1 /*- 2 * Copyright (c) 2003-2007 Tim Kientzle 3 * Copyright (c) 2011-2012 Michihiro NAKAJIMA 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR 16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 18 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, 19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 25 */ 26 27 #include "archive_platform.h" 28 __FBSDID("$FreeBSD: head/lib/libarchive/archive_read_support_format_tar.c 201161 2009-12-29 05:44:39Z kientzle $"); 29 30 #ifdef HAVE_ERRNO_H 31 #include <errno.h> 32 #endif 33 #include <stddef.h> 34 #ifdef HAVE_STDLIB_H 35 #include <stdlib.h> 36 #endif 37 #ifdef HAVE_STRING_H 38 #include <string.h> 39 #endif 40 41 #include "archive.h" 42 #include "archive_acl_private.h" /* For ACL parsing routines. */ 43 #include "archive_entry.h" 44 #include "archive_entry_locale.h" 45 #include "archive_private.h" 46 #include "archive_read_private.h" 47 48 #define tar_min(a,b) ((a) < (b) ? (a) : (b)) 49 50 /* 51 * Layout of POSIX 'ustar' tar header. 52 */ 53 struct archive_entry_header_ustar { 54 char name[100]; 55 char mode[8]; 56 char uid[8]; 57 char gid[8]; 58 char size[12]; 59 char mtime[12]; 60 char checksum[8]; 61 char typeflag[1]; 62 char linkname[100]; /* "old format" header ends here */ 63 char magic[6]; /* For POSIX: "ustar\0" */ 64 char version[2]; /* For POSIX: "00" */ 65 char uname[32]; 66 char gname[32]; 67 char rdevmajor[8]; 68 char rdevminor[8]; 69 char prefix[155]; 70 }; 71 72 /* 73 * Structure of GNU tar header 74 */ 75 struct gnu_sparse { 76 char offset[12]; 77 char numbytes[12]; 78 }; 79 80 struct archive_entry_header_gnutar { 81 char name[100]; 82 char mode[8]; 83 char uid[8]; 84 char gid[8]; 85 char size[12]; 86 char mtime[12]; 87 char checksum[8]; 88 char typeflag[1]; 89 char linkname[100]; 90 char magic[8]; /* "ustar \0" (note blank/blank/null at end) */ 91 char uname[32]; 92 char gname[32]; 93 char rdevmajor[8]; 94 char rdevminor[8]; 95 char atime[12]; 96 char ctime[12]; 97 char offset[12]; 98 char longnames[4]; 99 char unused[1]; 100 struct gnu_sparse sparse[4]; 101 char isextended[1]; 102 char realsize[12]; 103 /* 104 * Old GNU format doesn't use POSIX 'prefix' field; they use 105 * the 'L' (longname) entry instead. 106 */ 107 }; 108 109 /* 110 * Data specific to this format. 111 */ 112 struct sparse_block { 113 struct sparse_block *next; 114 int64_t offset; 115 int64_t remaining; 116 int hole; 117 }; 118 119 struct tar { 120 struct archive_string acl_text; 121 struct archive_string entry_pathname; 122 /* For "GNU.sparse.name" and other similar path extensions. */ 123 struct archive_string entry_pathname_override; 124 struct archive_string entry_linkpath; 125 struct archive_string entry_uname; 126 struct archive_string entry_gname; 127 struct archive_string longlink; 128 struct archive_string longname; 129 struct archive_string pax_header; 130 struct archive_string pax_global; 131 struct archive_string line; 132 int pax_hdrcharset_binary; 133 int header_recursion_depth; 134 int64_t entry_bytes_remaining; 135 int64_t entry_offset; 136 int64_t entry_padding; 137 int64_t entry_bytes_unconsumed; 138 int64_t realsize; 139 struct sparse_block *sparse_list; 140 struct sparse_block *sparse_last; 141 int64_t sparse_offset; 142 int64_t sparse_numbytes; 143 int sparse_gnu_major; 144 int sparse_gnu_minor; 145 char sparse_gnu_pending; 146 147 struct archive_string localname; 148 struct archive_string_conv *opt_sconv; 149 struct archive_string_conv *sconv; 150 struct archive_string_conv *sconv_acl; 151 struct archive_string_conv *sconv_default; 152 int init_default_conversion; 153 int compat_2x; 154 int process_mac_extensions; 155 int read_concatenated_archives; 156 }; 157 158 static int archive_block_is_null(const char *p); 159 static char *base64_decode(const char *, size_t, size_t *); 160 static int gnu_add_sparse_entry(struct archive_read *, struct tar *, 161 int64_t offset, int64_t remaining); 162 163 static void gnu_clear_sparse_list(struct tar *); 164 static int gnu_sparse_old_read(struct archive_read *, struct tar *, 165 const struct archive_entry_header_gnutar *header, size_t *); 166 static int gnu_sparse_old_parse(struct archive_read *, struct tar *, 167 const struct gnu_sparse *sparse, int length); 168 static int gnu_sparse_01_parse(struct archive_read *, struct tar *, 169 const char *); 170 static ssize_t gnu_sparse_10_read(struct archive_read *, struct tar *, 171 size_t *); 172 static int header_Solaris_ACL(struct archive_read *, struct tar *, 173 struct archive_entry *, const void *, size_t *); 174 static int header_common(struct archive_read *, struct tar *, 175 struct archive_entry *, const void *); 176 static int header_old_tar(struct archive_read *, struct tar *, 177 struct archive_entry *, const void *); 178 static int header_pax_extensions(struct archive_read *, struct tar *, 179 struct archive_entry *, const void *, size_t *); 180 static int header_pax_global(struct archive_read *, struct tar *, 181 struct archive_entry *, const void *h, size_t *); 182 static int header_longlink(struct archive_read *, struct tar *, 183 struct archive_entry *, const void *h, size_t *); 184 static int header_longname(struct archive_read *, struct tar *, 185 struct archive_entry *, const void *h, size_t *); 186 static int read_mac_metadata_blob(struct archive_read *, struct tar *, 187 struct archive_entry *, const void *h, size_t *); 188 static int header_volume(struct archive_read *, struct tar *, 189 struct archive_entry *, const void *h, size_t *); 190 static int header_ustar(struct archive_read *, struct tar *, 191 struct archive_entry *, const void *h); 192 static int header_gnutar(struct archive_read *, struct tar *, 193 struct archive_entry *, const void *h, size_t *); 194 static int archive_read_format_tar_bid(struct archive_read *, int); 195 static int archive_read_format_tar_options(struct archive_read *, 196 const char *, const char *); 197 static int archive_read_format_tar_cleanup(struct archive_read *); 198 static int archive_read_format_tar_read_data(struct archive_read *a, 199 const void **buff, size_t *size, int64_t *offset); 200 static int archive_read_format_tar_skip(struct archive_read *a); 201 static int archive_read_format_tar_read_header(struct archive_read *, 202 struct archive_entry *); 203 static int checksum(struct archive_read *, const void *); 204 static int pax_attribute(struct archive_read *, struct tar *, 205 struct archive_entry *, char *key, char *value); 206 static int pax_header(struct archive_read *, struct tar *, 207 struct archive_entry *, char *attr); 208 static void pax_time(const char *, int64_t *sec, long *nanos); 209 static ssize_t readline(struct archive_read *, struct tar *, const char **, 210 ssize_t limit, size_t *); 211 static int read_body_to_string(struct archive_read *, struct tar *, 212 struct archive_string *, const void *h, size_t *); 213 static int solaris_sparse_parse(struct archive_read *, struct tar *, 214 struct archive_entry *, const char *); 215 static int64_t tar_atol(const char *, size_t); 216 static int64_t tar_atol10(const char *, size_t); 217 static int64_t tar_atol256(const char *, size_t); 218 static int64_t tar_atol8(const char *, size_t); 219 static int tar_read_header(struct archive_read *, struct tar *, 220 struct archive_entry *, size_t *); 221 static int tohex(int c); 222 static char *url_decode(const char *); 223 static void tar_flush_unconsumed(struct archive_read *, size_t *); 224 225 226 int 227 archive_read_support_format_gnutar(struct archive *a) 228 { 229 archive_check_magic(a, ARCHIVE_READ_MAGIC, 230 ARCHIVE_STATE_NEW, "archive_read_support_format_gnutar"); 231 return (archive_read_support_format_tar(a)); 232 } 233 234 235 int 236 archive_read_support_format_tar(struct archive *_a) 237 { 238 struct archive_read *a = (struct archive_read *)_a; 239 struct tar *tar; 240 int r; 241 242 archive_check_magic(_a, ARCHIVE_READ_MAGIC, 243 ARCHIVE_STATE_NEW, "archive_read_support_format_tar"); 244 245 tar = (struct tar *)calloc(1, sizeof(*tar)); 246 #ifdef HAVE_COPYFILE_H 247 /* Set this by default on Mac OS. */ 248 tar->process_mac_extensions = 1; 249 #endif 250 if (tar == NULL) { 251 archive_set_error(&a->archive, ENOMEM, 252 "Can't allocate tar data"); 253 return (ARCHIVE_FATAL); 254 } 255 256 r = __archive_read_register_format(a, tar, "tar", 257 archive_read_format_tar_bid, 258 archive_read_format_tar_options, 259 archive_read_format_tar_read_header, 260 archive_read_format_tar_read_data, 261 archive_read_format_tar_skip, 262 NULL, 263 archive_read_format_tar_cleanup, 264 NULL, 265 NULL); 266 267 if (r != ARCHIVE_OK) 268 free(tar); 269 return (ARCHIVE_OK); 270 } 271 272 static int 273 archive_read_format_tar_cleanup(struct archive_read *a) 274 { 275 struct tar *tar; 276 277 tar = (struct tar *)(a->format->data); 278 gnu_clear_sparse_list(tar); 279 archive_string_free(&tar->acl_text); 280 archive_string_free(&tar->entry_pathname); 281 archive_string_free(&tar->entry_pathname_override); 282 archive_string_free(&tar->entry_linkpath); 283 archive_string_free(&tar->entry_uname); 284 archive_string_free(&tar->entry_gname); 285 archive_string_free(&tar->line); 286 archive_string_free(&tar->pax_global); 287 archive_string_free(&tar->pax_header); 288 archive_string_free(&tar->longname); 289 archive_string_free(&tar->longlink); 290 archive_string_free(&tar->localname); 291 free(tar); 292 (a->format->data) = NULL; 293 return (ARCHIVE_OK); 294 } 295 296 297 static int 298 archive_read_format_tar_bid(struct archive_read *a, int best_bid) 299 { 300 int bid; 301 const char *h; 302 const struct archive_entry_header_ustar *header; 303 304 (void)best_bid; /* UNUSED */ 305 306 bid = 0; 307 308 /* Now let's look at the actual header and see if it matches. */ 309 h = __archive_read_ahead(a, 512, NULL); 310 if (h == NULL) 311 return (-1); 312 313 /* If it's an end-of-archive mark, we can handle it. */ 314 if (h[0] == 0 && archive_block_is_null(h)) { 315 /* 316 * Usually, I bid the number of bits verified, but 317 * in this case, 4096 seems excessive so I picked 10 as 318 * an arbitrary but reasonable-seeming value. 319 */ 320 return (10); 321 } 322 323 /* If it's not an end-of-archive mark, it must have a valid checksum.*/ 324 if (!checksum(a, h)) 325 return (0); 326 bid += 48; /* Checksum is usually 6 octal digits. */ 327 328 header = (const struct archive_entry_header_ustar *)h; 329 330 /* Recognize POSIX formats. */ 331 if ((memcmp(header->magic, "ustar\0", 6) == 0) 332 && (memcmp(header->version, "00", 2) == 0)) 333 bid += 56; 334 335 /* Recognize GNU tar format. */ 336 if ((memcmp(header->magic, "ustar ", 6) == 0) 337 && (memcmp(header->version, " \0", 2) == 0)) 338 bid += 56; 339 340 /* Type flag must be null, digit or A-Z, a-z. */ 341 if (header->typeflag[0] != 0 && 342 !( header->typeflag[0] >= '0' && header->typeflag[0] <= '9') && 343 !( header->typeflag[0] >= 'A' && header->typeflag[0] <= 'Z') && 344 !( header->typeflag[0] >= 'a' && header->typeflag[0] <= 'z') ) 345 return (0); 346 bid += 2; /* 6 bits of variation in an 8-bit field leaves 2 bits. */ 347 348 /* Sanity check: Look at first byte of mode field. */ 349 switch (255 & (unsigned)header->mode[0]) { 350 case 0: case 255: 351 /* Base-256 value: No further verification possible! */ 352 break; 353 case ' ': /* Not recommended, but not illegal, either. */ 354 break; 355 case '0': case '1': case '2': case '3': 356 case '4': case '5': case '6': case '7': 357 /* Octal Value. */ 358 /* TODO: Check format of remainder of this field. */ 359 break; 360 default: 361 /* Not a valid mode; bail out here. */ 362 return (0); 363 } 364 /* TODO: Sanity test uid/gid/size/mtime/rdevmajor/rdevminor fields. */ 365 366 return (bid); 367 } 368 369 static int 370 archive_read_format_tar_options(struct archive_read *a, 371 const char *key, const char *val) 372 { 373 struct tar *tar; 374 int ret = ARCHIVE_FAILED; 375 376 tar = (struct tar *)(a->format->data); 377 if (strcmp(key, "compat-2x") == 0) { 378 /* Handle UTF-8 filnames as libarchive 2.x */ 379 tar->compat_2x = (val != NULL && val[0] != 0); 380 tar->init_default_conversion = tar->compat_2x; 381 return (ARCHIVE_OK); 382 } else if (strcmp(key, "hdrcharset") == 0) { 383 if (val == NULL || val[0] == 0) 384 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 385 "tar: hdrcharset option needs a character-set name"); 386 else { 387 tar->opt_sconv = 388 archive_string_conversion_from_charset( 389 &a->archive, val, 0); 390 if (tar->opt_sconv != NULL) 391 ret = ARCHIVE_OK; 392 else 393 ret = ARCHIVE_FATAL; 394 } 395 return (ret); 396 } else if (strcmp(key, "mac-ext") == 0) { 397 tar->process_mac_extensions = (val != NULL && val[0] != 0); 398 return (ARCHIVE_OK); 399 } else if (strcmp(key, "read_concatenated_archives") == 0) { 400 tar->read_concatenated_archives = (val != NULL && val[0] != 0); 401 return (ARCHIVE_OK); 402 } 403 404 /* Note: The "warn" return is just to inform the options 405 * supervisor that we didn't handle it. It will generate 406 * a suitable error if no one used this option. */ 407 return (ARCHIVE_WARN); 408 } 409 410 /* utility function- this exists to centralize the logic of tracking 411 * how much unconsumed data we have floating around, and to consume 412 * anything outstanding since we're going to do read_aheads 413 */ 414 static void 415 tar_flush_unconsumed(struct archive_read *a, size_t *unconsumed) 416 { 417 if (*unconsumed) { 418 /* 419 void *data = (void *)__archive_read_ahead(a, *unconsumed, NULL); 420 * this block of code is to poison claimed unconsumed space, ensuring 421 * things break if it is in use still. 422 * currently it WILL break things, so enable it only for debugging this issue 423 if (data) { 424 memset(data, 0xff, *unconsumed); 425 } 426 */ 427 __archive_read_consume(a, *unconsumed); 428 *unconsumed = 0; 429 } 430 } 431 432 /* 433 * The function invoked by archive_read_next_header(). This 434 * just sets up a few things and then calls the internal 435 * tar_read_header() function below. 436 */ 437 static int 438 archive_read_format_tar_read_header(struct archive_read *a, 439 struct archive_entry *entry) 440 { 441 /* 442 * When converting tar archives to cpio archives, it is 443 * essential that each distinct file have a distinct inode 444 * number. To simplify this, we keep a static count here to 445 * assign fake dev/inode numbers to each tar entry. Note that 446 * pax format archives may overwrite this with something more 447 * useful. 448 * 449 * Ideally, we would track every file read from the archive so 450 * that we could assign the same dev/ino pair to hardlinks, 451 * but the memory required to store a complete lookup table is 452 * probably not worthwhile just to support the relatively 453 * obscure tar->cpio conversion case. 454 */ 455 static int default_inode; 456 static int default_dev; 457 struct tar *tar; 458 const char *p; 459 const wchar_t *wp; 460 int r; 461 size_t l, unconsumed = 0; 462 463 /* Assign default device/inode values. */ 464 archive_entry_set_dev(entry, 1 + default_dev); /* Don't use zero. */ 465 archive_entry_set_ino(entry, ++default_inode); /* Don't use zero. */ 466 /* Limit generated st_ino number to 16 bits. */ 467 if (default_inode >= 0xffff) { 468 ++default_dev; 469 default_inode = 0; 470 } 471 472 tar = (struct tar *)(a->format->data); 473 tar->entry_offset = 0; 474 gnu_clear_sparse_list(tar); 475 tar->realsize = -1; /* Mark this as "unset" */ 476 477 /* Setup default string conversion. */ 478 tar->sconv = tar->opt_sconv; 479 if (tar->sconv == NULL) { 480 if (!tar->init_default_conversion) { 481 tar->sconv_default = 482 archive_string_default_conversion_for_read(&(a->archive)); 483 tar->init_default_conversion = 1; 484 } 485 tar->sconv = tar->sconv_default; 486 } 487 488 r = tar_read_header(a, tar, entry, &unconsumed); 489 490 tar_flush_unconsumed(a, &unconsumed); 491 492 /* 493 * "non-sparse" files are really just sparse files with 494 * a single block. 495 */ 496 if (tar->sparse_list == NULL) { 497 if (gnu_add_sparse_entry(a, tar, 0, tar->entry_bytes_remaining) 498 != ARCHIVE_OK) 499 return (ARCHIVE_FATAL); 500 } else { 501 struct sparse_block *sb; 502 503 for (sb = tar->sparse_list; sb != NULL; sb = sb->next) { 504 if (!sb->hole) 505 archive_entry_sparse_add_entry(entry, 506 sb->offset, sb->remaining); 507 } 508 } 509 510 if (r == ARCHIVE_OK && archive_entry_filetype(entry) == AE_IFREG) { 511 /* 512 * "Regular" entry with trailing '/' is really 513 * directory: This is needed for certain old tar 514 * variants and even for some broken newer ones. 515 */ 516 if ((wp = archive_entry_pathname_w(entry)) != NULL) { 517 l = wcslen(wp); 518 if (l > 0 && wp[l - 1] == L'/') { 519 archive_entry_set_filetype(entry, AE_IFDIR); 520 } 521 } else if ((p = archive_entry_pathname(entry)) != NULL) { 522 l = strlen(p); 523 if (l > 0 && p[l - 1] == '/') { 524 archive_entry_set_filetype(entry, AE_IFDIR); 525 } 526 } 527 } 528 return (r); 529 } 530 531 static int 532 archive_read_format_tar_read_data(struct archive_read *a, 533 const void **buff, size_t *size, int64_t *offset) 534 { 535 ssize_t bytes_read; 536 struct tar *tar; 537 struct sparse_block *p; 538 539 tar = (struct tar *)(a->format->data); 540 541 for (;;) { 542 /* Remove exhausted entries from sparse list. */ 543 while (tar->sparse_list != NULL && 544 tar->sparse_list->remaining == 0) { 545 p = tar->sparse_list; 546 tar->sparse_list = p->next; 547 free(p); 548 } 549 550 if (tar->entry_bytes_unconsumed) { 551 __archive_read_consume(a, tar->entry_bytes_unconsumed); 552 tar->entry_bytes_unconsumed = 0; 553 } 554 555 /* If we're at end of file, return EOF. */ 556 if (tar->sparse_list == NULL || 557 tar->entry_bytes_remaining == 0) { 558 if (__archive_read_consume(a, tar->entry_padding) < 0) 559 return (ARCHIVE_FATAL); 560 tar->entry_padding = 0; 561 *buff = NULL; 562 *size = 0; 563 *offset = tar->realsize; 564 return (ARCHIVE_EOF); 565 } 566 567 *buff = __archive_read_ahead(a, 1, &bytes_read); 568 if (bytes_read < 0) 569 return (ARCHIVE_FATAL); 570 if (*buff == NULL) { 571 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 572 "Truncated tar archive"); 573 return (ARCHIVE_FATAL); 574 } 575 if (bytes_read > tar->entry_bytes_remaining) 576 bytes_read = (ssize_t)tar->entry_bytes_remaining; 577 /* Don't read more than is available in the 578 * current sparse block. */ 579 if (tar->sparse_list->remaining < bytes_read) 580 bytes_read = (ssize_t)tar->sparse_list->remaining; 581 *size = bytes_read; 582 *offset = tar->sparse_list->offset; 583 tar->sparse_list->remaining -= bytes_read; 584 tar->sparse_list->offset += bytes_read; 585 tar->entry_bytes_remaining -= bytes_read; 586 tar->entry_bytes_unconsumed = bytes_read; 587 588 if (!tar->sparse_list->hole) 589 return (ARCHIVE_OK); 590 /* Current is hole data and skip this. */ 591 } 592 } 593 594 static int 595 archive_read_format_tar_skip(struct archive_read *a) 596 { 597 int64_t bytes_skipped; 598 int64_t request; 599 struct sparse_block *p; 600 struct tar* tar; 601 602 tar = (struct tar *)(a->format->data); 603 604 /* Do not consume the hole of a sparse file. */ 605 request = 0; 606 for (p = tar->sparse_list; p != NULL; p = p->next) { 607 if (!p->hole) { 608 if (p->remaining >= INT64_MAX - request) { 609 return ARCHIVE_FATAL; 610 } 611 request += p->remaining; 612 } 613 } 614 if (request > tar->entry_bytes_remaining) 615 request = tar->entry_bytes_remaining; 616 request += tar->entry_padding + tar->entry_bytes_unconsumed; 617 618 bytes_skipped = __archive_read_consume(a, request); 619 if (bytes_skipped < 0) 620 return (ARCHIVE_FATAL); 621 622 tar->entry_bytes_remaining = 0; 623 tar->entry_bytes_unconsumed = 0; 624 tar->entry_padding = 0; 625 626 /* Free the sparse list. */ 627 gnu_clear_sparse_list(tar); 628 629 return (ARCHIVE_OK); 630 } 631 632 /* 633 * This function recursively interprets all of the headers associated 634 * with a single entry. 635 */ 636 static int 637 tar_read_header(struct archive_read *a, struct tar *tar, 638 struct archive_entry *entry, size_t *unconsumed) 639 { 640 ssize_t bytes; 641 int err; 642 const char *h; 643 const struct archive_entry_header_ustar *header; 644 const struct archive_entry_header_gnutar *gnuheader; 645 646 /* Loop until we find a workable header record. */ 647 for (;;) { 648 tar_flush_unconsumed(a, unconsumed); 649 650 /* Read 512-byte header record */ 651 h = __archive_read_ahead(a, 512, &bytes); 652 if (bytes < 0) 653 return ((int)bytes); 654 if (bytes == 0) { /* EOF at a block boundary. */ 655 /* Some writers do omit the block of nulls. <sigh> */ 656 return (ARCHIVE_EOF); 657 } 658 if (bytes < 512) { /* Short block at EOF; this is bad. */ 659 archive_set_error(&a->archive, 660 ARCHIVE_ERRNO_FILE_FORMAT, 661 "Truncated tar archive"); 662 return (ARCHIVE_FATAL); 663 } 664 *unconsumed = 512; 665 666 /* Header is workable if it's not an end-of-archive mark. */ 667 if (h[0] != 0 || !archive_block_is_null(h)) 668 break; 669 670 /* Ensure format is set for archives with only null blocks. */ 671 if (a->archive.archive_format_name == NULL) { 672 a->archive.archive_format = ARCHIVE_FORMAT_TAR; 673 a->archive.archive_format_name = "tar"; 674 } 675 676 if (!tar->read_concatenated_archives) { 677 /* Try to consume a second all-null record, as well. */ 678 tar_flush_unconsumed(a, unconsumed); 679 h = __archive_read_ahead(a, 512, NULL); 680 if (h != NULL && h[0] == 0 && archive_block_is_null(h)) 681 __archive_read_consume(a, 512); 682 archive_clear_error(&a->archive); 683 return (ARCHIVE_EOF); 684 } 685 686 /* 687 * We're reading concatenated archives, ignore this block and 688 * loop to get the next. 689 */ 690 } 691 692 /* 693 * Note: If the checksum fails and we return ARCHIVE_RETRY, 694 * then the client is likely to just retry. This is a very 695 * crude way to search for the next valid header! 696 * 697 * TODO: Improve this by implementing a real header scan. 698 */ 699 if (!checksum(a, h)) { 700 tar_flush_unconsumed(a, unconsumed); 701 archive_set_error(&a->archive, EINVAL, "Damaged tar archive"); 702 return (ARCHIVE_RETRY); /* Retryable: Invalid header */ 703 } 704 705 if (++tar->header_recursion_depth > 32) { 706 tar_flush_unconsumed(a, unconsumed); 707 archive_set_error(&a->archive, EINVAL, "Too many special headers"); 708 return (ARCHIVE_WARN); 709 } 710 711 /* Determine the format variant. */ 712 header = (const struct archive_entry_header_ustar *)h; 713 714 switch(header->typeflag[0]) { 715 case 'A': /* Solaris tar ACL */ 716 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 717 a->archive.archive_format_name = "Solaris tar"; 718 err = header_Solaris_ACL(a, tar, entry, h, unconsumed); 719 break; 720 case 'g': /* POSIX-standard 'g' header. */ 721 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 722 a->archive.archive_format_name = "POSIX pax interchange format"; 723 err = header_pax_global(a, tar, entry, h, unconsumed); 724 if (err == ARCHIVE_EOF) 725 return (err); 726 break; 727 case 'K': /* Long link name (GNU tar, others) */ 728 err = header_longlink(a, tar, entry, h, unconsumed); 729 break; 730 case 'L': /* Long filename (GNU tar, others) */ 731 err = header_longname(a, tar, entry, h, unconsumed); 732 break; 733 case 'V': /* GNU volume header */ 734 err = header_volume(a, tar, entry, h, unconsumed); 735 break; 736 case 'X': /* Used by SUN tar; same as 'x'. */ 737 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 738 a->archive.archive_format_name = 739 "POSIX pax interchange format (Sun variant)"; 740 err = header_pax_extensions(a, tar, entry, h, unconsumed); 741 break; 742 case 'x': /* POSIX-standard 'x' header. */ 743 a->archive.archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE; 744 a->archive.archive_format_name = "POSIX pax interchange format"; 745 err = header_pax_extensions(a, tar, entry, h, unconsumed); 746 break; 747 default: 748 gnuheader = (const struct archive_entry_header_gnutar *)h; 749 if (memcmp(gnuheader->magic, "ustar \0", 8) == 0) { 750 a->archive.archive_format = ARCHIVE_FORMAT_TAR_GNUTAR; 751 a->archive.archive_format_name = "GNU tar format"; 752 err = header_gnutar(a, tar, entry, h, unconsumed); 753 } else if (memcmp(header->magic, "ustar", 5) == 0) { 754 if (a->archive.archive_format != ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) { 755 a->archive.archive_format = ARCHIVE_FORMAT_TAR_USTAR; 756 a->archive.archive_format_name = "POSIX ustar format"; 757 } 758 err = header_ustar(a, tar, entry, h); 759 } else { 760 a->archive.archive_format = ARCHIVE_FORMAT_TAR; 761 a->archive.archive_format_name = "tar (non-POSIX)"; 762 err = header_old_tar(a, tar, entry, h); 763 } 764 } 765 if (err == ARCHIVE_FATAL) 766 return (err); 767 768 tar_flush_unconsumed(a, unconsumed); 769 770 h = NULL; 771 header = NULL; 772 773 --tar->header_recursion_depth; 774 /* Yuck. Apple's design here ends up storing long pathname 775 * extensions for both the AppleDouble extension entry and the 776 * regular entry. 777 */ 778 if ((err == ARCHIVE_WARN || err == ARCHIVE_OK) && 779 tar->header_recursion_depth == 0 && 780 tar->process_mac_extensions) { 781 int err2 = read_mac_metadata_blob(a, tar, entry, h, unconsumed); 782 if (err2 < err) 783 err = err2; 784 } 785 786 /* We return warnings or success as-is. Anything else is fatal. */ 787 if (err == ARCHIVE_WARN || err == ARCHIVE_OK) { 788 if (tar->sparse_gnu_pending) { 789 if (tar->sparse_gnu_major == 1 && 790 tar->sparse_gnu_minor == 0) { 791 ssize_t bytes_read; 792 793 tar->sparse_gnu_pending = 0; 794 /* Read initial sparse map. */ 795 bytes_read = gnu_sparse_10_read(a, tar, unconsumed); 796 tar->entry_bytes_remaining -= bytes_read; 797 if (bytes_read < 0) 798 return ((int)bytes_read); 799 } else { 800 archive_set_error(&a->archive, 801 ARCHIVE_ERRNO_MISC, 802 "Unrecognized GNU sparse file format"); 803 return (ARCHIVE_WARN); 804 } 805 tar->sparse_gnu_pending = 0; 806 } 807 return (err); 808 } 809 if (err == ARCHIVE_EOF) 810 /* EOF when recursively reading a header is bad. */ 811 archive_set_error(&a->archive, EINVAL, "Damaged tar archive"); 812 return (ARCHIVE_FATAL); 813 } 814 815 /* 816 * Return true if block checksum is correct. 817 */ 818 static int 819 checksum(struct archive_read *a, const void *h) 820 { 821 const unsigned char *bytes; 822 const struct archive_entry_header_ustar *header; 823 int check, sum; 824 size_t i; 825 826 (void)a; /* UNUSED */ 827 bytes = (const unsigned char *)h; 828 header = (const struct archive_entry_header_ustar *)h; 829 830 /* Checksum field must hold an octal number */ 831 for (i = 0; i < sizeof(header->checksum); ++i) { 832 char c = header->checksum[i]; 833 if (c != ' ' && c != '\0' && (c < '0' || c > '7')) 834 return 0; 835 } 836 837 /* 838 * Test the checksum. Note that POSIX specifies _unsigned_ 839 * bytes for this calculation. 840 */ 841 sum = (int)tar_atol(header->checksum, sizeof(header->checksum)); 842 check = 0; 843 for (i = 0; i < 148; i++) 844 check += (unsigned char)bytes[i]; 845 for (; i < 156; i++) 846 check += 32; 847 for (; i < 512; i++) 848 check += (unsigned char)bytes[i]; 849 if (sum == check) 850 return (1); 851 852 /* 853 * Repeat test with _signed_ bytes, just in case this archive 854 * was created by an old BSD, Solaris, or HP-UX tar with a 855 * broken checksum calculation. 856 */ 857 check = 0; 858 for (i = 0; i < 148; i++) 859 check += (signed char)bytes[i]; 860 for (; i < 156; i++) 861 check += 32; 862 for (; i < 512; i++) 863 check += (signed char)bytes[i]; 864 if (sum == check) 865 return (1); 866 867 return (0); 868 } 869 870 /* 871 * Return true if this block contains only nulls. 872 */ 873 static int 874 archive_block_is_null(const char *p) 875 { 876 unsigned i; 877 878 for (i = 0; i < 512; i++) 879 if (*p++) 880 return (0); 881 return (1); 882 } 883 884 /* 885 * Interpret 'A' Solaris ACL header 886 */ 887 static int 888 header_Solaris_ACL(struct archive_read *a, struct tar *tar, 889 struct archive_entry *entry, const void *h, size_t *unconsumed) 890 { 891 const struct archive_entry_header_ustar *header; 892 size_t size; 893 int err; 894 int64_t type; 895 char *acl, *p; 896 897 /* 898 * read_body_to_string adds a NUL terminator, but we need a little 899 * more to make sure that we don't overrun acl_text later. 900 */ 901 header = (const struct archive_entry_header_ustar *)h; 902 size = (size_t)tar_atol(header->size, sizeof(header->size)); 903 err = read_body_to_string(a, tar, &(tar->acl_text), h, unconsumed); 904 if (err != ARCHIVE_OK) 905 return (err); 906 907 /* Recursively read next header */ 908 err = tar_read_header(a, tar, entry, unconsumed); 909 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 910 return (err); 911 912 /* TODO: Examine the first characters to see if this 913 * is an AIX ACL descriptor. We'll likely never support 914 * them, but it would be polite to recognize and warn when 915 * we do see them. */ 916 917 /* Leading octal number indicates ACL type and number of entries. */ 918 p = acl = tar->acl_text.s; 919 type = 0; 920 while (*p != '\0' && p < acl + size) { 921 if (*p < '0' || *p > '7') { 922 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 923 "Malformed Solaris ACL attribute (invalid digit)"); 924 return(ARCHIVE_WARN); 925 } 926 type <<= 3; 927 type += *p - '0'; 928 if (type > 077777777) { 929 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 930 "Malformed Solaris ACL attribute (count too large)"); 931 return (ARCHIVE_WARN); 932 } 933 p++; 934 } 935 switch ((int)type & ~0777777) { 936 case 01000000: 937 /* POSIX.1e ACL */ 938 break; 939 case 03000000: 940 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 941 "Solaris NFSv4 ACLs not supported"); 942 return (ARCHIVE_WARN); 943 default: 944 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 945 "Malformed Solaris ACL attribute (unsupported type %o)", 946 (int)type); 947 return (ARCHIVE_WARN); 948 } 949 p++; 950 951 if (p >= acl + size) { 952 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 953 "Malformed Solaris ACL attribute (body overflow)"); 954 return(ARCHIVE_WARN); 955 } 956 957 /* ACL text is null-terminated; find the end. */ 958 size -= (p - acl); 959 acl = p; 960 961 while (*p != '\0' && p < acl + size) 962 p++; 963 964 if (tar->sconv_acl == NULL) { 965 tar->sconv_acl = archive_string_conversion_from_charset( 966 &(a->archive), "UTF-8", 1); 967 if (tar->sconv_acl == NULL) 968 return (ARCHIVE_FATAL); 969 } 970 archive_strncpy(&(tar->localname), acl, p - acl); 971 err = archive_acl_parse_l(archive_entry_acl(entry), 972 tar->localname.s, ARCHIVE_ENTRY_ACL_TYPE_ACCESS, tar->sconv_acl); 973 if (err != ARCHIVE_OK) { 974 if (errno == ENOMEM) { 975 archive_set_error(&a->archive, ENOMEM, 976 "Can't allocate memory for ACL"); 977 } else 978 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 979 "Malformed Solaris ACL attribute (unparsable)"); 980 } 981 return (err); 982 } 983 984 /* 985 * Interpret 'K' long linkname header. 986 */ 987 static int 988 header_longlink(struct archive_read *a, struct tar *tar, 989 struct archive_entry *entry, const void *h, size_t *unconsumed) 990 { 991 int err; 992 993 err = read_body_to_string(a, tar, &(tar->longlink), h, unconsumed); 994 if (err != ARCHIVE_OK) 995 return (err); 996 err = tar_read_header(a, tar, entry, unconsumed); 997 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 998 return (err); 999 /* Set symlink if symlink already set, else hardlink. */ 1000 archive_entry_copy_link(entry, tar->longlink.s); 1001 return (ARCHIVE_OK); 1002 } 1003 1004 static int 1005 set_conversion_failed_error(struct archive_read *a, 1006 struct archive_string_conv *sconv, const char *name) 1007 { 1008 if (errno == ENOMEM) { 1009 archive_set_error(&a->archive, ENOMEM, 1010 "Can't allocate memory for %s", name); 1011 return (ARCHIVE_FATAL); 1012 } 1013 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 1014 "%s can't be converted from %s to current locale.", 1015 name, archive_string_conversion_charset_name(sconv)); 1016 return (ARCHIVE_WARN); 1017 } 1018 1019 /* 1020 * Interpret 'L' long filename header. 1021 */ 1022 static int 1023 header_longname(struct archive_read *a, struct tar *tar, 1024 struct archive_entry *entry, const void *h, size_t *unconsumed) 1025 { 1026 int err; 1027 1028 err = read_body_to_string(a, tar, &(tar->longname), h, unconsumed); 1029 if (err != ARCHIVE_OK) 1030 return (err); 1031 /* Read and parse "real" header, then override name. */ 1032 err = tar_read_header(a, tar, entry, unconsumed); 1033 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 1034 return (err); 1035 if (archive_entry_copy_pathname_l(entry, tar->longname.s, 1036 archive_strlen(&(tar->longname)), tar->sconv) != 0) 1037 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 1038 return (err); 1039 } 1040 1041 1042 /* 1043 * Interpret 'V' GNU tar volume header. 1044 */ 1045 static int 1046 header_volume(struct archive_read *a, struct tar *tar, 1047 struct archive_entry *entry, const void *h, size_t *unconsumed) 1048 { 1049 (void)h; 1050 1051 /* Just skip this and read the next header. */ 1052 return (tar_read_header(a, tar, entry, unconsumed)); 1053 } 1054 1055 /* 1056 * Read body of an archive entry into an archive_string object. 1057 */ 1058 static int 1059 read_body_to_string(struct archive_read *a, struct tar *tar, 1060 struct archive_string *as, const void *h, size_t *unconsumed) 1061 { 1062 int64_t size; 1063 const struct archive_entry_header_ustar *header; 1064 const void *src; 1065 1066 (void)tar; /* UNUSED */ 1067 header = (const struct archive_entry_header_ustar *)h; 1068 size = tar_atol(header->size, sizeof(header->size)); 1069 if ((size > 1048576) || (size < 0)) { 1070 archive_set_error(&a->archive, EINVAL, 1071 "Special header too large"); 1072 return (ARCHIVE_FATAL); 1073 } 1074 1075 /* Fail if we can't make our buffer big enough. */ 1076 if (archive_string_ensure(as, (size_t)size+1) == NULL) { 1077 archive_set_error(&a->archive, ENOMEM, 1078 "No memory"); 1079 return (ARCHIVE_FATAL); 1080 } 1081 1082 tar_flush_unconsumed(a, unconsumed); 1083 1084 /* Read the body into the string. */ 1085 *unconsumed = (size_t)((size + 511) & ~ 511); 1086 src = __archive_read_ahead(a, *unconsumed, NULL); 1087 if (src == NULL) { 1088 *unconsumed = 0; 1089 return (ARCHIVE_FATAL); 1090 } 1091 memcpy(as->s, src, (size_t)size); 1092 as->s[size] = '\0'; 1093 as->length = (size_t)size; 1094 return (ARCHIVE_OK); 1095 } 1096 1097 /* 1098 * Parse out common header elements. 1099 * 1100 * This would be the same as header_old_tar, except that the 1101 * filename is handled slightly differently for old and POSIX 1102 * entries (POSIX entries support a 'prefix'). This factoring 1103 * allows header_old_tar and header_ustar 1104 * to handle filenames differently, while still putting most of the 1105 * common parsing into one place. 1106 */ 1107 static int 1108 header_common(struct archive_read *a, struct tar *tar, 1109 struct archive_entry *entry, const void *h) 1110 { 1111 const struct archive_entry_header_ustar *header; 1112 char tartype; 1113 int err = ARCHIVE_OK; 1114 1115 header = (const struct archive_entry_header_ustar *)h; 1116 if (header->linkname[0]) 1117 archive_strncpy(&(tar->entry_linkpath), 1118 header->linkname, sizeof(header->linkname)); 1119 else 1120 archive_string_empty(&(tar->entry_linkpath)); 1121 1122 /* Parse out the numeric fields (all are octal) */ 1123 archive_entry_set_mode(entry, 1124 (mode_t)tar_atol(header->mode, sizeof(header->mode))); 1125 archive_entry_set_uid(entry, tar_atol(header->uid, sizeof(header->uid))); 1126 archive_entry_set_gid(entry, tar_atol(header->gid, sizeof(header->gid))); 1127 tar->entry_bytes_remaining = tar_atol(header->size, sizeof(header->size)); 1128 if (tar->entry_bytes_remaining < 0) { 1129 tar->entry_bytes_remaining = 0; 1130 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1131 "Tar entry has negative size?"); 1132 err = ARCHIVE_WARN; 1133 } 1134 tar->realsize = tar->entry_bytes_remaining; 1135 archive_entry_set_size(entry, tar->entry_bytes_remaining); 1136 archive_entry_set_mtime(entry, tar_atol(header->mtime, sizeof(header->mtime)), 0); 1137 1138 /* Handle the tar type flag appropriately. */ 1139 tartype = header->typeflag[0]; 1140 1141 switch (tartype) { 1142 case '1': /* Hard link */ 1143 if (archive_entry_copy_hardlink_l(entry, tar->entry_linkpath.s, 1144 archive_strlen(&(tar->entry_linkpath)), tar->sconv) != 0) { 1145 err = set_conversion_failed_error(a, tar->sconv, 1146 "Linkname"); 1147 if (err == ARCHIVE_FATAL) 1148 return (err); 1149 } 1150 /* 1151 * The following may seem odd, but: Technically, tar 1152 * does not store the file type for a "hard link" 1153 * entry, only the fact that it is a hard link. So, I 1154 * leave the type zero normally. But, pax interchange 1155 * format allows hard links to have data, which 1156 * implies that the underlying entry is a regular 1157 * file. 1158 */ 1159 if (archive_entry_size(entry) > 0) 1160 archive_entry_set_filetype(entry, AE_IFREG); 1161 1162 /* 1163 * A tricky point: Traditionally, tar readers have 1164 * ignored the size field when reading hardlink 1165 * entries, and some writers put non-zero sizes even 1166 * though the body is empty. POSIX blessed this 1167 * convention in the 1988 standard, but broke with 1168 * this tradition in 2001 by permitting hardlink 1169 * entries to store valid bodies in pax interchange 1170 * format, but not in ustar format. Since there is no 1171 * hard and fast way to distinguish pax interchange 1172 * from earlier archives (the 'x' and 'g' entries are 1173 * optional, after all), we need a heuristic. 1174 */ 1175 if (archive_entry_size(entry) == 0) { 1176 /* If the size is already zero, we're done. */ 1177 } else if (a->archive.archive_format 1178 == ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE) { 1179 /* Definitely pax extended; must obey hardlink size. */ 1180 } else if (a->archive.archive_format == ARCHIVE_FORMAT_TAR 1181 || a->archive.archive_format == ARCHIVE_FORMAT_TAR_GNUTAR) 1182 { 1183 /* Old-style or GNU tar: we must ignore the size. */ 1184 archive_entry_set_size(entry, 0); 1185 tar->entry_bytes_remaining = 0; 1186 } else if (archive_read_format_tar_bid(a, 50) > 50) { 1187 /* 1188 * We don't know if it's pax: If the bid 1189 * function sees a valid ustar header 1190 * immediately following, then let's ignore 1191 * the hardlink size. 1192 */ 1193 archive_entry_set_size(entry, 0); 1194 tar->entry_bytes_remaining = 0; 1195 } 1196 /* 1197 * TODO: There are still two cases I'd like to handle: 1198 * = a ustar non-pax archive with a hardlink entry at 1199 * end-of-archive. (Look for block of nulls following?) 1200 * = a pax archive that has not seen any pax headers 1201 * and has an entry which is a hardlink entry storing 1202 * a body containing an uncompressed tar archive. 1203 * The first is worth addressing; I don't see any reliable 1204 * way to deal with the second possibility. 1205 */ 1206 break; 1207 case '2': /* Symlink */ 1208 archive_entry_set_filetype(entry, AE_IFLNK); 1209 archive_entry_set_size(entry, 0); 1210 tar->entry_bytes_remaining = 0; 1211 if (archive_entry_copy_symlink_l(entry, tar->entry_linkpath.s, 1212 archive_strlen(&(tar->entry_linkpath)), tar->sconv) != 0) { 1213 err = set_conversion_failed_error(a, tar->sconv, 1214 "Linkname"); 1215 if (err == ARCHIVE_FATAL) 1216 return (err); 1217 } 1218 break; 1219 case '3': /* Character device */ 1220 archive_entry_set_filetype(entry, AE_IFCHR); 1221 archive_entry_set_size(entry, 0); 1222 tar->entry_bytes_remaining = 0; 1223 break; 1224 case '4': /* Block device */ 1225 archive_entry_set_filetype(entry, AE_IFBLK); 1226 archive_entry_set_size(entry, 0); 1227 tar->entry_bytes_remaining = 0; 1228 break; 1229 case '5': /* Dir */ 1230 archive_entry_set_filetype(entry, AE_IFDIR); 1231 archive_entry_set_size(entry, 0); 1232 tar->entry_bytes_remaining = 0; 1233 break; 1234 case '6': /* FIFO device */ 1235 archive_entry_set_filetype(entry, AE_IFIFO); 1236 archive_entry_set_size(entry, 0); 1237 tar->entry_bytes_remaining = 0; 1238 break; 1239 case 'D': /* GNU incremental directory type */ 1240 /* 1241 * No special handling is actually required here. 1242 * It might be nice someday to preprocess the file list and 1243 * provide it to the client, though. 1244 */ 1245 archive_entry_set_filetype(entry, AE_IFDIR); 1246 break; 1247 case 'M': /* GNU "Multi-volume" (remainder of file from last archive)*/ 1248 /* 1249 * As far as I can tell, this is just like a regular file 1250 * entry, except that the contents should be _appended_ to 1251 * the indicated file at the indicated offset. This may 1252 * require some API work to fully support. 1253 */ 1254 break; 1255 case 'N': /* Old GNU "long filename" entry. */ 1256 /* The body of this entry is a script for renaming 1257 * previously-extracted entries. Ugh. It will never 1258 * be supported by libarchive. */ 1259 archive_entry_set_filetype(entry, AE_IFREG); 1260 break; 1261 case 'S': /* GNU sparse files */ 1262 /* 1263 * Sparse files are really just regular files with 1264 * sparse information in the extended area. 1265 */ 1266 /* FALLTHROUGH */ 1267 default: /* Regular file and non-standard types */ 1268 /* 1269 * Per POSIX: non-recognized types should always be 1270 * treated as regular files. 1271 */ 1272 archive_entry_set_filetype(entry, AE_IFREG); 1273 break; 1274 } 1275 return (err); 1276 } 1277 1278 /* 1279 * Parse out header elements for "old-style" tar archives. 1280 */ 1281 static int 1282 header_old_tar(struct archive_read *a, struct tar *tar, 1283 struct archive_entry *entry, const void *h) 1284 { 1285 const struct archive_entry_header_ustar *header; 1286 int err = ARCHIVE_OK, err2; 1287 1288 /* Copy filename over (to ensure null termination). */ 1289 header = (const struct archive_entry_header_ustar *)h; 1290 if (archive_entry_copy_pathname_l(entry, 1291 header->name, sizeof(header->name), tar->sconv) != 0) { 1292 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 1293 if (err == ARCHIVE_FATAL) 1294 return (err); 1295 } 1296 1297 /* Grab rest of common fields */ 1298 err2 = header_common(a, tar, entry, h); 1299 if (err > err2) 1300 err = err2; 1301 1302 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 1303 return (err); 1304 } 1305 1306 /* 1307 * Read a Mac AppleDouble-encoded blob of file metadata, 1308 * if there is one. 1309 */ 1310 static int 1311 read_mac_metadata_blob(struct archive_read *a, struct tar *tar, 1312 struct archive_entry *entry, const void *h, size_t *unconsumed) 1313 { 1314 int64_t size; 1315 const void *data; 1316 const char *p, *name; 1317 const wchar_t *wp, *wname; 1318 1319 (void)h; /* UNUSED */ 1320 1321 wname = wp = archive_entry_pathname_w(entry); 1322 if (wp != NULL) { 1323 /* Find the last path element. */ 1324 for (; *wp != L'\0'; ++wp) { 1325 if (wp[0] == '/' && wp[1] != L'\0') 1326 wname = wp + 1; 1327 } 1328 /* 1329 * If last path element starts with "._", then 1330 * this is a Mac extension. 1331 */ 1332 if (wname[0] != L'.' || wname[1] != L'_' || wname[2] == L'\0') 1333 return ARCHIVE_OK; 1334 } else { 1335 /* Find the last path element. */ 1336 name = p = archive_entry_pathname(entry); 1337 if (p == NULL) 1338 return (ARCHIVE_FAILED); 1339 for (; *p != '\0'; ++p) { 1340 if (p[0] == '/' && p[1] != '\0') 1341 name = p + 1; 1342 } 1343 /* 1344 * If last path element starts with "._", then 1345 * this is a Mac extension. 1346 */ 1347 if (name[0] != '.' || name[1] != '_' || name[2] == '\0') 1348 return ARCHIVE_OK; 1349 } 1350 1351 /* Read the body as a Mac OS metadata blob. */ 1352 size = archive_entry_size(entry); 1353 1354 /* 1355 * TODO: Look beyond the body here to peek at the next header. 1356 * If it's a regular header (not an extension header) 1357 * that has the wrong name, just return the current 1358 * entry as-is, without consuming the body here. 1359 * That would reduce the risk of us mis-identifying 1360 * an ordinary file that just happened to have 1361 * a name starting with "._". 1362 * 1363 * Q: Is the above idea really possible? Even 1364 * when there are GNU or pax extension entries? 1365 */ 1366 data = __archive_read_ahead(a, (size_t)size, NULL); 1367 if (data == NULL) { 1368 *unconsumed = 0; 1369 return (ARCHIVE_FATAL); 1370 } 1371 archive_entry_copy_mac_metadata(entry, data, (size_t)size); 1372 *unconsumed = (size_t)((size + 511) & ~ 511); 1373 tar_flush_unconsumed(a, unconsumed); 1374 return (tar_read_header(a, tar, entry, unconsumed)); 1375 } 1376 1377 /* 1378 * Parse a file header for a pax extended archive entry. 1379 */ 1380 static int 1381 header_pax_global(struct archive_read *a, struct tar *tar, 1382 struct archive_entry *entry, const void *h, size_t *unconsumed) 1383 { 1384 int err; 1385 1386 err = read_body_to_string(a, tar, &(tar->pax_global), h, unconsumed); 1387 if (err != ARCHIVE_OK) 1388 return (err); 1389 err = tar_read_header(a, tar, entry, unconsumed); 1390 return (err); 1391 } 1392 1393 static int 1394 header_pax_extensions(struct archive_read *a, struct tar *tar, 1395 struct archive_entry *entry, const void *h, size_t *unconsumed) 1396 { 1397 int err, err2; 1398 1399 err = read_body_to_string(a, tar, &(tar->pax_header), h, unconsumed); 1400 if (err != ARCHIVE_OK) 1401 return (err); 1402 1403 /* Parse the next header. */ 1404 err = tar_read_header(a, tar, entry, unconsumed); 1405 if ((err != ARCHIVE_OK) && (err != ARCHIVE_WARN)) 1406 return (err); 1407 1408 /* 1409 * TODO: Parse global/default options into 'entry' struct here 1410 * before handling file-specific options. 1411 * 1412 * This design (parse standard header, then overwrite with pax 1413 * extended attribute data) usually works well, but isn't ideal; 1414 * it would be better to parse the pax extended attributes first 1415 * and then skip any fields in the standard header that were 1416 * defined in the pax header. 1417 */ 1418 err2 = pax_header(a, tar, entry, tar->pax_header.s); 1419 err = err_combine(err, err2); 1420 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 1421 return (err); 1422 } 1423 1424 1425 /* 1426 * Parse a file header for a Posix "ustar" archive entry. This also 1427 * handles "pax" or "extended ustar" entries. 1428 */ 1429 static int 1430 header_ustar(struct archive_read *a, struct tar *tar, 1431 struct archive_entry *entry, const void *h) 1432 { 1433 const struct archive_entry_header_ustar *header; 1434 struct archive_string *as; 1435 int err = ARCHIVE_OK, r; 1436 1437 header = (const struct archive_entry_header_ustar *)h; 1438 1439 /* Copy name into an internal buffer to ensure null-termination. */ 1440 as = &(tar->entry_pathname); 1441 if (header->prefix[0]) { 1442 archive_strncpy(as, header->prefix, sizeof(header->prefix)); 1443 if (as->s[archive_strlen(as) - 1] != '/') 1444 archive_strappend_char(as, '/'); 1445 archive_strncat(as, header->name, sizeof(header->name)); 1446 } else { 1447 archive_strncpy(as, header->name, sizeof(header->name)); 1448 } 1449 if (archive_entry_copy_pathname_l(entry, as->s, archive_strlen(as), 1450 tar->sconv) != 0) { 1451 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 1452 if (err == ARCHIVE_FATAL) 1453 return (err); 1454 } 1455 1456 /* Handle rest of common fields. */ 1457 r = header_common(a, tar, entry, h); 1458 if (r == ARCHIVE_FATAL) 1459 return (r); 1460 if (r < err) 1461 err = r; 1462 1463 /* Handle POSIX ustar fields. */ 1464 if (archive_entry_copy_uname_l(entry, 1465 header->uname, sizeof(header->uname), tar->sconv) != 0) { 1466 err = set_conversion_failed_error(a, tar->sconv, "Uname"); 1467 if (err == ARCHIVE_FATAL) 1468 return (err); 1469 } 1470 1471 if (archive_entry_copy_gname_l(entry, 1472 header->gname, sizeof(header->gname), tar->sconv) != 0) { 1473 err = set_conversion_failed_error(a, tar->sconv, "Gname"); 1474 if (err == ARCHIVE_FATAL) 1475 return (err); 1476 } 1477 1478 /* Parse out device numbers only for char and block specials. */ 1479 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') { 1480 archive_entry_set_rdevmajor(entry, (dev_t) 1481 tar_atol(header->rdevmajor, sizeof(header->rdevmajor))); 1482 archive_entry_set_rdevminor(entry, (dev_t) 1483 tar_atol(header->rdevminor, sizeof(header->rdevminor))); 1484 } 1485 1486 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 1487 1488 return (err); 1489 } 1490 1491 1492 /* 1493 * Parse the pax extended attributes record. 1494 * 1495 * Returns non-zero if there's an error in the data. 1496 */ 1497 static int 1498 pax_header(struct archive_read *a, struct tar *tar, 1499 struct archive_entry *entry, char *attr) 1500 { 1501 size_t attr_length, l, line_length; 1502 char *p; 1503 char *key, *value; 1504 struct archive_string *as; 1505 struct archive_string_conv *sconv; 1506 int err, err2; 1507 1508 attr_length = strlen(attr); 1509 tar->pax_hdrcharset_binary = 0; 1510 archive_string_empty(&(tar->entry_gname)); 1511 archive_string_empty(&(tar->entry_linkpath)); 1512 archive_string_empty(&(tar->entry_pathname)); 1513 archive_string_empty(&(tar->entry_pathname_override)); 1514 archive_string_empty(&(tar->entry_uname)); 1515 err = ARCHIVE_OK; 1516 while (attr_length > 0) { 1517 /* Parse decimal length field at start of line. */ 1518 line_length = 0; 1519 l = attr_length; 1520 p = attr; /* Record start of line. */ 1521 while (l>0) { 1522 if (*p == ' ') { 1523 p++; 1524 l--; 1525 break; 1526 } 1527 if (*p < '0' || *p > '9') { 1528 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1529 "Ignoring malformed pax extended attributes"); 1530 return (ARCHIVE_WARN); 1531 } 1532 line_length *= 10; 1533 line_length += *p - '0'; 1534 if (line_length > 999999) { 1535 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1536 "Rejecting pax extended attribute > 1MB"); 1537 return (ARCHIVE_WARN); 1538 } 1539 p++; 1540 l--; 1541 } 1542 1543 /* 1544 * Parsed length must be no bigger than available data, 1545 * at least 1, and the last character of the line must 1546 * be '\n'. 1547 */ 1548 if (line_length > attr_length 1549 || line_length < 1 1550 || attr[line_length - 1] != '\n') 1551 { 1552 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1553 "Ignoring malformed pax extended attribute"); 1554 return (ARCHIVE_WARN); 1555 } 1556 1557 /* Null-terminate the line. */ 1558 attr[line_length - 1] = '\0'; 1559 1560 /* Find end of key and null terminate it. */ 1561 key = p; 1562 if (key[0] == '=') 1563 return (-1); 1564 while (*p && *p != '=') 1565 ++p; 1566 if (*p == '\0') { 1567 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, 1568 "Invalid pax extended attributes"); 1569 return (ARCHIVE_WARN); 1570 } 1571 *p = '\0'; 1572 1573 /* Identify null-terminated 'value' portion. */ 1574 value = p + 1; 1575 1576 /* Identify this attribute and set it in the entry. */ 1577 err2 = pax_attribute(a, tar, entry, key, value); 1578 if (err2 == ARCHIVE_FATAL) 1579 return (err2); 1580 err = err_combine(err, err2); 1581 1582 /* Skip to next line */ 1583 attr += line_length; 1584 attr_length -= line_length; 1585 } 1586 1587 /* 1588 * PAX format uses UTF-8 as default charset for its metadata 1589 * unless hdrcharset=BINARY is present in its header. 1590 * We apply the charset specified by the hdrcharset option only 1591 * when the hdrcharset attribute(in PAX header) is BINARY because 1592 * we respect the charset described in PAX header and BINARY also 1593 * means that metadata(filename,uname and gname) character-set 1594 * is unknown. 1595 */ 1596 if (tar->pax_hdrcharset_binary) 1597 sconv = tar->opt_sconv; 1598 else { 1599 sconv = archive_string_conversion_from_charset( 1600 &(a->archive), "UTF-8", 1); 1601 if (sconv == NULL) 1602 return (ARCHIVE_FATAL); 1603 if (tar->compat_2x) 1604 archive_string_conversion_set_opt(sconv, 1605 SCONV_SET_OPT_UTF8_LIBARCHIVE2X); 1606 } 1607 1608 if (archive_strlen(&(tar->entry_gname)) > 0) { 1609 if (archive_entry_copy_gname_l(entry, tar->entry_gname.s, 1610 archive_strlen(&(tar->entry_gname)), sconv) != 0) { 1611 err = set_conversion_failed_error(a, sconv, "Gname"); 1612 if (err == ARCHIVE_FATAL) 1613 return (err); 1614 /* Use a converted an original name. */ 1615 archive_entry_copy_gname(entry, tar->entry_gname.s); 1616 } 1617 } 1618 if (archive_strlen(&(tar->entry_linkpath)) > 0) { 1619 if (archive_entry_copy_link_l(entry, tar->entry_linkpath.s, 1620 archive_strlen(&(tar->entry_linkpath)), sconv) != 0) { 1621 err = set_conversion_failed_error(a, sconv, "Linkname"); 1622 if (err == ARCHIVE_FATAL) 1623 return (err); 1624 /* Use a converted an original name. */ 1625 archive_entry_copy_link(entry, tar->entry_linkpath.s); 1626 } 1627 } 1628 /* 1629 * Some extensions (such as the GNU sparse file extensions) 1630 * deliberately store a synthetic name under the regular 'path' 1631 * attribute and the real file name under a different attribute. 1632 * Since we're supposed to not care about the order, we 1633 * have no choice but to store all of the various filenames 1634 * we find and figure it all out afterwards. This is the 1635 * figuring out part. 1636 */ 1637 as = NULL; 1638 if (archive_strlen(&(tar->entry_pathname_override)) > 0) 1639 as = &(tar->entry_pathname_override); 1640 else if (archive_strlen(&(tar->entry_pathname)) > 0) 1641 as = &(tar->entry_pathname); 1642 if (as != NULL) { 1643 if (archive_entry_copy_pathname_l(entry, as->s, 1644 archive_strlen(as), sconv) != 0) { 1645 err = set_conversion_failed_error(a, sconv, "Pathname"); 1646 if (err == ARCHIVE_FATAL) 1647 return (err); 1648 /* Use a converted an original name. */ 1649 archive_entry_copy_pathname(entry, as->s); 1650 } 1651 } 1652 if (archive_strlen(&(tar->entry_uname)) > 0) { 1653 if (archive_entry_copy_uname_l(entry, tar->entry_uname.s, 1654 archive_strlen(&(tar->entry_uname)), sconv) != 0) { 1655 err = set_conversion_failed_error(a, sconv, "Uname"); 1656 if (err == ARCHIVE_FATAL) 1657 return (err); 1658 /* Use a converted an original name. */ 1659 archive_entry_copy_uname(entry, tar->entry_uname.s); 1660 } 1661 } 1662 return (err); 1663 } 1664 1665 static int 1666 pax_attribute_xattr(struct archive_entry *entry, 1667 char *name, char *value) 1668 { 1669 char *name_decoded; 1670 void *value_decoded; 1671 size_t value_len; 1672 1673 if (strlen(name) < 18 || (memcmp(name, "LIBARCHIVE.xattr.", 17)) != 0) 1674 return 3; 1675 1676 name += 17; 1677 1678 /* URL-decode name */ 1679 name_decoded = url_decode(name); 1680 if (name_decoded == NULL) 1681 return 2; 1682 1683 /* Base-64 decode value */ 1684 value_decoded = base64_decode(value, strlen(value), &value_len); 1685 if (value_decoded == NULL) { 1686 free(name_decoded); 1687 return 1; 1688 } 1689 1690 archive_entry_xattr_add_entry(entry, name_decoded, 1691 value_decoded, value_len); 1692 1693 free(name_decoded); 1694 free(value_decoded); 1695 return 0; 1696 } 1697 1698 /* 1699 * Parse a single key=value attribute. key/value pointers are 1700 * assumed to point into reasonably long-lived storage. 1701 * 1702 * Note that POSIX reserves all-lowercase keywords. Vendor-specific 1703 * extensions should always have keywords of the form "VENDOR.attribute" 1704 * In particular, it's quite feasible to support many different 1705 * vendor extensions here. I'm using "LIBARCHIVE" for extensions 1706 * unique to this library. 1707 * 1708 * Investigate other vendor-specific extensions and see if 1709 * any of them look useful. 1710 */ 1711 static int 1712 pax_attribute(struct archive_read *a, struct tar *tar, 1713 struct archive_entry *entry, char *key, char *value) 1714 { 1715 int64_t s; 1716 long n; 1717 int err = ARCHIVE_OK, r; 1718 1719 if (value == NULL) 1720 value = ""; /* Disable compiler warning; do not pass 1721 * NULL pointer to strlen(). */ 1722 switch (key[0]) { 1723 case 'G': 1724 /* GNU "0.0" sparse pax format. */ 1725 if (strcmp(key, "GNU.sparse.numblocks") == 0) { 1726 tar->sparse_offset = -1; 1727 tar->sparse_numbytes = -1; 1728 tar->sparse_gnu_major = 0; 1729 tar->sparse_gnu_minor = 0; 1730 } 1731 if (strcmp(key, "GNU.sparse.offset") == 0) { 1732 tar->sparse_offset = tar_atol10(value, strlen(value)); 1733 if (tar->sparse_numbytes != -1) { 1734 if (gnu_add_sparse_entry(a, tar, 1735 tar->sparse_offset, tar->sparse_numbytes) 1736 != ARCHIVE_OK) 1737 return (ARCHIVE_FATAL); 1738 tar->sparse_offset = -1; 1739 tar->sparse_numbytes = -1; 1740 } 1741 } 1742 if (strcmp(key, "GNU.sparse.numbytes") == 0) { 1743 tar->sparse_numbytes = tar_atol10(value, strlen(value)); 1744 if (tar->sparse_numbytes != -1) { 1745 if (gnu_add_sparse_entry(a, tar, 1746 tar->sparse_offset, tar->sparse_numbytes) 1747 != ARCHIVE_OK) 1748 return (ARCHIVE_FATAL); 1749 tar->sparse_offset = -1; 1750 tar->sparse_numbytes = -1; 1751 } 1752 } 1753 if (strcmp(key, "GNU.sparse.size") == 0) { 1754 tar->realsize = tar_atol10(value, strlen(value)); 1755 archive_entry_set_size(entry, tar->realsize); 1756 } 1757 1758 /* GNU "0.1" sparse pax format. */ 1759 if (strcmp(key, "GNU.sparse.map") == 0) { 1760 tar->sparse_gnu_major = 0; 1761 tar->sparse_gnu_minor = 1; 1762 if (gnu_sparse_01_parse(a, tar, value) != ARCHIVE_OK) 1763 return (ARCHIVE_WARN); 1764 } 1765 1766 /* GNU "1.0" sparse pax format */ 1767 if (strcmp(key, "GNU.sparse.major") == 0) { 1768 tar->sparse_gnu_major = (int)tar_atol10(value, strlen(value)); 1769 tar->sparse_gnu_pending = 1; 1770 } 1771 if (strcmp(key, "GNU.sparse.minor") == 0) { 1772 tar->sparse_gnu_minor = (int)tar_atol10(value, strlen(value)); 1773 tar->sparse_gnu_pending = 1; 1774 } 1775 if (strcmp(key, "GNU.sparse.name") == 0) { 1776 /* 1777 * The real filename; when storing sparse 1778 * files, GNU tar puts a synthesized name into 1779 * the regular 'path' attribute in an attempt 1780 * to limit confusion. ;-) 1781 */ 1782 archive_strcpy(&(tar->entry_pathname_override), value); 1783 } 1784 if (strcmp(key, "GNU.sparse.realsize") == 0) { 1785 tar->realsize = tar_atol10(value, strlen(value)); 1786 archive_entry_set_size(entry, tar->realsize); 1787 } 1788 break; 1789 case 'L': 1790 /* Our extensions */ 1791 /* TODO: Handle arbitrary extended attributes... */ 1792 /* 1793 if (strcmp(key, "LIBARCHIVE.xxxxxxx") == 0) 1794 archive_entry_set_xxxxxx(entry, value); 1795 */ 1796 if (strcmp(key, "LIBARCHIVE.creationtime") == 0) { 1797 pax_time(value, &s, &n); 1798 archive_entry_set_birthtime(entry, s, n); 1799 } 1800 if (memcmp(key, "LIBARCHIVE.xattr.", 17) == 0) 1801 pax_attribute_xattr(entry, key, value); 1802 break; 1803 case 'S': 1804 /* We support some keys used by the "star" archiver */ 1805 if (strcmp(key, "SCHILY.acl.access") == 0) { 1806 if (tar->sconv_acl == NULL) { 1807 tar->sconv_acl = 1808 archive_string_conversion_from_charset( 1809 &(a->archive), "UTF-8", 1); 1810 if (tar->sconv_acl == NULL) 1811 return (ARCHIVE_FATAL); 1812 } 1813 1814 r = archive_acl_parse_l(archive_entry_acl(entry), 1815 value, ARCHIVE_ENTRY_ACL_TYPE_ACCESS, 1816 tar->sconv_acl); 1817 if (r != ARCHIVE_OK) { 1818 err = r; 1819 if (err == ARCHIVE_FATAL) { 1820 archive_set_error(&a->archive, ENOMEM, 1821 "Can't allocate memory for " 1822 "SCHILY.acl.access"); 1823 return (err); 1824 } 1825 archive_set_error(&a->archive, 1826 ARCHIVE_ERRNO_MISC, 1827 "Parse error: SCHILY.acl.access"); 1828 } 1829 } else if (strcmp(key, "SCHILY.acl.default") == 0) { 1830 if (tar->sconv_acl == NULL) { 1831 tar->sconv_acl = 1832 archive_string_conversion_from_charset( 1833 &(a->archive), "UTF-8", 1); 1834 if (tar->sconv_acl == NULL) 1835 return (ARCHIVE_FATAL); 1836 } 1837 1838 r = archive_acl_parse_l(archive_entry_acl(entry), 1839 value, ARCHIVE_ENTRY_ACL_TYPE_DEFAULT, 1840 tar->sconv_acl); 1841 if (r != ARCHIVE_OK) { 1842 err = r; 1843 if (err == ARCHIVE_FATAL) { 1844 archive_set_error(&a->archive, ENOMEM, 1845 "Can't allocate memory for " 1846 "SCHILY.acl.default"); 1847 return (err); 1848 } 1849 archive_set_error(&a->archive, 1850 ARCHIVE_ERRNO_MISC, 1851 "Parse error: SCHILY.acl.default"); 1852 } 1853 } else if (strcmp(key, "SCHILY.devmajor") == 0) { 1854 archive_entry_set_rdevmajor(entry, 1855 (dev_t)tar_atol10(value, strlen(value))); 1856 } else if (strcmp(key, "SCHILY.devminor") == 0) { 1857 archive_entry_set_rdevminor(entry, 1858 (dev_t)tar_atol10(value, strlen(value))); 1859 } else if (strcmp(key, "SCHILY.fflags") == 0) { 1860 archive_entry_copy_fflags_text(entry, value); 1861 } else if (strcmp(key, "SCHILY.dev") == 0) { 1862 archive_entry_set_dev(entry, 1863 (dev_t)tar_atol10(value, strlen(value))); 1864 } else if (strcmp(key, "SCHILY.ino") == 0) { 1865 archive_entry_set_ino(entry, 1866 tar_atol10(value, strlen(value))); 1867 } else if (strcmp(key, "SCHILY.nlink") == 0) { 1868 archive_entry_set_nlink(entry, (unsigned) 1869 tar_atol10(value, strlen(value))); 1870 } else if (strcmp(key, "SCHILY.realsize") == 0) { 1871 tar->realsize = tar_atol10(value, strlen(value)); 1872 archive_entry_set_size(entry, tar->realsize); 1873 } else if (strcmp(key, "SUN.holesdata") == 0) { 1874 /* A Solaris extension for sparse. */ 1875 r = solaris_sparse_parse(a, tar, entry, value); 1876 if (r < err) { 1877 if (r == ARCHIVE_FATAL) 1878 return (r); 1879 err = r; 1880 archive_set_error(&a->archive, 1881 ARCHIVE_ERRNO_MISC, 1882 "Parse error: SUN.holesdata"); 1883 } 1884 } 1885 break; 1886 case 'a': 1887 if (strcmp(key, "atime") == 0) { 1888 pax_time(value, &s, &n); 1889 archive_entry_set_atime(entry, s, n); 1890 } 1891 break; 1892 case 'c': 1893 if (strcmp(key, "ctime") == 0) { 1894 pax_time(value, &s, &n); 1895 archive_entry_set_ctime(entry, s, n); 1896 } else if (strcmp(key, "charset") == 0) { 1897 /* TODO: Publish charset information in entry. */ 1898 } else if (strcmp(key, "comment") == 0) { 1899 /* TODO: Publish comment in entry. */ 1900 } 1901 break; 1902 case 'g': 1903 if (strcmp(key, "gid") == 0) { 1904 archive_entry_set_gid(entry, 1905 tar_atol10(value, strlen(value))); 1906 } else if (strcmp(key, "gname") == 0) { 1907 archive_strcpy(&(tar->entry_gname), value); 1908 } 1909 break; 1910 case 'h': 1911 if (strcmp(key, "hdrcharset") == 0) { 1912 if (strcmp(value, "BINARY") == 0) 1913 /* Binary mode. */ 1914 tar->pax_hdrcharset_binary = 1; 1915 else if (strcmp(value, "ISO-IR 10646 2000 UTF-8") == 0) 1916 tar->pax_hdrcharset_binary = 0; 1917 } 1918 break; 1919 case 'l': 1920 /* pax interchange doesn't distinguish hardlink vs. symlink. */ 1921 if (strcmp(key, "linkpath") == 0) { 1922 archive_strcpy(&(tar->entry_linkpath), value); 1923 } 1924 break; 1925 case 'm': 1926 if (strcmp(key, "mtime") == 0) { 1927 pax_time(value, &s, &n); 1928 archive_entry_set_mtime(entry, s, n); 1929 } 1930 break; 1931 case 'p': 1932 if (strcmp(key, "path") == 0) { 1933 archive_strcpy(&(tar->entry_pathname), value); 1934 } 1935 break; 1936 case 'r': 1937 /* POSIX has reserved 'realtime.*' */ 1938 break; 1939 case 's': 1940 /* POSIX has reserved 'security.*' */ 1941 /* Someday: if (strcmp(key, "security.acl") == 0) { ... } */ 1942 if (strcmp(key, "size") == 0) { 1943 /* "size" is the size of the data in the entry. */ 1944 tar->entry_bytes_remaining 1945 = tar_atol10(value, strlen(value)); 1946 /* 1947 * But, "size" is not necessarily the size of 1948 * the file on disk; if this is a sparse file, 1949 * the disk size may have already been set from 1950 * GNU.sparse.realsize or GNU.sparse.size or 1951 * an old GNU header field or SCHILY.realsize 1952 * or .... 1953 */ 1954 if (tar->realsize < 0) { 1955 archive_entry_set_size(entry, 1956 tar->entry_bytes_remaining); 1957 tar->realsize 1958 = tar->entry_bytes_remaining; 1959 } 1960 } 1961 break; 1962 case 'u': 1963 if (strcmp(key, "uid") == 0) { 1964 archive_entry_set_uid(entry, 1965 tar_atol10(value, strlen(value))); 1966 } else if (strcmp(key, "uname") == 0) { 1967 archive_strcpy(&(tar->entry_uname), value); 1968 } 1969 break; 1970 } 1971 return (err); 1972 } 1973 1974 1975 1976 /* 1977 * parse a decimal time value, which may include a fractional portion 1978 */ 1979 static void 1980 pax_time(const char *p, int64_t *ps, long *pn) 1981 { 1982 char digit; 1983 int64_t s; 1984 unsigned long l; 1985 int sign; 1986 int64_t limit, last_digit_limit; 1987 1988 limit = INT64_MAX / 10; 1989 last_digit_limit = INT64_MAX % 10; 1990 1991 s = 0; 1992 sign = 1; 1993 if (*p == '-') { 1994 sign = -1; 1995 p++; 1996 } 1997 while (*p >= '0' && *p <= '9') { 1998 digit = *p - '0'; 1999 if (s > limit || 2000 (s == limit && digit > last_digit_limit)) { 2001 s = INT64_MAX; 2002 break; 2003 } 2004 s = (s * 10) + digit; 2005 ++p; 2006 } 2007 2008 *ps = s * sign; 2009 2010 /* Calculate nanoseconds. */ 2011 *pn = 0; 2012 2013 if (*p != '.') 2014 return; 2015 2016 l = 100000000UL; 2017 do { 2018 ++p; 2019 if (*p >= '0' && *p <= '9') 2020 *pn += (*p - '0') * l; 2021 else 2022 break; 2023 } while (l /= 10); 2024 } 2025 2026 /* 2027 * Parse GNU tar header 2028 */ 2029 static int 2030 header_gnutar(struct archive_read *a, struct tar *tar, 2031 struct archive_entry *entry, const void *h, size_t *unconsumed) 2032 { 2033 const struct archive_entry_header_gnutar *header; 2034 int64_t t; 2035 int err = ARCHIVE_OK; 2036 2037 /* 2038 * GNU header is like POSIX ustar, except 'prefix' is 2039 * replaced with some other fields. This also means the 2040 * filename is stored as in old-style archives. 2041 */ 2042 2043 /* Grab fields common to all tar variants. */ 2044 err = header_common(a, tar, entry, h); 2045 if (err == ARCHIVE_FATAL) 2046 return (err); 2047 2048 /* Copy filename over (to ensure null termination). */ 2049 header = (const struct archive_entry_header_gnutar *)h; 2050 if (archive_entry_copy_pathname_l(entry, 2051 header->name, sizeof(header->name), tar->sconv) != 0) { 2052 err = set_conversion_failed_error(a, tar->sconv, "Pathname"); 2053 if (err == ARCHIVE_FATAL) 2054 return (err); 2055 } 2056 2057 /* Fields common to ustar and GNU */ 2058 /* XXX Can the following be factored out since it's common 2059 * to ustar and gnu tar? Is it okay to move it down into 2060 * header_common, perhaps? */ 2061 if (archive_entry_copy_uname_l(entry, 2062 header->uname, sizeof(header->uname), tar->sconv) != 0) { 2063 err = set_conversion_failed_error(a, tar->sconv, "Uname"); 2064 if (err == ARCHIVE_FATAL) 2065 return (err); 2066 } 2067 2068 if (archive_entry_copy_gname_l(entry, 2069 header->gname, sizeof(header->gname), tar->sconv) != 0) { 2070 err = set_conversion_failed_error(a, tar->sconv, "Gname"); 2071 if (err == ARCHIVE_FATAL) 2072 return (err); 2073 } 2074 2075 /* Parse out device numbers only for char and block specials */ 2076 if (header->typeflag[0] == '3' || header->typeflag[0] == '4') { 2077 archive_entry_set_rdevmajor(entry, (dev_t) 2078 tar_atol(header->rdevmajor, sizeof(header->rdevmajor))); 2079 archive_entry_set_rdevminor(entry, (dev_t) 2080 tar_atol(header->rdevminor, sizeof(header->rdevminor))); 2081 } else 2082 archive_entry_set_rdev(entry, 0); 2083 2084 tar->entry_padding = 0x1ff & (-tar->entry_bytes_remaining); 2085 2086 /* Grab GNU-specific fields. */ 2087 t = tar_atol(header->atime, sizeof(header->atime)); 2088 if (t > 0) 2089 archive_entry_set_atime(entry, t, 0); 2090 t = tar_atol(header->ctime, sizeof(header->ctime)); 2091 if (t > 0) 2092 archive_entry_set_ctime(entry, t, 0); 2093 2094 if (header->realsize[0] != 0) { 2095 tar->realsize 2096 = tar_atol(header->realsize, sizeof(header->realsize)); 2097 archive_entry_set_size(entry, tar->realsize); 2098 } 2099 2100 if (header->sparse[0].offset[0] != 0) { 2101 if (gnu_sparse_old_read(a, tar, header, unconsumed) 2102 != ARCHIVE_OK) 2103 return (ARCHIVE_FATAL); 2104 } else { 2105 if (header->isextended[0] != 0) { 2106 /* XXX WTF? XXX */ 2107 } 2108 } 2109 2110 return (err); 2111 } 2112 2113 static int 2114 gnu_add_sparse_entry(struct archive_read *a, struct tar *tar, 2115 int64_t offset, int64_t remaining) 2116 { 2117 struct sparse_block *p; 2118 2119 p = (struct sparse_block *)malloc(sizeof(*p)); 2120 if (p == NULL) { 2121 archive_set_error(&a->archive, ENOMEM, "Out of memory"); 2122 return (ARCHIVE_FATAL); 2123 } 2124 memset(p, 0, sizeof(*p)); 2125 if (tar->sparse_last != NULL) 2126 tar->sparse_last->next = p; 2127 else 2128 tar->sparse_list = p; 2129 tar->sparse_last = p; 2130 if (remaining < 0 || offset < 0) { 2131 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Malformed sparse map data"); 2132 return (ARCHIVE_FATAL); 2133 } 2134 p->offset = offset; 2135 p->remaining = remaining; 2136 return (ARCHIVE_OK); 2137 } 2138 2139 static void 2140 gnu_clear_sparse_list(struct tar *tar) 2141 { 2142 struct sparse_block *p; 2143 2144 while (tar->sparse_list != NULL) { 2145 p = tar->sparse_list; 2146 tar->sparse_list = p->next; 2147 free(p); 2148 } 2149 tar->sparse_last = NULL; 2150 } 2151 2152 /* 2153 * GNU tar old-format sparse data. 2154 * 2155 * GNU old-format sparse data is stored in a fixed-field 2156 * format. Offset/size values are 11-byte octal fields (same 2157 * format as 'size' field in ustart header). These are 2158 * stored in the header, allocating subsequent header blocks 2159 * as needed. Extending the header in this way is a pretty 2160 * severe POSIX violation; this design has earned GNU tar a 2161 * lot of criticism. 2162 */ 2163 2164 static int 2165 gnu_sparse_old_read(struct archive_read *a, struct tar *tar, 2166 const struct archive_entry_header_gnutar *header, size_t *unconsumed) 2167 { 2168 ssize_t bytes_read; 2169 const void *data; 2170 struct extended { 2171 struct gnu_sparse sparse[21]; 2172 char isextended[1]; 2173 char padding[7]; 2174 }; 2175 const struct extended *ext; 2176 2177 if (gnu_sparse_old_parse(a, tar, header->sparse, 4) != ARCHIVE_OK) 2178 return (ARCHIVE_FATAL); 2179 if (header->isextended[0] == 0) 2180 return (ARCHIVE_OK); 2181 2182 do { 2183 tar_flush_unconsumed(a, unconsumed); 2184 data = __archive_read_ahead(a, 512, &bytes_read); 2185 if (bytes_read < 0) 2186 return (ARCHIVE_FATAL); 2187 if (bytes_read < 512) { 2188 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 2189 "Truncated tar archive " 2190 "detected while reading sparse file data"); 2191 return (ARCHIVE_FATAL); 2192 } 2193 *unconsumed = 512; 2194 ext = (const struct extended *)data; 2195 if (gnu_sparse_old_parse(a, tar, ext->sparse, 21) != ARCHIVE_OK) 2196 return (ARCHIVE_FATAL); 2197 } while (ext->isextended[0] != 0); 2198 if (tar->sparse_list != NULL) 2199 tar->entry_offset = tar->sparse_list->offset; 2200 return (ARCHIVE_OK); 2201 } 2202 2203 static int 2204 gnu_sparse_old_parse(struct archive_read *a, struct tar *tar, 2205 const struct gnu_sparse *sparse, int length) 2206 { 2207 while (length > 0 && sparse->offset[0] != 0) { 2208 if (gnu_add_sparse_entry(a, tar, 2209 tar_atol(sparse->offset, sizeof(sparse->offset)), 2210 tar_atol(sparse->numbytes, sizeof(sparse->numbytes))) 2211 != ARCHIVE_OK) 2212 return (ARCHIVE_FATAL); 2213 sparse++; 2214 length--; 2215 } 2216 return (ARCHIVE_OK); 2217 } 2218 2219 /* 2220 * GNU tar sparse format 0.0 2221 * 2222 * Beginning with GNU tar 1.15, sparse files are stored using 2223 * information in the pax extended header. The GNU tar maintainers 2224 * have gone through a number of variations in the process of working 2225 * out this scheme; fortunately, they're all numbered. 2226 * 2227 * Sparse format 0.0 uses attribute GNU.sparse.numblocks to store the 2228 * number of blocks, and GNU.sparse.offset/GNU.sparse.numbytes to 2229 * store offset/size for each block. The repeated instances of these 2230 * latter fields violate the pax specification (which frowns on 2231 * duplicate keys), so this format was quickly replaced. 2232 */ 2233 2234 /* 2235 * GNU tar sparse format 0.1 2236 * 2237 * This version replaced the offset/numbytes attributes with 2238 * a single "map" attribute that stored a list of integers. This 2239 * format had two problems: First, the "map" attribute could be very 2240 * long, which caused problems for some implementations. More 2241 * importantly, the sparse data was lost when extracted by archivers 2242 * that didn't recognize this extension. 2243 */ 2244 2245 static int 2246 gnu_sparse_01_parse(struct archive_read *a, struct tar *tar, const char *p) 2247 { 2248 const char *e; 2249 int64_t offset = -1, size = -1; 2250 2251 for (;;) { 2252 e = p; 2253 while (*e != '\0' && *e != ',') { 2254 if (*e < '0' || *e > '9') 2255 return (ARCHIVE_WARN); 2256 e++; 2257 } 2258 if (offset < 0) { 2259 offset = tar_atol10(p, e - p); 2260 if (offset < 0) 2261 return (ARCHIVE_WARN); 2262 } else { 2263 size = tar_atol10(p, e - p); 2264 if (size < 0) 2265 return (ARCHIVE_WARN); 2266 if (gnu_add_sparse_entry(a, tar, offset, size) 2267 != ARCHIVE_OK) 2268 return (ARCHIVE_FATAL); 2269 offset = -1; 2270 } 2271 if (*e == '\0') 2272 return (ARCHIVE_OK); 2273 p = e + 1; 2274 } 2275 } 2276 2277 /* 2278 * GNU tar sparse format 1.0 2279 * 2280 * The idea: The offset/size data is stored as a series of base-10 2281 * ASCII numbers prepended to the file data, so that dearchivers that 2282 * don't support this format will extract the block map along with the 2283 * data and a separate post-process can restore the sparseness. 2284 * 2285 * Unfortunately, GNU tar 1.16 had a bug that added unnecessary 2286 * padding to the body of the file when using this format. GNU tar 2287 * 1.17 corrected this bug without bumping the version number, so 2288 * it's not possible to support both variants. This code supports 2289 * the later variant at the expense of not supporting the former. 2290 * 2291 * This variant also replaced GNU.sparse.size with GNU.sparse.realsize 2292 * and introduced the GNU.sparse.major/GNU.sparse.minor attributes. 2293 */ 2294 2295 /* 2296 * Read the next line from the input, and parse it as a decimal 2297 * integer followed by '\n'. Returns positive integer value or 2298 * negative on error. 2299 */ 2300 static int64_t 2301 gnu_sparse_10_atol(struct archive_read *a, struct tar *tar, 2302 int64_t *remaining, size_t *unconsumed) 2303 { 2304 int64_t l, limit, last_digit_limit; 2305 const char *p; 2306 ssize_t bytes_read; 2307 int base, digit; 2308 2309 base = 10; 2310 limit = INT64_MAX / base; 2311 last_digit_limit = INT64_MAX % base; 2312 2313 /* 2314 * Skip any lines starting with '#'; GNU tar specs 2315 * don't require this, but they should. 2316 */ 2317 do { 2318 bytes_read = readline(a, tar, &p, 2319 (ssize_t)tar_min(*remaining, 100), unconsumed); 2320 if (bytes_read <= 0) 2321 return (ARCHIVE_FATAL); 2322 *remaining -= bytes_read; 2323 } while (p[0] == '#'); 2324 2325 l = 0; 2326 while (bytes_read > 0) { 2327 if (*p == '\n') 2328 return (l); 2329 if (*p < '0' || *p >= '0' + base) 2330 return (ARCHIVE_WARN); 2331 digit = *p - '0'; 2332 if (l > limit || (l == limit && digit > last_digit_limit)) 2333 l = INT64_MAX; /* Truncate on overflow. */ 2334 else 2335 l = (l * base) + digit; 2336 p++; 2337 bytes_read--; 2338 } 2339 /* TODO: Error message. */ 2340 return (ARCHIVE_WARN); 2341 } 2342 2343 /* 2344 * Returns length (in bytes) of the sparse data description 2345 * that was read. 2346 */ 2347 static ssize_t 2348 gnu_sparse_10_read(struct archive_read *a, struct tar *tar, size_t *unconsumed) 2349 { 2350 ssize_t bytes_read; 2351 int entries; 2352 int64_t offset, size, to_skip, remaining; 2353 2354 /* Clear out the existing sparse list. */ 2355 gnu_clear_sparse_list(tar); 2356 2357 remaining = tar->entry_bytes_remaining; 2358 2359 /* Parse entries. */ 2360 entries = (int)gnu_sparse_10_atol(a, tar, &remaining, unconsumed); 2361 if (entries < 0) 2362 return (ARCHIVE_FATAL); 2363 /* Parse the individual entries. */ 2364 while (entries-- > 0) { 2365 /* Parse offset/size */ 2366 offset = gnu_sparse_10_atol(a, tar, &remaining, unconsumed); 2367 if (offset < 0) 2368 return (ARCHIVE_FATAL); 2369 size = gnu_sparse_10_atol(a, tar, &remaining, unconsumed); 2370 if (size < 0) 2371 return (ARCHIVE_FATAL); 2372 /* Add a new sparse entry. */ 2373 if (gnu_add_sparse_entry(a, tar, offset, size) != ARCHIVE_OK) 2374 return (ARCHIVE_FATAL); 2375 } 2376 /* Skip rest of block... */ 2377 tar_flush_unconsumed(a, unconsumed); 2378 bytes_read = (ssize_t)(tar->entry_bytes_remaining - remaining); 2379 to_skip = 0x1ff & -bytes_read; 2380 if (to_skip != __archive_read_consume(a, to_skip)) 2381 return (ARCHIVE_FATAL); 2382 return ((ssize_t)(bytes_read + to_skip)); 2383 } 2384 2385 /* 2386 * Solaris pax extension for a sparse file. This is recorded with the 2387 * data and hole pairs. The way recording sparse information by Solaris' 2388 * pax simply indicates where data and sparse are, so the stored contents 2389 * consist of both data and hole. 2390 */ 2391 static int 2392 solaris_sparse_parse(struct archive_read *a, struct tar *tar, 2393 struct archive_entry *entry, const char *p) 2394 { 2395 const char *e; 2396 int64_t start, end; 2397 int hole = 1; 2398 2399 (void)entry; /* UNUSED */ 2400 2401 end = 0; 2402 if (*p == ' ') 2403 p++; 2404 else 2405 return (ARCHIVE_WARN); 2406 for (;;) { 2407 e = p; 2408 while (*e != '\0' && *e != ' ') { 2409 if (*e < '0' || *e > '9') 2410 return (ARCHIVE_WARN); 2411 e++; 2412 } 2413 start = end; 2414 end = tar_atol10(p, e - p); 2415 if (end < 0) 2416 return (ARCHIVE_WARN); 2417 if (start < end) { 2418 if (gnu_add_sparse_entry(a, tar, start, 2419 end - start) != ARCHIVE_OK) 2420 return (ARCHIVE_FATAL); 2421 tar->sparse_last->hole = hole; 2422 } 2423 if (*e == '\0') 2424 return (ARCHIVE_OK); 2425 p = e + 1; 2426 hole = hole == 0; 2427 } 2428 } 2429 2430 /*- 2431 * Convert text->integer. 2432 * 2433 * Traditional tar formats (including POSIX) specify base-8 for 2434 * all of the standard numeric fields. This is a significant limitation 2435 * in practice: 2436 * = file size is limited to 8GB 2437 * = rdevmajor and rdevminor are limited to 21 bits 2438 * = uid/gid are limited to 21 bits 2439 * 2440 * There are two workarounds for this: 2441 * = pax extended headers, which use variable-length string fields 2442 * = GNU tar and STAR both allow either base-8 or base-256 in 2443 * most fields. The high bit is set to indicate base-256. 2444 * 2445 * On read, this implementation supports both extensions. 2446 */ 2447 static int64_t 2448 tar_atol(const char *p, size_t char_cnt) 2449 { 2450 /* 2451 * Technically, GNU tar considers a field to be in base-256 2452 * only if the first byte is 0xff or 0x80. 2453 */ 2454 if (*p & 0x80) 2455 return (tar_atol256(p, char_cnt)); 2456 return (tar_atol8(p, char_cnt)); 2457 } 2458 2459 /* 2460 * Note that this implementation does not (and should not!) obey 2461 * locale settings; you cannot simply substitute strtol here, since 2462 * it does obey locale. 2463 */ 2464 static int64_t 2465 tar_atol_base_n(const char *p, size_t char_cnt, int base) 2466 { 2467 int64_t l, maxval, limit, last_digit_limit; 2468 int digit, sign; 2469 2470 maxval = INT64_MAX; 2471 limit = INT64_MAX / base; 2472 last_digit_limit = INT64_MAX % base; 2473 2474 /* the pointer will not be dereferenced if char_cnt is zero 2475 * due to the way the && operator is evaulated. 2476 */ 2477 while (char_cnt != 0 && (*p == ' ' || *p == '\t')) { 2478 p++; 2479 char_cnt--; 2480 } 2481 2482 sign = 1; 2483 if (char_cnt != 0 && *p == '-') { 2484 sign = -1; 2485 p++; 2486 char_cnt--; 2487 2488 maxval = INT64_MIN; 2489 limit = -(INT64_MIN / base); 2490 last_digit_limit = INT64_MIN % base; 2491 } 2492 2493 l = 0; 2494 if (char_cnt != 0) { 2495 digit = *p - '0'; 2496 while (digit >= 0 && digit < base && char_cnt != 0) { 2497 if (l>limit || (l == limit && digit > last_digit_limit)) { 2498 return maxval; /* Truncate on overflow. */ 2499 } 2500 l = (l * base) + digit; 2501 digit = *++p - '0'; 2502 char_cnt--; 2503 } 2504 } 2505 return (sign < 0) ? -l : l; 2506 } 2507 2508 static int64_t 2509 tar_atol8(const char *p, size_t char_cnt) 2510 { 2511 return tar_atol_base_n(p, char_cnt, 8); 2512 } 2513 2514 static int64_t 2515 tar_atol10(const char *p, size_t char_cnt) 2516 { 2517 return tar_atol_base_n(p, char_cnt, 10); 2518 } 2519 2520 /* 2521 * Parse a base-256 integer. This is just a variable-length 2522 * twos-complement signed binary value in big-endian order, except 2523 * that the high-order bit is ignored. The values here can be up to 2524 * 12 bytes, so we need to be careful about overflowing 64-bit 2525 * (8-byte) integers. 2526 * 2527 * This code unashamedly assumes that the local machine uses 8-bit 2528 * bytes and twos-complement arithmetic. 2529 */ 2530 static int64_t 2531 tar_atol256(const char *_p, size_t char_cnt) 2532 { 2533 uint64_t l; 2534 const unsigned char *p = (const unsigned char *)_p; 2535 unsigned char c, neg; 2536 2537 /* Extend 7-bit 2s-comp to 8-bit 2s-comp, decide sign. */ 2538 c = *p; 2539 if (c & 0x40) { 2540 neg = 0xff; 2541 c |= 0x80; 2542 l = ~ARCHIVE_LITERAL_ULL(0); 2543 } else { 2544 neg = 0; 2545 c &= 0x7f; 2546 l = 0; 2547 } 2548 2549 /* If more than 8 bytes, check that we can ignore 2550 * high-order bits without overflow. */ 2551 while (char_cnt > sizeof(int64_t)) { 2552 --char_cnt; 2553 if (c != neg) 2554 return neg ? INT64_MIN : INT64_MAX; 2555 c = *++p; 2556 } 2557 2558 /* c is first byte that fits; if sign mismatch, return overflow */ 2559 if ((c ^ neg) & 0x80) { 2560 return neg ? INT64_MIN : INT64_MAX; 2561 } 2562 2563 /* Accumulate remaining bytes. */ 2564 while (--char_cnt > 0) { 2565 l = (l << 8) | c; 2566 c = *++p; 2567 } 2568 l = (l << 8) | c; 2569 /* Return signed twos-complement value. */ 2570 return (int64_t)(l); 2571 } 2572 2573 /* 2574 * Returns length of line (including trailing newline) 2575 * or negative on error. 'start' argument is updated to 2576 * point to first character of line. This avoids copying 2577 * when possible. 2578 */ 2579 static ssize_t 2580 readline(struct archive_read *a, struct tar *tar, const char **start, 2581 ssize_t limit, size_t *unconsumed) 2582 { 2583 ssize_t bytes_read; 2584 ssize_t total_size = 0; 2585 const void *t; 2586 const char *s; 2587 void *p; 2588 2589 tar_flush_unconsumed(a, unconsumed); 2590 2591 t = __archive_read_ahead(a, 1, &bytes_read); 2592 if (bytes_read <= 0) 2593 return (ARCHIVE_FATAL); 2594 s = t; /* Start of line? */ 2595 p = memchr(t, '\n', bytes_read); 2596 /* If we found '\n' in the read buffer, return pointer to that. */ 2597 if (p != NULL) { 2598 bytes_read = 1 + ((const char *)p) - s; 2599 if (bytes_read > limit) { 2600 archive_set_error(&a->archive, 2601 ARCHIVE_ERRNO_FILE_FORMAT, 2602 "Line too long"); 2603 return (ARCHIVE_FATAL); 2604 } 2605 *unconsumed = bytes_read; 2606 *start = s; 2607 return (bytes_read); 2608 } 2609 *unconsumed = bytes_read; 2610 /* Otherwise, we need to accumulate in a line buffer. */ 2611 for (;;) { 2612 if (total_size + bytes_read > limit) { 2613 archive_set_error(&a->archive, 2614 ARCHIVE_ERRNO_FILE_FORMAT, 2615 "Line too long"); 2616 return (ARCHIVE_FATAL); 2617 } 2618 if (archive_string_ensure(&tar->line, total_size + bytes_read) == NULL) { 2619 archive_set_error(&a->archive, ENOMEM, 2620 "Can't allocate working buffer"); 2621 return (ARCHIVE_FATAL); 2622 } 2623 memcpy(tar->line.s + total_size, t, bytes_read); 2624 tar_flush_unconsumed(a, unconsumed); 2625 total_size += bytes_read; 2626 /* If we found '\n', clean up and return. */ 2627 if (p != NULL) { 2628 *start = tar->line.s; 2629 return (total_size); 2630 } 2631 /* Read some more. */ 2632 t = __archive_read_ahead(a, 1, &bytes_read); 2633 if (bytes_read <= 0) 2634 return (ARCHIVE_FATAL); 2635 s = t; /* Start of line? */ 2636 p = memchr(t, '\n', bytes_read); 2637 /* If we found '\n', trim the read. */ 2638 if (p != NULL) { 2639 bytes_read = 1 + ((const char *)p) - s; 2640 } 2641 *unconsumed = bytes_read; 2642 } 2643 } 2644 2645 /* 2646 * base64_decode - Base64 decode 2647 * 2648 * This accepts most variations of base-64 encoding, including: 2649 * * with or without line breaks 2650 * * with or without the final group padded with '=' or '_' characters 2651 * (The most economical Base-64 variant does not pad the last group and 2652 * omits line breaks; RFC1341 used for MIME requires both.) 2653 */ 2654 static char * 2655 base64_decode(const char *s, size_t len, size_t *out_len) 2656 { 2657 static const unsigned char digits[64] = { 2658 'A','B','C','D','E','F','G','H','I','J','K','L','M','N', 2659 'O','P','Q','R','S','T','U','V','W','X','Y','Z','a','b', 2660 'c','d','e','f','g','h','i','j','k','l','m','n','o','p', 2661 'q','r','s','t','u','v','w','x','y','z','0','1','2','3', 2662 '4','5','6','7','8','9','+','/' }; 2663 static unsigned char decode_table[128]; 2664 char *out, *d; 2665 const unsigned char *src = (const unsigned char *)s; 2666 2667 /* If the decode table is not yet initialized, prepare it. */ 2668 if (decode_table[digits[1]] != 1) { 2669 unsigned i; 2670 memset(decode_table, 0xff, sizeof(decode_table)); 2671 for (i = 0; i < sizeof(digits); i++) 2672 decode_table[digits[i]] = i; 2673 } 2674 2675 /* Allocate enough space to hold the entire output. */ 2676 /* Note that we may not use all of this... */ 2677 out = (char *)malloc(len - len / 4 + 1); 2678 if (out == NULL) { 2679 *out_len = 0; 2680 return (NULL); 2681 } 2682 d = out; 2683 2684 while (len > 0) { 2685 /* Collect the next group of (up to) four characters. */ 2686 int v = 0; 2687 int group_size = 0; 2688 while (group_size < 4 && len > 0) { 2689 /* '=' or '_' padding indicates final group. */ 2690 if (*src == '=' || *src == '_') { 2691 len = 0; 2692 break; 2693 } 2694 /* Skip illegal characters (including line breaks) */ 2695 if (*src > 127 || *src < 32 2696 || decode_table[*src] == 0xff) { 2697 len--; 2698 src++; 2699 continue; 2700 } 2701 v <<= 6; 2702 v |= decode_table[*src++]; 2703 len --; 2704 group_size++; 2705 } 2706 /* Align a short group properly. */ 2707 v <<= 6 * (4 - group_size); 2708 /* Unpack the group we just collected. */ 2709 switch (group_size) { 2710 case 4: d[2] = v & 0xff; 2711 /* FALLTHROUGH */ 2712 case 3: d[1] = (v >> 8) & 0xff; 2713 /* FALLTHROUGH */ 2714 case 2: d[0] = (v >> 16) & 0xff; 2715 break; 2716 case 1: /* this is invalid! */ 2717 break; 2718 } 2719 d += group_size * 3 / 4; 2720 } 2721 2722 *out_len = d - out; 2723 return (out); 2724 } 2725 2726 static char * 2727 url_decode(const char *in) 2728 { 2729 char *out, *d; 2730 const char *s; 2731 2732 out = (char *)malloc(strlen(in) + 1); 2733 if (out == NULL) 2734 return (NULL); 2735 for (s = in, d = out; *s != '\0'; ) { 2736 if (s[0] == '%' && s[1] != '\0' && s[2] != '\0') { 2737 /* Try to convert % escape */ 2738 int digit1 = tohex(s[1]); 2739 int digit2 = tohex(s[2]); 2740 if (digit1 >= 0 && digit2 >= 0) { 2741 /* Looks good, consume three chars */ 2742 s += 3; 2743 /* Convert output */ 2744 *d++ = ((digit1 << 4) | digit2); 2745 continue; 2746 } 2747 /* Else fall through and treat '%' as normal char */ 2748 } 2749 *d++ = *s++; 2750 } 2751 *d = '\0'; 2752 return (out); 2753 } 2754 2755 static int 2756 tohex(int c) 2757 { 2758 if (c >= '0' && c <= '9') 2759 return (c - '0'); 2760 else if (c >= 'A' && c <= 'F') 2761 return (c - 'A' + 10); 2762 else if (c >= 'a' && c <= 'f') 2763 return (c - 'a' + 10); 2764 else 2765 return (-1); 2766 } 2767