1 /*- 2 * Copyright (c) 2003-2007 Tim Kientzle 3 * Copyright (c) 2010-2012 Michihiro NAKAJIMA 4 * Copyright (c) 2016 Martin Matuska 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, 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(S) ``AS IS'' AND ANY EXPRESS OR 17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 19 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, 20 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 21 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 22 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 23 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 25 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26 */ 27 28 #include "archive_platform.h" 29 30 #ifdef HAVE_ERRNO_H 31 #include <errno.h> 32 #endif 33 #ifdef HAVE_STDLIB_H 34 #include <stdlib.h> 35 #endif 36 #ifdef HAVE_STRING_H 37 #include <string.h> 38 #endif 39 40 #include "archive.h" 41 #include "archive_entry.h" 42 #include "archive_entry_locale.h" 43 #include "archive_private.h" 44 #include "archive_write_private.h" 45 #include "archive_write_set_format_private.h" 46 47 struct sparse_block { 48 struct sparse_block *next; 49 int is_hole; 50 uint64_t offset; 51 uint64_t remaining; 52 }; 53 54 struct pax { 55 uint64_t entry_bytes_remaining; 56 uint64_t entry_padding; 57 struct archive_string l_url_encoded_name; 58 struct archive_string pax_header; 59 struct archive_string sparse_map; 60 size_t sparse_map_padding; 61 struct sparse_block *sparse_list; 62 struct sparse_block *sparse_tail; 63 struct archive_string_conv *sconv_utf8; 64 int opt_binary; 65 66 unsigned flags; 67 #define WRITE_SCHILY_XATTR (1 << 0) 68 #define WRITE_LIBARCHIVE_XATTR (1 << 1) 69 }; 70 71 static void add_pax_attr(struct archive_string *, const char *key, 72 const char *value); 73 static void add_pax_attr_binary(struct archive_string *, 74 const char *key, 75 const char *value, size_t value_len); 76 static void add_pax_attr_int(struct archive_string *, 77 const char *key, int64_t value); 78 static void add_pax_attr_time(struct archive_string *, 79 const char *key, int64_t sec, 80 unsigned long nanos); 81 static int add_pax_acl(struct archive_write *, 82 struct archive_entry *, struct pax *, int); 83 static ssize_t archive_write_pax_data(struct archive_write *, 84 const void *, size_t); 85 static int archive_write_pax_close(struct archive_write *); 86 static int archive_write_pax_free(struct archive_write *); 87 static int archive_write_pax_finish_entry(struct archive_write *); 88 static int archive_write_pax_header(struct archive_write *, 89 struct archive_entry *); 90 static int archive_write_pax_options(struct archive_write *, 91 const char *, const char *); 92 static char *base64_encode(const char *src, size_t len); 93 static char *build_gnu_sparse_name(char *dest, const char *src); 94 static char *build_pax_attribute_name(char *dest, const char *src); 95 static char *build_ustar_entry_name(char *dest, const char *src, 96 size_t src_length, const char *insert); 97 static char *format_int(char *dest, int64_t); 98 static int has_non_ASCII(const char *); 99 static void sparse_list_clear(struct pax *); 100 static int sparse_list_add(struct pax *, int64_t, int64_t); 101 static char *url_encode(const char *in); 102 static time_t get_ustar_max_mtime(void); 103 104 /* 105 * Set output format to 'restricted pax' format. 106 * 107 * This is the same as normal 'pax', but tries to suppress 108 * the pax header whenever possible. This is the default for 109 * bsdtar, for instance. 110 */ 111 int 112 archive_write_set_format_pax_restricted(struct archive *_a) 113 { 114 struct archive_write *a = (struct archive_write *)_a; 115 int r; 116 117 archive_check_magic(_a, ARCHIVE_WRITE_MAGIC, 118 ARCHIVE_STATE_NEW, "archive_write_set_format_pax_restricted"); 119 120 r = archive_write_set_format_pax(&a->archive); 121 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_RESTRICTED; 122 a->archive.archive_format_name = "restricted POSIX pax interchange"; 123 return (r); 124 } 125 126 /* 127 * Set output format to 'pax' format. 128 */ 129 int 130 archive_write_set_format_pax(struct archive *_a) 131 { 132 struct archive_write *a = (struct archive_write *)_a; 133 struct pax *pax; 134 135 archive_check_magic(_a, ARCHIVE_WRITE_MAGIC, 136 ARCHIVE_STATE_NEW, "archive_write_set_format_pax"); 137 138 if (a->format_free != NULL) 139 (a->format_free)(a); 140 141 pax = calloc(1, sizeof(*pax)); 142 if (pax == NULL) { 143 archive_set_error(&a->archive, ENOMEM, 144 "Can't allocate pax data"); 145 return (ARCHIVE_FATAL); 146 } 147 pax->flags = WRITE_LIBARCHIVE_XATTR | WRITE_SCHILY_XATTR; 148 149 a->format_data = pax; 150 a->format_name = "pax"; 151 a->format_options = archive_write_pax_options; 152 a->format_write_header = archive_write_pax_header; 153 a->format_write_data = archive_write_pax_data; 154 a->format_close = archive_write_pax_close; 155 a->format_free = archive_write_pax_free; 156 a->format_finish_entry = archive_write_pax_finish_entry; 157 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 158 a->archive.archive_format_name = "POSIX pax interchange"; 159 return (ARCHIVE_OK); 160 } 161 162 static int 163 archive_write_pax_options(struct archive_write *a, const char *key, 164 const char *val) 165 { 166 struct pax *pax = (struct pax *)a->format_data; 167 int ret = ARCHIVE_FAILED; 168 169 if (strcmp(key, "hdrcharset") == 0) { 170 /* 171 * The character-set we can use are defined in 172 * IEEE Std 1003.1-2001 173 */ 174 if (val == NULL || val[0] == 0) 175 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 176 "pax: hdrcharset option needs a character-set name"); 177 else if (strcmp(val, "BINARY") == 0 || 178 strcmp(val, "binary") == 0) { 179 /* 180 * Specify binary mode. We will not convert 181 * filenames, uname and gname to any charsets. 182 */ 183 pax->opt_binary = 1; 184 ret = ARCHIVE_OK; 185 } else if (strcmp(val, "UTF-8") == 0) { 186 /* 187 * Specify UTF-8 character-set to be used for 188 * filenames. This is almost the test that 189 * running platform supports the string conversion. 190 * Especially libarchive_test needs this trick for 191 * its test. 192 */ 193 pax->sconv_utf8 = archive_string_conversion_to_charset( 194 &(a->archive), "UTF-8", 0); 195 if (pax->sconv_utf8 == NULL) 196 ret = ARCHIVE_FATAL; 197 else 198 ret = ARCHIVE_OK; 199 } else 200 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 201 "pax: invalid charset name"); 202 return (ret); 203 } else if (strcmp(key, "xattrheader") == 0) { 204 if (val == NULL || val[0] == 0) { 205 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 206 "pax: xattrheader requires a value"); 207 } else if (strcmp(val, "ALL") == 0 || 208 strcmp(val, "all") == 0) { 209 pax->flags |= WRITE_LIBARCHIVE_XATTR | WRITE_SCHILY_XATTR; 210 ret = ARCHIVE_OK; 211 } else if (strcmp(val, "SCHILY") == 0 || 212 strcmp(val, "schily") == 0) { 213 pax->flags |= WRITE_SCHILY_XATTR; 214 pax->flags &= ~WRITE_LIBARCHIVE_XATTR; 215 ret = ARCHIVE_OK; 216 } else if (strcmp(val, "LIBARCHIVE") == 0 || 217 strcmp(val, "libarchive") == 0) { 218 pax->flags |= WRITE_LIBARCHIVE_XATTR; 219 pax->flags &= ~WRITE_SCHILY_XATTR; 220 ret = ARCHIVE_OK; 221 } else 222 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 223 "pax: invalid xattr header name"); 224 return (ret); 225 } 226 227 /* Note: The "warn" return is just to inform the options 228 * supervisor that we didn't handle it. It will generate 229 * a suitable error if no one used this option. */ 230 return (ARCHIVE_WARN); 231 } 232 233 /* 234 * Note: This code assumes that 'nanos' has the same sign as 'sec', 235 * which implies that sec=-1, nanos=200000000 represents -1.2 seconds 236 * and not -0.8 seconds. This is a pretty pedantic point, as we're 237 * unlikely to encounter many real files created before Jan 1, 1970, 238 * much less ones with timestamps recorded to sub-second resolution. 239 */ 240 static void 241 add_pax_attr_time(struct archive_string *as, const char *key, 242 int64_t sec, unsigned long nanos) 243 { 244 int digit, i; 245 char *t; 246 /* 247 * Note that each byte contributes fewer than 3 base-10 248 * digits, so this will always be big enough. 249 */ 250 char tmp[1 + 3*sizeof(sec) + 1 + 3*sizeof(nanos)]; 251 252 tmp[sizeof(tmp) - 1] = 0; 253 t = tmp + sizeof(tmp) - 1; 254 255 /* Skip trailing zeros in the fractional part. */ 256 for (digit = 0, i = 10; i > 0 && digit == 0; i--) { 257 digit = nanos % 10; 258 nanos /= 10; 259 } 260 261 /* Only format the fraction if it's non-zero. */ 262 if (i > 0) { 263 while (i > 0) { 264 *--t = "0123456789"[digit]; 265 digit = nanos % 10; 266 nanos /= 10; 267 i--; 268 } 269 *--t = '.'; 270 } 271 t = format_int(t, sec); 272 273 add_pax_attr(as, key, t); 274 } 275 276 static char * 277 format_int(char *t, int64_t i) 278 { 279 uint64_t ui; 280 281 if (i < 0) 282 ui = (i == INT64_MIN) ? (uint64_t)(INT64_MAX) + 1 : (uint64_t)(-i); 283 else 284 ui = i; 285 286 do { 287 *--t = "0123456789"[ui % 10]; 288 } while (ui /= 10); 289 if (i < 0) 290 *--t = '-'; 291 return (t); 292 } 293 294 static void 295 add_pax_attr_int(struct archive_string *as, const char *key, int64_t value) 296 { 297 char tmp[1 + 3 * sizeof(value)]; 298 299 tmp[sizeof(tmp) - 1] = 0; 300 add_pax_attr(as, key, format_int(tmp + sizeof(tmp) - 1, value)); 301 } 302 303 /* 304 * Add a key/value attribute to the pax header. This function handles 305 * the length field and various other syntactic requirements. 306 */ 307 static void 308 add_pax_attr(struct archive_string *as, const char *key, const char *value) 309 { 310 add_pax_attr_binary(as, key, value, strlen(value)); 311 } 312 313 /* 314 * Add a key/value attribute to the pax header. This function handles 315 * binary values. 316 */ 317 static void 318 add_pax_attr_binary(struct archive_string *as, const char *key, 319 const char *value, size_t value_len) 320 { 321 int digits, i, len, next_ten; 322 char tmp[1 + 3 * sizeof(int)]; /* < 3 base-10 digits per byte */ 323 324 /*- 325 * PAX attributes have the following layout: 326 * <len> <space> <key> <=> <value> <nl> 327 */ 328 len = 1 + (int)strlen(key) + 1 + (int)value_len + 1; 329 330 /* 331 * The <len> field includes the length of the <len> field, so 332 * computing the correct length is tricky. I start by 333 * counting the number of base-10 digits in 'len' and 334 * computing the next higher power of 10. 335 */ 336 next_ten = 1; 337 digits = 0; 338 i = len; 339 while (i > 0) { 340 i = i / 10; 341 digits++; 342 next_ten = next_ten * 10; 343 } 344 /* 345 * For example, if string without the length field is 99 346 * chars, then adding the 2 digit length "99" will force the 347 * total length past 100, requiring an extra digit. The next 348 * statement adjusts for this effect. 349 */ 350 if (len + digits >= next_ten) 351 digits++; 352 353 /* Now, we have the right length so we can build the line. */ 354 tmp[sizeof(tmp) - 1] = 0; /* Null-terminate the work area. */ 355 archive_strcat(as, format_int(tmp + sizeof(tmp) - 1, len + digits)); 356 archive_strappend_char(as, ' '); 357 archive_strcat(as, key); 358 archive_strappend_char(as, '='); 359 archive_array_append(as, value, value_len); 360 archive_strappend_char(as, '\n'); 361 } 362 363 static void 364 archive_write_pax_header_xattr(struct pax *pax, const char *encoded_name, 365 const void *value, size_t value_len) 366 { 367 struct archive_string s; 368 char *encoded_value; 369 370 if (encoded_name == NULL) 371 return; 372 373 if (pax->flags & WRITE_LIBARCHIVE_XATTR) { 374 encoded_value = base64_encode((const char *)value, value_len); 375 if (encoded_value != NULL) { 376 archive_string_init(&s); 377 archive_strcpy(&s, "LIBARCHIVE.xattr."); 378 archive_strcat(&s, encoded_name); 379 add_pax_attr(&(pax->pax_header), s.s, encoded_value); 380 archive_string_free(&s); 381 } 382 free(encoded_value); 383 } 384 if (pax->flags & WRITE_SCHILY_XATTR) { 385 archive_string_init(&s); 386 archive_strcpy(&s, "SCHILY.xattr."); 387 archive_strcat(&s, encoded_name); 388 add_pax_attr_binary(&(pax->pax_header), s.s, value, value_len); 389 archive_string_free(&s); 390 } 391 } 392 393 static int 394 archive_write_pax_header_xattrs(struct archive_write *a, 395 struct pax *pax, struct archive_entry *entry) 396 { 397 int i = archive_entry_xattr_reset(entry); 398 399 while (i--) { 400 const char *name; 401 const void *value; 402 char *url_encoded_name = NULL, *encoded_name = NULL; 403 size_t size; 404 int r; 405 406 archive_entry_xattr_next(entry, &name, &value, &size); 407 url_encoded_name = url_encode(name); 408 if (url_encoded_name == NULL) 409 goto malloc_error; 410 else { 411 /* Convert narrow-character to UTF-8. */ 412 r = archive_strcpy_l(&(pax->l_url_encoded_name), 413 url_encoded_name, pax->sconv_utf8); 414 free(url_encoded_name); /* Done with this. */ 415 if (r == 0) 416 encoded_name = pax->l_url_encoded_name.s; 417 else if (r == -1) 418 goto malloc_error; 419 else { 420 archive_set_error(&a->archive, 421 ARCHIVE_ERRNO_MISC, 422 "Error encoding pax extended attribute"); 423 return (ARCHIVE_FAILED); 424 } 425 } 426 427 archive_write_pax_header_xattr(pax, encoded_name, 428 value, size); 429 430 } 431 return (ARCHIVE_OK); 432 malloc_error: 433 archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); 434 return (ARCHIVE_FATAL); 435 } 436 437 static int 438 get_entry_hardlink(struct archive_write *a, struct archive_entry *entry, 439 const char **name, size_t *length, struct archive_string_conv *sc) 440 { 441 int r; 442 443 r = archive_entry_hardlink_l(entry, name, length, sc); 444 if (r != 0) { 445 if (errno == ENOMEM) { 446 archive_set_error(&a->archive, ENOMEM, 447 "Can't allocate memory for Linkname"); 448 return (ARCHIVE_FATAL); 449 } 450 return (ARCHIVE_WARN); 451 } 452 return (ARCHIVE_OK); 453 } 454 455 static int 456 get_entry_pathname(struct archive_write *a, struct archive_entry *entry, 457 const char **name, size_t *length, struct archive_string_conv *sc) 458 { 459 int r; 460 461 r = archive_entry_pathname_l(entry, name, length, sc); 462 if (r != 0) { 463 if (errno == ENOMEM) { 464 archive_set_error(&a->archive, ENOMEM, 465 "Can't allocate memory for Pathname"); 466 return (ARCHIVE_FATAL); 467 } 468 return (ARCHIVE_WARN); 469 } 470 return (ARCHIVE_OK); 471 } 472 473 static int 474 get_entry_uname(struct archive_write *a, struct archive_entry *entry, 475 const char **name, size_t *length, struct archive_string_conv *sc) 476 { 477 int r; 478 479 r = archive_entry_uname_l(entry, name, length, sc); 480 if (r != 0) { 481 if (errno == ENOMEM) { 482 archive_set_error(&a->archive, ENOMEM, 483 "Can't allocate memory for Uname"); 484 return (ARCHIVE_FATAL); 485 } 486 return (ARCHIVE_WARN); 487 } 488 return (ARCHIVE_OK); 489 } 490 491 static int 492 get_entry_gname(struct archive_write *a, struct archive_entry *entry, 493 const char **name, size_t *length, struct archive_string_conv *sc) 494 { 495 int r; 496 497 r = archive_entry_gname_l(entry, name, length, sc); 498 if (r != 0) { 499 if (errno == ENOMEM) { 500 archive_set_error(&a->archive, ENOMEM, 501 "Can't allocate memory for Gname"); 502 return (ARCHIVE_FATAL); 503 } 504 return (ARCHIVE_WARN); 505 } 506 return (ARCHIVE_OK); 507 } 508 509 static int 510 get_entry_symlink(struct archive_write *a, struct archive_entry *entry, 511 const char **name, size_t *length, struct archive_string_conv *sc) 512 { 513 int r; 514 515 r = archive_entry_symlink_l(entry, name, length, sc); 516 if (r != 0) { 517 if (errno == ENOMEM) { 518 archive_set_error(&a->archive, ENOMEM, 519 "Can't allocate memory for Linkname"); 520 return (ARCHIVE_FATAL); 521 } 522 return (ARCHIVE_WARN); 523 } 524 return (ARCHIVE_OK); 525 } 526 527 /* Add ACL to pax header */ 528 static int 529 add_pax_acl(struct archive_write *a, 530 struct archive_entry *entry, struct pax *pax, int flags) 531 { 532 char *p; 533 const char *attr; 534 int acl_types; 535 536 acl_types = archive_entry_acl_types(entry); 537 538 if ((acl_types & ARCHIVE_ENTRY_ACL_TYPE_NFS4) != 0) 539 attr = "SCHILY.acl.ace"; 540 else if ((flags & ARCHIVE_ENTRY_ACL_TYPE_ACCESS) != 0) 541 attr = "SCHILY.acl.access"; 542 else if ((flags & ARCHIVE_ENTRY_ACL_TYPE_DEFAULT) != 0) 543 attr = "SCHILY.acl.default"; 544 else 545 return (ARCHIVE_FATAL); 546 547 p = archive_entry_acl_to_text_l(entry, NULL, flags, pax->sconv_utf8); 548 if (p == NULL) { 549 if (errno == ENOMEM) { 550 archive_set_error(&a->archive, ENOMEM, "%s %s", 551 "Can't allocate memory for ", attr); 552 return (ARCHIVE_FATAL); 553 } 554 archive_set_error(&a->archive, 555 ARCHIVE_ERRNO_FILE_FORMAT, "%s %s %s", 556 "Can't translate ", attr, " to UTF-8"); 557 return(ARCHIVE_WARN); 558 } 559 560 if (*p != '\0') { 561 add_pax_attr(&(pax->pax_header), 562 attr, p); 563 } 564 free(p); 565 return(ARCHIVE_OK); 566 } 567 568 /* 569 * TODO: Consider adding 'comment' and 'charset' fields to 570 * archive_entry so that clients can specify them. Also, consider 571 * adding generic key/value tags so clients can add arbitrary 572 * key/value data. 573 * 574 * TODO: Break up this 700-line function!!!! Yowza! 575 */ 576 static int 577 archive_write_pax_header(struct archive_write *a, 578 struct archive_entry *entry_original) 579 { 580 struct archive_entry *entry_main; 581 const char *p; 582 const char *suffix; 583 int need_extension, r, ret; 584 int acl_types; 585 int sparse_count; 586 uint64_t sparse_total, real_size; 587 struct pax *pax; 588 const char *hardlink; 589 const char *path = NULL, *linkpath = NULL; 590 const char *uname = NULL, *gname = NULL; 591 const void *mac_metadata; 592 size_t mac_metadata_size; 593 struct archive_string_conv *sconv; 594 size_t hardlink_length, path_length, linkpath_length; 595 size_t uname_length, gname_length; 596 597 char paxbuff[512]; 598 char ustarbuff[512]; 599 char ustar_entry_name[256]; 600 char pax_entry_name[256]; 601 char gnu_sparse_name[256]; 602 struct archive_string entry_name; 603 604 ret = ARCHIVE_OK; 605 need_extension = 0; 606 pax = (struct pax *)a->format_data; 607 608 const time_t ustar_max_mtime = get_ustar_max_mtime(); 609 610 /* Sanity check. */ 611 #if defined(_WIN32) && !defined(__CYGWIN__) 612 /* NOTE: If the caller supplied a pathname that fails WCS conversion (e.g. 613 * if it is invalid UTF-8), we are expected to return ARCHIVE_WARN later on 614 * in execution, hence the check for both pointers */ 615 if ((archive_entry_pathname_w(entry_original) == NULL) && 616 (archive_entry_pathname(entry_original) == NULL)) { 617 #else 618 if (archive_entry_pathname(entry_original) == NULL) { 619 #endif 620 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 621 "Can't record entry in tar file without pathname"); 622 return (ARCHIVE_FAILED); 623 } 624 625 /* 626 * Choose a header encoding. 627 */ 628 if (pax->opt_binary) 629 sconv = NULL;/* Binary mode. */ 630 else { 631 /* Header encoding is UTF-8. */ 632 if (pax->sconv_utf8 == NULL) { 633 /* Initialize the string conversion object 634 * we must need */ 635 pax->sconv_utf8 = archive_string_conversion_to_charset( 636 &(a->archive), "UTF-8", 1); 637 if (pax->sconv_utf8 == NULL) 638 /* Couldn't allocate memory */ 639 return (ARCHIVE_FAILED); 640 } 641 sconv = pax->sconv_utf8; 642 } 643 644 r = get_entry_hardlink(a, entry_original, &hardlink, 645 &hardlink_length, sconv); 646 if (r == ARCHIVE_FATAL) 647 return (r); 648 else if (r != ARCHIVE_OK) { 649 r = get_entry_hardlink(a, entry_original, &hardlink, 650 &hardlink_length, NULL); 651 if (r == ARCHIVE_FATAL) 652 return (r); 653 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 654 "Can't translate linkname '%s' to %s", hardlink, 655 archive_string_conversion_charset_name(sconv)); 656 ret = ARCHIVE_WARN; 657 sconv = NULL;/* The header charset switches to binary mode. */ 658 } 659 660 /* Make sure this is a type of entry that we can handle here */ 661 if (hardlink == NULL) { 662 switch (archive_entry_filetype(entry_original)) { 663 case AE_IFBLK: 664 case AE_IFCHR: 665 case AE_IFIFO: 666 case AE_IFLNK: 667 case AE_IFREG: 668 break; 669 case AE_IFDIR: 670 { 671 /* 672 * Ensure a trailing '/'. Modify the original 673 * entry so the client sees the change. 674 */ 675 #if defined(_WIN32) && !defined(__CYGWIN__) 676 const wchar_t *wp; 677 678 wp = archive_entry_pathname_w(entry_original); 679 if (wp != NULL && wp[wcslen(wp) -1] != L'/') { 680 struct archive_wstring ws; 681 682 archive_string_init(&ws); 683 path_length = wcslen(wp); 684 if (archive_wstring_ensure(&ws, 685 path_length + 2) == NULL) { 686 archive_set_error(&a->archive, ENOMEM, 687 "Can't allocate pax data"); 688 archive_wstring_free(&ws); 689 return(ARCHIVE_FATAL); 690 } 691 /* Should we keep '\' ? */ 692 if (wp[path_length -1] == L'\\') 693 path_length--; 694 archive_wstrncpy(&ws, wp, path_length); 695 archive_wstrappend_wchar(&ws, L'/'); 696 archive_entry_copy_pathname_w( 697 entry_original, ws.s); 698 archive_wstring_free(&ws); 699 p = NULL; 700 } else 701 #endif 702 p = archive_entry_pathname(entry_original); 703 /* 704 * On Windows, this is a backup operation just in 705 * case getting WCS failed. On POSIX, this is a 706 * normal operation. 707 */ 708 if (p != NULL && p[0] != '\0' && p[strlen(p) - 1] != '/') { 709 struct archive_string as; 710 711 archive_string_init(&as); 712 path_length = strlen(p); 713 if (archive_string_ensure(&as, 714 path_length + 2) == NULL) { 715 archive_set_error(&a->archive, ENOMEM, 716 "Can't allocate pax data"); 717 archive_string_free(&as); 718 return(ARCHIVE_FATAL); 719 } 720 #if defined(_WIN32) && !defined(__CYGWIN__) 721 /* NOTE: This might break the pathname 722 * if the current code page is CP932 and 723 * the pathname includes a character '\' 724 * as a part of its multibyte pathname. */ 725 if (p[strlen(p) -1] == '\\') 726 path_length--; 727 else 728 #endif 729 archive_strncpy(&as, p, path_length); 730 archive_strappend_char(&as, '/'); 731 archive_entry_copy_pathname( 732 entry_original, as.s); 733 archive_string_free(&as); 734 } 735 break; 736 } 737 default: /* AE_IFSOCK and unknown */ 738 __archive_write_entry_filetype_unsupported( 739 &a->archive, entry_original, "pax"); 740 return (ARCHIVE_FAILED); 741 } 742 } 743 744 /* 745 * If Mac OS metadata blob is here, recurse to write that 746 * as a separate entry. This is really a pretty poor design: 747 * In particular, it doubles the overhead for long filenames. 748 * TODO: Help Apple folks design something better and figure 749 * out how to transition from this legacy format. 750 * 751 * Note that this code is present on every platform; clients 752 * on non-Mac are unlikely to ever provide this data, but 753 * applications that copy entries from one archive to another 754 * should not lose data just because the local filesystem 755 * can't store it. 756 */ 757 mac_metadata = 758 archive_entry_mac_metadata(entry_original, &mac_metadata_size); 759 if (mac_metadata != NULL) { 760 const char *oname; 761 char *name, *bname; 762 size_t name_length; 763 struct archive_entry *extra = archive_entry_new2(&a->archive); 764 765 oname = archive_entry_pathname(entry_original); 766 name_length = strlen(oname); 767 name = malloc(name_length + 3); 768 if (name == NULL || extra == NULL) { 769 /* XXX error message */ 770 archive_entry_free(extra); 771 free(name); 772 return (ARCHIVE_FAILED); 773 } 774 strcpy(name, oname); 775 /* Find last '/'; strip trailing '/' characters */ 776 bname = strrchr(name, '/'); 777 while (bname != NULL && bname[1] == '\0') { 778 *bname = '\0'; 779 bname = strrchr(name, '/'); 780 } 781 if (bname == NULL) { 782 memmove(name + 2, name, name_length + 1); 783 memmove(name, "._", 2); 784 } else { 785 bname += 1; 786 memmove(bname + 2, bname, strlen(bname) + 1); 787 memmove(bname, "._", 2); 788 } 789 archive_entry_copy_pathname(extra, name); 790 free(name); 791 792 archive_entry_set_size(extra, mac_metadata_size); 793 archive_entry_set_filetype(extra, AE_IFREG); 794 archive_entry_set_perm(extra, 795 archive_entry_perm(entry_original)); 796 archive_entry_set_mtime(extra, 797 archive_entry_mtime(entry_original), 798 archive_entry_mtime_nsec(entry_original)); 799 archive_entry_set_gid(extra, 800 archive_entry_gid(entry_original)); 801 archive_entry_set_gname(extra, 802 archive_entry_gname(entry_original)); 803 archive_entry_set_uid(extra, 804 archive_entry_uid(entry_original)); 805 archive_entry_set_uname(extra, 806 archive_entry_uname(entry_original)); 807 808 /* Recurse to write the special copyfile entry. */ 809 r = archive_write_pax_header(a, extra); 810 archive_entry_free(extra); 811 if (r < ARCHIVE_WARN) 812 return (r); 813 if (r < ret) 814 ret = r; 815 r = (int)archive_write_pax_data(a, mac_metadata, 816 mac_metadata_size); 817 if (r < ARCHIVE_WARN) 818 return (r); 819 if (r < ret) 820 ret = r; 821 r = archive_write_pax_finish_entry(a); 822 if (r < ARCHIVE_WARN) 823 return (r); 824 if (r < ret) 825 ret = r; 826 } 827 828 /* Copy entry so we can modify it as needed. */ 829 #if defined(_WIN32) && !defined(__CYGWIN__) 830 /* Make sure the path separators in pathname, hardlink and symlink 831 * are all slash '/', not the Windows path separator '\'. */ 832 entry_main = __la_win_entry_in_posix_pathseparator(entry_original); 833 if (entry_main == entry_original) 834 entry_main = archive_entry_clone(entry_original); 835 #else 836 entry_main = archive_entry_clone(entry_original); 837 #endif 838 if (entry_main == NULL) { 839 archive_set_error(&a->archive, ENOMEM, 840 "Can't allocate pax data"); 841 return(ARCHIVE_FATAL); 842 } 843 archive_string_empty(&(pax->pax_header)); /* Blank our work area. */ 844 archive_string_empty(&(pax->sparse_map)); 845 sparse_total = 0; 846 sparse_list_clear(pax); 847 848 if (hardlink == NULL && 849 archive_entry_filetype(entry_main) == AE_IFREG) 850 sparse_count = archive_entry_sparse_reset(entry_main); 851 else 852 sparse_count = 0; 853 if (sparse_count) { 854 int64_t offset, length, last_offset = 0; 855 /* Get the last entry of sparse block. */ 856 while (archive_entry_sparse_next( 857 entry_main, &offset, &length) == ARCHIVE_OK) 858 last_offset = offset + length; 859 860 /* If the last sparse block does not reach the end of file, 861 * We have to add a empty sparse block as the last entry to 862 * manage storing file data. */ 863 if (last_offset < archive_entry_size(entry_main)) 864 archive_entry_sparse_add_entry(entry_main, 865 archive_entry_size(entry_main), 0); 866 sparse_count = archive_entry_sparse_reset(entry_main); 867 } 868 869 /* 870 * First, check the name fields and see if any of them 871 * require binary coding. If any of them does, then all of 872 * them do. 873 */ 874 r = get_entry_pathname(a, entry_main, &path, &path_length, sconv); 875 if (r == ARCHIVE_FATAL) { 876 archive_entry_free(entry_main); 877 return (r); 878 } else if (r != ARCHIVE_OK) { 879 r = get_entry_pathname(a, entry_main, &path, 880 &path_length, NULL); 881 if (r == ARCHIVE_FATAL) { 882 archive_entry_free(entry_main); 883 return (r); 884 } 885 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 886 "Can't translate pathname '%s' to %s", path, 887 archive_string_conversion_charset_name(sconv)); 888 ret = ARCHIVE_WARN; 889 sconv = NULL;/* The header charset switches to binary mode. */ 890 } 891 r = get_entry_uname(a, entry_main, &uname, &uname_length, sconv); 892 if (r == ARCHIVE_FATAL) { 893 archive_entry_free(entry_main); 894 return (r); 895 } else if (r != ARCHIVE_OK) { 896 r = get_entry_uname(a, entry_main, &uname, &uname_length, NULL); 897 if (r == ARCHIVE_FATAL) { 898 archive_entry_free(entry_main); 899 return (r); 900 } 901 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 902 "Can't translate uname '%s' to %s", uname, 903 archive_string_conversion_charset_name(sconv)); 904 ret = ARCHIVE_WARN; 905 sconv = NULL;/* The header charset switches to binary mode. */ 906 } 907 r = get_entry_gname(a, entry_main, &gname, &gname_length, sconv); 908 if (r == ARCHIVE_FATAL) { 909 archive_entry_free(entry_main); 910 return (r); 911 } else if (r != ARCHIVE_OK) { 912 r = get_entry_gname(a, entry_main, &gname, &gname_length, NULL); 913 if (r == ARCHIVE_FATAL) { 914 archive_entry_free(entry_main); 915 return (r); 916 } 917 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 918 "Can't translate gname '%s' to %s", gname, 919 archive_string_conversion_charset_name(sconv)); 920 ret = ARCHIVE_WARN; 921 sconv = NULL;/* The header charset switches to binary mode. */ 922 } 923 linkpath = hardlink; 924 linkpath_length = hardlink_length; 925 if (linkpath == NULL) { 926 r = get_entry_symlink(a, entry_main, &linkpath, 927 &linkpath_length, sconv); 928 if (r == ARCHIVE_FATAL) { 929 archive_entry_free(entry_main); 930 return (r); 931 } else if (r != ARCHIVE_OK) { 932 r = get_entry_symlink(a, entry_main, &linkpath, 933 &linkpath_length, NULL); 934 if (r == ARCHIVE_FATAL) { 935 archive_entry_free(entry_main); 936 return (r); 937 } 938 archive_set_error(&a->archive, 939 ARCHIVE_ERRNO_FILE_FORMAT, 940 "Can't translate linkname '%s' to %s", linkpath, 941 archive_string_conversion_charset_name(sconv)); 942 ret = ARCHIVE_WARN; 943 sconv = NULL; 944 } 945 } 946 947 /* If any string conversions failed, get all attributes 948 * in binary-mode. */ 949 if (sconv == NULL && !pax->opt_binary) { 950 if (hardlink != NULL) { 951 r = get_entry_hardlink(a, entry_main, &hardlink, 952 &hardlink_length, NULL); 953 if (r == ARCHIVE_FATAL) { 954 archive_entry_free(entry_main); 955 return (r); 956 } 957 linkpath = hardlink; 958 linkpath_length = hardlink_length; 959 } 960 r = get_entry_pathname(a, entry_main, &path, 961 &path_length, NULL); 962 if (r == ARCHIVE_FATAL) { 963 archive_entry_free(entry_main); 964 return (r); 965 } 966 r = get_entry_uname(a, entry_main, &uname, &uname_length, NULL); 967 if (r == ARCHIVE_FATAL) { 968 archive_entry_free(entry_main); 969 return (r); 970 } 971 r = get_entry_gname(a, entry_main, &gname, &gname_length, NULL); 972 if (r == ARCHIVE_FATAL) { 973 archive_entry_free(entry_main); 974 return (r); 975 } 976 } 977 978 /* Store the header encoding first, to be nice to readers. */ 979 if (sconv == NULL) 980 add_pax_attr(&(pax->pax_header), "hdrcharset", "BINARY"); 981 982 983 /* 984 * If name is too long, or has non-ASCII characters, add 985 * 'path' to pax extended attrs. (Note that an unconvertible 986 * name must have non-ASCII characters.) 987 */ 988 if (has_non_ASCII(path)) { 989 /* We have non-ASCII characters. */ 990 add_pax_attr(&(pax->pax_header), "path", path); 991 archive_entry_set_pathname(entry_main, 992 build_ustar_entry_name(ustar_entry_name, 993 path, path_length, NULL)); 994 need_extension = 1; 995 } else { 996 /* We have an all-ASCII path; we'd like to just store 997 * it in the ustar header if it will fit. Yes, this 998 * duplicates some of the logic in 999 * archive_write_set_format_ustar.c 1000 */ 1001 if (path_length <= 100) { 1002 /* Fits in the old 100-char tar name field. */ 1003 } else { 1004 /* Find largest suffix that will fit. */ 1005 /* Note: strlen() > 100, so strlen() - 100 - 1 >= 0 */ 1006 suffix = strchr(path + path_length - 100 - 1, '/'); 1007 /* Don't attempt an empty prefix. */ 1008 if (suffix == path) 1009 suffix = strchr(suffix + 1, '/'); 1010 /* We can put it in the ustar header if it's 1011 * all ASCII and it's either <= 100 characters 1012 * or can be split at a '/' into a prefix <= 1013 * 155 chars and a suffix <= 100 chars. (Note 1014 * the strchr() above will return NULL exactly 1015 * when the path can't be split.) 1016 */ 1017 if (suffix == NULL /* Suffix > 100 chars. */ 1018 || suffix[1] == '\0' /* empty suffix */ 1019 || suffix - path > 155) /* Prefix > 155 chars */ 1020 { 1021 add_pax_attr(&(pax->pax_header), "path", path); 1022 archive_entry_set_pathname(entry_main, 1023 build_ustar_entry_name(ustar_entry_name, 1024 path, path_length, NULL)); 1025 need_extension = 1; 1026 } 1027 } 1028 } 1029 1030 if (linkpath != NULL) { 1031 /* If link name is too long or has non-ASCII characters, add 1032 * 'linkpath' to pax extended attrs. */ 1033 if (linkpath_length > 100 || has_non_ASCII(linkpath)) { 1034 add_pax_attr(&(pax->pax_header), "linkpath", linkpath); 1035 if (linkpath_length > 100) { 1036 if (hardlink != NULL) 1037 archive_entry_set_hardlink(entry_main, 1038 "././@LongHardLink"); 1039 else 1040 archive_entry_set_symlink(entry_main, 1041 "././@LongSymLink"); 1042 } 1043 else { 1044 /* Otherwise, has non-ASCII characters; update the paths to 1045 * however they got decoded above */ 1046 if (hardlink != NULL) 1047 archive_entry_set_hardlink(entry_main, linkpath); 1048 else 1049 archive_entry_set_symlink(entry_main, linkpath); 1050 } 1051 need_extension = 1; 1052 } 1053 } 1054 /* Save a pathname since it will be renamed if `entry_main` has 1055 * sparse blocks. */ 1056 archive_string_init(&entry_name); 1057 archive_strcpy(&entry_name, archive_entry_pathname(entry_main)); 1058 1059 /* If file size is too large, we need pax extended attrs. */ 1060 if (archive_entry_size(entry_main) >= (((int64_t)1) << 33)) { 1061 need_extension = 1; 1062 } 1063 1064 /* If numeric GID is too large, add 'gid' to pax extended attrs. */ 1065 if (archive_entry_gid(entry_main) >= (1 << 18)) { 1066 add_pax_attr_int(&(pax->pax_header), "gid", 1067 archive_entry_gid(entry_main)); 1068 need_extension = 1; 1069 } 1070 1071 /* If group name is too large or has non-ASCII characters, add 1072 * 'gname' to pax extended attrs. */ 1073 if (gname != NULL) { 1074 if (gname_length > 31 || has_non_ASCII(gname)) { 1075 add_pax_attr(&(pax->pax_header), "gname", gname); 1076 need_extension = 1; 1077 } 1078 } 1079 1080 /* If numeric UID is too large, add 'uid' to pax extended attrs. */ 1081 if (archive_entry_uid(entry_main) >= (1 << 18)) { 1082 add_pax_attr_int(&(pax->pax_header), "uid", 1083 archive_entry_uid(entry_main)); 1084 need_extension = 1; 1085 } 1086 1087 /* Add 'uname' to pax extended attrs if necessary. */ 1088 if (uname != NULL) { 1089 if (uname_length > 31 || has_non_ASCII(uname)) { 1090 add_pax_attr(&(pax->pax_header), "uname", uname); 1091 need_extension = 1; 1092 } 1093 } 1094 1095 /* 1096 * POSIX/SUSv3 doesn't provide a standard key for large device 1097 * numbers. I use the same keys here that Joerg Schilling 1098 * used for 'star.' (Which, somewhat confusingly, are called 1099 * "devXXX" even though they code "rdev" values.) No doubt, 1100 * other implementations use other keys. Note that there's no 1101 * reason we can't write the same information into a number of 1102 * different keys. 1103 * 1104 * Of course, this is only needed for block or char device entries. 1105 */ 1106 if (archive_entry_filetype(entry_main) == AE_IFBLK 1107 || archive_entry_filetype(entry_main) == AE_IFCHR) { 1108 /* 1109 * If rdevmajor is too large, add 'SCHILY.devmajor' to 1110 * extended attributes. 1111 */ 1112 int rdevmajor, rdevminor; 1113 rdevmajor = archive_entry_rdevmajor(entry_main); 1114 rdevminor = archive_entry_rdevminor(entry_main); 1115 if (rdevmajor >= (1 << 18)) { 1116 add_pax_attr_int(&(pax->pax_header), "SCHILY.devmajor", 1117 rdevmajor); 1118 /* 1119 * Non-strict formatting below means we don't 1120 * have to truncate here. Not truncating improves 1121 * the chance that some more modern tar archivers 1122 * (such as GNU tar 1.13) can restore the full 1123 * value even if they don't understand the pax 1124 * extended attributes. See my rant below about 1125 * file size fields for additional details. 1126 */ 1127 /* archive_entry_set_rdevmajor(entry_main, 1128 rdevmajor & ((1 << 18) - 1)); */ 1129 need_extension = 1; 1130 } 1131 1132 /* 1133 * If devminor is too large, add 'SCHILY.devminor' to 1134 * extended attributes. 1135 */ 1136 if (rdevminor >= (1 << 18)) { 1137 add_pax_attr_int(&(pax->pax_header), "SCHILY.devminor", 1138 rdevminor); 1139 /* Truncation is not necessary here, either. */ 1140 /* archive_entry_set_rdevminor(entry_main, 1141 rdevminor & ((1 << 18) - 1)); */ 1142 need_extension = 1; 1143 } 1144 } 1145 1146 /* 1147 * Yes, this check is duplicated just below; this helps to 1148 * avoid writing an mtime attribute just to handle a 1149 * high-resolution timestamp in "restricted pax" mode. 1150 */ 1151 if (!need_extension && 1152 ((archive_entry_mtime(entry_main) < 0) 1153 || (archive_entry_mtime(entry_main) >= ustar_max_mtime))) 1154 need_extension = 1; 1155 1156 /* I use a star-compatible file flag attribute. */ 1157 p = archive_entry_fflags_text(entry_main); 1158 if (!need_extension && p != NULL && *p != '\0') 1159 need_extension = 1; 1160 1161 /* If there are extended attributes, we need an extension */ 1162 if (!need_extension && archive_entry_xattr_count(entry_original) > 0) 1163 need_extension = 1; 1164 1165 /* If there are sparse info, we need an extension */ 1166 if (!need_extension && sparse_count > 0) 1167 need_extension = 1; 1168 1169 acl_types = archive_entry_acl_types(entry_original); 1170 1171 /* If there are any ACL entries, we need an extension */ 1172 if (!need_extension && acl_types != 0) 1173 need_extension = 1; 1174 1175 /* If the symlink type is defined, we need an extension */ 1176 if (!need_extension && archive_entry_symlink_type(entry_main) > 0) 1177 need_extension = 1; 1178 1179 /* 1180 * Libarchive used to include these in extended headers for 1181 * restricted pax format, but that confused people who 1182 * expected ustar-like time semantics. So now we only include 1183 * them in full pax format. 1184 */ 1185 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_RESTRICTED) { 1186 if (archive_entry_ctime(entry_main) != 0 || 1187 archive_entry_ctime_nsec(entry_main) != 0) 1188 add_pax_attr_time(&(pax->pax_header), "ctime", 1189 archive_entry_ctime(entry_main), 1190 archive_entry_ctime_nsec(entry_main)); 1191 1192 if (archive_entry_atime(entry_main) != 0 || 1193 archive_entry_atime_nsec(entry_main) != 0) 1194 add_pax_attr_time(&(pax->pax_header), "atime", 1195 archive_entry_atime(entry_main), 1196 archive_entry_atime_nsec(entry_main)); 1197 1198 /* Store birth/creationtime only if it's earlier than mtime */ 1199 if (archive_entry_birthtime_is_set(entry_main) && 1200 archive_entry_birthtime(entry_main) 1201 < archive_entry_mtime(entry_main)) 1202 add_pax_attr_time(&(pax->pax_header), 1203 "LIBARCHIVE.creationtime", 1204 archive_entry_birthtime(entry_main), 1205 archive_entry_birthtime_nsec(entry_main)); 1206 } 1207 1208 /* 1209 * The following items are handled differently in "pax 1210 * restricted" format. In particular, in "pax restricted" 1211 * format they won't be added unless need_extension is 1212 * already set (we're already generating an extended header, so 1213 * may as well include these). 1214 */ 1215 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_RESTRICTED || 1216 need_extension) { 1217 if (archive_entry_mtime(entry_main) < 0 || 1218 archive_entry_mtime(entry_main) >= ustar_max_mtime || 1219 archive_entry_mtime_nsec(entry_main) != 0) 1220 add_pax_attr_time(&(pax->pax_header), "mtime", 1221 archive_entry_mtime(entry_main), 1222 archive_entry_mtime_nsec(entry_main)); 1223 1224 /* I use a star-compatible file flag attribute. */ 1225 p = archive_entry_fflags_text(entry_main); 1226 if (p != NULL && *p != '\0') 1227 add_pax_attr(&(pax->pax_header), "SCHILY.fflags", p); 1228 1229 /* I use star-compatible ACL attributes. */ 1230 if ((acl_types & ARCHIVE_ENTRY_ACL_TYPE_NFS4) != 0) { 1231 ret = add_pax_acl(a, entry_original, pax, 1232 ARCHIVE_ENTRY_ACL_STYLE_EXTRA_ID | 1233 ARCHIVE_ENTRY_ACL_STYLE_SEPARATOR_COMMA | 1234 ARCHIVE_ENTRY_ACL_STYLE_COMPACT); 1235 if (ret == ARCHIVE_FATAL) { 1236 archive_entry_free(entry_main); 1237 archive_string_free(&entry_name); 1238 return (ARCHIVE_FATAL); 1239 } 1240 } 1241 if (acl_types & ARCHIVE_ENTRY_ACL_TYPE_ACCESS) { 1242 ret = add_pax_acl(a, entry_original, pax, 1243 ARCHIVE_ENTRY_ACL_TYPE_ACCESS | 1244 ARCHIVE_ENTRY_ACL_STYLE_EXTRA_ID | 1245 ARCHIVE_ENTRY_ACL_STYLE_SEPARATOR_COMMA); 1246 if (ret == ARCHIVE_FATAL) { 1247 archive_entry_free(entry_main); 1248 archive_string_free(&entry_name); 1249 return (ARCHIVE_FATAL); 1250 } 1251 } 1252 if (acl_types & ARCHIVE_ENTRY_ACL_TYPE_DEFAULT) { 1253 ret = add_pax_acl(a, entry_original, pax, 1254 ARCHIVE_ENTRY_ACL_TYPE_DEFAULT | 1255 ARCHIVE_ENTRY_ACL_STYLE_EXTRA_ID | 1256 ARCHIVE_ENTRY_ACL_STYLE_SEPARATOR_COMMA); 1257 if (ret == ARCHIVE_FATAL) { 1258 archive_entry_free(entry_main); 1259 archive_string_free(&entry_name); 1260 return (ARCHIVE_FATAL); 1261 } 1262 } 1263 1264 /* We use GNU-tar-compatible sparse attributes. */ 1265 if (sparse_count > 0) { 1266 int64_t soffset, slength; 1267 1268 add_pax_attr_int(&(pax->pax_header), 1269 "GNU.sparse.major", 1); 1270 add_pax_attr_int(&(pax->pax_header), 1271 "GNU.sparse.minor", 0); 1272 /* 1273 * Make sure to store the original path, since 1274 * truncation to ustar limit happened already. 1275 */ 1276 add_pax_attr(&(pax->pax_header), 1277 "GNU.sparse.name", path); 1278 add_pax_attr_int(&(pax->pax_header), 1279 "GNU.sparse.realsize", 1280 archive_entry_size(entry_main)); 1281 1282 /* Rename the file name which will be used for 1283 * ustar header to a special name, which GNU 1284 * PAX Format 1.0 requires */ 1285 archive_entry_set_pathname(entry_main, 1286 build_gnu_sparse_name(gnu_sparse_name, 1287 entry_name.s)); 1288 1289 /* 1290 * - Make a sparse map, which will precede a file data. 1291 * - Get the total size of available data of sparse. 1292 */ 1293 archive_string_sprintf(&(pax->sparse_map), "%d\n", 1294 sparse_count); 1295 while (archive_entry_sparse_next(entry_main, 1296 &soffset, &slength) == ARCHIVE_OK) { 1297 archive_string_sprintf(&(pax->sparse_map), 1298 "%jd\n%jd\n", 1299 (intmax_t)soffset, 1300 (intmax_t)slength); 1301 sparse_total += slength; 1302 if (sparse_list_add(pax, soffset, slength) 1303 != ARCHIVE_OK) { 1304 archive_set_error(&a->archive, 1305 ENOMEM, 1306 "Can't allocate memory"); 1307 archive_entry_free(entry_main); 1308 archive_string_free(&entry_name); 1309 return (ARCHIVE_FATAL); 1310 } 1311 } 1312 } 1313 1314 /* Store extended attributes */ 1315 if (archive_write_pax_header_xattrs(a, pax, entry_original) 1316 == ARCHIVE_FATAL) { 1317 archive_entry_free(entry_main); 1318 archive_string_free(&entry_name); 1319 return (ARCHIVE_FATAL); 1320 } 1321 1322 /* Store extended symlink information */ 1323 if (archive_entry_symlink_type(entry_main) == 1324 AE_SYMLINK_TYPE_FILE) { 1325 add_pax_attr(&(pax->pax_header), 1326 "LIBARCHIVE.symlinktype", "file"); 1327 } else if (archive_entry_symlink_type(entry_main) == 1328 AE_SYMLINK_TYPE_DIRECTORY) { 1329 add_pax_attr(&(pax->pax_header), 1330 "LIBARCHIVE.symlinktype", "dir"); 1331 } 1332 } 1333 1334 /* Only regular files have data. */ 1335 if (archive_entry_filetype(entry_main) != AE_IFREG) 1336 archive_entry_set_size(entry_main, 0); 1337 1338 /* 1339 * Pax-restricted does not store data for hardlinks, in order 1340 * to improve compatibility with ustar. 1341 */ 1342 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE && 1343 hardlink != NULL) 1344 archive_entry_set_size(entry_main, 0); 1345 1346 /* 1347 * XXX Full pax interchange format does permit a hardlink 1348 * entry to have data associated with it. I'm not supporting 1349 * that here because the client expects me to tell them whether 1350 * or not this format expects data for hardlinks. If I 1351 * don't check here, then every pax archive will end up with 1352 * duplicated data for hardlinks. Someday, there may be 1353 * need to select this behavior, in which case the following 1354 * will need to be revisited. XXX 1355 */ 1356 if (hardlink != NULL) 1357 archive_entry_set_size(entry_main, 0); 1358 1359 /* Save a real file size. */ 1360 real_size = archive_entry_size(entry_main); 1361 /* 1362 * Overwrite a file size by the total size of sparse blocks and 1363 * the size of sparse map info. That file size is the length of 1364 * the data, which we will exactly store into an archive file. 1365 */ 1366 if (archive_strlen(&(pax->sparse_map))) { 1367 size_t mapsize = archive_strlen(&(pax->sparse_map)); 1368 pax->sparse_map_padding = 0x1ff & (-(ssize_t)mapsize); 1369 archive_entry_set_size(entry_main, 1370 mapsize + pax->sparse_map_padding + sparse_total); 1371 } 1372 1373 /* If file size is too large, add 'size' to pax extended attrs. */ 1374 if (archive_entry_size(entry_main) >= (((int64_t)1) << 33)) { 1375 add_pax_attr_int(&(pax->pax_header), "size", 1376 archive_entry_size(entry_main)); 1377 } 1378 1379 /* Format 'ustar' header for main entry. 1380 * 1381 * The trouble with file size: If the reader can't understand 1382 * the file size, they may not be able to locate the next 1383 * entry and the rest of the archive is toast. Pax-compliant 1384 * readers are supposed to ignore the file size in the main 1385 * header, so the question becomes how to maximize portability 1386 * for readers that don't support pax attribute extensions. 1387 * For maximum compatibility, I permit numeric extensions in 1388 * the main header so that the file size stored will always be 1389 * correct, even if it's in a format that only some 1390 * implementations understand. The technique used here is: 1391 * 1392 * a) If possible, follow the standard exactly. This handles 1393 * files up to 8 gigabytes minus 1. 1394 * 1395 * b) If that fails, try octal but omit the field terminator. 1396 * That handles files up to 64 gigabytes minus 1. 1397 * 1398 * c) Otherwise, use base-256 extensions. That handles files 1399 * up to 2^63 in this implementation, with the potential to 1400 * go up to 2^94. That should hold us for a while. ;-) 1401 * 1402 * The non-strict formatter uses similar logic for other 1403 * numeric fields, though they're less critical. 1404 */ 1405 if (__archive_write_format_header_ustar(a, ustarbuff, entry_main, -1, 0, 1406 NULL) == ARCHIVE_FATAL) { 1407 archive_entry_free(entry_main); 1408 archive_string_free(&entry_name); 1409 return (ARCHIVE_FATAL); 1410 } 1411 1412 /* If we built any extended attributes, write that entry first. */ 1413 if (archive_strlen(&(pax->pax_header)) > 0) { 1414 struct archive_entry *pax_attr_entry; 1415 time_t s; 1416 int64_t uid, gid; 1417 __LA_MODE_T mode; 1418 1419 pax_attr_entry = archive_entry_new2(&a->archive); 1420 p = entry_name.s; 1421 archive_entry_set_pathname(pax_attr_entry, 1422 build_pax_attribute_name(pax_entry_name, p)); 1423 archive_entry_set_size(pax_attr_entry, 1424 archive_strlen(&(pax->pax_header))); 1425 /* Copy uid/gid (but clip to ustar limits). */ 1426 uid = archive_entry_uid(entry_main); 1427 if (uid >= 1 << 18) 1428 uid = (1 << 18) - 1; 1429 archive_entry_set_uid(pax_attr_entry, uid); 1430 gid = archive_entry_gid(entry_main); 1431 if (gid >= 1 << 18) 1432 gid = (1 << 18) - 1; 1433 archive_entry_set_gid(pax_attr_entry, gid); 1434 /* Copy mode over (but not setuid/setgid bits) */ 1435 mode = archive_entry_mode(entry_main); 1436 #ifdef S_ISUID 1437 mode &= ~S_ISUID; 1438 #endif 1439 #ifdef S_ISGID 1440 mode &= ~S_ISGID; 1441 #endif 1442 #ifdef S_ISVTX 1443 mode &= ~S_ISVTX; 1444 #endif 1445 archive_entry_set_mode(pax_attr_entry, mode); 1446 1447 /* Copy uname/gname. */ 1448 archive_entry_set_uname(pax_attr_entry, 1449 archive_entry_uname(entry_main)); 1450 archive_entry_set_gname(pax_attr_entry, 1451 archive_entry_gname(entry_main)); 1452 1453 /* Copy mtime, but clip to ustar limits. */ 1454 s = archive_entry_mtime(entry_main); 1455 if (s < 0) { s = 0; } 1456 if (s > ustar_max_mtime) { s = ustar_max_mtime; } 1457 archive_entry_set_mtime(pax_attr_entry, s, 0); 1458 1459 /* Standard ustar doesn't support atime. */ 1460 archive_entry_set_atime(pax_attr_entry, 0, 0); 1461 1462 /* Standard ustar doesn't support ctime. */ 1463 archive_entry_set_ctime(pax_attr_entry, 0, 0); 1464 1465 r = __archive_write_format_header_ustar(a, paxbuff, 1466 pax_attr_entry, 'x', 1, NULL); 1467 1468 archive_entry_free(pax_attr_entry); 1469 1470 /* Note that the 'x' header shouldn't ever fail to format */ 1471 if (r < ARCHIVE_WARN) { 1472 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1473 "archive_write_pax_header: " 1474 "'x' header failed?! This can't happen"); 1475 archive_entry_free(entry_main); 1476 archive_string_free(&entry_name); 1477 return (ARCHIVE_FATAL); 1478 } else if (r < ret) 1479 ret = r; 1480 r = __archive_write_output(a, paxbuff, 512); 1481 if (r != ARCHIVE_OK) { 1482 sparse_list_clear(pax); 1483 pax->entry_bytes_remaining = 0; 1484 pax->entry_padding = 0; 1485 archive_entry_free(entry_main); 1486 archive_string_free(&entry_name); 1487 return (ARCHIVE_FATAL); 1488 } 1489 1490 pax->entry_bytes_remaining = archive_strlen(&(pax->pax_header)); 1491 pax->entry_padding = 1492 0x1ff & (-(int64_t)pax->entry_bytes_remaining); 1493 1494 r = __archive_write_output(a, pax->pax_header.s, 1495 archive_strlen(&(pax->pax_header))); 1496 if (r != ARCHIVE_OK) { 1497 /* If a write fails, we're pretty much toast. */ 1498 archive_entry_free(entry_main); 1499 archive_string_free(&entry_name); 1500 return (ARCHIVE_FATAL); 1501 } 1502 /* Pad out the end of the entry. */ 1503 r = __archive_write_nulls(a, (size_t)pax->entry_padding); 1504 if (r != ARCHIVE_OK) { 1505 /* If a write fails, we're pretty much toast. */ 1506 archive_entry_free(entry_main); 1507 archive_string_free(&entry_name); 1508 return (ARCHIVE_FATAL); 1509 } 1510 pax->entry_bytes_remaining = pax->entry_padding = 0; 1511 } 1512 1513 /* Write the header for main entry. */ 1514 r = __archive_write_output(a, ustarbuff, 512); 1515 if (r != ARCHIVE_OK) { 1516 archive_entry_free(entry_main); 1517 archive_string_free(&entry_name); 1518 return (r); 1519 } 1520 1521 /* 1522 * Inform the client of the on-disk size we're using, so 1523 * they can avoid unnecessarily writing a body for something 1524 * that we're just going to ignore. 1525 */ 1526 archive_entry_set_size(entry_original, real_size); 1527 if (pax->sparse_list == NULL && real_size > 0) { 1528 /* This is not a sparse file but we handle its data as 1529 * a sparse block. */ 1530 sparse_list_add(pax, 0, real_size); 1531 sparse_total = real_size; 1532 } 1533 pax->entry_padding = 0x1ff & (-(int64_t)sparse_total); 1534 archive_entry_free(entry_main); 1535 archive_string_free(&entry_name); 1536 1537 return (ret); 1538 } 1539 1540 /* 1541 * We need a valid name for the regular 'ustar' entry. This routine 1542 * tries to hack something more-or-less reasonable. 1543 * 1544 * The approach here tries to preserve leading dir names. We do so by 1545 * working with four sections: 1546 * 1) "prefix" directory names, 1547 * 2) "suffix" directory names, 1548 * 3) inserted dir name (optional), 1549 * 4) filename. 1550 * 1551 * These sections must satisfy the following requirements: 1552 * * Parts 1 & 2 together form an initial portion of the dir name. 1553 * * Part 3 is specified by the caller. (It should not contain a leading 1554 * or trailing '/'.) 1555 * * Part 4 forms an initial portion of the base filename. 1556 * * The filename must be <= 99 chars to fit the ustar 'name' field. 1557 * * Parts 2, 3, 4 together must be <= 99 chars to fit the ustar 'name' fld. 1558 * * Part 1 must be <= 155 chars to fit the ustar 'prefix' field. 1559 * * If the original name ends in a '/', the new name must also end in a '/' 1560 * * Trailing '/.' sequences may be stripped. 1561 * 1562 * Note: Recall that the ustar format does not store the '/' separating 1563 * parts 1 & 2, but does store the '/' separating parts 2 & 3. 1564 */ 1565 static char * 1566 build_ustar_entry_name(char *dest, const char *src, size_t src_length, 1567 const char *insert) 1568 { 1569 const char *prefix, *prefix_end; 1570 const char *suffix, *suffix_end; 1571 const char *filename, *filename_end; 1572 char *p; 1573 int need_slash = 0; /* Was there a trailing slash? */ 1574 size_t suffix_length = 98; /* 99 - 1 for trailing slash */ 1575 size_t insert_length; 1576 1577 /* Length of additional dir element to be added. */ 1578 if (insert == NULL) 1579 insert_length = 0; 1580 else 1581 /* +2 here allows for '/' before and after the insert. */ 1582 insert_length = strlen(insert) + 2; 1583 1584 /* Step 0: Quick bailout in a common case. */ 1585 if (src_length < 100 && insert == NULL) { 1586 strncpy(dest, src, src_length); 1587 dest[src_length] = '\0'; 1588 return (dest); 1589 } 1590 1591 /* Step 1: Locate filename and enforce the length restriction. */ 1592 filename_end = src + src_length; 1593 /* Remove trailing '/' chars and '/.' pairs. */ 1594 for (;;) { 1595 if (filename_end > src && filename_end[-1] == '/') { 1596 filename_end --; 1597 need_slash = 1; /* Remember to restore trailing '/'. */ 1598 continue; 1599 } 1600 if (filename_end > src + 1 && filename_end[-1] == '.' 1601 && filename_end[-2] == '/') { 1602 filename_end -= 2; 1603 need_slash = 1; /* "foo/." will become "foo/" */ 1604 continue; 1605 } 1606 break; 1607 } 1608 if (need_slash) 1609 suffix_length--; 1610 /* Find start of filename. */ 1611 filename = filename_end - 1; 1612 while ((filename > src) && (*filename != '/')) 1613 filename --; 1614 if ((*filename == '/') && (filename < filename_end - 1)) 1615 filename ++; 1616 /* Adjust filename_end so that filename + insert fits in 99 chars. */ 1617 suffix_length -= insert_length; 1618 if (filename_end > filename + suffix_length) 1619 filename_end = filename + suffix_length; 1620 /* Calculate max size for "suffix" section (#3 above). */ 1621 suffix_length -= filename_end - filename; 1622 1623 /* Step 2: Locate the "prefix" section of the dirname, including 1624 * trailing '/'. */ 1625 prefix = src; 1626 prefix_end = prefix + 154 /* 155 - 1 for trailing / */; 1627 if (prefix_end > filename) 1628 prefix_end = filename; 1629 while (prefix_end > prefix && *prefix_end != '/') 1630 prefix_end--; 1631 if ((prefix_end < filename) && (*prefix_end == '/')) 1632 prefix_end++; 1633 1634 /* Step 3: Locate the "suffix" section of the dirname, 1635 * including trailing '/'. */ 1636 suffix = prefix_end; 1637 suffix_end = suffix + suffix_length; /* Enforce limit. */ 1638 if (suffix_end > filename) 1639 suffix_end = filename; 1640 if (suffix_end < suffix) 1641 suffix_end = suffix; 1642 while (suffix_end > suffix && *suffix_end != '/') 1643 suffix_end--; 1644 if ((suffix_end < filename) && (*suffix_end == '/')) 1645 suffix_end++; 1646 1647 /* Step 4: Build the new name. */ 1648 /* The OpenBSD strlcpy function is safer, but less portable. */ 1649 /* Rather than maintain two versions, just use the strncpy version. */ 1650 p = dest; 1651 if (prefix_end > prefix) { 1652 strncpy(p, prefix, prefix_end - prefix); 1653 p += prefix_end - prefix; 1654 } 1655 if (suffix_end > suffix) { 1656 strncpy(p, suffix, suffix_end - suffix); 1657 p += suffix_end - suffix; 1658 } 1659 if (insert != NULL) { 1660 /* Note: assume insert does not have leading or trailing '/' */ 1661 strcpy(p, insert); 1662 p += strlen(insert); 1663 *p++ = '/'; 1664 } 1665 strncpy(p, filename, filename_end - filename); 1666 p += filename_end - filename; 1667 if (need_slash) 1668 *p++ = '/'; 1669 *p = '\0'; 1670 1671 return (dest); 1672 } 1673 1674 /* 1675 * The ustar header for the pax extended attributes must have a 1676 * reasonable name: SUSv3 requires 'dirname'/PaxHeader.'pid'/'filename' 1677 * where 'pid' is the PID of the archiving process. Unfortunately, 1678 * that makes testing a pain since the output varies for each run, 1679 * so I'm sticking with the simpler 'dirname'/PaxHeader/'filename' 1680 * for now. (Someday, I'll make this settable. Then I can use the 1681 * SUS recommendation as default and test harnesses can override it 1682 * to get predictable results.) 1683 * 1684 * Joerg Schilling has argued that this is unnecessary because, in 1685 * practice, if the pax extended attributes get extracted as regular 1686 * files, no one is going to bother reading those attributes to 1687 * manually restore them. Based on this, 'star' uses 1688 * /tmp/PaxHeader/'basename' as the ustar header name. This is a 1689 * tempting argument, in part because it's simpler than the SUSv3 1690 * recommendation, but I'm not entirely convinced. I'm also 1691 * uncomfortable with the fact that "/tmp" is a Unix-ism. 1692 * 1693 * The following routine leverages build_ustar_entry_name() above and 1694 * so is simpler than you might think. It just needs to provide the 1695 * additional path element and handle a few pathological cases). 1696 */ 1697 static char * 1698 build_pax_attribute_name(char *dest, const char *src) 1699 { 1700 char buff[64]; 1701 const char *p; 1702 1703 /* Handle the null filename case. */ 1704 if (src == NULL || *src == '\0') { 1705 strcpy(dest, "PaxHeader/blank"); 1706 return (dest); 1707 } 1708 1709 /* Prune final '/' and other unwanted final elements. */ 1710 p = src + strlen(src); 1711 for (;;) { 1712 /* Ends in "/", remove the '/' */ 1713 if (p > src && p[-1] == '/') { 1714 --p; 1715 continue; 1716 } 1717 /* Ends in "/.", remove the '.' */ 1718 if (p > src + 1 && p[-1] == '.' 1719 && p[-2] == '/') { 1720 --p; 1721 continue; 1722 } 1723 break; 1724 } 1725 1726 /* Pathological case: After above, there was nothing left. 1727 * This includes "/." "/./." "/.//./." etc. */ 1728 if (p == src) { 1729 strcpy(dest, "/PaxHeader/rootdir"); 1730 return (dest); 1731 } 1732 1733 /* Convert unadorned "." into a suitable filename. */ 1734 if (*src == '.' && p == src + 1) { 1735 strcpy(dest, "PaxHeader/currentdir"); 1736 return (dest); 1737 } 1738 1739 /* 1740 * TODO: Push this string into the 'pax' structure to avoid 1741 * recomputing it every time. That will also open the door 1742 * to having clients override it. 1743 */ 1744 #if HAVE_GETPID && 0 /* Disable this for now; see above comment. */ 1745 snprintf(buff, sizeof(buff), "PaxHeader.%d", getpid()); 1746 #else 1747 /* If the platform can't fetch the pid, don't include it. */ 1748 strcpy(buff, "PaxHeader"); 1749 #endif 1750 /* General case: build a ustar-compatible name adding 1751 * "/PaxHeader/". */ 1752 build_ustar_entry_name(dest, src, p - src, buff); 1753 1754 return (dest); 1755 } 1756 1757 /* 1758 * GNU PAX Format 1.0 requires the special name, which pattern is: 1759 * <dir>/GNUSparseFile.<pid>/<original file name> 1760 * 1761 * Since reproducible archives are more important, use 0 as pid. 1762 * 1763 * This function is used for only Sparse file, a file type of which 1764 * is regular file. 1765 */ 1766 static char * 1767 build_gnu_sparse_name(char *dest, const char *src) 1768 { 1769 const char *p; 1770 1771 /* Handle the null filename case. */ 1772 if (src == NULL || *src == '\0') { 1773 strcpy(dest, "GNUSparseFile/blank"); 1774 return (dest); 1775 } 1776 1777 /* Prune final '/' and other unwanted final elements. */ 1778 p = src + strlen(src); 1779 for (;;) { 1780 /* Ends in "/", remove the '/' */ 1781 if (p > src && p[-1] == '/') { 1782 --p; 1783 continue; 1784 } 1785 /* Ends in "/.", remove the '.' */ 1786 if (p > src + 1 && p[-1] == '.' 1787 && p[-2] == '/') { 1788 --p; 1789 continue; 1790 } 1791 break; 1792 } 1793 1794 /* General case: build a ustar-compatible name adding 1795 * "/GNUSparseFile/". */ 1796 build_ustar_entry_name(dest, src, p - src, "GNUSparseFile.0"); 1797 1798 return (dest); 1799 } 1800 1801 /* Write two null blocks for the end of archive */ 1802 static int 1803 archive_write_pax_close(struct archive_write *a) 1804 { 1805 return (__archive_write_nulls(a, 512 * 2)); 1806 } 1807 1808 static int 1809 archive_write_pax_free(struct archive_write *a) 1810 { 1811 struct pax *pax; 1812 1813 pax = (struct pax *)a->format_data; 1814 if (pax == NULL) 1815 return (ARCHIVE_OK); 1816 1817 archive_string_free(&pax->pax_header); 1818 archive_string_free(&pax->sparse_map); 1819 archive_string_free(&pax->l_url_encoded_name); 1820 sparse_list_clear(pax); 1821 free(pax); 1822 a->format_data = NULL; 1823 return (ARCHIVE_OK); 1824 } 1825 1826 static int 1827 archive_write_pax_finish_entry(struct archive_write *a) 1828 { 1829 struct pax *pax; 1830 uint64_t remaining; 1831 int ret; 1832 1833 pax = (struct pax *)a->format_data; 1834 remaining = pax->entry_bytes_remaining; 1835 if (remaining == 0) { 1836 while (pax->sparse_list) { 1837 struct sparse_block *sb; 1838 if (!pax->sparse_list->is_hole) 1839 remaining += pax->sparse_list->remaining; 1840 sb = pax->sparse_list->next; 1841 free(pax->sparse_list); 1842 pax->sparse_list = sb; 1843 } 1844 } 1845 ret = __archive_write_nulls(a, (size_t)(remaining + pax->entry_padding)); 1846 pax->entry_bytes_remaining = pax->entry_padding = 0; 1847 return (ret); 1848 } 1849 1850 static ssize_t 1851 archive_write_pax_data(struct archive_write *a, const void *buff, size_t s) 1852 { 1853 struct pax *pax; 1854 size_t ws; 1855 size_t total; 1856 int ret; 1857 1858 pax = (struct pax *)a->format_data; 1859 1860 /* 1861 * According to GNU PAX format 1.0, write a sparse map 1862 * before the body. 1863 */ 1864 if (archive_strlen(&(pax->sparse_map))) { 1865 ret = __archive_write_output(a, pax->sparse_map.s, 1866 archive_strlen(&(pax->sparse_map))); 1867 if (ret != ARCHIVE_OK) 1868 return (ret); 1869 ret = __archive_write_nulls(a, pax->sparse_map_padding); 1870 if (ret != ARCHIVE_OK) 1871 return (ret); 1872 archive_string_empty(&(pax->sparse_map)); 1873 } 1874 1875 total = 0; 1876 while (total < s) { 1877 const unsigned char *p; 1878 1879 while (pax->sparse_list != NULL && 1880 pax->sparse_list->remaining == 0) { 1881 struct sparse_block *sb = pax->sparse_list->next; 1882 free(pax->sparse_list); 1883 pax->sparse_list = sb; 1884 } 1885 1886 if (pax->sparse_list == NULL) 1887 return (total); 1888 1889 p = ((const unsigned char *)buff) + total; 1890 ws = s - total; 1891 if (ws > pax->sparse_list->remaining) 1892 ws = (size_t)pax->sparse_list->remaining; 1893 1894 if (pax->sparse_list->is_hole) { 1895 /* Current block is hole thus we do not write 1896 * the body. */ 1897 pax->sparse_list->remaining -= ws; 1898 total += ws; 1899 continue; 1900 } 1901 1902 ret = __archive_write_output(a, p, ws); 1903 pax->sparse_list->remaining -= ws; 1904 total += ws; 1905 if (ret != ARCHIVE_OK) 1906 return (ret); 1907 } 1908 return (total); 1909 } 1910 1911 static int 1912 has_non_ASCII(const char *_p) 1913 { 1914 const unsigned char *p = (const unsigned char *)_p; 1915 1916 if (p == NULL) 1917 return (1); 1918 while (*p != '\0' && *p < 128) 1919 p++; 1920 return (*p != '\0'); 1921 } 1922 1923 /* 1924 * Used by extended attribute support; encodes the name 1925 * so that there will be no '=' characters in the result. 1926 */ 1927 static char * 1928 url_encode(const char *in) 1929 { 1930 const char *s; 1931 char *d; 1932 size_t out_len = 0; 1933 char *out; 1934 1935 for (s = in; *s != '\0'; s++) { 1936 if (*s < 33 || *s > 126 || *s == '%' || *s == '=') { 1937 if (SIZE_MAX - out_len < 4) 1938 return (NULL); 1939 out_len += 3; 1940 } else { 1941 if (SIZE_MAX - out_len < 2) 1942 return (NULL); 1943 out_len++; 1944 } 1945 } 1946 1947 out = malloc(out_len + 1); 1948 if (out == NULL) 1949 return (NULL); 1950 1951 for (s = in, d = out; *s != '\0'; s++) { 1952 /* encode any non-printable ASCII character or '%' or '=' */ 1953 if (*s < 33 || *s > 126 || *s == '%' || *s == '=') { 1954 /* URL encoding is '%' followed by two hex digits */ 1955 *d++ = '%'; 1956 *d++ = "0123456789ABCDEF"[0x0f & (*s >> 4)]; 1957 *d++ = "0123456789ABCDEF"[0x0f & *s]; 1958 } else { 1959 *d++ = *s; 1960 } 1961 } 1962 *d = '\0'; 1963 return (out); 1964 } 1965 1966 /* 1967 * Encode a sequence of bytes into a C string using base-64 encoding. 1968 * 1969 * Returns a null-terminated C string allocated with malloc(); caller 1970 * is responsible for freeing the result. 1971 */ 1972 static char * 1973 base64_encode(const char *s, size_t len) 1974 { 1975 static const char digits[64] = 1976 { 'A','B','C','D','E','F','G','H','I','J','K','L','M','N','O', 1977 'P','Q','R','S','T','U','V','W','X','Y','Z','a','b','c','d', 1978 'e','f','g','h','i','j','k','l','m','n','o','p','q','r','s', 1979 't','u','v','w','x','y','z','0','1','2','3','4','5','6','7', 1980 '8','9','+','/' }; 1981 int v; 1982 char *d, *out; 1983 1984 /* 3 bytes becomes 4 chars, but round up and allow for trailing NUL */ 1985 out = malloc((len * 4 + 2) / 3 + 1); 1986 if (out == NULL) 1987 return (NULL); 1988 d = out; 1989 1990 /* Convert each group of 3 bytes into 4 characters. */ 1991 while (len >= 3) { 1992 v = (((int)s[0] << 16) & 0xff0000) 1993 | (((int)s[1] << 8) & 0xff00) 1994 | (((int)s[2]) & 0x00ff); 1995 s += 3; 1996 len -= 3; 1997 *d++ = digits[(v >> 18) & 0x3f]; 1998 *d++ = digits[(v >> 12) & 0x3f]; 1999 *d++ = digits[(v >> 6) & 0x3f]; 2000 *d++ = digits[(v) & 0x3f]; 2001 } 2002 /* Handle final group of 1 byte (2 chars) or 2 bytes (3 chars). */ 2003 switch (len) { 2004 case 0: break; 2005 case 1: 2006 v = (((int)s[0] << 16) & 0xff0000); 2007 *d++ = digits[(v >> 18) & 0x3f]; 2008 *d++ = digits[(v >> 12) & 0x3f]; 2009 break; 2010 case 2: 2011 v = (((int)s[0] << 16) & 0xff0000) 2012 | (((int)s[1] << 8) & 0xff00); 2013 *d++ = digits[(v >> 18) & 0x3f]; 2014 *d++ = digits[(v >> 12) & 0x3f]; 2015 *d++ = digits[(v >> 6) & 0x3f]; 2016 break; 2017 } 2018 /* Add trailing NUL character so output is a valid C string. */ 2019 *d = '\0'; 2020 return (out); 2021 } 2022 2023 static void 2024 sparse_list_clear(struct pax *pax) 2025 { 2026 while (pax->sparse_list != NULL) { 2027 struct sparse_block *sb = pax->sparse_list; 2028 pax->sparse_list = sb->next; 2029 free(sb); 2030 } 2031 pax->sparse_tail = NULL; 2032 } 2033 2034 static int 2035 _sparse_list_add_block(struct pax *pax, int64_t offset, int64_t length, 2036 int is_hole) 2037 { 2038 struct sparse_block *sb; 2039 2040 sb = malloc(sizeof(*sb)); 2041 if (sb == NULL) 2042 return (ARCHIVE_FATAL); 2043 sb->next = NULL; 2044 sb->is_hole = is_hole; 2045 sb->offset = offset; 2046 sb->remaining = length; 2047 if (pax->sparse_list == NULL || pax->sparse_tail == NULL) 2048 pax->sparse_list = pax->sparse_tail = sb; 2049 else { 2050 pax->sparse_tail->next = sb; 2051 pax->sparse_tail = sb; 2052 } 2053 return (ARCHIVE_OK); 2054 } 2055 2056 static int 2057 sparse_list_add(struct pax *pax, int64_t offset, int64_t length) 2058 { 2059 int64_t last_offset; 2060 int r; 2061 2062 if (pax->sparse_tail == NULL) 2063 last_offset = 0; 2064 else { 2065 last_offset = pax->sparse_tail->offset + 2066 pax->sparse_tail->remaining; 2067 } 2068 if (last_offset < offset) { 2069 /* Add a hole block. */ 2070 r = _sparse_list_add_block(pax, last_offset, 2071 offset - last_offset, 1); 2072 if (r != ARCHIVE_OK) 2073 return (r); 2074 } 2075 /* Add data block. */ 2076 return (_sparse_list_add_block(pax, offset, length, 0)); 2077 } 2078 2079 static time_t 2080 get_ustar_max_mtime(void) 2081 { 2082 /* 2083 * Technically, the mtime field in the ustar header can 2084 * support 33 bits. We are using all of them to keep 2085 * tar/test/test_option_C_mtree.c simple and passing after 2038. 2086 * For platforms that use signed 32-bit time values we 2087 * use the 32-bit maximum. 2088 */ 2089 if (sizeof(time_t) > sizeof(int32_t)) 2090 return (time_t)0x1ffffffff; 2091 else 2092 return (time_t)0x7fffffff; 2093 } 2094
Indexes created Thu Jul 16 00:26:27 UTC 2026