1 /*- 2 * Copyright (c) 2003-2011 Tim Kientzle 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR 15 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 16 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 17 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, 18 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 19 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 20 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 21 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 22 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 23 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 24 */ 25 26 /* 27 * This file contains the "essential" portions of the read API, that 28 * is, stuff that will probably always be used by any client that 29 * actually needs to read an archive. Optional pieces have been, as 30 * far as possible, separated out into separate files to avoid 31 * needlessly bloating statically-linked clients. 32 */ 33 34 #include "archive_platform.h" 35 __FBSDID("$FreeBSD$"); 36 37 #ifdef HAVE_ERRNO_H 38 #include <errno.h> 39 #endif 40 #include <stdio.h> 41 #ifdef HAVE_STDLIB_H 42 #include <stdlib.h> 43 #endif 44 #ifdef HAVE_STRING_H 45 #include <string.h> 46 #endif 47 #ifdef HAVE_UNISTD_H 48 #include <unistd.h> 49 #endif 50 51 #include "archive.h" 52 #include "archive_entry.h" 53 #include "archive_private.h" 54 #include "archive_read_private.h" 55 56 #define minimum(a, b) (a < b ? a : b) 57 58 static int choose_filters(struct archive_read *); 59 static int choose_format(struct archive_read *); 60 static int close_filters(struct archive_read *); 61 static struct archive_vtable *archive_read_vtable(void); 62 static int64_t _archive_filter_bytes(struct archive *, int); 63 static int _archive_filter_code(struct archive *, int); 64 static const char *_archive_filter_name(struct archive *, int); 65 static int _archive_filter_count(struct archive *); 66 static int _archive_read_close(struct archive *); 67 static int _archive_read_data_block(struct archive *, 68 const void **, size_t *, int64_t *); 69 static int _archive_read_free(struct archive *); 70 static int _archive_read_next_header(struct archive *, 71 struct archive_entry **); 72 static int _archive_read_next_header2(struct archive *, 73 struct archive_entry *); 74 static int64_t advance_file_pointer(struct archive_read_filter *, int64_t); 75 76 static struct archive_vtable * 77 archive_read_vtable(void) 78 { 79 static struct archive_vtable av; 80 static int inited = 0; 81 82 if (!inited) { 83 av.archive_filter_bytes = _archive_filter_bytes; 84 av.archive_filter_code = _archive_filter_code; 85 av.archive_filter_name = _archive_filter_name; 86 av.archive_filter_count = _archive_filter_count; 87 av.archive_read_data_block = _archive_read_data_block; 88 av.archive_read_next_header = _archive_read_next_header; 89 av.archive_read_next_header2 = _archive_read_next_header2; 90 av.archive_free = _archive_read_free; 91 av.archive_close = _archive_read_close; 92 inited = 1; 93 } 94 return (&av); 95 } 96 97 /* 98 * Allocate, initialize and return a struct archive object. 99 */ 100 struct archive * 101 archive_read_new(void) 102 { 103 struct archive_read *a; 104 105 a = (struct archive_read *)calloc(1, sizeof(*a)); 106 if (a == NULL) 107 return (NULL); 108 a->archive.magic = ARCHIVE_READ_MAGIC; 109 110 a->archive.state = ARCHIVE_STATE_NEW; 111 a->entry = archive_entry_new2(&a->archive); 112 a->archive.vtable = archive_read_vtable(); 113 114 a->passphrases.last = &a->passphrases.first; 115 116 return (&a->archive); 117 } 118 119 /* 120 * Record the do-not-extract-to file. This belongs in archive_read_extract.c. 121 */ 122 void 123 archive_read_extract_set_skip_file(struct archive *_a, la_int64_t d, 124 la_int64_t i) 125 { 126 struct archive_read *a = (struct archive_read *)_a; 127 128 if (ARCHIVE_OK != __archive_check_magic(_a, ARCHIVE_READ_MAGIC, 129 ARCHIVE_STATE_ANY, "archive_read_extract_set_skip_file")) 130 return; 131 a->skip_file_set = 1; 132 a->skip_file_dev = d; 133 a->skip_file_ino = i; 134 } 135 136 /* 137 * Open the archive 138 */ 139 int 140 archive_read_open(struct archive *a, void *client_data, 141 archive_open_callback *client_opener, archive_read_callback *client_reader, 142 archive_close_callback *client_closer) 143 { 144 /* Old archive_read_open() is just a thin shell around 145 * archive_read_open1. */ 146 archive_read_set_open_callback(a, client_opener); 147 archive_read_set_read_callback(a, client_reader); 148 archive_read_set_close_callback(a, client_closer); 149 archive_read_set_callback_data(a, client_data); 150 return archive_read_open1(a); 151 } 152 153 154 int 155 archive_read_open2(struct archive *a, void *client_data, 156 archive_open_callback *client_opener, 157 archive_read_callback *client_reader, 158 archive_skip_callback *client_skipper, 159 archive_close_callback *client_closer) 160 { 161 /* Old archive_read_open2() is just a thin shell around 162 * archive_read_open1. */ 163 archive_read_set_callback_data(a, client_data); 164 archive_read_set_open_callback(a, client_opener); 165 archive_read_set_read_callback(a, client_reader); 166 archive_read_set_skip_callback(a, client_skipper); 167 archive_read_set_close_callback(a, client_closer); 168 return archive_read_open1(a); 169 } 170 171 static ssize_t 172 client_read_proxy(struct archive_read_filter *self, const void **buff) 173 { 174 ssize_t r; 175 r = (self->archive->client.reader)(&self->archive->archive, 176 self->data, buff); 177 return (r); 178 } 179 180 static int64_t 181 client_skip_proxy(struct archive_read_filter *self, int64_t request) 182 { 183 if (request < 0) 184 __archive_errx(1, "Negative skip requested."); 185 if (request == 0) 186 return 0; 187 188 if (self->archive->client.skipper != NULL) { 189 /* Seek requests over 1GiB are broken down into 190 * multiple seeks. This avoids overflows when the 191 * requests get passed through 32-bit arguments. */ 192 int64_t skip_limit = (int64_t)1 << 30; 193 int64_t total = 0; 194 for (;;) { 195 int64_t get, ask = request; 196 if (ask > skip_limit) 197 ask = skip_limit; 198 get = (self->archive->client.skipper) 199 (&self->archive->archive, self->data, ask); 200 total += get; 201 if (get == 0 || get == request) 202 return (total); 203 if (get > request) 204 return ARCHIVE_FATAL; 205 request -= get; 206 } 207 } else if (self->archive->client.seeker != NULL 208 && request > 64 * 1024) { 209 /* If the client provided a seeker but not a skipper, 210 * we can use the seeker to skip forward. 211 * 212 * Note: This isn't always a good idea. The client 213 * skipper is allowed to skip by less than requested 214 * if it needs to maintain block alignment. The 215 * seeker is not allowed to play such games, so using 216 * the seeker here may be a performance loss compared 217 * to just reading and discarding. That's why we 218 * only do this for skips of over 64k. 219 */ 220 int64_t before = self->position; 221 int64_t after = (self->archive->client.seeker) 222 (&self->archive->archive, self->data, request, SEEK_CUR); 223 if (after != before + request) 224 return ARCHIVE_FATAL; 225 return after - before; 226 } 227 return 0; 228 } 229 230 static int64_t 231 client_seek_proxy(struct archive_read_filter *self, int64_t offset, int whence) 232 { 233 /* DO NOT use the skipper here! If we transparently handled 234 * forward seek here by using the skipper, that will break 235 * other libarchive code that assumes a successful forward 236 * seek means it can also seek backwards. 237 */ 238 if (self->archive->client.seeker == NULL) { 239 archive_set_error(&self->archive->archive, ARCHIVE_ERRNO_MISC, 240 "Current client reader does not support seeking a device"); 241 return (ARCHIVE_FAILED); 242 } 243 return (self->archive->client.seeker)(&self->archive->archive, 244 self->data, offset, whence); 245 } 246 247 static int 248 client_close_proxy(struct archive_read_filter *self) 249 { 250 int r = ARCHIVE_OK, r2; 251 unsigned int i; 252 253 if (self->archive->client.closer == NULL) 254 return (r); 255 for (i = 0; i < self->archive->client.nodes; i++) 256 { 257 r2 = (self->archive->client.closer) 258 ((struct archive *)self->archive, 259 self->archive->client.dataset[i].data); 260 if (r > r2) 261 r = r2; 262 } 263 return (r); 264 } 265 266 static int 267 client_open_proxy(struct archive_read_filter *self) 268 { 269 int r = ARCHIVE_OK; 270 if (self->archive->client.opener != NULL) 271 r = (self->archive->client.opener)( 272 (struct archive *)self->archive, self->data); 273 return (r); 274 } 275 276 static int 277 client_switch_proxy(struct archive_read_filter *self, unsigned int iindex) 278 { 279 int r1 = ARCHIVE_OK, r2 = ARCHIVE_OK; 280 void *data2 = NULL; 281 282 /* Don't do anything if already in the specified data node */ 283 if (self->archive->client.cursor == iindex) 284 return (ARCHIVE_OK); 285 286 self->archive->client.cursor = iindex; 287 data2 = self->archive->client.dataset[self->archive->client.cursor].data; 288 if (self->archive->client.switcher != NULL) 289 { 290 r1 = r2 = (self->archive->client.switcher) 291 ((struct archive *)self->archive, self->data, data2); 292 self->data = data2; 293 } 294 else 295 { 296 /* Attempt to call close and open instead */ 297 if (self->archive->client.closer != NULL) 298 r1 = (self->archive->client.closer) 299 ((struct archive *)self->archive, self->data); 300 self->data = data2; 301 if (self->archive->client.opener != NULL) 302 r2 = (self->archive->client.opener) 303 ((struct archive *)self->archive, self->data); 304 } 305 return (r1 < r2) ? r1 : r2; 306 } 307 308 int 309 archive_read_set_open_callback(struct archive *_a, 310 archive_open_callback *client_opener) 311 { 312 struct archive_read *a = (struct archive_read *)_a; 313 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW, 314 "archive_read_set_open_callback"); 315 a->client.opener = client_opener; 316 return ARCHIVE_OK; 317 } 318 319 int 320 archive_read_set_read_callback(struct archive *_a, 321 archive_read_callback *client_reader) 322 { 323 struct archive_read *a = (struct archive_read *)_a; 324 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW, 325 "archive_read_set_read_callback"); 326 a->client.reader = client_reader; 327 return ARCHIVE_OK; 328 } 329 330 int 331 archive_read_set_skip_callback(struct archive *_a, 332 archive_skip_callback *client_skipper) 333 { 334 struct archive_read *a = (struct archive_read *)_a; 335 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW, 336 "archive_read_set_skip_callback"); 337 a->client.skipper = client_skipper; 338 return ARCHIVE_OK; 339 } 340 341 int 342 archive_read_set_seek_callback(struct archive *_a, 343 archive_seek_callback *client_seeker) 344 { 345 struct archive_read *a = (struct archive_read *)_a; 346 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW, 347 "archive_read_set_seek_callback"); 348 a->client.seeker = client_seeker; 349 return ARCHIVE_OK; 350 } 351 352 int 353 archive_read_set_close_callback(struct archive *_a, 354 archive_close_callback *client_closer) 355 { 356 struct archive_read *a = (struct archive_read *)_a; 357 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW, 358 "archive_read_set_close_callback"); 359 a->client.closer = client_closer; 360 return ARCHIVE_OK; 361 } 362 363 int 364 archive_read_set_switch_callback(struct archive *_a, 365 archive_switch_callback *client_switcher) 366 { 367 struct archive_read *a = (struct archive_read *)_a; 368 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW, 369 "archive_read_set_switch_callback"); 370 a->client.switcher = client_switcher; 371 return ARCHIVE_OK; 372 } 373 374 int 375 archive_read_set_callback_data(struct archive *_a, void *client_data) 376 { 377 return archive_read_set_callback_data2(_a, client_data, 0); 378 } 379 380 int 381 archive_read_set_callback_data2(struct archive *_a, void *client_data, 382 unsigned int iindex) 383 { 384 struct archive_read *a = (struct archive_read *)_a; 385 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW, 386 "archive_read_set_callback_data2"); 387 388 if (a->client.nodes == 0) 389 { 390 a->client.dataset = (struct archive_read_data_node *) 391 calloc(1, sizeof(*a->client.dataset)); 392 if (a->client.dataset == NULL) 393 { 394 archive_set_error(&a->archive, ENOMEM, 395 "No memory."); 396 return ARCHIVE_FATAL; 397 } 398 a->client.nodes = 1; 399 } 400 401 if (iindex > a->client.nodes - 1) 402 { 403 archive_set_error(&a->archive, EINVAL, 404 "Invalid index specified."); 405 return ARCHIVE_FATAL; 406 } 407 a->client.dataset[iindex].data = client_data; 408 a->client.dataset[iindex].begin_position = -1; 409 a->client.dataset[iindex].total_size = -1; 410 return ARCHIVE_OK; 411 } 412 413 int 414 archive_read_add_callback_data(struct archive *_a, void *client_data, 415 unsigned int iindex) 416 { 417 struct archive_read *a = (struct archive_read *)_a; 418 void *p; 419 unsigned int i; 420 421 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW, 422 "archive_read_add_callback_data"); 423 if (iindex > a->client.nodes) { 424 archive_set_error(&a->archive, EINVAL, 425 "Invalid index specified."); 426 return ARCHIVE_FATAL; 427 } 428 p = realloc(a->client.dataset, sizeof(*a->client.dataset) 429 * (++(a->client.nodes))); 430 if (p == NULL) { 431 archive_set_error(&a->archive, ENOMEM, 432 "No memory."); 433 return ARCHIVE_FATAL; 434 } 435 a->client.dataset = (struct archive_read_data_node *)p; 436 for (i = a->client.nodes - 1; i > iindex; i--) { 437 a->client.dataset[i].data = a->client.dataset[i-1].data; 438 a->client.dataset[i].begin_position = -1; 439 a->client.dataset[i].total_size = -1; 440 } 441 a->client.dataset[iindex].data = client_data; 442 a->client.dataset[iindex].begin_position = -1; 443 a->client.dataset[iindex].total_size = -1; 444 return ARCHIVE_OK; 445 } 446 447 int 448 archive_read_append_callback_data(struct archive *_a, void *client_data) 449 { 450 struct archive_read *a = (struct archive_read *)_a; 451 return archive_read_add_callback_data(_a, client_data, a->client.nodes); 452 } 453 454 int 455 archive_read_prepend_callback_data(struct archive *_a, void *client_data) 456 { 457 return archive_read_add_callback_data(_a, client_data, 0); 458 } 459 460 int 461 archive_read_open1(struct archive *_a) 462 { 463 struct archive_read *a = (struct archive_read *)_a; 464 struct archive_read_filter *filter, *tmp; 465 int slot, e = ARCHIVE_OK; 466 unsigned int i; 467 468 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW, 469 "archive_read_open"); 470 archive_clear_error(&a->archive); 471 472 if (a->client.reader == NULL) { 473 archive_set_error(&a->archive, EINVAL, 474 "No reader function provided to archive_read_open"); 475 a->archive.state = ARCHIVE_STATE_FATAL; 476 return (ARCHIVE_FATAL); 477 } 478 479 /* Open data source. */ 480 if (a->client.opener != NULL) { 481 e = (a->client.opener)(&a->archive, a->client.dataset[0].data); 482 if (e != 0) { 483 /* If the open failed, call the closer to clean up. */ 484 if (a->client.closer) { 485 for (i = 0; i < a->client.nodes; i++) 486 (a->client.closer)(&a->archive, 487 a->client.dataset[i].data); 488 } 489 return (e); 490 } 491 } 492 493 filter = calloc(1, sizeof(*filter)); 494 if (filter == NULL) 495 return (ARCHIVE_FATAL); 496 filter->bidder = NULL; 497 filter->upstream = NULL; 498 filter->archive = a; 499 filter->data = a->client.dataset[0].data; 500 filter->open = client_open_proxy; 501 filter->read = client_read_proxy; 502 filter->skip = client_skip_proxy; 503 filter->seek = client_seek_proxy; 504 filter->close = client_close_proxy; 505 filter->sswitch = client_switch_proxy; 506 filter->name = "none"; 507 filter->code = ARCHIVE_FILTER_NONE; 508 509 a->client.dataset[0].begin_position = 0; 510 if (!a->filter || !a->bypass_filter_bidding) 511 { 512 a->filter = filter; 513 /* Build out the input pipeline. */ 514 e = choose_filters(a); 515 if (e < ARCHIVE_WARN) { 516 a->archive.state = ARCHIVE_STATE_FATAL; 517 return (ARCHIVE_FATAL); 518 } 519 } 520 else 521 { 522 /* Need to add "NONE" type filter at the end of the filter chain */ 523 tmp = a->filter; 524 while (tmp->upstream) 525 tmp = tmp->upstream; 526 tmp->upstream = filter; 527 } 528 529 if (!a->format) 530 { 531 slot = choose_format(a); 532 if (slot < 0) { 533 close_filters(a); 534 a->archive.state = ARCHIVE_STATE_FATAL; 535 return (ARCHIVE_FATAL); 536 } 537 a->format = &(a->formats[slot]); 538 } 539 540 a->archive.state = ARCHIVE_STATE_HEADER; 541 542 /* Ensure libarchive starts from the first node in a multivolume set */ 543 client_switch_proxy(a->filter, 0); 544 return (e); 545 } 546 547 /* 548 * Allow each registered stream transform to bid on whether 549 * it wants to handle this stream. Repeat until we've finished 550 * building the pipeline. 551 */ 552 553 /* We won't build a filter pipeline with more stages than this. */ 554 #define MAX_NUMBER_FILTERS 25 555 556 static int 557 choose_filters(struct archive_read *a) 558 { 559 int number_bidders, i, bid, best_bid, number_filters; 560 struct archive_read_filter_bidder *bidder, *best_bidder; 561 struct archive_read_filter *filter; 562 ssize_t avail; 563 int r; 564 565 for (number_filters = 0; number_filters < MAX_NUMBER_FILTERS; ++number_filters) { 566 number_bidders = sizeof(a->bidders) / sizeof(a->bidders[0]); 567 568 best_bid = 0; 569 best_bidder = NULL; 570 571 bidder = a->bidders; 572 for (i = 0; i < number_bidders; i++, bidder++) { 573 if (bidder->bid != NULL) { 574 bid = (bidder->bid)(bidder, a->filter); 575 if (bid > best_bid) { 576 best_bid = bid; 577 best_bidder = bidder; 578 } 579 } 580 } 581 582 /* If no bidder, we're done. */ 583 if (best_bidder == NULL) { 584 /* Verify the filter by asking it for some data. */ 585 __archive_read_filter_ahead(a->filter, 1, &avail); 586 if (avail < 0) { 587 __archive_read_free_filters(a); 588 return (ARCHIVE_FATAL); 589 } 590 a->archive.compression_name = a->filter->name; 591 a->archive.compression_code = a->filter->code; 592 return (ARCHIVE_OK); 593 } 594 595 filter 596 = (struct archive_read_filter *)calloc(1, sizeof(*filter)); 597 if (filter == NULL) 598 return (ARCHIVE_FATAL); 599 filter->bidder = best_bidder; 600 filter->archive = a; 601 filter->upstream = a->filter; 602 a->filter = filter; 603 r = (best_bidder->init)(a->filter); 604 if (r != ARCHIVE_OK) { 605 __archive_read_free_filters(a); 606 return (ARCHIVE_FATAL); 607 } 608 } 609 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 610 "Input requires too many filters for decoding"); 611 return (ARCHIVE_FATAL); 612 } 613 614 int 615 __archive_read_header(struct archive_read *a, struct archive_entry *entry) 616 { 617 if (a->filter->read_header) 618 return a->filter->read_header(a->filter, entry); 619 else 620 return (ARCHIVE_OK); 621 } 622 623 /* 624 * Read header of next entry. 625 */ 626 static int 627 _archive_read_next_header2(struct archive *_a, struct archive_entry *entry) 628 { 629 struct archive_read *a = (struct archive_read *)_a; 630 int r1 = ARCHIVE_OK, r2; 631 632 archive_check_magic(_a, ARCHIVE_READ_MAGIC, 633 ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, 634 "archive_read_next_header"); 635 636 archive_entry_clear(entry); 637 archive_clear_error(&a->archive); 638 639 /* 640 * If client didn't consume entire data, skip any remainder 641 * (This is especially important for GNU incremental directories.) 642 */ 643 if (a->archive.state == ARCHIVE_STATE_DATA) { 644 r1 = archive_read_data_skip(&a->archive); 645 if (r1 == ARCHIVE_EOF) 646 archive_set_error(&a->archive, EIO, 647 "Premature end-of-file."); 648 if (r1 == ARCHIVE_EOF || r1 == ARCHIVE_FATAL) { 649 a->archive.state = ARCHIVE_STATE_FATAL; 650 return (ARCHIVE_FATAL); 651 } 652 } 653 654 /* Record start-of-header offset in uncompressed stream. */ 655 a->header_position = a->filter->position; 656 657 ++_a->file_count; 658 r2 = (a->format->read_header)(a, entry); 659 660 /* 661 * EOF and FATAL are persistent at this layer. By 662 * modifying the state, we guarantee that future calls to 663 * read a header or read data will fail. 664 */ 665 switch (r2) { 666 case ARCHIVE_EOF: 667 a->archive.state = ARCHIVE_STATE_EOF; 668 --_a->file_count;/* Revert a file counter. */ 669 break; 670 case ARCHIVE_OK: 671 a->archive.state = ARCHIVE_STATE_DATA; 672 break; 673 case ARCHIVE_WARN: 674 a->archive.state = ARCHIVE_STATE_DATA; 675 break; 676 case ARCHIVE_RETRY: 677 break; 678 case ARCHIVE_FATAL: 679 a->archive.state = ARCHIVE_STATE_FATAL; 680 break; 681 } 682 683 __archive_reset_read_data(&a->archive); 684 685 a->data_start_node = a->client.cursor; 686 /* EOF always wins; otherwise return the worst error. */ 687 return (r2 < r1 || r2 == ARCHIVE_EOF) ? r2 : r1; 688 } 689 690 static int 691 _archive_read_next_header(struct archive *_a, struct archive_entry **entryp) 692 { 693 int ret; 694 struct archive_read *a = (struct archive_read *)_a; 695 *entryp = NULL; 696 ret = _archive_read_next_header2(_a, a->entry); 697 *entryp = a->entry; 698 return ret; 699 } 700 701 /* 702 * Allow each registered format to bid on whether it wants to handle 703 * the next entry. Return index of winning bidder. 704 */ 705 static int 706 choose_format(struct archive_read *a) 707 { 708 int slots; 709 int i; 710 int bid, best_bid; 711 int best_bid_slot; 712 713 slots = sizeof(a->formats) / sizeof(a->formats[0]); 714 best_bid = -1; 715 best_bid_slot = -1; 716 717 /* Set up a->format for convenience of bidders. */ 718 a->format = &(a->formats[0]); 719 for (i = 0; i < slots; i++, a->format++) { 720 if (a->format->bid) { 721 bid = (a->format->bid)(a, best_bid); 722 if (bid == ARCHIVE_FATAL) 723 return (ARCHIVE_FATAL); 724 if (a->filter->position != 0) 725 __archive_read_seek(a, 0, SEEK_SET); 726 if ((bid > best_bid) || (best_bid_slot < 0)) { 727 best_bid = bid; 728 best_bid_slot = i; 729 } 730 } 731 } 732 733 /* 734 * There were no bidders; this is a serious programmer error 735 * and demands a quick and definitive abort. 736 */ 737 if (best_bid_slot < 0) { 738 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 739 "No formats registered"); 740 return (ARCHIVE_FATAL); 741 } 742 743 /* 744 * There were bidders, but no non-zero bids; this means we 745 * can't support this stream. 746 */ 747 if (best_bid < 1) { 748 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, 749 "Unrecognized archive format"); 750 return (ARCHIVE_FATAL); 751 } 752 753 return (best_bid_slot); 754 } 755 756 /* 757 * Return the file offset (within the uncompressed data stream) where 758 * the last header started. 759 */ 760 la_int64_t 761 archive_read_header_position(struct archive *_a) 762 { 763 struct archive_read *a = (struct archive_read *)_a; 764 archive_check_magic(_a, ARCHIVE_READ_MAGIC, 765 ARCHIVE_STATE_ANY, "archive_read_header_position"); 766 return (a->header_position); 767 } 768 769 /* 770 * Returns 1 if the archive contains at least one encrypted entry. 771 * If the archive format not support encryption at all 772 * ARCHIVE_READ_FORMAT_ENCRYPTION_UNSUPPORTED is returned. 773 * If for any other reason (e.g. not enough data read so far) 774 * we cannot say whether there are encrypted entries, then 775 * ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW is returned. 776 * In general, this function will return values below zero when the 777 * reader is uncertain or totally incapable of encryption support. 778 * When this function returns 0 you can be sure that the reader 779 * supports encryption detection but no encrypted entries have 780 * been found yet. 781 * 782 * NOTE: If the metadata/header of an archive is also encrypted, you 783 * cannot rely on the number of encrypted entries. That is why this 784 * function does not return the number of encrypted entries but# 785 * just shows that there are some. 786 */ 787 int 788 archive_read_has_encrypted_entries(struct archive *_a) 789 { 790 struct archive_read *a = (struct archive_read *)_a; 791 int format_supports_encryption = archive_read_format_capabilities(_a) 792 & (ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_DATA | ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_METADATA); 793 794 if (!_a || !format_supports_encryption) { 795 /* Format in general doesn't support encryption */ 796 return ARCHIVE_READ_FORMAT_ENCRYPTION_UNSUPPORTED; 797 } 798 799 /* A reader potentially has read enough data now. */ 800 if (a->format && a->format->has_encrypted_entries) { 801 return (a->format->has_encrypted_entries)(a); 802 } 803 804 /* For any other reason we cannot say how many entries are there. */ 805 return ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW; 806 } 807 808 /* 809 * Returns a bitmask of capabilities that are supported by the archive format reader. 810 * If the reader has no special capabilities, ARCHIVE_READ_FORMAT_CAPS_NONE is returned. 811 */ 812 int 813 archive_read_format_capabilities(struct archive *_a) 814 { 815 struct archive_read *a = (struct archive_read *)_a; 816 if (a && a->format && a->format->format_capabilties) { 817 return (a->format->format_capabilties)(a); 818 } 819 return ARCHIVE_READ_FORMAT_CAPS_NONE; 820 } 821 822 /* 823 * Read data from an archive entry, using a read(2)-style interface. 824 * This is a convenience routine that just calls 825 * archive_read_data_block and copies the results into the client 826 * buffer, filling any gaps with zero bytes. Clients using this 827 * API can be completely ignorant of sparse-file issues; sparse files 828 * will simply be padded with nulls. 829 * 830 * DO NOT intermingle calls to this function and archive_read_data_block 831 * to read a single entry body. 832 */ 833 la_ssize_t 834 archive_read_data(struct archive *_a, void *buff, size_t s) 835 { 836 struct archive *a = (struct archive *)_a; 837 char *dest; 838 const void *read_buf; 839 size_t bytes_read; 840 size_t len; 841 int r; 842 843 bytes_read = 0; 844 dest = (char *)buff; 845 846 while (s > 0) { 847 if (a->read_data_offset == a->read_data_output_offset && 848 a->read_data_remaining == 0) { 849 read_buf = a->read_data_block; 850 a->read_data_is_posix_read = 1; 851 a->read_data_requested = s; 852 r = archive_read_data_block(a, &read_buf, 853 &a->read_data_remaining, &a->read_data_offset); 854 a->read_data_block = read_buf; 855 if (r == ARCHIVE_EOF) 856 return (bytes_read); 857 /* 858 * Error codes are all negative, so the status 859 * return here cannot be confused with a valid 860 * byte count. (ARCHIVE_OK is zero.) 861 */ 862 if (r < ARCHIVE_OK) 863 return (r); 864 } 865 866 if (a->read_data_offset < a->read_data_output_offset) { 867 archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT, 868 "Encountered out-of-order sparse blocks"); 869 return (ARCHIVE_RETRY); 870 } 871 872 /* Compute the amount of zero padding needed. */ 873 if (a->read_data_output_offset + (int64_t)s < 874 a->read_data_offset) { 875 len = s; 876 } else if (a->read_data_output_offset < 877 a->read_data_offset) { 878 len = (size_t)(a->read_data_offset - 879 a->read_data_output_offset); 880 } else 881 len = 0; 882 883 /* Add zeroes. */ 884 memset(dest, 0, len); 885 s -= len; 886 a->read_data_output_offset += len; 887 dest += len; 888 bytes_read += len; 889 890 /* Copy data if there is any space left. */ 891 if (s > 0) { 892 len = a->read_data_remaining; 893 if (len > s) 894 len = s; 895 if (len) { 896 memcpy(dest, a->read_data_block, len); 897 s -= len; 898 a->read_data_block += len; 899 a->read_data_remaining -= len; 900 a->read_data_output_offset += len; 901 a->read_data_offset += len; 902 dest += len; 903 bytes_read += len; 904 } 905 } 906 } 907 a->read_data_is_posix_read = 0; 908 a->read_data_requested = 0; 909 return (bytes_read); 910 } 911 912 /* 913 * Reset the read_data_* variables, used for starting a new entry. 914 */ 915 void __archive_reset_read_data(struct archive * a) 916 { 917 a->read_data_output_offset = 0; 918 a->read_data_remaining = 0; 919 a->read_data_is_posix_read = 0; 920 a->read_data_requested = 0; 921 922 /* extra resets, from rar.c */ 923 a->read_data_block = NULL; 924 a->read_data_offset = 0; 925 } 926 927 /* 928 * Skip over all remaining data in this entry. 929 */ 930 int 931 archive_read_data_skip(struct archive *_a) 932 { 933 struct archive_read *a = (struct archive_read *)_a; 934 int r; 935 const void *buff; 936 size_t size; 937 int64_t offset; 938 939 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA, 940 "archive_read_data_skip"); 941 942 if (a->format->read_data_skip != NULL) 943 r = (a->format->read_data_skip)(a); 944 else { 945 while ((r = archive_read_data_block(&a->archive, 946 &buff, &size, &offset)) 947 == ARCHIVE_OK) 948 ; 949 } 950 951 if (r == ARCHIVE_EOF) 952 r = ARCHIVE_OK; 953 954 a->archive.state = ARCHIVE_STATE_HEADER; 955 return (r); 956 } 957 958 la_int64_t 959 archive_seek_data(struct archive *_a, int64_t offset, int whence) 960 { 961 struct archive_read *a = (struct archive_read *)_a; 962 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA, 963 "archive_seek_data_block"); 964 965 if (a->format->seek_data == NULL) { 966 archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER, 967 "Internal error: " 968 "No format_seek_data_block function registered"); 969 return (ARCHIVE_FATAL); 970 } 971 972 return (a->format->seek_data)(a, offset, whence); 973 } 974 975 /* 976 * Read the next block of entry data from the archive. 977 * This is a zero-copy interface; the client receives a pointer, 978 * size, and file offset of the next available block of data. 979 * 980 * Returns ARCHIVE_OK if the operation is successful, ARCHIVE_EOF if 981 * the end of entry is encountered. 982 */ 983 static int 984 _archive_read_data_block(struct archive *_a, 985 const void **buff, size_t *size, int64_t *offset) 986 { 987 struct archive_read *a = (struct archive_read *)_a; 988 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA, 989 "archive_read_data_block"); 990 991 if (a->format->read_data == NULL) { 992 archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER, 993 "Internal error: " 994 "No format->read_data function registered"); 995 return (ARCHIVE_FATAL); 996 } 997 998 return (a->format->read_data)(a, buff, size, offset); 999 } 1000 1001 static int 1002 close_filters(struct archive_read *a) 1003 { 1004 struct archive_read_filter *f = a->filter; 1005 int r = ARCHIVE_OK; 1006 /* Close each filter in the pipeline. */ 1007 while (f != NULL) { 1008 struct archive_read_filter *t = f->upstream; 1009 if (!f->closed && f->close != NULL) { 1010 int r1 = (f->close)(f); 1011 f->closed = 1; 1012 if (r1 < r) 1013 r = r1; 1014 } 1015 free(f->buffer); 1016 f->buffer = NULL; 1017 f = t; 1018 } 1019 return r; 1020 } 1021 1022 void 1023 __archive_read_free_filters(struct archive_read *a) 1024 { 1025 /* Make sure filters are closed and their buffers are freed */ 1026 close_filters(a); 1027 1028 while (a->filter != NULL) { 1029 struct archive_read_filter *t = a->filter->upstream; 1030 free(a->filter); 1031 a->filter = t; 1032 } 1033 } 1034 1035 /* 1036 * return the count of # of filters in use 1037 */ 1038 static int 1039 _archive_filter_count(struct archive *_a) 1040 { 1041 struct archive_read *a = (struct archive_read *)_a; 1042 struct archive_read_filter *p = a->filter; 1043 int count = 0; 1044 while(p) { 1045 count++; 1046 p = p->upstream; 1047 } 1048 return count; 1049 } 1050 1051 /* 1052 * Close the file and all I/O. 1053 */ 1054 static int 1055 _archive_read_close(struct archive *_a) 1056 { 1057 struct archive_read *a = (struct archive_read *)_a; 1058 int r = ARCHIVE_OK, r1 = ARCHIVE_OK; 1059 1060 archive_check_magic(&a->archive, ARCHIVE_READ_MAGIC, 1061 ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_close"); 1062 if (a->archive.state == ARCHIVE_STATE_CLOSED) 1063 return (ARCHIVE_OK); 1064 archive_clear_error(&a->archive); 1065 a->archive.state = ARCHIVE_STATE_CLOSED; 1066 1067 /* TODO: Clean up the formatters. */ 1068 1069 /* Release the filter objects. */ 1070 r1 = close_filters(a); 1071 if (r1 < r) 1072 r = r1; 1073 1074 return (r); 1075 } 1076 1077 /* 1078 * Release memory and other resources. 1079 */ 1080 static int 1081 _archive_read_free(struct archive *_a) 1082 { 1083 struct archive_read *a = (struct archive_read *)_a; 1084 struct archive_read_passphrase *p; 1085 int i, n; 1086 int slots; 1087 int r = ARCHIVE_OK; 1088 1089 if (_a == NULL) 1090 return (ARCHIVE_OK); 1091 archive_check_magic(_a, ARCHIVE_READ_MAGIC, 1092 ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_free"); 1093 if (a->archive.state != ARCHIVE_STATE_CLOSED 1094 && a->archive.state != ARCHIVE_STATE_FATAL) 1095 r = archive_read_close(&a->archive); 1096 1097 /* Call cleanup functions registered by optional components. */ 1098 if (a->cleanup_archive_extract != NULL) 1099 r = (a->cleanup_archive_extract)(a); 1100 1101 /* Cleanup format-specific data. */ 1102 slots = sizeof(a->formats) / sizeof(a->formats[0]); 1103 for (i = 0; i < slots; i++) { 1104 a->format = &(a->formats[i]); 1105 if (a->formats[i].cleanup) 1106 (a->formats[i].cleanup)(a); 1107 } 1108 1109 /* Free the filters */ 1110 __archive_read_free_filters(a); 1111 1112 /* Release the bidder objects. */ 1113 n = sizeof(a->bidders)/sizeof(a->bidders[0]); 1114 for (i = 0; i < n; i++) { 1115 if (a->bidders[i].free != NULL) { 1116 int r1 = (a->bidders[i].free)(&a->bidders[i]); 1117 if (r1 < r) 1118 r = r1; 1119 } 1120 } 1121 1122 /* Release passphrase list. */ 1123 p = a->passphrases.first; 1124 while (p != NULL) { 1125 struct archive_read_passphrase *np = p->next; 1126 1127 /* A passphrase should be cleaned. */ 1128 memset(p->passphrase, 0, strlen(p->passphrase)); 1129 free(p->passphrase); 1130 free(p); 1131 p = np; 1132 } 1133 1134 archive_string_free(&a->archive.error_string); 1135 archive_entry_free(a->entry); 1136 a->archive.magic = 0; 1137 __archive_clean(&a->archive); 1138 free(a->client.dataset); 1139 free(a); 1140 return (r); 1141 } 1142 1143 static struct archive_read_filter * 1144 get_filter(struct archive *_a, int n) 1145 { 1146 struct archive_read *a = (struct archive_read *)_a; 1147 struct archive_read_filter *f = a->filter; 1148 /* We use n == -1 for 'the last filter', which is always the 1149 * client proxy. */ 1150 if (n == -1 && f != NULL) { 1151 struct archive_read_filter *last = f; 1152 f = f->upstream; 1153 while (f != NULL) { 1154 last = f; 1155 f = f->upstream; 1156 } 1157 return (last); 1158 } 1159 if (n < 0) 1160 return NULL; 1161 while (n > 0 && f != NULL) { 1162 f = f->upstream; 1163 --n; 1164 } 1165 return (f); 1166 } 1167 1168 static int 1169 _archive_filter_code(struct archive *_a, int n) 1170 { 1171 struct archive_read_filter *f = get_filter(_a, n); 1172 return f == NULL ? -1 : f->code; 1173 } 1174 1175 static const char * 1176 _archive_filter_name(struct archive *_a, int n) 1177 { 1178 struct archive_read_filter *f = get_filter(_a, n); 1179 return f != NULL ? f->name : NULL; 1180 } 1181 1182 static int64_t 1183 _archive_filter_bytes(struct archive *_a, int n) 1184 { 1185 struct archive_read_filter *f = get_filter(_a, n); 1186 return f == NULL ? -1 : f->position; 1187 } 1188 1189 /* 1190 * Used internally by read format handlers to register their bid and 1191 * initialization functions. 1192 */ 1193 int 1194 __archive_read_register_format(struct archive_read *a, 1195 void *format_data, 1196 const char *name, 1197 int (*bid)(struct archive_read *, int), 1198 int (*options)(struct archive_read *, const char *, const char *), 1199 int (*read_header)(struct archive_read *, struct archive_entry *), 1200 int (*read_data)(struct archive_read *, const void **, size_t *, int64_t *), 1201 int (*read_data_skip)(struct archive_read *), 1202 int64_t (*seek_data)(struct archive_read *, int64_t, int), 1203 int (*cleanup)(struct archive_read *), 1204 int (*format_capabilities)(struct archive_read *), 1205 int (*has_encrypted_entries)(struct archive_read *)) 1206 { 1207 int i, number_slots; 1208 1209 archive_check_magic(&a->archive, 1210 ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW, 1211 "__archive_read_register_format"); 1212 1213 number_slots = sizeof(a->formats) / sizeof(a->formats[0]); 1214 1215 for (i = 0; i < number_slots; i++) { 1216 if (a->formats[i].bid == bid) 1217 return (ARCHIVE_WARN); /* We've already installed */ 1218 if (a->formats[i].bid == NULL) { 1219 a->formats[i].bid = bid; 1220 a->formats[i].options = options; 1221 a->formats[i].read_header = read_header; 1222 a->formats[i].read_data = read_data; 1223 a->formats[i].read_data_skip = read_data_skip; 1224 a->formats[i].seek_data = seek_data; 1225 a->formats[i].cleanup = cleanup; 1226 a->formats[i].data = format_data; 1227 a->formats[i].name = name; 1228 a->formats[i].format_capabilties = format_capabilities; 1229 a->formats[i].has_encrypted_entries = has_encrypted_entries; 1230 return (ARCHIVE_OK); 1231 } 1232 } 1233 1234 archive_set_error(&a->archive, ENOMEM, 1235 "Not enough slots for format registration"); 1236 return (ARCHIVE_FATAL); 1237 } 1238 1239 /* 1240 * Used internally by decompression routines to register their bid and 1241 * initialization functions. 1242 */ 1243 int 1244 __archive_read_get_bidder(struct archive_read *a, 1245 struct archive_read_filter_bidder **bidder) 1246 { 1247 int i, number_slots; 1248 1249 number_slots = sizeof(a->bidders) / sizeof(a->bidders[0]); 1250 1251 for (i = 0; i < number_slots; i++) { 1252 if (a->bidders[i].bid == NULL) { 1253 memset(a->bidders + i, 0, sizeof(a->bidders[0])); 1254 *bidder = (a->bidders + i); 1255 return (ARCHIVE_OK); 1256 } 1257 } 1258 1259 archive_set_error(&a->archive, ENOMEM, 1260 "Not enough slots for filter registration"); 1261 return (ARCHIVE_FATAL); 1262 } 1263 1264 /* 1265 * The next section implements the peek/consume internal I/O 1266 * system used by archive readers. This system allows simple 1267 * read-ahead for consumers while preserving zero-copy operation 1268 * most of the time. 1269 * 1270 * The two key operations: 1271 * * The read-ahead function returns a pointer to a block of data 1272 * that satisfies a minimum request. 1273 * * The consume function advances the file pointer. 1274 * 1275 * In the ideal case, filters generate blocks of data 1276 * and __archive_read_ahead() just returns pointers directly into 1277 * those blocks. Then __archive_read_consume() just bumps those 1278 * pointers. Only if your request would span blocks does the I/O 1279 * layer use a copy buffer to provide you with a contiguous block of 1280 * data. 1281 * 1282 * A couple of useful idioms: 1283 * * "I just want some data." Ask for 1 byte and pay attention to 1284 * the "number of bytes available" from __archive_read_ahead(). 1285 * Consume whatever you actually use. 1286 * * "I want to output a large block of data." As above, ask for 1 byte, 1287 * emit all that's available (up to whatever limit you have), consume 1288 * it all, then repeat until you're done. This effectively means that 1289 * you're passing along the blocks that came from your provider. 1290 * * "I want to peek ahead by a large amount." Ask for 4k or so, then 1291 * double and repeat until you get an error or have enough. Note 1292 * that the I/O layer will likely end up expanding its copy buffer 1293 * to fit your request, so use this technique cautiously. This 1294 * technique is used, for example, by some of the format tasting 1295 * code that has uncertain look-ahead needs. 1296 */ 1297 1298 /* 1299 * Looks ahead in the input stream: 1300 * * If 'avail' pointer is provided, that returns number of bytes available 1301 * in the current buffer, which may be much larger than requested. 1302 * * If end-of-file, *avail gets set to zero. 1303 * * If error, *avail gets error code. 1304 * * If request can be met, returns pointer to data. 1305 * * If minimum request cannot be met, returns NULL. 1306 * 1307 * Note: If you just want "some data", ask for 1 byte and pay attention 1308 * to *avail, which will have the actual amount available. If you 1309 * know exactly how many bytes you need, just ask for that and treat 1310 * a NULL return as an error. 1311 * 1312 * Important: This does NOT move the file pointer. See 1313 * __archive_read_consume() below. 1314 */ 1315 const void * 1316 __archive_read_ahead(struct archive_read *a, size_t min, ssize_t *avail) 1317 { 1318 return (__archive_read_filter_ahead(a->filter, min, avail)); 1319 } 1320 1321 const void * 1322 __archive_read_filter_ahead(struct archive_read_filter *filter, 1323 size_t min, ssize_t *avail) 1324 { 1325 ssize_t bytes_read; 1326 size_t tocopy; 1327 1328 if (filter->fatal) { 1329 if (avail) 1330 *avail = ARCHIVE_FATAL; 1331 return (NULL); 1332 } 1333 1334 /* 1335 * Keep pulling more data until we can satisfy the request. 1336 */ 1337 for (;;) { 1338 1339 /* 1340 * If we can satisfy from the copy buffer (and the 1341 * copy buffer isn't empty), we're done. In particular, 1342 * note that min == 0 is a perfectly well-defined 1343 * request. 1344 */ 1345 if (filter->avail >= min && filter->avail > 0) { 1346 if (avail != NULL) 1347 *avail = filter->avail; 1348 return (filter->next); 1349 } 1350 1351 /* 1352 * We can satisfy directly from client buffer if everything 1353 * currently in the copy buffer is still in the client buffer. 1354 */ 1355 if (filter->client_total >= filter->client_avail + filter->avail 1356 && filter->client_avail + filter->avail >= min) { 1357 /* "Roll back" to client buffer. */ 1358 filter->client_avail += filter->avail; 1359 filter->client_next -= filter->avail; 1360 /* Copy buffer is now empty. */ 1361 filter->avail = 0; 1362 filter->next = filter->buffer; 1363 /* Return data from client buffer. */ 1364 if (avail != NULL) 1365 *avail = filter->client_avail; 1366 return (filter->client_next); 1367 } 1368 1369 /* Move data forward in copy buffer if necessary. */ 1370 if (filter->next > filter->buffer && 1371 filter->next + min > filter->buffer + filter->buffer_size) { 1372 if (filter->avail > 0) 1373 memmove(filter->buffer, filter->next, 1374 filter->avail); 1375 filter->next = filter->buffer; 1376 } 1377 1378 /* If we've used up the client data, get more. */ 1379 if (filter->client_avail <= 0) { 1380 if (filter->end_of_file) { 1381 if (avail != NULL) 1382 *avail = 0; 1383 return (NULL); 1384 } 1385 bytes_read = (filter->read)(filter, 1386 &filter->client_buff); 1387 if (bytes_read < 0) { /* Read error. */ 1388 filter->client_total = filter->client_avail = 0; 1389 filter->client_next = 1390 filter->client_buff = NULL; 1391 filter->fatal = 1; 1392 if (avail != NULL) 1393 *avail = ARCHIVE_FATAL; 1394 return (NULL); 1395 } 1396 if (bytes_read == 0) { 1397 /* Check for another client object first */ 1398 if (filter->archive->client.cursor != 1399 filter->archive->client.nodes - 1) { 1400 if (client_switch_proxy(filter, 1401 filter->archive->client.cursor + 1) 1402 == ARCHIVE_OK) 1403 continue; 1404 } 1405 /* Premature end-of-file. */ 1406 filter->client_total = filter->client_avail = 0; 1407 filter->client_next = 1408 filter->client_buff = NULL; 1409 filter->end_of_file = 1; 1410 /* Return whatever we do have. */ 1411 if (avail != NULL) 1412 *avail = filter->avail; 1413 return (NULL); 1414 } 1415 filter->client_total = bytes_read; 1416 filter->client_avail = filter->client_total; 1417 filter->client_next = filter->client_buff; 1418 } else { 1419 /* 1420 * We can't satisfy the request from the copy 1421 * buffer or the existing client data, so we 1422 * need to copy more client data over to the 1423 * copy buffer. 1424 */ 1425 1426 /* Ensure the buffer is big enough. */ 1427 if (min > filter->buffer_size) { 1428 size_t s, t; 1429 char *p; 1430 1431 /* Double the buffer; watch for overflow. */ 1432 s = t = filter->buffer_size; 1433 if (s == 0) 1434 s = min; 1435 while (s < min) { 1436 t *= 2; 1437 if (t <= s) { /* Integer overflow! */ 1438 archive_set_error( 1439 &filter->archive->archive, 1440 ENOMEM, 1441 "Unable to allocate copy" 1442 " buffer"); 1443 filter->fatal = 1; 1444 if (avail != NULL) 1445 *avail = ARCHIVE_FATAL; 1446 return (NULL); 1447 } 1448 s = t; 1449 } 1450 /* Now s >= min, so allocate a new buffer. */ 1451 p = (char *)malloc(s); 1452 if (p == NULL) { 1453 archive_set_error( 1454 &filter->archive->archive, 1455 ENOMEM, 1456 "Unable to allocate copy buffer"); 1457 filter->fatal = 1; 1458 if (avail != NULL) 1459 *avail = ARCHIVE_FATAL; 1460 return (NULL); 1461 } 1462 /* Move data into newly-enlarged buffer. */ 1463 if (filter->avail > 0) 1464 memmove(p, filter->next, filter->avail); 1465 free(filter->buffer); 1466 filter->next = filter->buffer = p; 1467 filter->buffer_size = s; 1468 } 1469 1470 /* We can add client data to copy buffer. */ 1471 /* First estimate: copy to fill rest of buffer. */ 1472 tocopy = (filter->buffer + filter->buffer_size) 1473 - (filter->next + filter->avail); 1474 /* Don't waste time buffering more than we need to. */ 1475 if (tocopy + filter->avail > min) 1476 tocopy = min - filter->avail; 1477 /* Don't copy more than is available. */ 1478 if (tocopy > filter->client_avail) 1479 tocopy = filter->client_avail; 1480 1481 memcpy(filter->next + filter->avail, 1482 filter->client_next, tocopy); 1483 /* Remove this data from client buffer. */ 1484 filter->client_next += tocopy; 1485 filter->client_avail -= tocopy; 1486 /* add it to copy buffer. */ 1487 filter->avail += tocopy; 1488 } 1489 } 1490 } 1491 1492 /* 1493 * Move the file pointer forward. 1494 */ 1495 int64_t 1496 __archive_read_consume(struct archive_read *a, int64_t request) 1497 { 1498 return (__archive_read_filter_consume(a->filter, request)); 1499 } 1500 1501 int64_t 1502 __archive_read_filter_consume(struct archive_read_filter * filter, 1503 int64_t request) 1504 { 1505 int64_t skipped; 1506 1507 if (request < 0) 1508 return ARCHIVE_FATAL; 1509 if (request == 0) 1510 return 0; 1511 1512 skipped = advance_file_pointer(filter, request); 1513 if (skipped == request) 1514 return (skipped); 1515 /* We hit EOF before we satisfied the skip request. */ 1516 if (skipped < 0) /* Map error code to 0 for error message below. */ 1517 skipped = 0; 1518 archive_set_error(&filter->archive->archive, 1519 ARCHIVE_ERRNO_MISC, 1520 "Truncated input file (needed %jd bytes, only %jd available)", 1521 (intmax_t)request, (intmax_t)skipped); 1522 return (ARCHIVE_FATAL); 1523 } 1524 1525 /* 1526 * Advance the file pointer by the amount requested. 1527 * Returns the amount actually advanced, which may be less than the 1528 * request if EOF is encountered first. 1529 * Returns a negative value if there's an I/O error. 1530 */ 1531 static int64_t 1532 advance_file_pointer(struct archive_read_filter *filter, int64_t request) 1533 { 1534 int64_t bytes_skipped, total_bytes_skipped = 0; 1535 ssize_t bytes_read; 1536 size_t min; 1537 1538 if (filter->fatal) 1539 return (-1); 1540 1541 /* Use up the copy buffer first. */ 1542 if (filter->avail > 0) { 1543 min = (size_t)minimum(request, (int64_t)filter->avail); 1544 filter->next += min; 1545 filter->avail -= min; 1546 request -= min; 1547 filter->position += min; 1548 total_bytes_skipped += min; 1549 } 1550 1551 /* Then use up the client buffer. */ 1552 if (filter->client_avail > 0) { 1553 min = (size_t)minimum(request, (int64_t)filter->client_avail); 1554 filter->client_next += min; 1555 filter->client_avail -= min; 1556 request -= min; 1557 filter->position += min; 1558 total_bytes_skipped += min; 1559 } 1560 if (request == 0) 1561 return (total_bytes_skipped); 1562 1563 /* If there's an optimized skip function, use it. */ 1564 if (filter->skip != NULL) { 1565 bytes_skipped = (filter->skip)(filter, request); 1566 if (bytes_skipped < 0) { /* error */ 1567 filter->fatal = 1; 1568 return (bytes_skipped); 1569 } 1570 filter->position += bytes_skipped; 1571 total_bytes_skipped += bytes_skipped; 1572 request -= bytes_skipped; 1573 if (request == 0) 1574 return (total_bytes_skipped); 1575 } 1576 1577 /* Use ordinary reads as necessary to complete the request. */ 1578 for (;;) { 1579 bytes_read = (filter->read)(filter, &filter->client_buff); 1580 if (bytes_read < 0) { 1581 filter->client_buff = NULL; 1582 filter->fatal = 1; 1583 return (bytes_read); 1584 } 1585 1586 if (bytes_read == 0) { 1587 if (filter->archive->client.cursor != 1588 filter->archive->client.nodes - 1) { 1589 if (client_switch_proxy(filter, 1590 filter->archive->client.cursor + 1) 1591 == ARCHIVE_OK) 1592 continue; 1593 } 1594 filter->client_buff = NULL; 1595 filter->end_of_file = 1; 1596 return (total_bytes_skipped); 1597 } 1598 1599 if (bytes_read >= request) { 1600 filter->client_next = 1601 ((const char *)filter->client_buff) + request; 1602 filter->client_avail = (size_t)(bytes_read - request); 1603 filter->client_total = bytes_read; 1604 total_bytes_skipped += request; 1605 filter->position += request; 1606 return (total_bytes_skipped); 1607 } 1608 1609 filter->position += bytes_read; 1610 total_bytes_skipped += bytes_read; 1611 request -= bytes_read; 1612 } 1613 } 1614 1615 /** 1616 * Returns ARCHIVE_FAILED if seeking isn't supported. 1617 */ 1618 int64_t 1619 __archive_read_seek(struct archive_read *a, int64_t offset, int whence) 1620 { 1621 return __archive_read_filter_seek(a->filter, offset, whence); 1622 } 1623 1624 int64_t 1625 __archive_read_filter_seek(struct archive_read_filter *filter, int64_t offset, 1626 int whence) 1627 { 1628 struct archive_read_client *client; 1629 int64_t r; 1630 unsigned int cursor; 1631 1632 if (filter->closed || filter->fatal) 1633 return (ARCHIVE_FATAL); 1634 if (filter->seek == NULL) 1635 return (ARCHIVE_FAILED); 1636 1637 client = &(filter->archive->client); 1638 switch (whence) { 1639 case SEEK_CUR: 1640 /* Adjust the offset and use SEEK_SET instead */ 1641 offset += filter->position; 1642 __LA_FALLTHROUGH; 1643 case SEEK_SET: 1644 cursor = 0; 1645 while (1) 1646 { 1647 if (client->dataset[cursor].begin_position < 0 || 1648 client->dataset[cursor].total_size < 0 || 1649 client->dataset[cursor].begin_position + 1650 client->dataset[cursor].total_size - 1 > offset || 1651 cursor + 1 >= client->nodes) 1652 break; 1653 r = client->dataset[cursor].begin_position + 1654 client->dataset[cursor].total_size; 1655 client->dataset[++cursor].begin_position = r; 1656 } 1657 while (1) { 1658 r = client_switch_proxy(filter, cursor); 1659 if (r != ARCHIVE_OK) 1660 return r; 1661 if ((r = client_seek_proxy(filter, 0, SEEK_END)) < 0) 1662 return r; 1663 client->dataset[cursor].total_size = r; 1664 if (client->dataset[cursor].begin_position + 1665 client->dataset[cursor].total_size - 1 > offset || 1666 cursor + 1 >= client->nodes) 1667 break; 1668 r = client->dataset[cursor].begin_position + 1669 client->dataset[cursor].total_size; 1670 client->dataset[++cursor].begin_position = r; 1671 } 1672 offset -= client->dataset[cursor].begin_position; 1673 if (offset < 0 1674 || offset > client->dataset[cursor].total_size) 1675 return ARCHIVE_FATAL; 1676 if ((r = client_seek_proxy(filter, offset, SEEK_SET)) < 0) 1677 return r; 1678 break; 1679 1680 case SEEK_END: 1681 cursor = 0; 1682 while (1) { 1683 if (client->dataset[cursor].begin_position < 0 || 1684 client->dataset[cursor].total_size < 0 || 1685 cursor + 1 >= client->nodes) 1686 break; 1687 r = client->dataset[cursor].begin_position + 1688 client->dataset[cursor].total_size; 1689 client->dataset[++cursor].begin_position = r; 1690 } 1691 while (1) { 1692 r = client_switch_proxy(filter, cursor); 1693 if (r != ARCHIVE_OK) 1694 return r; 1695 if ((r = client_seek_proxy(filter, 0, SEEK_END)) < 0) 1696 return r; 1697 client->dataset[cursor].total_size = r; 1698 r = client->dataset[cursor].begin_position + 1699 client->dataset[cursor].total_size; 1700 if (cursor + 1 >= client->nodes) 1701 break; 1702 client->dataset[++cursor].begin_position = r; 1703 } 1704 while (1) { 1705 if (r + offset >= 1706 client->dataset[cursor].begin_position) 1707 break; 1708 offset += client->dataset[cursor].total_size; 1709 if (cursor == 0) 1710 break; 1711 cursor--; 1712 r = client->dataset[cursor].begin_position + 1713 client->dataset[cursor].total_size; 1714 } 1715 offset = (r + offset) - client->dataset[cursor].begin_position; 1716 if ((r = client_switch_proxy(filter, cursor)) != ARCHIVE_OK) 1717 return r; 1718 r = client_seek_proxy(filter, offset, SEEK_SET); 1719 if (r < ARCHIVE_OK) 1720 return r; 1721 break; 1722 1723 default: 1724 return (ARCHIVE_FATAL); 1725 } 1726 r += client->dataset[cursor].begin_position; 1727 1728 if (r >= 0) { 1729 /* 1730 * Ouch. Clearing the buffer like this hurts, especially 1731 * at bid time. A lot of our efficiency at bid time comes 1732 * from having bidders reuse the data we've already read. 1733 * 1734 * TODO: If the seek request is in data we already 1735 * have, then don't call the seek callback. 1736 * 1737 * TODO: Zip seeks to end-of-file at bid time. If 1738 * other formats also start doing this, we may need to 1739 * find a way for clients to fudge the seek offset to 1740 * a block boundary. 1741 * 1742 * Hmmm... If whence was SEEK_END, we know the file 1743 * size is (r - offset). Can we use that to simplify 1744 * the TODO items above? 1745 */ 1746 filter->avail = filter->client_avail = 0; 1747 filter->next = filter->buffer; 1748 filter->position = r; 1749 filter->end_of_file = 0; 1750 } 1751 return r; 1752 } 1753