1 /* 2 * Copyright (c) 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from software contributed to Berkeley by 6 * Guido van Rossum. 7 * 8 * Copyright (c) 2011 The FreeBSD Foundation 9 * All rights reserved. 10 * Portions of this software were developed by David Chisnall 11 * under sponsorship from the FreeBSD Foundation. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 4. Neither the name of the University nor the names of its contributors 22 * may be used to endorse or promote products derived from this software 23 * without specific prior written permission. 24 * 25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 35 * SUCH DAMAGE. 36 */ 37 38 #if defined(LIBC_SCCS) && !defined(lint) 39 static char sccsid[] = "@(#)glob.c 8.3 (Berkeley) 10/13/93"; 40 #endif /* LIBC_SCCS and not lint */ 41 #include <sys/cdefs.h> 42 __FBSDID("$FreeBSD$"); 43 44 /* 45 * glob(3) -- a superset of the one defined in POSIX 1003.2. 46 * 47 * The [!...] convention to negate a range is supported (SysV, Posix, ksh). 48 * 49 * Optional extra services, controlled by flags not defined by POSIX: 50 * 51 * GLOB_QUOTE: 52 * Escaping convention: \ inhibits any special meaning the following 53 * character might have (except \ at end of string is retained). 54 * GLOB_MAGCHAR: 55 * Set in gl_flags if pattern contained a globbing character. 56 * GLOB_NOMAGIC: 57 * Same as GLOB_NOCHECK, but it will only append pattern if it did 58 * not contain any magic characters. [Used in csh style globbing] 59 * GLOB_ALTDIRFUNC: 60 * Use alternately specified directory access functions. 61 * GLOB_TILDE: 62 * expand ~user/foo to the /home/dir/of/user/foo 63 * GLOB_BRACE: 64 * expand {1,2}{a,b} to 1a 1b 2a 2b 65 * gl_matchc: 66 * Number of matches in the current invocation of glob. 67 */ 68 69 /* 70 * Some notes on multibyte character support: 71 * 1. Patterns with illegal byte sequences match nothing - even if 72 * GLOB_NOCHECK is specified. 73 * 2. Illegal byte sequences in filenames are handled by treating them as 74 * single-byte characters with a values of such bytes of the sequence 75 * cast to wchar_t. 76 * 3. State-dependent encodings are not currently supported. 77 */ 78 79 #include <sys/param.h> 80 #include <sys/stat.h> 81 82 #include <ctype.h> 83 #include <dirent.h> 84 #include <errno.h> 85 #include <glob.h> 86 #include <limits.h> 87 #include <pwd.h> 88 #include <stdint.h> 89 #include <stdio.h> 90 #include <stdlib.h> 91 #include <string.h> 92 #include <unistd.h> 93 #include <wchar.h> 94 95 #include "collate.h" 96 97 /* 98 * glob(3) expansion limits. Stop the expansion if any of these limits 99 * is reached. This caps the runtime in the face of DoS attacks. See 100 * also CVE-2010-2632 101 */ 102 #define GLOB_LIMIT_BRACE 128 /* number of brace calls */ 103 #define GLOB_LIMIT_PATH 65536 /* number of path elements */ 104 #define GLOB_LIMIT_READDIR 16384 /* number of readdirs */ 105 #define GLOB_LIMIT_STAT 1024 /* number of stat system calls */ 106 #define GLOB_LIMIT_STRING ARG_MAX /* maximum total size for paths */ 107 108 struct glob_limit { 109 size_t l_brace_cnt; 110 size_t l_path_lim; 111 size_t l_readdir_cnt; 112 size_t l_stat_cnt; 113 size_t l_string_cnt; 114 }; 115 116 #define DOT L'.' 117 #define EOS L'\0' 118 #define LBRACKET L'[' 119 #define NOT L'!' 120 #define QUESTION L'?' 121 #define QUOTE L'\\' 122 #define RANGE L'-' 123 #define RBRACKET L']' 124 #define SEP L'/' 125 #define STAR L'*' 126 #define TILDE L'~' 127 #define LBRACE L'{' 128 #define RBRACE L'}' 129 #define SLASH L'/' 130 #define COMMA L',' 131 132 #ifndef DEBUG 133 134 #define M_QUOTE 0x8000000000ULL 135 #define M_PROTECT 0x4000000000ULL 136 #define M_MASK 0xffffffffffULL 137 #define M_CHAR 0x00ffffffffULL 138 139 typedef uint_fast64_t Char; 140 141 #else 142 143 #define M_QUOTE 0x80 144 #define M_PROTECT 0x40 145 #define M_MASK 0xff 146 #define M_CHAR 0x7f 147 148 typedef char Char; 149 150 #endif 151 152 153 #define CHAR(c) ((Char)((c)&M_CHAR)) 154 #define META(c) ((Char)((c)|M_QUOTE)) 155 #define M_ALL META(L'*') 156 #define M_END META(L']') 157 #define M_NOT META(L'!') 158 #define M_ONE META(L'?') 159 #define M_RNG META(L'-') 160 #define M_SET META(L'[') 161 #define ismeta(c) (((c)&M_QUOTE) != 0) 162 163 164 static int compare(const void *, const void *); 165 static int g_Ctoc(const Char *, char *, size_t); 166 static int g_lstat(Char *, struct stat *, glob_t *); 167 static DIR *g_opendir(Char *, glob_t *); 168 static const Char *g_strchr(const Char *, wchar_t); 169 #ifdef notdef 170 static Char *g_strcat(Char *, const Char *); 171 #endif 172 static int g_stat(Char *, struct stat *, glob_t *); 173 static int glob0(const Char *, glob_t *, struct glob_limit *); 174 static int glob1(Char *, glob_t *, struct glob_limit *); 175 static int glob2(Char *, Char *, Char *, Char *, glob_t *, 176 struct glob_limit *); 177 static int glob3(Char *, Char *, Char *, Char *, Char *, glob_t *, 178 struct glob_limit *); 179 static int globextend(const Char *, glob_t *, struct glob_limit *); 180 static const Char * 181 globtilde(const Char *, Char *, size_t, glob_t *); 182 static int globexp1(const Char *, glob_t *, struct glob_limit *); 183 static int globexp2(const Char *, const Char *, glob_t *, int *, 184 struct glob_limit *); 185 static int match(Char *, Char *, Char *); 186 #ifdef DEBUG 187 static void qprintf(const char *, Char *); 188 #endif 189 190 int 191 glob(const char * __restrict pattern, int flags, 192 int (*errfunc)(const char *, int), glob_t * __restrict pglob) 193 { 194 struct glob_limit limit = { 0, 0, 0, 0, 0 }; 195 const char *patnext; 196 Char *bufnext, *bufend, patbuf[MAXPATHLEN], prot; 197 mbstate_t mbs; 198 wchar_t wc; 199 size_t clen; 200 201 patnext = pattern; 202 if (!(flags & GLOB_APPEND)) { 203 pglob->gl_pathc = 0; 204 pglob->gl_pathv = NULL; 205 if (!(flags & GLOB_DOOFFS)) 206 pglob->gl_offs = 0; 207 } 208 if (flags & GLOB_LIMIT) { 209 limit.l_path_lim = pglob->gl_matchc; 210 if (limit.l_path_lim == 0) 211 limit.l_path_lim = GLOB_LIMIT_PATH; 212 } 213 pglob->gl_flags = flags & ~GLOB_MAGCHAR; 214 pglob->gl_errfunc = errfunc; 215 pglob->gl_matchc = 0; 216 217 bufnext = patbuf; 218 bufend = bufnext + MAXPATHLEN - 1; 219 if (flags & GLOB_NOESCAPE) { 220 memset(&mbs, 0, sizeof(mbs)); 221 while (bufend - bufnext >= MB_CUR_MAX) { 222 clen = mbrtowc(&wc, patnext, MB_LEN_MAX, &mbs); 223 if (clen == (size_t)-1 || clen == (size_t)-2) 224 return (GLOB_NOMATCH); 225 else if (clen == 0) 226 break; 227 *bufnext++ = wc; 228 patnext += clen; 229 } 230 } else { 231 /* Protect the quoted characters. */ 232 memset(&mbs, 0, sizeof(mbs)); 233 while (bufend - bufnext >= MB_CUR_MAX) { 234 if (*patnext == '\\') { 235 if (*++patnext == '\0') { 236 *bufnext++ = QUOTE | M_PROTECT; 237 continue; 238 } 239 prot = M_PROTECT; 240 } else 241 prot = 0; 242 clen = mbrtowc(&wc, patnext, MB_LEN_MAX, &mbs); 243 if (clen == (size_t)-1 || clen == (size_t)-2) 244 return (GLOB_NOMATCH); 245 else if (clen == 0) 246 break; 247 *bufnext++ = wc | ((wc != DOT && wc != SEP) ? 248 prot : 0); 249 patnext += clen; 250 } 251 } 252 *bufnext = EOS; 253 254 if (flags & GLOB_BRACE) 255 return (globexp1(patbuf, pglob, &limit)); 256 else 257 return (glob0(patbuf, pglob, &limit)); 258 } 259 260 /* 261 * Expand recursively a glob {} pattern. When there is no more expansion 262 * invoke the standard globbing routine to glob the rest of the magic 263 * characters 264 */ 265 static int 266 globexp1(const Char *pattern, glob_t *pglob, struct glob_limit *limit) 267 { 268 const Char* ptr = pattern; 269 int rv; 270 271 if ((pglob->gl_flags & GLOB_LIMIT) && 272 limit->l_brace_cnt++ >= GLOB_LIMIT_BRACE) { 273 errno = 0; 274 return (GLOB_NOSPACE); 275 } 276 277 /* Protect a single {}, for find(1), like csh */ 278 if (pattern[0] == LBRACE && pattern[1] == RBRACE && pattern[2] == EOS) 279 return glob0(pattern, pglob, limit); 280 281 while ((ptr = g_strchr(ptr, LBRACE)) != NULL) 282 if (!globexp2(ptr, pattern, pglob, &rv, limit)) 283 return rv; 284 285 return glob0(pattern, pglob, limit); 286 } 287 288 289 /* 290 * Recursive brace globbing helper. Tries to expand a single brace. 291 * If it succeeds then it invokes globexp1 with the new pattern. 292 * If it fails then it tries to glob the rest of the pattern and returns. 293 */ 294 static int 295 globexp2(const Char *ptr, const Char *pattern, glob_t *pglob, int *rv, 296 struct glob_limit *limit) 297 { 298 int i; 299 Char *lm, *ls; 300 const Char *pe, *pm, *pm1, *pl; 301 Char patbuf[MAXPATHLEN]; 302 303 /* copy part up to the brace */ 304 for (lm = patbuf, pm = pattern; pm != ptr; *lm++ = *pm++) 305 continue; 306 *lm = EOS; 307 ls = lm; 308 309 /* Find the balanced brace */ 310 for (i = 0, pe = ++ptr; *pe; pe++) 311 if (*pe == LBRACKET) { 312 /* Ignore everything between [] */ 313 for (pm = pe++; *pe != RBRACKET && *pe != EOS; pe++) 314 continue; 315 if (*pe == EOS) { 316 /* 317 * We could not find a matching RBRACKET. 318 * Ignore and just look for RBRACE 319 */ 320 pe = pm; 321 } 322 } 323 else if (*pe == LBRACE) 324 i++; 325 else if (*pe == RBRACE) { 326 if (i == 0) 327 break; 328 i--; 329 } 330 331 /* Non matching braces; just glob the pattern */ 332 if (i != 0 || *pe == EOS) { 333 *rv = glob0(patbuf, pglob, limit); 334 return (0); 335 } 336 337 for (i = 0, pl = pm = ptr; pm <= pe; pm++) 338 switch (*pm) { 339 case LBRACKET: 340 /* Ignore everything between [] */ 341 for (pm1 = pm++; *pm != RBRACKET && *pm != EOS; pm++) 342 continue; 343 if (*pm == EOS) { 344 /* 345 * We could not find a matching RBRACKET. 346 * Ignore and just look for RBRACE 347 */ 348 pm = pm1; 349 } 350 break; 351 352 case LBRACE: 353 i++; 354 break; 355 356 case RBRACE: 357 if (i) { 358 i--; 359 break; 360 } 361 /* FALLTHROUGH */ 362 case COMMA: 363 if (i && *pm == COMMA) 364 break; 365 else { 366 /* Append the current string */ 367 for (lm = ls; (pl < pm); *lm++ = *pl++) 368 continue; 369 /* 370 * Append the rest of the pattern after the 371 * closing brace 372 */ 373 for (pl = pe + 1; (*lm++ = *pl++) != EOS;) 374 continue; 375 376 /* Expand the current pattern */ 377 #ifdef DEBUG 378 qprintf("globexp2:", patbuf); 379 #endif 380 *rv = globexp1(patbuf, pglob, limit); 381 382 /* move after the comma, to the next string */ 383 pl = pm + 1; 384 } 385 break; 386 387 default: 388 break; 389 } 390 *rv = 0; 391 return (0); 392 } 393 394 395 396 /* 397 * expand tilde from the passwd file. 398 */ 399 static const Char * 400 globtilde(const Char *pattern, Char *patbuf, size_t patbuf_len, glob_t *pglob) 401 { 402 struct passwd *pwd; 403 char *h, *sc; 404 const Char *p; 405 Char *b, *eb; 406 wchar_t wc; 407 wchar_t wbuf[MAXPATHLEN]; 408 wchar_t *wbufend, *dc; 409 size_t clen; 410 mbstate_t mbs; 411 int too_long; 412 413 if (*pattern != TILDE || !(pglob->gl_flags & GLOB_TILDE)) 414 return (pattern); 415 416 /* 417 * Copy up to the end of the string or / 418 */ 419 eb = &patbuf[patbuf_len - 1]; 420 for (p = pattern + 1, b = patbuf; 421 b < eb && *p != EOS && *p != SLASH; *b++ = *p++) 422 continue; 423 424 if (*p != EOS && *p != SLASH) 425 return (NULL); 426 427 *b = EOS; 428 h = NULL; 429 430 if (patbuf[0] == EOS) { 431 /* 432 * handle a plain ~ or ~/ by expanding $HOME first (iff 433 * we're not running setuid or setgid) and then trying 434 * the password file 435 */ 436 if (issetugid() != 0 || 437 (h = getenv("HOME")) == NULL) { 438 if (((h = getlogin()) != NULL && 439 (pwd = getpwnam(h)) != NULL) || 440 (pwd = getpwuid(getuid())) != NULL) 441 h = pwd->pw_dir; 442 else 443 return (pattern); 444 } 445 } 446 else { 447 /* 448 * Expand a ~user 449 */ 450 if (g_Ctoc(patbuf, (char *)wbuf, sizeof(wbuf))) 451 return (NULL); 452 if ((pwd = getpwnam((char *)wbuf)) == NULL) 453 return (pattern); 454 else 455 h = pwd->pw_dir; 456 } 457 458 /* Copy the home directory */ 459 dc = wbuf; 460 sc = h; 461 wbufend = wbuf + MAXPATHLEN - 1; 462 too_long = 1; 463 memset(&mbs, 0, sizeof(mbs)); 464 while (dc <= wbufend) { 465 clen = mbrtowc(&wc, sc, MB_LEN_MAX, &mbs); 466 if (clen == (size_t)-1 || clen == (size_t)-2) { 467 /* XXX See initial comment #2. */ 468 wc = (unsigned char)*sc; 469 clen = 1; 470 memset(&mbs, 0, sizeof(mbs)); 471 } 472 if ((*dc++ = wc) == EOS) { 473 too_long = 0; 474 break; 475 } 476 sc += clen; 477 } 478 if (too_long) 479 return (NULL); 480 481 dc = wbuf; 482 for (b = patbuf; b < eb && *dc != EOS; b++, dc++) 483 *b = *dc | ((*dc != DOT && *dc != SEP) ? M_PROTECT : 0); 484 if (*dc != EOS) 485 return (NULL); 486 487 /* Append the rest of the pattern */ 488 if (*p != EOS) { 489 too_long = 1; 490 while (b <= eb) { 491 if ((*b++ = *p++) == EOS) { 492 too_long = 0; 493 break; 494 } 495 } 496 if (too_long) 497 return (NULL); 498 } else 499 *b = EOS; 500 501 return (patbuf); 502 } 503 504 505 /* 506 * The main glob() routine: compiles the pattern (optionally processing 507 * quotes), calls glob1() to do the real pattern matching, and finally 508 * sorts the list (unless unsorted operation is requested). Returns 0 509 * if things went well, nonzero if errors occurred. 510 */ 511 static int 512 glob0(const Char *pattern, glob_t *pglob, struct glob_limit *limit) 513 { 514 const Char *qpatnext; 515 int err; 516 size_t oldpathc; 517 Char *bufnext, c, patbuf[MAXPATHLEN]; 518 519 qpatnext = globtilde(pattern, patbuf, MAXPATHLEN, pglob); 520 if (qpatnext == NULL) { 521 errno = 0; 522 return (GLOB_NOSPACE); 523 } 524 oldpathc = pglob->gl_pathc; 525 bufnext = patbuf; 526 527 /* We don't need to check for buffer overflow any more. */ 528 while ((c = *qpatnext++) != EOS) { 529 switch (c) { 530 case LBRACKET: 531 c = *qpatnext; 532 if (c == NOT) 533 ++qpatnext; 534 if (*qpatnext == EOS || 535 g_strchr(qpatnext+1, RBRACKET) == NULL) { 536 *bufnext++ = LBRACKET; 537 if (c == NOT) 538 --qpatnext; 539 break; 540 } 541 *bufnext++ = M_SET; 542 if (c == NOT) 543 *bufnext++ = M_NOT; 544 c = *qpatnext++; 545 do { 546 *bufnext++ = CHAR(c); 547 if (*qpatnext == RANGE && 548 (c = qpatnext[1]) != RBRACKET) { 549 *bufnext++ = M_RNG; 550 *bufnext++ = CHAR(c); 551 qpatnext += 2; 552 } 553 } while ((c = *qpatnext++) != RBRACKET); 554 pglob->gl_flags |= GLOB_MAGCHAR; 555 *bufnext++ = M_END; 556 break; 557 case QUESTION: 558 pglob->gl_flags |= GLOB_MAGCHAR; 559 *bufnext++ = M_ONE; 560 break; 561 case STAR: 562 pglob->gl_flags |= GLOB_MAGCHAR; 563 /* collapse adjacent stars to one, 564 * to avoid exponential behavior 565 */ 566 if (bufnext == patbuf || bufnext[-1] != M_ALL) 567 *bufnext++ = M_ALL; 568 break; 569 default: 570 *bufnext++ = CHAR(c); 571 break; 572 } 573 } 574 *bufnext = EOS; 575 #ifdef DEBUG 576 qprintf("glob0:", patbuf); 577 #endif 578 579 if ((err = glob1(patbuf, pglob, limit)) != 0) 580 return(err); 581 582 /* 583 * If there was no match we are going to append the pattern 584 * if GLOB_NOCHECK was specified or if GLOB_NOMAGIC was specified 585 * and the pattern did not contain any magic characters 586 * GLOB_NOMAGIC is there just for compatibility with csh. 587 */ 588 if (pglob->gl_pathc == oldpathc) { 589 if (((pglob->gl_flags & GLOB_NOCHECK) || 590 ((pglob->gl_flags & GLOB_NOMAGIC) && 591 !(pglob->gl_flags & GLOB_MAGCHAR)))) 592 return (globextend(pattern, pglob, limit)); 593 else 594 return (GLOB_NOMATCH); 595 } 596 if (!(pglob->gl_flags & GLOB_NOSORT)) 597 qsort(pglob->gl_pathv + pglob->gl_offs + oldpathc, 598 pglob->gl_pathc - oldpathc, sizeof(char *), compare); 599 return (0); 600 } 601 602 static int 603 compare(const void *p, const void *q) 604 { 605 return (strcoll(*(char **)p, *(char **)q)); 606 } 607 608 static int 609 glob1(Char *pattern, glob_t *pglob, struct glob_limit *limit) 610 { 611 Char pathbuf[MAXPATHLEN]; 612 613 /* A null pathname is invalid -- POSIX 1003.1 sect. 2.4. */ 614 if (*pattern == EOS) 615 return (0); 616 return (glob2(pathbuf, pathbuf, pathbuf + MAXPATHLEN - 1, 617 pattern, pglob, limit)); 618 } 619 620 /* 621 * The functions glob2 and glob3 are mutually recursive; there is one level 622 * of recursion for each segment in the pattern that contains one or more 623 * meta characters. 624 */ 625 static int 626 glob2(Char *pathbuf, Char *pathend, Char *pathend_last, Char *pattern, 627 glob_t *pglob, struct glob_limit *limit) 628 { 629 struct stat sb; 630 Char *p, *q; 631 int anymeta; 632 633 /* 634 * Loop over pattern segments until end of pattern or until 635 * segment with meta character found. 636 */ 637 for (anymeta = 0;;) { 638 if (*pattern == EOS) { /* End of pattern? */ 639 *pathend = EOS; 640 if (g_lstat(pathbuf, &sb, pglob)) 641 return (0); 642 643 if ((pglob->gl_flags & GLOB_LIMIT) && 644 limit->l_stat_cnt++ >= GLOB_LIMIT_STAT) { 645 errno = 0; 646 return (GLOB_NOSPACE); 647 } 648 if (((pglob->gl_flags & GLOB_MARK) && 649 pathend[-1] != SEP) && (S_ISDIR(sb.st_mode) 650 || (S_ISLNK(sb.st_mode) && 651 (g_stat(pathbuf, &sb, pglob) == 0) && 652 S_ISDIR(sb.st_mode)))) { 653 if (pathend + 1 > pathend_last) { 654 errno = 0; 655 return (GLOB_NOSPACE); 656 } 657 *pathend++ = SEP; 658 *pathend = EOS; 659 } 660 ++pglob->gl_matchc; 661 return (globextend(pathbuf, pglob, limit)); 662 } 663 664 /* Find end of next segment, copy tentatively to pathend. */ 665 q = pathend; 666 p = pattern; 667 while (*p != EOS && *p != SEP) { 668 if (ismeta(*p)) 669 anymeta = 1; 670 if (q + 1 > pathend_last) { 671 errno = 0; 672 return (GLOB_NOSPACE); 673 } 674 *q++ = *p++; 675 } 676 677 if (!anymeta) { /* No expansion, do next segment. */ 678 pathend = q; 679 pattern = p; 680 while (*pattern == SEP) { 681 if (pathend + 1 > pathend_last) { 682 errno = 0; 683 return (GLOB_NOSPACE); 684 } 685 *pathend++ = *pattern++; 686 } 687 } else /* Need expansion, recurse. */ 688 return (glob3(pathbuf, pathend, pathend_last, pattern, 689 p, pglob, limit)); 690 } 691 /* NOTREACHED */ 692 } 693 694 static int 695 glob3(Char *pathbuf, Char *pathend, Char *pathend_last, 696 Char *pattern, Char *restpattern, 697 glob_t *pglob, struct glob_limit *limit) 698 { 699 struct dirent *dp; 700 DIR *dirp; 701 int err; 702 char buf[MAXPATHLEN + MB_LEN_MAX - 1]; 703 704 struct dirent *(*readdirfunc)(DIR *); 705 706 errno = 0; 707 if (pathend > pathend_last) 708 return (GLOB_NOSPACE); 709 *pathend = EOS; 710 711 if ((dirp = g_opendir(pathbuf, pglob)) == NULL) { 712 if (errno == ENOENT || errno == ENOTDIR) 713 return (0); 714 if (pglob->gl_errfunc != NULL) { 715 if (g_Ctoc(pathbuf, buf, sizeof(buf))) { 716 errno = 0; 717 return (GLOB_NOSPACE); 718 } 719 if (pglob->gl_errfunc(buf, errno)) 720 return (GLOB_ABORTED); 721 } 722 if (pglob->gl_flags & GLOB_ERR) 723 return (GLOB_ABORTED); 724 return (0); 725 } 726 727 err = 0; 728 729 /* pglob->gl_readdir takes a void *, fix this manually */ 730 if (pglob->gl_flags & GLOB_ALTDIRFUNC) 731 readdirfunc = (struct dirent *(*)(DIR *))pglob->gl_readdir; 732 else 733 readdirfunc = readdir; 734 735 /* Search directory for matching names. */ 736 while ((dp = (*readdirfunc)(dirp)) != NULL) { 737 char *sc; 738 Char *dc; 739 wchar_t wc; 740 size_t clen; 741 mbstate_t mbs; 742 743 if ((pglob->gl_flags & GLOB_LIMIT) && 744 limit->l_readdir_cnt++ >= GLOB_LIMIT_READDIR) { 745 errno = 0; 746 if (pathend + 1 > pathend_last) 747 err = GLOB_NOSPACE; 748 else { 749 *pathend++ = SEP; 750 *pathend = EOS; 751 err = GLOB_NOSPACE; 752 } 753 break; 754 } 755 756 /* Initial DOT must be matched literally. */ 757 if (dp->d_name[0] == '.' && *pattern != DOT) 758 continue; 759 memset(&mbs, 0, sizeof(mbs)); 760 dc = pathend; 761 sc = dp->d_name; 762 while (dc < pathend_last) { 763 clen = mbrtowc(&wc, sc, MB_LEN_MAX, &mbs); 764 if (clen == (size_t)-1 || clen == (size_t)-2) { 765 /* XXX See initial comment #2. */ 766 wc = (unsigned char)*sc; 767 clen = 1; 768 memset(&mbs, 0, sizeof(mbs)); 769 } 770 if ((*dc++ = wc) == EOS) 771 break; 772 sc += clen; 773 } 774 if (!match(pathend, pattern, restpattern)) { 775 *pathend = EOS; 776 continue; 777 } 778 err = glob2(pathbuf, --dc, pathend_last, restpattern, 779 pglob, limit); 780 if (err) 781 break; 782 } 783 784 if (pglob->gl_flags & GLOB_ALTDIRFUNC) 785 (*pglob->gl_closedir)(dirp); 786 else 787 closedir(dirp); 788 return (err); 789 } 790 791 792 /* 793 * Extend the gl_pathv member of a glob_t structure to accommodate a new item, 794 * add the new item, and update gl_pathc. 795 * 796 * This assumes the BSD realloc, which only copies the block when its size 797 * crosses a power-of-two boundary; for v7 realloc, this would cause quadratic 798 * behavior. 799 * 800 * Return 0 if new item added, error code if memory couldn't be allocated. 801 * 802 * Invariant of the glob_t structure: 803 * Either gl_pathc is zero and gl_pathv is NULL; or gl_pathc > 0 and 804 * gl_pathv points to (gl_offs + gl_pathc + 1) items. 805 */ 806 static int 807 globextend(const Char *path, glob_t *pglob, struct glob_limit *limit) 808 { 809 char **pathv; 810 size_t i, newsize, len; 811 char *copy; 812 const Char *p; 813 814 if ((pglob->gl_flags & GLOB_LIMIT) && 815 pglob->gl_matchc > limit->l_path_lim) { 816 errno = 0; 817 return (GLOB_NOSPACE); 818 } 819 820 newsize = sizeof(*pathv) * (2 + pglob->gl_pathc + pglob->gl_offs); 821 /* realloc(NULL, newsize) is equivalent to malloc(newsize). */ 822 pathv = realloc((void *)pglob->gl_pathv, newsize); 823 if (pathv == NULL) 824 return (GLOB_NOSPACE); 825 826 if (pglob->gl_pathv == NULL && pglob->gl_offs > 0) { 827 /* first time around -- clear initial gl_offs items */ 828 pathv += pglob->gl_offs; 829 for (i = pglob->gl_offs + 1; --i > 0; ) 830 *--pathv = NULL; 831 } 832 pglob->gl_pathv = pathv; 833 834 for (p = path; *p++;) 835 continue; 836 len = MB_CUR_MAX * (size_t)(p - path); /* XXX overallocation */ 837 limit->l_string_cnt += len; 838 if ((pglob->gl_flags & GLOB_LIMIT) && 839 limit->l_string_cnt >= GLOB_LIMIT_STRING) { 840 errno = 0; 841 return (GLOB_NOSPACE); 842 } 843 if ((copy = malloc(len)) != NULL) { 844 if (g_Ctoc(path, copy, len)) { 845 free(copy); 846 errno = 0; 847 return (GLOB_NOSPACE); 848 } 849 pathv[pglob->gl_offs + pglob->gl_pathc++] = copy; 850 } 851 pathv[pglob->gl_offs + pglob->gl_pathc] = NULL; 852 return (copy == NULL ? GLOB_NOSPACE : 0); 853 } 854 855 /* 856 * pattern matching function for filenames. Each occurrence of the * 857 * pattern causes a recursion level. 858 */ 859 static int 860 match(Char *name, Char *pat, Char *patend) 861 { 862 int ok, negate_range; 863 Char c, k; 864 struct xlocale_collate *table = 865 (struct xlocale_collate*)__get_locale()->components[XLC_COLLATE]; 866 867 while (pat < patend) { 868 c = *pat++; 869 switch (c & M_MASK) { 870 case M_ALL: 871 if (pat == patend) 872 return (1); 873 do 874 if (match(name, pat, patend)) 875 return (1); 876 while (*name++ != EOS); 877 return (0); 878 case M_ONE: 879 if (*name++ == EOS) 880 return (0); 881 break; 882 case M_SET: 883 ok = 0; 884 if ((k = *name++) == EOS) 885 return (0); 886 if ((negate_range = ((*pat & M_MASK) == M_NOT)) != 0) 887 ++pat; 888 while (((c = *pat++) & M_MASK) != M_END) 889 if ((*pat & M_MASK) == M_RNG) { 890 if (table->__collate_load_error ? 891 CHAR(c) <= CHAR(k) && 892 CHAR(k) <= CHAR(pat[1]) : 893 __wcollate_range_cmp(CHAR(c), 894 CHAR(k)) <= 0 && 895 __wcollate_range_cmp(CHAR(k), 896 CHAR(pat[1])) <= 0) 897 ok = 1; 898 pat += 2; 899 } else if (c == k) 900 ok = 1; 901 if (ok == negate_range) 902 return (0); 903 break; 904 default: 905 if (*name++ != c) 906 return (0); 907 break; 908 } 909 } 910 return (*name == EOS); 911 } 912 913 /* Free allocated data belonging to a glob_t structure. */ 914 void 915 globfree(glob_t *pglob) 916 { 917 size_t i; 918 char **pp; 919 920 if (pglob->gl_pathv != NULL) { 921 pp = pglob->gl_pathv + pglob->gl_offs; 922 for (i = pglob->gl_pathc; i--; ++pp) 923 if (*pp) 924 free(*pp); 925 free(pglob->gl_pathv); 926 pglob->gl_pathv = NULL; 927 } 928 } 929 930 static DIR * 931 g_opendir(Char *str, glob_t *pglob) 932 { 933 char buf[MAXPATHLEN + MB_LEN_MAX - 1]; 934 935 if (*str == EOS) 936 strcpy(buf, "."); 937 else { 938 if (g_Ctoc(str, buf, sizeof(buf))) { 939 errno = ENAMETOOLONG; 940 return (NULL); 941 } 942 } 943 944 if (pglob->gl_flags & GLOB_ALTDIRFUNC) 945 return ((*pglob->gl_opendir)(buf)); 946 947 return (opendir(buf)); 948 } 949 950 static int 951 g_lstat(Char *fn, struct stat *sb, glob_t *pglob) 952 { 953 char buf[MAXPATHLEN + MB_LEN_MAX - 1]; 954 955 if (g_Ctoc(fn, buf, sizeof(buf))) { 956 errno = ENAMETOOLONG; 957 return (-1); 958 } 959 if (pglob->gl_flags & GLOB_ALTDIRFUNC) 960 return((*pglob->gl_lstat)(buf, sb)); 961 return (lstat(buf, sb)); 962 } 963 964 static int 965 g_stat(Char *fn, struct stat *sb, glob_t *pglob) 966 { 967 char buf[MAXPATHLEN + MB_LEN_MAX - 1]; 968 969 if (g_Ctoc(fn, buf, sizeof(buf))) { 970 errno = ENAMETOOLONG; 971 return (-1); 972 } 973 if (pglob->gl_flags & GLOB_ALTDIRFUNC) 974 return ((*pglob->gl_stat)(buf, sb)); 975 return (stat(buf, sb)); 976 } 977 978 static const Char * 979 g_strchr(const Char *str, wchar_t ch) 980 { 981 982 do { 983 if (*str == ch) 984 return (str); 985 } while (*str++); 986 return (NULL); 987 } 988 989 static int 990 g_Ctoc(const Char *str, char *buf, size_t len) 991 { 992 mbstate_t mbs; 993 size_t clen; 994 995 memset(&mbs, 0, sizeof(mbs)); 996 while (len >= MB_CUR_MAX) { 997 clen = wcrtomb(buf, CHAR(*str), &mbs); 998 if (clen == (size_t)-1) { 999 /* XXX See initial comment #2. */ 1000 *buf = (char)CHAR(*str); 1001 clen = 1; 1002 memset(&mbs, 0, sizeof(mbs)); 1003 } 1004 if (CHAR(*str) == EOS) 1005 return (0); 1006 str++; 1007 buf += clen; 1008 len -= clen; 1009 } 1010 return (1); 1011 } 1012 1013 #ifdef DEBUG 1014 static void 1015 qprintf(const char *str, Char *s) 1016 { 1017 Char *p; 1018 1019 (void)printf("%s:\n", str); 1020 for (p = s; *p; p++) 1021 (void)printf("%c", CHAR(*p)); 1022 (void)printf("\n"); 1023 for (p = s; *p; p++) 1024 (void)printf("%c", *p & M_PROTECT ? '"' : ' '); 1025 (void)printf("\n"); 1026 for (p = s; *p; p++) 1027 (void)printf("%c", ismeta(*p) ? '_' : ' '); 1028 (void)printf("\n"); 1029 } 1030 #endif 1031