archive_write_set_format_pax.c revision 348607
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__FBSDID("$FreeBSD: stable/11/contrib/libarchive/libarchive/archive_write_set_format_pax.c 348607 2019-06-04 10:35:54Z mm $"); 30 31#ifdef HAVE_ERRNO_H 32#include <errno.h> 33#endif 34#ifdef HAVE_STDLIB_H 35#include <stdlib.h> 36#endif 37#ifdef HAVE_STRING_H 38#include <string.h> 39#endif 40 41#include "archive.h" 42#include "archive_entry.h" 43#include "archive_entry_locale.h" 44#include "archive_private.h" 45#include "archive_write_private.h" 46 47struct sparse_block { 48 struct sparse_block *next; 49 int is_hole; 50 uint64_t offset; 51 uint64_t remaining; 52}; 53 54struct 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 71static void add_pax_attr(struct archive_string *, const char *key, 72 const char *value); 73static void add_pax_attr_binary(struct archive_string *, 74 const char *key, 75 const char *value, size_t value_len); 76static void add_pax_attr_int(struct archive_string *, 77 const char *key, int64_t value); 78static void add_pax_attr_time(struct archive_string *, 79 const char *key, int64_t sec, 80 unsigned long nanos); 81static int add_pax_acl(struct archive_write *, 82 struct archive_entry *, struct pax *, int); 83static ssize_t archive_write_pax_data(struct archive_write *, 84 const void *, size_t); 85static int archive_write_pax_close(struct archive_write *); 86static int archive_write_pax_free(struct archive_write *); 87static int archive_write_pax_finish_entry(struct archive_write *); 88static int archive_write_pax_header(struct archive_write *, 89 struct archive_entry *); 90static int archive_write_pax_options(struct archive_write *, 91 const char *, const char *); 92static char *base64_encode(const char *src, size_t len); 93static char *build_gnu_sparse_name(char *dest, const char *src); 94static char *build_pax_attribute_name(char *dest, const char *src); 95static char *build_ustar_entry_name(char *dest, const char *src, 96 size_t src_length, const char *insert); 97static char *format_int(char *dest, int64_t); 98static int has_non_ASCII(const char *); 99static void sparse_list_clear(struct pax *); 100static int sparse_list_add(struct pax *, int64_t, int64_t); 101static char *url_encode(const char *in); 102 103/* 104 * Set output format to 'restricted pax' format. 105 * 106 * This is the same as normal 'pax', but tries to suppress 107 * the pax header whenever possible. This is the default for 108 * bsdtar, for instance. 109 */ 110int 111archive_write_set_format_pax_restricted(struct archive *_a) 112{ 113 struct archive_write *a = (struct archive_write *)_a; 114 int r; 115 116 archive_check_magic(_a, ARCHIVE_WRITE_MAGIC, 117 ARCHIVE_STATE_NEW, "archive_write_set_format_pax_restricted"); 118 119 r = archive_write_set_format_pax(&a->archive); 120 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_RESTRICTED; 121 a->archive.archive_format_name = "restricted POSIX pax interchange"; 122 return (r); 123} 124 125/* 126 * Set output format to 'pax' format. 127 */ 128int 129archive_write_set_format_pax(struct archive *_a) 130{ 131 struct archive_write *a = (struct archive_write *)_a; 132 struct pax *pax; 133 134 archive_check_magic(_a, ARCHIVE_WRITE_MAGIC, 135 ARCHIVE_STATE_NEW, "archive_write_set_format_pax"); 136 137 if (a->format_free != NULL) 138 (a->format_free)(a); 139 140 pax = (struct pax *)calloc(1, sizeof(*pax)); 141 if (pax == NULL) { 142 archive_set_error(&a->archive, ENOMEM, 143 "Can't allocate pax data"); 144 return (ARCHIVE_FATAL); 145 } 146 pax->flags = WRITE_LIBARCHIVE_XATTR | WRITE_SCHILY_XATTR; 147 148 a->format_data = pax; 149 a->format_name = "pax"; 150 a->format_options = archive_write_pax_options; 151 a->format_write_header = archive_write_pax_header; 152 a->format_write_data = archive_write_pax_data; 153 a->format_close = archive_write_pax_close; 154 a->format_free = archive_write_pax_free; 155 a->format_finish_entry = archive_write_pax_finish_entry; 156 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 157 a->archive.archive_format_name = "POSIX pax interchange"; 158 return (ARCHIVE_OK); 159} 160 161static int 162archive_write_pax_options(struct archive_write *a, const char *key, 163 const char *val) 164{ 165 struct pax *pax = (struct pax *)a->format_data; 166 int ret = ARCHIVE_FAILED; 167 168 if (strcmp(key, "hdrcharset") == 0) { 169 /* 170 * The character-set we can use are defined in 171 * IEEE Std 1003.1-2001 172 */ 173 if (val == NULL || val[0] == 0) 174 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 175 "pax: hdrcharset option needs a character-set name"); 176 else if (strcmp(val, "BINARY") == 0 || 177 strcmp(val, "binary") == 0) { 178 /* 179 * Specify binary mode. We will not convert 180 * filenames, uname and gname to any charsets. 181 */ 182 pax->opt_binary = 1; 183 ret = ARCHIVE_OK; 184 } else if (strcmp(val, "UTF-8") == 0) { 185 /* 186 * Specify UTF-8 character-set to be used for 187 * filenames. This is almost the test that 188 * running platform supports the string conversion. 189 * Especially libarchive_test needs this trick for 190 * its test. 191 */ 192 pax->sconv_utf8 = archive_string_conversion_to_charset( 193 &(a->archive), "UTF-8", 0); 194 if (pax->sconv_utf8 == NULL) 195 ret = ARCHIVE_FATAL; 196 else 197 ret = ARCHIVE_OK; 198 } else 199 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 200 "pax: invalid charset name"); 201 return (ret); 202 } 203 204 /* Note: The "warn" return is just to inform the options 205 * supervisor that we didn't handle it. It will generate 206 * a suitable error if no one used this option. */ 207 return (ARCHIVE_WARN); 208} 209 210/* 211 * Note: This code assumes that 'nanos' has the same sign as 'sec', 212 * which implies that sec=-1, nanos=200000000 represents -1.2 seconds 213 * and not -0.8 seconds. This is a pretty pedantic point, as we're 214 * unlikely to encounter many real files created before Jan 1, 1970, 215 * much less ones with timestamps recorded to sub-second resolution. 216 */ 217static void 218add_pax_attr_time(struct archive_string *as, const char *key, 219 int64_t sec, unsigned long nanos) 220{ 221 int digit, i; 222 char *t; 223 /* 224 * Note that each byte contributes fewer than 3 base-10 225 * digits, so this will always be big enough. 226 */ 227 char tmp[1 + 3*sizeof(sec) + 1 + 3*sizeof(nanos)]; 228 229 tmp[sizeof(tmp) - 1] = 0; 230 t = tmp + sizeof(tmp) - 1; 231 232 /* Skip trailing zeros in the fractional part. */ 233 for (digit = 0, i = 10; i > 0 && digit == 0; i--) { 234 digit = nanos % 10; 235 nanos /= 10; 236 } 237 238 /* Only format the fraction if it's non-zero. */ 239 if (i > 0) { 240 while (i > 0) { 241 *--t = "0123456789"[digit]; 242 digit = nanos % 10; 243 nanos /= 10; 244 i--; 245 } 246 *--t = '.'; 247 } 248 t = format_int(t, sec); 249 250 add_pax_attr(as, key, t); 251} 252 253static char * 254format_int(char *t, int64_t i) 255{ 256 uint64_t ui; 257 258 if (i < 0) 259 ui = (i == INT64_MIN) ? (uint64_t)(INT64_MAX) + 1 : (uint64_t)(-i); 260 else 261 ui = i; 262 263 do { 264 *--t = "0123456789"[ui % 10]; 265 } while (ui /= 10); 266 if (i < 0) 267 *--t = '-'; 268 return (t); 269} 270 271static void 272add_pax_attr_int(struct archive_string *as, const char *key, int64_t value) 273{ 274 char tmp[1 + 3 * sizeof(value)]; 275 276 tmp[sizeof(tmp) - 1] = 0; 277 add_pax_attr(as, key, format_int(tmp + sizeof(tmp) - 1, value)); 278} 279 280/* 281 * Add a key/value attribute to the pax header. This function handles 282 * the length field and various other syntactic requirements. 283 */ 284static void 285add_pax_attr(struct archive_string *as, const char *key, const char *value) 286{ 287 add_pax_attr_binary(as, key, value, strlen(value)); 288} 289 290/* 291 * Add a key/value attribute to the pax header. This function handles 292 * binary values. 293 */ 294static void 295add_pax_attr_binary(struct archive_string *as, const char *key, 296 const char *value, size_t value_len) 297{ 298 int digits, i, len, next_ten; 299 char tmp[1 + 3 * sizeof(int)]; /* < 3 base-10 digits per byte */ 300 301 /*- 302 * PAX attributes have the following layout: 303 * <len> <space> <key> <=> <value> <nl> 304 */ 305 len = 1 + (int)strlen(key) + 1 + (int)value_len + 1; 306 307 /* 308 * The <len> field includes the length of the <len> field, so 309 * computing the correct length is tricky. I start by 310 * counting the number of base-10 digits in 'len' and 311 * computing the next higher power of 10. 312 */ 313 next_ten = 1; 314 digits = 0; 315 i = len; 316 while (i > 0) { 317 i = i / 10; 318 digits++; 319 next_ten = next_ten * 10; 320 } 321 /* 322 * For example, if string without the length field is 99 323 * chars, then adding the 2 digit length "99" will force the 324 * total length past 100, requiring an extra digit. The next 325 * statement adjusts for this effect. 326 */ 327 if (len + digits >= next_ten) 328 digits++; 329 330 /* Now, we have the right length so we can build the line. */ 331 tmp[sizeof(tmp) - 1] = 0; /* Null-terminate the work area. */ 332 archive_strcat(as, format_int(tmp + sizeof(tmp) - 1, len + digits)); 333 archive_strappend_char(as, ' '); 334 archive_strcat(as, key); 335 archive_strappend_char(as, '='); 336 archive_array_append(as, value, value_len); 337 archive_strappend_char(as, '\n'); 338} 339 340static void 341archive_write_pax_header_xattr(struct pax *pax, const char *encoded_name, 342 const void *value, size_t value_len) 343{ 344 struct archive_string s; 345 char *encoded_value; 346 347 if (pax->flags & WRITE_LIBARCHIVE_XATTR) { 348 encoded_value = base64_encode((const char *)value, value_len); 349 350 if (encoded_name != NULL && encoded_value != NULL) { 351 archive_string_init(&s); 352 archive_strcpy(&s, "LIBARCHIVE.xattr."); 353 archive_strcat(&s, encoded_name); 354 add_pax_attr(&(pax->pax_header), s.s, encoded_value); 355 archive_string_free(&s); 356 } 357 free(encoded_value); 358 } 359 if (pax->flags & WRITE_SCHILY_XATTR) { 360 archive_string_init(&s); 361 archive_strcpy(&s, "SCHILY.xattr."); 362 archive_strcat(&s, encoded_name); 363 add_pax_attr_binary(&(pax->pax_header), s.s, value, value_len); 364 archive_string_free(&s); 365 } 366} 367 368static int 369archive_write_pax_header_xattrs(struct archive_write *a, 370 struct pax *pax, struct archive_entry *entry) 371{ 372 int i = archive_entry_xattr_reset(entry); 373 374 while (i--) { 375 const char *name; 376 const void *value; 377 char *url_encoded_name = NULL, *encoded_name = NULL; 378 size_t size; 379 int r; 380 381 archive_entry_xattr_next(entry, &name, &value, &size); 382 url_encoded_name = url_encode(name); 383 if (url_encoded_name != NULL) { 384 /* Convert narrow-character to UTF-8. */ 385 r = archive_strcpy_l(&(pax->l_url_encoded_name), 386 url_encoded_name, pax->sconv_utf8); 387 free(url_encoded_name); /* Done with this. */ 388 if (r == 0) 389 encoded_name = pax->l_url_encoded_name.s; 390 else if (errno == ENOMEM) { 391 archive_set_error(&a->archive, ENOMEM, 392 "Can't allocate memory for Linkname"); 393 return (ARCHIVE_FATAL); 394 } 395 } 396 397 archive_write_pax_header_xattr(pax, encoded_name, 398 value, size); 399 400 } 401 return (ARCHIVE_OK); 402} 403 404static int 405get_entry_hardlink(struct archive_write *a, struct archive_entry *entry, 406 const char **name, size_t *length, struct archive_string_conv *sc) 407{ 408 int r; 409 410 r = archive_entry_hardlink_l(entry, name, length, sc); 411 if (r != 0) { 412 if (errno == ENOMEM) { 413 archive_set_error(&a->archive, ENOMEM, 414 "Can't allocate memory for Linkname"); 415 return (ARCHIVE_FATAL); 416 } 417 return (ARCHIVE_WARN); 418 } 419 return (ARCHIVE_OK); 420} 421 422static int 423get_entry_pathname(struct archive_write *a, struct archive_entry *entry, 424 const char **name, size_t *length, struct archive_string_conv *sc) 425{ 426 int r; 427 428 r = archive_entry_pathname_l(entry, name, length, sc); 429 if (r != 0) { 430 if (errno == ENOMEM) { 431 archive_set_error(&a->archive, ENOMEM, 432 "Can't allocate memory for Pathname"); 433 return (ARCHIVE_FATAL); 434 } 435 return (ARCHIVE_WARN); 436 } 437 return (ARCHIVE_OK); 438} 439 440static int 441get_entry_uname(struct archive_write *a, struct archive_entry *entry, 442 const char **name, size_t *length, struct archive_string_conv *sc) 443{ 444 int r; 445 446 r = archive_entry_uname_l(entry, name, length, sc); 447 if (r != 0) { 448 if (errno == ENOMEM) { 449 archive_set_error(&a->archive, ENOMEM, 450 "Can't allocate memory for Uname"); 451 return (ARCHIVE_FATAL); 452 } 453 return (ARCHIVE_WARN); 454 } 455 return (ARCHIVE_OK); 456} 457 458static int 459get_entry_gname(struct archive_write *a, struct archive_entry *entry, 460 const char **name, size_t *length, struct archive_string_conv *sc) 461{ 462 int r; 463 464 r = archive_entry_gname_l(entry, name, length, sc); 465 if (r != 0) { 466 if (errno == ENOMEM) { 467 archive_set_error(&a->archive, ENOMEM, 468 "Can't allocate memory for Gname"); 469 return (ARCHIVE_FATAL); 470 } 471 return (ARCHIVE_WARN); 472 } 473 return (ARCHIVE_OK); 474} 475 476static int 477get_entry_symlink(struct archive_write *a, struct archive_entry *entry, 478 const char **name, size_t *length, struct archive_string_conv *sc) 479{ 480 int r; 481 482 r = archive_entry_symlink_l(entry, name, length, sc); 483 if (r != 0) { 484 if (errno == ENOMEM) { 485 archive_set_error(&a->archive, ENOMEM, 486 "Can't allocate memory for Linkname"); 487 return (ARCHIVE_FATAL); 488 } 489 return (ARCHIVE_WARN); 490 } 491 return (ARCHIVE_OK); 492} 493 494/* Add ACL to pax header */ 495static int 496add_pax_acl(struct archive_write *a, 497 struct archive_entry *entry, struct pax *pax, int flags) 498{ 499 char *p; 500 const char *attr; 501 int acl_types; 502 503 acl_types = archive_entry_acl_types(entry); 504 505 if ((acl_types & ARCHIVE_ENTRY_ACL_TYPE_NFS4) != 0) 506 attr = "SCHILY.acl.ace"; 507 else if ((flags & ARCHIVE_ENTRY_ACL_TYPE_ACCESS) != 0) 508 attr = "SCHILY.acl.access"; 509 else if ((flags & ARCHIVE_ENTRY_ACL_TYPE_DEFAULT) != 0) 510 attr = "SCHILY.acl.default"; 511 else 512 return (ARCHIVE_FATAL); 513 514 p = archive_entry_acl_to_text_l(entry, NULL, flags, pax->sconv_utf8); 515 if (p == NULL) { 516 if (errno == ENOMEM) { 517 archive_set_error(&a->archive, ENOMEM, "%s %s", 518 "Can't allocate memory for ", attr); 519 return (ARCHIVE_FATAL); 520 } 521 archive_set_error(&a->archive, 522 ARCHIVE_ERRNO_FILE_FORMAT, "%s %s %s", 523 "Can't translate ", attr, " to UTF-8"); 524 return(ARCHIVE_WARN); 525 } 526 527 if (*p != '\0') { 528 add_pax_attr(&(pax->pax_header), 529 attr, p); 530 } 531 free(p); 532 return(ARCHIVE_OK); 533} 534 535/* 536 * TODO: Consider adding 'comment' and 'charset' fields to 537 * archive_entry so that clients can specify them. Also, consider 538 * adding generic key/value tags so clients can add arbitrary 539 * key/value data. 540 * 541 * TODO: Break up this 700-line function!!!! Yowza! 542 */ 543static int 544archive_write_pax_header(struct archive_write *a, 545 struct archive_entry *entry_original) 546{ 547 struct archive_entry *entry_main; 548 const char *p; 549 const char *suffix; 550 int need_extension, r, ret; 551 int acl_types; 552 int sparse_count; 553 uint64_t sparse_total, real_size; 554 struct pax *pax; 555 const char *hardlink; 556 const char *path = NULL, *linkpath = NULL; 557 const char *uname = NULL, *gname = NULL; 558 const void *mac_metadata; 559 size_t mac_metadata_size; 560 struct archive_string_conv *sconv; 561 size_t hardlink_length, path_length, linkpath_length; 562 size_t uname_length, gname_length; 563 564 char paxbuff[512]; 565 char ustarbuff[512]; 566 char ustar_entry_name[256]; 567 char pax_entry_name[256]; 568 char gnu_sparse_name[256]; 569 struct archive_string entry_name; 570 571 ret = ARCHIVE_OK; 572 need_extension = 0; 573 pax = (struct pax *)a->format_data; 574 575 /* Sanity check. */ 576 if (archive_entry_pathname(entry_original) == NULL) { 577 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 578 "Can't record entry in tar file without pathname"); 579 return (ARCHIVE_FAILED); 580 } 581 582 /* 583 * Choose a header encoding. 584 */ 585 if (pax->opt_binary) 586 sconv = NULL;/* Binary mode. */ 587 else { 588 /* Header encoding is UTF-8. */ 589 if (pax->sconv_utf8 == NULL) { 590 /* Initialize the string conversion object 591 * we must need */ 592 pax->sconv_utf8 = archive_string_conversion_to_charset( 593 &(a->archive), "UTF-8", 1); 594 if (pax->sconv_utf8 == NULL) 595 /* Couldn't allocate memory */ 596 return (ARCHIVE_FAILED); 597 } 598 sconv = pax->sconv_utf8; 599 } 600 601 r = get_entry_hardlink(a, entry_original, &hardlink, 602 &hardlink_length, sconv); 603 if (r == ARCHIVE_FATAL) 604 return (r); 605 else if (r != ARCHIVE_OK) { 606 r = get_entry_hardlink(a, entry_original, &hardlink, 607 &hardlink_length, NULL); 608 if (r == ARCHIVE_FATAL) 609 return (r); 610 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 611 "Can't translate linkname '%s' to %s", hardlink, 612 archive_string_conversion_charset_name(sconv)); 613 ret = ARCHIVE_WARN; 614 sconv = NULL;/* The header charset switches to binary mode. */ 615 } 616 617 /* Make sure this is a type of entry that we can handle here */ 618 if (hardlink == NULL) { 619 switch (archive_entry_filetype(entry_original)) { 620 case AE_IFBLK: 621 case AE_IFCHR: 622 case AE_IFIFO: 623 case AE_IFLNK: 624 case AE_IFREG: 625 break; 626 case AE_IFDIR: 627 { 628 /* 629 * Ensure a trailing '/'. Modify the original 630 * entry so the client sees the change. 631 */ 632#if defined(_WIN32) && !defined(__CYGWIN__) 633 const wchar_t *wp; 634 635 wp = archive_entry_pathname_w(entry_original); 636 if (wp != NULL && wp[wcslen(wp) -1] != L'/') { 637 struct archive_wstring ws; 638 639 archive_string_init(&ws); 640 path_length = wcslen(wp); 641 if (archive_wstring_ensure(&ws, 642 path_length + 2) == NULL) { 643 archive_set_error(&a->archive, ENOMEM, 644 "Can't allocate pax data"); 645 archive_wstring_free(&ws); 646 return(ARCHIVE_FATAL); 647 } 648 /* Should we keep '\' ? */ 649 if (wp[path_length -1] == L'\\') 650 path_length--; 651 archive_wstrncpy(&ws, wp, path_length); 652 archive_wstrappend_wchar(&ws, L'/'); 653 archive_entry_copy_pathname_w( 654 entry_original, ws.s); 655 archive_wstring_free(&ws); 656 p = NULL; 657 } else 658#endif 659 p = archive_entry_pathname(entry_original); 660 /* 661 * On Windows, this is a backup operation just in 662 * case getting WCS failed. On POSIX, this is a 663 * normal operation. 664 */ 665 if (p != NULL && p[0] != '\0' && p[strlen(p) - 1] != '/') { 666 struct archive_string as; 667 668 archive_string_init(&as); 669 path_length = strlen(p); 670 if (archive_string_ensure(&as, 671 path_length + 2) == NULL) { 672 archive_set_error(&a->archive, ENOMEM, 673 "Can't allocate pax data"); 674 archive_string_free(&as); 675 return(ARCHIVE_FATAL); 676 } 677#if defined(_WIN32) && !defined(__CYGWIN__) 678 /* NOTE: This might break the pathname 679 * if the current code page is CP932 and 680 * the pathname includes a character '\' 681 * as a part of its multibyte pathname. */ 682 if (p[strlen(p) -1] == '\\') 683 path_length--; 684 else 685#endif 686 archive_strncpy(&as, p, path_length); 687 archive_strappend_char(&as, '/'); 688 archive_entry_copy_pathname( 689 entry_original, as.s); 690 archive_string_free(&as); 691 } 692 break; 693 } 694 case AE_IFSOCK: 695 archive_set_error(&a->archive, 696 ARCHIVE_ERRNO_FILE_FORMAT, 697 "tar format cannot archive socket"); 698 return (ARCHIVE_FAILED); 699 default: 700 archive_set_error(&a->archive, 701 ARCHIVE_ERRNO_FILE_FORMAT, 702 "tar format cannot archive this (type=0%lo)", 703 (unsigned long) 704 archive_entry_filetype(entry_original)); 705 return (ARCHIVE_FAILED); 706 } 707 } 708 709 /* 710 * If Mac OS metadata blob is here, recurse to write that 711 * as a separate entry. This is really a pretty poor design: 712 * In particular, it doubles the overhead for long filenames. 713 * TODO: Help Apple folks design something better and figure 714 * out how to transition from this legacy format. 715 * 716 * Note that this code is present on every platform; clients 717 * on non-Mac are unlikely to ever provide this data, but 718 * applications that copy entries from one archive to another 719 * should not lose data just because the local filesystem 720 * can't store it. 721 */ 722 mac_metadata = 723 archive_entry_mac_metadata(entry_original, &mac_metadata_size); 724 if (mac_metadata != NULL) { 725 const char *oname; 726 char *name, *bname; 727 size_t name_length; 728 struct archive_entry *extra = archive_entry_new2(&a->archive); 729 730 oname = archive_entry_pathname(entry_original); 731 name_length = strlen(oname); 732 name = malloc(name_length + 3); 733 if (name == NULL || extra == NULL) { 734 /* XXX error message */ 735 archive_entry_free(extra); 736 free(name); 737 return (ARCHIVE_FAILED); 738 } 739 strcpy(name, oname); 740 /* Find last '/'; strip trailing '/' characters */ 741 bname = strrchr(name, '/'); 742 while (bname != NULL && bname[1] == '\0') { 743 *bname = '\0'; 744 bname = strrchr(name, '/'); 745 } 746 if (bname == NULL) { 747 memmove(name + 2, name, name_length + 1); 748 memmove(name, "._", 2); 749 } else { 750 bname += 1; 751 memmove(bname + 2, bname, strlen(bname) + 1); 752 memmove(bname, "._", 2); 753 } 754 archive_entry_copy_pathname(extra, name); 755 free(name); 756 757 archive_entry_set_size(extra, mac_metadata_size); 758 archive_entry_set_filetype(extra, AE_IFREG); 759 archive_entry_set_perm(extra, 760 archive_entry_perm(entry_original)); 761 archive_entry_set_mtime(extra, 762 archive_entry_mtime(entry_original), 763 archive_entry_mtime_nsec(entry_original)); 764 archive_entry_set_gid(extra, 765 archive_entry_gid(entry_original)); 766 archive_entry_set_gname(extra, 767 archive_entry_gname(entry_original)); 768 archive_entry_set_uid(extra, 769 archive_entry_uid(entry_original)); 770 archive_entry_set_uname(extra, 771 archive_entry_uname(entry_original)); 772 773 /* Recurse to write the special copyfile entry. */ 774 r = archive_write_pax_header(a, extra); 775 archive_entry_free(extra); 776 if (r < ARCHIVE_WARN) 777 return (r); 778 if (r < ret) 779 ret = r; 780 r = (int)archive_write_pax_data(a, mac_metadata, 781 mac_metadata_size); 782 if (r < ARCHIVE_WARN) 783 return (r); 784 if (r < ret) 785 ret = r; 786 r = archive_write_pax_finish_entry(a); 787 if (r < ARCHIVE_WARN) 788 return (r); 789 if (r < ret) 790 ret = r; 791 } 792 793 /* Copy entry so we can modify it as needed. */ 794#if defined(_WIN32) && !defined(__CYGWIN__) 795 /* Make sure the path separators in pathname, hardlink and symlink 796 * are all slash '/', not the Windows path separator '\'. */ 797 entry_main = __la_win_entry_in_posix_pathseparator(entry_original); 798 if (entry_main == entry_original) 799 entry_main = archive_entry_clone(entry_original); 800#else 801 entry_main = archive_entry_clone(entry_original); 802#endif 803 if (entry_main == NULL) { 804 archive_set_error(&a->archive, ENOMEM, 805 "Can't allocate pax data"); 806 return(ARCHIVE_FATAL); 807 } 808 archive_string_empty(&(pax->pax_header)); /* Blank our work area. */ 809 archive_string_empty(&(pax->sparse_map)); 810 sparse_total = 0; 811 sparse_list_clear(pax); 812 813 if (hardlink == NULL && 814 archive_entry_filetype(entry_main) == AE_IFREG) 815 sparse_count = archive_entry_sparse_reset(entry_main); 816 else 817 sparse_count = 0; 818 if (sparse_count) { 819 int64_t offset, length, last_offset = 0; 820 /* Get the last entry of sparse block. */ 821 while (archive_entry_sparse_next( 822 entry_main, &offset, &length) == ARCHIVE_OK) 823 last_offset = offset + length; 824 825 /* If the last sparse block does not reach the end of file, 826 * We have to add a empty sparse block as the last entry to 827 * manage storing file data. */ 828 if (last_offset < archive_entry_size(entry_main)) 829 archive_entry_sparse_add_entry(entry_main, 830 archive_entry_size(entry_main), 0); 831 sparse_count = archive_entry_sparse_reset(entry_main); 832 } 833 834 /* 835 * First, check the name fields and see if any of them 836 * require binary coding. If any of them does, then all of 837 * them do. 838 */ 839 r = get_entry_pathname(a, entry_main, &path, &path_length, sconv); 840 if (r == ARCHIVE_FATAL) 841 return (r); 842 else if (r != ARCHIVE_OK) { 843 r = get_entry_pathname(a, entry_main, &path, 844 &path_length, NULL); 845 if (r == ARCHIVE_FATAL) 846 return (r); 847 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 848 "Can't translate pathname '%s' to %s", path, 849 archive_string_conversion_charset_name(sconv)); 850 ret = ARCHIVE_WARN; 851 sconv = NULL;/* The header charset switches to binary mode. */ 852 } 853 r = get_entry_uname(a, entry_main, &uname, &uname_length, sconv); 854 if (r == ARCHIVE_FATAL) 855 return (r); 856 else if (r != ARCHIVE_OK) { 857 r = get_entry_uname(a, entry_main, &uname, &uname_length, NULL); 858 if (r == ARCHIVE_FATAL) 859 return (r); 860 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 861 "Can't translate uname '%s' to %s", uname, 862 archive_string_conversion_charset_name(sconv)); 863 ret = ARCHIVE_WARN; 864 sconv = NULL;/* The header charset switches to binary mode. */ 865 } 866 r = get_entry_gname(a, entry_main, &gname, &gname_length, sconv); 867 if (r == ARCHIVE_FATAL) 868 return (r); 869 else if (r != ARCHIVE_OK) { 870 r = get_entry_gname(a, entry_main, &gname, &gname_length, NULL); 871 if (r == ARCHIVE_FATAL) 872 return (r); 873 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 874 "Can't translate gname '%s' to %s", gname, 875 archive_string_conversion_charset_name(sconv)); 876 ret = ARCHIVE_WARN; 877 sconv = NULL;/* The header charset switches to binary mode. */ 878 } 879 linkpath = hardlink; 880 linkpath_length = hardlink_length; 881 if (linkpath == NULL) { 882 r = get_entry_symlink(a, entry_main, &linkpath, 883 &linkpath_length, sconv); 884 if (r == ARCHIVE_FATAL) 885 return (r); 886 else if (r != ARCHIVE_OK) { 887 r = get_entry_symlink(a, entry_main, &linkpath, 888 &linkpath_length, NULL); 889 if (r == ARCHIVE_FATAL) 890 return (r); 891 archive_set_error(&a->archive, 892 ARCHIVE_ERRNO_FILE_FORMAT, 893 "Can't translate linkname '%s' to %s", linkpath, 894 archive_string_conversion_charset_name(sconv)); 895 ret = ARCHIVE_WARN; 896 sconv = NULL; 897 } 898 } 899 900 /* If any string conversions failed, get all attributes 901 * in binary-mode. */ 902 if (sconv == NULL && !pax->opt_binary) { 903 if (hardlink != NULL) { 904 r = get_entry_hardlink(a, entry_main, &hardlink, 905 &hardlink_length, NULL); 906 if (r == ARCHIVE_FATAL) 907 return (r); 908 linkpath = hardlink; 909 linkpath_length = hardlink_length; 910 } 911 r = get_entry_pathname(a, entry_main, &path, 912 &path_length, NULL); 913 if (r == ARCHIVE_FATAL) 914 return (r); 915 r = get_entry_uname(a, entry_main, &uname, &uname_length, NULL); 916 if (r == ARCHIVE_FATAL) 917 return (r); 918 r = get_entry_gname(a, entry_main, &gname, &gname_length, NULL); 919 if (r == ARCHIVE_FATAL) 920 return (r); 921 } 922 923 /* Store the header encoding first, to be nice to readers. */ 924 if (sconv == NULL) 925 add_pax_attr(&(pax->pax_header), "hdrcharset", "BINARY"); 926 927 928 /* 929 * If name is too long, or has non-ASCII characters, add 930 * 'path' to pax extended attrs. (Note that an unconvertible 931 * name must have non-ASCII characters.) 932 */ 933 if (has_non_ASCII(path)) { 934 /* We have non-ASCII characters. */ 935 add_pax_attr(&(pax->pax_header), "path", path); 936 archive_entry_set_pathname(entry_main, 937 build_ustar_entry_name(ustar_entry_name, 938 path, path_length, NULL)); 939 need_extension = 1; 940 } else { 941 /* We have an all-ASCII path; we'd like to just store 942 * it in the ustar header if it will fit. Yes, this 943 * duplicates some of the logic in 944 * archive_write_set_format_ustar.c 945 */ 946 if (path_length <= 100) { 947 /* Fits in the old 100-char tar name field. */ 948 } else { 949 /* Find largest suffix that will fit. */ 950 /* Note: strlen() > 100, so strlen() - 100 - 1 >= 0 */ 951 suffix = strchr(path + path_length - 100 - 1, '/'); 952 /* Don't attempt an empty prefix. */ 953 if (suffix == path) 954 suffix = strchr(suffix + 1, '/'); 955 /* We can put it in the ustar header if it's 956 * all ASCII and it's either <= 100 characters 957 * or can be split at a '/' into a prefix <= 958 * 155 chars and a suffix <= 100 chars. (Note 959 * the strchr() above will return NULL exactly 960 * when the path can't be split.) 961 */ 962 if (suffix == NULL /* Suffix > 100 chars. */ 963 || suffix[1] == '\0' /* empty suffix */ 964 || suffix - path > 155) /* Prefix > 155 chars */ 965 { 966 add_pax_attr(&(pax->pax_header), "path", path); 967 archive_entry_set_pathname(entry_main, 968 build_ustar_entry_name(ustar_entry_name, 969 path, path_length, NULL)); 970 need_extension = 1; 971 } 972 } 973 } 974 975 if (linkpath != NULL) { 976 /* If link name is too long or has non-ASCII characters, add 977 * 'linkpath' to pax extended attrs. */ 978 if (linkpath_length > 100 || has_non_ASCII(linkpath)) { 979 add_pax_attr(&(pax->pax_header), "linkpath", linkpath); 980 if (linkpath_length > 100) { 981 if (hardlink != NULL) 982 archive_entry_set_hardlink(entry_main, 983 "././@LongHardLink"); 984 else 985 archive_entry_set_symlink(entry_main, 986 "././@LongSymLink"); 987 } 988 need_extension = 1; 989 } 990 } 991 /* Save a pathname since it will be renamed if `entry_main` has 992 * sparse blocks. */ 993 archive_string_init(&entry_name); 994 archive_strcpy(&entry_name, archive_entry_pathname(entry_main)); 995 996 /* If file size is too large, add 'size' to pax extended attrs. */ 997 if (archive_entry_size(entry_main) >= (((int64_t)1) << 33)) { 998 add_pax_attr_int(&(pax->pax_header), "size", 999 archive_entry_size(entry_main)); 1000 need_extension = 1; 1001 } 1002 1003 /* If numeric GID is too large, add 'gid' to pax extended attrs. */ 1004 if ((unsigned int)archive_entry_gid(entry_main) >= (1 << 18)) { 1005 add_pax_attr_int(&(pax->pax_header), "gid", 1006 archive_entry_gid(entry_main)); 1007 need_extension = 1; 1008 } 1009 1010 /* If group name is too large or has non-ASCII characters, add 1011 * 'gname' to pax extended attrs. */ 1012 if (gname != NULL) { 1013 if (gname_length > 31 || has_non_ASCII(gname)) { 1014 add_pax_attr(&(pax->pax_header), "gname", gname); 1015 need_extension = 1; 1016 } 1017 } 1018 1019 /* If numeric UID is too large, add 'uid' to pax extended attrs. */ 1020 if ((unsigned int)archive_entry_uid(entry_main) >= (1 << 18)) { 1021 add_pax_attr_int(&(pax->pax_header), "uid", 1022 archive_entry_uid(entry_main)); 1023 need_extension = 1; 1024 } 1025 1026 /* Add 'uname' to pax extended attrs if necessary. */ 1027 if (uname != NULL) { 1028 if (uname_length > 31 || has_non_ASCII(uname)) { 1029 add_pax_attr(&(pax->pax_header), "uname", uname); 1030 need_extension = 1; 1031 } 1032 } 1033 1034 /* 1035 * POSIX/SUSv3 doesn't provide a standard key for large device 1036 * numbers. I use the same keys here that Joerg Schilling 1037 * used for 'star.' (Which, somewhat confusingly, are called 1038 * "devXXX" even though they code "rdev" values.) No doubt, 1039 * other implementations use other keys. Note that there's no 1040 * reason we can't write the same information into a number of 1041 * different keys. 1042 * 1043 * Of course, this is only needed for block or char device entries. 1044 */ 1045 if (archive_entry_filetype(entry_main) == AE_IFBLK 1046 || archive_entry_filetype(entry_main) == AE_IFCHR) { 1047 /* 1048 * If rdevmajor is too large, add 'SCHILY.devmajor' to 1049 * extended attributes. 1050 */ 1051 int rdevmajor, rdevminor; 1052 rdevmajor = archive_entry_rdevmajor(entry_main); 1053 rdevminor = archive_entry_rdevminor(entry_main); 1054 if (rdevmajor >= (1 << 18)) { 1055 add_pax_attr_int(&(pax->pax_header), "SCHILY.devmajor", 1056 rdevmajor); 1057 /* 1058 * Non-strict formatting below means we don't 1059 * have to truncate here. Not truncating improves 1060 * the chance that some more modern tar archivers 1061 * (such as GNU tar 1.13) can restore the full 1062 * value even if they don't understand the pax 1063 * extended attributes. See my rant below about 1064 * file size fields for additional details. 1065 */ 1066 /* archive_entry_set_rdevmajor(entry_main, 1067 rdevmajor & ((1 << 18) - 1)); */ 1068 need_extension = 1; 1069 } 1070 1071 /* 1072 * If devminor is too large, add 'SCHILY.devminor' to 1073 * extended attributes. 1074 */ 1075 if (rdevminor >= (1 << 18)) { 1076 add_pax_attr_int(&(pax->pax_header), "SCHILY.devminor", 1077 rdevminor); 1078 /* Truncation is not necessary here, either. */ 1079 /* archive_entry_set_rdevminor(entry_main, 1080 rdevminor & ((1 << 18) - 1)); */ 1081 need_extension = 1; 1082 } 1083 } 1084 1085 /* 1086 * Technically, the mtime field in the ustar header can 1087 * support 33 bits, but many platforms use signed 32-bit time 1088 * values. The cutoff of 0x7fffffff here is a compromise. 1089 * Yes, this check is duplicated just below; this helps to 1090 * avoid writing an mtime attribute just to handle a 1091 * high-resolution timestamp in "restricted pax" mode. 1092 */ 1093 if (!need_extension && 1094 ((archive_entry_mtime(entry_main) < 0) 1095 || (archive_entry_mtime(entry_main) >= 0x7fffffff))) 1096 need_extension = 1; 1097 1098 /* I use a star-compatible file flag attribute. */ 1099 p = archive_entry_fflags_text(entry_main); 1100 if (!need_extension && p != NULL && *p != '\0') 1101 need_extension = 1; 1102 1103 /* If there are extended attributes, we need an extension */ 1104 if (!need_extension && archive_entry_xattr_count(entry_original) > 0) 1105 need_extension = 1; 1106 1107 /* If there are sparse info, we need an extension */ 1108 if (!need_extension && sparse_count > 0) 1109 need_extension = 1; 1110 1111 acl_types = archive_entry_acl_types(entry_original); 1112 1113 /* If there are any ACL entries, we need an extension */ 1114 if (!need_extension && acl_types != 0) 1115 need_extension = 1; 1116 1117 /* If the symlink type is defined, we need an extension */ 1118 if (!need_extension && archive_entry_symlink_type(entry_main) > 0) 1119 need_extension = 1; 1120 1121 /* 1122 * Libarchive used to include these in extended headers for 1123 * restricted pax format, but that confused people who 1124 * expected ustar-like time semantics. So now we only include 1125 * them in full pax format. 1126 */ 1127 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_RESTRICTED) { 1128 if (archive_entry_ctime(entry_main) != 0 || 1129 archive_entry_ctime_nsec(entry_main) != 0) 1130 add_pax_attr_time(&(pax->pax_header), "ctime", 1131 archive_entry_ctime(entry_main), 1132 archive_entry_ctime_nsec(entry_main)); 1133 1134 if (archive_entry_atime(entry_main) != 0 || 1135 archive_entry_atime_nsec(entry_main) != 0) 1136 add_pax_attr_time(&(pax->pax_header), "atime", 1137 archive_entry_atime(entry_main), 1138 archive_entry_atime_nsec(entry_main)); 1139 1140 /* Store birth/creationtime only if it's earlier than mtime */ 1141 if (archive_entry_birthtime_is_set(entry_main) && 1142 archive_entry_birthtime(entry_main) 1143 < archive_entry_mtime(entry_main)) 1144 add_pax_attr_time(&(pax->pax_header), 1145 "LIBARCHIVE.creationtime", 1146 archive_entry_birthtime(entry_main), 1147 archive_entry_birthtime_nsec(entry_main)); 1148 } 1149 1150 /* 1151 * The following items are handled differently in "pax 1152 * restricted" format. In particular, in "pax restricted" 1153 * format they won't be added unless need_extension is 1154 * already set (we're already generating an extended header, so 1155 * may as well include these). 1156 */ 1157 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_RESTRICTED || 1158 need_extension) { 1159 if (archive_entry_mtime(entry_main) < 0 || 1160 archive_entry_mtime(entry_main) >= 0x7fffffff || 1161 archive_entry_mtime_nsec(entry_main) != 0) 1162 add_pax_attr_time(&(pax->pax_header), "mtime", 1163 archive_entry_mtime(entry_main), 1164 archive_entry_mtime_nsec(entry_main)); 1165 1166 /* I use a star-compatible file flag attribute. */ 1167 p = archive_entry_fflags_text(entry_main); 1168 if (p != NULL && *p != '\0') 1169 add_pax_attr(&(pax->pax_header), "SCHILY.fflags", p); 1170 1171 /* I use star-compatible ACL attributes. */ 1172 if ((acl_types & ARCHIVE_ENTRY_ACL_TYPE_NFS4) != 0) { 1173 ret = add_pax_acl(a, entry_original, pax, 1174 ARCHIVE_ENTRY_ACL_STYLE_EXTRA_ID | 1175 ARCHIVE_ENTRY_ACL_STYLE_SEPARATOR_COMMA | 1176 ARCHIVE_ENTRY_ACL_STYLE_COMPACT); 1177 if (ret == ARCHIVE_FATAL) 1178 return (ARCHIVE_FATAL); 1179 } 1180 if (acl_types & ARCHIVE_ENTRY_ACL_TYPE_ACCESS) { 1181 ret = add_pax_acl(a, entry_original, pax, 1182 ARCHIVE_ENTRY_ACL_TYPE_ACCESS | 1183 ARCHIVE_ENTRY_ACL_STYLE_EXTRA_ID | 1184 ARCHIVE_ENTRY_ACL_STYLE_SEPARATOR_COMMA); 1185 if (ret == ARCHIVE_FATAL) 1186 return (ARCHIVE_FATAL); 1187 } 1188 if (acl_types & ARCHIVE_ENTRY_ACL_TYPE_DEFAULT) { 1189 ret = add_pax_acl(a, entry_original, pax, 1190 ARCHIVE_ENTRY_ACL_TYPE_DEFAULT | 1191 ARCHIVE_ENTRY_ACL_STYLE_EXTRA_ID | 1192 ARCHIVE_ENTRY_ACL_STYLE_SEPARATOR_COMMA); 1193 if (ret == ARCHIVE_FATAL) 1194 return (ARCHIVE_FATAL); 1195 } 1196 1197 /* We use GNU-tar-compatible sparse attributes. */ 1198 if (sparse_count > 0) { 1199 int64_t soffset, slength; 1200 1201 add_pax_attr_int(&(pax->pax_header), 1202 "GNU.sparse.major", 1); 1203 add_pax_attr_int(&(pax->pax_header), 1204 "GNU.sparse.minor", 0); 1205 /* 1206 * Make sure to store the original path, since 1207 * truncation to ustar limit happened already. 1208 */ 1209 add_pax_attr(&(pax->pax_header), 1210 "GNU.sparse.name", path); 1211 add_pax_attr_int(&(pax->pax_header), 1212 "GNU.sparse.realsize", 1213 archive_entry_size(entry_main)); 1214 1215 /* Rename the file name which will be used for 1216 * ustar header to a special name, which GNU 1217 * PAX Format 1.0 requires */ 1218 archive_entry_set_pathname(entry_main, 1219 build_gnu_sparse_name(gnu_sparse_name, 1220 entry_name.s)); 1221 1222 /* 1223 * - Make a sparse map, which will precede a file data. 1224 * - Get the total size of available data of sparse. 1225 */ 1226 archive_string_sprintf(&(pax->sparse_map), "%d\n", 1227 sparse_count); 1228 while (archive_entry_sparse_next(entry_main, 1229 &soffset, &slength) == ARCHIVE_OK) { 1230 archive_string_sprintf(&(pax->sparse_map), 1231 "%jd\n%jd\n", 1232 (intmax_t)soffset, 1233 (intmax_t)slength); 1234 sparse_total += slength; 1235 if (sparse_list_add(pax, soffset, slength) 1236 != ARCHIVE_OK) { 1237 archive_set_error(&a->archive, 1238 ENOMEM, 1239 "Can't allocate memory"); 1240 archive_entry_free(entry_main); 1241 archive_string_free(&entry_name); 1242 return (ARCHIVE_FATAL); 1243 } 1244 } 1245 } 1246 1247 /* Store extended attributes */ 1248 if (archive_write_pax_header_xattrs(a, pax, entry_original) 1249 == ARCHIVE_FATAL) { 1250 archive_entry_free(entry_main); 1251 archive_string_free(&entry_name); 1252 return (ARCHIVE_FATAL); 1253 } 1254 1255 /* Store extended symlink information */ 1256 if (archive_entry_symlink_type(entry_main) == 1257 AE_SYMLINK_TYPE_FILE) { 1258 add_pax_attr(&(pax->pax_header), 1259 "LIBARCHIVE.symlinktype", "file"); 1260 } else if (archive_entry_symlink_type(entry_main) == 1261 AE_SYMLINK_TYPE_DIRECTORY) { 1262 add_pax_attr(&(pax->pax_header), 1263 "LIBARCHIVE.symlinktype", "dir"); 1264 } 1265 } 1266 1267 /* Only regular files have data. */ 1268 if (archive_entry_filetype(entry_main) != AE_IFREG) 1269 archive_entry_set_size(entry_main, 0); 1270 1271 /* 1272 * Pax-restricted does not store data for hardlinks, in order 1273 * to improve compatibility with ustar. 1274 */ 1275 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE && 1276 hardlink != NULL) 1277 archive_entry_set_size(entry_main, 0); 1278 1279 /* 1280 * XXX Full pax interchange format does permit a hardlink 1281 * entry to have data associated with it. I'm not supporting 1282 * that here because the client expects me to tell them whether 1283 * or not this format expects data for hardlinks. If I 1284 * don't check here, then every pax archive will end up with 1285 * duplicated data for hardlinks. Someday, there may be 1286 * need to select this behavior, in which case the following 1287 * will need to be revisited. XXX 1288 */ 1289 if (hardlink != NULL) 1290 archive_entry_set_size(entry_main, 0); 1291 1292 /* Save a real file size. */ 1293 real_size = archive_entry_size(entry_main); 1294 /* 1295 * Overwrite a file size by the total size of sparse blocks and 1296 * the size of sparse map info. That file size is the length of 1297 * the data, which we will exactly store into an archive file. 1298 */ 1299 if (archive_strlen(&(pax->sparse_map))) { 1300 size_t mapsize = archive_strlen(&(pax->sparse_map)); 1301 pax->sparse_map_padding = 0x1ff & (-(ssize_t)mapsize); 1302 archive_entry_set_size(entry_main, 1303 mapsize + pax->sparse_map_padding + sparse_total); 1304 } 1305 1306 /* Format 'ustar' header for main entry. 1307 * 1308 * The trouble with file size: If the reader can't understand 1309 * the file size, they may not be able to locate the next 1310 * entry and the rest of the archive is toast. Pax-compliant 1311 * readers are supposed to ignore the file size in the main 1312 * header, so the question becomes how to maximize portability 1313 * for readers that don't support pax attribute extensions. 1314 * For maximum compatibility, I permit numeric extensions in 1315 * the main header so that the file size stored will always be 1316 * correct, even if it's in a format that only some 1317 * implementations understand. The technique used here is: 1318 * 1319 * a) If possible, follow the standard exactly. This handles 1320 * files up to 8 gigabytes minus 1. 1321 * 1322 * b) If that fails, try octal but omit the field terminator. 1323 * That handles files up to 64 gigabytes minus 1. 1324 * 1325 * c) Otherwise, use base-256 extensions. That handles files 1326 * up to 2^63 in this implementation, with the potential to 1327 * go up to 2^94. That should hold us for a while. ;-) 1328 * 1329 * The non-strict formatter uses similar logic for other 1330 * numeric fields, though they're less critical. 1331 */ 1332 if (__archive_write_format_header_ustar(a, ustarbuff, entry_main, -1, 0, 1333 NULL) == ARCHIVE_FATAL) 1334 return (ARCHIVE_FATAL); 1335 1336 /* If we built any extended attributes, write that entry first. */ 1337 if (archive_strlen(&(pax->pax_header)) > 0) { 1338 struct archive_entry *pax_attr_entry; 1339 time_t s; 1340 int64_t uid, gid; 1341 int mode; 1342 1343 pax_attr_entry = archive_entry_new2(&a->archive); 1344 p = entry_name.s; 1345 archive_entry_set_pathname(pax_attr_entry, 1346 build_pax_attribute_name(pax_entry_name, p)); 1347 archive_entry_set_size(pax_attr_entry, 1348 archive_strlen(&(pax->pax_header))); 1349 /* Copy uid/gid (but clip to ustar limits). */ 1350 uid = archive_entry_uid(entry_main); 1351 if (uid >= 1 << 18) 1352 uid = (1 << 18) - 1; 1353 archive_entry_set_uid(pax_attr_entry, uid); 1354 gid = archive_entry_gid(entry_main); 1355 if (gid >= 1 << 18) 1356 gid = (1 << 18) - 1; 1357 archive_entry_set_gid(pax_attr_entry, gid); 1358 /* Copy mode over (but not setuid/setgid bits) */ 1359 mode = archive_entry_mode(entry_main); 1360#ifdef S_ISUID 1361 mode &= ~S_ISUID; 1362#endif 1363#ifdef S_ISGID 1364 mode &= ~S_ISGID; 1365#endif 1366#ifdef S_ISVTX 1367 mode &= ~S_ISVTX; 1368#endif 1369 archive_entry_set_mode(pax_attr_entry, mode); 1370 1371 /* Copy uname/gname. */ 1372 archive_entry_set_uname(pax_attr_entry, 1373 archive_entry_uname(entry_main)); 1374 archive_entry_set_gname(pax_attr_entry, 1375 archive_entry_gname(entry_main)); 1376 1377 /* Copy mtime, but clip to ustar limits. */ 1378 s = archive_entry_mtime(entry_main); 1379 if (s < 0) { s = 0; } 1380 if (s >= 0x7fffffff) { s = 0x7fffffff; } 1381 archive_entry_set_mtime(pax_attr_entry, s, 0); 1382 1383 /* Standard ustar doesn't support atime. */ 1384 archive_entry_set_atime(pax_attr_entry, 0, 0); 1385 1386 /* Standard ustar doesn't support ctime. */ 1387 archive_entry_set_ctime(pax_attr_entry, 0, 0); 1388 1389 r = __archive_write_format_header_ustar(a, paxbuff, 1390 pax_attr_entry, 'x', 1, NULL); 1391 1392 archive_entry_free(pax_attr_entry); 1393 1394 /* Note that the 'x' header shouldn't ever fail to format */ 1395 if (r < ARCHIVE_WARN) { 1396 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1397 "archive_write_pax_header: " 1398 "'x' header failed?! This can't happen.\n"); 1399 return (ARCHIVE_FATAL); 1400 } else if (r < ret) 1401 ret = r; 1402 r = __archive_write_output(a, paxbuff, 512); 1403 if (r != ARCHIVE_OK) { 1404 sparse_list_clear(pax); 1405 pax->entry_bytes_remaining = 0; 1406 pax->entry_padding = 0; 1407 return (ARCHIVE_FATAL); 1408 } 1409 1410 pax->entry_bytes_remaining = archive_strlen(&(pax->pax_header)); 1411 pax->entry_padding = 1412 0x1ff & (-(int64_t)pax->entry_bytes_remaining); 1413 1414 r = __archive_write_output(a, pax->pax_header.s, 1415 archive_strlen(&(pax->pax_header))); 1416 if (r != ARCHIVE_OK) { 1417 /* If a write fails, we're pretty much toast. */ 1418 return (ARCHIVE_FATAL); 1419 } 1420 /* Pad out the end of the entry. */ 1421 r = __archive_write_nulls(a, (size_t)pax->entry_padding); 1422 if (r != ARCHIVE_OK) { 1423 /* If a write fails, we're pretty much toast. */ 1424 return (ARCHIVE_FATAL); 1425 } 1426 pax->entry_bytes_remaining = pax->entry_padding = 0; 1427 } 1428 1429 /* Write the header for main entry. */ 1430 r = __archive_write_output(a, ustarbuff, 512); 1431 if (r != ARCHIVE_OK) 1432 return (r); 1433 1434 /* 1435 * Inform the client of the on-disk size we're using, so 1436 * they can avoid unnecessarily writing a body for something 1437 * that we're just going to ignore. 1438 */ 1439 archive_entry_set_size(entry_original, real_size); 1440 if (pax->sparse_list == NULL && real_size > 0) { 1441 /* This is not a sparse file but we handle its data as 1442 * a sparse block. */ 1443 sparse_list_add(pax, 0, real_size); 1444 sparse_total = real_size; 1445 } 1446 pax->entry_padding = 0x1ff & (-(int64_t)sparse_total); 1447 archive_entry_free(entry_main); 1448 archive_string_free(&entry_name); 1449 1450 return (ret); 1451} 1452 1453/* 1454 * We need a valid name for the regular 'ustar' entry. This routine 1455 * tries to hack something more-or-less reasonable. 1456 * 1457 * The approach here tries to preserve leading dir names. We do so by 1458 * working with four sections: 1459 * 1) "prefix" directory names, 1460 * 2) "suffix" directory names, 1461 * 3) inserted dir name (optional), 1462 * 4) filename. 1463 * 1464 * These sections must satisfy the following requirements: 1465 * * Parts 1 & 2 together form an initial portion of the dir name. 1466 * * Part 3 is specified by the caller. (It should not contain a leading 1467 * or trailing '/'.) 1468 * * Part 4 forms an initial portion of the base filename. 1469 * * The filename must be <= 99 chars to fit the ustar 'name' field. 1470 * * Parts 2, 3, 4 together must be <= 99 chars to fit the ustar 'name' fld. 1471 * * Part 1 must be <= 155 chars to fit the ustar 'prefix' field. 1472 * * If the original name ends in a '/', the new name must also end in a '/' 1473 * * Trailing '/.' sequences may be stripped. 1474 * 1475 * Note: Recall that the ustar format does not store the '/' separating 1476 * parts 1 & 2, but does store the '/' separating parts 2 & 3. 1477 */ 1478static char * 1479build_ustar_entry_name(char *dest, const char *src, size_t src_length, 1480 const char *insert) 1481{ 1482 const char *prefix, *prefix_end; 1483 const char *suffix, *suffix_end; 1484 const char *filename, *filename_end; 1485 char *p; 1486 int need_slash = 0; /* Was there a trailing slash? */ 1487 size_t suffix_length = 99; 1488 size_t insert_length; 1489 1490 /* Length of additional dir element to be added. */ 1491 if (insert == NULL) 1492 insert_length = 0; 1493 else 1494 /* +2 here allows for '/' before and after the insert. */ 1495 insert_length = strlen(insert) + 2; 1496 1497 /* Step 0: Quick bailout in a common case. */ 1498 if (src_length < 100 && insert == NULL) { 1499 strncpy(dest, src, src_length); 1500 dest[src_length] = '\0'; 1501 return (dest); 1502 } 1503 1504 /* Step 1: Locate filename and enforce the length restriction. */ 1505 filename_end = src + src_length; 1506 /* Remove trailing '/' chars and '/.' pairs. */ 1507 for (;;) { 1508 if (filename_end > src && filename_end[-1] == '/') { 1509 filename_end --; 1510 need_slash = 1; /* Remember to restore trailing '/'. */ 1511 continue; 1512 } 1513 if (filename_end > src + 1 && filename_end[-1] == '.' 1514 && filename_end[-2] == '/') { 1515 filename_end -= 2; 1516 need_slash = 1; /* "foo/." will become "foo/" */ 1517 continue; 1518 } 1519 break; 1520 } 1521 if (need_slash) 1522 suffix_length--; 1523 /* Find start of filename. */ 1524 filename = filename_end - 1; 1525 while ((filename > src) && (*filename != '/')) 1526 filename --; 1527 if ((*filename == '/') && (filename < filename_end - 1)) 1528 filename ++; 1529 /* Adjust filename_end so that filename + insert fits in 99 chars. */ 1530 suffix_length -= insert_length; 1531 if (filename_end > filename + suffix_length) 1532 filename_end = filename + suffix_length; 1533 /* Calculate max size for "suffix" section (#3 above). */ 1534 suffix_length -= filename_end - filename; 1535 1536 /* Step 2: Locate the "prefix" section of the dirname, including 1537 * trailing '/'. */ 1538 prefix = src; 1539 prefix_end = prefix + 155; 1540 if (prefix_end > filename) 1541 prefix_end = filename; 1542 while (prefix_end > prefix && *prefix_end != '/') 1543 prefix_end--; 1544 if ((prefix_end < filename) && (*prefix_end == '/')) 1545 prefix_end++; 1546 1547 /* Step 3: Locate the "suffix" section of the dirname, 1548 * including trailing '/'. */ 1549 suffix = prefix_end; 1550 suffix_end = suffix + suffix_length; /* Enforce limit. */ 1551 if (suffix_end > filename) 1552 suffix_end = filename; 1553 if (suffix_end < suffix) 1554 suffix_end = suffix; 1555 while (suffix_end > suffix && *suffix_end != '/') 1556 suffix_end--; 1557 if ((suffix_end < filename) && (*suffix_end == '/')) 1558 suffix_end++; 1559 1560 /* Step 4: Build the new name. */ 1561 /* The OpenBSD strlcpy function is safer, but less portable. */ 1562 /* Rather than maintain two versions, just use the strncpy version. */ 1563 p = dest; 1564 if (prefix_end > prefix) { 1565 strncpy(p, prefix, prefix_end - prefix); 1566 p += prefix_end - prefix; 1567 } 1568 if (suffix_end > suffix) { 1569 strncpy(p, suffix, suffix_end - suffix); 1570 p += suffix_end - suffix; 1571 } 1572 if (insert != NULL) { 1573 /* Note: assume insert does not have leading or trailing '/' */ 1574 strcpy(p, insert); 1575 p += strlen(insert); 1576 *p++ = '/'; 1577 } 1578 strncpy(p, filename, filename_end - filename); 1579 p += filename_end - filename; 1580 if (need_slash) 1581 *p++ = '/'; 1582 *p = '\0'; 1583 1584 return (dest); 1585} 1586 1587/* 1588 * The ustar header for the pax extended attributes must have a 1589 * reasonable name: SUSv3 requires 'dirname'/PaxHeader.'pid'/'filename' 1590 * where 'pid' is the PID of the archiving process. Unfortunately, 1591 * that makes testing a pain since the output varies for each run, 1592 * so I'm sticking with the simpler 'dirname'/PaxHeader/'filename' 1593 * for now. (Someday, I'll make this settable. Then I can use the 1594 * SUS recommendation as default and test harnesses can override it 1595 * to get predictable results.) 1596 * 1597 * Joerg Schilling has argued that this is unnecessary because, in 1598 * practice, if the pax extended attributes get extracted as regular 1599 * files, no one is going to bother reading those attributes to 1600 * manually restore them. Based on this, 'star' uses 1601 * /tmp/PaxHeader/'basename' as the ustar header name. This is a 1602 * tempting argument, in part because it's simpler than the SUSv3 1603 * recommendation, but I'm not entirely convinced. I'm also 1604 * uncomfortable with the fact that "/tmp" is a Unix-ism. 1605 * 1606 * The following routine leverages build_ustar_entry_name() above and 1607 * so is simpler than you might think. It just needs to provide the 1608 * additional path element and handle a few pathological cases). 1609 */ 1610static char * 1611build_pax_attribute_name(char *dest, const char *src) 1612{ 1613 char buff[64]; 1614 const char *p; 1615 1616 /* Handle the null filename case. */ 1617 if (src == NULL || *src == '\0') { 1618 strcpy(dest, "PaxHeader/blank"); 1619 return (dest); 1620 } 1621 1622 /* Prune final '/' and other unwanted final elements. */ 1623 p = src + strlen(src); 1624 for (;;) { 1625 /* Ends in "/", remove the '/' */ 1626 if (p > src && p[-1] == '/') { 1627 --p; 1628 continue; 1629 } 1630 /* Ends in "/.", remove the '.' */ 1631 if (p > src + 1 && p[-1] == '.' 1632 && p[-2] == '/') { 1633 --p; 1634 continue; 1635 } 1636 break; 1637 } 1638 1639 /* Pathological case: After above, there was nothing left. 1640 * This includes "/." "/./." "/.//./." etc. */ 1641 if (p == src) { 1642 strcpy(dest, "/PaxHeader/rootdir"); 1643 return (dest); 1644 } 1645 1646 /* Convert unadorned "." into a suitable filename. */ 1647 if (*src == '.' && p == src + 1) { 1648 strcpy(dest, "PaxHeader/currentdir"); 1649 return (dest); 1650 } 1651 1652 /* 1653 * TODO: Push this string into the 'pax' structure to avoid 1654 * recomputing it every time. That will also open the door 1655 * to having clients override it. 1656 */ 1657#if HAVE_GETPID && 0 /* Disable this for now; see above comment. */ 1658 sprintf(buff, "PaxHeader.%d", getpid()); 1659#else 1660 /* If the platform can't fetch the pid, don't include it. */ 1661 strcpy(buff, "PaxHeader"); 1662#endif 1663 /* General case: build a ustar-compatible name adding 1664 * "/PaxHeader/". */ 1665 build_ustar_entry_name(dest, src, p - src, buff); 1666 1667 return (dest); 1668} 1669 1670/* 1671 * GNU PAX Format 1.0 requires the special name, which pattern is: 1672 * <dir>/GNUSparseFile.<pid>/<original file name> 1673 * 1674 * Since reproducible archives are more important, use 0 as pid. 1675 * 1676 * This function is used for only Sparse file, a file type of which 1677 * is regular file. 1678 */ 1679static char * 1680build_gnu_sparse_name(char *dest, const char *src) 1681{ 1682 const char *p; 1683 1684 /* Handle the null filename case. */ 1685 if (src == NULL || *src == '\0') { 1686 strcpy(dest, "GNUSparseFile/blank"); 1687 return (dest); 1688 } 1689 1690 /* Prune final '/' and other unwanted final elements. */ 1691 p = src + strlen(src); 1692 for (;;) { 1693 /* Ends in "/", remove the '/' */ 1694 if (p > src && p[-1] == '/') { 1695 --p; 1696 continue; 1697 } 1698 /* Ends in "/.", remove the '.' */ 1699 if (p > src + 1 && p[-1] == '.' 1700 && p[-2] == '/') { 1701 --p; 1702 continue; 1703 } 1704 break; 1705 } 1706 1707 /* General case: build a ustar-compatible name adding 1708 * "/GNUSparseFile/". */ 1709 build_ustar_entry_name(dest, src, p - src, "GNUSparseFile.0"); 1710 1711 return (dest); 1712} 1713 1714/* Write two null blocks for the end of archive */ 1715static int 1716archive_write_pax_close(struct archive_write *a) 1717{ 1718 return (__archive_write_nulls(a, 512 * 2)); 1719} 1720 1721static int 1722archive_write_pax_free(struct archive_write *a) 1723{ 1724 struct pax *pax; 1725 1726 pax = (struct pax *)a->format_data; 1727 if (pax == NULL) 1728 return (ARCHIVE_OK); 1729 1730 archive_string_free(&pax->pax_header); 1731 archive_string_free(&pax->sparse_map); 1732 archive_string_free(&pax->l_url_encoded_name); 1733 sparse_list_clear(pax); 1734 free(pax); 1735 a->format_data = NULL; 1736 return (ARCHIVE_OK); 1737} 1738 1739static int 1740archive_write_pax_finish_entry(struct archive_write *a) 1741{ 1742 struct pax *pax; 1743 uint64_t remaining; 1744 int ret; 1745 1746 pax = (struct pax *)a->format_data; 1747 remaining = pax->entry_bytes_remaining; 1748 if (remaining == 0) { 1749 while (pax->sparse_list) { 1750 struct sparse_block *sb; 1751 if (!pax->sparse_list->is_hole) 1752 remaining += pax->sparse_list->remaining; 1753 sb = pax->sparse_list->next; 1754 free(pax->sparse_list); 1755 pax->sparse_list = sb; 1756 } 1757 } 1758 ret = __archive_write_nulls(a, (size_t)(remaining + pax->entry_padding)); 1759 pax->entry_bytes_remaining = pax->entry_padding = 0; 1760 return (ret); 1761} 1762 1763static ssize_t 1764archive_write_pax_data(struct archive_write *a, const void *buff, size_t s) 1765{ 1766 struct pax *pax; 1767 size_t ws; 1768 size_t total; 1769 int ret; 1770 1771 pax = (struct pax *)a->format_data; 1772 1773 /* 1774 * According to GNU PAX format 1.0, write a sparse map 1775 * before the body. 1776 */ 1777 if (archive_strlen(&(pax->sparse_map))) { 1778 ret = __archive_write_output(a, pax->sparse_map.s, 1779 archive_strlen(&(pax->sparse_map))); 1780 if (ret != ARCHIVE_OK) 1781 return (ret); 1782 ret = __archive_write_nulls(a, pax->sparse_map_padding); 1783 if (ret != ARCHIVE_OK) 1784 return (ret); 1785 archive_string_empty(&(pax->sparse_map)); 1786 } 1787 1788 total = 0; 1789 while (total < s) { 1790 const unsigned char *p; 1791 1792 while (pax->sparse_list != NULL && 1793 pax->sparse_list->remaining == 0) { 1794 struct sparse_block *sb = pax->sparse_list->next; 1795 free(pax->sparse_list); 1796 pax->sparse_list = sb; 1797 } 1798 1799 if (pax->sparse_list == NULL) 1800 return (total); 1801 1802 p = ((const unsigned char *)buff) + total; 1803 ws = s - total; 1804 if (ws > pax->sparse_list->remaining) 1805 ws = (size_t)pax->sparse_list->remaining; 1806 1807 if (pax->sparse_list->is_hole) { 1808 /* Current block is hole thus we do not write 1809 * the body. */ 1810 pax->sparse_list->remaining -= ws; 1811 total += ws; 1812 continue; 1813 } 1814 1815 ret = __archive_write_output(a, p, ws); 1816 pax->sparse_list->remaining -= ws; 1817 total += ws; 1818 if (ret != ARCHIVE_OK) 1819 return (ret); 1820 } 1821 return (total); 1822} 1823 1824static int 1825has_non_ASCII(const char *_p) 1826{ 1827 const unsigned char *p = (const unsigned char *)_p; 1828 1829 if (p == NULL) 1830 return (1); 1831 while (*p != '\0' && *p < 128) 1832 p++; 1833 return (*p != '\0'); 1834} 1835 1836/* 1837 * Used by extended attribute support; encodes the name 1838 * so that there will be no '=' characters in the result. 1839 */ 1840static char * 1841url_encode(const char *in) 1842{ 1843 const char *s; 1844 char *d; 1845 int out_len = 0; 1846 char *out; 1847 1848 for (s = in; *s != '\0'; s++) { 1849 if (*s < 33 || *s > 126 || *s == '%' || *s == '=') 1850 out_len += 3; 1851 else 1852 out_len++; 1853 } 1854 1855 out = (char *)malloc(out_len + 1); 1856 if (out == NULL) 1857 return (NULL); 1858 1859 for (s = in, d = out; *s != '\0'; s++) { 1860 /* encode any non-printable ASCII character or '%' or '=' */ 1861 if (*s < 33 || *s > 126 || *s == '%' || *s == '=') { 1862 /* URL encoding is '%' followed by two hex digits */ 1863 *d++ = '%'; 1864 *d++ = "0123456789ABCDEF"[0x0f & (*s >> 4)]; 1865 *d++ = "0123456789ABCDEF"[0x0f & *s]; 1866 } else { 1867 *d++ = *s; 1868 } 1869 } 1870 *d = '\0'; 1871 return (out); 1872} 1873 1874/* 1875 * Encode a sequence of bytes into a C string using base-64 encoding. 1876 * 1877 * Returns a null-terminated C string allocated with malloc(); caller 1878 * is responsible for freeing the result. 1879 */ 1880static char * 1881base64_encode(const char *s, size_t len) 1882{ 1883 static const char digits[64] = 1884 { 'A','B','C','D','E','F','G','H','I','J','K','L','M','N','O', 1885 'P','Q','R','S','T','U','V','W','X','Y','Z','a','b','c','d', 1886 'e','f','g','h','i','j','k','l','m','n','o','p','q','r','s', 1887 't','u','v','w','x','y','z','0','1','2','3','4','5','6','7', 1888 '8','9','+','/' }; 1889 int v; 1890 char *d, *out; 1891 1892 /* 3 bytes becomes 4 chars, but round up and allow for trailing NUL */ 1893 out = (char *)malloc((len * 4 + 2) / 3 + 1); 1894 if (out == NULL) 1895 return (NULL); 1896 d = out; 1897 1898 /* Convert each group of 3 bytes into 4 characters. */ 1899 while (len >= 3) { 1900 v = (((int)s[0] << 16) & 0xff0000) 1901 | (((int)s[1] << 8) & 0xff00) 1902 | (((int)s[2]) & 0x00ff); 1903 s += 3; 1904 len -= 3; 1905 *d++ = digits[(v >> 18) & 0x3f]; 1906 *d++ = digits[(v >> 12) & 0x3f]; 1907 *d++ = digits[(v >> 6) & 0x3f]; 1908 *d++ = digits[(v) & 0x3f]; 1909 } 1910 /* Handle final group of 1 byte (2 chars) or 2 bytes (3 chars). */ 1911 switch (len) { 1912 case 0: break; 1913 case 1: 1914 v = (((int)s[0] << 16) & 0xff0000); 1915 *d++ = digits[(v >> 18) & 0x3f]; 1916 *d++ = digits[(v >> 12) & 0x3f]; 1917 break; 1918 case 2: 1919 v = (((int)s[0] << 16) & 0xff0000) 1920 | (((int)s[1] << 8) & 0xff00); 1921 *d++ = digits[(v >> 18) & 0x3f]; 1922 *d++ = digits[(v >> 12) & 0x3f]; 1923 *d++ = digits[(v >> 6) & 0x3f]; 1924 break; 1925 } 1926 /* Add trailing NUL character so output is a valid C string. */ 1927 *d = '\0'; 1928 return (out); 1929} 1930 1931static void 1932sparse_list_clear(struct pax *pax) 1933{ 1934 while (pax->sparse_list != NULL) { 1935 struct sparse_block *sb = pax->sparse_list; 1936 pax->sparse_list = sb->next; 1937 free(sb); 1938 } 1939 pax->sparse_tail = NULL; 1940} 1941 1942static int 1943_sparse_list_add_block(struct pax *pax, int64_t offset, int64_t length, 1944 int is_hole) 1945{ 1946 struct sparse_block *sb; 1947 1948 sb = (struct sparse_block *)malloc(sizeof(*sb)); 1949 if (sb == NULL) 1950 return (ARCHIVE_FATAL); 1951 sb->next = NULL; 1952 sb->is_hole = is_hole; 1953 sb->offset = offset; 1954 sb->remaining = length; 1955 if (pax->sparse_list == NULL || pax->sparse_tail == NULL) 1956 pax->sparse_list = pax->sparse_tail = sb; 1957 else { 1958 pax->sparse_tail->next = sb; 1959 pax->sparse_tail = sb; 1960 } 1961 return (ARCHIVE_OK); 1962} 1963 1964static int 1965sparse_list_add(struct pax *pax, int64_t offset, int64_t length) 1966{ 1967 int64_t last_offset; 1968 int r; 1969 1970 if (pax->sparse_tail == NULL) 1971 last_offset = 0; 1972 else { 1973 last_offset = pax->sparse_tail->offset + 1974 pax->sparse_tail->remaining; 1975 } 1976 if (last_offset < offset) { 1977 /* Add a hole block. */ 1978 r = _sparse_list_add_block(pax, last_offset, 1979 offset - last_offset, 1); 1980 if (r != ARCHIVE_OK) 1981 return (r); 1982 } 1983 /* Add data block. */ 1984 return (_sparse_list_add_block(pax, offset, length, 0)); 1985} 1986 1987