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_time_private.h"
75 #include "archive_ppmd8_private.h"
76
77 #ifndef HAVE_ZLIB_H
78 #include "archive_crc32.h"
79 #endif
80
81 /* length of local file header, not including filename and extra */
82 #define ZIP_LOCHDR_LEN 30U
83
84 /* maximum length of Mac metadata in MiB */
85 #define ZIP_MAX_METADATA 10U
86
87 struct zip_entry {
88 struct archive_rb_node node;
89 struct zip_entry *next;
90 int64_t local_header_offset;
91 int64_t compressed_size;
92 int64_t uncompressed_size;
93 int64_t gid;
94 int64_t uid;
95 struct archive_string rsrcname;
96 time_t mtime;
97 time_t atime;
98 time_t ctime;
99 uint32_t crc32;
100 uint16_t mode;
101 uint16_t zip_flags; /* From GP Flags Field */
102 unsigned char compression;
103 unsigned char system; /* From "version written by" */
104 unsigned char flags; /* Our extra markers. */
105 unsigned char decdat;/* Used for Decryption check */
106
107 /* WinZip AES encryption extra field should be available
108 * when compression is 99. */
109 struct {
110 /* Vendor version: AE-1 - 0x0001, AE-2 - 0x0002 */
111 unsigned vendor;
112 #define AES_VENDOR_AE_1 0x0001
113 #define AES_VENDOR_AE_2 0x0002
114 /* AES encryption strength:
115 * 1 - 128 bits, 2 - 192 bits, 2 - 256 bits. */
116 unsigned strength;
117 /* Actual compression method. */
118 unsigned char compression;
119 } aes_extra;
120 };
121
122 struct trad_enc_ctx {
123 uint32_t keys[3];
124 };
125
126 /* Bits used in zip_flags. */
127 #define ZIP_ENCRYPTED (1 << 0)
128 #define ZIP_LENGTH_AT_END (1 << 3) /* Also called "Streaming bit" */
129 #define ZIP_STRONG_ENCRYPTED (1 << 6)
130 #define ZIP_UTF8_NAME (1 << 11)
131 /* See "7.2 Single Password Symmetric Encryption Method"
132 in http://www.pkware.com/documents/casestudies/APPNOTE.TXT */
133 #define ZIP_CENTRAL_DIRECTORY_ENCRYPTED (1 << 13)
134
135 /* Bits used in flags. */
136 #define LA_USED_ZIP64 (1 << 0)
137 #define LA_FROM_CENTRAL_DIRECTORY (1 << 1)
138
139 /*
140 * See "WinZip - AES Encryption Information"
141 * http://www.winzip.com/aes_info.htm
142 */
143 /* Value used in compression method. */
144 #define WINZIP_AES_ENCRYPTION 99
145 /* Authentication code size. */
146 #define AUTH_CODE_SIZE 10
147 /**/
148 #define MAX_DERIVED_KEY_BUF_SIZE (AES_MAX_KEY_SIZE * 2 + 2)
149
150 struct zip {
151 /* Structural information about the archive. */
152 struct archive_string format_name;
153 int64_t central_directory_offset;
154 int64_t central_directory_offset_adjusted;
155 int has_encrypted_entries;
156
157 /* List of entries (seekable Zip only) */
158 struct zip_entry *zip_entries;
159 struct archive_rb_tree tree;
160 struct archive_rb_tree tree_rsrc;
161
162 /* Bytes read but not yet consumed via __archive_read_consume() */
163 size_t unconsumed;
164
165 /* Information about entry we're currently reading. */
166 struct zip_entry *entry;
167 int64_t entry_bytes_remaining;
168
169 /* These count the number of bytes actually read for the entry. */
170 int64_t entry_compressed_bytes_read;
171 int64_t entry_uncompressed_bytes_read;
172
173 /* Running CRC32 of the decompressed and decrypted data */
174 unsigned long computed_crc32;
175 unsigned long (*crc32func)(unsigned long, const void *,
176 size_t);
177 char ignore_crc32;
178
179 /* Flags to mark progress of decompression. */
180 char decompress_init;
181 char end_of_entry;
182
183 unsigned char *uncompressed_buffer;
184 size_t uncompressed_buffer_size;
185
186 #ifdef HAVE_ZLIB_H
187 z_stream stream;
188 char stream_valid;
189 #endif
190
191 #if HAVE_LZMA_H && HAVE_LIBLZMA
192 lzma_stream zipx_lzma_stream;
193 char zipx_lzma_valid;
194 #endif
195
196 #ifdef HAVE_BZLIB_H
197 bz_stream bzstream;
198 char bzstream_valid;
199 #endif
200
201 #if HAVE_ZSTD_H && HAVE_LIBZSTD
202 ZSTD_DStream *zstdstream;
203 char zstdstream_valid;
204 #endif
205
206 IByteIn zipx_ppmd_stream;
207 ssize_t zipx_ppmd_read_compressed;
208 CPpmd8 ppmd8;
209 char ppmd8_valid;
210 char ppmd8_stream_failed;
211
212 struct archive_string_conv *sconv;
213 struct archive_string_conv *sconv_default;
214 struct archive_string_conv *sconv_utf8;
215 int init_default_conversion;
216 int process_mac_extensions;
217
218 char init_decryption;
219
220 /* Decryption buffer. */
221 /*
222 * The decrypted data starts at decrypted_ptr and
223 * extends for decrypted_bytes_remaining. Decryption
224 * adds new data to the end of this block, data is returned
225 * to clients from the beginning. When the block hits the
226 * end of decrypted_buffer, it has to be shuffled back to
227 * the beginning of the buffer.
228 */
229 unsigned char *decrypted_buffer;
230 unsigned char *decrypted_ptr;
231 size_t decrypted_buffer_size;
232 size_t decrypted_bytes_remaining;
233 size_t decrypted_unconsumed_bytes;
234
235 /* Traditional PKWARE decryption. */
236 struct trad_enc_ctx tctx;
237 char tctx_valid;
238
239 /* WinZip AES decryption. */
240 /* Contexts used for AES decryption. */
241 archive_crypto_ctx cctx;
242 char cctx_valid;
243 archive_hmac_sha1_ctx hctx;
244 char hctx_valid;
245
246 /* Strong encryption's decryption header information. */
247 unsigned iv_size;
248 unsigned alg_id;
249 unsigned bit_len;
250 unsigned flags;
251 unsigned erd_size;
252 unsigned v_size;
253 unsigned v_crc32;
254 uint8_t *iv;
255 uint8_t *erd;
256 uint8_t *v_data;
257 };
258
259 /* Many systems define min or MIN, but not all. */
260 #define zipmin(a,b) ((a) < (b) ? (a) : (b))
261
262 /* True if this entry has declared a specific uncompressed size. */
263 static int
zip_entry_size_is_set(const struct zip_entry * zip_entry)264 zip_entry_size_is_set(const struct zip_entry *zip_entry)
265 {
266 return (0 == (zip_entry->zip_flags & ZIP_LENGTH_AT_END)
267 || (zip_entry->uncompressed_size > 0
268 && zip_entry->uncompressed_size != 0xffffffff));
269 }
270
271 #ifdef HAVE_ZLIB_H
272 static int
273 zip_read_data_deflate(struct archive_read *a, const void **buff,
274 size_t *size, int64_t *offset);
275 #endif
276 #if HAVE_LZMA_H && HAVE_LIBLZMA
277 static int
278 zip_read_data_zipx_lzma_alone(struct archive_read *a, const void **buff,
279 size_t *size, int64_t *offset);
280 #endif
281
282 static void
283 trad_enc_decrypt_update(struct trad_enc_ctx *, const uint8_t *, size_t,
284 uint8_t *, size_t);
285
286 /* This function is used by Ppmd8_DecodeSymbol during decompression of Ppmd8
287 * streams inside ZIP files. It has 2 purposes: one is to fetch the next
288 * compressed byte from the stream, second one is to increase the counter how
289 * many compressed bytes were read. */
290 static Byte
ppmd_read(void * p)291 ppmd_read(void* p) {
292 /* Get the handle to current decompression context. */
293 struct archive_read *a = ((IByteIn*)p)->a;
294 struct zip *zip = a->format->data;
295
296 /* Fetch next byte. */
297 const uint8_t* data = __archive_read_ahead(a, 1, NULL);
298 if(data == NULL) {
299 zip->ppmd8_stream_failed = 1;
300 return 0;
301 }
302
303 if (zip->tctx_valid || zip->cctx_valid) {
304 uint8_t val;
305 if (zip->tctx_valid) {
306 trad_enc_decrypt_update(&zip->tctx,
307 data, 1, &val, 1);
308 } else {
309 size_t dsize = 1;
310 archive_decrypto_aes_ctr_update(&zip->cctx,
311 data, 1, &val, &dsize);
312 }
313 if (zip->hctx_valid)
314 archive_hmac_sha1_update(&zip->hctx, data, 1);
315
316 __archive_read_consume(a, 1);
317 ++zip->zipx_ppmd_read_compressed;
318 return val;
319 }
320
321 __archive_read_consume(a, 1);
322
323 /* Increment the counter. */
324 ++zip->zipx_ppmd_read_compressed;
325
326 /* Return the next compressed byte. */
327 return data[0];
328 }
329
330 /* ------------------------------------------------------------------------ */
331
332 /*
333 Traditional PKWARE Decryption functions.
334 */
335
336 static void
trad_enc_update_keys(struct trad_enc_ctx * ctx,uint8_t c)337 trad_enc_update_keys(struct trad_enc_ctx *ctx, uint8_t c)
338 {
339 uint8_t t;
340 #define CRC32(c, b) (crc32(c ^ 0xffffffffUL, &b, 1) ^ 0xffffffffUL)
341
342 ctx->keys[0] = CRC32(ctx->keys[0], c);
343 ctx->keys[1] = (ctx->keys[1] + (ctx->keys[0] & 0xff)) * 134775813L + 1;
344 t = (ctx->keys[1] >> 24) & 0xff;
345 ctx->keys[2] = CRC32(ctx->keys[2], t);
346 #undef CRC32
347 }
348
349 static uint8_t
trad_enc_decrypt_byte(struct trad_enc_ctx * ctx)350 trad_enc_decrypt_byte(struct trad_enc_ctx *ctx)
351 {
352 unsigned temp = ctx->keys[2] | 2;
353 return (uint8_t)((temp * (temp ^ 1)) >> 8) & 0xff;
354 }
355
356 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)357 trad_enc_decrypt_update(struct trad_enc_ctx *ctx, const uint8_t *in,
358 size_t in_len, uint8_t *out, size_t out_len)
359 {
360 unsigned i, max;
361
362 max = (unsigned)((in_len < out_len)? in_len: out_len);
363
364 for (i = 0; i < max; i++) {
365 uint8_t t = in[i] ^ trad_enc_decrypt_byte(ctx);
366 out[i] = t;
367 trad_enc_update_keys(ctx, t);
368 }
369 }
370
371 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)372 trad_enc_init(struct trad_enc_ctx *ctx, const char *pw, size_t pw_len,
373 const uint8_t *key, size_t key_len, uint8_t *crcchk)
374 {
375 uint8_t header[12];
376
377 if (key_len < 12) {
378 *crcchk = 0xff;
379 return -1;
380 }
381
382 ctx->keys[0] = 305419896L;
383 ctx->keys[1] = 591751049L;
384 ctx->keys[2] = 878082192L;
385
386 for (;pw_len; --pw_len)
387 trad_enc_update_keys(ctx, *pw++);
388
389 trad_enc_decrypt_update(ctx, key, 12, header, 12);
390 /* Return the last byte for CRC check. */
391 *crcchk = header[11];
392 return 0;
393 }
394
395 #if 0
396 static void
397 crypt_derive_key_sha1(const void *p, int size, unsigned char *key,
398 int key_size)
399 {
400 #define MD_SIZE 20
401 archive_sha1_ctx ctx;
402 unsigned char md1[MD_SIZE];
403 unsigned char md2[MD_SIZE * 2];
404 unsigned char mkb[64];
405 int i;
406
407 archive_sha1_init(&ctx);
408 archive_sha1_update(&ctx, p, size);
409 archive_sha1_final(&ctx, md1);
410
411 memset(mkb, 0x36, sizeof(mkb));
412 for (i = 0; i < MD_SIZE; i++)
413 mkb[i] ^= md1[i];
414 archive_sha1_init(&ctx);
415 archive_sha1_update(&ctx, mkb, sizeof(mkb));
416 archive_sha1_final(&ctx, md2);
417
418 memset(mkb, 0x5C, sizeof(mkb));
419 for (i = 0; i < MD_SIZE; i++)
420 mkb[i] ^= md1[i];
421 archive_sha1_init(&ctx);
422 archive_sha1_update(&ctx, mkb, sizeof(mkb));
423 archive_sha1_final(&ctx, md2 + MD_SIZE);
424
425 if (key_size > 32)
426 key_size = 32;
427 memcpy(key, md2, key_size);
428 #undef MD_SIZE
429 }
430 #endif
431
432 /* Read and decrypt bytes for zipx init headers.
433 * Used by format-specific init functions (lzma, ppmd) that need to
434 * read a small header from the compressed stream. When encryption is
435 * active the bytes are decrypted in-place into the decryption buffer. */
436 static int
zipx_read_header_and_decrypt(struct archive_read * a,const void ** buf,size_t in_len,size_t * out_len,size_t * consumed)437 zipx_read_header_and_decrypt(struct archive_read *a, const void **buf, size_t in_len,
438 size_t *out_len, size_t *consumed)
439 {
440 struct zip *zip = a->format->data;
441 const void *raw;
442 ssize_t bytes_avail;
443 size_t to_decrypt;
444
445 raw = __archive_read_ahead(a, in_len, &bytes_avail);
446 if (raw == NULL || bytes_avail < (ssize_t)in_len) {
447 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
448 "Truncated ZIP file data");
449 return (ARCHIVE_FATAL);
450 }
451
452 if (zip->tctx_valid || zip->cctx_valid) {
453 to_decrypt = in_len;
454 if (to_decrypt > zip->decrypted_buffer_size)
455 to_decrypt = zip->decrypted_buffer_size;
456
457 if (zip->tctx_valid) {
458 trad_enc_decrypt_update(&zip->tctx,
459 raw, to_decrypt,
460 zip->decrypted_buffer, to_decrypt);
461 } else {
462 size_t dsize = to_decrypt;
463 archive_decrypto_aes_ctr_update(&zip->cctx,
464 raw, to_decrypt,
465 zip->decrypted_buffer, &dsize);
466 }
467 if (zip->hctx_valid)
468 archive_hmac_sha1_update(&zip->hctx,
469 raw, to_decrypt);
470
471 *buf = zip->decrypted_buffer;
472 *out_len = to_decrypt;
473 *consumed = to_decrypt;
474 } else {
475 *buf = raw;
476 *out_len = in_len;
477 *consumed = in_len;
478 }
479 return (ARCHIVE_OK);
480 }
481
482 /* Decrypt bulk compressed data for zipx decompression.
483 * Manages the decryption buffer, handles partial fills, and returns decrypted
484 * data pointer + length. `sp` is set to the raw pointer for HMAC accounting. */
485 static void
zip_read_decrypt(struct zip * zip,const void * compressed_buff,ssize_t bytes_avail,const void ** result_buff,ssize_t * result_avail,const void ** sp)486 zip_read_decrypt(struct zip *zip, const void *compressed_buff,
487 ssize_t bytes_avail, const void **result_buff, ssize_t *result_avail,
488 const void **sp)
489 {
490 *sp = compressed_buff;
491
492 /* Safety check to prevent potential OOB reads if something went wrong
493 * previously. We should not have a negative bytes_avail count here.
494 * If we do, set them to zero so that reading the ZIP will fail later,
495 * safely as corrupted instead of crashing. */
496 if (bytes_avail < 0)
497 bytes_avail = 0;
498
499 if (zip->tctx_valid || zip->cctx_valid) {
500 if (zip->decrypted_bytes_remaining < (size_t)bytes_avail) {
501 size_t buff_remaining =
502 (zip->decrypted_buffer +
503 zip->decrypted_buffer_size)
504 - (zip->decrypted_ptr +
505 zip->decrypted_bytes_remaining);
506 /* The new bytes to decrypt start after decrypted_bytes_remaining
507 * in the raw stream: those leading bytes were already
508 * decrypted on a previous call but have not yet been consumed. */
509 size_t new_bytes =
510 (size_t)bytes_avail
511 - zip->decrypted_bytes_remaining;
512
513 if (buff_remaining > new_bytes)
514 buff_remaining = new_bytes;
515
516 if (0 == (zip->entry->zip_flags & ZIP_LENGTH_AT_END) &&
517 zip->entry_bytes_remaining > 0) {
518 if ((int64_t)(zip->decrypted_bytes_remaining
519 + buff_remaining)
520 > zip->entry_bytes_remaining) {
521 if (zip->entry_bytes_remaining <
522 (int64_t)zip->decrypted_bytes_remaining)
523 buff_remaining = 0;
524 else
525 buff_remaining =
526 (size_t)zip->entry_bytes_remaining
527 - zip->decrypted_bytes_remaining;
528 }
529 }
530 if (buff_remaining > 0) {
531 if (zip->tctx_valid) {
532 trad_enc_decrypt_update(&zip->tctx,
533 (const uint8_t *)compressed_buff
534 + zip->decrypted_bytes_remaining,
535 buff_remaining,
536 zip->decrypted_ptr
537 + zip->decrypted_bytes_remaining,
538 buff_remaining);
539 } else {
540 size_t dsize = buff_remaining;
541 archive_decrypto_aes_ctr_update(
542 &zip->cctx,
543 (const uint8_t *)compressed_buff
544 + zip->decrypted_bytes_remaining,
545 buff_remaining,
546 zip->decrypted_ptr
547 + zip->decrypted_bytes_remaining,
548 &dsize);
549 }
550 zip->decrypted_bytes_remaining +=
551 buff_remaining;
552 }
553 }
554 *result_avail = zip->decrypted_bytes_remaining;
555 *result_buff = (const char *)zip->decrypted_ptr;
556 } else {
557 *result_buff = compressed_buff;
558 *result_avail = bytes_avail;
559 }
560 }
561
562 /* Post-decompression decrypt state update.
563 * Updates decrypt buffer pointers and HMAC after the decompressor
564 * has consumed `to_consume` bytes. */
565 static void
zip_read_decrypt_update(struct zip * zip,ssize_t to_consume,const void * sp)566 zip_read_decrypt_update(struct zip *zip, ssize_t to_consume, const void *sp)
567 {
568 if (zip->tctx_valid || zip->cctx_valid) {
569 zip->decrypted_bytes_remaining -= to_consume;
570 if (zip->decrypted_bytes_remaining == 0)
571 zip->decrypted_ptr = zip->decrypted_buffer;
572 else
573 zip->decrypted_ptr += to_consume;
574 }
575 if (zip->hctx_valid)
576 archive_hmac_sha1_update(&zip->hctx, sp, to_consume);
577 }
578
579 /*
580 * Common code for streaming or seeking modes.
581 *
582 * Includes code to read local file headers, decompress data
583 * from entry bodies, and common API.
584 */
585
586 static unsigned long
real_crc32(unsigned long crc,const void * buff,size_t len)587 real_crc32(unsigned long crc, const void *buff, size_t len)
588 {
589 return crc32(crc, buff, (unsigned int)len);
590 }
591
592 /* Used by "ignorecrc32" option to speed up tests. */
593 static unsigned long
fake_crc32(unsigned long crc,const void * buff,size_t len)594 fake_crc32(unsigned long crc, const void *buff, size_t len)
595 {
596 (void)crc; /* UNUSED */
597 (void)buff; /* UNUSED */
598 (void)len; /* UNUSED */
599 return 0;
600 }
601
602 static const struct {
603 int id;
604 const char * name;
605 } compression_methods[] = {
606 {0, "uncompressed"}, /* The file is stored (no compression) */
607 {1, "shrinking"}, /* The file is Shrunk */
608 {2, "reduced-1"}, /* The file is Reduced with compression factor 1 */
609 {3, "reduced-2"}, /* The file is Reduced with compression factor 2 */
610 {4, "reduced-3"}, /* The file is Reduced with compression factor 3 */
611 {5, "reduced-4"}, /* The file is Reduced with compression factor 4 */
612 {6, "imploded"}, /* The file is Imploded */
613 {7, "reserved"}, /* Reserved for Tokenizing compression algorithm */
614 {8, "deflation"}, /* The file is Deflated */
615 {9, "deflation-64-bit"}, /* Enhanced Deflating using Deflate64(tm) */
616 {10, "ibm-terse"},/* PKWARE Data Compression Library Imploding
617 * (old IBM TERSE) */
618 {11, "reserved"}, /* Reserved by PKWARE */
619 {12, "bzip"}, /* File is compressed using BZIP2 algorithm */
620 {13, "reserved"}, /* Reserved by PKWARE */
621 {14, "lzma"}, /* LZMA (EFS) */
622 {15, "reserved"}, /* Reserved by PKWARE */
623 {16, "reserved"}, /* Reserved by PKWARE */
624 {17, "reserved"}, /* Reserved by PKWARE */
625 {18, "ibm-terse-new"}, /* File is compressed using IBM TERSE (new) */
626 {19, "ibm-lz777"},/* IBM LZ77 z Architecture (PFS) */
627 {93, "zstd"}, /* Zstandard (zstd) Compression */
628 {95, "xz"}, /* XZ compressed data */
629 {96, "jpeg"}, /* JPEG compressed data */
630 {97, "wav-pack"}, /* WavPack compressed data */
631 {98, "ppmd-1"}, /* PPMd version I, Rev 1 */
632 {99, "aes"} /* WinZip AES encryption */
633 };
634
635 static const char *
compression_name(const int compression)636 compression_name(const int compression)
637 {
638 static const int num_compression_methods =
639 sizeof(compression_methods)/sizeof(compression_methods[0]);
640 int i=0;
641
642 while(compression >= 0 && i < num_compression_methods) {
643 if (compression_methods[i].id == compression)
644 return compression_methods[i].name;
645 i++;
646 }
647 return "??";
648 }
649
650 /*
651 * The extra data is stored as a list of
652 * id1+size1+data1 + id2+size2+data2 ...
653 * triplets. id and size are 2 bytes each.
654 */
655 static int
process_extra(struct archive_read * a,struct archive_entry * entry,const char * p,size_t extra_length,struct zip_entry * zip_entry)656 process_extra(struct archive_read *a, struct archive_entry *entry,
657 const char *p, size_t extra_length, struct zip_entry* zip_entry)
658 {
659 struct zip *zip = a->format->data;
660 unsigned offset = 0;
661
662 if (extra_length == 0) {
663 return ARCHIVE_OK;
664 }
665
666 if (extra_length < 4) {
667 size_t i = 0;
668 /* Some ZIP files may have trailing 0 bytes. Let's check they
669 * are all 0 and ignore them instead of returning an error.
670 *
671 * This is not technically correct, but some ZIP files look
672 * like this and other tools support those files - so let's
673 * also support them.
674 */
675 for (; i < extra_length; i++) {
676 if (p[i] != 0) {
677 archive_set_error(&a->archive,
678 ARCHIVE_ERRNO_FILE_FORMAT,
679 "Too-small extra data: "
680 "Need at least 4 bytes, "
681 "but only found %d bytes",
682 (int)extra_length);
683 return ARCHIVE_FAILED;
684 }
685 }
686
687 return ARCHIVE_OK;
688 }
689
690 while (offset <= extra_length - 4) {
691 unsigned short headerid = archive_le16dec(p + offset);
692 unsigned short datasize = archive_le16dec(p + offset + 2);
693
694 offset += 4;
695 if (offset + datasize > extra_length) {
696 archive_set_error(&a->archive,
697 ARCHIVE_ERRNO_FILE_FORMAT, "Extra data overflow: "
698 "Need %d bytes but only found %d bytes",
699 (int)datasize, (int)(extra_length - offset));
700 return ARCHIVE_FAILED;
701 }
702 #ifdef DEBUG
703 fprintf(stderr, "Header id 0x%04x, length %d\n",
704 headerid, datasize);
705 #endif
706 switch (headerid) {
707 case 0x0001:
708 /* Zip64 extended information extra field. */
709 zip_entry->flags |= LA_USED_ZIP64;
710 if (zip_entry->uncompressed_size == 0xffffffff) {
711 uint64_t t = 0;
712 if (datasize < 8
713 || (t = archive_le64dec(p + offset)) >
714 INT64_MAX) {
715 archive_set_error(&a->archive,
716 ARCHIVE_ERRNO_FILE_FORMAT,
717 "Malformed 64-bit "
718 "uncompressed size");
719 return ARCHIVE_FAILED;
720 }
721 zip_entry->uncompressed_size = t;
722 offset += 8;
723 datasize -= 8;
724 }
725 if (zip_entry->compressed_size == 0xffffffff) {
726 uint64_t t = 0;
727 if (datasize < 8
728 || (t = archive_le64dec(p + offset)) >
729 INT64_MAX) {
730 archive_set_error(&a->archive,
731 ARCHIVE_ERRNO_FILE_FORMAT,
732 "Malformed 64-bit "
733 "compressed size");
734 return ARCHIVE_FAILED;
735 }
736 zip_entry->compressed_size = t;
737 offset += 8;
738 datasize -= 8;
739 }
740 if (zip_entry->local_header_offset == 0xffffffff) {
741 uint64_t t = 0;
742 if (datasize < 8
743 || (t = archive_le64dec(p + offset)) >
744 INT64_MAX) {
745 archive_set_error(&a->archive,
746 ARCHIVE_ERRNO_FILE_FORMAT,
747 "Malformed 64-bit "
748 "local header offset");
749 return ARCHIVE_FAILED;
750 }
751 zip_entry->local_header_offset = t;
752 offset += 8;
753 datasize -= 8;
754 }
755 /* archive_le32dec(p + offset) gives disk
756 * on which file starts, but we don't handle
757 * multi-volume Zip files. */
758 break;
759 case 0x000d:
760 /* PKWARE Unix Extra Field fixed metadata. */
761 if (datasize >= 12) {
762 zip_entry->atime = archive_le32dec(p + offset);
763 zip_entry->mtime =
764 archive_le32dec(p + offset + 4);
765 zip_entry->uid =
766 archive_le16dec(p + offset + 8);
767 zip_entry->gid =
768 archive_le16dec(p + offset + 10);
769 /*
770 * APPNOTE.TXT also defines additional data after
771 * this fixed metadata, depending on file type.
772 */
773 }
774 break;
775 #ifdef DEBUG
776 case 0x0017:
777 {
778 /* Strong encryption field. */
779 if (archive_le16dec(p + offset) == 2) {
780 unsigned algId =
781 archive_le16dec(p + offset + 2);
782 unsigned bitLen =
783 archive_le16dec(p + offset + 4);
784 int flags =
785 archive_le16dec(p + offset + 6);
786 fprintf(stderr, "algId=0x%04x, bitLen=%u, "
787 "flgas=%d\n", algId, bitLen,flags);
788 }
789 break;
790 }
791 #endif
792 case 0x5455:
793 {
794 /* Extended time field "UT". */
795 int flags;
796 if (datasize == 0) {
797 archive_set_error(&a->archive,
798 ARCHIVE_ERRNO_FILE_FORMAT,
799 "Incomplete extended time field");
800 return ARCHIVE_FAILED;
801 }
802 flags = p[offset];
803 offset++;
804 datasize--;
805 /* Flag bits indicate which dates are present. */
806 if (flags & 0x01)
807 {
808 #ifdef DEBUG
809 fprintf(stderr, "mtime: %lld -> %d\n",
810 (long long)zip_entry->mtime,
811 archive_le32dec(p + offset));
812 #endif
813 if (datasize < 4)
814 break;
815 zip_entry->mtime = archive_le32dec(p + offset);
816 offset += 4;
817 datasize -= 4;
818 }
819 if (flags & 0x02)
820 {
821 if (datasize < 4)
822 break;
823 zip_entry->atime = archive_le32dec(p + offset);
824 offset += 4;
825 datasize -= 4;
826 }
827 if (flags & 0x04)
828 {
829 if (datasize < 4)
830 break;
831 zip_entry->ctime = archive_le32dec(p + offset);
832 offset += 4;
833 datasize -= 4;
834 }
835 break;
836 }
837 case 0x5855:
838 {
839 /* Info-ZIP Unix Extra Field (old version) "UX". */
840 if (datasize >= 8) {
841 zip_entry->atime = archive_le32dec(p + offset);
842 zip_entry->mtime =
843 archive_le32dec(p + offset + 4);
844 }
845 if (datasize >= 12) {
846 zip_entry->uid =
847 archive_le16dec(p + offset + 8);
848 zip_entry->gid =
849 archive_le16dec(p + offset + 10);
850 }
851 break;
852 }
853 case 0x6c78:
854 {
855 /* Experimental 'xl' field */
856 /*
857 * Introduced Dec 2013 to provide a way to
858 * include external file attributes (and other
859 * fields that ordinarily appear only in
860 * central directory) in local file header.
861 * This provides file type and permission
862 * information necessary to support full
863 * streaming extraction. Currently being
864 * discussed with other Zip developers
865 * ... subject to change.
866 *
867 * Format:
868 * The field starts with a bitmap that specifies
869 * which additional fields are included. The
870 * bitmap is variable length and can be extended in
871 * the future.
872 *
873 * n bytes - feature bitmap: first byte has low-order
874 * 7 bits. If high-order bit is set, a subsequent
875 * byte holds the next 7 bits, etc.
876 *
877 * if bitmap & 1, 2 byte "version made by"
878 * if bitmap & 2, 2 byte "internal file attributes"
879 * if bitmap & 4, 4 byte "external file attributes"
880 * if bitmap & 8, 2 byte comment length + n byte
881 * comment
882 */
883 int bitmap, bitmap_last;
884
885 if (datasize < 1)
886 break;
887 bitmap_last = bitmap = 0xff & p[offset];
888 offset += 1;
889 datasize -= 1;
890
891 /* We only support first 7 bits of bitmap; skip rest. */
892 while ((bitmap_last & 0x80) != 0
893 && datasize >= 1) {
894 bitmap_last = p[offset];
895 offset += 1;
896 datasize -= 1;
897 }
898
899 if (bitmap & 1) {
900 /* 2 byte "version made by" */
901 if (datasize < 2)
902 break;
903 zip_entry->system
904 = archive_le16dec(p + offset) >> 8;
905 offset += 2;
906 datasize -= 2;
907 }
908 if (bitmap & 2) {
909 /* 2 byte "internal file attributes" */
910 uint32_t internal_attributes;
911 if (datasize < 2)
912 break;
913 internal_attributes
914 = archive_le16dec(p + offset);
915 /* Not used by libarchive at present. */
916 (void)internal_attributes; /* UNUSED */
917 offset += 2;
918 datasize -= 2;
919 }
920 if (bitmap & 4) {
921 /* 4 byte "external file attributes" */
922 uint32_t external_attributes;
923 if (datasize < 4)
924 break;
925 external_attributes
926 = archive_le32dec(p + offset);
927 if (zip_entry->system == 3) {
928 zip_entry->mode
929 = external_attributes >> 16;
930 } else if (zip_entry->system == 0) {
931 // Interpret MSDOS directory bit
932 if (0x10 == (external_attributes &
933 0x10)) {
934 zip_entry->mode =
935 AE_IFDIR | 0775;
936 } else {
937 zip_entry->mode =
938 AE_IFREG | 0664;
939 }
940 if (0x01 == (external_attributes &
941 0x01)) {
942 /* Read-only bit;
943 * strip write permissions */
944 zip_entry->mode &= 0555;
945 }
946 } else {
947 zip_entry->mode = 0;
948 }
949 offset += 4;
950 datasize -= 4;
951 }
952 if (bitmap & 8) {
953 /* 2 byte comment length + comment */
954 uint32_t comment_length;
955 if (datasize < 2)
956 break;
957 comment_length
958 = archive_le16dec(p + offset);
959 offset += 2;
960 datasize -= 2;
961
962 if (datasize < comment_length)
963 break;
964 /* Comment is not supported by libarchive */
965 offset += comment_length;
966 datasize -= comment_length;
967 }
968 break;
969 }
970 case 0x7075:
971 {
972 /* Info-ZIP Unicode Path Extra Field. */
973 if (datasize < 5 || entry == NULL)
974 break;
975 offset += 5;
976 datasize -= 5;
977
978 /* The path name in this field is always encoded
979 * in UTF-8. */
980 if (zip->sconv_utf8 == NULL) {
981 zip->sconv_utf8 =
982 archive_string_conversion_from_charset(
983 &a->archive, "UTF-8", 1);
984 /* If the converter from UTF-8 is not
985 * available, then the path name from the main
986 * field will more likely be correct. */
987 if (zip->sconv_utf8 == NULL)
988 break;
989 }
990
991 /* Make sure the CRC32 of the filename matches. */
992 if (!zip->ignore_crc32) {
993 const char *cp = archive_entry_pathname(entry);
994 if (cp) {
995 unsigned long file_crc =
996 zip->crc32func(0, cp, strlen(cp));
997 unsigned long utf_crc =
998 archive_le32dec(p + offset - 4);
999 if (file_crc != utf_crc) {
1000 #ifdef DEBUG
1001 fprintf(stderr,
1002 "CRC filename mismatch; "
1003 "CDE is %lx, but UTF8 "
1004 "is outdated with %lx\n",
1005 file_crc, utf_crc);
1006 #endif
1007 break;
1008 }
1009 }
1010 }
1011
1012 if (archive_entry_copy_pathname_l(entry,
1013 p + offset, datasize, zip->sconv_utf8) != 0) {
1014 /* Ignore the error, and fallback to the path
1015 * name from the main field. */
1016 #ifdef DEBUG
1017 fprintf(stderr, "Failed to read the ZIP "
1018 "0x7075 extra field path.\n");
1019 #endif
1020 }
1021 break;
1022 }
1023 case 0x7855:
1024 /* Info-ZIP Unix Extra Field (type 2) "Ux". */
1025 #ifdef DEBUG
1026 fprintf(stderr, "uid %d gid %d\n",
1027 archive_le16dec(p + offset),
1028 archive_le16dec(p + offset + 2));
1029 #endif
1030 if (datasize >= 2)
1031 zip_entry->uid = archive_le16dec(p + offset);
1032 if (datasize >= 4)
1033 zip_entry->gid =
1034 archive_le16dec(p + offset + 2);
1035 break;
1036 case 0x7875:
1037 {
1038 /* Info-Zip Unix Extra Field (type 3) "ux". */
1039 int uidsize = 0, gidsize = 0;
1040
1041 /* TODO: support arbitrary uidsize/gidsize. */
1042 if (datasize >= 1 && p[offset] == 1) {/* version=1 */
1043 if (datasize >= 4) {
1044 /* get a uid size. */
1045 uidsize = 0xff & (int)p[offset+1];
1046 if (uidsize == 2)
1047 zip_entry->uid =
1048 archive_le16dec(
1049 p + offset + 2);
1050 else if (uidsize == 4 && datasize >= 6)
1051 zip_entry->uid =
1052 archive_le32dec(
1053 p + offset + 2);
1054 }
1055 if (datasize >= (2 + uidsize + 3)) {
1056 /* get a gid size. */
1057 gidsize = 0xff &
1058 (int)p[offset+2+uidsize];
1059 if (gidsize == 2)
1060 zip_entry->gid =
1061 archive_le16dec(
1062 p+offset+2+uidsize+1);
1063 else if (gidsize == 4 &&
1064 datasize >= (2 + uidsize + 5))
1065 zip_entry->gid =
1066 archive_le32dec(
1067 p+offset+2+uidsize+1);
1068 }
1069 }
1070 break;
1071 }
1072 case 0x9901:
1073 /* WinZip AES extra data field. */
1074 if (datasize < 6) {
1075 archive_set_error(&a->archive,
1076 ARCHIVE_ERRNO_FILE_FORMAT,
1077 "Incomplete AES field");
1078 return ARCHIVE_FAILED;
1079 }
1080 if (p[offset + 2] == 'A' && p[offset + 3] == 'E') {
1081 /* Vendor version. */
1082 zip_entry->aes_extra.vendor =
1083 archive_le16dec(p + offset);
1084 /* AES encryption strength. */
1085 zip_entry->aes_extra.strength = p[offset + 4];
1086 /* Actual compression method. */
1087 zip_entry->aes_extra.compression =
1088 p[offset + 5];
1089 }
1090 break;
1091 default:
1092 break;
1093 }
1094 offset += datasize;
1095 }
1096 return ARCHIVE_OK;
1097 }
1098
1099 /*
1100 * Assumes file pointer is at beginning of local file header.
1101 */
1102 static int
zip_read_local_file_header(struct archive_read * a,struct archive_entry * entry,struct zip * zip)1103 zip_read_local_file_header(struct archive_read *a, struct archive_entry *entry,
1104 struct zip *zip)
1105 {
1106 const char *p;
1107 const void *h;
1108 const wchar_t *wp;
1109 const char *cp;
1110 size_t len, filename_length, extra_length;
1111 struct archive_string_conv *sconv;
1112 struct zip_entry *zip_entry = zip->entry;
1113 struct zip_entry zip_entry_central_dir;
1114 int ret = ARCHIVE_OK;
1115 char version;
1116
1117 /* Save a copy of the original for consistency checks. */
1118 zip_entry_central_dir = *zip_entry;
1119
1120 zip->decompress_init = 0;
1121 zip->end_of_entry = 0;
1122 zip->entry_uncompressed_bytes_read = 0;
1123 zip->entry_compressed_bytes_read = 0;
1124 zip->computed_crc32 = zip->crc32func(0, NULL, 0);
1125
1126 /* Setup default conversion. */
1127 if (zip->sconv == NULL && !zip->init_default_conversion) {
1128 zip->sconv_default =
1129 archive_string_default_conversion_for_read(&(a->archive));
1130 zip->init_default_conversion = 1;
1131 }
1132
1133 if ((p = __archive_read_ahead(a, ZIP_LOCHDR_LEN, NULL)) == NULL) {
1134 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1135 "Truncated ZIP file header");
1136 return (ARCHIVE_FATAL);
1137 }
1138
1139 if (memcmp(p, "PK\003\004", 4) != 0) {
1140 archive_set_error(&a->archive, -1, "Damaged Zip archive");
1141 return ARCHIVE_FATAL;
1142 }
1143 version = p[4];
1144 zip_entry->system = p[5];
1145 zip_entry->zip_flags = archive_le16dec(p + 6);
1146 if (zip_entry->zip_flags & (ZIP_ENCRYPTED | ZIP_STRONG_ENCRYPTED)) {
1147 zip->has_encrypted_entries = 1;
1148 archive_entry_set_is_data_encrypted(entry, 1);
1149 if (zip_entry->zip_flags & ZIP_CENTRAL_DIRECTORY_ENCRYPTED &&
1150 zip_entry->zip_flags & ZIP_ENCRYPTED &&
1151 zip_entry->zip_flags & ZIP_STRONG_ENCRYPTED) {
1152 archive_entry_set_is_metadata_encrypted(entry, 1);
1153 return ARCHIVE_FATAL;
1154 }
1155 }
1156 zip->init_decryption = (zip_entry->zip_flags & ZIP_ENCRYPTED);
1157 zip_entry->compression = (char)archive_le16dec(p + 8);
1158 zip_entry->mtime = dos_to_unix(archive_le32dec(p + 10));
1159 zip_entry->crc32 = archive_le32dec(p + 14);
1160 if (zip_entry->zip_flags & ZIP_LENGTH_AT_END)
1161 zip_entry->decdat = p[11];
1162 else
1163 zip_entry->decdat = p[17];
1164 zip_entry->compressed_size = archive_le32dec(p + 18);
1165 zip_entry->uncompressed_size = archive_le32dec(p + 22);
1166 filename_length = archive_le16dec(p + 26);
1167 extra_length = archive_le16dec(p + 28);
1168
1169 __archive_read_consume(a, ZIP_LOCHDR_LEN);
1170
1171 /* Read the filename. */
1172 if ((h = __archive_read_ahead(a, filename_length, NULL)) == NULL) {
1173 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1174 "Truncated ZIP file header");
1175 return (ARCHIVE_FATAL);
1176 }
1177 if (zip_entry->zip_flags & ZIP_UTF8_NAME) {
1178 /* The filename is stored to be UTF-8. */
1179 if (zip->sconv_utf8 == NULL) {
1180 zip->sconv_utf8 =
1181 archive_string_conversion_from_charset(
1182 &a->archive, "UTF-8", 1);
1183 if (zip->sconv_utf8 == NULL)
1184 return (ARCHIVE_FATAL);
1185 }
1186 sconv = zip->sconv_utf8;
1187 } else if (zip->sconv != NULL)
1188 sconv = zip->sconv;
1189 else
1190 sconv = zip->sconv_default;
1191
1192 if (archive_entry_copy_pathname_l(entry,
1193 h, filename_length, sconv) != 0) {
1194 if (errno == ENOMEM) {
1195 archive_set_error(&a->archive, ENOMEM,
1196 "Can't allocate memory for Pathname");
1197 return (ARCHIVE_FATAL);
1198 }
1199 archive_set_error(&a->archive,
1200 ARCHIVE_ERRNO_FILE_FORMAT,
1201 "Pathname cannot be converted "
1202 "from %s to current locale",
1203 archive_string_conversion_charset_name(sconv));
1204 ret = ARCHIVE_WARN;
1205 }
1206 __archive_read_consume(a, filename_length);
1207
1208 /* Read the extra data. */
1209 if ((h = __archive_read_ahead(a, extra_length, NULL)) == NULL) {
1210 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1211 "Truncated ZIP file header");
1212 return (ARCHIVE_FATAL);
1213 }
1214
1215 if (ARCHIVE_OK != process_extra(a, entry, h, extra_length,
1216 zip_entry)) {
1217 return ARCHIVE_FATAL;
1218 }
1219 __archive_read_consume(a, extra_length);
1220
1221 /* Work around a bug in Info-Zip: When reading from a pipe, it
1222 * stats the pipe instead of synthesizing a file entry. */
1223 if ((zip_entry->mode & AE_IFMT) == AE_IFIFO) {
1224 zip_entry->mode &= ~ AE_IFMT;
1225 zip_entry->mode |= AE_IFREG;
1226 }
1227
1228 /* If the mode is totally empty, set some sane default. */
1229 if (zip_entry->mode == 0) {
1230 zip_entry->mode |= 0664;
1231 }
1232
1233 /* Windows archivers sometimes use backslash as the directory
1234 * separator. Normalize to slash. */
1235 if (zip_entry->system == 0 &&
1236 (wp = archive_entry_pathname_w(entry)) != NULL) {
1237 if (wcschr(wp, L'/') == NULL && wcschr(wp, L'\\') != NULL) {
1238 size_t i;
1239 struct archive_wstring s;
1240 archive_string_init(&s);
1241 archive_wstrcpy(&s, wp);
1242 for (i = 0; i < archive_strlen(&s); i++) {
1243 if (s.s[i] == '\\')
1244 s.s[i] = '/';
1245 }
1246 archive_entry_copy_pathname_w(entry, s.s);
1247 archive_wstring_free(&s);
1248 }
1249 }
1250
1251 /* Make sure that entries with a trailing '/' are marked as directories
1252 * even if the External File Attributes contains bogus values. If this
1253 * is not a directory and there is no type, assume a regular file. */
1254 if ((zip_entry->mode & AE_IFMT) != AE_IFDIR) {
1255 int has_slash;
1256
1257 wp = archive_entry_pathname_w(entry);
1258 if (wp != NULL) {
1259 len = wcslen(wp);
1260 has_slash = len > 0 && wp[len - 1] == L'/';
1261 } else {
1262 cp = archive_entry_pathname(entry);
1263 len = (cp != NULL)?strlen(cp):0;
1264 has_slash = len > 0 && cp[len - 1] == '/';
1265 }
1266 /* Correct file type as needed. */
1267 if (has_slash) {
1268 zip_entry->mode &= ~AE_IFMT;
1269 zip_entry->mode |= AE_IFDIR;
1270 zip_entry->mode |= 0111;
1271 } else if ((zip_entry->mode & AE_IFMT) == 0) {
1272 zip_entry->mode |= AE_IFREG;
1273 }
1274 }
1275
1276 /* Make sure directories end in '/' */
1277 if ((zip_entry->mode & AE_IFMT) == AE_IFDIR) {
1278 wp = archive_entry_pathname_w(entry);
1279 if (wp != NULL) {
1280 len = wcslen(wp);
1281 if (len > 0 && wp[len - 1] != L'/') {
1282 struct archive_wstring s;
1283 archive_string_init(&s);
1284 archive_wstrcat(&s, wp);
1285 archive_wstrappend_wchar(&s, L'/');
1286 archive_entry_copy_pathname_w(entry, s.s);
1287 archive_wstring_free(&s);
1288 }
1289 } else {
1290 cp = archive_entry_pathname(entry);
1291 len = (cp != NULL)?strlen(cp):0;
1292 if (len > 0 && cp[len - 1] != '/') {
1293 struct archive_string s;
1294 archive_string_init(&s);
1295 archive_strcat(&s, cp);
1296 archive_strappend_char(&s, '/');
1297 archive_entry_set_pathname(entry, s.s);
1298 archive_string_free(&s);
1299 }
1300 }
1301 }
1302
1303 if (zip_entry->flags & LA_FROM_CENTRAL_DIRECTORY) {
1304 /* If this came from the central dir, its size info
1305 * is definitive, so ignore the length-at-end flag. */
1306 zip_entry->zip_flags &= ~ZIP_LENGTH_AT_END;
1307 /* If local header is missing a value, use the one from
1308 the central directory. If both have it, warn about
1309 mismatches. */
1310 if (zip_entry->crc32 == 0) {
1311 zip_entry->crc32 = zip_entry_central_dir.crc32;
1312 } else if (!zip->ignore_crc32
1313 && zip_entry->crc32 != zip_entry_central_dir.crc32) {
1314 archive_set_error(&a->archive,
1315 ARCHIVE_ERRNO_FILE_FORMAT,
1316 "Inconsistent CRC32 values");
1317 ret = ARCHIVE_WARN;
1318 }
1319 if (zip_entry->compressed_size == 0
1320 || zip_entry->compressed_size == 0xffffffff) {
1321 zip_entry->compressed_size
1322 = zip_entry_central_dir.compressed_size;
1323 } else if (zip_entry->compressed_size
1324 != zip_entry_central_dir.compressed_size) {
1325 archive_set_error(&a->archive,
1326 ARCHIVE_ERRNO_FILE_FORMAT,
1327 "Inconsistent compressed size: "
1328 "%jd in central directory, %jd in local header",
1329 (intmax_t)zip_entry_central_dir.compressed_size,
1330 (intmax_t)zip_entry->compressed_size);
1331 ret = ARCHIVE_WARN;
1332 }
1333 if (zip_entry->uncompressed_size == 0 ||
1334 zip_entry->uncompressed_size == 0xffffffff) {
1335 zip_entry->uncompressed_size
1336 = zip_entry_central_dir.uncompressed_size;
1337 } else if (zip_entry->uncompressed_size
1338 != zip_entry_central_dir.uncompressed_size) {
1339 archive_set_error(&a->archive,
1340 ARCHIVE_ERRNO_FILE_FORMAT,
1341 "Inconsistent uncompressed size: "
1342 "%jd in central directory, %jd in local header",
1343 (intmax_t)zip_entry_central_dir.uncompressed_size,
1344 (intmax_t)zip_entry->uncompressed_size);
1345 ret = ARCHIVE_WARN;
1346 }
1347 }
1348
1349 /* Populate some additional entry fields: */
1350 archive_entry_set_mode(entry, zip_entry->mode);
1351 archive_entry_set_uid(entry, zip_entry->uid);
1352 archive_entry_set_gid(entry, zip_entry->gid);
1353 archive_entry_set_mtime(entry, zip_entry->mtime, 0);
1354 archive_entry_set_ctime(entry, zip_entry->ctime, 0);
1355 archive_entry_set_atime(entry, zip_entry->atime, 0);
1356
1357 if ((zip->entry->mode & AE_IFMT) == AE_IFLNK) {
1358 size_t linkname_length;
1359
1360 if (zip_entry->compressed_size > 64 * 1024) {
1361 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1362 "Zip file with oversized link entry");
1363 return ARCHIVE_FATAL;
1364 }
1365
1366 linkname_length = (size_t)zip_entry->compressed_size;
1367
1368 archive_entry_set_size(entry, 0);
1369
1370 // take into account link compression if any
1371 size_t linkname_full_length = linkname_length;
1372 if (zip->entry->compression != 0)
1373 {
1374 // symlink target string appeared to be compressed
1375 int status = ARCHIVE_FATAL;
1376 const void *uncompressed_buffer = NULL;
1377
1378 switch (zip->entry->compression)
1379 {
1380 #if HAVE_ZLIB_H
1381 case 8: /* Deflate compression. */
1382 zip->entry_bytes_remaining = zip_entry->compressed_size;
1383 status = zip_read_data_deflate(a, &uncompressed_buffer,
1384 &linkname_full_length, NULL);
1385 break;
1386 #endif
1387 #if HAVE_LZMA_H && HAVE_LIBLZMA
1388 case 14: /* ZIPx LZMA compression. */
1389 /*(see zip file format specification, section 4.4.5)*/
1390 zip->entry_bytes_remaining = zip_entry->compressed_size;
1391 status = zip_read_data_zipx_lzma_alone(a, &uncompressed_buffer,
1392 &linkname_full_length, NULL);
1393 break;
1394 #endif
1395 default: /* Unsupported compression. */
1396 break;
1397 }
1398 if (status == ARCHIVE_OK)
1399 {
1400 p = uncompressed_buffer;
1401 }
1402 else
1403 {
1404 archive_set_error(&a->archive,
1405 ARCHIVE_ERRNO_FILE_FORMAT,
1406 "Unsupported ZIP compression method "
1407 "during decompression of link entry (%d: %s)",
1408 zip->entry->compression,
1409 compression_name(zip->entry->compression));
1410 return ARCHIVE_FAILED;
1411 }
1412 }
1413 else
1414 {
1415 p = __archive_read_ahead(a, linkname_length, NULL);
1416 }
1417
1418 if (p == NULL) {
1419 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1420 "Truncated Zip file");
1421 return ARCHIVE_FATAL;
1422 }
1423
1424 sconv = zip->sconv;
1425 if (sconv == NULL && (zip->entry->zip_flags & ZIP_UTF8_NAME))
1426 sconv = zip->sconv_utf8;
1427 if (sconv == NULL)
1428 sconv = zip->sconv_default;
1429 if (archive_entry_copy_symlink_l(entry, p, linkname_full_length,
1430 sconv) != 0) {
1431 if (errno != ENOMEM && sconv == zip->sconv_utf8 &&
1432 (zip->entry->zip_flags & ZIP_UTF8_NAME))
1433 archive_entry_copy_symlink_l(entry, p,
1434 linkname_full_length, NULL);
1435 if (errno == ENOMEM) {
1436 archive_set_error(&a->archive, ENOMEM,
1437 "Can't allocate memory for Symlink");
1438 return (ARCHIVE_FATAL);
1439 }
1440 /*
1441 * Since there is no character-set regulation for
1442 * symlink name, do not report the conversion error
1443 * in an automatic conversion.
1444 */
1445 if (sconv != zip->sconv_utf8 ||
1446 (zip->entry->zip_flags & ZIP_UTF8_NAME) == 0) {
1447 archive_set_error(&a->archive,
1448 ARCHIVE_ERRNO_FILE_FORMAT,
1449 "Symlink cannot be converted "
1450 "from %s to current locale",
1451 archive_string_conversion_charset_name(
1452 sconv));
1453 ret = ARCHIVE_WARN;
1454 }
1455 }
1456 zip_entry->uncompressed_size = zip_entry->compressed_size = 0;
1457
1458 if (__archive_read_consume(a, linkname_length) < 0) {
1459 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1460 "Read error skipping symlink target name");
1461 return ARCHIVE_FATAL;
1462 }
1463 } else if (zip_entry_size_is_set(zip_entry)) {
1464 archive_entry_set_size(entry, zip_entry->uncompressed_size);
1465 }
1466 zip->entry_bytes_remaining = zip_entry->compressed_size;
1467
1468 /* If there's no body, force read_data() to return EOF immediately. */
1469 if (0 == (zip_entry->zip_flags & ZIP_LENGTH_AT_END)
1470 && zip->entry_bytes_remaining < 1)
1471 zip->end_of_entry = 1;
1472
1473 /* Set up a more descriptive format name. */
1474 archive_string_empty(&zip->format_name);
1475 archive_string_sprintf(&zip->format_name, "ZIP %d.%d (%s)",
1476 version / 10, version % 10,
1477 compression_name(zip->entry->compression));
1478 a->archive.archive_format_name = zip->format_name.s;
1479
1480 return (ret);
1481 }
1482
1483 static int
check_authentication_code(struct archive_read * a,const void * _p)1484 check_authentication_code(struct archive_read *a, const void *_p)
1485 {
1486 struct zip *zip = a->format->data;
1487
1488 /* Check authentication code. */
1489 if (zip->hctx_valid) {
1490 const void *p;
1491 uint8_t hmac[20];
1492 size_t hmac_len = 20;
1493 int cmp;
1494
1495 archive_hmac_sha1_final(&zip->hctx, hmac, &hmac_len);
1496 if (_p == NULL) {
1497 /* Read authentication code. */
1498 p = __archive_read_ahead(a, AUTH_CODE_SIZE, NULL);
1499 if (p == NULL) {
1500 archive_set_error(&a->archive,
1501 ARCHIVE_ERRNO_FILE_FORMAT,
1502 "Truncated ZIP file data");
1503 return (ARCHIVE_FATAL);
1504 }
1505 } else {
1506 p = _p;
1507 }
1508 cmp = memcmp(hmac, p, AUTH_CODE_SIZE);
1509 __archive_read_consume(a, AUTH_CODE_SIZE);
1510 if (cmp != 0) {
1511 archive_set_error(&a->archive,
1512 ARCHIVE_ERRNO_MISC,
1513 "ZIP bad Authentication code");
1514 return (ARCHIVE_WARN);
1515 }
1516 }
1517 return (ARCHIVE_OK);
1518 }
1519
1520 /*
1521 * The Zip end-of-file marker is inherently ambiguous. The specification
1522 * in APPNOTE.TXT allows any of four possible formats, and there is no
1523 * guaranteed-correct way for a reader to know a priori which one the writer
1524 * will have used. The four formats are:
1525 * 1. 32-bit format with an initial PK78 marker
1526 * 2. 32-bit format without that marker
1527 * 3. 64-bit format with the marker
1528 * 4. 64-bit format without the marker
1529 *
1530 * Mark Adler's `sunzip` streaming unzip program solved this ambiguity
1531 * by just looking at every possible combination and accepting the
1532 * longest one that matches the expected values. His approach always
1533 * consumes the longest possible matching EOF marker, based on an
1534 * analysis of all the possible failures and how the values could
1535 * overlap.
1536 *
1537 * For example, suppose both of the first two formats listed
1538 * above match. In that case, we know the next four
1539 * 32-bit words match this pattern:
1540 * ```
1541 * [PK\07\08] [CRC32] [compressed size] [uncompressed size]
1542 * ```
1543 * but we know they must also match this pattern:
1544 * ```
1545 * [CRC32] [compressed size] [uncompressed size] [other PK marker]
1546 * ```
1547 *
1548 * Since the first word here matches both the PK78 signature in the
1549 * first form and the CRC32 in the second, we know those two values
1550 * are equal, the CRC32 must be exactly 0x08074b50. Similarly, the
1551 * compressed and uncompressed size must also be exactly this value.
1552 * So we know these four words are all 0x08074b50. If we were to
1553 * accept the shorter pattern, it would be immediately followed by
1554 * another PK78 marker, which is not possible in a well-formed ZIP
1555 * archive unless there is garbage between entries. This implies we
1556 * should not accept the shorter form in such a case; we should accept
1557 * the longer form.
1558 *
1559 * If the second and third possibilities above both match, we
1560 * have a slightly different situation. The following words
1561 * must match both the 32-bit format
1562 * ```
1563 * [CRC32] [compressed size] [uncompressed size] [other PK marker]
1564 * ```
1565 * and the 64-bit format
1566 * ```
1567 * [CRC32] [compressed low] [compressed high] [uncompressed low] [uncompressed high] [other PK marker]
1568 * ```
1569 * Since the 32-bit and 64-bit compressed sizes both match, the
1570 * actual size must fit in 32 bits, which implies the high-order
1571 * word of the compressed size is zero. So we know the uncompressed
1572 * low word is zero, which again implies that if we accept the shorter
1573 * format, there will not be a valid PK marker following it.
1574 *
1575 * Similar considerations rule out the shorter form in every other
1576 * possibly-ambiguous pair. So if two of the four possible formats
1577 * match, we should accept the longer option.
1578 *
1579 * If none of the four formats matches, we know the archive must be
1580 * corrupted in some fashion. In particular, it's possible that the
1581 * length-at-end bit was incorrect and we should not really be looking
1582 * for an EOF marker at all. To allow for this possibility, we
1583 * evaluate the following words to collect data for a later error
1584 * report but do not consume any bytes. We instead rely on the later
1585 * search for a new PK marker to re-sync to the next well-formed
1586 * entry.
1587 */
1588 static void
consume_end_of_file_marker(struct archive_read * a,struct zip * zip)1589 consume_end_of_file_marker(struct archive_read *a, struct zip *zip)
1590 {
1591 const char *marker;
1592 const char *p;
1593 uint64_t compressed32, uncompressed32;
1594 uint64_t compressed64, uncompressed64;
1595 uint64_t compressed_actual, uncompressed_actual;
1596 uint32_t crc32_actual;
1597 const uint32_t PK78 = 0x08074B50ULL;
1598 uint8_t crc32_ignored, crc32_may_be_zero;
1599
1600 /* If there shouldn't be a marker, don't consume it. */
1601 if ((zip->entry->zip_flags & ZIP_LENGTH_AT_END) == 0) {
1602 return;
1603 }
1604
1605 /* The longest Zip end-of-file record is 24 bytes. Since an
1606 * end-of-file record can never appear at the end of the
1607 * archive, we know 24 bytes will be available unless
1608 * the archive is severely truncated. */
1609 if (NULL == (marker = __archive_read_ahead(a, 24, NULL))) {
1610 return;
1611 }
1612 p = marker;
1613
1614 /* The end-of-file record comprises:
1615 * = Optional PK\007\010 marker
1616 * = 4-byte CRC32
1617 * = Compressed size
1618 * = Uncompressed size
1619 *
1620 * The last two fields are either both 32 bits or both 64
1621 * bits. We check all possible layouts and accept any one
1622 * that gives us a complete match, else we make a best-effort
1623 * attempt to parse out the pieces.
1624 */
1625
1626 /* CRC32 checking can be tricky:
1627 * * Test suites sometimes ignore the CRC32
1628 * * AES AE-2 always writes zero for the CRC32
1629 * * AES AE-1 sometimes writes zero for the CRC32
1630 */
1631 crc32_ignored = zip->ignore_crc32;
1632 crc32_may_be_zero = 0;
1633 crc32_actual = zip->computed_crc32;
1634 if (zip->hctx_valid) {
1635 switch (zip->entry->aes_extra.vendor) {
1636 case AES_VENDOR_AE_2:
1637 crc32_actual = 0;
1638 break;
1639 case AES_VENDOR_AE_1:
1640 default:
1641 crc32_may_be_zero = 1;
1642 break;
1643 }
1644 }
1645
1646 /* Values computed from the actual data in the archive. */
1647 compressed_actual = (uint64_t)zip->entry_compressed_bytes_read;
1648 uncompressed_actual = (uint64_t)zip->entry_uncompressed_bytes_read;
1649
1650
1651 /* Longest: PK78 marker, all 64-bit fields (24 bytes total) */
1652 if (archive_le32dec(p) == PK78
1653 && ((archive_le32dec(p + 4) == crc32_actual)
1654 || (crc32_may_be_zero && (archive_le32dec(p + 4) == 0))
1655 || crc32_ignored)
1656 && (archive_le64dec(p + 8) == compressed_actual)
1657 && (archive_le64dec(p + 16) == uncompressed_actual)) {
1658 if (!crc32_ignored) {
1659 zip->entry->crc32 = crc32_actual;
1660 }
1661 zip->entry->compressed_size = compressed_actual;
1662 zip->entry->uncompressed_size = uncompressed_actual;
1663 zip->unconsumed += 24;
1664 return;
1665 }
1666
1667 /* No PK78 marker, 64-bit fields (20 bytes total) */
1668 if (((archive_le32dec(p) == crc32_actual)
1669 || (crc32_may_be_zero && (archive_le32dec(p + 4) == 0))
1670 || crc32_ignored)
1671 && (archive_le64dec(p + 4) == compressed_actual)
1672 && (archive_le64dec(p + 12) == uncompressed_actual)) {
1673 if (!crc32_ignored) {
1674 zip->entry->crc32 = crc32_actual;
1675 }
1676 zip->entry->compressed_size = compressed_actual;
1677 zip->entry->uncompressed_size = uncompressed_actual;
1678 zip->unconsumed += 20;
1679 return;
1680 }
1681
1682 /* PK78 marker and 32-bit fields (16 bytes total) */
1683 if (archive_le32dec(p) == PK78
1684 && ((archive_le32dec(p + 4) == crc32_actual)
1685 || (crc32_may_be_zero && (archive_le32dec(p + 4) == 0))
1686 || crc32_ignored)
1687 && (archive_le32dec(p + 8) == compressed_actual)
1688 && (archive_le32dec(p + 12) == uncompressed_actual)) {
1689 if (!crc32_ignored) {
1690 zip->entry->crc32 = crc32_actual;
1691 }
1692 zip->entry->compressed_size = compressed_actual;
1693 zip->entry->uncompressed_size = uncompressed_actual;
1694 zip->unconsumed += 16;
1695 return;
1696 }
1697
1698 /* Shortest: No PK78 marker, all 32-bit fields (12 bytes total) */
1699 if (((archive_le32dec(p) == crc32_actual)
1700 || (crc32_may_be_zero && (archive_le32dec(p + 4) == 0))
1701 || crc32_ignored)
1702 && (archive_le32dec(p + 4) == compressed_actual)
1703 && (archive_le32dec(p + 8) == uncompressed_actual)) {
1704 if (!crc32_ignored) {
1705 zip->entry->crc32 = crc32_actual;
1706 }
1707 zip->entry->compressed_size = compressed_actual;
1708 zip->entry->uncompressed_size = uncompressed_actual;
1709 zip->unconsumed += 12;
1710 return;
1711 }
1712
1713 /* None of the exact patterns matched. If entry size was unknown
1714 * (ZIP_LENGTH_AT_END flag), before treating this as
1715 * corruption, check whether the next ZIP record follows the data
1716 * immediately: a length-at-end entry whose compression format has
1717 * its own end-of-stream marker (e.g. PPMd) may be written with no
1718 * data descriptor at all. In that case the byte counts we measured
1719 * during decompression are authoritative, so trust them and leave
1720 * the stream untouched. */
1721 if (zip->entry->zip_flags & ZIP_LENGTH_AT_END)
1722 {
1723 const uint32_t sig = archive_le32dec(p);
1724 if (sig == 0x04034b50U /* Local file header */
1725 || sig == 0x02014b50U /* Central directory record */
1726 || sig == 0x06054b50U) /* End of central directory */ {
1727 zip->entry->compressed_size = compressed_actual;
1728 zip->entry->uncompressed_size = uncompressed_actual;
1729 return;
1730 }
1731 }
1732
1733 /* If none of the above patterns gives us a full exact match,
1734 * then there's something definitely amiss. The fallback code
1735 * below will parse out some plausible values for error
1736 * reporting purposes. Note that this won't actually
1737 * consume anything:
1738 *
1739 * = If there really is a marker here, the logic to resync to
1740 * the next entry will suffice to skip it.
1741 *
1742 * = There might not really be a marker: Corruption or bugs
1743 * may have set the length-at-end bit without a marker ever
1744 * having actually been written. In this case, we
1745 * explicitly should not consume any bytes, since that would
1746 * prevent us from correctly reading the next entry.
1747 */
1748 if (archive_le32dec(p) == PK78) {
1749 p += 4; /* Ignore PK78 if it appears to be present */
1750 }
1751 zip->entry->crc32 = archive_le32dec(p); /* Parse CRC32 */
1752 p += 4;
1753
1754 /* Consider both 32- and 64-bit interpretations */
1755 compressed32 = archive_le32dec(p);
1756 uncompressed32 = archive_le32dec(p + 4);
1757 compressed64 = archive_le64dec(p);
1758 uncompressed64 = archive_le64dec(p + 8);
1759
1760 /* The earlier patterns may have failed because of CRC32
1761 * mismatch, so it's still possible that both sizes match.
1762 * Try to match as many as we can...
1763 */
1764 if (compressed32 == compressed_actual
1765 && uncompressed32 == uncompressed_actual) {
1766 /* Both 32-bit fields match */
1767 zip->entry->compressed_size = compressed32;
1768 zip->entry->uncompressed_size = uncompressed32;
1769 } else if (compressed64 == compressed_actual
1770 || uncompressed64 == uncompressed_actual) {
1771 /* One or both 64-bit fields match */
1772 zip->entry->compressed_size = compressed64;
1773 zip->entry->uncompressed_size = uncompressed64;
1774 } else {
1775 /* Zero or one 32-bit fields match */
1776 zip->entry->compressed_size = compressed32;
1777 zip->entry->uncompressed_size = uncompressed32;
1778 }
1779 }
1780
1781 /*
1782 * Read "uncompressed" data.
1783 *
1784 * This is straightforward if we know the size of the data. This is
1785 * always true for the seeking reader (we've examined the Central
1786 * Directory already), and will often be true for the streaming reader
1787 * (the writer was writing uncompressed so probably knows the size).
1788 *
1789 * If we don't know the size, then life is more interesting. Note
1790 * that a careful reading of the Zip specification says that a writer
1791 * must use ZIP_LENGTH_AT_END if it cannot write the CRC into the
1792 * local header. And if it uses ZIP_LENGTH_AT_END, then it is
1793 * prohibited from storing the sizes in the local header. This
1794 * prevents fully-compliant streaming writers from providing any size
1795 * clues to a streaming reader. In this case, we have to scan the
1796 * data as we read to try to locate the end-of-file marker.
1797 *
1798 * We assume here that the end-of-file marker always has the
1799 * PK\007\010 signature. Although it's technically optional, newer
1800 * writers seem to provide it pretty consistently, and it's not clear
1801 * how to efficiently recognize an end-of-file marker that lacks it.
1802 *
1803 * Returns ARCHIVE_OK if successful, ARCHIVE_FATAL otherwise, sets
1804 * zip->end_of_entry if it consumes all of the data.
1805 */
1806 static int
zip_read_data_none(struct archive_read * a,const void ** _buff,size_t * size,int64_t * offset)1807 zip_read_data_none(struct archive_read *a, const void **_buff,
1808 size_t *size, int64_t *offset)
1809 {
1810 struct zip *zip = a->format->data;
1811 const char *buff;
1812 ssize_t bytes_avail;
1813 ssize_t trailing_extra;
1814 int r;
1815
1816 (void)offset; /* UNUSED */
1817
1818 trailing_extra = zip->hctx_valid ? AUTH_CODE_SIZE : 0;
1819
1820 if (zip->entry->zip_flags & ZIP_LENGTH_AT_END) {
1821 const char *p;
1822 ssize_t grabbing_bytes = 24 + trailing_extra;
1823
1824 /* Grab at least 24 bytes. */
1825 buff = __archive_read_ahead(a, grabbing_bytes, &bytes_avail);
1826 if (bytes_avail < grabbing_bytes) {
1827 /* Zip archives have end-of-archive markers
1828 that are longer than this, so a failure to get at
1829 least 24 bytes really does indicate a truncated
1830 file. */
1831 archive_set_error(&a->archive,
1832 ARCHIVE_ERRNO_FILE_FORMAT,
1833 "Truncated ZIP file data");
1834 return (ARCHIVE_FATAL);
1835 }
1836 /* Check for a complete PK\007\010 signature, followed
1837 * by the correct 4-byte CRC. */
1838 p = buff + trailing_extra;
1839 if (p[0] == 'P' && p[1] == 'K'
1840 && p[2] == '\007' && p[3] == '\010'
1841 && (archive_le32dec(p + 4) == zip->computed_crc32
1842 || zip->ignore_crc32
1843 || (zip->hctx_valid
1844 && zip->entry->aes_extra.vendor == AES_VENDOR_AE_2))) {
1845 zip->end_of_entry = 1;
1846 if (zip->hctx_valid) {
1847 r = check_authentication_code(a, buff);
1848 if (r != ARCHIVE_OK)
1849 return (r);
1850 }
1851 return (ARCHIVE_OK);
1852 }
1853 /* If not at EOF, ensure we consume at least one byte. */
1854 ++p;
1855
1856 /* Scan forward until we see where a PK\007\010 signature
1857 * might be. */
1858 /* Return bytes up until that point. On the next call,
1859 * the code above will verify the data descriptor. */
1860 while (p < buff + bytes_avail - 4) {
1861 if (p[3] == 'P') { p += 3; }
1862 else if (p[3] == 'K') { p += 2; }
1863 else if (p[3] == '\007') { p += 1; }
1864 else if (p[3] == '\010' && p[2] == '\007'
1865 && p[1] == 'K' && p[0] == 'P') {
1866 break;
1867 } else { p += 4; }
1868 }
1869 p -= trailing_extra;
1870 bytes_avail = p - buff;
1871 } else {
1872 if (zip->entry_bytes_remaining == 0) {
1873 zip->end_of_entry = 1;
1874 if (zip->hctx_valid) {
1875 r = check_authentication_code(a, NULL);
1876 if (r != ARCHIVE_OK)
1877 return (r);
1878 }
1879 return (ARCHIVE_OK);
1880 }
1881 /* Grab a bunch of bytes. */
1882 buff = __archive_read_ahead(a, 1, &bytes_avail);
1883 if (bytes_avail <= 0) {
1884 archive_set_error(&a->archive,
1885 ARCHIVE_ERRNO_FILE_FORMAT,
1886 "Truncated ZIP file data");
1887 return (ARCHIVE_FATAL);
1888 }
1889 if (bytes_avail > zip->entry_bytes_remaining)
1890 bytes_avail = (ssize_t)zip->entry_bytes_remaining;
1891 }
1892 if (zip->tctx_valid || zip->cctx_valid) {
1893 size_t dec_size = bytes_avail;
1894
1895 if (dec_size > zip->decrypted_buffer_size)
1896 dec_size = zip->decrypted_buffer_size;
1897 if (zip->tctx_valid) {
1898 trad_enc_decrypt_update(&zip->tctx,
1899 (const uint8_t *)buff, dec_size,
1900 zip->decrypted_buffer, dec_size);
1901 } else {
1902 size_t dsize = dec_size;
1903 archive_hmac_sha1_update(&zip->hctx,
1904 (const uint8_t *)buff, dec_size);
1905 archive_decrypto_aes_ctr_update(&zip->cctx,
1906 (const uint8_t *)buff, dec_size,
1907 zip->decrypted_buffer, &dsize);
1908 }
1909 bytes_avail = dec_size;
1910 buff = (const char *)zip->decrypted_buffer;
1911 }
1912 zip->entry_bytes_remaining -= bytes_avail;
1913 zip->entry_uncompressed_bytes_read += bytes_avail;
1914 zip->entry_compressed_bytes_read += bytes_avail;
1915 zip->unconsumed += bytes_avail;
1916 *size = bytes_avail;
1917 *_buff = buff;
1918 return (ARCHIVE_OK);
1919 }
1920
1921 #if HAVE_LZMA_H && HAVE_LIBLZMA
1922 static int
zipx_xz_init(struct archive_read * a,struct zip * zip)1923 zipx_xz_init(struct archive_read *a, struct zip *zip)
1924 {
1925 lzma_ret r;
1926
1927 if(zip->zipx_lzma_valid) {
1928 lzma_end(&zip->zipx_lzma_stream);
1929 zip->zipx_lzma_valid = 0;
1930 }
1931
1932 memset(&zip->zipx_lzma_stream, 0, sizeof(zip->zipx_lzma_stream));
1933 r = lzma_stream_decoder(&zip->zipx_lzma_stream, UINT64_MAX, 0);
1934 if (r != LZMA_OK) {
1935 archive_set_error(&(a->archive), ARCHIVE_ERRNO_MISC,
1936 "xz initialization failed (%d)",
1937 r);
1938
1939 return (ARCHIVE_FAILED);
1940 }
1941
1942 zip->zipx_lzma_valid = 1;
1943
1944 free(zip->uncompressed_buffer);
1945
1946 zip->uncompressed_buffer_size = 256 * 1024;
1947 zip->uncompressed_buffer = malloc(zip->uncompressed_buffer_size);
1948 if (zip->uncompressed_buffer == NULL) {
1949 archive_set_error(&a->archive, ENOMEM,
1950 "No memory for xz decompression");
1951 return (ARCHIVE_FATAL);
1952 }
1953
1954 zip->decompress_init = 1;
1955 return (ARCHIVE_OK);
1956 }
1957
1958 static int
zipx_lzma_alone_init(struct archive_read * a,struct zip * zip)1959 zipx_lzma_alone_init(struct archive_read *a, struct zip *zip)
1960 {
1961 lzma_ret r;
1962 const uint8_t* p;
1963
1964 #pragma pack(push)
1965 #pragma pack(1)
1966 struct _alone_header {
1967 uint8_t bytes[5];
1968 uint64_t uncompressed_size;
1969 } alone_header;
1970 #pragma pack(pop)
1971
1972 if(zip->zipx_lzma_valid) {
1973 lzma_end(&zip->zipx_lzma_stream);
1974 zip->zipx_lzma_valid = 0;
1975 }
1976
1977 /* To unpack ZIPX's "LZMA" (id 14) stream we can use standard liblzma
1978 * that is a part of XZ Utils. The stream format stored inside ZIPX
1979 * file is a modified "lzma alone" file format, that was used by the
1980 * `lzma` utility which was later deprecated in favour of `xz` utility.
1981 * Since those formats are nearly the same, we can use a standard
1982 * "lzma alone" decoder from XZ Utils. */
1983
1984 memset(&zip->zipx_lzma_stream, 0, sizeof(zip->zipx_lzma_stream));
1985 r = lzma_alone_decoder(&zip->zipx_lzma_stream, 576 * ((uint64_t)1 << 20));
1986 if (r != LZMA_OK) {
1987 archive_set_error(&(a->archive), ARCHIVE_ERRNO_MISC,
1988 "lzma initialization failed (%d)", r);
1989
1990 return (ARCHIVE_FAILED);
1991 }
1992
1993 /* Flag the cleanup function that we want our lzma-related structures
1994 * to be freed later. */
1995 zip->zipx_lzma_valid = 1;
1996
1997 /* The "lzma alone" file format and the stream format inside ZIPx are
1998 * almost the same. Here's an example of a structure of "lzma alone"
1999 * format:
2000 *
2001 * $ cat /bin/ls | lzma | xxd | head -n 1
2002 * 00000000: 5d00 0080 00ff ffff ffff ffff ff00 2814
2003 *
2004 * 5 bytes 8 bytes n bytes
2005 * <lzma_params><uncompressed_size><data...>
2006 *
2007 * lzma_params is a 5-byte blob that has to be decoded to extract
2008 * parameters of this LZMA stream. The uncompressed_size field is an
2009 * uint64_t value that contains information about the size of the
2010 * uncompressed file, or UINT64_MAX if this value is unknown.
2011 * The <data...> part is the actual lzma-compressed data stream.
2012 *
2013 * Now here's the structure of the stream inside the ZIPX file:
2014 *
2015 * $ cat stream_inside_zipx | xxd | head -n 1
2016 * 00000000: 0914 0500 5d00 8000 0000 2814 .... ....
2017 *
2018 * 2byte 2byte 5 bytes n bytes
2019 * <magic1><magic2><lzma_params><data...>
2020 *
2021 * This means that the ZIPX file contains an additional magic1 and
2022 * magic2 headers, the lzma_params field contains the same parameter
2023 * set as in the "lzma alone" format, and the <data...> field is the
2024 * same as in the "lzma alone" format as well. Note that also the zipx
2025 * format is missing the uncompressed_size field.
2026 *
2027 * So, in order to use the "lzma alone" decoder for the zipx lzma
2028 * stream, we simply need to shuffle around some fields, prepare a new
2029 * lzma alone header, feed it into lzma alone decoder so it will
2030 * initialize itself properly, and then we can start feeding normal
2031 * zipx lzma stream into the decoder.
2032 */
2033
2034 /* Read magic1,magic2,lzma_params from the ZIPX stream. */
2035 /* When the compressed size is unknown (e.g. ZIP_LENGTH_AT_END read
2036 * from a non-seekable source), entry_bytes_remaining is 0 or negative
2037 * here. We can still attempt to read the 9-byte header; if the data
2038 * is truly truncated, the __archive_read_ahead calls below will catch
2039 * it. */
2040 if(zip->entry_bytes_remaining > 0
2041 && zip->entry_bytes_remaining < 9) {
2042 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2043 "Truncated lzma data");
2044 return (ARCHIVE_FATAL);
2045 }
2046
2047 if (zip->tctx_valid || zip->cctx_valid) {
2048 const void *decrypted;
2049 size_t out_len;
2050 size_t consumed;
2051 int ret;
2052
2053 ret = zipx_read_header_and_decrypt(a, &decrypted, 9, &out_len, &consumed);
2054 if (ret != ARCHIVE_OK)
2055 return ret;
2056 p = decrypted;
2057 } else {
2058 p = __archive_read_ahead(a, 9, NULL);
2059 if (p == NULL) {
2060 archive_set_error(&a->archive,
2061 ARCHIVE_ERRNO_FILE_FORMAT,
2062 "Truncated lzma data");
2063 return (ARCHIVE_FATAL);
2064 }
2065 }
2066
2067 if(p[2] != 0x05 || p[3] != 0x00) {
2068 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2069 "Invalid lzma data");
2070 return (ARCHIVE_FATAL);
2071 }
2072
2073 /* Prepare an lzma alone header: copy the lzma_params blob into
2074 * a proper place into the lzma alone header. */
2075 memcpy(&alone_header.bytes[0], p + 4, 5);
2076
2077 /* Initialize the 'uncompressed size' field to unknown; we'll manually
2078 * monitor how many bytes there are still to be uncompressed. */
2079 alone_header.uncompressed_size = UINT64_MAX;
2080
2081 if(!zip->uncompressed_buffer) {
2082 zip->uncompressed_buffer_size = 256 * 1024;
2083 zip->uncompressed_buffer = malloc(zip->uncompressed_buffer_size);
2084
2085 if (zip->uncompressed_buffer == NULL) {
2086 archive_set_error(&a->archive, ENOMEM,
2087 "No memory for lzma decompression");
2088 return (ARCHIVE_FATAL);
2089 }
2090 }
2091
2092 zip->zipx_lzma_stream.next_in = (void*) &alone_header;
2093 zip->zipx_lzma_stream.avail_in = sizeof(alone_header);
2094 zip->zipx_lzma_stream.total_in = 0;
2095 zip->zipx_lzma_stream.next_out = zip->uncompressed_buffer;
2096 zip->zipx_lzma_stream.avail_out = zip->uncompressed_buffer_size;
2097 zip->zipx_lzma_stream.total_out = 0;
2098
2099 /* Feed only the header into the lzma alone decoder. This will
2100 * effectively initialize the decoder, and will not produce any
2101 * output bytes yet. */
2102 r = lzma_code(&zip->zipx_lzma_stream, LZMA_RUN);
2103 if (r != LZMA_OK) {
2104 if (r == LZMA_MEMLIMIT_ERROR)
2105 archive_set_error(&a->archive, ENOMEM,
2106 "lzma stream requires too much memory");
2107 else
2108 archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
2109 "lzma stream initialization error");
2110 return ARCHIVE_FATAL;
2111 }
2112
2113 /* We've already consumed some bytes, so take this into account. */
2114 __archive_read_consume(a, 9);
2115 zip->entry_compressed_bytes_read += 9;
2116 if (zip->entry_bytes_remaining > 0) {
2117 zip->entry_bytes_remaining -= 9;
2118 }
2119
2120 zip->decompress_init = 1;
2121 return (ARCHIVE_OK);
2122 }
2123
2124 static int
zip_read_data_zipx_xz(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)2125 zip_read_data_zipx_xz(struct archive_read *a, const void **buff,
2126 size_t *size, int64_t *offset)
2127 {
2128 struct zip *zip = a->format->data;
2129 int ret;
2130 lzma_ret lz_ret;
2131 const void* compressed_buf;
2132 const void* sp;
2133 ssize_t bytes_avail, to_consume = 0;
2134
2135 (void) offset; /* UNUSED */
2136
2137 /* Initialize decompressor if not yet initialized. */
2138 if (!zip->decompress_init) {
2139 ret = zipx_xz_init(a, zip);
2140 if (ret != ARCHIVE_OK)
2141 return (ret);
2142 }
2143
2144 compressed_buf = sp = __archive_read_ahead(a, 1, &bytes_avail);
2145 if (0 == (zip->entry->zip_flags & ZIP_LENGTH_AT_END)
2146 && bytes_avail > zip->entry_bytes_remaining) {
2147 bytes_avail = (ssize_t)zip->entry_bytes_remaining;
2148 }
2149 if (bytes_avail < 0) {
2150 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2151 "Truncated xz file body");
2152 return (ARCHIVE_FATAL);
2153 }
2154
2155 zip_read_decrypt(zip, compressed_buf, bytes_avail,
2156 &compressed_buf, &bytes_avail, &sp);
2157
2158 zip->zipx_lzma_stream.next_in = compressed_buf;
2159 zip->zipx_lzma_stream.avail_in = bytes_avail;
2160 zip->zipx_lzma_stream.total_in = 0;
2161 zip->zipx_lzma_stream.next_out = zip->uncompressed_buffer;
2162 zip->zipx_lzma_stream.avail_out = zip->uncompressed_buffer_size;
2163 zip->zipx_lzma_stream.total_out = 0;
2164
2165 /* Perform the decompression. */
2166 lz_ret = lzma_code(&zip->zipx_lzma_stream, LZMA_RUN);
2167 switch(lz_ret) {
2168 case LZMA_DATA_ERROR:
2169 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2170 "xz data error (%d)", (int) lz_ret);
2171 return (ARCHIVE_FATAL);
2172
2173 case LZMA_NO_CHECK:
2174 case LZMA_OK:
2175 break;
2176
2177 default:
2178 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2179 "xz unknown error (%d)", (int) lz_ret);
2180 return (ARCHIVE_FATAL);
2181
2182 case LZMA_STREAM_END:
2183 lzma_end(&zip->zipx_lzma_stream);
2184 zip->zipx_lzma_valid = 0;
2185
2186 /* This assertion is only possible if the size of the compressed data
2187 * stream is known -> !ZIP_LENGTH_AT_END */
2188 if((int64_t) zip->zipx_lzma_stream.total_in != zip->entry_bytes_remaining
2189 && !(zip->entry->zip_flags & ZIP_LENGTH_AT_END))
2190 {
2191 archive_set_error(&a->archive,
2192 ARCHIVE_ERRNO_MISC,
2193 "xz premature end of stream");
2194 return (ARCHIVE_FATAL);
2195 }
2196
2197 zip->end_of_entry = 1;
2198 break;
2199 }
2200
2201 to_consume = (ssize_t)zip->zipx_lzma_stream.total_in;
2202
2203 __archive_read_consume(a, to_consume);
2204 zip->entry_bytes_remaining -= to_consume;
2205 zip->entry_compressed_bytes_read += to_consume;
2206 zip->entry_uncompressed_bytes_read += zip->zipx_lzma_stream.total_out;
2207
2208 zip_read_decrypt_update(zip, to_consume, sp);
2209
2210 if (zip->end_of_entry) {
2211 if (zip->hctx_valid) {
2212 ret = check_authentication_code(a, NULL);
2213 if (ret != ARCHIVE_OK)
2214 return ret;
2215 }
2216 }
2217
2218 *size = (size_t)zip->zipx_lzma_stream.total_out;
2219 *buff = zip->uncompressed_buffer;
2220
2221 return (ARCHIVE_OK);
2222 }
2223
2224 static int
zip_read_data_zipx_lzma_alone(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)2225 zip_read_data_zipx_lzma_alone(struct archive_read *a, const void **buff,
2226 size_t *size, int64_t *offset)
2227 {
2228 struct zip *zip = a->format->data;
2229 int ret;
2230 lzma_ret lz_ret;
2231 const void* compressed_buf;
2232 const void* sp;
2233 ssize_t bytes_avail, to_consume;
2234
2235 (void) offset; /* UNUSED */
2236
2237 /* Initialize decompressor if not yet initialized. */
2238 if (!zip->decompress_init) {
2239 ret = zipx_lzma_alone_init(a, zip);
2240 if (ret != ARCHIVE_OK)
2241 return (ret);
2242 }
2243
2244 /* Fetch more compressed data. The same note as in deflate handler
2245 * applies here as well:
2246 *
2247 * Note: '1' here is a performance optimization. Recall that the
2248 * decompression layer returns a count of available bytes; asking for
2249 * more than that forces the decompressor to combine reads by copying
2250 * data.
2251 */
2252 compressed_buf = __archive_read_ahead(a, 1, &bytes_avail);
2253 if (zip->entry_bytes_remaining > 0
2254 && bytes_avail > zip->entry_bytes_remaining) {
2255 bytes_avail = (ssize_t)zip->entry_bytes_remaining;
2256 }
2257 if (bytes_avail < 0) {
2258 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2259 "Truncated lzma file body");
2260 return (ARCHIVE_FATAL);
2261 }
2262
2263 zip_read_decrypt(zip, compressed_buf, bytes_avail,
2264 &compressed_buf, &bytes_avail, &sp);
2265
2266 zip->zipx_lzma_stream.next_in = compressed_buf;
2267 zip->zipx_lzma_stream.avail_in = bytes_avail;
2268 zip->zipx_lzma_stream.total_in = 0;
2269 zip->zipx_lzma_stream.next_out = zip->uncompressed_buffer;
2270 /* These lzma_alone streams lack an end of stream marker in some
2271 * cases, so when the uncompressed size is known we cap avail_out to
2272 * make sure the unpacker won't try to unpack more than it's supposed
2273 * to. When the compressed size is unknown (entry_bytes_remaining <= 0,
2274 * e.g. ZIP_LENGTH_AT_END from a non-seekable source) we must use the
2275 * full buffer and rely on the LZMA stream end marker to detect the end
2276 * of the entry. */
2277 if (zip->entry_bytes_remaining <= 0) {
2278 zip->zipx_lzma_stream.avail_out = zip->uncompressed_buffer_size;
2279 } else {
2280 zip->zipx_lzma_stream.avail_out =
2281 (size_t)zipmin((int64_t) zip->uncompressed_buffer_size,
2282 zip->entry->uncompressed_size -
2283 zip->entry_uncompressed_bytes_read);
2284 }
2285 zip->zipx_lzma_stream.total_out = 0;
2286
2287 /* Perform the decompression. */
2288 lz_ret = lzma_code(&zip->zipx_lzma_stream, LZMA_RUN);
2289 switch(lz_ret) {
2290 case LZMA_DATA_ERROR:
2291 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2292 "lzma data error (%d)", (int) lz_ret);
2293 return (ARCHIVE_FATAL);
2294
2295 /* This case is optional in lzma alone format. It can happen,
2296 * but most of the files don't have it. (GitHub #1257) */
2297 case LZMA_STREAM_END:
2298 /* This assertion is only possible if the size of the
2299 * compressed data stream is known. */
2300 if((int64_t) zip->zipx_lzma_stream.total_in !=
2301 zip->entry_bytes_remaining
2302 && zip->entry_bytes_remaining > 0)
2303 {
2304 archive_set_error(&a->archive,
2305 ARCHIVE_ERRNO_MISC,
2306 "lzma alone premature end of stream");
2307 return (ARCHIVE_FATAL);
2308 }
2309
2310 zip->end_of_entry = 1;
2311 break;
2312
2313 case LZMA_OK:
2314 break;
2315
2316 case LZMA_BUF_ERROR:
2317 if (zip->zipx_lzma_stream.avail_out == 0) {
2318 /* The output buffer was filled exactly. When
2319 * the uncompressed size is known this means we
2320 * have decompressed all expected bytes. When
2321 * the size is unknown a full buffer just means
2322 * we need another iteration. */
2323 if (zip->entry_bytes_remaining > 0)
2324 zip->end_of_entry = 1;
2325 break;
2326 }
2327 /* FALL THROUGH */
2328 default:
2329 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2330 "lzma unknown error (%d)", (int) lz_ret);
2331 return (ARCHIVE_FATAL);
2332 }
2333
2334 to_consume = (ssize_t)zip->zipx_lzma_stream.total_in;
2335
2336 /* Update pointers. */
2337 __archive_read_consume(a, to_consume);
2338 zip->entry_compressed_bytes_read += to_consume;
2339 zip->entry_uncompressed_bytes_read += zip->zipx_lzma_stream.total_out;
2340
2341 zip_read_decrypt_update(zip, to_consume, sp);
2342
2343 if(zip->entry_bytes_remaining > 0) {
2344 zip->entry_bytes_remaining -= to_consume;
2345 if(zip->entry_bytes_remaining == 0) {
2346 zip->end_of_entry = 1;
2347 }
2348 }
2349
2350 if(zip->end_of_entry && zip->entry_bytes_remaining > 0) {
2351 ssize_t remaining = (ssize_t)zip->entry_bytes_remaining;
2352 const void *p = __archive_read_ahead(a, remaining, NULL);
2353 if (p != NULL) {
2354 if (zip->hctx_valid)
2355 archive_hmac_sha1_update(&zip->hctx,
2356 p, remaining);
2357 __archive_read_consume(a, remaining);
2358 zip->entry_compressed_bytes_read += remaining;
2359 zip->entry_bytes_remaining = 0;
2360 }
2361 }
2362
2363 /* Free lzma decoder handle because we'll no longer need it. */
2364 /* This cannot be folded into LZMA_STREAM_END handling above
2365 * because the stream end marker is not required in this format. */
2366 if(zip->end_of_entry) {
2367 lzma_end(&zip->zipx_lzma_stream);
2368 zip->zipx_lzma_valid = 0;
2369
2370 if (zip->hctx_valid) {
2371 ret = check_authentication_code(a, NULL);
2372 if (ret != ARCHIVE_OK)
2373 return ret;
2374 }
2375 }
2376
2377 /* Return values. */
2378 *size = (size_t)zip->zipx_lzma_stream.total_out;
2379 *buff = zip->uncompressed_buffer;
2380
2381 /* If we're here, then we're good! */
2382 return (ARCHIVE_OK);
2383 }
2384 #endif /* HAVE_LZMA_H && HAVE_LIBLZMA */
2385
2386 static int
zipx_ppmd8_init(struct archive_read * a,struct zip * zip)2387 zipx_ppmd8_init(struct archive_read *a, struct zip *zip)
2388 {
2389 const void* p;
2390 uint32_t val;
2391 uint32_t order;
2392 uint32_t mem;
2393 uint32_t restore_method;
2394
2395 /* Remove previous decompression context if it exists. */
2396 if(zip->ppmd8_valid) {
2397 __archive_ppmd8_functions.Ppmd8_Free(&zip->ppmd8);
2398 zip->ppmd8_valid = 0;
2399 }
2400
2401 /* Create a new decompression context. */
2402 __archive_ppmd8_functions.Ppmd8_Construct(&zip->ppmd8);
2403 zip->ppmd8_stream_failed = 0;
2404
2405 /* Setup function pointers required by Ppmd8 decompressor. The
2406 * 'ppmd_read' function will feed new bytes to the decompressor,
2407 * and will increment the 'zip->zipx_ppmd_read_compressed' counter. */
2408 zip->ppmd8.Stream.In = &zip->zipx_ppmd_stream;
2409 zip->zipx_ppmd_stream.a = a;
2410 zip->zipx_ppmd_stream.Read = &ppmd_read;
2411
2412 /* Reset number of read bytes to 0. */
2413 zip->zipx_ppmd_read_compressed = 0;
2414
2415 /* Read Ppmd8 header (2 bytes). */
2416 if (zip->tctx_valid || zip->cctx_valid) {
2417 size_t out_len;
2418 size_t consumed;
2419 int ret;
2420
2421 ret = zipx_read_header_and_decrypt(a, &p, 2, &out_len, &consumed);
2422 if (ret != ARCHIVE_OK)
2423 return ret;
2424 } else {
2425 p = __archive_read_ahead(a, 2, NULL);
2426 if(!p) {
2427 archive_set_error(&a->archive,
2428 ARCHIVE_ERRNO_FILE_FORMAT,
2429 "Truncated file data in PPMd8 stream");
2430 return (ARCHIVE_FATAL);
2431 }
2432 }
2433 __archive_read_consume(a, 2);
2434
2435 /* Decode the stream's compression parameters. */
2436 val = archive_le16dec(p);
2437 order = (val & 15) + 1;
2438 mem = ((val >> 4) & 0xff) + 1;
2439 restore_method = (val >> 12);
2440
2441 if(order < 2 || restore_method > 2) {
2442 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2443 "Invalid parameter set in PPMd8 stream (order=%" PRIu32 ", "
2444 "restore=%" PRIu32 ")", order, restore_method);
2445 return (ARCHIVE_FAILED);
2446 }
2447
2448 /* Allocate the memory needed to properly decompress the file. */
2449 if(!__archive_ppmd8_functions.Ppmd8_Alloc(&zip->ppmd8, mem << 20)) {
2450 archive_set_error(&a->archive, ENOMEM,
2451 "Unable to allocate memory for PPMd8 stream: %" PRIu32 " bytes",
2452 mem << 20);
2453 return (ARCHIVE_FATAL);
2454 }
2455
2456 /* Signal the cleanup function to release Ppmd8 context in the
2457 * cleanup phase. */
2458 zip->ppmd8_valid = 1;
2459
2460 /* Perform further Ppmd8 initialization. */
2461 if(!__archive_ppmd8_functions.Ppmd8_RangeDec_Init(&zip->ppmd8)) {
2462 archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
2463 "PPMd8 stream range decoder initialization error");
2464 return (ARCHIVE_FATAL);
2465 }
2466
2467 __archive_ppmd8_functions.Ppmd8_Init(&zip->ppmd8, order,
2468 restore_method);
2469
2470 /* Allocate the buffer that will hold uncompressed data. */
2471 free(zip->uncompressed_buffer);
2472
2473 zip->uncompressed_buffer_size = 256 * 1024;
2474 zip->uncompressed_buffer = malloc(zip->uncompressed_buffer_size);
2475
2476 if(zip->uncompressed_buffer == NULL) {
2477 archive_set_error(&a->archive, ENOMEM,
2478 "No memory for PPMd8 decompression");
2479 return ARCHIVE_FATAL;
2480 }
2481
2482 /* Ppmd8 initialization is done. */
2483 zip->decompress_init = 1;
2484
2485 /* We've already read 2 bytes in the output stream. Additionally,
2486 * Ppmd8 initialization code could read some data as well. So we
2487 * are advancing the stream by 2 bytes plus whatever number of
2488 * bytes Ppmd8 init function used. */
2489 zip->entry_compressed_bytes_read += 2 + zip->zipx_ppmd_read_compressed;
2490
2491 return ARCHIVE_OK;
2492 }
2493
2494 static int
zip_read_data_zipx_ppmd(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)2495 zip_read_data_zipx_ppmd(struct archive_read *a, const void **buff,
2496 size_t *size, int64_t *offset)
2497 {
2498 struct zip *zip = a->format->data;
2499 int ret;
2500 size_t consumed_bytes = 0;
2501
2502 (void) offset; /* UNUSED */
2503
2504 /* If we're here for the first time, initialize Ppmd8 decompression
2505 * context first. */
2506 if(!zip->decompress_init) {
2507 ret = zipx_ppmd8_init(a, zip);
2508 if(ret != ARCHIVE_OK)
2509 return ret;
2510 }
2511
2512 /* Fetch for more data. We're reading 1 byte here, but libarchive
2513 * should prefetch more bytes. */
2514 if(__archive_read_ahead(a, 1, NULL) == NULL) {
2515 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2516 "Truncated PPMd8 file body");
2517 return (ARCHIVE_FATAL);
2518 }
2519
2520 /* This counter will be updated inside ppmd_read(), which at one
2521 * point will be called by Ppmd8_DecodeSymbol. */
2522 zip->zipx_ppmd_read_compressed = 0;
2523
2524 /* Decompression loop. */
2525 do {
2526 int sym = __archive_ppmd8_functions.Ppmd8_DecodeSymbol(
2527 &zip->ppmd8);
2528 if(sym < 0) {
2529 zip->end_of_entry = 1;
2530 break;
2531 }
2532
2533 /* This field is set by ppmd_read() when there was no more data
2534 * to be read. */
2535 if(zip->ppmd8_stream_failed) {
2536 archive_set_error(&a->archive,
2537 ARCHIVE_ERRNO_FILE_FORMAT,
2538 "Truncated PPMd8 file body");
2539 return (ARCHIVE_FATAL);
2540 }
2541
2542 zip->uncompressed_buffer[consumed_bytes] = (uint8_t) sym;
2543 ++consumed_bytes;
2544 } while(consumed_bytes < zip->uncompressed_buffer_size);
2545
2546 /* Update pointers so we can continue decompression in another call. */
2547 zip->entry_bytes_remaining -= zip->zipx_ppmd_read_compressed;
2548 zip->entry_compressed_bytes_read += zip->zipx_ppmd_read_compressed;
2549 zip->entry_uncompressed_bytes_read += consumed_bytes;
2550
2551 /* If we're at the end of stream, deinitialize Ppmd8 context. */
2552 if(zip->end_of_entry) {
2553 __archive_ppmd8_functions.Ppmd8_Free(&zip->ppmd8);
2554 zip->ppmd8_valid = 0;
2555
2556 if (zip->hctx_valid) {
2557 int r = check_authentication_code(a, NULL);
2558 if (r != ARCHIVE_OK)
2559 return (r);
2560 }
2561 }
2562
2563 /* Update pointers for libarchive. */
2564 *buff = zip->uncompressed_buffer;
2565 *size = consumed_bytes;
2566
2567 return ARCHIVE_OK;
2568 }
2569
2570 #ifdef HAVE_BZLIB_H
2571 static int
zipx_bzip2_init(struct archive_read * a,struct zip * zip)2572 zipx_bzip2_init(struct archive_read *a, struct zip *zip)
2573 {
2574 int r;
2575
2576 /* Deallocate already existing BZ2 decompression context if it
2577 * exists. */
2578 if(zip->bzstream_valid) {
2579 BZ2_bzDecompressEnd(&zip->bzstream);
2580 zip->bzstream_valid = 0;
2581 }
2582
2583 /* Allocate a new BZ2 decompression context. */
2584 memset(&zip->bzstream, 0, sizeof(bz_stream));
2585 r = BZ2_bzDecompressInit(&zip->bzstream, 0, 1);
2586 if(r != BZ_OK) {
2587 archive_set_error(&(a->archive), ARCHIVE_ERRNO_MISC,
2588 "bzip2 initialization failed (%d)",
2589 r);
2590
2591 return ARCHIVE_FAILED;
2592 }
2593
2594 /* Mark the bzstream field to be released in cleanup phase. */
2595 zip->bzstream_valid = 1;
2596
2597 /* (Re)allocate the buffer that will contain decompressed bytes. */
2598 free(zip->uncompressed_buffer);
2599
2600 zip->uncompressed_buffer_size = 256 * 1024;
2601 zip->uncompressed_buffer = malloc(zip->uncompressed_buffer_size);
2602 if (zip->uncompressed_buffer == NULL) {
2603 archive_set_error(&a->archive, ENOMEM,
2604 "No memory for bzip2 decompression");
2605 return ARCHIVE_FATAL;
2606 }
2607
2608 /* Initialization done. */
2609 zip->decompress_init = 1;
2610 return ARCHIVE_OK;
2611 }
2612
2613 static int
zip_read_data_zipx_bzip2(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)2614 zip_read_data_zipx_bzip2(struct archive_read *a, const void **buff,
2615 size_t *size, int64_t *offset)
2616 {
2617 struct zip *zip = a->format->data;
2618 ssize_t bytes_avail = 0, to_consume;
2619 const void *compressed_buff;
2620 const void *sp;
2621 int r;
2622 uint64_t total_out;
2623
2624 (void) offset; /* UNUSED */
2625
2626 /* Initialize decompression context if we're here for the first time. */
2627 if(!zip->decompress_init) {
2628 r = zipx_bzip2_init(a, zip);
2629 if(r != ARCHIVE_OK)
2630 return r;
2631 }
2632
2633 /* Fetch more compressed bytes. */
2634 compressed_buff = __archive_read_ahead(a, 1, &bytes_avail);
2635 if (0 == (zip->entry->zip_flags & ZIP_LENGTH_AT_END)
2636 && bytes_avail > zip->entry_bytes_remaining) {
2637 bytes_avail = (ssize_t)zip->entry_bytes_remaining;
2638 }
2639 if(bytes_avail < 1) {
2640 /* libbz2 doesn't complain when caller feeds avail_in == 0.
2641 * It will actually return success in this case, which is
2642 * undesirable. This is why we need to make this check
2643 * manually. */
2644 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2645 "Truncated bzip2 file body");
2646 return (ARCHIVE_FATAL);
2647 }
2648
2649 zip_read_decrypt(zip, compressed_buff, bytes_avail,
2650 &compressed_buff, &bytes_avail, &sp);
2651
2652 /* Setup buffer boundaries. */
2653 zip->bzstream.next_in = (char*)(uintptr_t) compressed_buff;
2654 zip->bzstream.avail_in = (uint32_t)bytes_avail;
2655 zip->bzstream.total_in_hi32 = 0;
2656 zip->bzstream.total_in_lo32 = 0;
2657 zip->bzstream.next_out = (char*) zip->uncompressed_buffer;
2658 zip->bzstream.avail_out = (uint32_t)zip->uncompressed_buffer_size;
2659 zip->bzstream.total_out_hi32 = 0;
2660 zip->bzstream.total_out_lo32 = 0;
2661
2662 /* Perform the decompression. */
2663 r = BZ2_bzDecompress(&zip->bzstream);
2664 switch(r) {
2665 case BZ_STREAM_END:
2666 /* If we're at the end of the stream, deinitialize the
2667 * decompression context now. */
2668 switch(BZ2_bzDecompressEnd(&zip->bzstream)) {
2669 case BZ_OK:
2670 break;
2671 default:
2672 archive_set_error(&a->archive,
2673 ARCHIVE_ERRNO_MISC,
2674 "Failed to clean up bzip2 "
2675 "decompressor");
2676 return ARCHIVE_FATAL;
2677 }
2678
2679 zip->end_of_entry = 1;
2680 break;
2681 case BZ_OK:
2682 /* The decompressor has successfully decoded this
2683 * chunk of data, but more data is still in queue. */
2684 break;
2685 default:
2686 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2687 "bzip2 decompression failed");
2688 return ARCHIVE_FATAL;
2689 }
2690
2691 /* Update the pointers so decompressor can continue decoding. */
2692 to_consume = zip->bzstream.total_in_lo32;
2693 __archive_read_consume(a, to_consume);
2694
2695 total_out = ((uint64_t) zip->bzstream.total_out_hi32 << 32) |
2696 zip->bzstream.total_out_lo32;
2697
2698 zip->entry_bytes_remaining -= to_consume;
2699 zip->entry_compressed_bytes_read += to_consume;
2700 zip->entry_uncompressed_bytes_read += total_out;
2701
2702 zip_read_decrypt_update(zip, to_consume, sp);
2703
2704 if (zip->end_of_entry && zip->hctx_valid) {
2705 r = check_authentication_code(a, NULL);
2706 if (r != ARCHIVE_OK)
2707 return r;
2708 }
2709
2710 /* Give libarchive its due. */
2711 *size = (size_t)total_out;
2712 *buff = zip->uncompressed_buffer;
2713
2714 return ARCHIVE_OK;
2715 }
2716
2717 #endif
2718
2719 #if HAVE_ZSTD_H && HAVE_LIBZSTD
2720 static int
zipx_zstd_init(struct archive_read * a,struct zip * zip)2721 zipx_zstd_init(struct archive_read *a, struct zip *zip)
2722 {
2723 size_t r;
2724
2725 /* Deallocate already existing Zstd decompression context if it
2726 * exists. */
2727 if(zip->zstdstream_valid) {
2728 ZSTD_freeDStream(zip->zstdstream);
2729 zip->zstdstream_valid = 0;
2730 }
2731
2732 /* Allocate a new Zstd decompression context. */
2733 zip->zstdstream = ZSTD_createDStream();
2734
2735 r = ZSTD_initDStream(zip->zstdstream);
2736 if (ZSTD_isError(r)) {
2737 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2738 "Error initializing zstd decompressor: %s",
2739 ZSTD_getErrorName(r));
2740
2741 return ARCHIVE_FAILED;
2742 }
2743
2744 /* Mark the zstdstream field to be released in cleanup phase. */
2745 zip->zstdstream_valid = 1;
2746
2747 /* (Re)allocate the buffer that will contain decompressed bytes. */
2748 free(zip->uncompressed_buffer);
2749
2750 zip->uncompressed_buffer_size = ZSTD_DStreamOutSize();
2751 zip->uncompressed_buffer = malloc(zip->uncompressed_buffer_size);
2752 if (zip->uncompressed_buffer == NULL) {
2753 archive_set_error(&a->archive, ENOMEM,
2754 "No memory for Zstd decompression");
2755
2756 return ARCHIVE_FATAL;
2757 }
2758
2759 /* Initialization done. */
2760 zip->decompress_init = 1;
2761 return ARCHIVE_OK;
2762 }
2763
2764 static int
zip_read_data_zipx_zstd(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)2765 zip_read_data_zipx_zstd(struct archive_read *a, const void **buff,
2766 size_t *size, int64_t *offset)
2767 {
2768 struct zip *zip = a->format->data;
2769 ssize_t bytes_avail = 0, to_consume;
2770 const void *compressed_buff;
2771 const void *sp;
2772 int r;
2773 size_t ret;
2774 uint64_t total_out;
2775 ZSTD_outBuffer out;
2776 ZSTD_inBuffer in;
2777
2778 (void) offset; /* UNUSED */
2779
2780 /* Initialize decompression context if we're here for the first time. */
2781 if(!zip->decompress_init) {
2782 r = zipx_zstd_init(a, zip);
2783 if(r != ARCHIVE_OK)
2784 return r;
2785 }
2786
2787 /* Fetch more compressed bytes */
2788 compressed_buff = sp = __archive_read_ahead(a, 1, &bytes_avail);
2789 if (0 == (zip->entry->zip_flags & ZIP_LENGTH_AT_END)
2790 && bytes_avail > zip->entry_bytes_remaining) {
2791 bytes_avail = (ssize_t)zip->entry_bytes_remaining;
2792 }
2793 if(bytes_avail < 1) {
2794 /* zstd doesn't complain when caller feeds avail_in == 0.
2795 * It will actually return success in this case, which is
2796 * undesirable. This is why we need to make this check
2797 * manually. */
2798 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2799 "Truncated zstd file body");
2800 return (ARCHIVE_FATAL);
2801 }
2802
2803 zip_read_decrypt(zip, compressed_buff, bytes_avail,
2804 &compressed_buff, &bytes_avail, &sp);
2805
2806 /* Setup buffer boundaries */
2807 in.src = compressed_buff;
2808 in.size = bytes_avail;
2809 in.pos = 0;
2810 out = (ZSTD_outBuffer) { zip->uncompressed_buffer, zip->uncompressed_buffer_size, 0 };
2811
2812 /* Perform the decompression. */
2813 ret = ZSTD_decompressStream(zip->zstdstream, &out, &in);
2814 if (ZSTD_isError(ret)) {
2815 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2816 "Error during zstd decompression: %s",
2817 ZSTD_getErrorName(ret));
2818 return (ARCHIVE_FATAL);
2819 }
2820 /* End of stream handling for zips with ZIP_LENGTH_AT_END flag */
2821 if (ret == 0 && (zip->entry->zip_flags & ZIP_LENGTH_AT_END)) {
2822 zip->end_of_entry = 1;
2823 }
2824
2825 /* Check end of the stream. */
2826 if (ret == 0) {
2827 if ((in.pos == in.size) && (out.pos < out.size)) {
2828 zip->end_of_entry = 1;
2829 ZSTD_freeDStream(zip->zstdstream);
2830 zip->zstdstream_valid = 0;
2831 }
2832 }
2833
2834 /* Update the pointers so decompressor can continue decoding. */
2835 to_consume = in.pos;
2836 __archive_read_consume(a, to_consume);
2837
2838 total_out = out.pos;
2839
2840 zip->entry_bytes_remaining -= to_consume;
2841 zip->entry_compressed_bytes_read += to_consume;
2842 zip->entry_uncompressed_bytes_read += total_out;
2843
2844 zip_read_decrypt_update(zip, to_consume, sp);
2845
2846 if (zip->end_of_entry && zip->hctx_valid) {
2847 r = check_authentication_code(a, NULL);
2848 if (r != ARCHIVE_OK)
2849 return r;
2850 }
2851
2852 /* Give libarchive its due. */
2853 *size = (size_t)total_out;
2854 *buff = zip->uncompressed_buffer;
2855
2856 return ARCHIVE_OK;
2857 }
2858 #endif
2859
2860 #ifdef HAVE_ZLIB_H
2861 static int
zip_deflate_init(struct archive_read * a,struct zip * zip)2862 zip_deflate_init(struct archive_read *a, struct zip *zip)
2863 {
2864 int r;
2865
2866 /* If we haven't yet read any data, initialize the decompressor. */
2867 if (!zip->decompress_init) {
2868 if (zip->stream_valid)
2869 r = inflateReset(&zip->stream);
2870 else
2871 r = inflateInit2(&zip->stream,
2872 -15 /* Don't check for zlib header */);
2873 if (r != Z_OK) {
2874 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2875 "Can't initialize ZIP decompression");
2876 return (ARCHIVE_FATAL);
2877 }
2878 /* Stream structure has been set up. */
2879 zip->stream_valid = 1;
2880 /* We've initialized decompression for this stream. */
2881 zip->decompress_init = 1;
2882 }
2883 return (ARCHIVE_OK);
2884 }
2885
2886 static int
zip_read_data_deflate(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)2887 zip_read_data_deflate(struct archive_read *a, const void **buff,
2888 size_t *size, int64_t *offset)
2889 {
2890 struct zip *zip = a->format->data;
2891 ssize_t bytes_avail, to_consume = 0;
2892 const void *compressed_buff;
2893 const void *sp;
2894 int r;
2895
2896 (void)offset; /* UNUSED */
2897
2898 /* If the buffer hasn't been allocated, allocate it now. */
2899 if (zip->uncompressed_buffer == NULL) {
2900 zip->uncompressed_buffer_size = 256 * 1024;
2901 zip->uncompressed_buffer
2902 = malloc(zip->uncompressed_buffer_size);
2903 if (zip->uncompressed_buffer == NULL) {
2904 archive_set_error(&a->archive, ENOMEM,
2905 "No memory for ZIP decompression");
2906 return (ARCHIVE_FATAL);
2907 }
2908 }
2909
2910 r = zip_deflate_init(a, zip);
2911 if (r != ARCHIVE_OK)
2912 return (r);
2913
2914 /*
2915 * Note: '1' here is a performance optimization.
2916 * Recall that the decompression layer returns a count of
2917 * available bytes; asking for more than that forces the
2918 * decompressor to combine reads by copying data.
2919 */
2920 compressed_buff = sp = __archive_read_ahead(a, 1, &bytes_avail);
2921 if (0 == (zip->entry->zip_flags & ZIP_LENGTH_AT_END)
2922 && bytes_avail > zip->entry_bytes_remaining) {
2923 bytes_avail = (ssize_t)zip->entry_bytes_remaining;
2924 }
2925 if (bytes_avail < 0) {
2926 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
2927 "Truncated ZIP file body");
2928 return (ARCHIVE_FATAL);
2929 }
2930
2931 zip_read_decrypt(zip, compressed_buff, bytes_avail,
2932 &compressed_buff, &bytes_avail, &sp);
2933
2934 /*
2935 * A bug in zlib.h: stream.next_in should be marked 'const'
2936 * but isn't (the library never alters data through the
2937 * next_in pointer, only reads it). The result: this ugly
2938 * cast to remove 'const'.
2939 */
2940 zip->stream.next_in = (Bytef *)(uintptr_t)(const void *)compressed_buff;
2941 zip->stream.avail_in = (uInt)bytes_avail;
2942 zip->stream.total_in = 0;
2943 zip->stream.next_out = zip->uncompressed_buffer;
2944 zip->stream.avail_out = (uInt)zip->uncompressed_buffer_size;
2945 zip->stream.total_out = 0;
2946
2947 r = inflate(&zip->stream, 0);
2948 switch (r) {
2949 case Z_OK:
2950 break;
2951 case Z_STREAM_END:
2952 zip->end_of_entry = 1;
2953 break;
2954 case Z_MEM_ERROR:
2955 archive_set_error(&a->archive, ENOMEM,
2956 "Out of memory for ZIP decompression");
2957 return (ARCHIVE_FATAL);
2958 default:
2959 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
2960 "ZIP decompression failed (%d)", r);
2961 return (ARCHIVE_FATAL);
2962 }
2963
2964 /* Consume as much as the compressor actually used. */
2965 to_consume = zip->stream.total_in;
2966 __archive_read_consume(a, to_consume);
2967 zip->entry_bytes_remaining -= to_consume;
2968 zip->entry_compressed_bytes_read += to_consume;
2969 zip->entry_uncompressed_bytes_read += zip->stream.total_out;
2970
2971 zip_read_decrypt_update(zip, to_consume, sp);
2972
2973 if (zip->end_of_entry && zip->hctx_valid) {
2974 r = check_authentication_code(a, NULL);
2975 if (r != ARCHIVE_OK) {
2976 return r;
2977 }
2978 }
2979
2980 *size = zip->stream.total_out;
2981 *buff = zip->uncompressed_buffer;
2982
2983 return (ARCHIVE_OK);
2984 }
2985 #endif
2986
2987 static int
read_decryption_header(struct archive_read * a)2988 read_decryption_header(struct archive_read *a)
2989 {
2990 struct zip *zip = a->format->data;
2991 const char *p;
2992 unsigned int remaining_size;
2993 unsigned int ts;
2994
2995 /*
2996 * Read an initialization vector data field.
2997 */
2998 p = __archive_read_ahead(a, 2, NULL);
2999 if (p == NULL)
3000 goto truncated;
3001 ts = zip->iv_size;
3002 zip->iv_size = archive_le16dec(p);
3003 __archive_read_consume(a, 2);
3004 if (ts < zip->iv_size) {
3005 free(zip->iv);
3006 zip->iv = NULL;
3007 }
3008 p = __archive_read_ahead(a, zip->iv_size, NULL);
3009 if (p == NULL)
3010 goto truncated;
3011 if (zip->iv == NULL) {
3012 zip->iv = malloc(zip->iv_size);
3013 if (zip->iv == NULL)
3014 goto nomem;
3015 }
3016 memcpy(zip->iv, p, zip->iv_size);
3017 __archive_read_consume(a, zip->iv_size);
3018
3019 /*
3020 * Read a size of remaining decryption header field.
3021 */
3022 p = __archive_read_ahead(a, 14, NULL);
3023 if (p == NULL)
3024 goto truncated;
3025 remaining_size = archive_le32dec(p);
3026 if (remaining_size < 16 || remaining_size > (1 << 18))
3027 goto corrupted;
3028
3029 /* Check if format version is supported. */
3030 if (archive_le16dec(p+4) != 3) {
3031 archive_set_error(&a->archive,
3032 ARCHIVE_ERRNO_FILE_FORMAT,
3033 "Unsupported encryption format version: %u",
3034 archive_le16dec(p+4));
3035 return (ARCHIVE_FAILED);
3036 }
3037
3038 /*
3039 * Read an encryption algorithm field.
3040 */
3041 zip->alg_id = archive_le16dec(p+6);
3042 switch (zip->alg_id) {
3043 case 0x6601:/* DES */
3044 case 0x6602:/* RC2 */
3045 case 0x6603:/* 3DES 168 */
3046 case 0x6609:/* 3DES 112 */
3047 case 0x660E:/* AES 128 */
3048 case 0x660F:/* AES 192 */
3049 case 0x6610:/* AES 256 */
3050 case 0x6702:/* RC2 (version >= 5.2) */
3051 case 0x6720:/* Blowfish */
3052 case 0x6721:/* Twofish */
3053 case 0x6801:/* RC4 */
3054 /* Supported encryption algorithm. */
3055 break;
3056 default:
3057 archive_set_error(&a->archive,
3058 ARCHIVE_ERRNO_FILE_FORMAT,
3059 "Unknown encryption algorithm: %u", zip->alg_id);
3060 return (ARCHIVE_FAILED);
3061 }
3062
3063 /*
3064 * Read a bit length field.
3065 */
3066 zip->bit_len = archive_le16dec(p+8);
3067
3068 /*
3069 * Read a flags field.
3070 */
3071 zip->flags = archive_le16dec(p+10);
3072 switch (zip->flags & 0xf000) {
3073 case 0x0001: /* Password is required to decrypt. */
3074 case 0x0002: /* Certificates only. */
3075 case 0x0003: /* Password or certificate required to decrypt. */
3076 break;
3077 default:
3078 archive_set_error(&a->archive,
3079 ARCHIVE_ERRNO_FILE_FORMAT,
3080 "Unknown encryption flag: %u", zip->flags);
3081 return (ARCHIVE_FAILED);
3082 }
3083 if ((zip->flags & 0xf000) == 0 ||
3084 (zip->flags & 0xf000) == 0x4000) {
3085 archive_set_error(&a->archive,
3086 ARCHIVE_ERRNO_FILE_FORMAT,
3087 "Unknown encryption flag: %u", zip->flags);
3088 return (ARCHIVE_FAILED);
3089 }
3090
3091 /*
3092 * Read an encrypted random data field.
3093 */
3094 ts = zip->erd_size;
3095 zip->erd_size = archive_le16dec(p+12);
3096 __archive_read_consume(a, 14);
3097 if ((zip->erd_size & 0xf) != 0 ||
3098 (zip->erd_size + 16) > remaining_size ||
3099 (zip->erd_size + 16) < zip->erd_size)
3100 goto corrupted;
3101
3102 if (ts < zip->erd_size) {
3103 free(zip->erd);
3104 zip->erd = NULL;
3105 }
3106 p = __archive_read_ahead(a, zip->erd_size, NULL);
3107 if (p == NULL)
3108 goto truncated;
3109 if (zip->erd == NULL) {
3110 zip->erd = malloc(zip->erd_size);
3111 if (zip->erd == NULL)
3112 goto nomem;
3113 }
3114 memcpy(zip->erd, p, zip->erd_size);
3115 __archive_read_consume(a, zip->erd_size);
3116
3117 /*
3118 * Read a reserved data field.
3119 */
3120 p = __archive_read_ahead(a, 4, NULL);
3121 if (p == NULL)
3122 goto truncated;
3123 /* Reserved data size should be zero. */
3124 if (archive_le32dec(p) != 0)
3125 goto corrupted;
3126 __archive_read_consume(a, 4);
3127
3128 /*
3129 * Read a password validation data field.
3130 */
3131 p = __archive_read_ahead(a, 2, NULL);
3132 if (p == NULL)
3133 goto truncated;
3134 ts = zip->v_size;
3135 zip->v_size = archive_le16dec(p);
3136 __archive_read_consume(a, 2);
3137 if ((zip->v_size & 0x0f) != 0 ||
3138 (zip->erd_size + zip->v_size + 16) > remaining_size ||
3139 (zip->erd_size + zip->v_size + 16) < (zip->erd_size + zip->v_size))
3140 goto corrupted;
3141 if (ts < zip->v_size) {
3142 free(zip->v_data);
3143 zip->v_data = NULL;
3144 }
3145 p = __archive_read_ahead(a, zip->v_size, NULL);
3146 if (p == NULL)
3147 goto truncated;
3148 if (zip->v_data == NULL) {
3149 zip->v_data = malloc(zip->v_size);
3150 if (zip->v_data == NULL)
3151 goto nomem;
3152 }
3153 memcpy(zip->v_data, p, zip->v_size);
3154 __archive_read_consume(a, zip->v_size);
3155
3156 p = __archive_read_ahead(a, 4, NULL);
3157 if (p == NULL)
3158 goto truncated;
3159 zip->v_crc32 = archive_le32dec(p);
3160 __archive_read_consume(a, 4);
3161
3162 /*return (ARCHIVE_OK);
3163 * This is not fully implemented yet.*/
3164 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
3165 "Encrypted file is unsupported");
3166 return (ARCHIVE_FAILED);
3167 truncated:
3168 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
3169 "Truncated ZIP file data");
3170 return (ARCHIVE_FATAL);
3171 corrupted:
3172 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
3173 "Corrupted ZIP file data");
3174 return (ARCHIVE_FATAL);
3175 nomem:
3176 archive_set_error(&a->archive, ENOMEM,
3177 "No memory for ZIP decryption");
3178 return (ARCHIVE_FATAL);
3179 }
3180
3181 static int
zip_alloc_decryption_buffer(struct archive_read * a)3182 zip_alloc_decryption_buffer(struct archive_read *a)
3183 {
3184 struct zip *zip = a->format->data;
3185 size_t bs = 256 * 1024;
3186
3187 if (zip->decrypted_buffer == NULL) {
3188 zip->decrypted_buffer_size = bs;
3189 zip->decrypted_buffer = malloc(bs);
3190 if (zip->decrypted_buffer == NULL) {
3191 archive_set_error(&a->archive, ENOMEM,
3192 "No memory for ZIP decryption");
3193 return (ARCHIVE_FATAL);
3194 }
3195 }
3196 zip->decrypted_ptr = zip->decrypted_buffer;
3197 return (ARCHIVE_OK);
3198 }
3199
3200 static int
init_traditional_PKWARE_decryption(struct archive_read * a)3201 init_traditional_PKWARE_decryption(struct archive_read *a)
3202 {
3203 struct zip *zip = a->format->data;
3204 const void *p;
3205 int retry;
3206 int r;
3207
3208 if (zip->tctx_valid)
3209 return (ARCHIVE_OK);
3210
3211 /*
3212 Read the 12 bytes encryption header stored at
3213 the start of the data area.
3214 */
3215 #define ENC_HEADER_SIZE 12
3216 if (0 == (zip->entry->zip_flags & ZIP_LENGTH_AT_END)
3217 && zip->entry_bytes_remaining < ENC_HEADER_SIZE) {
3218 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
3219 "Truncated Zip encrypted body: only %jd bytes available",
3220 (intmax_t)zip->entry_bytes_remaining);
3221 return (ARCHIVE_FATAL);
3222 }
3223
3224 p = __archive_read_ahead(a, ENC_HEADER_SIZE, NULL);
3225 if (p == NULL) {
3226 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
3227 "Truncated ZIP file data");
3228 return (ARCHIVE_FATAL);
3229 }
3230
3231 for (retry = 0;; retry++) {
3232 const char *passphrase;
3233 uint8_t crcchk;
3234
3235 passphrase = __archive_read_next_passphrase(a);
3236 if (passphrase == NULL) {
3237 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3238 (retry > 0)?
3239 "Incorrect passphrase":
3240 "Passphrase required for this entry");
3241 return (ARCHIVE_FAILED);
3242 }
3243
3244 /*
3245 * Initialize ctx for Traditional PKWARE Decryption.
3246 */
3247 r = trad_enc_init(&zip->tctx, passphrase, strlen(passphrase),
3248 p, ENC_HEADER_SIZE, &crcchk);
3249 if (r == 0 && crcchk == zip->entry->decdat)
3250 break;/* The passphrase is OK. */
3251 if (retry > 10000) {
3252 /* Avoid infinity loop. */
3253 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3254 "Too many incorrect passphrases");
3255 return (ARCHIVE_FAILED);
3256 }
3257 }
3258
3259 __archive_read_consume(a, ENC_HEADER_SIZE);
3260 zip->tctx_valid = 1;
3261 if (0 == (zip->entry->zip_flags & ZIP_LENGTH_AT_END)) {
3262 zip->entry_bytes_remaining -= ENC_HEADER_SIZE;
3263 }
3264 /*zip->entry_uncompressed_bytes_read += ENC_HEADER_SIZE;*/
3265 zip->entry_compressed_bytes_read += ENC_HEADER_SIZE;
3266 zip->decrypted_bytes_remaining = 0;
3267
3268 return (zip_alloc_decryption_buffer(a));
3269 #undef ENC_HEADER_SIZE
3270 }
3271
3272 static int
init_WinZip_AES_decryption(struct archive_read * a)3273 init_WinZip_AES_decryption(struct archive_read *a)
3274 {
3275 struct zip *zip = a->format->data;
3276 const void *p;
3277 const uint8_t *pv;
3278 size_t key_len, salt_len;
3279 uint8_t derived_key[MAX_DERIVED_KEY_BUF_SIZE];
3280 int retry;
3281 int r;
3282
3283 if (zip->cctx_valid || zip->hctx_valid)
3284 return (ARCHIVE_OK);
3285
3286 switch (zip->entry->aes_extra.strength) {
3287 case 1: salt_len = 8; key_len = 16; break;
3288 case 2: salt_len = 12; key_len = 24; break;
3289 case 3: salt_len = 16; key_len = 32; break;
3290 default: goto corrupted;
3291 }
3292 p = __archive_read_ahead(a, salt_len + 2, NULL);
3293 if (p == NULL)
3294 goto truncated;
3295
3296 for (retry = 0;; retry++) {
3297 const char *passphrase;
3298
3299 passphrase = __archive_read_next_passphrase(a);
3300 if (passphrase == NULL) {
3301 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3302 (retry > 0)?
3303 "Incorrect passphrase":
3304 "Passphrase required for this entry");
3305 return (ARCHIVE_FAILED);
3306 }
3307 memset(derived_key, 0, sizeof(derived_key));
3308 r = archive_pbkdf2_sha1(passphrase, strlen(passphrase),
3309 p, salt_len, 1000, derived_key, key_len * 2 + 2);
3310 if (r != 0) {
3311 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3312 r == CRYPTOR_STUB_FUNCTION ? "Decryption is unsupported due "
3313 "to lack of crypto library" : "Failed to process passphrase");
3314 return (ARCHIVE_FAILED);
3315 }
3316
3317 /* Check password verification value. */
3318 pv = ((const uint8_t *)p) + salt_len;
3319 if (derived_key[key_len * 2] == pv[0] &&
3320 derived_key[key_len * 2 + 1] == pv[1])
3321 break;/* The passphrase is OK. */
3322 if (retry > 10000) {
3323 /* Avoid infinity loop. */
3324 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3325 "Too many incorrect passphrases");
3326 return (ARCHIVE_FAILED);
3327 }
3328 }
3329
3330 r = archive_decrypto_aes_ctr_init(&zip->cctx, derived_key, key_len);
3331 if (r != 0) {
3332 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3333 "Decryption is unsupported due to lack of crypto library");
3334 return (ARCHIVE_FAILED);
3335 }
3336 r = archive_hmac_sha1_init(&zip->hctx, derived_key + key_len, key_len);
3337 if (r != 0) {
3338 archive_decrypto_aes_ctr_release(&zip->cctx);
3339 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3340 "Failed to initialize HMAC-SHA1");
3341 return (ARCHIVE_FAILED);
3342 }
3343 zip->cctx_valid = zip->hctx_valid = 1;
3344 __archive_read_consume(a, salt_len + 2);
3345 zip->entry_bytes_remaining -= salt_len + 2 + AUTH_CODE_SIZE;
3346 if (0 == (zip->entry->zip_flags & ZIP_LENGTH_AT_END)
3347 && zip->entry_bytes_remaining < 0)
3348 goto corrupted;
3349 zip->entry_compressed_bytes_read += salt_len + 2 + AUTH_CODE_SIZE;
3350 zip->decrypted_bytes_remaining = 0;
3351
3352 zip->entry->compression = zip->entry->aes_extra.compression;
3353 return (zip_alloc_decryption_buffer(a));
3354
3355 truncated:
3356 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
3357 "Truncated ZIP file data");
3358 return (ARCHIVE_FATAL);
3359 corrupted:
3360 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
3361 "Corrupted ZIP file data");
3362 return (ARCHIVE_FATAL);
3363 }
3364
3365 static int
archive_read_format_zip_read_data(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)3366 archive_read_format_zip_read_data(struct archive_read *a,
3367 const void **buff, size_t *size, int64_t *offset)
3368 {
3369 struct zip *zip = a->format->data;
3370 int r;
3371
3372 if (zip->has_encrypted_entries ==
3373 ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW) {
3374 zip->has_encrypted_entries = 0;
3375 }
3376
3377 *offset = zip->entry_uncompressed_bytes_read;
3378 *size = 0;
3379 *buff = NULL;
3380
3381 /* If we hit end-of-entry last time, return ARCHIVE_EOF. */
3382 if (zip->end_of_entry)
3383 return (ARCHIVE_EOF);
3384
3385 /* Return EOF immediately if this is a non-regular file. */
3386 if (AE_IFREG != (zip->entry->mode & AE_IFMT))
3387 return (ARCHIVE_EOF);
3388
3389 __archive_read_consume(a, zip->unconsumed);
3390 zip->unconsumed = 0;
3391
3392 if (zip->init_decryption) {
3393 zip->has_encrypted_entries = 1;
3394 if (zip->entry->zip_flags & ZIP_STRONG_ENCRYPTED)
3395 r = read_decryption_header(a);
3396 else if (zip->entry->compression == WINZIP_AES_ENCRYPTION)
3397 r = init_WinZip_AES_decryption(a);
3398 else
3399 r = init_traditional_PKWARE_decryption(a);
3400 if (r != ARCHIVE_OK)
3401 return (r);
3402 zip->init_decryption = 0;
3403 }
3404
3405 switch(zip->entry->compression) {
3406 case 0: /* No compression. */
3407 r = zip_read_data_none(a, buff, size, offset);
3408 break;
3409 #ifdef HAVE_BZLIB_H
3410 case 12: /* ZIPx bzip2 compression. */
3411 r = zip_read_data_zipx_bzip2(a, buff, size, offset);
3412 break;
3413 #endif
3414 #if HAVE_LZMA_H && HAVE_LIBLZMA
3415 case 14: /* ZIPx LZMA compression. */
3416 r = zip_read_data_zipx_lzma_alone(a, buff, size, offset);
3417 break;
3418 case 95: /* ZIPx XZ compression. */
3419 r = zip_read_data_zipx_xz(a, buff, size, offset);
3420 break;
3421 #endif
3422 #if HAVE_ZSTD_H && HAVE_LIBZSTD
3423 case 93: /* ZIPx Zstd compression. */
3424 r = zip_read_data_zipx_zstd(a, buff, size, offset);
3425 break;
3426 #endif
3427 /* PPMd support is built-in, so we don't need any #if guards. */
3428 case 98: /* ZIPx PPMd compression. */
3429 r = zip_read_data_zipx_ppmd(a, buff, size, offset);
3430 break;
3431
3432 #ifdef HAVE_ZLIB_H
3433 case 8: /* Deflate compression. */
3434 r = zip_read_data_deflate(a, buff, size, offset);
3435 break;
3436 #endif
3437 default: /* Unsupported compression. */
3438 /* Return a warning. */
3439 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
3440 "Unsupported ZIP compression method (%d: %s)",
3441 zip->entry->compression, compression_name(zip->entry->compression));
3442 /* We can't decompress this entry, but we will
3443 * be able to skip() it and try the next entry. */
3444 return (ARCHIVE_FAILED);
3445 }
3446 if (r != ARCHIVE_OK)
3447 return (r);
3448
3449 /*
3450 * FAIL if there are more uncompressed bytes than were
3451 * initially advertised. The end-of-entry check below also
3452 * compares these values, but only once decoding reaches its
3453 * own natural end. Fail the entry so a caller relying on
3454 * archive_entry_size() as a hard boundary is never misled;
3455 * later entries in the archive can still be read normally.
3456 */
3457 if (*size > 0 && zip_entry_size_is_set(zip->entry) &&
3458 zip->entry_uncompressed_bytes_read > zip->entry->uncompressed_size) {
3459 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3460 "ZIP uncompressed data is larger than the declared "
3461 "entry size (read at least %jd, expected %jd)",
3462 (intmax_t)zip->entry_uncompressed_bytes_read,
3463 (intmax_t)zip->entry->uncompressed_size);
3464 return (ARCHIVE_FAILED);
3465 }
3466
3467 if (*size > 0) {
3468 zip->computed_crc32 = zip->crc32func(zip->computed_crc32, *buff,
3469 (unsigned)*size);
3470 }
3471 /* If we hit the end, swallow any end-of-data marker and
3472 * verify the final check values. */
3473 if (zip->end_of_entry) {
3474 consume_end_of_file_marker(a, zip);
3475
3476 /* Check computed CRC against header */
3477 if ((!zip->hctx_valid ||
3478 zip->entry->aes_extra.vendor != AES_VENDOR_AE_2) &&
3479 zip->entry->crc32 != zip->computed_crc32
3480 && !zip->ignore_crc32) {
3481 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3482 "ZIP bad CRC: 0x%lx should be 0x%lx",
3483 (unsigned long)zip->computed_crc32,
3484 (unsigned long)zip->entry->crc32);
3485 return (ARCHIVE_FAILED);
3486 }
3487 /* Check file size against header. */
3488 if (zip->entry->compressed_size !=
3489 zip->entry_compressed_bytes_read) {
3490 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3491 "ZIP compressed data is wrong size "
3492 "(read %jd, expected %jd)",
3493 (intmax_t)zip->entry_compressed_bytes_read,
3494 (intmax_t)zip->entry->compressed_size);
3495 return (ARCHIVE_FAILED);
3496 }
3497 if (zip->entry->uncompressed_size !=
3498 zip->entry_uncompressed_bytes_read) {
3499 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3500 "ZIP uncompressed data is wrong size "
3501 "(read %jd, expected %jd)",
3502 (intmax_t)zip->entry_uncompressed_bytes_read,
3503 (intmax_t)zip->entry->uncompressed_size);
3504 return (ARCHIVE_FAILED);
3505 }
3506 }
3507
3508 return (ARCHIVE_OK);
3509 }
3510
3511 static int
archive_read_format_zip_cleanup(struct archive_read * a)3512 archive_read_format_zip_cleanup(struct archive_read *a)
3513 {
3514 struct zip *zip = a->format->data;
3515 struct zip_entry *zip_entry, *next_zip_entry;
3516
3517 #ifdef HAVE_ZLIB_H
3518 if (zip->stream_valid)
3519 inflateEnd(&zip->stream);
3520 #endif
3521
3522 #if HAVE_LZMA_H && HAVE_LIBLZMA
3523 if (zip->zipx_lzma_valid) {
3524 lzma_end(&zip->zipx_lzma_stream);
3525 }
3526 #endif
3527
3528 #ifdef HAVE_BZLIB_H
3529 if (zip->bzstream_valid) {
3530 BZ2_bzDecompressEnd(&zip->bzstream);
3531 }
3532 #endif
3533
3534 #if HAVE_ZSTD_H && HAVE_LIBZSTD
3535 if (zip->zstdstream_valid) {
3536 ZSTD_freeDStream(zip->zstdstream);
3537 }
3538 #endif
3539
3540 free(zip->uncompressed_buffer);
3541
3542 if (zip->ppmd8_valid)
3543 __archive_ppmd8_functions.Ppmd8_Free(&zip->ppmd8);
3544
3545 if (zip->zip_entries) {
3546 zip_entry = zip->zip_entries;
3547 while (zip_entry != NULL) {
3548 next_zip_entry = zip_entry->next;
3549 archive_string_free(&zip_entry->rsrcname);
3550 free(zip_entry);
3551 zip_entry = next_zip_entry;
3552 }
3553 }
3554 free(zip->decrypted_buffer);
3555 if (zip->cctx_valid)
3556 archive_decrypto_aes_ctr_release(&zip->cctx);
3557 if (zip->hctx_valid)
3558 archive_hmac_sha1_cleanup(&zip->hctx);
3559 free(zip->iv);
3560 free(zip->erd);
3561 free(zip->v_data);
3562 archive_string_free(&zip->format_name);
3563 free(zip);
3564 a->format->data = NULL;
3565 return (ARCHIVE_OK);
3566 }
3567
3568 static int
archive_read_format_zip_has_encrypted_entries(struct archive_read * _a)3569 archive_read_format_zip_has_encrypted_entries(struct archive_read *_a)
3570 {
3571 if (_a && _a->format) {
3572 struct zip *zip = _a->format->data;
3573 if (zip) {
3574 return zip->has_encrypted_entries;
3575 }
3576 }
3577 return ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW;
3578 }
3579
3580 static int
archive_read_format_zip_options(struct archive_read * a,const char * key,const char * val)3581 archive_read_format_zip_options(struct archive_read *a,
3582 const char *key, const char *val)
3583 {
3584 struct zip *zip = a->format->data;
3585 int ret = ARCHIVE_FAILED;
3586
3587 if (strcmp(key, "compat-2x") == 0) {
3588 /* Handle filenames as libarchive 2.x */
3589 zip->init_default_conversion = (val != NULL) ? 1 : 0;
3590 return (ARCHIVE_OK);
3591 } else if (strcmp(key, "hdrcharset") == 0) {
3592 if (val == NULL || val[0] == 0)
3593 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
3594 "zip: hdrcharset option needs a character-set name"
3595 );
3596 else {
3597 zip->sconv = archive_string_conversion_from_charset(
3598 &a->archive, val, 0);
3599 if (zip->sconv != NULL) {
3600 if (strcmp(val, "UTF-8") == 0)
3601 zip->sconv_utf8 = zip->sconv;
3602 ret = ARCHIVE_OK;
3603 } else
3604 ret = ARCHIVE_FATAL;
3605 }
3606 return (ret);
3607 } else if (strcmp(key, "ignorecrc32") == 0) {
3608 /* Mostly useful for testing. */
3609 if (val == NULL || val[0] == 0) {
3610 zip->crc32func = real_crc32;
3611 zip->ignore_crc32 = 0;
3612 } else {
3613 zip->crc32func = fake_crc32;
3614 zip->ignore_crc32 = 1;
3615 }
3616 return (ARCHIVE_OK);
3617 } else if (strcmp(key, "mac-ext") == 0) {
3618 zip->process_mac_extensions = (val != NULL && val[0] != 0);
3619 return (ARCHIVE_OK);
3620 }
3621
3622 /* Note: The "warn" return is just to inform the options
3623 * supervisor that we didn't handle it. It will generate
3624 * a suitable error if no one used this option. */
3625 return (ARCHIVE_WARN);
3626 }
3627
3628 int
archive_read_support_format_zip(struct archive * a)3629 archive_read_support_format_zip(struct archive *a)
3630 {
3631 int r;
3632 r = archive_read_support_format_zip_streamable(a);
3633 if (r != ARCHIVE_OK)
3634 return r;
3635 return (archive_read_support_format_zip_seekable(a));
3636 }
3637
3638 /* ------------------------------------------------------------------------ */
3639
3640 /*
3641 * Streaming-mode support
3642 */
3643
3644
3645 static int
archive_read_support_format_zip_capabilities_streamable(struct archive_read * a)3646 archive_read_support_format_zip_capabilities_streamable(struct archive_read * a)
3647 {
3648 (void)a; /* UNUSED */
3649 return (ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_DATA |
3650 ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_METADATA);
3651 }
3652
3653 static int
archive_read_format_zip_streamable_bid(struct archive_read * a,int best_bid)3654 archive_read_format_zip_streamable_bid(struct archive_read *a, int best_bid)
3655 {
3656 const char *p;
3657
3658 (void)best_bid; /* UNUSED */
3659
3660 if ((p = __archive_read_ahead(a, 4, NULL)) == NULL)
3661 return (-1);
3662
3663 /*
3664 * Bid of 29 here comes from:
3665 * + 16 bits for "PK",
3666 * + next 16-bit field has 6 options so contributes
3667 * about 16 - log_2(6) ~= 16 - 2.6 ~= 13 bits
3668 *
3669 * So we've effectively verified ~29 total bits of check data.
3670 */
3671 if (p[0] == 'P' && p[1] == 'K') {
3672 if ((p[2] == '\001' && p[3] == '\002')
3673 || (p[2] == '\003' && p[3] == '\004')
3674 || (p[2] == '\005' && p[3] == '\006')
3675 || (p[2] == '\006' && p[3] == '\006')
3676 || (p[2] == '\007' && p[3] == '\010')
3677 || (p[2] == '0' && p[3] == '0'))
3678 return (29);
3679 }
3680
3681 /* TODO: It's worth looking ahead a little bit for a valid
3682 * PK signature. In particular, that would make it possible
3683 * to read some UUEncoded SFX files or SFX files coming from
3684 * a network socket. */
3685
3686 return (0);
3687 }
3688
3689 static int
archive_read_format_zip_streamable_read_header(struct archive_read * a,struct archive_entry * entry)3690 archive_read_format_zip_streamable_read_header(struct archive_read *a,
3691 struct archive_entry *entry)
3692 {
3693 struct zip *zip = a->format->data;
3694
3695 a->archive.archive_format = ARCHIVE_FORMAT_ZIP;
3696 if (a->archive.archive_format_name == NULL)
3697 a->archive.archive_format_name = "ZIP";
3698
3699 /*
3700 * It should be sufficient to call archive_read_next_header() for
3701 * a reader to determine if an entry is encrypted or not. If the
3702 * encryption of an entry is only detectable when calling
3703 * archive_read_data(), so be it. We'll do the same check there
3704 * as well.
3705 */
3706 if (zip->has_encrypted_entries ==
3707 ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW)
3708 zip->has_encrypted_entries = 0;
3709
3710 /* Make sure we have a zip_entry structure to use. */
3711 if (zip->zip_entries == NULL) {
3712 zip->zip_entries = malloc(sizeof(struct zip_entry));
3713 if (zip->zip_entries == NULL) {
3714 archive_set_error(&a->archive, ENOMEM,
3715 "Out of memory");
3716 return ARCHIVE_FATAL;
3717 }
3718 }
3719 zip->entry = zip->zip_entries;
3720 memset(zip->entry, 0, sizeof(struct zip_entry));
3721
3722 if (zip->cctx_valid)
3723 archive_decrypto_aes_ctr_release(&zip->cctx);
3724 if (zip->hctx_valid)
3725 archive_hmac_sha1_cleanup(&zip->hctx);
3726 zip->tctx_valid = zip->cctx_valid = zip->hctx_valid = 0;
3727 __archive_read_reset_passphrase(a);
3728
3729 /* Search ahead for the next local file header. */
3730 __archive_read_consume(a, zip->unconsumed);
3731 zip->unconsumed = 0;
3732 for (;;) {
3733 int64_t skipped = 0;
3734 const char *p, *end;
3735 ssize_t bytes;
3736
3737 p = __archive_read_ahead(a, 4, &bytes);
3738 if (p == NULL)
3739 return (ARCHIVE_FATAL);
3740 end = p + bytes;
3741
3742 while (p + 4 <= end) {
3743 if (p[0] == 'P' && p[1] == 'K') {
3744 if (p[2] == '\003' && p[3] == '\004') {
3745 /* Regular file entry. */
3746 __archive_read_consume(a, skipped);
3747 return zip_read_local_file_header(a,
3748 entry, zip);
3749 }
3750
3751 /*
3752 * TODO: We cannot restore permissions
3753 * based only on the local file headers.
3754 * Consider scanning the central
3755 * directory and returning additional
3756 * entries for at least directories.
3757 * This would allow us to properly set
3758 * directory permissions.
3759 *
3760 * This won't help us fix symlinks
3761 * and may not help with regular file
3762 * permissions, either. <sigh>
3763 */
3764 if (p[2] == '\001' && p[3] == '\002') {
3765 return (ARCHIVE_EOF);
3766 }
3767
3768 /* End of central directory? Must be an
3769 * empty archive. */
3770 if ((p[2] == '\005' && p[3] == '\006')
3771 || (p[2] == '\006' && p[3] == '\006'))
3772 return (ARCHIVE_EOF);
3773 }
3774 ++p;
3775 ++skipped;
3776 }
3777 __archive_read_consume(a, skipped);
3778 }
3779 }
3780
3781 static int
archive_read_format_zip_read_data_skip_streamable(struct archive_read * a)3782 archive_read_format_zip_read_data_skip_streamable(struct archive_read *a)
3783 {
3784 struct zip *zip = a->format->data;
3785 int64_t bytes_skipped;
3786
3787 bytes_skipped = __archive_read_consume(a, zip->unconsumed);
3788 zip->unconsumed = 0;
3789 if (bytes_skipped < 0)
3790 return (ARCHIVE_FATAL);
3791
3792 /* If we've already read to end of data, we're done. */
3793 if (zip->end_of_entry)
3794 return (ARCHIVE_OK);
3795
3796 /* So we know we're streaming... */
3797 if (0 == (zip->entry->zip_flags & ZIP_LENGTH_AT_END)
3798 || zip->entry->compressed_size > 0) {
3799 /* We know the compressed length, so we can just skip. */
3800 bytes_skipped = __archive_read_consume(a,
3801 zip->entry_bytes_remaining);
3802 if (bytes_skipped < 0)
3803 return (ARCHIVE_FATAL);
3804 return (ARCHIVE_OK);
3805 }
3806
3807 if (zip->init_decryption) {
3808 int r;
3809
3810 zip->has_encrypted_entries = 1;
3811 if (zip->entry->zip_flags & ZIP_STRONG_ENCRYPTED)
3812 r = read_decryption_header(a);
3813 else if (zip->entry->compression == WINZIP_AES_ENCRYPTION)
3814 r = init_WinZip_AES_decryption(a);
3815 else
3816 r = init_traditional_PKWARE_decryption(a);
3817 if (r != ARCHIVE_OK)
3818 return (r);
3819 zip->init_decryption = 0;
3820 }
3821
3822 /* We're streaming and we don't know the length. */
3823 /* If the body is compressed and we know the format, we can
3824 * find an exact end-of-entry by decompressing it. */
3825 switch (zip->entry->compression) {
3826 #ifdef HAVE_ZLIB_H
3827 case 8: /* Deflate compression. */
3828 while (!zip->end_of_entry) {
3829 int64_t offset = 0;
3830 const void *buff = NULL;
3831 size_t size = 0;
3832 int r;
3833 r = zip_read_data_deflate(a, &buff, &size, &offset);
3834 if (r != ARCHIVE_OK)
3835 return (r);
3836 }
3837 return ARCHIVE_OK;
3838 #endif
3839 default: /* Uncompressed or unknown. */
3840 /* Scan for a PK\007\010 signature. */
3841 for (;;) {
3842 const char *p, *buff;
3843 ssize_t bytes_avail;
3844 buff = __archive_read_ahead(a, 16, &bytes_avail);
3845 if (bytes_avail < 16) {
3846 archive_set_error(&a->archive,
3847 ARCHIVE_ERRNO_FILE_FORMAT,
3848 "Truncated ZIP file data");
3849 return (ARCHIVE_FATAL);
3850 }
3851 p = buff;
3852 while (p <= buff + bytes_avail - 16) {
3853 if (p[3] == 'P') { p += 3; }
3854 else if (p[3] == 'K') { p += 2; }
3855 else if (p[3] == '\007') { p += 1; }
3856 else if (p[3] == '\010' && p[2] == '\007'
3857 && p[1] == 'K' && p[0] == 'P') {
3858 if (zip->entry->flags & LA_USED_ZIP64)
3859 __archive_read_consume(a,
3860 p - buff + 24);
3861 else
3862 __archive_read_consume(a,
3863 p - buff + 16);
3864 return ARCHIVE_OK;
3865 } else { p += 4; }
3866 }
3867 __archive_read_consume(a, p - buff);
3868 }
3869 }
3870 }
3871
3872 int
archive_read_support_format_zip_streamable(struct archive * _a)3873 archive_read_support_format_zip_streamable(struct archive *_a)
3874 {
3875 struct archive_read *a = (struct archive_read *)_a;
3876 struct zip *zip;
3877 int r;
3878
3879 archive_check_magic(_a, ARCHIVE_READ_MAGIC,
3880 ARCHIVE_STATE_NEW, "archive_read_support_format_zip");
3881
3882 zip = calloc(1, sizeof(*zip));
3883 if (zip == NULL) {
3884 archive_set_error(&a->archive, ENOMEM,
3885 "Can't allocate zip data");
3886 return (ARCHIVE_FATAL);
3887 }
3888
3889 /* Streamable reader doesn't support mac extensions. */
3890 zip->process_mac_extensions = 0;
3891
3892 /*
3893 * Until enough data has been read, we cannot tell about
3894 * any encrypted entries yet.
3895 */
3896 zip->has_encrypted_entries = ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW;
3897 zip->crc32func = real_crc32;
3898
3899 r = __archive_read_register_format(a,
3900 zip,
3901 "zip",
3902 archive_read_format_zip_streamable_bid,
3903 archive_read_format_zip_options,
3904 archive_read_format_zip_streamable_read_header,
3905 archive_read_format_zip_read_data,
3906 archive_read_format_zip_read_data_skip_streamable,
3907 NULL,
3908 archive_read_format_zip_cleanup,
3909 archive_read_support_format_zip_capabilities_streamable,
3910 archive_read_format_zip_has_encrypted_entries);
3911
3912 if (r != ARCHIVE_OK)
3913 free(zip);
3914 return (ARCHIVE_OK);
3915 }
3916
3917 /* ------------------------------------------------------------------------ */
3918
3919 /*
3920 * Seeking-mode support
3921 */
3922
3923 static int
archive_read_support_format_zip_capabilities_seekable(struct archive_read * a)3924 archive_read_support_format_zip_capabilities_seekable(struct archive_read * a)
3925 {
3926 (void)a; /* UNUSED */
3927 return (ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_DATA |
3928 ARCHIVE_READ_FORMAT_CAPS_ENCRYPT_METADATA);
3929 }
3930
3931 /*
3932 * TODO: This is a performance sink because it forces the read core to
3933 * drop buffered data from the start of file, which will then have to
3934 * be re-read again if this bidder loses.
3935 *
3936 * We workaround this a little by passing in the best bid so far so
3937 * that later bidders can do nothing if they know they'll never
3938 * outbid. But we can certainly do better...
3939 */
3940 static int
read_eocd(struct zip * zip,const char * p,int64_t current_offset)3941 read_eocd(struct zip *zip, const char *p, int64_t current_offset)
3942 {
3943 uint16_t disk_num;
3944 uint32_t cd_size, cd_offset;
3945
3946 disk_num = archive_le16dec(p + 4);
3947 cd_size = archive_le32dec(p + 12);
3948 cd_offset = archive_le32dec(p + 16);
3949
3950 /* Sanity-check the EOCD we've found. */
3951
3952 /* This must be the first volume. */
3953 if (disk_num != 0)
3954 return 0;
3955 /* Central directory must be on this volume. */
3956 if (disk_num != archive_le16dec(p + 6))
3957 return 0;
3958 /* All central directory entries must be on this volume. */
3959 if (archive_le16dec(p + 10) != archive_le16dec(p + 8))
3960 return 0;
3961 /* Central directory can't extend beyond start of EOCD record. */
3962 if ((int64_t)cd_offset + cd_size > current_offset)
3963 return 0;
3964
3965 /* Save the central directory location for later use. */
3966 zip->central_directory_offset = cd_offset;
3967 zip->central_directory_offset_adjusted = current_offset - cd_size;
3968
3969 /* This is just a tiny bit higher than the maximum
3970 returned by the streaming Zip bidder. This ensures
3971 that the more accurate seeking Zip parser wins
3972 whenever seek is available. */
3973 return 32;
3974 }
3975
3976 /*
3977 * Examine Zip64 EOCD locator: If it's valid, store the information
3978 * from it.
3979 */
3980 static int
read_zip64_eocd(struct archive_read * a,struct zip * zip,const char * p)3981 read_zip64_eocd(struct archive_read *a, struct zip *zip, const char *p)
3982 {
3983 int64_t eocd64_offset;
3984 int64_t eocd64_size;
3985
3986 /* Sanity-check the locator record. */
3987
3988 /* Central dir must be on first volume. */
3989 if (archive_le32dec(p + 4) != 0)
3990 return 0;
3991 /* Must be only a single volume. */
3992 if (archive_le32dec(p + 16) != 1)
3993 return 0;
3994
3995 /* Find the Zip64 EOCD record. */
3996 eocd64_offset = archive_le64dec(p + 8);
3997 if (__archive_read_seek(a, eocd64_offset, SEEK_SET) < 0)
3998 return 0;
3999 if ((p = __archive_read_ahead(a, 56, NULL)) == NULL)
4000 return 0;
4001 /* Make sure we can read all of it. */
4002 eocd64_size = archive_le64dec(p + 4) + 12;
4003 if (eocd64_size < 56 || eocd64_size > 16384)
4004 return 0;
4005 if ((p = __archive_read_ahead(a, (size_t)eocd64_size, NULL)) == NULL)
4006 return 0;
4007
4008 /* Sanity-check the EOCD64 */
4009 if (archive_le32dec(p + 16) != 0) /* Must be disk #0 */
4010 return 0;
4011 if (archive_le32dec(p + 20) != 0) /* CD must be on disk #0 */
4012 return 0;
4013 /* CD can't be split. */
4014 if (archive_le64dec(p + 24) != archive_le64dec(p + 32))
4015 return 0;
4016
4017 /* Save the central directory offset for later use. */
4018 zip->central_directory_offset = archive_le64dec(p + 48);
4019 /* TODO: Needs scanning backwards to find the eocd64 instead of assuming */
4020 zip->central_directory_offset_adjusted = zip->central_directory_offset;
4021
4022 return 32;
4023 }
4024
4025 static int
archive_read_format_zip_seekable_bid(struct archive_read * a,int best_bid)4026 archive_read_format_zip_seekable_bid(struct archive_read *a, int best_bid)
4027 {
4028 struct zip *zip = a->format->data;
4029 int64_t file_size, current_offset;
4030 const char *p;
4031 int i, tail;
4032
4033 /* If someone has already bid more than 32, then avoid
4034 trashing the look-ahead buffers with a seek. */
4035 if (best_bid > 32)
4036 return (-1);
4037
4038 file_size = __archive_read_seek(a, 0, SEEK_END);
4039 if (file_size <= 0)
4040 return 0;
4041
4042 /* Search last 16k of file for end-of-central-directory
4043 * record (which starts with PK\005\006) */
4044 tail = (int)zipmin(1024 * 16, file_size);
4045 current_offset = __archive_read_seek(a, -tail, SEEK_END);
4046 if (current_offset < 0)
4047 return 0;
4048 if ((p = __archive_read_ahead(a, (size_t)tail, NULL)) == NULL)
4049 return 0;
4050 /* Boyer-Moore search backwards from the end, since we want
4051 * to match the last EOCD in the file (there can be more than
4052 * one if there is an uncompressed Zip archive as a member
4053 * within this Zip archive). */
4054 for (i = tail - 22; i > 0;) {
4055 switch (p[i]) {
4056 case 'P':
4057 if (memcmp(p + i, "PK\005\006", 4) == 0) {
4058 int ret = read_eocd(zip, p + i,
4059 current_offset + i);
4060 /* Zip64 EOCD locator precedes
4061 * regular EOCD if present. */
4062 if (i >= 20 && memcmp(p + i - 20, "PK\006\007", 4) == 0) {
4063 int ret_zip64 = read_zip64_eocd(a, zip, p + i - 20);
4064 if (ret_zip64 > ret)
4065 ret = ret_zip64;
4066 }
4067 return (ret);
4068 }
4069 i -= 4;
4070 break;
4071 case 'K': i -= 1; break;
4072 case 005: i -= 2; break;
4073 case 006: i -= 3; break;
4074 default: i -= 4; break;
4075 }
4076 }
4077 return 0;
4078 }
4079
4080 /* The red-black trees are only used in seeking mode to manage
4081 * the in-memory copy of the central directory. */
4082
4083 static int
cmp_node(const struct archive_rb_node * n1,const struct archive_rb_node * n2)4084 cmp_node(const struct archive_rb_node *n1, const struct archive_rb_node *n2)
4085 {
4086 const struct zip_entry *e1 = (const struct zip_entry *)n1;
4087 const struct zip_entry *e2 = (const struct zip_entry *)n2;
4088
4089 if (e1->local_header_offset > e2->local_header_offset)
4090 return -1;
4091 if (e1->local_header_offset < e2->local_header_offset)
4092 return 1;
4093 return 0;
4094 }
4095
4096 static int
cmp_key(const struct archive_rb_node * n,const void * key)4097 cmp_key(const struct archive_rb_node *n, const void *key)
4098 {
4099 /* This function won't be called */
4100 (void)n; /* UNUSED */
4101 (void)key; /* UNUSED */
4102 return 1;
4103 }
4104
4105 static const struct archive_rb_tree_ops rb_ops = {
4106 &cmp_node, &cmp_key
4107 };
4108
4109 static int
rsrc_cmp_node(const struct archive_rb_node * n1,const struct archive_rb_node * n2)4110 rsrc_cmp_node(const struct archive_rb_node *n1,
4111 const struct archive_rb_node *n2)
4112 {
4113 const struct zip_entry *e1 = (const struct zip_entry *)n1;
4114 const struct zip_entry *e2 = (const struct zip_entry *)n2;
4115
4116 return (strcmp(e2->rsrcname.s, e1->rsrcname.s));
4117 }
4118
4119 static int
rsrc_cmp_key(const struct archive_rb_node * n,const void * key)4120 rsrc_cmp_key(const struct archive_rb_node *n, const void *key)
4121 {
4122 const struct zip_entry *e = (const struct zip_entry *)n;
4123 return (strcmp((const char *)key, e->rsrcname.s));
4124 }
4125
4126 static const struct archive_rb_tree_ops rb_rsrc_ops = {
4127 &rsrc_cmp_node, &rsrc_cmp_key
4128 };
4129
4130 static const char *
rsrc_basename(const char * name,size_t name_length)4131 rsrc_basename(const char *name, size_t name_length)
4132 {
4133 const char *s, *r;
4134
4135 r = s = name;
4136 for (;;) {
4137 s = memchr(s, '/', name_length - (s - name));
4138 if (s == NULL)
4139 break;
4140 r = ++s;
4141 }
4142 return (r);
4143 }
4144
4145 static void
expose_parent_dirs(struct zip * zip,const char * name,size_t name_length)4146 expose_parent_dirs(struct zip *zip, const char *name, size_t name_length)
4147 {
4148 struct archive_string str;
4149 struct zip_entry *dir;
4150 char *s;
4151
4152 archive_string_init(&str);
4153 archive_strncpy(&str, name, name_length);
4154 for (;;) {
4155 s = strrchr(str.s, '/');
4156 if (s == NULL)
4157 break;
4158 *s = '\0';
4159 /* Transfer the parent directory from zip->tree_rsrc RB
4160 * tree to zip->tree RB tree to expose. */
4161 dir = (struct zip_entry *)
4162 __archive_rb_tree_find_node(&zip->tree_rsrc, str.s);
4163 if (dir == NULL)
4164 break;
4165 __archive_rb_tree_remove_node(&zip->tree_rsrc, &dir->node);
4166 archive_string_free(&dir->rsrcname);
4167 __archive_rb_tree_insert_node(&zip->tree, &dir->node);
4168 }
4169 archive_string_free(&str);
4170 }
4171
4172 static int
slurp_central_directory(struct archive_read * a,struct archive_entry * entry,struct zip * zip)4173 slurp_central_directory(struct archive_read *a, struct archive_entry* entry,
4174 struct zip *zip)
4175 {
4176 ssize_t i;
4177 unsigned found;
4178 int64_t correction;
4179 ssize_t bytes_avail;
4180 const char *p;
4181
4182 /*
4183 * Find the start of the central directory. The end-of-CD
4184 * record has our starting point, but there are lots of
4185 * Zip archives which have had other data prepended to the
4186 * file, which makes the recorded offsets all too small.
4187 * So we search forward from the specified offset until we
4188 * find the real start of the central directory. Then we
4189 * know the correction we need to apply to account for leading
4190 * padding.
4191 */
4192 if (__archive_read_seek(a, zip->central_directory_offset_adjusted, SEEK_SET)
4193 < 0)
4194 return ARCHIVE_FATAL;
4195
4196 found = 0;
4197 while (!found) {
4198 if ((p = __archive_read_ahead(a, 20, &bytes_avail)) == NULL)
4199 return ARCHIVE_FATAL;
4200 for (found = 0, i = 0; !found && i < bytes_avail - 4;) {
4201 switch (p[i + 3]) {
4202 case 'P': i += 3; break;
4203 case 'K': i += 2; break;
4204 case 001: i += 1; break;
4205 case 002:
4206 if (memcmp(p + i, "PK\001\002", 4) == 0) {
4207 p += i;
4208 found = 1;
4209 } else
4210 i += 4;
4211 break;
4212 case 005: i += 1; break;
4213 case 006:
4214 if (memcmp(p + i, "PK\005\006", 4) == 0) {
4215 p += i;
4216 found = 1;
4217 } else if (memcmp(p + i, "PK\006\006", 4) == 0) {
4218 p += i;
4219 found = 1;
4220 } else
4221 i += 1;
4222 break;
4223 default: i += 4; break;
4224 }
4225 }
4226 __archive_read_consume(a, i);
4227 }
4228 correction = archive_filter_bytes(&a->archive, 0)
4229 - zip->central_directory_offset;
4230
4231 __archive_rb_tree_init(&zip->tree, &rb_ops);
4232 __archive_rb_tree_init(&zip->tree_rsrc, &rb_rsrc_ops);
4233
4234 while (1) {
4235 struct zip_entry *zip_entry;
4236 size_t filename_length, extra_length, comment_length;
4237 uint32_t external_attributes;
4238 const char *name, *r;
4239
4240 if ((p = __archive_read_ahead(a, 4, NULL)) == NULL)
4241 return ARCHIVE_FATAL;
4242 if (memcmp(p, "PK\006\006", 4) == 0
4243 || memcmp(p, "PK\005\006", 4) == 0) {
4244 break;
4245 } else if (memcmp(p, "PK\001\002", 4) != 0) {
4246 archive_set_error(&a->archive,
4247 -1, "Invalid central directory signature");
4248 return ARCHIVE_FATAL;
4249 }
4250 if ((p = __archive_read_ahead(a, 46, NULL)) == NULL)
4251 return ARCHIVE_FATAL;
4252
4253 zip_entry = calloc(1, sizeof(struct zip_entry));
4254 if (zip_entry == NULL) {
4255 archive_set_error(&a->archive, ENOMEM,
4256 "Can't allocate zip entry");
4257 return ARCHIVE_FATAL;
4258 }
4259 zip_entry->next = zip->zip_entries;
4260 zip_entry->flags |= LA_FROM_CENTRAL_DIRECTORY;
4261 zip->zip_entries = zip_entry;
4262
4263 /* version = p[4]; */
4264 zip_entry->system = p[5];
4265 /* version_required = archive_le16dec(p + 6); */
4266 zip_entry->zip_flags = archive_le16dec(p + 8);
4267 if (zip_entry->zip_flags
4268 & (ZIP_ENCRYPTED | ZIP_STRONG_ENCRYPTED)){
4269 zip->has_encrypted_entries = 1;
4270 }
4271 zip_entry->compression = (char)archive_le16dec(p + 10);
4272 zip_entry->mtime = dos_to_unix(archive_le32dec(p + 12));
4273 zip_entry->crc32 = archive_le32dec(p + 16);
4274 if (zip_entry->zip_flags & ZIP_LENGTH_AT_END)
4275 zip_entry->decdat = p[13];
4276 else
4277 zip_entry->decdat = p[19];
4278 zip_entry->compressed_size = archive_le32dec(p + 20);
4279 zip_entry->uncompressed_size = archive_le32dec(p + 24);
4280 filename_length = archive_le16dec(p + 28);
4281 extra_length = archive_le16dec(p + 30);
4282 comment_length = archive_le16dec(p + 32);
4283 /* disk_start = archive_le16dec(p + 34);
4284 * Better be zero.
4285 * internal_attributes = archive_le16dec(p + 36);
4286 * text bit */
4287 external_attributes = archive_le32dec(p + 38);
4288 zip_entry->local_header_offset =
4289 archive_le32dec(p + 42) + correction;
4290
4291 /* If we can't guess the mode, leave it zero here;
4292 when we read the local file header we might get
4293 more information. */
4294 if (zip_entry->system == 3) {
4295 zip_entry->mode = external_attributes >> 16;
4296 } else if (zip_entry->system == 0) {
4297 // Interpret MSDOS directory bit
4298 if (0x10 == (external_attributes & 0x10)) {
4299 zip_entry->mode = AE_IFDIR | 0775;
4300 } else {
4301 zip_entry->mode = AE_IFREG | 0664;
4302 }
4303 if (0x01 == (external_attributes & 0x01)) {
4304 // Read-only bit; strip write permissions
4305 zip_entry->mode &= 0555;
4306 }
4307 } else {
4308 zip_entry->mode = 0;
4309 }
4310
4311 /* We're done with the regular data; get the filename and
4312 * extra data. */
4313 __archive_read_consume(a, 46);
4314 p = __archive_read_ahead(a, filename_length + extra_length,
4315 NULL);
4316 if (p == NULL) {
4317 archive_set_error(&a->archive,
4318 ARCHIVE_ERRNO_FILE_FORMAT,
4319 "Truncated ZIP file header");
4320 return ARCHIVE_FATAL;
4321 }
4322 if (ARCHIVE_OK != process_extra(a, entry, p + filename_length,
4323 extra_length, zip_entry)) {
4324 return ARCHIVE_FATAL;
4325 }
4326
4327 /*
4328 * Mac resource fork files are stored under the
4329 * "__MACOSX/" directory, so we should check if
4330 * it is.
4331 */
4332 if (!zip->process_mac_extensions) {
4333 /* Treat every entry as a regular entry. */
4334 __archive_rb_tree_insert_node(&zip->tree,
4335 &zip_entry->node);
4336 } else {
4337 name = p;
4338 r = rsrc_basename(name, filename_length);
4339 if (filename_length >= 9 &&
4340 strncmp("__MACOSX/", name, 9) == 0) {
4341 const char *name_end;
4342
4343 name_end = name + filename_length;
4344 /* If this file is not a resource fork nor
4345 * a directory. We should treat it as a non
4346 * resource fork file to expose it. */
4347 if (name[filename_length-1] != '/' &&
4348 (name_end - r < 2 || r[0] != '.' || r[1] != '_')) {
4349 __archive_rb_tree_insert_node(
4350 &zip->tree, &zip_entry->node);
4351 /* Expose its parent directories. */
4352 expose_parent_dirs(zip, name,
4353 filename_length);
4354 } else {
4355 /* This file is a resource fork file or
4356 * a directory. */
4357 archive_strncpy(&(zip_entry->rsrcname),
4358 name, filename_length);
4359 __archive_rb_tree_insert_node(
4360 &zip->tree_rsrc, &zip_entry->node);
4361 }
4362 } else {
4363 /* Generate resource fork name to find its
4364 * resource file at zip->tree_rsrc. */
4365
4366 /* If this is an entry ending with slash,
4367 * make the resource for name slash-less
4368 * as the actual resource fork doesn't end with '/'.
4369 */
4370 size_t tmp_length = filename_length;
4371 if (tmp_length > 0 && name[tmp_length - 1] == '/') {
4372 tmp_length--;
4373 r = rsrc_basename(name, tmp_length);
4374 }
4375
4376 archive_strcpy(&(zip_entry->rsrcname),
4377 "__MACOSX/");
4378 archive_strncat(&(zip_entry->rsrcname),
4379 name, r - name);
4380 archive_strcat(&(zip_entry->rsrcname), "._");
4381 archive_strncat(&(zip_entry->rsrcname),
4382 name + (r - name),
4383 tmp_length - (r - name));
4384 /* Register an entry to RB tree to sort it by
4385 * file offset. */
4386 __archive_rb_tree_insert_node(&zip->tree,
4387 &zip_entry->node);
4388 }
4389 }
4390
4391 /* Skip the comment too ... */
4392 __archive_read_consume(a,
4393 filename_length + extra_length + comment_length);
4394 }
4395
4396 return ARCHIVE_OK;
4397 }
4398
4399 static ssize_t
zip_get_local_file_header_size(struct archive_read * a,size_t extra)4400 zip_get_local_file_header_size(struct archive_read *a, size_t extra)
4401 {
4402 const char *p;
4403 ssize_t filename_length, extra_length;
4404
4405 if ((p = __archive_read_ahead(a, extra + ZIP_LOCHDR_LEN, NULL)) == NULL) {
4406 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
4407 "Truncated ZIP file header");
4408 return (ARCHIVE_WARN);
4409 }
4410 p += extra;
4411
4412 if (memcmp(p, "PK\003\004", 4) != 0) {
4413 archive_set_error(&a->archive, -1, "Damaged Zip archive");
4414 return ARCHIVE_WARN;
4415 }
4416 filename_length = archive_le16dec(p + 26);
4417 extra_length = archive_le16dec(p + 28);
4418
4419 return (ZIP_LOCHDR_LEN + filename_length + extra_length);
4420 }
4421
4422 static int
zip_read_mac_metadata(struct archive_read * a,struct archive_entry * entry,struct zip_entry * rsrc)4423 zip_read_mac_metadata(struct archive_read *a, struct archive_entry *entry,
4424 struct zip_entry *rsrc)
4425 {
4426 struct zip *zip = a->format->data;
4427 unsigned char *metadata, *mp;
4428 int64_t offset = archive_filter_bytes(&a->archive, 0);
4429 size_t remaining_bytes, metadata_bytes;
4430 ssize_t hsize;
4431 int ret = ARCHIVE_OK, eof;
4432
4433 switch(rsrc->compression) {
4434 case 0: /* No compression. */
4435 if (rsrc->uncompressed_size != rsrc->compressed_size) {
4436 archive_set_error(&a->archive,
4437 ARCHIVE_ERRNO_FILE_FORMAT,
4438 "Malformed OS X metadata entry: "
4439 "inconsistent size");
4440 return (ARCHIVE_FATAL);
4441 }
4442 #ifdef HAVE_ZLIB_H
4443 case 8: /* Deflate compression. */
4444 #endif
4445 break;
4446 default: /* Unsupported compression. */
4447 /* Return a warning. */
4448 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
4449 "Unsupported ZIP compression method (%s)",
4450 compression_name(rsrc->compression));
4451 /* We can't decompress this entry, but we will
4452 * be able to skip() it and try the next entry. */
4453 return (ARCHIVE_WARN);
4454 }
4455
4456 if (rsrc->uncompressed_size > ZIP_MAX_METADATA * 1048576U) {
4457 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
4458 "Mac metadata is too large: %jd > %u MiB",
4459 (intmax_t)rsrc->uncompressed_size, ZIP_MAX_METADATA);
4460 return (ARCHIVE_WARN);
4461 }
4462 if (rsrc->compressed_size > ZIP_MAX_METADATA * 1048576U) {
4463 archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
4464 "Mac metadata is too large: %jd > %u MiB",
4465 (intmax_t)rsrc->compressed_size, ZIP_MAX_METADATA);
4466 return (ARCHIVE_WARN);
4467 }
4468
4469 metadata = malloc((size_t)rsrc->uncompressed_size);
4470 if (metadata == NULL) {
4471 archive_set_error(&a->archive, ENOMEM,
4472 "Can't allocate memory for Mac metadata");
4473 return (ARCHIVE_FATAL);
4474 }
4475
4476 if (offset < rsrc->local_header_offset)
4477 __archive_read_consume(a, rsrc->local_header_offset - offset);
4478 else if (offset != rsrc->local_header_offset) {
4479 __archive_read_seek(a, rsrc->local_header_offset, SEEK_SET);
4480 }
4481
4482 hsize = zip_get_local_file_header_size(a, 0);
4483 __archive_read_consume(a, hsize);
4484
4485 remaining_bytes = (size_t)rsrc->compressed_size;
4486 metadata_bytes = (size_t)rsrc->uncompressed_size;
4487 mp = metadata;
4488 eof = 0;
4489 while (!eof && remaining_bytes) {
4490 const unsigned char *p;
4491 ssize_t bytes_avail;
4492 size_t bytes_used;
4493
4494 p = __archive_read_ahead(a, 1, &bytes_avail);
4495 if (p == NULL) {
4496 archive_set_error(&a->archive,
4497 ARCHIVE_ERRNO_FILE_FORMAT,
4498 "Truncated ZIP file header");
4499 ret = ARCHIVE_WARN;
4500 goto exit_mac_metadata;
4501 }
4502 if ((size_t)bytes_avail > remaining_bytes)
4503 bytes_avail = remaining_bytes;
4504 switch(rsrc->compression) {
4505 case 0: /* No compression. */
4506 if ((size_t)bytes_avail > metadata_bytes)
4507 bytes_avail = metadata_bytes;
4508 memcpy(mp, p, bytes_avail);
4509 bytes_used = (size_t)bytes_avail;
4510 metadata_bytes -= bytes_used;
4511 mp += bytes_used;
4512 if (metadata_bytes == 0)
4513 eof = 1;
4514 break;
4515 #ifdef HAVE_ZLIB_H
4516 case 8: /* Deflate compression. */
4517 {
4518 int r;
4519
4520 ret = zip_deflate_init(a, zip);
4521 if (ret != ARCHIVE_OK)
4522 goto exit_mac_metadata;
4523 zip->stream.next_in =
4524 (Bytef *)(uintptr_t)(const void *)p;
4525 zip->stream.avail_in = (uInt)bytes_avail;
4526 zip->stream.total_in = 0;
4527 zip->stream.next_out = mp;
4528 zip->stream.avail_out = (uInt)metadata_bytes;
4529 zip->stream.total_out = 0;
4530
4531 r = inflate(&zip->stream, 0);
4532 switch (r) {
4533 case Z_OK:
4534 break;
4535 case Z_STREAM_END:
4536 eof = 1;
4537 break;
4538 case Z_MEM_ERROR:
4539 archive_set_error(&a->archive, ENOMEM,
4540 "Out of memory for ZIP decompression");
4541 ret = ARCHIVE_FATAL;
4542 goto exit_mac_metadata;
4543 default:
4544 archive_set_error(&a->archive,
4545 ARCHIVE_ERRNO_MISC,
4546 "ZIP decompression failed (%d)", r);
4547 ret = ARCHIVE_FATAL;
4548 goto exit_mac_metadata;
4549 }
4550 bytes_used = zip->stream.total_in;
4551 metadata_bytes -= zip->stream.total_out;
4552 mp += zip->stream.total_out;
4553 break;
4554 }
4555 #endif
4556 default:
4557 bytes_used = 0;
4558 break;
4559 }
4560 __archive_read_consume(a, bytes_used);
4561 remaining_bytes -= bytes_used;
4562 }
4563 archive_entry_copy_mac_metadata(entry, metadata,
4564 (size_t)rsrc->uncompressed_size - metadata_bytes);
4565
4566 exit_mac_metadata:
4567 __archive_read_seek(a, offset, SEEK_SET);
4568 zip->decompress_init = 0;
4569 free(metadata);
4570 return (ret);
4571 }
4572
4573 static int
archive_read_format_zip_seekable_read_header(struct archive_read * a,struct archive_entry * entry)4574 archive_read_format_zip_seekable_read_header(struct archive_read *a,
4575 struct archive_entry *entry)
4576 {
4577 struct zip *zip = a->format->data;
4578 struct zip_entry *rsrc;
4579 int64_t offset;
4580 int r, ret = ARCHIVE_OK;
4581
4582 /*
4583 * It should be sufficient to call archive_read_next_header() for
4584 * a reader to determine if an entry is encrypted or not. If the
4585 * encryption of an entry is only detectable when calling
4586 * archive_read_data(), so be it. We'll do the same check there
4587 * as well.
4588 */
4589 if (zip->has_encrypted_entries ==
4590 ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW)
4591 zip->has_encrypted_entries = 0;
4592
4593 a->archive.archive_format = ARCHIVE_FORMAT_ZIP;
4594 if (a->archive.archive_format_name == NULL)
4595 a->archive.archive_format_name = "ZIP";
4596
4597 if (zip->zip_entries == NULL) {
4598 r = slurp_central_directory(a, entry, zip);
4599 if (r != ARCHIVE_OK)
4600 return r;
4601 /* Get first entry whose local header offset is lower than
4602 * other entries in the archive file. */
4603 zip->entry =
4604 (struct zip_entry *)ARCHIVE_RB_TREE_MIN(&zip->tree);
4605 } else if (zip->entry != NULL) {
4606 /* Get next entry in local header offset order. */
4607 zip->entry = (struct zip_entry *)__archive_rb_tree_iterate(
4608 &zip->tree, &zip->entry->node, ARCHIVE_RB_DIR_RIGHT);
4609 }
4610
4611 if (zip->entry == NULL)
4612 return ARCHIVE_EOF;
4613
4614 if (zip->entry->rsrcname.s)
4615 rsrc = (struct zip_entry *)__archive_rb_tree_find_node(
4616 &zip->tree_rsrc, zip->entry->rsrcname.s);
4617 else
4618 rsrc = NULL;
4619
4620 if (zip->cctx_valid)
4621 archive_decrypto_aes_ctr_release(&zip->cctx);
4622 if (zip->hctx_valid)
4623 archive_hmac_sha1_cleanup(&zip->hctx);
4624 zip->tctx_valid = zip->cctx_valid = zip->hctx_valid = 0;
4625 __archive_read_reset_passphrase(a);
4626
4627 /* File entries are sorted by the header offset, we should mostly
4628 * use __archive_read_consume to advance a read point to avoid
4629 * redundant data reading. */
4630 offset = archive_filter_bytes(&a->archive, 0);
4631 if (offset < zip->entry->local_header_offset)
4632 __archive_read_consume(a,
4633 zip->entry->local_header_offset - offset);
4634 else if (offset != zip->entry->local_header_offset) {
4635 __archive_read_seek(a, zip->entry->local_header_offset,
4636 SEEK_SET);
4637 }
4638 zip->unconsumed = 0;
4639 r = zip_read_local_file_header(a, entry, zip);
4640 if (r != ARCHIVE_OK)
4641 return r;
4642 if (rsrc) {
4643 int ret2 = zip_read_mac_metadata(a, entry, rsrc);
4644 if (ret2 < ret)
4645 ret = ret2;
4646 }
4647 return (ret);
4648 }
4649
4650 /*
4651 * We're going to seek for the next header anyway, so we don't
4652 * need to bother doing anything here.
4653 */
4654 static int
archive_read_format_zip_read_data_skip_seekable(struct archive_read * a)4655 archive_read_format_zip_read_data_skip_seekable(struct archive_read *a)
4656 {
4657 struct zip *zip = a->format->data;
4658
4659 zip->unconsumed = 0;
4660 return (ARCHIVE_OK);
4661 }
4662
4663 int
archive_read_support_format_zip_seekable(struct archive * _a)4664 archive_read_support_format_zip_seekable(struct archive *_a)
4665 {
4666 struct archive_read *a = (struct archive_read *)_a;
4667 struct zip *zip;
4668 int r;
4669
4670 archive_check_magic(_a, ARCHIVE_READ_MAGIC,
4671 ARCHIVE_STATE_NEW, "archive_read_support_format_zip_seekable");
4672
4673 zip = calloc(1, sizeof(*zip));
4674 if (zip == NULL) {
4675 archive_set_error(&a->archive, ENOMEM,
4676 "Can't allocate zip data");
4677 return (ARCHIVE_FATAL);
4678 }
4679
4680 #ifdef HAVE_COPYFILE_H
4681 /* Set this by default on Mac OS. */
4682 zip->process_mac_extensions = 1;
4683 #endif
4684
4685 /*
4686 * Until enough data has been read, we cannot tell about
4687 * any encrypted entries yet.
4688 */
4689 zip->has_encrypted_entries = ARCHIVE_READ_FORMAT_ENCRYPTION_DONT_KNOW;
4690 zip->crc32func = real_crc32;
4691
4692 r = __archive_read_register_format(a,
4693 zip,
4694 "zip",
4695 archive_read_format_zip_seekable_bid,
4696 archive_read_format_zip_options,
4697 archive_read_format_zip_seekable_read_header,
4698 archive_read_format_zip_read_data,
4699 archive_read_format_zip_read_data_skip_seekable,
4700 NULL,
4701 archive_read_format_zip_cleanup,
4702 archive_read_support_format_zip_capabilities_seekable,
4703 archive_read_format_zip_has_encrypted_entries);
4704
4705 if (r != ARCHIVE_OK)
4706 free(zip);
4707 return (ARCHIVE_OK);
4708 }
4709
4710 /*# vim:set noet:*/
4711