1 /*-
2  * Copyright (c) 2004 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 #include "archive_platform.h"
27 __FBSDID("$FreeBSD: src/lib/libarchive/archive_read_support_format_zip.c,v 1.28 2008/12/06 06:45:15 kientzle Exp $");
28 
29 #ifdef HAVE_ERRNO_H
30 #include <errno.h>
31 #endif
32 #include <stdio.h>
33 #ifdef HAVE_STDLIB_H
34 #include <stdlib.h>
35 #endif
36 #include <time.h>
37 #ifdef HAVE_ZLIB_H
38 #include <zlib.h>
39 #else
40 /* Hmmm... This is necessary, but means that we can't correctly extract
41  * even uncompressed entries on platforms that lack zlib. */
42 #define	crc32(crc, buf, len) (unsigned long)0
43 #endif
44 
45 #include "archive.h"
46 #include "archive_entry.h"
47 #include "archive_private.h"
48 #include "archive_read_private.h"
49 #include "archive_endian.h"
50 
51 struct zip {
52 	/* entry_bytes_remaining is the number of bytes we expect. */
53 	int64_t			entry_bytes_remaining;
54 	int64_t			entry_offset;
55 
56 	/* These count the number of bytes actually read for the entry. */
57 	int64_t			entry_compressed_bytes_read;
58 	int64_t			entry_uncompressed_bytes_read;
59 
60 	/* Running CRC32 of the decompressed data */
61 	unsigned long		entry_crc32;
62 
63 	unsigned		version;
64 	unsigned		system;
65 	unsigned		flags;
66 	unsigned		compression;
67 	const char *		compression_name;
68 	time_t			mtime;
69 	time_t			ctime;
70 	time_t			atime;
71 	mode_t			mode;
72 	uid_t			uid;
73 	gid_t			gid;
74 
75 	/* Flags to mark progress of decompression. */
76 	char			decompress_init;
77 	char			end_of_entry;
78 
79 	unsigned long		crc32;
80 	ssize_t			filename_length;
81 	ssize_t			extra_length;
82 	int64_t			uncompressed_size;
83 	int64_t			compressed_size;
84 
85 	unsigned char 		*uncompressed_buffer;
86 	size_t 			uncompressed_buffer_size;
87 #ifdef HAVE_ZLIB_H
88 	z_stream		stream;
89 	char			stream_valid;
90 #endif
91 
92 	struct archive_string	pathname;
93 	struct archive_string	extra;
94 	char	format_name[64];
95 };
96 
97 #define ZIP_LENGTH_AT_END	8
98 
99 struct zip_file_header {
100 	char	signature[4];
101 	char	version[2];
102 	char	flags[2];
103 	char	compression[2];
104 	char	timedate[4];
105 	char	crc32[4];
106 	char	compressed_size[4];
107 	char	uncompressed_size[4];
108 	char	filename_length[2];
109 	char	extra_length[2];
110 };
111 
112 static const char *compression_names[] = {
113 	"uncompressed",
114 	"shrinking",
115 	"reduced-1",
116 	"reduced-2",
117 	"reduced-3",
118 	"reduced-4",
119 	"imploded",
120 	"reserved",
121 	"deflation"
122 };
123 
124 static int	archive_read_format_zip_bid(struct archive_read *);
125 static int	archive_read_format_zip_cleanup(struct archive_read *);
126 static int	archive_read_format_zip_read_data(struct archive_read *,
127 		    const void **, size_t *, off_t *);
128 static int	archive_read_format_zip_read_data_skip(struct archive_read *a);
129 static int	archive_read_format_zip_read_header(struct archive_read *,
130 		    struct archive_entry *);
131 static int	zip_read_data_deflate(struct archive_read *a, const void **buff,
132 		    size_t *size, off_t *offset);
133 static int	zip_read_data_none(struct archive_read *a, const void **buff,
134 		    size_t *size, off_t *offset);
135 static int	zip_read_file_header(struct archive_read *a,
136 		    struct archive_entry *entry, struct zip *zip);
137 static time_t	zip_time(const char *);
138 static void process_extra(const void* extra, struct zip* zip);
139 
140 int
141 archive_read_support_format_zip(struct archive *_a)
142 {
143 	struct archive_read *a = (struct archive_read *)_a;
144 	struct zip *zip;
145 	int r;
146 
147 	zip = (struct zip *)malloc(sizeof(*zip));
148 	if (zip == NULL) {
149 		archive_set_error(&a->archive, ENOMEM, "Can't allocate zip data");
150 		return (ARCHIVE_FATAL);
151 	}
152 	memset(zip, 0, sizeof(*zip));
153 
154 	r = __archive_read_register_format(a,
155 	    zip,
156 	    "zip",
157 	    archive_read_format_zip_bid,
158 	    NULL,
159 	    archive_read_format_zip_read_header,
160 	    archive_read_format_zip_read_data,
161 	    archive_read_format_zip_read_data_skip,
162 	    archive_read_format_zip_cleanup);
163 
164 	if (r != ARCHIVE_OK)
165 		free(zip);
166 	return (ARCHIVE_OK);
167 }
168 
169 
170 static int
171 archive_read_format_zip_bid(struct archive_read *a)
172 {
173 	const char *p;
174 	const void *buff;
175 	ssize_t bytes_avail, offset;
176 
177 	if ((p = __archive_read_ahead(a, 4, NULL)) == NULL)
178 		return (-1);
179 
180 	/*
181 	 * Bid of 30 here is: 16 bits for "PK",
182 	 * next 16-bit field has four options (-2 bits).
183 	 * 16 + 16-2 = 30.
184 	 */
185 	if (p[0] == 'P' && p[1] == 'K') {
186 		if ((p[2] == '\001' && p[3] == '\002')
187 		    || (p[2] == '\003' && p[3] == '\004')
188 		    || (p[2] == '\005' && p[3] == '\006')
189 		    || (p[2] == '\007' && p[3] == '\010')
190 		    || (p[2] == '0' && p[3] == '0'))
191 			return (30);
192 	}
193 
194 	/*
195 	 * Attempt to handle self-extracting archives
196 	 * by noting a PE header and searching forward
197 	 * up to 128k for a 'PK\003\004' marker.
198 	 */
199 	if (p[0] == 'M' && p[1] == 'Z') {
200 		/*
201 		 * TODO: Optimize by initializing 'offset' to an
202 		 * estimate of the likely start of the archive data
203 		 * based on values in the PE header.  Note that we
204 		 * don't need to be exact, but we mustn't skip too
205 		 * far.  The search below will compensate if we
206 		 * undershoot.
207 		 */
208 		offset = 0;
209 		while (offset < 124000) {
210 			/* Get 4k of data beyond where we stopped. */
211 			buff = __archive_read_ahead(a, offset + 4096,
212 			    &bytes_avail);
213 			if (bytes_avail < offset + 1)
214 				break;
215 			p = (const char *)buff + offset;
216 			while (p + 9 < (const char *)buff + bytes_avail) {
217 				if (p[0] == 'P' && p[1] == 'K' /* signature */
218 				    && p[2] == 3 && p[3] == 4 /* File entry */
219 				    && p[8] == 8 /* compression == deflate */
220 				    && p[9] == 0 /* High byte of compression */
221 					)
222 				{
223 					return (30);
224 				}
225 				++p;
226 			}
227 			offset = p - (const char *)buff;
228 		}
229 	}
230 
231 	return (0);
232 }
233 
234 /*
235  * Search forward for a "PK\003\004" file header.  This handles the
236  * case of self-extracting archives, where there is an executable
237  * prepended to the ZIP archive.
238  */
239 static int
240 skip_sfx(struct archive_read *a)
241 {
242 	const void *h;
243 	const char *p, *q;
244 	size_t skip;
245 	ssize_t bytes;
246 
247 	/*
248 	 * TODO: We should be able to skip forward by a bunch
249 	 * by lifting some values from the PE header.  We don't
250 	 * need to be exact (we're still going to search forward
251 	 * to find the header), but it will speed things up and
252 	 * reduce the chance of a false positive.
253 	 */
254 	for (;;) {
255 		h = __archive_read_ahead(a, 4, &bytes);
256 		if (bytes < 4)
257 			return (ARCHIVE_FATAL);
258 		p = h;
259 		q = p + bytes;
260 
261 		/*
262 		 * Scan ahead until we find something that looks
263 		 * like the zip header.
264 		 */
265 		while (p + 4 < q) {
266 			switch (p[3]) {
267 			case '\004':
268 				/* TODO: Additional verification here. */
269 				if (memcmp("PK\003\004", p, 4) == 0) {
270 					skip = p - (const char *)h;
271 					__archive_read_consume(a, skip);
272 					return (ARCHIVE_OK);
273 				}
274 				p += 4;
275 				break;
276 			case '\003': p += 1; break;
277 			case 'K': p += 2; break;
278 			case 'P': p += 3; break;
279 			default: p += 4; break;
280 			}
281 		}
282 		skip = p - (const char *)h;
283 		__archive_read_consume(a, skip);
284 	}
285 }
286 
287 static int
288 archive_read_format_zip_read_header(struct archive_read *a,
289     struct archive_entry *entry)
290 {
291 	const void *h;
292 	const char *signature;
293 	struct zip *zip;
294 	int r = ARCHIVE_OK, r1;
295 
296 	a->archive.archive_format = ARCHIVE_FORMAT_ZIP;
297 	if (a->archive.archive_format_name == NULL)
298 		a->archive.archive_format_name = "ZIP";
299 
300 	zip = (struct zip *)(a->format->data);
301 	zip->decompress_init = 0;
302 	zip->end_of_entry = 0;
303 	zip->entry_uncompressed_bytes_read = 0;
304 	zip->entry_compressed_bytes_read = 0;
305 	zip->entry_crc32 = crc32(0, NULL, 0);
306 	if ((h = __archive_read_ahead(a, 4, NULL)) == NULL)
307 		return (ARCHIVE_FATAL);
308 
309 	signature = (const char *)h;
310 	if (signature[0] == 'M' && signature[1] == 'Z') {
311 		/* This is an executable?  Must be self-extracting... */
312 		r = skip_sfx(a);
313 		if (r < ARCHIVE_WARN)
314 			return (r);
315 		if ((h = __archive_read_ahead(a, 4, NULL)) == NULL)
316 			return (ARCHIVE_FATAL);
317 		signature = (const char *)h;
318 	}
319 
320 	if (signature[0] != 'P' || signature[1] != 'K') {
321 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
322 		    "Bad ZIP file");
323 		return (ARCHIVE_FATAL);
324 	}
325 
326 	/*
327 	 * "PK00" signature is used for "split" archives that
328 	 * only have a single segment.  This means we can just
329 	 * skip the PK00; the first real file header should follow.
330 	 */
331 	if (signature[2] == '0' && signature[3] == '0') {
332 		__archive_read_consume(a, 4);
333 		if ((h = __archive_read_ahead(a, 4, NULL)) == NULL)
334 			return (ARCHIVE_FATAL);
335 		signature = (const char *)h;
336 		if (signature[0] != 'P' || signature[1] != 'K') {
337 			archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
338 			    "Bad ZIP file");
339 			return (ARCHIVE_FATAL);
340 		}
341 	}
342 
343 	if (signature[2] == '\001' && signature[3] == '\002') {
344 		/* Beginning of central directory. */
345 		return (ARCHIVE_EOF);
346 	}
347 
348 	if (signature[2] == '\003' && signature[3] == '\004') {
349 		/* Regular file entry. */
350 		r1 = zip_read_file_header(a, entry, zip);
351 		if (r1 != ARCHIVE_OK)
352 			return (r1);
353 		return (r);
354 	}
355 
356 	if (signature[2] == '\005' && signature[3] == '\006') {
357 		/* End-of-archive record. */
358 		return (ARCHIVE_EOF);
359 	}
360 
361 	if (signature[2] == '\007' && signature[3] == '\010') {
362 		/*
363 		 * We should never encounter this record here;
364 		 * see ZIP_LENGTH_AT_END handling below for details.
365 		 */
366 		archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
367 		    "Bad ZIP file: Unexpected end-of-entry record");
368 		return (ARCHIVE_FATAL);
369 	}
370 
371 	archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
372 	    "Damaged ZIP file or unsupported format variant (%d,%d)",
373 	    signature[2], signature[3]);
374 	return (ARCHIVE_FATAL);
375 }
376 
377 static int
378 zip_read_file_header(struct archive_read *a, struct archive_entry *entry,
379     struct zip *zip)
380 {
381 	const struct zip_file_header *p;
382 	const void *h;
383 
384 	if ((p = __archive_read_ahead(a, sizeof *p, NULL)) == NULL) {
385 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
386 		    "Truncated ZIP file header");
387 		return (ARCHIVE_FATAL);
388 	}
389 
390 	zip->version = p->version[0];
391 	zip->system = p->version[1];
392 	zip->flags = archive_le16dec(p->flags);
393 	zip->compression = archive_le16dec(p->compression);
394 	if (zip->compression <
395 	    sizeof(compression_names)/sizeof(compression_names[0]))
396 		zip->compression_name = compression_names[zip->compression];
397 	else
398 		zip->compression_name = "??";
399 	zip->mtime = zip_time(p->timedate);
400 	zip->ctime = 0;
401 	zip->atime = 0;
402 	zip->mode = 0;
403 	zip->uid = 0;
404 	zip->gid = 0;
405 	zip->crc32 = archive_le32dec(p->crc32);
406 	zip->filename_length = archive_le16dec(p->filename_length);
407 	zip->extra_length = archive_le16dec(p->extra_length);
408 	zip->uncompressed_size = archive_le32dec(p->uncompressed_size);
409 	zip->compressed_size = archive_le32dec(p->compressed_size);
410 
411 	__archive_read_consume(a, sizeof(struct zip_file_header));
412 
413 
414 	/* Read the filename. */
415 	if ((h = __archive_read_ahead(a, zip->filename_length, NULL)) == NULL) {
416 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
417 		    "Truncated ZIP file header");
418 		return (ARCHIVE_FATAL);
419 	}
420 	if (archive_string_ensure(&zip->pathname, zip->filename_length) == NULL)
421 		__archive_errx(1, "Out of memory");
422 	archive_strncpy(&zip->pathname, h, zip->filename_length);
423 	__archive_read_consume(a, zip->filename_length);
424 	archive_entry_set_pathname(entry, zip->pathname.s);
425 
426 	if (zip->pathname.s[archive_strlen(&zip->pathname) - 1] == '/')
427 		zip->mode = AE_IFDIR | 0777;
428 	else
429 		zip->mode = AE_IFREG | 0777;
430 
431 	/* Read the extra data. */
432 	if ((h = __archive_read_ahead(a, zip->extra_length, NULL)) == NULL) {
433 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
434 		    "Truncated ZIP file header");
435 		return (ARCHIVE_FATAL);
436 	}
437 	process_extra(h, zip);
438 	__archive_read_consume(a, zip->extra_length);
439 
440 	/* Populate some additional entry fields: */
441 	archive_entry_set_mode(entry, zip->mode);
442 	archive_entry_set_uid(entry, zip->uid);
443 	archive_entry_set_gid(entry, zip->gid);
444 	archive_entry_set_mtime(entry, zip->mtime, 0);
445 	archive_entry_set_ctime(entry, zip->ctime, 0);
446 	archive_entry_set_atime(entry, zip->atime, 0);
447 	/* Set the size only if it's meaningful. */
448 	if (0 == (zip->flags & ZIP_LENGTH_AT_END))
449 		archive_entry_set_size(entry, zip->uncompressed_size);
450 
451 	zip->entry_bytes_remaining = zip->compressed_size;
452 	zip->entry_offset = 0;
453 
454 	/* If there's no body, force read_data() to return EOF immediately. */
455 	if (0 == (zip->flags & ZIP_LENGTH_AT_END)
456 	    && zip->entry_bytes_remaining < 1)
457 		zip->end_of_entry = 1;
458 
459 	/* Set up a more descriptive format name. */
460 	sprintf(zip->format_name, "ZIP %d.%d (%s)",
461 	    zip->version / 10, zip->version % 10,
462 	    zip->compression_name);
463 	a->archive.archive_format_name = zip->format_name;
464 
465 	return (ARCHIVE_OK);
466 }
467 
468 /* Convert an MSDOS-style date/time into Unix-style time. */
469 static time_t
470 zip_time(const char *p)
471 {
472 	int msTime, msDate;
473 	struct tm ts;
474 
475 	msTime = (0xff & (unsigned)p[0]) + 256 * (0xff & (unsigned)p[1]);
476 	msDate = (0xff & (unsigned)p[2]) + 256 * (0xff & (unsigned)p[3]);
477 
478 	memset(&ts, 0, sizeof(ts));
479 	ts.tm_year = ((msDate >> 9) & 0x7f) + 80; /* Years since 1900. */
480 	ts.tm_mon = ((msDate >> 5) & 0x0f) - 1; /* Month number. */
481 	ts.tm_mday = msDate & 0x1f; /* Day of month. */
482 	ts.tm_hour = (msTime >> 11) & 0x1f;
483 	ts.tm_min = (msTime >> 5) & 0x3f;
484 	ts.tm_sec = (msTime << 1) & 0x3e;
485 	ts.tm_isdst = -1;
486 	return mktime(&ts);
487 }
488 
489 static int
490 archive_read_format_zip_read_data(struct archive_read *a,
491     const void **buff, size_t *size, off_t *offset)
492 {
493 	int r;
494 	struct zip *zip;
495 
496 	zip = (struct zip *)(a->format->data);
497 
498 	/*
499 	 * If we hit end-of-entry last time, clean up and return
500 	 * ARCHIVE_EOF this time.
501 	 */
502 	if (zip->end_of_entry) {
503 		*offset = zip->entry_uncompressed_bytes_read;
504 		*size = 0;
505 		*buff = NULL;
506 		return (ARCHIVE_EOF);
507 	}
508 
509 	switch(zip->compression) {
510 	case 0:  /* No compression. */
511 		r =  zip_read_data_none(a, buff, size, offset);
512 		break;
513 	case 8: /* Deflate compression. */
514 		r =  zip_read_data_deflate(a, buff, size, offset);
515 		break;
516 	default: /* Unsupported compression. */
517 		*buff = NULL;
518 		*size = 0;
519 		*offset = 0;
520 		/* Return a warning. */
521 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
522 		    "Unsupported ZIP compression method (%s)",
523 		    zip->compression_name);
524 		if (zip->flags & ZIP_LENGTH_AT_END) {
525 			/*
526 			 * ZIP_LENGTH_AT_END requires us to
527 			 * decompress the entry in order to
528 			 * skip it, but we don't know this
529 			 * compression method, so we give up.
530 			 */
531 			r = ARCHIVE_FATAL;
532 		} else {
533 			/* We can't decompress this entry, but we will
534 			 * be able to skip() it and try the next entry. */
535 			r = ARCHIVE_WARN;
536 		}
537 		break;
538 	}
539 	if (r != ARCHIVE_OK)
540 		return (r);
541 	/* Update checksum */
542 	if (*size)
543 		zip->entry_crc32 =
544 		    crc32(zip->entry_crc32, *buff, *size);
545 	/* If we hit the end, swallow any end-of-data marker. */
546 	if (zip->end_of_entry) {
547 		if (zip->flags & ZIP_LENGTH_AT_END) {
548 			const char *p;
549 
550 			if ((p = __archive_read_ahead(a, 16, NULL)) == NULL) {
551 				archive_set_error(&a->archive,
552 				    ARCHIVE_ERRNO_FILE_FORMAT,
553 				    "Truncated ZIP end-of-file record");
554 				return (ARCHIVE_FATAL);
555 			}
556 			zip->crc32 = archive_le32dec(p + 4);
557 			zip->compressed_size = archive_le32dec(p + 8);
558 			zip->uncompressed_size = archive_le32dec(p + 12);
559 			__archive_read_consume(a, 16);
560 		}
561 		/* Check file size, CRC against these values. */
562 		if (zip->compressed_size != zip->entry_compressed_bytes_read) {
563 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
564 			    "ZIP compressed data is wrong size");
565 			return (ARCHIVE_WARN);
566 		}
567 		/* Size field only stores the lower 32 bits of the actual size. */
568 		if ((zip->uncompressed_size & UINT32_MAX)
569 		    != (zip->entry_uncompressed_bytes_read & UINT32_MAX)) {
570 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
571 			    "ZIP uncompressed data is wrong size");
572 			return (ARCHIVE_WARN);
573 		}
574 		/* Check computed CRC against header */
575 		if (zip->crc32 != zip->entry_crc32) {
576 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
577 			    "ZIP bad CRC: 0x%lx should be 0x%lx",
578 			    zip->entry_crc32, zip->crc32);
579 			return (ARCHIVE_WARN);
580 		}
581 	}
582 
583 	/* Return EOF immediately if this is a non-regular file. */
584 	if (AE_IFREG != (zip->mode & AE_IFMT))
585 		return (ARCHIVE_EOF);
586 	return (ARCHIVE_OK);
587 }
588 
589 /*
590  * Read "uncompressed" data.  According to the current specification,
591  * if ZIP_LENGTH_AT_END is specified, then the size fields in the
592  * initial file header are supposed to be set to zero.  This would, of
593  * course, make it impossible for us to read the archive, since we
594  * couldn't determine the end of the file data.  Info-ZIP seems to
595  * include the real size fields both before and after the data in this
596  * case (the CRC only appears afterwards), so this works as you would
597  * expect.
598  *
599  * Returns ARCHIVE_OK if successful, ARCHIVE_FATAL otherwise, sets
600  * zip->end_of_entry if it consumes all of the data.
601  */
602 static int
603 zip_read_data_none(struct archive_read *a, const void **buff,
604     size_t *size, off_t *offset)
605 {
606 	struct zip *zip;
607 	ssize_t bytes_avail;
608 
609 	zip = (struct zip *)(a->format->data);
610 
611 	if (zip->entry_bytes_remaining == 0) {
612 		*buff = NULL;
613 		*size = 0;
614 		*offset = zip->entry_offset;
615 		zip->end_of_entry = 1;
616 		return (ARCHIVE_OK);
617 	}
618 	/*
619 	 * Note: '1' here is a performance optimization.
620 	 * Recall that the decompression layer returns a count of
621 	 * available bytes; asking for more than that forces the
622 	 * decompressor to combine reads by copying data.
623 	 */
624 	*buff = __archive_read_ahead(a, 1, &bytes_avail);
625 	if (bytes_avail <= 0) {
626 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
627 		    "Truncated ZIP file data");
628 		return (ARCHIVE_FATAL);
629 	}
630 	if (bytes_avail > zip->entry_bytes_remaining)
631 		bytes_avail = zip->entry_bytes_remaining;
632 	__archive_read_consume(a, bytes_avail);
633 	*size = bytes_avail;
634 	*offset = zip->entry_offset;
635 	zip->entry_offset += *size;
636 	zip->entry_bytes_remaining -= *size;
637 	zip->entry_uncompressed_bytes_read += *size;
638 	zip->entry_compressed_bytes_read += *size;
639 	return (ARCHIVE_OK);
640 }
641 
642 #ifdef HAVE_ZLIB_H
643 static int
644 zip_read_data_deflate(struct archive_read *a, const void **buff,
645     size_t *size, off_t *offset)
646 {
647 	struct zip *zip;
648 	ssize_t bytes_avail;
649 	const void *compressed_buff;
650 	int r;
651 
652 	zip = (struct zip *)(a->format->data);
653 
654 	/* If the buffer hasn't been allocated, allocate it now. */
655 	if (zip->uncompressed_buffer == NULL) {
656 		zip->uncompressed_buffer_size = 32 * 1024;
657 		zip->uncompressed_buffer
658 		    = (unsigned char *)malloc(zip->uncompressed_buffer_size);
659 		if (zip->uncompressed_buffer == NULL) {
660 			archive_set_error(&a->archive, ENOMEM,
661 			    "No memory for ZIP decompression");
662 			return (ARCHIVE_FATAL);
663 		}
664 	}
665 
666 	/* If we haven't yet read any data, initialize the decompressor. */
667 	if (!zip->decompress_init) {
668 		if (zip->stream_valid)
669 			r = inflateReset(&zip->stream);
670 		else
671 			r = inflateInit2(&zip->stream,
672 			    -15 /* Don't check for zlib header */);
673 		if (r != Z_OK) {
674 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
675 			    "Can't initialize ZIP decompression.");
676 			return (ARCHIVE_FATAL);
677 		}
678 		/* Stream structure has been set up. */
679 		zip->stream_valid = 1;
680 		/* We've initialized decompression for this stream. */
681 		zip->decompress_init = 1;
682 	}
683 
684 	/*
685 	 * Note: '1' here is a performance optimization.
686 	 * Recall that the decompression layer returns a count of
687 	 * available bytes; asking for more than that forces the
688 	 * decompressor to combine reads by copying data.
689 	 */
690 	compressed_buff = __archive_read_ahead(a, 1, &bytes_avail);
691 	if (bytes_avail <= 0) {
692 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
693 		    "Truncated ZIP file body");
694 		return (ARCHIVE_FATAL);
695 	}
696 
697 	/*
698 	 * A bug in zlib.h: stream.next_in should be marked 'const'
699 	 * but isn't (the library never alters data through the
700 	 * next_in pointer, only reads it).  The result: this ugly
701 	 * cast to remove 'const'.
702 	 */
703 	zip->stream.next_in = (Bytef *)(uintptr_t)(const void *)compressed_buff;
704 	zip->stream.avail_in = bytes_avail;
705 	zip->stream.total_in = 0;
706 	zip->stream.next_out = zip->uncompressed_buffer;
707 	zip->stream.avail_out = zip->uncompressed_buffer_size;
708 	zip->stream.total_out = 0;
709 
710 	r = inflate(&zip->stream, 0);
711 	switch (r) {
712 	case Z_OK:
713 		break;
714 	case Z_STREAM_END:
715 		zip->end_of_entry = 1;
716 		break;
717 	case Z_MEM_ERROR:
718 		archive_set_error(&a->archive, ENOMEM,
719 		    "Out of memory for ZIP decompression");
720 		return (ARCHIVE_FATAL);
721 	default:
722 		archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
723 		    "ZIP decompression failed (%d)", r);
724 		return (ARCHIVE_FATAL);
725 	}
726 
727 	/* Consume as much as the compressor actually used. */
728 	bytes_avail = zip->stream.total_in;
729 	__archive_read_consume(a, bytes_avail);
730 	zip->entry_bytes_remaining -= bytes_avail;
731 	zip->entry_compressed_bytes_read += bytes_avail;
732 
733 	*offset = zip->entry_offset;
734 	*size = zip->stream.total_out;
735 	zip->entry_uncompressed_bytes_read += *size;
736 	*buff = zip->uncompressed_buffer;
737 	zip->entry_offset += *size;
738 	return (ARCHIVE_OK);
739 }
740 #else
741 static int
742 zip_read_data_deflate(struct archive_read *a, const void **buff,
743     size_t *size, off_t *offset)
744 {
745 	*buff = NULL;
746 	*size = 0;
747 	*offset = 0;
748 	archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
749 	    "libarchive compiled without deflate support (no libz)");
750 	return (ARCHIVE_FATAL);
751 }
752 #endif
753 
754 static int
755 archive_read_format_zip_read_data_skip(struct archive_read *a)
756 {
757 	struct zip *zip;
758 	const void *buff = NULL;
759 	off_t bytes_skipped;
760 
761 	zip = (struct zip *)(a->format->data);
762 
763 	/* If we've already read to end of data, we're done. */
764 	if (zip->end_of_entry)
765 		return (ARCHIVE_OK);
766 
767 	/*
768 	 * If the length is at the end, we have no choice but
769 	 * to decompress all the data to find the end marker.
770 	 */
771 	if (zip->flags & ZIP_LENGTH_AT_END) {
772 		size_t size;
773 		off_t offset;
774 		int r;
775 		do {
776 			r = archive_read_format_zip_read_data(a, &buff,
777 			    &size, &offset);
778 		} while (r == ARCHIVE_OK);
779 		return (r);
780 	}
781 
782 	/*
783 	 * If the length is at the beginning, we can skip the
784 	 * compressed data much more quickly.
785 	 */
786 	bytes_skipped = __archive_read_skip(a, zip->entry_bytes_remaining);
787 	if (bytes_skipped < 0)
788 		return (ARCHIVE_FATAL);
789 
790 	/* This entry is finished and done. */
791 	zip->end_of_entry = 1;
792 	return (ARCHIVE_OK);
793 }
794 
795 static int
796 archive_read_format_zip_cleanup(struct archive_read *a)
797 {
798 	struct zip *zip;
799 
800 	zip = (struct zip *)(a->format->data);
801 #ifdef HAVE_ZLIB_H
802 	if (zip->stream_valid)
803 		inflateEnd(&zip->stream);
804 #endif
805 	free(zip->uncompressed_buffer);
806 	archive_string_free(&(zip->pathname));
807 	archive_string_free(&(zip->extra));
808 	free(zip);
809 	(a->format->data) = NULL;
810 	return (ARCHIVE_OK);
811 }
812 
813 /*
814  * The extra data is stored as a list of
815  *	id1+size1+data1 + id2+size2+data2 ...
816  *  triplets.  id and size are 2 bytes each.
817  */
818 static void
819 process_extra(const void* extra, struct zip* zip)
820 {
821 	int offset = 0;
822 	const char *p = (const char *)extra;
823 	while (offset < zip->extra_length - 4)
824 	{
825 		unsigned short headerid = archive_le16dec(p + offset);
826 		unsigned short datasize = archive_le16dec(p + offset + 2);
827 		offset += 4;
828 		if (offset + datasize > zip->extra_length)
829 			break;
830 #ifdef DEBUG
831 		fprintf(stderr, "Header id 0x%04x, length %d\n",
832 		    headerid, datasize);
833 #endif
834 		switch (headerid) {
835 		case 0x0001:
836 			/* Zip64 extended information extra field. */
837 			if (datasize >= 8)
838 				zip->uncompressed_size = archive_le64dec(p + offset);
839 			if (datasize >= 16)
840 				zip->compressed_size = archive_le64dec(p + offset + 8);
841 			break;
842 		case 0x5455:
843 		{
844 			/* Extended time field "UT". */
845 			int flags = p[offset];
846 			offset++;
847 			datasize--;
848 			/* Flag bits indicate which dates are present. */
849 			if (flags & 0x01)
850 			{
851 #ifdef DEBUG
852 				fprintf(stderr, "mtime: %lld -> %d\n",
853 				    (long long)zip->mtime,
854 				    archive_le32dec(p + offset));
855 #endif
856 				if (datasize < 4)
857 					break;
858 				zip->mtime = archive_le32dec(p + offset);
859 				offset += 4;
860 				datasize -= 4;
861 			}
862 			if (flags & 0x02)
863 			{
864 				if (datasize < 4)
865 					break;
866 				zip->atime = archive_le32dec(p + offset);
867 				offset += 4;
868 				datasize -= 4;
869 			}
870 			if (flags & 0x04)
871 			{
872 				if (datasize < 4)
873 					break;
874 				zip->ctime = archive_le32dec(p + offset);
875 				offset += 4;
876 				datasize -= 4;
877 			}
878 			break;
879 		}
880 		case 0x7855:
881 			/* Info-ZIP Unix Extra Field (type 2) "Ux". */
882 #ifdef DEBUG
883 			fprintf(stderr, "uid %d gid %d\n",
884 			    archive_le16dec(p + offset),
885 			    archive_le16dec(p + offset + 2));
886 #endif
887 			if (datasize >= 2)
888 				zip->uid = archive_le16dec(p + offset);
889 			if (datasize >= 4)
890 				zip->gid = archive_le16dec(p + offset + 2);
891 			break;
892 		default:
893 			break;
894 		}
895 		offset += datasize;
896 	}
897 #ifdef DEBUG
898 	if (offset != zip->extra_length)
899 	{
900 		fprintf(stderr,
901 		    "Extra data field contents do not match reported size!");
902 	}
903 #endif
904 }
905