1 /*- 2 * Copyright (c) 2003-2007 Tim Kientzle 3 * Copyright (c) 2011-2012 Michihiro NAKAJIMA 4 * Copyright (c) 2016 Martin Matuska 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR 17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 19 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, 20 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 21 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 22 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 23 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 25 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26 */ 27 28 #include "archive_platform.h" 29 __FBSDID("$FreeBSD$"); 30 31 #ifdef HAVE_ERRNO_H 32 #include <errno.h> 33 #endif 34 #include <stddef.h> 35 #ifdef HAVE_STDLIB_H 36 #include <stdlib.h> 37 #endif 38 #ifdef HAVE_STRING_H 39 #include <string.h> 40 #endif 41 42 #include "archive.h" 43 #include "archive_acl_private.h" /* For ACL parsing routines. */ 44 #include "archive_entry.h" 45 #include "archive_entry_locale.h" 46 #include "archive_private.h" 47 #include "archive_read_private.h" 48 49 #define tar_min(a,b) ((a) < (b) ? (a) : (b)) 50 51 /* 52 * Layout of POSIX 'ustar' tar header. 53 */ 54 struct archive_entry_header_ustar { 55 char name[100]; 56 char mode[8]; 57 char uid[8]; 58 char gid[8]; 59 char size[12]; 60 char mtime[12]; 61 char checksum[8]; 62 char typeflag[1]; 63 char linkname[100]; /* "old format" header ends here */ 64 char magic[6]; /* For POSIX: "ustar\0" */ 65 char version[2]; /* For POSIX: "00" */ 66 char uname[32]; 67 char gname[32]; 68 char rdevmajor[8]; 69 char rdevminor[8]; 70 char prefix[155]; 71 }; 72 73 /* 74 * Structure of GNU tar header 75 */ 76 struct gnu_sparse { 77 char offset[12]; 78 char numbytes[12]; 79 }; 80 81 struct archive_entry_header_gnutar { 82 char name[100]; 83 char mode[8]; 84 char uid[8]; 85 char gid[8]; 86 char size[12]; 87 char mtime[12]; 88 char checksum[8]; 89 char typeflag[1]; 90 char linkname[100]; 91 char magic[8]; /* "ustar \0" (note blank/blank/null at end) */ 92 char uname[32]; 93 char gname[32]; 94 char rdevmajor[8]; 95 char rdevminor[8]; 96 char atime[12]; 97 char ctime[12]; 98 char offset[12]; 99 char longnames[4]; 100 char unused[1]; 101 struct gnu_sparse sparse[4]; 102 char isextended[1]; 103 char realsize[12]; 104 /* 105 * Old GNU format doesn't use POSIX 'prefix' field; they use 106 * the 'L' (longname) entry instead. 107 */ 108 }; 109 110 /* 111 * Data specific to this format. 112 */ 113 struct sparse_block { 114 struct sparse_block *next; 115 int64_t offset; 116 int64_t remaining; 117 int hole; 118 }; 119 120 struct tar { 121 struct archive_string acl_text; 122 struct archive_string entry_pathname; 123 /* For "GNU.sparse.name" and other similar path extensions. */ 124 struct archive_string entry_pathname_override; 125 struct archive_string entry_linkpath; 126 struct archive_string entry_uname; 127 struct archive_string entry_gname; 128 struct archive_string longlink; 129 struct archive_string longname; 130 struct archive_string pax_header; 131 struct archive_string pax_global; 132 struct archive_string line; 133 int pax_hdrcharset_binary; 134 int header_recursion_depth; 135 int64_t entry_bytes_remaining; 136 int64_t entry_offset; 137 int64_t entry_padding; 138 int64_t entry_bytes_unconsumed; 139 int64_t realsize; 140 int sparse_allowed; 141 struct sparse_block *sparse_list; 142 struct sparse_block *sparse_last; 143 int64_t sparse_offset; 144 int64_t sparse_numbytes; 145 int sparse_gnu_major; 146 int sparse_gnu_minor; 147 char sparse_gnu_pending; 148 149 struct archive_string localname; 150 struct archive_string_conv *opt_sconv; 151 struct archive_string_conv *sconv; 152 struct archive_string_conv *sconv_acl; 153 struct archive_string_conv *sconv_default; 154 int init_default_conversion; 155 int compat_2x; 156 int process_mac_extensions; 157 int read_concatenated_archives; 158 int realsize_override; 159 }; 160 161 static int archive_block_is_null(const char *p); 162 static char *base64_decode(const char *, size_t, size_t *); 163 static int gnu_add_sparse_entry(struct archive_read *, struct tar *, 164 int64_t offset, int64_t remaining); 165 166 static void gnu_clear_sparse_list(struct tar *); 167 static int gnu_sparse_old_read(struct archive_read *, struct tar *, 168 const struct archive_entry_header_gnutar *header, size_t *); 169 static int gnu_sparse_old_parse(struct archive_read *, struct tar *, 170 const struct gnu_sparse *sparse, int length); 171 static int gnu_sparse_01_parse(struct archive_read *, struct tar *, 172 const char *); 173 static ssize_t gnu_sparse_10_read(struct archive_read *, struct tar *, 174 size_t *); 175 static int header_Solaris_ACL(struct archive_read *, struct tar *, 176 struct archive_entry *, const void *, size_t *); 177 static int header_common(struct archive_read *, struct tar *, 178 struct archive_entry *, const void *); 179 static int header_old_tar(struct archive_read *, struct tar *, 180 struct archive_entry *, const void *); 181 static int header_pax_extensions(struct archive_read *, struct tar *, 182 struct archive_entry *, const void *, size_t *); 183 static int header_pax_global(struct archive_read *, struct tar *, 184 struct archive_entry *, const void *h, size_t *); 185 static int header_longlink(struct archive_read *, struct tar *, 186 struct archive_entry *, const void *h, size_t *); 187 static int header_longname(struct archive_read *, struct tar *, 188 struct archive_entry *, const void *h, size_t *); 189 static int read_mac_metadata_blob(struct archive_read *, struct tar *, 190 struct archive_entry *, const void *h, size_t *); 191 static int header_volume(struct archive_read *, struct tar *, 192 struct archive_entry *, const void *h, size_t *); 193 static int header_ustar(struct archive_read *, struct tar *, 194 struct archive_entry *, const void *h); 195 static int header_gnutar(struct archive_read *, struct tar *, 196 struct archive_entry *, const void *h, size_t *); 197 static int archive_read_format_tar_bid(struct archive_read *, int); 198 static int archive_read_format_tar_options(struct archive_read *, 199 const char *, const char *); 200 static int archive_read_format_tar_cleanup(struct archive_read *); 201 static int archive_read_format_tar_read_data(struct archive_read *a, 202 const void **buff, size_t *size, int64_t *offset); 203 static int archive_read_format_tar_skip(struct archive_read *a); 204 static int archive_read_format_tar_read_header(struct archive_read *, 205 struct archive_entry *); 206 static int checksum(struct archive_read *, const void *); 207 static int pax_attribute(struct archive_read *, struct tar *, 208 struct archive_entry *, const char *key, const char *value, 209 size_t value_length); 210 static int pax_attribute_acl(struct archive_read *, struct tar *, 211 struct archive_entry *, const char *, int); 212 static int pax_attribute_xattr(struct archive_entry *, const char *, 213 const char *); 214 static int pax_header(struct archive_read *, struct tar *, 215 struct archive_entry *, struct archive_string *); 216 static void pax_time(const char *, int64_t *sec, long *nanos); 217 static ssize_t readline(struct archive_read *, struct tar *, const char **, 218 ssize_t limit, size_t *); 219 static int read_body_to_string(struct archive_read *, struct tar *, 220 struct archive_string *, const void *h, size_t *); 221 static int solaris_sparse_parse(struct archive_read *, struct tar *, 222 struct archive_entry *, const char *); 223 static int64_t tar_atol(const char *, size_t); 224 static int64_t tar_atol10(const char *, size_t); 225 static int64_t tar_atol256(const char *, size_t); 226 static int64_t tar_atol8(const char *, size_t); 227 static int tar_read_header(struct archive_read *, struct tar *, 228 struct archive_entry *, size_t *); 229 static int tohex(int c); 230 static char *url_decode(const char *); 231 static void tar_flush_unconsumed(struct archive_read *, size_t *); 232 233 234 int 235 archive_read_support_format_gnutar(struct archive *a) 236 { 237 archive_check_magic(a, ARCHIVE_READ_MAGIC, 238 ARCHIVE_STATE_NEW, "archive_read_support_format_gnutar"); 239 return (archive_read_support_format_tar(a)); 240 } 241 242 243 int 244 archive_read_support_format_tar(struct archive *_a) 245 { 246 struct archive_read *a = (struct archive_read *)_a; 247 struct tar *tar; 248 int r; 249 250 archive_check_magic(_a, ARCHIVE_READ_MAGIC, 251 ARCHIVE_STATE_NEW, "archive_read_support_format_tar"); 252 253 tar = (struct tar *)calloc(1, sizeof(*tar)); 254 if (tar == NULL) { 255 archive_set_error(&a->archive, ENOMEM, 256 "Can't allocate tar data"); 257 return (ARCHIVE_FATAL); 258 } 259 #ifdef HAVE_COPYFILE_H 260 /* Set this by default on Mac OS. */ 261 tar->process_mac_extensions = 1; 262 #endif 263 264 r = __archive_read_register_format(a, tar, "tar", 265 archive_read_format_tar_bid, 266 archive_read_format_tar_options, 267 archive_read_format_tar_read_header, 268 archive_read_format_tar_read_data, 269 archive_read_format_tar_skip, 270 NULL, 271 archive_read_format_tar_cleanup, 272 NULL, 273 NULL); 274 275 if (r != ARCHIVE_OK) 276 free(tar); 277 return (ARCHIVE_OK); 278 } 279 280 static int 281 archive_read_format_tar_cleanup(struct archive_read *a) 282 { 283 struct tar *tar; 284 285 tar = (struct tar *)(a->format->data); 286 gnu_clear_sparse_list(tar); 287 archive_string_free(&tar->acl_text); 288 archive_string_free(&tar->entry_pathname); 289 archive_string_free(&tar->entry_pathname_override); 290 archive_string_free(&tar->entry_linkpath); 291 archive_string_free(&tar->entry_uname); 292 archive_string_free(&tar->entry_gname); 293 archive_string_free(&tar->line); 294 archive_string_free(&tar->pax_global); 295 archive_string_free(&tar->pax_header); 296 archive_string_free(&tar->longname); 297 archive_string_free(&tar->longlink); 298 archive_string_free(&tar->localname); 299 free(tar); 300 (a->format->data) = NULL; 301 return (ARCHIVE_OK); 302 } 303 304 /* 305 * Validate number field 306 * 307 * This has to be pretty lenient in order to accommodate the enormous 308 * variety of tar writers in the world: 309 * = POSIX (IEEE Std 1003.1-1988) ustar requires octal values with leading 310 * zeros and allows fields to be terminated with space or null characters 311 * = Many writers use different termination (in particular, libarchive 312 * omits terminator bytes to squeeze one or two more digits) 313 * = Many writers pad with space and omit leading zeros 314 * = GNU tar and star write base-256 values if numbers are too 315 * big to be represented in octal 316 * 317 * Examples of specific tar headers that we should support: 318 * = Perl Archive::Tar terminates uid, gid, devminor and devmajor with two 319 * null bytes, pads size with spaces and other numeric fields with zeroes 320 * = plexus-archiver prior to 2.6.3 (before switching to commons-compress) 321 * may have uid and gid fields filled with spaces without any octal digits 322 * at all and pads all numeric fields with spaces 323 * 324 * This should tolerate all variants in use. It will reject a field 325 * where the writer just left garbage after a trailing NUL. 326 */ 327 static int 328 validate_number_field(const char* p_field, size_t i_size) 329 { 330 unsigned char marker = (unsigned char)p_field[0]; 331 if (marker == 128 || marker == 255 || marker == 0) { 332 /* Base-256 marker, there's nothing we can check. */ 333 return 1; 334 } else { 335 /* Must be octal */ 336 size_t i = 0; 337 /* Skip any leading spaces */ 338 while (i < i_size && p_field[i] == ' ') { 339 ++i; 340 } 341 /* Skip octal digits. */ 342 while (i < i_size && p_field[i] >= '0' && p_field[i] <= '7') { 343 ++i; 344 } 345 /* Any remaining characters must be space or NUL padding. */ 346 while (i < i_size) { 347 if (p_field[i] != ' ' && p_field[i] != 0) { 348 return 0; 349 } 350 ++i; 351 } 352 return 1; 353 } 354 } 355 356 static int 357 archive_read_format_tar_bid(struct archive_read *a, int best_bid) 358 { 359 int bid; 360 const char *h; 361 const struct archive_entry_header_ustar *header; 362 363 (void)best_bid; /* UNUSED */ 364 365 bid = 0; 366 367 /* Now let's look at the actual header and see if it matches. */ 368 h = __archive_read_ahead(a, 512, NULL); 369 if (h == NULL) 370 return (-1); 371 372 /* If it's an end-of-archive mark, we can handle it. */ 373 if (h[0] == 0 && archive_block_is_null(h)) { 374 /* 375 * Usually, I bid the number of bits verified, but 376 * in this case, 4096 seems excessive so I picked 10 as 377 * an arbitrary but reasonable-seeming value. 378 */ 379 return (10); 380 } 381 382 /* If it's not an end-of-archive mark, it must have a valid checksum.*/ 383 if (!checksum(a, h)) 384 return (0); 385 bid += 48; /* Checksum is usually 6 octal digits. */ 386 387 header = (const struct archive_entry_header_ustar *)h; 388 389 /* Recognize POSIX formats. */ 390 if ((memcmp(header->magic, "ustar\0", 6) == 0) 391 && (memcmp(header->version, "00", 2) == 0)) 392 bid += 56; 393 394 /* Recognize GNU tar format. */ 395 if ((memcmp(header->magic, "ustar ", 6) == 0) 396 && (memcmp(header->version, " \0", 2) == 0)) 397 bid += 56; 398 399 /* Type flag must be null, digit or A-Z, a-z. */ 400 if (header->typeflag[0] != 0 && 401 !( header->typeflag[0] >= '0' && header->typeflag[0] <= '9') && 402 !( header->typeflag[0] >= 'A' && header->typeflag[0] <= 'Z') && 403 !( header->typeflag[0] >= 'a' && header->typeflag[0] <= 'z') ) 404 return (0); 405 bid += 2; /* 6 bits of variation in an 8-bit field leaves 2 bits. */ 406 407 /* 408 * Check format of mode/uid/gid/mtime/size/rdevmajor/rdevminor fields. 409 */ 410 if (validate_number_field(header->mode, sizeof(header->mode)) == 0 411 || validate_number_field(header->uid, sizeof(header->uid)) == 0 412 || validate_number_field(header->gid, sizeof(header->gid)) == 0 413 || validate_number_field(header->mtime, sizeof(header->mtime)) == 0 414 || validate_number_field(header->size, sizeof(header->size)) == 0 415 || validate_number_field(header->rdevmajor, sizeof(header->rdevmajor)) == 0 416 || validate_number_field(header->rdevminor, sizeof(header->rdevminor)) == 0) { 417 bid = 0; 418 } 419 420 return (bid); 421 } 422 423 static int 424 archive_read_format_tar_options(struct archive_read *a, 425 const char *key, const char *val) 426 { 427 struct tar *tar; 428 int ret = ARCHIVE_FAILED; 429 430 tar = (struct tar *)(a->format->data); 431 if (strcmp(key, "compat-2x") == 0) { 432 /* Handle UTF-8 filenames as libarchive 2.x */ 433 tar->compat_2x = (val != NULL && val[0] != 0); 434 tar->init_default_conversion = tar->compat_2x; 435 return (ARCHIVE_OK); 436 } else if (strcmp(key, "hdrcharset") == 0) { 437 if (val == NULL || val[0] == 0) 438 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 439 "tar: hdrcharset option needs a character-set name"); 440 else { 441 tar->opt_sconv = 442 archive_string_conversion_from_charset( 443 &a->archive, val, 0); 444 if (tar->opt_sconv != NULL) 445 ret = ARCHIVE_OK; 446 else 447 ret = ARCHIVE_FATAL; 448 } 449 return (ret); 450 } else if (strcmp(key, "mac-ext") == 0) { 451 tar->process_mac_extensions = (val != NULL && val[0] != 0); 452 return (ARCHIVE_OK); 453 } else if (strcmp(key, "read_concatenated_archives") == 0) { 454 tar->read_concatenated_archives = (val != NULL && val[0] != 0); 455 return (ARCHIVE_OK); 456 } 457 458 /* Note: The "warn" return is just to inform the options 459 * supervisor that we didn't handle it. It will generate 460 * a suitable error if no one used this option. */ 461 return (ARCHIVE_WARN); 462 } 463 464 /* utility function- this exists to centralize the logic of tracking 465 * how much unconsumed data we have floating around, and to consume 466 * anything outstanding since we're going to do read_aheads 467 */ 468 static void 469 tar_flush_unconsumed(struct archive_read *a, size_t *unconsumed) 470 { 471 if (*unconsumed) { 472 /* 473 void *data = (void *)__archive_read_ahead(a, *unconsumed, NULL); 474 * this block of code is to poison claimed unconsumed space, ensuring 475 * things break if it is in use still. 476 * currently it WILL break things, so enable it only for debugging this issue 477 if (data) { 478 memset(data, 0xff, *unconsumed); 479 } 480 */ 481 __archive_read_consume(a, *unconsumed); 482 *unconsumed = 0; 483 } 484 } 485 486 /* 487 * The function invoked by archive_read_next_header(). This 488 * just sets up a few things and then calls the internal 489 * tar_read_header() function below. 490 */ 491 static int 492 archive_read_format_tar_read_header(struct archive_read *a, 493 struct archive_entry *entry) 494 { 495 /* 496 * When converting tar archives to cpio archives, it is 497 * essential that each distinct file have a distinct inode 498 * number. To simplify this, we keep a static count here to 499 * assign fake dev/inode numbers to each tar entry. Note that 500 * pax format archives may overwrite this with something more 501 * useful. 502 * 503 * Ideally, we would track every file read from the archive so 504 * that we could assign the same dev/ino pair to hardlinks, 505 * but the memory required to store a complete lookup table is 506 * probably not worthwhile just to support the relatively 507 * obscure tar->cpio conversion case. 508 */ 509 static int default_inode; 510 static int default_dev; 511 struct tar *tar; 512 const char *p; 513 const wchar_t *wp; 514 int r; 515 size_t l, unconsumed = 0; 516 517 /* Assign default device/inode values. */ 518 archive_entry_set_dev(entry, 1 + default_dev); /* Don't use zero. */ 519 archive_entry_set_ino(entry, ++default_inode); /* Don't use zero. */ 520 /* Limit generated st_ino number to 16 bits. */ 521 if (default_inode >= 0xffff) { 522 ++default_dev; 523 default_inode = 0; 524 } 525 526 tar = (struct tar *)(a->format->data); 527 tar->entry_offset = 0; 528 gnu_clear_sparse_list(tar); 529 tar->realsize = -1; /* Mark this as "unset" */ 530 tar->realsize_override = 0; 531 532 /* Setup default string conversion. */ 533 tar->sconv = tar->opt_sconv; 534 if (tar->sconv == NULL) { 535 if (!tar->init_default_conversion) { 536 tar->sconv_default = 537 archive_string_default_conversion_for_read(&(a->archive)); 538 tar->init_default_conversion = 1; 539 } 540 tar->sconv = tar->sconv_default; 541 } 542 543 r = tar_read_header(a, tar, entry, &unconsumed); 544 545 tar_flush_unconsumed(a, &unconsumed); 546 547 /* 548 * "non-sparse" files are really just sparse files with 549 * a single block. 550 */ 551 if (tar->sparse_list == NULL) { 552 if (gnu_add_sparse_entry(a, tar, 0, tar->entry_bytes_remaining) 553 != ARCHIVE_OK) 554 return (ARCHIVE_FATAL); 555 } else { 556 struct sparse_block *sb; 557 558 for (sb = tar->sparse_list; sb != NULL; sb = sb->next) { 559 if (!sb->hole) 560 archive_entry_sparse_add_entry(entry, 561 sb->offset, sb->remaining); 562 } 563 } 564 565 if (r == ARCHIVE_OK && archive_entry_filetype(entry) == AE_IFREG) { 566 /* 567 * "Regular" entry with trailing '/' is really 568 * directory: This is needed for certain old tar 569 * variants and even for some broken newer ones. 570 */ 571 if ((wp = archive_entry_pathname_w(entry)) != NULL) { 572 l = wcslen(wp); 573 if (l > 0 && wp[l - 1] == L'/') { 574 archive_entry_set_filetype(entry, AE_IFDIR); 575 tar->entry_bytes_remaining = 0; 576 tar->entry_padding = 0; 577 } 578 } else if ((p = archive_entry_pathname(entry)) != NULL) { 579 l = strlen(p); 580 if (l > 0 && p[l - 1] == '/') { 581 archive_entry_set_filetype(entry, AE_IFDIR); 582 tar->entry_bytes_remaining = 0; 583 tar->entry_padding = 0; 584 } 585 } 586 } 587 return (r); 588 } 589 590 static int 591 archive_read_format_tar_read_data(struct archive_read *a, 592 const void **buff, size_t *size, int64_t *offset) 593 { 594 ssize_t bytes_read; 595 struct tar *tar; 596 struct sparse_block *p; 597 598 tar = (struct tar *)(a->format->data); 599 600 for (;;) { 601 /* Remove exhausted entries from sparse list. */ 602 while (tar->sparse_list != NULL && 603 tar->sparse_list->remaining == 0) { 604 p = tar->sparse_list; 605 tar->sparse_list = p->next; 606 free(p); 607 } 608 609 if (tar->entry_bytes_unconsumed) { 610 __archive_read_consume(a, tar->entry_bytes_unconsumed); 611 tar->entry_bytes_unconsumed = 0; 612 } 613 614 /* If we're at end of file, return EOF. */ 615 if (tar->sparse_list == NULL || 616 tar->entry_bytes_remaining == 0) { 617 if (__archive_read_consume(a, tar->entry_padding) < 0) 618 return (ARCHIVE_FATAL); 619 tar->entry_padding = 0; 620 *buff = NULL; 621 *size = 0; 622 *offset = tar->realsize; 623 return (ARCHIVE_EOF); 624 } 625 626 *buff = __archive_read_ahead(a, 1, &bytes_read); 627 if (bytes_read < 0) 628 return (ARCHIVE_FATAL); 629 if (*buff == NULL) { 630 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 631 "Truncated tar archive"); 632 return (ARCHIVE_FATAL); 633 } 634 if (bytes_read > tar->entry_bytes_remaining) 635 bytes_read = (ssize_t)tar->entry_bytes_remaining; 636 /* Don't read more than is available in the 637 * current sparse block. */ 638 if (tar->sparse_list->remaining < bytes_read) 639 bytes_read = (ssize_t)tar->sparse_list->remaining; 640 *size = bytes_read; 641 *offset = tar->sparse_list->offset; 642 tar->sparse_list->remaining -= bytes_read; 643 tar->sparse_list->offset += bytes_read; 644 tar->entry_bytes_remaining -= bytes_read; 645 tar->entry_bytes_unconsumed = bytes_read; 646 647 if (!tar->sparse_list->hole) 648 return (ARCHIVE_OK); 649 /* Current is hole data and skip this. */ 650 } 651 } 652 653 static int 654 archive_read_format_tar_skip(struct archive_read *a) 655 { 656 int64_t bytes_skipped; 657 int64_t request; 658 struct sparse_block *p; 659 struct tar* tar; 660 661 tar = (struct tar *)(a->format->data); 662 663 /* Do not consume the hole of a sparse file. */ 664 request = 0; 665 for (p = tar->sparse_list; p != NULL; p = p->next) { 666 if (!p->hole) { 667 if (p->remaining >= INT64_MAX - request) { 668 return ARCHIVE_FATAL; 669 } 670 request += p->remaining; 671 } 672 } 673 if (request > tar->entry_bytes_remaining) 674 request = tar->entry_bytes_remaining; 675 request += tar->entry_padding + tar->entry_bytes_unconsumed; 676 677 bytes_skipped = __archive_read_consume(a, request); 678 if (bytes_skipped < 0) 679 return (ARCHIVE_FATAL); 680 681 tar->entry_bytes_remaining = 0; 682 tar->entry_bytes_unconsumed = 0; 683 tar->entry_padding = 0; 684 685 /* Free the sparse list. */ 686 gnu_clear_sparse_list(tar); 687 688 return (ARCHIVE_OK); 689 } 690 691 /* 692 * This function recursively interprets all of the headers associated 693 * with a single entry. 694 */ 695 static int 696 tar_read_header(struct archive_read *a, struct tar *tar, 697 struct archive_entry *entry, size_t *unconsumed) 698 { 699 ssize_t bytes; 700 int err, eof_vol_header; 701 const char *h; 702 const struct archive_entry_header_ustar *header; 703 const struct archive_entry_header_gnutar *gnuheader; 704 705 eof_vol_header = 0; 706 707 /* Loop until we find a workable header record. */ 708 for (;;) { 709 tar_flush_unconsumed(a, unconsumed); 710 711 /* Read 512-byte header record */ 712 h = __archive_read_ahead(a, 512, &bytes); 713 if (bytes < 0) 714 return ((int)bytes); 715 if (bytes == 0) { /* EOF at a block boundary. */ 716 /* Some writers do omit the block of nulls. <sigh> */ 717 return (ARCHIVE_EOF); 718 } 719 if (bytes < 512) { /* Short block at EOF; this is bad. */ 720 archive_set_error(&a->archive, 721 ARCHIVE_ERRNO_FILE_FORMAT, 722 "Truncated tar archive"); 723 return (ARCHIVE_FATAL); 724 } 725 *unconsumed = 512; 726 727 /* Header is workable if it's not an end-of-archive mark. */ 728 if (h[0] != 0 || !archive_block_is_null(h)) 729 break; 730 731 /* Ensure format is set for archives with only null blocks. */ 732 if (a->archive.archive_format_name == NULL) { 733 a->archive.archive_format = ARCHIVE_FORMAT_TAR; 734 a->archive.archive_format_name = "tar"; 735 } 736 737 if (!tar->read_concatenated_archives) { 738 /* Try to consume a second all-null record, as well. */ 739 tar_flush_unconsumed(a, unconsumed); 740 h = __archive_read_ahead(a, 512, NULL); 741 if (h != NULL && h[0] == 0 && archive_block_is_null(h)) 742 __archive_read_consume(a, 512); 743 archive_clear_error(&a->archive); 744 return (ARCHIVE_EOF); 745 } 746 747 /* 748 * We're reading concatenated archives, ignore this block and 749 * loop to get the next. 750 */ 751 } 752 753 /* 754 * Note: If the checksum fails and we return ARCHIVE_RETRY, 755 * then the client is likely to just retry. This is a very 756 * crude way to search for the next valid header! 757 * 758 * TODO: Improve this by implementing a real header scan. 759 */ 760 if (!checksum(a, h)) { 761 tar_flush_unconsumed(a, unconsumed); 762 archive_set_error(&a->archive, EINVAL, "Damaged tar archive"); 763 return (ARCHIVE_RETRY); /* Retryable: Invalid header */ 764 } 765 766 if (++tar->header_recursion_depth > 32) { 767 tar_flush_unconsumed(a, unconsumed); 768 archive_set_error(&a->archive, EINVAL, "Too many special headers"); 769 return (ARCHIVE_WARN); 770 } 771 772 /* Determine the format variant. */ 773 header = (const struct archive_entry_header_ustar *)h; 774 775 switch(header->typeflag[0]) { 776 case 'A': /* Solaris tar ACL */ 777 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 778 a->archive.archive_format_name = "Solaris tar"; 779 err = header_Solaris_ACL(a, tar, entry, h, unconsumed); 780 break; 781 case 'g': /* POSIX-standard 'g' header. */ 782 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 783 a->archive.archive_format_name = "POSIX pax interchange format"; 784 err = header_pax_global(a, tar, entry, h, unconsumed); 785 if (err == ARCHIVE_EOF) 786 return (err); 787 break; 788 case 'K': /* Long link name (GNU tar, others) */ 789 err = header_longlink(a, tar, entry, h, unconsumed); 790 break; 791 case 'L': /* Long filename (GNU tar, others) */ 792 err = header_longname(a, tar, entry, h, unconsumed); 793 break; 794 case 'V': /* GNU volume header */ 795 err = header_volume(a, tar, entry, h, unconsumed); 796 if (err == ARCHIVE_EOF) 797 eof_vol_header = 1; 798 break; 799 case 'X': /* Used by SUN tar; same as 'x'. */ 800 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 801 a->archive.archive_format_name = 802 "POSIX pax interchange format (Sun variant)"; 803 err = header_pax_extensions(a, tar, entry, h, unconsumed); 804 break; 805 case 'x': /* POSIX-standard 'x' header. */ 806 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 807 a->archive.archive_format_name = "POSIX pax interchange format"; 808 err = header_pax_extensions(a, tar, entry, h, unconsumed); 809 break; 810 default: 811 gnuheader = (const struct archive_entry_header_gnutar *)h; 812 if (memcmp(gnuheader->magic, "ustar \0", 8) == 0) { 813 a->archive.archive_format = ARCHIVE_FORMAT_TAR_GNUTAR; 814 a->archive.archive_format_name = "GNU tar format"; 815 err = header_gnutar(a, tar, entry, h, unconsumed); 816 } else if (memcmp(header->magic, "ustar", 5) == 0) { 817 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) { 818 a->archive.archive_format = ARCHIVE_FORMAT_TAR_USTAR; 819 a->archive.archive_format_name = "POSIX ustar format"; 820 } 821 err = header_ustar(a, tar, entry, h); 822 } else { 823 a->archive.archive_format = ARCHIVE_FORMAT_TAR; 824 a->archive.archive_format_name = "tar (non-POSIX)"; 825 err = header_old_tar(a, tar, entry, h); 826 } 827 } 828 if (err == ARCHIVE_FATAL) 829 return (err); 830 831 tar_flush_unconsumed(a, unconsumed); 832 833 h = NULL; 834 header = NULL; 835 836 --tar->header_recursion_depth; 837 /* Yuck. Apple's design here ends up storing long pathname 838 * extensions for both the AppleDouble extension entry and the 839 * regular entry. 840 */ 841 if ((err == ARCHIVE_WARN || err == ARCHIVE_OK) && 842 tar->header_recursion_depth == 0 && 843 tar->process_mac_extensions) { 844 int err2 = read_mac_metadata_blob(a, tar, entry, h, unconsumed); 845 if (err2 < err) 846 err = err2; 847 } 848 849 /* We return warnings or success as-is. Anything else is fatal. */ 850 if (err == ARCHIVE_WARN || err == ARCHIVE_OK) { 851 if (tar->sparse_gnu_pending) { 852 if (tar->sparse_gnu_major == 1 && 853 tar->sparse_gnu_minor == 0) { 854 ssize_t bytes_read; 855 856 tar->sparse_gnu_pending = 0; 857 /* Read initial sparse map. */ 858 bytes_read = gnu_sparse_10_read(a, tar, unconsumed); 859 if (bytes_read < 0) 860 return ((int)bytes_read); 861 tar->entry_bytes_remaining -= bytes_read; 862 } else { 863 archive_set_error(&a->archive, 864 ARCHIVE_ERRNO_MISC, 865 "Unrecognized GNU sparse file format"); 866 return (ARCHIVE_WARN); 867 } 868 tar->sparse_gnu_pending = 0; 869 } 870 return (err); 871 } 872 if (err == ARCHIVE_EOF) { 873 if (!eof_vol_header) { 874 /* EOF when recursively reading a header is bad. */ 875 archive_set_error(&a->archive, EINVAL, 876 "Damaged tar archive"); 877 } else { 878 /* If we encounter just a GNU volume header treat 879 * this situation as an empty archive */ 880 return (ARCHIVE_EOF); 881 } 882 } 883 return (ARCHIVE_FATAL); 884 } 885 886 /* 887 * Return true if block checksum is correct. 888 */ 889 static int 890 checksum(struct archive_read *a, const void *h) 891 { 892 const unsigned char *bytes; 893 const struct archive_entry_header_ustar *header; 894 int check, sum; 895 size_t i; 896 897 (void)a; /* UNUSED */ 898 bytes = (const unsigned char *)h; 899 header = (const struct archive_entry_header_ustar *)h; 900 901 /* Checksum field must hold an octal number */ 902 for (i = 0; i < sizeof(header->checksum); ++i) { 903 char c = header->checksum[i]; 904 if (c != ' ' && c != '\0' && (c < '0' || c > '7')) 905 return 0; 906 } 907 908 /* 909 * Test the checksum. Note that POSIX specifies _unsigned_ 910 * bytes for this calculation. 911 */ 912 sum = (int)tar_atol(header->checksum, sizeof(header->checksum)); 913 check = 0; 914 for (i = 0; i < 148; i++) 915 check += (unsigned char)bytes[i]; 916 for (; i < 156; i++) 917 check += 32; 918 for (; i < 512; i++) 919 check += (unsigned char)bytes[i]; 920 if (sum == check) 921 return (1); 922 923 /* 924 * Repeat test with _signed_ bytes, just in case this archive 925 * was created by an old BSD, Solaris, or HP-UX tar with a 926 * broken checksum calculation. 927 */ 928 check = 0; 929 for (i = 0; i < 148; i++) 930 check += (signed char)bytes[i]; 931 for (; i < 156; i++) 932 check += 32; 933 for (; i < 512; i++) 934 check += (signed char)bytes[i]; 935 if (sum == check) 936 return (1); 937 938 return (0); 939 } 940 941 /* 942 * Return true if this block contains only nulls. 943 */ 944 static int 945 archive_block_is_null(const char *p) 946 { 947 unsigned i; 948 949 for (i = 0; i < 512; i++) 950 if (*p++) 951 return (0); 952 return (1); 953 } 954 955 /* 956 * Interpret 'A' Solaris ACL header 957 */ 958 static int 959 header_Solaris_ACL(struct archive_read *a, struct tar *tar, 960 struct archive_entry *entry, const void *h, size_t *unconsumed) 961 { 962 const struct archive_entry_header_ustar *header; 963 size_t size; 964 int err, acl_type; 965 int64_t type; 966 char *acl, *p; 967 968 /* 969 * read_body_to_string adds a NUL terminator, but we need a little 970 * more to make sure that we don't overrun acl_text later. 971 */ 972 header = (const struct archive_entry_header_ustar *)h; 973 size = (size_t)tar_atol(header->size, sizeof(header->size)); 974 err = read_body_to_string(a, tar, &(tar->acl_text), h, unconsumed); 975 if (err != ARCHIVE_OK) 976 return (err); 977 978 /* Recursively read next header */ 979 err = tar_read_header(a, tar, entry, unconsumed); 980 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 981 return (err); 982 983 /* TODO: Examine the first characters to see if this 984 * is an AIX ACL descriptor. We'll likely never support 985 * them, but it would be polite to recognize and warn when 986 * we do see them. */ 987 988 /* Leading octal number indicates ACL type and number of entries. */ 989 p = acl = tar->acl_text.s; 990 type = 0; 991 while (*p != '\0' && p < acl + size) { 992 if (*p < '0' || *p > '7') { 993 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 994 "Malformed Solaris ACL attribute (invalid digit)"); 995 return(ARCHIVE_WARN); 996 } 997 type <<= 3; 998 type += *p - '0'; 999 if (type > 077777777) { 1000 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1001 "Malformed Solaris ACL attribute (count too large)"); 1002 return (ARCHIVE_WARN); 1003 } 1004 p++; 1005 } 1006 switch ((int)type & ~0777777) { 1007 case 01000000: 1008 /* POSIX.1e ACL */ 1009 acl_type = ARCHIVE_ENTRY_ACL_TYPE_ACCESS; 1010 break; 1011 case 03000000: 1012 /* NFSv4 ACL */ 1013 acl_type = ARCHIVE_ENTRY_ACL_TYPE_NFS4; 1014 break; 1015 default: 1016 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1017 "Malformed Solaris ACL attribute (unsupported type %o)", 1018 (int)type); 1019 return (ARCHIVE_WARN); 1020 } 1021 p++; 1022 1023 if (p >= acl + size) { 1024 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1025 "Malformed Solaris ACL attribute (body overflow)"); 1026 return(ARCHIVE_WARN); 1027 } 1028 1029 /* ACL text is null-terminated; find the end. */ 1030 size -= (p - acl); 1031 acl = p; 1032 1033 while (*p != '\0' && p < acl + size) 1034 p++; 1035 1036 if (tar->sconv_acl == NULL) { 1037 tar->sconv_acl = archive_string_conversion_from_charset( 1038 &(a->archive), "UTF-8", 1); 1039 if (tar->sconv_acl == NULL) 1040 return (ARCHIVE_FATAL); 1041 } 1042 archive_strncpy(&(tar->localname), acl, p - acl); 1043 err = archive_acl_from_text_l(archive_entry_acl(entry), 1044 tar->localname.s, acl_type, tar->sconv_acl); 1045 if (err != ARCHIVE_OK) { 1046 if (errno == ENOMEM) { 1047 archive_set_error(&a->archive, ENOMEM, 1048 "Can't allocate memory for ACL"); 1049 } else 1050 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1051 "Malformed Solaris ACL attribute (unparsable)"); 1052 } 1053 return (err); 1054 } 1055 1056 /* 1057 * Interpret 'K' long linkname header. 1058 */ 1059 static int 1060 header_longlink(struct archive_read *a, struct tar *tar, 1061 struct archive_entry *entry, const void *h, size_t *unconsumed) 1062 { 1063 int err; 1064 1065 err = read_body_to_string(a, tar, &(tar->longlink), h, unconsumed); 1066 if (err != ARCHIVE_OK) 1067 return (err); 1068 err = tar_read_header(a, tar, entry, unconsumed); 1069 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 1070 return (err); 1071 /* Set symlink if symlink already set, else hardlink. */ 1072 archive_entry_copy_link(entry, tar->longlink.s); 1073 return (ARCHIVE_OK); 1074 } 1075 1076 static int 1077 set_conversion_failed_error(struct archive_read *a, 1078 struct archive_string_conv *sconv, const char *name) 1079 { 1080 if (errno == ENOMEM) { 1081 archive_set_error(&a->archive, ENOMEM, 1082 "Can't allocate memory for %s", name); 1083 return (ARCHIVE_FATAL); 1084 } 1085 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 1086 "%s can't be converted from %s to current locale.", 1087 name, archive_string_conversion_charset_name(sconv)); 1088 return (ARCHIVE_WARN); 1089 } 1090 1091 /* 1092 * Interpret 'L' long filename header. 1093 */ 1094 static int 1095 header_longname(struct archive_read *a, struct tar *tar, 1096 struct archive_entry *entry, const void *h, size_t *unconsumed) 1097 { 1098 int err; 1099 1100 err = read_body_to_string(a, tar, &(tar->longname), h, unconsumed); 1101 if (err != ARCHIVE_OK) 1102 return (err); 1103 /* Read and parse "real" header, then override name. */ 1104 err = tar_read_header(a, tar, entry, unconsumed); 1105 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 1106 return (err); 1107 if (archive_entry_copy_pathname_l(entry, tar->longname.s, 1108 archive_strlen(&(tar->longname)), tar->sconv) != 0) 1109 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 1110 return (err); 1111 } 1112 1113 1114 /* 1115 * Interpret 'V' GNU tar volume header. 1116 */ 1117 static int 1118 header_volume(struct archive_read *a, struct tar *tar, 1119 struct archive_entry *entry, const void *h, size_t *unconsumed) 1120 { 1121 (void)h; 1122 1123 /* Just skip this and read the next header. */ 1124 return (tar_read_header(a, tar, entry, unconsumed)); 1125 } 1126 1127 /* 1128 * Read body of an archive entry into an archive_string object. 1129 */ 1130 static int 1131 read_body_to_string(struct archive_read *a, struct tar *tar, 1132 struct archive_string *as, const void *h, size_t *unconsumed) 1133 { 1134 int64_t size; 1135 const struct archive_entry_header_ustar *header; 1136 const void *src; 1137 1138 (void)tar; /* UNUSED */ 1139 header = (const struct archive_entry_header_ustar *)h; 1140 size = tar_atol(header->size, sizeof(header->size)); 1141 if ((size > 1048576) || (size < 0)) { 1142 archive_set_error(&a->archive, EINVAL, 1143 "Special header too large"); 1144 return (ARCHIVE_FATAL); 1145 } 1146 1147 /* Fail if we can't make our buffer big enough. */ 1148 if (archive_string_ensure(as, (size_t)size+1) == NULL) { 1149 archive_set_error(&a->archive, ENOMEM, 1150 "No memory"); 1151 return (ARCHIVE_FATAL); 1152 } 1153 1154 tar_flush_unconsumed(a, unconsumed); 1155 1156 /* Read the body into the string. */ 1157 *unconsumed = (size_t)((size + 511) & ~ 511); 1158 src = __archive_read_ahead(a, *unconsumed, NULL); 1159 if (src == NULL) { 1160 *unconsumed = 0; 1161 return (ARCHIVE_FATAL); 1162 } 1163 memcpy(as->s, src, (size_t)size); 1164 as->s[size] = '\0'; 1165 as->length = (size_t)size; 1166 return (ARCHIVE_OK); 1167 } 1168 1169 /* 1170 * Parse out common header elements. 1171 * 1172 * This would be the same as header_old_tar, except that the 1173 * filename is handled slightly differently for old and POSIX 1174 * entries (POSIX entries support a 'prefix'). This factoring 1175 * allows header_old_tar and header_ustar 1176 * to handle filenames differently, while still putting most of the 1177 * common parsing into one place. 1178 */ 1179 static int 1180 header_common(struct archive_read *a, struct tar *tar, 1181 struct archive_entry *entry, const void *h) 1182 { 1183 const struct archive_entry_header_ustar *header; 1184 char tartype; 1185 int err = ARCHIVE_OK; 1186 1187 header = (const struct archive_entry_header_ustar *)h; 1188 if (header->linkname[0]) 1189 archive_strncpy(&(tar->entry_linkpath), 1190 header->linkname, sizeof(header->linkname)); 1191 else 1192 archive_string_empty(&(tar->entry_linkpath)); 1193 1194 /* Parse out the numeric fields (all are octal) */ 1195 archive_entry_set_mode(entry, 1196 (mode_t)tar_atol(header->mode, sizeof(header->mode))); 1197 archive_entry_set_uid(entry, tar_atol(header->uid, sizeof(header->uid))); 1198 archive_entry_set_gid(entry, tar_atol(header->gid, sizeof(header->gid))); 1199 tar->entry_bytes_remaining = tar_atol(header->size, sizeof(header->size)); 1200 if (tar->entry_bytes_remaining < 0) { 1201 tar->entry_bytes_remaining = 0; 1202 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1203 "Tar entry has negative size"); 1204 return (ARCHIVE_FATAL); 1205 } 1206 if (tar->entry_bytes_remaining == INT64_MAX) { 1207 /* Note: tar_atol returns INT64_MAX on overflow */ 1208 tar->entry_bytes_remaining = 0; 1209 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1210 "Tar entry size overflow"); 1211 return (ARCHIVE_FATAL); 1212 } 1213 tar->realsize = tar->entry_bytes_remaining; 1214 archive_entry_set_size(entry, tar->entry_bytes_remaining); 1215 archive_entry_set_mtime(entry, tar_atol(header->mtime, sizeof(header->mtime)), 0); 1216 1217 /* Handle the tar type flag appropriately. */ 1218 tartype = header->typeflag[0]; 1219 1220 switch (tartype) { 1221 case '1': /* Hard link */ 1222 if (archive_entry_copy_hardlink_l(entry, tar->entry_linkpath.s, 1223 archive_strlen(&(tar->entry_linkpath)), tar->sconv) != 0) { 1224 err = set_conversion_failed_error(a, tar->sconv, 1225 "Linkname"); 1226 if (err == ARCHIVE_FATAL) 1227 return (err); 1228 } 1229 /* 1230 * The following may seem odd, but: Technically, tar 1231 * does not store the file type for a "hard link" 1232 * entry, only the fact that it is a hard link. So, I 1233 * leave the type zero normally. But, pax interchange 1234 * format allows hard links to have data, which 1235 * implies that the underlying entry is a regular 1236 * file. 1237 */ 1238 if (archive_entry_size(entry) > 0) 1239 archive_entry_set_filetype(entry, AE_IFREG); 1240 1241 /* 1242 * A tricky point: Traditionally, tar readers have 1243 * ignored the size field when reading hardlink 1244 * entries, and some writers put non-zero sizes even 1245 * though the body is empty. POSIX blessed this 1246 * convention in the 1988 standard, but broke with 1247 * this tradition in 2001 by permitting hardlink 1248 * entries to store valid bodies in pax interchange 1249 * format, but not in ustar format. Since there is no 1250 * hard and fast way to distinguish pax interchange 1251 * from earlier archives (the 'x' and 'g' entries are 1252 * optional, after all), we need a heuristic. 1253 */ 1254 if (archive_entry_size(entry) == 0) { 1255 /* If the size is already zero, we're done. */ 1256 } else if (a->archive.archive_format 1257 == ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) { 1258 /* Definitely pax extended; must obey hardlink size. */ 1259 } else if (a->archive.archive_format == ARCHIVE_FORMAT_TAR 1260 || a->archive.archive_format == ARCHIVE_FORMAT_TAR_GNUTAR) 1261 { 1262 /* Old-style or GNU tar: we must ignore the size. */ 1263 archive_entry_set_size(entry, 0); 1264 tar->entry_bytes_remaining = 0; 1265 } else if (archive_read_format_tar_bid(a, 50) > 50) { 1266 /* 1267 * We don't know if it's pax: If the bid 1268 * function sees a valid ustar header 1269 * immediately following, then let's ignore 1270 * the hardlink size. 1271 */ 1272 archive_entry_set_size(entry, 0); 1273 tar->entry_bytes_remaining = 0; 1274 } 1275 /* 1276 * TODO: There are still two cases I'd like to handle: 1277 * = a ustar non-pax archive with a hardlink entry at 1278 * end-of-archive. (Look for block of nulls following?) 1279 * = a pax archive that has not seen any pax headers 1280 * and has an entry which is a hardlink entry storing 1281 * a body containing an uncompressed tar archive. 1282 * The first is worth addressing; I don't see any reliable 1283 * way to deal with the second possibility. 1284 */ 1285 break; 1286 case '2': /* Symlink */ 1287 archive_entry_set_filetype(entry, AE_IFLNK); 1288 archive_entry_set_size(entry, 0); 1289 tar->entry_bytes_remaining = 0; 1290 if (archive_entry_copy_symlink_l(entry, tar->entry_linkpath.s, 1291 archive_strlen(&(tar->entry_linkpath)), tar->sconv) != 0) { 1292 err = set_conversion_failed_error(a, tar->sconv, 1293 "Linkname"); 1294 if (err == ARCHIVE_FATAL) 1295 return (err); 1296 } 1297 break; 1298 case '3': /* Character device */ 1299 archive_entry_set_filetype(entry, AE_IFCHR); 1300 archive_entry_set_size(entry, 0); 1301 tar->entry_bytes_remaining = 0; 1302 break; 1303 case '4': /* Block device */ 1304 archive_entry_set_filetype(entry, AE_IFBLK); 1305 archive_entry_set_size(entry, 0); 1306 tar->entry_bytes_remaining = 0; 1307 break; 1308 case '5': /* Dir */ 1309 archive_entry_set_filetype(entry, AE_IFDIR); 1310 archive_entry_set_size(entry, 0); 1311 tar->entry_bytes_remaining = 0; 1312 break; 1313 case '6': /* FIFO device */ 1314 archive_entry_set_filetype(entry, AE_IFIFO); 1315 archive_entry_set_size(entry, 0); 1316 tar->entry_bytes_remaining = 0; 1317 break; 1318 case 'D': /* GNU incremental directory type */ 1319 /* 1320 * No special handling is actually required here. 1321 * It might be nice someday to preprocess the file list and 1322 * provide it to the client, though. 1323 */ 1324 archive_entry_set_filetype(entry, AE_IFDIR); 1325 break; 1326 case 'M': /* GNU "Multi-volume" (remainder of file from last archive)*/ 1327 /* 1328 * As far as I can tell, this is just like a regular file 1329 * entry, except that the contents should be _appended_ to 1330 * the indicated file at the indicated offset. This may 1331 * require some API work to fully support. 1332 */ 1333 break; 1334 case 'N': /* Old GNU "long filename" entry. */ 1335 /* The body of this entry is a script for renaming 1336 * previously-extracted entries. Ugh. It will never 1337 * be supported by libarchive. */ 1338 archive_entry_set_filetype(entry, AE_IFREG); 1339 break; 1340 case 'S': /* GNU sparse files */ 1341 /* 1342 * Sparse files are really just regular files with 1343 * sparse information in the extended area. 1344 */ 1345 /* FALLTHROUGH */ 1346 case '0': 1347 /* 1348 * Enable sparse file "read" support only for regular 1349 * files and explicit GNU sparse files. However, we 1350 * don't allow non-standard file types to be sparse. 1351 */ 1352 tar->sparse_allowed = 1; 1353 /* FALLTHROUGH */ 1354 default: /* Regular file and non-standard types */ 1355 /* 1356 * Per POSIX: non-recognized types should always be 1357 * treated as regular files. 1358 */ 1359 archive_entry_set_filetype(entry, AE_IFREG); 1360 break; 1361 } 1362 return (err); 1363 } 1364 1365 /* 1366 * Parse out header elements for "old-style" tar archives. 1367 */ 1368 static int 1369 header_old_tar(struct archive_read *a, struct tar *tar, 1370 struct archive_entry *entry, const void *h) 1371 { 1372 const struct archive_entry_header_ustar *header; 1373 int err = ARCHIVE_OK, err2; 1374 1375 /* Copy filename over (to ensure null termination). */ 1376 header = (const struct archive_entry_header_ustar *)h; 1377 if (archive_entry_copy_pathname_l(entry, 1378 header->name, sizeof(header->name), tar->sconv) != 0) { 1379 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 1380 if (err == ARCHIVE_FATAL) 1381 return (err); 1382 } 1383 1384 /* Grab rest of common fields */ 1385 err2 = header_common(a, tar, entry, h); 1386 if (err > err2) 1387 err = err2; 1388 1389 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 1390 return (err); 1391 } 1392 1393 /* 1394 * Read a Mac AppleDouble-encoded blob of file metadata, 1395 * if there is one. 1396 */ 1397 static int 1398 read_mac_metadata_blob(struct archive_read *a, struct tar *tar, 1399 struct archive_entry *entry, const void *h, size_t *unconsumed) 1400 { 1401 int64_t size; 1402 size_t msize; 1403 const void *data; 1404 const char *p, *name; 1405 const wchar_t *wp, *wname; 1406 1407 (void)h; /* UNUSED */ 1408 1409 wname = wp = archive_entry_pathname_w(entry); 1410 if (wp != NULL) { 1411 /* Find the last path element. */ 1412 for (; *wp != L'\0'; ++wp) { 1413 if (wp[0] == '/' && wp[1] != L'\0') 1414 wname = wp + 1; 1415 } 1416 /* 1417 * If last path element starts with "._", then 1418 * this is a Mac extension. 1419 */ 1420 if (wname[0] != L'.' || wname[1] != L'_' || wname[2] == L'\0') 1421 return ARCHIVE_OK; 1422 } else { 1423 /* Find the last path element. */ 1424 name = p = archive_entry_pathname(entry); 1425 if (p == NULL) 1426 return (ARCHIVE_FAILED); 1427 for (; *p != '\0'; ++p) { 1428 if (p[0] == '/' && p[1] != '\0') 1429 name = p + 1; 1430 } 1431 /* 1432 * If last path element starts with "._", then 1433 * this is a Mac extension. 1434 */ 1435 if (name[0] != '.' || name[1] != '_' || name[2] == '\0') 1436 return ARCHIVE_OK; 1437 } 1438 1439 /* Read the body as a Mac OS metadata blob. */ 1440 size = archive_entry_size(entry); 1441 msize = (size_t)size; 1442 if (size < 0 || (uintmax_t)msize != (uintmax_t)size) { 1443 *unconsumed = 0; 1444 return (ARCHIVE_FATAL); 1445 } 1446 1447 /* 1448 * TODO: Look beyond the body here to peek at the next header. 1449 * If it's a regular header (not an extension header) 1450 * that has the wrong name, just return the current 1451 * entry as-is, without consuming the body here. 1452 * That would reduce the risk of us mis-identifying 1453 * an ordinary file that just happened to have 1454 * a name starting with "._". 1455 * 1456 * Q: Is the above idea really possible? Even 1457 * when there are GNU or pax extension entries? 1458 */ 1459 data = __archive_read_ahead(a, msize, NULL); 1460 if (data == NULL) { 1461 *unconsumed = 0; 1462 return (ARCHIVE_FATAL); 1463 } 1464 archive_entry_copy_mac_metadata(entry, data, msize); 1465 *unconsumed = (msize + 511) & ~ 511; 1466 tar_flush_unconsumed(a, unconsumed); 1467 return (tar_read_header(a, tar, entry, unconsumed)); 1468 } 1469 1470 /* 1471 * Parse a file header for a pax extended archive entry. 1472 */ 1473 static int 1474 header_pax_global(struct archive_read *a, struct tar *tar, 1475 struct archive_entry *entry, const void *h, size_t *unconsumed) 1476 { 1477 int err; 1478 1479 err = read_body_to_string(a, tar, &(tar->pax_global), h, unconsumed); 1480 if (err != ARCHIVE_OK) 1481 return (err); 1482 err = tar_read_header(a, tar, entry, unconsumed); 1483 return (err); 1484 } 1485 1486 static int 1487 header_pax_extensions(struct archive_read *a, struct tar *tar, 1488 struct archive_entry *entry, const void *h, size_t *unconsumed) 1489 { 1490 int err, err2; 1491 1492 err = read_body_to_string(a, tar, &(tar->pax_header), h, unconsumed); 1493 if (err != ARCHIVE_OK) 1494 return (err); 1495 1496 /* Parse the next header. */ 1497 err = tar_read_header(a, tar, entry, unconsumed); 1498 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 1499 return (err); 1500 1501 /* 1502 * TODO: Parse global/default options into 'entry' struct here 1503 * before handling file-specific options. 1504 * 1505 * This design (parse standard header, then overwrite with pax 1506 * extended attribute data) usually works well, but isn't ideal; 1507 * it would be better to parse the pax extended attributes first 1508 * and then skip any fields in the standard header that were 1509 * defined in the pax header. 1510 */ 1511 err2 = pax_header(a, tar, entry, &tar->pax_header); 1512 err = err_combine(err, err2); 1513 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 1514 return (err); 1515 } 1516 1517 1518 /* 1519 * Parse a file header for a Posix "ustar" archive entry. This also 1520 * handles "pax" or "extended ustar" entries. 1521 */ 1522 static int 1523 header_ustar(struct archive_read *a, struct tar *tar, 1524 struct archive_entry *entry, const void *h) 1525 { 1526 const struct archive_entry_header_ustar *header; 1527 struct archive_string *as; 1528 int err = ARCHIVE_OK, r; 1529 1530 header = (const struct archive_entry_header_ustar *)h; 1531 1532 /* Copy name into an internal buffer to ensure null-termination. */ 1533 as = &(tar->entry_pathname); 1534 if (header->prefix[0]) { 1535 archive_strncpy(as, header->prefix, sizeof(header->prefix)); 1536 if (as->s[archive_strlen(as) - 1] != '/') 1537 archive_strappend_char(as, '/'); 1538 archive_strncat(as, header->name, sizeof(header->name)); 1539 } else { 1540 archive_strncpy(as, header->name, sizeof(header->name)); 1541 } 1542 if (archive_entry_copy_pathname_l(entry, as->s, archive_strlen(as), 1543 tar->sconv) != 0) { 1544 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 1545 if (err == ARCHIVE_FATAL) 1546 return (err); 1547 } 1548 1549 /* Handle rest of common fields. */ 1550 r = header_common(a, tar, entry, h); 1551 if (r == ARCHIVE_FATAL) 1552 return (r); 1553 if (r < err) 1554 err = r; 1555 1556 /* Handle POSIX ustar fields. */ 1557 if (archive_entry_copy_uname_l(entry, 1558 header->uname, sizeof(header->uname), tar->sconv) != 0) { 1559 err = set_conversion_failed_error(a, tar->sconv, "Uname"); 1560 if (err == ARCHIVE_FATAL) 1561 return (err); 1562 } 1563 1564 if (archive_entry_copy_gname_l(entry, 1565 header->gname, sizeof(header->gname), tar->sconv) != 0) { 1566 err = set_conversion_failed_error(a, tar->sconv, "Gname"); 1567 if (err == ARCHIVE_FATAL) 1568 return (err); 1569 } 1570 1571 /* Parse out device numbers only for char and block specials. */ 1572 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') { 1573 archive_entry_set_rdevmajor(entry, (dev_t) 1574 tar_atol(header->rdevmajor, sizeof(header->rdevmajor))); 1575 archive_entry_set_rdevminor(entry, (dev_t) 1576 tar_atol(header->rdevminor, sizeof(header->rdevminor))); 1577 } 1578 1579 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 1580 1581 return (err); 1582 } 1583 1584 1585 /* 1586 * Parse the pax extended attributes record. 1587 * 1588 * Returns non-zero if there's an error in the data. 1589 */ 1590 static int 1591 pax_header(struct archive_read *a, struct tar *tar, 1592 struct archive_entry *entry, struct archive_string *in_as) 1593 { 1594 size_t attr_length, l, line_length, value_length; 1595 char *p; 1596 char *key, *value; 1597 struct archive_string *as; 1598 struct archive_string_conv *sconv; 1599 int err, err2; 1600 char *attr = in_as->s; 1601 1602 attr_length = in_as->length; 1603 tar->pax_hdrcharset_binary = 0; 1604 archive_string_empty(&(tar->entry_gname)); 1605 archive_string_empty(&(tar->entry_linkpath)); 1606 archive_string_empty(&(tar->entry_pathname)); 1607 archive_string_empty(&(tar->entry_pathname_override)); 1608 archive_string_empty(&(tar->entry_uname)); 1609 err = ARCHIVE_OK; 1610 while (attr_length > 0) { 1611 /* Parse decimal length field at start of line. */ 1612 line_length = 0; 1613 l = attr_length; 1614 p = attr; /* Record start of line. */ 1615 while (l>0) { 1616 if (*p == ' ') { 1617 p++; 1618 l--; 1619 break; 1620 } 1621 if (*p < '0' || *p > '9') { 1622 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1623 "Ignoring malformed pax extended attributes"); 1624 return (ARCHIVE_WARN); 1625 } 1626 line_length *= 10; 1627 line_length += *p - '0'; 1628 if (line_length > 999999) { 1629 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1630 "Rejecting pax extended attribute > 1MB"); 1631 return (ARCHIVE_WARN); 1632 } 1633 p++; 1634 l--; 1635 } 1636 1637 /* 1638 * Parsed length must be no bigger than available data, 1639 * at least 1, and the last character of the line must 1640 * be '\n'. 1641 */ 1642 if (line_length > attr_length 1643 || line_length < 1 1644 || attr[line_length - 1] != '\n') 1645 { 1646 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1647 "Ignoring malformed pax extended attribute"); 1648 return (ARCHIVE_WARN); 1649 } 1650 1651 /* Null-terminate the line. */ 1652 attr[line_length - 1] = '\0'; 1653 1654 /* Find end of key and null terminate it. */ 1655 key = p; 1656 if (key[0] == '=') 1657 return (-1); 1658 while (*p && *p != '=') 1659 ++p; 1660 if (*p == '\0') { 1661 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1662 "Invalid pax extended attributes"); 1663 return (ARCHIVE_WARN); 1664 } 1665 *p = '\0'; 1666 1667 value = p + 1; 1668 1669 /* Some values may be binary data */ 1670 value_length = attr + line_length - 1 - value; 1671 1672 /* Identify this attribute and set it in the entry. */ 1673 err2 = pax_attribute(a, tar, entry, key, value, value_length); 1674 if (err2 == ARCHIVE_FATAL) 1675 return (err2); 1676 err = err_combine(err, err2); 1677 1678 /* Skip to next line */ 1679 attr += line_length; 1680 attr_length -= line_length; 1681 } 1682 1683 /* 1684 * PAX format uses UTF-8 as default charset for its metadata 1685 * unless hdrcharset=BINARY is present in its header. 1686 * We apply the charset specified by the hdrcharset option only 1687 * when the hdrcharset attribute(in PAX header) is BINARY because 1688 * we respect the charset described in PAX header and BINARY also 1689 * means that metadata(filename,uname and gname) character-set 1690 * is unknown. 1691 */ 1692 if (tar->pax_hdrcharset_binary) 1693 sconv = tar->opt_sconv; 1694 else { 1695 sconv = archive_string_conversion_from_charset( 1696 &(a->archive), "UTF-8", 1); 1697 if (sconv == NULL) 1698 return (ARCHIVE_FATAL); 1699 if (tar->compat_2x) 1700 archive_string_conversion_set_opt(sconv, 1701 SCONV_SET_OPT_UTF8_LIBARCHIVE2X); 1702 } 1703 1704 if (archive_strlen(&(tar->entry_gname)) > 0) { 1705 if (archive_entry_copy_gname_l(entry, tar->entry_gname.s, 1706 archive_strlen(&(tar->entry_gname)), sconv) != 0) { 1707 err = set_conversion_failed_error(a, sconv, "Gname"); 1708 if (err == ARCHIVE_FATAL) 1709 return (err); 1710 /* Use a converted an original name. */ 1711 archive_entry_copy_gname(entry, tar->entry_gname.s); 1712 } 1713 } 1714 if (archive_strlen(&(tar->entry_linkpath)) > 0) { 1715 if (archive_entry_copy_link_l(entry, tar->entry_linkpath.s, 1716 archive_strlen(&(tar->entry_linkpath)), sconv) != 0) { 1717 err = set_conversion_failed_error(a, sconv, "Linkname"); 1718 if (err == ARCHIVE_FATAL) 1719 return (err); 1720 /* Use a converted an original name. */ 1721 archive_entry_copy_link(entry, tar->entry_linkpath.s); 1722 } 1723 } 1724 /* 1725 * Some extensions (such as the GNU sparse file extensions) 1726 * deliberately store a synthetic name under the regular 'path' 1727 * attribute and the real file name under a different attribute. 1728 * Since we're supposed to not care about the order, we 1729 * have no choice but to store all of the various filenames 1730 * we find and figure it all out afterwards. This is the 1731 * figuring out part. 1732 */ 1733 as = NULL; 1734 if (archive_strlen(&(tar->entry_pathname_override)) > 0) 1735 as = &(tar->entry_pathname_override); 1736 else if (archive_strlen(&(tar->entry_pathname)) > 0) 1737 as = &(tar->entry_pathname); 1738 if (as != NULL) { 1739 if (archive_entry_copy_pathname_l(entry, as->s, 1740 archive_strlen(as), sconv) != 0) { 1741 err = set_conversion_failed_error(a, sconv, "Pathname"); 1742 if (err == ARCHIVE_FATAL) 1743 return (err); 1744 /* Use a converted an original name. */ 1745 archive_entry_copy_pathname(entry, as->s); 1746 } 1747 } 1748 if (archive_strlen(&(tar->entry_uname)) > 0) { 1749 if (archive_entry_copy_uname_l(entry, tar->entry_uname.s, 1750 archive_strlen(&(tar->entry_uname)), sconv) != 0) { 1751 err = set_conversion_failed_error(a, sconv, "Uname"); 1752 if (err == ARCHIVE_FATAL) 1753 return (err); 1754 /* Use a converted an original name. */ 1755 archive_entry_copy_uname(entry, tar->entry_uname.s); 1756 } 1757 } 1758 return (err); 1759 } 1760 1761 static int 1762 pax_attribute_xattr(struct archive_entry *entry, 1763 const char *name, const char *value) 1764 { 1765 char *name_decoded; 1766 void *value_decoded; 1767 size_t value_len; 1768 1769 if (strlen(name) < 18 || (memcmp(name, "LIBARCHIVE.xattr.", 17)) != 0) 1770 return 3; 1771 1772 name += 17; 1773 1774 /* URL-decode name */ 1775 name_decoded = url_decode(name); 1776 if (name_decoded == NULL) 1777 return 2; 1778 1779 /* Base-64 decode value */ 1780 value_decoded = base64_decode(value, strlen(value), &value_len); 1781 if (value_decoded == NULL) { 1782 free(name_decoded); 1783 return 1; 1784 } 1785 1786 archive_entry_xattr_add_entry(entry, name_decoded, 1787 value_decoded, value_len); 1788 1789 free(name_decoded); 1790 free(value_decoded); 1791 return 0; 1792 } 1793 1794 static int 1795 pax_attribute_schily_xattr(struct archive_entry *entry, 1796 const char *name, const char *value, size_t value_length) 1797 { 1798 if (strlen(name) < 14 || (memcmp(name, "SCHILY.xattr.", 13)) != 0) 1799 return 1; 1800 1801 name += 13; 1802 1803 archive_entry_xattr_add_entry(entry, name, value, value_length); 1804 1805 return 0; 1806 } 1807 1808 static int 1809 pax_attribute_rht_security_selinux(struct archive_entry *entry, 1810 const char *value, size_t value_length) 1811 { 1812 archive_entry_xattr_add_entry(entry, "security.selinux", 1813 value, value_length); 1814 1815 return 0; 1816 } 1817 1818 static int 1819 pax_attribute_acl(struct archive_read *a, struct tar *tar, 1820 struct archive_entry *entry, const char *value, int type) 1821 { 1822 int r; 1823 const char* errstr; 1824 1825 switch (type) { 1826 case ARCHIVE_ENTRY_ACL_TYPE_ACCESS: 1827 errstr = "SCHILY.acl.access"; 1828 break; 1829 case ARCHIVE_ENTRY_ACL_TYPE_DEFAULT: 1830 errstr = "SCHILY.acl.default"; 1831 break; 1832 case ARCHIVE_ENTRY_ACL_TYPE_NFS4: 1833 errstr = "SCHILY.acl.ace"; 1834 break; 1835 default: 1836 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1837 "Unknown ACL type: %d", type); 1838 return(ARCHIVE_FATAL); 1839 } 1840 1841 if (tar->sconv_acl == NULL) { 1842 tar->sconv_acl = 1843 archive_string_conversion_from_charset( 1844 &(a->archive), "UTF-8", 1); 1845 if (tar->sconv_acl == NULL) 1846 return (ARCHIVE_FATAL); 1847 } 1848 1849 r = archive_acl_from_text_l(archive_entry_acl(entry), value, type, 1850 tar->sconv_acl); 1851 if (r != ARCHIVE_OK) { 1852 if (r == ARCHIVE_FATAL) { 1853 archive_set_error(&a->archive, ENOMEM, 1854 "%s %s", "Can't allocate memory for ", 1855 errstr); 1856 return (r); 1857 } 1858 archive_set_error(&a->archive, 1859 ARCHIVE_ERRNO_MISC, "%s %s", "Parse error: ", errstr); 1860 } 1861 return (r); 1862 } 1863 1864 /* 1865 * Parse a single key=value attribute. key/value pointers are 1866 * assumed to point into reasonably long-lived storage. 1867 * 1868 * Note that POSIX reserves all-lowercase keywords. Vendor-specific 1869 * extensions should always have keywords of the form "VENDOR.attribute" 1870 * In particular, it's quite feasible to support many different 1871 * vendor extensions here. I'm using "LIBARCHIVE" for extensions 1872 * unique to this library. 1873 * 1874 * Investigate other vendor-specific extensions and see if 1875 * any of them look useful. 1876 */ 1877 static int 1878 pax_attribute(struct archive_read *a, struct tar *tar, 1879 struct archive_entry *entry, const char *key, const char *value, size_t value_length) 1880 { 1881 int64_t s; 1882 long n; 1883 int err = ARCHIVE_OK, r; 1884 1885 #ifndef __FreeBSD__ 1886 if (value == NULL) 1887 value = ""; /* Disable compiler warning; do not pass 1888 * NULL pointer to strlen(). */ 1889 #endif 1890 switch (key[0]) { 1891 case 'G': 1892 /* Reject GNU.sparse.* headers on non-regular files. */ 1893 if (strncmp(key, "GNU.sparse", 10) == 0 && 1894 !tar->sparse_allowed) { 1895 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1896 "Non-regular file cannot be sparse"); 1897 return (ARCHIVE_FATAL); 1898 } 1899 1900 /* GNU "0.0" sparse pax format. */ 1901 if (strcmp(key, "GNU.sparse.numblocks") == 0) { 1902 tar->sparse_offset = -1; 1903 tar->sparse_numbytes = -1; 1904 tar->sparse_gnu_major = 0; 1905 tar->sparse_gnu_minor = 0; 1906 } 1907 if (strcmp(key, "GNU.sparse.offset") == 0) { 1908 tar->sparse_offset = tar_atol10(value, strlen(value)); 1909 if (tar->sparse_numbytes != -1) { 1910 if (gnu_add_sparse_entry(a, tar, 1911 tar->sparse_offset, tar->sparse_numbytes) 1912 != ARCHIVE_OK) 1913 return (ARCHIVE_FATAL); 1914 tar->sparse_offset = -1; 1915 tar->sparse_numbytes = -1; 1916 } 1917 } 1918 if (strcmp(key, "GNU.sparse.numbytes") == 0) { 1919 tar->sparse_numbytes = tar_atol10(value, strlen(value)); 1920 if (tar->sparse_offset != -1) { 1921 if (gnu_add_sparse_entry(a, tar, 1922 tar->sparse_offset, tar->sparse_numbytes) 1923 != ARCHIVE_OK) 1924 return (ARCHIVE_FATAL); 1925 tar->sparse_offset = -1; 1926 tar->sparse_numbytes = -1; 1927 } 1928 } 1929 if (strcmp(key, "GNU.sparse.size") == 0) { 1930 tar->realsize = tar_atol10(value, strlen(value)); 1931 archive_entry_set_size(entry, tar->realsize); 1932 tar->realsize_override = 1; 1933 } 1934 1935 /* GNU "0.1" sparse pax format. */ 1936 if (strcmp(key, "GNU.sparse.map") == 0) { 1937 tar->sparse_gnu_major = 0; 1938 tar->sparse_gnu_minor = 1; 1939 if (gnu_sparse_01_parse(a, tar, value) != ARCHIVE_OK) 1940 return (ARCHIVE_WARN); 1941 } 1942 1943 /* GNU "1.0" sparse pax format */ 1944 if (strcmp(key, "GNU.sparse.major") == 0) { 1945 tar->sparse_gnu_major = (int)tar_atol10(value, strlen(value)); 1946 tar->sparse_gnu_pending = 1; 1947 } 1948 if (strcmp(key, "GNU.sparse.minor") == 0) { 1949 tar->sparse_gnu_minor = (int)tar_atol10(value, strlen(value)); 1950 tar->sparse_gnu_pending = 1; 1951 } 1952 if (strcmp(key, "GNU.sparse.name") == 0) { 1953 /* 1954 * The real filename; when storing sparse 1955 * files, GNU tar puts a synthesized name into 1956 * the regular 'path' attribute in an attempt 1957 * to limit confusion. ;-) 1958 */ 1959 archive_strcpy(&(tar->entry_pathname_override), value); 1960 } 1961 if (strcmp(key, "GNU.sparse.realsize") == 0) { 1962 tar->realsize = tar_atol10(value, strlen(value)); 1963 archive_entry_set_size(entry, tar->realsize); 1964 tar->realsize_override = 1; 1965 } 1966 break; 1967 case 'L': 1968 /* Our extensions */ 1969 /* TODO: Handle arbitrary extended attributes... */ 1970 /* 1971 if (strcmp(key, "LIBARCHIVE.xxxxxxx") == 0) 1972 archive_entry_set_xxxxxx(entry, value); 1973 */ 1974 if (strcmp(key, "LIBARCHIVE.creationtime") == 0) { 1975 pax_time(value, &s, &n); 1976 archive_entry_set_birthtime(entry, s, n); 1977 } 1978 if (strcmp(key, "LIBARCHIVE.symlinktype") == 0) { 1979 if (strcmp(value, "file") == 0) { 1980 archive_entry_set_symlink_type(entry, 1981 AE_SYMLINK_TYPE_FILE); 1982 } else if (strcmp(value, "dir") == 0) { 1983 archive_entry_set_symlink_type(entry, 1984 AE_SYMLINK_TYPE_DIRECTORY); 1985 } 1986 } 1987 if (memcmp(key, "LIBARCHIVE.xattr.", 17) == 0) 1988 pax_attribute_xattr(entry, key, value); 1989 break; 1990 case 'R': 1991 /* GNU tar uses RHT.security header to store SELinux xattrs 1992 * SCHILY.xattr.security.selinux == RHT.security.selinux */ 1993 if (strcmp(key, "RHT.security.selinux") == 0) { 1994 pax_attribute_rht_security_selinux(entry, value, 1995 value_length); 1996 } 1997 break; 1998 case 'S': 1999 /* We support some keys used by the "star" archiver */ 2000 if (strcmp(key, "SCHILY.acl.access") == 0) { 2001 r = pax_attribute_acl(a, tar, entry, value, 2002 ARCHIVE_ENTRY_ACL_TYPE_ACCESS); 2003 if (r == ARCHIVE_FATAL) 2004 return (r); 2005 } else if (strcmp(key, "SCHILY.acl.default") == 0) { 2006 r = pax_attribute_acl(a, tar, entry, value, 2007 ARCHIVE_ENTRY_ACL_TYPE_DEFAULT); 2008 if (r == ARCHIVE_FATAL) 2009 return (r); 2010 } else if (strcmp(key, "SCHILY.acl.ace") == 0) { 2011 r = pax_attribute_acl(a, tar, entry, value, 2012 ARCHIVE_ENTRY_ACL_TYPE_NFS4); 2013 if (r == ARCHIVE_FATAL) 2014 return (r); 2015 } else if (strcmp(key, "SCHILY.devmajor") == 0) { 2016 archive_entry_set_rdevmajor(entry, 2017 (dev_t)tar_atol10(value, strlen(value))); 2018 } else if (strcmp(key, "SCHILY.devminor") == 0) { 2019 archive_entry_set_rdevminor(entry, 2020 (dev_t)tar_atol10(value, strlen(value))); 2021 } else if (strcmp(key, "SCHILY.fflags") == 0) { 2022 archive_entry_copy_fflags_text(entry, value); 2023 } else if (strcmp(key, "SCHILY.dev") == 0) { 2024 archive_entry_set_dev(entry, 2025 (dev_t)tar_atol10(value, strlen(value))); 2026 } else if (strcmp(key, "SCHILY.ino") == 0) { 2027 archive_entry_set_ino(entry, 2028 tar_atol10(value, strlen(value))); 2029 } else if (strcmp(key, "SCHILY.nlink") == 0) { 2030 archive_entry_set_nlink(entry, (unsigned) 2031 tar_atol10(value, strlen(value))); 2032 } else if (strcmp(key, "SCHILY.realsize") == 0) { 2033 tar->realsize = tar_atol10(value, strlen(value)); 2034 tar->realsize_override = 1; 2035 archive_entry_set_size(entry, tar->realsize); 2036 } else if (strncmp(key, "SCHILY.xattr.", 13) == 0) { 2037 pax_attribute_schily_xattr(entry, key, value, 2038 value_length); 2039 } else if (strcmp(key, "SUN.holesdata") == 0) { 2040 /* A Solaris extension for sparse. */ 2041 r = solaris_sparse_parse(a, tar, entry, value); 2042 if (r < err) { 2043 if (r == ARCHIVE_FATAL) 2044 return (r); 2045 err = r; 2046 archive_set_error(&a->archive, 2047 ARCHIVE_ERRNO_MISC, 2048 "Parse error: SUN.holesdata"); 2049 } 2050 } 2051 break; 2052 case 'a': 2053 if (strcmp(key, "atime") == 0) { 2054 pax_time(value, &s, &n); 2055 archive_entry_set_atime(entry, s, n); 2056 } 2057 break; 2058 case 'c': 2059 if (strcmp(key, "ctime") == 0) { 2060 pax_time(value, &s, &n); 2061 archive_entry_set_ctime(entry, s, n); 2062 } else if (strcmp(key, "charset") == 0) { 2063 /* TODO: Publish charset information in entry. */ 2064 } else if (strcmp(key, "comment") == 0) { 2065 /* TODO: Publish comment in entry. */ 2066 } 2067 break; 2068 case 'g': 2069 if (strcmp(key, "gid") == 0) { 2070 archive_entry_set_gid(entry, 2071 tar_atol10(value, strlen(value))); 2072 } else if (strcmp(key, "gname") == 0) { 2073 archive_strcpy(&(tar->entry_gname), value); 2074 } 2075 break; 2076 case 'h': 2077 if (strcmp(key, "hdrcharset") == 0) { 2078 if (strcmp(value, "BINARY") == 0) 2079 /* Binary mode. */ 2080 tar->pax_hdrcharset_binary = 1; 2081 else if (strcmp(value, "ISO-IR 10646 2000 UTF-8") == 0) 2082 tar->pax_hdrcharset_binary = 0; 2083 } 2084 break; 2085 case 'l': 2086 /* pax interchange doesn't distinguish hardlink vs. symlink. */ 2087 if (strcmp(key, "linkpath") == 0) { 2088 archive_strcpy(&(tar->entry_linkpath), value); 2089 } 2090 break; 2091 case 'm': 2092 if (strcmp(key, "mtime") == 0) { 2093 pax_time(value, &s, &n); 2094 archive_entry_set_mtime(entry, s, n); 2095 } 2096 break; 2097 case 'p': 2098 if (strcmp(key, "path") == 0) { 2099 archive_strcpy(&(tar->entry_pathname), value); 2100 } 2101 break; 2102 case 'r': 2103 /* POSIX has reserved 'realtime.*' */ 2104 break; 2105 case 's': 2106 /* POSIX has reserved 'security.*' */ 2107 /* Someday: if (strcmp(key, "security.acl") == 0) { ... } */ 2108 if (strcmp(key, "size") == 0) { 2109 /* "size" is the size of the data in the entry. */ 2110 tar->entry_bytes_remaining 2111 = tar_atol10(value, strlen(value)); 2112 if (tar->entry_bytes_remaining < 0) { 2113 tar->entry_bytes_remaining = 0; 2114 archive_set_error(&a->archive, 2115 ARCHIVE_ERRNO_MISC, 2116 "Tar size attribute is negative"); 2117 return (ARCHIVE_FATAL); 2118 } 2119 if (tar->entry_bytes_remaining == INT64_MAX) { 2120 /* Note: tar_atol returns INT64_MAX on overflow */ 2121 tar->entry_bytes_remaining = 0; 2122 archive_set_error(&a->archive, 2123 ARCHIVE_ERRNO_MISC, 2124 "Tar size attribute overflow"); 2125 return (ARCHIVE_FATAL); 2126 } 2127 /* 2128 * The "size" pax header keyword always overrides the 2129 * "size" field in the tar header. 2130 * GNU.sparse.realsize, GNU.sparse.size and 2131 * SCHILY.realsize override this value. 2132 */ 2133 if (!tar->realsize_override) { 2134 archive_entry_set_size(entry, 2135 tar->entry_bytes_remaining); 2136 tar->realsize 2137 = tar->entry_bytes_remaining; 2138 } 2139 } 2140 break; 2141 case 'u': 2142 if (strcmp(key, "uid") == 0) { 2143 archive_entry_set_uid(entry, 2144 tar_atol10(value, strlen(value))); 2145 } else if (strcmp(key, "uname") == 0) { 2146 archive_strcpy(&(tar->entry_uname), value); 2147 } 2148 break; 2149 } 2150 return (err); 2151 } 2152 2153 2154 2155 /* 2156 * parse a decimal time value, which may include a fractional portion 2157 */ 2158 static void 2159 pax_time(const char *p, int64_t *ps, long *pn) 2160 { 2161 char digit; 2162 int64_t s; 2163 unsigned long l; 2164 int sign; 2165 int64_t limit, last_digit_limit; 2166 2167 limit = INT64_MAX / 10; 2168 last_digit_limit = INT64_MAX % 10; 2169 2170 s = 0; 2171 sign = 1; 2172 if (*p == '-') { 2173 sign = -1; 2174 p++; 2175 } 2176 while (*p >= '0' && *p <= '9') { 2177 digit = *p - '0'; 2178 if (s > limit || 2179 (s == limit && digit > last_digit_limit)) { 2180 s = INT64_MAX; 2181 break; 2182 } 2183 s = (s * 10) + digit; 2184 ++p; 2185 } 2186 2187 *ps = s * sign; 2188 2189 /* Calculate nanoseconds. */ 2190 *pn = 0; 2191 2192 if (*p != '.') 2193 return; 2194 2195 l = 100000000UL; 2196 do { 2197 ++p; 2198 if (*p >= '0' && *p <= '9') 2199 *pn += (*p - '0') * l; 2200 else 2201 break; 2202 } while (l /= 10); 2203 } 2204 2205 /* 2206 * Parse GNU tar header 2207 */ 2208 static int 2209 header_gnutar(struct archive_read *a, struct tar *tar, 2210 struct archive_entry *entry, const void *h, size_t *unconsumed) 2211 { 2212 const struct archive_entry_header_gnutar *header; 2213 int64_t t; 2214 int err = ARCHIVE_OK; 2215 2216 /* 2217 * GNU header is like POSIX ustar, except 'prefix' is 2218 * replaced with some other fields. This also means the 2219 * filename is stored as in old-style archives. 2220 */ 2221 2222 /* Grab fields common to all tar variants. */ 2223 err = header_common(a, tar, entry, h); 2224 if (err == ARCHIVE_FATAL) 2225 return (err); 2226 2227 /* Copy filename over (to ensure null termination). */ 2228 header = (const struct archive_entry_header_gnutar *)h; 2229 if (archive_entry_copy_pathname_l(entry, 2230 header->name, sizeof(header->name), tar->sconv) != 0) { 2231 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 2232 if (err == ARCHIVE_FATAL) 2233 return (err); 2234 } 2235 2236 /* Fields common to ustar and GNU */ 2237 /* XXX Can the following be factored out since it's common 2238 * to ustar and gnu tar? Is it okay to move it down into 2239 * header_common, perhaps? */ 2240 if (archive_entry_copy_uname_l(entry, 2241 header->uname, sizeof(header->uname), tar->sconv) != 0) { 2242 err = set_conversion_failed_error(a, tar->sconv, "Uname"); 2243 if (err == ARCHIVE_FATAL) 2244 return (err); 2245 } 2246 2247 if (archive_entry_copy_gname_l(entry, 2248 header->gname, sizeof(header->gname), tar->sconv) != 0) { 2249 err = set_conversion_failed_error(a, tar->sconv, "Gname"); 2250 if (err == ARCHIVE_FATAL) 2251 return (err); 2252 } 2253 2254 /* Parse out device numbers only for char and block specials */ 2255 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') { 2256 archive_entry_set_rdevmajor(entry, (dev_t) 2257 tar_atol(header->rdevmajor, sizeof(header->rdevmajor))); 2258 archive_entry_set_rdevminor(entry, (dev_t) 2259 tar_atol(header->rdevminor, sizeof(header->rdevminor))); 2260 } else 2261 archive_entry_set_rdev(entry, 0); 2262 2263 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 2264 2265 /* Grab GNU-specific fields. */ 2266 t = tar_atol(header->atime, sizeof(header->atime)); 2267 if (t > 0) 2268 archive_entry_set_atime(entry, t, 0); 2269 t = tar_atol(header->ctime, sizeof(header->ctime)); 2270 if (t > 0) 2271 archive_entry_set_ctime(entry, t, 0); 2272 2273 if (header->realsize[0] != 0) { 2274 tar->realsize 2275 = tar_atol(header->realsize, sizeof(header->realsize)); 2276 archive_entry_set_size(entry, tar->realsize); 2277 tar->realsize_override = 1; 2278 } 2279 2280 if (header->sparse[0].offset[0] != 0) { 2281 if (gnu_sparse_old_read(a, tar, header, unconsumed) 2282 != ARCHIVE_OK) 2283 return (ARCHIVE_FATAL); 2284 } else { 2285 if (header->isextended[0] != 0) { 2286 /* XXX WTF? XXX */ 2287 } 2288 } 2289 2290 return (err); 2291 } 2292 2293 static int 2294 gnu_add_sparse_entry(struct archive_read *a, struct tar *tar, 2295 int64_t offset, int64_t remaining) 2296 { 2297 struct sparse_block *p; 2298 2299 p = (struct sparse_block *)calloc(1, sizeof(*p)); 2300 if (p == NULL) { 2301 archive_set_error(&a->archive, ENOMEM, "Out of memory"); 2302 return (ARCHIVE_FATAL); 2303 } 2304 if (tar->sparse_last != NULL) 2305 tar->sparse_last->next = p; 2306 else 2307 tar->sparse_list = p; 2308 tar->sparse_last = p; 2309 if (remaining < 0 || offset < 0 || offset > INT64_MAX - remaining) { 2310 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Malformed sparse map data"); 2311 return (ARCHIVE_FATAL); 2312 } 2313 p->offset = offset; 2314 p->remaining = remaining; 2315 return (ARCHIVE_OK); 2316 } 2317 2318 static void 2319 gnu_clear_sparse_list(struct tar *tar) 2320 { 2321 struct sparse_block *p; 2322 2323 while (tar->sparse_list != NULL) { 2324 p = tar->sparse_list; 2325 tar->sparse_list = p->next; 2326 free(p); 2327 } 2328 tar->sparse_last = NULL; 2329 } 2330 2331 /* 2332 * GNU tar old-format sparse data. 2333 * 2334 * GNU old-format sparse data is stored in a fixed-field 2335 * format. Offset/size values are 11-byte octal fields (same 2336 * format as 'size' field in ustart header). These are 2337 * stored in the header, allocating subsequent header blocks 2338 * as needed. Extending the header in this way is a pretty 2339 * severe POSIX violation; this design has earned GNU tar a 2340 * lot of criticism. 2341 */ 2342 2343 static int 2344 gnu_sparse_old_read(struct archive_read *a, struct tar *tar, 2345 const struct archive_entry_header_gnutar *header, size_t *unconsumed) 2346 { 2347 ssize_t bytes_read; 2348 const void *data; 2349 struct extended { 2350 struct gnu_sparse sparse[21]; 2351 char isextended[1]; 2352 char padding[7]; 2353 }; 2354 const struct extended *ext; 2355 2356 if (gnu_sparse_old_parse(a, tar, header->sparse, 4) != ARCHIVE_OK) 2357 return (ARCHIVE_FATAL); 2358 if (header->isextended[0] == 0) 2359 return (ARCHIVE_OK); 2360 2361 do { 2362 tar_flush_unconsumed(a, unconsumed); 2363 data = __archive_read_ahead(a, 512, &bytes_read); 2364 if (bytes_read < 0) 2365 return (ARCHIVE_FATAL); 2366 if (bytes_read < 512) { 2367 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 2368 "Truncated tar archive " 2369 "detected while reading sparse file data"); 2370 return (ARCHIVE_FATAL); 2371 } 2372 *unconsumed = 512; 2373 ext = (const struct extended *)data; 2374 if (gnu_sparse_old_parse(a, tar, ext->sparse, 21) != ARCHIVE_OK) 2375 return (ARCHIVE_FATAL); 2376 } while (ext->isextended[0] != 0); 2377 if (tar->sparse_list != NULL) 2378 tar->entry_offset = tar->sparse_list->offset; 2379 return (ARCHIVE_OK); 2380 } 2381 2382 static int 2383 gnu_sparse_old_parse(struct archive_read *a, struct tar *tar, 2384 const struct gnu_sparse *sparse, int length) 2385 { 2386 while (length > 0 && sparse->offset[0] != 0) { 2387 if (gnu_add_sparse_entry(a, tar, 2388 tar_atol(sparse->offset, sizeof(sparse->offset)), 2389 tar_atol(sparse->numbytes, sizeof(sparse->numbytes))) 2390 != ARCHIVE_OK) 2391 return (ARCHIVE_FATAL); 2392 sparse++; 2393 length--; 2394 } 2395 return (ARCHIVE_OK); 2396 } 2397 2398 /* 2399 * GNU tar sparse format 0.0 2400 * 2401 * Beginning with GNU tar 1.15, sparse files are stored using 2402 * information in the pax extended header. The GNU tar maintainers 2403 * have gone through a number of variations in the process of working 2404 * out this scheme; fortunately, they're all numbered. 2405 * 2406 * Sparse format 0.0 uses attribute GNU.sparse.numblocks to store the 2407 * number of blocks, and GNU.sparse.offset/GNU.sparse.numbytes to 2408 * store offset/size for each block. The repeated instances of these 2409 * latter fields violate the pax specification (which frowns on 2410 * duplicate keys), so this format was quickly replaced. 2411 */ 2412 2413 /* 2414 * GNU tar sparse format 0.1 2415 * 2416 * This version replaced the offset/numbytes attributes with 2417 * a single "map" attribute that stored a list of integers. This 2418 * format had two problems: First, the "map" attribute could be very 2419 * long, which caused problems for some implementations. More 2420 * importantly, the sparse data was lost when extracted by archivers 2421 * that didn't recognize this extension. 2422 */ 2423 2424 static int 2425 gnu_sparse_01_parse(struct archive_read *a, struct tar *tar, const char *p) 2426 { 2427 const char *e; 2428 int64_t offset = -1, size = -1; 2429 2430 for (;;) { 2431 e = p; 2432 while (*e != '\0' && *e != ',') { 2433 if (*e < '0' || *e > '9') 2434 return (ARCHIVE_WARN); 2435 e++; 2436 } 2437 if (offset < 0) { 2438 offset = tar_atol10(p, e - p); 2439 if (offset < 0) 2440 return (ARCHIVE_WARN); 2441 } else { 2442 size = tar_atol10(p, e - p); 2443 if (size < 0) 2444 return (ARCHIVE_WARN); 2445 if (gnu_add_sparse_entry(a, tar, offset, size) 2446 != ARCHIVE_OK) 2447 return (ARCHIVE_FATAL); 2448 offset = -1; 2449 } 2450 if (*e == '\0') 2451 return (ARCHIVE_OK); 2452 p = e + 1; 2453 } 2454 } 2455 2456 /* 2457 * GNU tar sparse format 1.0 2458 * 2459 * The idea: The offset/size data is stored as a series of base-10 2460 * ASCII numbers prepended to the file data, so that dearchivers that 2461 * don't support this format will extract the block map along with the 2462 * data and a separate post-process can restore the sparseness. 2463 * 2464 * Unfortunately, GNU tar 1.16 had a bug that added unnecessary 2465 * padding to the body of the file when using this format. GNU tar 2466 * 1.17 corrected this bug without bumping the version number, so 2467 * it's not possible to support both variants. This code supports 2468 * the later variant at the expense of not supporting the former. 2469 * 2470 * This variant also replaced GNU.sparse.size with GNU.sparse.realsize 2471 * and introduced the GNU.sparse.major/GNU.sparse.minor attributes. 2472 */ 2473 2474 /* 2475 * Read the next line from the input, and parse it as a decimal 2476 * integer followed by '\n'. Returns positive integer value or 2477 * negative on error. 2478 */ 2479 static int64_t 2480 gnu_sparse_10_atol(struct archive_read *a, struct tar *tar, 2481 int64_t *remaining, size_t *unconsumed) 2482 { 2483 int64_t l, limit, last_digit_limit; 2484 const char *p; 2485 ssize_t bytes_read; 2486 int base, digit; 2487 2488 base = 10; 2489 limit = INT64_MAX / base; 2490 last_digit_limit = INT64_MAX % base; 2491 2492 /* 2493 * Skip any lines starting with '#'; GNU tar specs 2494 * don't require this, but they should. 2495 */ 2496 do { 2497 bytes_read = readline(a, tar, &p, 2498 (ssize_t)tar_min(*remaining, 100), unconsumed); 2499 if (bytes_read <= 0) 2500 return (ARCHIVE_FATAL); 2501 *remaining -= bytes_read; 2502 } while (p[0] == '#'); 2503 2504 l = 0; 2505 while (bytes_read > 0) { 2506 if (*p == '\n') 2507 return (l); 2508 if (*p < '0' || *p >= '0' + base) 2509 return (ARCHIVE_WARN); 2510 digit = *p - '0'; 2511 if (l > limit || (l == limit && digit > last_digit_limit)) 2512 l = INT64_MAX; /* Truncate on overflow. */ 2513 else 2514 l = (l * base) + digit; 2515 p++; 2516 bytes_read--; 2517 } 2518 /* TODO: Error message. */ 2519 return (ARCHIVE_WARN); 2520 } 2521 2522 /* 2523 * Returns length (in bytes) of the sparse data description 2524 * that was read. 2525 */ 2526 static ssize_t 2527 gnu_sparse_10_read(struct archive_read *a, struct tar *tar, size_t *unconsumed) 2528 { 2529 ssize_t bytes_read; 2530 int entries; 2531 int64_t offset, size, to_skip, remaining; 2532 2533 /* Clear out the existing sparse list. */ 2534 gnu_clear_sparse_list(tar); 2535 2536 remaining = tar->entry_bytes_remaining; 2537 2538 /* Parse entries. */ 2539 entries = (int)gnu_sparse_10_atol(a, tar, &remaining, unconsumed); 2540 if (entries < 0) 2541 return (ARCHIVE_FATAL); 2542 /* Parse the individual entries. */ 2543 while (entries-- > 0) { 2544 /* Parse offset/size */ 2545 offset = gnu_sparse_10_atol(a, tar, &remaining, unconsumed); 2546 if (offset < 0) 2547 return (ARCHIVE_FATAL); 2548 size = gnu_sparse_10_atol(a, tar, &remaining, unconsumed); 2549 if (size < 0) 2550 return (ARCHIVE_FATAL); 2551 /* Add a new sparse entry. */ 2552 if (gnu_add_sparse_entry(a, tar, offset, size) != ARCHIVE_OK) 2553 return (ARCHIVE_FATAL); 2554 } 2555 /* Skip rest of block... */ 2556 tar_flush_unconsumed(a, unconsumed); 2557 bytes_read = (ssize_t)(tar->entry_bytes_remaining - remaining); 2558 to_skip = 0x1ff & -bytes_read; 2559 /* Fail if tar->entry_bytes_remaing would get negative */ 2560 if (to_skip > remaining) 2561 return (ARCHIVE_FATAL); 2562 if (to_skip != __archive_read_consume(a, to_skip)) 2563 return (ARCHIVE_FATAL); 2564 return ((ssize_t)(bytes_read + to_skip)); 2565 } 2566 2567 /* 2568 * Solaris pax extension for a sparse file. This is recorded with the 2569 * data and hole pairs. The way recording sparse information by Solaris' 2570 * pax simply indicates where data and sparse are, so the stored contents 2571 * consist of both data and hole. 2572 */ 2573 static int 2574 solaris_sparse_parse(struct archive_read *a, struct tar *tar, 2575 struct archive_entry *entry, const char *p) 2576 { 2577 const char *e; 2578 int64_t start, end; 2579 int hole = 1; 2580 2581 (void)entry; /* UNUSED */ 2582 2583 end = 0; 2584 if (*p == ' ') 2585 p++; 2586 else 2587 return (ARCHIVE_WARN); 2588 for (;;) { 2589 e = p; 2590 while (*e != '\0' && *e != ' ') { 2591 if (*e < '0' || *e > '9') 2592 return (ARCHIVE_WARN); 2593 e++; 2594 } 2595 start = end; 2596 end = tar_atol10(p, e - p); 2597 if (end < 0) 2598 return (ARCHIVE_WARN); 2599 if (start < end) { 2600 if (gnu_add_sparse_entry(a, tar, start, 2601 end - start) != ARCHIVE_OK) 2602 return (ARCHIVE_FATAL); 2603 tar->sparse_last->hole = hole; 2604 } 2605 if (*e == '\0') 2606 return (ARCHIVE_OK); 2607 p = e + 1; 2608 hole = hole == 0; 2609 } 2610 } 2611 2612 /*- 2613 * Convert text->integer. 2614 * 2615 * Traditional tar formats (including POSIX) specify base-8 for 2616 * all of the standard numeric fields. This is a significant limitation 2617 * in practice: 2618 * = file size is limited to 8GB 2619 * = rdevmajor and rdevminor are limited to 21 bits 2620 * = uid/gid are limited to 21 bits 2621 * 2622 * There are two workarounds for this: 2623 * = pax extended headers, which use variable-length string fields 2624 * = GNU tar and STAR both allow either base-8 or base-256 in 2625 * most fields. The high bit is set to indicate base-256. 2626 * 2627 * On read, this implementation supports both extensions. 2628 */ 2629 static int64_t 2630 tar_atol(const char *p, size_t char_cnt) 2631 { 2632 /* 2633 * Technically, GNU tar considers a field to be in base-256 2634 * only if the first byte is 0xff or 0x80. 2635 */ 2636 if (*p & 0x80) 2637 return (tar_atol256(p, char_cnt)); 2638 return (tar_atol8(p, char_cnt)); 2639 } 2640 2641 /* 2642 * Note that this implementation does not (and should not!) obey 2643 * locale settings; you cannot simply substitute strtol here, since 2644 * it does obey locale. 2645 */ 2646 static int64_t 2647 tar_atol_base_n(const char *p, size_t char_cnt, int base) 2648 { 2649 int64_t l, maxval, limit, last_digit_limit; 2650 int digit, sign; 2651 2652 maxval = INT64_MAX; 2653 limit = INT64_MAX / base; 2654 last_digit_limit = INT64_MAX % base; 2655 2656 /* the pointer will not be dereferenced if char_cnt is zero 2657 * due to the way the && operator is evaluated. 2658 */ 2659 while (char_cnt != 0 && (*p == ' ' || *p == '\t')) { 2660 p++; 2661 char_cnt--; 2662 } 2663 2664 sign = 1; 2665 if (char_cnt != 0 && *p == '-') { 2666 sign = -1; 2667 p++; 2668 char_cnt--; 2669 2670 maxval = INT64_MIN; 2671 limit = -(INT64_MIN / base); 2672 last_digit_limit = -(INT64_MIN % base); 2673 } 2674 2675 l = 0; 2676 if (char_cnt != 0) { 2677 digit = *p - '0'; 2678 while (digit >= 0 && digit < base && char_cnt != 0) { 2679 if (l>limit || (l == limit && digit >= last_digit_limit)) { 2680 return maxval; /* Truncate on overflow. */ 2681 } 2682 l = (l * base) + digit; 2683 digit = *++p - '0'; 2684 char_cnt--; 2685 } 2686 } 2687 return (sign < 0) ? -l : l; 2688 } 2689 2690 static int64_t 2691 tar_atol8(const char *p, size_t char_cnt) 2692 { 2693 return tar_atol_base_n(p, char_cnt, 8); 2694 } 2695 2696 static int64_t 2697 tar_atol10(const char *p, size_t char_cnt) 2698 { 2699 return tar_atol_base_n(p, char_cnt, 10); 2700 } 2701 2702 /* 2703 * Parse a base-256 integer. This is just a variable-length 2704 * twos-complement signed binary value in big-endian order, except 2705 * that the high-order bit is ignored. The values here can be up to 2706 * 12 bytes, so we need to be careful about overflowing 64-bit 2707 * (8-byte) integers. 2708 * 2709 * This code unashamedly assumes that the local machine uses 8-bit 2710 * bytes and twos-complement arithmetic. 2711 */ 2712 static int64_t 2713 tar_atol256(const char *_p, size_t char_cnt) 2714 { 2715 uint64_t l; 2716 const unsigned char *p = (const unsigned char *)_p; 2717 unsigned char c, neg; 2718 2719 /* Extend 7-bit 2s-comp to 8-bit 2s-comp, decide sign. */ 2720 c = *p; 2721 if (c & 0x40) { 2722 neg = 0xff; 2723 c |= 0x80; 2724 l = ~ARCHIVE_LITERAL_ULL(0); 2725 } else { 2726 neg = 0; 2727 c &= 0x7f; 2728 l = 0; 2729 } 2730 2731 /* If more than 8 bytes, check that we can ignore 2732 * high-order bits without overflow. */ 2733 while (char_cnt > sizeof(int64_t)) { 2734 --char_cnt; 2735 if (c != neg) 2736 return neg ? INT64_MIN : INT64_MAX; 2737 c = *++p; 2738 } 2739 2740 /* c is first byte that fits; if sign mismatch, return overflow */ 2741 if ((c ^ neg) & 0x80) { 2742 return neg ? INT64_MIN : INT64_MAX; 2743 } 2744 2745 /* Accumulate remaining bytes. */ 2746 while (--char_cnt > 0) { 2747 l = (l << 8) | c; 2748 c = *++p; 2749 } 2750 l = (l << 8) | c; 2751 /* Return signed twos-complement value. */ 2752 return (int64_t)(l); 2753 } 2754 2755 /* 2756 * Returns length of line (including trailing newline) 2757 * or negative on error. 'start' argument is updated to 2758 * point to first character of line. This avoids copying 2759 * when possible. 2760 */ 2761 static ssize_t 2762 readline(struct archive_read *a, struct tar *tar, const char **start, 2763 ssize_t limit, size_t *unconsumed) 2764 { 2765 ssize_t bytes_read; 2766 ssize_t total_size = 0; 2767 const void *t; 2768 const char *s; 2769 void *p; 2770 2771 tar_flush_unconsumed(a, unconsumed); 2772 2773 t = __archive_read_ahead(a, 1, &bytes_read); 2774 if (bytes_read <= 0) 2775 return (ARCHIVE_FATAL); 2776 s = t; /* Start of line? */ 2777 p = memchr(t, '\n', bytes_read); 2778 /* If we found '\n' in the read buffer, return pointer to that. */ 2779 if (p != NULL) { 2780 bytes_read = 1 + ((const char *)p) - s; 2781 if (bytes_read > limit) { 2782 archive_set_error(&a->archive, 2783 ARCHIVE_ERRNO_FILE_FORMAT, 2784 "Line too long"); 2785 return (ARCHIVE_FATAL); 2786 } 2787 *unconsumed = bytes_read; 2788 *start = s; 2789 return (bytes_read); 2790 } 2791 *unconsumed = bytes_read; 2792 /* Otherwise, we need to accumulate in a line buffer. */ 2793 for (;;) { 2794 if (total_size + bytes_read > limit) { 2795 archive_set_error(&a->archive, 2796 ARCHIVE_ERRNO_FILE_FORMAT, 2797 "Line too long"); 2798 return (ARCHIVE_FATAL); 2799 } 2800 if (archive_string_ensure(&tar->line, total_size + bytes_read) == NULL) { 2801 archive_set_error(&a->archive, ENOMEM, 2802 "Can't allocate working buffer"); 2803 return (ARCHIVE_FATAL); 2804 } 2805 memcpy(tar->line.s + total_size, t, bytes_read); 2806 tar_flush_unconsumed(a, unconsumed); 2807 total_size += bytes_read; 2808 /* If we found '\n', clean up and return. */ 2809 if (p != NULL) { 2810 *start = tar->line.s; 2811 return (total_size); 2812 } 2813 /* Read some more. */ 2814 t = __archive_read_ahead(a, 1, &bytes_read); 2815 if (bytes_read <= 0) 2816 return (ARCHIVE_FATAL); 2817 s = t; /* Start of line? */ 2818 p = memchr(t, '\n', bytes_read); 2819 /* If we found '\n', trim the read. */ 2820 if (p != NULL) { 2821 bytes_read = 1 + ((const char *)p) - s; 2822 } 2823 *unconsumed = bytes_read; 2824 } 2825 } 2826 2827 /* 2828 * base64_decode - Base64 decode 2829 * 2830 * This accepts most variations of base-64 encoding, including: 2831 * * with or without line breaks 2832 * * with or without the final group padded with '=' or '_' characters 2833 * (The most economical Base-64 variant does not pad the last group and 2834 * omits line breaks; RFC1341 used for MIME requires both.) 2835 */ 2836 static char * 2837 base64_decode(const char *s, size_t len, size_t *out_len) 2838 { 2839 static const unsigned char digits[64] = { 2840 'A','B','C','D','E','F','G','H','I','J','K','L','M','N', 2841 'O','P','Q','R','S','T','U','V','W','X','Y','Z','a','b', 2842 'c','d','e','f','g','h','i','j','k','l','m','n','o','p', 2843 'q','r','s','t','u','v','w','x','y','z','0','1','2','3', 2844 '4','5','6','7','8','9','+','/' }; 2845 static unsigned char decode_table[128]; 2846 char *out, *d; 2847 const unsigned char *src = (const unsigned char *)s; 2848 2849 /* If the decode table is not yet initialized, prepare it. */ 2850 if (decode_table[digits[1]] != 1) { 2851 unsigned i; 2852 memset(decode_table, 0xff, sizeof(decode_table)); 2853 for (i = 0; i < sizeof(digits); i++) 2854 decode_table[digits[i]] = i; 2855 } 2856 2857 /* Allocate enough space to hold the entire output. */ 2858 /* Note that we may not use all of this... */ 2859 out = (char *)malloc(len - len / 4 + 1); 2860 if (out == NULL) { 2861 *out_len = 0; 2862 return (NULL); 2863 } 2864 d = out; 2865 2866 while (len > 0) { 2867 /* Collect the next group of (up to) four characters. */ 2868 int v = 0; 2869 int group_size = 0; 2870 while (group_size < 4 && len > 0) { 2871 /* '=' or '_' padding indicates final group. */ 2872 if (*src == '=' || *src == '_') { 2873 len = 0; 2874 break; 2875 } 2876 /* Skip illegal characters (including line breaks) */ 2877 if (*src > 127 || *src < 32 2878 || decode_table[*src] == 0xff) { 2879 len--; 2880 src++; 2881 continue; 2882 } 2883 v <<= 6; 2884 v |= decode_table[*src++]; 2885 len --; 2886 group_size++; 2887 } 2888 /* Align a short group properly. */ 2889 v <<= 6 * (4 - group_size); 2890 /* Unpack the group we just collected. */ 2891 switch (group_size) { 2892 case 4: d[2] = v & 0xff; 2893 /* FALLTHROUGH */ 2894 case 3: d[1] = (v >> 8) & 0xff; 2895 /* FALLTHROUGH */ 2896 case 2: d[0] = (v >> 16) & 0xff; 2897 break; 2898 case 1: /* this is invalid! */ 2899 break; 2900 } 2901 d += group_size * 3 / 4; 2902 } 2903 2904 *out_len = d - out; 2905 return (out); 2906 } 2907 2908 static char * 2909 url_decode(const char *in) 2910 { 2911 char *out, *d; 2912 const char *s; 2913 2914 out = (char *)malloc(strlen(in) + 1); 2915 if (out == NULL) 2916 return (NULL); 2917 for (s = in, d = out; *s != '\0'; ) { 2918 if (s[0] == '%' && s[1] != '\0' && s[2] != '\0') { 2919 /* Try to convert % escape */ 2920 int digit1 = tohex(s[1]); 2921 int digit2 = tohex(s[2]); 2922 if (digit1 >= 0 && digit2 >= 0) { 2923 /* Looks good, consume three chars */ 2924 s += 3; 2925 /* Convert output */ 2926 *d++ = ((digit1 << 4) | digit2); 2927 continue; 2928 } 2929 /* Else fall through and treat '%' as normal char */ 2930 } 2931 *d++ = *s++; 2932 } 2933 *d = '\0'; 2934 return (out); 2935 } 2936 2937 static int 2938 tohex(int c) 2939 { 2940 if (c >= '0' && c <= '9') 2941 return (c - '0'); 2942 else if (c >= 'A' && c <= 'F') 2943 return (c - 'A' + 10); 2944 else if (c >= 'a' && c <= 'f') 2945 return (c - 'a' + 10); 2946 else 2947 return (-1); 2948 } 2949