magic.c revision 1.16
1/* $NetBSD: magic.c,v 1.16 2021/04/09 19:11:42 christos Exp $ */ 2 3/* 4 * Copyright (c) Christos Zoulas 2003. 5 * All Rights Reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice immediately at the beginning of the file, without modification, 12 * this list of conditions, and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR 21 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 */ 29 30#ifdef WIN32 31#include <windows.h> 32#include <shlwapi.h> 33#endif 34 35#include "file.h" 36 37#ifndef lint 38#if 0 39FILE_RCSID("@(#)$File: magic.c,v 1.114 2021/02/05 21:33:49 christos Exp $") 40#else 41__RCSID("$NetBSD: magic.c,v 1.16 2021/04/09 19:11:42 christos Exp $"); 42#endif 43#endif /* lint */ 44 45#include "magic.h" 46 47#include <stdlib.h> 48#include <unistd.h> 49#include <string.h> 50#ifdef QUICK 51#include <sys/mman.h> 52#endif 53#include <limits.h> /* for PIPE_BUF */ 54 55#if defined(HAVE_UTIMES) 56# include <sys/time.h> 57#elif defined(HAVE_UTIME) 58# if defined(HAVE_SYS_UTIME_H) 59# include <sys/utime.h> 60# elif defined(HAVE_UTIME_H) 61# include <utime.h> 62# endif 63#endif 64 65#ifdef HAVE_UNISTD_H 66#include <unistd.h> /* for read() */ 67#endif 68 69#ifndef PIPE_BUF 70/* Get the PIPE_BUF from pathconf */ 71#ifdef _PC_PIPE_BUF 72#define PIPE_BUF pathconf(".", _PC_PIPE_BUF) 73#else 74#define PIPE_BUF 512 75#endif 76#endif 77 78private void close_and_restore(const struct magic_set *, const char *, int, 79 const struct stat *); 80private int unreadable_info(struct magic_set *, mode_t, const char *); 81private const char* get_default_magic(void); 82#ifndef COMPILE_ONLY 83private const char *file_or_fd(struct magic_set *, const char *, int); 84#endif 85 86#ifndef STDIN_FILENO 87#define STDIN_FILENO 0 88#endif 89 90#ifdef WIN32 91/* HINSTANCE of this shared library. Needed for get_default_magic() */ 92static HINSTANCE _w32_dll_instance = NULL; 93 94static void 95_w32_append_path(char **hmagicpath, const char *fmt, ...) 96{ 97 char *tmppath; 98 char *newpath; 99 va_list ap; 100 101 va_start(ap, fmt); 102 if (vasprintf(&tmppath, fmt, ap) < 0) { 103 va_end(ap); 104 return; 105 } 106 va_end(ap); 107 108 if (access(tmppath, R_OK) == -1) 109 goto out; 110 111 if (*hmagicpath == NULL) { 112 *hmagicpath = tmppath; 113 return; 114 } 115 116 if (asprintf(&newpath, "%s%c%s", *hmagicpath, PATHSEP, tmppath) < 0) 117 goto out; 118 119 free(*hmagicpath); 120 free(tmppath); 121 *hmagicpath = newpath; 122 return; 123out: 124 free(tmppath); 125} 126 127static void 128_w32_get_magic_relative_to(char **hmagicpath, HINSTANCE module) 129{ 130 static const char *trypaths[] = { 131 "%s/share/misc/magic.mgc", 132 "%s/magic.mgc", 133 }; 134 LPSTR dllpath; 135 size_t sp; 136 137 dllpath = calloc(MAX_PATH + 1, sizeof(*dllpath)); 138 139 if (!GetModuleFileNameA(module, dllpath, MAX_PATH)) 140 goto out; 141 142 PathRemoveFileSpecA(dllpath); 143 144 if (module) { 145 char exepath[MAX_PATH]; 146 GetModuleFileNameA(NULL, exepath, MAX_PATH); 147 PathRemoveFileSpecA(exepath); 148 if (stricmp(exepath, dllpath) == 0) 149 goto out; 150 } 151 152 sp = strlen(dllpath); 153 if (sp > 3 && stricmp(&dllpath[sp - 3], "bin") == 0) { 154 _w32_append_path(hmagicpath, 155 "%s/../share/misc/magic.mgc", dllpath); 156 goto out; 157 } 158 159 for (sp = 0; sp < __arraycount(trypaths); sp++) 160 _w32_append_path(hmagicpath, trypaths[sp], dllpath); 161out: 162 free(dllpath); 163} 164 165/* Placate GCC by offering a sacrificial previous prototype */ 166BOOL WINAPI DllMain(HINSTANCE, DWORD, LPVOID); 167 168BOOL WINAPI 169DllMain(HINSTANCE hinstDLL, DWORD fdwReason, 170 LPVOID lpvReserved __attribute__((__unused__))) 171{ 172 if (fdwReason == DLL_PROCESS_ATTACH) 173 _w32_dll_instance = hinstDLL; 174 return 1; 175} 176#endif 177 178private const char * 179get_default_magic(void) 180{ 181 static const char hmagic[] = "/.magic/magic.mgc"; 182 static char *default_magic; 183 char *home, *hmagicpath; 184 185#ifndef WIN32 186 struct stat st; 187 188 if (default_magic) { 189 free(default_magic); 190 default_magic = NULL; 191 } 192 if ((home = getenv("HOME")) == NULL) 193 return MAGIC; 194 195 if (asprintf(&hmagicpath, "%s/.magic.mgc", home) < 0) 196 return MAGIC; 197 if (stat(hmagicpath, &st) == -1) { 198 free(hmagicpath); 199 if (asprintf(&hmagicpath, "%s/.magic", home) < 0) 200 return MAGIC; 201 if (stat(hmagicpath, &st) == -1) 202 goto out; 203 if (S_ISDIR(st.st_mode)) { 204 free(hmagicpath); 205 if (asprintf(&hmagicpath, "%s/%s", home, hmagic) < 0) 206 return MAGIC; 207 if (access(hmagicpath, R_OK) == -1) 208 goto out; 209 } 210 } 211 212 if (asprintf(&default_magic, "%s:%s", hmagicpath, MAGIC) < 0) 213 goto out; 214 free(hmagicpath); 215 return default_magic; 216out: 217 default_magic = NULL; 218 free(hmagicpath); 219 return MAGIC; 220#else 221 hmagicpath = NULL; 222 223 if (default_magic) { 224 free(default_magic); 225 default_magic = NULL; 226 } 227 228 /* First, try to get a magic file from user-application data */ 229 if ((home = getenv("LOCALAPPDATA")) != NULL) 230 _w32_append_path(&hmagicpath, "%s%s", home, hmagic); 231 232 /* Second, try to get a magic file from the user profile data */ 233 if ((home = getenv("USERPROFILE")) != NULL) 234 _w32_append_path(&hmagicpath, 235 "%s/Local Settings/Application Data%s", home, hmagic); 236 237 /* Third, try to get a magic file from Common Files */ 238 if ((home = getenv("COMMONPROGRAMFILES")) != NULL) 239 _w32_append_path(&hmagicpath, "%s%s", home, hmagic); 240 241 /* Fourth, try to get magic file relative to exe location */ 242 _w32_get_magic_relative_to(&hmagicpath, NULL); 243 244 /* Fifth, try to get magic file relative to dll location */ 245 _w32_get_magic_relative_to(&hmagicpath, _w32_dll_instance); 246 247 /* Avoid MAGIC constant - it likely points to a file within MSys tree */ 248 default_magic = hmagicpath; 249 return default_magic; 250#endif 251} 252 253public const char * 254magic_getpath(const char *magicfile, int action) 255{ 256 if (magicfile != NULL) 257 return magicfile; 258 259 magicfile = getenv("MAGIC"); 260 if (magicfile != NULL) 261 return magicfile; 262 263 return action == FILE_LOAD ? get_default_magic() : MAGIC; 264} 265 266public struct magic_set * 267magic_open(int flags) 268{ 269 return file_ms_alloc(flags); 270} 271 272private int 273unreadable_info(struct magic_set *ms, mode_t md, const char *file) 274{ 275 if (file) { 276 /* We cannot open it, but we were able to stat it. */ 277 if (access(file, W_OK) == 0) 278 if (file_printf(ms, "writable, ") == -1) 279 return -1; 280 if (access(file, X_OK) == 0) 281 if (file_printf(ms, "executable, ") == -1) 282 return -1; 283 } 284 if (S_ISREG(md)) 285 if (file_printf(ms, "regular file, ") == -1) 286 return -1; 287 if (file_printf(ms, "no read permission") == -1) 288 return -1; 289 return 0; 290} 291 292public void 293magic_close(struct magic_set *ms) 294{ 295 if (ms == NULL) 296 return; 297 file_ms_free(ms); 298} 299 300/* 301 * load a magic file 302 */ 303public int 304magic_load(struct magic_set *ms, const char *magicfile) 305{ 306 if (ms == NULL) 307 return -1; 308 return file_apprentice(ms, magicfile, FILE_LOAD); 309} 310 311#ifndef COMPILE_ONLY 312/* 313 * Install a set of compiled magic buffers. 314 */ 315public int 316magic_load_buffers(struct magic_set *ms, void **bufs, size_t *sizes, 317 size_t nbufs) 318{ 319 if (ms == NULL) 320 return -1; 321 return buffer_apprentice(ms, RCAST(struct magic **, bufs), 322 sizes, nbufs); 323} 324#endif 325 326public int 327magic_compile(struct magic_set *ms, const char *magicfile) 328{ 329 if (ms == NULL) 330 return -1; 331 return file_apprentice(ms, magicfile, FILE_COMPILE); 332} 333 334public int 335magic_check(struct magic_set *ms, const char *magicfile) 336{ 337 if (ms == NULL) 338 return -1; 339 return file_apprentice(ms, magicfile, FILE_CHECK); 340} 341 342public int 343magic_list(struct magic_set *ms, const char *magicfile) 344{ 345 if (ms == NULL) 346 return -1; 347 return file_apprentice(ms, magicfile, FILE_LIST); 348} 349 350private void 351close_and_restore(const struct magic_set *ms, const char *name, int fd, 352 const struct stat *sb) 353{ 354 if (fd == STDIN_FILENO || name == NULL) 355 return; 356 (void) close(fd); 357 358 if ((ms->flags & MAGIC_PRESERVE_ATIME) != 0) { 359 /* 360 * Try to restore access, modification times if read it. 361 * This is really *bad* because it will modify the status 362 * time of the file... And of course this will affect 363 * backup programs 364 */ 365#ifdef HAVE_UTIMES 366 struct timeval utsbuf[2]; 367 (void)memset(utsbuf, 0, sizeof(utsbuf)); 368 utsbuf[0].tv_sec = sb->st_atime; 369 utsbuf[1].tv_sec = sb->st_mtime; 370 371 (void) utimes(name, utsbuf); /* don't care if loses */ 372#elif defined(HAVE_UTIME_H) || defined(HAVE_SYS_UTIME_H) 373 struct utimbuf utbuf; 374 375 (void)memset(&utbuf, 0, sizeof(utbuf)); 376 utbuf.actime = sb->st_atime; 377 utbuf.modtime = sb->st_mtime; 378 (void) utime(name, &utbuf); /* don't care if loses */ 379#endif 380 } 381} 382 383#ifndef COMPILE_ONLY 384 385/* 386 * find type of descriptor 387 */ 388public const char * 389magic_descriptor(struct magic_set *ms, int fd) 390{ 391 if (ms == NULL) 392 return NULL; 393 return file_or_fd(ms, NULL, fd); 394} 395 396/* 397 * find type of named file 398 */ 399public const char * 400magic_file(struct magic_set *ms, const char *inname) 401{ 402 if (ms == NULL) 403 return NULL; 404 return file_or_fd(ms, inname, STDIN_FILENO); 405} 406 407private const char * 408file_or_fd(struct magic_set *ms, const char *inname, int fd) 409{ 410 int rv = -1; 411 unsigned char *buf; 412 struct stat sb; 413 ssize_t nbytes = 0; /* number of bytes read from a datafile */ 414 int ispipe = 0; 415 int okstat = 0; 416 off_t pos = CAST(off_t, -1); 417 418 if (file_reset(ms, 1) == -1) 419 goto out; 420 421 /* 422 * one extra for terminating '\0', and 423 * some overlapping space for matches near EOF 424 */ 425#define SLOP (1 + sizeof(union VALUETYPE)) 426 if ((buf = CAST(unsigned char *, malloc(ms->bytes_max + SLOP))) == NULL) 427 return NULL; 428 429 switch (file_fsmagic(ms, inname, &sb)) { 430 case -1: /* error */ 431 goto done; 432 case 0: /* nothing found */ 433 break; 434 default: /* matched it and printed type */ 435 rv = 0; 436 goto done; 437 } 438 439#ifdef WIN32 440 /* Place stdin in binary mode, so EOF (Ctrl+Z) doesn't stop early. */ 441 if (fd == STDIN_FILENO) 442 _setmode(STDIN_FILENO, O_BINARY); 443#endif 444 if (inname != NULL) { 445 int flags = O_RDONLY|O_BINARY|O_NONBLOCK|O_CLOEXEC; 446 errno = 0; 447 if ((fd = open(inname, flags)) < 0) { 448 okstat = stat(inname, &sb) == 0; 449 if (okstat && S_ISFIFO(sb.st_mode)) 450 ispipe = 1; 451#ifdef WIN32 452 /* 453 * Can't stat, can't open. It may have been opened in 454 * fsmagic, so if the user doesn't have read permission, 455 * allow it to say so; otherwise an error was probably 456 * displayed in fsmagic. 457 */ 458 if (!okstat && errno == EACCES) { 459 sb.st_mode = S_IFBLK; 460 okstat = 1; 461 } 462#endif 463 if (okstat && 464 unreadable_info(ms, sb.st_mode, inname) == -1) 465 goto done; 466 rv = 0; 467 goto done; 468 } 469#if O_CLOEXEC == 0 470 (void)fcntl(fd, F_SETFD, FD_CLOEXEC); 471#endif 472 } 473 474 if (fd != -1) { 475 okstat = fstat(fd, &sb) == 0; 476 if (okstat && S_ISFIFO(sb.st_mode)) 477 ispipe = 1; 478 if (inname == NULL) 479 pos = lseek(fd, CAST(off_t, 0), SEEK_CUR); 480 } 481 482 /* 483 * try looking at the first ms->bytes_max bytes 484 */ 485 if (ispipe) { 486 if (fd != -1) { 487 ssize_t r = 0; 488 489 while ((r = sread(fd, RCAST(void *, &buf[nbytes]), 490 CAST(size_t, ms->bytes_max - nbytes), 1)) > 0) { 491 nbytes += r; 492 if (r < PIPE_BUF) break; 493 } 494 } 495 496 if (nbytes == 0 && inname) { 497 /* We can not read it, but we were able to stat it. */ 498 if (unreadable_info(ms, sb.st_mode, inname) == -1) 499 goto done; 500 rv = 0; 501 goto done; 502 } 503 504 } else if (fd != -1) { 505 /* Windows refuses to read from a big console buffer. */ 506 size_t howmany = 507#if defined(WIN32) 508 _isatty(fd) ? 8 * 1024 : 509#endif 510 ms->bytes_max; 511 if ((nbytes = read(fd, RCAST(void *, buf), howmany)) == -1) { 512 if (inname == NULL && fd != STDIN_FILENO) 513 file_error(ms, errno, "cannot read fd %d", fd); 514 else 515 file_error(ms, errno, "cannot read `%s'", 516 inname == NULL ? "/dev/stdin" : inname); 517 goto done; 518 } 519 } 520 521 (void)memset(buf + nbytes, 0, SLOP); /* NUL terminate */ 522 if (file_buffer(ms, fd, okstat ? &sb : NULL, inname, buf, CAST(size_t, nbytes)) == -1) 523 goto done; 524 rv = 0; 525done: 526 free(buf); 527 if (fd != -1) { 528 if (pos != CAST(off_t, -1)) 529 (void)lseek(fd, pos, SEEK_SET); 530 close_and_restore(ms, inname, fd, &sb); 531 } 532out: 533 return rv == 0 ? file_getbuffer(ms) : NULL; 534} 535 536 537public const char * 538magic_buffer(struct magic_set *ms, const void *buf, size_t nb) 539{ 540 if (ms == NULL) 541 return NULL; 542 if (file_reset(ms, 1) == -1) 543 return NULL; 544 /* 545 * The main work is done here! 546 * We have the file name and/or the data buffer to be identified. 547 */ 548 if (file_buffer(ms, -1, NULL, NULL, buf, nb) == -1) { 549 return NULL; 550 } 551 return file_getbuffer(ms); 552} 553#endif 554 555public const char * 556magic_error(struct magic_set *ms) 557{ 558 if (ms == NULL) 559 return "Magic database is not open"; 560 return (ms->event_flags & EVENT_HAD_ERR) ? ms->o.buf : NULL; 561} 562 563public int 564magic_errno(struct magic_set *ms) 565{ 566 if (ms == NULL) 567 return EINVAL; 568 return (ms->event_flags & EVENT_HAD_ERR) ? ms->error : 0; 569} 570 571public int 572magic_getflags(struct magic_set *ms) 573{ 574 if (ms == NULL) 575 return -1; 576 577 return ms->flags; 578} 579 580public int 581magic_setflags(struct magic_set *ms, int flags) 582{ 583 if (ms == NULL) 584 return -1; 585#if !defined(HAVE_UTIME) && !defined(HAVE_UTIMES) 586 if (flags & MAGIC_PRESERVE_ATIME) 587 return -1; 588#endif 589 ms->flags = flags; 590 return 0; 591} 592 593public int 594magic_version(void) 595{ 596 return MAGIC_VERSION; 597} 598 599public int 600magic_setparam(struct magic_set *ms, int param, const void *val) 601{ 602 if (ms == NULL) 603 return -1; 604 switch (param) { 605 case MAGIC_PARAM_INDIR_MAX: 606 ms->indir_max = CAST(uint16_t, *CAST(const size_t *, val)); 607 return 0; 608 case MAGIC_PARAM_NAME_MAX: 609 ms->name_max = CAST(uint16_t, *CAST(const size_t *, val)); 610 return 0; 611 case MAGIC_PARAM_ELF_PHNUM_MAX: 612 ms->elf_phnum_max = CAST(uint16_t, *CAST(const size_t *, val)); 613 return 0; 614 case MAGIC_PARAM_ELF_SHNUM_MAX: 615 ms->elf_shnum_max = CAST(uint16_t, *CAST(const size_t *, val)); 616 return 0; 617 case MAGIC_PARAM_ELF_NOTES_MAX: 618 ms->elf_notes_max = CAST(uint16_t, *CAST(const size_t *, val)); 619 return 0; 620 case MAGIC_PARAM_REGEX_MAX: 621 ms->regex_max = CAST(uint16_t, *CAST(const size_t *, val)); 622 return 0; 623 case MAGIC_PARAM_BYTES_MAX: 624 ms->bytes_max = *CAST(const size_t *, val); 625 return 0; 626 case MAGIC_PARAM_ENCODING_MAX: 627 ms->encoding_max = *CAST(const size_t *, val); 628 return 0; 629 default: 630 errno = EINVAL; 631 return -1; 632 } 633} 634 635public int 636magic_getparam(struct magic_set *ms, int param, void *val) 637{ 638 if (ms == NULL) 639 return -1; 640 switch (param) { 641 case MAGIC_PARAM_INDIR_MAX: 642 *CAST(size_t *, val) = ms->indir_max; 643 return 0; 644 case MAGIC_PARAM_NAME_MAX: 645 *CAST(size_t *, val) = ms->name_max; 646 return 0; 647 case MAGIC_PARAM_ELF_PHNUM_MAX: 648 *CAST(size_t *, val) = ms->elf_phnum_max; 649 return 0; 650 case MAGIC_PARAM_ELF_SHNUM_MAX: 651 *CAST(size_t *, val) = ms->elf_shnum_max; 652 return 0; 653 case MAGIC_PARAM_ELF_NOTES_MAX: 654 *CAST(size_t *, val) = ms->elf_notes_max; 655 return 0; 656 case MAGIC_PARAM_REGEX_MAX: 657 *CAST(size_t *, val) = ms->regex_max; 658 return 0; 659 case MAGIC_PARAM_BYTES_MAX: 660 *CAST(size_t *, val) = ms->bytes_max; 661 return 0; 662 case MAGIC_PARAM_ENCODING_MAX: 663 *CAST(size_t *, val) = ms->encoding_max; 664 return 0; 665 default: 666 errno = EINVAL; 667 return -1; 668 } 669} 670