1874 1875 while (!isspace(CAST(unsigned char, *++l))) 1876 switch (*l) { 1877 case CHAR_INDIRECT_RELATIVE: 1878 m->str_flags |= INDIRECT_RELATIVE; 1879 break; 1880 default: 1881 if (ms->flags & MAGIC_CHECK) 1882 file_magwarn(ms, "indirect modifier `%c' " 1883 "invalid", *l); 1884 *lp = l; 1885 return -1; 1886 } 1887 *lp = l; 1888 return 0; 1889} 1890 1891file_private void 1892parse_op_modifier(struct magic_set *ms, struct magic *m, const char **lp, 1893 int op) 1894{ 1895 const char *l = *lp; 1896 char *t; 1897 uint64_t val; 1898 1899 ++l; 1900 m->mask_op |= op; 1901 val = CAST(uint64_t, strtoull(l, &t, 0)); 1902 l = t; 1903 m->num_mask = file_signextend(ms, m, val); 1904 eatsize(&l); 1905 *lp = l; 1906} 1907 1908file_private int 1909parse_string_modifier(struct magic_set *ms, struct magic *m, const char **lp) 1910{ 1911 const char *l = *lp; 1912 char *t; 1913 int have_range = 0; 1914 1915 while (!isspace(CAST(unsigned char, *++l))) { 1916 switch (*l) { 1917 case '0': case '1': case '2': 1918 case '3': case '4': case '5': 1919 case '6': case '7': case '8': 1920 case '9': 1921 if (have_range && (ms->flags & MAGIC_CHECK)) 1922 file_magwarn(ms, "multiple ranges"); 1923 have_range = 1; 1924 m->str_range = CAST(uint32_t, strtoul(l, &t, 0)); 1925 if (m->str_range == 0) 1926 file_magwarn(ms, "zero range"); 1927 l = t - 1; 1928 break; 1929 case CHAR_COMPACT_WHITESPACE: 1930 m->str_flags |= STRING_COMPACT_WHITESPACE; 1931 break; 1932 case CHAR_COMPACT_OPTIONAL_WHITESPACE: 1933 m->str_flags |= STRING_COMPACT_OPTIONAL_WHITESPACE; 1934 break; 1935 case CHAR_IGNORE_LOWERCASE: 1936 m->str_flags |= STRING_IGNORE_LOWERCASE; 1937 break; 1938 case CHAR_IGNORE_UPPERCASE: 1939 m->str_flags |= STRING_IGNORE_UPPERCASE; 1940 break; 1941 case CHAR_REGEX_OFFSET_START: 1942 m->str_flags |= REGEX_OFFSET_START; 1943 break; 1944 case CHAR_BINTEST: 1945 m->str_flags |= STRING_BINTEST; 1946 break; 1947 case CHAR_TEXTTEST: 1948 m->str_flags |= STRING_TEXTTEST; 1949 break; 1950 case CHAR_TRIM: 1951 m->str_flags |= STRING_TRIM; 1952 break; 1953 case CHAR_FULL_WORD: 1954 m->str_flags |= STRING_FULL_WORD; 1955 break; 1956 case CHAR_PSTRING_1_LE: 1957#define SET_LENGTH(a) m->str_flags = (m->str_flags & ~PSTRING_LEN) | (a) 1958 if (m->type != FILE_PSTRING) 1959 goto bad; 1960 SET_LENGTH(PSTRING_1_LE); 1961 break; 1962 case CHAR_PSTRING_2_BE: 1963 if (m->type != FILE_PSTRING) 1964 goto bad; 1965 SET_LENGTH(PSTRING_2_BE); 1966 break; 1967 case CHAR_PSTRING_2_LE: 1968 if (m->type != FILE_PSTRING) 1969 goto bad; 1970 SET_LENGTH(PSTRING_2_LE); 1971 break; 1972 case CHAR_PSTRING_4_BE: 1973 if (m->type != FILE_PSTRING) 1974 goto bad; 1975 SET_LENGTH(PSTRING_4_BE); 1976 break; 1977 case CHAR_PSTRING_4_LE: 1978 switch (m->type) { 1979 case FILE_PSTRING: 1980 case FILE_REGEX: 1981 break; 1982 default: 1983 goto bad; 1984 } 1985 SET_LENGTH(PSTRING_4_LE); 1986 break; 1987 case CHAR_PSTRING_LENGTH_INCLUDES_ITSELF: 1988 if (m->type != FILE_PSTRING) 1989 goto bad; 1990 m->str_flags |= PSTRING_LENGTH_INCLUDES_ITSELF; 1991 break; 1992 default: 1993 bad: 1994 if (ms->flags & MAGIC_CHECK) 1995 file_magwarn(ms, "string modifier `%c' " 1996 "invalid", *l); 1997 goto out; 1998 } 1999 /* allow multiple '/' for readability */ 2000 if (l[1] == '/' && !isspace(CAST(unsigned char, l[2]))) 2001 l++; 2002 } 2003 if (string_modifier_check(ms, m) == -1) 2004 goto out; 2005 *lp = l; 2006 return 0; 2007out: 2008 *lp = l; 2009 return -1; 2010} 2011 2012/* 2013 * parse one line from magic file, put into magic[index++] if valid 2014 */ 2015file_private int 2016parse(struct magic_set *ms, struct magic_entry *me, const char *line, 2017 size_t lineno, int action) 2018{ 2019#ifdef ENABLE_CONDITIONALS 2020 static uint32_t last_cont_level = 0; 2021#endif 2022 size_t i; 2023 struct magic *m; 2024 const char *l = line; 2025 char *t; 2026 int op; 2027 uint32_t cont_level; 2028 int32_t diff; 2029 2030 cont_level = 0; 2031 2032 /* 2033 * Parse the offset. 2034 */ 2035 while (*l == '>') { 2036 ++l; /* step over */ 2037 cont_level++; 2038 } 2039#ifdef ENABLE_CONDITIONALS 2040 if (cont_level == 0 || cont_level > last_cont_level) 2041 if (file_check_mem(ms, cont_level) == -1) 2042 return -1; 2043 last_cont_level = cont_level; 2044#endif 2045 if (cont_level != 0) { 2046 if (me->mp == NULL) { 2047 file_magerror(ms, "No current entry for continuation"); 2048 return -1; 2049 } 2050 if (me->cont_count == 0) { 2051 file_magerror(ms, "Continuations present with 0 count"); 2052 return -1; 2053 } 2054 m = &me->mp[me->cont_count - 1]; 2055 diff = CAST(int32_t, cont_level) - CAST(int32_t, m->cont_level); 2056 if (diff > 1) 2057 file_magwarn(ms, "New continuation level %u is more " 2058 "than one larger than current level %u", cont_level, 2059 m->cont_level); 2060 if (me->cont_count == me->max_count) { 2061 struct magic *nm; 2062 size_t cnt = me->max_count + ALLOC_CHUNK; 2063 if ((nm = CAST(struct magic *, realloc(me->mp, 2064 sizeof(*nm) * cnt))) == NULL) { 2065 file_oomem(ms, sizeof(*nm) * cnt); 2066 return -1; 2067 } 2068 me->mp = nm; 2069 me->max_count = CAST(uint32_t, cnt); 2070 } 2071 m = &me->mp[me->cont_count++]; 2072 (void)memset(m, 0, sizeof(*m)); 2073 m->cont_level = cont_level; 2074 } else { 2075 static const size_t len = sizeof(*m) * ALLOC_CHUNK; 2076 if (me->mp != NULL) 2077 return 1; 2078 if ((m = CAST(struct magic *, malloc(len))) == NULL) { 2079 file_oomem(ms, len); 2080 return -1; 2081 } 2082 me->mp = m; 2083 me->max_count = ALLOC_CHUNK; 2084 (void)memset(m, 0, sizeof(*m)); 2085 m->factor_op = FILE_FACTOR_OP_NONE; 2086 m->cont_level = 0; 2087 me->cont_count = 1; 2088 } 2089 m->lineno = CAST(uint32_t, lineno); 2090 2091 if (*l == '&') { /* m->cont_level == 0 checked below. */ 2092 ++l; /* step over */ 2093 m->flag |= OFFADD; 2094 } 2095 if (*l == '(') { 2096 ++l; /* step over */ 2097 m->flag |= INDIR; 2098 if (m->flag & OFFADD) 2099 m->flag = (m->flag & ~OFFADD) | INDIROFFADD; 2100 2101 if (*l == '&') { /* m->cont_level == 0 checked below */ 2102 ++l; /* step over */ 2103 m->flag |= OFFADD; 2104 } 2105 } 2106 /* Indirect offsets are not valid at level 0. */ 2107 if (m->cont_level == 0 && (m->flag & (OFFADD | INDIROFFADD))) { 2108 if (ms->flags & MAGIC_CHECK) 2109 file_magwarn(ms, "relative offset at level 0"); 2110 return -1; 2111 } 2112 2113 /* get offset, then skip over it */ 2114 if (*l == '-') { 2115 ++l; /* step over */ 2116 m->flag |= OFFNEGATIVE; 2117 } 2118 m->offset = CAST(int32_t, strtol(l, &t, 0)); 2119 if (l == t) { 2120 if (ms->flags & MAGIC_CHECK) 2121 file_magwarn(ms, "offset `%s' invalid", l); 2122 return -1; 2123 } 2124 2125 l = t; 2126 2127 if (m->flag & INDIR) { 2128 m->in_type = FILE_LONG; 2129 m->in_offset = 0; 2130 m->in_op = 0; 2131 /* 2132 * read [.,lbs][+-]nnnnn) 2133 */ 2134 if (*l == '.' || *l == ',') { 2135 if (*l == ',') 2136 m->in_op |= FILE_OPSIGNED; 2137 l++; 2138 switch (*l) { 2139 case 'l': 2140 m->in_type = FILE_LELONG; 2141 break; 2142 case 'L': 2143 m->in_type = FILE_BELONG; 2144 break; 2145 case 'm': 2146 m->in_type = FILE_MELONG; 2147 break; 2148 case 'h': 2149 case 's': 2150 m->in_type = FILE_LESHORT; 2151 break; 2152 case 'H': 2153 case 'S': 2154 m->in_type = FILE_BESHORT; 2155 break; 2156 case 'c': 2157 case 'b': 2158 case 'C': 2159 case 'B': 2160 m->in_type = FILE_BYTE; 2161 break; 2162 case 'e': 2163 case 'f': 2164 case 'g': 2165 m->in_type = FILE_LEDOUBLE; 2166 break; 2167 case 'E': 2168 case 'F': 2169 case 'G': 2170 m->in_type = FILE_BEDOUBLE; 2171 break; 2172 case 'i': 2173 m->in_type = FILE_LEID3; 2174 break; 2175 case 'I': 2176 m->in_type = FILE_BEID3; 2177 break; 2178 case 'o': 2179 m->in_type = FILE_OCTAL; 2180 break; 2181 case 'q': 2182 m->in_type = FILE_LEQUAD; 2183 break; 2184 case 'Q': 2185 m->in_type = FILE_BEQUAD; 2186 break; 2187 default: 2188 if (ms->flags & MAGIC_CHECK) 2189 file_magwarn(ms, 2190 "indirect offset type `%c' invalid", 2191 *l); 2192 return -1; 2193 } 2194 l++; 2195 } 2196 2197 if (*l == '~') { 2198 m->in_op |= FILE_OPINVERSE; 2199 l++; 2200 } 2201 if ((op = get_op(*l)) != -1) { 2202 m->in_op |= op; 2203 l++; 2204 } 2205 if (*l == '(') { 2206 m->in_op |= FILE_OPINDIRECT; 2207 l++; 2208 } 2209 if (isdigit(CAST(unsigned char, *l)) || *l == '-') { 2210 m->in_offset = CAST(int32_t, strtol(l, &t, 0)); 2211 if (l == t) { 2212 if (ms->flags & MAGIC_CHECK) 2213 file_magwarn(ms, 2214 "in_offset `%s' invalid", l); 2215 return -1; 2216 } 2217 l = t; 2218 } 2219 if (*l++ != ')' || 2220 ((m->in_op & FILE_OPINDIRECT) && *l++ != ')')) { 2221 if (ms->flags & MAGIC_CHECK) 2222 file_magwarn(ms, 2223 "missing ')' in indirect offset"); 2224 return -1; 2225 } 2226 } 2227 EATAB; 2228 2229#ifdef ENABLE_CONDITIONALS 2230 m->cond = get_cond(l, &l); 2231 if (check_cond(ms, m->cond, cont_level) == -1) 2232 return -1; 2233 2234 EATAB; 2235#endif 2236 2237 /* 2238 * Parse the type. 2239 */ 2240 if (*l == 'u') { 2241 /* 2242 * Try it as a keyword type prefixed by "u"; match what 2243 * follows the "u". If that fails, try it as an SUS 2244 * integer type. 2245 */ 2246 m->type = get_type(type_tbl, l + 1, &l); 2247 if (m->type == FILE_INVALID) { 2248 /* 2249 * Not a keyword type; parse it as an SUS type, 2250 * 'u' possibly followed by a number or C/S/L. 2251 */ 2252 m->type = get_standard_integer_type(l, &l); 2253 } 2254 /* It's unsigned. */ 2255 if (m->type != FILE_INVALID) 2256 m->flag |= UNSIGNED; 2257 } else { 2258 /* 2259 * Try it as a keyword type. If that fails, try it as 2260 * an SUS integer type if it begins with "d" or as an 2261 * SUS string type if it begins with "s". In any case, 2262 * it's not unsigned. 2263 */ 2264 m->type = get_type(type_tbl, l, &l); 2265 if (m->type == FILE_INVALID) { 2266 /* 2267 * Not a keyword type; parse it as an SUS type, 2268 * either 'd' possibly followed by a number or 2269 * C/S/L, or just 's'. 2270 */ 2271 if (*l == 'd') 2272 m->type = get_standard_integer_type(l, &l); 2273 else if (*l == 's' 2274 && !isalpha(CAST(unsigned char, l[1]))) { 2275 m->type = FILE_STRING; 2276 ++l; 2277 } 2278 } 2279 } 2280 2281 if (m->type == FILE_INVALID) { 2282 /* Not found - try it as a special keyword. */ 2283 m->type = get_type(special_tbl, l, &l); 2284 } 2285 2286 if (m->type == FILE_INVALID) { 2287 if (ms->flags & MAGIC_CHECK) 2288 file_magwarn(ms, "type `%s' invalid", l); 2289 return -1; 2290 } 2291 2292 if (m->type == FILE_NAME && cont_level != 0) { 2293 if (ms->flags & MAGIC_CHECK) 2294 file_magwarn(ms, "`name%s' entries can only be " 2295 "declared at top level", l); 2296 return -1; 2297 } 2298 2299 /* New-style anding: "0 byte&0x80 =0x80 dynamically linked" */ 2300 /* New and improved: ~ & | ^ + - * / % -- exciting, isn't it? */ 2301 2302 m->mask_op = 0; 2303 if (*l == '~') { 2304 if (!IS_STRING(m->type)) 2305 m->mask_op |= FILE_OPINVERSE; 2306 else if (ms->flags & MAGIC_CHECK) 2307 file_magwarn(ms, "'~' invalid for string types"); 2308 ++l; 2309 } 2310 m->str_range = 0; 2311 m->str_flags = m->type == FILE_PSTRING ? PSTRING_1_LE : 0; 2312 if ((op = get_op(*l)) != -1) { 2313 if (IS_STRING(m->type)) { 2314 int r; 2315 2316 if (op != FILE_OPDIVIDE) { 2317 if (ms->flags & MAGIC_CHECK) 2318 file_magwarn(ms, 2319 "invalid string/indirect op: " 2320 "`%c'", *t); 2321 return -1; 2322 } 2323 2324 if (m->type == FILE_INDIRECT) 2325 r = parse_indirect_modifier(ms, m, &l); 2326 else 2327 r = parse_string_modifier(ms, m, &l); 2328 if (r == -1) 2329 return -1; 2330 } else 2331 parse_op_modifier(ms, m, &l, op); 2332 } 2333 2334 /* 2335 * We used to set mask to all 1's here, instead let's just not do 2336 * anything if mask = 0 (unless you have a better idea) 2337 */ 2338 EATAB; 2339 2340 switch (*l) { 2341 case '>': 2342 case '<': 2343 m->reln = *l; 2344 ++l; 2345 if (*l == '=') { 2346 if (ms->flags & MAGIC_CHECK) { 2347 file_magwarn(ms, "%c= not supported", 2348 m->reln); 2349 return -1; 2350 } 2351 ++l; 2352 } 2353 break; 2354 /* Old-style anding: "0 byte &0x80 dynamically linked" */ 2355 case '&': 2356 case '^': 2357 case '=': 2358 m->reln = *l; 2359 ++l; 2360 if (*l == '=') { 2361 /* HP compat: ignore &= etc. */ 2362 ++l; 2363 } 2364 break; 2365 case '!': 2366 m->reln = *l; 2367 ++l; 2368 break; 2369 default: 2370 m->reln = '='; /* the default relation */ 2371 if (*l == 'x' && ((isascii(CAST(unsigned char, l[1])) && 2372 isspace(CAST(unsigned char, l[1]))) || !l[1])) { 2373 m->reln = *l; 2374 ++l; 2375 } 2376 break; 2377 } 2378 /* 2379 * Grab the value part, except for an 'x' reln. 2380 */ 2381 if (m->reln != 'x' && getvalue(ms, m, &l, action)) 2382 return -1; 2383 2384 /* 2385 * TODO finish this macro and start using it! 2386 * #define offsetcheck {if (offset > ms->bytes_max -1) 2387 * magwarn("offset too big"); } 2388 */ 2389 2390 /* 2391 * Now get last part - the description 2392 */ 2393 EATAB; 2394 if (l[0] == '\b') { 2395 ++l; 2396 m->flag |= NOSPACE; 2397 } else if ((l[0] == '\\') && (l[1] == 'b')) { 2398 ++l; 2399 ++l; 2400 m->flag |= NOSPACE; 2401 } 2402 for (i = 0; (m->desc[i++] = *l++) != '\0' && i < sizeof(m->desc); ) 2403 continue; 2404 if (i == sizeof(m->desc)) { 2405 m->desc[sizeof(m->desc) - 1] = '\0'; 2406 if (ms->flags & MAGIC_CHECK) 2407 file_magwarn(ms, "description `%s' truncated", m->desc); 2408 } 2409 2410 /* 2411 * We only do this check while compiling, or if any of the magic 2412 * files were not compiled. 2413 */ 2414 if (ms->flags & MAGIC_CHECK) { 2415 if (check_format(ms, m) == -1) 2416 return -1; 2417 } 2418#ifndef COMPILE_ONLY 2419 if (action == FILE_CHECK) { 2420 file_mdump(m); 2421 } 2422#endif 2423 m->mimetype[0] = '\0'; /* initialise MIME type to none */ 2424 return 0; 2425} 2426 2427/* 2428 * parse a STRENGTH annotation line from magic file, put into magic[index - 1] 2429 * if valid 2430 */ 2431/*ARGSUSED*/ 2432file_private int 2433parse_strength(struct magic_set *ms, struct magic_entry *me, const char *line, 2434 size_t len __attribute__((__unused__))) 2435{ 2436 const char *l = line; 2437 char *el; 2438 unsigned long factor; 2439 char sbuf[512]; 2440 struct magic *m = &me->mp[0]; 2441 2442 if (m->factor_op != FILE_FACTOR_OP_NONE) { 2443 file_magwarn(ms, 2444 "Current entry already has a strength type: %c %d", 2445 m->factor_op, m->factor); 2446 return -1; 2447 } 2448 if (m->type == FILE_NAME) { 2449 file_magwarn(ms, "%s: Strength setting is not supported in " 2450 "\"name\" magic entries", 2451 file_printable(ms, sbuf, sizeof(sbuf), m->value.s, 2452 sizeof(m->value.s))); 2453 return -1; 2454 } 2455 EATAB; 2456 switch (*l) { 2457 case FILE_FACTOR_OP_NONE: 2458 break; 2459 case FILE_FACTOR_OP_PLUS: 2460 case FILE_FACTOR_OP_MINUS: 2461 case FILE_FACTOR_OP_TIMES: 2462 case FILE_FACTOR_OP_DIV: 2463 m->factor_op = *l++; 2464 break; 2465 default: 2466 file_magwarn(ms, "Unknown factor op `%c'", *l); 2467 return -1; 2468 } 2469 EATAB; 2470 factor = strtoul(l, &el, 0); 2471 if (factor > 255) { 2472 file_magwarn(ms, "Too large factor `%lu'", factor); 2473 goto out; 2474 } 2475 if (*el && !isspace(CAST(unsigned char, *el))) { 2476 file_magwarn(ms, "Bad factor `%s'", l); 2477 goto out; 2478 } 2479 m->factor = CAST(uint8_t, factor); 2480 if (m->factor == 0 && m->factor_op == FILE_FACTOR_OP_DIV) { 2481 file_magwarn(ms, "Cannot have factor op `%c' and factor %u", 2482 m->factor_op, m->factor); 2483 goto out; 2484 } 2485 return 0; 2486out: 2487 m->factor_op = FILE_FACTOR_OP_NONE; 2488 m->factor = 0; 2489 return -1; 2490} 2491 2492file_private int 2493goodchar(unsigned char x, const char *extra) 2494{ 2495 return (isascii(x) && isalnum(x)) || strchr(extra, x); 2496} 2497 2498file_private int 2499parse_extra(struct magic_set *ms, struct magic_entry *me, const char *line, 2500 size_t llen, off_t off, size_t len, const char *name, const char *extra, 2501 int nt) 2502{ 2503 size_t i; 2504 const char *l = line; 2505 struct magic *m = &me->mp[me->cont_count == 0 ? 0 : me->cont_count - 1]; 2506 char *buf = CAST(char *, CAST(void *, m)) + off; 2507 2508 if (buf[0] != '\0') { 2509 len = nt ? strlen(buf) : len; 2510 file_magwarn(ms, "Current entry already has a %s type " 2511 "`%.*s', new type `%s'", name, CAST(int, len), buf, l); 2512 return -1; 2513 } 2514 2515 if (*m->desc == '\0') { 2516 file_magwarn(ms, "Current entry does not yet have a " 2517 "description for adding a %s type", name); 2518 return -1; 2519 } 2520 2521 EATAB; 2522 for (i = 0; *l && i < llen && i < len && goodchar(*l, extra); 2523 buf[i++] = *l++) 2524 continue; 2525 2526 if (i == len && *l) { 2527 if (nt) 2528 buf[len - 1] = '\0'; 2529 if (ms->flags & MAGIC_CHECK) 2530 file_magwarn(ms, "%s type `%s' truncated %" 2531 SIZE_T_FORMAT "u", name, line, i); 2532 } else { 2533 if (!isspace(CAST(unsigned char, *l)) && !goodchar(*l, extra)) 2534 file_magwarn(ms, "%s type `%s' has bad char '%c'", 2535 name, line, *l); 2536 if (nt) 2537 buf[i] = '\0'; 2538 } 2539 2540 if (i > 0) 2541 return 0; 2542 2543 file_magerror(ms, "Bad magic entry '%s'", line); 2544 return -1; 2545} 2546 2547/* 2548 * Parse an Apple CREATOR/TYPE annotation from magic file and put it into 2549 * magic[index - 1] 2550 */ 2551file_private int 2552parse_apple(struct magic_set *ms, struct magic_entry *me, const char *line, 2553 size_t len) 2554{ 2555 return parse_extra(ms, me, line, len, 2556 CAST(off_t, offsetof(struct magic, apple)), 2557 sizeof(me->mp[0].apple), "APPLE", "!+-./?", 0); 2558} 2559 2560/* 2561 * Parse a comma-separated list of extensions 2562 */ 2563file_private int 2564parse_ext(struct magic_set *ms, struct magic_entry *me, const char *line, 2565 size_t len) 2566{ 2567 return parse_extra(ms, me, line, len, 2568 CAST(off_t, offsetof(struct magic, ext)), 2569 sizeof(me->mp[0].ext), "EXTENSION", ",!+-/@?_$&~", 0); 2570 /* & for b&w */ 2571 /* ~ for journal~ */ 2572} 2573 2574/* 2575 * parse a MIME annotation line from magic file, put into magic[index - 1] 2576 * if valid 2577 */ 2578file_private int 2579parse_mime(struct magic_set *ms, struct magic_entry *me, const char *line, 2580 size_t len) 2581{ 2582 return parse_extra(ms, me, line, len, 2583 CAST(off_t, offsetof(struct magic, mimetype)), 2584 sizeof(me->mp[0].mimetype), "MIME", "+-/.$?:{}", 1); 2585} 2586 2587file_private int 2588check_format_type(const char *ptr, int type, const char **estr) 2589{ 2590 int quad = 0, h; 2591 size_t len, cnt; 2592 if (*ptr == '\0') { 2593 /* Missing format string; bad */ 2594 *estr = "missing format spec"; 2595 return -1; 2596 } 2597 2598 switch (file_formats[type]) { 2599 case FILE_FMT_QUAD: 2600 quad = 1; 2601 /*FALLTHROUGH*/ 2602 case FILE_FMT_NUM: 2603 if (quad == 0) { 2604 switch (type) { 2605 case FILE_BYTE: 2606 h = 2; 2607 break; 2608 case FILE_SHORT: 2609 case FILE_BESHORT: 2610 case FILE_LESHORT: 2611 h = 1; 2612 break; 2613 case FILE_LONG: 2614 case FILE_BELONG: 2615 case FILE_LELONG: 2616 case FILE_MELONG: 2617 case FILE_LEID3: 2618 case FILE_BEID3: 2619 case FILE_INDIRECT: 2620 h = 0; 2621 break; 2622 default: 2623 fprintf(stderr, "Bad number format %d", type); 2624 abort(); 2625 } 2626 } else 2627 h = 0; 2628 while (*ptr && strchr("-.#", *ptr) != NULL) 2629 ptr++; 2630#define CHECKLEN() do { \ 2631 for (len = cnt = 0; isdigit(CAST(unsigned char, *ptr)); ptr++, cnt++) \ 2632 len = len * 10 + (*ptr - '0'); \ 2633 if (cnt > 5 || len > 1024) \ 2634 goto toolong; \ 2635} while (/*CONSTCOND*/0) 2636 2637 CHECKLEN(); 2638 if (*ptr == '.') 2639 ptr++; 2640 CHECKLEN(); 2641 if (quad) { 2642 if (*ptr++ != 'l') 2643 goto invalid; 2644 if (*ptr++ != 'l') 2645 goto invalid; 2646 } 2647 2648 switch (*ptr++) { 2649#ifdef STRICT_FORMAT /* "long" formats are int formats for us */ 2650 /* so don't accept the 'l' modifier */ 2651 case 'l': 2652 switch (*ptr++) { 2653 case 'i': 2654 case 'd': 2655 case 'u': 2656 case 'o': 2657 case 'x': 2658 case 'X': 2659 if (h == 0) 2660 return 0; 2661 /*FALLTHROUGH*/ 2662 default: 2663 goto invalid; 2664 } 2665 2666 /* 2667 * Don't accept h and hh modifiers. They make writing 2668 * magic entries more complicated, for very little benefit 2669 */ 2670 case 'h': 2671 if (h-- <= 0) 2672 goto invalid; 2673 switch (*ptr++) { 2674 case 'h': 2675 if (h-- <= 0) 2676 goto invalid; 2677 switch (*ptr++) { 2678 case 'i': 2679 case 'd': 2680 case 'u': 2681 case 'o': 2682 case 'x': 2683 case 'X': 2684 return 0; 2685 default: 2686 goto invalid; 2687 } 2688 case 'i': 2689 case 'd': 2690 case 'u': 2691 case 'o': 2692 case 'x': 2693 case 'X': 2694 if (h == 0) 2695 return 0; 2696 /*FALLTHROUGH*/ 2697 default: 2698 goto invalid; 2699 } 2700#endif 2701 case 'c': 2702 if (h == 2) 2703 return 0; 2704 goto invalid; 2705 case 'i': 2706 case 'd': 2707 case 'u': 2708 case 'o': 2709 case 'x': 2710 case 'X': 2711#ifdef STRICT_FORMAT 2712 if (h == 0) 2713 return 0; 2714 /*FALLTHROUGH*/ 2715#else 2716 return 0; 2717#endif 2718 default: 2719 goto invalid; 2720 } 2721 2722 case FILE_FMT_FLOAT: 2723 case FILE_FMT_DOUBLE: 2724 if (*ptr == '-') 2725 ptr++; 2726 if (*ptr == '.') 2727 ptr++; 2728 CHECKLEN(); 2729 if (*ptr == '.') 2730 ptr++; 2731 CHECKLEN(); 2732 switch (*ptr++) { 2733 case 'e': 2734 case 'E': 2735 case 'f': 2736 case 'F': 2737 case 'g': 2738 case 'G': 2739 return 0; 2740 2741 default: 2742 goto invalid; 2743 } 2744 2745 2746 case FILE_FMT_STR: 2747 if (*ptr == '-') 2748 ptr++; 2749 while (isdigit(CAST(unsigned char, *ptr))) 2750 ptr++; 2751 if (*ptr == '.') { 2752 ptr++; 2753 while (isdigit(CAST(unsigned char , *ptr))) 2754 ptr++; 2755 } 2756 2757 switch (*ptr++) { 2758 case 's': 2759 return 0; 2760 default: 2761 goto invalid; 2762 } 2763 2764 default: 2765 /* internal error */ 2766 fprintf(stderr, "Bad file format %d", type); 2767 abort(); 2768 } 2769invalid: 2770 *estr = "not valid"; 2771 return -1; 2772toolong: 2773 *estr = "too long"; 2774 return -1; 2775} 2776 2777/* 2778 * Check that the optional printf format in description matches 2779 * the type of the magic. 2780 */ 2781file_private int 2782check_format(struct magic_set *ms, struct magic *m) 2783{ 2784 char *ptr; 2785 const char *estr; 2786 2787 for (ptr = m->desc; *ptr; ptr++) 2788 if (*ptr == '%') 2789 break; 2790 if (*ptr == '\0') { 2791 /* No format string; ok */ 2792 return 1; 2793 } 2794 2795 assert(file_nformats == file_nnames); 2796 2797 if (m->type >= file_nformats) { 2798 file_magwarn(ms, "Internal error inconsistency between " 2799 "m->type and format strings"); 2800 return -1; 2801 } 2802 if (file_formats[m->type] == FILE_FMT_NONE) { 2803 file_magwarn(ms, "No format string for `%s' with description " 2804 "`%s'", m->desc, file_names[m->type]); 2805 return -1; 2806 } 2807 2808 ptr++; 2809 if (check_format_type(ptr, m->type, &estr) == -1) { 2810 /* 2811 * TODO: this error message is unhelpful if the format 2812 * string is not one character long 2813 */ 2814 file_magwarn(ms, "Printf format is %s for type " 2815 "`%s' in description `%s'", estr, 2816 file_names[m->type], m->desc); 2817 return -1; 2818 } 2819 2820 for (; *ptr; ptr++) { 2821 if (*ptr == '%') { 2822 file_magwarn(ms, 2823 "Too many format strings (should have at most one) " 2824 "for `%s' with description `%s'", 2825 file_names[m->type], m->desc); 2826 return -1; 2827 } 2828 } 2829 return 0; 2830} 2831 2832/* 2833 * Read a numeric value from a pointer, into the value union of a magic 2834 * pointer, according to the magic type. Update the string pointer to point 2835 * just after the number read. Return 0 for success, non-zero for failure. 2836 */ 2837file_private int 2838getvalue(struct magic_set *ms, struct magic *m, const char **p, int action) 2839{ 2840 char *ep; 2841 uint64_t ull; 2842 int y; 2843 2844 switch (m->type) { 2845 case FILE_BESTRING16: 2846 case FILE_LESTRING16: 2847 case FILE_STRING: 2848 case FILE_PSTRING: 2849 case FILE_REGEX: 2850 case FILE_SEARCH: 2851 case FILE_NAME: 2852 case FILE_USE: 2853 case FILE_DER: 2854 case FILE_OCTAL: 2855 *p = getstr(ms, m, *p, action == FILE_COMPILE); 2856 if (*p == NULL) { 2857 if (ms->flags & MAGIC_CHECK) 2858 file_magwarn(ms, "cannot get string from `%s'", 2859 m->value.s); 2860 return -1; 2861 } 2862 if (m->type == FILE_REGEX) { 2863 file_regex_t rx; 2864 int rc = 2865 file_regcomp(ms, &rx, m->value.s, REG_EXTENDED); 2866 if (rc == 0) { 2867 file_regfree(&rx); 2868 } 2869 return rc ? -1 : 0; 2870 } 2871 return 0; 2872 default: 2873 if (m->reln == 'x') 2874 return 0; 2875 break; 2876 } 2877 2878 switch (m->type) { 2879 case FILE_FLOAT: 2880 case FILE_BEFLOAT: 2881 case FILE_LEFLOAT: 2882 errno = 0; 2883#ifdef HAVE_STRTOF 2884 m->value.f = strtof(*p, &ep); 2885#else 2886 m->value.f = (float)strtod(*p, &ep); 2887#endif 2888 if (errno == 0) 2889 *p = ep; 2890 return 0; 2891 case FILE_DOUBLE: 2892 case FILE_BEDOUBLE: 2893 case FILE_LEDOUBLE: 2894 errno = 0; 2895 m->value.d = strtod(*p, &ep); 2896 if (errno == 0) 2897 *p = ep; 2898 return 0; 2899 case FILE_GUID: 2900 if (file_parse_guid(*p, m->value.guid) == -1) 2901 return -1; 2902 *p += FILE_GUID_SIZE - 1; 2903 return 0; 2904 default: 2905 errno = 0; 2906 ull = CAST(uint64_t, strtoull(*p, &ep, 0)); 2907 m->value.q = file_signextend(ms, m, ull); 2908 if (*p == ep) { 2909 file_magwarn(ms, "Unparsable number `%s'", *p); 2910 return -1; 2911 } else { 2912 size_t ts = typesize(m->type); 2913 uint64_t x; 2914 const char *q; 2915 2916 if (ts == FILE_BADSIZE) { 2917 file_magwarn(ms, 2918 "Expected numeric type got `%s'", 2919 type_tbl[m->type].name); 2920 return -1; 2921 } 2922 for (q = *p; isspace(CAST(unsigned char, *q)); q++) 2923 continue; 2924 if (*q == '-' && ull != UINT64_MAX) 2925 ull = -CAST(int64_t, ull); 2926 switch (ts) { 2927 case 1: 2928 x = CAST(uint64_t, ull & ~0xffULL); 2929 y = (x & ~0xffULL) != ~0xffULL; 2930 break; 2931 case 2: 2932 x = CAST(uint64_t, ull & ~0xffffULL); 2933 y = (x & ~0xffffULL) != ~0xffffULL; 2934 break; 2935 case 4: 2936 x = CAST(uint64_t, ull & ~0xffffffffULL); 2937 y = (x & ~0xffffffffULL) != ~0xffffffffULL; 2938 break; 2939 case 8: 2940 x = 0; 2941 y = 0; 2942 break; 2943 default: 2944 fprintf(stderr, "Bad width %zu", ts); 2945 abort(); 2946 } 2947 if (x && y) { 2948 file_magwarn(ms, "Overflow for numeric" 2949 " type `%s' value %#" PRIx64, 2950 type_tbl[m->type].name, ull); 2951 return -1; 2952 } 2953 } 2954 if (errno == 0) { 2955 *p = ep; 2956 eatsize(p); 2957 } 2958 return 0; 2959 } 2960} 2961 2962/* 2963 * Convert a string containing C character escapes. Stop at an unescaped 2964 * space or tab. 2965 * Copy the converted version to "m->value.s", and the length in m->vallen. 2966 * Return updated scan pointer as function result. Warn if set. 2967 */ 2968file_private const char * 2969getstr(struct magic_set *ms, struct magic *m, const char *s, int warn) 2970{ 2971 const char *origs = s; 2972 char *p = m->value.s; 2973 size_t plen = sizeof(m->value.s); 2974 char *origp = p; 2975 char *pmax = p + plen - 1; 2976 int c; 2977 int val; 2978 size_t bracket_nesting = 0; 2979 2980 while ((c = *s++) != '\0') { 2981 if (isspace(CAST(unsigned char, c))) 2982 break; 2983 if (p >= pmax) { 2984 file_error(ms, 0, "string too long: `%s'", origs); 2985 return NULL; 2986 } 2987 if (c != '\\') { 2988 if (c == '[') { 2989 bracket_nesting++; 2990 } 2991 if (c == ']' && bracket_nesting > 0) { 2992 bracket_nesting--; 2993 } 2994 *p++ = CAST(char, c); 2995 continue; 2996 } 2997 switch(c = *s++) { 2998 2999 case '\0': 3000 if (warn) 3001 file_magwarn(ms, "incomplete escape"); 3002 s--; 3003 goto out; 3004 case '.': 3005 if (m->type == FILE_REGEX && 3006 bracket_nesting == 0 && warn) { 3007 file_magwarn(ms, "escaped dot ('.') found, " 3008 "use \\\\. instead"); 3009 } 3010 warn = 0; /* already did */ 3011 /*FALLTHROUGH*/ 3012 case '\t': 3013 if (warn) { 3014 file_magwarn(ms, 3015 "escaped tab found, use \\\\t instead"); 3016 warn = 0; /* already did */ 3017 } 3018 /*FALLTHROUGH*/ 3019 default: 3020 if (warn) { 3021 if (isprint(CAST(unsigned char, c))) { 3022 /* Allow escaping of 3023 * ``relations'' */ 3024 if (strchr("<>&^=!", c) == NULL 3025 && (m->type != FILE_REGEX || 3026 strchr("[]().*?^$|{}", c) 3027 == NULL)) { 3028 file_magwarn(ms, "no " 3029 "need to escape " 3030 "`%c'", c); 3031 } 3032 } else { 3033 file_magwarn(ms, 3034 "unknown escape sequence: " 3035 "\\%03o", c); 3036 } 3037 } 3038 /*FALLTHROUGH*/ 3039 /* space, perhaps force people to use \040? */ 3040 case ' ': 3041#if 0 3042 /* 3043 * Other things people escape, but shouldn't need to, 3044 * so we disallow them 3045 */ 3046 case '\'': 3047 case '"': 3048 case '?': 3049#endif 3050 /* Relations */ 3051 case '>': 3052 case '<': 3053 case '&': 3054 case '^': 3055 case '=': 3056 case '!': 3057 /* and backslash itself */ 3058 case '\\': 3059 *p++ = CAST(char, c); 3060 break; 3061 3062 case 'a': 3063 *p++ = '\a'; 3064 break; 3065 3066 case 'b': 3067 *p++ = '\b'; 3068 break; 3069 3070 case 'f': 3071 *p++ = '\f'; 3072 break; 3073 3074 case 'n': 3075 *p++ = '\n'; 3076 break; 3077 3078 case 'r': 3079 *p++ = '\r'; 3080 break; 3081 3082 case 't': 3083 *p++ = '\t'; 3084 break; 3085 3086 case 'v': 3087 *p++ = '\v'; 3088 break; 3089 3090 /* \ and up to 3 octal digits */ 3091 case '0': 3092 case '1': 3093 case '2': 3094 case '3': 3095 case '4': 3096 case '5': 3097 case '6': 3098 case '7': 3099 val = c - '0'; 3100 c = *s++; /* try for 2 */ 3101 if (c >= '0' && c <= '7') { 3102 val = (val << 3) | (c - '0'); 3103 c = *s++; /* try for 3 */ 3104 if (c >= '0' && c <= '7') 3105 val = (val << 3) | (c-'0'); 3106 else 3107 --s; 3108 } 3109 else 3110 --s; 3111 *p++ = CAST(char, val); 3112 break; 3113 3114 /* \x and up to 2 hex digits */ 3115 case 'x': 3116 val = 'x'; /* Default if no digits */ 3117 c = hextoint(*s++); /* Get next char */ 3118 if (c >= 0) { 3119 val = c; 3120 c = hextoint(*s++); 3121 if (c >= 0) 3122 val = (val << 4) + c; 3123 else 3124 --s; 3125 } else 3126 --s; 3127 *p++ = CAST(char, val); 3128 break; 3129 } 3130 } 3131 --s; 3132out: 3133 *p = '\0'; 3134 m->vallen = CAST(unsigned char, (p - origp)); 3135 if (m->type == FILE_PSTRING) { 3136 size_t l = file_pstring_length_size(ms, m); 3137 if (l == FILE_BADSIZE) 3138 return NULL; 3139 m->vallen += CAST(unsigned char, l); 3140 } 3141 return s; 3142} 3143 3144 3145/* Single hex char to int; -1 if not a hex char. */ 3146file_private int 3147hextoint(int c) 3148{ 3149 if (!isascii(CAST(unsigned char, c))) 3150 return -1; 3151 if (isdigit(CAST(unsigned char, c))) 3152 return c - '0'; 3153 if ((c >= 'a') && (c <= 'f')) 3154 return c + 10 - 'a'; 3155 if (( c>= 'A') && (c <= 'F')) 3156 return c + 10 - 'A'; 3157 return -1; 3158} 3159 3160 3161/* 3162 * Print a string containing C character escapes. 3163 */ 3164file_protected void 3165file_showstr(FILE *fp, const char *s, size_t len) 3166{ 3167 char c; 3168 3169 for (;;) { 3170 if (len == FILE_BADSIZE) { 3171 c = *s++; 3172 if (c == '\0') 3173 break; 3174 } 3175 else { 3176 if (len-- == 0) 3177 break; 3178 c = *s++; 3179 } 3180 if (c >= 040 && c <= 0176) /* TODO isprint && !iscntrl */ 3181 (void) fputc(c, fp); 3182 else { 3183 (void) fputc('\\', fp); 3184 switch (c) { 3185 case '\a': 3186 (void) fputc('a', fp); 3187 break; 3188 3189 case '\b': 3190 (void) fputc('b', fp); 3191 break; 3192 3193 case '\f': 3194 (void) fputc('f', fp); 3195 break; 3196 3197 case '\n': 3198 (void) fputc('n', fp); 3199 break; 3200 3201 case '\r': 3202 (void) fputc('r', fp); 3203 break; 3204 3205 case '\t': 3206 (void) fputc('t', fp); 3207 break; 3208 3209 case '\v': 3210 (void) fputc('v', fp); 3211 break; 3212 3213 default: 3214 (void) fprintf(fp, "%.3o", c & 0377); 3215 break; 3216 } 3217 } 3218 } 3219} 3220 3221/* 3222 * eatsize(): Eat the size spec from a number [eg. 10UL] 3223 */ 3224file_private void 3225eatsize(const char **p) 3226{ 3227 const char *l = *p; 3228 3229 if (LOWCASE(*l) == 'u') 3230 l++; 3231 3232 switch (LOWCASE(*l)) { 3233 case 'l': /* long */ 3234 case 's': /* short */ 3235 case 'h': /* short */ 3236 case 'b': /* char/byte */ 3237 case 'c': /* char/byte */ 3238 l++; 3239 /*FALLTHROUGH*/ 3240 default: 3241 break; 3242 } 3243 3244 *p = l; 3245} 3246 3247/* 3248 * handle a buffer containing a compiled file. 3249 */ 3250file_private struct magic_map * 3251apprentice_buf(struct magic_set *ms, struct magic *buf, size_t len) 3252{ 3253 struct magic_map *map; 3254 3255 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 3256 file_oomem(ms, sizeof(*map)); 3257 return NULL; 3258 } 3259 map->len = len; 3260 map->p = buf; 3261 map->type = MAP_TYPE_USER; 3262 if (check_buffer(ms, map, "buffer") != 0) { 3263 apprentice_unmap(map); 3264 return NULL; 3265 } 3266 return map; 3267} 3268 3269/* 3270 * handle a compiled file. 3271 */ 3272 3273file_private struct magic_map * 3274apprentice_map(struct magic_set *ms, const char *fn) 3275{ 3276 int fd; 3277 struct stat st; 3278 char *dbname = NULL; 3279 struct magic_map *map; 3280 struct magic_map *rv = NULL; 3281 3282 fd = -1; 3283 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) { 3284 file_oomem(ms, sizeof(*map)); 3285 goto error; 3286 } 3287 map->type = MAP_TYPE_USER; /* unspecified */ 3288 3289 dbname = mkdbname(ms, fn, 0); 3290 if (dbname == NULL) 3291 goto error; 3292 3293 if ((fd = open(dbname, O_RDONLY|O_BINARY)) == -1) 3294 goto error; 3295 3296 if (fstat(fd, &st) == -1) { 3297 file_error(ms, errno, "cannot stat `%s'", dbname); 3298 goto error; 3299 } 3300 if (st.st_size < 8 || st.st_size > maxoff_t()) { 3301 file_error(ms, 0, "file `%s' is too %s", dbname, 3302 st.st_size < 8 ? "small" : "large"); 3303 goto error; 3304 } 3305 3306 map->len = CAST(size_t, st.st_size); 3307#ifdef QUICK 3308 map->type = MAP_TYPE_MMAP; 3309 if ((map->p = mmap(0, CAST(size_t, st.st_size), PROT_READ|PROT_WRITE, 3310 MAP_PRIVATE|MAP_FILE, fd, CAST(off_t, 0))) == MAP_FAILED) { 3311 file_error(ms, errno, "cannot map `%s'", dbname); 3312 goto error; 3313 } 3314#else 3315 map->type = MAP_TYPE_MALLOC; 3316 if ((map->p = CAST(void *, malloc(map->len))) == NULL) { 3317 file_oomem(ms, map->len); 3318 goto error; 3319 } 3320 if (read(fd, map->p, map->len) != (ssize_t)map->len) { 3321 file_badread(ms); 3322 goto error; 3323 } 3324#endif 3325 (void)close(fd); 3326 fd = -1; 3327 3328 if (check_buffer(ms, map, dbname) != 0) { 3329 goto error; 3330 } 3331#ifdef QUICK 3332 if (mprotect(map->p, CAST(size_t, st.st_size), PROT_READ) == -1) { 3333 file_error(ms, errno, "cannot mprotect `%s'", dbname); 3334 goto error; 3335 } 3336#endif 3337 3338 free(dbname); 3339 return map; 3340 3341error: 3342 if (fd != -1) 3343 (void)close(fd); 3344 apprentice_unmap(map); 3345 free(dbname); 3346 return rv; 3347} 3348 3349file_private int 3350check_buffer(struct magic_set *ms, struct magic_map *map, const char *dbname) 3351{ 3352 uint32_t *ptr; 3353 uint32_t entries, nentries; 3354 uint32_t version; 3355 int i, needsbyteswap; 3356 3357 ptr = CAST(uint32_t *, map->p); 3358 if (*ptr != MAGICNO) { 3359 if (swap4(*ptr) != MAGICNO) { 3360 file_error(ms, 0, "bad magic in `%s'", dbname); 3361 return -1; 3362 } 3363 needsbyteswap = 1; 3364 } else 3365 needsbyteswap = 0; 3366 if (needsbyteswap) 3367 version = swap4(ptr[1]); 3368 else 3369 version = ptr[1]; 3370 if (version != VERSIONNO) { 3371 file_error(ms, 0, "File %s supports only version %d magic " 3372 "files. `%s' is version %d", VERSION, 3373 VERSIONNO, dbname, version); 3374 return -1; 3375 } 3376 entries = CAST(uint32_t, map->len / sizeof(struct magic)); 3377 if ((entries * sizeof(struct magic)) != map->len) { 3378 file_error(ms, 0, "Size of `%s' %" SIZE_T_FORMAT "u is not " 3379 "a multiple of %" SIZE_T_FORMAT "u", 3380 dbname, map->len, sizeof(struct magic)); 3381 return -1; 3382 } 3383 map->magic[0] = CAST(struct magic *, map->p) + 1; 3384 nentries = 0; 3385 for (i = 0; i < MAGIC_SETS; i++) { 3386 if (needsbyteswap) 3387 map->nmagic[i] = swap4(ptr[i + 2]); 3388 else 3389 map->nmagic[i] = ptr[i + 2]; 3390 if (i != MAGIC_SETS - 1) 3391 map->magic[i + 1] = map->magic[i] + map->nmagic[i]; 3392 nentries += map->nmagic[i]; 3393 } 3394 if (entries != nentries + 1) { 3395 file_error(ms, 0, "Inconsistent entries in `%s' %u != %u", 3396 dbname, entries, nentries + 1); 3397 return -1; 3398 } 3399 if (needsbyteswap) 3400 for (i = 0; i < MAGIC_SETS; i++) 3401 byteswap(map->magic[i], map->nmagic[i]); 3402 return 0; 3403} 3404 3405/* 3406 * handle an mmaped file. 3407 */ 3408file_private int 3409apprentice_compile(struct magic_set *ms, struct magic_map *map, const char *fn) 3410{ 3411 static const size_t nm = sizeof(*map->nmagic) * MAGIC_SETS; 3412 static const size_t m = sizeof(**map->magic); 3413 int fd = -1; 3414 size_t len; 3415 char *dbname; 3416 int rv = -1; 3417 uint32_t i; 3418 union { 3419 struct magic m; 3420 uint32_t h[2 + MAGIC_SETS]; 3421 } hdr; 3422 3423 dbname = mkdbname(ms, fn, 1); 3424 3425 if (dbname == NULL) 3426 goto out; 3427 3428 if ((fd = open(dbname, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY, 0644)) == -1) 3429 { 3430 file_error(ms, errno, "cannot open `%s'", dbname); 3431 goto out; 3432 } 3433 memset(&hdr, 0, sizeof(hdr)); 3434 hdr.h[0] = MAGICNO; 3435 hdr.h[1] = VERSIONNO; 3436 memcpy(hdr.h + 2, map->nmagic, nm); 3437 3438 if (write(fd, &hdr, sizeof(hdr)) != CAST(ssize_t, sizeof(hdr))) { 3439 file_error(ms, errno, "error writing `%s'", dbname); 3440 goto out2; 3441 } 3442 3443 for (i = 0; i < MAGIC_SETS; i++) { 3444 len = m * map->nmagic[i]; 3445 if (write(fd, map->magic[i], len) != CAST(ssize_t, len)) { 3446 file_error(ms, errno, "error writing `%s'", dbname); 3447 goto out2; 3448 } 3449 } 3450 3451 rv = 0; 3452out2: 3453 if (fd != -1) 3454 (void)close(fd); 3455out: 3456 apprentice_unmap(map); 3457 free(dbname); 3458 return rv; 3459} 3460 3461file_private const char ext[] = ".mgc"; 3462/* 3463 * make a dbname 3464 */ 3465file_private char * 3466mkdbname(struct magic_set *ms, const char *fn, int strip) 3467{ 3468 const char *p, *q; 3469 char *buf; 3470 3471 if (strip) { 3472 if ((p = strrchr(fn, '/')) != NULL) 3473 fn = ++p; 3474 } 3475 3476 for (q = fn; *q; q++) 3477 continue; 3478 /* Look for .mgc */ 3479 for (p = ext + sizeof(ext) - 1; p >= ext && q >= fn; p--, q--) 3480 if (*p != *q) 3481 break; 3482 3483 /* Did not find .mgc, restore q */ 3484 if (p >= ext) 3485 while (*q) 3486 q++; 3487 3488 q++; 3489 /* Compatibility with old code that looked in .mime */ 3490 if (ms->flags & MAGIC_MIME) { 3491 if (asprintf(&buf, "%.*s.mime%s", CAST(int, q - fn), fn, ext) 3492 < 0) 3493 return NULL; 3494 if (access(buf, R_OK) != -1) { 3495 ms->flags &= MAGIC_MIME_TYPE; 3496 return buf; 3497 } 3498 free(buf); 3499 } 3500 if (asprintf(&buf, "%.*s%s", CAST(int, q - fn), fn, ext) < 0) 3501 return NULL; 3502 3503 /* Compatibility with old code that looked in .mime */ 3504 if (strstr(fn, ".mime") != NULL) 3505 ms->flags &= MAGIC_MIME_TYPE; 3506 return buf; 3507} 3508 3509/* 3510 * Byteswap an mmap'ed file if needed 3511 */ 3512file_private void 3513byteswap(struct magic *magic, uint32_t nmagic) 3514{ 3515 uint32_t i; 3516 for (i = 0; i < nmagic; i++) 3517 bs1(&magic[i]); 3518} 3519 3520#if !defined(HAVE_BYTESWAP_H) && !defined(HAVE_SYS_BSWAP_H) 3521/* 3522 * swap a short 3523 */ 3524file_private uint16_t 3525swap2(uint16_t sv) 3526{ 3527 uint16_t rv; 3528 uint8_t *s = RCAST(uint8_t *, RCAST(void *, &sv)); 3529 uint8_t *d = RCAST(uint8_t *, RCAST(void *, &rv)); 3530 d[0] = s[1]; 3531 d[1] = s[0]; 3532 return rv; 3533} 3534 3535/* 3536 * swap an int 3537 */ 3538file_private uint32_t 3539swap4(uint32_t sv) 3540{ 3541 uint32_t rv; 3542 uint8_t *s = RCAST(uint8_t *, RCAST(void *, &sv)); 3543 uint8_t *d = RCAST(uint8_t *, RCAST(void *, &rv)); 3544 d[0] = s[3]; 3545 d[1] = s[2]; 3546 d[2] = s[1]; 3547 d[3] = s[0]; 3548 return rv; 3549} 3550 3551/* 3552 * swap a quad 3553 */ 3554file_private uint64_t 3555swap8(uint64_t sv) 3556{ 3557 uint64_t rv; 3558 uint8_t *s = RCAST(uint8_t *, RCAST(void *, &sv)); 3559 uint8_t *d = RCAST(uint8_t *, RCAST(void *, &rv)); 3560# if 0 3561 d[0] = s[3]; 3562 d[1] = s[2]; 3563 d[2] = s[1]; 3564 d[3] = s[0]; 3565 d[4] = s[7]; 3566 d[5] = s[6]; 3567 d[6] = s[5]; 3568 d[7] = s[4]; 3569# else 3570 d[0] = s[7]; 3571 d[1] = s[6]; 3572 d[2] = s[5]; 3573 d[3] = s[4]; 3574 d[4] = s[3]; 3575 d[5] = s[2]; 3576 d[6] = s[1]; 3577 d[7] = s[0]; 3578# endif 3579 return rv; 3580} 3581#endif 3582 3583file_protected uintmax_t 3584file_varint2uintmax_t(const unsigned char *us, int t, size_t *l) 3585{ 3586 uintmax_t x = 0; 3587 const unsigned char *c; 3588 if (t == FILE_LEVARINT) { 3589 for (c = us; *c; c++) { 3590 if ((*c & 0x80) == 0) 3591 break; 3592 } 3593 if (l) 3594 *l = c - us + 1; 3595 for (; c >= us; c--) { 3596 x |= *c & 0x7f; 3597 x <<= 7; 3598 } 3599 } else { 3600 for (c = us; *c; c++) { 3601 x |= *c & 0x7f; 3602 if ((*c & 0x80) == 0) 3603 break; 3604 x <<= 7; 3605 } 3606 if (l) 3607 *l = c - us + 1; 3608 } 3609 return x; 3610} 3611 3612 3613/* 3614 * byteswap a single magic entry 3615 */ 3616file_private void 3617bs1(struct magic *m) 3618{ 3619 m->cont_level = swap2(m->cont_level); 3620 m->offset = swap4(CAST(uint32_t, m->offset)); 3621 m->in_offset = swap4(CAST(uint32_t, m->in_offset)); 3622 m->lineno = swap4(CAST(uint32_t, m->lineno)); 3623 if (IS_STRING(m->type)) { 3624 m->str_range = swap4(m->str_range); 3625 m->str_flags = swap4(m->str_flags); 3626 } 3627 else { 3628 m->value.q = swap8(m->value.q); 3629 m->num_mask = swap8(m->num_mask); 3630 } 3631} 3632 3633file_protected size_t 3634file_pstring_length_size(struct magic_set *ms, const struct magic *m) 3635{ 3636 switch (m->str_flags & PSTRING_LEN) { 3637 case PSTRING_1_LE: 3638 return 1; 3639 case PSTRING_2_LE: 3640 case PSTRING_2_BE: 3641 return 2; 3642 case PSTRING_4_LE: 3643 case PSTRING_4_BE: 3644 return 4; 3645 default: 3646 file_error(ms, 0, "corrupt magic file " 3647 "(bad pascal string length %d)", 3648 m->str_flags & PSTRING_LEN); 3649 return FILE_BADSIZE; 3650 } 3651} 3652file_protected size_t 3653file_pstring_get_length(struct magic_set *ms, const struct magic *m, 3654 const char *ss) 3655{ 3656 size_t len = 0; 3657 const unsigned char *s = RCAST(const unsigned char *, ss); 3658 unsigned int s3, s2, s1, s0; 3659 3660 switch (m->str_flags & PSTRING_LEN) { 3661 case PSTRING_1_LE: 3662 len = *s; 3663 break; 3664 case PSTRING_2_LE: 3665 s0 = s[0]; 3666 s1 = s[1]; 3667 len = (s1 << 8) | s0; 3668 break; 3669 case PSTRING_2_BE: 3670 s0 = s[0]; 3671 s1 = s[1]; 3672 len = (s0 << 8) | s1; 3673 break; 3674 case PSTRING_4_LE: 3675 s0 = s[0]; 3676 s1 = s[1]; 3677 s2 = s[2]; 3678 s3 = s[3]; 3679 len = (s3 << 24) | (s2 << 16) | (s1 << 8) | s0; 3680 break; 3681 case PSTRING_4_BE: 3682 s0 = s[0]; 3683 s1 = s[1]; 3684 s2 = s[2]; 3685 s3 = s[3]; 3686 len = (s0 << 24) | (s1 << 16) | (s2 << 8) | s3; 3687 break; 3688 default: 3689 file_error(ms, 0, "corrupt magic file " 3690 "(bad pascal string length %d)", 3691 m->str_flags & PSTRING_LEN); 3692 return FILE_BADSIZE; 3693 } 3694 3695 if (m->str_flags & PSTRING_LENGTH_INCLUDES_ITSELF) { 3696 size_t l = file_pstring_length_size(ms, m); 3697 if (l == FILE_BADSIZE) 3698 return l; 3699 len -= l; 3700 } 3701 3702 return len; 3703} 3704 3705file_protected int 3706file_magicfind(struct magic_set *ms, const char *name, struct mlist *v) 3707{ 3708 uint32_t i, j; 3709 struct mlist *mlist, *ml; 3710 3711 mlist = ms->mlist[1]; 3712 3713 for (ml = mlist->next; ml != mlist; ml = ml->next) { 3714 struct magic *ma = ml->magic; 3715 for (i = 0; i < ml->nmagic; i++) { 3716 if (ma[i].type != FILE_NAME) 3717 continue; 3718 if (strcmp(ma[i].value.s, name) == 0) { 3719 v->magic = &ma[i]; 3720 v->magic_rxcomp = &(ml->magic_rxcomp[i]); 3721 for (j = i + 1; j < ml->nmagic; j++) 3722 if (ma[j].cont_level == 0) 3723 break; 3724 v->nmagic = j - i; 3725 return 0; 3726 } 3727 } 3728 } 3729 return -1; 3730} 3731