blob: 100c833fe51fa332bdab0df0d9c35c7bbaea0e01 [file] [log] [blame]
The Android Open Source Projectc24a8e62009-03-03 19:28:42 -08001/*
2 * Copyright 2006 The Android Open Source Project
3 *
4 * Simple Zip file support.
5 */
6#include "safe_iop.h"
7#include "zlib.h"
8
9#include <errno.h>
10#include <fcntl.h>
11#include <limits.h>
12#include <stdint.h> // for uintptr_t
13#include <stdlib.h>
14#include <sys/stat.h> // for S_ISLNK()
15#include <unistd.h>
16
17#define LOG_TAG "minzip"
18#include "Zip.h"
19#include "Bits.h"
20#include "Log.h"
21#include "DirUtil.h"
22
23#undef NDEBUG // do this after including Log.h
24#include <assert.h>
25
26#define SORT_ENTRIES 1
27
28/*
29 * Offset and length constants (java.util.zip naming convention).
30 */
31enum {
32 CENSIG = 0x02014b50, // PK12
33 CENHDR = 46,
34
35 CENVEM = 4,
36 CENVER = 6,
37 CENFLG = 8,
38 CENHOW = 10,
39 CENTIM = 12,
40 CENCRC = 16,
41 CENSIZ = 20,
42 CENLEN = 24,
43 CENNAM = 28,
44 CENEXT = 30,
45 CENCOM = 32,
46 CENDSK = 34,
47 CENATT = 36,
48 CENATX = 38,
49 CENOFF = 42,
50
51 ENDSIG = 0x06054b50, // PK56
52 ENDHDR = 22,
53
54 ENDSUB = 8,
55 ENDTOT = 10,
56 ENDSIZ = 12,
57 ENDOFF = 16,
58 ENDCOM = 20,
59
60 EXTSIG = 0x08074b50, // PK78
61 EXTHDR = 16,
62
63 EXTCRC = 4,
64 EXTSIZ = 8,
65 EXTLEN = 12,
66
67 LOCSIG = 0x04034b50, // PK34
68 LOCHDR = 30,
69
70 LOCVER = 4,
71 LOCFLG = 6,
72 LOCHOW = 8,
73 LOCTIM = 10,
74 LOCCRC = 14,
75 LOCSIZ = 18,
76 LOCLEN = 22,
77 LOCNAM = 26,
78 LOCEXT = 28,
79
80 STORED = 0,
81 DEFLATED = 8,
82
83 CENVEM_UNIX = 3 << 8, // the high byte of CENVEM
84};
85
86
87/*
88 * For debugging, dump the contents of a ZipEntry.
89 */
90#if 0
91static void dumpEntry(const ZipEntry* pEntry)
92{
93 LOGI(" %p '%.*s'\n", pEntry->fileName,pEntry->fileNameLen,pEntry->fileName);
94 LOGI(" off=%ld comp=%ld uncomp=%ld how=%d\n", pEntry->offset,
95 pEntry->compLen, pEntry->uncompLen, pEntry->compression);
96}
97#endif
98
99/*
100 * (This is a mzHashTableLookup callback.)
101 *
102 * Compare two ZipEntry structs, by name.
103 */
104static int hashcmpZipEntry(const void* ventry1, const void* ventry2)
105{
106 const ZipEntry* entry1 = (const ZipEntry*) ventry1;
107 const ZipEntry* entry2 = (const ZipEntry*) ventry2;
108
109 if (entry1->fileNameLen != entry2->fileNameLen)
110 return entry1->fileNameLen - entry2->fileNameLen;
111 return memcmp(entry1->fileName, entry2->fileName, entry1->fileNameLen);
112}
113
114/*
115 * (This is a mzHashTableLookup callback.)
116 *
117 * find a ZipEntry struct by name.
118 */
119static int hashcmpZipName(const void* ventry, const void* vname)
120{
121 const ZipEntry* entry = (const ZipEntry*) ventry;
122 const char* name = (const char*) vname;
123 unsigned int nameLen = strlen(name);
124
125 if (entry->fileNameLen != nameLen)
126 return entry->fileNameLen - nameLen;
127 return memcmp(entry->fileName, name, nameLen);
128}
129
130/*
131 * Compute the hash code for a ZipEntry filename.
132 *
133 * Not expected to be compatible with any other hash function, so we init
134 * to 2 to ensure it doesn't happen to match.
135 */
136static unsigned int computeHash(const char* name, int nameLen)
137{
138 unsigned int hash = 2;
139
140 while (nameLen--)
141 hash = hash * 31 + *name++;
142
143 return hash;
144}
145
146static void addEntryToHashTable(HashTable* pHash, ZipEntry* pEntry)
147{
148 unsigned int itemHash = computeHash(pEntry->fileName, pEntry->fileNameLen);
149 const ZipEntry* found;
150
151 found = (const ZipEntry*)mzHashTableLookup(pHash,
152 itemHash, pEntry, hashcmpZipEntry, true);
153 if (found != pEntry) {
154 LOGW("WARNING: duplicate entry '%.*s' in Zip\n",
155 found->fileNameLen, found->fileName);
156 /* keep going */
157 }
158}
159
160static int validFilename(const char *fileName, unsigned int fileNameLen)
161{
162 // Forbid super long filenames.
163 if (fileNameLen >= PATH_MAX) {
164 LOGW("Filename too long (%d chatacters)\n", fileNameLen);
165 return 0;
166 }
167
168 // Require all characters to be printable ASCII (no NUL, no UTF-8, etc).
169 unsigned int i;
170 for (i = 0; i < fileNameLen; ++i) {
171 if (fileName[i] < 32 || fileName[i] >= 127) {
172 LOGW("Filename contains invalid character '\%03o'\n", fileName[i]);
173 return 0;
174 }
175 }
176
177 return 1;
178}
179
180/*
181 * Parse the contents of a Zip archive. After confirming that the file
182 * is in fact a Zip, we scan out the contents of the central directory and
183 * store it in a hash table.
184 *
185 * Returns "true" on success.
186 */
187static bool parseZipArchive(ZipArchive* pArchive, const MemMapping* pMap)
188{
189 bool result = false;
190 const unsigned char* ptr;
191 unsigned int i, numEntries, cdOffset;
192 unsigned int val;
193
194 /*
195 * The first 4 bytes of the file will either be the local header
196 * signature for the first file (LOCSIG) or, if the archive doesn't
197 * have any files in it, the end-of-central-directory signature (ENDSIG).
198 */
199 val = get4LE(pMap->addr);
200 if (val == ENDSIG) {
201 LOGI("Found Zip archive, but it looks empty\n");
202 goto bail;
203 } else if (val != LOCSIG) {
204 LOGV("Not a Zip archive (found 0x%08x)\n", val);
205 goto bail;
206 }
207
208 /*
209 * Find the EOCD. We'll find it immediately unless they have a file
210 * comment.
211 */
212 ptr = pMap->addr + pMap->length - ENDHDR;
213
214 while (ptr >= (const unsigned char*) pMap->addr) {
215 if (*ptr == (ENDSIG & 0xff) && get4LE(ptr) == ENDSIG)
216 break;
217 ptr--;
218 }
219 if (ptr < (const unsigned char*) pMap->addr) {
220 LOGI("Could not find end-of-central-directory in Zip\n");
221 goto bail;
222 }
223
224 /*
225 * There are two interesting items in the EOCD block: the number of
226 * entries in the file, and the file offset of the start of the
227 * central directory.
228 */
229 numEntries = get2LE(ptr + ENDSUB);
230 cdOffset = get4LE(ptr + ENDOFF);
231
232 LOGVV("numEntries=%d cdOffset=%d\n", numEntries, cdOffset);
233 if (numEntries == 0 || cdOffset >= pMap->length) {
234 LOGW("Invalid entries=%d offset=%d (len=%zd)\n",
235 numEntries, cdOffset, pMap->length);
236 goto bail;
237 }
238
239 /*
240 * Create data structures to hold entries.
241 */
242 pArchive->numEntries = numEntries;
243 pArchive->pEntries = (ZipEntry*) calloc(numEntries, sizeof(ZipEntry));
244 pArchive->pHash = mzHashTableCreate(mzHashSize(numEntries), NULL);
245 if (pArchive->pEntries == NULL || pArchive->pHash == NULL)
246 goto bail;
247
248 ptr = pMap->addr + cdOffset;
249 for (i = 0; i < numEntries; i++) {
250 ZipEntry* pEntry;
251 unsigned int fileNameLen, extraLen, commentLen, localHdrOffset;
252 const unsigned char* localHdr;
253 const char *fileName;
254
255 if (ptr + CENHDR > (const unsigned char*)pMap->addr + pMap->length) {
256 LOGW("Ran off the end (at %d)\n", i);
257 goto bail;
258 }
259 if (get4LE(ptr) != CENSIG) {
260 LOGW("Missed a central dir sig (at %d)\n", i);
261 goto bail;
262 }
263
264 localHdrOffset = get4LE(ptr + CENOFF);
265 fileNameLen = get2LE(ptr + CENNAM);
266 extraLen = get2LE(ptr + CENEXT);
267 commentLen = get2LE(ptr + CENCOM);
268 fileName = (const char*)ptr + CENHDR;
269 if (fileName + fileNameLen > (const char*)pMap->addr + pMap->length) {
270 LOGW("Filename ran off the end (at %d)\n", i);
271 goto bail;
272 }
273 if (!validFilename(fileName, fileNameLen)) {
274 LOGW("Invalid filename (at %d)\n", i);
275 goto bail;
276 }
277
278#if SORT_ENTRIES
279 /* Figure out where this entry should go (binary search).
280 */
281 if (i > 0) {
282 int low, high;
283
284 low = 0;
285 high = i - 1;
286 while (low <= high) {
287 int mid;
288 int diff;
289 int diffLen;
290
291 mid = low + ((high - low) / 2); // avoid overflow
292
293 if (pArchive->pEntries[mid].fileNameLen < fileNameLen) {
294 diffLen = pArchive->pEntries[mid].fileNameLen;
295 } else {
296 diffLen = fileNameLen;
297 }
298 diff = strncmp(pArchive->pEntries[mid].fileName, fileName,
299 diffLen);
300 if (diff == 0) {
301 diff = pArchive->pEntries[mid].fileNameLen - fileNameLen;
302 }
303 if (diff < 0) {
304 low = mid + 1;
305 } else if (diff > 0) {
306 high = mid - 1;
307 } else {
308 high = mid;
309 break;
310 }
311 }
312
313 unsigned int target = high + 1;
314 assert(target <= i);
315 if (target != i) {
316 /* It belongs somewhere other than at the end of
317 * the list. Make some room at [target].
318 */
319 memmove(pArchive->pEntries + target + 1,
320 pArchive->pEntries + target,
321 (i - target) * sizeof(ZipEntry));
322 }
323 pEntry = &pArchive->pEntries[target];
324 } else {
325 pEntry = &pArchive->pEntries[0];
326 }
327#else
328 pEntry = &pArchive->pEntries[i];
329#endif
330
331 //LOGI("%d: localHdr=%d fnl=%d el=%d cl=%d\n",
332 // i, localHdrOffset, fileNameLen, extraLen, commentLen);
333
334 pEntry->fileNameLen = fileNameLen;
335 pEntry->fileName = fileName;
336
337 pEntry->compLen = get4LE(ptr + CENSIZ);
338 pEntry->uncompLen = get4LE(ptr + CENLEN);
339 pEntry->compression = get2LE(ptr + CENHOW);
340 pEntry->modTime = get4LE(ptr + CENTIM);
341 pEntry->crc32 = get4LE(ptr + CENCRC);
342
343 /* These two are necessary for finding the mode of the file.
344 */
345 pEntry->versionMadeBy = get2LE(ptr + CENVEM);
346 if ((pEntry->versionMadeBy & 0xff00) != 0 &&
347 (pEntry->versionMadeBy & 0xff00) != CENVEM_UNIX)
348 {
349 LOGW("Incompatible \"version made by\": 0x%02x (at %d)\n",
350 pEntry->versionMadeBy >> 8, i);
351 goto bail;
352 }
353 pEntry->externalFileAttributes = get4LE(ptr + CENATX);
354
355 // Perform pMap->addr + localHdrOffset, ensuring that it won't
356 // overflow. This is needed because localHdrOffset is untrusted.
357 if (!safe_add((uintptr_t *)&localHdr, (uintptr_t)pMap->addr,
358 (uintptr_t)localHdrOffset)) {
359 LOGW("Integer overflow adding in parseZipArchive\n");
360 goto bail;
361 }
362 if ((uintptr_t)localHdr + LOCHDR >
363 (uintptr_t)pMap->addr + pMap->length) {
364 LOGW("Bad offset to local header: %d (at %d)\n", localHdrOffset, i);
365 goto bail;
366 }
367 if (get4LE(localHdr) != LOCSIG) {
368 LOGW("Missed a local header sig (at %d)\n", i);
369 goto bail;
370 }
371 pEntry->offset = localHdrOffset + LOCHDR
372 + get2LE(localHdr + LOCNAM) + get2LE(localHdr + LOCEXT);
373 if (!safe_add(NULL, pEntry->offset, pEntry->compLen)) {
374 LOGW("Integer overflow adding in parseZipArchive\n");
375 goto bail;
376 }
377 if ((size_t)pEntry->offset + pEntry->compLen > pMap->length) {
378 LOGW("Data ran off the end (at %d)\n", i);
379 goto bail;
380 }
381
382#if !SORT_ENTRIES
383 /* Add to hash table; no need to lock here.
384 * Can't do this now if we're sorting, because entries
385 * will move around.
386 */
387 addEntryToHashTable(pArchive->pHash, pEntry);
388#endif
389
390 //dumpEntry(pEntry);
391 ptr += CENHDR + fileNameLen + extraLen + commentLen;
392 }
393
394#if SORT_ENTRIES
395 /* If we're sorting, we have to wait until all entries
396 * are in their final places, otherwise the pointers will
397 * probably point to the wrong things.
398 */
399 for (i = 0; i < numEntries; i++) {
400 /* Add to hash table; no need to lock here.
401 */
402 addEntryToHashTable(pArchive->pHash, &pArchive->pEntries[i]);
403 }
404#endif
405
406 result = true;
407
408bail:
409 if (!result) {
410 mzHashTableFree(pArchive->pHash);
411 pArchive->pHash = NULL;
412 }
413 return result;
414}
415
416/*
417 * Open a Zip archive and scan out the contents.
418 *
419 * The easiest way to do this is to mmap() the whole thing and do the
420 * traditional backward scan for central directory. Since the EOCD is
421 * a relatively small bit at the end, we should end up only touching a
422 * small set of pages.
423 *
424 * This will be called on non-Zip files, especially during startup, so
425 * we don't want to be too noisy about failures. (Do we want a "quiet"
426 * flag?)
427 *
428 * On success, we fill out the contents of "pArchive".
429 */
430int mzOpenZipArchive(const char* fileName, ZipArchive* pArchive)
431{
432 MemMapping map;
433 int err;
434
435 LOGV("Opening archive '%s' %p\n", fileName, pArchive);
436
437 map.addr = NULL;
438 memset(pArchive, 0, sizeof(*pArchive));
439
440 pArchive->fd = open(fileName, O_RDONLY, 0);
441 if (pArchive->fd < 0) {
442 err = errno ? errno : -1;
443 LOGV("Unable to open '%s': %s\n", fileName, strerror(err));
444 goto bail;
445 }
446
447 if (sysMapFileInShmem(pArchive->fd, &map) != 0) {
448 err = -1;
449 LOGW("Map of '%s' failed\n", fileName);
450 goto bail;
451 }
452
453 if (map.length < ENDHDR) {
454 err = -1;
455 LOGV("File '%s' too small to be zip (%zd)\n", fileName, map.length);
456 goto bail;
457 }
458
459 if (!parseZipArchive(pArchive, &map)) {
460 err = -1;
461 LOGV("Parsing '%s' failed\n", fileName);
462 goto bail;
463 }
464
465 err = 0;
466 sysCopyMap(&pArchive->map, &map);
467 map.addr = NULL;
468
469bail:
470 if (err != 0)
471 mzCloseZipArchive(pArchive);
472 if (map.addr != NULL)
473 sysReleaseShmem(&map);
474 return err;
475}
476
477/*
478 * Close a ZipArchive, closing the file and freeing the contents.
479 *
480 * NOTE: the ZipArchive may not have been fully created.
481 */
482void mzCloseZipArchive(ZipArchive* pArchive)
483{
484 LOGV("Closing archive %p\n", pArchive);
485
486 if (pArchive->fd >= 0)
487 close(pArchive->fd);
488 if (pArchive->map.addr != NULL)
489 sysReleaseShmem(&pArchive->map);
490
491 free(pArchive->pEntries);
492
493 mzHashTableFree(pArchive->pHash);
494
495 pArchive->fd = -1;
496 pArchive->pHash = NULL;
497 pArchive->pEntries = NULL;
498}
499
500/*
501 * Find a matching entry.
502 *
503 * Returns NULL if no matching entry found.
504 */
505const ZipEntry* mzFindZipEntry(const ZipArchive* pArchive,
506 const char* entryName)
507{
508 unsigned int itemHash = computeHash(entryName, strlen(entryName));
509
510 return (const ZipEntry*)mzHashTableLookup(pArchive->pHash,
511 itemHash, (char*) entryName, hashcmpZipName, false);
512}
513
514/*
515 * Return true if the entry is a symbolic link.
516 */
517bool mzIsZipEntrySymlink(const ZipEntry* pEntry)
518{
519 if ((pEntry->versionMadeBy & 0xff00) == CENVEM_UNIX) {
520 return S_ISLNK(pEntry->externalFileAttributes >> 16);
521 }
522 return false;
523}
524
525/* Call processFunction on the uncompressed data of a STORED entry.
526 */
527static bool processStoredEntry(const ZipArchive *pArchive,
528 const ZipEntry *pEntry, ProcessZipEntryContentsFunction processFunction,
529 void *cookie)
530{
531 size_t bytesLeft = pEntry->compLen;
532 while (bytesLeft > 0) {
533 unsigned char buf[32 * 1024];
534 ssize_t n;
535 size_t count;
536 bool ret;
537
538 count = bytesLeft;
539 if (count > sizeof(buf)) {
540 count = sizeof(buf);
541 }
542 n = read(pArchive->fd, buf, count);
543 if (n < 0 || (size_t)n != count) {
544 LOGE("Can't read %zu bytes from zip file: %ld\n", count, n);
545 return false;
546 }
547 ret = processFunction(buf, n, cookie);
548 if (!ret) {
549 return false;
550 }
551 bytesLeft -= count;
552 }
553 return true;
554}
555
556static bool processDeflatedEntry(const ZipArchive *pArchive,
557 const ZipEntry *pEntry, ProcessZipEntryContentsFunction processFunction,
558 void *cookie)
559{
560 long result = -1;
561 unsigned char readBuf[32 * 1024];
562 unsigned char procBuf[32 * 1024];
563 z_stream zstream;
564 int zerr;
565 long compRemaining;
566
567 compRemaining = pEntry->compLen;
568
569 /*
570 * Initialize the zlib stream.
571 */
572 memset(&zstream, 0, sizeof(zstream));
573 zstream.zalloc = Z_NULL;
574 zstream.zfree = Z_NULL;
575 zstream.opaque = Z_NULL;
576 zstream.next_in = NULL;
577 zstream.avail_in = 0;
578 zstream.next_out = (Bytef*) procBuf;
579 zstream.avail_out = sizeof(procBuf);
580 zstream.data_type = Z_UNKNOWN;
581
582 /*
583 * Use the undocumented "negative window bits" feature to tell zlib
584 * that there's no zlib header waiting for it.
585 */
586 zerr = inflateInit2(&zstream, -MAX_WBITS);
587 if (zerr != Z_OK) {
588 if (zerr == Z_VERSION_ERROR) {
589 LOGE("Installed zlib is not compatible with linked version (%s)\n",
590 ZLIB_VERSION);
591 } else {
592 LOGE("Call to inflateInit2 failed (zerr=%d)\n", zerr);
593 }
594 goto bail;
595 }
596
597 /*
598 * Loop while we have data.
599 */
600 do {
601 /* read as much as we can */
602 if (zstream.avail_in == 0) {
603 long getSize = (compRemaining > (long)sizeof(readBuf)) ?
604 (long)sizeof(readBuf) : compRemaining;
605 LOGVV("+++ reading %ld bytes (%ld left)\n",
606 getSize, compRemaining);
607
608 int cc = read(pArchive->fd, readBuf, getSize);
609 if (cc != (int) getSize) {
610 LOGW("inflate read failed (%d vs %ld)\n", cc, getSize);
611 goto z_bail;
612 }
613
614 compRemaining -= getSize;
615
616 zstream.next_in = readBuf;
617 zstream.avail_in = getSize;
618 }
619
620 /* uncompress the data */
621 zerr = inflate(&zstream, Z_NO_FLUSH);
622 if (zerr != Z_OK && zerr != Z_STREAM_END) {
623 LOGD("zlib inflate call failed (zerr=%d)\n", zerr);
624 goto z_bail;
625 }
626
627 /* write when we're full or when we're done */
628 if (zstream.avail_out == 0 ||
629 (zerr == Z_STREAM_END && zstream.avail_out != sizeof(procBuf)))
630 {
631 long procSize = zstream.next_out - procBuf;
632 LOGVV("+++ processing %d bytes\n", (int) procSize);
633 bool ret = processFunction(procBuf, procSize, cookie);
634 if (!ret) {
635 LOGW("Process function elected to fail (in inflate)\n");
636 goto z_bail;
637 }
638
639 zstream.next_out = procBuf;
640 zstream.avail_out = sizeof(procBuf);
641 }
642 } while (zerr == Z_OK);
643
644 assert(zerr == Z_STREAM_END); /* other errors should've been caught */
645
646 // success!
647 result = zstream.total_out;
648
649z_bail:
650 inflateEnd(&zstream); /* free up any allocated structures */
651
652bail:
653 if (result != pEntry->uncompLen) {
654 if (result != -1) // error already shown?
655 LOGW("Size mismatch on inflated file (%ld vs %ld)\n",
656 result, pEntry->uncompLen);
657 return false;
658 }
659 return true;
660}
661
662/*
663 * Stream the uncompressed data through the supplied function,
664 * passing cookie to it each time it gets called. processFunction
665 * may be called more than once.
666 *
667 * If processFunction returns false, the operation is abandoned and
668 * mzProcessZipEntryContents() immediately returns false.
669 *
670 * This is useful for calculating the hash of an entry's uncompressed contents.
671 */
672bool mzProcessZipEntryContents(const ZipArchive *pArchive,
673 const ZipEntry *pEntry, ProcessZipEntryContentsFunction processFunction,
674 void *cookie)
675{
676 bool ret = false;
677 off_t oldOff;
678
679 /* save current offset */
680 oldOff = lseek(pArchive->fd, 0, SEEK_CUR);
681
682 /* Seek to the beginning of the entry's compressed data. */
683 lseek(pArchive->fd, pEntry->offset, SEEK_SET);
684
685 switch (pEntry->compression) {
686 case STORED:
687 ret = processStoredEntry(pArchive, pEntry, processFunction, cookie);
688 break;
689 case DEFLATED:
690 ret = processDeflatedEntry(pArchive, pEntry, processFunction, cookie);
691 break;
692 default:
693 LOGE("Unsupported compression type %d for entry '%s'\n",
694 pEntry->compression, pEntry->fileName);
695 break;
696 }
697
698 /* restore file offset */
699 lseek(pArchive->fd, oldOff, SEEK_SET);
700 return ret;
701}
702
703static bool crcProcessFunction(const unsigned char *data, int dataLen,
704 void *crc)
705{
706 *(unsigned long *)crc = crc32(*(unsigned long *)crc, data, dataLen);
707 return true;
708}
709
710/*
711 * Check the CRC on this entry; return true if it is correct.
712 * May do other internal checks as well.
713 */
714bool mzIsZipEntryIntact(const ZipArchive *pArchive, const ZipEntry *pEntry)
715{
716 unsigned long crc;
717 bool ret;
718
719 crc = crc32(0L, Z_NULL, 0);
720 ret = mzProcessZipEntryContents(pArchive, pEntry, crcProcessFunction,
721 (void *)&crc);
722 if (!ret) {
723 LOGE("Can't calculate CRC for entry\n");
724 return false;
725 }
726 if (crc != (unsigned long)pEntry->crc32) {
727 LOGW("CRC for entry %.*s (0x%08lx) != expected (0x%08lx)\n",
728 pEntry->fileNameLen, pEntry->fileName, crc, pEntry->crc32);
729 return false;
730 }
731 return true;
732}
733
734typedef struct {
735 char *buf;
736 int bufLen;
737} CopyProcessArgs;
738
739static bool copyProcessFunction(const unsigned char *data, int dataLen,
740 void *cookie)
741{
742 CopyProcessArgs *args = (CopyProcessArgs *)cookie;
743 if (dataLen <= args->bufLen) {
744 memcpy(args->buf, data, dataLen);
745 args->buf += dataLen;
746 args->bufLen -= dataLen;
747 return true;
748 }
749 return false;
750}
751
752/*
753 * Read an entry into a buffer allocated by the caller.
754 */
755bool mzReadZipEntry(const ZipArchive* pArchive, const ZipEntry* pEntry,
756 char *buf, int bufLen)
757{
758 CopyProcessArgs args;
759 bool ret;
760
761 args.buf = buf;
762 args.bufLen = bufLen;
763 ret = mzProcessZipEntryContents(pArchive, pEntry, copyProcessFunction,
764 (void *)&args);
765 if (!ret) {
766 LOGE("Can't extract entry to buffer.\n");
767 return false;
768 }
769 return true;
770}
771
772static bool writeProcessFunction(const unsigned char *data, int dataLen,
773 void *fd)
774{
775 ssize_t n = write((int)fd, data, dataLen);
776 if (n != dataLen) {
777 LOGE("Can't write %d bytes (only %ld) from zip file: %s\n",
778 dataLen, n, strerror(errno));
779 return false;
780 }
781 return true;
782}
783
784/*
785 * Uncompress "pEntry" in "pArchive" to "fd" at the current offset.
786 */
787bool mzExtractZipEntryToFile(const ZipArchive *pArchive,
788 const ZipEntry *pEntry, int fd)
789{
790 bool ret = mzProcessZipEntryContents(pArchive, pEntry, writeProcessFunction,
791 (void *)fd);
792 if (!ret) {
793 LOGE("Can't extract entry to file.\n");
794 return false;
795 }
796 return true;
797}
798
799/* Helper state to make path translation easier and less malloc-happy.
800 */
801typedef struct {
802 const char *targetDir;
803 const char *zipDir;
804 char *buf;
805 int targetDirLen;
806 int zipDirLen;
807 int bufLen;
808} MzPathHelper;
809
810/* Given the values of targetDir and zipDir in the helper,
811 * return the target filename of the provided entry.
812 * The helper must be initialized first.
813 */
814static const char *targetEntryPath(MzPathHelper *helper, ZipEntry *pEntry)
815{
816 int needLen;
817 bool firstTime = (helper->buf == NULL);
818
819 /* target file <-- targetDir + / + entry[zipDirLen:]
820 */
821 needLen = helper->targetDirLen + 1 +
822 pEntry->fileNameLen - helper->zipDirLen + 1;
823 if (needLen > helper->bufLen) {
824 char *newBuf;
825
826 needLen *= 2;
827 newBuf = (char *)realloc(helper->buf, needLen);
828 if (newBuf == NULL) {
829 return NULL;
830 }
831 helper->buf = newBuf;
832 helper->bufLen = needLen;
833 }
834
835 /* Every path will start with the target path and a slash.
836 */
837 if (firstTime) {
838 char *p = helper->buf;
839 memcpy(p, helper->targetDir, helper->targetDirLen);
840 p += helper->targetDirLen;
841 if (p == helper->buf || p[-1] != '/') {
842 helper->targetDirLen += 1;
843 *p++ = '/';
844 }
845 }
846
847 /* Replace the custom part of the path with the appropriate
848 * part of the entry's path.
849 */
850 char *epath = helper->buf + helper->targetDirLen;
851 memcpy(epath, pEntry->fileName + helper->zipDirLen,
852 pEntry->fileNameLen - helper->zipDirLen);
853 epath += pEntry->fileNameLen - helper->zipDirLen;
854 *epath = '\0';
855
856 return helper->buf;
857}
858
859/*
860 * Inflate all entries under zipDir to the directory specified by
861 * targetDir, which must exist and be a writable directory.
862 *
863 * The immediate children of zipDir will become the immediate
864 * children of targetDir; e.g., if the archive contains the entries
865 *
866 * a/b/c/one
867 * a/b/c/two
868 * a/b/c/d/three
869 *
870 * and mzExtractRecursive(a, "a/b/c", "/tmp") is called, the resulting
871 * files will be
872 *
873 * /tmp/one
874 * /tmp/two
875 * /tmp/d/three
876 *
877 * Returns true on success, false on failure.
878 */
879bool mzExtractRecursive(const ZipArchive *pArchive,
880 const char *zipDir, const char *targetDir,
881 int flags, const struct utimbuf *timestamp,
882 void (*callback)(const char *fn, void *), void *cookie)
883{
884 if (zipDir[0] == '/') {
885 LOGE("mzExtractRecursive(): zipDir must be a relative path.\n");
886 return false;
887 }
888 if (targetDir[0] != '/') {
889 LOGE("mzExtractRecursive(): targetDir must be an absolute path.\n");
890 return false;
891 }
892
893 unsigned int zipDirLen;
894 char *zpath;
895
896 zipDirLen = strlen(zipDir);
897 zpath = (char *)malloc(zipDirLen + 2);
898 if (zpath == NULL) {
899 LOGE("Can't allocate %d bytes for zip path\n", zipDirLen + 2);
900 return false;
901 }
902 /* If zipDir is empty, we'll extract the entire zip file.
903 * Otherwise, canonicalize the path.
904 */
905 if (zipDirLen > 0) {
906 /* Make sure there's (hopefully, exactly one) slash at the
907 * end of the path. This way we don't need to worry about
908 * accidentally extracting "one/twothree" when a path like
909 * "one/two" is specified.
910 */
911 memcpy(zpath, zipDir, zipDirLen);
912 if (zpath[zipDirLen-1] != '/') {
913 zpath[zipDirLen++] = '/';
914 }
915 }
916 zpath[zipDirLen] = '\0';
917
918 /* Set up the helper structure that we'll use to assemble paths.
919 */
920 MzPathHelper helper;
921 helper.targetDir = targetDir;
922 helper.targetDirLen = strlen(helper.targetDir);
923 helper.zipDir = zpath;
924 helper.zipDirLen = strlen(helper.zipDir);
925 helper.buf = NULL;
926 helper.bufLen = 0;
927
928 /* Walk through the entries and extract anything whose path begins
929 * with zpath.
930//TODO: since the entries are sorted, binary search for the first match
931// and stop after the first non-match.
932 */
933 unsigned int i;
934 bool seenMatch = false;
935 int ok = true;
936 for (i = 0; i < pArchive->numEntries; i++) {
937 ZipEntry *pEntry = pArchive->pEntries + i;
938 if (pEntry->fileNameLen < zipDirLen) {
939//TODO: look out for a single empty directory entry that matches zpath, but
940// missing the trailing slash. Most zip files seem to include
941// the trailing slash, but I think it's legal to leave it off.
942// e.g., zpath "a/b/", entry "a/b", with no children of the entry.
943 /* No chance of matching.
944 */
945#if SORT_ENTRIES
946 if (seenMatch) {
947 /* Since the entries are sorted, we can give up
948 * on the first mismatch after the first match.
949 */
950 break;
951 }
952#endif
953 continue;
954 }
955 /* If zpath is empty, this strncmp() will match everything,
956 * which is what we want.
957 */
958 if (strncmp(pEntry->fileName, zpath, zipDirLen) != 0) {
959#if SORT_ENTRIES
960 if (seenMatch) {
961 /* Since the entries are sorted, we can give up
962 * on the first mismatch after the first match.
963 */
964 break;
965 }
966#endif
967 continue;
968 }
969 /* This entry begins with zipDir, so we'll extract it.
970 */
971 seenMatch = true;
972
973 /* Find the target location of the entry.
974 */
975 const char *targetFile = targetEntryPath(&helper, pEntry);
976 if (targetFile == NULL) {
977 LOGE("Can't assemble target path for \"%.*s\"\n",
978 pEntry->fileNameLen, pEntry->fileName);
979 ok = false;
980 break;
981 }
982
983 /* With DRY_RUN set, invoke the callback but don't do anything else.
984 */
985 if (flags & MZ_EXTRACT_DRY_RUN) {
986 if (callback != NULL) callback(targetFile, cookie);
987 continue;
988 }
989
990 /* Create the file or directory.
991 */
992#define UNZIP_DIRMODE 0755
993#define UNZIP_FILEMODE 0644
994 if (pEntry->fileName[pEntry->fileNameLen-1] == '/') {
995 if (!(flags & MZ_EXTRACT_FILES_ONLY)) {
996 int ret = dirCreateHierarchy(
997 targetFile, UNZIP_DIRMODE, timestamp, false);
998 if (ret != 0) {
999 LOGE("Can't create containing directory for \"%s\": %s\n",
1000 targetFile, strerror(errno));
1001 ok = false;
1002 break;
1003 }
1004 LOGD("Extracted dir \"%s\"\n", targetFile);
1005 }
1006 } else {
1007 /* This is not a directory. First, make sure that
1008 * the containing directory exists.
1009 */
1010 int ret = dirCreateHierarchy(
1011 targetFile, UNZIP_DIRMODE, timestamp, true);
1012 if (ret != 0) {
1013 LOGE("Can't create containing directory for \"%s\": %s\n",
1014 targetFile, strerror(errno));
1015 ok = false;
1016 break;
1017 }
1018
1019 /* With FILES_ONLY set, we need to ignore metadata entirely,
1020 * so treat symlinks as regular files.
1021 */
1022 if (!(flags & MZ_EXTRACT_FILES_ONLY) && mzIsZipEntrySymlink(pEntry)) {
1023 /* The entry is a symbolic link.
1024 * The relative target of the symlink is in the
1025 * data section of this entry.
1026 */
1027 if (pEntry->uncompLen == 0) {
1028 LOGE("Symlink entry \"%s\" has no target\n",
1029 targetFile);
1030 ok = false;
1031 break;
1032 }
1033 char *linkTarget = malloc(pEntry->uncompLen + 1);
1034 if (linkTarget == NULL) {
1035 ok = false;
1036 break;
1037 }
1038 ok = mzReadZipEntry(pArchive, pEntry, linkTarget,
1039 pEntry->uncompLen);
1040 if (!ok) {
1041 LOGE("Can't read symlink target for \"%s\"\n",
1042 targetFile);
1043 free(linkTarget);
1044 break;
1045 }
1046 linkTarget[pEntry->uncompLen] = '\0';
1047
1048 /* Make the link.
1049 */
1050 ret = symlink(linkTarget, targetFile);
1051 if (ret != 0) {
1052 LOGE("Can't symlink \"%s\" to \"%s\": %s\n",
1053 targetFile, linkTarget, strerror(errno));
1054 free(linkTarget);
1055 ok = false;
1056 break;
1057 }
1058 LOGD("Extracted symlink \"%s\" -> \"%s\"\n",
1059 targetFile, linkTarget);
1060 free(linkTarget);
1061 } else {
1062 /* The entry is a regular file.
1063 * Open the target for writing.
1064 */
1065 int fd = creat(targetFile, UNZIP_FILEMODE);
1066 if (fd < 0) {
1067 LOGE("Can't create target file \"%s\": %s\n",
1068 targetFile, strerror(errno));
1069 ok = false;
1070 break;
1071 }
1072
1073 bool ok = mzExtractZipEntryToFile(pArchive, pEntry, fd);
1074 close(fd);
1075 if (!ok) {
1076 LOGE("Error extracting \"%s\"\n", targetFile);
1077 ok = false;
1078 break;
1079 }
1080
1081 if (timestamp != NULL && utime(targetFile, timestamp)) {
1082 LOGE("Error touching \"%s\"\n", targetFile);
1083 ok = false;
1084 break;
1085 }
1086
1087 LOGD("Extracted file \"%s\"\n", targetFile);
1088 }
1089 }
1090
1091 if (callback != NULL) callback(targetFile, cookie);
1092 }
1093
1094 free(helper.buf);
1095 free(zpath);
1096
1097 return ok;
1098}