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
36 #ifdef HAVE_ERRNO_H
37 #include <errno.h>
38 #endif
39 #ifdef HAVE_LIMITS_H
40 #include <limits.h>
41 #endif
42 #include <stdio.h>
43 #ifdef HAVE_STDLIB_H
44 #include <stdlib.h>
45 #endif
46 #ifdef HAVE_STRING_H
47 #include <string.h>
48 #endif
49 #ifdef HAVE_UNISTD_H
50 #include <unistd.h>
51 #endif
52
53 #include "archive.h"
54 #include "archive_entry.h"
55 #include "archive_integer.h"
56 #include "archive_private.h"
57 #include "archive_read_private.h"
58
59 #define minimum(a, b) (a < b ? a : b)
60
61 static int choose_filters(struct archive_read *);
62 static int choose_format(struct archive_read *);
63 static int close_filters(struct archive_read *);
64 static int64_t _archive_filter_bytes(struct archive *, int);
65 static int _archive_filter_code(struct archive *, int);
66 static const char *_archive_filter_name(struct archive *, int);
67 static int _archive_filter_count(struct archive *);
68 static int _archive_read_close(struct archive *);
69 static int _archive_read_data_block(struct archive *,
70 const void **, size_t *, int64_t *);
71 static int _archive_read_free(struct archive *);
72 static int _archive_read_next_header(struct archive *,
73 struct archive_entry **);
74 static int _archive_read_next_header2(struct archive *,
75 struct archive_entry *);
76 static int64_t advance_file_pointer(struct archive_read_filter *, int64_t);
77
78 static const struct archive_vtable
79 archive_read_vtable = {
80 .archive_filter_bytes = _archive_filter_bytes,
81 .archive_filter_code = _archive_filter_code,
82 .archive_filter_name = _archive_filter_name,
83 .archive_filter_count = _archive_filter_count,
84 .archive_read_data_block = _archive_read_data_block,
85 .archive_read_next_header = _archive_read_next_header,
86 .archive_read_next_header2 = _archive_read_next_header2,
87 .archive_free = _archive_read_free,
88 .archive_close = _archive_read_close,
89 };
90
91 /*
92 * Allocate, initialize and return a struct archive object.
93 */
94 struct archive *
archive_read_new(void)95 archive_read_new(void)
96 {
97 struct archive_read *a;
98
99 a = calloc(1, sizeof(*a));
100 if (a == NULL)
101 return (NULL);
102 a->archive.magic = ARCHIVE_READ_MAGIC;
103
104 a->archive.state = ARCHIVE_STATE_NEW;
105 a->entry = archive_entry_new2(&a->archive);
106 if (a->entry == NULL) {
107 free(a);
108 return (NULL);
109 }
110 a->archive.vtable = &archive_read_vtable;
111 a->entry_bytes_declared = -1;
112
113 a->passphrases.last = &a->passphrases.first;
114
115 return (&a->archive);
116 }
117
118 /*
119 * Record the do-not-extract-to file. This belongs in archive_read_extract.c.
120 */
121 void
archive_read_extract_set_skip_file(struct archive * _a,la_int64_t d,la_int64_t i)122 archive_read_extract_set_skip_file(struct archive *_a, la_int64_t d,
123 la_int64_t i)
124 {
125 struct archive_read *a = (struct archive_read *)_a;
126
127 if (ARCHIVE_OK != __archive_check_magic(_a, ARCHIVE_READ_MAGIC,
128 ARCHIVE_STATE_ANY, "archive_read_extract_set_skip_file"))
129 return;
130 a->skip_file_set = 1;
131 a->skip_file_dev = d;
132 a->skip_file_ino = i;
133 }
134
135 /*
136 * Open the archive
137 */
138 int
archive_read_open(struct archive * a,void * client_data,archive_open_callback * client_opener,archive_read_callback * client_reader,archive_close_callback * client_closer)139 archive_read_open(struct archive *a, void *client_data,
140 archive_open_callback *client_opener, archive_read_callback *client_reader,
141 archive_close_callback *client_closer)
142 {
143 int r;
144
145 /* Old archive_read_open() is just a thin shell around
146 * archive_read_open1. */
147 archive_read_set_open_callback(a, client_opener);
148 archive_read_set_read_callback(a, client_reader);
149 archive_read_set_close_callback(a, client_closer);
150 r = archive_read_set_callback_data(a, client_data);
151 if (r < 0)
152 return (r);
153 return archive_read_open1(a);
154 }
155
156
157 int
archive_read_open2(struct archive * a,void * client_data,archive_open_callback * client_opener,archive_read_callback * client_reader,archive_skip_callback * client_skipper,archive_close_callback * client_closer)158 archive_read_open2(struct archive *a, void *client_data,
159 archive_open_callback *client_opener,
160 archive_read_callback *client_reader,
161 archive_skip_callback *client_skipper,
162 archive_close_callback *client_closer)
163 {
164 int r;
165
166 /* Old archive_read_open2() is just a thin shell around
167 * archive_read_open1. */
168 r = archive_read_set_callback_data(a, client_data);
169 if (r < 0)
170 return (r);
171 archive_read_set_open_callback(a, client_opener);
172 archive_read_set_read_callback(a, client_reader);
173 archive_read_set_skip_callback(a, client_skipper);
174 archive_read_set_close_callback(a, client_closer);
175 return archive_read_open1(a);
176 }
177
178 static ssize_t
client_read_proxy(struct archive_read_filter * f,const void ** buff)179 client_read_proxy(struct archive_read_filter *f, const void **buff)
180 {
181 ssize_t r;
182 r = (f->archive->client.reader)(&f->archive->archive,
183 f->data, buff);
184 return (r);
185 }
186
187 static int64_t
client_skip_proxy(struct archive_read_filter * f,int64_t request)188 client_skip_proxy(struct archive_read_filter *f, int64_t request)
189 {
190 if (request < 0)
191 __archive_errx(1, "Negative skip requested");
192 if (request == 0)
193 return 0;
194
195 if (f->archive->client.skipper != NULL) {
196 int64_t total = 0;
197 for (;;) {
198 int64_t get, ask = request;
199 get = (f->archive->client.skipper)
200 (&f->archive->archive, f->data, ask);
201 total += get;
202 if (get == 0 || get == request)
203 return (total);
204 if (get > request)
205 return ARCHIVE_FATAL;
206 request -= get;
207 }
208 } else if (f->archive->client.seeker != NULL
209 && request > 64 * 1024) {
210 /* If the client provided a seeker but not a skipper,
211 * we can use the seeker to skip forward.
212 *
213 * Note: This isn't always a good idea. The client
214 * skipper is allowed to skip by less than requested
215 * if it needs to maintain block alignment. The
216 * seeker is not allowed to play such games, so using
217 * the seeker here may be a performance loss compared
218 * to just reading and discarding. That's why we
219 * only do this for skips of over 64k.
220 */
221 int64_t before = f->position;
222 int64_t after = (f->archive->client.seeker)
223 (&f->archive->archive, f->data, request, SEEK_CUR);
224 if (after != before + request)
225 return ARCHIVE_FATAL;
226 return after - before;
227 }
228 return 0;
229 }
230
231 static int64_t
client_seek_proxy(struct archive_read_filter * f,int64_t offset,int whence)232 client_seek_proxy(struct archive_read_filter *f, int64_t offset, int whence)
233 {
234 /* DO NOT use the skipper here! If we transparently handled
235 * forward seek here by using the skipper, that will break
236 * other libarchive code that assumes a successful forward
237 * seek means it can also seek backwards.
238 */
239 if (f->archive->client.seeker == NULL) {
240 archive_set_error(&f->archive->archive, ARCHIVE_ERRNO_MISC,
241 "Current client reader does not support seeking a device");
242 return (ARCHIVE_FAILED);
243 }
244 return (f->archive->client.seeker)(&f->archive->archive,
245 f->data, offset, whence);
246 }
247
248 static int
read_client_close_proxy(struct archive_read * a)249 read_client_close_proxy(struct archive_read *a)
250 {
251 int r = ARCHIVE_OK, r2;
252 unsigned int i;
253
254 if (a->client.closer == NULL)
255 return (r);
256 for (i = 0; i < a->client.nodes; i++)
257 {
258 r2 = (a->client.closer)
259 ((struct archive *)a, a->client.dataset[i].data);
260 if (r > r2)
261 r = r2;
262 }
263 return (r);
264 }
265
266 static int
client_close_proxy(struct archive_read_filter * f)267 client_close_proxy(struct archive_read_filter *f)
268 {
269 return read_client_close_proxy(f->archive);
270 }
271
272 static int
client_switch_proxy(struct archive_read_filter * f,unsigned int iindex)273 client_switch_proxy(struct archive_read_filter *f, unsigned int iindex)
274 {
275 struct archive_read *a;
276 int r1 = ARCHIVE_OK, r2 = ARCHIVE_OK;
277 void *data2;
278
279 while (f->upstream != NULL)
280 f = f->upstream;
281 a = f->archive;
282
283 /* Don't do anything if already in the specified data node */
284 if (a->client.cursor == iindex)
285 return (ARCHIVE_OK);
286
287 a->client.cursor = iindex;
288 data2 = a->client.dataset[a->client.cursor].data;
289 if (a->client.switcher != NULL)
290 {
291 r1 = r2 = (a->client.switcher)
292 ((struct archive *)a, f->data, data2);
293 f->data = data2;
294 }
295 else
296 {
297 /* Attempt to call close and open instead */
298 if (a->client.closer != NULL)
299 r1 = (a->client.closer)
300 ((struct archive *)a, f->data);
301 f->data = data2;
302 if (a->client.opener != NULL)
303 r2 = (a->client.opener)
304 ((struct archive *)a, f->data);
305 }
306 return (r1 < r2) ? r1 : r2;
307 }
308
309 int
archive_read_set_open_callback(struct archive * _a,archive_open_callback * client_opener)310 archive_read_set_open_callback(struct archive *_a,
311 archive_open_callback *client_opener)
312 {
313 struct archive_read *a = (struct archive_read *)_a;
314 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
315 "archive_read_set_open_callback");
316 a->client.opener = client_opener;
317 return ARCHIVE_OK;
318 }
319
320 int
archive_read_set_read_callback(struct archive * _a,archive_read_callback * client_reader)321 archive_read_set_read_callback(struct archive *_a,
322 archive_read_callback *client_reader)
323 {
324 struct archive_read *a = (struct archive_read *)_a;
325 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
326 "archive_read_set_read_callback");
327 a->client.reader = client_reader;
328 return ARCHIVE_OK;
329 }
330
331 int
archive_read_set_skip_callback(struct archive * _a,archive_skip_callback * client_skipper)332 archive_read_set_skip_callback(struct archive *_a,
333 archive_skip_callback *client_skipper)
334 {
335 struct archive_read *a = (struct archive_read *)_a;
336 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
337 "archive_read_set_skip_callback");
338 a->client.skipper = client_skipper;
339 return ARCHIVE_OK;
340 }
341
342 int
archive_read_set_seek_callback(struct archive * _a,archive_seek_callback * client_seeker)343 archive_read_set_seek_callback(struct archive *_a,
344 archive_seek_callback *client_seeker)
345 {
346 struct archive_read *a = (struct archive_read *)_a;
347 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
348 "archive_read_set_seek_callback");
349 a->client.seeker = client_seeker;
350 return ARCHIVE_OK;
351 }
352
353 int
archive_read_set_close_callback(struct archive * _a,archive_close_callback * client_closer)354 archive_read_set_close_callback(struct archive *_a,
355 archive_close_callback *client_closer)
356 {
357 struct archive_read *a = (struct archive_read *)_a;
358 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
359 "archive_read_set_close_callback");
360 a->client.closer = client_closer;
361 return ARCHIVE_OK;
362 }
363
364 int
archive_read_set_switch_callback(struct archive * _a,archive_switch_callback * client_switcher)365 archive_read_set_switch_callback(struct archive *_a,
366 archive_switch_callback *client_switcher)
367 {
368 struct archive_read *a = (struct archive_read *)_a;
369 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
370 "archive_read_set_switch_callback");
371 a->client.switcher = client_switcher;
372 return ARCHIVE_OK;
373 }
374
375 int
archive_read_set_callback_data(struct archive * _a,void * client_data)376 archive_read_set_callback_data(struct archive *_a, void *client_data)
377 {
378 return archive_read_set_callback_data2(_a, client_data, 0);
379 }
380
381 int
archive_read_set_callback_data2(struct archive * _a,void * client_data,unsigned int iindex)382 archive_read_set_callback_data2(struct archive *_a, void *client_data,
383 unsigned int iindex)
384 {
385 struct archive_read *a = (struct archive_read *)_a;
386 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
387 "archive_read_set_callback_data2");
388
389 if (a->client.nodes == 0)
390 {
391 a->client.dataset = (struct archive_read_data_node *)
392 calloc(1, sizeof(*a->client.dataset));
393 if (a->client.dataset == NULL)
394 {
395 archive_set_error(&a->archive, ENOMEM,
396 "No memory");
397 return ARCHIVE_FATAL;
398 }
399 a->client.nodes = 1;
400 }
401
402 if (iindex > a->client.nodes - 1)
403 {
404 archive_set_error(&a->archive, EINVAL,
405 "Invalid index specified");
406 return ARCHIVE_FATAL;
407 }
408 a->client.dataset[iindex].data = client_data;
409 a->client.dataset[iindex].begin_position = -1;
410 a->client.dataset[iindex].total_size = -1;
411 return ARCHIVE_OK;
412 }
413
414 int
archive_read_add_callback_data(struct archive * _a,void * client_data,unsigned int iindex)415 archive_read_add_callback_data(struct archive *_a, void *client_data,
416 unsigned int iindex)
417 {
418 struct archive_read *a = (struct archive_read *)_a;
419 void *p;
420 size_t alloc_size;
421 unsigned int i;
422 unsigned int nodes;
423
424 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
425 "archive_read_add_callback_data");
426 if (iindex > a->client.nodes) {
427 archive_set_error(&a->archive, EINVAL,
428 "Invalid index specified");
429 return ARCHIVE_FATAL;
430 }
431
432 if (a->client.nodes == UINT_MAX ||
433 archive_ckd_mul_size(&alloc_size,
434 (size_t)a->client.nodes + 1, sizeof(*a->client.dataset))) {
435 archive_set_error(&a->archive, ENOMEM,
436 "No memory");
437 return ARCHIVE_FATAL;
438 }
439
440 nodes = a->client.nodes + 1;
441 p = realloc(a->client.dataset, alloc_size);
442 if (p == NULL) {
443 archive_set_error(&a->archive, ENOMEM,
444 "No memory");
445 return ARCHIVE_FATAL;
446 }
447
448 a->client.dataset = (struct archive_read_data_node *)p;
449 a->client.nodes = nodes;
450
451 for (i = a->client.nodes - 1; i > iindex; i--) {
452 a->client.dataset[i].data = a->client.dataset[i-1].data;
453 a->client.dataset[i].begin_position = -1;
454 a->client.dataset[i].total_size = -1;
455 }
456 a->client.dataset[iindex].data = client_data;
457 a->client.dataset[iindex].begin_position = -1;
458 a->client.dataset[iindex].total_size = -1;
459 return ARCHIVE_OK;
460 }
461
462 int
archive_read_append_callback_data(struct archive * _a,void * client_data)463 archive_read_append_callback_data(struct archive *_a, void *client_data)
464 {
465 struct archive_read *a = (struct archive_read *)_a;
466 return archive_read_add_callback_data(_a, client_data, a->client.nodes);
467 }
468
469 int
archive_read_prepend_callback_data(struct archive * _a,void * client_data)470 archive_read_prepend_callback_data(struct archive *_a, void *client_data)
471 {
472 return archive_read_add_callback_data(_a, client_data, 0);
473 }
474
475 static const struct archive_read_filter_vtable
476 none_reader_vtable = {
477 .read = client_read_proxy,
478 .close = client_close_proxy,
479 };
480
481 int
archive_read_open1(struct archive * _a)482 archive_read_open1(struct archive *_a)
483 {
484 struct archive_read *a = (struct archive_read *)_a;
485 struct archive_read_filter *f, *tmp;
486 int slot, e = ARCHIVE_OK;
487
488 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
489 "archive_read_open");
490 archive_clear_error(&a->archive);
491
492 if (a->client.reader == NULL) {
493 archive_set_error(&a->archive, EINVAL,
494 "No reader function provided to archive_read_open");
495 a->archive.state = ARCHIVE_STATE_FATAL;
496 return (ARCHIVE_FATAL);
497 }
498
499 /* Open data source. */
500 if (a->client.opener != NULL) {
501 e = (a->client.opener)(&a->archive, a->client.dataset[0].data);
502 if (e != 0) {
503 /* If the open failed, call the closer to clean up. */
504 read_client_close_proxy(a);
505 return (e);
506 }
507 }
508
509 f = calloc(1, sizeof(*f));
510 if (f == NULL)
511 return (ARCHIVE_FATAL);
512 f->bidder = NULL;
513 f->upstream = NULL;
514 f->archive = a;
515 f->data = a->client.dataset[0].data;
516 f->vtable = &none_reader_vtable;
517 f->name = "none";
518 f->code = ARCHIVE_FILTER_NONE;
519 f->can_skip = 1;
520 f->can_seek = 1;
521
522 a->client.dataset[0].begin_position = 0;
523 if (!a->filter || !a->bypass_filter_bidding)
524 {
525 a->filter = f;
526 /* Build out the input pipeline. */
527 e = choose_filters(a);
528 if (e < ARCHIVE_WARN) {
529 a->archive.state = ARCHIVE_STATE_FATAL;
530 return (ARCHIVE_FATAL);
531 }
532 }
533 else
534 {
535 /* Need to add "NONE" type filter at the end of the filter chain */
536 tmp = a->filter;
537 while (tmp->upstream)
538 tmp = tmp->upstream;
539 tmp->upstream = f;
540 }
541
542 if (!a->format)
543 {
544 slot = choose_format(a);
545 if (slot < 0) {
546 close_filters(a);
547 a->archive.state = ARCHIVE_STATE_FATAL;
548 return (ARCHIVE_FATAL);
549 }
550 a->format = &(a->formats[slot]);
551 }
552
553 a->archive.state = ARCHIVE_STATE_HEADER;
554
555 /* Ensure libarchive starts from the first node in a multivolume set */
556 client_switch_proxy(a->filter, 0);
557 return (e);
558 }
559
560 /*
561 * Allow each registered stream transform to bid on whether
562 * it wants to handle this stream. Repeat until we've finished
563 * building the pipeline.
564 */
565
566 /* We won't build a filter pipeline with more stages than this. */
567 #define MAX_NUMBER_FILTERS 25
568
569 static int
choose_filters(struct archive_read * a)570 choose_filters(struct archive_read *a)
571 {
572 int number_bidders, i, bid, best_bid, number_filters;
573 struct archive_read_filter_bidder *bidder, *best_bidder;
574 struct archive_read_filter *f;
575 ssize_t avail;
576 int r;
577
578 for (number_filters = 0; number_filters < MAX_NUMBER_FILTERS; ++number_filters) {
579 number_bidders = sizeof(a->bidders) / sizeof(a->bidders[0]);
580
581 best_bid = 0;
582 best_bidder = NULL;
583
584 bidder = a->bidders;
585 for (i = 0; i < number_bidders; i++, bidder++) {
586 if (bidder->vtable == NULL)
587 continue;
588 bid = (bidder->vtable->bid)(bidder, a->filter);
589 if (bid > best_bid) {
590 best_bid = bid;
591 best_bidder = bidder;
592 }
593 }
594
595 /* If no bidder, we're done. */
596 if (best_bidder == NULL) {
597 /* Verify the filter by asking it for some data. */
598 __archive_read_filter_ahead(a->filter, 1, &avail);
599 if (avail < 0) {
600 __archive_read_free_filters(a);
601 return (ARCHIVE_FATAL);
602 }
603 return (ARCHIVE_OK);
604 }
605
606 f = calloc(1, sizeof(*f));
607 if (f == NULL)
608 return (ARCHIVE_FATAL);
609 f->bidder = best_bidder;
610 f->archive = a;
611 f->upstream = a->filter;
612 a->filter = f;
613 r = (best_bidder->vtable->init)(a->filter);
614 if (r != ARCHIVE_OK) {
615 __archive_read_free_filters(a);
616 return (ARCHIVE_FATAL);
617 }
618 }
619 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
620 "Input requires too many filters for decoding");
621 return (ARCHIVE_FATAL);
622 }
623
624 int
__archive_read_header(struct archive_read * a,struct archive_entry * entry)625 __archive_read_header(struct archive_read *a, struct archive_entry *entry)
626 {
627 if (!a->filter->vtable->read_header)
628 return (ARCHIVE_OK);
629 return a->filter->vtable->read_header(a->filter, entry);
630 }
631
632 /*
633 * Read header of next entry.
634 */
635 static int
_archive_read_next_header2(struct archive * _a,struct archive_entry * entry)636 _archive_read_next_header2(struct archive *_a, struct archive_entry *entry)
637 {
638 struct archive_read *a = (struct archive_read *)_a;
639 int r1 = ARCHIVE_OK, r2;
640
641 archive_check_magic(_a, ARCHIVE_READ_MAGIC,
642 ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA |
643 ARCHIVE_STATE_DATA_RECOVERY,
644 "archive_read_next_header");
645
646 archive_entry_clear(entry);
647 archive_clear_error(&a->archive);
648
649 /*
650 * If client didn't consume entire data, skip any remainder
651 * (This is especially important for GNU incremental directories.)
652 * A header that failed to parse (DATA_RECOVERY) still needs the
653 * same treatment: whatever of its body the format reader left
654 * unconsumed must be skipped before we can read the next header.
655 */
656 if (a->archive.state == ARCHIVE_STATE_DATA ||
657 a->archive.state == ARCHIVE_STATE_DATA_RECOVERY) {
658 r1 = archive_read_data_skip(&a->archive);
659 if (r1 == ARCHIVE_EOF)
660 archive_set_error(&a->archive, EIO,
661 "Premature end-of-file");
662 if (r1 == ARCHIVE_EOF || r1 == ARCHIVE_FATAL) {
663 a->archive.state = ARCHIVE_STATE_FATAL;
664 return (ARCHIVE_FATAL);
665 }
666 }
667
668 /* Record start-of-header offset in uncompressed stream. */
669 a->header_position = a->filter->position;
670
671 ++_a->file_count;
672 r2 = (a->format->read_header)(a, entry);
673
674 /*
675 * EOF and FATAL are persistent at this layer. By
676 * modifying the state, we guarantee that future calls to
677 * read a header or read data will fail.
678 */
679 switch (r2) {
680 case ARCHIVE_EOF:
681 a->archive.state = ARCHIVE_STATE_EOF;
682 --_a->file_count;/* Revert a file counter. */
683 break;
684 case ARCHIVE_OK:
685 a->archive.state = ARCHIVE_STATE_DATA;
686 break;
687 case ARCHIVE_WARN:
688 a->archive.state = ARCHIVE_STATE_DATA;
689 break;
690 case ARCHIVE_RETRY:
691 break;
692 case ARCHIVE_FATAL:
693 a->archive.state = ARCHIVE_STATE_FATAL;
694 break;
695 case ARCHIVE_FAILED:
696 /*
697 * This entry's header could not be parsed, so its metadata
698 * cannot be trusted. ARCHIVE_STATE_DATA_RECOVERY still
699 * permits skipping past it (the format reader is
700 * responsible for ensuring that's actually possible), but
701 * blocks archive_read_data() and friends, which all check
702 * for ARCHIVE_STATE_DATA specifically and would otherwise
703 * return content for an entry we don't actually understand.
704 */
705 a->archive.state = ARCHIVE_STATE_DATA_RECOVERY;
706 break;
707 }
708
709 if (r2 == ARCHIVE_OK || r2 == ARCHIVE_WARN)
710 a->entry_bytes_declared = archive_entry_size_is_set(entry)
711 ? archive_entry_size(entry) : -1;
712 else
713 a->entry_bytes_declared = -1;
714
715 __archive_reset_read_data(&a->archive);
716
717 a->data_start_node = a->client.cursor;
718 /* EOF always wins; otherwise return the worst error. */
719 return (r2 < r1 || r2 == ARCHIVE_EOF) ? r2 : r1;
720 }
721
722 static int
_archive_read_next_header(struct archive * _a,struct archive_entry ** entryp)723 _archive_read_next_header(struct archive *_a, struct archive_entry **entryp)
724 {
725 int ret;
726 struct archive_read *a = (struct archive_read *)_a;
727 *entryp = NULL;
728 ret = _archive_read_next_header2(_a, a->entry);
729 *entryp = a->entry;
730 return ret;
731 }
732
733 /*
734 * Allow each registered format to bid on whether it wants to handle
735 * the next entry. Return index of winning bidder.
736 */
737 static int
choose_format(struct archive_read * a)738 choose_format(struct archive_read *a)
739 {
740 int slots;
741 int i;
742 int bid, best_bid;
743 int best_bid_slot;
744
745 slots = sizeof(a->formats) / sizeof(a->formats[0]);
746 best_bid = -1;
747 best_bid_slot = -1;
748
749 /* Set up a->format for convenience of bidders. */
750 a->format = &(a->formats[0]);
751 for (i = 0; i < slots; i++, a->format++) {
752 if (a->format->bid) {
753 bid = (a->format->bid)(a, best_bid);
754 if (bid == ARCHIVE_FATAL)
755 return (ARCHIVE_FATAL);
756 if (a->filter->position != 0)
757 __archive_read_seek(a, 0, SEEK_SET);
758 if ((bid > best_bid) || (best_bid_slot < 0)) {
759 best_bid = bid;
760 best_bid_slot = i;
761 }
762 }
763 }
764
765 /*
766 * There were no bidders; this is a serious programmer error
767 * and demands a quick and definitive abort.
768 */
769 if (best_bid_slot < 0) {
770 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
771 "No formats registered");
772 return (ARCHIVE_FATAL);
773 }
774
775 /*
776 * There were bidders, but no non-zero bids; this means we
777 * can't support this stream.
778 */
779 if (best_bid < 1) {
780 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
781 "Unrecognized archive format");
782 return (ARCHIVE_FATAL);
783 }
784
785 return (best_bid_slot);
786 }
787
788 /*
789 * Return the file offset (within the uncompressed data stream) where
790 * the last header started.
791 */
792 la_int64_t
archive_read_header_position(struct archive * _a)793 archive_read_header_position(struct archive *_a)
794 {
795 struct archive_read *a = (struct archive_read *)_a;
796 archive_check_magic(_a, ARCHIVE_READ_MAGIC,
797 ARCHIVE_STATE_ANY, "archive_read_header_position");
798 return (a->header_position);
799 }
800
801 /*
802 * Returns 1 if the archive contains at least one encrypted entry.
803 * If the archive format does not support encryption at all,
804 * ARCHIVE_READ_FORMAT_ENCRYPTION_UNSUPPORTED is returned.
805 * If for any other reason (e.g. not enough data read so far)
806 * we cannot say whether there are encrypted entries, then
807 * ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW is returned.
808 * In general, this function will return values below zero when the
809 * reader is uncertain or totally incapable of encryption support.
810 * When this function returns 0 you can be sure that the reader
811 * supports encryption detection but no encrypted entries have
812 * been found yet.
813 *
814 * NOTE: If the metadata/header of an archive is also encrypted, you
815 * cannot rely on the number of encrypted entries. That is why this
816 * function does not return the number of encrypted entries but#
817 * just shows that there are some.
818 */
819 int
archive_read_has_encrypted_entries(struct archive * _a)820 archive_read_has_encrypted_entries(struct archive *_a)
821 {
822 struct archive_read *a = (struct archive_read *)_a;
823 int format_supports_encryption = archive_read_format_capabilities(_a)
824 & (ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_DATA | ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_METADATA);
825
826 if (!_a || !format_supports_encryption) {
827 /* Format in general doesn't support encryption */
828 return ARCHIVE_READ_FORMAT_ENCRYPTION_UNSUPPORTED;
829 }
830
831 /* A reader potentially has read enough data now. */
832 if (a->format && a->format->has_encrypted_entries) {
833 return (a->format->has_encrypted_entries)(a);
834 }
835
836 /* For any other reason we cannot say how many entries are there. */
837 return ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW;
838 }
839
840 /*
841 * Returns a bitmask of capabilities that are supported by the archive format reader.
842 * If the reader has no special capabilities, ARCHIVE_READ_FORMAT_CAPS_NONE is returned.
843 */
844 int
archive_read_format_capabilities(struct archive * _a)845 archive_read_format_capabilities(struct archive *_a)
846 {
847 struct archive_read *a = (struct archive_read *)_a;
848 if (a && a->format && a->format->format_capabilties) {
849 return (a->format->format_capabilties)(a);
850 }
851 return ARCHIVE_READ_FORMAT_CAPS_NONE;
852 }
853
854 /*
855 * Read data from an archive entry, using a read(2)-style interface.
856 * This is a convenience routine that just calls
857 * archive_read_data_block and copies the results into the client
858 * buffer, filling any gaps with zero bytes. Clients using this
859 * API can be completely ignorant of sparse-file issues; sparse files
860 * will simply be padded with nulls.
861 *
862 * DO NOT intermingle calls to this function and archive_read_data_block
863 * to read a single entry body.
864 */
865 la_ssize_t
archive_read_data(struct archive * _a,void * buff,size_t s)866 archive_read_data(struct archive *_a, void *buff, size_t s)
867 {
868 struct archive *a = (struct archive *)_a;
869 char *dest;
870 const void *read_buf;
871 size_t bytes_read;
872 size_t len;
873 int r;
874
875 bytes_read = 0;
876 dest = (char *)buff;
877
878 while (s > 0) {
879 if (a->read_data_offset == a->read_data_output_offset &&
880 a->read_data_remaining == 0) {
881 read_buf = a->read_data_block;
882 a->read_data_is_posix_read = 1;
883 a->read_data_requested = s;
884 r = archive_read_data_block(a, &read_buf,
885 &a->read_data_remaining, &a->read_data_offset);
886 a->read_data_block = read_buf;
887 if (r == ARCHIVE_EOF &&
888 a->read_data_offset == a->read_data_output_offset &&
889 a->read_data_remaining == 0)
890 return (bytes_read);
891 /*
892 * Error codes are all negative, so the status
893 * return here cannot be confused with a valid
894 * byte count. (ARCHIVE_OK is zero.)
895 */
896 if (r < ARCHIVE_OK)
897 return (r);
898 }
899
900 if (a->read_data_offset < a->read_data_output_offset) {
901 archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
902 "Encountered out-of-order sparse blocks");
903 return (ARCHIVE_RETRY);
904 }
905
906 /* Compute the amount of zero padding needed. */
907 if (a->read_data_output_offset + (int64_t)s <
908 a->read_data_offset) {
909 len = s;
910 } else if (a->read_data_output_offset <
911 a->read_data_offset) {
912 len = (size_t)(a->read_data_offset -
913 a->read_data_output_offset);
914 } else
915 len = 0;
916
917 /* Add zeroes. */
918 memset(dest, 0, len);
919 s -= len;
920 a->read_data_output_offset += len;
921 dest += len;
922 bytes_read += len;
923
924 /* Copy data if there is any space left. */
925 if (s > 0) {
926 len = a->read_data_remaining;
927 if (len > s)
928 len = s;
929 if (len) {
930 memcpy(dest, a->read_data_block, len);
931 s -= len;
932 a->read_data_block += len;
933 a->read_data_remaining -= len;
934 a->read_data_output_offset += len;
935 a->read_data_offset += len;
936 dest += len;
937 bytes_read += len;
938 }
939 }
940 }
941 a->read_data_is_posix_read = 0;
942 a->read_data_requested = 0;
943 return (bytes_read);
944 }
945
946 /*
947 * Reset the read_data_* variables, used for starting a new entry.
948 */
__archive_reset_read_data(struct archive * a)949 void __archive_reset_read_data(struct archive * a)
950 {
951 a->read_data_output_offset = 0;
952 a->read_data_remaining = 0;
953 a->read_data_is_posix_read = 0;
954 a->read_data_requested = 0;
955
956 /* extra resets, from rar.c */
957 a->read_data_block = NULL;
958 a->read_data_offset = 0;
959 }
960
961 /*
962 * Skip over all remaining data in this entry.
963 */
964 int
archive_read_data_skip(struct archive * _a)965 archive_read_data_skip(struct archive *_a)
966 {
967 struct archive_read *a = (struct archive_read *)_a;
968 int r;
969 const void *buff;
970 size_t size;
971 int64_t offset;
972
973 archive_check_magic(_a, ARCHIVE_READ_MAGIC,
974 ARCHIVE_STATE_DATA | ARCHIVE_STATE_DATA_RECOVERY,
975 "archive_read_data_skip");
976
977 if (a->format->read_data_skip != NULL)
978 r = (a->format->read_data_skip)(a);
979 else {
980 while ((r = archive_read_data_block(&a->archive,
981 &buff, &size, &offset))
982 == ARCHIVE_OK)
983 ;
984 }
985
986 if (r == ARCHIVE_EOF)
987 r = ARCHIVE_OK;
988
989 if (r == ARCHIVE_FATAL)
990 a->archive.state = ARCHIVE_STATE_FATAL;
991 else
992 a->archive.state = ARCHIVE_STATE_HEADER;
993 return (r);
994 }
995
996 la_int64_t
archive_seek_data(struct archive * _a,int64_t offset,int whence)997 archive_seek_data(struct archive *_a, int64_t offset, int whence)
998 {
999 struct archive_read *a = (struct archive_read *)_a;
1000 la_int64_t r;
1001
1002 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA,
1003 "archive_seek_data_block");
1004
1005 if (a->format->seek_data == NULL) {
1006 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1007 "Cannot seek data with this format");
1008 return (ARCHIVE_FAILED);
1009 }
1010
1011 r = (a->format->seek_data)(a, offset, whence);
1012 if (r == ARCHIVE_FATAL)
1013 a->archive.state = ARCHIVE_STATE_FATAL;
1014 return (r);
1015 }
1016
1017 /*
1018 * Read the next block of entry data from the archive.
1019 * This is a zero-copy interface; the client receives a pointer,
1020 * size, and file offset of the next available block of data.
1021 *
1022 * Returns ARCHIVE_OK if the operation is successful, ARCHIVE_EOF if
1023 * the end of entry is encountered.
1024 */
1025 static int
_archive_read_data_block(struct archive * _a,const void ** buff,size_t * size,int64_t * offset)1026 _archive_read_data_block(struct archive *_a,
1027 const void **buff, size_t *size, int64_t *offset)
1028 {
1029 struct archive_read *a = (struct archive_read *)_a;
1030 int r;
1031
1032 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA,
1033 "archive_read_data_block");
1034
1035 if (a->format->read_data == NULL) {
1036 archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
1037 "Internal error: "
1038 "No format->read_data function registered");
1039 a->archive.state = ARCHIVE_STATE_FATAL;
1040 return (ARCHIVE_FATAL);
1041 }
1042
1043 r = (a->format->read_data)(a, buff, size, offset);
1044 if (r == ARCHIVE_FATAL)
1045 a->archive.state = ARCHIVE_STATE_FATAL;
1046 return (r);
1047 }
1048
1049 static int
close_filters(struct archive_read * a)1050 close_filters(struct archive_read *a)
1051 {
1052 struct archive_read_filter *f = a->filter;
1053 int r = ARCHIVE_OK;
1054 /* Close each filter in the pipeline. */
1055 while (f != NULL) {
1056 struct archive_read_filter *t = f->upstream;
1057 if (!f->closed && f->vtable != NULL) {
1058 int r1 = (f->vtable->close)(f);
1059 f->closed = 1;
1060 if (r1 < r)
1061 r = r1;
1062 }
1063 free(f->buffer);
1064 f->buffer = NULL;
1065 f = t;
1066 }
1067 return r;
1068 }
1069
1070 void
__archive_read_free_filters(struct archive_read * a)1071 __archive_read_free_filters(struct archive_read *a)
1072 {
1073 /* Make sure filters are closed and their buffers are freed */
1074 close_filters(a);
1075
1076 while (a->filter != NULL) {
1077 struct archive_read_filter *t = a->filter->upstream;
1078 free(a->filter);
1079 a->filter = t;
1080 }
1081 }
1082
1083 /*
1084 * return the count of # of filters in use
1085 */
1086 static int
_archive_filter_count(struct archive * _a)1087 _archive_filter_count(struct archive *_a)
1088 {
1089 struct archive_read *a = (struct archive_read *)_a;
1090 struct archive_read_filter *p = a->filter;
1091 int count = 0;
1092 while(p) {
1093 count++;
1094 p = p->upstream;
1095 }
1096 return count;
1097 }
1098
1099 /*
1100 * Close the file and all I/O.
1101 */
1102 static int
_archive_read_close(struct archive * _a)1103 _archive_read_close(struct archive *_a)
1104 {
1105 struct archive_read *a = (struct archive_read *)_a;
1106 int r = ARCHIVE_OK, r1 = ARCHIVE_OK;
1107
1108 archive_check_magic(&a->archive, ARCHIVE_READ_MAGIC,
1109 ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_close");
1110 if (a->archive.state == ARCHIVE_STATE_CLOSED)
1111 return (ARCHIVE_OK);
1112 archive_clear_error(&a->archive);
1113 a->archive.state = ARCHIVE_STATE_CLOSED;
1114
1115 /* TODO: Clean up the formatters. */
1116
1117 /* Release the filter objects. */
1118 r1 = close_filters(a);
1119 if (r1 < r)
1120 r = r1;
1121
1122 return (r);
1123 }
1124
1125 /*
1126 * Release memory and other resources.
1127 */
1128 static int
_archive_read_free(struct archive * _a)1129 _archive_read_free(struct archive *_a)
1130 {
1131 struct archive_read *a = (struct archive_read *)_a;
1132 struct archive_read_passphrase *p;
1133 int i, n;
1134 int slots;
1135 int r = ARCHIVE_OK;
1136
1137 if (_a == NULL)
1138 return (ARCHIVE_OK);
1139 archive_check_magic(_a, ARCHIVE_READ_MAGIC,
1140 ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_free");
1141 if (a->archive.state != ARCHIVE_STATE_CLOSED
1142 && a->archive.state != ARCHIVE_STATE_FATAL)
1143 r = archive_read_close(&a->archive);
1144
1145 /* Call cleanup functions registered by optional components. */
1146 if (a->cleanup_archive_extract != NULL)
1147 r = (a->cleanup_archive_extract)(a);
1148
1149 /* Cleanup format-specific data. */
1150 slots = sizeof(a->formats) / sizeof(a->formats[0]);
1151 for (i = 0; i < slots; i++) {
1152 a->format = &(a->formats[i]);
1153 if (a->formats[i].cleanup)
1154 (a->formats[i].cleanup)(a);
1155 }
1156
1157 /* Free the filters */
1158 __archive_read_free_filters(a);
1159
1160 /* Release the bidder objects. */
1161 n = sizeof(a->bidders)/sizeof(a->bidders[0]);
1162 for (i = 0; i < n; i++) {
1163 if (a->bidders[i].vtable == NULL ||
1164 a->bidders[i].vtable->free == NULL)
1165 continue;
1166 (a->bidders[i].vtable->free)(&a->bidders[i]);
1167 }
1168
1169 /* Release passphrase list. */
1170 p = a->passphrases.first;
1171 while (p != NULL) {
1172 struct archive_read_passphrase *np = p->next;
1173
1174 /* A passphrase should be cleaned. */
1175 memset(p->passphrase, 0, strlen(p->passphrase));
1176 free(p->passphrase);
1177 free(p);
1178 p = np;
1179 }
1180
1181 archive_string_free(&a->archive.error_string);
1182 archive_entry_free(a->entry);
1183 a->archive.magic = 0;
1184 __archive_clean(&a->archive);
1185 free(a->client.dataset);
1186 free(a);
1187 return (r);
1188 }
1189
1190 static struct archive_read_filter *
get_filter(struct archive * _a,int n)1191 get_filter(struct archive *_a, int n)
1192 {
1193 struct archive_read *a = (struct archive_read *)_a;
1194 struct archive_read_filter *f = a->filter;
1195 /* We use n == -1 for 'the last filter', which is always the
1196 * client proxy. */
1197 if (n == -1 && f != NULL) {
1198 struct archive_read_filter *last = f;
1199 f = f->upstream;
1200 while (f != NULL) {
1201 last = f;
1202 f = f->upstream;
1203 }
1204 return (last);
1205 }
1206 if (n < 0)
1207 return NULL;
1208 while (n > 0 && f != NULL) {
1209 f = f->upstream;
1210 --n;
1211 }
1212 return (f);
1213 }
1214
1215 static int
_archive_filter_code(struct archive * _a,int n)1216 _archive_filter_code(struct archive *_a, int n)
1217 {
1218 struct archive_read_filter *f = get_filter(_a, n);
1219 return f == NULL ? -1 : f->code;
1220 }
1221
1222 static const char *
_archive_filter_name(struct archive * _a,int n)1223 _archive_filter_name(struct archive *_a, int n)
1224 {
1225 struct archive_read_filter *f = get_filter(_a, n);
1226 return f != NULL ? f->name : NULL;
1227 }
1228
1229 static int64_t
_archive_filter_bytes(struct archive * _a,int n)1230 _archive_filter_bytes(struct archive *_a, int n)
1231 {
1232 struct archive_read_filter *f = get_filter(_a, n);
1233 return f == NULL ? -1 : f->position;
1234 }
1235
1236 /*
1237 * Used internally by read format handlers to register their bid and
1238 * initialization functions.
1239 */
1240 int
__archive_read_register_format(struct archive_read * a,void * format_data,const char * name,int (* bid)(struct archive_read *,int),int (* options)(struct archive_read *,const char *,const char *),int (* read_header)(struct archive_read *,struct archive_entry *),int (* read_data)(struct archive_read *,const void **,size_t *,int64_t *),int (* read_data_skip)(struct archive_read *),int64_t (* seek_data)(struct archive_read *,int64_t,int),int (* cleanup)(struct archive_read *),int (* format_capabilities)(struct archive_read *),int (* has_encrypted_entries)(struct archive_read *))1241 __archive_read_register_format(struct archive_read *a,
1242 void *format_data,
1243 const char *name,
1244 int (*bid)(struct archive_read *, int),
1245 int (*options)(struct archive_read *, const char *, const char *),
1246 int (*read_header)(struct archive_read *, struct archive_entry *),
1247 int (*read_data)(struct archive_read *, const void **, size_t *, int64_t *),
1248 int (*read_data_skip)(struct archive_read *),
1249 int64_t (*seek_data)(struct archive_read *, int64_t, int),
1250 int (*cleanup)(struct archive_read *),
1251 int (*format_capabilities)(struct archive_read *),
1252 int (*has_encrypted_entries)(struct archive_read *))
1253 {
1254 int i, number_slots;
1255
1256 archive_check_magic(&a->archive,
1257 ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
1258 "__archive_read_register_format");
1259
1260 number_slots = sizeof(a->formats) / sizeof(a->formats[0]);
1261
1262 for (i = 0; i < number_slots; i++) {
1263 if (a->formats[i].bid == bid)
1264 return (ARCHIVE_WARN); /* We've already installed */
1265 if (a->formats[i].bid == NULL) {
1266 a->formats[i].bid = bid;
1267 a->formats[i].options = options;
1268 a->formats[i].read_header = read_header;
1269 a->formats[i].read_data = read_data;
1270 a->formats[i].read_data_skip = read_data_skip;
1271 a->formats[i].seek_data = seek_data;
1272 a->formats[i].cleanup = cleanup;
1273 a->formats[i].data = format_data;
1274 a->formats[i].name = name;
1275 a->formats[i].format_capabilties = format_capabilities;
1276 a->formats[i].has_encrypted_entries = has_encrypted_entries;
1277 return (ARCHIVE_OK);
1278 }
1279 }
1280
1281 archive_set_error(&a->archive, ENOMEM,
1282 "Not enough slots for format registration");
1283 return (ARCHIVE_FATAL);
1284 }
1285
1286 /*
1287 * Used internally by decompression routines to register their bid and
1288 * initialization functions.
1289 */
1290 int
__archive_read_register_bidder(struct archive_read * a,void * bidder_data,const char * name,const struct archive_read_filter_bidder_vtable * vtable)1291 __archive_read_register_bidder(struct archive_read *a,
1292 void *bidder_data,
1293 const char *name,
1294 const struct archive_read_filter_bidder_vtable *vtable)
1295 {
1296 struct archive_read_filter_bidder *bidder;
1297 int i, number_slots;
1298
1299 archive_check_magic(&a->archive, ARCHIVE_READ_MAGIC,
1300 ARCHIVE_STATE_NEW, "__archive_read_register_bidder");
1301
1302 number_slots = sizeof(a->bidders) / sizeof(a->bidders[0]);
1303
1304 for (i = 0; i < number_slots; i++) {
1305 if (a->bidders[i].vtable != NULL)
1306 continue;
1307 memset(a->bidders + i, 0, sizeof(a->bidders[0]));
1308 bidder = (a->bidders + i);
1309 bidder->data = bidder_data;
1310 bidder->name = name;
1311 bidder->vtable = vtable;
1312 if (bidder->vtable->bid == NULL || bidder->vtable->init == NULL) {
1313 archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
1314 "Internal error: "
1315 "no bid/init for filter bidder");
1316 return (ARCHIVE_FATAL);
1317 }
1318
1319 return (ARCHIVE_OK);
1320 }
1321
1322 archive_set_error(&a->archive, ENOMEM,
1323 "Not enough slots for filter registration");
1324 return (ARCHIVE_FATAL);
1325 }
1326
1327 /*
1328 * The next section implements the peek/consume internal I/O
1329 * system used by archive readers. This system allows simple
1330 * read-ahead for consumers while preserving zero-copy operation
1331 * most of the time.
1332 *
1333 * The two key operations:
1334 * * The read-ahead function returns a pointer to a block of data
1335 * that satisfies a minimum request.
1336 * * The consume function advances the file pointer.
1337 *
1338 * In the ideal case, filters generate blocks of data
1339 * and __archive_read_ahead() just returns pointers directly into
1340 * those blocks. Then __archive_read_consume() just bumps those
1341 * pointers. Only if your request would span blocks does the I/O
1342 * layer use a copy buffer to provide you with a contiguous block of
1343 * data.
1344 *
1345 * A couple of useful idioms:
1346 * * "I just want some data." Ask for 1 byte and pay attention to
1347 * the "number of bytes available" from __archive_read_ahead().
1348 * Consume whatever you actually use.
1349 * * "I want to output a large block of data." As above, ask for 1 byte,
1350 * emit all that's available (up to whatever limit you have), consume
1351 * it all, then repeat until you're done. This effectively means that
1352 * you're passing along the blocks that came from your provider.
1353 * * "I want to peek ahead by a large amount." Ask for 4k or so, then
1354 * double and repeat until you get an error or have enough. Note
1355 * that the I/O layer will likely end up expanding its copy buffer
1356 * to fit your request, so use this technique cautiously. This
1357 * technique is used, for example, by some of the format tasting
1358 * code that has uncertain look-ahead needs.
1359 */
1360
1361 /*
1362 * Looks ahead in the input stream:
1363 * * If 'avail' pointer is provided, that returns number of bytes available
1364 * in the current buffer, which may be much larger than requested.
1365 * * If end-of-file, *avail gets set to zero.
1366 * * If error, *avail gets error code.
1367 * * If request can be met, returns pointer to data.
1368 * * If minimum request cannot be met, returns NULL.
1369 *
1370 * Note: If you just want "some data", ask for 1 byte and pay attention
1371 * to *avail, which will have the actual amount available. If you
1372 * know exactly how many bytes you need, just ask for that and treat
1373 * a NULL return as an error.
1374 *
1375 * Important: This does NOT move the file pointer. See
1376 * __archive_read_consume() below.
1377 */
1378 const void *
__archive_read_ahead(struct archive_read * a,size_t min,ssize_t * avail)1379 __archive_read_ahead(struct archive_read *a, size_t min, ssize_t *avail)
1380 {
1381 return (__archive_read_filter_ahead(a->filter, min, avail));
1382 }
1383
1384 const void *
__archive_read_filter_ahead(struct archive_read_filter * f,size_t min,ssize_t * avail)1385 __archive_read_filter_ahead(struct archive_read_filter *f,
1386 size_t min, ssize_t *avail)
1387 {
1388 ssize_t bytes_read;
1389 size_t tocopy;
1390
1391 if (f->fatal) {
1392 if (avail)
1393 *avail = ARCHIVE_FATAL;
1394 return (NULL);
1395 }
1396
1397 /*
1398 * Keep pulling more data until we can satisfy the request.
1399 */
1400 for (;;) {
1401
1402 /*
1403 * If we can satisfy from the copy buffer (and the
1404 * copy buffer isn't empty), we're done. In particular,
1405 * note that min == 0 is a perfectly well-defined
1406 * request.
1407 */
1408 if (f->avail >= min && f->avail > 0) {
1409 if (avail != NULL)
1410 *avail = f->avail;
1411 return (f->next);
1412 }
1413
1414 /*
1415 * We can satisfy directly from client buffer if everything
1416 * currently in the copy buffer is still in the client buffer.
1417 */
1418 if (f->client_total >= f->client_avail + f->avail
1419 && f->client_avail + f->avail >= min) {
1420 /* "Roll back" to client buffer. */
1421 f->client_avail += f->avail;
1422 f->client_next -= f->avail;
1423 /* Copy buffer is now empty. */
1424 f->avail = 0;
1425 f->next = f->buffer;
1426 /* Return data from client buffer. */
1427 if (avail != NULL)
1428 *avail = f->client_avail;
1429 return (f->client_next);
1430 }
1431
1432 /* Move data forward in copy buffer if necessary. */
1433 if (f->next > f->buffer &&
1434 min > f->buffer_size - (f->next - f->buffer)) {
1435 if (f->avail > 0)
1436 memmove(f->buffer, f->next,
1437 f->avail);
1438 f->next = f->buffer;
1439 }
1440
1441 /* If we've used up the client data, get more. */
1442 if (f->client_avail <= 0) {
1443 if (f->end_of_file) {
1444 if (avail != NULL)
1445 *avail = f->avail;
1446 return (NULL);
1447 }
1448 bytes_read = (f->vtable->read)(f,
1449 &f->client_buff);
1450 if (bytes_read < 0) { /* Read error. */
1451 f->client_total = f->client_avail = 0;
1452 f->client_next =
1453 f->client_buff = NULL;
1454 f->fatal = 1;
1455 if (avail != NULL)
1456 *avail = ARCHIVE_FATAL;
1457 return (NULL);
1458 }
1459 if (bytes_read == 0) {
1460 /* Check for another client object first */
1461 if (f->archive->client.cursor !=
1462 f->archive->client.nodes - 1) {
1463 if (client_switch_proxy(f,
1464 f->archive->client.cursor + 1)
1465 == ARCHIVE_OK)
1466 continue;
1467 }
1468 /* Premature end-of-file. */
1469 f->client_total = f->client_avail = 0;
1470 f->client_next =
1471 f->client_buff = NULL;
1472 f->end_of_file = 1;
1473 /* Return whatever we do have. */
1474 if (avail != NULL)
1475 *avail = f->avail;
1476 return (NULL);
1477 }
1478 f->client_total = bytes_read;
1479 f->client_avail = f->client_total;
1480 f->client_next = f->client_buff;
1481 } else {
1482 /*
1483 * We can't satisfy the request from the copy
1484 * buffer or the existing client data, so we
1485 * need to copy more client data over to the
1486 * copy buffer.
1487 */
1488
1489 /* Ensure the buffer is big enough. */
1490 if (min > f->buffer_size) {
1491 size_t s;
1492 char *p;
1493
1494 /* Double the buffer; watch for overflow. */
1495 s = f->buffer_size;
1496 if (s == 0)
1497 s = min;
1498 while (s < min) {
1499 if (archive_ckd_mul_size(&s, s, 2)) {
1500 /* Integer overflow! */
1501 archive_set_error(
1502 &f->archive->archive,
1503 ENOMEM,
1504 "Unable to allocate copy"
1505 " buffer");
1506 f->fatal = 1;
1507 if (avail != NULL)
1508 *avail = ARCHIVE_FATAL;
1509 return (NULL);
1510 }
1511 }
1512 /* Now s >= min, so allocate a new buffer. */
1513 p = malloc(s);
1514 if (p == NULL) {
1515 archive_set_error(
1516 &f->archive->archive,
1517 ENOMEM,
1518 "Unable to allocate copy buffer");
1519 f->fatal = 1;
1520 if (avail != NULL)
1521 *avail = ARCHIVE_FATAL;
1522 return (NULL);
1523 }
1524 /* Move data into newly-enlarged buffer. */
1525 if (f->avail > 0)
1526 memmove(p, f->next, f->avail);
1527 free(f->buffer);
1528 f->next = f->buffer = p;
1529 f->buffer_size = s;
1530 }
1531
1532 /* We can add client data to copy buffer. */
1533 /* First estimate: copy to fill rest of buffer. */
1534 tocopy = (f->buffer + f->buffer_size)
1535 - (f->next + f->avail);
1536 /* Don't waste time buffering more than we need to. */
1537 if (tocopy + f->avail > min)
1538 tocopy = min - f->avail;
1539 /* Don't copy more than is available. */
1540 if (tocopy > f->client_avail)
1541 tocopy = f->client_avail;
1542
1543 memcpy(f->next + f->avail,
1544 f->client_next, tocopy);
1545 /* Remove this data from client buffer. */
1546 f->client_next += tocopy;
1547 f->client_avail -= tocopy;
1548 /* add it to copy buffer. */
1549 f->avail += tocopy;
1550 }
1551 }
1552 }
1553
1554 /*
1555 * Move the file pointer forward.
1556 */
1557 int64_t
__archive_read_consume(struct archive_read * a,int64_t request)1558 __archive_read_consume(struct archive_read *a, int64_t request)
1559 {
1560 return (__archive_read_filter_consume(a->filter, request));
1561 }
1562
1563 int64_t
__archive_read_filter_consume(struct archive_read_filter * f,int64_t request)1564 __archive_read_filter_consume(struct archive_read_filter *f,
1565 int64_t request)
1566 {
1567 int64_t skipped;
1568
1569 if (request < 0)
1570 return ARCHIVE_FATAL;
1571 if (request == 0)
1572 return 0;
1573
1574 skipped = advance_file_pointer(f, request);
1575 if (skipped == request)
1576 return (skipped);
1577 /* We hit EOF before we satisfied the skip request. */
1578 if (skipped < 0) /* Map error code to 0 for error message below. */
1579 skipped = 0;
1580 archive_set_error(&f->archive->archive,
1581 ARCHIVE_ERRNO_MISC,
1582 "Truncated input file (needed %jd bytes, only %jd available)",
1583 (intmax_t)request, (intmax_t)skipped);
1584 return (ARCHIVE_FATAL);
1585 }
1586
1587 /*
1588 * Advance the file pointer by the amount requested.
1589 * Returns the amount actually advanced, which may be less than the
1590 * request if EOF is encountered first.
1591 * Returns a negative value if there's an I/O error.
1592 */
1593 static int64_t
advance_file_pointer(struct archive_read_filter * f,int64_t request)1594 advance_file_pointer(struct archive_read_filter *f, int64_t request)
1595 {
1596 int64_t bytes_skipped, total_bytes_skipped = 0;
1597 ssize_t bytes_read;
1598 size_t min;
1599
1600 if (f->fatal)
1601 return (-1);
1602
1603 /* Use up the copy buffer first. */
1604 if (f->avail > 0) {
1605 min = (size_t)minimum(request, (int64_t)f->avail);
1606 f->next += min;
1607 f->avail -= min;
1608 request -= min;
1609 f->position += min;
1610 total_bytes_skipped += min;
1611 }
1612
1613 /* Then use up the client buffer. */
1614 if (f->client_avail > 0) {
1615 min = (size_t)minimum(request, (int64_t)f->client_avail);
1616 f->client_next += min;
1617 f->client_avail -= min;
1618 request -= min;
1619 f->position += min;
1620 total_bytes_skipped += min;
1621 }
1622 if (request == 0)
1623 return (total_bytes_skipped);
1624
1625 /* If there's an optimized skip function, use it. */
1626 if (f->can_skip != 0) {
1627 bytes_skipped = client_skip_proxy(f, request);
1628 if (bytes_skipped < 0) { /* error */
1629 f->fatal = 1;
1630 return (bytes_skipped);
1631 }
1632 f->position += bytes_skipped;
1633 total_bytes_skipped += bytes_skipped;
1634 request -= bytes_skipped;
1635 if (request == 0)
1636 return (total_bytes_skipped);
1637 }
1638
1639 /* Use ordinary reads as necessary to complete the request. */
1640 for (;;) {
1641 bytes_read = (f->vtable->read)(f, &f->client_buff);
1642 if (bytes_read < 0) {
1643 f->client_buff = NULL;
1644 f->fatal = 1;
1645 return (bytes_read);
1646 }
1647
1648 if (bytes_read == 0) {
1649 if (f->archive->client.cursor !=
1650 f->archive->client.nodes - 1) {
1651 if (client_switch_proxy(f,
1652 f->archive->client.cursor + 1)
1653 == ARCHIVE_OK)
1654 continue;
1655 }
1656 f->client_buff = NULL;
1657 f->end_of_file = 1;
1658 return (total_bytes_skipped);
1659 }
1660
1661 if (bytes_read >= request) {
1662 f->client_next =
1663 ((const char *)f->client_buff) + request;
1664 f->client_avail = (size_t)(bytes_read - request);
1665 f->client_total = bytes_read;
1666 total_bytes_skipped += request;
1667 f->position += request;
1668 return (total_bytes_skipped);
1669 }
1670
1671 f->position += bytes_read;
1672 total_bytes_skipped += bytes_read;
1673 request -= bytes_read;
1674 }
1675 }
1676
1677 /**
1678 * Returns ARCHIVE_FAILED if seeking isn't supported.
1679 */
1680 int64_t
__archive_read_seek(struct archive_read * a,int64_t offset,int whence)1681 __archive_read_seek(struct archive_read *a, int64_t offset, int whence)
1682 {
1683 return __archive_read_filter_seek(a->filter, offset, whence);
1684 }
1685
1686 int64_t
__archive_read_filter_seek(struct archive_read_filter * f,int64_t offset,int whence)1687 __archive_read_filter_seek(struct archive_read_filter *f, int64_t offset,
1688 int whence)
1689 {
1690 struct archive_read_client *client;
1691 int64_t r;
1692 unsigned int cursor;
1693
1694 if (f->closed || f->fatal)
1695 return (ARCHIVE_FATAL);
1696 if (f->can_seek == 0)
1697 return (ARCHIVE_FAILED);
1698
1699 client = &(f->archive->client);
1700 switch (whence) {
1701 case SEEK_CUR:
1702 /* Adjust the offset and use SEEK_SET instead */
1703 offset += f->position;
1704 __LA_FALLTHROUGH;
1705 case SEEK_SET:
1706 cursor = 0;
1707 while (1)
1708 {
1709 if (client->dataset[cursor].begin_position < 0 ||
1710 client->dataset[cursor].total_size < 0 ||
1711 client->dataset[cursor].begin_position +
1712 client->dataset[cursor].total_size - 1 > offset ||
1713 cursor + 1 >= client->nodes)
1714 break;
1715 r = client->dataset[cursor].begin_position +
1716 client->dataset[cursor].total_size;
1717 client->dataset[++cursor].begin_position = r;
1718 }
1719 while (1) {
1720 r = client_switch_proxy(f, cursor);
1721 if (r != ARCHIVE_OK)
1722 return r;
1723 if ((r = client_seek_proxy(f, 0, SEEK_END)) < 0)
1724 return r;
1725 client->dataset[cursor].total_size = r;
1726 if (client->dataset[cursor].begin_position +
1727 client->dataset[cursor].total_size - 1 > offset ||
1728 cursor + 1 >= client->nodes)
1729 break;
1730 r = client->dataset[cursor].begin_position +
1731 client->dataset[cursor].total_size;
1732 client->dataset[++cursor].begin_position = r;
1733 }
1734 offset -= client->dataset[cursor].begin_position;
1735 if (offset < 0
1736 || offset > client->dataset[cursor].total_size)
1737 return ARCHIVE_FATAL;
1738 if ((r = client_seek_proxy(f, offset, SEEK_SET)) < 0)
1739 return r;
1740 break;
1741
1742 case SEEK_END:
1743 cursor = 0;
1744 while (1) {
1745 if (client->dataset[cursor].begin_position < 0 ||
1746 client->dataset[cursor].total_size < 0 ||
1747 cursor + 1 >= client->nodes)
1748 break;
1749 r = client->dataset[cursor].begin_position +
1750 client->dataset[cursor].total_size;
1751 client->dataset[++cursor].begin_position = r;
1752 }
1753 while (1) {
1754 r = client_switch_proxy(f, cursor);
1755 if (r != ARCHIVE_OK)
1756 return r;
1757 if ((r = client_seek_proxy(f, 0, SEEK_END)) < 0)
1758 return r;
1759 client->dataset[cursor].total_size = r;
1760 r = client->dataset[cursor].begin_position +
1761 client->dataset[cursor].total_size;
1762 if (cursor + 1 >= client->nodes)
1763 break;
1764 client->dataset[++cursor].begin_position = r;
1765 }
1766 while (1) {
1767 if (r + offset >=
1768 client->dataset[cursor].begin_position)
1769 break;
1770 offset += client->dataset[cursor].total_size;
1771 if (cursor == 0)
1772 break;
1773 cursor--;
1774 r = client->dataset[cursor].begin_position +
1775 client->dataset[cursor].total_size;
1776 }
1777 offset = (r + offset) - client->dataset[cursor].begin_position;
1778 if ((r = client_switch_proxy(f, cursor)) != ARCHIVE_OK)
1779 return r;
1780 r = client_seek_proxy(f, offset, SEEK_SET);
1781 if (r < ARCHIVE_OK)
1782 return r;
1783 break;
1784
1785 default:
1786 return (ARCHIVE_FATAL);
1787 }
1788 r += client->dataset[cursor].begin_position;
1789
1790 if (r >= 0) {
1791 /*
1792 * Ouch. Clearing the buffer like this hurts, especially
1793 * at bid time. A lot of our efficiency at bid time comes
1794 * from having bidders reuse the data we've already read.
1795 *
1796 * TODO: If the seek request is in data we already
1797 * have, then don't call the seek callback.
1798 *
1799 * TODO: Zip seeks to end-of-file at bid time. If
1800 * other formats also start doing this, we may need to
1801 * find a way for clients to fudge the seek offset to
1802 * a block boundary.
1803 *
1804 * Hmmm... If whence was SEEK_END, we know the file
1805 * size is (r - offset). Can we use that to simplify
1806 * the TODO items above?
1807 */
1808 f->avail = f->client_avail = 0;
1809 f->next = f->buffer;
1810 f->position = r;
1811 f->end_of_file = 0;
1812 }
1813 return r;
1814 }
1815