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: head/lib/libarchive/archive_read.c 201157 2009-12-29 05:30:23Z kientzle $");
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 *
archive_read_vtable(void)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 *
archive_read_new(void)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
archive_read_extract_set_skip_file(struct archive * _a,la_int64_t d,la_int64_t i)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
archive_read_open(struct archive * a,void * client_data,archive_open_callback * client_opener,archive_read_callback * client_reader,archive_close_callback * client_closer)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
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)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
client_read_proxy(struct archive_read_filter * self,const void ** buff)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
client_skip_proxy(struct archive_read_filter * self,int64_t request)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
client_seek_proxy(struct archive_read_filter * self,int64_t offset,int whence)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
client_close_proxy(struct archive_read_filter * self)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
client_open_proxy(struct archive_read_filter * self)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
client_switch_proxy(struct archive_read_filter * self,unsigned int iindex)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
archive_read_set_open_callback(struct archive * _a,archive_open_callback * client_opener)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
archive_read_set_read_callback(struct archive * _a,archive_read_callback * client_reader)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
archive_read_set_skip_callback(struct archive * _a,archive_skip_callback * client_skipper)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
archive_read_set_seek_callback(struct archive * _a,archive_seek_callback * client_seeker)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
archive_read_set_close_callback(struct archive * _a,archive_close_callback * client_closer)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
archive_read_set_switch_callback(struct archive * _a,archive_switch_callback * client_switcher)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
archive_read_set_callback_data(struct archive * _a,void * client_data)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
archive_read_set_callback_data2(struct archive * _a,void * client_data,unsigned int iindex)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
archive_read_add_callback_data(struct archive * _a,void * client_data,unsigned int iindex)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 > 0; 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
archive_read_append_callback_data(struct archive * _a,void * client_data)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
archive_read_prepend_callback_data(struct archive * _a,void * client_data)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
archive_read_open1(struct archive * _a)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
choose_filters(struct archive_read * a)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
__archive_read_header(struct archive_read * a,struct archive_entry * entry)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
_archive_read_next_header2(struct archive * _a,struct archive_entry * entry)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
_archive_read_next_header(struct archive * _a,struct archive_entry ** entryp)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
choose_format(struct archive_read * a)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
archive_read_header_position(struct archive * _a)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
archive_read_has_encrypted_entries(struct archive * _a)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
archive_read_format_capabilities(struct archive * _a)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
archive_read_data(struct archive * _a,void * buff,size_t s)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_remaining == 0) {
848 read_buf = a->read_data_block;
849 a->read_data_is_posix_read = 1;
850 a->read_data_requested = s;
851 r = archive_read_data_block(a, &read_buf,
852 &a->read_data_remaining, &a->read_data_offset);
853 a->read_data_block = read_buf;
854 if (r == ARCHIVE_EOF)
855 return (bytes_read);
856 /*
857 * Error codes are all negative, so the status
858 * return here cannot be confused with a valid
859 * byte count. (ARCHIVE_OK is zero.)
860 */
861 if (r < ARCHIVE_OK)
862 return (r);
863 }
864
865 if (a->read_data_offset < a->read_data_output_offset) {
866 archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
867 "Encountered out-of-order sparse blocks");
868 return (ARCHIVE_RETRY);
869 }
870
871 /* Compute the amount of zero padding needed. */
872 if (a->read_data_output_offset + (int64_t)s <
873 a->read_data_offset) {
874 len = s;
875 } else if (a->read_data_output_offset <
876 a->read_data_offset) {
877 len = (size_t)(a->read_data_offset -
878 a->read_data_output_offset);
879 } else
880 len = 0;
881
882 /* Add zeroes. */
883 memset(dest, 0, len);
884 s -= len;
885 a->read_data_output_offset += len;
886 dest += len;
887 bytes_read += len;
888
889 /* Copy data if there is any space left. */
890 if (s > 0) {
891 len = a->read_data_remaining;
892 if (len > s)
893 len = s;
894 if (len) {
895 memcpy(dest, a->read_data_block, len);
896 s -= len;
897 a->read_data_block += len;
898 a->read_data_remaining -= len;
899 a->read_data_output_offset += len;
900 a->read_data_offset += len;
901 dest += len;
902 bytes_read += len;
903 }
904 }
905 }
906 a->read_data_is_posix_read = 0;
907 a->read_data_requested = 0;
908 return (bytes_read);
909 }
910
911 /*
912 * Reset the read_data_* variables, used for starting a new entry.
913 */
__archive_reset_read_data(struct archive * a)914 void __archive_reset_read_data(struct archive * a)
915 {
916 a->read_data_output_offset = 0;
917 a->read_data_remaining = 0;
918 a->read_data_is_posix_read = 0;
919 a->read_data_requested = 0;
920
921 /* extra resets, from rar.c */
922 a->read_data_block = NULL;
923 a->read_data_offset = 0;
924 }
925
926 /*
927 * Skip over all remaining data in this entry.
928 */
929 int
archive_read_data_skip(struct archive * _a)930 archive_read_data_skip(struct archive *_a)
931 {
932 struct archive_read *a = (struct archive_read *)_a;
933 int r;
934 const void *buff;
935 size_t size;
936 int64_t offset;
937
938 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA,
939 "archive_read_data_skip");
940
941 if (a->format->read_data_skip != NULL)
942 r = (a->format->read_data_skip)(a);
943 else {
944 while ((r = archive_read_data_block(&a->archive,
945 &buff, &size, &offset))
946 == ARCHIVE_OK)
947 ;
948 }
949
950 if (r == ARCHIVE_EOF)
951 r = ARCHIVE_OK;
952
953 a->archive.state = ARCHIVE_STATE_HEADER;
954 return (r);
955 }
956
957 la_int64_t
archive_seek_data(struct archive * _a,int64_t offset,int whence)958 archive_seek_data(struct archive *_a, int64_t offset, int whence)
959 {
960 struct archive_read *a = (struct archive_read *)_a;
961 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA,
962 "archive_seek_data_block");
963
964 if (a->format->seek_data == NULL) {
965 archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
966 "Internal error: "
967 "No format_seek_data_block function registered");
968 return (ARCHIVE_FATAL);
969 }
970
971 return (a->format->seek_data)(a, offset, whence);
972 }
973
974 /*
975 * Read the next block of entry data from the archive.
976 * This is a zero-copy interface; the client receives a pointer,
977 * size, and file offset of the next available block of data.
978 *
979 * Returns ARCHIVE_OK if the operation is successful, ARCHIVE_EOF if
980 * the end of entry is encountered.
981 */
982 static int
_archive_read_data_block(struct archive * _a,const void ** buff,size_t * size,int64_t * offset)983 _archive_read_data_block(struct archive *_a,
984 const void **buff, size_t *size, int64_t *offset)
985 {
986 struct archive_read *a = (struct archive_read *)_a;
987 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA,
988 "archive_read_data_block");
989
990 if (a->format->read_data == NULL) {
991 archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
992 "Internal error: "
993 "No format->read_data function registered");
994 return (ARCHIVE_FATAL);
995 }
996
997 return (a->format->read_data)(a, buff, size, offset);
998 }
999
1000 static int
close_filters(struct archive_read * a)1001 close_filters(struct archive_read *a)
1002 {
1003 struct archive_read_filter *f = a->filter;
1004 int r = ARCHIVE_OK;
1005 /* Close each filter in the pipeline. */
1006 while (f != NULL) {
1007 struct archive_read_filter *t = f->upstream;
1008 if (!f->closed && f->close != NULL) {
1009 int r1 = (f->close)(f);
1010 f->closed = 1;
1011 if (r1 < r)
1012 r = r1;
1013 }
1014 free(f->buffer);
1015 f->buffer = NULL;
1016 f = t;
1017 }
1018 return r;
1019 }
1020
1021 void
__archive_read_free_filters(struct archive_read * a)1022 __archive_read_free_filters(struct archive_read *a)
1023 {
1024 /* Make sure filters are closed and their buffers are freed */
1025 close_filters(a);
1026
1027 while (a->filter != NULL) {
1028 struct archive_read_filter *t = a->filter->upstream;
1029 free(a->filter);
1030 a->filter = t;
1031 }
1032 }
1033
1034 /*
1035 * return the count of # of filters in use
1036 */
1037 static int
_archive_filter_count(struct archive * _a)1038 _archive_filter_count(struct archive *_a)
1039 {
1040 struct archive_read *a = (struct archive_read *)_a;
1041 struct archive_read_filter *p = a->filter;
1042 int count = 0;
1043 while(p) {
1044 count++;
1045 p = p->upstream;
1046 }
1047 return count;
1048 }
1049
1050 /*
1051 * Close the file and all I/O.
1052 */
1053 static int
_archive_read_close(struct archive * _a)1054 _archive_read_close(struct archive *_a)
1055 {
1056 struct archive_read *a = (struct archive_read *)_a;
1057 int r = ARCHIVE_OK, r1 = ARCHIVE_OK;
1058
1059 archive_check_magic(&a->archive, ARCHIVE_READ_MAGIC,
1060 ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_close");
1061 if (a->archive.state == ARCHIVE_STATE_CLOSED)
1062 return (ARCHIVE_OK);
1063 archive_clear_error(&a->archive);
1064 a->archive.state = ARCHIVE_STATE_CLOSED;
1065
1066 /* TODO: Clean up the formatters. */
1067
1068 /* Release the filter objects. */
1069 r1 = close_filters(a);
1070 if (r1 < r)
1071 r = r1;
1072
1073 return (r);
1074 }
1075
1076 /*
1077 * Release memory and other resources.
1078 */
1079 static int
_archive_read_free(struct archive * _a)1080 _archive_read_free(struct archive *_a)
1081 {
1082 struct archive_read *a = (struct archive_read *)_a;
1083 struct archive_read_passphrase *p;
1084 int i, n;
1085 int slots;
1086 int r = ARCHIVE_OK;
1087
1088 if (_a == NULL)
1089 return (ARCHIVE_OK);
1090 archive_check_magic(_a, ARCHIVE_READ_MAGIC,
1091 ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_free");
1092 if (a->archive.state != ARCHIVE_STATE_CLOSED
1093 && a->archive.state != ARCHIVE_STATE_FATAL)
1094 r = archive_read_close(&a->archive);
1095
1096 /* Call cleanup functions registered by optional components. */
1097 if (a->cleanup_archive_extract != NULL)
1098 r = (a->cleanup_archive_extract)(a);
1099
1100 /* Cleanup format-specific data. */
1101 slots = sizeof(a->formats) / sizeof(a->formats[0]);
1102 for (i = 0; i < slots; i++) {
1103 a->format = &(a->formats[i]);
1104 if (a->formats[i].cleanup)
1105 (a->formats[i].cleanup)(a);
1106 }
1107
1108 /* Free the filters */
1109 __archive_read_free_filters(a);
1110
1111 /* Release the bidder objects. */
1112 n = sizeof(a->bidders)/sizeof(a->bidders[0]);
1113 for (i = 0; i < n; i++) {
1114 if (a->bidders[i].free != NULL) {
1115 int r1 = (a->bidders[i].free)(&a->bidders[i]);
1116 if (r1 < r)
1117 r = r1;
1118 }
1119 }
1120
1121 /* Release passphrase list. */
1122 p = a->passphrases.first;
1123 while (p != NULL) {
1124 struct archive_read_passphrase *np = p->next;
1125
1126 /* A passphrase should be cleaned. */
1127 memset(p->passphrase, 0, strlen(p->passphrase));
1128 free(p->passphrase);
1129 free(p);
1130 p = np;
1131 }
1132
1133 archive_string_free(&a->archive.error_string);
1134 archive_entry_free(a->entry);
1135 a->archive.magic = 0;
1136 __archive_clean(&a->archive);
1137 free(a->client.dataset);
1138 free(a);
1139 return (r);
1140 }
1141
1142 static struct archive_read_filter *
get_filter(struct archive * _a,int n)1143 get_filter(struct archive *_a, int n)
1144 {
1145 struct archive_read *a = (struct archive_read *)_a;
1146 struct archive_read_filter *f = a->filter;
1147 /* We use n == -1 for 'the last filter', which is always the
1148 * client proxy. */
1149 if (n == -1 && f != NULL) {
1150 struct archive_read_filter *last = f;
1151 f = f->upstream;
1152 while (f != NULL) {
1153 last = f;
1154 f = f->upstream;
1155 }
1156 return (last);
1157 }
1158 if (n < 0)
1159 return NULL;
1160 while (n > 0 && f != NULL) {
1161 f = f->upstream;
1162 --n;
1163 }
1164 return (f);
1165 }
1166
1167 static int
_archive_filter_code(struct archive * _a,int n)1168 _archive_filter_code(struct archive *_a, int n)
1169 {
1170 struct archive_read_filter *f = get_filter(_a, n);
1171 return f == NULL ? -1 : f->code;
1172 }
1173
1174 static const char *
_archive_filter_name(struct archive * _a,int n)1175 _archive_filter_name(struct archive *_a, int n)
1176 {
1177 struct archive_read_filter *f = get_filter(_a, n);
1178 return f != NULL ? f->name : NULL;
1179 }
1180
1181 static int64_t
_archive_filter_bytes(struct archive * _a,int n)1182 _archive_filter_bytes(struct archive *_a, int n)
1183 {
1184 struct archive_read_filter *f = get_filter(_a, n);
1185 return f == NULL ? -1 : f->position;
1186 }
1187
1188 /*
1189 * Used internally by read format handlers to register their bid and
1190 * initialization functions.
1191 */
1192 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 *))1193 __archive_read_register_format(struct archive_read *a,
1194 void *format_data,
1195 const char *name,
1196 int (*bid)(struct archive_read *, int),
1197 int (*options)(struct archive_read *, const char *, const char *),
1198 int (*read_header)(struct archive_read *, struct archive_entry *),
1199 int (*read_data)(struct archive_read *, const void **, size_t *, int64_t *),
1200 int (*read_data_skip)(struct archive_read *),
1201 int64_t (*seek_data)(struct archive_read *, int64_t, int),
1202 int (*cleanup)(struct archive_read *),
1203 int (*format_capabilities)(struct archive_read *),
1204 int (*has_encrypted_entries)(struct archive_read *))
1205 {
1206 int i, number_slots;
1207
1208 archive_check_magic(&a->archive,
1209 ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
1210 "__archive_read_register_format");
1211
1212 number_slots = sizeof(a->formats) / sizeof(a->formats[0]);
1213
1214 for (i = 0; i < number_slots; i++) {
1215 if (a->formats[i].bid == bid)
1216 return (ARCHIVE_WARN); /* We've already installed */
1217 if (a->formats[i].bid == NULL) {
1218 a->formats[i].bid = bid;
1219 a->formats[i].options = options;
1220 a->formats[i].read_header = read_header;
1221 a->formats[i].read_data = read_data;
1222 a->formats[i].read_data_skip = read_data_skip;
1223 a->formats[i].seek_data = seek_data;
1224 a->formats[i].cleanup = cleanup;
1225 a->formats[i].data = format_data;
1226 a->formats[i].name = name;
1227 a->formats[i].format_capabilties = format_capabilities;
1228 a->formats[i].has_encrypted_entries = has_encrypted_entries;
1229 return (ARCHIVE_OK);
1230 }
1231 }
1232
1233 archive_set_error(&a->archive, ENOMEM,
1234 "Not enough slots for format registration");
1235 return (ARCHIVE_FATAL);
1236 }
1237
1238 /*
1239 * Used internally by decompression routines to register their bid and
1240 * initialization functions.
1241 */
1242 int
__archive_read_get_bidder(struct archive_read * a,struct archive_read_filter_bidder ** bidder)1243 __archive_read_get_bidder(struct archive_read *a,
1244 struct archive_read_filter_bidder **bidder)
1245 {
1246 int i, number_slots;
1247
1248 number_slots = sizeof(a->bidders) / sizeof(a->bidders[0]);
1249
1250 for (i = 0; i < number_slots; i++) {
1251 if (a->bidders[i].bid == NULL) {
1252 memset(a->bidders + i, 0, sizeof(a->bidders[0]));
1253 *bidder = (a->bidders + i);
1254 return (ARCHIVE_OK);
1255 }
1256 }
1257
1258 archive_set_error(&a->archive, ENOMEM,
1259 "Not enough slots for filter registration");
1260 return (ARCHIVE_FATAL);
1261 }
1262
1263 /*
1264 * The next section implements the peek/consume internal I/O
1265 * system used by archive readers. This system allows simple
1266 * read-ahead for consumers while preserving zero-copy operation
1267 * most of the time.
1268 *
1269 * The two key operations:
1270 * * The read-ahead function returns a pointer to a block of data
1271 * that satisfies a minimum request.
1272 * * The consume function advances the file pointer.
1273 *
1274 * In the ideal case, filters generate blocks of data
1275 * and __archive_read_ahead() just returns pointers directly into
1276 * those blocks. Then __archive_read_consume() just bumps those
1277 * pointers. Only if your request would span blocks does the I/O
1278 * layer use a copy buffer to provide you with a contiguous block of
1279 * data.
1280 *
1281 * A couple of useful idioms:
1282 * * "I just want some data." Ask for 1 byte and pay attention to
1283 * the "number of bytes available" from __archive_read_ahead().
1284 * Consume whatever you actually use.
1285 * * "I want to output a large block of data." As above, ask for 1 byte,
1286 * emit all that's available (up to whatever limit you have), consume
1287 * it all, then repeat until you're done. This effectively means that
1288 * you're passing along the blocks that came from your provider.
1289 * * "I want to peek ahead by a large amount." Ask for 4k or so, then
1290 * double and repeat until you get an error or have enough. Note
1291 * that the I/O layer will likely end up expanding its copy buffer
1292 * to fit your request, so use this technique cautiously. This
1293 * technique is used, for example, by some of the format tasting
1294 * code that has uncertain look-ahead needs.
1295 */
1296
1297 /*
1298 * Looks ahead in the input stream:
1299 * * If 'avail' pointer is provided, that returns number of bytes available
1300 * in the current buffer, which may be much larger than requested.
1301 * * If end-of-file, *avail gets set to zero.
1302 * * If error, *avail gets error code.
1303 * * If request can be met, returns pointer to data.
1304 * * If minimum request cannot be met, returns NULL.
1305 *
1306 * Note: If you just want "some data", ask for 1 byte and pay attention
1307 * to *avail, which will have the actual amount available. If you
1308 * know exactly how many bytes you need, just ask for that and treat
1309 * a NULL return as an error.
1310 *
1311 * Important: This does NOT move the file pointer. See
1312 * __archive_read_consume() below.
1313 */
1314 const void *
__archive_read_ahead(struct archive_read * a,size_t min,ssize_t * avail)1315 __archive_read_ahead(struct archive_read *a, size_t min, ssize_t *avail)
1316 {
1317 return (__archive_read_filter_ahead(a->filter, min, avail));
1318 }
1319
1320 const void *
__archive_read_filter_ahead(struct archive_read_filter * filter,size_t min,ssize_t * avail)1321 __archive_read_filter_ahead(struct archive_read_filter *filter,
1322 size_t min, ssize_t *avail)
1323 {
1324 ssize_t bytes_read;
1325 size_t tocopy;
1326
1327 if (filter->fatal) {
1328 if (avail)
1329 *avail = ARCHIVE_FATAL;
1330 return (NULL);
1331 }
1332
1333 /*
1334 * Keep pulling more data until we can satisfy the request.
1335 */
1336 for (;;) {
1337
1338 /*
1339 * If we can satisfy from the copy buffer (and the
1340 * copy buffer isn't empty), we're done. In particular,
1341 * note that min == 0 is a perfectly well-defined
1342 * request.
1343 */
1344 if (filter->avail >= min && filter->avail > 0) {
1345 if (avail != NULL)
1346 *avail = filter->avail;
1347 return (filter->next);
1348 }
1349
1350 /*
1351 * We can satisfy directly from client buffer if everything
1352 * currently in the copy buffer is still in the client buffer.
1353 */
1354 if (filter->client_total >= filter->client_avail + filter->avail
1355 && filter->client_avail + filter->avail >= min) {
1356 /* "Roll back" to client buffer. */
1357 filter->client_avail += filter->avail;
1358 filter->client_next -= filter->avail;
1359 /* Copy buffer is now empty. */
1360 filter->avail = 0;
1361 filter->next = filter->buffer;
1362 /* Return data from client buffer. */
1363 if (avail != NULL)
1364 *avail = filter->client_avail;
1365 return (filter->client_next);
1366 }
1367
1368 /* Move data forward in copy buffer if necessary. */
1369 if (filter->next > filter->buffer &&
1370 filter->next + min > filter->buffer + filter->buffer_size) {
1371 if (filter->avail > 0)
1372 memmove(filter->buffer, filter->next,
1373 filter->avail);
1374 filter->next = filter->buffer;
1375 }
1376
1377 /* If we've used up the client data, get more. */
1378 if (filter->client_avail <= 0) {
1379 if (filter->end_of_file) {
1380 if (avail != NULL)
1381 *avail = 0;
1382 return (NULL);
1383 }
1384 bytes_read = (filter->read)(filter,
1385 &filter->client_buff);
1386 if (bytes_read < 0) { /* Read error. */
1387 filter->client_total = filter->client_avail = 0;
1388 filter->client_next =
1389 filter->client_buff = NULL;
1390 filter->fatal = 1;
1391 if (avail != NULL)
1392 *avail = ARCHIVE_FATAL;
1393 return (NULL);
1394 }
1395 if (bytes_read == 0) {
1396 /* Check for another client object first */
1397 if (filter->archive->client.cursor !=
1398 filter->archive->client.nodes - 1) {
1399 if (client_switch_proxy(filter,
1400 filter->archive->client.cursor + 1)
1401 == ARCHIVE_OK)
1402 continue;
1403 }
1404 /* Premature end-of-file. */
1405 filter->client_total = filter->client_avail = 0;
1406 filter->client_next =
1407 filter->client_buff = NULL;
1408 filter->end_of_file = 1;
1409 /* Return whatever we do have. */
1410 if (avail != NULL)
1411 *avail = filter->avail;
1412 return (NULL);
1413 }
1414 filter->client_total = bytes_read;
1415 filter->client_avail = filter->client_total;
1416 filter->client_next = filter->client_buff;
1417 } else {
1418 /*
1419 * We can't satisfy the request from the copy
1420 * buffer or the existing client data, so we
1421 * need to copy more client data over to the
1422 * copy buffer.
1423 */
1424
1425 /* Ensure the buffer is big enough. */
1426 if (min > filter->buffer_size) {
1427 size_t s, t;
1428 char *p;
1429
1430 /* Double the buffer; watch for overflow. */
1431 s = t = filter->buffer_size;
1432 if (s == 0)
1433 s = min;
1434 while (s < min) {
1435 t *= 2;
1436 if (t <= s) { /* Integer overflow! */
1437 archive_set_error(
1438 &filter->archive->archive,
1439 ENOMEM,
1440 "Unable to allocate copy"
1441 " buffer");
1442 filter->fatal = 1;
1443 if (avail != NULL)
1444 *avail = ARCHIVE_FATAL;
1445 return (NULL);
1446 }
1447 s = t;
1448 }
1449 /* Now s >= min, so allocate a new buffer. */
1450 p = (char *)malloc(s);
1451 if (p == NULL) {
1452 archive_set_error(
1453 &filter->archive->archive,
1454 ENOMEM,
1455 "Unable to allocate copy buffer");
1456 filter->fatal = 1;
1457 if (avail != NULL)
1458 *avail = ARCHIVE_FATAL;
1459 return (NULL);
1460 }
1461 /* Move data into newly-enlarged buffer. */
1462 if (filter->avail > 0)
1463 memmove(p, filter->next, filter->avail);
1464 free(filter->buffer);
1465 filter->next = filter->buffer = p;
1466 filter->buffer_size = s;
1467 }
1468
1469 /* We can add client data to copy buffer. */
1470 /* First estimate: copy to fill rest of buffer. */
1471 tocopy = (filter->buffer + filter->buffer_size)
1472 - (filter->next + filter->avail);
1473 /* Don't waste time buffering more than we need to. */
1474 if (tocopy + filter->avail > min)
1475 tocopy = min - filter->avail;
1476 /* Don't copy more than is available. */
1477 if (tocopy > filter->client_avail)
1478 tocopy = filter->client_avail;
1479
1480 memcpy(filter->next + filter->avail,
1481 filter->client_next, tocopy);
1482 /* Remove this data from client buffer. */
1483 filter->client_next += tocopy;
1484 filter->client_avail -= tocopy;
1485 /* add it to copy buffer. */
1486 filter->avail += tocopy;
1487 }
1488 }
1489 }
1490
1491 /*
1492 * Move the file pointer forward.
1493 */
1494 int64_t
__archive_read_consume(struct archive_read * a,int64_t request)1495 __archive_read_consume(struct archive_read *a, int64_t request)
1496 {
1497 return (__archive_read_filter_consume(a->filter, request));
1498 }
1499
1500 int64_t
__archive_read_filter_consume(struct archive_read_filter * filter,int64_t request)1501 __archive_read_filter_consume(struct archive_read_filter * filter,
1502 int64_t request)
1503 {
1504 int64_t skipped;
1505
1506 if (request < 0)
1507 return ARCHIVE_FATAL;
1508 if (request == 0)
1509 return 0;
1510
1511 skipped = advance_file_pointer(filter, request);
1512 if (skipped == request)
1513 return (skipped);
1514 /* We hit EOF before we satisfied the skip request. */
1515 if (skipped < 0) /* Map error code to 0 for error message below. */
1516 skipped = 0;
1517 archive_set_error(&filter->archive->archive,
1518 ARCHIVE_ERRNO_MISC,
1519 "Truncated input file (needed %jd bytes, only %jd available)",
1520 (intmax_t)request, (intmax_t)skipped);
1521 return (ARCHIVE_FATAL);
1522 }
1523
1524 /*
1525 * Advance the file pointer by the amount requested.
1526 * Returns the amount actually advanced, which may be less than the
1527 * request if EOF is encountered first.
1528 * Returns a negative value if there's an I/O error.
1529 */
1530 static int64_t
advance_file_pointer(struct archive_read_filter * filter,int64_t request)1531 advance_file_pointer(struct archive_read_filter *filter, int64_t request)
1532 {
1533 int64_t bytes_skipped, total_bytes_skipped = 0;
1534 ssize_t bytes_read;
1535 size_t min;
1536
1537 if (filter->fatal)
1538 return (-1);
1539
1540 /* Use up the copy buffer first. */
1541 if (filter->avail > 0) {
1542 min = (size_t)minimum(request, (int64_t)filter->avail);
1543 filter->next += min;
1544 filter->avail -= min;
1545 request -= min;
1546 filter->position += min;
1547 total_bytes_skipped += min;
1548 }
1549
1550 /* Then use up the client buffer. */
1551 if (filter->client_avail > 0) {
1552 min = (size_t)minimum(request, (int64_t)filter->client_avail);
1553 filter->client_next += min;
1554 filter->client_avail -= min;
1555 request -= min;
1556 filter->position += min;
1557 total_bytes_skipped += min;
1558 }
1559 if (request == 0)
1560 return (total_bytes_skipped);
1561
1562 /* If there's an optimized skip function, use it. */
1563 if (filter->skip != NULL) {
1564 bytes_skipped = (filter->skip)(filter, request);
1565 if (bytes_skipped < 0) { /* error */
1566 filter->fatal = 1;
1567 return (bytes_skipped);
1568 }
1569 filter->position += bytes_skipped;
1570 total_bytes_skipped += bytes_skipped;
1571 request -= bytes_skipped;
1572 if (request == 0)
1573 return (total_bytes_skipped);
1574 }
1575
1576 /* Use ordinary reads as necessary to complete the request. */
1577 for (;;) {
1578 bytes_read = (filter->read)(filter, &filter->client_buff);
1579 if (bytes_read < 0) {
1580 filter->client_buff = NULL;
1581 filter->fatal = 1;
1582 return (bytes_read);
1583 }
1584
1585 if (bytes_read == 0) {
1586 if (filter->archive->client.cursor !=
1587 filter->archive->client.nodes - 1) {
1588 if (client_switch_proxy(filter,
1589 filter->archive->client.cursor + 1)
1590 == ARCHIVE_OK)
1591 continue;
1592 }
1593 filter->client_buff = NULL;
1594 filter->end_of_file = 1;
1595 return (total_bytes_skipped);
1596 }
1597
1598 if (bytes_read >= request) {
1599 filter->client_next =
1600 ((const char *)filter->client_buff) + request;
1601 filter->client_avail = (size_t)(bytes_read - request);
1602 filter->client_total = bytes_read;
1603 total_bytes_skipped += request;
1604 filter->position += request;
1605 return (total_bytes_skipped);
1606 }
1607
1608 filter->position += bytes_read;
1609 total_bytes_skipped += bytes_read;
1610 request -= bytes_read;
1611 }
1612 }
1613
1614 /**
1615 * Returns ARCHIVE_FAILED if seeking isn't supported.
1616 */
1617 int64_t
__archive_read_seek(struct archive_read * a,int64_t offset,int whence)1618 __archive_read_seek(struct archive_read *a, int64_t offset, int whence)
1619 {
1620 return __archive_read_filter_seek(a->filter, offset, whence);
1621 }
1622
1623 int64_t
__archive_read_filter_seek(struct archive_read_filter * filter,int64_t offset,int whence)1624 __archive_read_filter_seek(struct archive_read_filter *filter, int64_t offset,
1625 int whence)
1626 {
1627 struct archive_read_client *client;
1628 int64_t r;
1629 unsigned int cursor;
1630
1631 if (filter->closed || filter->fatal)
1632 return (ARCHIVE_FATAL);
1633 if (filter->seek == NULL)
1634 return (ARCHIVE_FAILED);
1635
1636 client = &(filter->archive->client);
1637 switch (whence) {
1638 case SEEK_CUR:
1639 /* Adjust the offset and use SEEK_SET instead */
1640 offset += filter->position;
1641 __LA_FALLTHROUGH;
1642 case SEEK_SET:
1643 cursor = 0;
1644 while (1)
1645 {
1646 if (client->dataset[cursor].begin_position < 0 ||
1647 client->dataset[cursor].total_size < 0 ||
1648 client->dataset[cursor].begin_position +
1649 client->dataset[cursor].total_size - 1 > offset ||
1650 cursor + 1 >= client->nodes)
1651 break;
1652 r = client->dataset[cursor].begin_position +
1653 client->dataset[cursor].total_size;
1654 client->dataset[++cursor].begin_position = r;
1655 }
1656 while (1) {
1657 r = client_switch_proxy(filter, cursor);
1658 if (r != ARCHIVE_OK)
1659 return r;
1660 if ((r = client_seek_proxy(filter, 0, SEEK_END)) < 0)
1661 return r;
1662 client->dataset[cursor].total_size = r;
1663 if (client->dataset[cursor].begin_position +
1664 client->dataset[cursor].total_size - 1 > offset ||
1665 cursor + 1 >= client->nodes)
1666 break;
1667 r = client->dataset[cursor].begin_position +
1668 client->dataset[cursor].total_size;
1669 client->dataset[++cursor].begin_position = r;
1670 }
1671 offset -= client->dataset[cursor].begin_position;
1672 if (offset < 0
1673 || offset > client->dataset[cursor].total_size)
1674 return ARCHIVE_FATAL;
1675 if ((r = client_seek_proxy(filter, offset, SEEK_SET)) < 0)
1676 return r;
1677 break;
1678
1679 case SEEK_END:
1680 cursor = 0;
1681 while (1) {
1682 if (client->dataset[cursor].begin_position < 0 ||
1683 client->dataset[cursor].total_size < 0 ||
1684 cursor + 1 >= client->nodes)
1685 break;
1686 r = client->dataset[cursor].begin_position +
1687 client->dataset[cursor].total_size;
1688 client->dataset[++cursor].begin_position = r;
1689 }
1690 while (1) {
1691 r = client_switch_proxy(filter, cursor);
1692 if (r != ARCHIVE_OK)
1693 return r;
1694 if ((r = client_seek_proxy(filter, 0, SEEK_END)) < 0)
1695 return r;
1696 client->dataset[cursor].total_size = r;
1697 r = client->dataset[cursor].begin_position +
1698 client->dataset[cursor].total_size;
1699 if (cursor + 1 >= client->nodes)
1700 break;
1701 client->dataset[++cursor].begin_position = r;
1702 }
1703 while (1) {
1704 if (r + offset >=
1705 client->dataset[cursor].begin_position)
1706 break;
1707 offset += client->dataset[cursor].total_size;
1708 if (cursor == 0)
1709 break;
1710 cursor--;
1711 r = client->dataset[cursor].begin_position +
1712 client->dataset[cursor].total_size;
1713 }
1714 offset = (r + offset) - client->dataset[cursor].begin_position;
1715 if ((r = client_switch_proxy(filter, cursor)) != ARCHIVE_OK)
1716 return r;
1717 r = client_seek_proxy(filter, offset, SEEK_SET);
1718 if (r < ARCHIVE_OK)
1719 return r;
1720 break;
1721
1722 default:
1723 return (ARCHIVE_FATAL);
1724 }
1725 r += client->dataset[cursor].begin_position;
1726
1727 if (r >= 0) {
1728 /*
1729 * Ouch. Clearing the buffer like this hurts, especially
1730 * at bid time. A lot of our efficiency at bid time comes
1731 * from having bidders reuse the data we've already read.
1732 *
1733 * TODO: If the seek request is in data we already
1734 * have, then don't call the seek callback.
1735 *
1736 * TODO: Zip seeks to end-of-file at bid time. If
1737 * other formats also start doing this, we may need to
1738 * find a way for clients to fudge the seek offset to
1739 * a block boundary.
1740 *
1741 * Hmmm... If whence was SEEK_END, we know the file
1742 * size is (r - offset). Can we use that to simplify
1743 * the TODO items above?
1744 */
1745 filter->avail = filter->client_avail = 0;
1746 filter->next = filter->buffer;
1747 filter->position = r;
1748 filter->end_of_file = 0;
1749 }
1750 return r;
1751 }
1752