1 /*- 2 * Copyright (c) 2003-2007 Tim Kientzle 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR 15 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 16 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 17 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, 18 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 19 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 20 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 21 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 22 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 23 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 24 */ 25 26 #include "bsdtar_platform.h" 27 __FBSDID("$FreeBSD: src/usr.bin/tar/util.c,v 1.23 2008/12/15 06:00:25 kientzle Exp $"); 28 29 #ifdef HAVE_SYS_STAT_H 30 #include <sys/stat.h> 31 #endif 32 #ifdef HAVE_SYS_TYPES_H 33 #include <sys/types.h> /* Linux doesn't define mode_t, etc. in sys/stat.h. */ 34 #endif 35 #include <ctype.h> 36 #ifdef HAVE_ERRNO_H 37 #include <errno.h> 38 #endif 39 #ifdef HAVE_IO_H 40 #include <io.h> 41 #endif 42 #ifdef HAVE_STDARG_H 43 #include <stdarg.h> 44 #endif 45 #ifdef HAVE_STDINT_H 46 #include <stdint.h> 47 #endif 48 #include <stdio.h> 49 #ifdef HAVE_STDLIB_H 50 #include <stdlib.h> 51 #endif 52 #ifdef HAVE_STRING_H 53 #include <string.h> 54 #endif 55 #ifdef HAVE_WCTYPE_H 56 #include <wctype.h> 57 #else 58 /* If we don't have wctype, we need to hack up some version of iswprint(). */ 59 #define iswprint isprint 60 #endif 61 62 #include "bsdtar.h" 63 #include "err.h" 64 65 static size_t bsdtar_expand_char(char *, size_t, char); 66 static const char *strip_components(const char *path, int elements); 67 68 #if defined(_WIN32) && !defined(__CYGWIN__) 69 #define read _read 70 #endif 71 72 /* TODO: Hack up a version of mbtowc for platforms with no wide 73 * character support at all. I think the following might suffice, 74 * but it needs careful testing. 75 * #if !HAVE_MBTOWC 76 * #define mbtowc(wcp, p, n) ((*wcp = *p), 1) 77 * #endif 78 */ 79 80 /* 81 * Print a string, taking care with any non-printable characters. 82 * 83 * Note that we use a stack-allocated buffer to receive the formatted 84 * string if we can. This is partly performance (avoiding a call to 85 * malloc()), partly out of expedience (we have to call vsnprintf() 86 * before malloc() anyway to find out how big a buffer we need; we may 87 * as well point that first call at a small local buffer in case it 88 * works), but mostly for safety (so we can use this to print messages 89 * about out-of-memory conditions). 90 */ 91 92 void 93 safe_fprintf(FILE *f, const char *fmt, ...) 94 { 95 char fmtbuff_stack[256]; /* Place to format the printf() string. */ 96 char outbuff[256]; /* Buffer for outgoing characters. */ 97 char *fmtbuff_heap; /* If fmtbuff_stack is too small, we use malloc */ 98 char *fmtbuff; /* Pointer to fmtbuff_stack or fmtbuff_heap. */ 99 int fmtbuff_length; 100 int length, n; 101 va_list ap; 102 const char *p; 103 unsigned i; 104 wchar_t wc; 105 char try_wc; 106 107 /* Use a stack-allocated buffer if we can, for speed and safety. */ 108 fmtbuff_heap = NULL; 109 fmtbuff_length = sizeof(fmtbuff_stack); 110 fmtbuff = fmtbuff_stack; 111 112 /* Try formatting into the stack buffer. */ 113 va_start(ap, fmt); 114 length = vsnprintf(fmtbuff, fmtbuff_length, fmt, ap); 115 va_end(ap); 116 117 /* If the result was too large, allocate a buffer on the heap. */ 118 while (length < 0 || length >= fmtbuff_length) { 119 if (length >= fmtbuff_length) 120 fmtbuff_length = length+1; 121 else if (fmtbuff_length < 8192) 122 fmtbuff_length *= 2; 123 else { 124 int old_length = fmtbuff_length; 125 fmtbuff_length += fmtbuff_length / 4; 126 if (old_length > fmtbuff_length) { 127 length = old_length; 128 fmtbuff_heap[length-1] = '\0'; 129 break; 130 } 131 } 132 free(fmtbuff_heap); 133 fmtbuff_heap = malloc(fmtbuff_length); 134 135 /* Reformat the result into the heap buffer if we can. */ 136 if (fmtbuff_heap != NULL) { 137 fmtbuff = fmtbuff_heap; 138 va_start(ap, fmt); 139 length = vsnprintf(fmtbuff, fmtbuff_length, fmt, ap); 140 va_end(ap); 141 } else { 142 /* Leave fmtbuff pointing to the truncated 143 * string in fmtbuff_stack. */ 144 length = sizeof(fmtbuff_stack) - 1; 145 break; 146 } 147 } 148 149 /* Note: mbrtowc() has a cleaner API, but mbtowc() seems a bit 150 * more portable, so we use that here instead. */ 151 n = mbtowc(NULL, NULL, 1); /* Reset the shift state. */ 152 153 /* Write data, expanding unprintable characters. */ 154 p = fmtbuff; 155 i = 0; 156 try_wc = 1; 157 while (*p != '\0') { 158 159 /* Convert to wide char, test if the wide 160 * char is printable in the current locale. */ 161 if (try_wc && (n = mbtowc(&wc, p, length)) != -1) { 162 length -= n; 163 if (iswprint(wc) && wc != L'\\') { 164 /* Printable, copy the bytes through. */ 165 while (n-- > 0) 166 outbuff[i++] = *p++; 167 } else { 168 /* Not printable, format the bytes. */ 169 while (n-- > 0) 170 i += (unsigned)bsdtar_expand_char( 171 outbuff, i, *p++); 172 } 173 } else { 174 /* After any conversion failure, don't bother 175 * trying to convert the rest. */ 176 i += (unsigned)bsdtar_expand_char(outbuff, i, *p++); 177 try_wc = 0; 178 } 179 180 /* If our output buffer is full, dump it and keep going. */ 181 if (i > (sizeof(outbuff) - 20)) { 182 outbuff[i] = '\0'; 183 fprintf(f, "%s", outbuff); 184 i = 0; 185 } 186 } 187 outbuff[i] = '\0'; 188 fprintf(f, "%s", outbuff); 189 190 /* If we allocated a heap-based formatting buffer, free it now. */ 191 if (fmtbuff_heap != NULL) 192 free(fmtbuff_heap); 193 } 194 195 /* 196 * Render an arbitrary sequence of bytes into printable ASCII characters. 197 */ 198 static size_t 199 bsdtar_expand_char(char *buff, size_t offset, char c) 200 { 201 size_t i = offset; 202 203 if (isprint((unsigned char)c) && c != '\\') 204 buff[i++] = c; 205 else { 206 buff[i++] = '\\'; 207 switch (c) { 208 case '\a': buff[i++] = 'a'; break; 209 case '\b': buff[i++] = 'b'; break; 210 case '\f': buff[i++] = 'f'; break; 211 case '\n': buff[i++] = 'n'; break; 212 #if '\r' != '\n' 213 /* On some platforms, \n and \r are the same. */ 214 case '\r': buff[i++] = 'r'; break; 215 #endif 216 case '\t': buff[i++] = 't'; break; 217 case '\v': buff[i++] = 'v'; break; 218 case '\\': buff[i++] = '\\'; break; 219 default: 220 sprintf(buff + i, "%03o", 0xFF & (int)c); 221 i += 3; 222 } 223 } 224 225 return (i - offset); 226 } 227 228 int 229 yes(const char *fmt, ...) 230 { 231 char buff[32]; 232 char *p; 233 ssize_t l; 234 235 va_list ap; 236 va_start(ap, fmt); 237 vfprintf(stderr, fmt, ap); 238 va_end(ap); 239 fprintf(stderr, " (y/N)? "); 240 fflush(stderr); 241 242 l = read(2, buff, sizeof(buff) - 1); 243 if (l < 0) { 244 fprintf(stderr, "Keyboard read failed\n"); 245 exit(1); 246 } 247 if (l == 0) 248 return (0); 249 buff[l] = 0; 250 251 for (p = buff; *p != '\0'; p++) { 252 if (isspace((unsigned char)*p)) 253 continue; 254 switch(*p) { 255 case 'y': case 'Y': 256 return (1); 257 case 'n': case 'N': 258 return (0); 259 default: 260 return (0); 261 } 262 } 263 264 return (0); 265 } 266 267 /*- 268 * The logic here for -C <dir> attempts to avoid 269 * chdir() as long as possible. For example: 270 * "-C /foo -C /bar file" needs chdir("/bar") but not chdir("/foo") 271 * "-C /foo -C bar file" needs chdir("/foo/bar") 272 * "-C /foo -C bar /file1" does not need chdir() 273 * "-C /foo -C bar /file1 file2" needs chdir("/foo/bar") before file2 274 * 275 * The only correct way to handle this is to record a "pending" chdir 276 * request and combine multiple requests intelligently until we 277 * need to process a non-absolute file. set_chdir() adds the new dir 278 * to the pending list; do_chdir() actually executes any pending chdir. 279 * 280 * This way, programs that build tar command lines don't have to worry 281 * about -C with non-existent directories; such requests will only 282 * fail if the directory must be accessed. 283 * 284 */ 285 void 286 set_chdir(struct bsdtar *bsdtar, const char *newdir) 287 { 288 #if defined(_WIN32) && !defined(__CYGWIN__) 289 if (newdir[0] == '/' || newdir[0] == '\\' || 290 /* Detect this type, for example, "C:\" or "C:/" */ 291 (((newdir[0] >= 'a' && newdir[0] <= 'z') || 292 (newdir[0] >= 'A' && newdir[0] <= 'Z')) && 293 newdir[1] == ':' && (newdir[2] == '/' || newdir[2] == '\\'))) { 294 #else 295 if (newdir[0] == '/') { 296 #endif 297 /* The -C /foo -C /bar case; dump first one. */ 298 free(bsdtar->pending_chdir); 299 bsdtar->pending_chdir = NULL; 300 } 301 if (bsdtar->pending_chdir == NULL) 302 /* Easy case: no previously-saved dir. */ 303 bsdtar->pending_chdir = strdup(newdir); 304 else { 305 /* The -C /foo -C bar case; concatenate */ 306 char *old_pending = bsdtar->pending_chdir; 307 size_t old_len = strlen(old_pending); 308 bsdtar->pending_chdir = malloc(old_len + strlen(newdir) + 2); 309 if (old_pending[old_len - 1] == '/') 310 old_pending[old_len - 1] = '\0'; 311 if (bsdtar->pending_chdir != NULL) 312 sprintf(bsdtar->pending_chdir, "%s/%s", 313 old_pending, newdir); 314 free(old_pending); 315 } 316 if (bsdtar->pending_chdir == NULL) 317 lafe_errc(1, errno, "No memory"); 318 } 319 320 void 321 do_chdir(struct bsdtar *bsdtar) 322 { 323 if (bsdtar->pending_chdir == NULL) 324 return; 325 326 if (chdir(bsdtar->pending_chdir) != 0) { 327 lafe_errc(1, 0, "could not chdir to '%s'\n", 328 bsdtar->pending_chdir); 329 } 330 free(bsdtar->pending_chdir); 331 bsdtar->pending_chdir = NULL; 332 } 333 334 static const char * 335 strip_components(const char *p, int elements) 336 { 337 /* Skip as many elements as necessary. */ 338 while (elements > 0) { 339 switch (*p++) { 340 case '/': 341 #if defined(_WIN32) && !defined(__CYGWIN__) 342 case '\\': /* Support \ path sep on Windows ONLY. */ 343 #endif 344 elements--; 345 break; 346 case '\0': 347 /* Path is too short, skip it. */ 348 return (NULL); 349 } 350 } 351 352 /* Skip any / characters. This handles short paths that have 353 * additional / termination. This also handles the case where 354 * the logic above stops in the middle of a duplicate // 355 * sequence (which would otherwise get converted to an 356 * absolute path). */ 357 for (;;) { 358 switch (*p) { 359 case '/': 360 #if defined(_WIN32) && !defined(__CYGWIN__) 361 case '\\': /* Support \ path sep on Windows ONLY. */ 362 #endif 363 ++p; 364 break; 365 case '\0': 366 return (NULL); 367 default: 368 return (p); 369 } 370 } 371 } 372 373 /* 374 * Handle --strip-components and any future path-rewriting options. 375 * Returns non-zero if the pathname should not be extracted. 376 * 377 * TODO: Support pax-style regex path rewrites. 378 */ 379 int 380 edit_pathname(struct bsdtar *bsdtar, struct archive_entry *entry) 381 { 382 const char *name = archive_entry_pathname(entry); 383 #if HAVE_REGEX_H 384 char *subst_name; 385 int r; 386 387 r = apply_substitution(bsdtar, name, &subst_name, 0, 0); 388 if (r == -1) { 389 lafe_warnc(0, "Invalid substitution, skipping entry"); 390 return 1; 391 } 392 if (r == 1) { 393 archive_entry_copy_pathname(entry, subst_name); 394 if (*subst_name == '\0') { 395 free(subst_name); 396 return -1; 397 } else 398 free(subst_name); 399 name = archive_entry_pathname(entry); 400 } 401 402 if (archive_entry_hardlink(entry)) { 403 r = apply_substitution(bsdtar, archive_entry_hardlink(entry), &subst_name, 0, 1); 404 if (r == -1) { 405 lafe_warnc(0, "Invalid substitution, skipping entry"); 406 return 1; 407 } 408 if (r == 1) { 409 archive_entry_copy_hardlink(entry, subst_name); 410 free(subst_name); 411 } 412 } 413 if (archive_entry_symlink(entry) != NULL) { 414 r = apply_substitution(bsdtar, archive_entry_symlink(entry), &subst_name, 1, 0); 415 if (r == -1) { 416 lafe_warnc(0, "Invalid substitution, skipping entry"); 417 return 1; 418 } 419 if (r == 1) { 420 archive_entry_copy_symlink(entry, subst_name); 421 free(subst_name); 422 } 423 } 424 #endif 425 426 /* Strip leading dir names as per --strip-components option. */ 427 if (bsdtar->strip_components > 0) { 428 const char *linkname = archive_entry_hardlink(entry); 429 430 name = strip_components(name, bsdtar->strip_components); 431 if (name == NULL) 432 return (1); 433 434 if (linkname != NULL) { 435 linkname = strip_components(linkname, 436 bsdtar->strip_components); 437 if (linkname == NULL) 438 return (1); 439 archive_entry_copy_hardlink(entry, linkname); 440 } 441 } 442 443 /* By default, don't write or restore absolute pathnames. */ 444 if (!bsdtar->option_absolute_paths) { 445 const char *rp, *p = name; 446 int slashonly = 1; 447 448 /* Remove leading "//./" or "//?/" or "//?/UNC/" 449 * (absolute path prefixes used by Windows API) */ 450 if ((p[0] == '/' || p[0] == '\\') && 451 (p[1] == '/' || p[1] == '\\') && 452 (p[2] == '.' || p[2] == '?') && 453 (p[3] == '/' || p[3] == '\\')) 454 { 455 if (p[2] == '?' && 456 (p[4] == 'U' || p[4] == 'u') && 457 (p[5] == 'N' || p[5] == 'n') && 458 (p[6] == 'C' || p[6] == 'c') && 459 (p[7] == '/' || p[7] == '\\')) 460 p += 8; 461 else 462 p += 4; 463 slashonly = 0; 464 } 465 do { 466 rp = p; 467 /* Remove leading drive letter from archives created 468 * on Windows. */ 469 if (((p[0] >= 'a' && p[0] <= 'z') || 470 (p[0] >= 'A' && p[0] <= 'Z')) && 471 p[1] == ':') { 472 p += 2; 473 slashonly = 0; 474 } 475 /* Remove leading "/../", "//", etc. */ 476 while (p[0] == '/' || p[0] == '\\') { 477 if (p[1] == '.' && p[2] == '.' && 478 (p[3] == '/' || p[3] == '\\')) { 479 p += 3; /* Remove "/..", leave "/" 480 * for next pass. */ 481 slashonly = 0; 482 } else 483 p += 1; /* Remove "/". */ 484 } 485 } while (rp != p); 486 487 if (p != name && !bsdtar->warned_lead_slash) { 488 /* Generate a warning the first time this happens. */ 489 if (slashonly) 490 lafe_warnc(0, 491 "Removing leading '%c' from member names", 492 name[0]); 493 else 494 lafe_warnc(0, 495 "Removing leading drive letter from " 496 "member names"); 497 bsdtar->warned_lead_slash = 1; 498 } 499 500 /* Special case: Stripping everything yields ".". */ 501 if (*p == '\0') 502 name = "."; 503 else 504 name = p; 505 } else { 506 /* Strip redundant leading '/' characters. */ 507 while (name[0] == '/' && name[1] == '/') 508 name++; 509 } 510 511 /* Safely replace name in archive_entry. */ 512 if (name != archive_entry_pathname(entry)) { 513 char *q = strdup(name); 514 archive_entry_copy_pathname(entry, q); 515 free(q); 516 } 517 return (0); 518 } 519 520 /* 521 * It would be nice to just use printf() for formatting large numbers, 522 * but the compatibility problems are quite a headache. Hence the 523 * following simple utility function. 524 */ 525 const char * 526 tar_i64toa(int64_t n0) 527 { 528 static char buff[24]; 529 uint64_t n = n0 < 0 ? -n0 : n0; 530 char *p = buff + sizeof(buff); 531 532 *--p = '\0'; 533 do { 534 *--p = '0' + (int)(n % 10); 535 } while (n /= 10); 536 if (n0 < 0) 537 *--p = '-'; 538 return p; 539 } 540 541 /* 542 * Like strcmp(), but try to be a little more aware of the fact that 543 * we're comparing two paths. Right now, it just handles leading 544 * "./" and trailing '/' specially, so that "a/b/" == "./a/b" 545 * 546 * TODO: Make this better, so that "./a//b/./c/" == "a/b/c" 547 * TODO: After this works, push it down into libarchive. 548 * TODO: Publish the path normalization routines in libarchive so 549 * that bsdtar can normalize paths and use fast strcmp() instead 550 * of this. 551 * 552 * Note: This is currently only used within write.c, so should 553 * not handle \ path separators. 554 */ 555 556 int 557 pathcmp(const char *a, const char *b) 558 { 559 /* Skip leading './' */ 560 if (a[0] == '.' && a[1] == '/' && a[2] != '\0') 561 a += 2; 562 if (b[0] == '.' && b[1] == '/' && b[2] != '\0') 563 b += 2; 564 /* Find the first difference, or return (0) if none. */ 565 while (*a == *b) { 566 if (*a == '\0') 567 return (0); 568 a++; 569 b++; 570 } 571 /* 572 * If one ends in '/' and the other one doesn't, 573 * they're the same. 574 */ 575 if (a[0] == '/' && a[1] == '\0' && b[0] == '\0') 576 return (0); 577 if (a[0] == '\0' && b[0] == '/' && b[1] == '\0') 578 return (0); 579 /* They're really different, return the correct sign. */ 580 return (*(const unsigned char *)a - *(const unsigned char *)b); 581 } 582