2 * Copyright (c) 2003-2007 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("$FreeBSD$");
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 build_stream(struct archive_read *);
59 static int choose_format(struct archive_read *);
60 static int cleanup_filters(struct archive_read *);
61 static struct archive_vtable *archive_read_vtable(void);
62 static int _archive_read_close(struct archive *);
63 static int _archive_read_free(struct archive *);
65 static struct archive_vtable *
66 archive_read_vtable(void)
68 static struct archive_vtable av;
69 static int inited = 0;
72 av.archive_free = _archive_read_free;
73 av.archive_close = _archive_read_close;
79 * Allocate, initialize and return a struct archive object.
82 archive_read_new(void)
84 struct archive_read *a;
86 a = (struct archive_read *)malloc(sizeof(*a));
89 memset(a, 0, sizeof(*a));
90 a->archive.magic = ARCHIVE_READ_MAGIC;
92 a->archive.state = ARCHIVE_STATE_NEW;
93 a->entry = archive_entry_new();
94 a->archive.vtable = archive_read_vtable();
100 * Record the do-not-extract-to file. This belongs in archive_read_extract.c.
103 archive_read_extract_set_skip_file(struct archive *_a, dev_t d, ino_t i)
105 struct archive_read *a = (struct archive_read *)_a;
106 __archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_ANY,
107 "archive_read_extract_set_skip_file");
108 a->skip_file_dev = d;
109 a->skip_file_ino = i;
113 * Set read options for the format.
116 archive_read_set_format_options(struct archive *_a, const char *s)
118 struct archive_read *a;
119 struct archive_format_descriptor *format;
120 char key[64], val[64];
125 __archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
126 "archive_read_set_format_options");
128 if (s == NULL || *s == '\0')
130 a = (struct archive_read *)_a;
131 __archive_check_magic(&a->archive, ARCHIVE_READ_MAGIC,
132 ARCHIVE_STATE_NEW, "archive_read_set_format_options");
134 for (i = 0; i < sizeof(a->formats)/sizeof(a->formats[0]); i++) {
135 format = &a->formats[i];
136 if (format == NULL || format->options == NULL ||
137 format->name == NULL)
138 /* This format does not support option. */
141 while ((len = __archive_parse_options(s, format->name,
142 sizeof(key), key, sizeof(val), val)) > 0) {
143 valp = val[0] == '\0' ? NULL : val;
145 r = format->options(a, key, valp);
147 if (r == ARCHIVE_FATAL)
153 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
154 "Illegal format options.");
155 return (ARCHIVE_WARN);
161 * Set read options for the filter.
164 archive_read_set_filter_options(struct archive *_a, const char *s)
166 struct archive_read *a;
167 struct archive_read_filter *filter;
168 struct archive_read_filter_bidder *bidder;
169 char key[64], val[64];
172 __archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
173 "archive_read_set_filter_options");
175 if (s == NULL || *s == '\0')
177 a = (struct archive_read *)_a;
178 __archive_check_magic(&a->archive, ARCHIVE_READ_MAGIC,
179 ARCHIVE_STATE_NEW, "archive_read_set_filter_options");
181 for (filter = a->filter; filter != NULL; filter = filter->upstream) {
182 bidder = filter->bidder;
185 if (bidder->options == NULL)
186 /* This bidder does not support option */
188 while ((len = __archive_parse_options(s, filter->name,
189 sizeof(key), key, sizeof(val), val)) > 0) {
191 r = bidder->options(bidder, key, NULL);
193 r = bidder->options(bidder, key, val);
194 if (r == ARCHIVE_FATAL)
200 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
201 "Illegal format options.");
202 return (ARCHIVE_WARN);
208 * Set read options for the format and the filter.
211 archive_read_set_options(struct archive *_a, const char *s)
215 __archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
216 "archive_read_set_options");
217 archive_clear_error(_a);
219 r = archive_read_set_format_options(_a, s);
222 r = archive_read_set_filter_options(_a, s);
232 archive_read_open(struct archive *a, void *client_data,
233 archive_open_callback *client_opener, archive_read_callback *client_reader,
234 archive_close_callback *client_closer)
236 /* Old archive_read_open() is just a thin shell around
237 * archive_read_open2. */
238 return archive_read_open2(a, client_data, client_opener,
239 client_reader, NULL, client_closer);
243 client_read_proxy(struct archive_read_filter *self, const void **buff)
246 r = (self->archive->client.reader)(&self->archive->archive,
248 self->archive->archive.raw_position += r;
253 client_skip_proxy(struct archive_read_filter *self, int64_t request)
255 int64_t ask, get, total;
256 /* Limit our maximum seek request to 1GB on platforms
257 * with 32-bit off_t (such as Windows). */
258 int64_t skip_limit = ((int64_t)1) << (sizeof(off_t) * 8 - 2);
260 if (self->archive->client.skipper == NULL)
265 if (ask > skip_limit)
267 get = (self->archive->client.skipper)(&self->archive->archive,
272 self->archive->archive.raw_position += get;
278 client_close_proxy(struct archive_read_filter *self)
282 if (self->archive->client.closer != NULL)
283 r = (self->archive->client.closer)((struct archive *)self->archive,
291 archive_read_open2(struct archive *_a, void *client_data,
292 archive_open_callback *client_opener,
293 archive_read_callback *client_reader,
294 archive_skip_callback *client_skipper,
295 archive_close_callback *client_closer)
297 struct archive_read *a = (struct archive_read *)_a;
298 struct archive_read_filter *filter;
301 __archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
302 "archive_read_open");
303 archive_clear_error(&a->archive);
305 if (client_reader == NULL)
307 "No reader function provided to archive_read_open");
309 /* Open data source. */
310 if (client_opener != NULL) {
311 e =(client_opener)(&a->archive, client_data);
313 /* If the open failed, call the closer to clean up. */
315 (client_closer)(&a->archive, client_data);
320 /* Save the client functions and mock up the initial source. */
321 a->client.reader = client_reader;
322 a->client.skipper = client_skipper;
323 a->client.closer = client_closer;
325 filter = calloc(1, sizeof(*filter));
327 return (ARCHIVE_FATAL);
328 filter->bidder = NULL;
329 filter->upstream = NULL;
331 filter->data = client_data;
332 filter->read = client_read_proxy;
333 filter->skip = client_skip_proxy;
334 filter->close = client_close_proxy;
335 filter->name = "none";
336 filter->code = ARCHIVE_COMPRESSION_NONE;
339 /* Build out the input pipeline. */
342 a->archive.state = ARCHIVE_STATE_HEADER;
348 * Allow each registered stream transform to bid on whether
349 * it wants to handle this stream. Repeat until we've finished
350 * building the pipeline.
353 build_stream(struct archive_read *a)
355 int number_bidders, i, bid, best_bid;
356 struct archive_read_filter_bidder *bidder, *best_bidder;
357 struct archive_read_filter *filter;
362 number_bidders = sizeof(a->bidders) / sizeof(a->bidders[0]);
368 for (i = 0; i < number_bidders; i++, bidder++) {
369 if (bidder->bid != NULL) {
370 bid = (bidder->bid)(bidder, a->filter);
371 if (bid > best_bid) {
373 best_bidder = bidder;
378 /* If no bidder, we're done. */
379 if (best_bidder == NULL) {
380 /* Verify the final pipelin by asking it for some data. */
381 __archive_read_filter_ahead(a->filter, 1, &avail);
384 return (ARCHIVE_FATAL);
386 a->archive.compression_name = a->filter->name;
387 a->archive.compression_code = a->filter->code;
392 = (struct archive_read_filter *)calloc(1, sizeof(*filter));
394 return (ARCHIVE_FATAL);
395 filter->bidder = best_bidder;
397 filter->upstream = a->filter;
399 r = (best_bidder->init)(a->filter);
400 if (r != ARCHIVE_OK) {
408 * Read header of next entry.
411 archive_read_next_header2(struct archive *_a, struct archive_entry *entry)
413 struct archive_read *a = (struct archive_read *)_a;
416 __archive_check_magic(_a, ARCHIVE_READ_MAGIC,
417 ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA,
418 "archive_read_next_header");
421 archive_entry_clear(entry);
422 archive_clear_error(&a->archive);
425 * If no format has yet been chosen, choose one.
427 if (a->format == NULL) {
428 slot = choose_format(a);
430 a->archive.state = ARCHIVE_STATE_FATAL;
431 return (ARCHIVE_FATAL);
433 a->format = &(a->formats[slot]);
437 * If client didn't consume entire data, skip any remainder
438 * (This is especially important for GNU incremental directories.)
440 if (a->archive.state == ARCHIVE_STATE_DATA) {
441 ret = archive_read_data_skip(&a->archive);
442 if (ret == ARCHIVE_EOF) {
443 archive_set_error(&a->archive, EIO, "Premature end-of-file.");
444 a->archive.state = ARCHIVE_STATE_FATAL;
445 return (ARCHIVE_FATAL);
447 if (ret != ARCHIVE_OK)
451 /* Record start-of-header. */
452 a->header_position = a->archive.file_position;
454 ret = (a->format->read_header)(a, entry);
457 * EOF and FATAL are persistent at this layer. By
458 * modifying the state, we guarantee that future calls to
459 * read a header or read data will fail.
463 a->archive.state = ARCHIVE_STATE_EOF;
466 a->archive.state = ARCHIVE_STATE_DATA;
469 a->archive.state = ARCHIVE_STATE_DATA;
474 a->archive.state = ARCHIVE_STATE_FATAL;
478 a->read_data_output_offset = 0;
479 a->read_data_remaining = 0;
484 archive_read_next_header(struct archive *_a, struct archive_entry **entryp)
487 struct archive_read *a = (struct archive_read *)_a;
489 ret = archive_read_next_header2(_a, a->entry);
495 * Allow each registered format to bid on whether it wants to handle
496 * the next entry. Return index of winning bidder.
499 choose_format(struct archive_read *a)
506 slots = sizeof(a->formats) / sizeof(a->formats[0]);
510 /* Set up a->format and a->pformat_data for convenience of bidders. */
511 a->format = &(a->formats[0]);
512 for (i = 0; i < slots; i++, a->format++) {
513 if (a->format->bid) {
514 bid = (a->format->bid)(a);
515 if (bid == ARCHIVE_FATAL)
516 return (ARCHIVE_FATAL);
517 if ((bid > best_bid) || (best_bid_slot < 0)) {
525 * There were no bidders; this is a serious programmer error
526 * and demands a quick and definitive abort.
528 if (best_bid_slot < 0)
529 __archive_errx(1, "No formats were registered; you must "
530 "invoke at least one archive_read_support_format_XXX "
531 "function in order to successfully read an archive.");
534 * There were bidders, but no non-zero bids; this means we
535 * can't support this stream.
538 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
539 "Unrecognized archive format");
540 return (ARCHIVE_FATAL);
543 return (best_bid_slot);
547 * Return the file offset (within the uncompressed data stream) where
548 * the last header started.
551 archive_read_header_position(struct archive *_a)
553 struct archive_read *a = (struct archive_read *)_a;
554 __archive_check_magic(_a, ARCHIVE_READ_MAGIC,
555 ARCHIVE_STATE_ANY, "archive_read_header_position");
556 return (a->header_position);
560 * Read data from an archive entry, using a read(2)-style interface.
561 * This is a convenience routine that just calls
562 * archive_read_data_block and copies the results into the client
563 * buffer, filling any gaps with zero bytes. Clients using this
564 * API can be completely ignorant of sparse-file issues; sparse files
565 * will simply be padded with nulls.
567 * DO NOT intermingle calls to this function and archive_read_data_block
568 * to read a single entry body.
571 archive_read_data(struct archive *_a, void *buff, size_t s)
573 struct archive_read *a = (struct archive_read *)_a;
575 const void *read_buf;
584 if (a->read_data_remaining == 0) {
585 read_buf = a->read_data_block;
586 r = archive_read_data_block(&a->archive, &read_buf,
587 &a->read_data_remaining, &a->read_data_offset);
588 a->read_data_block = read_buf;
589 if (r == ARCHIVE_EOF)
592 * Error codes are all negative, so the status
593 * return here cannot be confused with a valid
594 * byte count. (ARCHIVE_OK is zero.)
600 if (a->read_data_offset < a->read_data_output_offset) {
601 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
602 "Encountered out-of-order sparse blocks");
603 return (ARCHIVE_RETRY);
606 /* Compute the amount of zero padding needed. */
607 if (a->read_data_output_offset + (off_t)s <
608 a->read_data_offset) {
610 } else if (a->read_data_output_offset <
611 a->read_data_offset) {
612 len = a->read_data_offset -
613 a->read_data_output_offset;
618 memset(dest, 0, len);
620 a->read_data_output_offset += len;
624 /* Copy data if there is any space left. */
626 len = a->read_data_remaining;
629 memcpy(dest, a->read_data_block, len);
631 a->read_data_block += len;
632 a->read_data_remaining -= len;
633 a->read_data_output_offset += len;
634 a->read_data_offset += len;
642 #if ARCHIVE_API_VERSION < 3
644 * Obsolete function provided for compatibility only. Note that the API
645 * of this function doesn't allow the caller to detect if the remaining
646 * data from the archive entry is shorter than the buffer provided, or
647 * even if an error occurred while reading data.
650 archive_read_data_into_buffer(struct archive *a, void *d, ssize_t len)
653 archive_read_data(a, d, len);
659 * Skip over all remaining data in this entry.
662 archive_read_data_skip(struct archive *_a)
664 struct archive_read *a = (struct archive_read *)_a;
670 __archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA,
671 "archive_read_data_skip");
673 if (a->format->read_data_skip != NULL)
674 r = (a->format->read_data_skip)(a);
676 while ((r = archive_read_data_block(&a->archive,
677 &buff, &size, &offset))
682 if (r == ARCHIVE_EOF)
685 a->archive.state = ARCHIVE_STATE_HEADER;
690 * Read the next block of entry data from the archive.
691 * This is a zero-copy interface; the client receives a pointer,
692 * size, and file offset of the next available block of data.
694 * Returns ARCHIVE_OK if the operation is successful, ARCHIVE_EOF if
695 * the end of entry is encountered.
698 archive_read_data_block(struct archive *_a,
699 const void **buff, size_t *size, off_t *offset)
701 struct archive_read *a = (struct archive_read *)_a;
702 __archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_DATA,
703 "archive_read_data_block");
705 if (a->format->read_data == NULL) {
706 archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
708 "No format_read_data_block function registered");
709 return (ARCHIVE_FATAL);
712 return (a->format->read_data)(a, buff, size, offset);
716 * Close the file and release most resources.
718 * Be careful: client might just call read_new and then read_finish.
719 * Don't assume we actually read anything or performed any non-trivial
723 _archive_read_close(struct archive *_a)
725 struct archive_read *a = (struct archive_read *)_a;
726 int r = ARCHIVE_OK, r1 = ARCHIVE_OK;
729 __archive_check_magic(&a->archive, ARCHIVE_READ_MAGIC,
730 ARCHIVE_STATE_ANY, "archive_read_close");
731 archive_clear_error(&a->archive);
732 a->archive.state = ARCHIVE_STATE_CLOSED;
735 /* Call cleanup functions registered by optional components. */
736 if (a->cleanup_archive_extract != NULL)
737 r = (a->cleanup_archive_extract)(a);
739 /* TODO: Clean up the formatters. */
741 /* Release the filter objects. */
742 r1 = cleanup_filters(a);
746 /* Release the bidder objects. */
747 n = sizeof(a->bidders)/sizeof(a->bidders[0]);
748 for (i = 0; i < n; i++) {
749 if (a->bidders[i].free != NULL) {
750 r1 = (a->bidders[i].free)(&a->bidders[i]);
760 cleanup_filters(struct archive_read *a)
763 /* Clean up the filter pipeline. */
764 while (a->filter != NULL) {
765 struct archive_read_filter *t = a->filter->upstream;
766 if (a->filter->close != NULL) {
767 int r1 = (a->filter->close)(a->filter);
771 free(a->filter->buffer);
779 * Release memory and other resources.
782 _archive_read_free(struct archive *_a)
784 struct archive_read *a = (struct archive_read *)_a;
789 __archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_ANY,
790 "archive_read_free");
791 if (a->archive.state != ARCHIVE_STATE_CLOSED)
792 r = archive_read_close(&a->archive);
794 /* Cleanup format-specific data. */
795 slots = sizeof(a->formats) / sizeof(a->formats[0]);
796 for (i = 0; i < slots; i++) {
797 a->format = &(a->formats[i]);
798 if (a->formats[i].cleanup)
799 (a->formats[i].cleanup)(a);
802 archive_string_free(&a->archive.error_string);
804 archive_entry_free(a->entry);
805 a->archive.magic = 0;
807 #if ARCHIVE_API_VERSION > 1
813 * Used internally by read format handlers to register their bid and
814 * initialization functions.
817 __archive_read_register_format(struct archive_read *a,
820 int (*bid)(struct archive_read *),
821 int (*options)(struct archive_read *, const char *, const char *),
822 int (*read_header)(struct archive_read *, struct archive_entry *),
823 int (*read_data)(struct archive_read *, const void **, size_t *, off_t *),
824 int (*read_data_skip)(struct archive_read *),
825 int (*cleanup)(struct archive_read *))
829 __archive_check_magic(&a->archive,
830 ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
831 "__archive_read_register_format");
833 number_slots = sizeof(a->formats) / sizeof(a->formats[0]);
835 for (i = 0; i < number_slots; i++) {
836 if (a->formats[i].bid == bid)
837 return (ARCHIVE_WARN); /* We've already installed */
838 if (a->formats[i].bid == NULL) {
839 a->formats[i].bid = bid;
840 a->formats[i].options = options;
841 a->formats[i].read_header = read_header;
842 a->formats[i].read_data = read_data;
843 a->formats[i].read_data_skip = read_data_skip;
844 a->formats[i].cleanup = cleanup;
845 a->formats[i].data = format_data;
846 a->formats[i].name = name;
851 __archive_errx(1, "Not enough slots for format registration");
852 return (ARCHIVE_FATAL); /* Never actually called. */
856 * Used internally by decompression routines to register their bid and
857 * initialization functions.
859 struct archive_read_filter_bidder *
860 __archive_read_get_bidder(struct archive_read *a)
864 __archive_check_magic(&a->archive,
865 ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW,
866 "__archive_read_get_bidder");
868 number_slots = sizeof(a->bidders) / sizeof(a->bidders[0]);
870 for (i = 0; i < number_slots; i++) {
871 if (a->bidders[i].bid == NULL) {
872 memset(a->bidders + i, 0, sizeof(a->bidders[0]));
873 return (a->bidders + i);
877 __archive_errx(1, "Not enough slots for compression registration");
878 return (NULL); /* Never actually executed. */
882 * The next three functions comprise the peek/consume internal I/O
883 * system used by archive format readers. This system allows fairly
884 * flexible read-ahead and allows the I/O code to operate in a
885 * zero-copy manner most of the time.
887 * In the ideal case, filters generate blocks of data
888 * and __archive_read_ahead() just returns pointers directly into
889 * those blocks. Then __archive_read_consume() just bumps those
890 * pointers. Only if your request would span blocks does the I/O
891 * layer use a copy buffer to provide you with a contiguous block of
892 * data. The __archive_read_skip() is an optimization; it scans ahead
893 * very quickly (it usually translates into a seek() operation if
894 * you're reading uncompressed disk files).
896 * A couple of useful idioms:
897 * * "I just want some data." Ask for 1 byte and pay attention to
898 * the "number of bytes available" from __archive_read_ahead().
899 * You can consume more than you asked for; you just can't consume
900 * more than is available. If you consume everything that's
901 * immediately available, the next read_ahead() call will pull
903 * * "I want to output a large block of data." As above, ask for 1 byte,
904 * emit all that's available (up to whatever limit you have), then
905 * repeat until you're done.
906 * * "I want to peek ahead by a large amount." Ask for 4k or so, then
907 * double and repeat until you get an error or have enough. Note
908 * that the I/O layer will likely end up expanding its copy buffer
909 * to fit your request, so use this technique cautiously. This
910 * technique is used, for example, by some of the format tasting
911 * code that has uncertain look-ahead needs.
913 * TODO: Someday, provide a more generic __archive_read_seek() for
914 * those cases where it's useful. This is tricky because there are lots
915 * of cases where seek() is not available (reading gzip data from a
916 * network socket, for instance), so there needs to be a good way to
917 * communicate whether seek() is available and users of that interface
918 * need to use non-seeking strategies whenever seek() is not available.
922 * Looks ahead in the input stream:
923 * * If 'avail' pointer is provided, that returns number of bytes available
924 * in the current buffer, which may be much larger than requested.
925 * * If end-of-file, *avail gets set to zero.
926 * * If error, *avail gets error code.
927 * * If request can be met, returns pointer to data, returns NULL
928 * if request is not met.
930 * Note: If you just want "some data", ask for 1 byte and pay attention
931 * to *avail, which will have the actual amount available. If you
932 * know exactly how many bytes you need, just ask for that and treat
933 * a NULL return as an error.
935 * Important: This does NOT move the file pointer. See
936 * __archive_read_consume() below.
940 * This is tricky. We need to provide our clients with pointers to
941 * contiguous blocks of memory but we want to avoid copying whenever
944 * Mostly, this code returns pointers directly into the block of data
945 * provided by the client_read routine. It can do this unless the
946 * request would split across blocks. In that case, we have to copy
947 * into an internal buffer to combine reads.
950 __archive_read_ahead(struct archive_read *a, size_t min, ssize_t *avail)
952 return (__archive_read_filter_ahead(a->filter, min, avail));
956 __archive_read_filter_ahead(struct archive_read_filter *filter,
957 size_t min, ssize_t *avail)
964 *avail = ARCHIVE_FATAL;
969 * Keep pulling more data until we can satisfy the request.
974 * If we can satisfy from the copy buffer (and the
975 * copy buffer isn't empty), we're done. In particular,
976 * note that min == 0 is a perfectly well-defined
979 if (filter->avail >= min && filter->avail > 0) {
981 *avail = filter->avail;
982 return (filter->next);
986 * We can satisfy directly from client buffer if everything
987 * currently in the copy buffer is still in the client buffer.
989 if (filter->client_total >= filter->client_avail + filter->avail
990 && filter->client_avail + filter->avail >= min) {
991 /* "Roll back" to client buffer. */
992 filter->client_avail += filter->avail;
993 filter->client_next -= filter->avail;
994 /* Copy buffer is now empty. */
996 filter->next = filter->buffer;
997 /* Return data from client buffer. */
999 *avail = filter->client_avail;
1000 return (filter->client_next);
1003 /* Move data forward in copy buffer if necessary. */
1004 if (filter->next > filter->buffer &&
1005 filter->next + min > filter->buffer + filter->buffer_size) {
1006 if (filter->avail > 0)
1007 memmove(filter->buffer, filter->next, filter->avail);
1008 filter->next = filter->buffer;
1011 /* If we've used up the client data, get more. */
1012 if (filter->client_avail <= 0) {
1013 if (filter->end_of_file) {
1018 bytes_read = (filter->read)(filter,
1019 &filter->client_buff);
1020 if (bytes_read < 0) { /* Read error. */
1021 filter->client_total = filter->client_avail = 0;
1022 filter->client_next = filter->client_buff = NULL;
1025 *avail = ARCHIVE_FATAL;
1028 if (bytes_read == 0) { /* Premature end-of-file. */
1029 filter->client_total = filter->client_avail = 0;
1030 filter->client_next = filter->client_buff = NULL;
1031 filter->end_of_file = 1;
1032 /* Return whatever we do have. */
1034 *avail = filter->avail;
1037 filter->position += bytes_read;
1038 filter->client_total = bytes_read;
1039 filter->client_avail = filter->client_total;
1040 filter->client_next = filter->client_buff;
1045 * We can't satisfy the request from the copy
1046 * buffer or the existing client data, so we
1047 * need to copy more client data over to the
1051 /* Ensure the buffer is big enough. */
1052 if (min > filter->buffer_size) {
1056 /* Double the buffer; watch for overflow. */
1057 s = t = filter->buffer_size;
1062 if (t <= s) { /* Integer overflow! */
1064 &filter->archive->archive,
1066 "Unable to allocate copy buffer");
1069 *avail = ARCHIVE_FATAL;
1074 /* Now s >= min, so allocate a new buffer. */
1075 p = (char *)malloc(s);
1078 &filter->archive->archive,
1080 "Unable to allocate copy buffer");
1083 *avail = ARCHIVE_FATAL;
1086 /* Move data into newly-enlarged buffer. */
1087 if (filter->avail > 0)
1088 memmove(p, filter->next, filter->avail);
1089 free(filter->buffer);
1090 filter->next = filter->buffer = p;
1091 filter->buffer_size = s;
1094 /* We can add client data to copy buffer. */
1095 /* First estimate: copy to fill rest of buffer. */
1096 tocopy = (filter->buffer + filter->buffer_size)
1097 - (filter->next + filter->avail);
1098 /* Don't waste time buffering more than we need to. */
1099 if (tocopy + filter->avail > min)
1100 tocopy = min - filter->avail;
1101 /* Don't copy more than is available. */
1102 if (tocopy > filter->client_avail)
1103 tocopy = filter->client_avail;
1105 memcpy(filter->next + filter->avail, filter->client_next,
1107 /* Remove this data from client buffer. */
1108 filter->client_next += tocopy;
1109 filter->client_avail -= tocopy;
1110 /* add it to copy buffer. */
1111 filter->avail += tocopy;
1117 * Move the file pointer forward. This should be called after
1118 * __archive_read_ahead() returns data to you. Don't try to move
1119 * ahead by more than the amount of data available according to
1120 * __archive_read_ahead().
1123 * Mark the appropriate data as used. Note that the request here will
1124 * often be much smaller than the size of the previous read_ahead
1128 __archive_read_consume(struct archive_read *a, size_t request)
1131 r = __archive_read_filter_consume(a->filter, request);
1132 a->archive.file_position += r;
1137 __archive_read_filter_consume(struct archive_read_filter * filter,
1140 if (filter->avail > 0) {
1141 /* Read came from copy buffer. */
1142 filter->next += request;
1143 filter->avail -= request;
1145 /* Read came from client buffer. */
1146 filter->client_next += request;
1147 filter->client_avail -= request;
1153 * Move the file pointer ahead by an arbitrary amount. If you're
1154 * reading uncompressed data from a disk file, this will actually
1155 * translate into a seek() operation. Even in cases where seek()
1156 * isn't feasible, this at least pushes the read-and-discard loop
1157 * down closer to the data source.
1160 __archive_read_skip(struct archive_read *a, int64_t request)
1162 int64_t skipped = __archive_read_skip_lenient(a, request);
1163 if (skipped == request)
1165 /* We hit EOF before we satisfied the skip request. */
1166 if (skipped < 0) // Map error code to 0 for error message below.
1168 archive_set_error(&a->archive,
1170 "Truncated input file (needed %jd bytes, only %jd available)",
1171 (intmax_t)request, (intmax_t)skipped);
1172 return (ARCHIVE_FATAL);
1176 __archive_read_skip_lenient(struct archive_read *a, int64_t request)
1178 int64_t skipped = __archive_read_filter_skip(a->filter, request);
1180 a->archive.file_position += skipped;
1185 __archive_read_filter_skip(struct archive_read_filter *filter, int64_t request)
1187 int64_t bytes_skipped, total_bytes_skipped = 0;
1193 * If there is data in the buffers already, use that first.
1195 if (filter->avail > 0) {
1196 min = minimum(request, (off_t)filter->avail);
1197 bytes_skipped = __archive_read_filter_consume(filter, min);
1198 request -= bytes_skipped;
1199 total_bytes_skipped += bytes_skipped;
1201 if (filter->client_avail > 0) {
1202 min = minimum(request, (int64_t)filter->client_avail);
1203 bytes_skipped = __archive_read_filter_consume(filter, min);
1204 request -= bytes_skipped;
1205 total_bytes_skipped += bytes_skipped;
1208 return (total_bytes_skipped);
1210 * If a client_skipper was provided, try that first.
1212 #if ARCHIVE_API_VERSION < 2
1213 if ((filter->skip != NULL) && (request < SSIZE_MAX)) {
1215 if (filter->skip != NULL) {
1217 bytes_skipped = (filter->skip)(filter, request);
1218 if (bytes_skipped < 0) { /* error */
1219 filter->client_total = filter->client_avail = 0;
1220 filter->client_next = filter->client_buff = NULL;
1222 return (bytes_skipped);
1224 total_bytes_skipped += bytes_skipped;
1225 request -= bytes_skipped;
1226 filter->client_next = filter->client_buff;
1227 filter->client_avail = filter->client_total = 0;
1230 * Note that client_skipper will usually not satisfy the
1231 * full request (due to low-level blocking concerns),
1232 * so even if client_skipper is provided, we may still
1233 * have to use ordinary reads to finish out the request.
1235 while (request > 0) {
1237 (void)__archive_read_filter_ahead(filter, 1, &bytes_read);
1239 return (bytes_read);
1240 if (bytes_read == 0) {
1241 return (total_bytes_skipped);
1243 min = (size_t)(minimum(bytes_read, request));
1244 bytes_read = __archive_read_filter_consume(filter, min);
1245 total_bytes_skipped += bytes_read;
1246 request -= bytes_read;
1248 return (total_bytes_skipped);