2 * Copyright (c) 2003-2011 Tim Kientzle
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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.
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.
34 #include "archive_platform.h"
35 __FBSDID("$FreeBSD: head/lib/libarchive/archive_read.c 201157 2009-12-29 05:30:23Z kientzle $");
52 #include "archive_entry.h"
53 #include "archive_private.h"
54 #include "archive_read_private.h"
56 #define minimum(a, b) (a < b ? a : b)
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);
76 static struct archive_vtable *
77 archive_read_vtable(void)
79 static struct archive_vtable av;
80 static int inited = 0;
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;
98 * Allocate, initialize and return a struct archive object.
101 archive_read_new(void)
103 struct archive_read *a;
105 a = (struct archive_read *)calloc(1, sizeof(*a));
108 a->archive.magic = ARCHIVE_READ_MAGIC;
110 a->archive.state = ARCHIVE_STATE_NEW;
111 a->entry = archive_entry_new2(&a->archive);
112 a->archive.vtable = archive_read_vtable();
114 a->passphrases.last = &a->passphrases.first;
116 return (&a->archive);
120 * Record the do-not-extract-to file. This belongs in archive_read_extract.c.
123 archive_read_extract_set_skip_file(struct archive *_a, la_int64_t d,
126 struct archive_read *a = (struct archive_read *)_a;
128 if (ARCHIVE_OK != __archive_check_magic(_a, ARCHIVE_READ_MAGIC,
129 ARCHIVE_STATE_ANY, "archive_read_extract_set_skip_file"))
131 a->skip_file_set = 1;
132 a->skip_file_dev = d;
133 a->skip_file_ino = i;
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)
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);
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)
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);
172 client_read_proxy(struct archive_read_filter *self, const void **buff)
175 r = (self->archive->client.reader)(&self->archive->archive,
181 client_skip_proxy(struct archive_read_filter *self, int64_t request)
184 __archive_errx(1, "Negative skip requested.");
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;
195 int64_t get, ask = request;
196 if (ask > skip_limit)
198 get = (self->archive->client.skipper)
199 (&self->archive->archive, self->data, ask);
201 if (get == 0 || get == request)
204 return ARCHIVE_FATAL;
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.
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.
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;
231 client_seek_proxy(struct archive_read_filter *self, int64_t offset, int whence)
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.
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);
243 return (self->archive->client.seeker)(&self->archive->archive,
244 self->data, offset, whence);
248 client_close_proxy(struct archive_read_filter *self)
250 int r = ARCHIVE_OK, r2;
253 if (self->archive->client.closer == NULL)
255 for (i = 0; i < self->archive->client.nodes; i++)
257 r2 = (self->archive->client.closer)
258 ((struct archive *)self->archive,
259 self->archive->client.dataset[i].data);
267 client_open_proxy(struct archive_read_filter *self)
270 if (self->archive->client.opener != NULL)
271 r = (self->archive->client.opener)(
272 (struct archive *)self->archive, self->data);
277 client_switch_proxy(struct archive_read_filter *self, unsigned int iindex)
279 int r1 = ARCHIVE_OK, r2 = ARCHIVE_OK;
282 /* Don't do anything if already in the specified data node */
283 if (self->archive->client.cursor == iindex)
286 self->archive->client.cursor = iindex;
287 data2 = self->archive->client.dataset[self->archive->client.cursor].data;
288 if (self->archive->client.switcher != NULL)
290 r1 = r2 = (self->archive->client.switcher)
291 ((struct archive *)self->archive, self->data, data2);
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);
301 if (self->archive->client.opener != NULL)
302 r2 = (self->archive->client.opener)
303 ((struct archive *)self->archive, self->data);
305 return (r1 < r2) ? r1 : r2;
309 archive_read_set_open_callback(struct archive *_a,
310 archive_open_callback *client_opener)
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;
320 archive_read_set_read_callback(struct archive *_a,
321 archive_read_callback *client_reader)
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;
331 archive_read_set_skip_callback(struct archive *_a,
332 archive_skip_callback *client_skipper)
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;
342 archive_read_set_seek_callback(struct archive *_a,
343 archive_seek_callback *client_seeker)
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;
353 archive_read_set_close_callback(struct archive *_a,
354 archive_close_callback *client_closer)
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;
364 archive_read_set_switch_callback(struct archive *_a,
365 archive_switch_callback *client_switcher)
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;
375 archive_read_set_callback_data(struct archive *_a, void *client_data)
377 return archive_read_set_callback_data2(_a, client_data, 0);
381 archive_read_set_callback_data2(struct archive *_a, void *client_data,
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");
388 if (a->client.nodes == 0)
390 a->client.dataset = (struct archive_read_data_node *)
391 calloc(1, sizeof(*a->client.dataset));
392 if (a->client.dataset == NULL)
394 archive_set_error(&a->archive, ENOMEM,
396 return ARCHIVE_FATAL;
401 if (iindex > a->client.nodes - 1)
403 archive_set_error(&a->archive, EINVAL,
404 "Invalid index specified.");
405 return ARCHIVE_FATAL;
407 a->client.dataset[iindex].data = client_data;
408 a->client.dataset[iindex].begin_position = -1;
409 a->client.dataset[iindex].total_size = -1;
414 archive_read_add_callback_data(struct archive *_a, void *client_data,
417 struct archive_read *a = (struct archive_read *)_a;
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;
428 p = realloc(a->client.dataset, sizeof(*a->client.dataset)
429 * (++(a->client.nodes)));
431 archive_set_error(&a->archive, ENOMEM,
433 return ARCHIVE_FATAL;
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;
441 a->client.dataset[iindex].data = client_data;
442 a->client.dataset[iindex].begin_position = -1;
443 a->client.dataset[iindex].total_size = -1;
448 archive_read_append_callback_data(struct archive *_a, void *client_data)
450 struct archive_read *a = (struct archive_read *)_a;
451 return archive_read_add_callback_data(_a, client_data, a->client.nodes);
455 archive_read_prepend_callback_data(struct archive *_a, void *client_data)
457 return archive_read_add_callback_data(_a, client_data, 0);
461 archive_read_open1(struct archive *_a)
463 struct archive_read *a = (struct archive_read *)_a;
464 struct archive_read_filter *filter, *tmp;
465 int slot, e = ARCHIVE_OK;
468 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
469 "archive_read_open");
470 archive_clear_error(&a->archive);
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);
479 /* Open data source. */
480 if (a->client.opener != NULL) {
481 e = (a->client.opener)(&a->archive, a->client.dataset[0].data);
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);
493 filter = calloc(1, sizeof(*filter));
495 return (ARCHIVE_FATAL);
496 filter->bidder = NULL;
497 filter->upstream = NULL;
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;
509 a->client.dataset[0].begin_position = 0;
510 if (!a->filter || !a->bypass_filter_bidding)
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);
522 /* Need to add "NONE" type filter at the end of the filter chain */
524 while (tmp->upstream)
526 tmp->upstream = filter;
531 slot = choose_format(a);
534 a->archive.state = ARCHIVE_STATE_FATAL;
535 return (ARCHIVE_FATAL);
537 a->format = &(a->formats[slot]);
540 a->archive.state = ARCHIVE_STATE_HEADER;
542 /* Ensure libarchive starts from the first node in a multivolume set */
543 client_switch_proxy(a->filter, 0);
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.
553 /* We won't build a filter pipeline with more stages than this. */
554 #define MAX_NUMBER_FILTERS 25
557 choose_filters(struct archive_read *a)
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;
565 for (number_filters = 0; number_filters < MAX_NUMBER_FILTERS; ++number_filters) {
566 number_bidders = sizeof(a->bidders) / sizeof(a->bidders[0]);
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) {
577 best_bidder = bidder;
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);
587 __archive_read_free_filters(a);
588 return (ARCHIVE_FATAL);
590 a->archive.compression_name = a->filter->name;
591 a->archive.compression_code = a->filter->code;
596 = (struct archive_read_filter *)calloc(1, sizeof(*filter));
598 return (ARCHIVE_FATAL);
599 filter->bidder = best_bidder;
601 filter->upstream = a->filter;
603 r = (best_bidder->init)(a->filter);
604 if (r != ARCHIVE_OK) {
605 __archive_read_free_filters(a);
606 return (ARCHIVE_FATAL);
609 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
610 "Input requires too many filters for decoding");
611 return (ARCHIVE_FATAL);
615 * Read header of next entry.
618 _archive_read_next_header2(struct archive *_a, struct archive_entry *entry)
620 struct archive_read *a = (struct archive_read *)_a;
621 int r1 = ARCHIVE_OK, r2;
623 archive_check_magic(_a, ARCHIVE_READ_MAGIC,
624 ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA,
625 "archive_read_next_header");
627 archive_entry_clear(entry);
628 archive_clear_error(&a->archive);
631 * If client didn't consume entire data, skip any remainder
632 * (This is especially important for GNU incremental directories.)
634 if (a->archive.state == ARCHIVE_STATE_DATA) {
635 r1 = archive_read_data_skip(&a->archive);
636 if (r1 == ARCHIVE_EOF)
637 archive_set_error(&a->archive, EIO,
638 "Premature end-of-file.");
639 if (r1 == ARCHIVE_EOF || r1 == ARCHIVE_FATAL) {
640 a->archive.state = ARCHIVE_STATE_FATAL;
641 return (ARCHIVE_FATAL);
645 /* Record start-of-header offset in uncompressed stream. */
646 a->header_position = a->filter->position;
649 r2 = (a->format->read_header)(a, entry);
652 * EOF and FATAL are persistent at this layer. By
653 * modifying the state, we guarantee that future calls to
654 * read a header or read data will fail.
658 a->archive.state = ARCHIVE_STATE_EOF;
659 --_a->file_count;/* Revert a file counter. */
662 a->archive.state = ARCHIVE_STATE_DATA;
665 a->archive.state = ARCHIVE_STATE_DATA;
670 a->archive.state = ARCHIVE_STATE_FATAL;
674 __archive_reset_read_data(&a->archive);
676 a->data_start_node = a->client.cursor;
677 /* EOF always wins; otherwise return the worst error. */
678 return (r2 < r1 || r2 == ARCHIVE_EOF) ? r2 : r1;
682 _archive_read_next_header(struct archive *_a, struct archive_entry **entryp)
685 struct archive_read *a = (struct archive_read *)_a;
687 ret = _archive_read_next_header2(_a, a->entry);
693 * Allow each registered format to bid on whether it wants to handle
694 * the next entry. Return index of winning bidder.
697 choose_format(struct archive_read *a)
704 slots = sizeof(a->formats) / sizeof(a->formats[0]);
708 /* Set up a->format for convenience of bidders. */
709 a->format = &(a->formats[0]);
710 for (i = 0; i < slots; i++, a->format++) {
711 if (a->format->bid) {
712 bid = (a->format->bid)(a, best_bid);
713 if (bid == ARCHIVE_FATAL)
714 return (ARCHIVE_FATAL);
715 if (a->filter->position != 0)
716 __archive_read_seek(a, 0, SEEK_SET);
717 if ((bid > best_bid) || (best_bid_slot < 0)) {
725 * There were no bidders; this is a serious programmer error
726 * and demands a quick and definitive abort.
728 if (best_bid_slot < 0) {
729 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
730 "No formats registered");
731 return (ARCHIVE_FATAL);
735 * There were bidders, but no non-zero bids; this means we
736 * can't support this stream.
739 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
740 "Unrecognized archive format");
741 return (ARCHIVE_FATAL);
744 return (best_bid_slot);
748 * Return the file offset (within the uncompressed data stream) where
749 * the last header started.
752 archive_read_header_position(struct archive *_a)
754 struct archive_read *a = (struct archive_read *)_a;
755 archive_check_magic(_a, ARCHIVE_READ_MAGIC,
756 ARCHIVE_STATE_ANY, "archive_read_header_position");
757 return (a->header_position);
761 * Returns 1 if the archive contains at least one encrypted entry.
762 * If the archive format not support encryption at all
763 * ARCHIVE_READ_FORMAT_ENCRYPTION_UNSUPPORTED is returned.
764 * If for any other reason (e.g. not enough data read so far)
765 * we cannot say whether there are encrypted entries, then
766 * ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW is returned.
767 * In general, this function will return values below zero when the
768 * reader is uncertain or totally incapable of encryption support.
769 * When this function returns 0 you can be sure that the reader
770 * supports encryption detection but no encrypted entries have
773 * NOTE: If the metadata/header of an archive is also encrypted, you
774 * cannot rely on the number of encrypted entries. That is why this
775 * function does not return the number of encrypted entries but#
776 * just shows that there are some.
779 archive_read_has_encrypted_entries(struct archive *_a)
781 struct archive_read *a = (struct archive_read *)_a;
782 int format_supports_encryption = archive_read_format_capabilities(_a)
783 & (ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_DATA | ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_METADATA);
785 if (!_a || !format_supports_encryption) {
786 /* Format in general doesn't support encryption */
787 return ARCHIVE_READ_FORMAT_ENCRYPTION_UNSUPPORTED;
790 /* A reader potentially has read enough data now. */
791 if (a->format && a->format->has_encrypted_entries) {
792 return (a->format->has_encrypted_entries)(a);
795 /* For any other reason we cannot say how many entries are there. */
796 return ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW;
800 * Returns a bitmask of capabilities that are supported by the archive format reader.
801 * If the reader has no special capabilities, ARCHIVE_READ_FORMAT_CAPS_NONE is returned.
804 archive_read_format_capabilities(struct archive *_a)
806 struct archive_read *a = (struct archive_read *)_a;
807 if (a && a->format && a->format->format_capabilties) {
808 return (a->format->format_capabilties)(a);
810 return ARCHIVE_READ_FORMAT_CAPS_NONE;
814 * Read data from an archive entry, using a read(2)-style interface.
815 * This is a convenience routine that just calls
816 * archive_read_data_block and copies the results into the client
817 * buffer, filling any gaps with zero bytes. Clients using this
818 * API can be completely ignorant of sparse-file issues; sparse files
819 * will simply be padded with nulls.
821 * DO NOT intermingle calls to this function and archive_read_data_block
822 * to read a single entry body.
825 archive_read_data(struct archive *_a, void *buff, size_t s)
827 struct archive *a = (struct archive *)_a;
829 const void *read_buf;
838 if (a->read_data_remaining == 0) {
839 read_buf = a->read_data_block;
840 a->read_data_is_posix_read = 1;
841 a->read_data_requested = s;
842 r = archive_read_data_block(a, &read_buf,
843 &a->read_data_remaining, &a->read_data_offset);
844 a->read_data_block = read_buf;
845 if (r == ARCHIVE_EOF)
848 * Error codes are all negative, so the status
849 * return here cannot be confused with a valid
850 * byte count. (ARCHIVE_OK is zero.)
856 if (a->read_data_offset < a->read_data_output_offset) {
857 archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
858 "Encountered out-of-order sparse blocks");
859 return (ARCHIVE_RETRY);
862 /* Compute the amount of zero padding needed. */
863 if (a->read_data_output_offset + (int64_t)s <
864 a->read_data_offset) {
866 } else if (a->read_data_output_offset <
867 a->read_data_offset) {
868 len = (size_t)(a->read_data_offset -
869 a->read_data_output_offset);
874 memset(dest, 0, len);
876 a->read_data_output_offset += len;
880 /* Copy data if there is any space left. */
882 len = a->read_data_remaining;
886 memcpy(dest, a->read_data_block, len);
888 a->read_data_block += len;
889 a->read_data_remaining -= len;
890 a->read_data_output_offset += len;
891 a->read_data_offset += len;
896 a->read_data_is_posix_read = 0;
897 a->read_data_requested = 0;
902 * Reset the read_data_* variables, used for starting a new entry.
904 void __archive_reset_read_data(struct archive * a)
906 a->read_data_output_offset = 0;
907 a->read_data_remaining = 0;
908 a->read_data_is_posix_read = 0;
909 a->read_data_requested = 0;
911 /* extra resets, from rar.c */
912 a->read_data_block = NULL;
913 a->read_data_offset = 0;
917 * Skip over all remaining data in this entry.
920 archive_read_data_skip(struct archive *_a)
922 struct archive_read *a = (struct archive_read *)_a;
928 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA,
929 "archive_read_data_skip");
931 if (a->format->read_data_skip != NULL)
932 r = (a->format->read_data_skip)(a);
934 while ((r = archive_read_data_block(&a->archive,
935 &buff, &size, &offset))
940 if (r == ARCHIVE_EOF)
943 a->archive.state = ARCHIVE_STATE_HEADER;
948 archive_seek_data(struct archive *_a, int64_t offset, int whence)
950 struct archive_read *a = (struct archive_read *)_a;
951 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA,
952 "archive_seek_data_block");
954 if (a->format->seek_data == NULL) {
955 archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
957 "No format_seek_data_block function registered");
958 return (ARCHIVE_FATAL);
961 return (a->format->seek_data)(a, offset, whence);
965 * Read the next block of entry data from the archive.
966 * This is a zero-copy interface; the client receives a pointer,
967 * size, and file offset of the next available block of data.
969 * Returns ARCHIVE_OK if the operation is successful, ARCHIVE_EOF if
970 * the end of entry is encountered.
973 _archive_read_data_block(struct archive *_a,
974 const void **buff, size_t *size, int64_t *offset)
976 struct archive_read *a = (struct archive_read *)_a;
977 archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA,
978 "archive_read_data_block");
980 if (a->format->read_data == NULL) {
981 archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
983 "No format->read_data function registered");
984 return (ARCHIVE_FATAL);
987 return (a->format->read_data)(a, buff, size, offset);
991 close_filters(struct archive_read *a)
993 struct archive_read_filter *f = a->filter;
995 /* Close each filter in the pipeline. */
997 struct archive_read_filter *t = f->upstream;
998 if (!f->closed && f->close != NULL) {
999 int r1 = (f->close)(f);
1012 __archive_read_free_filters(struct archive_read *a)
1014 /* Make sure filters are closed and their buffers are freed */
1017 while (a->filter != NULL) {
1018 struct archive_read_filter *t = a->filter->upstream;
1025 * return the count of # of filters in use
1028 _archive_filter_count(struct archive *_a)
1030 struct archive_read *a = (struct archive_read *)_a;
1031 struct archive_read_filter *p = a->filter;
1041 * Close the file and all I/O.
1044 _archive_read_close(struct archive *_a)
1046 struct archive_read *a = (struct archive_read *)_a;
1047 int r = ARCHIVE_OK, r1 = ARCHIVE_OK;
1049 archive_check_magic(&a->archive, ARCHIVE_READ_MAGIC,
1050 ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_close");
1051 if (a->archive.state == ARCHIVE_STATE_CLOSED)
1052 return (ARCHIVE_OK);
1053 archive_clear_error(&a->archive);
1054 a->archive.state = ARCHIVE_STATE_CLOSED;
1056 /* TODO: Clean up the formatters. */
1058 /* Release the filter objects. */
1059 r1 = close_filters(a);
1067 * Release memory and other resources.
1070 _archive_read_free(struct archive *_a)
1072 struct archive_read *a = (struct archive_read *)_a;
1073 struct archive_read_passphrase *p;
1079 return (ARCHIVE_OK);
1080 archive_check_magic(_a, ARCHIVE_READ_MAGIC,
1081 ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_free");
1082 if (a->archive.state != ARCHIVE_STATE_CLOSED
1083 && a->archive.state != ARCHIVE_STATE_FATAL)
1084 r = archive_read_close(&a->archive);
1086 /* Call cleanup functions registered by optional components. */
1087 if (a->cleanup_archive_extract != NULL)
1088 r = (a->cleanup_archive_extract)(a);
1090 /* Cleanup format-specific data. */
1091 slots = sizeof(a->formats) / sizeof(a->formats[0]);
1092 for (i = 0; i < slots; i++) {
1093 a->format = &(a->formats[i]);
1094 if (a->formats[i].cleanup)
1095 (a->formats[i].cleanup)(a);
1098 /* Free the filters */
1099 __archive_read_free_filters(a);
1101 /* Release the bidder objects. */
1102 n = sizeof(a->bidders)/sizeof(a->bidders[0]);
1103 for (i = 0; i < n; i++) {
1104 if (a->bidders[i].free != NULL) {
1105 int r1 = (a->bidders[i].free)(&a->bidders[i]);
1111 /* Release passphrase list. */
1112 p = a->passphrases.first;
1114 struct archive_read_passphrase *np = p->next;
1116 /* A passphrase should be cleaned. */
1117 memset(p->passphrase, 0, strlen(p->passphrase));
1118 free(p->passphrase);
1123 archive_string_free(&a->archive.error_string);
1124 archive_entry_free(a->entry);
1125 a->archive.magic = 0;
1126 __archive_clean(&a->archive);
1127 free(a->client.dataset);
1132 static struct archive_read_filter *
1133 get_filter(struct archive *_a, int n)
1135 struct archive_read *a = (struct archive_read *)_a;
1136 struct archive_read_filter *f = a->filter;
1137 /* We use n == -1 for 'the last filter', which is always the
1139 if (n == -1 && f != NULL) {
1140 struct archive_read_filter *last = f;
1150 while (n > 0 && f != NULL) {
1158 _archive_filter_code(struct archive *_a, int n)
1160 struct archive_read_filter *f = get_filter(_a, n);
1161 return f == NULL ? -1 : f->code;
1165 _archive_filter_name(struct archive *_a, int n)
1167 struct archive_read_filter *f = get_filter(_a, n);
1168 return f != NULL ? f->name : NULL;
1172 _archive_filter_bytes(struct archive *_a, int n)
1174 struct archive_read_filter *f = get_filter(_a, n);
1175 return f == NULL ? -1 : f->position;
1179 * Used internally by read format handlers to register their bid and
1180 * initialization functions.
1183 __archive_read_register_format(struct archive_read *a,
1186 int (*bid)(struct archive_read *, int),
1187 int (*options)(struct archive_read *, const char *, const char *),
1188 int (*read_header)(struct archive_read *, struct archive_entry *),
1189 int (*read_data)(struct archive_read *, const void **, size_t *, int64_t *),
1190 int (*read_data_skip)(struct archive_read *),
1191 int64_t (*seek_data)(struct archive_read *, int64_t, int),
1192 int (*cleanup)(struct archive_read *),
1193 int (*format_capabilities)(struct archive_read *),
1194 int (*has_encrypted_entries)(struct archive_read *))
1196 int i, number_slots;
1198 archive_check_magic(&a->archive,
1199 ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
1200 "__archive_read_register_format");
1202 number_slots = sizeof(a->formats) / sizeof(a->formats[0]);
1204 for (i = 0; i < number_slots; i++) {
1205 if (a->formats[i].bid == bid)
1206 return (ARCHIVE_WARN); /* We've already installed */
1207 if (a->formats[i].bid == NULL) {
1208 a->formats[i].bid = bid;
1209 a->formats[i].options = options;
1210 a->formats[i].read_header = read_header;
1211 a->formats[i].read_data = read_data;
1212 a->formats[i].read_data_skip = read_data_skip;
1213 a->formats[i].seek_data = seek_data;
1214 a->formats[i].cleanup = cleanup;
1215 a->formats[i].data = format_data;
1216 a->formats[i].name = name;
1217 a->formats[i].format_capabilties = format_capabilities;
1218 a->formats[i].has_encrypted_entries = has_encrypted_entries;
1219 return (ARCHIVE_OK);
1223 archive_set_error(&a->archive, ENOMEM,
1224 "Not enough slots for format registration");
1225 return (ARCHIVE_FATAL);
1229 * Used internally by decompression routines to register their bid and
1230 * initialization functions.
1233 __archive_read_get_bidder(struct archive_read *a,
1234 struct archive_read_filter_bidder **bidder)
1236 int i, number_slots;
1238 number_slots = sizeof(a->bidders) / sizeof(a->bidders[0]);
1240 for (i = 0; i < number_slots; i++) {
1241 if (a->bidders[i].bid == NULL) {
1242 memset(a->bidders + i, 0, sizeof(a->bidders[0]));
1243 *bidder = (a->bidders + i);
1244 return (ARCHIVE_OK);
1248 archive_set_error(&a->archive, ENOMEM,
1249 "Not enough slots for filter registration");
1250 return (ARCHIVE_FATAL);
1254 * The next section implements the peek/consume internal I/O
1255 * system used by archive readers. This system allows simple
1256 * read-ahead for consumers while preserving zero-copy operation
1259 * The two key operations:
1260 * * The read-ahead function returns a pointer to a block of data
1261 * that satisfies a minimum request.
1262 * * The consume function advances the file pointer.
1264 * In the ideal case, filters generate blocks of data
1265 * and __archive_read_ahead() just returns pointers directly into
1266 * those blocks. Then __archive_read_consume() just bumps those
1267 * pointers. Only if your request would span blocks does the I/O
1268 * layer use a copy buffer to provide you with a contiguous block of
1271 * A couple of useful idioms:
1272 * * "I just want some data." Ask for 1 byte and pay attention to
1273 * the "number of bytes available" from __archive_read_ahead().
1274 * Consume whatever you actually use.
1275 * * "I want to output a large block of data." As above, ask for 1 byte,
1276 * emit all that's available (up to whatever limit you have), consume
1277 * it all, then repeat until you're done. This effectively means that
1278 * you're passing along the blocks that came from your provider.
1279 * * "I want to peek ahead by a large amount." Ask for 4k or so, then
1280 * double and repeat until you get an error or have enough. Note
1281 * that the I/O layer will likely end up expanding its copy buffer
1282 * to fit your request, so use this technique cautiously. This
1283 * technique is used, for example, by some of the format tasting
1284 * code that has uncertain look-ahead needs.
1288 * Looks ahead in the input stream:
1289 * * If 'avail' pointer is provided, that returns number of bytes available
1290 * in the current buffer, which may be much larger than requested.
1291 * * If end-of-file, *avail gets set to zero.
1292 * * If error, *avail gets error code.
1293 * * If request can be met, returns pointer to data.
1294 * * If minimum request cannot be met, returns NULL.
1296 * Note: If you just want "some data", ask for 1 byte and pay attention
1297 * to *avail, which will have the actual amount available. If you
1298 * know exactly how many bytes you need, just ask for that and treat
1299 * a NULL return as an error.
1301 * Important: This does NOT move the file pointer. See
1302 * __archive_read_consume() below.
1305 __archive_read_ahead(struct archive_read *a, size_t min, ssize_t *avail)
1307 return (__archive_read_filter_ahead(a->filter, min, avail));
1311 __archive_read_filter_ahead(struct archive_read_filter *filter,
1312 size_t min, ssize_t *avail)
1317 if (filter->fatal) {
1319 *avail = ARCHIVE_FATAL;
1324 * Keep pulling more data until we can satisfy the request.
1329 * If we can satisfy from the copy buffer (and the
1330 * copy buffer isn't empty), we're done. In particular,
1331 * note that min == 0 is a perfectly well-defined
1334 if (filter->avail >= min && filter->avail > 0) {
1336 *avail = filter->avail;
1337 return (filter->next);
1341 * We can satisfy directly from client buffer if everything
1342 * currently in the copy buffer is still in the client buffer.
1344 if (filter->client_total >= filter->client_avail + filter->avail
1345 && filter->client_avail + filter->avail >= min) {
1346 /* "Roll back" to client buffer. */
1347 filter->client_avail += filter->avail;
1348 filter->client_next -= filter->avail;
1349 /* Copy buffer is now empty. */
1351 filter->next = filter->buffer;
1352 /* Return data from client buffer. */
1354 *avail = filter->client_avail;
1355 return (filter->client_next);
1358 /* Move data forward in copy buffer if necessary. */
1359 if (filter->next > filter->buffer &&
1360 filter->next + min > filter->buffer + filter->buffer_size) {
1361 if (filter->avail > 0)
1362 memmove(filter->buffer, filter->next,
1364 filter->next = filter->buffer;
1367 /* If we've used up the client data, get more. */
1368 if (filter->client_avail <= 0) {
1369 if (filter->end_of_file) {
1374 bytes_read = (filter->read)(filter,
1375 &filter->client_buff);
1376 if (bytes_read < 0) { /* Read error. */
1377 filter->client_total = filter->client_avail = 0;
1378 filter->client_next =
1379 filter->client_buff = NULL;
1382 *avail = ARCHIVE_FATAL;
1385 if (bytes_read == 0) {
1386 /* Check for another client object first */
1387 if (filter->archive->client.cursor !=
1388 filter->archive->client.nodes - 1) {
1389 if (client_switch_proxy(filter,
1390 filter->archive->client.cursor + 1)
1394 /* Premature end-of-file. */
1395 filter->client_total = filter->client_avail = 0;
1396 filter->client_next =
1397 filter->client_buff = NULL;
1398 filter->end_of_file = 1;
1399 /* Return whatever we do have. */
1401 *avail = filter->avail;
1404 filter->client_total = bytes_read;
1405 filter->client_avail = filter->client_total;
1406 filter->client_next = filter->client_buff;
1409 * We can't satisfy the request from the copy
1410 * buffer or the existing client data, so we
1411 * need to copy more client data over to the
1415 /* Ensure the buffer is big enough. */
1416 if (min > filter->buffer_size) {
1420 /* Double the buffer; watch for overflow. */
1421 s = t = filter->buffer_size;
1426 if (t <= s) { /* Integer overflow! */
1428 &filter->archive->archive,
1430 "Unable to allocate copy"
1434 *avail = ARCHIVE_FATAL;
1439 /* Now s >= min, so allocate a new buffer. */
1440 p = (char *)malloc(s);
1443 &filter->archive->archive,
1445 "Unable to allocate copy buffer");
1448 *avail = ARCHIVE_FATAL;
1451 /* Move data into newly-enlarged buffer. */
1452 if (filter->avail > 0)
1453 memmove(p, filter->next, filter->avail);
1454 free(filter->buffer);
1455 filter->next = filter->buffer = p;
1456 filter->buffer_size = s;
1459 /* We can add client data to copy buffer. */
1460 /* First estimate: copy to fill rest of buffer. */
1461 tocopy = (filter->buffer + filter->buffer_size)
1462 - (filter->next + filter->avail);
1463 /* Don't waste time buffering more than we need to. */
1464 if (tocopy + filter->avail > min)
1465 tocopy = min - filter->avail;
1466 /* Don't copy more than is available. */
1467 if (tocopy > filter->client_avail)
1468 tocopy = filter->client_avail;
1470 memcpy(filter->next + filter->avail,
1471 filter->client_next, tocopy);
1472 /* Remove this data from client buffer. */
1473 filter->client_next += tocopy;
1474 filter->client_avail -= tocopy;
1475 /* add it to copy buffer. */
1476 filter->avail += tocopy;
1482 * Move the file pointer forward.
1485 __archive_read_consume(struct archive_read *a, int64_t request)
1487 return (__archive_read_filter_consume(a->filter, request));
1491 __archive_read_filter_consume(struct archive_read_filter * filter,
1497 return ARCHIVE_FATAL;
1501 skipped = advance_file_pointer(filter, request);
1502 if (skipped == request)
1504 /* We hit EOF before we satisfied the skip request. */
1505 if (skipped < 0) /* Map error code to 0 for error message below. */
1507 archive_set_error(&filter->archive->archive,
1509 "Truncated input file (needed %jd bytes, only %jd available)",
1510 (intmax_t)request, (intmax_t)skipped);
1511 return (ARCHIVE_FATAL);
1515 * Advance the file pointer by the amount requested.
1516 * Returns the amount actually advanced, which may be less than the
1517 * request if EOF is encountered first.
1518 * Returns a negative value if there's an I/O error.
1521 advance_file_pointer(struct archive_read_filter *filter, int64_t request)
1523 int64_t bytes_skipped, total_bytes_skipped = 0;
1530 /* Use up the copy buffer first. */
1531 if (filter->avail > 0) {
1532 min = (size_t)minimum(request, (int64_t)filter->avail);
1533 filter->next += min;
1534 filter->avail -= min;
1536 filter->position += min;
1537 total_bytes_skipped += min;
1540 /* Then use up the client buffer. */
1541 if (filter->client_avail > 0) {
1542 min = (size_t)minimum(request, (int64_t)filter->client_avail);
1543 filter->client_next += min;
1544 filter->client_avail -= min;
1546 filter->position += min;
1547 total_bytes_skipped += min;
1550 return (total_bytes_skipped);
1552 /* If there's an optimized skip function, use it. */
1553 if (filter->skip != NULL) {
1554 bytes_skipped = (filter->skip)(filter, request);
1555 if (bytes_skipped < 0) { /* error */
1557 return (bytes_skipped);
1559 filter->position += bytes_skipped;
1560 total_bytes_skipped += bytes_skipped;
1561 request -= bytes_skipped;
1563 return (total_bytes_skipped);
1566 /* Use ordinary reads as necessary to complete the request. */
1568 bytes_read = (filter->read)(filter, &filter->client_buff);
1569 if (bytes_read < 0) {
1570 filter->client_buff = NULL;
1572 return (bytes_read);
1575 if (bytes_read == 0) {
1576 if (filter->archive->client.cursor !=
1577 filter->archive->client.nodes - 1) {
1578 if (client_switch_proxy(filter,
1579 filter->archive->client.cursor + 1)
1583 filter->client_buff = NULL;
1584 filter->end_of_file = 1;
1585 return (total_bytes_skipped);
1588 if (bytes_read >= request) {
1589 filter->client_next =
1590 ((const char *)filter->client_buff) + request;
1591 filter->client_avail = (size_t)(bytes_read - request);
1592 filter->client_total = bytes_read;
1593 total_bytes_skipped += request;
1594 filter->position += request;
1595 return (total_bytes_skipped);
1598 filter->position += bytes_read;
1599 total_bytes_skipped += bytes_read;
1600 request -= bytes_read;
1605 * Returns ARCHIVE_FAILED if seeking isn't supported.
1608 __archive_read_seek(struct archive_read *a, int64_t offset, int whence)
1610 return __archive_read_filter_seek(a->filter, offset, whence);
1614 __archive_read_filter_seek(struct archive_read_filter *filter, int64_t offset,
1617 struct archive_read_client *client;
1619 unsigned int cursor;
1621 if (filter->closed || filter->fatal)
1622 return (ARCHIVE_FATAL);
1623 if (filter->seek == NULL)
1624 return (ARCHIVE_FAILED);
1626 client = &(filter->archive->client);
1629 /* Adjust the offset and use SEEK_SET instead */
1630 offset += filter->position;
1636 if (client->dataset[cursor].begin_position < 0 ||
1637 client->dataset[cursor].total_size < 0 ||
1638 client->dataset[cursor].begin_position +
1639 client->dataset[cursor].total_size - 1 > offset ||
1640 cursor + 1 >= client->nodes)
1642 r = client->dataset[cursor].begin_position +
1643 client->dataset[cursor].total_size;
1644 client->dataset[++cursor].begin_position = r;
1647 r = client_switch_proxy(filter, cursor);
1648 if (r != ARCHIVE_OK)
1650 if ((r = client_seek_proxy(filter, 0, SEEK_END)) < 0)
1652 client->dataset[cursor].total_size = r;
1653 if (client->dataset[cursor].begin_position +
1654 client->dataset[cursor].total_size - 1 > offset ||
1655 cursor + 1 >= client->nodes)
1657 r = client->dataset[cursor].begin_position +
1658 client->dataset[cursor].total_size;
1659 client->dataset[++cursor].begin_position = r;
1661 offset -= client->dataset[cursor].begin_position;
1663 || offset > client->dataset[cursor].total_size)
1664 return ARCHIVE_FATAL;
1665 if ((r = client_seek_proxy(filter, offset, SEEK_SET)) < 0)
1672 if (client->dataset[cursor].begin_position < 0 ||
1673 client->dataset[cursor].total_size < 0 ||
1674 cursor + 1 >= client->nodes)
1676 r = client->dataset[cursor].begin_position +
1677 client->dataset[cursor].total_size;
1678 client->dataset[++cursor].begin_position = r;
1681 r = client_switch_proxy(filter, cursor);
1682 if (r != ARCHIVE_OK)
1684 if ((r = client_seek_proxy(filter, 0, SEEK_END)) < 0)
1686 client->dataset[cursor].total_size = r;
1687 r = client->dataset[cursor].begin_position +
1688 client->dataset[cursor].total_size;
1689 if (cursor + 1 >= client->nodes)
1691 client->dataset[++cursor].begin_position = r;
1695 client->dataset[cursor].begin_position)
1697 offset += client->dataset[cursor].total_size;
1701 r = client->dataset[cursor].begin_position +
1702 client->dataset[cursor].total_size;
1704 offset = (r + offset) - client->dataset[cursor].begin_position;
1705 if ((r = client_switch_proxy(filter, cursor)) != ARCHIVE_OK)
1707 r = client_seek_proxy(filter, offset, SEEK_SET);
1713 return (ARCHIVE_FATAL);
1715 r += client->dataset[cursor].begin_position;
1719 * Ouch. Clearing the buffer like this hurts, especially
1720 * at bid time. A lot of our efficiency at bid time comes
1721 * from having bidders reuse the data we've already read.
1723 * TODO: If the seek request is in data we already
1724 * have, then don't call the seek callback.
1726 * TODO: Zip seeks to end-of-file at bid time. If
1727 * other formats also start doing this, we may need to
1728 * find a way for clients to fudge the seek offset to
1731 * Hmmm... If whence was SEEK_END, we know the file
1732 * size is (r - offset). Can we use that to simplify
1733 * the TODO items above?
1735 filter->avail = filter->client_avail = 0;
1736 filter->next = filter->buffer;
1737 filter->position = r;
1738 filter->end_of_file = 0;