archive_read_support_format_tar.c revision 305192
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/10/contrib/libarchive/libarchive/archive_read_support_format_tar.c 305192 2016-09-01 12:01:23Z 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 struct sparse_block *sparse_list; 140 struct sparse_block *sparse_last; 141 int64_t sparse_offset; 142 int64_t sparse_numbytes; 143 int sparse_gnu_major; 144 int sparse_gnu_minor; 145 char sparse_gnu_pending; 146 147 struct archive_string localname; 148 struct archive_string_conv *opt_sconv; 149 struct archive_string_conv *sconv; 150 struct archive_string_conv *sconv_acl; 151 struct archive_string_conv *sconv_default; 152 int init_default_conversion; 153 int compat_2x; 154 int process_mac_extensions; 155 int read_concatenated_archives; 156}; 157 158static int archive_block_is_null(const char *p); 159static char *base64_decode(const char *, size_t, size_t *); 160static int gnu_add_sparse_entry(struct archive_read *, struct tar *, 161 int64_t offset, int64_t remaining); 162 163static void gnu_clear_sparse_list(struct tar *); 164static int gnu_sparse_old_read(struct archive_read *, struct tar *, 165 const struct archive_entry_header_gnutar *header, size_t *); 166static int gnu_sparse_old_parse(struct archive_read *, struct tar *, 167 const struct gnu_sparse *sparse, int length); 168static int gnu_sparse_01_parse(struct archive_read *, struct tar *, 169 const char *); 170static ssize_t gnu_sparse_10_read(struct archive_read *, struct tar *, 171 size_t *); 172static int header_Solaris_ACL(struct archive_read *, struct tar *, 173 struct archive_entry *, const void *, size_t *); 174static int header_common(struct archive_read *, struct tar *, 175 struct archive_entry *, const void *); 176static int header_old_tar(struct archive_read *, struct tar *, 177 struct archive_entry *, const void *); 178static int header_pax_extensions(struct archive_read *, struct tar *, 179 struct archive_entry *, const void *, size_t *); 180static int header_pax_global(struct archive_read *, struct tar *, 181 struct archive_entry *, const void *h, size_t *); 182static int header_longlink(struct archive_read *, struct tar *, 183 struct archive_entry *, const void *h, size_t *); 184static int header_longname(struct archive_read *, struct tar *, 185 struct archive_entry *, const void *h, size_t *); 186static int read_mac_metadata_blob(struct archive_read *, struct tar *, 187 struct archive_entry *, const void *h, size_t *); 188static int header_volume(struct archive_read *, struct tar *, 189 struct archive_entry *, const void *h, size_t *); 190static int header_ustar(struct archive_read *, struct tar *, 191 struct archive_entry *, const void *h); 192static int header_gnutar(struct archive_read *, struct tar *, 193 struct archive_entry *, const void *h, size_t *); 194static int archive_read_format_tar_bid(struct archive_read *, int); 195static int archive_read_format_tar_options(struct archive_read *, 196 const char *, const char *); 197static int archive_read_format_tar_cleanup(struct archive_read *); 198static int archive_read_format_tar_read_data(struct archive_read *a, 199 const void **buff, size_t *size, int64_t *offset); 200static int archive_read_format_tar_skip(struct archive_read *a); 201static int archive_read_format_tar_read_header(struct archive_read *, 202 struct archive_entry *); 203static int checksum(struct archive_read *, const void *); 204static int pax_attribute(struct archive_read *, struct tar *, 205 struct archive_entry *, const char *key, const char *value); 206static int pax_header(struct archive_read *, struct tar *, 207 struct archive_entry *, char *attr); 208static void pax_time(const char *, int64_t *sec, long *nanos); 209static ssize_t readline(struct archive_read *, struct tar *, const char **, 210 ssize_t limit, size_t *); 211static int read_body_to_string(struct archive_read *, struct tar *, 212 struct archive_string *, const void *h, size_t *); 213static int solaris_sparse_parse(struct archive_read *, struct tar *, 214 struct archive_entry *, const char *); 215static int64_t tar_atol(const char *, size_t); 216static int64_t tar_atol10(const char *, size_t); 217static int64_t tar_atol256(const char *, size_t); 218static int64_t tar_atol8(const char *, size_t); 219static int tar_read_header(struct archive_read *, struct tar *, 220 struct archive_entry *, size_t *); 221static int tohex(int c); 222static char *url_decode(const char *); 223static void tar_flush_unconsumed(struct archive_read *, size_t *); 224 225 226int 227archive_read_support_format_gnutar(struct archive *a) 228{ 229 archive_check_magic(a, ARCHIVE_READ_MAGIC, 230 ARCHIVE_STATE_NEW, "archive_read_support_format_gnutar"); 231 return (archive_read_support_format_tar(a)); 232} 233 234 235int 236archive_read_support_format_tar(struct archive *_a) 237{ 238 struct archive_read *a = (struct archive_read *)_a; 239 struct tar *tar; 240 int r; 241 242 archive_check_magic(_a, ARCHIVE_READ_MAGIC, 243 ARCHIVE_STATE_NEW, "archive_read_support_format_tar"); 244 245 tar = (struct tar *)calloc(1, sizeof(*tar)); 246#ifdef HAVE_COPYFILE_H 247 /* Set this by default on Mac OS. */ 248 tar->process_mac_extensions = 1; 249#endif 250 if (tar == NULL) { 251 archive_set_error(&a->archive, ENOMEM, 252 "Can't allocate tar data"); 253 return (ARCHIVE_FATAL); 254 } 255 256 r = __archive_read_register_format(a, tar, "tar", 257 archive_read_format_tar_bid, 258 archive_read_format_tar_options, 259 archive_read_format_tar_read_header, 260 archive_read_format_tar_read_data, 261 archive_read_format_tar_skip, 262 NULL, 263 archive_read_format_tar_cleanup, 264 NULL, 265 NULL); 266 267 if (r != ARCHIVE_OK) 268 free(tar); 269 return (ARCHIVE_OK); 270} 271 272static int 273archive_read_format_tar_cleanup(struct archive_read *a) 274{ 275 struct tar *tar; 276 277 tar = (struct tar *)(a->format->data); 278 gnu_clear_sparse_list(tar); 279 archive_string_free(&tar->acl_text); 280 archive_string_free(&tar->entry_pathname); 281 archive_string_free(&tar->entry_pathname_override); 282 archive_string_free(&tar->entry_linkpath); 283 archive_string_free(&tar->entry_uname); 284 archive_string_free(&tar->entry_gname); 285 archive_string_free(&tar->line); 286 archive_string_free(&tar->pax_global); 287 archive_string_free(&tar->pax_header); 288 archive_string_free(&tar->longname); 289 archive_string_free(&tar->longlink); 290 archive_string_free(&tar->localname); 291 free(tar); 292 (a->format->data) = NULL; 293 return (ARCHIVE_OK); 294} 295 296 297static int 298archive_read_format_tar_bid(struct archive_read *a, int best_bid) 299{ 300 int bid; 301 const char *h; 302 const struct archive_entry_header_ustar *header; 303 304 (void)best_bid; /* UNUSED */ 305 306 bid = 0; 307 308 /* Now let's look at the actual header and see if it matches. */ 309 h = __archive_read_ahead(a, 512, NULL); 310 if (h == NULL) 311 return (-1); 312 313 /* If it's an end-of-archive mark, we can handle it. */ 314 if (h[0] == 0 && archive_block_is_null(h)) { 315 /* 316 * Usually, I bid the number of bits verified, but 317 * in this case, 4096 seems excessive so I picked 10 as 318 * an arbitrary but reasonable-seeming value. 319 */ 320 return (10); 321 } 322 323 /* If it's not an end-of-archive mark, it must have a valid checksum.*/ 324 if (!checksum(a, h)) 325 return (0); 326 bid += 48; /* Checksum is usually 6 octal digits. */ 327 328 header = (const struct archive_entry_header_ustar *)h; 329 330 /* Recognize POSIX formats. */ 331 if ((memcmp(header->magic, "ustar\0", 6) == 0) 332 && (memcmp(header->version, "00", 2) == 0)) 333 bid += 56; 334 335 /* Recognize GNU tar format. */ 336 if ((memcmp(header->magic, "ustar ", 6) == 0) 337 && (memcmp(header->version, " \0", 2) == 0)) 338 bid += 56; 339 340 /* Type flag must be null, digit or A-Z, a-z. */ 341 if (header->typeflag[0] != 0 && 342 !( header->typeflag[0] >= '0' && header->typeflag[0] <= '9') && 343 !( header->typeflag[0] >= 'A' && header->typeflag[0] <= 'Z') && 344 !( header->typeflag[0] >= 'a' && header->typeflag[0] <= 'z') ) 345 return (0); 346 bid += 2; /* 6 bits of variation in an 8-bit field leaves 2 bits. */ 347 348 /* Sanity check: Look at first byte of mode field. */ 349 switch (255 & (unsigned)header->mode[0]) { 350 case 0: case 255: 351 /* Base-256 value: No further verification possible! */ 352 break; 353 case ' ': /* Not recommended, but not illegal, either. */ 354 break; 355 case '0': case '1': case '2': case '3': 356 case '4': case '5': case '6': case '7': 357 /* Octal Value. */ 358 /* TODO: Check format of remainder of this field. */ 359 break; 360 default: 361 /* Not a valid mode; bail out here. */ 362 return (0); 363 } 364 /* TODO: Sanity test uid/gid/size/mtime/rdevmajor/rdevminor fields. */ 365 366 return (bid); 367} 368 369static int 370archive_read_format_tar_options(struct archive_read *a, 371 const char *key, const char *val) 372{ 373 struct tar *tar; 374 int ret = ARCHIVE_FAILED; 375 376 tar = (struct tar *)(a->format->data); 377 if (strcmp(key, "compat-2x") == 0) { 378 /* Handle UTF-8 filnames as libarchive 2.x */ 379 tar->compat_2x = (val != NULL && val[0] != 0); 380 tar->init_default_conversion = tar->compat_2x; 381 return (ARCHIVE_OK); 382 } else if (strcmp(key, "hdrcharset") == 0) { 383 if (val == NULL || val[0] == 0) 384 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 385 "tar: hdrcharset option needs a character-set name"); 386 else { 387 tar->opt_sconv = 388 archive_string_conversion_from_charset( 389 &a->archive, val, 0); 390 if (tar->opt_sconv != NULL) 391 ret = ARCHIVE_OK; 392 else 393 ret = ARCHIVE_FATAL; 394 } 395 return (ret); 396 } else if (strcmp(key, "mac-ext") == 0) { 397 tar->process_mac_extensions = (val != NULL && val[0] != 0); 398 return (ARCHIVE_OK); 399 } else if (strcmp(key, "read_concatenated_archives") == 0) { 400 tar->read_concatenated_archives = (val != NULL && val[0] != 0); 401 return (ARCHIVE_OK); 402 } 403 404 /* Note: The "warn" return is just to inform the options 405 * supervisor that we didn't handle it. It will generate 406 * a suitable error if no one used this option. */ 407 return (ARCHIVE_WARN); 408} 409 410/* utility function- this exists to centralize the logic of tracking 411 * how much unconsumed data we have floating around, and to consume 412 * anything outstanding since we're going to do read_aheads 413 */ 414static void 415tar_flush_unconsumed(struct archive_read *a, size_t *unconsumed) 416{ 417 if (*unconsumed) { 418/* 419 void *data = (void *)__archive_read_ahead(a, *unconsumed, NULL); 420 * this block of code is to poison claimed unconsumed space, ensuring 421 * things break if it is in use still. 422 * currently it WILL break things, so enable it only for debugging this issue 423 if (data) { 424 memset(data, 0xff, *unconsumed); 425 } 426*/ 427 __archive_read_consume(a, *unconsumed); 428 *unconsumed = 0; 429 } 430} 431 432/* 433 * The function invoked by archive_read_next_header(). This 434 * just sets up a few things and then calls the internal 435 * tar_read_header() function below. 436 */ 437static int 438archive_read_format_tar_read_header(struct archive_read *a, 439 struct archive_entry *entry) 440{ 441 /* 442 * When converting tar archives to cpio archives, it is 443 * essential that each distinct file have a distinct inode 444 * number. To simplify this, we keep a static count here to 445 * assign fake dev/inode numbers to each tar entry. Note that 446 * pax format archives may overwrite this with something more 447 * useful. 448 * 449 * Ideally, we would track every file read from the archive so 450 * that we could assign the same dev/ino pair to hardlinks, 451 * but the memory required to store a complete lookup table is 452 * probably not worthwhile just to support the relatively 453 * obscure tar->cpio conversion case. 454 */ 455 static int default_inode; 456 static int default_dev; 457 struct tar *tar; 458 const char *p; 459 const wchar_t *wp; 460 int r; 461 size_t l, unconsumed = 0; 462 463 /* Assign default device/inode values. */ 464 archive_entry_set_dev(entry, 1 + default_dev); /* Don't use zero. */ 465 archive_entry_set_ino(entry, ++default_inode); /* Don't use zero. */ 466 /* Limit generated st_ino number to 16 bits. */ 467 if (default_inode >= 0xffff) { 468 ++default_dev; 469 default_inode = 0; 470 } 471 472 tar = (struct tar *)(a->format->data); 473 tar->entry_offset = 0; 474 gnu_clear_sparse_list(tar); 475 tar->realsize = -1; /* Mark this as "unset" */ 476 477 /* Setup default string conversion. */ 478 tar->sconv = tar->opt_sconv; 479 if (tar->sconv == NULL) { 480 if (!tar->init_default_conversion) { 481 tar->sconv_default = 482 archive_string_default_conversion_for_read(&(a->archive)); 483 tar->init_default_conversion = 1; 484 } 485 tar->sconv = tar->sconv_default; 486 } 487 488 r = tar_read_header(a, tar, entry, &unconsumed); 489 490 tar_flush_unconsumed(a, &unconsumed); 491 492 /* 493 * "non-sparse" files are really just sparse files with 494 * a single block. 495 */ 496 if (tar->sparse_list == NULL) { 497 if (gnu_add_sparse_entry(a, tar, 0, tar->entry_bytes_remaining) 498 != ARCHIVE_OK) 499 return (ARCHIVE_FATAL); 500 } else { 501 struct sparse_block *sb; 502 503 for (sb = tar->sparse_list; sb != NULL; sb = sb->next) { 504 if (!sb->hole) 505 archive_entry_sparse_add_entry(entry, 506 sb->offset, sb->remaining); 507 } 508 } 509 510 if (r == ARCHIVE_OK && archive_entry_filetype(entry) == AE_IFREG) { 511 /* 512 * "Regular" entry with trailing '/' is really 513 * directory: This is needed for certain old tar 514 * variants and even for some broken newer ones. 515 */ 516 if ((wp = archive_entry_pathname_w(entry)) != NULL) { 517 l = wcslen(wp); 518 if (l > 0 && wp[l - 1] == L'/') { 519 archive_entry_set_filetype(entry, AE_IFDIR); 520 } 521 } else if ((p = archive_entry_pathname(entry)) != NULL) { 522 l = strlen(p); 523 if (l > 0 && p[l - 1] == '/') { 524 archive_entry_set_filetype(entry, AE_IFDIR); 525 } 526 } 527 } 528 return (r); 529} 530 531static int 532archive_read_format_tar_read_data(struct archive_read *a, 533 const void **buff, size_t *size, int64_t *offset) 534{ 535 ssize_t bytes_read; 536 struct tar *tar; 537 struct sparse_block *p; 538 539 tar = (struct tar *)(a->format->data); 540 541 for (;;) { 542 /* Remove exhausted entries from sparse list. */ 543 while (tar->sparse_list != NULL && 544 tar->sparse_list->remaining == 0) { 545 p = tar->sparse_list; 546 tar->sparse_list = p->next; 547 free(p); 548 } 549 550 if (tar->entry_bytes_unconsumed) { 551 __archive_read_consume(a, tar->entry_bytes_unconsumed); 552 tar->entry_bytes_unconsumed = 0; 553 } 554 555 /* If we're at end of file, return EOF. */ 556 if (tar->sparse_list == NULL || 557 tar->entry_bytes_remaining == 0) { 558 if (__archive_read_consume(a, tar->entry_padding) < 0) 559 return (ARCHIVE_FATAL); 560 tar->entry_padding = 0; 561 *buff = NULL; 562 *size = 0; 563 *offset = tar->realsize; 564 return (ARCHIVE_EOF); 565 } 566 567 *buff = __archive_read_ahead(a, 1, &bytes_read); 568 if (bytes_read < 0) 569 return (ARCHIVE_FATAL); 570 if (*buff == NULL) { 571 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 572 "Truncated tar archive"); 573 return (ARCHIVE_FATAL); 574 } 575 if (bytes_read > tar->entry_bytes_remaining) 576 bytes_read = (ssize_t)tar->entry_bytes_remaining; 577 /* Don't read more than is available in the 578 * current sparse block. */ 579 if (tar->sparse_list->remaining < bytes_read) 580 bytes_read = (ssize_t)tar->sparse_list->remaining; 581 *size = bytes_read; 582 *offset = tar->sparse_list->offset; 583 tar->sparse_list->remaining -= bytes_read; 584 tar->sparse_list->offset += bytes_read; 585 tar->entry_bytes_remaining -= bytes_read; 586 tar->entry_bytes_unconsumed = bytes_read; 587 588 if (!tar->sparse_list->hole) 589 return (ARCHIVE_OK); 590 /* Current is hole data and skip this. */ 591 } 592} 593 594static int 595archive_read_format_tar_skip(struct archive_read *a) 596{ 597 int64_t bytes_skipped; 598 int64_t request; 599 struct sparse_block *p; 600 struct tar* tar; 601 602 tar = (struct tar *)(a->format->data); 603 604 /* Do not consume the hole of a sparse file. */ 605 request = 0; 606 for (p = tar->sparse_list; p != NULL; p = p->next) { 607 if (!p->hole) { 608 if (p->remaining >= INT64_MAX - request) { 609 return ARCHIVE_FATAL; 610 } 611 request += p->remaining; 612 } 613 } 614 if (request > tar->entry_bytes_remaining) 615 request = tar->entry_bytes_remaining; 616 request += tar->entry_padding + tar->entry_bytes_unconsumed; 617 618 bytes_skipped = __archive_read_consume(a, request); 619 if (bytes_skipped < 0) 620 return (ARCHIVE_FATAL); 621 622 tar->entry_bytes_remaining = 0; 623 tar->entry_bytes_unconsumed = 0; 624 tar->entry_padding = 0; 625 626 /* Free the sparse list. */ 627 gnu_clear_sparse_list(tar); 628 629 return (ARCHIVE_OK); 630} 631 632/* 633 * This function recursively interprets all of the headers associated 634 * with a single entry. 635 */ 636static int 637tar_read_header(struct archive_read *a, struct tar *tar, 638 struct archive_entry *entry, size_t *unconsumed) 639{ 640 ssize_t bytes; 641 int err; 642 const char *h; 643 const struct archive_entry_header_ustar *header; 644 const struct archive_entry_header_gnutar *gnuheader; 645 646 /* Loop until we find a workable header record. */ 647 for (;;) { 648 tar_flush_unconsumed(a, unconsumed); 649 650 /* Read 512-byte header record */ 651 h = __archive_read_ahead(a, 512, &bytes); 652 if (bytes < 0) 653 return ((int)bytes); 654 if (bytes == 0) { /* EOF at a block boundary. */ 655 /* Some writers do omit the block of nulls. <sigh> */ 656 return (ARCHIVE_EOF); 657 } 658 if (bytes < 512) { /* Short block at EOF; this is bad. */ 659 archive_set_error(&a->archive, 660 ARCHIVE_ERRNO_FILE_FORMAT, 661 "Truncated tar archive"); 662 return (ARCHIVE_FATAL); 663 } 664 *unconsumed = 512; 665 666 /* Header is workable if it's not an end-of-archive mark. */ 667 if (h[0] != 0 || !archive_block_is_null(h)) 668 break; 669 670 /* Ensure format is set for archives with only null blocks. */ 671 if (a->archive.archive_format_name == NULL) { 672 a->archive.archive_format = ARCHIVE_FORMAT_TAR; 673 a->archive.archive_format_name = "tar"; 674 } 675 676 if (!tar->read_concatenated_archives) { 677 /* Try to consume a second all-null record, as well. */ 678 tar_flush_unconsumed(a, unconsumed); 679 h = __archive_read_ahead(a, 512, NULL); 680 if (h != NULL && h[0] == 0 && archive_block_is_null(h)) 681 __archive_read_consume(a, 512); 682 archive_clear_error(&a->archive); 683 return (ARCHIVE_EOF); 684 } 685 686 /* 687 * We're reading concatenated archives, ignore this block and 688 * loop to get the next. 689 */ 690 } 691 692 /* 693 * Note: If the checksum fails and we return ARCHIVE_RETRY, 694 * then the client is likely to just retry. This is a very 695 * crude way to search for the next valid header! 696 * 697 * TODO: Improve this by implementing a real header scan. 698 */ 699 if (!checksum(a, h)) { 700 tar_flush_unconsumed(a, unconsumed); 701 archive_set_error(&a->archive, EINVAL, "Damaged tar archive"); 702 return (ARCHIVE_RETRY); /* Retryable: Invalid header */ 703 } 704 705 if (++tar->header_recursion_depth > 32) { 706 tar_flush_unconsumed(a, unconsumed); 707 archive_set_error(&a->archive, EINVAL, "Too many special headers"); 708 return (ARCHIVE_WARN); 709 } 710 711 /* Determine the format variant. */ 712 header = (const struct archive_entry_header_ustar *)h; 713 714 switch(header->typeflag[0]) { 715 case 'A': /* Solaris tar ACL */ 716 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 717 a->archive.archive_format_name = "Solaris tar"; 718 err = header_Solaris_ACL(a, tar, entry, h, unconsumed); 719 break; 720 case 'g': /* POSIX-standard 'g' header. */ 721 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 722 a->archive.archive_format_name = "POSIX pax interchange format"; 723 err = header_pax_global(a, tar, entry, h, unconsumed); 724 if (err == ARCHIVE_EOF) 725 return (err); 726 break; 727 case 'K': /* Long link name (GNU tar, others) */ 728 err = header_longlink(a, tar, entry, h, unconsumed); 729 break; 730 case 'L': /* Long filename (GNU tar, others) */ 731 err = header_longname(a, tar, entry, h, unconsumed); 732 break; 733 case 'V': /* GNU volume header */ 734 err = header_volume(a, tar, entry, h, unconsumed); 735 break; 736 case 'X': /* Used by SUN tar; same as 'x'. */ 737 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 738 a->archive.archive_format_name = 739 "POSIX pax interchange format (Sun variant)"; 740 err = header_pax_extensions(a, tar, entry, h, unconsumed); 741 break; 742 case 'x': /* POSIX-standard 'x' header. */ 743 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 744 a->archive.archive_format_name = "POSIX pax interchange format"; 745 err = header_pax_extensions(a, tar, entry, h, unconsumed); 746 break; 747 default: 748 gnuheader = (const struct archive_entry_header_gnutar *)h; 749 if (memcmp(gnuheader->magic, "ustar \0", 8) == 0) { 750 a->archive.archive_format = ARCHIVE_FORMAT_TAR_GNUTAR; 751 a->archive.archive_format_name = "GNU tar format"; 752 err = header_gnutar(a, tar, entry, h, unconsumed); 753 } else if (memcmp(header->magic, "ustar", 5) == 0) { 754 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) { 755 a->archive.archive_format = ARCHIVE_FORMAT_TAR_USTAR; 756 a->archive.archive_format_name = "POSIX ustar format"; 757 } 758 err = header_ustar(a, tar, entry, h); 759 } else { 760 a->archive.archive_format = ARCHIVE_FORMAT_TAR; 761 a->archive.archive_format_name = "tar (non-POSIX)"; 762 err = header_old_tar(a, tar, entry, h); 763 } 764 } 765 if (err == ARCHIVE_FATAL) 766 return (err); 767 768 tar_flush_unconsumed(a, unconsumed); 769 770 h = NULL; 771 header = NULL; 772 773 --tar->header_recursion_depth; 774 /* Yuck. Apple's design here ends up storing long pathname 775 * extensions for both the AppleDouble extension entry and the 776 * regular entry. 777 */ 778 if ((err == ARCHIVE_WARN || err == ARCHIVE_OK) && 779 tar->header_recursion_depth == 0 && 780 tar->process_mac_extensions) { 781 int err2 = read_mac_metadata_blob(a, tar, entry, h, unconsumed); 782 if (err2 < err) 783 err = err2; 784 } 785 786 /* We return warnings or success as-is. Anything else is fatal. */ 787 if (err == ARCHIVE_WARN || err == ARCHIVE_OK) { 788 if (tar->sparse_gnu_pending) { 789 if (tar->sparse_gnu_major == 1 && 790 tar->sparse_gnu_minor == 0) { 791 ssize_t bytes_read; 792 793 tar->sparse_gnu_pending = 0; 794 /* Read initial sparse map. */ 795 bytes_read = gnu_sparse_10_read(a, tar, unconsumed); 796 tar->entry_bytes_remaining -= bytes_read; 797 if (bytes_read < 0) 798 return ((int)bytes_read); 799 } else { 800 archive_set_error(&a->archive, 801 ARCHIVE_ERRNO_MISC, 802 "Unrecognized GNU sparse file format"); 803 return (ARCHIVE_WARN); 804 } 805 tar->sparse_gnu_pending = 0; 806 } 807 return (err); 808 } 809 if (err == ARCHIVE_EOF) 810 /* EOF when recursively reading a header is bad. */ 811 archive_set_error(&a->archive, EINVAL, "Damaged tar archive"); 812 return (ARCHIVE_FATAL); 813} 814 815/* 816 * Return true if block checksum is correct. 817 */ 818static int 819checksum(struct archive_read *a, const void *h) 820{ 821 const unsigned char *bytes; 822 const struct archive_entry_header_ustar *header; 823 int check, sum; 824 size_t i; 825 826 (void)a; /* UNUSED */ 827 bytes = (const unsigned char *)h; 828 header = (const struct archive_entry_header_ustar *)h; 829 830 /* Checksum field must hold an octal number */ 831 for (i = 0; i < sizeof(header->checksum); ++i) { 832 char c = header->checksum[i]; 833 if (c != ' ' && c != '\0' && (c < '0' || c > '7')) 834 return 0; 835 } 836 837 /* 838 * Test the checksum. Note that POSIX specifies _unsigned_ 839 * bytes for this calculation. 840 */ 841 sum = (int)tar_atol(header->checksum, sizeof(header->checksum)); 842 check = 0; 843 for (i = 0; i < 148; i++) 844 check += (unsigned char)bytes[i]; 845 for (; i < 156; i++) 846 check += 32; 847 for (; i < 512; i++) 848 check += (unsigned char)bytes[i]; 849 if (sum == check) 850 return (1); 851 852 /* 853 * Repeat test with _signed_ bytes, just in case this archive 854 * was created by an old BSD, Solaris, or HP-UX tar with a 855 * broken checksum calculation. 856 */ 857 check = 0; 858 for (i = 0; i < 148; i++) 859 check += (signed char)bytes[i]; 860 for (; i < 156; i++) 861 check += 32; 862 for (; i < 512; i++) 863 check += (signed char)bytes[i]; 864 if (sum == check) 865 return (1); 866 867 return (0); 868} 869 870/* 871 * Return true if this block contains only nulls. 872 */ 873static int 874archive_block_is_null(const char *p) 875{ 876 unsigned i; 877 878 for (i = 0; i < 512; i++) 879 if (*p++) 880 return (0); 881 return (1); 882} 883 884/* 885 * Interpret 'A' Solaris ACL header 886 */ 887static int 888header_Solaris_ACL(struct archive_read *a, struct tar *tar, 889 struct archive_entry *entry, const void *h, size_t *unconsumed) 890{ 891 const struct archive_entry_header_ustar *header; 892 size_t size; 893 int err; 894 int64_t type; 895 char *acl, *p; 896 897 /* 898 * read_body_to_string adds a NUL terminator, but we need a little 899 * more to make sure that we don't overrun acl_text later. 900 */ 901 header = (const struct archive_entry_header_ustar *)h; 902 size = (size_t)tar_atol(header->size, sizeof(header->size)); 903 err = read_body_to_string(a, tar, &(tar->acl_text), h, unconsumed); 904 if (err != ARCHIVE_OK) 905 return (err); 906 907 /* Recursively read next header */ 908 err = tar_read_header(a, tar, entry, unconsumed); 909 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 910 return (err); 911 912 /* TODO: Examine the first characters to see if this 913 * is an AIX ACL descriptor. We'll likely never support 914 * them, but it would be polite to recognize and warn when 915 * we do see them. */ 916 917 /* Leading octal number indicates ACL type and number of entries. */ 918 p = acl = tar->acl_text.s; 919 type = 0; 920 while (*p != '\0' && p < acl + size) { 921 if (*p < '0' || *p > '7') { 922 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 923 "Malformed Solaris ACL attribute (invalid digit)"); 924 return(ARCHIVE_WARN); 925 } 926 type <<= 3; 927 type += *p - '0'; 928 if (type > 077777777) { 929 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 930 "Malformed Solaris ACL attribute (count too large)"); 931 return (ARCHIVE_WARN); 932 } 933 p++; 934 } 935 switch ((int)type & ~0777777) { 936 case 01000000: 937 /* POSIX.1e ACL */ 938 break; 939 case 03000000: 940 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 941 "Solaris NFSv4 ACLs not supported"); 942 return (ARCHIVE_WARN); 943 default: 944 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 945 "Malformed Solaris ACL attribute (unsupported type %o)", 946 (int)type); 947 return (ARCHIVE_WARN); 948 } 949 p++; 950 951 if (p >= acl + size) { 952 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 953 "Malformed Solaris ACL attribute (body overflow)"); 954 return(ARCHIVE_WARN); 955 } 956 957 /* ACL text is null-terminated; find the end. */ 958 size -= (p - acl); 959 acl = p; 960 961 while (*p != '\0' && p < acl + size) 962 p++; 963 964 if (tar->sconv_acl == NULL) { 965 tar->sconv_acl = archive_string_conversion_from_charset( 966 &(a->archive), "UTF-8", 1); 967 if (tar->sconv_acl == NULL) 968 return (ARCHIVE_FATAL); 969 } 970 archive_strncpy(&(tar->localname), acl, p - acl); 971 err = archive_acl_parse_l(archive_entry_acl(entry), 972 tar->localname.s, ARCHIVE_ENTRY_ACL_TYPE_ACCESS, tar->sconv_acl); 973 if (err != ARCHIVE_OK) { 974 if (errno == ENOMEM) { 975 archive_set_error(&a->archive, ENOMEM, 976 "Can't allocate memory for ACL"); 977 } else 978 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 979 "Malformed Solaris ACL attribute (unparsable)"); 980 } 981 return (err); 982} 983 984/* 985 * Interpret 'K' long linkname header. 986 */ 987static int 988header_longlink(struct archive_read *a, struct tar *tar, 989 struct archive_entry *entry, const void *h, size_t *unconsumed) 990{ 991 int err; 992 993 err = read_body_to_string(a, tar, &(tar->longlink), h, unconsumed); 994 if (err != ARCHIVE_OK) 995 return (err); 996 err = tar_read_header(a, tar, entry, unconsumed); 997 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 998 return (err); 999 /* Set symlink if symlink already set, else hardlink. */ 1000 archive_entry_copy_link(entry, tar->longlink.s); 1001 return (ARCHIVE_OK); 1002} 1003 1004static int 1005set_conversion_failed_error(struct archive_read *a, 1006 struct archive_string_conv *sconv, const char *name) 1007{ 1008 if (errno == ENOMEM) { 1009 archive_set_error(&a->archive, ENOMEM, 1010 "Can't allocate memory for %s", name); 1011 return (ARCHIVE_FATAL); 1012 } 1013 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 1014 "%s can't be converted from %s to current locale.", 1015 name, archive_string_conversion_charset_name(sconv)); 1016 return (ARCHIVE_WARN); 1017} 1018 1019/* 1020 * Interpret 'L' long filename header. 1021 */ 1022static int 1023header_longname(struct archive_read *a, struct tar *tar, 1024 struct archive_entry *entry, const void *h, size_t *unconsumed) 1025{ 1026 int err; 1027 1028 err = read_body_to_string(a, tar, &(tar->longname), h, unconsumed); 1029 if (err != ARCHIVE_OK) 1030 return (err); 1031 /* Read and parse "real" header, then override name. */ 1032 err = tar_read_header(a, tar, entry, unconsumed); 1033 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 1034 return (err); 1035 if (archive_entry_copy_pathname_l(entry, tar->longname.s, 1036 archive_strlen(&(tar->longname)), tar->sconv) != 0) 1037 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 1038 return (err); 1039} 1040 1041 1042/* 1043 * Interpret 'V' GNU tar volume header. 1044 */ 1045static int 1046header_volume(struct archive_read *a, struct tar *tar, 1047 struct archive_entry *entry, const void *h, size_t *unconsumed) 1048{ 1049 (void)h; 1050 1051 /* Just skip this and read the next header. */ 1052 return (tar_read_header(a, tar, entry, unconsumed)); 1053} 1054 1055/* 1056 * Read body of an archive entry into an archive_string object. 1057 */ 1058static int 1059read_body_to_string(struct archive_read *a, struct tar *tar, 1060 struct archive_string *as, const void *h, size_t *unconsumed) 1061{ 1062 int64_t size; 1063 const struct archive_entry_header_ustar *header; 1064 const void *src; 1065 1066 (void)tar; /* UNUSED */ 1067 header = (const struct archive_entry_header_ustar *)h; 1068 size = tar_atol(header->size, sizeof(header->size)); 1069 if ((size > 1048576) || (size < 0)) { 1070 archive_set_error(&a->archive, EINVAL, 1071 "Special header too large"); 1072 return (ARCHIVE_FATAL); 1073 } 1074 1075 /* Fail if we can't make our buffer big enough. */ 1076 if (archive_string_ensure(as, (size_t)size+1) == NULL) { 1077 archive_set_error(&a->archive, ENOMEM, 1078 "No memory"); 1079 return (ARCHIVE_FATAL); 1080 } 1081 1082 tar_flush_unconsumed(a, unconsumed); 1083 1084 /* Read the body into the string. */ 1085 *unconsumed = (size_t)((size + 511) & ~ 511); 1086 src = __archive_read_ahead(a, *unconsumed, NULL); 1087 if (src == NULL) { 1088 *unconsumed = 0; 1089 return (ARCHIVE_FATAL); 1090 } 1091 memcpy(as->s, src, (size_t)size); 1092 as->s[size] = '\0'; 1093 as->length = (size_t)size; 1094 return (ARCHIVE_OK); 1095} 1096 1097/* 1098 * Parse out common header elements. 1099 * 1100 * This would be the same as header_old_tar, except that the 1101 * filename is handled slightly differently for old and POSIX 1102 * entries (POSIX entries support a 'prefix'). This factoring 1103 * allows header_old_tar and header_ustar 1104 * to handle filenames differently, while still putting most of the 1105 * common parsing into one place. 1106 */ 1107static int 1108header_common(struct archive_read *a, struct tar *tar, 1109 struct archive_entry *entry, const void *h) 1110{ 1111 const struct archive_entry_header_ustar *header; 1112 char tartype; 1113 int err = ARCHIVE_OK; 1114 1115 header = (const struct archive_entry_header_ustar *)h; 1116 if (header->linkname[0]) 1117 archive_strncpy(&(tar->entry_linkpath), 1118 header->linkname, sizeof(header->linkname)); 1119 else 1120 archive_string_empty(&(tar->entry_linkpath)); 1121 1122 /* Parse out the numeric fields (all are octal) */ 1123 archive_entry_set_mode(entry, 1124 (mode_t)tar_atol(header->mode, sizeof(header->mode))); 1125 archive_entry_set_uid(entry, tar_atol(header->uid, sizeof(header->uid))); 1126 archive_entry_set_gid(entry, tar_atol(header->gid, sizeof(header->gid))); 1127 tar->entry_bytes_remaining = tar_atol(header->size, sizeof(header->size)); 1128 if (tar->entry_bytes_remaining < 0) { 1129 tar->entry_bytes_remaining = 0; 1130 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1131 "Tar entry has negative size"); 1132 return (ARCHIVE_FATAL); 1133 } 1134 if (tar->entry_bytes_remaining == INT64_MAX) { 1135 /* Note: tar_atol returns INT64_MAX on overflow */ 1136 tar->entry_bytes_remaining = 0; 1137 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1138 "Tar entry size overflow"); 1139 return (ARCHIVE_FATAL); 1140 } 1141 tar->realsize = tar->entry_bytes_remaining; 1142 archive_entry_set_size(entry, tar->entry_bytes_remaining); 1143 archive_entry_set_mtime(entry, tar_atol(header->mtime, sizeof(header->mtime)), 0); 1144 1145 /* Handle the tar type flag appropriately. */ 1146 tartype = header->typeflag[0]; 1147 1148 switch (tartype) { 1149 case '1': /* Hard link */ 1150 if (archive_entry_copy_hardlink_l(entry, tar->entry_linkpath.s, 1151 archive_strlen(&(tar->entry_linkpath)), tar->sconv) != 0) { 1152 err = set_conversion_failed_error(a, tar->sconv, 1153 "Linkname"); 1154 if (err == ARCHIVE_FATAL) 1155 return (err); 1156 } 1157 /* 1158 * The following may seem odd, but: Technically, tar 1159 * does not store the file type for a "hard link" 1160 * entry, only the fact that it is a hard link. So, I 1161 * leave the type zero normally. But, pax interchange 1162 * format allows hard links to have data, which 1163 * implies that the underlying entry is a regular 1164 * file. 1165 */ 1166 if (archive_entry_size(entry) > 0) 1167 archive_entry_set_filetype(entry, AE_IFREG); 1168 1169 /* 1170 * A tricky point: Traditionally, tar readers have 1171 * ignored the size field when reading hardlink 1172 * entries, and some writers put non-zero sizes even 1173 * though the body is empty. POSIX blessed this 1174 * convention in the 1988 standard, but broke with 1175 * this tradition in 2001 by permitting hardlink 1176 * entries to store valid bodies in pax interchange 1177 * format, but not in ustar format. Since there is no 1178 * hard and fast way to distinguish pax interchange 1179 * from earlier archives (the 'x' and 'g' entries are 1180 * optional, after all), we need a heuristic. 1181 */ 1182 if (archive_entry_size(entry) == 0) { 1183 /* If the size is already zero, we're done. */ 1184 } else if (a->archive.archive_format 1185 == ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) { 1186 /* Definitely pax extended; must obey hardlink size. */ 1187 } else if (a->archive.archive_format == ARCHIVE_FORMAT_TAR 1188 || a->archive.archive_format == ARCHIVE_FORMAT_TAR_GNUTAR) 1189 { 1190 /* Old-style or GNU tar: we must ignore the size. */ 1191 archive_entry_set_size(entry, 0); 1192 tar->entry_bytes_remaining = 0; 1193 } else if (archive_read_format_tar_bid(a, 50) > 50) { 1194 /* 1195 * We don't know if it's pax: If the bid 1196 * function sees a valid ustar header 1197 * immediately following, then let's ignore 1198 * the hardlink size. 1199 */ 1200 archive_entry_set_size(entry, 0); 1201 tar->entry_bytes_remaining = 0; 1202 } 1203 /* 1204 * TODO: There are still two cases I'd like to handle: 1205 * = a ustar non-pax archive with a hardlink entry at 1206 * end-of-archive. (Look for block of nulls following?) 1207 * = a pax archive that has not seen any pax headers 1208 * and has an entry which is a hardlink entry storing 1209 * a body containing an uncompressed tar archive. 1210 * The first is worth addressing; I don't see any reliable 1211 * way to deal with the second possibility. 1212 */ 1213 break; 1214 case '2': /* Symlink */ 1215 archive_entry_set_filetype(entry, AE_IFLNK); 1216 archive_entry_set_size(entry, 0); 1217 tar->entry_bytes_remaining = 0; 1218 if (archive_entry_copy_symlink_l(entry, tar->entry_linkpath.s, 1219 archive_strlen(&(tar->entry_linkpath)), tar->sconv) != 0) { 1220 err = set_conversion_failed_error(a, tar->sconv, 1221 "Linkname"); 1222 if (err == ARCHIVE_FATAL) 1223 return (err); 1224 } 1225 break; 1226 case '3': /* Character device */ 1227 archive_entry_set_filetype(entry, AE_IFCHR); 1228 archive_entry_set_size(entry, 0); 1229 tar->entry_bytes_remaining = 0; 1230 break; 1231 case '4': /* Block device */ 1232 archive_entry_set_filetype(entry, AE_IFBLK); 1233 archive_entry_set_size(entry, 0); 1234 tar->entry_bytes_remaining = 0; 1235 break; 1236 case '5': /* Dir */ 1237 archive_entry_set_filetype(entry, AE_IFDIR); 1238 archive_entry_set_size(entry, 0); 1239 tar->entry_bytes_remaining = 0; 1240 break; 1241 case '6': /* FIFO device */ 1242 archive_entry_set_filetype(entry, AE_IFIFO); 1243 archive_entry_set_size(entry, 0); 1244 tar->entry_bytes_remaining = 0; 1245 break; 1246 case 'D': /* GNU incremental directory type */ 1247 /* 1248 * No special handling is actually required here. 1249 * It might be nice someday to preprocess the file list and 1250 * provide it to the client, though. 1251 */ 1252 archive_entry_set_filetype(entry, AE_IFDIR); 1253 break; 1254 case 'M': /* GNU "Multi-volume" (remainder of file from last archive)*/ 1255 /* 1256 * As far as I can tell, this is just like a regular file 1257 * entry, except that the contents should be _appended_ to 1258 * the indicated file at the indicated offset. This may 1259 * require some API work to fully support. 1260 */ 1261 break; 1262 case 'N': /* Old GNU "long filename" entry. */ 1263 /* The body of this entry is a script for renaming 1264 * previously-extracted entries. Ugh. It will never 1265 * be supported by libarchive. */ 1266 archive_entry_set_filetype(entry, AE_IFREG); 1267 break; 1268 case 'S': /* GNU sparse files */ 1269 /* 1270 * Sparse files are really just regular files with 1271 * sparse information in the extended area. 1272 */ 1273 /* FALLTHROUGH */ 1274 default: /* Regular file and non-standard types */ 1275 /* 1276 * Per POSIX: non-recognized types should always be 1277 * treated as regular files. 1278 */ 1279 archive_entry_set_filetype(entry, AE_IFREG); 1280 break; 1281 } 1282 return (err); 1283} 1284 1285/* 1286 * Parse out header elements for "old-style" tar archives. 1287 */ 1288static int 1289header_old_tar(struct archive_read *a, struct tar *tar, 1290 struct archive_entry *entry, const void *h) 1291{ 1292 const struct archive_entry_header_ustar *header; 1293 int err = ARCHIVE_OK, err2; 1294 1295 /* Copy filename over (to ensure null termination). */ 1296 header = (const struct archive_entry_header_ustar *)h; 1297 if (archive_entry_copy_pathname_l(entry, 1298 header->name, sizeof(header->name), tar->sconv) != 0) { 1299 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 1300 if (err == ARCHIVE_FATAL) 1301 return (err); 1302 } 1303 1304 /* Grab rest of common fields */ 1305 err2 = header_common(a, tar, entry, h); 1306 if (err > err2) 1307 err = err2; 1308 1309 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 1310 return (err); 1311} 1312 1313/* 1314 * Read a Mac AppleDouble-encoded blob of file metadata, 1315 * if there is one. 1316 */ 1317static int 1318read_mac_metadata_blob(struct archive_read *a, struct tar *tar, 1319 struct archive_entry *entry, const void *h, size_t *unconsumed) 1320{ 1321 int64_t size; 1322 const void *data; 1323 const char *p, *name; 1324 const wchar_t *wp, *wname; 1325 1326 (void)h; /* UNUSED */ 1327 1328 wname = wp = archive_entry_pathname_w(entry); 1329 if (wp != NULL) { 1330 /* Find the last path element. */ 1331 for (; *wp != L'\0'; ++wp) { 1332 if (wp[0] == '/' && wp[1] != L'\0') 1333 wname = wp + 1; 1334 } 1335 /* 1336 * If last path element starts with "._", then 1337 * this is a Mac extension. 1338 */ 1339 if (wname[0] != L'.' || wname[1] != L'_' || wname[2] == L'\0') 1340 return ARCHIVE_OK; 1341 } else { 1342 /* Find the last path element. */ 1343 name = p = archive_entry_pathname(entry); 1344 if (p == NULL) 1345 return (ARCHIVE_FAILED); 1346 for (; *p != '\0'; ++p) { 1347 if (p[0] == '/' && p[1] != '\0') 1348 name = p + 1; 1349 } 1350 /* 1351 * If last path element starts with "._", then 1352 * this is a Mac extension. 1353 */ 1354 if (name[0] != '.' || name[1] != '_' || name[2] == '\0') 1355 return ARCHIVE_OK; 1356 } 1357 1358 /* Read the body as a Mac OS metadata blob. */ 1359 size = archive_entry_size(entry); 1360 1361 /* 1362 * TODO: Look beyond the body here to peek at the next header. 1363 * If it's a regular header (not an extension header) 1364 * that has the wrong name, just return the current 1365 * entry as-is, without consuming the body here. 1366 * That would reduce the risk of us mis-identifying 1367 * an ordinary file that just happened to have 1368 * a name starting with "._". 1369 * 1370 * Q: Is the above idea really possible? Even 1371 * when there are GNU or pax extension entries? 1372 */ 1373 data = __archive_read_ahead(a, (size_t)size, NULL); 1374 if (data == NULL) { 1375 *unconsumed = 0; 1376 return (ARCHIVE_FATAL); 1377 } 1378 archive_entry_copy_mac_metadata(entry, data, (size_t)size); 1379 *unconsumed = (size_t)((size + 511) & ~ 511); 1380 tar_flush_unconsumed(a, unconsumed); 1381 return (tar_read_header(a, tar, entry, unconsumed)); 1382} 1383 1384/* 1385 * Parse a file header for a pax extended archive entry. 1386 */ 1387static int 1388header_pax_global(struct archive_read *a, struct tar *tar, 1389 struct archive_entry *entry, const void *h, size_t *unconsumed) 1390{ 1391 int err; 1392 1393 err = read_body_to_string(a, tar, &(tar->pax_global), h, unconsumed); 1394 if (err != ARCHIVE_OK) 1395 return (err); 1396 err = tar_read_header(a, tar, entry, unconsumed); 1397 return (err); 1398} 1399 1400static int 1401header_pax_extensions(struct archive_read *a, struct tar *tar, 1402 struct archive_entry *entry, const void *h, size_t *unconsumed) 1403{ 1404 int err, err2; 1405 1406 err = read_body_to_string(a, tar, &(tar->pax_header), h, unconsumed); 1407 if (err != ARCHIVE_OK) 1408 return (err); 1409 1410 /* Parse the next header. */ 1411 err = tar_read_header(a, tar, entry, unconsumed); 1412 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 1413 return (err); 1414 1415 /* 1416 * TODO: Parse global/default options into 'entry' struct here 1417 * before handling file-specific options. 1418 * 1419 * This design (parse standard header, then overwrite with pax 1420 * extended attribute data) usually works well, but isn't ideal; 1421 * it would be better to parse the pax extended attributes first 1422 * and then skip any fields in the standard header that were 1423 * defined in the pax header. 1424 */ 1425 err2 = pax_header(a, tar, entry, tar->pax_header.s); 1426 err = err_combine(err, err2); 1427 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 1428 return (err); 1429} 1430 1431 1432/* 1433 * Parse a file header for a Posix "ustar" archive entry. This also 1434 * handles "pax" or "extended ustar" entries. 1435 */ 1436static int 1437header_ustar(struct archive_read *a, struct tar *tar, 1438 struct archive_entry *entry, const void *h) 1439{ 1440 const struct archive_entry_header_ustar *header; 1441 struct archive_string *as; 1442 int err = ARCHIVE_OK, r; 1443 1444 header = (const struct archive_entry_header_ustar *)h; 1445 1446 /* Copy name into an internal buffer to ensure null-termination. */ 1447 as = &(tar->entry_pathname); 1448 if (header->prefix[0]) { 1449 archive_strncpy(as, header->prefix, sizeof(header->prefix)); 1450 if (as->s[archive_strlen(as) - 1] != '/') 1451 archive_strappend_char(as, '/'); 1452 archive_strncat(as, header->name, sizeof(header->name)); 1453 } else { 1454 archive_strncpy(as, header->name, sizeof(header->name)); 1455 } 1456 if (archive_entry_copy_pathname_l(entry, as->s, archive_strlen(as), 1457 tar->sconv) != 0) { 1458 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 1459 if (err == ARCHIVE_FATAL) 1460 return (err); 1461 } 1462 1463 /* Handle rest of common fields. */ 1464 r = header_common(a, tar, entry, h); 1465 if (r == ARCHIVE_FATAL) 1466 return (r); 1467 if (r < err) 1468 err = r; 1469 1470 /* Handle POSIX ustar fields. */ 1471 if (archive_entry_copy_uname_l(entry, 1472 header->uname, sizeof(header->uname), tar->sconv) != 0) { 1473 err = set_conversion_failed_error(a, tar->sconv, "Uname"); 1474 if (err == ARCHIVE_FATAL) 1475 return (err); 1476 } 1477 1478 if (archive_entry_copy_gname_l(entry, 1479 header->gname, sizeof(header->gname), tar->sconv) != 0) { 1480 err = set_conversion_failed_error(a, tar->sconv, "Gname"); 1481 if (err == ARCHIVE_FATAL) 1482 return (err); 1483 } 1484 1485 /* Parse out device numbers only for char and block specials. */ 1486 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') { 1487 archive_entry_set_rdevmajor(entry, (dev_t) 1488 tar_atol(header->rdevmajor, sizeof(header->rdevmajor))); 1489 archive_entry_set_rdevminor(entry, (dev_t) 1490 tar_atol(header->rdevminor, sizeof(header->rdevminor))); 1491 } 1492 1493 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 1494 1495 return (err); 1496} 1497 1498 1499/* 1500 * Parse the pax extended attributes record. 1501 * 1502 * Returns non-zero if there's an error in the data. 1503 */ 1504static int 1505pax_header(struct archive_read *a, struct tar *tar, 1506 struct archive_entry *entry, char *attr) 1507{ 1508 size_t attr_length, l, line_length; 1509 char *p; 1510 char *key, *value; 1511 struct archive_string *as; 1512 struct archive_string_conv *sconv; 1513 int err, err2; 1514 1515 attr_length = strlen(attr); 1516 tar->pax_hdrcharset_binary = 0; 1517 archive_string_empty(&(tar->entry_gname)); 1518 archive_string_empty(&(tar->entry_linkpath)); 1519 archive_string_empty(&(tar->entry_pathname)); 1520 archive_string_empty(&(tar->entry_pathname_override)); 1521 archive_string_empty(&(tar->entry_uname)); 1522 err = ARCHIVE_OK; 1523 while (attr_length > 0) { 1524 /* Parse decimal length field at start of line. */ 1525 line_length = 0; 1526 l = attr_length; 1527 p = attr; /* Record start of line. */ 1528 while (l>0) { 1529 if (*p == ' ') { 1530 p++; 1531 l--; 1532 break; 1533 } 1534 if (*p < '0' || *p > '9') { 1535 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1536 "Ignoring malformed pax extended attributes"); 1537 return (ARCHIVE_WARN); 1538 } 1539 line_length *= 10; 1540 line_length += *p - '0'; 1541 if (line_length > 999999) { 1542 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1543 "Rejecting pax extended attribute > 1MB"); 1544 return (ARCHIVE_WARN); 1545 } 1546 p++; 1547 l--; 1548 } 1549 1550 /* 1551 * Parsed length must be no bigger than available data, 1552 * at least 1, and the last character of the line must 1553 * be '\n'. 1554 */ 1555 if (line_length > attr_length 1556 || line_length < 1 1557 || attr[line_length - 1] != '\n') 1558 { 1559 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1560 "Ignoring malformed pax extended attribute"); 1561 return (ARCHIVE_WARN); 1562 } 1563 1564 /* Null-terminate the line. */ 1565 attr[line_length - 1] = '\0'; 1566 1567 /* Find end of key and null terminate it. */ 1568 key = p; 1569 if (key[0] == '=') 1570 return (-1); 1571 while (*p && *p != '=') 1572 ++p; 1573 if (*p == '\0') { 1574 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1575 "Invalid pax extended attributes"); 1576 return (ARCHIVE_WARN); 1577 } 1578 *p = '\0'; 1579 1580 /* Identify null-terminated 'value' portion. */ 1581 value = p + 1; 1582 1583 /* Identify this attribute and set it in the entry. */ 1584 err2 = pax_attribute(a, tar, entry, key, value); 1585 if (err2 == ARCHIVE_FATAL) 1586 return (err2); 1587 err = err_combine(err, err2); 1588 1589 /* Skip to next line */ 1590 attr += line_length; 1591 attr_length -= line_length; 1592 } 1593 1594 /* 1595 * PAX format uses UTF-8 as default charset for its metadata 1596 * unless hdrcharset=BINARY is present in its header. 1597 * We apply the charset specified by the hdrcharset option only 1598 * when the hdrcharset attribute(in PAX header) is BINARY because 1599 * we respect the charset described in PAX header and BINARY also 1600 * means that metadata(filename,uname and gname) character-set 1601 * is unknown. 1602 */ 1603 if (tar->pax_hdrcharset_binary) 1604 sconv = tar->opt_sconv; 1605 else { 1606 sconv = archive_string_conversion_from_charset( 1607 &(a->archive), "UTF-8", 1); 1608 if (sconv == NULL) 1609 return (ARCHIVE_FATAL); 1610 if (tar->compat_2x) 1611 archive_string_conversion_set_opt(sconv, 1612 SCONV_SET_OPT_UTF8_LIBARCHIVE2X); 1613 } 1614 1615 if (archive_strlen(&(tar->entry_gname)) > 0) { 1616 if (archive_entry_copy_gname_l(entry, tar->entry_gname.s, 1617 archive_strlen(&(tar->entry_gname)), sconv) != 0) { 1618 err = set_conversion_failed_error(a, sconv, "Gname"); 1619 if (err == ARCHIVE_FATAL) 1620 return (err); 1621 /* Use a converted an original name. */ 1622 archive_entry_copy_gname(entry, tar->entry_gname.s); 1623 } 1624 } 1625 if (archive_strlen(&(tar->entry_linkpath)) > 0) { 1626 if (archive_entry_copy_link_l(entry, tar->entry_linkpath.s, 1627 archive_strlen(&(tar->entry_linkpath)), sconv) != 0) { 1628 err = set_conversion_failed_error(a, sconv, "Linkname"); 1629 if (err == ARCHIVE_FATAL) 1630 return (err); 1631 /* Use a converted an original name. */ 1632 archive_entry_copy_link(entry, tar->entry_linkpath.s); 1633 } 1634 } 1635 /* 1636 * Some extensions (such as the GNU sparse file extensions) 1637 * deliberately store a synthetic name under the regular 'path' 1638 * attribute and the real file name under a different attribute. 1639 * Since we're supposed to not care about the order, we 1640 * have no choice but to store all of the various filenames 1641 * we find and figure it all out afterwards. This is the 1642 * figuring out part. 1643 */ 1644 as = NULL; 1645 if (archive_strlen(&(tar->entry_pathname_override)) > 0) 1646 as = &(tar->entry_pathname_override); 1647 else if (archive_strlen(&(tar->entry_pathname)) > 0) 1648 as = &(tar->entry_pathname); 1649 if (as != NULL) { 1650 if (archive_entry_copy_pathname_l(entry, as->s, 1651 archive_strlen(as), sconv) != 0) { 1652 err = set_conversion_failed_error(a, sconv, "Pathname"); 1653 if (err == ARCHIVE_FATAL) 1654 return (err); 1655 /* Use a converted an original name. */ 1656 archive_entry_copy_pathname(entry, as->s); 1657 } 1658 } 1659 if (archive_strlen(&(tar->entry_uname)) > 0) { 1660 if (archive_entry_copy_uname_l(entry, tar->entry_uname.s, 1661 archive_strlen(&(tar->entry_uname)), sconv) != 0) { 1662 err = set_conversion_failed_error(a, sconv, "Uname"); 1663 if (err == ARCHIVE_FATAL) 1664 return (err); 1665 /* Use a converted an original name. */ 1666 archive_entry_copy_uname(entry, tar->entry_uname.s); 1667 } 1668 } 1669 return (err); 1670} 1671 1672static int 1673pax_attribute_xattr(struct archive_entry *entry, 1674 const char *name, const char *value) 1675{ 1676 char *name_decoded; 1677 void *value_decoded; 1678 size_t value_len; 1679 1680 if (strlen(name) < 18 || (memcmp(name, "LIBARCHIVE.xattr.", 17)) != 0) 1681 return 3; 1682 1683 name += 17; 1684 1685 /* URL-decode name */ 1686 name_decoded = url_decode(name); 1687 if (name_decoded == NULL) 1688 return 2; 1689 1690 /* Base-64 decode value */ 1691 value_decoded = base64_decode(value, strlen(value), &value_len); 1692 if (value_decoded == NULL) { 1693 free(name_decoded); 1694 return 1; 1695 } 1696 1697 archive_entry_xattr_add_entry(entry, name_decoded, 1698 value_decoded, value_len); 1699 1700 free(name_decoded); 1701 free(value_decoded); 1702 return 0; 1703} 1704 1705/* 1706 * Parse a single key=value attribute. key/value pointers are 1707 * assumed to point into reasonably long-lived storage. 1708 * 1709 * Note that POSIX reserves all-lowercase keywords. Vendor-specific 1710 * extensions should always have keywords of the form "VENDOR.attribute" 1711 * In particular, it's quite feasible to support many different 1712 * vendor extensions here. I'm using "LIBARCHIVE" for extensions 1713 * unique to this library. 1714 * 1715 * Investigate other vendor-specific extensions and see if 1716 * any of them look useful. 1717 */ 1718static int 1719pax_attribute(struct archive_read *a, struct tar *tar, 1720 struct archive_entry *entry, const char *key, const char *value) 1721{ 1722 int64_t s; 1723 long n; 1724 int err = ARCHIVE_OK, r; 1725 1726#ifndef __FreeBSD__ 1727 if (value == NULL) 1728 value = ""; /* Disable compiler warning; do not pass 1729 * NULL pointer to strlen(). */ 1730#endif 1731 switch (key[0]) { 1732 case 'G': 1733 /* GNU "0.0" sparse pax format. */ 1734 if (strcmp(key, "GNU.sparse.numblocks") == 0) { 1735 tar->sparse_offset = -1; 1736 tar->sparse_numbytes = -1; 1737 tar->sparse_gnu_major = 0; 1738 tar->sparse_gnu_minor = 0; 1739 } 1740 if (strcmp(key, "GNU.sparse.offset") == 0) { 1741 tar->sparse_offset = tar_atol10(value, strlen(value)); 1742 if (tar->sparse_numbytes != -1) { 1743 if (gnu_add_sparse_entry(a, tar, 1744 tar->sparse_offset, tar->sparse_numbytes) 1745 != ARCHIVE_OK) 1746 return (ARCHIVE_FATAL); 1747 tar->sparse_offset = -1; 1748 tar->sparse_numbytes = -1; 1749 } 1750 } 1751 if (strcmp(key, "GNU.sparse.numbytes") == 0) { 1752 tar->sparse_numbytes = tar_atol10(value, strlen(value)); 1753 if (tar->sparse_numbytes != -1) { 1754 if (gnu_add_sparse_entry(a, tar, 1755 tar->sparse_offset, tar->sparse_numbytes) 1756 != ARCHIVE_OK) 1757 return (ARCHIVE_FATAL); 1758 tar->sparse_offset = -1; 1759 tar->sparse_numbytes = -1; 1760 } 1761 } 1762 if (strcmp(key, "GNU.sparse.size") == 0) { 1763 tar->realsize = tar_atol10(value, strlen(value)); 1764 archive_entry_set_size(entry, tar->realsize); 1765 } 1766 1767 /* GNU "0.1" sparse pax format. */ 1768 if (strcmp(key, "GNU.sparse.map") == 0) { 1769 tar->sparse_gnu_major = 0; 1770 tar->sparse_gnu_minor = 1; 1771 if (gnu_sparse_01_parse(a, tar, value) != ARCHIVE_OK) 1772 return (ARCHIVE_WARN); 1773 } 1774 1775 /* GNU "1.0" sparse pax format */ 1776 if (strcmp(key, "GNU.sparse.major") == 0) { 1777 tar->sparse_gnu_major = (int)tar_atol10(value, strlen(value)); 1778 tar->sparse_gnu_pending = 1; 1779 } 1780 if (strcmp(key, "GNU.sparse.minor") == 0) { 1781 tar->sparse_gnu_minor = (int)tar_atol10(value, strlen(value)); 1782 tar->sparse_gnu_pending = 1; 1783 } 1784 if (strcmp(key, "GNU.sparse.name") == 0) { 1785 /* 1786 * The real filename; when storing sparse 1787 * files, GNU tar puts a synthesized name into 1788 * the regular 'path' attribute in an attempt 1789 * to limit confusion. ;-) 1790 */ 1791 archive_strcpy(&(tar->entry_pathname_override), value); 1792 } 1793 if (strcmp(key, "GNU.sparse.realsize") == 0) { 1794 tar->realsize = tar_atol10(value, strlen(value)); 1795 archive_entry_set_size(entry, tar->realsize); 1796 } 1797 break; 1798 case 'L': 1799 /* Our extensions */ 1800/* TODO: Handle arbitrary extended attributes... */ 1801/* 1802 if (strcmp(key, "LIBARCHIVE.xxxxxxx") == 0) 1803 archive_entry_set_xxxxxx(entry, value); 1804*/ 1805 if (strcmp(key, "LIBARCHIVE.creationtime") == 0) { 1806 pax_time(value, &s, &n); 1807 archive_entry_set_birthtime(entry, s, n); 1808 } 1809 if (memcmp(key, "LIBARCHIVE.xattr.", 17) == 0) 1810 pax_attribute_xattr(entry, key, value); 1811 break; 1812 case 'S': 1813 /* We support some keys used by the "star" archiver */ 1814 if (strcmp(key, "SCHILY.acl.access") == 0) { 1815 if (tar->sconv_acl == NULL) { 1816 tar->sconv_acl = 1817 archive_string_conversion_from_charset( 1818 &(a->archive), "UTF-8", 1); 1819 if (tar->sconv_acl == NULL) 1820 return (ARCHIVE_FATAL); 1821 } 1822 1823 r = archive_acl_parse_l(archive_entry_acl(entry), 1824 value, ARCHIVE_ENTRY_ACL_TYPE_ACCESS, 1825 tar->sconv_acl); 1826 if (r != ARCHIVE_OK) { 1827 err = r; 1828 if (err == ARCHIVE_FATAL) { 1829 archive_set_error(&a->archive, ENOMEM, 1830 "Can't allocate memory for " 1831 "SCHILY.acl.access"); 1832 return (err); 1833 } 1834 archive_set_error(&a->archive, 1835 ARCHIVE_ERRNO_MISC, 1836 "Parse error: SCHILY.acl.access"); 1837 } 1838 } else if (strcmp(key, "SCHILY.acl.default") == 0) { 1839 if (tar->sconv_acl == NULL) { 1840 tar->sconv_acl = 1841 archive_string_conversion_from_charset( 1842 &(a->archive), "UTF-8", 1); 1843 if (tar->sconv_acl == NULL) 1844 return (ARCHIVE_FATAL); 1845 } 1846 1847 r = archive_acl_parse_l(archive_entry_acl(entry), 1848 value, ARCHIVE_ENTRY_ACL_TYPE_DEFAULT, 1849 tar->sconv_acl); 1850 if (r != ARCHIVE_OK) { 1851 err = r; 1852 if (err == ARCHIVE_FATAL) { 1853 archive_set_error(&a->archive, ENOMEM, 1854 "Can't allocate memory for " 1855 "SCHILY.acl.default"); 1856 return (err); 1857 } 1858 archive_set_error(&a->archive, 1859 ARCHIVE_ERRNO_MISC, 1860 "Parse error: SCHILY.acl.default"); 1861 } 1862 } else if (strcmp(key, "SCHILY.devmajor") == 0) { 1863 archive_entry_set_rdevmajor(entry, 1864 (dev_t)tar_atol10(value, strlen(value))); 1865 } else if (strcmp(key, "SCHILY.devminor") == 0) { 1866 archive_entry_set_rdevminor(entry, 1867 (dev_t)tar_atol10(value, strlen(value))); 1868 } else if (strcmp(key, "SCHILY.fflags") == 0) { 1869 archive_entry_copy_fflags_text(entry, value); 1870 } else if (strcmp(key, "SCHILY.dev") == 0) { 1871 archive_entry_set_dev(entry, 1872 (dev_t)tar_atol10(value, strlen(value))); 1873 } else if (strcmp(key, "SCHILY.ino") == 0) { 1874 archive_entry_set_ino(entry, 1875 tar_atol10(value, strlen(value))); 1876 } else if (strcmp(key, "SCHILY.nlink") == 0) { 1877 archive_entry_set_nlink(entry, (unsigned) 1878 tar_atol10(value, strlen(value))); 1879 } else if (strcmp(key, "SCHILY.realsize") == 0) { 1880 tar->realsize = tar_atol10(value, strlen(value)); 1881 archive_entry_set_size(entry, tar->realsize); 1882 } else if (strcmp(key, "SUN.holesdata") == 0) { 1883 /* A Solaris extension for sparse. */ 1884 r = solaris_sparse_parse(a, tar, entry, value); 1885 if (r < err) { 1886 if (r == ARCHIVE_FATAL) 1887 return (r); 1888 err = r; 1889 archive_set_error(&a->archive, 1890 ARCHIVE_ERRNO_MISC, 1891 "Parse error: SUN.holesdata"); 1892 } 1893 } 1894 break; 1895 case 'a': 1896 if (strcmp(key, "atime") == 0) { 1897 pax_time(value, &s, &n); 1898 archive_entry_set_atime(entry, s, n); 1899 } 1900 break; 1901 case 'c': 1902 if (strcmp(key, "ctime") == 0) { 1903 pax_time(value, &s, &n); 1904 archive_entry_set_ctime(entry, s, n); 1905 } else if (strcmp(key, "charset") == 0) { 1906 /* TODO: Publish charset information in entry. */ 1907 } else if (strcmp(key, "comment") == 0) { 1908 /* TODO: Publish comment in entry. */ 1909 } 1910 break; 1911 case 'g': 1912 if (strcmp(key, "gid") == 0) { 1913 archive_entry_set_gid(entry, 1914 tar_atol10(value, strlen(value))); 1915 } else if (strcmp(key, "gname") == 0) { 1916 archive_strcpy(&(tar->entry_gname), value); 1917 } 1918 break; 1919 case 'h': 1920 if (strcmp(key, "hdrcharset") == 0) { 1921 if (strcmp(value, "BINARY") == 0) 1922 /* Binary mode. */ 1923 tar->pax_hdrcharset_binary = 1; 1924 else if (strcmp(value, "ISO-IR 10646 2000 UTF-8") == 0) 1925 tar->pax_hdrcharset_binary = 0; 1926 } 1927 break; 1928 case 'l': 1929 /* pax interchange doesn't distinguish hardlink vs. symlink. */ 1930 if (strcmp(key, "linkpath") == 0) { 1931 archive_strcpy(&(tar->entry_linkpath), value); 1932 } 1933 break; 1934 case 'm': 1935 if (strcmp(key, "mtime") == 0) { 1936 pax_time(value, &s, &n); 1937 archive_entry_set_mtime(entry, s, n); 1938 } 1939 break; 1940 case 'p': 1941 if (strcmp(key, "path") == 0) { 1942 archive_strcpy(&(tar->entry_pathname), value); 1943 } 1944 break; 1945 case 'r': 1946 /* POSIX has reserved 'realtime.*' */ 1947 break; 1948 case 's': 1949 /* POSIX has reserved 'security.*' */ 1950 /* Someday: if (strcmp(key, "security.acl") == 0) { ... } */ 1951 if (strcmp(key, "size") == 0) { 1952 /* "size" is the size of the data in the entry. */ 1953 tar->entry_bytes_remaining 1954 = tar_atol10(value, strlen(value)); 1955 /* 1956 * But, "size" is not necessarily the size of 1957 * the file on disk; if this is a sparse file, 1958 * the disk size may have already been set from 1959 * GNU.sparse.realsize or GNU.sparse.size or 1960 * an old GNU header field or SCHILY.realsize 1961 * or .... 1962 */ 1963 if (tar->realsize < 0) { 1964 archive_entry_set_size(entry, 1965 tar->entry_bytes_remaining); 1966 tar->realsize 1967 = tar->entry_bytes_remaining; 1968 } 1969 } 1970 break; 1971 case 'u': 1972 if (strcmp(key, "uid") == 0) { 1973 archive_entry_set_uid(entry, 1974 tar_atol10(value, strlen(value))); 1975 } else if (strcmp(key, "uname") == 0) { 1976 archive_strcpy(&(tar->entry_uname), value); 1977 } 1978 break; 1979 } 1980 return (err); 1981} 1982 1983 1984 1985/* 1986 * parse a decimal time value, which may include a fractional portion 1987 */ 1988static void 1989pax_time(const char *p, int64_t *ps, long *pn) 1990{ 1991 char digit; 1992 int64_t s; 1993 unsigned long l; 1994 int sign; 1995 int64_t limit, last_digit_limit; 1996 1997 limit = INT64_MAX / 10; 1998 last_digit_limit = INT64_MAX % 10; 1999 2000 s = 0; 2001 sign = 1; 2002 if (*p == '-') { 2003 sign = -1; 2004 p++; 2005 } 2006 while (*p >= '0' && *p <= '9') { 2007 digit = *p - '0'; 2008 if (s > limit || 2009 (s == limit && digit > last_digit_limit)) { 2010 s = INT64_MAX; 2011 break; 2012 } 2013 s = (s * 10) + digit; 2014 ++p; 2015 } 2016 2017 *ps = s * sign; 2018 2019 /* Calculate nanoseconds. */ 2020 *pn = 0; 2021 2022 if (*p != '.') 2023 return; 2024 2025 l = 100000000UL; 2026 do { 2027 ++p; 2028 if (*p >= '0' && *p <= '9') 2029 *pn += (*p - '0') * l; 2030 else 2031 break; 2032 } while (l /= 10); 2033} 2034 2035/* 2036 * Parse GNU tar header 2037 */ 2038static int 2039header_gnutar(struct archive_read *a, struct tar *tar, 2040 struct archive_entry *entry, const void *h, size_t *unconsumed) 2041{ 2042 const struct archive_entry_header_gnutar *header; 2043 int64_t t; 2044 int err = ARCHIVE_OK; 2045 2046 /* 2047 * GNU header is like POSIX ustar, except 'prefix' is 2048 * replaced with some other fields. This also means the 2049 * filename is stored as in old-style archives. 2050 */ 2051 2052 /* Grab fields common to all tar variants. */ 2053 err = header_common(a, tar, entry, h); 2054 if (err == ARCHIVE_FATAL) 2055 return (err); 2056 2057 /* Copy filename over (to ensure null termination). */ 2058 header = (const struct archive_entry_header_gnutar *)h; 2059 if (archive_entry_copy_pathname_l(entry, 2060 header->name, sizeof(header->name), tar->sconv) != 0) { 2061 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 2062 if (err == ARCHIVE_FATAL) 2063 return (err); 2064 } 2065 2066 /* Fields common to ustar and GNU */ 2067 /* XXX Can the following be factored out since it's common 2068 * to ustar and gnu tar? Is it okay to move it down into 2069 * header_common, perhaps? */ 2070 if (archive_entry_copy_uname_l(entry, 2071 header->uname, sizeof(header->uname), tar->sconv) != 0) { 2072 err = set_conversion_failed_error(a, tar->sconv, "Uname"); 2073 if (err == ARCHIVE_FATAL) 2074 return (err); 2075 } 2076 2077 if (archive_entry_copy_gname_l(entry, 2078 header->gname, sizeof(header->gname), tar->sconv) != 0) { 2079 err = set_conversion_failed_error(a, tar->sconv, "Gname"); 2080 if (err == ARCHIVE_FATAL) 2081 return (err); 2082 } 2083 2084 /* Parse out device numbers only for char and block specials */ 2085 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') { 2086 archive_entry_set_rdevmajor(entry, (dev_t) 2087 tar_atol(header->rdevmajor, sizeof(header->rdevmajor))); 2088 archive_entry_set_rdevminor(entry, (dev_t) 2089 tar_atol(header->rdevminor, sizeof(header->rdevminor))); 2090 } else 2091 archive_entry_set_rdev(entry, 0); 2092 2093 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 2094 2095 /* Grab GNU-specific fields. */ 2096 t = tar_atol(header->atime, sizeof(header->atime)); 2097 if (t > 0) 2098 archive_entry_set_atime(entry, t, 0); 2099 t = tar_atol(header->ctime, sizeof(header->ctime)); 2100 if (t > 0) 2101 archive_entry_set_ctime(entry, t, 0); 2102 2103 if (header->realsize[0] != 0) { 2104 tar->realsize 2105 = tar_atol(header->realsize, sizeof(header->realsize)); 2106 archive_entry_set_size(entry, tar->realsize); 2107 } 2108 2109 if (header->sparse[0].offset[0] != 0) { 2110 if (gnu_sparse_old_read(a, tar, header, unconsumed) 2111 != ARCHIVE_OK) 2112 return (ARCHIVE_FATAL); 2113 } else { 2114 if (header->isextended[0] != 0) { 2115 /* XXX WTF? XXX */ 2116 } 2117 } 2118 2119 return (err); 2120} 2121 2122static int 2123gnu_add_sparse_entry(struct archive_read *a, struct tar *tar, 2124 int64_t offset, int64_t remaining) 2125{ 2126 struct sparse_block *p; 2127 2128 p = (struct sparse_block *)malloc(sizeof(*p)); 2129 if (p == NULL) { 2130 archive_set_error(&a->archive, ENOMEM, "Out of memory"); 2131 return (ARCHIVE_FATAL); 2132 } 2133 memset(p, 0, sizeof(*p)); 2134 if (tar->sparse_last != NULL) 2135 tar->sparse_last->next = p; 2136 else 2137 tar->sparse_list = p; 2138 tar->sparse_last = p; 2139 if (remaining < 0 || offset < 0) { 2140 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Malformed sparse map data"); 2141 return (ARCHIVE_FATAL); 2142 } 2143 p->offset = offset; 2144 p->remaining = remaining; 2145 return (ARCHIVE_OK); 2146} 2147 2148static void 2149gnu_clear_sparse_list(struct tar *tar) 2150{ 2151 struct sparse_block *p; 2152 2153 while (tar->sparse_list != NULL) { 2154 p = tar->sparse_list; 2155 tar->sparse_list = p->next; 2156 free(p); 2157 } 2158 tar->sparse_last = NULL; 2159} 2160 2161/* 2162 * GNU tar old-format sparse data. 2163 * 2164 * GNU old-format sparse data is stored in a fixed-field 2165 * format. Offset/size values are 11-byte octal fields (same 2166 * format as 'size' field in ustart header). These are 2167 * stored in the header, allocating subsequent header blocks 2168 * as needed. Extending the header in this way is a pretty 2169 * severe POSIX violation; this design has earned GNU tar a 2170 * lot of criticism. 2171 */ 2172 2173static int 2174gnu_sparse_old_read(struct archive_read *a, struct tar *tar, 2175 const struct archive_entry_header_gnutar *header, size_t *unconsumed) 2176{ 2177 ssize_t bytes_read; 2178 const void *data; 2179 struct extended { 2180 struct gnu_sparse sparse[21]; 2181 char isextended[1]; 2182 char padding[7]; 2183 }; 2184 const struct extended *ext; 2185 2186 if (gnu_sparse_old_parse(a, tar, header->sparse, 4) != ARCHIVE_OK) 2187 return (ARCHIVE_FATAL); 2188 if (header->isextended[0] == 0) 2189 return (ARCHIVE_OK); 2190 2191 do { 2192 tar_flush_unconsumed(a, unconsumed); 2193 data = __archive_read_ahead(a, 512, &bytes_read); 2194 if (bytes_read < 0) 2195 return (ARCHIVE_FATAL); 2196 if (bytes_read < 512) { 2197 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 2198 "Truncated tar archive " 2199 "detected while reading sparse file data"); 2200 return (ARCHIVE_FATAL); 2201 } 2202 *unconsumed = 512; 2203 ext = (const struct extended *)data; 2204 if (gnu_sparse_old_parse(a, tar, ext->sparse, 21) != ARCHIVE_OK) 2205 return (ARCHIVE_FATAL); 2206 } while (ext->isextended[0] != 0); 2207 if (tar->sparse_list != NULL) 2208 tar->entry_offset = tar->sparse_list->offset; 2209 return (ARCHIVE_OK); 2210} 2211 2212static int 2213gnu_sparse_old_parse(struct archive_read *a, struct tar *tar, 2214 const struct gnu_sparse *sparse, int length) 2215{ 2216 while (length > 0 && sparse->offset[0] != 0) { 2217 if (gnu_add_sparse_entry(a, tar, 2218 tar_atol(sparse->offset, sizeof(sparse->offset)), 2219 tar_atol(sparse->numbytes, sizeof(sparse->numbytes))) 2220 != ARCHIVE_OK) 2221 return (ARCHIVE_FATAL); 2222 sparse++; 2223 length--; 2224 } 2225 return (ARCHIVE_OK); 2226} 2227 2228/* 2229 * GNU tar sparse format 0.0 2230 * 2231 * Beginning with GNU tar 1.15, sparse files are stored using 2232 * information in the pax extended header. The GNU tar maintainers 2233 * have gone through a number of variations in the process of working 2234 * out this scheme; fortunately, they're all numbered. 2235 * 2236 * Sparse format 0.0 uses attribute GNU.sparse.numblocks to store the 2237 * number of blocks, and GNU.sparse.offset/GNU.sparse.numbytes to 2238 * store offset/size for each block. The repeated instances of these 2239 * latter fields violate the pax specification (which frowns on 2240 * duplicate keys), so this format was quickly replaced. 2241 */ 2242 2243/* 2244 * GNU tar sparse format 0.1 2245 * 2246 * This version replaced the offset/numbytes attributes with 2247 * a single "map" attribute that stored a list of integers. This 2248 * format had two problems: First, the "map" attribute could be very 2249 * long, which caused problems for some implementations. More 2250 * importantly, the sparse data was lost when extracted by archivers 2251 * that didn't recognize this extension. 2252 */ 2253 2254static int 2255gnu_sparse_01_parse(struct archive_read *a, struct tar *tar, const char *p) 2256{ 2257 const char *e; 2258 int64_t offset = -1, size = -1; 2259 2260 for (;;) { 2261 e = p; 2262 while (*e != '\0' && *e != ',') { 2263 if (*e < '0' || *e > '9') 2264 return (ARCHIVE_WARN); 2265 e++; 2266 } 2267 if (offset < 0) { 2268 offset = tar_atol10(p, e - p); 2269 if (offset < 0) 2270 return (ARCHIVE_WARN); 2271 } else { 2272 size = tar_atol10(p, e - p); 2273 if (size < 0) 2274 return (ARCHIVE_WARN); 2275 if (gnu_add_sparse_entry(a, tar, offset, size) 2276 != ARCHIVE_OK) 2277 return (ARCHIVE_FATAL); 2278 offset = -1; 2279 } 2280 if (*e == '\0') 2281 return (ARCHIVE_OK); 2282 p = e + 1; 2283 } 2284} 2285 2286/* 2287 * GNU tar sparse format 1.0 2288 * 2289 * The idea: The offset/size data is stored as a series of base-10 2290 * ASCII numbers prepended to the file data, so that dearchivers that 2291 * don't support this format will extract the block map along with the 2292 * data and a separate post-process can restore the sparseness. 2293 * 2294 * Unfortunately, GNU tar 1.16 had a bug that added unnecessary 2295 * padding to the body of the file when using this format. GNU tar 2296 * 1.17 corrected this bug without bumping the version number, so 2297 * it's not possible to support both variants. This code supports 2298 * the later variant at the expense of not supporting the former. 2299 * 2300 * This variant also replaced GNU.sparse.size with GNU.sparse.realsize 2301 * and introduced the GNU.sparse.major/GNU.sparse.minor attributes. 2302 */ 2303 2304/* 2305 * Read the next line from the input, and parse it as a decimal 2306 * integer followed by '\n'. Returns positive integer value or 2307 * negative on error. 2308 */ 2309static int64_t 2310gnu_sparse_10_atol(struct archive_read *a, struct tar *tar, 2311 int64_t *remaining, size_t *unconsumed) 2312{ 2313 int64_t l, limit, last_digit_limit; 2314 const char *p; 2315 ssize_t bytes_read; 2316 int base, digit; 2317 2318 base = 10; 2319 limit = INT64_MAX / base; 2320 last_digit_limit = INT64_MAX % base; 2321 2322 /* 2323 * Skip any lines starting with '#'; GNU tar specs 2324 * don't require this, but they should. 2325 */ 2326 do { 2327 bytes_read = readline(a, tar, &p, 2328 (ssize_t)tar_min(*remaining, 100), unconsumed); 2329 if (bytes_read <= 0) 2330 return (ARCHIVE_FATAL); 2331 *remaining -= bytes_read; 2332 } while (p[0] == '#'); 2333 2334 l = 0; 2335 while (bytes_read > 0) { 2336 if (*p == '\n') 2337 return (l); 2338 if (*p < '0' || *p >= '0' + base) 2339 return (ARCHIVE_WARN); 2340 digit = *p - '0'; 2341 if (l > limit || (l == limit && digit > last_digit_limit)) 2342 l = INT64_MAX; /* Truncate on overflow. */ 2343 else 2344 l = (l * base) + digit; 2345 p++; 2346 bytes_read--; 2347 } 2348 /* TODO: Error message. */ 2349 return (ARCHIVE_WARN); 2350} 2351 2352/* 2353 * Returns length (in bytes) of the sparse data description 2354 * that was read. 2355 */ 2356static ssize_t 2357gnu_sparse_10_read(struct archive_read *a, struct tar *tar, size_t *unconsumed) 2358{ 2359 ssize_t bytes_read; 2360 int entries; 2361 int64_t offset, size, to_skip, remaining; 2362 2363 /* Clear out the existing sparse list. */ 2364 gnu_clear_sparse_list(tar); 2365 2366 remaining = tar->entry_bytes_remaining; 2367 2368 /* Parse entries. */ 2369 entries = (int)gnu_sparse_10_atol(a, tar, &remaining, unconsumed); 2370 if (entries < 0) 2371 return (ARCHIVE_FATAL); 2372 /* Parse the individual entries. */ 2373 while (entries-- > 0) { 2374 /* Parse offset/size */ 2375 offset = gnu_sparse_10_atol(a, tar, &remaining, unconsumed); 2376 if (offset < 0) 2377 return (ARCHIVE_FATAL); 2378 size = gnu_sparse_10_atol(a, tar, &remaining, unconsumed); 2379 if (size < 0) 2380 return (ARCHIVE_FATAL); 2381 /* Add a new sparse entry. */ 2382 if (gnu_add_sparse_entry(a, tar, offset, size) != ARCHIVE_OK) 2383 return (ARCHIVE_FATAL); 2384 } 2385 /* Skip rest of block... */ 2386 tar_flush_unconsumed(a, unconsumed); 2387 bytes_read = (ssize_t)(tar->entry_bytes_remaining - remaining); 2388 to_skip = 0x1ff & -bytes_read; 2389 if (to_skip != __archive_read_consume(a, to_skip)) 2390 return (ARCHIVE_FATAL); 2391 return ((ssize_t)(bytes_read + to_skip)); 2392} 2393 2394/* 2395 * Solaris pax extension for a sparse file. This is recorded with the 2396 * data and hole pairs. The way recording sparse information by Solaris' 2397 * pax simply indicates where data and sparse are, so the stored contents 2398 * consist of both data and hole. 2399 */ 2400static int 2401solaris_sparse_parse(struct archive_read *a, struct tar *tar, 2402 struct archive_entry *entry, const char *p) 2403{ 2404 const char *e; 2405 int64_t start, end; 2406 int hole = 1; 2407 2408 (void)entry; /* UNUSED */ 2409 2410 end = 0; 2411 if (*p == ' ') 2412 p++; 2413 else 2414 return (ARCHIVE_WARN); 2415 for (;;) { 2416 e = p; 2417 while (*e != '\0' && *e != ' ') { 2418 if (*e < '0' || *e > '9') 2419 return (ARCHIVE_WARN); 2420 e++; 2421 } 2422 start = end; 2423 end = tar_atol10(p, e - p); 2424 if (end < 0) 2425 return (ARCHIVE_WARN); 2426 if (start < end) { 2427 if (gnu_add_sparse_entry(a, tar, start, 2428 end - start) != ARCHIVE_OK) 2429 return (ARCHIVE_FATAL); 2430 tar->sparse_last->hole = hole; 2431 } 2432 if (*e == '\0') 2433 return (ARCHIVE_OK); 2434 p = e + 1; 2435 hole = hole == 0; 2436 } 2437} 2438 2439/*- 2440 * Convert text->integer. 2441 * 2442 * Traditional tar formats (including POSIX) specify base-8 for 2443 * all of the standard numeric fields. This is a significant limitation 2444 * in practice: 2445 * = file size is limited to 8GB 2446 * = rdevmajor and rdevminor are limited to 21 bits 2447 * = uid/gid are limited to 21 bits 2448 * 2449 * There are two workarounds for this: 2450 * = pax extended headers, which use variable-length string fields 2451 * = GNU tar and STAR both allow either base-8 or base-256 in 2452 * most fields. The high bit is set to indicate base-256. 2453 * 2454 * On read, this implementation supports both extensions. 2455 */ 2456static int64_t 2457tar_atol(const char *p, size_t char_cnt) 2458{ 2459 /* 2460 * Technically, GNU tar considers a field to be in base-256 2461 * only if the first byte is 0xff or 0x80. 2462 */ 2463 if (*p & 0x80) 2464 return (tar_atol256(p, char_cnt)); 2465 return (tar_atol8(p, char_cnt)); 2466} 2467 2468/* 2469 * Note that this implementation does not (and should not!) obey 2470 * locale settings; you cannot simply substitute strtol here, since 2471 * it does obey locale. 2472 */ 2473static int64_t 2474tar_atol_base_n(const char *p, size_t char_cnt, int base) 2475{ 2476 int64_t l, maxval, limit, last_digit_limit; 2477 int digit, sign; 2478 2479 maxval = INT64_MAX; 2480 limit = INT64_MAX / base; 2481 last_digit_limit = INT64_MAX % base; 2482 2483 /* the pointer will not be dereferenced if char_cnt is zero 2484 * due to the way the && operator is evaulated. 2485 */ 2486 while (char_cnt != 0 && (*p == ' ' || *p == '\t')) { 2487 p++; 2488 char_cnt--; 2489 } 2490 2491 sign = 1; 2492 if (char_cnt != 0 && *p == '-') { 2493 sign = -1; 2494 p++; 2495 char_cnt--; 2496 2497 maxval = INT64_MIN; 2498 limit = -(INT64_MIN / base); 2499 last_digit_limit = INT64_MIN % base; 2500 } 2501 2502 l = 0; 2503 if (char_cnt != 0) { 2504 digit = *p - '0'; 2505 while (digit >= 0 && digit < base && char_cnt != 0) { 2506 if (l>limit || (l == limit && digit > last_digit_limit)) { 2507 return maxval; /* Truncate on overflow. */ 2508 } 2509 l = (l * base) + digit; 2510 digit = *++p - '0'; 2511 char_cnt--; 2512 } 2513 } 2514 return (sign < 0) ? -l : l; 2515} 2516 2517static int64_t 2518tar_atol8(const char *p, size_t char_cnt) 2519{ 2520 return tar_atol_base_n(p, char_cnt, 8); 2521} 2522 2523static int64_t 2524tar_atol10(const char *p, size_t char_cnt) 2525{ 2526 return tar_atol_base_n(p, char_cnt, 10); 2527} 2528 2529/* 2530 * Parse a base-256 integer. This is just a variable-length 2531 * twos-complement signed binary value in big-endian order, except 2532 * that the high-order bit is ignored. The values here can be up to 2533 * 12 bytes, so we need to be careful about overflowing 64-bit 2534 * (8-byte) integers. 2535 * 2536 * This code unashamedly assumes that the local machine uses 8-bit 2537 * bytes and twos-complement arithmetic. 2538 */ 2539static int64_t 2540tar_atol256(const char *_p, size_t char_cnt) 2541{ 2542 uint64_t l; 2543 const unsigned char *p = (const unsigned char *)_p; 2544 unsigned char c, neg; 2545 2546 /* Extend 7-bit 2s-comp to 8-bit 2s-comp, decide sign. */ 2547 c = *p; 2548 if (c & 0x40) { 2549 neg = 0xff; 2550 c |= 0x80; 2551 l = ~ARCHIVE_LITERAL_ULL(0); 2552 } else { 2553 neg = 0; 2554 c &= 0x7f; 2555 l = 0; 2556 } 2557 2558 /* If more than 8 bytes, check that we can ignore 2559 * high-order bits without overflow. */ 2560 while (char_cnt > sizeof(int64_t)) { 2561 --char_cnt; 2562 if (c != neg) 2563 return neg ? INT64_MIN : INT64_MAX; 2564 c = *++p; 2565 } 2566 2567 /* c is first byte that fits; if sign mismatch, return overflow */ 2568 if ((c ^ neg) & 0x80) { 2569 return neg ? INT64_MIN : INT64_MAX; 2570 } 2571 2572 /* Accumulate remaining bytes. */ 2573 while (--char_cnt > 0) { 2574 l = (l << 8) | c; 2575 c = *++p; 2576 } 2577 l = (l << 8) | c; 2578 /* Return signed twos-complement value. */ 2579 return (int64_t)(l); 2580} 2581 2582/* 2583 * Returns length of line (including trailing newline) 2584 * or negative on error. 'start' argument is updated to 2585 * point to first character of line. This avoids copying 2586 * when possible. 2587 */ 2588static ssize_t 2589readline(struct archive_read *a, struct tar *tar, const char **start, 2590 ssize_t limit, size_t *unconsumed) 2591{ 2592 ssize_t bytes_read; 2593 ssize_t total_size = 0; 2594 const void *t; 2595 const char *s; 2596 void *p; 2597 2598 tar_flush_unconsumed(a, unconsumed); 2599 2600 t = __archive_read_ahead(a, 1, &bytes_read); 2601 if (bytes_read <= 0) 2602 return (ARCHIVE_FATAL); 2603 s = t; /* Start of line? */ 2604 p = memchr(t, '\n', bytes_read); 2605 /* If we found '\n' in the read buffer, return pointer to that. */ 2606 if (p != NULL) { 2607 bytes_read = 1 + ((const char *)p) - s; 2608 if (bytes_read > limit) { 2609 archive_set_error(&a->archive, 2610 ARCHIVE_ERRNO_FILE_FORMAT, 2611 "Line too long"); 2612 return (ARCHIVE_FATAL); 2613 } 2614 *unconsumed = bytes_read; 2615 *start = s; 2616 return (bytes_read); 2617 } 2618 *unconsumed = bytes_read; 2619 /* Otherwise, we need to accumulate in a line buffer. */ 2620 for (;;) { 2621 if (total_size + bytes_read > limit) { 2622 archive_set_error(&a->archive, 2623 ARCHIVE_ERRNO_FILE_FORMAT, 2624 "Line too long"); 2625 return (ARCHIVE_FATAL); 2626 } 2627 if (archive_string_ensure(&tar->line, total_size + bytes_read) == NULL) { 2628 archive_set_error(&a->archive, ENOMEM, 2629 "Can't allocate working buffer"); 2630 return (ARCHIVE_FATAL); 2631 } 2632 memcpy(tar->line.s + total_size, t, bytes_read); 2633 tar_flush_unconsumed(a, unconsumed); 2634 total_size += bytes_read; 2635 /* If we found '\n', clean up and return. */ 2636 if (p != NULL) { 2637 *start = tar->line.s; 2638 return (total_size); 2639 } 2640 /* Read some more. */ 2641 t = __archive_read_ahead(a, 1, &bytes_read); 2642 if (bytes_read <= 0) 2643 return (ARCHIVE_FATAL); 2644 s = t; /* Start of line? */ 2645 p = memchr(t, '\n', bytes_read); 2646 /* If we found '\n', trim the read. */ 2647 if (p != NULL) { 2648 bytes_read = 1 + ((const char *)p) - s; 2649 } 2650 *unconsumed = bytes_read; 2651 } 2652} 2653 2654/* 2655 * base64_decode - Base64 decode 2656 * 2657 * This accepts most variations of base-64 encoding, including: 2658 * * with or without line breaks 2659 * * with or without the final group padded with '=' or '_' characters 2660 * (The most economical Base-64 variant does not pad the last group and 2661 * omits line breaks; RFC1341 used for MIME requires both.) 2662 */ 2663static char * 2664base64_decode(const char *s, size_t len, size_t *out_len) 2665{ 2666 static const unsigned char digits[64] = { 2667 'A','B','C','D','E','F','G','H','I','J','K','L','M','N', 2668 'O','P','Q','R','S','T','U','V','W','X','Y','Z','a','b', 2669 'c','d','e','f','g','h','i','j','k','l','m','n','o','p', 2670 'q','r','s','t','u','v','w','x','y','z','0','1','2','3', 2671 '4','5','6','7','8','9','+','/' }; 2672 static unsigned char decode_table[128]; 2673 char *out, *d; 2674 const unsigned char *src = (const unsigned char *)s; 2675 2676 /* If the decode table is not yet initialized, prepare it. */ 2677 if (decode_table[digits[1]] != 1) { 2678 unsigned i; 2679 memset(decode_table, 0xff, sizeof(decode_table)); 2680 for (i = 0; i < sizeof(digits); i++) 2681 decode_table[digits[i]] = i; 2682 } 2683 2684 /* Allocate enough space to hold the entire output. */ 2685 /* Note that we may not use all of this... */ 2686 out = (char *)malloc(len - len / 4 + 1); 2687 if (out == NULL) { 2688 *out_len = 0; 2689 return (NULL); 2690 } 2691 d = out; 2692 2693 while (len > 0) { 2694 /* Collect the next group of (up to) four characters. */ 2695 int v = 0; 2696 int group_size = 0; 2697 while (group_size < 4 && len > 0) { 2698 /* '=' or '_' padding indicates final group. */ 2699 if (*src == '=' || *src == '_') { 2700 len = 0; 2701 break; 2702 } 2703 /* Skip illegal characters (including line breaks) */ 2704 if (*src > 127 || *src < 32 2705 || decode_table[*src] == 0xff) { 2706 len--; 2707 src++; 2708 continue; 2709 } 2710 v <<= 6; 2711 v |= decode_table[*src++]; 2712 len --; 2713 group_size++; 2714 } 2715 /* Align a short group properly. */ 2716 v <<= 6 * (4 - group_size); 2717 /* Unpack the group we just collected. */ 2718 switch (group_size) { 2719 case 4: d[2] = v & 0xff; 2720 /* FALLTHROUGH */ 2721 case 3: d[1] = (v >> 8) & 0xff; 2722 /* FALLTHROUGH */ 2723 case 2: d[0] = (v >> 16) & 0xff; 2724 break; 2725 case 1: /* this is invalid! */ 2726 break; 2727 } 2728 d += group_size * 3 / 4; 2729 } 2730 2731 *out_len = d - out; 2732 return (out); 2733} 2734 2735static char * 2736url_decode(const char *in) 2737{ 2738 char *out, *d; 2739 const char *s; 2740 2741 out = (char *)malloc(strlen(in) + 1); 2742 if (out == NULL) 2743 return (NULL); 2744 for (s = in, d = out; *s != '\0'; ) { 2745 if (s[0] == '%' && s[1] != '\0' && s[2] != '\0') { 2746 /* Try to convert % escape */ 2747 int digit1 = tohex(s[1]); 2748 int digit2 = tohex(s[2]); 2749 if (digit1 >= 0 && digit2 >= 0) { 2750 /* Looks good, consume three chars */ 2751 s += 3; 2752 /* Convert output */ 2753 *d++ = ((digit1 << 4) | digit2); 2754 continue; 2755 } 2756 /* Else fall through and treat '%' as normal char */ 2757 } 2758 *d++ = *s++; 2759 } 2760 *d = '\0'; 2761 return (out); 2762} 2763 2764static int 2765tohex(int c) 2766{ 2767 if (c >= '0' && c <= '9') 2768 return (c - '0'); 2769 else if (c >= 'A' && c <= 'F') 2770 return (c - 'A' + 10); 2771 else if (c >= 'a' && c <= 'f') 2772 return (c - 'a' + 10); 2773 else 2774 return (-1); 2775} 2776