1 /*-
2  * Copyright (c) 2004-2013 Tim Kientzle
3  * Copyright (c) 2011-2012,2014 Michihiro NAKAJIMA
4  * Copyright (c) 2013 Konrad Kleine
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
17  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
18  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
19  * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
20  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
21  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
22  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
23  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
25  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26  */
27 
28 #include "archive_platform.h"
29 
30 /*
31  * The definitive documentation of the Zip file format is:
32  *   http://www.pkware.com/documents/casestudies/APPNOTE.TXT
33  *
34  * The Info-Zip project has pioneered various extensions to better
35  * support Zip on Unix, including the 0x5455 "UT", 0x5855 "UX", 0x7855
36  * "Ux", and 0x7875 "ux" extensions for time and ownership
37  * information.
38  *
39  * History of this code: The streaming Zip reader was first added to
40  * libarchive in January 2005.  Support for seekable input sources was
41  * added in Nov 2011.  Zip64 support (including a significant code
42  * refactoring) was added in 2014.
43  */
44 
45 #ifdef HAVE_ERRNO_H
46 #include <errno.h>
47 #endif
48 #ifdef HAVE_STDLIB_H
49 #include <stdlib.h>
50 #endif
51 #ifdef HAVE_ZLIB_H
52 #include <zlib.h>
53 #endif
54 #ifdef HAVE_BZLIB_H
55 #include <bzlib.h>
56 #endif
57 #ifdef HAVE_LZMA_H
58 #include <lzma.h>
59 #endif
60 #ifdef HAVE_ZSTD_H
61 #include <zstd.h>
62 #endif
63 
64 #include "archive.h"
65 #include "archive_digest_private.h"
66 #include "archive_cryptor_private.h"
67 #include "archive_endian.h"
68 #include "archive_entry.h"
69 #include "archive_entry_locale.h"
70 #include "archive_hmac_private.h"
71 #include "archive_private.h"
72 #include "archive_rb.h"
73 #include "archive_read_private.h"
74 #include "archive_ppmd8_private.h"
75 
76 #ifndef HAVE_ZLIB_H
77 #include "archive_crc32.h"
78 #endif
79 
80 struct zip_entry {
81 	struct archive_rb_node	node;
82 	struct zip_entry	*next;
83 	int64_t			local_header_offset;
84 	int64_t			compressed_size;
85 	int64_t			uncompressed_size;
86 	int64_t			gid;
87 	int64_t			uid;
88 	struct archive_string	rsrcname;
89 	time_t			mtime;
90 	time_t			atime;
91 	time_t			ctime;
92 	uint32_t		crc32;
93 	uint16_t		mode;
94 	uint16_t		zip_flags; /* From GP Flags Field */
95 	unsigned char		compression;
96 	unsigned char		system; /* From "version written by" */
97 	unsigned char		flags; /* Our extra markers. */
98 	unsigned char		decdat;/* Used for Decryption check */
99 
100 	/* WinZip AES encryption extra field should be available
101 	 * when compression is 99. */
102 	struct {
103 		/* Vendor version: AE-1 - 0x0001, AE-2 - 0x0002 */
104 		unsigned	vendor;
105 #define AES_VENDOR_AE_1	0x0001
106 #define AES_VENDOR_AE_2	0x0002
107 		/* AES encryption strength:
108 		 * 1 - 128 bits, 2 - 192 bits, 2 - 256 bits. */
109 		unsigned	strength;
110 		/* Actual compression method. */
111 		unsigned char	compression;
112 	}			aes_extra;
113 };
114 
115 struct trad_enc_ctx {
116 	uint32_t	keys[3];
117 };
118 
119 /* Bits used in zip_flags. */
120 #define ZIP_ENCRYPTED	(1 << 0)
121 #define ZIP_LENGTH_AT_END	(1 << 3) /* Also called "Streaming bit" */
122 #define ZIP_STRONG_ENCRYPTED	(1 << 6)
123 #define ZIP_UTF8_NAME	(1 << 11)
124 /* See "7.2 Single Password Symmetric Encryption Method"
125    in http://www.pkware.com/documents/casestudies/APPNOTE.TXT */
126 #define ZIP_CENTRAL_DIRECTORY_ENCRYPTED	(1 << 13)
127 
128 /* Bits used in flags. */
129 #define LA_USED_ZIP64	(1 << 0)
130 #define LA_FROM_CENTRAL_DIRECTORY (1 << 1)
131 
132 /*
133  * See "WinZip - AES Encryption Information"
134  *     http://www.winzip.com/aes_info.htm
135  */
136 /* Value used in compression method. */
137 #define WINZIP_AES_ENCRYPTION	99
138 /* Authentication code size. */
139 #define AUTH_CODE_SIZE	10
140 /**/
141 #define MAX_DERIVED_KEY_BUF_SIZE	(AES_MAX_KEY_SIZE * 2 + 2)
142 
143 struct zip {
144 	/* Structural information about the archive. */
145 	struct archive_string	format_name;
146 	int64_t			central_directory_offset;
147 	int64_t			central_directory_offset_adjusted;
148 	size_t			central_directory_entries_total;
149 	size_t			central_directory_entries_on_this_disk;
150 	int			has_encrypted_entries;
151 
152 	/* List of entries (seekable Zip only) */
153 	struct zip_entry	*zip_entries;
154 	struct archive_rb_tree	tree;
155 	struct archive_rb_tree	tree_rsrc;
156 
157 	/* Bytes read but not yet consumed via __archive_read_consume() */
158 	size_t			unconsumed;
159 
160 	/* Information about entry we're currently reading. */
161 	struct zip_entry	*entry;
162 	int64_t			entry_bytes_remaining;
163 
164 	/* These count the number of bytes actually read for the entry. */
165 	int64_t			entry_compressed_bytes_read;
166 	int64_t			entry_uncompressed_bytes_read;
167 
168 	/* Running CRC32 of the decompressed and decrypted data */
169 	unsigned long		computed_crc32;
170 	unsigned long		(*crc32func)(unsigned long, const void *,
171 				    size_t);
172 	char			ignore_crc32;
173 
174 	/* Flags to mark progress of decompression. */
175 	char			decompress_init;
176 	char			end_of_entry;
177 
178 	unsigned char 		*uncompressed_buffer;
179 	size_t 			uncompressed_buffer_size;
180 
181 #ifdef HAVE_ZLIB_H
182 	z_stream		stream;
183 	char			stream_valid;
184 #endif
185 
186 #if HAVE_LZMA_H && HAVE_LIBLZMA
187 	lzma_stream		zipx_lzma_stream;
188 	char            zipx_lzma_valid;
189 #endif
190 
191 #ifdef HAVE_BZLIB_H
192 	bz_stream		bzstream;
193 	char            bzstream_valid;
194 #endif
195 
196 #if HAVE_ZSTD_H && HAVE_LIBZSTD
197 	ZSTD_DStream	*zstdstream;
198 	char            zstdstream_valid;
199 #endif
200 
201 	IByteIn			zipx_ppmd_stream;
202 	ssize_t			zipx_ppmd_read_compressed;
203 	CPpmd8			ppmd8;
204 	char			ppmd8_valid;
205 	char			ppmd8_stream_failed;
206 
207 	struct archive_string_conv *sconv;
208 	struct archive_string_conv *sconv_default;
209 	struct archive_string_conv *sconv_utf8;
210 	int			init_default_conversion;
211 	int			process_mac_extensions;
212 
213 	char			init_decryption;
214 
215 	/* Decryption buffer. */
216 	/*
217 	 * The decrypted data starts at decrypted_ptr and
218 	 * extends for decrypted_bytes_remaining.  Decryption
219 	 * adds new data to the end of this block, data is returned
220 	 * to clients from the beginning.  When the block hits the
221 	 * end of decrypted_buffer, it has to be shuffled back to
222 	 * the beginning of the buffer.
223 	 */
224 	unsigned char 		*decrypted_buffer;
225 	unsigned char 		*decrypted_ptr;
226 	size_t 			decrypted_buffer_size;
227 	size_t 			decrypted_bytes_remaining;
228 	size_t 			decrypted_unconsumed_bytes;
229 
230 	/* Traditional PKWARE decryption. */
231 	struct trad_enc_ctx	tctx;
232 	char			tctx_valid;
233 
234 	/* WinZip AES decryption. */
235 	/* Contexts used for AES decryption. */
236 	archive_crypto_ctx	cctx;
237 	char			cctx_valid;
238 	archive_hmac_sha1_ctx	hctx;
239 	char			hctx_valid;
240 
241 	/* Strong encryption's decryption header information. */
242 	unsigned		iv_size;
243 	unsigned		alg_id;
244 	unsigned		bit_len;
245 	unsigned		flags;
246 	unsigned		erd_size;
247 	unsigned		v_size;
248 	unsigned		v_crc32;
249 	uint8_t			*iv;
250 	uint8_t			*erd;
251 	uint8_t			*v_data;
252 };
253 
254 /* Many systems define min or MIN, but not all. */
255 #define	zipmin(a,b) ((a) < (b) ? (a) : (b))
256 
257 #ifdef HAVE_ZLIB_H
258 static int
259 zip_read_data_deflate(struct archive_read *a, const void **buff,
260 	size_t *size, int64_t *offset);
261 #endif
262 #if HAVE_LZMA_H && HAVE_LIBLZMA
263 static int
264 zip_read_data_zipx_lzma_alone(struct archive_read *a, const void **buff,
265 	size_t *size, int64_t *offset);
266 #endif
267 
268 /* This function is used by Ppmd8_DecodeSymbol during decompression of Ppmd8
269  * streams inside ZIP files. It has 2 purposes: one is to fetch the next
270  * compressed byte from the stream, second one is to increase the counter how
271  * many compressed bytes were read. */
272 static Byte
ppmd_read(void * p)273 ppmd_read(void* p) {
274 	/* Get the handle to current decompression context. */
275 	struct archive_read *a = ((IByteIn*)p)->a;
276 	struct zip *zip = (struct zip*) a->format->data;
277 	ssize_t bytes_avail = 0;
278 
279 	/* Fetch next byte. */
280 	const uint8_t* data = __archive_read_ahead(a, 1, &bytes_avail);
281 	if(bytes_avail < 1) {
282 		zip->ppmd8_stream_failed = 1;
283 		return 0;
284 	}
285 
286 	__archive_read_consume(a, 1);
287 
288 	/* Increment the counter. */
289 	++zip->zipx_ppmd_read_compressed;
290 
291 	/* Return the next compressed byte. */
292 	return data[0];
293 }
294 
295 /* ------------------------------------------------------------------------ */
296 
297 /*
298   Traditional PKWARE Decryption functions.
299  */
300 
301 static void
trad_enc_update_keys(struct trad_enc_ctx * ctx,uint8_t c)302 trad_enc_update_keys(struct trad_enc_ctx *ctx, uint8_t c)
303 {
304 	uint8_t t;
305 #define CRC32(c, b) (crc32(c ^ 0xffffffffUL, &b, 1) ^ 0xffffffffUL)
306 
307 	ctx->keys[0] = CRC32(ctx->keys[0], c);
308 	ctx->keys[1] = (ctx->keys[1] + (ctx->keys[0] & 0xff)) * 134775813L + 1;
309 	t = (ctx->keys[1] >> 24) & 0xff;
310 	ctx->keys[2] = CRC32(ctx->keys[2], t);
311 #undef CRC32
312 }
313 
314 static uint8_t
trad_enc_decrypt_byte(struct trad_enc_ctx * ctx)315 trad_enc_decrypt_byte(struct trad_enc_ctx *ctx)
316 {
317 	unsigned temp = ctx->keys[2] | 2;
318 	return (uint8_t)((temp * (temp ^ 1)) >> 8) & 0xff;
319 }
320 
321 static void
trad_enc_decrypt_update(struct trad_enc_ctx * ctx,const uint8_t * in,size_t in_len,uint8_t * out,size_t out_len)322 trad_enc_decrypt_update(struct trad_enc_ctx *ctx, const uint8_t *in,
323     size_t in_len, uint8_t *out, size_t out_len)
324 {
325 	unsigned i, max;
326 
327 	max = (unsigned)((in_len < out_len)? in_len: out_len);
328 
329 	for (i = 0; i < max; i++) {
330 		uint8_t t = in[i] ^ trad_enc_decrypt_byte(ctx);
331 		out[i] = t;
332 		trad_enc_update_keys(ctx, t);
333 	}
334 }
335 
336 static int
trad_enc_init(struct trad_enc_ctx * ctx,const char * pw,size_t pw_len,const uint8_t * key,size_t key_len,uint8_t * crcchk)337 trad_enc_init(struct trad_enc_ctx *ctx, const char *pw, size_t pw_len,
338     const uint8_t *key, size_t key_len, uint8_t *crcchk)
339 {
340 	uint8_t header[12];
341 
342 	if (key_len < 12) {
343 		*crcchk = 0xff;
344 		return -1;
345 	}
346 
347 	ctx->keys[0] = 305419896L;
348 	ctx->keys[1] = 591751049L;
349 	ctx->keys[2] = 878082192L;
350 
351 	for (;pw_len; --pw_len)
352 		trad_enc_update_keys(ctx, *pw++);
353 
354 	trad_enc_decrypt_update(ctx, key, 12, header, 12);
355 	/* Return the last byte for CRC check. */
356 	*crcchk = header[11];
357 	return 0;
358 }
359 
360 #if 0
361 static void
362 crypt_derive_key_sha1(const void *p, int size, unsigned char *key,
363     int key_size)
364 {
365 #define MD_SIZE 20
366 	archive_sha1_ctx ctx;
367 	unsigned char md1[MD_SIZE];
368 	unsigned char md2[MD_SIZE * 2];
369 	unsigned char mkb[64];
370 	int i;
371 
372 	archive_sha1_init(&ctx);
373 	archive_sha1_update(&ctx, p, size);
374 	archive_sha1_final(&ctx, md1);
375 
376 	memset(mkb, 0x36, sizeof(mkb));
377 	for (i = 0; i < MD_SIZE; i++)
378 		mkb[i] ^= md1[i];
379 	archive_sha1_init(&ctx);
380 	archive_sha1_update(&ctx, mkb, sizeof(mkb));
381 	archive_sha1_final(&ctx, md2);
382 
383 	memset(mkb, 0x5C, sizeof(mkb));
384 	for (i = 0; i < MD_SIZE; i++)
385 		mkb[i] ^= md1[i];
386 	archive_sha1_init(&ctx);
387 	archive_sha1_update(&ctx, mkb, sizeof(mkb));
388 	archive_sha1_final(&ctx, md2 + MD_SIZE);
389 
390 	if (key_size > 32)
391 		key_size = 32;
392 	memcpy(key, md2, key_size);
393 #undef MD_SIZE
394 }
395 #endif
396 
397 /*
398  * Common code for streaming or seeking modes.
399  *
400  * Includes code to read local file headers, decompress data
401  * from entry bodies, and common API.
402  */
403 
404 static unsigned long
real_crc32(unsigned long crc,const void * buff,size_t len)405 real_crc32(unsigned long crc, const void *buff, size_t len)
406 {
407 	return crc32(crc, buff, (unsigned int)len);
408 }
409 
410 /* Used by "ignorecrc32" option to speed up tests. */
411 static unsigned long
fake_crc32(unsigned long crc,const void * buff,size_t len)412 fake_crc32(unsigned long crc, const void *buff, size_t len)
413 {
414 	(void)crc; /* UNUSED */
415 	(void)buff; /* UNUSED */
416 	(void)len; /* UNUSED */
417 	return 0;
418 }
419 
420 static const struct {
421 	int id;
422 	const char * name;
423 } compression_methods[] = {
424 	{0, "uncompressed"}, /* The file is stored (no compression) */
425 	{1, "shrinking"}, /* The file is Shrunk */
426 	{2, "reduced-1"}, /* The file is Reduced with compression factor 1 */
427 	{3, "reduced-2"}, /* The file is Reduced with compression factor 2 */
428 	{4, "reduced-3"}, /* The file is Reduced with compression factor 3 */
429 	{5, "reduced-4"}, /* The file is Reduced with compression factor 4 */
430 	{6, "imploded"},  /* The file is Imploded */
431 	{7, "reserved"},  /* Reserved for Tokenizing compression algorithm */
432 	{8, "deflation"}, /* The file is Deflated */
433 	{9, "deflation-64-bit"}, /* Enhanced Deflating using Deflate64(tm) */
434 	{10, "ibm-terse"},/* PKWARE Data Compression Library Imploding
435 			   * (old IBM TERSE) */
436 	{11, "reserved"}, /* Reserved by PKWARE */
437 	{12, "bzip"},     /* File is compressed using BZIP2 algorithm */
438 	{13, "reserved"}, /* Reserved by PKWARE */
439 	{14, "lzma"},     /* LZMA (EFS) */
440 	{15, "reserved"}, /* Reserved by PKWARE */
441 	{16, "reserved"}, /* Reserved by PKWARE */
442 	{17, "reserved"}, /* Reserved by PKWARE */
443 	{18, "ibm-terse-new"}, /* File is compressed using IBM TERSE (new) */
444 	{19, "ibm-lz777"},/* IBM LZ77 z Architecture (PFS) */
445 	{93, "zstd"},     /*  Zstandard (zstd) Compression */
446 	{95, "xz"},       /* XZ compressed data */
447 	{96, "jpeg"},     /* JPEG compressed data */
448 	{97, "wav-pack"}, /* WavPack compressed data */
449 	{98, "ppmd-1"},   /* PPMd version I, Rev 1 */
450 	{99, "aes"}       /* WinZip AES encryption  */
451 };
452 
453 static const char *
compression_name(const int compression)454 compression_name(const int compression)
455 {
456 	static const int num_compression_methods =
457 		sizeof(compression_methods)/sizeof(compression_methods[0]);
458 	int i=0;
459 
460 	while(compression >= 0 && i < num_compression_methods) {
461 		if (compression_methods[i].id == compression)
462 			return compression_methods[i].name;
463 		i++;
464 	}
465 	return "??";
466 }
467 
468 /* Convert an MSDOS-style date/time into Unix-style time. */
469 static time_t
zip_time(const char * p)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 /*
490  * The extra data is stored as a list of
491  *	id1+size1+data1 + id2+size2+data2 ...
492  *  triplets.  id and size are 2 bytes each.
493  */
494 static int
process_extra(struct archive_read * a,struct archive_entry * entry,const char * p,size_t extra_length,struct zip_entry * zip_entry)495 process_extra(struct archive_read *a, struct archive_entry *entry,
496      const char *p, size_t extra_length, struct zip_entry* zip_entry)
497 {
498 	unsigned offset = 0;
499 	struct zip *zip = (struct zip *)(a->format->data);
500 
501 	if (extra_length == 0) {
502 		return ARCHIVE_OK;
503 	}
504 
505 	if (extra_length < 4) {
506 		size_t i = 0;
507 		/* Some ZIP files may have trailing 0 bytes. Let's check they
508 		 * are all 0 and ignore them instead of returning an error.
509 		 *
510 		 * This is not technically correct, but some ZIP files look
511 		 * like this and other tools support those files - so let's
512 		 * also  support them.
513 		 */
514 		for (; i < extra_length; i++) {
515 			if (p[i] != 0) {
516 				archive_set_error(&a->archive,
517 				    ARCHIVE_ERRNO_FILE_FORMAT,
518 				    "Too-small extra data: "
519 				    "Need at least 4 bytes, "
520 				    "but only found %d bytes",
521 				    (int)extra_length);
522 				return ARCHIVE_FAILED;
523 			}
524 		}
525 
526 		return ARCHIVE_OK;
527 	}
528 
529 	while (offset <= extra_length - 4) {
530 		unsigned short headerid = archive_le16dec(p + offset);
531 		unsigned short datasize = archive_le16dec(p + offset + 2);
532 
533 		offset += 4;
534 		if (offset + datasize > extra_length) {
535 			archive_set_error(&a->archive,
536 			    ARCHIVE_ERRNO_FILE_FORMAT, "Extra data overflow: "
537 			    "Need %d bytes but only found %d bytes",
538 			    (int)datasize, (int)(extra_length - offset));
539 			return ARCHIVE_FAILED;
540 		}
541 #ifdef DEBUG
542 		fprintf(stderr, "Header id 0x%04x, length %d\n",
543 		    headerid, datasize);
544 #endif
545 		switch (headerid) {
546 		case 0x0001:
547 			/* Zip64 extended information extra field. */
548 			zip_entry->flags |= LA_USED_ZIP64;
549 			if (zip_entry->uncompressed_size == 0xffffffff) {
550 				uint64_t t = 0;
551 				if (datasize < 8
552 				    || (t = archive_le64dec(p + offset)) >
553 				    INT64_MAX) {
554 					archive_set_error(&a->archive,
555 					    ARCHIVE_ERRNO_FILE_FORMAT,
556 					    "Malformed 64-bit "
557 					    "uncompressed size");
558 					return ARCHIVE_FAILED;
559 				}
560 				zip_entry->uncompressed_size = t;
561 				offset += 8;
562 				datasize -= 8;
563 			}
564 			if (zip_entry->compressed_size == 0xffffffff) {
565 				uint64_t t = 0;
566 				if (datasize < 8
567 				    || (t = archive_le64dec(p + offset)) >
568 				    INT64_MAX) {
569 					archive_set_error(&a->archive,
570 					    ARCHIVE_ERRNO_FILE_FORMAT,
571 					    "Malformed 64-bit "
572 					    "compressed size");
573 					return ARCHIVE_FAILED;
574 				}
575 				zip_entry->compressed_size = t;
576 				offset += 8;
577 				datasize -= 8;
578 			}
579 			if (zip_entry->local_header_offset == 0xffffffff) {
580 				uint64_t t = 0;
581 				if (datasize < 8
582 				    || (t = archive_le64dec(p + offset)) >
583 				    INT64_MAX) {
584 					archive_set_error(&a->archive,
585 					    ARCHIVE_ERRNO_FILE_FORMAT,
586 					    "Malformed 64-bit "
587 					    "local header offset");
588 					return ARCHIVE_FAILED;
589 				}
590 				zip_entry->local_header_offset = t;
591 				offset += 8;
592 				datasize -= 8;
593 			}
594 			/* archive_le32dec(p + offset) gives disk
595 			 * on which file starts, but we don't handle
596 			 * multi-volume Zip files. */
597 			break;
598 #ifdef DEBUG
599 		case 0x0017:
600 		{
601 			/* Strong encryption field. */
602 			if (archive_le16dec(p + offset) == 2) {
603 				unsigned algId =
604 					archive_le16dec(p + offset + 2);
605 				unsigned bitLen =
606 					archive_le16dec(p + offset + 4);
607 				int	 flags =
608 					archive_le16dec(p + offset + 6);
609 				fprintf(stderr, "algId=0x%04x, bitLen=%u, "
610 				    "flgas=%d\n", algId, bitLen,flags);
611 			}
612 			break;
613 		}
614 #endif
615 		case 0x5455:
616 		{
617 			/* Extended time field "UT". */
618 			int flags;
619 			if (datasize == 0) {
620 				archive_set_error(&a->archive,
621 				    ARCHIVE_ERRNO_FILE_FORMAT,
622 				    "Incomplete extended time field");
623 				return ARCHIVE_FAILED;
624 			}
625 			flags = p[offset];
626 			offset++;
627 			datasize--;
628 			/* Flag bits indicate which dates are present. */
629 			if (flags & 0x01)
630 			{
631 #ifdef DEBUG
632 				fprintf(stderr, "mtime: %lld -> %d\n",
633 				    (long long)zip_entry->mtime,
634 				    archive_le32dec(p + offset));
635 #endif
636 				if (datasize < 4)
637 					break;
638 				zip_entry->mtime = archive_le32dec(p + offset);
639 				offset += 4;
640 				datasize -= 4;
641 			}
642 			if (flags & 0x02)
643 			{
644 				if (datasize < 4)
645 					break;
646 				zip_entry->atime = archive_le32dec(p + offset);
647 				offset += 4;
648 				datasize -= 4;
649 			}
650 			if (flags & 0x04)
651 			{
652 				if (datasize < 4)
653 					break;
654 				zip_entry->ctime = archive_le32dec(p + offset);
655 				offset += 4;
656 				datasize -= 4;
657 			}
658 			break;
659 		}
660 		case 0x5855:
661 		{
662 			/* Info-ZIP Unix Extra Field (old version) "UX". */
663 			if (datasize >= 8) {
664 				zip_entry->atime = archive_le32dec(p + offset);
665 				zip_entry->mtime =
666 				    archive_le32dec(p + offset + 4);
667 			}
668 			if (datasize >= 12) {
669 				zip_entry->uid =
670 				    archive_le16dec(p + offset + 8);
671 				zip_entry->gid =
672 				    archive_le16dec(p + offset + 10);
673 			}
674 			break;
675 		}
676 		case 0x6c78:
677 		{
678 			/* Experimental 'xl' field */
679 			/*
680 			 * Introduced Dec 2013 to provide a way to
681 			 * include external file attributes (and other
682 			 * fields that ordinarily appear only in
683 			 * central directory) in local file header.
684 			 * This provides file type and permission
685 			 * information necessary to support full
686 			 * streaming extraction.  Currently being
687 			 * discussed with other Zip developers
688 			 * ... subject to change.
689 			 *
690 			 * Format:
691 			 *  The field starts with a bitmap that specifies
692 			 *  which additional fields are included.  The
693 			 *  bitmap is variable length and can be extended in
694 			 *  the future.
695 			 *
696 			 *  n bytes - feature bitmap: first byte has low-order
697 			 *    7 bits.  If high-order bit is set, a subsequent
698 			 *    byte holds the next 7 bits, etc.
699 			 *
700 			 *  if bitmap & 1, 2 byte "version made by"
701 			 *  if bitmap & 2, 2 byte "internal file attributes"
702 			 *  if bitmap & 4, 4 byte "external file attributes"
703 			 *  if bitmap & 8, 2 byte comment length + n byte
704 			 *  comment
705 			 */
706 			int bitmap, bitmap_last;
707 
708 			if (datasize < 1)
709 				break;
710 			bitmap_last = bitmap = 0xff & p[offset];
711 			offset += 1;
712 			datasize -= 1;
713 
714 			/* We only support first 7 bits of bitmap; skip rest. */
715 			while ((bitmap_last & 0x80) != 0
716 			    && datasize >= 1) {
717 				bitmap_last = p[offset];
718 				offset += 1;
719 				datasize -= 1;
720 			}
721 
722 			if (bitmap & 1) {
723 				/* 2 byte "version made by" */
724 				if (datasize < 2)
725 					break;
726 				zip_entry->system
727 				    = archive_le16dec(p + offset) >> 8;
728 				offset += 2;
729 				datasize -= 2;
730 			}
731 			if (bitmap & 2) {
732 				/* 2 byte "internal file attributes" */
733 				uint32_t internal_attributes;
734 				if (datasize < 2)
735 					break;
736 				internal_attributes
737 				    = archive_le16dec(p + offset);
738 				/* Not used by libarchive at present. */
739 				(void)internal_attributes; /* UNUSED */
740 				offset += 2;
741 				datasize -= 2;
742 			}
743 			if (bitmap & 4) {
744 				/* 4 byte "external file attributes" */
745 				uint32_t external_attributes;
746 				if (datasize < 4)
747 					break;
748 				external_attributes
749 				    = archive_le32dec(p + offset);
750 				if (zip_entry->system == 3) {
751 					zip_entry->mode
752 					    = external_attributes >> 16;
753 				} else if (zip_entry->system == 0) {
754 					// Interpret MSDOS directory bit
755 					if (0x10 == (external_attributes &
756 					    0x10)) {
757 						zip_entry->mode =
758 						    AE_IFDIR | 0775;
759 					} else {
760 						zip_entry->mode =
761 						    AE_IFREG | 0664;
762 					}
763 					if (0x01 == (external_attributes &
764 					    0x01)) {
765 						/* Read-only bit;
766 						 * strip write permissions */
767 						zip_entry->mode &= 0555;
768 					}
769 				} else {
770 					zip_entry->mode = 0;
771 				}
772 				offset += 4;
773 				datasize -= 4;
774 			}
775 			if (bitmap & 8) {
776 				/* 2 byte comment length + comment */
777 				uint32_t comment_length;
778 				if (datasize < 2)
779 					break;
780 				comment_length
781 				    = archive_le16dec(p + offset);
782 				offset += 2;
783 				datasize -= 2;
784 
785 				if (datasize < comment_length)
786 					break;
787 				/* Comment is not supported by libarchive */
788 				offset += comment_length;
789 				datasize -= comment_length;
790 			}
791 			break;
792 		}
793 		case 0x7075:
794 		{
795 			/* Info-ZIP Unicode Path Extra Field. */
796 			if (datasize < 5 || entry == NULL)
797 				break;
798 			offset += 5;
799 			datasize -= 5;
800 
801 			/* The path name in this field is always encoded
802 			 * in UTF-8. */
803 			if (zip->sconv_utf8 == NULL) {
804 				zip->sconv_utf8 =
805 					archive_string_conversion_from_charset(
806 					&a->archive, "UTF-8", 1);
807 				/* If the converter from UTF-8 is not
808 				 * available, then the path name from the main
809 				 * field will more likely be correct. */
810 				if (zip->sconv_utf8 == NULL)
811 					break;
812 			}
813 
814 			/* Make sure the CRC32 of the filename matches. */
815 			if (!zip->ignore_crc32) {
816 				const char *cp = archive_entry_pathname(entry);
817 				if (cp) {
818 					unsigned long file_crc =
819 					    zip->crc32func(0, cp, strlen(cp));
820 					unsigned long utf_crc =
821 					    archive_le32dec(p + offset - 4);
822 					if (file_crc != utf_crc) {
823 #ifdef DEBUG
824 						fprintf(stderr,
825 						    "CRC filename mismatch; "
826 						    "CDE is %lx, but UTF8 "
827 						    "is outdated with %lx\n",
828 						    file_crc, utf_crc);
829 #endif
830 						break;
831 					}
832 				}
833 			}
834 
835 			if (archive_entry_copy_pathname_l(entry,
836 			    p + offset, datasize, zip->sconv_utf8) != 0) {
837 				/* Ignore the error, and fallback to the path
838 				 * name from the main field. */
839 #ifdef DEBUG
840 				fprintf(stderr, "Failed to read the ZIP "
841 				    "0x7075 extra field path.\n");
842 #endif
843 			}
844 			break;
845 		}
846 		case 0x7855:
847 			/* Info-ZIP Unix Extra Field (type 2) "Ux". */
848 #ifdef DEBUG
849 			fprintf(stderr, "uid %d gid %d\n",
850 			    archive_le16dec(p + offset),
851 			    archive_le16dec(p + offset + 2));
852 #endif
853 			if (datasize >= 2)
854 				zip_entry->uid = archive_le16dec(p + offset);
855 			if (datasize >= 4)
856 				zip_entry->gid =
857 				    archive_le16dec(p + offset + 2);
858 			break;
859 		case 0x7875:
860 		{
861 			/* Info-Zip Unix Extra Field (type 3) "ux". */
862 			int uidsize = 0, gidsize = 0;
863 
864 			/* TODO: support arbitrary uidsize/gidsize. */
865 			if (datasize >= 1 && p[offset] == 1) {/* version=1 */
866 				if (datasize >= 4) {
867 					/* get a uid size. */
868 					uidsize = 0xff & (int)p[offset+1];
869 					if (uidsize == 2)
870 						zip_entry->uid =
871 						    archive_le16dec(
872 						        p + offset + 2);
873 					else if (uidsize == 4 && datasize >= 6)
874 						zip_entry->uid =
875 						    archive_le32dec(
876 						        p + offset + 2);
877 				}
878 				if (datasize >= (2 + uidsize + 3)) {
879 					/* get a gid size. */
880 					gidsize = 0xff &
881 					    (int)p[offset+2+uidsize];
882 					if (gidsize == 2)
883 						zip_entry->gid =
884 						    archive_le16dec(
885 						        p+offset+2+uidsize+1);
886 					else if (gidsize == 4 &&
887 					    datasize >= (2 + uidsize + 5))
888 						zip_entry->gid =
889 						    archive_le32dec(
890 						        p+offset+2+uidsize+1);
891 				}
892 			}
893 			break;
894 		}
895 		case 0x9901:
896 			/* WinZip AES extra data field. */
897 			if (datasize < 6) {
898 				archive_set_error(&a->archive,
899 				    ARCHIVE_ERRNO_FILE_FORMAT,
900 				    "Incomplete AES field");
901 				return ARCHIVE_FAILED;
902 			}
903 			if (p[offset + 2] == 'A' && p[offset + 3] == 'E') {
904 				/* Vendor version. */
905 				zip_entry->aes_extra.vendor =
906 				    archive_le16dec(p + offset);
907 				/* AES encryption strength. */
908 				zip_entry->aes_extra.strength = p[offset + 4];
909 				/* Actual compression method. */
910 				zip_entry->aes_extra.compression =
911 				    p[offset + 5];
912 			}
913 			break;
914 		default:
915 			break;
916 		}
917 		offset += datasize;
918 	}
919 	return ARCHIVE_OK;
920 }
921 
922 /*
923  * Assumes file pointer is at beginning of local file header.
924  */
925 static int
zip_read_local_file_header(struct archive_read * a,struct archive_entry * entry,struct zip * zip)926 zip_read_local_file_header(struct archive_read *a, struct archive_entry *entry,
927     struct zip *zip)
928 {
929 	const char *p;
930 	const void *h;
931 	const wchar_t *wp;
932 	const char *cp;
933 	size_t len, filename_length, extra_length;
934 	struct archive_string_conv *sconv;
935 	struct zip_entry *zip_entry = zip->entry;
936 	struct zip_entry zip_entry_central_dir;
937 	int ret = ARCHIVE_OK;
938 	char version;
939 
940 	/* Save a copy of the original for consistency checks. */
941 	zip_entry_central_dir = *zip_entry;
942 
943 	zip->decompress_init = 0;
944 	zip->end_of_entry = 0;
945 	zip->entry_uncompressed_bytes_read = 0;
946 	zip->entry_compressed_bytes_read = 0;
947 	zip->computed_crc32 = zip->crc32func(0, NULL, 0);
948 
949 	/* Setup default conversion. */
950 	if (zip->sconv == NULL && !zip->init_default_conversion) {
951 		zip->sconv_default =
952 		    archive_string_default_conversion_for_read(&(a->archive));
953 		zip->init_default_conversion = 1;
954 	}
955 
956 	if ((p = __archive_read_ahead(a, 30, NULL)) == NULL) {
957 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
958 		    "Truncated ZIP file header");
959 		return (ARCHIVE_FATAL);
960 	}
961 
962 	if (memcmp(p, "PK\003\004", 4) != 0) {
963 		archive_set_error(&a->archive, -1, "Damaged Zip archive");
964 		return ARCHIVE_FATAL;
965 	}
966 	version = p[4];
967 	zip_entry->system = p[5];
968 	zip_entry->zip_flags = archive_le16dec(p + 6);
969 	if (zip_entry->zip_flags & (ZIP_ENCRYPTED | ZIP_STRONG_ENCRYPTED)) {
970 		zip->has_encrypted_entries = 1;
971 		archive_entry_set_is_data_encrypted(entry, 1);
972 		if (zip_entry->zip_flags & ZIP_CENTRAL_DIRECTORY_ENCRYPTED &&
973 			zip_entry->zip_flags & ZIP_ENCRYPTED &&
974 			zip_entry->zip_flags & ZIP_STRONG_ENCRYPTED) {
975 			archive_entry_set_is_metadata_encrypted(entry, 1);
976 			return ARCHIVE_FATAL;
977 		}
978 	}
979 	zip->init_decryption = (zip_entry->zip_flags & ZIP_ENCRYPTED);
980 	zip_entry->compression = (char)archive_le16dec(p + 8);
981 	zip_entry->mtime = zip_time(p + 10);
982 	zip_entry->crc32 = archive_le32dec(p + 14);
983 	if (zip_entry->zip_flags & ZIP_LENGTH_AT_END)
984 		zip_entry->decdat = p[11];
985 	else
986 		zip_entry->decdat = p[17];
987 	zip_entry->compressed_size = archive_le32dec(p + 18);
988 	zip_entry->uncompressed_size = archive_le32dec(p + 22);
989 	filename_length = archive_le16dec(p + 26);
990 	extra_length = archive_le16dec(p + 28);
991 
992 	__archive_read_consume(a, 30);
993 
994 	/* Read the filename. */
995 	if ((h = __archive_read_ahead(a, filename_length, NULL)) == NULL) {
996 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
997 		    "Truncated ZIP file header");
998 		return (ARCHIVE_FATAL);
999 	}
1000 	if (zip_entry->zip_flags & ZIP_UTF8_NAME) {
1001 		/* The filename is stored to be UTF-8. */
1002 		if (zip->sconv_utf8 == NULL) {
1003 			zip->sconv_utf8 =
1004 			    archive_string_conversion_from_charset(
1005 				&a->archive, "UTF-8", 1);
1006 			if (zip->sconv_utf8 == NULL)
1007 				return (ARCHIVE_FATAL);
1008 		}
1009 		sconv = zip->sconv_utf8;
1010 	} else if (zip->sconv != NULL)
1011 		sconv = zip->sconv;
1012 	else
1013 		sconv = zip->sconv_default;
1014 
1015 	if (archive_entry_copy_pathname_l(entry,
1016 	    h, filename_length, sconv) != 0) {
1017 		if (errno == ENOMEM) {
1018 			archive_set_error(&a->archive, ENOMEM,
1019 			    "Can't allocate memory for Pathname");
1020 			return (ARCHIVE_FATAL);
1021 		}
1022 		archive_set_error(&a->archive,
1023 		    ARCHIVE_ERRNO_FILE_FORMAT,
1024 		    "Pathname cannot be converted "
1025 		    "from %s to current locale.",
1026 		    archive_string_conversion_charset_name(sconv));
1027 		ret = ARCHIVE_WARN;
1028 	}
1029 	__archive_read_consume(a, filename_length);
1030 
1031 	/* Read the extra data. */
1032 	if ((h = __archive_read_ahead(a, extra_length, NULL)) == NULL) {
1033 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1034 		    "Truncated ZIP file header");
1035 		return (ARCHIVE_FATAL);
1036 	}
1037 
1038 	if (ARCHIVE_OK != process_extra(a, entry, h, extra_length,
1039 	    zip_entry)) {
1040 		return ARCHIVE_FATAL;
1041 	}
1042 	__archive_read_consume(a, extra_length);
1043 
1044 	/* Work around a bug in Info-Zip: When reading from a pipe, it
1045 	 * stats the pipe instead of synthesizing a file entry. */
1046 	if ((zip_entry->mode & AE_IFMT) == AE_IFIFO) {
1047 		zip_entry->mode &= ~ AE_IFMT;
1048 		zip_entry->mode |= AE_IFREG;
1049 	}
1050 
1051 	/* If the mode is totally empty, set some sane default. */
1052 	if (zip_entry->mode == 0) {
1053 		zip_entry->mode |= 0664;
1054 	}
1055 
1056 	/* Windows archivers sometimes use backslash as the directory
1057 	 * separator. Normalize to slash. */
1058 	if (zip_entry->system == 0 &&
1059 	    (wp = archive_entry_pathname_w(entry)) != NULL) {
1060 		if (wcschr(wp, L'/') == NULL && wcschr(wp, L'\\') != NULL) {
1061 			size_t i;
1062 			struct archive_wstring s;
1063 			archive_string_init(&s);
1064 			archive_wstrcpy(&s, wp);
1065 			for (i = 0; i < archive_strlen(&s); i++) {
1066 				if (s.s[i] == '\\')
1067 					s.s[i] = '/';
1068 			}
1069 			archive_entry_copy_pathname_w(entry, s.s);
1070 			archive_wstring_free(&s);
1071 		}
1072 	}
1073 
1074 	/* Make sure that entries with a trailing '/' are marked as directories
1075 	 * even if the External File Attributes contains bogus values.  If this
1076 	 * is not a directory and there is no type, assume a regular file. */
1077 	if ((zip_entry->mode & AE_IFMT) != AE_IFDIR) {
1078 		int has_slash;
1079 
1080 		wp = archive_entry_pathname_w(entry);
1081 		if (wp != NULL) {
1082 			len = wcslen(wp);
1083 			has_slash = len > 0 && wp[len - 1] == L'/';
1084 		} else {
1085 			cp = archive_entry_pathname(entry);
1086 			len = (cp != NULL)?strlen(cp):0;
1087 			has_slash = len > 0 && cp[len - 1] == '/';
1088 		}
1089 		/* Correct file type as needed. */
1090 		if (has_slash) {
1091 			zip_entry->mode &= ~AE_IFMT;
1092 			zip_entry->mode |= AE_IFDIR;
1093 			zip_entry->mode |= 0111;
1094 		} else if ((zip_entry->mode & AE_IFMT) == 0) {
1095 			zip_entry->mode |= AE_IFREG;
1096 		}
1097 	}
1098 
1099 	/* Make sure directories end in '/' */
1100 	if ((zip_entry->mode & AE_IFMT) == AE_IFDIR) {
1101 		wp = archive_entry_pathname_w(entry);
1102 		if (wp != NULL) {
1103 			len = wcslen(wp);
1104 			if (len > 0 && wp[len - 1] != L'/') {
1105 				struct archive_wstring s;
1106 				archive_string_init(&s);
1107 				archive_wstrcat(&s, wp);
1108 				archive_wstrappend_wchar(&s, L'/');
1109 				archive_entry_copy_pathname_w(entry, s.s);
1110 				archive_wstring_free(&s);
1111 			}
1112 		} else {
1113 			cp = archive_entry_pathname(entry);
1114 			len = (cp != NULL)?strlen(cp):0;
1115 			if (len > 0 && cp[len - 1] != '/') {
1116 				struct archive_string s;
1117 				archive_string_init(&s);
1118 				archive_strcat(&s, cp);
1119 				archive_strappend_char(&s, '/');
1120 				archive_entry_set_pathname(entry, s.s);
1121 				archive_string_free(&s);
1122 			}
1123 		}
1124 	}
1125 
1126 	if (zip_entry->flags & LA_FROM_CENTRAL_DIRECTORY) {
1127 		/* If this came from the central dir, its size info
1128 		 * is definitive, so ignore the length-at-end flag. */
1129 		zip_entry->zip_flags &= ~ZIP_LENGTH_AT_END;
1130 		/* If local header is missing a value, use the one from
1131 		   the central directory.  If both have it, warn about
1132 		   mismatches. */
1133 		if (zip_entry->crc32 == 0) {
1134 			zip_entry->crc32 = zip_entry_central_dir.crc32;
1135 		} else if (!zip->ignore_crc32
1136 		    && zip_entry->crc32 != zip_entry_central_dir.crc32) {
1137 			archive_set_error(&a->archive,
1138 			    ARCHIVE_ERRNO_FILE_FORMAT,
1139 			    "Inconsistent CRC32 values");
1140 			ret = ARCHIVE_WARN;
1141 		}
1142 		if (zip_entry->compressed_size == 0
1143 		    || zip_entry->compressed_size == 0xffffffff) {
1144 			zip_entry->compressed_size
1145 			    = zip_entry_central_dir.compressed_size;
1146 		} else if (zip_entry->compressed_size
1147 		    != zip_entry_central_dir.compressed_size) {
1148 			archive_set_error(&a->archive,
1149 			    ARCHIVE_ERRNO_FILE_FORMAT,
1150 			    "Inconsistent compressed size: "
1151 			    "%jd in central directory, %jd in local header",
1152 			    (intmax_t)zip_entry_central_dir.compressed_size,
1153 			    (intmax_t)zip_entry->compressed_size);
1154 			ret = ARCHIVE_WARN;
1155 		}
1156 		if (zip_entry->uncompressed_size == 0 ||
1157 			zip_entry->uncompressed_size == 0xffffffff) {
1158 			zip_entry->uncompressed_size
1159 			    = zip_entry_central_dir.uncompressed_size;
1160 		} else if (zip_entry->uncompressed_size
1161 		    != zip_entry_central_dir.uncompressed_size) {
1162 			archive_set_error(&a->archive,
1163 			    ARCHIVE_ERRNO_FILE_FORMAT,
1164 			    "Inconsistent uncompressed size: "
1165 			    "%jd in central directory, %jd in local header",
1166 			    (intmax_t)zip_entry_central_dir.uncompressed_size,
1167 			    (intmax_t)zip_entry->uncompressed_size);
1168 			ret = ARCHIVE_WARN;
1169 		}
1170 	}
1171 
1172 	/* Populate some additional entry fields: */
1173 	archive_entry_set_mode(entry, zip_entry->mode);
1174 	archive_entry_set_uid(entry, zip_entry->uid);
1175 	archive_entry_set_gid(entry, zip_entry->gid);
1176 	archive_entry_set_mtime(entry, zip_entry->mtime, 0);
1177 	archive_entry_set_ctime(entry, zip_entry->ctime, 0);
1178 	archive_entry_set_atime(entry, zip_entry->atime, 0);
1179 
1180 	if ((zip->entry->mode & AE_IFMT) == AE_IFLNK) {
1181 		size_t linkname_length;
1182 
1183 		if (zip_entry->compressed_size > 64 * 1024) {
1184 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1185 			    "Zip file with oversized link entry");
1186 			return ARCHIVE_FATAL;
1187 		}
1188 
1189 		linkname_length = (size_t)zip_entry->compressed_size;
1190 
1191 		archive_entry_set_size(entry, 0);
1192 
1193 		// take into account link compression if any
1194 		size_t linkname_full_length = linkname_length;
1195 		if (zip->entry->compression != 0)
1196 		{
1197 			// symlink target string appeared to be compressed
1198 			int status = ARCHIVE_FATAL;
1199 			const void *uncompressed_buffer = NULL;
1200 
1201 			switch (zip->entry->compression)
1202 			{
1203 #if HAVE_ZLIB_H
1204 				case 8: /* Deflate compression. */
1205 					zip->entry_bytes_remaining = zip_entry->compressed_size;
1206 					status = zip_read_data_deflate(a, &uncompressed_buffer,
1207 						&linkname_full_length, NULL);
1208 					break;
1209 #endif
1210 #if HAVE_LZMA_H && HAVE_LIBLZMA
1211 				case 14: /* ZIPx LZMA compression. */
1212 					/*(see zip file format specification, section 4.4.5)*/
1213 					zip->entry_bytes_remaining = zip_entry->compressed_size;
1214 					status = zip_read_data_zipx_lzma_alone(a, &uncompressed_buffer,
1215 						&linkname_full_length, NULL);
1216 					break;
1217 #endif
1218 				default: /* Unsupported compression. */
1219 					break;
1220 			}
1221 			if (status == ARCHIVE_OK)
1222 			{
1223 				p = uncompressed_buffer;
1224 			}
1225 			else
1226 			{
1227 				archive_set_error(&a->archive,
1228 					ARCHIVE_ERRNO_FILE_FORMAT,
1229 					"Unsupported ZIP compression method "
1230 					"during decompression of link entry (%d: %s)",
1231 					zip->entry->compression,
1232 					compression_name(zip->entry->compression));
1233 				return ARCHIVE_FAILED;
1234 			}
1235 		}
1236 		else
1237 		{
1238 			p = __archive_read_ahead(a, linkname_length, NULL);
1239 		}
1240 
1241 		if (p == NULL) {
1242 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1243 			    "Truncated Zip file");
1244 			return ARCHIVE_FATAL;
1245 		}
1246 
1247 		sconv = zip->sconv;
1248 		if (sconv == NULL && (zip->entry->zip_flags & ZIP_UTF8_NAME))
1249 			sconv = zip->sconv_utf8;
1250 		if (sconv == NULL)
1251 			sconv = zip->sconv_default;
1252 		if (archive_entry_copy_symlink_l(entry, p, linkname_full_length,
1253 		    sconv) != 0) {
1254 			if (errno != ENOMEM && sconv == zip->sconv_utf8 &&
1255 			    (zip->entry->zip_flags & ZIP_UTF8_NAME))
1256 			    archive_entry_copy_symlink_l(entry, p,
1257 				linkname_full_length, NULL);
1258 			if (errno == ENOMEM) {
1259 				archive_set_error(&a->archive, ENOMEM,
1260 				    "Can't allocate memory for Symlink");
1261 				return (ARCHIVE_FATAL);
1262 			}
1263 			/*
1264 			 * Since there is no character-set regulation for
1265 			 * symlink name, do not report the conversion error
1266 			 * in an automatic conversion.
1267 			 */
1268 			if (sconv != zip->sconv_utf8 ||
1269 			    (zip->entry->zip_flags & ZIP_UTF8_NAME) == 0) {
1270 				archive_set_error(&a->archive,
1271 				    ARCHIVE_ERRNO_FILE_FORMAT,
1272 				    "Symlink cannot be converted "
1273 				    "from %s to current locale.",
1274 				    archive_string_conversion_charset_name(
1275 					sconv));
1276 				ret = ARCHIVE_WARN;
1277 			}
1278 		}
1279 		zip_entry->uncompressed_size = zip_entry->compressed_size = 0;
1280 
1281 		if (__archive_read_consume(a, linkname_length) < 0) {
1282 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1283 			    "Read error skipping symlink target name");
1284 			return ARCHIVE_FATAL;
1285 		}
1286 	} else if (0 == (zip_entry->zip_flags & ZIP_LENGTH_AT_END)
1287 	   || (zip_entry->uncompressed_size > 0
1288 	       && zip_entry->uncompressed_size != 0xffffffff)) {
1289 		/* Set the size only if it's meaningful. */
1290 		archive_entry_set_size(entry, zip_entry->uncompressed_size);
1291 	}
1292 	zip->entry_bytes_remaining = zip_entry->compressed_size;
1293 
1294 	/* If there's no body, force read_data() to return EOF immediately. */
1295 	if (0 == (zip_entry->zip_flags & ZIP_LENGTH_AT_END)
1296 	    && zip->entry_bytes_remaining < 1)
1297 		zip->end_of_entry = 1;
1298 
1299 	/* Set up a more descriptive format name. */
1300         archive_string_empty(&zip->format_name);
1301 	archive_string_sprintf(&zip->format_name, "ZIP %d.%d (%s)",
1302 	    version / 10, version % 10,
1303 	    compression_name(zip->entry->compression));
1304 	a->archive.archive_format_name = zip->format_name.s;
1305 
1306 	return (ret);
1307 }
1308 
1309 static int
check_authentication_code(struct archive_read * a,const void * _p)1310 check_authentication_code(struct archive_read *a, const void *_p)
1311 {
1312 	struct zip *zip = (struct zip *)(a->format->data);
1313 
1314 	/* Check authentication code. */
1315 	if (zip->hctx_valid) {
1316 		const void *p;
1317 		uint8_t hmac[20];
1318 		size_t hmac_len = 20;
1319 		int cmp;
1320 
1321 		archive_hmac_sha1_final(&zip->hctx, hmac, &hmac_len);
1322 		if (_p == NULL) {
1323 			/* Read authentication code. */
1324 			p = __archive_read_ahead(a, AUTH_CODE_SIZE, NULL);
1325 			if (p == NULL) {
1326 				archive_set_error(&a->archive,
1327 				    ARCHIVE_ERRNO_FILE_FORMAT,
1328 				    "Truncated ZIP file data");
1329 				return (ARCHIVE_FATAL);
1330 			}
1331 		} else {
1332 			p = _p;
1333 		}
1334 		cmp = memcmp(hmac, p, AUTH_CODE_SIZE);
1335 		__archive_read_consume(a, AUTH_CODE_SIZE);
1336 		if (cmp != 0) {
1337 			archive_set_error(&a->archive,
1338 			    ARCHIVE_ERRNO_MISC,
1339 			    "ZIP bad Authentication code");
1340 			return (ARCHIVE_WARN);
1341 		}
1342 	}
1343 	return (ARCHIVE_OK);
1344 }
1345 
1346 /*
1347  * The Zip end-of-file marker is inherently ambiguous.  The specification
1348  * in APPNOTE.TXT allows any of four possible formats, and there is no
1349  * guaranteed-correct way for a reader to know a priori which one the writer
1350  * will have used.  The four formats are:
1351  * 1. 32-bit format with an initial PK78 marker
1352  * 2. 32-bit format without that marker
1353  * 3. 64-bit format with the marker
1354  * 4. 64-bit format without the marker
1355  *
1356  * Mark Adler's `sunzip` streaming unzip program solved this ambiguity
1357  * by just looking at every possible combination and accepting the
1358  * longest one that matches the expected values.  His approach always
1359  * consumes the longest possible matching EOF marker, based on an
1360  * analysis of all the possible failures and how the values could
1361  * overlap.
1362  *
1363  * For example, suppose both of the first two formats listed
1364  * above match.  In that case, we know the next four
1365  * 32-bit words match this pattern:
1366  * ```
1367  *  [PK\07\08] [CRC32]        [compressed size]   [uncompressed size]
1368  * ```
1369  * but we know they must also match this pattern:
1370  * ```
1371  *  [CRC32] [compressed size] [uncompressed size] [other PK marker]
1372  * ```
1373  *
1374  * Since the first word here matches both the PK78 signature in the
1375  * first form and the CRC32 in the second, we know those two values
1376  * are equal, the CRC32 must be exactly 0x08074b50.  Similarly, the
1377  * compressed and uncompressed size must also be exactly this value.
1378  * So we know these four words are all 0x08074b50.  If we were to
1379  * accept the shorter pattern, it would be immediately followed by
1380  * another PK78 marker, which is not possible in a well-formed ZIP
1381  * archive unless there is garbage between entries. This implies we
1382  * should not accept the shorter form in such a case; we should accept
1383  * the longer form.
1384  *
1385  * If the second and third possibilities above both match, we
1386  * have a slightly different situation.  The following words
1387  * must match both the 32-bit format
1388  * ```
1389  *  [CRC32] [compressed size] [uncompressed size] [other PK marker]
1390  * ```
1391  * and the 64-bit format
1392  * ```
1393  *  [CRC32] [compressed low] [compressed high] [uncompressed low] [uncompressed high] [other PK marker]
1394  * ```
1395  * Since the 32-bit and 64-bit compressed sizes both match, the
1396  * actual size must fit in 32 bits, which implies the high-order
1397  * word of the compressed size is zero.  So we know the uncompressed
1398  * low word is zero, which again implies that if we accept the shorter
1399  * format, there will not be a valid PK marker following it.
1400  *
1401  * Similar considerations rule out the shorter form in every other
1402  * possibly-ambiguous pair.  So if two of the four possible formats
1403  * match, we should accept the longer option.
1404  *
1405  * If none of the four formats matches, we know the archive must be
1406  * corrupted in some fashion.  In particular, it's possible that the
1407  * length-at-end bit was incorrect and we should not really be looking
1408  * for an EOF marker at all.  To allow for this possibility, we
1409  * evaluate the following words to collect data for a later error
1410  * report but do not consume any bytes.  We instead rely on the later
1411  * search for a new PK marker to re-sync to the next well-formed
1412  * entry.
1413  */
1414 static void
consume_end_of_file_marker(struct archive_read * a,struct zip * zip)1415 consume_end_of_file_marker(struct archive_read *a, struct zip *zip)
1416 {
1417 	const char *marker;
1418 	const char *p;
1419 	uint64_t compressed32, uncompressed32;
1420 	uint64_t compressed64, uncompressed64;
1421 	uint64_t compressed_actual, uncompressed_actual;
1422 	uint32_t crc32_actual;
1423 	const uint32_t PK78 = 0x08074B50ULL;
1424 	uint8_t crc32_ignored, crc32_may_be_zero;
1425 
1426 	/* If there shouldn't be a marker, don't consume it. */
1427 	if ((zip->entry->zip_flags & ZIP_LENGTH_AT_END) == 0) {
1428 		return;
1429 	}
1430 
1431 	/* The longest Zip end-of-file record is 24 bytes.  Since an
1432 	 * end-of-file record can never appear at the end of the
1433 	 * archive, we know 24 bytes will be available unless
1434 	 * the archive is severely truncated. */
1435 	if (NULL == (marker = __archive_read_ahead(a, 24, NULL))) {
1436 		return;
1437 	}
1438 	p = marker;
1439 
1440 	/* The end-of-file record comprises:
1441 	 * = Optional PK\007\010 marker
1442 	 * = 4-byte CRC32
1443 	 * = Compressed size
1444 	 * = Uncompressed size
1445 	 *
1446 	 * The last two fields are either both 32 bits or both 64
1447 	 * bits.  We check all possible layouts and accept any one
1448 	 * that gives us a complete match, else we make a best-effort
1449 	 * attempt to parse out the pieces.
1450 	 */
1451 
1452 	/* CRC32 checking can be tricky:
1453 	 * * Test suites sometimes ignore the CRC32
1454 	 * * AES AE-2 always writes zero for the CRC32
1455 	 * * AES AE-1 sometimes writes zero for the CRC32
1456 	 */
1457 	crc32_ignored = zip->ignore_crc32;
1458 	crc32_may_be_zero = 0;
1459 	crc32_actual = zip->computed_crc32;
1460 	if (zip->hctx_valid) {
1461 	  switch (zip->entry->aes_extra.vendor) {
1462 	  case AES_VENDOR_AE_2:
1463 	    crc32_actual = 0;
1464 	    break;
1465 	  case AES_VENDOR_AE_1:
1466 	  default:
1467 	    crc32_may_be_zero = 1;
1468 	    break;
1469 	  }
1470 	}
1471 
1472 	/* Values computed from the actual data in the archive. */
1473 	compressed_actual = (uint64_t)zip->entry_compressed_bytes_read;
1474 	uncompressed_actual = (uint64_t)zip->entry_uncompressed_bytes_read;
1475 
1476 
1477 	/* Longest: PK78 marker, all 64-bit fields (24 bytes total) */
1478 	if (archive_le32dec(p) == PK78
1479 	    && ((archive_le32dec(p + 4) == crc32_actual)
1480 		|| (crc32_may_be_zero && (archive_le32dec(p + 4) == 0))
1481 		|| crc32_ignored)
1482 	    && (archive_le64dec(p + 8) == compressed_actual)
1483 	    && (archive_le64dec(p + 16) == uncompressed_actual)) {
1484 		if (!crc32_ignored) {
1485 			zip->entry->crc32 = crc32_actual;
1486 		}
1487 		zip->entry->compressed_size = compressed_actual;
1488 		zip->entry->uncompressed_size = uncompressed_actual;
1489 		zip->unconsumed += 24;
1490 		return;
1491 	}
1492 
1493 	/* No PK78 marker, 64-bit fields (20 bytes total) */
1494 	if (((archive_le32dec(p) == crc32_actual)
1495 	     || (crc32_may_be_zero && (archive_le32dec(p + 4) == 0))
1496 	     || crc32_ignored)
1497 	    && (archive_le64dec(p + 4) == compressed_actual)
1498 	    && (archive_le64dec(p + 12) == uncompressed_actual)) {
1499 	        if (!crc32_ignored) {
1500 			zip->entry->crc32 = crc32_actual;
1501 		}
1502 		zip->entry->compressed_size = compressed_actual;
1503 		zip->entry->uncompressed_size = uncompressed_actual;
1504 		zip->unconsumed += 20;
1505 		return;
1506 	}
1507 
1508 	/* PK78 marker and 32-bit fields (16 bytes total) */
1509 	if (archive_le32dec(p) == PK78
1510 	    && ((archive_le32dec(p + 4) == crc32_actual)
1511 		|| (crc32_may_be_zero && (archive_le32dec(p + 4) == 0))
1512 		|| crc32_ignored)
1513 	    && (archive_le32dec(p + 8) == compressed_actual)
1514 	    && (archive_le32dec(p + 12) == uncompressed_actual)) {
1515 		if (!crc32_ignored) {
1516 			zip->entry->crc32 = crc32_actual;
1517 		}
1518 		zip->entry->compressed_size = compressed_actual;
1519 		zip->entry->uncompressed_size = uncompressed_actual;
1520 		zip->unconsumed += 16;
1521 		return;
1522 	}
1523 
1524 	/* Shortest: No PK78 marker, all 32-bit fields (12 bytes total) */
1525 	if (((archive_le32dec(p) == crc32_actual)
1526 	     || (crc32_may_be_zero && (archive_le32dec(p + 4) == 0))
1527 	     || crc32_ignored)
1528 	    && (archive_le32dec(p + 4) == compressed_actual)
1529 	    && (archive_le32dec(p + 8) == uncompressed_actual)) {
1530 		if (!crc32_ignored) {
1531 			zip->entry->crc32 = crc32_actual;
1532 		}
1533 		zip->entry->compressed_size = compressed_actual;
1534 		zip->entry->uncompressed_size = uncompressed_actual;
1535 		zip->unconsumed += 12;
1536 		return;
1537 	}
1538 
1539 	/* If none of the above patterns gives us a full exact match,
1540 	 * then there's something definitely amiss.  The fallback code
1541 	 * below will parse out some plausible values for error
1542 	 * reporting purposes.  Note that this won't actually
1543 	 * consume anything:
1544 	 *
1545 	 * = If there really is a marker here, the logic to resync to
1546 	 *   the next entry will suffice to skip it.
1547 	 *
1548 	 * = There might not really be a marker: Corruption or bugs
1549 	 *   may have set the length-at-end bit without a marker ever
1550 	 *   having actually been written. In this case, we
1551 	 *   explicitly should not consume any bytes, since that would
1552 	 *   prevent us from correctly reading the next entry.
1553 	 */
1554 	if (archive_le32dec(p) == PK78) {
1555 		p += 4; /* Ignore PK78 if it appears to be present */
1556 	}
1557 	zip->entry->crc32 = archive_le32dec(p);  /* Parse CRC32 */
1558 	p += 4;
1559 
1560 	/* Consider both 32- and 64-bit interpretations */
1561 	compressed32 = archive_le32dec(p);
1562 	uncompressed32 = archive_le32dec(p + 4);
1563 	compressed64 = archive_le64dec(p);
1564 	uncompressed64 = archive_le64dec(p + 8);
1565 
1566 	/* The earlier patterns may have failed because of CRC32
1567 	 * mismatch, so it's still possible that both sizes match.
1568 	 * Try to match as many as we can...
1569 	 */
1570 	if (compressed32 == compressed_actual
1571 	    && uncompressed32 == uncompressed_actual) {
1572 		/* Both 32-bit fields match */
1573 		zip->entry->compressed_size = compressed32;
1574 		zip->entry->uncompressed_size = uncompressed32;
1575 	} else if (compressed64 == compressed_actual
1576 		   || uncompressed64 == uncompressed_actual) {
1577 		/* One or both 64-bit fields match */
1578 		zip->entry->compressed_size = compressed64;
1579 		zip->entry->uncompressed_size = uncompressed64;
1580 	} else {
1581 		/* Zero or one 32-bit fields match */
1582 		zip->entry->compressed_size = compressed32;
1583 		zip->entry->uncompressed_size = uncompressed32;
1584 	}
1585 }
1586 
1587 /*
1588  * Read "uncompressed" data.
1589  *
1590  * This is straightforward if we know the size of the data.  This is
1591  * always true for the seeking reader (we've examined the Central
1592  * Directory already), and will often be true for the streaming reader
1593  * (the writer was writing uncompressed so probably knows the size).
1594  *
1595  * If we don't know the size, then life is more interesting.  Note
1596  * that a careful reading of the Zip specification says that a writer
1597  * must use ZIP_LENGTH_AT_END if it cannot write the CRC into the
1598  * local header.  And if it uses ZIP_LENGTH_AT_END, then it is
1599  * prohibited from storing the sizes in the local header.  This
1600  * prevents fully-compliant streaming writers from providing any size
1601  * clues to a streaming reader.  In this case, we have to scan the
1602  * data as we read to try to locate the end-of-file marker.
1603  *
1604  * We assume here that the end-of-file marker always has the
1605  * PK\007\010 signature.  Although it's technically optional, newer
1606  * writers seem to provide it pretty consistently, and it's not clear
1607  * how to efficiently recognize an end-of-file marker that lacks it.
1608  *
1609  * Returns ARCHIVE_OK if successful, ARCHIVE_FATAL otherwise, sets
1610  * zip->end_of_entry if it consumes all of the data.
1611  */
1612 static int
zip_read_data_none(struct archive_read * a,const void ** _buff,size_t * size,int64_t * offset)1613 zip_read_data_none(struct archive_read *a, const void **_buff,
1614     size_t *size, int64_t *offset)
1615 {
1616 	struct zip *zip;
1617 	const char *buff;
1618 	ssize_t bytes_avail;
1619 	ssize_t trailing_extra;
1620 	int r;
1621 
1622 	(void)offset; /* UNUSED */
1623 
1624 	zip = (struct zip *)(a->format->data);
1625 	trailing_extra = zip->hctx_valid ? AUTH_CODE_SIZE : 0;
1626 
1627 	if (zip->entry->zip_flags & ZIP_LENGTH_AT_END) {
1628 		const char *p;
1629 		ssize_t grabbing_bytes = 24 + trailing_extra;
1630 
1631 		/* Grab at least 24 bytes. */
1632 		buff = __archive_read_ahead(a, grabbing_bytes, &bytes_avail);
1633 		if (bytes_avail < grabbing_bytes) {
1634 			/* Zip archives have end-of-archive markers
1635 			   that are longer than this, so a failure to get at
1636 			   least 24 bytes really does indicate a truncated
1637 			   file. */
1638 			archive_set_error(&a->archive,
1639 			    ARCHIVE_ERRNO_FILE_FORMAT,
1640 			    "Truncated ZIP file data");
1641 			return (ARCHIVE_FATAL);
1642 		}
1643 		/* Check for a complete PK\007\010 signature, followed
1644 		 * by the correct 4-byte CRC. */
1645 		p = buff + trailing_extra;
1646 		if (p[0] == 'P' && p[1] == 'K'
1647 		    && p[2] == '\007' && p[3] == '\010'
1648 		    && (archive_le32dec(p + 4) == zip->computed_crc32
1649 			|| zip->ignore_crc32
1650 			|| (zip->hctx_valid
1651 			 && zip->entry->aes_extra.vendor == AES_VENDOR_AE_2))) {
1652 			zip->end_of_entry = 1;
1653 			if (zip->hctx_valid) {
1654 				r = check_authentication_code(a, buff);
1655 				if (r != ARCHIVE_OK)
1656 					return (r);
1657 			}
1658 			return (ARCHIVE_OK);
1659 		}
1660 		/* If not at EOF, ensure we consume at least one byte. */
1661 		++p;
1662 
1663 		/* Scan forward until we see where a PK\007\010 signature
1664 		 * might be. */
1665 		/* Return bytes up until that point.  On the next call,
1666 		 * the code above will verify the data descriptor. */
1667 		while (p < buff + bytes_avail - 4) {
1668 			if (p[3] == 'P') { p += 3; }
1669 			else if (p[3] == 'K') { p += 2; }
1670 			else if (p[3] == '\007') { p += 1; }
1671 			else if (p[3] == '\010' && p[2] == '\007'
1672 			    && p[1] == 'K' && p[0] == 'P') {
1673 				break;
1674 			} else { p += 4; }
1675 		}
1676 		p -= trailing_extra;
1677 		bytes_avail = p - buff;
1678 	} else {
1679 		if (zip->entry_bytes_remaining == 0) {
1680 			zip->end_of_entry = 1;
1681 			if (zip->hctx_valid) {
1682 				r = check_authentication_code(a, NULL);
1683 				if (r != ARCHIVE_OK)
1684 					return (r);
1685 			}
1686 			return (ARCHIVE_OK);
1687 		}
1688 		/* Grab a bunch of bytes. */
1689 		buff = __archive_read_ahead(a, 1, &bytes_avail);
1690 		if (bytes_avail <= 0) {
1691 			archive_set_error(&a->archive,
1692 			    ARCHIVE_ERRNO_FILE_FORMAT,
1693 			    "Truncated ZIP file data");
1694 			return (ARCHIVE_FATAL);
1695 		}
1696 		if (bytes_avail > zip->entry_bytes_remaining)
1697 			bytes_avail = (ssize_t)zip->entry_bytes_remaining;
1698 	}
1699 	if (zip->tctx_valid || zip->cctx_valid) {
1700 		size_t dec_size = bytes_avail;
1701 
1702 		if (dec_size > zip->decrypted_buffer_size)
1703 			dec_size = zip->decrypted_buffer_size;
1704 		if (zip->tctx_valid) {
1705 			trad_enc_decrypt_update(&zip->tctx,
1706 			    (const uint8_t *)buff, dec_size,
1707 			    zip->decrypted_buffer, dec_size);
1708 		} else {
1709 			size_t dsize = dec_size;
1710 			archive_hmac_sha1_update(&zip->hctx,
1711 			    (const uint8_t *)buff, dec_size);
1712 			archive_decrypto_aes_ctr_update(&zip->cctx,
1713 			    (const uint8_t *)buff, dec_size,
1714 			    zip->decrypted_buffer, &dsize);
1715 		}
1716 		bytes_avail = dec_size;
1717 		buff = (const char *)zip->decrypted_buffer;
1718 	}
1719 	zip->entry_bytes_remaining -= bytes_avail;
1720 	zip->entry_uncompressed_bytes_read += bytes_avail;
1721 	zip->entry_compressed_bytes_read += bytes_avail;
1722 	zip->unconsumed += bytes_avail;
1723 	*size = bytes_avail;
1724 	*_buff = buff;
1725 	return (ARCHIVE_OK);
1726 }
1727 
1728 #if HAVE_LZMA_H && HAVE_LIBLZMA
1729 static int
zipx_xz_init(struct archive_read * a,struct zip * zip)1730 zipx_xz_init(struct archive_read *a, struct zip *zip)
1731 {
1732 	lzma_ret r;
1733 
1734 	if(zip->zipx_lzma_valid) {
1735 		lzma_end(&zip->zipx_lzma_stream);
1736 		zip->zipx_lzma_valid = 0;
1737 	}
1738 
1739 	memset(&zip->zipx_lzma_stream, 0, sizeof(zip->zipx_lzma_stream));
1740 	r = lzma_stream_decoder(&zip->zipx_lzma_stream, UINT64_MAX, 0);
1741 	if (r != LZMA_OK) {
1742 		archive_set_error(&(a->archive), ARCHIVE_ERRNO_MISC,
1743 		    "xz initialization failed(%d)",
1744 		    r);
1745 
1746 		return (ARCHIVE_FAILED);
1747 	}
1748 
1749 	zip->zipx_lzma_valid = 1;
1750 
1751 	free(zip->uncompressed_buffer);
1752 
1753 	zip->uncompressed_buffer_size = 256 * 1024;
1754 	zip->uncompressed_buffer =
1755 	    (uint8_t*) malloc(zip->uncompressed_buffer_size);
1756 	if (zip->uncompressed_buffer == NULL) {
1757 		archive_set_error(&a->archive, ENOMEM,
1758 		    "No memory for xz decompression");
1759 		    return (ARCHIVE_FATAL);
1760 	}
1761 
1762 	zip->decompress_init = 1;
1763 	return (ARCHIVE_OK);
1764 }
1765 
1766 static int
zipx_lzma_alone_init(struct archive_read * a,struct zip * zip)1767 zipx_lzma_alone_init(struct archive_read *a, struct zip *zip)
1768 {
1769 	lzma_ret r;
1770 	const uint8_t* p;
1771 
1772 #pragma pack(push)
1773 #pragma pack(1)
1774 	struct _alone_header {
1775 	    uint8_t bytes[5];
1776 	    uint64_t uncompressed_size;
1777 	} alone_header;
1778 #pragma pack(pop)
1779 
1780 	if(zip->zipx_lzma_valid) {
1781 		lzma_end(&zip->zipx_lzma_stream);
1782 		zip->zipx_lzma_valid = 0;
1783 	}
1784 
1785 	/* To unpack ZIPX's "LZMA" (id 14) stream we can use standard liblzma
1786 	 * that is a part of XZ Utils. The stream format stored inside ZIPX
1787 	 * file is a modified "lzma alone" file format, that was used by the
1788 	 * `lzma` utility which was later deprecated in favour of `xz` utility.
1789  	 * Since those formats are nearly the same, we can use a standard
1790 	 * "lzma alone" decoder from XZ Utils. */
1791 
1792 	memset(&zip->zipx_lzma_stream, 0, sizeof(zip->zipx_lzma_stream));
1793 	r = lzma_alone_decoder(&zip->zipx_lzma_stream, UINT64_MAX);
1794 	if (r != LZMA_OK) {
1795 		archive_set_error(&(a->archive), ARCHIVE_ERRNO_MISC,
1796 		    "lzma initialization failed(%d)", r);
1797 
1798 		return (ARCHIVE_FAILED);
1799 	}
1800 
1801 	/* Flag the cleanup function that we want our lzma-related structures
1802 	 * to be freed later. */
1803 	zip->zipx_lzma_valid = 1;
1804 
1805 	/* The "lzma alone" file format and the stream format inside ZIPx are
1806 	 * almost the same. Here's an example of a structure of "lzma alone"
1807 	 * format:
1808 	 *
1809 	 * $ cat /bin/ls | lzma | xxd | head -n 1
1810 	 * 00000000: 5d00 0080 00ff ffff ffff ffff ff00 2814
1811 	 *
1812 	 *    5 bytes        8 bytes        n bytes
1813 	 * <lzma_params><uncompressed_size><data...>
1814 	 *
1815 	 * lzma_params is a 5-byte blob that has to be decoded to extract
1816 	 * parameters of this LZMA stream. The uncompressed_size field is an
1817 	 * uint64_t value that contains information about the size of the
1818 	 * uncompressed file, or UINT64_MAX if this value is unknown.
1819 	 * The <data...> part is the actual lzma-compressed data stream.
1820 	 *
1821 	 * Now here's the structure of the stream inside the ZIPX file:
1822 	 *
1823 	 * $ cat stream_inside_zipx | xxd | head -n 1
1824 	 * 00000000: 0914 0500 5d00 8000 0000 2814 .... ....
1825 	 *
1826 	 *  2byte   2byte    5 bytes     n bytes
1827 	 * <magic1><magic2><lzma_params><data...>
1828 	 *
1829 	 * This means that the ZIPX file contains an additional magic1 and
1830 	 * magic2 headers, the lzma_params field contains the same parameter
1831 	 * set as in the "lzma alone" format, and the <data...> field is the
1832 	 * same as in the "lzma alone" format as well. Note that also the zipx
1833 	 * format is missing the uncompressed_size field.
1834 	 *
1835 	 * So, in order to use the "lzma alone" decoder for the zipx lzma
1836 	 * stream, we simply need to shuffle around some fields, prepare a new
1837 	 * lzma alone header, feed it into lzma alone decoder so it will
1838 	 * initialize itself properly, and then we can start feeding normal
1839 	 * zipx lzma stream into the decoder.
1840 	 */
1841 
1842 	/* Read magic1,magic2,lzma_params from the ZIPX stream. */
1843 	if(zip->entry_bytes_remaining < 9 || (p = __archive_read_ahead(a, 9, NULL)) == NULL) {
1844 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1845 		    "Truncated lzma data");
1846 		return (ARCHIVE_FATAL);
1847 	}
1848 
1849 	if(p[2] != 0x05 || p[3] != 0x00) {
1850 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1851 		    "Invalid lzma data");
1852 		return (ARCHIVE_FATAL);
1853 	}
1854 
1855 	/* Prepare an lzma alone header: copy the lzma_params blob into
1856 	 * a proper place into the lzma alone header. */
1857 	memcpy(&alone_header.bytes[0], p + 4, 5);
1858 
1859 	/* Initialize the 'uncompressed size' field to unknown; we'll manually
1860 	 * monitor how many bytes there are still to be uncompressed. */
1861 	alone_header.uncompressed_size = UINT64_MAX;
1862 
1863 	if(!zip->uncompressed_buffer) {
1864 		zip->uncompressed_buffer_size = 256 * 1024;
1865 		zip->uncompressed_buffer =
1866 			(uint8_t*) malloc(zip->uncompressed_buffer_size);
1867 
1868 		if (zip->uncompressed_buffer == NULL) {
1869 			archive_set_error(&a->archive, ENOMEM,
1870 			    "No memory for lzma decompression");
1871 			return (ARCHIVE_FATAL);
1872 		}
1873 	}
1874 
1875 	zip->zipx_lzma_stream.next_in = (void*) &alone_header;
1876 	zip->zipx_lzma_stream.avail_in = sizeof(alone_header);
1877 	zip->zipx_lzma_stream.total_in = 0;
1878 	zip->zipx_lzma_stream.next_out = zip->uncompressed_buffer;
1879 	zip->zipx_lzma_stream.avail_out = zip->uncompressed_buffer_size;
1880 	zip->zipx_lzma_stream.total_out = 0;
1881 
1882 	/* Feed only the header into the lzma alone decoder. This will
1883 	 * effectively initialize the decoder, and will not produce any
1884 	 * output bytes yet. */
1885 	r = lzma_code(&zip->zipx_lzma_stream, LZMA_RUN);
1886 	if (r != LZMA_OK) {
1887 		archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
1888 		    "lzma stream initialization error");
1889 		return ARCHIVE_FATAL;
1890 	}
1891 
1892 	/* We've already consumed some bytes, so take this into account. */
1893 	__archive_read_consume(a, 9);
1894 	zip->entry_bytes_remaining -= 9;
1895 	zip->entry_compressed_bytes_read += 9;
1896 
1897 	zip->decompress_init = 1;
1898 	return (ARCHIVE_OK);
1899 }
1900 
1901 static int
zip_read_data_zipx_xz(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)1902 zip_read_data_zipx_xz(struct archive_read *a, const void **buff,
1903 	size_t *size, int64_t *offset)
1904 {
1905 	struct zip* zip = (struct zip *)(a->format->data);
1906 	int ret;
1907 	lzma_ret lz_ret;
1908 	const void* compressed_buf;
1909 	ssize_t bytes_avail, in_bytes, to_consume = 0;
1910 
1911 	(void) offset; /* UNUSED */
1912 
1913 	/* Initialize decompressor if not yet initialized. */
1914 	if (!zip->decompress_init) {
1915 		ret = zipx_xz_init(a, zip);
1916 		if (ret != ARCHIVE_OK)
1917 			return (ret);
1918 	}
1919 
1920 	compressed_buf = __archive_read_ahead(a, 1, &bytes_avail);
1921 	if (bytes_avail < 0) {
1922 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1923 		    "Truncated xz file body");
1924 		return (ARCHIVE_FATAL);
1925 	}
1926 
1927 	in_bytes = zipmin(zip->entry_bytes_remaining, bytes_avail);
1928 	zip->zipx_lzma_stream.next_in = compressed_buf;
1929 	zip->zipx_lzma_stream.avail_in = in_bytes;
1930 	zip->zipx_lzma_stream.total_in = 0;
1931 	zip->zipx_lzma_stream.next_out = zip->uncompressed_buffer;
1932 	zip->zipx_lzma_stream.avail_out = zip->uncompressed_buffer_size;
1933 	zip->zipx_lzma_stream.total_out = 0;
1934 
1935 	/* Perform the decompression. */
1936 	lz_ret = lzma_code(&zip->zipx_lzma_stream, LZMA_RUN);
1937 	switch(lz_ret) {
1938 		case LZMA_DATA_ERROR:
1939 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1940 			    "xz data error (error %d)", (int) lz_ret);
1941 			return (ARCHIVE_FATAL);
1942 
1943 		case LZMA_NO_CHECK:
1944 		case LZMA_OK:
1945 			break;
1946 
1947 		default:
1948 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1949 			    "xz unknown error %d", (int) lz_ret);
1950 			return (ARCHIVE_FATAL);
1951 
1952 		case LZMA_STREAM_END:
1953 			lzma_end(&zip->zipx_lzma_stream);
1954 			zip->zipx_lzma_valid = 0;
1955 
1956 			if((int64_t) zip->zipx_lzma_stream.total_in !=
1957 			    zip->entry_bytes_remaining)
1958 			{
1959 				archive_set_error(&a->archive,
1960 				    ARCHIVE_ERRNO_MISC,
1961 				    "xz premature end of stream");
1962 				return (ARCHIVE_FATAL);
1963 			}
1964 
1965 			zip->end_of_entry = 1;
1966 			break;
1967 	}
1968 
1969 	to_consume = zip->zipx_lzma_stream.total_in;
1970 
1971 	__archive_read_consume(a, to_consume);
1972 	zip->entry_bytes_remaining -= to_consume;
1973 	zip->entry_compressed_bytes_read += to_consume;
1974 	zip->entry_uncompressed_bytes_read += zip->zipx_lzma_stream.total_out;
1975 
1976 	*size = zip->zipx_lzma_stream.total_out;
1977 	*buff = zip->uncompressed_buffer;
1978 
1979 	return (ARCHIVE_OK);
1980 }
1981 
1982 static int
zip_read_data_zipx_lzma_alone(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)1983 zip_read_data_zipx_lzma_alone(struct archive_read *a, const void **buff,
1984     size_t *size, int64_t *offset)
1985 {
1986 	struct zip* zip = (struct zip *)(a->format->data);
1987 	int ret;
1988 	lzma_ret lz_ret;
1989 	const void* compressed_buf;
1990 	ssize_t bytes_avail, in_bytes, to_consume;
1991 
1992 	(void) offset; /* UNUSED */
1993 
1994 	/* Initialize decompressor if not yet initialized. */
1995 	if (!zip->decompress_init) {
1996 		ret = zipx_lzma_alone_init(a, zip);
1997 		if (ret != ARCHIVE_OK)
1998 			return (ret);
1999 	}
2000 
2001 	/* Fetch more compressed data. The same note as in deflate handler
2002 	 * applies here as well:
2003 	 *
2004 	 * Note: '1' here is a performance optimization. Recall that the
2005 	 * decompression layer returns a count of available bytes; asking for
2006 	 * more than that forces the decompressor to combine reads by copying
2007 	 * data.
2008 	 */
2009 	compressed_buf = __archive_read_ahead(a, 1, &bytes_avail);
2010 	if (bytes_avail < 0) {
2011 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2012 		    "Truncated lzma file body");
2013 		return (ARCHIVE_FATAL);
2014 	}
2015 
2016 	/* Set decompressor parameters. */
2017 	in_bytes = zipmin(zip->entry_bytes_remaining, bytes_avail);
2018 
2019 	zip->zipx_lzma_stream.next_in = compressed_buf;
2020 	zip->zipx_lzma_stream.avail_in = in_bytes;
2021 	zip->zipx_lzma_stream.total_in = 0;
2022 	zip->zipx_lzma_stream.next_out = zip->uncompressed_buffer;
2023 	zip->zipx_lzma_stream.avail_out =
2024 		/* These lzma_alone streams lack end of stream marker, so let's
2025 		 * make sure the unpacker won't try to unpack more than it's
2026 		 * supposed to. */
2027 		zipmin((int64_t) zip->uncompressed_buffer_size,
2028 		    zip->entry->uncompressed_size -
2029 		    zip->entry_uncompressed_bytes_read);
2030 	zip->zipx_lzma_stream.total_out = 0;
2031 
2032 	/* Perform the decompression. */
2033 	lz_ret = lzma_code(&zip->zipx_lzma_stream, LZMA_RUN);
2034 	switch(lz_ret) {
2035 		case LZMA_DATA_ERROR:
2036 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2037 			    "lzma data error (error %d)", (int) lz_ret);
2038 			return (ARCHIVE_FATAL);
2039 
2040 		/* This case is optional in lzma alone format. It can happen,
2041 		 * but most of the files don't have it. (GitHub #1257) */
2042 		case LZMA_STREAM_END:
2043 			if((int64_t) zip->zipx_lzma_stream.total_in !=
2044 			    zip->entry_bytes_remaining)
2045 			{
2046 				archive_set_error(&a->archive,
2047 				    ARCHIVE_ERRNO_MISC,
2048 				    "lzma alone premature end of stream");
2049 				return (ARCHIVE_FATAL);
2050 			}
2051 
2052 			zip->end_of_entry = 1;
2053 			break;
2054 
2055 		case LZMA_OK:
2056 			break;
2057 
2058 		default:
2059 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2060 			    "lzma unknown error %d", (int) lz_ret);
2061 			return (ARCHIVE_FATAL);
2062 	}
2063 
2064 	to_consume = zip->zipx_lzma_stream.total_in;
2065 
2066 	/* Update pointers. */
2067 	__archive_read_consume(a, to_consume);
2068 	zip->entry_bytes_remaining -= to_consume;
2069 	zip->entry_compressed_bytes_read += to_consume;
2070 	zip->entry_uncompressed_bytes_read += zip->zipx_lzma_stream.total_out;
2071 
2072 	if(zip->entry_bytes_remaining == 0) {
2073 		zip->end_of_entry = 1;
2074 	}
2075 
2076 	/* Free lzma decoder handle because we'll no longer need it. */
2077 	/* This cannot be folded into LZMA_STREAM_END handling above
2078 	 * because the stream end marker is not required in this format. */
2079 	if(zip->end_of_entry) {
2080 		lzma_end(&zip->zipx_lzma_stream);
2081 		zip->zipx_lzma_valid = 0;
2082 	}
2083 
2084 	/* Return values. */
2085 	*size = zip->zipx_lzma_stream.total_out;
2086 	*buff = zip->uncompressed_buffer;
2087 
2088 	/* If we're here, then we're good! */
2089 	return (ARCHIVE_OK);
2090 }
2091 #endif /* HAVE_LZMA_H && HAVE_LIBLZMA */
2092 
2093 static int
zipx_ppmd8_init(struct archive_read * a,struct zip * zip)2094 zipx_ppmd8_init(struct archive_read *a, struct zip *zip)
2095 {
2096 	const void* p;
2097 	uint32_t val;
2098 	uint32_t order;
2099 	uint32_t mem;
2100 	uint32_t restore_method;
2101 
2102 	/* Remove previous decompression context if it exists. */
2103 	if(zip->ppmd8_valid) {
2104 		__archive_ppmd8_functions.Ppmd8_Free(&zip->ppmd8);
2105 		zip->ppmd8_valid = 0;
2106 	}
2107 
2108 	/* Create a new decompression context. */
2109 	__archive_ppmd8_functions.Ppmd8_Construct(&zip->ppmd8);
2110 	zip->ppmd8_stream_failed = 0;
2111 
2112 	/* Setup function pointers required by Ppmd8 decompressor. The
2113 	 * 'ppmd_read' function will feed new bytes to the decompressor,
2114 	 * and will increment the 'zip->zipx_ppmd_read_compressed' counter. */
2115 	zip->ppmd8.Stream.In = &zip->zipx_ppmd_stream;
2116 	zip->zipx_ppmd_stream.a = a;
2117 	zip->zipx_ppmd_stream.Read = &ppmd_read;
2118 
2119 	/* Reset number of read bytes to 0. */
2120 	zip->zipx_ppmd_read_compressed = 0;
2121 
2122 	/* Read Ppmd8 header (2 bytes). */
2123 	p = __archive_read_ahead(a, 2, NULL);
2124 	if(!p) {
2125 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2126 		    "Truncated file data in PPMd8 stream");
2127 		return (ARCHIVE_FATAL);
2128 	}
2129 	__archive_read_consume(a, 2);
2130 
2131 	/* Decode the stream's compression parameters. */
2132 	val = archive_le16dec(p);
2133 	order = (val & 15) + 1;
2134 	mem = ((val >> 4) & 0xff) + 1;
2135 	restore_method = (val >> 12);
2136 
2137 	if(order < 2 || restore_method > 2) {
2138 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2139 		    "Invalid parameter set in PPMd8 stream (order=%" PRId32 ", "
2140 		    "restore=%" PRId32 ")", order, restore_method);
2141 		return (ARCHIVE_FAILED);
2142 	}
2143 
2144 	/* Allocate the memory needed to properly decompress the file. */
2145 	if(!__archive_ppmd8_functions.Ppmd8_Alloc(&zip->ppmd8, mem << 20)) {
2146 		archive_set_error(&a->archive, ENOMEM,
2147 		    "Unable to allocate memory for PPMd8 stream: %" PRId32 " bytes",
2148 		    mem << 20);
2149 		return (ARCHIVE_FATAL);
2150 	}
2151 
2152 	/* Signal the cleanup function to release Ppmd8 context in the
2153 	 * cleanup phase. */
2154 	zip->ppmd8_valid = 1;
2155 
2156 	/* Perform further Ppmd8 initialization. */
2157 	if(!__archive_ppmd8_functions.Ppmd8_RangeDec_Init(&zip->ppmd8)) {
2158 		archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
2159 		    "PPMd8 stream range decoder initialization error");
2160 		return (ARCHIVE_FATAL);
2161 	}
2162 
2163 	__archive_ppmd8_functions.Ppmd8_Init(&zip->ppmd8, order,
2164 	    restore_method);
2165 
2166 	/* Allocate the buffer that will hold uncompressed data. */
2167 	free(zip->uncompressed_buffer);
2168 
2169 	zip->uncompressed_buffer_size = 256 * 1024;
2170 	zip->uncompressed_buffer =
2171 	    (uint8_t*) malloc(zip->uncompressed_buffer_size);
2172 
2173 	if(zip->uncompressed_buffer == NULL) {
2174 		archive_set_error(&a->archive, ENOMEM,
2175 		    "No memory for PPMd8 decompression");
2176 		return ARCHIVE_FATAL;
2177 	}
2178 
2179 	/* Ppmd8 initialization is done. */
2180 	zip->decompress_init = 1;
2181 
2182 	/* We've already read 2 bytes in the output stream. Additionally,
2183 	 * Ppmd8 initialization code could read some data as well. So we
2184 	 * are advancing the stream by 2 bytes plus whatever number of
2185 	 * bytes Ppmd8 init function used. */
2186 	zip->entry_compressed_bytes_read += 2 + zip->zipx_ppmd_read_compressed;
2187 
2188 	return ARCHIVE_OK;
2189 }
2190 
2191 static int
zip_read_data_zipx_ppmd(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)2192 zip_read_data_zipx_ppmd(struct archive_read *a, const void **buff,
2193     size_t *size, int64_t *offset)
2194 {
2195 	struct zip* zip = (struct zip *)(a->format->data);
2196 	int ret;
2197 	size_t consumed_bytes = 0;
2198 	ssize_t bytes_avail = 0;
2199 
2200 	(void) offset; /* UNUSED */
2201 
2202 	/* If we're here for the first time, initialize Ppmd8 decompression
2203 	 * context first. */
2204 	if(!zip->decompress_init) {
2205 		ret = zipx_ppmd8_init(a, zip);
2206 		if(ret != ARCHIVE_OK)
2207 			return ret;
2208 	}
2209 
2210 	/* Fetch for more data. We're reading 1 byte here, but libarchive
2211 	 * should prefetch more bytes. */
2212 	(void) __archive_read_ahead(a, 1, &bytes_avail);
2213 	if(bytes_avail < 0) {
2214 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2215 		    "Truncated PPMd8 file body");
2216 		return (ARCHIVE_FATAL);
2217 	}
2218 
2219 	/* This counter will be updated inside ppmd_read(), which at one
2220 	 * point will be called by Ppmd8_DecodeSymbol. */
2221 	zip->zipx_ppmd_read_compressed = 0;
2222 
2223 	/* Decompression loop. */
2224 	do {
2225 		int sym = __archive_ppmd8_functions.Ppmd8_DecodeSymbol(
2226 		    &zip->ppmd8);
2227 		if(sym < 0) {
2228 			zip->end_of_entry = 1;
2229 			break;
2230 		}
2231 
2232 		/* This field is set by ppmd_read() when there was no more data
2233 		 * to be read. */
2234 		if(zip->ppmd8_stream_failed) {
2235 			archive_set_error(&a->archive,
2236 			    ARCHIVE_ERRNO_FILE_FORMAT,
2237 			    "Truncated PPMd8 file body");
2238 			return (ARCHIVE_FATAL);
2239 		}
2240 
2241 		zip->uncompressed_buffer[consumed_bytes] = (uint8_t) sym;
2242 		++consumed_bytes;
2243 	} while(consumed_bytes < zip->uncompressed_buffer_size);
2244 
2245 	/* Update pointers so we can continue decompression in another call. */
2246 	zip->entry_bytes_remaining -= zip->zipx_ppmd_read_compressed;
2247 	zip->entry_compressed_bytes_read += zip->zipx_ppmd_read_compressed;
2248 	zip->entry_uncompressed_bytes_read += consumed_bytes;
2249 
2250 	/* If we're at the end of stream, deinitialize Ppmd8 context. */
2251 	if(zip->end_of_entry) {
2252 		__archive_ppmd8_functions.Ppmd8_Free(&zip->ppmd8);
2253 		zip->ppmd8_valid = 0;
2254 	}
2255 
2256 	/* Update pointers for libarchive. */
2257 	*buff = zip->uncompressed_buffer;
2258 	*size = consumed_bytes;
2259 
2260 	return ARCHIVE_OK;
2261 }
2262 
2263 #ifdef HAVE_BZLIB_H
2264 static int
zipx_bzip2_init(struct archive_read * a,struct zip * zip)2265 zipx_bzip2_init(struct archive_read *a, struct zip *zip)
2266 {
2267 	int r;
2268 
2269 	/* Deallocate already existing BZ2 decompression context if it
2270 	 * exists. */
2271 	if(zip->bzstream_valid) {
2272 		BZ2_bzDecompressEnd(&zip->bzstream);
2273 		zip->bzstream_valid = 0;
2274 	}
2275 
2276 	/* Allocate a new BZ2 decompression context. */
2277 	memset(&zip->bzstream, 0, sizeof(bz_stream));
2278 	r = BZ2_bzDecompressInit(&zip->bzstream, 0, 1);
2279 	if(r != BZ_OK) {
2280 		archive_set_error(&(a->archive), ARCHIVE_ERRNO_MISC,
2281 		    "bzip2 initialization failed(%d)",
2282 		    r);
2283 
2284 		return ARCHIVE_FAILED;
2285 	}
2286 
2287 	/* Mark the bzstream field to be released in cleanup phase. */
2288 	zip->bzstream_valid = 1;
2289 
2290 	/* (Re)allocate the buffer that will contain decompressed bytes. */
2291 	free(zip->uncompressed_buffer);
2292 
2293 	zip->uncompressed_buffer_size = 256 * 1024;
2294 	zip->uncompressed_buffer =
2295 	    (uint8_t*) malloc(zip->uncompressed_buffer_size);
2296 	if (zip->uncompressed_buffer == NULL) {
2297 		archive_set_error(&a->archive, ENOMEM,
2298 		    "No memory for bzip2 decompression");
2299 		    return ARCHIVE_FATAL;
2300 	}
2301 
2302 	/* Initialization done. */
2303 	zip->decompress_init = 1;
2304 	return ARCHIVE_OK;
2305 }
2306 
2307 static int
zip_read_data_zipx_bzip2(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)2308 zip_read_data_zipx_bzip2(struct archive_read *a, const void **buff,
2309     size_t *size, int64_t *offset)
2310 {
2311 	struct zip *zip = (struct zip *)(a->format->data);
2312 	ssize_t bytes_avail = 0, in_bytes, to_consume;
2313 	const void *compressed_buff;
2314 	int r;
2315 	uint64_t total_out;
2316 
2317 	(void) offset; /* UNUSED */
2318 
2319 	/* Initialize decompression context if we're here for the first time. */
2320 	if(!zip->decompress_init) {
2321 		r = zipx_bzip2_init(a, zip);
2322 		if(r != ARCHIVE_OK)
2323 			return r;
2324 	}
2325 
2326 	/* Fetch more compressed bytes. */
2327 	compressed_buff = __archive_read_ahead(a, 1, &bytes_avail);
2328 	if(bytes_avail < 0) {
2329 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2330 		    "Truncated bzip2 file body");
2331 		return (ARCHIVE_FATAL);
2332 	}
2333 
2334 	in_bytes = zipmin(zip->entry_bytes_remaining, bytes_avail);
2335 	if(in_bytes < 1) {
2336 		/* libbz2 doesn't complain when caller feeds avail_in == 0.
2337 		 * It will actually return success in this case, which is
2338 		 * undesirable. This is why we need to make this check
2339 		 * manually. */
2340 
2341 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2342 		    "Truncated bzip2 file body");
2343 		return (ARCHIVE_FATAL);
2344 	}
2345 
2346 	/* Setup buffer boundaries. */
2347 	zip->bzstream.next_in = (char*)(uintptr_t) compressed_buff;
2348 	zip->bzstream.avail_in = (uint32_t)in_bytes;
2349 	zip->bzstream.total_in_hi32 = 0;
2350 	zip->bzstream.total_in_lo32 = 0;
2351 	zip->bzstream.next_out = (char*) zip->uncompressed_buffer;
2352 	zip->bzstream.avail_out = (uint32_t)zip->uncompressed_buffer_size;
2353 	zip->bzstream.total_out_hi32 = 0;
2354 	zip->bzstream.total_out_lo32 = 0;
2355 
2356 	/* Perform the decompression. */
2357 	r = BZ2_bzDecompress(&zip->bzstream);
2358 	switch(r) {
2359 		case BZ_STREAM_END:
2360 			/* If we're at the end of the stream, deinitialize the
2361 			 * decompression context now. */
2362 			switch(BZ2_bzDecompressEnd(&zip->bzstream)) {
2363 				case BZ_OK:
2364 					break;
2365 				default:
2366 					archive_set_error(&a->archive,
2367 					    ARCHIVE_ERRNO_MISC,
2368 					    "Failed to clean up bzip2 "
2369 					    "decompressor");
2370 					return ARCHIVE_FATAL;
2371 			}
2372 
2373 			zip->end_of_entry = 1;
2374 			break;
2375 		case BZ_OK:
2376 			/* The decompressor has successfully decoded this
2377 			 * chunk of data, but more data is still in queue. */
2378 			break;
2379 		default:
2380 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2381 			    "bzip2 decompression failed");
2382 			return ARCHIVE_FATAL;
2383 	}
2384 
2385 	/* Update the pointers so decompressor can continue decoding. */
2386 	to_consume = zip->bzstream.total_in_lo32;
2387 	__archive_read_consume(a, to_consume);
2388 
2389 	total_out = ((uint64_t) zip->bzstream.total_out_hi32 << 32) |
2390 	    zip->bzstream.total_out_lo32;
2391 
2392 	zip->entry_bytes_remaining -= to_consume;
2393 	zip->entry_compressed_bytes_read += to_consume;
2394 	zip->entry_uncompressed_bytes_read += total_out;
2395 
2396 	/* Give libarchive its due. */
2397 	*size = total_out;
2398 	*buff = zip->uncompressed_buffer;
2399 
2400 	return ARCHIVE_OK;
2401 }
2402 
2403 #endif
2404 
2405 #if HAVE_ZSTD_H && HAVE_LIBZSTD
2406 static int
zipx_zstd_init(struct archive_read * a,struct zip * zip)2407 zipx_zstd_init(struct archive_read *a, struct zip *zip)
2408 {
2409 	size_t r;
2410 
2411 	/* Deallocate already existing Zstd decompression context if it
2412 	 * exists. */
2413 	if(zip->zstdstream_valid) {
2414 		ZSTD_freeDStream(zip->zstdstream);
2415 		zip->zstdstream_valid = 0;
2416 	}
2417 
2418 	/* Allocate a new Zstd decompression context. */
2419 	zip->zstdstream = ZSTD_createDStream();
2420 
2421 	r = ZSTD_initDStream(zip->zstdstream);
2422 	if (ZSTD_isError(r)) {
2423 		 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2424 			"Error initializing zstd decompressor: %s",
2425 			ZSTD_getErrorName(r));
2426 
2427 		return ARCHIVE_FAILED;
2428 	}
2429 
2430 	/* Mark the zstdstream field to be released in cleanup phase. */
2431 	zip->zstdstream_valid = 1;
2432 
2433 	/* (Re)allocate the buffer that will contain decompressed bytes. */
2434 	free(zip->uncompressed_buffer);
2435 
2436 	zip->uncompressed_buffer_size = ZSTD_DStreamOutSize();
2437 	zip->uncompressed_buffer =
2438 	    (uint8_t*) malloc(zip->uncompressed_buffer_size);
2439 	if (zip->uncompressed_buffer == NULL) {
2440 		archive_set_error(&a->archive, ENOMEM,
2441 			"No memory for Zstd decompression");
2442 
2443 		return ARCHIVE_FATAL;
2444 	}
2445 
2446 	/* Initialization done. */
2447 	zip->decompress_init = 1;
2448 	return ARCHIVE_OK;
2449 }
2450 
2451 static int
zip_read_data_zipx_zstd(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)2452 zip_read_data_zipx_zstd(struct archive_read *a, const void **buff,
2453     size_t *size, int64_t *offset)
2454 {
2455 	struct zip *zip = (struct zip *)(a->format->data);
2456 	ssize_t bytes_avail = 0, in_bytes, to_consume;
2457 	const void *compressed_buff;
2458 	int r;
2459 	size_t ret;
2460 	uint64_t total_out;
2461 	ZSTD_outBuffer out;
2462 	ZSTD_inBuffer in;
2463 
2464 	(void) offset; /* UNUSED */
2465 
2466 	/* Initialize decompression context if we're here for the first time. */
2467 	if(!zip->decompress_init) {
2468 		r = zipx_zstd_init(a, zip);
2469 		if(r != ARCHIVE_OK)
2470 			return r;
2471 	}
2472 
2473 	/* Fetch more compressed bytes */
2474 	compressed_buff = __archive_read_ahead(a, 1, &bytes_avail);
2475 	if(bytes_avail < 0) {
2476 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2477 		    "Truncated zstd file body");
2478 		return (ARCHIVE_FATAL);
2479 	}
2480 
2481 	in_bytes = zipmin(zip->entry_bytes_remaining, bytes_avail);
2482 	if(in_bytes < 1) {
2483 		/* zstd doesn't complain when caller feeds avail_in == 0.
2484 		 * It will actually return success in this case, which is
2485 		 * undesirable. This is why we need to make this check
2486 		 * manually. */
2487 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2488 		    "Truncated zstd file body");
2489 		return (ARCHIVE_FATAL);
2490 	}
2491 
2492 	/* Setup buffer boundaries */
2493 	in.src = compressed_buff;
2494 	in.size = in_bytes;
2495 	in.pos = 0;
2496 	out = (ZSTD_outBuffer) { zip->uncompressed_buffer, zip->uncompressed_buffer_size, 0 };
2497 
2498 	/* Perform the decompression. */
2499 	ret = ZSTD_decompressStream(zip->zstdstream, &out, &in);
2500 	if (ZSTD_isError(ret)) {
2501 		archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2502 			"Error during zstd decompression: %s",
2503 			ZSTD_getErrorName(ret));
2504 		return (ARCHIVE_FATAL);
2505 	}
2506 
2507 	/* Check end of the stream. */
2508 	if (ret == 0) {
2509 		if ((in.pos == in.size) && (out.pos < out.size)) {
2510 			zip->end_of_entry = 1;
2511 			ZSTD_freeDStream(zip->zstdstream);
2512 			zip->zstdstream_valid = 0;
2513 		}
2514 	}
2515 
2516 	/* Update the pointers so decompressor can continue decoding. */
2517 	to_consume = in.pos;
2518 	__archive_read_consume(a, to_consume);
2519 
2520 	total_out = out.pos;
2521 
2522 	zip->entry_bytes_remaining -= to_consume;
2523 	zip->entry_compressed_bytes_read += to_consume;
2524 	zip->entry_uncompressed_bytes_read += total_out;
2525 
2526 	/* Give libarchive its due. */
2527 	*size = total_out;
2528 	*buff = zip->uncompressed_buffer;
2529 
2530 	return ARCHIVE_OK;
2531 }
2532 #endif
2533 
2534 #ifdef HAVE_ZLIB_H
2535 static int
zip_deflate_init(struct archive_read * a,struct zip * zip)2536 zip_deflate_init(struct archive_read *a, struct zip *zip)
2537 {
2538 	int r;
2539 
2540 	/* If we haven't yet read any data, initialize the decompressor. */
2541 	if (!zip->decompress_init) {
2542 		if (zip->stream_valid)
2543 			r = inflateReset(&zip->stream);
2544 		else
2545 			r = inflateInit2(&zip->stream,
2546 			    -15 /* Don't check for zlib header */);
2547 		if (r != Z_OK) {
2548 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2549 			    "Can't initialize ZIP decompression.");
2550 			return (ARCHIVE_FATAL);
2551 		}
2552 		/* Stream structure has been set up. */
2553 		zip->stream_valid = 1;
2554 		/* We've initialized decompression for this stream. */
2555 		zip->decompress_init = 1;
2556 	}
2557 	return (ARCHIVE_OK);
2558 }
2559 
2560 static int
zip_read_data_deflate(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)2561 zip_read_data_deflate(struct archive_read *a, const void **buff,
2562     size_t *size, int64_t *offset)
2563 {
2564 	struct zip *zip;
2565 	ssize_t bytes_avail, to_consume = 0;
2566 	const void *compressed_buff, *sp;
2567 	int r;
2568 
2569 	(void)offset; /* UNUSED */
2570 
2571 	zip = (struct zip *)(a->format->data);
2572 
2573 	/* If the buffer hasn't been allocated, allocate it now. */
2574 	if (zip->uncompressed_buffer == NULL) {
2575 		zip->uncompressed_buffer_size = 256 * 1024;
2576 		zip->uncompressed_buffer
2577 		    = (unsigned char *)malloc(zip->uncompressed_buffer_size);
2578 		if (zip->uncompressed_buffer == NULL) {
2579 			archive_set_error(&a->archive, ENOMEM,
2580 			    "No memory for ZIP decompression");
2581 			return (ARCHIVE_FATAL);
2582 		}
2583 	}
2584 
2585 	r = zip_deflate_init(a, zip);
2586 	if (r != ARCHIVE_OK)
2587 		return (r);
2588 
2589 	/*
2590 	 * Note: '1' here is a performance optimization.
2591 	 * Recall that the decompression layer returns a count of
2592 	 * available bytes; asking for more than that forces the
2593 	 * decompressor to combine reads by copying data.
2594 	 */
2595 	compressed_buff = sp = __archive_read_ahead(a, 1, &bytes_avail);
2596 	if (0 == (zip->entry->zip_flags & ZIP_LENGTH_AT_END)
2597 	    && bytes_avail > zip->entry_bytes_remaining) {
2598 		bytes_avail = (ssize_t)zip->entry_bytes_remaining;
2599 	}
2600 	if (bytes_avail < 0) {
2601 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2602 		    "Truncated ZIP file body");
2603 		return (ARCHIVE_FATAL);
2604 	}
2605 
2606 	if (zip->tctx_valid || zip->cctx_valid) {
2607 		if (zip->decrypted_bytes_remaining < (size_t)bytes_avail) {
2608 			size_t buff_remaining =
2609 			    (zip->decrypted_buffer +
2610 			    zip->decrypted_buffer_size)
2611 			    - (zip->decrypted_ptr +
2612 			    zip->decrypted_bytes_remaining);
2613 
2614 			if (buff_remaining > (size_t)bytes_avail)
2615 				buff_remaining = (size_t)bytes_avail;
2616 
2617 			if (0 == (zip->entry->zip_flags & ZIP_LENGTH_AT_END) &&
2618 			      zip->entry_bytes_remaining > 0) {
2619 				if ((int64_t)(zip->decrypted_bytes_remaining
2620 				    + buff_remaining)
2621 				      > zip->entry_bytes_remaining) {
2622 					if (zip->entry_bytes_remaining <
2623 					    (int64_t)zip->decrypted_bytes_remaining)
2624 						buff_remaining = 0;
2625 					else
2626 						buff_remaining =
2627 						    (size_t)zip->entry_bytes_remaining
2628 						    - zip->decrypted_bytes_remaining;
2629 				}
2630 			}
2631 			if (buff_remaining > 0) {
2632 				if (zip->tctx_valid) {
2633 					trad_enc_decrypt_update(&zip->tctx,
2634 					    compressed_buff, buff_remaining,
2635 					    zip->decrypted_ptr
2636 					      + zip->decrypted_bytes_remaining,
2637 					    buff_remaining);
2638 				} else {
2639 					size_t dsize = buff_remaining;
2640 					archive_decrypto_aes_ctr_update(
2641 					    &zip->cctx,
2642 					    compressed_buff, buff_remaining,
2643 					    zip->decrypted_ptr
2644 					      + zip->decrypted_bytes_remaining,
2645 					    &dsize);
2646 				}
2647 				zip->decrypted_bytes_remaining +=
2648 				    buff_remaining;
2649 			}
2650 		}
2651 		bytes_avail = zip->decrypted_bytes_remaining;
2652 		compressed_buff = (const char *)zip->decrypted_ptr;
2653 	}
2654 
2655 	/*
2656 	 * A bug in zlib.h: stream.next_in should be marked 'const'
2657 	 * but isn't (the library never alters data through the
2658 	 * next_in pointer, only reads it).  The result: this ugly
2659 	 * cast to remove 'const'.
2660 	 */
2661 	zip->stream.next_in = (Bytef *)(uintptr_t)(const void *)compressed_buff;
2662 	zip->stream.avail_in = (uInt)bytes_avail;
2663 	zip->stream.total_in = 0;
2664 	zip->stream.next_out = zip->uncompressed_buffer;
2665 	zip->stream.avail_out = (uInt)zip->uncompressed_buffer_size;
2666 	zip->stream.total_out = 0;
2667 
2668 	r = inflate(&zip->stream, 0);
2669 	switch (r) {
2670 	case Z_OK:
2671 		break;
2672 	case Z_STREAM_END:
2673 		zip->end_of_entry = 1;
2674 		break;
2675 	case Z_MEM_ERROR:
2676 		archive_set_error(&a->archive, ENOMEM,
2677 		    "Out of memory for ZIP decompression");
2678 		return (ARCHIVE_FATAL);
2679 	default:
2680 		archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2681 		    "ZIP decompression failed (%d)", r);
2682 		return (ARCHIVE_FATAL);
2683 	}
2684 
2685 	/* Consume as much as the compressor actually used. */
2686 	to_consume = zip->stream.total_in;
2687 	__archive_read_consume(a, to_consume);
2688 	zip->entry_bytes_remaining -= to_consume;
2689 	zip->entry_compressed_bytes_read += to_consume;
2690 	zip->entry_uncompressed_bytes_read += zip->stream.total_out;
2691 
2692 	if (zip->tctx_valid || zip->cctx_valid) {
2693 		zip->decrypted_bytes_remaining -= to_consume;
2694 		if (zip->decrypted_bytes_remaining == 0)
2695 			zip->decrypted_ptr = zip->decrypted_buffer;
2696 		else
2697 			zip->decrypted_ptr += to_consume;
2698 	}
2699 	if (zip->hctx_valid)
2700 		archive_hmac_sha1_update(&zip->hctx, sp, to_consume);
2701 
2702 	if (zip->end_of_entry) {
2703 		if (zip->hctx_valid) {
2704 			r = check_authentication_code(a, NULL);
2705 			if (r != ARCHIVE_OK) {
2706 				return (r);
2707 			}
2708 		}
2709 	}
2710 
2711 	*size = zip->stream.total_out;
2712 	*buff = zip->uncompressed_buffer;
2713 
2714 	return (ARCHIVE_OK);
2715 }
2716 #endif
2717 
2718 static int
read_decryption_header(struct archive_read * a)2719 read_decryption_header(struct archive_read *a)
2720 {
2721 	struct zip *zip = (struct zip *)(a->format->data);
2722 	const char *p;
2723 	unsigned int remaining_size;
2724 	unsigned int ts;
2725 
2726 	/*
2727 	 * Read an initialization vector data field.
2728 	 */
2729 	p = __archive_read_ahead(a, 2, NULL);
2730 	if (p == NULL)
2731 		goto truncated;
2732 	ts = zip->iv_size;
2733 	zip->iv_size = archive_le16dec(p);
2734 	__archive_read_consume(a, 2);
2735 	if (ts < zip->iv_size) {
2736 		free(zip->iv);
2737 		zip->iv = NULL;
2738 	}
2739 	p = __archive_read_ahead(a, zip->iv_size, NULL);
2740 	if (p == NULL)
2741 		goto truncated;
2742 	if (zip->iv == NULL) {
2743 		zip->iv = malloc(zip->iv_size);
2744 		if (zip->iv == NULL)
2745 			goto nomem;
2746 	}
2747 	memcpy(zip->iv, p, zip->iv_size);
2748 	__archive_read_consume(a, zip->iv_size);
2749 
2750 	/*
2751 	 * Read a size of remaining decryption header field.
2752 	 */
2753 	p = __archive_read_ahead(a, 14, NULL);
2754 	if (p == NULL)
2755 		goto truncated;
2756 	remaining_size = archive_le32dec(p);
2757 	if (remaining_size < 16 || remaining_size > (1 << 18))
2758 		goto corrupted;
2759 
2760 	/* Check if format version is supported. */
2761 	if (archive_le16dec(p+4) != 3) {
2762 		archive_set_error(&a->archive,
2763 		    ARCHIVE_ERRNO_FILE_FORMAT,
2764 		    "Unsupported encryption format version: %u",
2765 		    archive_le16dec(p+4));
2766 		return (ARCHIVE_FAILED);
2767 	}
2768 
2769 	/*
2770 	 * Read an encryption algorithm field.
2771 	 */
2772 	zip->alg_id = archive_le16dec(p+6);
2773 	switch (zip->alg_id) {
2774 	case 0x6601:/* DES */
2775 	case 0x6602:/* RC2 */
2776 	case 0x6603:/* 3DES 168 */
2777 	case 0x6609:/* 3DES 112 */
2778 	case 0x660E:/* AES 128 */
2779 	case 0x660F:/* AES 192 */
2780 	case 0x6610:/* AES 256 */
2781 	case 0x6702:/* RC2 (version >= 5.2) */
2782 	case 0x6720:/* Blowfish */
2783 	case 0x6721:/* Twofish */
2784 	case 0x6801:/* RC4 */
2785 		/* Supported encryption algorithm. */
2786 		break;
2787 	default:
2788 		archive_set_error(&a->archive,
2789 		    ARCHIVE_ERRNO_FILE_FORMAT,
2790 		    "Unknown encryption algorithm: %u", zip->alg_id);
2791 		return (ARCHIVE_FAILED);
2792 	}
2793 
2794 	/*
2795 	 * Read a bit length field.
2796 	 */
2797 	zip->bit_len = archive_le16dec(p+8);
2798 
2799 	/*
2800 	 * Read a flags field.
2801 	 */
2802 	zip->flags = archive_le16dec(p+10);
2803 	switch (zip->flags & 0xf000) {
2804 	case 0x0001: /* Password is required to decrypt. */
2805 	case 0x0002: /* Certificates only. */
2806 	case 0x0003: /* Password or certificate required to decrypt. */
2807 		break;
2808 	default:
2809 		archive_set_error(&a->archive,
2810 		    ARCHIVE_ERRNO_FILE_FORMAT,
2811 		    "Unknown encryption flag: %u", zip->flags);
2812 		return (ARCHIVE_FAILED);
2813 	}
2814 	if ((zip->flags & 0xf000) == 0 ||
2815 	    (zip->flags & 0xf000) == 0x4000) {
2816 		archive_set_error(&a->archive,
2817 		    ARCHIVE_ERRNO_FILE_FORMAT,
2818 		    "Unknown encryption flag: %u", zip->flags);
2819 		return (ARCHIVE_FAILED);
2820 	}
2821 
2822 	/*
2823 	 * Read an encrypted random data field.
2824 	 */
2825 	ts = zip->erd_size;
2826 	zip->erd_size = archive_le16dec(p+12);
2827 	__archive_read_consume(a, 14);
2828 	if ((zip->erd_size & 0xf) != 0 ||
2829 	    (zip->erd_size + 16) > remaining_size ||
2830 	    (zip->erd_size + 16) < zip->erd_size)
2831 		goto corrupted;
2832 
2833 	if (ts < zip->erd_size) {
2834 		free(zip->erd);
2835 		zip->erd = NULL;
2836 	}
2837 	p = __archive_read_ahead(a, zip->erd_size, NULL);
2838 	if (p == NULL)
2839 		goto truncated;
2840 	if (zip->erd == NULL) {
2841 		zip->erd = malloc(zip->erd_size);
2842 		if (zip->erd == NULL)
2843 			goto nomem;
2844 	}
2845 	memcpy(zip->erd, p, zip->erd_size);
2846 	__archive_read_consume(a, zip->erd_size);
2847 
2848 	/*
2849 	 * Read a reserved data field.
2850 	 */
2851 	p = __archive_read_ahead(a, 4, NULL);
2852 	if (p == NULL)
2853 		goto truncated;
2854 	/* Reserved data size should be zero. */
2855 	if (archive_le32dec(p) != 0)
2856 		goto corrupted;
2857 	__archive_read_consume(a, 4);
2858 
2859 	/*
2860 	 * Read a password validation data field.
2861 	 */
2862 	p = __archive_read_ahead(a, 2, NULL);
2863 	if (p == NULL)
2864 		goto truncated;
2865 	ts = zip->v_size;
2866 	zip->v_size = archive_le16dec(p);
2867 	__archive_read_consume(a, 2);
2868 	if ((zip->v_size & 0x0f) != 0 ||
2869 	    (zip->erd_size + zip->v_size + 16) > remaining_size ||
2870 	    (zip->erd_size + zip->v_size + 16) < (zip->erd_size + zip->v_size))
2871 		goto corrupted;
2872 	if (ts < zip->v_size) {
2873 		free(zip->v_data);
2874 		zip->v_data = NULL;
2875 	}
2876 	p = __archive_read_ahead(a, zip->v_size, NULL);
2877 	if (p == NULL)
2878 		goto truncated;
2879 	if (zip->v_data == NULL) {
2880 		zip->v_data = malloc(zip->v_size);
2881 		if (zip->v_data == NULL)
2882 			goto nomem;
2883 	}
2884 	memcpy(zip->v_data, p, zip->v_size);
2885 	__archive_read_consume(a, zip->v_size);
2886 
2887 	p = __archive_read_ahead(a, 4, NULL);
2888 	if (p == NULL)
2889 		goto truncated;
2890 	zip->v_crc32 = archive_le32dec(p);
2891 	__archive_read_consume(a, 4);
2892 
2893 	/*return (ARCHIVE_OK);
2894 	 * This is not fully implemented yet.*/
2895 	archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2896 	    "Encrypted file is unsupported");
2897 	return (ARCHIVE_FAILED);
2898 truncated:
2899 	archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2900 	    "Truncated ZIP file data");
2901 	return (ARCHIVE_FATAL);
2902 corrupted:
2903 	archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2904 	    "Corrupted ZIP file data");
2905 	return (ARCHIVE_FATAL);
2906 nomem:
2907 	archive_set_error(&a->archive, ENOMEM,
2908 	    "No memory for ZIP decryption");
2909 	return (ARCHIVE_FATAL);
2910 }
2911 
2912 static int
zip_alloc_decryption_buffer(struct archive_read * a)2913 zip_alloc_decryption_buffer(struct archive_read *a)
2914 {
2915 	struct zip *zip = (struct zip *)(a->format->data);
2916 	size_t bs = 256 * 1024;
2917 
2918 	if (zip->decrypted_buffer == NULL) {
2919 		zip->decrypted_buffer_size = bs;
2920 		zip->decrypted_buffer = malloc(bs);
2921 		if (zip->decrypted_buffer == NULL) {
2922 			archive_set_error(&a->archive, ENOMEM,
2923 			    "No memory for ZIP decryption");
2924 			return (ARCHIVE_FATAL);
2925 		}
2926 	}
2927 	zip->decrypted_ptr = zip->decrypted_buffer;
2928 	return (ARCHIVE_OK);
2929 }
2930 
2931 static int
init_traditional_PKWARE_decryption(struct archive_read * a)2932 init_traditional_PKWARE_decryption(struct archive_read *a)
2933 {
2934 	struct zip *zip = (struct zip *)(a->format->data);
2935 	const void *p;
2936 	int retry;
2937 	int r;
2938 
2939 	if (zip->tctx_valid)
2940 		return (ARCHIVE_OK);
2941 
2942 	/*
2943 	   Read the 12 bytes encryption header stored at
2944 	   the start of the data area.
2945 	 */
2946 #define ENC_HEADER_SIZE	12
2947 	if (0 == (zip->entry->zip_flags & ZIP_LENGTH_AT_END)
2948 	    && zip->entry_bytes_remaining < ENC_HEADER_SIZE) {
2949 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2950 		    "Truncated Zip encrypted body: only %jd bytes available",
2951 		    (intmax_t)zip->entry_bytes_remaining);
2952 		return (ARCHIVE_FATAL);
2953 	}
2954 
2955 	p = __archive_read_ahead(a, ENC_HEADER_SIZE, NULL);
2956 	if (p == NULL) {
2957 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2958 		    "Truncated ZIP file data");
2959 		return (ARCHIVE_FATAL);
2960 	}
2961 
2962 	for (retry = 0;; retry++) {
2963 		const char *passphrase;
2964 		uint8_t crcchk;
2965 
2966 		passphrase = __archive_read_next_passphrase(a);
2967 		if (passphrase == NULL) {
2968 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2969 			    (retry > 0)?
2970 				"Incorrect passphrase":
2971 				"Passphrase required for this entry");
2972 			return (ARCHIVE_FAILED);
2973 		}
2974 
2975 		/*
2976 		 * Initialize ctx for Traditional PKWARE Decryption.
2977 		 */
2978 		r = trad_enc_init(&zip->tctx, passphrase, strlen(passphrase),
2979 			p, ENC_HEADER_SIZE, &crcchk);
2980 		if (r == 0 && crcchk == zip->entry->decdat)
2981 			break;/* The passphrase is OK. */
2982 		if (retry > 10000) {
2983 			/* Avoid infinity loop. */
2984 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2985 			    "Too many incorrect passphrases");
2986 			return (ARCHIVE_FAILED);
2987 		}
2988 	}
2989 
2990 	__archive_read_consume(a, ENC_HEADER_SIZE);
2991 	zip->tctx_valid = 1;
2992 	if (0 == (zip->entry->zip_flags & ZIP_LENGTH_AT_END)) {
2993 	    zip->entry_bytes_remaining -= ENC_HEADER_SIZE;
2994 	}
2995 	/*zip->entry_uncompressed_bytes_read += ENC_HEADER_SIZE;*/
2996 	zip->entry_compressed_bytes_read += ENC_HEADER_SIZE;
2997 	zip->decrypted_bytes_remaining = 0;
2998 
2999 	return (zip_alloc_decryption_buffer(a));
3000 #undef ENC_HEADER_SIZE
3001 }
3002 
3003 static int
init_WinZip_AES_decryption(struct archive_read * a)3004 init_WinZip_AES_decryption(struct archive_read *a)
3005 {
3006 	struct zip *zip = (struct zip *)(a->format->data);
3007 	const void *p;
3008 	const uint8_t *pv;
3009 	size_t key_len, salt_len;
3010 	uint8_t derived_key[MAX_DERIVED_KEY_BUF_SIZE];
3011 	int retry;
3012 	int r;
3013 
3014 	if (zip->cctx_valid || zip->hctx_valid)
3015 		return (ARCHIVE_OK);
3016 
3017 	switch (zip->entry->aes_extra.strength) {
3018 	case 1: salt_len = 8;  key_len = 16; break;
3019 	case 2: salt_len = 12; key_len = 24; break;
3020 	case 3: salt_len = 16; key_len = 32; break;
3021 	default: goto corrupted;
3022 	}
3023 	p = __archive_read_ahead(a, salt_len + 2, NULL);
3024 	if (p == NULL)
3025 		goto truncated;
3026 
3027 	for (retry = 0;; retry++) {
3028 		const char *passphrase;
3029 
3030 		passphrase = __archive_read_next_passphrase(a);
3031 		if (passphrase == NULL) {
3032 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3033 			    (retry > 0)?
3034 				"Incorrect passphrase":
3035 				"Passphrase required for this entry");
3036 			return (ARCHIVE_FAILED);
3037 		}
3038 		memset(derived_key, 0, sizeof(derived_key));
3039 		r = archive_pbkdf2_sha1(passphrase, strlen(passphrase),
3040 		    p, salt_len, 1000, derived_key, key_len * 2 + 2);
3041 		if (r != 0) {
3042 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3043 			    "Decryption is unsupported due to lack of "
3044 			    "crypto library");
3045 			return (ARCHIVE_FAILED);
3046 		}
3047 
3048 		/* Check password verification value. */
3049 		pv = ((const uint8_t *)p) + salt_len;
3050 		if (derived_key[key_len * 2] == pv[0] &&
3051 		    derived_key[key_len * 2 + 1] == pv[1])
3052 			break;/* The passphrase is OK. */
3053 		if (retry > 10000) {
3054 			/* Avoid infinity loop. */
3055 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3056 			    "Too many incorrect passphrases");
3057 			return (ARCHIVE_FAILED);
3058 		}
3059 	}
3060 
3061 	r = archive_decrypto_aes_ctr_init(&zip->cctx, derived_key, key_len);
3062 	if (r != 0) {
3063 		archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3064 		    "Decryption is unsupported due to lack of crypto library");
3065 		return (ARCHIVE_FAILED);
3066 	}
3067 	r = archive_hmac_sha1_init(&zip->hctx, derived_key + key_len, key_len);
3068 	if (r != 0) {
3069 		archive_decrypto_aes_ctr_release(&zip->cctx);
3070 		archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3071 		    "Failed to initialize HMAC-SHA1");
3072 		return (ARCHIVE_FAILED);
3073 	}
3074 	zip->cctx_valid = zip->hctx_valid = 1;
3075 	__archive_read_consume(a, salt_len + 2);
3076 	zip->entry_bytes_remaining -= salt_len + 2 + AUTH_CODE_SIZE;
3077 	if (0 == (zip->entry->zip_flags & ZIP_LENGTH_AT_END)
3078 	    && zip->entry_bytes_remaining < 0)
3079 		goto corrupted;
3080 	zip->entry_compressed_bytes_read += salt_len + 2 + AUTH_CODE_SIZE;
3081 	zip->decrypted_bytes_remaining = 0;
3082 
3083 	zip->entry->compression = zip->entry->aes_extra.compression;
3084 	return (zip_alloc_decryption_buffer(a));
3085 
3086 truncated:
3087 	archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
3088 	    "Truncated ZIP file data");
3089 	return (ARCHIVE_FATAL);
3090 corrupted:
3091 	archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
3092 	    "Corrupted ZIP file data");
3093 	return (ARCHIVE_FATAL);
3094 }
3095 
3096 static int
archive_read_format_zip_read_data(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)3097 archive_read_format_zip_read_data(struct archive_read *a,
3098     const void **buff, size_t *size, int64_t *offset)
3099 {
3100 	int r;
3101 	struct zip *zip = (struct zip *)(a->format->data);
3102 
3103 	if (zip->has_encrypted_entries ==
3104 			ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW) {
3105 		zip->has_encrypted_entries = 0;
3106 	}
3107 
3108 	*offset = zip->entry_uncompressed_bytes_read;
3109 	*size = 0;
3110 	*buff = NULL;
3111 
3112 	/* If we hit end-of-entry last time, return ARCHIVE_EOF. */
3113 	if (zip->end_of_entry)
3114 		return (ARCHIVE_EOF);
3115 
3116 	/* Return EOF immediately if this is a non-regular file. */
3117 	if (AE_IFREG != (zip->entry->mode & AE_IFMT))
3118 		return (ARCHIVE_EOF);
3119 
3120 	__archive_read_consume(a, zip->unconsumed);
3121 	zip->unconsumed = 0;
3122 
3123 	if (zip->init_decryption) {
3124 		zip->has_encrypted_entries = 1;
3125 		if (zip->entry->zip_flags & ZIP_STRONG_ENCRYPTED)
3126 			r = read_decryption_header(a);
3127 		else if (zip->entry->compression == WINZIP_AES_ENCRYPTION)
3128 			r = init_WinZip_AES_decryption(a);
3129 		else
3130 			r = init_traditional_PKWARE_decryption(a);
3131 		if (r != ARCHIVE_OK)
3132 			return (r);
3133 		zip->init_decryption = 0;
3134 	}
3135 
3136 	switch(zip->entry->compression) {
3137 	case 0:  /* No compression. */
3138 		r =  zip_read_data_none(a, buff, size, offset);
3139 		break;
3140 #ifdef HAVE_BZLIB_H
3141 	case 12: /* ZIPx bzip2 compression. */
3142 		r = zip_read_data_zipx_bzip2(a, buff, size, offset);
3143 		break;
3144 #endif
3145 #if HAVE_LZMA_H && HAVE_LIBLZMA
3146 	case 14: /* ZIPx LZMA compression. */
3147 		r = zip_read_data_zipx_lzma_alone(a, buff, size, offset);
3148 		break;
3149 	case 95: /* ZIPx XZ compression. */
3150 		r = zip_read_data_zipx_xz(a, buff, size, offset);
3151 		break;
3152 #endif
3153 #if HAVE_ZSTD_H && HAVE_LIBZSTD
3154 	case 93: /* ZIPx Zstd compression. */
3155 		r = zip_read_data_zipx_zstd(a, buff, size, offset);
3156 		break;
3157 #endif
3158 	/* PPMd support is built-in, so we don't need any #if guards. */
3159 	case 98: /* ZIPx PPMd compression. */
3160 		r = zip_read_data_zipx_ppmd(a, buff, size, offset);
3161 		break;
3162 
3163 #ifdef HAVE_ZLIB_H
3164 	case 8: /* Deflate compression. */
3165 		r =  zip_read_data_deflate(a, buff, size, offset);
3166 		break;
3167 #endif
3168 	default: /* Unsupported compression. */
3169 		/* Return a warning. */
3170 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
3171 		    "Unsupported ZIP compression method (%d: %s)",
3172 		    zip->entry->compression, compression_name(zip->entry->compression));
3173 		/* We can't decompress this entry, but we will
3174 		 * be able to skip() it and try the next entry. */
3175 		return (ARCHIVE_FAILED);
3176 		break;
3177 	}
3178 	if (r != ARCHIVE_OK)
3179 		return (r);
3180 	if (*size > 0) {
3181 		zip->computed_crc32 = zip->crc32func(zip->computed_crc32, *buff,
3182 						     (unsigned)*size);
3183 	}
3184 	/* If we hit the end, swallow any end-of-data marker and
3185 	 * verify the final check values. */
3186 	if (zip->end_of_entry) {
3187 		consume_end_of_file_marker(a, zip);
3188 
3189 		/* Check computed CRC against header */
3190 		if ((!zip->hctx_valid ||
3191 		      zip->entry->aes_extra.vendor != AES_VENDOR_AE_2) &&
3192 		   zip->entry->crc32 != zip->computed_crc32
3193 		    && !zip->ignore_crc32) {
3194 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3195 			    "ZIP bad CRC: 0x%lx should be 0x%lx",
3196 			    (unsigned long)zip->computed_crc32,
3197 			    (unsigned long)zip->entry->crc32);
3198 			return (ARCHIVE_FAILED);
3199 		}
3200 		/* Check file size against header. */
3201 		if (zip->entry->compressed_size !=
3202 		    zip->entry_compressed_bytes_read) {
3203 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3204 			    "ZIP compressed data is wrong size "
3205 			    "(read %jd, expected %jd)",
3206 			    (intmax_t)zip->entry_compressed_bytes_read,
3207 			    (intmax_t)zip->entry->compressed_size);
3208 			return (ARCHIVE_FAILED);
3209 		}
3210 		/* Size field only stores the lower 32 bits of the actual
3211 		 * size. */
3212 		if ((zip->entry->uncompressed_size & UINT32_MAX)
3213 		    != (zip->entry_uncompressed_bytes_read & UINT32_MAX)) {
3214 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3215 			    "ZIP uncompressed data is wrong size "
3216 			    "(read %jd, expected %jd)\n",
3217 			    (intmax_t)zip->entry_uncompressed_bytes_read,
3218 			    (intmax_t)zip->entry->uncompressed_size);
3219 			return (ARCHIVE_FAILED);
3220 		}
3221 	}
3222 
3223 	return (ARCHIVE_OK);
3224 }
3225 
3226 static int
archive_read_format_zip_cleanup(struct archive_read * a)3227 archive_read_format_zip_cleanup(struct archive_read *a)
3228 {
3229 	struct zip *zip;
3230 	struct zip_entry *zip_entry, *next_zip_entry;
3231 
3232 	zip = (struct zip *)(a->format->data);
3233 
3234 #ifdef HAVE_ZLIB_H
3235 	if (zip->stream_valid)
3236 		inflateEnd(&zip->stream);
3237 #endif
3238 
3239 #if HAVE_LZMA_H && HAVE_LIBLZMA
3240     if (zip->zipx_lzma_valid) {
3241 		lzma_end(&zip->zipx_lzma_stream);
3242 	}
3243 #endif
3244 
3245 #ifdef HAVE_BZLIB_H
3246 	if (zip->bzstream_valid) {
3247 		BZ2_bzDecompressEnd(&zip->bzstream);
3248 	}
3249 #endif
3250 
3251 #if HAVE_ZSTD_H && HAVE_LIBZSTD
3252 	if (zip->zstdstream_valid) {
3253 		ZSTD_freeDStream(zip->zstdstream);
3254 	}
3255 #endif
3256 
3257 	free(zip->uncompressed_buffer);
3258 
3259 	if (zip->ppmd8_valid)
3260 		__archive_ppmd8_functions.Ppmd8_Free(&zip->ppmd8);
3261 
3262 	if (zip->zip_entries) {
3263 		zip_entry = zip->zip_entries;
3264 		while (zip_entry != NULL) {
3265 			next_zip_entry = zip_entry->next;
3266 			archive_string_free(&zip_entry->rsrcname);
3267 			free(zip_entry);
3268 			zip_entry = next_zip_entry;
3269 		}
3270 	}
3271 	free(zip->decrypted_buffer);
3272 	if (zip->cctx_valid)
3273 		archive_decrypto_aes_ctr_release(&zip->cctx);
3274 	if (zip->hctx_valid)
3275 		archive_hmac_sha1_cleanup(&zip->hctx);
3276 	free(zip->iv);
3277 	free(zip->erd);
3278 	free(zip->v_data);
3279 	archive_string_free(&zip->format_name);
3280 	free(zip);
3281 	(a->format->data) = NULL;
3282 	return (ARCHIVE_OK);
3283 }
3284 
3285 static int
archive_read_format_zip_has_encrypted_entries(struct archive_read * _a)3286 archive_read_format_zip_has_encrypted_entries(struct archive_read *_a)
3287 {
3288 	if (_a && _a->format) {
3289 		struct zip * zip = (struct zip *)_a->format->data;
3290 		if (zip) {
3291 			return zip->has_encrypted_entries;
3292 		}
3293 	}
3294 	return ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW;
3295 }
3296 
3297 static int
archive_read_format_zip_options(struct archive_read * a,const char * key,const char * val)3298 archive_read_format_zip_options(struct archive_read *a,
3299     const char *key, const char *val)
3300 {
3301 	struct zip *zip;
3302 	int ret = ARCHIVE_FAILED;
3303 
3304 	zip = (struct zip *)(a->format->data);
3305 	if (strcmp(key, "compat-2x")  == 0) {
3306 		/* Handle filenames as libarchive 2.x */
3307 		zip->init_default_conversion = (val != NULL) ? 1 : 0;
3308 		return (ARCHIVE_OK);
3309 	} else if (strcmp(key, "hdrcharset")  == 0) {
3310 		if (val == NULL || val[0] == 0)
3311 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3312 			    "zip: hdrcharset option needs a character-set name"
3313 			);
3314 		else {
3315 			zip->sconv = archive_string_conversion_from_charset(
3316 			    &a->archive, val, 0);
3317 			if (zip->sconv != NULL) {
3318 				if (strcmp(val, "UTF-8") == 0)
3319 					zip->sconv_utf8 = zip->sconv;
3320 				ret = ARCHIVE_OK;
3321 			} else
3322 				ret = ARCHIVE_FATAL;
3323 		}
3324 		return (ret);
3325 	} else if (strcmp(key, "ignorecrc32") == 0) {
3326 		/* Mostly useful for testing. */
3327 		if (val == NULL || val[0] == 0) {
3328 			zip->crc32func = real_crc32;
3329 			zip->ignore_crc32 = 0;
3330 		} else {
3331 			zip->crc32func = fake_crc32;
3332 			zip->ignore_crc32 = 1;
3333 		}
3334 		return (ARCHIVE_OK);
3335 	} else if (strcmp(key, "mac-ext") == 0) {
3336 		zip->process_mac_extensions = (val != NULL && val[0] != 0);
3337 		return (ARCHIVE_OK);
3338 	}
3339 
3340 	/* Note: The "warn" return is just to inform the options
3341 	 * supervisor that we didn't handle it.  It will generate
3342 	 * a suitable error if no one used this option. */
3343 	return (ARCHIVE_WARN);
3344 }
3345 
3346 int
archive_read_support_format_zip(struct archive * a)3347 archive_read_support_format_zip(struct archive *a)
3348 {
3349 	int r;
3350 	r = archive_read_support_format_zip_streamable(a);
3351 	if (r != ARCHIVE_OK)
3352 		return r;
3353 	return (archive_read_support_format_zip_seekable(a));
3354 }
3355 
3356 /* ------------------------------------------------------------------------ */
3357 
3358 /*
3359  * Streaming-mode support
3360  */
3361 
3362 
3363 static int
archive_read_support_format_zip_capabilities_streamable(struct archive_read * a)3364 archive_read_support_format_zip_capabilities_streamable(struct archive_read * a)
3365 {
3366 	(void)a; /* UNUSED */
3367 	return (ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_DATA |
3368 		ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_METADATA);
3369 }
3370 
3371 static int
archive_read_format_zip_streamable_bid(struct archive_read * a,int best_bid)3372 archive_read_format_zip_streamable_bid(struct archive_read *a, int best_bid)
3373 {
3374 	const char *p;
3375 
3376 	(void)best_bid; /* UNUSED */
3377 
3378 	if ((p = __archive_read_ahead(a, 4, NULL)) == NULL)
3379 		return (-1);
3380 
3381 	/*
3382 	 * Bid of 29 here comes from:
3383 	 *  + 16 bits for "PK",
3384 	 *  + next 16-bit field has 6 options so contributes
3385 	 *    about 16 - log_2(6) ~= 16 - 2.6 ~= 13 bits
3386 	 *
3387 	 * So we've effectively verified ~29 total bits of check data.
3388 	 */
3389 	if (p[0] == 'P' && p[1] == 'K') {
3390 		if ((p[2] == '\001' && p[3] == '\002')
3391 		    || (p[2] == '\003' && p[3] == '\004')
3392 		    || (p[2] == '\005' && p[3] == '\006')
3393 		    || (p[2] == '\006' && p[3] == '\006')
3394 		    || (p[2] == '\007' && p[3] == '\010')
3395 		    || (p[2] == '0' && p[3] == '0'))
3396 			return (29);
3397 	}
3398 
3399 	/* TODO: It's worth looking ahead a little bit for a valid
3400 	 * PK signature.  In particular, that would make it possible
3401 	 * to read some UUEncoded SFX files or SFX files coming from
3402 	 * a network socket. */
3403 
3404 	return (0);
3405 }
3406 
3407 static int
archive_read_format_zip_streamable_read_header(struct archive_read * a,struct archive_entry * entry)3408 archive_read_format_zip_streamable_read_header(struct archive_read *a,
3409     struct archive_entry *entry)
3410 {
3411 	struct zip *zip;
3412 
3413 	a->archive.archive_format = ARCHIVE_FORMAT_ZIP;
3414 	if (a->archive.archive_format_name == NULL)
3415 		a->archive.archive_format_name = "ZIP";
3416 
3417 	zip = (struct zip *)(a->format->data);
3418 
3419 	/*
3420 	 * It should be sufficient to call archive_read_next_header() for
3421 	 * a reader to determine if an entry is encrypted or not. If the
3422 	 * encryption of an entry is only detectable when calling
3423 	 * archive_read_data(), so be it. We'll do the same check there
3424 	 * as well.
3425 	 */
3426 	if (zip->has_encrypted_entries ==
3427 			ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW)
3428 		zip->has_encrypted_entries = 0;
3429 
3430 	/* Make sure we have a zip_entry structure to use. */
3431 	if (zip->zip_entries == NULL) {
3432 		zip->zip_entries = malloc(sizeof(struct zip_entry));
3433 		if (zip->zip_entries == NULL) {
3434 			archive_set_error(&a->archive, ENOMEM,
3435 			    "Out  of memory");
3436 			return ARCHIVE_FATAL;
3437 		}
3438 	}
3439 	zip->entry = zip->zip_entries;
3440 	memset(zip->entry, 0, sizeof(struct zip_entry));
3441 
3442 	if (zip->cctx_valid)
3443 		archive_decrypto_aes_ctr_release(&zip->cctx);
3444 	if (zip->hctx_valid)
3445 		archive_hmac_sha1_cleanup(&zip->hctx);
3446 	zip->tctx_valid = zip->cctx_valid = zip->hctx_valid = 0;
3447 	__archive_read_reset_passphrase(a);
3448 
3449 	/* Search ahead for the next local file header. */
3450 	__archive_read_consume(a, zip->unconsumed);
3451 	zip->unconsumed = 0;
3452 	for (;;) {
3453 		int64_t skipped = 0;
3454 		const char *p, *end;
3455 		ssize_t bytes;
3456 
3457 		p = __archive_read_ahead(a, 4, &bytes);
3458 		if (p == NULL)
3459 			return (ARCHIVE_FATAL);
3460 		end = p + bytes;
3461 
3462 		while (p + 4 <= end) {
3463 			if (p[0] == 'P' && p[1] == 'K') {
3464 				if (p[2] == '\003' && p[3] == '\004') {
3465 					/* Regular file entry. */
3466 					__archive_read_consume(a, skipped);
3467 					return zip_read_local_file_header(a,
3468 					    entry, zip);
3469 				}
3470 
3471                               /*
3472                                * TODO: We cannot restore permissions
3473                                * based only on the local file headers.
3474                                * Consider scanning the central
3475                                * directory and returning additional
3476                                * entries for at least directories.
3477                                * This would allow us to properly set
3478                                * directory permissions.
3479 			       *
3480 			       * This won't help us fix symlinks
3481 			       * and may not help with regular file
3482 			       * permissions, either.  <sigh>
3483                                */
3484                               if (p[2] == '\001' && p[3] == '\002') {
3485                                       return (ARCHIVE_EOF);
3486                               }
3487 
3488                               /* End of central directory?  Must be an
3489                                * empty archive. */
3490                               if ((p[2] == '\005' && p[3] == '\006')
3491                                   || (p[2] == '\006' && p[3] == '\006'))
3492                                       return (ARCHIVE_EOF);
3493 			}
3494 			++p;
3495 			++skipped;
3496 		}
3497 		__archive_read_consume(a, skipped);
3498 	}
3499 }
3500 
3501 static int
archive_read_format_zip_read_data_skip_streamable(struct archive_read * a)3502 archive_read_format_zip_read_data_skip_streamable(struct archive_read *a)
3503 {
3504 	struct zip *zip;
3505 	int64_t bytes_skipped;
3506 
3507 	zip = (struct zip *)(a->format->data);
3508 	bytes_skipped = __archive_read_consume(a, zip->unconsumed);
3509 	zip->unconsumed = 0;
3510 	if (bytes_skipped < 0)
3511 		return (ARCHIVE_FATAL);
3512 
3513 	/* If we've already read to end of data, we're done. */
3514 	if (zip->end_of_entry)
3515 		return (ARCHIVE_OK);
3516 
3517 	/* So we know we're streaming... */
3518 	if (0 == (zip->entry->zip_flags & ZIP_LENGTH_AT_END)
3519 	    || zip->entry->compressed_size > 0) {
3520 		/* We know the compressed length, so we can just skip. */
3521 		bytes_skipped = __archive_read_consume(a,
3522 					zip->entry_bytes_remaining);
3523 		if (bytes_skipped < 0)
3524 			return (ARCHIVE_FATAL);
3525 		return (ARCHIVE_OK);
3526 	}
3527 
3528 	if (zip->init_decryption) {
3529 		int r;
3530 
3531 		zip->has_encrypted_entries = 1;
3532 		if (zip->entry->zip_flags & ZIP_STRONG_ENCRYPTED)
3533 			r = read_decryption_header(a);
3534 		else if (zip->entry->compression == WINZIP_AES_ENCRYPTION)
3535 			r = init_WinZip_AES_decryption(a);
3536 		else
3537 			r = init_traditional_PKWARE_decryption(a);
3538 		if (r != ARCHIVE_OK)
3539 			return (r);
3540 		zip->init_decryption = 0;
3541 	}
3542 
3543 	/* We're streaming and we don't know the length. */
3544 	/* If the body is compressed and we know the format, we can
3545 	 * find an exact end-of-entry by decompressing it. */
3546 	switch (zip->entry->compression) {
3547 #ifdef HAVE_ZLIB_H
3548 	case 8: /* Deflate compression. */
3549 		while (!zip->end_of_entry) {
3550 			int64_t offset = 0;
3551 			const void *buff = NULL;
3552 			size_t size = 0;
3553 			int r;
3554 			r =  zip_read_data_deflate(a, &buff, &size, &offset);
3555 			if (r != ARCHIVE_OK)
3556 				return (r);
3557 		}
3558 		return ARCHIVE_OK;
3559 #endif
3560 	default: /* Uncompressed or unknown. */
3561 		/* Scan for a PK\007\010 signature. */
3562 		for (;;) {
3563 			const char *p, *buff;
3564 			ssize_t bytes_avail;
3565 			buff = __archive_read_ahead(a, 16, &bytes_avail);
3566 			if (bytes_avail < 16) {
3567 				archive_set_error(&a->archive,
3568 				    ARCHIVE_ERRNO_FILE_FORMAT,
3569 				    "Truncated ZIP file data");
3570 				return (ARCHIVE_FATAL);
3571 			}
3572 			p = buff;
3573 			while (p <= buff + bytes_avail - 16) {
3574 				if (p[3] == 'P') { p += 3; }
3575 				else if (p[3] == 'K') { p += 2; }
3576 				else if (p[3] == '\007') { p += 1; }
3577 				else if (p[3] == '\010' && p[2] == '\007'
3578 				    && p[1] == 'K' && p[0] == 'P') {
3579 					if (zip->entry->flags & LA_USED_ZIP64)
3580 						__archive_read_consume(a,
3581 						    p - buff + 24);
3582 					else
3583 						__archive_read_consume(a,
3584 						    p - buff + 16);
3585 					return ARCHIVE_OK;
3586 				} else { p += 4; }
3587 			}
3588 			__archive_read_consume(a, p - buff);
3589 		}
3590 	}
3591 }
3592 
3593 int
archive_read_support_format_zip_streamable(struct archive * _a)3594 archive_read_support_format_zip_streamable(struct archive *_a)
3595 {
3596 	struct archive_read *a = (struct archive_read *)_a;
3597 	struct zip *zip;
3598 	int r;
3599 
3600 	archive_check_magic(_a, ARCHIVE_READ_MAGIC,
3601 	    ARCHIVE_STATE_NEW, "archive_read_support_format_zip");
3602 
3603 	zip = (struct zip *)calloc(1, sizeof(*zip));
3604 	if (zip == NULL) {
3605 		archive_set_error(&a->archive, ENOMEM,
3606 		    "Can't allocate zip data");
3607 		return (ARCHIVE_FATAL);
3608 	}
3609 
3610 	/* Streamable reader doesn't support mac extensions. */
3611 	zip->process_mac_extensions = 0;
3612 
3613 	/*
3614 	 * Until enough data has been read, we cannot tell about
3615 	 * any encrypted entries yet.
3616 	 */
3617 	zip->has_encrypted_entries = ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW;
3618 	zip->crc32func = real_crc32;
3619 
3620 	r = __archive_read_register_format(a,
3621 	    zip,
3622 	    "zip",
3623 	    archive_read_format_zip_streamable_bid,
3624 	    archive_read_format_zip_options,
3625 	    archive_read_format_zip_streamable_read_header,
3626 	    archive_read_format_zip_read_data,
3627 	    archive_read_format_zip_read_data_skip_streamable,
3628 	    NULL,
3629 	    archive_read_format_zip_cleanup,
3630 	    archive_read_support_format_zip_capabilities_streamable,
3631 	    archive_read_format_zip_has_encrypted_entries);
3632 
3633 	if (r != ARCHIVE_OK)
3634 		free(zip);
3635 	return (ARCHIVE_OK);
3636 }
3637 
3638 /* ------------------------------------------------------------------------ */
3639 
3640 /*
3641  * Seeking-mode support
3642  */
3643 
3644 static int
archive_read_support_format_zip_capabilities_seekable(struct archive_read * a)3645 archive_read_support_format_zip_capabilities_seekable(struct archive_read * a)
3646 {
3647 	(void)a; /* UNUSED */
3648 	return (ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_DATA |
3649 		ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_METADATA);
3650 }
3651 
3652 /*
3653  * TODO: This is a performance sink because it forces the read core to
3654  * drop buffered data from the start of file, which will then have to
3655  * be re-read again if this bidder loses.
3656  *
3657  * We workaround this a little by passing in the best bid so far so
3658  * that later bidders can do nothing if they know they'll never
3659  * outbid.  But we can certainly do better...
3660  */
3661 static int
read_eocd(struct zip * zip,const char * p,int64_t current_offset)3662 read_eocd(struct zip *zip, const char *p, int64_t current_offset)
3663 {
3664 	uint16_t disk_num;
3665 	uint32_t cd_size, cd_offset;
3666 
3667 	disk_num = archive_le16dec(p + 4);
3668 	cd_size = archive_le32dec(p + 12);
3669 	cd_offset = archive_le32dec(p + 16);
3670 
3671 	/* Sanity-check the EOCD we've found. */
3672 
3673 	/* This must be the first volume. */
3674 	if (disk_num != 0)
3675 		return 0;
3676 	/* Central directory must be on this volume. */
3677 	if (disk_num != archive_le16dec(p + 6))
3678 		return 0;
3679 	/* All central directory entries must be on this volume. */
3680 	if (archive_le16dec(p + 10) != archive_le16dec(p + 8))
3681 		return 0;
3682 	/* Central directory can't extend beyond start of EOCD record. */
3683 	if (cd_offset + cd_size > current_offset)
3684 		return 0;
3685 
3686 	/* Save the central directory location for later use. */
3687 	zip->central_directory_offset = cd_offset;
3688 	zip->central_directory_offset_adjusted = current_offset - cd_size;
3689 
3690 	/* This is just a tiny bit higher than the maximum
3691 	   returned by the streaming Zip bidder.  This ensures
3692 	   that the more accurate seeking Zip parser wins
3693 	   whenever seek is available. */
3694 	return 32;
3695 }
3696 
3697 /*
3698  * Examine Zip64 EOCD locator:  If it's valid, store the information
3699  * from it.
3700  */
3701 static int
read_zip64_eocd(struct archive_read * a,struct zip * zip,const char * p)3702 read_zip64_eocd(struct archive_read *a, struct zip *zip, const char *p)
3703 {
3704 	int64_t eocd64_offset;
3705 	int64_t eocd64_size;
3706 
3707 	/* Sanity-check the locator record. */
3708 
3709 	/* Central dir must be on first volume. */
3710 	if (archive_le32dec(p + 4) != 0)
3711 		return 0;
3712 	/* Must be only a single volume. */
3713 	if (archive_le32dec(p + 16) != 1)
3714 		return 0;
3715 
3716 	/* Find the Zip64 EOCD record. */
3717 	eocd64_offset = archive_le64dec(p + 8);
3718 	if (__archive_read_seek(a, eocd64_offset, SEEK_SET) < 0)
3719 		return 0;
3720 	if ((p = __archive_read_ahead(a, 56, NULL)) == NULL)
3721 		return 0;
3722 	/* Make sure we can read all of it. */
3723 	eocd64_size = archive_le64dec(p + 4) + 12;
3724 	if (eocd64_size < 56 || eocd64_size > 16384)
3725 		return 0;
3726 	if ((p = __archive_read_ahead(a, (size_t)eocd64_size, NULL)) == NULL)
3727 		return 0;
3728 
3729 	/* Sanity-check the EOCD64 */
3730 	if (archive_le32dec(p + 16) != 0) /* Must be disk #0 */
3731 		return 0;
3732 	if (archive_le32dec(p + 20) != 0) /* CD must be on disk #0 */
3733 		return 0;
3734 	/* CD can't be split. */
3735 	if (archive_le64dec(p + 24) != archive_le64dec(p + 32))
3736 		return 0;
3737 
3738 	/* Save the central directory offset for later use. */
3739 	zip->central_directory_offset = archive_le64dec(p + 48);
3740 	/* TODO: Needs scanning backwards to find the eocd64 instead of assuming */
3741 	zip->central_directory_offset_adjusted = zip->central_directory_offset;
3742 
3743 	return 32;
3744 }
3745 
3746 static int
archive_read_format_zip_seekable_bid(struct archive_read * a,int best_bid)3747 archive_read_format_zip_seekable_bid(struct archive_read *a, int best_bid)
3748 {
3749 	struct zip *zip = (struct zip *)a->format->data;
3750 	int64_t file_size, current_offset;
3751 	const char *p;
3752 	int i, tail;
3753 
3754 	/* If someone has already bid more than 32, then avoid
3755 	   trashing the look-ahead buffers with a seek. */
3756 	if (best_bid > 32)
3757 		return (-1);
3758 
3759 	file_size = __archive_read_seek(a, 0, SEEK_END);
3760 	if (file_size <= 0)
3761 		return 0;
3762 
3763 	/* Search last 16k of file for end-of-central-directory
3764 	 * record (which starts with PK\005\006) */
3765 	tail = (int)zipmin(1024 * 16, file_size);
3766 	current_offset = __archive_read_seek(a, -tail, SEEK_END);
3767 	if (current_offset < 0)
3768 		return 0;
3769 	if ((p = __archive_read_ahead(a, (size_t)tail, NULL)) == NULL)
3770 		return 0;
3771 	/* Boyer-Moore search backwards from the end, since we want
3772 	 * to match the last EOCD in the file (there can be more than
3773 	 * one if there is an uncompressed Zip archive as a member
3774 	 * within this Zip archive). */
3775 	for (i = tail - 22; i > 0;) {
3776 		switch (p[i]) {
3777 		case 'P':
3778 			if (memcmp(p + i, "PK\005\006", 4) == 0) {
3779 				int ret = read_eocd(zip, p + i,
3780 				    current_offset + i);
3781 				/* Zip64 EOCD locator precedes
3782 				 * regular EOCD if present. */
3783 				if (i >= 20 && memcmp(p + i - 20, "PK\006\007", 4) == 0) {
3784 					int ret_zip64 = read_zip64_eocd(a, zip, p + i - 20);
3785 					if (ret_zip64 > ret)
3786 						ret = ret_zip64;
3787 				}
3788 				return (ret);
3789 			}
3790 			i -= 4;
3791 			break;
3792 		case 'K': i -= 1; break;
3793 		case 005: i -= 2; break;
3794 		case 006: i -= 3; break;
3795 		default: i -= 4; break;
3796 		}
3797 	}
3798 	return 0;
3799 }
3800 
3801 /* The red-black trees are only used in seeking mode to manage
3802  * the in-memory copy of the central directory. */
3803 
3804 static int
cmp_node(const struct archive_rb_node * n1,const struct archive_rb_node * n2)3805 cmp_node(const struct archive_rb_node *n1, const struct archive_rb_node *n2)
3806 {
3807 	const struct zip_entry *e1 = (const struct zip_entry *)n1;
3808 	const struct zip_entry *e2 = (const struct zip_entry *)n2;
3809 
3810 	if (e1->local_header_offset > e2->local_header_offset)
3811 		return -1;
3812 	if (e1->local_header_offset < e2->local_header_offset)
3813 		return 1;
3814 	return 0;
3815 }
3816 
3817 static int
cmp_key(const struct archive_rb_node * n,const void * key)3818 cmp_key(const struct archive_rb_node *n, const void *key)
3819 {
3820 	/* This function won't be called */
3821 	(void)n; /* UNUSED */
3822 	(void)key; /* UNUSED */
3823 	return 1;
3824 }
3825 
3826 static const struct archive_rb_tree_ops rb_ops = {
3827 	&cmp_node, &cmp_key
3828 };
3829 
3830 static int
rsrc_cmp_node(const struct archive_rb_node * n1,const struct archive_rb_node * n2)3831 rsrc_cmp_node(const struct archive_rb_node *n1,
3832     const struct archive_rb_node *n2)
3833 {
3834 	const struct zip_entry *e1 = (const struct zip_entry *)n1;
3835 	const struct zip_entry *e2 = (const struct zip_entry *)n2;
3836 
3837 	return (strcmp(e2->rsrcname.s, e1->rsrcname.s));
3838 }
3839 
3840 static int
rsrc_cmp_key(const struct archive_rb_node * n,const void * key)3841 rsrc_cmp_key(const struct archive_rb_node *n, const void *key)
3842 {
3843 	const struct zip_entry *e = (const struct zip_entry *)n;
3844 	return (strcmp((const char *)key, e->rsrcname.s));
3845 }
3846 
3847 static const struct archive_rb_tree_ops rb_rsrc_ops = {
3848 	&rsrc_cmp_node, &rsrc_cmp_key
3849 };
3850 
3851 static const char *
rsrc_basename(const char * name,size_t name_length)3852 rsrc_basename(const char *name, size_t name_length)
3853 {
3854 	const char *s, *r;
3855 
3856 	r = s = name;
3857 	for (;;) {
3858 		s = memchr(s, '/', name_length - (s - name));
3859 		if (s == NULL)
3860 			break;
3861 		r = ++s;
3862 	}
3863 	return (r);
3864 }
3865 
3866 static void
expose_parent_dirs(struct zip * zip,const char * name,size_t name_length)3867 expose_parent_dirs(struct zip *zip, const char *name, size_t name_length)
3868 {
3869 	struct archive_string str;
3870 	struct zip_entry *dir;
3871 	char *s;
3872 
3873 	archive_string_init(&str);
3874 	archive_strncpy(&str, name, name_length);
3875 	for (;;) {
3876 		s = strrchr(str.s, '/');
3877 		if (s == NULL)
3878 			break;
3879 		*s = '\0';
3880 		/* Transfer the parent directory from zip->tree_rsrc RB
3881 		 * tree to zip->tree RB tree to expose. */
3882 		dir = (struct zip_entry *)
3883 		    __archive_rb_tree_find_node(&zip->tree_rsrc, str.s);
3884 		if (dir == NULL)
3885 			break;
3886 		__archive_rb_tree_remove_node(&zip->tree_rsrc, &dir->node);
3887 		archive_string_free(&dir->rsrcname);
3888 		__archive_rb_tree_insert_node(&zip->tree, &dir->node);
3889 	}
3890 	archive_string_free(&str);
3891 }
3892 
3893 static int
slurp_central_directory(struct archive_read * a,struct archive_entry * entry,struct zip * zip)3894 slurp_central_directory(struct archive_read *a, struct archive_entry* entry,
3895     struct zip *zip)
3896 {
3897 	ssize_t i;
3898 	unsigned found;
3899 	int64_t correction;
3900 	ssize_t bytes_avail;
3901 	const char *p;
3902 
3903 	/*
3904 	 * Find the start of the central directory.  The end-of-CD
3905 	 * record has our starting point, but there are lots of
3906 	 * Zip archives which have had other data prepended to the
3907 	 * file, which makes the recorded offsets all too small.
3908 	 * So we search forward from the specified offset until we
3909 	 * find the real start of the central directory.  Then we
3910 	 * know the correction we need to apply to account for leading
3911 	 * padding.
3912 	 */
3913 	if (__archive_read_seek(a, zip->central_directory_offset_adjusted, SEEK_SET)
3914 		< 0)
3915 		return ARCHIVE_FATAL;
3916 
3917 	found = 0;
3918 	while (!found) {
3919 		if ((p = __archive_read_ahead(a, 20, &bytes_avail)) == NULL)
3920 			return ARCHIVE_FATAL;
3921 		for (found = 0, i = 0; !found && i < bytes_avail - 4;) {
3922 			switch (p[i + 3]) {
3923 			case 'P': i += 3; break;
3924 			case 'K': i += 2; break;
3925 			case 001: i += 1; break;
3926 			case 002:
3927 				if (memcmp(p + i, "PK\001\002", 4) == 0) {
3928 					p += i;
3929 					found = 1;
3930 				} else
3931 					i += 4;
3932 				break;
3933 			case 005: i += 1; break;
3934 			case 006:
3935 				if (memcmp(p + i, "PK\005\006", 4) == 0) {
3936 					p += i;
3937 					found = 1;
3938 				} else if (memcmp(p + i, "PK\006\006", 4) == 0) {
3939 					p += i;
3940 					found = 1;
3941 				} else
3942 					i += 1;
3943 				break;
3944 			default: i += 4; break;
3945 			}
3946 		}
3947 		__archive_read_consume(a, i);
3948 	}
3949 	correction = archive_filter_bytes(&a->archive, 0)
3950 			- zip->central_directory_offset;
3951 
3952 	__archive_rb_tree_init(&zip->tree, &rb_ops);
3953 	__archive_rb_tree_init(&zip->tree_rsrc, &rb_rsrc_ops);
3954 
3955 	zip->central_directory_entries_total = 0;
3956 	while (1) {
3957 		struct zip_entry *zip_entry;
3958 		size_t filename_length, extra_length, comment_length;
3959 		uint32_t external_attributes;
3960 		const char *name, *r;
3961 
3962 		if ((p = __archive_read_ahead(a, 4, NULL)) == NULL)
3963 			return ARCHIVE_FATAL;
3964 		if (memcmp(p, "PK\006\006", 4) == 0
3965 		    || memcmp(p, "PK\005\006", 4) == 0) {
3966 			break;
3967 		} else if (memcmp(p, "PK\001\002", 4) != 0) {
3968 			archive_set_error(&a->archive,
3969 			    -1, "Invalid central directory signature");
3970 			return ARCHIVE_FATAL;
3971 		}
3972 		if ((p = __archive_read_ahead(a, 46, NULL)) == NULL)
3973 			return ARCHIVE_FATAL;
3974 
3975 		zip_entry = calloc(1, sizeof(struct zip_entry));
3976 		if (zip_entry == NULL) {
3977 			archive_set_error(&a->archive, ENOMEM,
3978 				"Can't allocate zip entry");
3979 			return ARCHIVE_FATAL;
3980 		}
3981 		zip_entry->next = zip->zip_entries;
3982 		zip_entry->flags |= LA_FROM_CENTRAL_DIRECTORY;
3983 		zip->zip_entries = zip_entry;
3984 		zip->central_directory_entries_total++;
3985 
3986 		/* version = p[4]; */
3987 		zip_entry->system = p[5];
3988 		/* version_required = archive_le16dec(p + 6); */
3989 		zip_entry->zip_flags = archive_le16dec(p + 8);
3990 		if (zip_entry->zip_flags
3991 		      & (ZIP_ENCRYPTED | ZIP_STRONG_ENCRYPTED)){
3992 			zip->has_encrypted_entries = 1;
3993 		}
3994 		zip_entry->compression = (char)archive_le16dec(p + 10);
3995 		zip_entry->mtime = zip_time(p + 12);
3996 		zip_entry->crc32 = archive_le32dec(p + 16);
3997 		if (zip_entry->zip_flags & ZIP_LENGTH_AT_END)
3998 			zip_entry->decdat = p[13];
3999 		else
4000 			zip_entry->decdat = p[19];
4001 		zip_entry->compressed_size = archive_le32dec(p + 20);
4002 		zip_entry->uncompressed_size = archive_le32dec(p + 24);
4003 		filename_length = archive_le16dec(p + 28);
4004 		extra_length = archive_le16dec(p + 30);
4005 		comment_length = archive_le16dec(p + 32);
4006 		/* disk_start = archive_le16dec(p + 34);
4007 		 *   Better be zero.
4008 		 * internal_attributes = archive_le16dec(p + 36);
4009 		 *   text bit */
4010 		external_attributes = archive_le32dec(p + 38);
4011 		zip_entry->local_header_offset =
4012 		    archive_le32dec(p + 42) + correction;
4013 
4014 		/* If we can't guess the mode, leave it zero here;
4015 		   when we read the local file header we might get
4016 		   more information. */
4017 		if (zip_entry->system == 3) {
4018 			zip_entry->mode = external_attributes >> 16;
4019 		} else if (zip_entry->system == 0) {
4020 			// Interpret MSDOS directory bit
4021 			if (0x10 == (external_attributes & 0x10)) {
4022 				zip_entry->mode = AE_IFDIR | 0775;
4023 			} else {
4024 				zip_entry->mode = AE_IFREG | 0664;
4025 			}
4026 			if (0x01 == (external_attributes & 0x01)) {
4027 				// Read-only bit; strip write permissions
4028 				zip_entry->mode &= 0555;
4029 			}
4030 		} else {
4031 			zip_entry->mode = 0;
4032 		}
4033 
4034 		/* We're done with the regular data; get the filename and
4035 		 * extra data. */
4036 		__archive_read_consume(a, 46);
4037 		p = __archive_read_ahead(a, filename_length + extra_length,
4038 			NULL);
4039 		if (p == NULL) {
4040 			archive_set_error(&a->archive,
4041 			    ARCHIVE_ERRNO_FILE_FORMAT,
4042 			    "Truncated ZIP file header");
4043 			return ARCHIVE_FATAL;
4044 		}
4045 		if (ARCHIVE_OK != process_extra(a, entry, p + filename_length,
4046 		    extra_length, zip_entry)) {
4047 			return ARCHIVE_FATAL;
4048 		}
4049 
4050 		/*
4051 		 * Mac resource fork files are stored under the
4052 		 * "__MACOSX/" directory, so we should check if
4053 		 * it is.
4054 		 */
4055 		if (!zip->process_mac_extensions) {
4056 			/* Treat every entry as a regular entry. */
4057 			__archive_rb_tree_insert_node(&zip->tree,
4058 			    &zip_entry->node);
4059 		} else {
4060 			name = p;
4061 			r = rsrc_basename(name, filename_length);
4062 			if (filename_length >= 9 &&
4063 			    strncmp("__MACOSX/", name, 9) == 0) {
4064 				/* If this file is not a resource fork nor
4065 				 * a directory. We should treat it as a non
4066 				 * resource fork file to expose it. */
4067 				if (name[filename_length-1] != '/' &&
4068 				    (r - name < 3 || r[0] != '.' ||
4069 				     r[1] != '_')) {
4070 					__archive_rb_tree_insert_node(
4071 					    &zip->tree, &zip_entry->node);
4072 					/* Expose its parent directories. */
4073 					expose_parent_dirs(zip, name,
4074 					    filename_length);
4075 				} else {
4076 					/* This file is a resource fork file or
4077 					 * a directory. */
4078 					archive_strncpy(&(zip_entry->rsrcname),
4079 					     name, filename_length);
4080 					__archive_rb_tree_insert_node(
4081 					    &zip->tree_rsrc, &zip_entry->node);
4082 				}
4083 			} else {
4084 				/* Generate resource fork name to find its
4085 				 * resource file at zip->tree_rsrc. */
4086 
4087 				/* If this is an entry ending with slash,
4088 				 * make the resource for name slash-less
4089 				 * as the actual resource fork doesn't end with '/'.
4090 				 */
4091 				size_t tmp_length = filename_length;
4092 				if (tmp_length > 0 && name[tmp_length - 1] == '/') {
4093 					tmp_length--;
4094 					r = rsrc_basename(name, tmp_length);
4095 				}
4096 
4097 				archive_strcpy(&(zip_entry->rsrcname),
4098 				    "__MACOSX/");
4099 				archive_strncat(&(zip_entry->rsrcname),
4100 				    name, r - name);
4101 				archive_strcat(&(zip_entry->rsrcname), "._");
4102 				archive_strncat(&(zip_entry->rsrcname),
4103 				    name + (r - name),
4104 				    tmp_length - (r - name));
4105 				/* Register an entry to RB tree to sort it by
4106 				 * file offset. */
4107 				__archive_rb_tree_insert_node(&zip->tree,
4108 				    &zip_entry->node);
4109 			}
4110 		}
4111 
4112 		/* Skip the comment too ... */
4113 		__archive_read_consume(a,
4114 		    filename_length + extra_length + comment_length);
4115 	}
4116 
4117 	return ARCHIVE_OK;
4118 }
4119 
4120 static ssize_t
zip_get_local_file_header_size(struct archive_read * a,size_t extra)4121 zip_get_local_file_header_size(struct archive_read *a, size_t extra)
4122 {
4123 	const char *p;
4124 	ssize_t filename_length, extra_length;
4125 
4126 	if ((p = __archive_read_ahead(a, extra + 30, NULL)) == NULL) {
4127 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
4128 		    "Truncated ZIP file header");
4129 		return (ARCHIVE_WARN);
4130 	}
4131 	p += extra;
4132 
4133 	if (memcmp(p, "PK\003\004", 4) != 0) {
4134 		archive_set_error(&a->archive, -1, "Damaged Zip archive");
4135 		return ARCHIVE_WARN;
4136 	}
4137 	filename_length = archive_le16dec(p + 26);
4138 	extra_length = archive_le16dec(p + 28);
4139 
4140 	return (30 + filename_length + extra_length);
4141 }
4142 
4143 static int
zip_read_mac_metadata(struct archive_read * a,struct archive_entry * entry,struct zip_entry * rsrc)4144 zip_read_mac_metadata(struct archive_read *a, struct archive_entry *entry,
4145     struct zip_entry *rsrc)
4146 {
4147 	struct zip *zip = (struct zip *)a->format->data;
4148 	unsigned char *metadata, *mp;
4149 	int64_t offset = archive_filter_bytes(&a->archive, 0);
4150 	size_t remaining_bytes, metadata_bytes;
4151 	ssize_t hsize;
4152 	int ret = ARCHIVE_OK, eof;
4153 
4154 	switch(rsrc->compression) {
4155 	case 0:  /* No compression. */
4156 		if (rsrc->uncompressed_size != rsrc->compressed_size) {
4157 			archive_set_error(&a->archive,
4158 			    ARCHIVE_ERRNO_FILE_FORMAT,
4159 			    "Malformed OS X metadata entry: "
4160 			    "inconsistent size");
4161 			return (ARCHIVE_FATAL);
4162 		}
4163 #ifdef HAVE_ZLIB_H
4164 	case 8: /* Deflate compression. */
4165 #endif
4166 		break;
4167 	default: /* Unsupported compression. */
4168 		/* Return a warning. */
4169 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
4170 		    "Unsupported ZIP compression method (%s)",
4171 		    compression_name(rsrc->compression));
4172 		/* We can't decompress this entry, but we will
4173 		 * be able to skip() it and try the next entry. */
4174 		return (ARCHIVE_WARN);
4175 	}
4176 
4177 	if (rsrc->uncompressed_size > (4 * 1024 * 1024)) {
4178 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
4179 		    "Mac metadata is too large: %jd > 4M bytes",
4180 		    (intmax_t)rsrc->uncompressed_size);
4181 		return (ARCHIVE_WARN);
4182 	}
4183 	if (rsrc->compressed_size > (4 * 1024 * 1024)) {
4184 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
4185 		    "Mac metadata is too large: %jd > 4M bytes",
4186 		    (intmax_t)rsrc->compressed_size);
4187 		return (ARCHIVE_WARN);
4188 	}
4189 
4190 	metadata = malloc((size_t)rsrc->uncompressed_size);
4191 	if (metadata == NULL) {
4192 		archive_set_error(&a->archive, ENOMEM,
4193 		    "Can't allocate memory for Mac metadata");
4194 		return (ARCHIVE_FATAL);
4195 	}
4196 
4197 	if (offset < rsrc->local_header_offset)
4198 		__archive_read_consume(a, rsrc->local_header_offset - offset);
4199 	else if (offset != rsrc->local_header_offset) {
4200 		__archive_read_seek(a, rsrc->local_header_offset, SEEK_SET);
4201 	}
4202 
4203 	hsize = zip_get_local_file_header_size(a, 0);
4204 	__archive_read_consume(a, hsize);
4205 
4206 	remaining_bytes = (size_t)rsrc->compressed_size;
4207 	metadata_bytes = (size_t)rsrc->uncompressed_size;
4208 	mp = metadata;
4209 	eof = 0;
4210 	while (!eof && remaining_bytes) {
4211 		const unsigned char *p;
4212 		ssize_t bytes_avail;
4213 		size_t bytes_used;
4214 
4215 		p = __archive_read_ahead(a, 1, &bytes_avail);
4216 		if (p == NULL) {
4217 			archive_set_error(&a->archive,
4218 			    ARCHIVE_ERRNO_FILE_FORMAT,
4219 			    "Truncated ZIP file header");
4220 			ret = ARCHIVE_WARN;
4221 			goto exit_mac_metadata;
4222 		}
4223 		if ((size_t)bytes_avail > remaining_bytes)
4224 			bytes_avail = remaining_bytes;
4225 		switch(rsrc->compression) {
4226 		case 0:  /* No compression. */
4227 			if ((size_t)bytes_avail > metadata_bytes)
4228 				bytes_avail = metadata_bytes;
4229 			memcpy(mp, p, bytes_avail);
4230 			bytes_used = (size_t)bytes_avail;
4231 			metadata_bytes -= bytes_used;
4232 			mp += bytes_used;
4233 			if (metadata_bytes == 0)
4234 				eof = 1;
4235 			break;
4236 #ifdef HAVE_ZLIB_H
4237 		case 8: /* Deflate compression. */
4238 		{
4239 			int r;
4240 
4241 			ret = zip_deflate_init(a, zip);
4242 			if (ret != ARCHIVE_OK)
4243 				goto exit_mac_metadata;
4244 			zip->stream.next_in =
4245 			    (Bytef *)(uintptr_t)(const void *)p;
4246 			zip->stream.avail_in = (uInt)bytes_avail;
4247 			zip->stream.total_in = 0;
4248 			zip->stream.next_out = mp;
4249 			zip->stream.avail_out = (uInt)metadata_bytes;
4250 			zip->stream.total_out = 0;
4251 
4252 			r = inflate(&zip->stream, 0);
4253 			switch (r) {
4254 			case Z_OK:
4255 				break;
4256 			case Z_STREAM_END:
4257 				eof = 1;
4258 				break;
4259 			case Z_MEM_ERROR:
4260 				archive_set_error(&a->archive, ENOMEM,
4261 				    "Out of memory for ZIP decompression");
4262 				ret = ARCHIVE_FATAL;
4263 				goto exit_mac_metadata;
4264 			default:
4265 				archive_set_error(&a->archive,
4266 				    ARCHIVE_ERRNO_MISC,
4267 				    "ZIP decompression failed (%d)", r);
4268 				ret = ARCHIVE_FATAL;
4269 				goto exit_mac_metadata;
4270 			}
4271 			bytes_used = zip->stream.total_in;
4272 			metadata_bytes -= zip->stream.total_out;
4273 			mp += zip->stream.total_out;
4274 			break;
4275 		}
4276 #endif
4277 		default:
4278 			bytes_used = 0;
4279 			break;
4280 		}
4281 		__archive_read_consume(a, bytes_used);
4282 		remaining_bytes -= bytes_used;
4283 	}
4284 	archive_entry_copy_mac_metadata(entry, metadata,
4285 	    (size_t)rsrc->uncompressed_size - metadata_bytes);
4286 
4287 exit_mac_metadata:
4288 	__archive_read_seek(a, offset, SEEK_SET);
4289 	zip->decompress_init = 0;
4290 	free(metadata);
4291 	return (ret);
4292 }
4293 
4294 static int
archive_read_format_zip_seekable_read_header(struct archive_read * a,struct archive_entry * entry)4295 archive_read_format_zip_seekable_read_header(struct archive_read *a,
4296 	struct archive_entry *entry)
4297 {
4298 	struct zip *zip = (struct zip *)a->format->data;
4299 	struct zip_entry *rsrc;
4300 	int64_t offset;
4301 	int r, ret = ARCHIVE_OK;
4302 
4303 	/*
4304 	 * It should be sufficient to call archive_read_next_header() for
4305 	 * a reader to determine if an entry is encrypted or not. If the
4306 	 * encryption of an entry is only detectable when calling
4307 	 * archive_read_data(), so be it. We'll do the same check there
4308 	 * as well.
4309 	 */
4310 	if (zip->has_encrypted_entries ==
4311 			ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW)
4312 		zip->has_encrypted_entries = 0;
4313 
4314 	a->archive.archive_format = ARCHIVE_FORMAT_ZIP;
4315 	if (a->archive.archive_format_name == NULL)
4316 		a->archive.archive_format_name = "ZIP";
4317 
4318 	if (zip->zip_entries == NULL) {
4319 		r = slurp_central_directory(a, entry, zip);
4320 		if (r != ARCHIVE_OK)
4321 			return r;
4322 		/* Get first entry whose local header offset is lower than
4323 		 * other entries in the archive file. */
4324 		zip->entry =
4325 		    (struct zip_entry *)ARCHIVE_RB_TREE_MIN(&zip->tree);
4326 	} else if (zip->entry != NULL) {
4327 		/* Get next entry in local header offset order. */
4328 		zip->entry = (struct zip_entry *)__archive_rb_tree_iterate(
4329 		    &zip->tree, &zip->entry->node, ARCHIVE_RB_DIR_RIGHT);
4330 	}
4331 
4332 	if (zip->entry == NULL)
4333 		return ARCHIVE_EOF;
4334 
4335 	if (zip->entry->rsrcname.s)
4336 		rsrc = (struct zip_entry *)__archive_rb_tree_find_node(
4337 		    &zip->tree_rsrc, zip->entry->rsrcname.s);
4338 	else
4339 		rsrc = NULL;
4340 
4341 	if (zip->cctx_valid)
4342 		archive_decrypto_aes_ctr_release(&zip->cctx);
4343 	if (zip->hctx_valid)
4344 		archive_hmac_sha1_cleanup(&zip->hctx);
4345 	zip->tctx_valid = zip->cctx_valid = zip->hctx_valid = 0;
4346 	__archive_read_reset_passphrase(a);
4347 
4348 	/* File entries are sorted by the header offset, we should mostly
4349 	 * use __archive_read_consume to advance a read point to avoid
4350 	 * redundant data reading.  */
4351 	offset = archive_filter_bytes(&a->archive, 0);
4352 	if (offset < zip->entry->local_header_offset)
4353 		__archive_read_consume(a,
4354 		    zip->entry->local_header_offset - offset);
4355 	else if (offset != zip->entry->local_header_offset) {
4356 		__archive_read_seek(a, zip->entry->local_header_offset,
4357 		    SEEK_SET);
4358 	}
4359 	zip->unconsumed = 0;
4360 	r = zip_read_local_file_header(a, entry, zip);
4361 	if (r != ARCHIVE_OK)
4362 		return r;
4363 	if (rsrc) {
4364 		int ret2 = zip_read_mac_metadata(a, entry, rsrc);
4365 		if (ret2 < ret)
4366 			ret = ret2;
4367 	}
4368 	return (ret);
4369 }
4370 
4371 /*
4372  * We're going to seek for the next header anyway, so we don't
4373  * need to bother doing anything here.
4374  */
4375 static int
archive_read_format_zip_read_data_skip_seekable(struct archive_read * a)4376 archive_read_format_zip_read_data_skip_seekable(struct archive_read *a)
4377 {
4378 	struct zip *zip;
4379 	zip = (struct zip *)(a->format->data);
4380 
4381 	zip->unconsumed = 0;
4382 	return (ARCHIVE_OK);
4383 }
4384 
4385 int
archive_read_support_format_zip_seekable(struct archive * _a)4386 archive_read_support_format_zip_seekable(struct archive *_a)
4387 {
4388 	struct archive_read *a = (struct archive_read *)_a;
4389 	struct zip *zip;
4390 	int r;
4391 
4392 	archive_check_magic(_a, ARCHIVE_READ_MAGIC,
4393 	    ARCHIVE_STATE_NEW, "archive_read_support_format_zip_seekable");
4394 
4395 	zip = (struct zip *)calloc(1, sizeof(*zip));
4396 	if (zip == NULL) {
4397 		archive_set_error(&a->archive, ENOMEM,
4398 		    "Can't allocate zip data");
4399 		return (ARCHIVE_FATAL);
4400 	}
4401 
4402 #ifdef HAVE_COPYFILE_H
4403 	/* Set this by default on Mac OS. */
4404 	zip->process_mac_extensions = 1;
4405 #endif
4406 
4407 	/*
4408 	 * Until enough data has been read, we cannot tell about
4409 	 * any encrypted entries yet.
4410 	 */
4411 	zip->has_encrypted_entries = ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW;
4412 	zip->crc32func = real_crc32;
4413 
4414 	r = __archive_read_register_format(a,
4415 	    zip,
4416 	    "zip",
4417 	    archive_read_format_zip_seekable_bid,
4418 	    archive_read_format_zip_options,
4419 	    archive_read_format_zip_seekable_read_header,
4420 	    archive_read_format_zip_read_data,
4421 	    archive_read_format_zip_read_data_skip_seekable,
4422 	    NULL,
4423 	    archive_read_format_zip_cleanup,
4424 	    archive_read_support_format_zip_capabilities_seekable,
4425 	    archive_read_format_zip_has_encrypted_entries);
4426 
4427 	if (r != ARCHIVE_OK)
4428 		free(zip);
4429 	return (ARCHIVE_OK);
4430 }
4431 
4432 /*# vim:set noet:*/
4433