archive_read_support_format_tar.c revision 306321
1/*- 2 * Copyright (c) 2003-2007 Tim Kientzle 3 * Copyright (c) 2011-2012 Michihiro NAKAJIMA 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR 16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 18 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, 19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 25 */ 26 27#include "archive_platform.h" 28__FBSDID("$FreeBSD: stable/11/contrib/libarchive/libarchive/archive_read_support_format_tar.c 306321 2016-09-25 22:02:27Z mm $"); 29 30#ifdef HAVE_ERRNO_H 31#include <errno.h> 32#endif 33#include <stddef.h> 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_acl_private.h" /* For ACL parsing routines. */ 43#include "archive_entry.h" 44#include "archive_entry_locale.h" 45#include "archive_private.h" 46#include "archive_read_private.h" 47 48#define tar_min(a,b) ((a) < (b) ? (a) : (b)) 49 50/* 51 * Layout of POSIX 'ustar' tar header. 52 */ 53struct archive_entry_header_ustar { 54 char name[100]; 55 char mode[8]; 56 char uid[8]; 57 char gid[8]; 58 char size[12]; 59 char mtime[12]; 60 char checksum[8]; 61 char typeflag[1]; 62 char linkname[100]; /* "old format" header ends here */ 63 char magic[6]; /* For POSIX: "ustar\0" */ 64 char version[2]; /* For POSIX: "00" */ 65 char uname[32]; 66 char gname[32]; 67 char rdevmajor[8]; 68 char rdevminor[8]; 69 char prefix[155]; 70}; 71 72/* 73 * Structure of GNU tar header 74 */ 75struct gnu_sparse { 76 char offset[12]; 77 char numbytes[12]; 78}; 79 80struct archive_entry_header_gnutar { 81 char name[100]; 82 char mode[8]; 83 char uid[8]; 84 char gid[8]; 85 char size[12]; 86 char mtime[12]; 87 char checksum[8]; 88 char typeflag[1]; 89 char linkname[100]; 90 char magic[8]; /* "ustar \0" (note blank/blank/null at end) */ 91 char uname[32]; 92 char gname[32]; 93 char rdevmajor[8]; 94 char rdevminor[8]; 95 char atime[12]; 96 char ctime[12]; 97 char offset[12]; 98 char longnames[4]; 99 char unused[1]; 100 struct gnu_sparse sparse[4]; 101 char isextended[1]; 102 char realsize[12]; 103 /* 104 * Old GNU format doesn't use POSIX 'prefix' field; they use 105 * the 'L' (longname) entry instead. 106 */ 107}; 108 109/* 110 * Data specific to this format. 111 */ 112struct sparse_block { 113 struct sparse_block *next; 114 int64_t offset; 115 int64_t remaining; 116 int hole; 117}; 118 119struct tar { 120 struct archive_string acl_text; 121 struct archive_string entry_pathname; 122 /* For "GNU.sparse.name" and other similar path extensions. */ 123 struct archive_string entry_pathname_override; 124 struct archive_string entry_linkpath; 125 struct archive_string entry_uname; 126 struct archive_string entry_gname; 127 struct archive_string longlink; 128 struct archive_string longname; 129 struct archive_string pax_header; 130 struct archive_string pax_global; 131 struct archive_string line; 132 int pax_hdrcharset_binary; 133 int header_recursion_depth; 134 int64_t entry_bytes_remaining; 135 int64_t entry_offset; 136 int64_t entry_padding; 137 int64_t entry_bytes_unconsumed; 138 int64_t realsize; 139 int sparse_allowed; 140 struct sparse_block *sparse_list; 141 struct sparse_block *sparse_last; 142 int64_t sparse_offset; 143 int64_t sparse_numbytes; 144 int sparse_gnu_major; 145 int sparse_gnu_minor; 146 char sparse_gnu_pending; 147 148 struct archive_string localname; 149 struct archive_string_conv *opt_sconv; 150 struct archive_string_conv *sconv; 151 struct archive_string_conv *sconv_acl; 152 struct archive_string_conv *sconv_default; 153 int init_default_conversion; 154 int compat_2x; 155 int process_mac_extensions; 156 int read_concatenated_archives; 157}; 158 159static int archive_block_is_null(const char *p); 160static char *base64_decode(const char *, size_t, size_t *); 161static int gnu_add_sparse_entry(struct archive_read *, struct tar *, 162 int64_t offset, int64_t remaining); 163 164static void gnu_clear_sparse_list(struct tar *); 165static int gnu_sparse_old_read(struct archive_read *, struct tar *, 166 const struct archive_entry_header_gnutar *header, size_t *); 167static int gnu_sparse_old_parse(struct archive_read *, struct tar *, 168 const struct gnu_sparse *sparse, int length); 169static int gnu_sparse_01_parse(struct archive_read *, struct tar *, 170 const char *); 171static ssize_t gnu_sparse_10_read(struct archive_read *, struct tar *, 172 size_t *); 173static int header_Solaris_ACL(struct archive_read *, struct tar *, 174 struct archive_entry *, const void *, size_t *); 175static int header_common(struct archive_read *, struct tar *, 176 struct archive_entry *, const void *); 177static int header_old_tar(struct archive_read *, struct tar *, 178 struct archive_entry *, const void *); 179static int header_pax_extensions(struct archive_read *, struct tar *, 180 struct archive_entry *, const void *, size_t *); 181static int header_pax_global(struct archive_read *, struct tar *, 182 struct archive_entry *, const void *h, size_t *); 183static int header_longlink(struct archive_read *, struct tar *, 184 struct archive_entry *, const void *h, size_t *); 185static int header_longname(struct archive_read *, struct tar *, 186 struct archive_entry *, const void *h, size_t *); 187static int read_mac_metadata_blob(struct archive_read *, struct tar *, 188 struct archive_entry *, const void *h, size_t *); 189static int header_volume(struct archive_read *, struct tar *, 190 struct archive_entry *, const void *h, size_t *); 191static int header_ustar(struct archive_read *, struct tar *, 192 struct archive_entry *, const void *h); 193static int header_gnutar(struct archive_read *, struct tar *, 194 struct archive_entry *, const void *h, size_t *); 195static int archive_read_format_tar_bid(struct archive_read *, int); 196static int archive_read_format_tar_options(struct archive_read *, 197 const char *, const char *); 198static int archive_read_format_tar_cleanup(struct archive_read *); 199static int archive_read_format_tar_read_data(struct archive_read *a, 200 const void **buff, size_t *size, int64_t *offset); 201static int archive_read_format_tar_skip(struct archive_read *a); 202static int archive_read_format_tar_read_header(struct archive_read *, 203 struct archive_entry *); 204static int checksum(struct archive_read *, const void *); 205static int pax_attribute(struct archive_read *, struct tar *, 206 struct archive_entry *, const char *key, const char *value); 207static int pax_header(struct archive_read *, struct tar *, 208 struct archive_entry *, char *attr); 209static void pax_time(const char *, int64_t *sec, long *nanos); 210static ssize_t readline(struct archive_read *, struct tar *, const char **, 211 ssize_t limit, size_t *); 212static int read_body_to_string(struct archive_read *, struct tar *, 213 struct archive_string *, const void *h, size_t *); 214static int solaris_sparse_parse(struct archive_read *, struct tar *, 215 struct archive_entry *, const char *); 216static int64_t tar_atol(const char *, size_t); 217static int64_t tar_atol10(const char *, size_t); 218static int64_t tar_atol256(const char *, size_t); 219static int64_t tar_atol8(const char *, size_t); 220static int tar_read_header(struct archive_read *, struct tar *, 221 struct archive_entry *, size_t *); 222static int tohex(int c); 223static char *url_decode(const char *); 224static void tar_flush_unconsumed(struct archive_read *, size_t *); 225 226 227int 228archive_read_support_format_gnutar(struct archive *a) 229{ 230 archive_check_magic(a, ARCHIVE_READ_MAGIC, 231 ARCHIVE_STATE_NEW, "archive_read_support_format_gnutar"); 232 return (archive_read_support_format_tar(a)); 233} 234 235 236int 237archive_read_support_format_tar(struct archive *_a) 238{ 239 struct archive_read *a = (struct archive_read *)_a; 240 struct tar *tar; 241 int r; 242 243 archive_check_magic(_a, ARCHIVE_READ_MAGIC, 244 ARCHIVE_STATE_NEW, "archive_read_support_format_tar"); 245 246 tar = (struct tar *)calloc(1, sizeof(*tar)); 247#ifdef HAVE_COPYFILE_H 248 /* Set this by default on Mac OS. */ 249 tar->process_mac_extensions = 1; 250#endif 251 if (tar == NULL) { 252 archive_set_error(&a->archive, ENOMEM, 253 "Can't allocate tar data"); 254 return (ARCHIVE_FATAL); 255 } 256 257 r = __archive_read_register_format(a, tar, "tar", 258 archive_read_format_tar_bid, 259 archive_read_format_tar_options, 260 archive_read_format_tar_read_header, 261 archive_read_format_tar_read_data, 262 archive_read_format_tar_skip, 263 NULL, 264 archive_read_format_tar_cleanup, 265 NULL, 266 NULL); 267 268 if (r != ARCHIVE_OK) 269 free(tar); 270 return (ARCHIVE_OK); 271} 272 273static int 274archive_read_format_tar_cleanup(struct archive_read *a) 275{ 276 struct tar *tar; 277 278 tar = (struct tar *)(a->format->data); 279 gnu_clear_sparse_list(tar); 280 archive_string_free(&tar->acl_text); 281 archive_string_free(&tar->entry_pathname); 282 archive_string_free(&tar->entry_pathname_override); 283 archive_string_free(&tar->entry_linkpath); 284 archive_string_free(&tar->entry_uname); 285 archive_string_free(&tar->entry_gname); 286 archive_string_free(&tar->line); 287 archive_string_free(&tar->pax_global); 288 archive_string_free(&tar->pax_header); 289 archive_string_free(&tar->longname); 290 archive_string_free(&tar->longlink); 291 archive_string_free(&tar->localname); 292 free(tar); 293 (a->format->data) = NULL; 294 return (ARCHIVE_OK); 295} 296 297 298static int 299archive_read_format_tar_bid(struct archive_read *a, int best_bid) 300{ 301 int bid; 302 const char *h; 303 const struct archive_entry_header_ustar *header; 304 305 (void)best_bid; /* UNUSED */ 306 307 bid = 0; 308 309 /* Now let's look at the actual header and see if it matches. */ 310 h = __archive_read_ahead(a, 512, NULL); 311 if (h == NULL) 312 return (-1); 313 314 /* If it's an end-of-archive mark, we can handle it. */ 315 if (h[0] == 0 && archive_block_is_null(h)) { 316 /* 317 * Usually, I bid the number of bits verified, but 318 * in this case, 4096 seems excessive so I picked 10 as 319 * an arbitrary but reasonable-seeming value. 320 */ 321 return (10); 322 } 323 324 /* If it's not an end-of-archive mark, it must have a valid checksum.*/ 325 if (!checksum(a, h)) 326 return (0); 327 bid += 48; /* Checksum is usually 6 octal digits. */ 328 329 header = (const struct archive_entry_header_ustar *)h; 330 331 /* Recognize POSIX formats. */ 332 if ((memcmp(header->magic, "ustar\0", 6) == 0) 333 && (memcmp(header->version, "00", 2) == 0)) 334 bid += 56; 335 336 /* Recognize GNU tar format. */ 337 if ((memcmp(header->magic, "ustar ", 6) == 0) 338 && (memcmp(header->version, " \0", 2) == 0)) 339 bid += 56; 340 341 /* Type flag must be null, digit or A-Z, a-z. */ 342 if (header->typeflag[0] != 0 && 343 !( header->typeflag[0] >= '0' && header->typeflag[0] <= '9') && 344 !( header->typeflag[0] >= 'A' && header->typeflag[0] <= 'Z') && 345 !( header->typeflag[0] >= 'a' && header->typeflag[0] <= 'z') ) 346 return (0); 347 bid += 2; /* 6 bits of variation in an 8-bit field leaves 2 bits. */ 348 349 /* Sanity check: Look at first byte of mode field. */ 350 switch (255 & (unsigned)header->mode[0]) { 351 case 0: case 255: 352 /* Base-256 value: No further verification possible! */ 353 break; 354 case ' ': /* Not recommended, but not illegal, either. */ 355 break; 356 case '0': case '1': case '2': case '3': 357 case '4': case '5': case '6': case '7': 358 /* Octal Value. */ 359 /* TODO: Check format of remainder of this field. */ 360 break; 361 default: 362 /* Not a valid mode; bail out here. */ 363 return (0); 364 } 365 /* TODO: Sanity test uid/gid/size/mtime/rdevmajor/rdevminor fields. */ 366 367 return (bid); 368} 369 370static int 371archive_read_format_tar_options(struct archive_read *a, 372 const char *key, const char *val) 373{ 374 struct tar *tar; 375 int ret = ARCHIVE_FAILED; 376 377 tar = (struct tar *)(a->format->data); 378 if (strcmp(key, "compat-2x") == 0) { 379 /* Handle UTF-8 filnames as libarchive 2.x */ 380 tar->compat_2x = (val != NULL && val[0] != 0); 381 tar->init_default_conversion = tar->compat_2x; 382 return (ARCHIVE_OK); 383 } else if (strcmp(key, "hdrcharset") == 0) { 384 if (val == NULL || val[0] == 0) 385 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 386 "tar: hdrcharset option needs a character-set name"); 387 else { 388 tar->opt_sconv = 389 archive_string_conversion_from_charset( 390 &a->archive, val, 0); 391 if (tar->opt_sconv != NULL) 392 ret = ARCHIVE_OK; 393 else 394 ret = ARCHIVE_FATAL; 395 } 396 return (ret); 397 } else if (strcmp(key, "mac-ext") == 0) { 398 tar->process_mac_extensions = (val != NULL && val[0] != 0); 399 return (ARCHIVE_OK); 400 } else if (strcmp(key, "read_concatenated_archives") == 0) { 401 tar->read_concatenated_archives = (val != NULL && val[0] != 0); 402 return (ARCHIVE_OK); 403 } 404 405 /* Note: The "warn" return is just to inform the options 406 * supervisor that we didn't handle it. It will generate 407 * a suitable error if no one used this option. */ 408 return (ARCHIVE_WARN); 409} 410 411/* utility function- this exists to centralize the logic of tracking 412 * how much unconsumed data we have floating around, and to consume 413 * anything outstanding since we're going to do read_aheads 414 */ 415static void 416tar_flush_unconsumed(struct archive_read *a, size_t *unconsumed) 417{ 418 if (*unconsumed) { 419/* 420 void *data = (void *)__archive_read_ahead(a, *unconsumed, NULL); 421 * this block of code is to poison claimed unconsumed space, ensuring 422 * things break if it is in use still. 423 * currently it WILL break things, so enable it only for debugging this issue 424 if (data) { 425 memset(data, 0xff, *unconsumed); 426 } 427*/ 428 __archive_read_consume(a, *unconsumed); 429 *unconsumed = 0; 430 } 431} 432 433/* 434 * The function invoked by archive_read_next_header(). This 435 * just sets up a few things and then calls the internal 436 * tar_read_header() function below. 437 */ 438static int 439archive_read_format_tar_read_header(struct archive_read *a, 440 struct archive_entry *entry) 441{ 442 /* 443 * When converting tar archives to cpio archives, it is 444 * essential that each distinct file have a distinct inode 445 * number. To simplify this, we keep a static count here to 446 * assign fake dev/inode numbers to each tar entry. Note that 447 * pax format archives may overwrite this with something more 448 * useful. 449 * 450 * Ideally, we would track every file read from the archive so 451 * that we could assign the same dev/ino pair to hardlinks, 452 * but the memory required to store a complete lookup table is 453 * probably not worthwhile just to support the relatively 454 * obscure tar->cpio conversion case. 455 */ 456 static int default_inode; 457 static int default_dev; 458 struct tar *tar; 459 const char *p; 460 const wchar_t *wp; 461 int r; 462 size_t l, unconsumed = 0; 463 464 /* Assign default device/inode values. */ 465 archive_entry_set_dev(entry, 1 + default_dev); /* Don't use zero. */ 466 archive_entry_set_ino(entry, ++default_inode); /* Don't use zero. */ 467 /* Limit generated st_ino number to 16 bits. */ 468 if (default_inode >= 0xffff) { 469 ++default_dev; 470 default_inode = 0; 471 } 472 473 tar = (struct tar *)(a->format->data); 474 tar->entry_offset = 0; 475 gnu_clear_sparse_list(tar); 476 tar->realsize = -1; /* Mark this as "unset" */ 477 478 /* Setup default string conversion. */ 479 tar->sconv = tar->opt_sconv; 480 if (tar->sconv == NULL) { 481 if (!tar->init_default_conversion) { 482 tar->sconv_default = 483 archive_string_default_conversion_for_read(&(a->archive)); 484 tar->init_default_conversion = 1; 485 } 486 tar->sconv = tar->sconv_default; 487 } 488 489 r = tar_read_header(a, tar, entry, &unconsumed); 490 491 tar_flush_unconsumed(a, &unconsumed); 492 493 /* 494 * "non-sparse" files are really just sparse files with 495 * a single block. 496 */ 497 if (tar->sparse_list == NULL) { 498 if (gnu_add_sparse_entry(a, tar, 0, tar->entry_bytes_remaining) 499 != ARCHIVE_OK) 500 return (ARCHIVE_FATAL); 501 } else { 502 struct sparse_block *sb; 503 504 for (sb = tar->sparse_list; sb != NULL; sb = sb->next) { 505 if (!sb->hole) 506 archive_entry_sparse_add_entry(entry, 507 sb->offset, sb->remaining); 508 } 509 } 510 511 if (r == ARCHIVE_OK && archive_entry_filetype(entry) == AE_IFREG) { 512 /* 513 * "Regular" entry with trailing '/' is really 514 * directory: This is needed for certain old tar 515 * variants and even for some broken newer ones. 516 */ 517 if ((wp = archive_entry_pathname_w(entry)) != NULL) { 518 l = wcslen(wp); 519 if (l > 0 && wp[l - 1] == L'/') { 520 archive_entry_set_filetype(entry, AE_IFDIR); 521 } 522 } else if ((p = archive_entry_pathname(entry)) != NULL) { 523 l = strlen(p); 524 if (l > 0 && p[l - 1] == '/') { 525 archive_entry_set_filetype(entry, AE_IFDIR); 526 } 527 } 528 } 529 return (r); 530} 531 532static int 533archive_read_format_tar_read_data(struct archive_read *a, 534 const void **buff, size_t *size, int64_t *offset) 535{ 536 ssize_t bytes_read; 537 struct tar *tar; 538 struct sparse_block *p; 539 540 tar = (struct tar *)(a->format->data); 541 542 for (;;) { 543 /* Remove exhausted entries from sparse list. */ 544 while (tar->sparse_list != NULL && 545 tar->sparse_list->remaining == 0) { 546 p = tar->sparse_list; 547 tar->sparse_list = p->next; 548 free(p); 549 } 550 551 if (tar->entry_bytes_unconsumed) { 552 __archive_read_consume(a, tar->entry_bytes_unconsumed); 553 tar->entry_bytes_unconsumed = 0; 554 } 555 556 /* If we're at end of file, return EOF. */ 557 if (tar->sparse_list == NULL || 558 tar->entry_bytes_remaining == 0) { 559 if (__archive_read_consume(a, tar->entry_padding) < 0) 560 return (ARCHIVE_FATAL); 561 tar->entry_padding = 0; 562 *buff = NULL; 563 *size = 0; 564 *offset = tar->realsize; 565 return (ARCHIVE_EOF); 566 } 567 568 *buff = __archive_read_ahead(a, 1, &bytes_read); 569 if (bytes_read < 0) 570 return (ARCHIVE_FATAL); 571 if (*buff == NULL) { 572 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 573 "Truncated tar archive"); 574 return (ARCHIVE_FATAL); 575 } 576 if (bytes_read > tar->entry_bytes_remaining) 577 bytes_read = (ssize_t)tar->entry_bytes_remaining; 578 /* Don't read more than is available in the 579 * current sparse block. */ 580 if (tar->sparse_list->remaining < bytes_read) 581 bytes_read = (ssize_t)tar->sparse_list->remaining; 582 *size = bytes_read; 583 *offset = tar->sparse_list->offset; 584 tar->sparse_list->remaining -= bytes_read; 585 tar->sparse_list->offset += bytes_read; 586 tar->entry_bytes_remaining -= bytes_read; 587 tar->entry_bytes_unconsumed = bytes_read; 588 589 if (!tar->sparse_list->hole) 590 return (ARCHIVE_OK); 591 /* Current is hole data and skip this. */ 592 } 593} 594 595static int 596archive_read_format_tar_skip(struct archive_read *a) 597{ 598 int64_t bytes_skipped; 599 int64_t request; 600 struct sparse_block *p; 601 struct tar* tar; 602 603 tar = (struct tar *)(a->format->data); 604 605 /* Do not consume the hole of a sparse file. */ 606 request = 0; 607 for (p = tar->sparse_list; p != NULL; p = p->next) { 608 if (!p->hole) { 609 if (p->remaining >= INT64_MAX - request) { 610 return ARCHIVE_FATAL; 611 } 612 request += p->remaining; 613 } 614 } 615 if (request > tar->entry_bytes_remaining) 616 request = tar->entry_bytes_remaining; 617 request += tar->entry_padding + tar->entry_bytes_unconsumed; 618 619 bytes_skipped = __archive_read_consume(a, request); 620 if (bytes_skipped < 0) 621 return (ARCHIVE_FATAL); 622 623 tar->entry_bytes_remaining = 0; 624 tar->entry_bytes_unconsumed = 0; 625 tar->entry_padding = 0; 626 627 /* Free the sparse list. */ 628 gnu_clear_sparse_list(tar); 629 630 return (ARCHIVE_OK); 631} 632 633/* 634 * This function recursively interprets all of the headers associated 635 * with a single entry. 636 */ 637static int 638tar_read_header(struct archive_read *a, struct tar *tar, 639 struct archive_entry *entry, size_t *unconsumed) 640{ 641 ssize_t bytes; 642 int err; 643 const char *h; 644 const struct archive_entry_header_ustar *header; 645 const struct archive_entry_header_gnutar *gnuheader; 646 647 /* Loop until we find a workable header record. */ 648 for (;;) { 649 tar_flush_unconsumed(a, unconsumed); 650 651 /* Read 512-byte header record */ 652 h = __archive_read_ahead(a, 512, &bytes); 653 if (bytes < 0) 654 return ((int)bytes); 655 if (bytes == 0) { /* EOF at a block boundary. */ 656 /* Some writers do omit the block of nulls. <sigh> */ 657 return (ARCHIVE_EOF); 658 } 659 if (bytes < 512) { /* Short block at EOF; this is bad. */ 660 archive_set_error(&a->archive, 661 ARCHIVE_ERRNO_FILE_FORMAT, 662 "Truncated tar archive"); 663 return (ARCHIVE_FATAL); 664 } 665 *unconsumed = 512; 666 667 /* Header is workable if it's not an end-of-archive mark. */ 668 if (h[0] != 0 || !archive_block_is_null(h)) 669 break; 670 671 /* Ensure format is set for archives with only null blocks. */ 672 if (a->archive.archive_format_name == NULL) { 673 a->archive.archive_format = ARCHIVE_FORMAT_TAR; 674 a->archive.archive_format_name = "tar"; 675 } 676 677 if (!tar->read_concatenated_archives) { 678 /* Try to consume a second all-null record, as well. */ 679 tar_flush_unconsumed(a, unconsumed); 680 h = __archive_read_ahead(a, 512, NULL); 681 if (h != NULL && h[0] == 0 && archive_block_is_null(h)) 682 __archive_read_consume(a, 512); 683 archive_clear_error(&a->archive); 684 return (ARCHIVE_EOF); 685 } 686 687 /* 688 * We're reading concatenated archives, ignore this block and 689 * loop to get the next. 690 */ 691 } 692 693 /* 694 * Note: If the checksum fails and we return ARCHIVE_RETRY, 695 * then the client is likely to just retry. This is a very 696 * crude way to search for the next valid header! 697 * 698 * TODO: Improve this by implementing a real header scan. 699 */ 700 if (!checksum(a, h)) { 701 tar_flush_unconsumed(a, unconsumed); 702 archive_set_error(&a->archive, EINVAL, "Damaged tar archive"); 703 return (ARCHIVE_RETRY); /* Retryable: Invalid header */ 704 } 705 706 if (++tar->header_recursion_depth > 32) { 707 tar_flush_unconsumed(a, unconsumed); 708 archive_set_error(&a->archive, EINVAL, "Too many special headers"); 709 return (ARCHIVE_WARN); 710 } 711 712 /* Determine the format variant. */ 713 header = (const struct archive_entry_header_ustar *)h; 714 715 switch(header->typeflag[0]) { 716 case 'A': /* Solaris tar ACL */ 717 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 718 a->archive.archive_format_name = "Solaris tar"; 719 err = header_Solaris_ACL(a, tar, entry, h, unconsumed); 720 break; 721 case 'g': /* POSIX-standard 'g' header. */ 722 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 723 a->archive.archive_format_name = "POSIX pax interchange format"; 724 err = header_pax_global(a, tar, entry, h, unconsumed); 725 if (err == ARCHIVE_EOF) 726 return (err); 727 break; 728 case 'K': /* Long link name (GNU tar, others) */ 729 err = header_longlink(a, tar, entry, h, unconsumed); 730 break; 731 case 'L': /* Long filename (GNU tar, others) */ 732 err = header_longname(a, tar, entry, h, unconsumed); 733 break; 734 case 'V': /* GNU volume header */ 735 err = header_volume(a, tar, entry, h, unconsumed); 736 break; 737 case 'X': /* Used by SUN tar; same as 'x'. */ 738 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 739 a->archive.archive_format_name = 740 "POSIX pax interchange format (Sun variant)"; 741 err = header_pax_extensions(a, tar, entry, h, unconsumed); 742 break; 743 case 'x': /* POSIX-standard 'x' header. */ 744 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 745 a->archive.archive_format_name = "POSIX pax interchange format"; 746 err = header_pax_extensions(a, tar, entry, h, unconsumed); 747 break; 748 default: 749 gnuheader = (const struct archive_entry_header_gnutar *)h; 750 if (memcmp(gnuheader->magic, "ustar \0", 8) == 0) { 751 a->archive.archive_format = ARCHIVE_FORMAT_TAR_GNUTAR; 752 a->archive.archive_format_name = "GNU tar format"; 753 err = header_gnutar(a, tar, entry, h, unconsumed); 754 } else if (memcmp(header->magic, "ustar", 5) == 0) { 755 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) { 756 a->archive.archive_format = ARCHIVE_FORMAT_TAR_USTAR; 757 a->archive.archive_format_name = "POSIX ustar format"; 758 } 759 err = header_ustar(a, tar, entry, h); 760 } else { 761 a->archive.archive_format = ARCHIVE_FORMAT_TAR; 762 a->archive.archive_format_name = "tar (non-POSIX)"; 763 err = header_old_tar(a, tar, entry, h); 764 } 765 } 766 if (err == ARCHIVE_FATAL) 767 return (err); 768 769 tar_flush_unconsumed(a, unconsumed); 770 771 h = NULL; 772 header = NULL; 773 774 --tar->header_recursion_depth; 775 /* Yuck. Apple's design here ends up storing long pathname 776 * extensions for both the AppleDouble extension entry and the 777 * regular entry. 778 */ 779 if ((err == ARCHIVE_WARN || err == ARCHIVE_OK) && 780 tar->header_recursion_depth == 0 && 781 tar->process_mac_extensions) { 782 int err2 = read_mac_metadata_blob(a, tar, entry, h, unconsumed); 783 if (err2 < err) 784 err = err2; 785 } 786 787 /* We return warnings or success as-is. Anything else is fatal. */ 788 if (err == ARCHIVE_WARN || err == ARCHIVE_OK) { 789 if (tar->sparse_gnu_pending) { 790 if (tar->sparse_gnu_major == 1 && 791 tar->sparse_gnu_minor == 0) { 792 ssize_t bytes_read; 793 794 tar->sparse_gnu_pending = 0; 795 /* Read initial sparse map. */ 796 bytes_read = gnu_sparse_10_read(a, tar, unconsumed); 797 tar->entry_bytes_remaining -= bytes_read; 798 if (bytes_read < 0) 799 return ((int)bytes_read); 800 } else { 801 archive_set_error(&a->archive, 802 ARCHIVE_ERRNO_MISC, 803 "Unrecognized GNU sparse file format"); 804 return (ARCHIVE_WARN); 805 } 806 tar->sparse_gnu_pending = 0; 807 } 808 return (err); 809 } 810 if (err == ARCHIVE_EOF) 811 /* EOF when recursively reading a header is bad. */ 812 archive_set_error(&a->archive, EINVAL, "Damaged tar archive"); 813 return (ARCHIVE_FATAL); 814} 815 816/* 817 * Return true if block checksum is correct. 818 */ 819static int 820checksum(struct archive_read *a, const void *h) 821{ 822 const unsigned char *bytes; 823 const struct archive_entry_header_ustar *header; 824 int check, sum; 825 size_t i; 826 827 (void)a; /* UNUSED */ 828 bytes = (const unsigned char *)h; 829 header = (const struct archive_entry_header_ustar *)h; 830 831 /* Checksum field must hold an octal number */ 832 for (i = 0; i < sizeof(header->checksum); ++i) { 833 char c = header->checksum[i]; 834 if (c != ' ' && c != '\0' && (c < '0' || c > '7')) 835 return 0; 836 } 837 838 /* 839 * Test the checksum. Note that POSIX specifies _unsigned_ 840 * bytes for this calculation. 841 */ 842 sum = (int)tar_atol(header->checksum, sizeof(header->checksum)); 843 check = 0; 844 for (i = 0; i < 148; i++) 845 check += (unsigned char)bytes[i]; 846 for (; i < 156; i++) 847 check += 32; 848 for (; i < 512; i++) 849 check += (unsigned char)bytes[i]; 850 if (sum == check) 851 return (1); 852 853 /* 854 * Repeat test with _signed_ bytes, just in case this archive 855 * was created by an old BSD, Solaris, or HP-UX tar with a 856 * broken checksum calculation. 857 */ 858 check = 0; 859 for (i = 0; i < 148; i++) 860 check += (signed char)bytes[i]; 861 for (; i < 156; i++) 862 check += 32; 863 for (; i < 512; i++) 864 check += (signed char)bytes[i]; 865 if (sum == check) 866 return (1); 867 868 return (0); 869} 870 871/* 872 * Return true if this block contains only nulls. 873 */ 874static int 875archive_block_is_null(const char *p) 876{ 877 unsigned i; 878 879 for (i = 0; i < 512; i++) 880 if (*p++) 881 return (0); 882 return (1); 883} 884 885/* 886 * Interpret 'A' Solaris ACL header 887 */ 888static int 889header_Solaris_ACL(struct archive_read *a, struct tar *tar, 890 struct archive_entry *entry, const void *h, size_t *unconsumed) 891{ 892 const struct archive_entry_header_ustar *header; 893 size_t size; 894 int err; 895 int64_t type; 896 char *acl, *p; 897 898 /* 899 * read_body_to_string adds a NUL terminator, but we need a little 900 * more to make sure that we don't overrun acl_text later. 901 */ 902 header = (const struct archive_entry_header_ustar *)h; 903 size = (size_t)tar_atol(header->size, sizeof(header->size)); 904 err = read_body_to_string(a, tar, &(tar->acl_text), h, unconsumed); 905 if (err != ARCHIVE_OK) 906 return (err); 907 908 /* Recursively read next header */ 909 err = tar_read_header(a, tar, entry, unconsumed); 910 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 911 return (err); 912 913 /* TODO: Examine the first characters to see if this 914 * is an AIX ACL descriptor. We'll likely never support 915 * them, but it would be polite to recognize and warn when 916 * we do see them. */ 917 918 /* Leading octal number indicates ACL type and number of entries. */ 919 p = acl = tar->acl_text.s; 920 type = 0; 921 while (*p != '\0' && p < acl + size) { 922 if (*p < '0' || *p > '7') { 923 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 924 "Malformed Solaris ACL attribute (invalid digit)"); 925 return(ARCHIVE_WARN); 926 } 927 type <<= 3; 928 type += *p - '0'; 929 if (type > 077777777) { 930 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 931 "Malformed Solaris ACL attribute (count too large)"); 932 return (ARCHIVE_WARN); 933 } 934 p++; 935 } 936 switch ((int)type & ~0777777) { 937 case 01000000: 938 /* POSIX.1e ACL */ 939 break; 940 case 03000000: 941 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 942 "Solaris NFSv4 ACLs not supported"); 943 return (ARCHIVE_WARN); 944 default: 945 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 946 "Malformed Solaris ACL attribute (unsupported type %o)", 947 (int)type); 948 return (ARCHIVE_WARN); 949 } 950 p++; 951 952 if (p >= acl + size) { 953 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 954 "Malformed Solaris ACL attribute (body overflow)"); 955 return(ARCHIVE_WARN); 956 } 957 958 /* ACL text is null-terminated; find the end. */ 959 size -= (p - acl); 960 acl = p; 961 962 while (*p != '\0' && p < acl + size) 963 p++; 964 965 if (tar->sconv_acl == NULL) { 966 tar->sconv_acl = archive_string_conversion_from_charset( 967 &(a->archive), "UTF-8", 1); 968 if (tar->sconv_acl == NULL) 969 return (ARCHIVE_FATAL); 970 } 971 archive_strncpy(&(tar->localname), acl, p - acl); 972 err = archive_acl_parse_l(archive_entry_acl(entry), 973 tar->localname.s, ARCHIVE_ENTRY_ACL_TYPE_ACCESS, tar->sconv_acl); 974 if (err != ARCHIVE_OK) { 975 if (errno == ENOMEM) { 976 archive_set_error(&a->archive, ENOMEM, 977 "Can't allocate memory for ACL"); 978 } else 979 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 980 "Malformed Solaris ACL attribute (unparsable)"); 981 } 982 return (err); 983} 984 985/* 986 * Interpret 'K' long linkname header. 987 */ 988static int 989header_longlink(struct archive_read *a, struct tar *tar, 990 struct archive_entry *entry, const void *h, size_t *unconsumed) 991{ 992 int err; 993 994 err = read_body_to_string(a, tar, &(tar->longlink), h, unconsumed); 995 if (err != ARCHIVE_OK) 996 return (err); 997 err = tar_read_header(a, tar, entry, unconsumed); 998 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 999 return (err); 1000 /* Set symlink if symlink already set, else hardlink. */ 1001 archive_entry_copy_link(entry, tar->longlink.s); 1002 return (ARCHIVE_OK); 1003} 1004 1005static int 1006set_conversion_failed_error(struct archive_read *a, 1007 struct archive_string_conv *sconv, const char *name) 1008{ 1009 if (errno == ENOMEM) { 1010 archive_set_error(&a->archive, ENOMEM, 1011 "Can't allocate memory for %s", name); 1012 return (ARCHIVE_FATAL); 1013 } 1014 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 1015 "%s can't be converted from %s to current locale.", 1016 name, archive_string_conversion_charset_name(sconv)); 1017 return (ARCHIVE_WARN); 1018} 1019 1020/* 1021 * Interpret 'L' long filename header. 1022 */ 1023static int 1024header_longname(struct archive_read *a, struct tar *tar, 1025 struct archive_entry *entry, const void *h, size_t *unconsumed) 1026{ 1027 int err; 1028 1029 err = read_body_to_string(a, tar, &(tar->longname), h, unconsumed); 1030 if (err != ARCHIVE_OK) 1031 return (err); 1032 /* Read and parse "real" header, then override name. */ 1033 err = tar_read_header(a, tar, entry, unconsumed); 1034 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 1035 return (err); 1036 if (archive_entry_copy_pathname_l(entry, tar->longname.s, 1037 archive_strlen(&(tar->longname)), tar->sconv) != 0) 1038 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 1039 return (err); 1040} 1041 1042 1043/* 1044 * Interpret 'V' GNU tar volume header. 1045 */ 1046static int 1047header_volume(struct archive_read *a, struct tar *tar, 1048 struct archive_entry *entry, const void *h, size_t *unconsumed) 1049{ 1050 (void)h; 1051 1052 /* Just skip this and read the next header. */ 1053 return (tar_read_header(a, tar, entry, unconsumed)); 1054} 1055 1056/* 1057 * Read body of an archive entry into an archive_string object. 1058 */ 1059static int 1060read_body_to_string(struct archive_read *a, struct tar *tar, 1061 struct archive_string *as, const void *h, size_t *unconsumed) 1062{ 1063 int64_t size; 1064 const struct archive_entry_header_ustar *header; 1065 const void *src; 1066 1067 (void)tar; /* UNUSED */ 1068 header = (const struct archive_entry_header_ustar *)h; 1069 size = tar_atol(header->size, sizeof(header->size)); 1070 if ((size > 1048576) || (size < 0)) { 1071 archive_set_error(&a->archive, EINVAL, 1072 "Special header too large"); 1073 return (ARCHIVE_FATAL); 1074 } 1075 1076 /* Fail if we can't make our buffer big enough. */ 1077 if (archive_string_ensure(as, (size_t)size+1) == NULL) { 1078 archive_set_error(&a->archive, ENOMEM, 1079 "No memory"); 1080 return (ARCHIVE_FATAL); 1081 } 1082 1083 tar_flush_unconsumed(a, unconsumed); 1084 1085 /* Read the body into the string. */ 1086 *unconsumed = (size_t)((size + 511) & ~ 511); 1087 src = __archive_read_ahead(a, *unconsumed, NULL); 1088 if (src == NULL) { 1089 *unconsumed = 0; 1090 return (ARCHIVE_FATAL); 1091 } 1092 memcpy(as->s, src, (size_t)size); 1093 as->s[size] = '\0'; 1094 as->length = (size_t)size; 1095 return (ARCHIVE_OK); 1096} 1097 1098/* 1099 * Parse out common header elements. 1100 * 1101 * This would be the same as header_old_tar, except that the 1102 * filename is handled slightly differently for old and POSIX 1103 * entries (POSIX entries support a 'prefix'). This factoring 1104 * allows header_old_tar and header_ustar 1105 * to handle filenames differently, while still putting most of the 1106 * common parsing into one place. 1107 */ 1108static int 1109header_common(struct archive_read *a, struct tar *tar, 1110 struct archive_entry *entry, const void *h) 1111{ 1112 const struct archive_entry_header_ustar *header; 1113 char tartype; 1114 int err = ARCHIVE_OK; 1115 1116 header = (const struct archive_entry_header_ustar *)h; 1117 if (header->linkname[0]) 1118 archive_strncpy(&(tar->entry_linkpath), 1119 header->linkname, sizeof(header->linkname)); 1120 else 1121 archive_string_empty(&(tar->entry_linkpath)); 1122 1123 /* Parse out the numeric fields (all are octal) */ 1124 archive_entry_set_mode(entry, 1125 (mode_t)tar_atol(header->mode, sizeof(header->mode))); 1126 archive_entry_set_uid(entry, tar_atol(header->uid, sizeof(header->uid))); 1127 archive_entry_set_gid(entry, tar_atol(header->gid, sizeof(header->gid))); 1128 tar->entry_bytes_remaining = tar_atol(header->size, sizeof(header->size)); 1129 if (tar->entry_bytes_remaining < 0) { 1130 tar->entry_bytes_remaining = 0; 1131 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1132 "Tar entry has negative size"); 1133 return (ARCHIVE_FATAL); 1134 } 1135 if (tar->entry_bytes_remaining == INT64_MAX) { 1136 /* Note: tar_atol returns INT64_MAX on overflow */ 1137 tar->entry_bytes_remaining = 0; 1138 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1139 "Tar entry size overflow"); 1140 return (ARCHIVE_FATAL); 1141 } 1142 tar->realsize = tar->entry_bytes_remaining; 1143 archive_entry_set_size(entry, tar->entry_bytes_remaining); 1144 archive_entry_set_mtime(entry, tar_atol(header->mtime, sizeof(header->mtime)), 0); 1145 1146 /* Handle the tar type flag appropriately. */ 1147 tartype = header->typeflag[0]; 1148 1149 switch (tartype) { 1150 case '1': /* Hard link */ 1151 if (archive_entry_copy_hardlink_l(entry, tar->entry_linkpath.s, 1152 archive_strlen(&(tar->entry_linkpath)), tar->sconv) != 0) { 1153 err = set_conversion_failed_error(a, tar->sconv, 1154 "Linkname"); 1155 if (err == ARCHIVE_FATAL) 1156 return (err); 1157 } 1158 /* 1159 * The following may seem odd, but: Technically, tar 1160 * does not store the file type for a "hard link" 1161 * entry, only the fact that it is a hard link. So, I 1162 * leave the type zero normally. But, pax interchange 1163 * format allows hard links to have data, which 1164 * implies that the underlying entry is a regular 1165 * file. 1166 */ 1167 if (archive_entry_size(entry) > 0) 1168 archive_entry_set_filetype(entry, AE_IFREG); 1169 1170 /* 1171 * A tricky point: Traditionally, tar readers have 1172 * ignored the size field when reading hardlink 1173 * entries, and some writers put non-zero sizes even 1174 * though the body is empty. POSIX blessed this 1175 * convention in the 1988 standard, but broke with 1176 * this tradition in 2001 by permitting hardlink 1177 * entries to store valid bodies in pax interchange 1178 * format, but not in ustar format. Since there is no 1179 * hard and fast way to distinguish pax interchange 1180 * from earlier archives (the 'x' and 'g' entries are 1181 * optional, after all), we need a heuristic. 1182 */ 1183 if (archive_entry_size(entry) == 0) { 1184 /* If the size is already zero, we're done. */ 1185 } else if (a->archive.archive_format 1186 == ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) { 1187 /* Definitely pax extended; must obey hardlink size. */ 1188 } else if (a->archive.archive_format == ARCHIVE_FORMAT_TAR 1189 || a->archive.archive_format == ARCHIVE_FORMAT_TAR_GNUTAR) 1190 { 1191 /* Old-style or GNU tar: we must ignore the size. */ 1192 archive_entry_set_size(entry, 0); 1193 tar->entry_bytes_remaining = 0; 1194 } else if (archive_read_format_tar_bid(a, 50) > 50) { 1195 /* 1196 * We don't know if it's pax: If the bid 1197 * function sees a valid ustar header 1198 * immediately following, then let's ignore 1199 * the hardlink size. 1200 */ 1201 archive_entry_set_size(entry, 0); 1202 tar->entry_bytes_remaining = 0; 1203 } 1204 /* 1205 * TODO: There are still two cases I'd like to handle: 1206 * = a ustar non-pax archive with a hardlink entry at 1207 * end-of-archive. (Look for block of nulls following?) 1208 * = a pax archive that has not seen any pax headers 1209 * and has an entry which is a hardlink entry storing 1210 * a body containing an uncompressed tar archive. 1211 * The first is worth addressing; I don't see any reliable 1212 * way to deal with the second possibility. 1213 */ 1214 break; 1215 case '2': /* Symlink */ 1216 archive_entry_set_filetype(entry, AE_IFLNK); 1217 archive_entry_set_size(entry, 0); 1218 tar->entry_bytes_remaining = 0; 1219 if (archive_entry_copy_symlink_l(entry, tar->entry_linkpath.s, 1220 archive_strlen(&(tar->entry_linkpath)), tar->sconv) != 0) { 1221 err = set_conversion_failed_error(a, tar->sconv, 1222 "Linkname"); 1223 if (err == ARCHIVE_FATAL) 1224 return (err); 1225 } 1226 break; 1227 case '3': /* Character device */ 1228 archive_entry_set_filetype(entry, AE_IFCHR); 1229 archive_entry_set_size(entry, 0); 1230 tar->entry_bytes_remaining = 0; 1231 break; 1232 case '4': /* Block device */ 1233 archive_entry_set_filetype(entry, AE_IFBLK); 1234 archive_entry_set_size(entry, 0); 1235 tar->entry_bytes_remaining = 0; 1236 break; 1237 case '5': /* Dir */ 1238 archive_entry_set_filetype(entry, AE_IFDIR); 1239 archive_entry_set_size(entry, 0); 1240 tar->entry_bytes_remaining = 0; 1241 break; 1242 case '6': /* FIFO device */ 1243 archive_entry_set_filetype(entry, AE_IFIFO); 1244 archive_entry_set_size(entry, 0); 1245 tar->entry_bytes_remaining = 0; 1246 break; 1247 case 'D': /* GNU incremental directory type */ 1248 /* 1249 * No special handling is actually required here. 1250 * It might be nice someday to preprocess the file list and 1251 * provide it to the client, though. 1252 */ 1253 archive_entry_set_filetype(entry, AE_IFDIR); 1254 break; 1255 case 'M': /* GNU "Multi-volume" (remainder of file from last archive)*/ 1256 /* 1257 * As far as I can tell, this is just like a regular file 1258 * entry, except that the contents should be _appended_ to 1259 * the indicated file at the indicated offset. This may 1260 * require some API work to fully support. 1261 */ 1262 break; 1263 case 'N': /* Old GNU "long filename" entry. */ 1264 /* The body of this entry is a script for renaming 1265 * previously-extracted entries. Ugh. It will never 1266 * be supported by libarchive. */ 1267 archive_entry_set_filetype(entry, AE_IFREG); 1268 break; 1269 case 'S': /* GNU sparse files */ 1270 /* 1271 * Sparse files are really just regular files with 1272 * sparse information in the extended area. 1273 */ 1274 /* FALLTHROUGH */ 1275 case '0': 1276 /* 1277 * Enable sparse file "read" support only for regular 1278 * files and explicit GNU sparse files. However, we 1279 * don't allow non-standard file types to be sparse. 1280 */ 1281 tar->sparse_allowed = 1; 1282 /* FALLTHROUGH */ 1283 default: /* Regular file and non-standard types */ 1284 /* 1285 * Per POSIX: non-recognized types should always be 1286 * treated as regular files. 1287 */ 1288 archive_entry_set_filetype(entry, AE_IFREG); 1289 break; 1290 } 1291 return (err); 1292} 1293 1294/* 1295 * Parse out header elements for "old-style" tar archives. 1296 */ 1297static int 1298header_old_tar(struct archive_read *a, struct tar *tar, 1299 struct archive_entry *entry, const void *h) 1300{ 1301 const struct archive_entry_header_ustar *header; 1302 int err = ARCHIVE_OK, err2; 1303 1304 /* Copy filename over (to ensure null termination). */ 1305 header = (const struct archive_entry_header_ustar *)h; 1306 if (archive_entry_copy_pathname_l(entry, 1307 header->name, sizeof(header->name), tar->sconv) != 0) { 1308 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 1309 if (err == ARCHIVE_FATAL) 1310 return (err); 1311 } 1312 1313 /* Grab rest of common fields */ 1314 err2 = header_common(a, tar, entry, h); 1315 if (err > err2) 1316 err = err2; 1317 1318 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 1319 return (err); 1320} 1321 1322/* 1323 * Read a Mac AppleDouble-encoded blob of file metadata, 1324 * if there is one. 1325 */ 1326static int 1327read_mac_metadata_blob(struct archive_read *a, struct tar *tar, 1328 struct archive_entry *entry, const void *h, size_t *unconsumed) 1329{ 1330 int64_t size; 1331 const void *data; 1332 const char *p, *name; 1333 const wchar_t *wp, *wname; 1334 1335 (void)h; /* UNUSED */ 1336 1337 wname = wp = archive_entry_pathname_w(entry); 1338 if (wp != NULL) { 1339 /* Find the last path element. */ 1340 for (; *wp != L'\0'; ++wp) { 1341 if (wp[0] == '/' && wp[1] != L'\0') 1342 wname = wp + 1; 1343 } 1344 /* 1345 * If last path element starts with "._", then 1346 * this is a Mac extension. 1347 */ 1348 if (wname[0] != L'.' || wname[1] != L'_' || wname[2] == L'\0') 1349 return ARCHIVE_OK; 1350 } else { 1351 /* Find the last path element. */ 1352 name = p = archive_entry_pathname(entry); 1353 if (p == NULL) 1354 return (ARCHIVE_FAILED); 1355 for (; *p != '\0'; ++p) { 1356 if (p[0] == '/' && p[1] != '\0') 1357 name = p + 1; 1358 } 1359 /* 1360 * If last path element starts with "._", then 1361 * this is a Mac extension. 1362 */ 1363 if (name[0] != '.' || name[1] != '_' || name[2] == '\0') 1364 return ARCHIVE_OK; 1365 } 1366 1367 /* Read the body as a Mac OS metadata blob. */ 1368 size = archive_entry_size(entry); 1369 1370 /* 1371 * TODO: Look beyond the body here to peek at the next header. 1372 * If it's a regular header (not an extension header) 1373 * that has the wrong name, just return the current 1374 * entry as-is, without consuming the body here. 1375 * That would reduce the risk of us mis-identifying 1376 * an ordinary file that just happened to have 1377 * a name starting with "._". 1378 * 1379 * Q: Is the above idea really possible? Even 1380 * when there are GNU or pax extension entries? 1381 */ 1382 data = __archive_read_ahead(a, (size_t)size, NULL); 1383 if (data == NULL) { 1384 *unconsumed = 0; 1385 return (ARCHIVE_FATAL); 1386 } 1387 archive_entry_copy_mac_metadata(entry, data, (size_t)size); 1388 *unconsumed = (size_t)((size + 511) & ~ 511); 1389 tar_flush_unconsumed(a, unconsumed); 1390 return (tar_read_header(a, tar, entry, unconsumed)); 1391} 1392 1393/* 1394 * Parse a file header for a pax extended archive entry. 1395 */ 1396static int 1397header_pax_global(struct archive_read *a, struct tar *tar, 1398 struct archive_entry *entry, const void *h, size_t *unconsumed) 1399{ 1400 int err; 1401 1402 err = read_body_to_string(a, tar, &(tar->pax_global), h, unconsumed); 1403 if (err != ARCHIVE_OK) 1404 return (err); 1405 err = tar_read_header(a, tar, entry, unconsumed); 1406 return (err); 1407} 1408 1409static int 1410header_pax_extensions(struct archive_read *a, struct tar *tar, 1411 struct archive_entry *entry, const void *h, size_t *unconsumed) 1412{ 1413 int err, err2; 1414 1415 err = read_body_to_string(a, tar, &(tar->pax_header), h, unconsumed); 1416 if (err != ARCHIVE_OK) 1417 return (err); 1418 1419 /* Parse the next header. */ 1420 err = tar_read_header(a, tar, entry, unconsumed); 1421 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 1422 return (err); 1423 1424 /* 1425 * TODO: Parse global/default options into 'entry' struct here 1426 * before handling file-specific options. 1427 * 1428 * This design (parse standard header, then overwrite with pax 1429 * extended attribute data) usually works well, but isn't ideal; 1430 * it would be better to parse the pax extended attributes first 1431 * and then skip any fields in the standard header that were 1432 * defined in the pax header. 1433 */ 1434 err2 = pax_header(a, tar, entry, tar->pax_header.s); 1435 err = err_combine(err, err2); 1436 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 1437 return (err); 1438} 1439 1440 1441/* 1442 * Parse a file header for a Posix "ustar" archive entry. This also 1443 * handles "pax" or "extended ustar" entries. 1444 */ 1445static int 1446header_ustar(struct archive_read *a, struct tar *tar, 1447 struct archive_entry *entry, const void *h) 1448{ 1449 const struct archive_entry_header_ustar *header; 1450 struct archive_string *as; 1451 int err = ARCHIVE_OK, r; 1452 1453 header = (const struct archive_entry_header_ustar *)h; 1454 1455 /* Copy name into an internal buffer to ensure null-termination. */ 1456 as = &(tar->entry_pathname); 1457 if (header->prefix[0]) { 1458 archive_strncpy(as, header->prefix, sizeof(header->prefix)); 1459 if (as->s[archive_strlen(as) - 1] != '/') 1460 archive_strappend_char(as, '/'); 1461 archive_strncat(as, header->name, sizeof(header->name)); 1462 } else { 1463 archive_strncpy(as, header->name, sizeof(header->name)); 1464 } 1465 if (archive_entry_copy_pathname_l(entry, as->s, archive_strlen(as), 1466 tar->sconv) != 0) { 1467 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 1468 if (err == ARCHIVE_FATAL) 1469 return (err); 1470 } 1471 1472 /* Handle rest of common fields. */ 1473 r = header_common(a, tar, entry, h); 1474 if (r == ARCHIVE_FATAL) 1475 return (r); 1476 if (r < err) 1477 err = r; 1478 1479 /* Handle POSIX ustar fields. */ 1480 if (archive_entry_copy_uname_l(entry, 1481 header->uname, sizeof(header->uname), tar->sconv) != 0) { 1482 err = set_conversion_failed_error(a, tar->sconv, "Uname"); 1483 if (err == ARCHIVE_FATAL) 1484 return (err); 1485 } 1486 1487 if (archive_entry_copy_gname_l(entry, 1488 header->gname, sizeof(header->gname), tar->sconv) != 0) { 1489 err = set_conversion_failed_error(a, tar->sconv, "Gname"); 1490 if (err == ARCHIVE_FATAL) 1491 return (err); 1492 } 1493 1494 /* Parse out device numbers only for char and block specials. */ 1495 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') { 1496 archive_entry_set_rdevmajor(entry, (dev_t) 1497 tar_atol(header->rdevmajor, sizeof(header->rdevmajor))); 1498 archive_entry_set_rdevminor(entry, (dev_t) 1499 tar_atol(header->rdevminor, sizeof(header->rdevminor))); 1500 } 1501 1502 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 1503 1504 return (err); 1505} 1506 1507 1508/* 1509 * Parse the pax extended attributes record. 1510 * 1511 * Returns non-zero if there's an error in the data. 1512 */ 1513static int 1514pax_header(struct archive_read *a, struct tar *tar, 1515 struct archive_entry *entry, char *attr) 1516{ 1517 size_t attr_length, l, line_length; 1518 char *p; 1519 char *key, *value; 1520 struct archive_string *as; 1521 struct archive_string_conv *sconv; 1522 int err, err2; 1523 1524 attr_length = strlen(attr); 1525 tar->pax_hdrcharset_binary = 0; 1526 archive_string_empty(&(tar->entry_gname)); 1527 archive_string_empty(&(tar->entry_linkpath)); 1528 archive_string_empty(&(tar->entry_pathname)); 1529 archive_string_empty(&(tar->entry_pathname_override)); 1530 archive_string_empty(&(tar->entry_uname)); 1531 err = ARCHIVE_OK; 1532 while (attr_length > 0) { 1533 /* Parse decimal length field at start of line. */ 1534 line_length = 0; 1535 l = attr_length; 1536 p = attr; /* Record start of line. */ 1537 while (l>0) { 1538 if (*p == ' ') { 1539 p++; 1540 l--; 1541 break; 1542 } 1543 if (*p < '0' || *p > '9') { 1544 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1545 "Ignoring malformed pax extended attributes"); 1546 return (ARCHIVE_WARN); 1547 } 1548 line_length *= 10; 1549 line_length += *p - '0'; 1550 if (line_length > 999999) { 1551 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1552 "Rejecting pax extended attribute > 1MB"); 1553 return (ARCHIVE_WARN); 1554 } 1555 p++; 1556 l--; 1557 } 1558 1559 /* 1560 * Parsed length must be no bigger than available data, 1561 * at least 1, and the last character of the line must 1562 * be '\n'. 1563 */ 1564 if (line_length > attr_length 1565 || line_length < 1 1566 || attr[line_length - 1] != '\n') 1567 { 1568 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1569 "Ignoring malformed pax extended attribute"); 1570 return (ARCHIVE_WARN); 1571 } 1572 1573 /* Null-terminate the line. */ 1574 attr[line_length - 1] = '\0'; 1575 1576 /* Find end of key and null terminate it. */ 1577 key = p; 1578 if (key[0] == '=') 1579 return (-1); 1580 while (*p && *p != '=') 1581 ++p; 1582 if (*p == '\0') { 1583 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1584 "Invalid pax extended attributes"); 1585 return (ARCHIVE_WARN); 1586 } 1587 *p = '\0'; 1588 1589 /* Identify null-terminated 'value' portion. */ 1590 value = p + 1; 1591 1592 /* Identify this attribute and set it in the entry. */ 1593 err2 = pax_attribute(a, tar, entry, key, value); 1594 if (err2 == ARCHIVE_FATAL) 1595 return (err2); 1596 err = err_combine(err, err2); 1597 1598 /* Skip to next line */ 1599 attr += line_length; 1600 attr_length -= line_length; 1601 } 1602 1603 /* 1604 * PAX format uses UTF-8 as default charset for its metadata 1605 * unless hdrcharset=BINARY is present in its header. 1606 * We apply the charset specified by the hdrcharset option only 1607 * when the hdrcharset attribute(in PAX header) is BINARY because 1608 * we respect the charset described in PAX header and BINARY also 1609 * means that metadata(filename,uname and gname) character-set 1610 * is unknown. 1611 */ 1612 if (tar->pax_hdrcharset_binary) 1613 sconv = tar->opt_sconv; 1614 else { 1615 sconv = archive_string_conversion_from_charset( 1616 &(a->archive), "UTF-8", 1); 1617 if (sconv == NULL) 1618 return (ARCHIVE_FATAL); 1619 if (tar->compat_2x) 1620 archive_string_conversion_set_opt(sconv, 1621 SCONV_SET_OPT_UTF8_LIBARCHIVE2X); 1622 } 1623 1624 if (archive_strlen(&(tar->entry_gname)) > 0) { 1625 if (archive_entry_copy_gname_l(entry, tar->entry_gname.s, 1626 archive_strlen(&(tar->entry_gname)), sconv) != 0) { 1627 err = set_conversion_failed_error(a, sconv, "Gname"); 1628 if (err == ARCHIVE_FATAL) 1629 return (err); 1630 /* Use a converted an original name. */ 1631 archive_entry_copy_gname(entry, tar->entry_gname.s); 1632 } 1633 } 1634 if (archive_strlen(&(tar->entry_linkpath)) > 0) { 1635 if (archive_entry_copy_link_l(entry, tar->entry_linkpath.s, 1636 archive_strlen(&(tar->entry_linkpath)), sconv) != 0) { 1637 err = set_conversion_failed_error(a, sconv, "Linkname"); 1638 if (err == ARCHIVE_FATAL) 1639 return (err); 1640 /* Use a converted an original name. */ 1641 archive_entry_copy_link(entry, tar->entry_linkpath.s); 1642 } 1643 } 1644 /* 1645 * Some extensions (such as the GNU sparse file extensions) 1646 * deliberately store a synthetic name under the regular 'path' 1647 * attribute and the real file name under a different attribute. 1648 * Since we're supposed to not care about the order, we 1649 * have no choice but to store all of the various filenames 1650 * we find and figure it all out afterwards. This is the 1651 * figuring out part. 1652 */ 1653 as = NULL; 1654 if (archive_strlen(&(tar->entry_pathname_override)) > 0) 1655 as = &(tar->entry_pathname_override); 1656 else if (archive_strlen(&(tar->entry_pathname)) > 0) 1657 as = &(tar->entry_pathname); 1658 if (as != NULL) { 1659 if (archive_entry_copy_pathname_l(entry, as->s, 1660 archive_strlen(as), sconv) != 0) { 1661 err = set_conversion_failed_error(a, sconv, "Pathname"); 1662 if (err == ARCHIVE_FATAL) 1663 return (err); 1664 /* Use a converted an original name. */ 1665 archive_entry_copy_pathname(entry, as->s); 1666 } 1667 } 1668 if (archive_strlen(&(tar->entry_uname)) > 0) { 1669 if (archive_entry_copy_uname_l(entry, tar->entry_uname.s, 1670 archive_strlen(&(tar->entry_uname)), sconv) != 0) { 1671 err = set_conversion_failed_error(a, sconv, "Uname"); 1672 if (err == ARCHIVE_FATAL) 1673 return (err); 1674 /* Use a converted an original name. */ 1675 archive_entry_copy_uname(entry, tar->entry_uname.s); 1676 } 1677 } 1678 return (err); 1679} 1680 1681static int 1682pax_attribute_xattr(struct archive_entry *entry, 1683 const char *name, const char *value) 1684{ 1685 char *name_decoded; 1686 void *value_decoded; 1687 size_t value_len; 1688 1689 if (strlen(name) < 18 || (memcmp(name, "LIBARCHIVE.xattr.", 17)) != 0) 1690 return 3; 1691 1692 name += 17; 1693 1694 /* URL-decode name */ 1695 name_decoded = url_decode(name); 1696 if (name_decoded == NULL) 1697 return 2; 1698 1699 /* Base-64 decode value */ 1700 value_decoded = base64_decode(value, strlen(value), &value_len); 1701 if (value_decoded == NULL) { 1702 free(name_decoded); 1703 return 1; 1704 } 1705 1706 archive_entry_xattr_add_entry(entry, name_decoded, 1707 value_decoded, value_len); 1708 1709 free(name_decoded); 1710 free(value_decoded); 1711 return 0; 1712} 1713 1714/* 1715 * Parse a single key=value attribute. key/value pointers are 1716 * assumed to point into reasonably long-lived storage. 1717 * 1718 * Note that POSIX reserves all-lowercase keywords. Vendor-specific 1719 * extensions should always have keywords of the form "VENDOR.attribute" 1720 * In particular, it's quite feasible to support many different 1721 * vendor extensions here. I'm using "LIBARCHIVE" for extensions 1722 * unique to this library. 1723 * 1724 * Investigate other vendor-specific extensions and see if 1725 * any of them look useful. 1726 */ 1727static int 1728pax_attribute(struct archive_read *a, struct tar *tar, 1729 struct archive_entry *entry, const char *key, const char *value) 1730{ 1731 int64_t s; 1732 long n; 1733 int err = ARCHIVE_OK, r; 1734 1735#ifndef __FreeBSD__ 1736 if (value == NULL) 1737 value = ""; /* Disable compiler warning; do not pass 1738 * NULL pointer to strlen(). */ 1739#endif 1740 switch (key[0]) { 1741 case 'G': 1742 /* Reject GNU.sparse.* headers on non-regular files. */ 1743 if (strncmp(key, "GNU.sparse", 10) == 0 && 1744 !tar->sparse_allowed) { 1745 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1746 "Non-regular file cannot be sparse"); 1747 return (ARCHIVE_FATAL); 1748 } 1749 1750 /* GNU "0.0" sparse pax format. */ 1751 if (strcmp(key, "GNU.sparse.numblocks") == 0) { 1752 tar->sparse_offset = -1; 1753 tar->sparse_numbytes = -1; 1754 tar->sparse_gnu_major = 0; 1755 tar->sparse_gnu_minor = 0; 1756 } 1757 if (strcmp(key, "GNU.sparse.offset") == 0) { 1758 tar->sparse_offset = tar_atol10(value, strlen(value)); 1759 if (tar->sparse_numbytes != -1) { 1760 if (gnu_add_sparse_entry(a, tar, 1761 tar->sparse_offset, tar->sparse_numbytes) 1762 != ARCHIVE_OK) 1763 return (ARCHIVE_FATAL); 1764 tar->sparse_offset = -1; 1765 tar->sparse_numbytes = -1; 1766 } 1767 } 1768 if (strcmp(key, "GNU.sparse.numbytes") == 0) { 1769 tar->sparse_numbytes = tar_atol10(value, strlen(value)); 1770 if (tar->sparse_numbytes != -1) { 1771 if (gnu_add_sparse_entry(a, tar, 1772 tar->sparse_offset, tar->sparse_numbytes) 1773 != ARCHIVE_OK) 1774 return (ARCHIVE_FATAL); 1775 tar->sparse_offset = -1; 1776 tar->sparse_numbytes = -1; 1777 } 1778 } 1779 if (strcmp(key, "GNU.sparse.size") == 0) { 1780 tar->realsize = tar_atol10(value, strlen(value)); 1781 archive_entry_set_size(entry, tar->realsize); 1782 } 1783 1784 /* GNU "0.1" sparse pax format. */ 1785 if (strcmp(key, "GNU.sparse.map") == 0) { 1786 tar->sparse_gnu_major = 0; 1787 tar->sparse_gnu_minor = 1; 1788 if (gnu_sparse_01_parse(a, tar, value) != ARCHIVE_OK) 1789 return (ARCHIVE_WARN); 1790 } 1791 1792 /* GNU "1.0" sparse pax format */ 1793 if (strcmp(key, "GNU.sparse.major") == 0) { 1794 tar->sparse_gnu_major = (int)tar_atol10(value, strlen(value)); 1795 tar->sparse_gnu_pending = 1; 1796 } 1797 if (strcmp(key, "GNU.sparse.minor") == 0) { 1798 tar->sparse_gnu_minor = (int)tar_atol10(value, strlen(value)); 1799 tar->sparse_gnu_pending = 1; 1800 } 1801 if (strcmp(key, "GNU.sparse.name") == 0) { 1802 /* 1803 * The real filename; when storing sparse 1804 * files, GNU tar puts a synthesized name into 1805 * the regular 'path' attribute in an attempt 1806 * to limit confusion. ;-) 1807 */ 1808 archive_strcpy(&(tar->entry_pathname_override), value); 1809 } 1810 if (strcmp(key, "GNU.sparse.realsize") == 0) { 1811 tar->realsize = tar_atol10(value, strlen(value)); 1812 archive_entry_set_size(entry, tar->realsize); 1813 } 1814 break; 1815 case 'L': 1816 /* Our extensions */ 1817/* TODO: Handle arbitrary extended attributes... */ 1818/* 1819 if (strcmp(key, "LIBARCHIVE.xxxxxxx") == 0) 1820 archive_entry_set_xxxxxx(entry, value); 1821*/ 1822 if (strcmp(key, "LIBARCHIVE.creationtime") == 0) { 1823 pax_time(value, &s, &n); 1824 archive_entry_set_birthtime(entry, s, n); 1825 } 1826 if (memcmp(key, "LIBARCHIVE.xattr.", 17) == 0) 1827 pax_attribute_xattr(entry, key, value); 1828 break; 1829 case 'S': 1830 /* We support some keys used by the "star" archiver */ 1831 if (strcmp(key, "SCHILY.acl.access") == 0) { 1832 if (tar->sconv_acl == NULL) { 1833 tar->sconv_acl = 1834 archive_string_conversion_from_charset( 1835 &(a->archive), "UTF-8", 1); 1836 if (tar->sconv_acl == NULL) 1837 return (ARCHIVE_FATAL); 1838 } 1839 1840 r = archive_acl_parse_l(archive_entry_acl(entry), 1841 value, ARCHIVE_ENTRY_ACL_TYPE_ACCESS, 1842 tar->sconv_acl); 1843 if (r != ARCHIVE_OK) { 1844 err = r; 1845 if (err == ARCHIVE_FATAL) { 1846 archive_set_error(&a->archive, ENOMEM, 1847 "Can't allocate memory for " 1848 "SCHILY.acl.access"); 1849 return (err); 1850 } 1851 archive_set_error(&a->archive, 1852 ARCHIVE_ERRNO_MISC, 1853 "Parse error: SCHILY.acl.access"); 1854 } 1855 } else if (strcmp(key, "SCHILY.acl.default") == 0) { 1856 if (tar->sconv_acl == NULL) { 1857 tar->sconv_acl = 1858 archive_string_conversion_from_charset( 1859 &(a->archive), "UTF-8", 1); 1860 if (tar->sconv_acl == NULL) 1861 return (ARCHIVE_FATAL); 1862 } 1863 1864 r = archive_acl_parse_l(archive_entry_acl(entry), 1865 value, ARCHIVE_ENTRY_ACL_TYPE_DEFAULT, 1866 tar->sconv_acl); 1867 if (r != ARCHIVE_OK) { 1868 err = r; 1869 if (err == ARCHIVE_FATAL) { 1870 archive_set_error(&a->archive, ENOMEM, 1871 "Can't allocate memory for " 1872 "SCHILY.acl.default"); 1873 return (err); 1874 } 1875 archive_set_error(&a->archive, 1876 ARCHIVE_ERRNO_MISC, 1877 "Parse error: SCHILY.acl.default"); 1878 } 1879 } else if (strcmp(key, "SCHILY.devmajor") == 0) { 1880 archive_entry_set_rdevmajor(entry, 1881 (dev_t)tar_atol10(value, strlen(value))); 1882 } else if (strcmp(key, "SCHILY.devminor") == 0) { 1883 archive_entry_set_rdevminor(entry, 1884 (dev_t)tar_atol10(value, strlen(value))); 1885 } else if (strcmp(key, "SCHILY.fflags") == 0) { 1886 archive_entry_copy_fflags_text(entry, value); 1887 } else if (strcmp(key, "SCHILY.dev") == 0) { 1888 archive_entry_set_dev(entry, 1889 (dev_t)tar_atol10(value, strlen(value))); 1890 } else if (strcmp(key, "SCHILY.ino") == 0) { 1891 archive_entry_set_ino(entry, 1892 tar_atol10(value, strlen(value))); 1893 } else if (strcmp(key, "SCHILY.nlink") == 0) { 1894 archive_entry_set_nlink(entry, (unsigned) 1895 tar_atol10(value, strlen(value))); 1896 } else if (strcmp(key, "SCHILY.realsize") == 0) { 1897 tar->realsize = tar_atol10(value, strlen(value)); 1898 archive_entry_set_size(entry, tar->realsize); 1899 } else if (strcmp(key, "SUN.holesdata") == 0) { 1900 /* A Solaris extension for sparse. */ 1901 r = solaris_sparse_parse(a, tar, entry, value); 1902 if (r < err) { 1903 if (r == ARCHIVE_FATAL) 1904 return (r); 1905 err = r; 1906 archive_set_error(&a->archive, 1907 ARCHIVE_ERRNO_MISC, 1908 "Parse error: SUN.holesdata"); 1909 } 1910 } 1911 break; 1912 case 'a': 1913 if (strcmp(key, "atime") == 0) { 1914 pax_time(value, &s, &n); 1915 archive_entry_set_atime(entry, s, n); 1916 } 1917 break; 1918 case 'c': 1919 if (strcmp(key, "ctime") == 0) { 1920 pax_time(value, &s, &n); 1921 archive_entry_set_ctime(entry, s, n); 1922 } else if (strcmp(key, "charset") == 0) { 1923 /* TODO: Publish charset information in entry. */ 1924 } else if (strcmp(key, "comment") == 0) { 1925 /* TODO: Publish comment in entry. */ 1926 } 1927 break; 1928 case 'g': 1929 if (strcmp(key, "gid") == 0) { 1930 archive_entry_set_gid(entry, 1931 tar_atol10(value, strlen(value))); 1932 } else if (strcmp(key, "gname") == 0) { 1933 archive_strcpy(&(tar->entry_gname), value); 1934 } 1935 break; 1936 case 'h': 1937 if (strcmp(key, "hdrcharset") == 0) { 1938 if (strcmp(value, "BINARY") == 0) 1939 /* Binary mode. */ 1940 tar->pax_hdrcharset_binary = 1; 1941 else if (strcmp(value, "ISO-IR 10646 2000 UTF-8") == 0) 1942 tar->pax_hdrcharset_binary = 0; 1943 } 1944 break; 1945 case 'l': 1946 /* pax interchange doesn't distinguish hardlink vs. symlink. */ 1947 if (strcmp(key, "linkpath") == 0) { 1948 archive_strcpy(&(tar->entry_linkpath), value); 1949 } 1950 break; 1951 case 'm': 1952 if (strcmp(key, "mtime") == 0) { 1953 pax_time(value, &s, &n); 1954 archive_entry_set_mtime(entry, s, n); 1955 } 1956 break; 1957 case 'p': 1958 if (strcmp(key, "path") == 0) { 1959 archive_strcpy(&(tar->entry_pathname), value); 1960 } 1961 break; 1962 case 'r': 1963 /* POSIX has reserved 'realtime.*' */ 1964 break; 1965 case 's': 1966 /* POSIX has reserved 'security.*' */ 1967 /* Someday: if (strcmp(key, "security.acl") == 0) { ... } */ 1968 if (strcmp(key, "size") == 0) { 1969 /* "size" is the size of the data in the entry. */ 1970 tar->entry_bytes_remaining 1971 = tar_atol10(value, strlen(value)); 1972 /* 1973 * But, "size" is not necessarily the size of 1974 * the file on disk; if this is a sparse file, 1975 * the disk size may have already been set from 1976 * GNU.sparse.realsize or GNU.sparse.size or 1977 * an old GNU header field or SCHILY.realsize 1978 * or .... 1979 */ 1980 if (tar->realsize < 0) { 1981 archive_entry_set_size(entry, 1982 tar->entry_bytes_remaining); 1983 tar->realsize 1984 = tar->entry_bytes_remaining; 1985 } 1986 } 1987 break; 1988 case 'u': 1989 if (strcmp(key, "uid") == 0) { 1990 archive_entry_set_uid(entry, 1991 tar_atol10(value, strlen(value))); 1992 } else if (strcmp(key, "uname") == 0) { 1993 archive_strcpy(&(tar->entry_uname), value); 1994 } 1995 break; 1996 } 1997 return (err); 1998} 1999 2000 2001 2002/* 2003 * parse a decimal time value, which may include a fractional portion 2004 */ 2005static void 2006pax_time(const char *p, int64_t *ps, long *pn) 2007{ 2008 char digit; 2009 int64_t s; 2010 unsigned long l; 2011 int sign; 2012 int64_t limit, last_digit_limit; 2013 2014 limit = INT64_MAX / 10; 2015 last_digit_limit = INT64_MAX % 10; 2016 2017 s = 0; 2018 sign = 1; 2019 if (*p == '-') { 2020 sign = -1; 2021 p++; 2022 } 2023 while (*p >= '0' && *p <= '9') { 2024 digit = *p - '0'; 2025 if (s > limit || 2026 (s == limit && digit > last_digit_limit)) { 2027 s = INT64_MAX; 2028 break; 2029 } 2030 s = (s * 10) + digit; 2031 ++p; 2032 } 2033 2034 *ps = s * sign; 2035 2036 /* Calculate nanoseconds. */ 2037 *pn = 0; 2038 2039 if (*p != '.') 2040 return; 2041 2042 l = 100000000UL; 2043 do { 2044 ++p; 2045 if (*p >= '0' && *p <= '9') 2046 *pn += (*p - '0') * l; 2047 else 2048 break; 2049 } while (l /= 10); 2050} 2051 2052/* 2053 * Parse GNU tar header 2054 */ 2055static int 2056header_gnutar(struct archive_read *a, struct tar *tar, 2057 struct archive_entry *entry, const void *h, size_t *unconsumed) 2058{ 2059 const struct archive_entry_header_gnutar *header; 2060 int64_t t; 2061 int err = ARCHIVE_OK; 2062 2063 /* 2064 * GNU header is like POSIX ustar, except 'prefix' is 2065 * replaced with some other fields. This also means the 2066 * filename is stored as in old-style archives. 2067 */ 2068 2069 /* Grab fields common to all tar variants. */ 2070 err = header_common(a, tar, entry, h); 2071 if (err == ARCHIVE_FATAL) 2072 return (err); 2073 2074 /* Copy filename over (to ensure null termination). */ 2075 header = (const struct archive_entry_header_gnutar *)h; 2076 if (archive_entry_copy_pathname_l(entry, 2077 header->name, sizeof(header->name), tar->sconv) != 0) { 2078 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 2079 if (err == ARCHIVE_FATAL) 2080 return (err); 2081 } 2082 2083 /* Fields common to ustar and GNU */ 2084 /* XXX Can the following be factored out since it's common 2085 * to ustar and gnu tar? Is it okay to move it down into 2086 * header_common, perhaps? */ 2087 if (archive_entry_copy_uname_l(entry, 2088 header->uname, sizeof(header->uname), tar->sconv) != 0) { 2089 err = set_conversion_failed_error(a, tar->sconv, "Uname"); 2090 if (err == ARCHIVE_FATAL) 2091 return (err); 2092 } 2093 2094 if (archive_entry_copy_gname_l(entry, 2095 header->gname, sizeof(header->gname), tar->sconv) != 0) { 2096 err = set_conversion_failed_error(a, tar->sconv, "Gname"); 2097 if (err == ARCHIVE_FATAL) 2098 return (err); 2099 } 2100 2101 /* Parse out device numbers only for char and block specials */ 2102 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') { 2103 archive_entry_set_rdevmajor(entry, (dev_t) 2104 tar_atol(header->rdevmajor, sizeof(header->rdevmajor))); 2105 archive_entry_set_rdevminor(entry, (dev_t) 2106 tar_atol(header->rdevminor, sizeof(header->rdevminor))); 2107 } else 2108 archive_entry_set_rdev(entry, 0); 2109 2110 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 2111 2112 /* Grab GNU-specific fields. */ 2113 t = tar_atol(header->atime, sizeof(header->atime)); 2114 if (t > 0) 2115 archive_entry_set_atime(entry, t, 0); 2116 t = tar_atol(header->ctime, sizeof(header->ctime)); 2117 if (t > 0) 2118 archive_entry_set_ctime(entry, t, 0); 2119 2120 if (header->realsize[0] != 0) { 2121 tar->realsize 2122 = tar_atol(header->realsize, sizeof(header->realsize)); 2123 archive_entry_set_size(entry, tar->realsize); 2124 } 2125 2126 if (header->sparse[0].offset[0] != 0) { 2127 if (gnu_sparse_old_read(a, tar, header, unconsumed) 2128 != ARCHIVE_OK) 2129 return (ARCHIVE_FATAL); 2130 } else { 2131 if (header->isextended[0] != 0) { 2132 /* XXX WTF? XXX */ 2133 } 2134 } 2135 2136 return (err); 2137} 2138 2139static int 2140gnu_add_sparse_entry(struct archive_read *a, struct tar *tar, 2141 int64_t offset, int64_t remaining) 2142{ 2143 struct sparse_block *p; 2144 2145 p = (struct sparse_block *)malloc(sizeof(*p)); 2146 if (p == NULL) { 2147 archive_set_error(&a->archive, ENOMEM, "Out of memory"); 2148 return (ARCHIVE_FATAL); 2149 } 2150 memset(p, 0, sizeof(*p)); 2151 if (tar->sparse_last != NULL) 2152 tar->sparse_last->next = p; 2153 else 2154 tar->sparse_list = p; 2155 tar->sparse_last = p; 2156 if (remaining < 0 || offset < 0) { 2157 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Malformed sparse map data"); 2158 return (ARCHIVE_FATAL); 2159 } 2160 p->offset = offset; 2161 p->remaining = remaining; 2162 return (ARCHIVE_OK); 2163} 2164 2165static void 2166gnu_clear_sparse_list(struct tar *tar) 2167{ 2168 struct sparse_block *p; 2169 2170 while (tar->sparse_list != NULL) { 2171 p = tar->sparse_list; 2172 tar->sparse_list = p->next; 2173 free(p); 2174 } 2175 tar->sparse_last = NULL; 2176} 2177 2178/* 2179 * GNU tar old-format sparse data. 2180 * 2181 * GNU old-format sparse data is stored in a fixed-field 2182 * format. Offset/size values are 11-byte octal fields (same 2183 * format as 'size' field in ustart header). These are 2184 * stored in the header, allocating subsequent header blocks 2185 * as needed. Extending the header in this way is a pretty 2186 * severe POSIX violation; this design has earned GNU tar a 2187 * lot of criticism. 2188 */ 2189 2190static int 2191gnu_sparse_old_read(struct archive_read *a, struct tar *tar, 2192 const struct archive_entry_header_gnutar *header, size_t *unconsumed) 2193{ 2194 ssize_t bytes_read; 2195 const void *data; 2196 struct extended { 2197 struct gnu_sparse sparse[21]; 2198 char isextended[1]; 2199 char padding[7]; 2200 }; 2201 const struct extended *ext; 2202 2203 if (gnu_sparse_old_parse(a, tar, header->sparse, 4) != ARCHIVE_OK) 2204 return (ARCHIVE_FATAL); 2205 if (header->isextended[0] == 0) 2206 return (ARCHIVE_OK); 2207 2208 do { 2209 tar_flush_unconsumed(a, unconsumed); 2210 data = __archive_read_ahead(a, 512, &bytes_read); 2211 if (bytes_read < 0) 2212 return (ARCHIVE_FATAL); 2213 if (bytes_read < 512) { 2214 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 2215 "Truncated tar archive " 2216 "detected while reading sparse file data"); 2217 return (ARCHIVE_FATAL); 2218 } 2219 *unconsumed = 512; 2220 ext = (const struct extended *)data; 2221 if (gnu_sparse_old_parse(a, tar, ext->sparse, 21) != ARCHIVE_OK) 2222 return (ARCHIVE_FATAL); 2223 } while (ext->isextended[0] != 0); 2224 if (tar->sparse_list != NULL) 2225 tar->entry_offset = tar->sparse_list->offset; 2226 return (ARCHIVE_OK); 2227} 2228 2229static int 2230gnu_sparse_old_parse(struct archive_read *a, struct tar *tar, 2231 const struct gnu_sparse *sparse, int length) 2232{ 2233 while (length > 0 && sparse->offset[0] != 0) { 2234 if (gnu_add_sparse_entry(a, tar, 2235 tar_atol(sparse->offset, sizeof(sparse->offset)), 2236 tar_atol(sparse->numbytes, sizeof(sparse->numbytes))) 2237 != ARCHIVE_OK) 2238 return (ARCHIVE_FATAL); 2239 sparse++; 2240 length--; 2241 } 2242 return (ARCHIVE_OK); 2243} 2244 2245/* 2246 * GNU tar sparse format 0.0 2247 * 2248 * Beginning with GNU tar 1.15, sparse files are stored using 2249 * information in the pax extended header. The GNU tar maintainers 2250 * have gone through a number of variations in the process of working 2251 * out this scheme; fortunately, they're all numbered. 2252 * 2253 * Sparse format 0.0 uses attribute GNU.sparse.numblocks to store the 2254 * number of blocks, and GNU.sparse.offset/GNU.sparse.numbytes to 2255 * store offset/size for each block. The repeated instances of these 2256 * latter fields violate the pax specification (which frowns on 2257 * duplicate keys), so this format was quickly replaced. 2258 */ 2259 2260/* 2261 * GNU tar sparse format 0.1 2262 * 2263 * This version replaced the offset/numbytes attributes with 2264 * a single "map" attribute that stored a list of integers. This 2265 * format had two problems: First, the "map" attribute could be very 2266 * long, which caused problems for some implementations. More 2267 * importantly, the sparse data was lost when extracted by archivers 2268 * that didn't recognize this extension. 2269 */ 2270 2271static int 2272gnu_sparse_01_parse(struct archive_read *a, struct tar *tar, const char *p) 2273{ 2274 const char *e; 2275 int64_t offset = -1, size = -1; 2276 2277 for (;;) { 2278 e = p; 2279 while (*e != '\0' && *e != ',') { 2280 if (*e < '0' || *e > '9') 2281 return (ARCHIVE_WARN); 2282 e++; 2283 } 2284 if (offset < 0) { 2285 offset = tar_atol10(p, e - p); 2286 if (offset < 0) 2287 return (ARCHIVE_WARN); 2288 } else { 2289 size = tar_atol10(p, e - p); 2290 if (size < 0) 2291 return (ARCHIVE_WARN); 2292 if (gnu_add_sparse_entry(a, tar, offset, size) 2293 != ARCHIVE_OK) 2294 return (ARCHIVE_FATAL); 2295 offset = -1; 2296 } 2297 if (*e == '\0') 2298 return (ARCHIVE_OK); 2299 p = e + 1; 2300 } 2301} 2302 2303/* 2304 * GNU tar sparse format 1.0 2305 * 2306 * The idea: The offset/size data is stored as a series of base-10 2307 * ASCII numbers prepended to the file data, so that dearchivers that 2308 * don't support this format will extract the block map along with the 2309 * data and a separate post-process can restore the sparseness. 2310 * 2311 * Unfortunately, GNU tar 1.16 had a bug that added unnecessary 2312 * padding to the body of the file when using this format. GNU tar 2313 * 1.17 corrected this bug without bumping the version number, so 2314 * it's not possible to support both variants. This code supports 2315 * the later variant at the expense of not supporting the former. 2316 * 2317 * This variant also replaced GNU.sparse.size with GNU.sparse.realsize 2318 * and introduced the GNU.sparse.major/GNU.sparse.minor attributes. 2319 */ 2320 2321/* 2322 * Read the next line from the input, and parse it as a decimal 2323 * integer followed by '\n'. Returns positive integer value or 2324 * negative on error. 2325 */ 2326static int64_t 2327gnu_sparse_10_atol(struct archive_read *a, struct tar *tar, 2328 int64_t *remaining, size_t *unconsumed) 2329{ 2330 int64_t l, limit, last_digit_limit; 2331 const char *p; 2332 ssize_t bytes_read; 2333 int base, digit; 2334 2335 base = 10; 2336 limit = INT64_MAX / base; 2337 last_digit_limit = INT64_MAX % base; 2338 2339 /* 2340 * Skip any lines starting with '#'; GNU tar specs 2341 * don't require this, but they should. 2342 */ 2343 do { 2344 bytes_read = readline(a, tar, &p, 2345 (ssize_t)tar_min(*remaining, 100), unconsumed); 2346 if (bytes_read <= 0) 2347 return (ARCHIVE_FATAL); 2348 *remaining -= bytes_read; 2349 } while (p[0] == '#'); 2350 2351 l = 0; 2352 while (bytes_read > 0) { 2353 if (*p == '\n') 2354 return (l); 2355 if (*p < '0' || *p >= '0' + base) 2356 return (ARCHIVE_WARN); 2357 digit = *p - '0'; 2358 if (l > limit || (l == limit && digit > last_digit_limit)) 2359 l = INT64_MAX; /* Truncate on overflow. */ 2360 else 2361 l = (l * base) + digit; 2362 p++; 2363 bytes_read--; 2364 } 2365 /* TODO: Error message. */ 2366 return (ARCHIVE_WARN); 2367} 2368 2369/* 2370 * Returns length (in bytes) of the sparse data description 2371 * that was read. 2372 */ 2373static ssize_t 2374gnu_sparse_10_read(struct archive_read *a, struct tar *tar, size_t *unconsumed) 2375{ 2376 ssize_t bytes_read; 2377 int entries; 2378 int64_t offset, size, to_skip, remaining; 2379 2380 /* Clear out the existing sparse list. */ 2381 gnu_clear_sparse_list(tar); 2382 2383 remaining = tar->entry_bytes_remaining; 2384 2385 /* Parse entries. */ 2386 entries = (int)gnu_sparse_10_atol(a, tar, &remaining, unconsumed); 2387 if (entries < 0) 2388 return (ARCHIVE_FATAL); 2389 /* Parse the individual entries. */ 2390 while (entries-- > 0) { 2391 /* Parse offset/size */ 2392 offset = gnu_sparse_10_atol(a, tar, &remaining, unconsumed); 2393 if (offset < 0) 2394 return (ARCHIVE_FATAL); 2395 size = gnu_sparse_10_atol(a, tar, &remaining, unconsumed); 2396 if (size < 0) 2397 return (ARCHIVE_FATAL); 2398 /* Add a new sparse entry. */ 2399 if (gnu_add_sparse_entry(a, tar, offset, size) != ARCHIVE_OK) 2400 return (ARCHIVE_FATAL); 2401 } 2402 /* Skip rest of block... */ 2403 tar_flush_unconsumed(a, unconsumed); 2404 bytes_read = (ssize_t)(tar->entry_bytes_remaining - remaining); 2405 to_skip = 0x1ff & -bytes_read; 2406 if (to_skip != __archive_read_consume(a, to_skip)) 2407 return (ARCHIVE_FATAL); 2408 return ((ssize_t)(bytes_read + to_skip)); 2409} 2410 2411/* 2412 * Solaris pax extension for a sparse file. This is recorded with the 2413 * data and hole pairs. The way recording sparse information by Solaris' 2414 * pax simply indicates where data and sparse are, so the stored contents 2415 * consist of both data and hole. 2416 */ 2417static int 2418solaris_sparse_parse(struct archive_read *a, struct tar *tar, 2419 struct archive_entry *entry, const char *p) 2420{ 2421 const char *e; 2422 int64_t start, end; 2423 int hole = 1; 2424 2425 (void)entry; /* UNUSED */ 2426 2427 end = 0; 2428 if (*p == ' ') 2429 p++; 2430 else 2431 return (ARCHIVE_WARN); 2432 for (;;) { 2433 e = p; 2434 while (*e != '\0' && *e != ' ') { 2435 if (*e < '0' || *e > '9') 2436 return (ARCHIVE_WARN); 2437 e++; 2438 } 2439 start = end; 2440 end = tar_atol10(p, e - p); 2441 if (end < 0) 2442 return (ARCHIVE_WARN); 2443 if (start < end) { 2444 if (gnu_add_sparse_entry(a, tar, start, 2445 end - start) != ARCHIVE_OK) 2446 return (ARCHIVE_FATAL); 2447 tar->sparse_last->hole = hole; 2448 } 2449 if (*e == '\0') 2450 return (ARCHIVE_OK); 2451 p = e + 1; 2452 hole = hole == 0; 2453 } 2454} 2455 2456/*- 2457 * Convert text->integer. 2458 * 2459 * Traditional tar formats (including POSIX) specify base-8 for 2460 * all of the standard numeric fields. This is a significant limitation 2461 * in practice: 2462 * = file size is limited to 8GB 2463 * = rdevmajor and rdevminor are limited to 21 bits 2464 * = uid/gid are limited to 21 bits 2465 * 2466 * There are two workarounds for this: 2467 * = pax extended headers, which use variable-length string fields 2468 * = GNU tar and STAR both allow either base-8 or base-256 in 2469 * most fields. The high bit is set to indicate base-256. 2470 * 2471 * On read, this implementation supports both extensions. 2472 */ 2473static int64_t 2474tar_atol(const char *p, size_t char_cnt) 2475{ 2476 /* 2477 * Technically, GNU tar considers a field to be in base-256 2478 * only if the first byte is 0xff or 0x80. 2479 */ 2480 if (*p & 0x80) 2481 return (tar_atol256(p, char_cnt)); 2482 return (tar_atol8(p, char_cnt)); 2483} 2484 2485/* 2486 * Note that this implementation does not (and should not!) obey 2487 * locale settings; you cannot simply substitute strtol here, since 2488 * it does obey locale. 2489 */ 2490static int64_t 2491tar_atol_base_n(const char *p, size_t char_cnt, int base) 2492{ 2493 int64_t l, maxval, limit, last_digit_limit; 2494 int digit, sign; 2495 2496 maxval = INT64_MAX; 2497 limit = INT64_MAX / base; 2498 last_digit_limit = INT64_MAX % base; 2499 2500 /* the pointer will not be dereferenced if char_cnt is zero 2501 * due to the way the && operator is evaulated. 2502 */ 2503 while (char_cnt != 0 && (*p == ' ' || *p == '\t')) { 2504 p++; 2505 char_cnt--; 2506 } 2507 2508 sign = 1; 2509 if (char_cnt != 0 && *p == '-') { 2510 sign = -1; 2511 p++; 2512 char_cnt--; 2513 2514 maxval = INT64_MIN; 2515 limit = -(INT64_MIN / base); 2516 last_digit_limit = INT64_MIN % base; 2517 } 2518 2519 l = 0; 2520 if (char_cnt != 0) { 2521 digit = *p - '0'; 2522 while (digit >= 0 && digit < base && char_cnt != 0) { 2523 if (l>limit || (l == limit && digit > last_digit_limit)) { 2524 return maxval; /* Truncate on overflow. */ 2525 } 2526 l = (l * base) + digit; 2527 digit = *++p - '0'; 2528 char_cnt--; 2529 } 2530 } 2531 return (sign < 0) ? -l : l; 2532} 2533 2534static int64_t 2535tar_atol8(const char *p, size_t char_cnt) 2536{ 2537 return tar_atol_base_n(p, char_cnt, 8); 2538} 2539 2540static int64_t 2541tar_atol10(const char *p, size_t char_cnt) 2542{ 2543 return tar_atol_base_n(p, char_cnt, 10); 2544} 2545 2546/* 2547 * Parse a base-256 integer. This is just a variable-length 2548 * twos-complement signed binary value in big-endian order, except 2549 * that the high-order bit is ignored. The values here can be up to 2550 * 12 bytes, so we need to be careful about overflowing 64-bit 2551 * (8-byte) integers. 2552 * 2553 * This code unashamedly assumes that the local machine uses 8-bit 2554 * bytes and twos-complement arithmetic. 2555 */ 2556static int64_t 2557tar_atol256(const char *_p, size_t char_cnt) 2558{ 2559 uint64_t l; 2560 const unsigned char *p = (const unsigned char *)_p; 2561 unsigned char c, neg; 2562 2563 /* Extend 7-bit 2s-comp to 8-bit 2s-comp, decide sign. */ 2564 c = *p; 2565 if (c & 0x40) { 2566 neg = 0xff; 2567 c |= 0x80; 2568 l = ~ARCHIVE_LITERAL_ULL(0); 2569 } else { 2570 neg = 0; 2571 c &= 0x7f; 2572 l = 0; 2573 } 2574 2575 /* If more than 8 bytes, check that we can ignore 2576 * high-order bits without overflow. */ 2577 while (char_cnt > sizeof(int64_t)) { 2578 --char_cnt; 2579 if (c != neg) 2580 return neg ? INT64_MIN : INT64_MAX; 2581 c = *++p; 2582 } 2583 2584 /* c is first byte that fits; if sign mismatch, return overflow */ 2585 if ((c ^ neg) & 0x80) { 2586 return neg ? INT64_MIN : INT64_MAX; 2587 } 2588 2589 /* Accumulate remaining bytes. */ 2590 while (--char_cnt > 0) { 2591 l = (l << 8) | c; 2592 c = *++p; 2593 } 2594 l = (l << 8) | c; 2595 /* Return signed twos-complement value. */ 2596 return (int64_t)(l); 2597} 2598 2599/* 2600 * Returns length of line (including trailing newline) 2601 * or negative on error. 'start' argument is updated to 2602 * point to first character of line. This avoids copying 2603 * when possible. 2604 */ 2605static ssize_t 2606readline(struct archive_read *a, struct tar *tar, const char **start, 2607 ssize_t limit, size_t *unconsumed) 2608{ 2609 ssize_t bytes_read; 2610 ssize_t total_size = 0; 2611 const void *t; 2612 const char *s; 2613 void *p; 2614 2615 tar_flush_unconsumed(a, unconsumed); 2616 2617 t = __archive_read_ahead(a, 1, &bytes_read); 2618 if (bytes_read <= 0) 2619 return (ARCHIVE_FATAL); 2620 s = t; /* Start of line? */ 2621 p = memchr(t, '\n', bytes_read); 2622 /* If we found '\n' in the read buffer, return pointer to that. */ 2623 if (p != NULL) { 2624 bytes_read = 1 + ((const char *)p) - s; 2625 if (bytes_read > limit) { 2626 archive_set_error(&a->archive, 2627 ARCHIVE_ERRNO_FILE_FORMAT, 2628 "Line too long"); 2629 return (ARCHIVE_FATAL); 2630 } 2631 *unconsumed = bytes_read; 2632 *start = s; 2633 return (bytes_read); 2634 } 2635 *unconsumed = bytes_read; 2636 /* Otherwise, we need to accumulate in a line buffer. */ 2637 for (;;) { 2638 if (total_size + bytes_read > limit) { 2639 archive_set_error(&a->archive, 2640 ARCHIVE_ERRNO_FILE_FORMAT, 2641 "Line too long"); 2642 return (ARCHIVE_FATAL); 2643 } 2644 if (archive_string_ensure(&tar->line, total_size + bytes_read) == NULL) { 2645 archive_set_error(&a->archive, ENOMEM, 2646 "Can't allocate working buffer"); 2647 return (ARCHIVE_FATAL); 2648 } 2649 memcpy(tar->line.s + total_size, t, bytes_read); 2650 tar_flush_unconsumed(a, unconsumed); 2651 total_size += bytes_read; 2652 /* If we found '\n', clean up and return. */ 2653 if (p != NULL) { 2654 *start = tar->line.s; 2655 return (total_size); 2656 } 2657 /* Read some more. */ 2658 t = __archive_read_ahead(a, 1, &bytes_read); 2659 if (bytes_read <= 0) 2660 return (ARCHIVE_FATAL); 2661 s = t; /* Start of line? */ 2662 p = memchr(t, '\n', bytes_read); 2663 /* If we found '\n', trim the read. */ 2664 if (p != NULL) { 2665 bytes_read = 1 + ((const char *)p) - s; 2666 } 2667 *unconsumed = bytes_read; 2668 } 2669} 2670 2671/* 2672 * base64_decode - Base64 decode 2673 * 2674 * This accepts most variations of base-64 encoding, including: 2675 * * with or without line breaks 2676 * * with or without the final group padded with '=' or '_' characters 2677 * (The most economical Base-64 variant does not pad the last group and 2678 * omits line breaks; RFC1341 used for MIME requires both.) 2679 */ 2680static char * 2681base64_decode(const char *s, size_t len, size_t *out_len) 2682{ 2683 static const unsigned char digits[64] = { 2684 'A','B','C','D','E','F','G','H','I','J','K','L','M','N', 2685 'O','P','Q','R','S','T','U','V','W','X','Y','Z','a','b', 2686 'c','d','e','f','g','h','i','j','k','l','m','n','o','p', 2687 'q','r','s','t','u','v','w','x','y','z','0','1','2','3', 2688 '4','5','6','7','8','9','+','/' }; 2689 static unsigned char decode_table[128]; 2690 char *out, *d; 2691 const unsigned char *src = (const unsigned char *)s; 2692 2693 /* If the decode table is not yet initialized, prepare it. */ 2694 if (decode_table[digits[1]] != 1) { 2695 unsigned i; 2696 memset(decode_table, 0xff, sizeof(decode_table)); 2697 for (i = 0; i < sizeof(digits); i++) 2698 decode_table[digits[i]] = i; 2699 } 2700 2701 /* Allocate enough space to hold the entire output. */ 2702 /* Note that we may not use all of this... */ 2703 out = (char *)malloc(len - len / 4 + 1); 2704 if (out == NULL) { 2705 *out_len = 0; 2706 return (NULL); 2707 } 2708 d = out; 2709 2710 while (len > 0) { 2711 /* Collect the next group of (up to) four characters. */ 2712 int v = 0; 2713 int group_size = 0; 2714 while (group_size < 4 && len > 0) { 2715 /* '=' or '_' padding indicates final group. */ 2716 if (*src == '=' || *src == '_') { 2717 len = 0; 2718 break; 2719 } 2720 /* Skip illegal characters (including line breaks) */ 2721 if (*src > 127 || *src < 32 2722 || decode_table[*src] == 0xff) { 2723 len--; 2724 src++; 2725 continue; 2726 } 2727 v <<= 6; 2728 v |= decode_table[*src++]; 2729 len --; 2730 group_size++; 2731 } 2732 /* Align a short group properly. */ 2733 v <<= 6 * (4 - group_size); 2734 /* Unpack the group we just collected. */ 2735 switch (group_size) { 2736 case 4: d[2] = v & 0xff; 2737 /* FALLTHROUGH */ 2738 case 3: d[1] = (v >> 8) & 0xff; 2739 /* FALLTHROUGH */ 2740 case 2: d[0] = (v >> 16) & 0xff; 2741 break; 2742 case 1: /* this is invalid! */ 2743 break; 2744 } 2745 d += group_size * 3 / 4; 2746 } 2747 2748 *out_len = d - out; 2749 return (out); 2750} 2751 2752static char * 2753url_decode(const char *in) 2754{ 2755 char *out, *d; 2756 const char *s; 2757 2758 out = (char *)malloc(strlen(in) + 1); 2759 if (out == NULL) 2760 return (NULL); 2761 for (s = in, d = out; *s != '\0'; ) { 2762 if (s[0] == '%' && s[1] != '\0' && s[2] != '\0') { 2763 /* Try to convert % escape */ 2764 int digit1 = tohex(s[1]); 2765 int digit2 = tohex(s[2]); 2766 if (digit1 >= 0 && digit2 >= 0) { 2767 /* Looks good, consume three chars */ 2768 s += 3; 2769 /* Convert output */ 2770 *d++ = ((digit1 << 4) | digit2); 2771 continue; 2772 } 2773 /* Else fall through and treat '%' as normal char */ 2774 } 2775 *d++ = *s++; 2776 } 2777 *d = '\0'; 2778 return (out); 2779} 2780 2781static int 2782tohex(int c) 2783{ 2784 if (c >= '0' && c <= '9') 2785 return (c - '0'); 2786 else if (c >= 'A' && c <= 'F') 2787 return (c - 'A' + 10); 2788 else if (c >= 'a' && c <= 'f') 2789 return (c - 'a' + 10); 2790 else 2791 return (-1); 2792} 2793