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 #include <stdio.h>
40 #ifdef HAVE_STDLIB_H
41 #include <stdlib.h>
42 #endif
43 #ifdef HAVE_STRING_H
44 #include <string.h>
45 #endif
46 #ifdef HAVE_UNISTD_H
47 #include <unistd.h>
48 #endif
49 
50 #include "archive.h"
51 #include "archive_entry.h"
52 #include "archive_private.h"
53 #include "archive_read_private.h"
54 
55 #define minimum(a, b) (a < b ? a : b)
56 
57 static int	choose_filters(struct archive_read *);
58 static int	choose_format(struct archive_read *);
59 static int	close_filters(struct archive_read *);
60 static int64_t	_archive_filter_bytes(struct archive *, int);
61 static int	_archive_filter_code(struct archive *, int);
62 static const char *_archive_filter_name(struct archive *, int);
63 static int  _archive_filter_count(struct archive *);
64 static int	_archive_read_close(struct archive *);
65 static int	_archive_read_data_block(struct archive *,
66 		    const void **, size_t *, int64_t *);
67 static int	_archive_read_free(struct archive *);
68 static int	_archive_read_next_header(struct archive *,
69 		    struct archive_entry **);
70 static int	_archive_read_next_header2(struct archive *,
71 		    struct archive_entry *);
72 static int64_t  advance_file_pointer(struct archive_read_filter *, int64_t);
73 
74 static const struct archive_vtable
75 archive_read_vtable = {
76 	.archive_filter_bytes = _archive_filter_bytes,
77 	.archive_filter_code = _archive_filter_code,
78 	.archive_filter_name = _archive_filter_name,
79 	.archive_filter_count = _archive_filter_count,
80 	.archive_read_data_block = _archive_read_data_block,
81 	.archive_read_next_header = _archive_read_next_header,
82 	.archive_read_next_header2 = _archive_read_next_header2,
83 	.archive_free = _archive_read_free,
84 	.archive_close = _archive_read_close,
85 };
86 
87 /*
88  * Allocate, initialize and return a struct archive object.
89  */
90 struct archive *
archive_read_new(void)91 archive_read_new(void)
92 {
93 	struct archive_read *a;
94 
95 	a = (struct archive_read *)calloc(1, sizeof(*a));
96 	if (a == NULL)
97 		return (NULL);
98 	a->archive.magic = ARCHIVE_READ_MAGIC;
99 
100 	a->archive.state = ARCHIVE_STATE_NEW;
101 	a->entry = archive_entry_new2(&a->archive);
102 	a->archive.vtable = &archive_read_vtable;
103 
104 	a->passphrases.last = &a->passphrases.first;
105 
106 	return (&a->archive);
107 }
108 
109 /*
110  * Record the do-not-extract-to file. This belongs in archive_read_extract.c.
111  */
112 void
archive_read_extract_set_skip_file(struct archive * _a,la_int64_t d,la_int64_t i)113 archive_read_extract_set_skip_file(struct archive *_a, la_int64_t d,
114     la_int64_t i)
115 {
116 	struct archive_read *a = (struct archive_read *)_a;
117 
118 	if (ARCHIVE_OK != __archive_check_magic(_a, ARCHIVE_READ_MAGIC,
119 		ARCHIVE_STATE_ANY, "archive_read_extract_set_skip_file"))
120 		return;
121 	a->skip_file_set = 1;
122 	a->skip_file_dev = d;
123 	a->skip_file_ino = i;
124 }
125 
126 /*
127  * Open the archive
128  */
129 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)130 archive_read_open(struct archive *a, void *client_data,
131     archive_open_callback *client_opener, archive_read_callback *client_reader,
132     archive_close_callback *client_closer)
133 {
134 	/* Old archive_read_open() is just a thin shell around
135 	 * archive_read_open1. */
136 	archive_read_set_open_callback(a, client_opener);
137 	archive_read_set_read_callback(a, client_reader);
138 	archive_read_set_close_callback(a, client_closer);
139 	archive_read_set_callback_data(a, client_data);
140 	return archive_read_open1(a);
141 }
142 
143 
144 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)145 archive_read_open2(struct archive *a, void *client_data,
146     archive_open_callback *client_opener,
147     archive_read_callback *client_reader,
148     archive_skip_callback *client_skipper,
149     archive_close_callback *client_closer)
150 {
151 	/* Old archive_read_open2() is just a thin shell around
152 	 * archive_read_open1. */
153 	archive_read_set_callback_data(a, client_data);
154 	archive_read_set_open_callback(a, client_opener);
155 	archive_read_set_read_callback(a, client_reader);
156 	archive_read_set_skip_callback(a, client_skipper);
157 	archive_read_set_close_callback(a, client_closer);
158 	return archive_read_open1(a);
159 }
160 
161 static ssize_t
client_read_proxy(struct archive_read_filter * self,const void ** buff)162 client_read_proxy(struct archive_read_filter *self, const void **buff)
163 {
164 	ssize_t r;
165 	r = (self->archive->client.reader)(&self->archive->archive,
166 	    self->data, buff);
167 	return (r);
168 }
169 
170 static int64_t
client_skip_proxy(struct archive_read_filter * self,int64_t request)171 client_skip_proxy(struct archive_read_filter *self, int64_t request)
172 {
173 	if (request < 0)
174 		__archive_errx(1, "Negative skip requested.");
175 	if (request == 0)
176 		return 0;
177 
178 	if (self->archive->client.skipper != NULL) {
179 		/* Seek requests over 1GiB are broken down into
180 		 * multiple seeks.  This avoids overflows when the
181 		 * requests get passed through 32-bit arguments. */
182 		int64_t skip_limit = (int64_t)1 << 30;
183 		int64_t total = 0;
184 		for (;;) {
185 			int64_t get, ask = request;
186 			if (ask > skip_limit)
187 				ask = skip_limit;
188 			get = (self->archive->client.skipper)
189 				(&self->archive->archive, self->data, ask);
190 			total += get;
191 			if (get == 0 || get == request)
192 				return (total);
193 			if (get > request)
194 				return ARCHIVE_FATAL;
195 			request -= get;
196 		}
197 	} else if (self->archive->client.seeker != NULL
198 		&& request > 64 * 1024) {
199 		/* If the client provided a seeker but not a skipper,
200 		 * we can use the seeker to skip forward.
201 		 *
202 		 * Note: This isn't always a good idea.  The client
203 		 * skipper is allowed to skip by less than requested
204 		 * if it needs to maintain block alignment.  The
205 		 * seeker is not allowed to play such games, so using
206 		 * the seeker here may be a performance loss compared
207 		 * to just reading and discarding.  That's why we
208 		 * only do this for skips of over 64k.
209 		 */
210 		int64_t before = self->position;
211 		int64_t after = (self->archive->client.seeker)
212 		    (&self->archive->archive, self->data, request, SEEK_CUR);
213 		if (after != before + request)
214 			return ARCHIVE_FATAL;
215 		return after - before;
216 	}
217 	return 0;
218 }
219 
220 static int64_t
client_seek_proxy(struct archive_read_filter * self,int64_t offset,int whence)221 client_seek_proxy(struct archive_read_filter *self, int64_t offset, int whence)
222 {
223 	/* DO NOT use the skipper here!  If we transparently handled
224 	 * forward seek here by using the skipper, that will break
225 	 * other libarchive code that assumes a successful forward
226 	 * seek means it can also seek backwards.
227 	 */
228 	if (self->archive->client.seeker == NULL) {
229 		archive_set_error(&self->archive->archive, ARCHIVE_ERRNO_MISC,
230 		    "Current client reader does not support seeking a device");
231 		return (ARCHIVE_FAILED);
232 	}
233 	return (self->archive->client.seeker)(&self->archive->archive,
234 	    self->data, offset, whence);
235 }
236 
237 static int
read_client_close_proxy(struct archive_read * a)238 read_client_close_proxy(struct archive_read *a)
239 {
240 	int r = ARCHIVE_OK, r2;
241 	unsigned int i;
242 
243 	if (a->client.closer == NULL)
244 		return (r);
245 	for (i = 0; i < a->client.nodes; i++)
246 	{
247 		r2 = (a->client.closer)
248 			((struct archive *)a, a->client.dataset[i].data);
249 		if (r > r2)
250 			r = r2;
251 	}
252 	return (r);
253 }
254 
255 static int
client_close_proxy(struct archive_read_filter * self)256 client_close_proxy(struct archive_read_filter *self)
257 {
258 	return read_client_close_proxy(self->archive);
259 }
260 
261 static int
client_open_proxy(struct archive_read_filter * self)262 client_open_proxy(struct archive_read_filter *self)
263 {
264   int r = ARCHIVE_OK;
265 	if (self->archive->client.opener != NULL)
266 		r = (self->archive->client.opener)(
267 		    (struct archive *)self->archive, self->data);
268 	return (r);
269 }
270 
271 static int
client_switch_proxy(struct archive_read_filter * self,unsigned int iindex)272 client_switch_proxy(struct archive_read_filter *self, unsigned int iindex)
273 {
274   int r1 = ARCHIVE_OK, r2 = ARCHIVE_OK;
275 	void *data2 = NULL;
276 
277 	/* Don't do anything if already in the specified data node */
278 	if (self->archive->client.cursor == iindex)
279 		return (ARCHIVE_OK);
280 
281 	self->archive->client.cursor = iindex;
282 	data2 = self->archive->client.dataset[self->archive->client.cursor].data;
283 	if (self->archive->client.switcher != NULL)
284 	{
285 		r1 = r2 = (self->archive->client.switcher)
286 			((struct archive *)self->archive, self->data, data2);
287 		self->data = data2;
288 	}
289 	else
290 	{
291 		/* Attempt to call close and open instead */
292 		if (self->archive->client.closer != NULL)
293 			r1 = (self->archive->client.closer)
294 				((struct archive *)self->archive, self->data);
295 		self->data = data2;
296 		r2 = client_open_proxy(self);
297 	}
298 	return (r1 < r2) ? r1 : r2;
299 }
300 
301 int
archive_read_set_open_callback(struct archive * _a,archive_open_callback * client_opener)302 archive_read_set_open_callback(struct archive *_a,
303     archive_open_callback *client_opener)
304 {
305 	struct archive_read *a = (struct archive_read *)_a;
306 	archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
307 	    "archive_read_set_open_callback");
308 	a->client.opener = client_opener;
309 	return ARCHIVE_OK;
310 }
311 
312 int
archive_read_set_read_callback(struct archive * _a,archive_read_callback * client_reader)313 archive_read_set_read_callback(struct archive *_a,
314     archive_read_callback *client_reader)
315 {
316 	struct archive_read *a = (struct archive_read *)_a;
317 	archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
318 	    "archive_read_set_read_callback");
319 	a->client.reader = client_reader;
320 	return ARCHIVE_OK;
321 }
322 
323 int
archive_read_set_skip_callback(struct archive * _a,archive_skip_callback * client_skipper)324 archive_read_set_skip_callback(struct archive *_a,
325     archive_skip_callback *client_skipper)
326 {
327 	struct archive_read *a = (struct archive_read *)_a;
328 	archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
329 	    "archive_read_set_skip_callback");
330 	a->client.skipper = client_skipper;
331 	return ARCHIVE_OK;
332 }
333 
334 int
archive_read_set_seek_callback(struct archive * _a,archive_seek_callback * client_seeker)335 archive_read_set_seek_callback(struct archive *_a,
336     archive_seek_callback *client_seeker)
337 {
338 	struct archive_read *a = (struct archive_read *)_a;
339 	archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
340 	    "archive_read_set_seek_callback");
341 	a->client.seeker = client_seeker;
342 	return ARCHIVE_OK;
343 }
344 
345 int
archive_read_set_close_callback(struct archive * _a,archive_close_callback * client_closer)346 archive_read_set_close_callback(struct archive *_a,
347     archive_close_callback *client_closer)
348 {
349 	struct archive_read *a = (struct archive_read *)_a;
350 	archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
351 	    "archive_read_set_close_callback");
352 	a->client.closer = client_closer;
353 	return ARCHIVE_OK;
354 }
355 
356 int
archive_read_set_switch_callback(struct archive * _a,archive_switch_callback * client_switcher)357 archive_read_set_switch_callback(struct archive *_a,
358     archive_switch_callback *client_switcher)
359 {
360 	struct archive_read *a = (struct archive_read *)_a;
361 	archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
362 	    "archive_read_set_switch_callback");
363 	a->client.switcher = client_switcher;
364 	return ARCHIVE_OK;
365 }
366 
367 int
archive_read_set_callback_data(struct archive * _a,void * client_data)368 archive_read_set_callback_data(struct archive *_a, void *client_data)
369 {
370 	return archive_read_set_callback_data2(_a, client_data, 0);
371 }
372 
373 int
archive_read_set_callback_data2(struct archive * _a,void * client_data,unsigned int iindex)374 archive_read_set_callback_data2(struct archive *_a, void *client_data,
375     unsigned int iindex)
376 {
377 	struct archive_read *a = (struct archive_read *)_a;
378 	archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
379 	    "archive_read_set_callback_data2");
380 
381 	if (a->client.nodes == 0)
382 	{
383 		a->client.dataset = (struct archive_read_data_node *)
384 		    calloc(1, sizeof(*a->client.dataset));
385 		if (a->client.dataset == NULL)
386 		{
387 			archive_set_error(&a->archive, ENOMEM,
388 				"No memory.");
389 			return ARCHIVE_FATAL;
390 		}
391 		a->client.nodes = 1;
392 	}
393 
394 	if (iindex > a->client.nodes - 1)
395 	{
396 		archive_set_error(&a->archive, EINVAL,
397 			"Invalid index specified.");
398 		return ARCHIVE_FATAL;
399 	}
400 	a->client.dataset[iindex].data = client_data;
401 	a->client.dataset[iindex].begin_position = -1;
402 	a->client.dataset[iindex].total_size = -1;
403 	return ARCHIVE_OK;
404 }
405 
406 int
archive_read_add_callback_data(struct archive * _a,void * client_data,unsigned int iindex)407 archive_read_add_callback_data(struct archive *_a, void *client_data,
408     unsigned int iindex)
409 {
410 	struct archive_read *a = (struct archive_read *)_a;
411 	void *p;
412 	unsigned int i;
413 
414 	archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
415 	    "archive_read_add_callback_data");
416 	if (iindex > a->client.nodes) {
417 		archive_set_error(&a->archive, EINVAL,
418 			"Invalid index specified.");
419 		return ARCHIVE_FATAL;
420 	}
421 	p = realloc(a->client.dataset, sizeof(*a->client.dataset)
422 		* (++(a->client.nodes)));
423 	if (p == NULL) {
424 		archive_set_error(&a->archive, ENOMEM,
425 			"No memory.");
426 		return ARCHIVE_FATAL;
427 	}
428 	a->client.dataset = (struct archive_read_data_node *)p;
429 	for (i = a->client.nodes - 1; i > iindex; i--) {
430 		a->client.dataset[i].data = a->client.dataset[i-1].data;
431 		a->client.dataset[i].begin_position = -1;
432 		a->client.dataset[i].total_size = -1;
433 	}
434 	a->client.dataset[iindex].data = client_data;
435 	a->client.dataset[iindex].begin_position = -1;
436 	a->client.dataset[iindex].total_size = -1;
437 	return ARCHIVE_OK;
438 }
439 
440 int
archive_read_append_callback_data(struct archive * _a,void * client_data)441 archive_read_append_callback_data(struct archive *_a, void *client_data)
442 {
443 	struct archive_read *a = (struct archive_read *)_a;
444 	return archive_read_add_callback_data(_a, client_data, a->client.nodes);
445 }
446 
447 int
archive_read_prepend_callback_data(struct archive * _a,void * client_data)448 archive_read_prepend_callback_data(struct archive *_a, void *client_data)
449 {
450 	return archive_read_add_callback_data(_a, client_data, 0);
451 }
452 
453 static const struct archive_read_filter_vtable
454 none_reader_vtable = {
455 	.read = client_read_proxy,
456 	.close = client_close_proxy,
457 };
458 
459 int
archive_read_open1(struct archive * _a)460 archive_read_open1(struct archive *_a)
461 {
462 	struct archive_read *a = (struct archive_read *)_a;
463 	struct archive_read_filter *filter, *tmp;
464 	int slot, e = ARCHIVE_OK;
465 
466 	archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
467 	    "archive_read_open");
468 	archive_clear_error(&a->archive);
469 
470 	if (a->client.reader == NULL) {
471 		archive_set_error(&a->archive, EINVAL,
472 		    "No reader function provided to archive_read_open");
473 		a->archive.state = ARCHIVE_STATE_FATAL;
474 		return (ARCHIVE_FATAL);
475 	}
476 
477 	/* Open data source. */
478 	if (a->client.opener != NULL) {
479 		e = (a->client.opener)(&a->archive, a->client.dataset[0].data);
480 		if (e != 0) {
481 			/* If the open failed, call the closer to clean up. */
482 			read_client_close_proxy(a);
483 			return (e);
484 		}
485 	}
486 
487 	filter = calloc(1, sizeof(*filter));
488 	if (filter == NULL)
489 		return (ARCHIVE_FATAL);
490 	filter->bidder = NULL;
491 	filter->upstream = NULL;
492 	filter->archive = a;
493 	filter->data = a->client.dataset[0].data;
494 	filter->vtable = &none_reader_vtable;
495 	filter->name = "none";
496 	filter->code = ARCHIVE_FILTER_NONE;
497 	filter->can_skip = 1;
498 	filter->can_seek = 1;
499 
500 	a->client.dataset[0].begin_position = 0;
501 	if (!a->filter || !a->bypass_filter_bidding)
502 	{
503 		a->filter = filter;
504 		/* Build out the input pipeline. */
505 		e = choose_filters(a);
506 		if (e < ARCHIVE_WARN) {
507 			a->archive.state = ARCHIVE_STATE_FATAL;
508 			return (ARCHIVE_FATAL);
509 		}
510 	}
511 	else
512 	{
513 		/* Need to add "NONE" type filter at the end of the filter chain */
514 		tmp = a->filter;
515 		while (tmp->upstream)
516 			tmp = tmp->upstream;
517 		tmp->upstream = filter;
518 	}
519 
520 	if (!a->format)
521 	{
522 		slot = choose_format(a);
523 		if (slot < 0) {
524 			close_filters(a);
525 			a->archive.state = ARCHIVE_STATE_FATAL;
526 			return (ARCHIVE_FATAL);
527 		}
528 		a->format = &(a->formats[slot]);
529 	}
530 
531 	a->archive.state = ARCHIVE_STATE_HEADER;
532 
533 	/* Ensure libarchive starts from the first node in a multivolume set */
534 	client_switch_proxy(a->filter, 0);
535 	return (e);
536 }
537 
538 /*
539  * Allow each registered stream transform to bid on whether
540  * it wants to handle this stream.  Repeat until we've finished
541  * building the pipeline.
542  */
543 
544 /* We won't build a filter pipeline with more stages than this. */
545 #define MAX_NUMBER_FILTERS 25
546 
547 static int
choose_filters(struct archive_read * a)548 choose_filters(struct archive_read *a)
549 {
550 	int number_bidders, i, bid, best_bid, number_filters;
551 	struct archive_read_filter_bidder *bidder, *best_bidder;
552 	struct archive_read_filter *filter;
553 	ssize_t avail;
554 	int r;
555 
556 	for (number_filters = 0; number_filters < MAX_NUMBER_FILTERS; ++number_filters) {
557 		number_bidders = sizeof(a->bidders) / sizeof(a->bidders[0]);
558 
559 		best_bid = 0;
560 		best_bidder = NULL;
561 
562 		bidder = a->bidders;
563 		for (i = 0; i < number_bidders; i++, bidder++) {
564 			if (bidder->vtable == NULL)
565 				continue;
566 			bid = (bidder->vtable->bid)(bidder, a->filter);
567 			if (bid > best_bid) {
568 				best_bid = bid;
569 				best_bidder = bidder;
570 			}
571 		}
572 
573 		/* If no bidder, we're done. */
574 		if (best_bidder == NULL) {
575 			/* Verify the filter by asking it for some data. */
576 			__archive_read_filter_ahead(a->filter, 1, &avail);
577 			if (avail < 0) {
578 				__archive_read_free_filters(a);
579 				return (ARCHIVE_FATAL);
580 			}
581 			return (ARCHIVE_OK);
582 		}
583 
584 		filter
585 		    = (struct archive_read_filter *)calloc(1, sizeof(*filter));
586 		if (filter == NULL)
587 			return (ARCHIVE_FATAL);
588 		filter->bidder = best_bidder;
589 		filter->archive = a;
590 		filter->upstream = a->filter;
591 		a->filter = filter;
592 		r = (best_bidder->vtable->init)(a->filter);
593 		if (r != ARCHIVE_OK) {
594 			__archive_read_free_filters(a);
595 			return (ARCHIVE_FATAL);
596 		}
597 	}
598 	archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
599 	    "Input requires too many filters for decoding");
600 	return (ARCHIVE_FATAL);
601 }
602 
603 int
__archive_read_header(struct archive_read * a,struct archive_entry * entry)604 __archive_read_header(struct archive_read *a, struct archive_entry *entry)
605 {
606 	if (!a->filter->vtable->read_header)
607 		return (ARCHIVE_OK);
608 	return a->filter->vtable->read_header(a->filter, entry);
609 }
610 
611 /*
612  * Read header of next entry.
613  */
614 static int
_archive_read_next_header2(struct archive * _a,struct archive_entry * entry)615 _archive_read_next_header2(struct archive *_a, struct archive_entry *entry)
616 {
617 	struct archive_read *a = (struct archive_read *)_a;
618 	int r1 = ARCHIVE_OK, r2;
619 
620 	archive_check_magic(_a, ARCHIVE_READ_MAGIC,
621 	    ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA,
622 	    "archive_read_next_header");
623 
624 	archive_entry_clear(entry);
625 	archive_clear_error(&a->archive);
626 
627 	/*
628 	 * If client didn't consume entire data, skip any remainder
629 	 * (This is especially important for GNU incremental directories.)
630 	 */
631 	if (a->archive.state == ARCHIVE_STATE_DATA) {
632 		r1 = archive_read_data_skip(&a->archive);
633 		if (r1 == ARCHIVE_EOF)
634 			archive_set_error(&a->archive, EIO,
635 			    "Premature end-of-file.");
636 		if (r1 == ARCHIVE_EOF || r1 == ARCHIVE_FATAL) {
637 			a->archive.state = ARCHIVE_STATE_FATAL;
638 			return (ARCHIVE_FATAL);
639 		}
640 	}
641 
642 	/* Record start-of-header offset in uncompressed stream. */
643 	a->header_position = a->filter->position;
644 
645 	++_a->file_count;
646 	r2 = (a->format->read_header)(a, entry);
647 
648 	/*
649 	 * EOF and FATAL are persistent at this layer.  By
650 	 * modifying the state, we guarantee that future calls to
651 	 * read a header or read data will fail.
652 	 */
653 	switch (r2) {
654 	case ARCHIVE_EOF:
655 		a->archive.state = ARCHIVE_STATE_EOF;
656 		--_a->file_count;/* Revert a file counter. */
657 		break;
658 	case ARCHIVE_OK:
659 		a->archive.state = ARCHIVE_STATE_DATA;
660 		break;
661 	case ARCHIVE_WARN:
662 		a->archive.state = ARCHIVE_STATE_DATA;
663 		break;
664 	case ARCHIVE_RETRY:
665 		break;
666 	case ARCHIVE_FATAL:
667 		a->archive.state = ARCHIVE_STATE_FATAL;
668 		break;
669 	}
670 
671 	__archive_reset_read_data(&a->archive);
672 
673 	a->data_start_node = a->client.cursor;
674 	/* EOF always wins; otherwise return the worst error. */
675 	return (r2 < r1 || r2 == ARCHIVE_EOF) ? r2 : r1;
676 }
677 
678 static int
_archive_read_next_header(struct archive * _a,struct archive_entry ** entryp)679 _archive_read_next_header(struct archive *_a, struct archive_entry **entryp)
680 {
681 	int ret;
682 	struct archive_read *a = (struct archive_read *)_a;
683 	*entryp = NULL;
684 	ret = _archive_read_next_header2(_a, a->entry);
685 	*entryp = a->entry;
686 	return ret;
687 }
688 
689 /*
690  * Allow each registered format to bid on whether it wants to handle
691  * the next entry.  Return index of winning bidder.
692  */
693 static int
choose_format(struct archive_read * a)694 choose_format(struct archive_read *a)
695 {
696 	int slots;
697 	int i;
698 	int bid, best_bid;
699 	int best_bid_slot;
700 
701 	slots = sizeof(a->formats) / sizeof(a->formats[0]);
702 	best_bid = -1;
703 	best_bid_slot = -1;
704 
705 	/* Set up a->format for convenience of bidders. */
706 	a->format = &(a->formats[0]);
707 	for (i = 0; i < slots; i++, a->format++) {
708 		if (a->format->bid) {
709 			bid = (a->format->bid)(a, best_bid);
710 			if (bid == ARCHIVE_FATAL)
711 				return (ARCHIVE_FATAL);
712 			if (a->filter->position != 0)
713 				__archive_read_seek(a, 0, SEEK_SET);
714 			if ((bid > best_bid) || (best_bid_slot < 0)) {
715 				best_bid = bid;
716 				best_bid_slot = i;
717 			}
718 		}
719 	}
720 
721 	/*
722 	 * There were no bidders; this is a serious programmer error
723 	 * and demands a quick and definitive abort.
724 	 */
725 	if (best_bid_slot < 0) {
726 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
727 		    "No formats registered");
728 		return (ARCHIVE_FATAL);
729 	}
730 
731 	/*
732 	 * There were bidders, but no non-zero bids; this means we
733 	 * can't support this stream.
734 	 */
735 	if (best_bid < 1) {
736 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
737 		    "Unrecognized archive format");
738 		return (ARCHIVE_FATAL);
739 	}
740 
741 	return (best_bid_slot);
742 }
743 
744 /*
745  * Return the file offset (within the uncompressed data stream) where
746  * the last header started.
747  */
748 la_int64_t
archive_read_header_position(struct archive * _a)749 archive_read_header_position(struct archive *_a)
750 {
751 	struct archive_read *a = (struct archive_read *)_a;
752 	archive_check_magic(_a, ARCHIVE_READ_MAGIC,
753 	    ARCHIVE_STATE_ANY, "archive_read_header_position");
754 	return (a->header_position);
755 }
756 
757 /*
758  * Returns 1 if the archive contains at least one encrypted entry.
759  * If the archive format not support encryption at all
760  * ARCHIVE_READ_FORMAT_ENCRYPTION_UNSUPPORTED is returned.
761  * If for any other reason (e.g. not enough data read so far)
762  * we cannot say whether there are encrypted entries, then
763  * ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW is returned.
764  * In general, this function will return values below zero when the
765  * reader is uncertain or totally incapable of encryption support.
766  * When this function returns 0 you can be sure that the reader
767  * supports encryption detection but no encrypted entries have
768  * been found yet.
769  *
770  * NOTE: If the metadata/header of an archive is also encrypted, you
771  * cannot rely on the number of encrypted entries. That is why this
772  * function does not return the number of encrypted entries but#
773  * just shows that there are some.
774  */
775 int
archive_read_has_encrypted_entries(struct archive * _a)776 archive_read_has_encrypted_entries(struct archive *_a)
777 {
778 	struct archive_read *a = (struct archive_read *)_a;
779 	int format_supports_encryption = archive_read_format_capabilities(_a)
780 			& (ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_DATA | ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_METADATA);
781 
782 	if (!_a || !format_supports_encryption) {
783 		/* Format in general doesn't support encryption */
784 		return ARCHIVE_READ_FORMAT_ENCRYPTION_UNSUPPORTED;
785 	}
786 
787 	/* A reader potentially has read enough data now. */
788 	if (a->format && a->format->has_encrypted_entries) {
789 		return (a->format->has_encrypted_entries)(a);
790 	}
791 
792 	/* For any other reason we cannot say how many entries are there. */
793 	return ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW;
794 }
795 
796 /*
797  * Returns a bitmask of capabilities that are supported by the archive format reader.
798  * If the reader has no special capabilities, ARCHIVE_READ_FORMAT_CAPS_NONE is returned.
799  */
800 int
archive_read_format_capabilities(struct archive * _a)801 archive_read_format_capabilities(struct archive *_a)
802 {
803 	struct archive_read *a = (struct archive_read *)_a;
804 	if (a && a->format && a->format->format_capabilties) {
805 		return (a->format->format_capabilties)(a);
806 	}
807 	return ARCHIVE_READ_FORMAT_CAPS_NONE;
808 }
809 
810 /*
811  * Read data from an archive entry, using a read(2)-style interface.
812  * This is a convenience routine that just calls
813  * archive_read_data_block and copies the results into the client
814  * buffer, filling any gaps with zero bytes.  Clients using this
815  * API can be completely ignorant of sparse-file issues; sparse files
816  * will simply be padded with nulls.
817  *
818  * DO NOT intermingle calls to this function and archive_read_data_block
819  * to read a single entry body.
820  */
821 la_ssize_t
archive_read_data(struct archive * _a,void * buff,size_t s)822 archive_read_data(struct archive *_a, void *buff, size_t s)
823 {
824 	struct archive *a = (struct archive *)_a;
825 	char	*dest;
826 	const void *read_buf;
827 	size_t	 bytes_read;
828 	size_t	 len;
829 	int	 r;
830 
831 	bytes_read = 0;
832 	dest = (char *)buff;
833 
834 	while (s > 0) {
835 		if (a->read_data_offset == a->read_data_output_offset &&
836 		    a->read_data_remaining == 0) {
837 			read_buf = a->read_data_block;
838 			a->read_data_is_posix_read = 1;
839 			a->read_data_requested = s;
840 			r = archive_read_data_block(a, &read_buf,
841 			    &a->read_data_remaining, &a->read_data_offset);
842 			a->read_data_block = read_buf;
843 			if (r == ARCHIVE_EOF)
844 				return (bytes_read);
845 			/*
846 			 * Error codes are all negative, so the status
847 			 * return here cannot be confused with a valid
848 			 * byte count.  (ARCHIVE_OK is zero.)
849 			 */
850 			if (r < ARCHIVE_OK)
851 				return (r);
852 		}
853 
854 		if (a->read_data_offset < a->read_data_output_offset) {
855 			archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
856 			    "Encountered out-of-order sparse blocks");
857 			return (ARCHIVE_RETRY);
858 		}
859 
860 		/* Compute the amount of zero padding needed. */
861 		if (a->read_data_output_offset + (int64_t)s <
862 		    a->read_data_offset) {
863 			len = s;
864 		} else if (a->read_data_output_offset <
865 		    a->read_data_offset) {
866 			len = (size_t)(a->read_data_offset -
867 			    a->read_data_output_offset);
868 		} else
869 			len = 0;
870 
871 		/* Add zeroes. */
872 		memset(dest, 0, len);
873 		s -= len;
874 		a->read_data_output_offset += len;
875 		dest += len;
876 		bytes_read += len;
877 
878 		/* Copy data if there is any space left. */
879 		if (s > 0) {
880 			len = a->read_data_remaining;
881 			if (len > s)
882 				len = s;
883 			if (len) {
884 				memcpy(dest, a->read_data_block, len);
885 				s -= len;
886 				a->read_data_block += len;
887 				a->read_data_remaining -= len;
888 				a->read_data_output_offset += len;
889 				a->read_data_offset += len;
890 				dest += len;
891 				bytes_read += len;
892 			}
893 		}
894 	}
895 	a->read_data_is_posix_read = 0;
896 	a->read_data_requested = 0;
897 	return (bytes_read);
898 }
899 
900 /*
901  * Reset the read_data_* variables, used for starting a new entry.
902  */
__archive_reset_read_data(struct archive * a)903 void __archive_reset_read_data(struct archive * a)
904 {
905 	a->read_data_output_offset = 0;
906 	a->read_data_remaining = 0;
907 	a->read_data_is_posix_read = 0;
908 	a->read_data_requested = 0;
909 
910    /* extra resets, from rar.c */
911    a->read_data_block = NULL;
912    a->read_data_offset = 0;
913 }
914 
915 /*
916  * Skip over all remaining data in this entry.
917  */
918 int
archive_read_data_skip(struct archive * _a)919 archive_read_data_skip(struct archive *_a)
920 {
921 	struct archive_read *a = (struct archive_read *)_a;
922 	int r;
923 	const void *buff;
924 	size_t size;
925 	int64_t offset;
926 
927 	archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA,
928 	    "archive_read_data_skip");
929 
930 	if (a->format->read_data_skip != NULL)
931 		r = (a->format->read_data_skip)(a);
932 	else {
933 		while ((r = archive_read_data_block(&a->archive,
934 			    &buff, &size, &offset))
935 		    == ARCHIVE_OK)
936 			;
937 	}
938 
939 	if (r == ARCHIVE_EOF)
940 		r = ARCHIVE_OK;
941 
942 	a->archive.state = ARCHIVE_STATE_HEADER;
943 	return (r);
944 }
945 
946 la_int64_t
archive_seek_data(struct archive * _a,int64_t offset,int whence)947 archive_seek_data(struct archive *_a, int64_t offset, int whence)
948 {
949 	struct archive_read *a = (struct archive_read *)_a;
950 	archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA,
951 	    "archive_seek_data_block");
952 
953 	if (a->format->seek_data == NULL) {
954 		archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
955 		    "Internal error: "
956 		    "No format_seek_data_block function registered");
957 		return (ARCHIVE_FATAL);
958 	}
959 
960 	return (a->format->seek_data)(a, offset, whence);
961 }
962 
963 /*
964  * Read the next block of entry data from the archive.
965  * This is a zero-copy interface; the client receives a pointer,
966  * size, and file offset of the next available block of data.
967  *
968  * Returns ARCHIVE_OK if the operation is successful, ARCHIVE_EOF if
969  * the end of entry is encountered.
970  */
971 static int
_archive_read_data_block(struct archive * _a,const void ** buff,size_t * size,int64_t * offset)972 _archive_read_data_block(struct archive *_a,
973     const void **buff, size_t *size, int64_t *offset)
974 {
975 	struct archive_read *a = (struct archive_read *)_a;
976 	archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA,
977 	    "archive_read_data_block");
978 
979 	if (a->format->read_data == NULL) {
980 		archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
981 		    "Internal error: "
982 		    "No format->read_data function registered");
983 		return (ARCHIVE_FATAL);
984 	}
985 
986 	return (a->format->read_data)(a, buff, size, offset);
987 }
988 
989 static int
close_filters(struct archive_read * a)990 close_filters(struct archive_read *a)
991 {
992 	struct archive_read_filter *f = a->filter;
993 	int r = ARCHIVE_OK;
994 	/* Close each filter in the pipeline. */
995 	while (f != NULL) {
996 		struct archive_read_filter *t = f->upstream;
997 		if (!f->closed && f->vtable != NULL) {
998 			int r1 = (f->vtable->close)(f);
999 			f->closed = 1;
1000 			if (r1 < r)
1001 				r = r1;
1002 		}
1003 		free(f->buffer);
1004 		f->buffer = NULL;
1005 		f = t;
1006 	}
1007 	return r;
1008 }
1009 
1010 void
__archive_read_free_filters(struct archive_read * a)1011 __archive_read_free_filters(struct archive_read *a)
1012 {
1013 	/* Make sure filters are closed and their buffers are freed */
1014 	close_filters(a);
1015 
1016 	while (a->filter != NULL) {
1017 		struct archive_read_filter *t = a->filter->upstream;
1018 		free(a->filter);
1019 		a->filter = t;
1020 	}
1021 }
1022 
1023 /*
1024  * return the count of # of filters in use
1025  */
1026 static int
_archive_filter_count(struct archive * _a)1027 _archive_filter_count(struct archive *_a)
1028 {
1029 	struct archive_read *a = (struct archive_read *)_a;
1030 	struct archive_read_filter *p = a->filter;
1031 	int count = 0;
1032 	while(p) {
1033 		count++;
1034 		p = p->upstream;
1035 	}
1036 	return count;
1037 }
1038 
1039 /*
1040  * Close the file and all I/O.
1041  */
1042 static int
_archive_read_close(struct archive * _a)1043 _archive_read_close(struct archive *_a)
1044 {
1045 	struct archive_read *a = (struct archive_read *)_a;
1046 	int r = ARCHIVE_OK, r1 = ARCHIVE_OK;
1047 
1048 	archive_check_magic(&a->archive, ARCHIVE_READ_MAGIC,
1049 	    ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_close");
1050 	if (a->archive.state == ARCHIVE_STATE_CLOSED)
1051 		return (ARCHIVE_OK);
1052 	archive_clear_error(&a->archive);
1053 	a->archive.state = ARCHIVE_STATE_CLOSED;
1054 
1055 	/* TODO: Clean up the formatters. */
1056 
1057 	/* Release the filter objects. */
1058 	r1 = close_filters(a);
1059 	if (r1 < r)
1060 		r = r1;
1061 
1062 	return (r);
1063 }
1064 
1065 /*
1066  * Release memory and other resources.
1067  */
1068 static int
_archive_read_free(struct archive * _a)1069 _archive_read_free(struct archive *_a)
1070 {
1071 	struct archive_read *a = (struct archive_read *)_a;
1072 	struct archive_read_passphrase *p;
1073 	int i, n;
1074 	int slots;
1075 	int r = ARCHIVE_OK;
1076 
1077 	if (_a == NULL)
1078 		return (ARCHIVE_OK);
1079 	archive_check_magic(_a, ARCHIVE_READ_MAGIC,
1080 	    ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_free");
1081 	if (a->archive.state != ARCHIVE_STATE_CLOSED
1082 	    && a->archive.state != ARCHIVE_STATE_FATAL)
1083 		r = archive_read_close(&a->archive);
1084 
1085 	/* Call cleanup functions registered by optional components. */
1086 	if (a->cleanup_archive_extract != NULL)
1087 		r = (a->cleanup_archive_extract)(a);
1088 
1089 	/* Cleanup format-specific data. */
1090 	slots = sizeof(a->formats) / sizeof(a->formats[0]);
1091 	for (i = 0; i < slots; i++) {
1092 		a->format = &(a->formats[i]);
1093 		if (a->formats[i].cleanup)
1094 			(a->formats[i].cleanup)(a);
1095 	}
1096 
1097 	/* Free the filters */
1098 	__archive_read_free_filters(a);
1099 
1100 	/* Release the bidder objects. */
1101 	n = sizeof(a->bidders)/sizeof(a->bidders[0]);
1102 	for (i = 0; i < n; i++) {
1103 		if (a->bidders[i].vtable == NULL ||
1104 		    a->bidders[i].vtable->free == NULL)
1105 			continue;
1106 		(a->bidders[i].vtable->free)(&a->bidders[i]);
1107 	}
1108 
1109 	/* Release passphrase list. */
1110 	p = a->passphrases.first;
1111 	while (p != NULL) {
1112 		struct archive_read_passphrase *np = p->next;
1113 
1114 		/* A passphrase should be cleaned. */
1115 		memset(p->passphrase, 0, strlen(p->passphrase));
1116 		free(p->passphrase);
1117 		free(p);
1118 		p = np;
1119 	}
1120 
1121 	archive_string_free(&a->archive.error_string);
1122 	archive_entry_free(a->entry);
1123 	a->archive.magic = 0;
1124 	__archive_clean(&a->archive);
1125 	free(a->client.dataset);
1126 	free(a);
1127 	return (r);
1128 }
1129 
1130 static struct archive_read_filter *
get_filter(struct archive * _a,int n)1131 get_filter(struct archive *_a, int n)
1132 {
1133 	struct archive_read *a = (struct archive_read *)_a;
1134 	struct archive_read_filter *f = a->filter;
1135 	/* We use n == -1 for 'the last filter', which is always the
1136 	 * client proxy. */
1137 	if (n == -1 && f != NULL) {
1138 		struct archive_read_filter *last = f;
1139 		f = f->upstream;
1140 		while (f != NULL) {
1141 			last = f;
1142 			f = f->upstream;
1143 		}
1144 		return (last);
1145 	}
1146 	if (n < 0)
1147 		return NULL;
1148 	while (n > 0 && f != NULL) {
1149 		f = f->upstream;
1150 		--n;
1151 	}
1152 	return (f);
1153 }
1154 
1155 static int
_archive_filter_code(struct archive * _a,int n)1156 _archive_filter_code(struct archive *_a, int n)
1157 {
1158 	struct archive_read_filter *f = get_filter(_a, n);
1159 	return f == NULL ? -1 : f->code;
1160 }
1161 
1162 static const char *
_archive_filter_name(struct archive * _a,int n)1163 _archive_filter_name(struct archive *_a, int n)
1164 {
1165 	struct archive_read_filter *f = get_filter(_a, n);
1166 	return f != NULL ? f->name : NULL;
1167 }
1168 
1169 static int64_t
_archive_filter_bytes(struct archive * _a,int n)1170 _archive_filter_bytes(struct archive *_a, int n)
1171 {
1172 	struct archive_read_filter *f = get_filter(_a, n);
1173 	return f == NULL ? -1 : f->position;
1174 }
1175 
1176 /*
1177  * Used internally by read format handlers to register their bid and
1178  * initialization functions.
1179  */
1180 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 *))1181 __archive_read_register_format(struct archive_read *a,
1182     void *format_data,
1183     const char *name,
1184     int (*bid)(struct archive_read *, int),
1185     int (*options)(struct archive_read *, const char *, const char *),
1186     int (*read_header)(struct archive_read *, struct archive_entry *),
1187     int (*read_data)(struct archive_read *, const void **, size_t *, int64_t *),
1188     int (*read_data_skip)(struct archive_read *),
1189     int64_t (*seek_data)(struct archive_read *, int64_t, int),
1190     int (*cleanup)(struct archive_read *),
1191     int (*format_capabilities)(struct archive_read *),
1192     int (*has_encrypted_entries)(struct archive_read *))
1193 {
1194 	int i, number_slots;
1195 
1196 	archive_check_magic(&a->archive,
1197 	    ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
1198 	    "__archive_read_register_format");
1199 
1200 	number_slots = sizeof(a->formats) / sizeof(a->formats[0]);
1201 
1202 	for (i = 0; i < number_slots; i++) {
1203 		if (a->formats[i].bid == bid)
1204 			return (ARCHIVE_WARN); /* We've already installed */
1205 		if (a->formats[i].bid == NULL) {
1206 			a->formats[i].bid = bid;
1207 			a->formats[i].options = options;
1208 			a->formats[i].read_header = read_header;
1209 			a->formats[i].read_data = read_data;
1210 			a->formats[i].read_data_skip = read_data_skip;
1211 			a->formats[i].seek_data = seek_data;
1212 			a->formats[i].cleanup = cleanup;
1213 			a->formats[i].data = format_data;
1214 			a->formats[i].name = name;
1215 			a->formats[i].format_capabilties = format_capabilities;
1216 			a->formats[i].has_encrypted_entries = has_encrypted_entries;
1217 			return (ARCHIVE_OK);
1218 		}
1219 	}
1220 
1221 	archive_set_error(&a->archive, ENOMEM,
1222 	    "Not enough slots for format registration");
1223 	return (ARCHIVE_FATAL);
1224 }
1225 
1226 /*
1227  * Used internally by decompression routines to register their bid and
1228  * initialization functions.
1229  */
1230 int
__archive_read_register_bidder(struct archive_read * a,void * bidder_data,const char * name,const struct archive_read_filter_bidder_vtable * vtable)1231 __archive_read_register_bidder(struct archive_read *a,
1232 	void *bidder_data,
1233 	const char *name,
1234 	const struct archive_read_filter_bidder_vtable *vtable)
1235 {
1236 	struct archive_read_filter_bidder *bidder;
1237 	int i, number_slots;
1238 
1239 	archive_check_magic(&a->archive, ARCHIVE_READ_MAGIC,
1240 	    ARCHIVE_STATE_NEW, "__archive_read_register_bidder");
1241 
1242 	number_slots = sizeof(a->bidders) / sizeof(a->bidders[0]);
1243 
1244 	for (i = 0; i < number_slots; i++) {
1245 		if (a->bidders[i].vtable != NULL)
1246 			continue;
1247 		memset(a->bidders + i, 0, sizeof(a->bidders[0]));
1248 		bidder = (a->bidders + i);
1249 		bidder->data = bidder_data;
1250 		bidder->name = name;
1251 		bidder->vtable = vtable;
1252 		if (bidder->vtable->bid == NULL || bidder->vtable->init == NULL) {
1253 			archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
1254 					"Internal error: "
1255 					"no bid/init for filter bidder");
1256 			return (ARCHIVE_FATAL);
1257 		}
1258 
1259 		return (ARCHIVE_OK);
1260 	}
1261 
1262 	archive_set_error(&a->archive, ENOMEM,
1263 	    "Not enough slots for filter registration");
1264 	return (ARCHIVE_FATAL);
1265 }
1266 
1267 /*
1268  * The next section implements the peek/consume internal I/O
1269  * system used by archive readers.  This system allows simple
1270  * read-ahead for consumers while preserving zero-copy operation
1271  * most of the time.
1272  *
1273  * The two key operations:
1274  *  * The read-ahead function returns a pointer to a block of data
1275  *    that satisfies a minimum request.
1276  *  * The consume function advances the file pointer.
1277  *
1278  * In the ideal case, filters generate blocks of data
1279  * and __archive_read_ahead() just returns pointers directly into
1280  * those blocks.  Then __archive_read_consume() just bumps those
1281  * pointers.  Only if your request would span blocks does the I/O
1282  * layer use a copy buffer to provide you with a contiguous block of
1283  * data.
1284  *
1285  * A couple of useful idioms:
1286  *  * "I just want some data."  Ask for 1 byte and pay attention to
1287  *    the "number of bytes available" from __archive_read_ahead().
1288  *    Consume whatever you actually use.
1289  *  * "I want to output a large block of data."  As above, ask for 1 byte,
1290  *    emit all that's available (up to whatever limit you have), consume
1291  *    it all, then repeat until you're done.  This effectively means that
1292  *    you're passing along the blocks that came from your provider.
1293  *  * "I want to peek ahead by a large amount."  Ask for 4k or so, then
1294  *    double and repeat until you get an error or have enough.  Note
1295  *    that the I/O layer will likely end up expanding its copy buffer
1296  *    to fit your request, so use this technique cautiously.  This
1297  *    technique is used, for example, by some of the format tasting
1298  *    code that has uncertain look-ahead needs.
1299  */
1300 
1301 /*
1302  * Looks ahead in the input stream:
1303  *  * If 'avail' pointer is provided, that returns number of bytes available
1304  *    in the current buffer, which may be much larger than requested.
1305  *  * If end-of-file, *avail gets set to zero.
1306  *  * If error, *avail gets error code.
1307  *  * If request can be met, returns pointer to data.
1308  *  * If minimum request cannot be met, returns NULL.
1309  *
1310  * Note: If you just want "some data", ask for 1 byte and pay attention
1311  * to *avail, which will have the actual amount available.  If you
1312  * know exactly how many bytes you need, just ask for that and treat
1313  * a NULL return as an error.
1314  *
1315  * Important:  This does NOT move the file pointer.  See
1316  * __archive_read_consume() below.
1317  */
1318 const void *
__archive_read_ahead(struct archive_read * a,size_t min,ssize_t * avail)1319 __archive_read_ahead(struct archive_read *a, size_t min, ssize_t *avail)
1320 {
1321 	return (__archive_read_filter_ahead(a->filter, min, avail));
1322 }
1323 
1324 const void *
__archive_read_filter_ahead(struct archive_read_filter * filter,size_t min,ssize_t * avail)1325 __archive_read_filter_ahead(struct archive_read_filter *filter,
1326     size_t min, ssize_t *avail)
1327 {
1328 	ssize_t bytes_read;
1329 	size_t tocopy;
1330 
1331 	if (filter->fatal) {
1332 		if (avail)
1333 			*avail = ARCHIVE_FATAL;
1334 		return (NULL);
1335 	}
1336 
1337 	/*
1338 	 * Keep pulling more data until we can satisfy the request.
1339 	 */
1340 	for (;;) {
1341 
1342 		/*
1343 		 * If we can satisfy from the copy buffer (and the
1344 		 * copy buffer isn't empty), we're done.  In particular,
1345 		 * note that min == 0 is a perfectly well-defined
1346 		 * request.
1347 		 */
1348 		if (filter->avail >= min && filter->avail > 0) {
1349 			if (avail != NULL)
1350 				*avail = filter->avail;
1351 			return (filter->next);
1352 		}
1353 
1354 		/*
1355 		 * We can satisfy directly from client buffer if everything
1356 		 * currently in the copy buffer is still in the client buffer.
1357 		 */
1358 		if (filter->client_total >= filter->client_avail + filter->avail
1359 		    && filter->client_avail + filter->avail >= min) {
1360 			/* "Roll back" to client buffer. */
1361 			filter->client_avail += filter->avail;
1362 			filter->client_next -= filter->avail;
1363 			/* Copy buffer is now empty. */
1364 			filter->avail = 0;
1365 			filter->next = filter->buffer;
1366 			/* Return data from client buffer. */
1367 			if (avail != NULL)
1368 				*avail = filter->client_avail;
1369 			return (filter->client_next);
1370 		}
1371 
1372 		/* Move data forward in copy buffer if necessary. */
1373 		if (filter->next > filter->buffer &&
1374 		    filter->next + min > filter->buffer + filter->buffer_size) {
1375 			if (filter->avail > 0)
1376 				memmove(filter->buffer, filter->next,
1377 				    filter->avail);
1378 			filter->next = filter->buffer;
1379 		}
1380 
1381 		/* If we've used up the client data, get more. */
1382 		if (filter->client_avail <= 0) {
1383 			if (filter->end_of_file) {
1384 				if (avail != NULL)
1385 					*avail = 0;
1386 				return (NULL);
1387 			}
1388 			bytes_read = (filter->vtable->read)(filter,
1389 			    &filter->client_buff);
1390 			if (bytes_read < 0) {		/* Read error. */
1391 				filter->client_total = filter->client_avail = 0;
1392 				filter->client_next =
1393 				    filter->client_buff = NULL;
1394 				filter->fatal = 1;
1395 				if (avail != NULL)
1396 					*avail = ARCHIVE_FATAL;
1397 				return (NULL);
1398 			}
1399 			if (bytes_read == 0) {
1400 				/* Check for another client object first */
1401 				if (filter->archive->client.cursor !=
1402 				      filter->archive->client.nodes - 1) {
1403 					if (client_switch_proxy(filter,
1404 					    filter->archive->client.cursor + 1)
1405 					    == ARCHIVE_OK)
1406 						continue;
1407 				}
1408 				/* Premature end-of-file. */
1409 				filter->client_total = filter->client_avail = 0;
1410 				filter->client_next =
1411 				    filter->client_buff = NULL;
1412 				filter->end_of_file = 1;
1413 				/* Return whatever we do have. */
1414 				if (avail != NULL)
1415 					*avail = filter->avail;
1416 				return (NULL);
1417 			}
1418 			filter->client_total = bytes_read;
1419 			filter->client_avail = filter->client_total;
1420 			filter->client_next = filter->client_buff;
1421 		} else {
1422 			/*
1423 			 * We can't satisfy the request from the copy
1424 			 * buffer or the existing client data, so we
1425 			 * need to copy more client data over to the
1426 			 * copy buffer.
1427 			 */
1428 
1429 			/* Ensure the buffer is big enough. */
1430 			if (min > filter->buffer_size) {
1431 				size_t s, t;
1432 				char *p;
1433 
1434 				/* Double the buffer; watch for overflow. */
1435 				s = t = filter->buffer_size;
1436 				if (s == 0)
1437 					s = min;
1438 				while (s < min) {
1439 					t *= 2;
1440 					if (t <= s) { /* Integer overflow! */
1441 						archive_set_error(
1442 						    &filter->archive->archive,
1443 						    ENOMEM,
1444 						    "Unable to allocate copy"
1445 						    " buffer");
1446 						filter->fatal = 1;
1447 						if (avail != NULL)
1448 							*avail = ARCHIVE_FATAL;
1449 						return (NULL);
1450 					}
1451 					s = t;
1452 				}
1453 				/* Now s >= min, so allocate a new buffer. */
1454 				p = (char *)malloc(s);
1455 				if (p == NULL) {
1456 					archive_set_error(
1457 						&filter->archive->archive,
1458 						ENOMEM,
1459 					    "Unable to allocate copy buffer");
1460 					filter->fatal = 1;
1461 					if (avail != NULL)
1462 						*avail = ARCHIVE_FATAL;
1463 					return (NULL);
1464 				}
1465 				/* Move data into newly-enlarged buffer. */
1466 				if (filter->avail > 0)
1467 					memmove(p, filter->next, filter->avail);
1468 				free(filter->buffer);
1469 				filter->next = filter->buffer = p;
1470 				filter->buffer_size = s;
1471 			}
1472 
1473 			/* We can add client data to copy buffer. */
1474 			/* First estimate: copy to fill rest of buffer. */
1475 			tocopy = (filter->buffer + filter->buffer_size)
1476 			    - (filter->next + filter->avail);
1477 			/* Don't waste time buffering more than we need to. */
1478 			if (tocopy + filter->avail > min)
1479 				tocopy = min - filter->avail;
1480 			/* Don't copy more than is available. */
1481 			if (tocopy > filter->client_avail)
1482 				tocopy = filter->client_avail;
1483 
1484 			memcpy(filter->next + filter->avail,
1485 			    filter->client_next, tocopy);
1486 			/* Remove this data from client buffer. */
1487 			filter->client_next += tocopy;
1488 			filter->client_avail -= tocopy;
1489 			/* add it to copy buffer. */
1490 			filter->avail += tocopy;
1491 		}
1492 	}
1493 }
1494 
1495 /*
1496  * Move the file pointer forward.
1497  */
1498 int64_t
__archive_read_consume(struct archive_read * a,int64_t request)1499 __archive_read_consume(struct archive_read *a, int64_t request)
1500 {
1501 	return (__archive_read_filter_consume(a->filter, request));
1502 }
1503 
1504 int64_t
__archive_read_filter_consume(struct archive_read_filter * filter,int64_t request)1505 __archive_read_filter_consume(struct archive_read_filter * filter,
1506     int64_t request)
1507 {
1508 	int64_t skipped;
1509 
1510 	if (request < 0)
1511 		return ARCHIVE_FATAL;
1512 	if (request == 0)
1513 		return 0;
1514 
1515 	skipped = advance_file_pointer(filter, request);
1516 	if (skipped == request)
1517 		return (skipped);
1518 	/* We hit EOF before we satisfied the skip request. */
1519 	if (skipped < 0)  /* Map error code to 0 for error message below. */
1520 		skipped = 0;
1521 	archive_set_error(&filter->archive->archive,
1522 	    ARCHIVE_ERRNO_MISC,
1523 	    "Truncated input file (needed %jd bytes, only %jd available)",
1524 	    (intmax_t)request, (intmax_t)skipped);
1525 	return (ARCHIVE_FATAL);
1526 }
1527 
1528 /*
1529  * Advance the file pointer by the amount requested.
1530  * Returns the amount actually advanced, which may be less than the
1531  * request if EOF is encountered first.
1532  * Returns a negative value if there's an I/O error.
1533  */
1534 static int64_t
advance_file_pointer(struct archive_read_filter * filter,int64_t request)1535 advance_file_pointer(struct archive_read_filter *filter, int64_t request)
1536 {
1537 	int64_t bytes_skipped, total_bytes_skipped = 0;
1538 	ssize_t bytes_read;
1539 	size_t min;
1540 
1541 	if (filter->fatal)
1542 		return (-1);
1543 
1544 	/* Use up the copy buffer first. */
1545 	if (filter->avail > 0) {
1546 		min = (size_t)minimum(request, (int64_t)filter->avail);
1547 		filter->next += min;
1548 		filter->avail -= min;
1549 		request -= min;
1550 		filter->position += min;
1551 		total_bytes_skipped += min;
1552 	}
1553 
1554 	/* Then use up the client buffer. */
1555 	if (filter->client_avail > 0) {
1556 		min = (size_t)minimum(request, (int64_t)filter->client_avail);
1557 		filter->client_next += min;
1558 		filter->client_avail -= min;
1559 		request -= min;
1560 		filter->position += min;
1561 		total_bytes_skipped += min;
1562 	}
1563 	if (request == 0)
1564 		return (total_bytes_skipped);
1565 
1566 	/* If there's an optimized skip function, use it. */
1567 	if (filter->can_skip != 0) {
1568 		bytes_skipped = client_skip_proxy(filter, request);
1569 		if (bytes_skipped < 0) {	/* error */
1570 			filter->fatal = 1;
1571 			return (bytes_skipped);
1572 		}
1573 		filter->position += bytes_skipped;
1574 		total_bytes_skipped += bytes_skipped;
1575 		request -= bytes_skipped;
1576 		if (request == 0)
1577 			return (total_bytes_skipped);
1578 	}
1579 
1580 	/* Use ordinary reads as necessary to complete the request. */
1581 	for (;;) {
1582 		bytes_read = (filter->vtable->read)(filter, &filter->client_buff);
1583 		if (bytes_read < 0) {
1584 			filter->client_buff = NULL;
1585 			filter->fatal = 1;
1586 			return (bytes_read);
1587 		}
1588 
1589 		if (bytes_read == 0) {
1590 			if (filter->archive->client.cursor !=
1591 			      filter->archive->client.nodes - 1) {
1592 				if (client_switch_proxy(filter,
1593 				    filter->archive->client.cursor + 1)
1594 				    == ARCHIVE_OK)
1595 					continue;
1596 			}
1597 			filter->client_buff = NULL;
1598 			filter->end_of_file = 1;
1599 			return (total_bytes_skipped);
1600 		}
1601 
1602 		if (bytes_read >= request) {
1603 			filter->client_next =
1604 			    ((const char *)filter->client_buff) + request;
1605 			filter->client_avail = (size_t)(bytes_read - request);
1606 			filter->client_total = bytes_read;
1607 			total_bytes_skipped += request;
1608 			filter->position += request;
1609 			return (total_bytes_skipped);
1610 		}
1611 
1612 		filter->position += bytes_read;
1613 		total_bytes_skipped += bytes_read;
1614 		request -= bytes_read;
1615 	}
1616 }
1617 
1618 /**
1619  * Returns ARCHIVE_FAILED if seeking isn't supported.
1620  */
1621 int64_t
__archive_read_seek(struct archive_read * a,int64_t offset,int whence)1622 __archive_read_seek(struct archive_read *a, int64_t offset, int whence)
1623 {
1624 	return __archive_read_filter_seek(a->filter, offset, whence);
1625 }
1626 
1627 int64_t
__archive_read_filter_seek(struct archive_read_filter * filter,int64_t offset,int whence)1628 __archive_read_filter_seek(struct archive_read_filter *filter, int64_t offset,
1629     int whence)
1630 {
1631 	struct archive_read_client *client;
1632 	int64_t r;
1633 	unsigned int cursor;
1634 
1635 	if (filter->closed || filter->fatal)
1636 		return (ARCHIVE_FATAL);
1637 	if (filter->can_seek == 0)
1638 		return (ARCHIVE_FAILED);
1639 
1640 	client = &(filter->archive->client);
1641 	switch (whence) {
1642 	case SEEK_CUR:
1643 		/* Adjust the offset and use SEEK_SET instead */
1644 		offset += filter->position;
1645 		__LA_FALLTHROUGH;
1646 	case SEEK_SET:
1647 		cursor = 0;
1648 		while (1)
1649 		{
1650 			if (client->dataset[cursor].begin_position < 0 ||
1651 			    client->dataset[cursor].total_size < 0 ||
1652 			    client->dataset[cursor].begin_position +
1653 			      client->dataset[cursor].total_size - 1 > offset ||
1654 			    cursor + 1 >= client->nodes)
1655 				break;
1656 			r = client->dataset[cursor].begin_position +
1657 				client->dataset[cursor].total_size;
1658 			client->dataset[++cursor].begin_position = r;
1659 		}
1660 		while (1) {
1661 			r = client_switch_proxy(filter, cursor);
1662 			if (r != ARCHIVE_OK)
1663 				return r;
1664 			if ((r = client_seek_proxy(filter, 0, SEEK_END)) < 0)
1665 				return r;
1666 			client->dataset[cursor].total_size = r;
1667 			if (client->dataset[cursor].begin_position +
1668 			    client->dataset[cursor].total_size - 1 > offset ||
1669 			    cursor + 1 >= client->nodes)
1670 				break;
1671 			r = client->dataset[cursor].begin_position +
1672 				client->dataset[cursor].total_size;
1673 			client->dataset[++cursor].begin_position = r;
1674 		}
1675 		offset -= client->dataset[cursor].begin_position;
1676 		if (offset < 0
1677 		    || offset > client->dataset[cursor].total_size)
1678 			return ARCHIVE_FATAL;
1679 		if ((r = client_seek_proxy(filter, offset, SEEK_SET)) < 0)
1680 			return r;
1681 		break;
1682 
1683 	case SEEK_END:
1684 		cursor = 0;
1685 		while (1) {
1686 			if (client->dataset[cursor].begin_position < 0 ||
1687 			    client->dataset[cursor].total_size < 0 ||
1688 			    cursor + 1 >= client->nodes)
1689 				break;
1690 			r = client->dataset[cursor].begin_position +
1691 				client->dataset[cursor].total_size;
1692 			client->dataset[++cursor].begin_position = r;
1693 		}
1694 		while (1) {
1695 			r = client_switch_proxy(filter, cursor);
1696 			if (r != ARCHIVE_OK)
1697 				return r;
1698 			if ((r = client_seek_proxy(filter, 0, SEEK_END)) < 0)
1699 				return r;
1700 			client->dataset[cursor].total_size = r;
1701 			r = client->dataset[cursor].begin_position +
1702 				client->dataset[cursor].total_size;
1703 			if (cursor + 1 >= client->nodes)
1704 				break;
1705 			client->dataset[++cursor].begin_position = r;
1706 		}
1707 		while (1) {
1708 			if (r + offset >=
1709 			    client->dataset[cursor].begin_position)
1710 				break;
1711 			offset += client->dataset[cursor].total_size;
1712 			if (cursor == 0)
1713 				break;
1714 			cursor--;
1715 			r = client->dataset[cursor].begin_position +
1716 				client->dataset[cursor].total_size;
1717 		}
1718 		offset = (r + offset) - client->dataset[cursor].begin_position;
1719 		if ((r = client_switch_proxy(filter, cursor)) != ARCHIVE_OK)
1720 			return r;
1721 		r = client_seek_proxy(filter, offset, SEEK_SET);
1722 		if (r < ARCHIVE_OK)
1723 			return r;
1724 		break;
1725 
1726 	default:
1727 		return (ARCHIVE_FATAL);
1728 	}
1729 	r += client->dataset[cursor].begin_position;
1730 
1731 	if (r >= 0) {
1732 		/*
1733 		 * Ouch.  Clearing the buffer like this hurts, especially
1734 		 * at bid time.  A lot of our efficiency at bid time comes
1735 		 * from having bidders reuse the data we've already read.
1736 		 *
1737 		 * TODO: If the seek request is in data we already
1738 		 * have, then don't call the seek callback.
1739 		 *
1740 		 * TODO: Zip seeks to end-of-file at bid time.  If
1741 		 * other formats also start doing this, we may need to
1742 		 * find a way for clients to fudge the seek offset to
1743 		 * a block boundary.
1744 		 *
1745 		 * Hmmm... If whence was SEEK_END, we know the file
1746 		 * size is (r - offset).  Can we use that to simplify
1747 		 * the TODO items above?
1748 		 */
1749 		filter->avail = filter->client_avail = 0;
1750 		filter->next = filter->buffer;
1751 		filter->position = r;
1752 		filter->end_of_file = 0;
1753 	}
1754 	return r;
1755 }
1756