ZipFileRO.h revision 1f5762e646bed2290934280464832782766ee68e
1/*
2 * Copyright (C) 2007 The Android Open Source Project
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 *      http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17/*
18 * Read-only access to Zip archives, with minimal heap allocation.
19 *
20 * This is similar to the more-complete ZipFile class, but no attempt
21 * has been made to make them interchangeable.  This class operates under
22 * a very different set of assumptions and constraints.
23 *
24 * One such assumption is that if you're getting file descriptors for
25 * use with this class as a child of a fork() operation, you must be on
26 * a pread() to guarantee correct operation. This is because pread() can
27 * atomically read at a file offset without worrying about a lock around an
28 * lseek() + read() pair.
29 */
30#ifndef __LIBS_ZIPFILERO_H
31#define __LIBS_ZIPFILERO_H
32
33#include <utils/Compat.h>
34#include <utils/Errors.h>
35#include <utils/FileMap.h>
36#include <utils/threads.h>
37
38#include <stdio.h>
39#include <stdlib.h>
40#include <unistd.h>
41#include <time.h>
42
43namespace android {
44
45/*
46 * Trivial typedef to ensure that ZipEntryRO is not treated as a simple
47 * integer.  We use NULL to indicate an invalid value.
48 */
49typedef void* ZipEntryRO;
50
51/*
52 * Open a Zip archive for reading.
53 *
54 * We want "open" and "find entry by name" to be fast operations, and we
55 * want to use as little memory as possible.  We memory-map the file,
56 * and load a hash table with pointers to the filenames (which aren't
57 * null-terminated).  The other fields are at a fixed offset from the
58 * filename, so we don't need to extract those (but we do need to byte-read
59 * and endian-swap them every time we want them).
60 *
61 * To speed comparisons when doing a lookup by name, we could make the mapping
62 * "private" (copy-on-write) and null-terminate the filenames after verifying
63 * the record structure.  However, this requires a private mapping of
64 * every page that the Central Directory touches.  Easier to tuck a copy
65 * of the string length into the hash table entry.
66 *
67 * NOTE: If this is used on file descriptors inherited from a fork() operation,
68 * you must be on a platform that implements pread() to guarantee correctness
69 * on the shared file descriptors.
70 */
71class ZipFileRO {
72public:
73    ZipFileRO()
74        : mFd(-1), mFileName(NULL), mFileLength(-1),
75          mDirectoryMap(NULL),
76          mNumEntries(-1), mDirectoryOffset(-1),
77          mHashTableSize(-1), mHashTable(NULL)
78        {}
79
80    ~ZipFileRO();
81
82    /*
83     * Open an archive.
84     */
85    status_t open(const char* zipFileName);
86
87    /*
88     * Find an entry, by name.  Returns the entry identifier, or NULL if
89     * not found.
90     *
91     * If two entries have the same name, one will be chosen at semi-random.
92     */
93    ZipEntryRO findEntryByName(const char* fileName) const;
94
95    /*
96     * Return the #of entries in the Zip archive.
97     */
98    int getNumEntries(void) const {
99        return mNumEntries;
100    }
101
102    /*
103     * Return the Nth entry.  Zip file entries are not stored in sorted
104     * order, and updated entries may appear at the end, so anyone walking
105     * the archive needs to avoid making ordering assumptions.  We take
106     * that further by returning the Nth non-empty entry in the hash table
107     * rather than the Nth entry in the archive.
108     *
109     * Valid values are [0..numEntries).
110     *
111     * [This is currently O(n).  If it needs to be fast we can allocate an
112     * additional data structure or provide an iterator interface.]
113     */
114    ZipEntryRO findEntryByIndex(int idx) const;
115
116    /*
117     * Copy the filename into the supplied buffer.  Returns 0 on success,
118     * -1 if "entry" is invalid, or the filename length if it didn't fit.  The
119     * length, and the returned string, include the null-termination.
120     */
121    int getEntryFileName(ZipEntryRO entry, char* buffer, int bufLen) const;
122
123    /*
124     * Get the vital stats for an entry.  Pass in NULL pointers for anything
125     * you don't need.
126     *
127     * "*pOffset" holds the Zip file offset of the entry's data.
128     *
129     * Returns "false" if "entry" is bogus or if the data in the Zip file
130     * appears to be bad.
131     */
132    bool getEntryInfo(ZipEntryRO entry, int* pMethod, size_t* pUncompLen,
133        size_t* pCompLen, off64_t* pOffset, long* pModWhen, long* pCrc32) const;
134
135    /*
136     * Create a new FileMap object that maps a subset of the archive.  For
137     * an uncompressed entry this effectively provides a pointer to the
138     * actual data, for a compressed entry this provides the input buffer
139     * for inflate().
140     */
141    FileMap* createEntryFileMap(ZipEntryRO entry) const;
142
143    /*
144     * Uncompress the data into a buffer.  Depending on the compression
145     * format, this is either an "inflate" operation or a memcpy.
146     *
147     * Use "uncompLen" from getEntryInfo() to determine the required
148     * buffer size.
149     *
150     * Returns "true" on success.
151     */
152    bool uncompressEntry(ZipEntryRO entry, void* buffer) const;
153
154    /*
155     * Uncompress the data to an open file descriptor.
156     */
157    bool uncompressEntry(ZipEntryRO entry, int fd) const;
158
159    /* Zip compression methods we support */
160    enum {
161        kCompressStored     = 0,        // no compression
162        kCompressDeflated   = 8,        // standard deflate
163    };
164
165    /*
166     * Utility function: uncompress deflated data, buffer to buffer.
167     */
168    static bool inflateBuffer(void* outBuf, const void* inBuf,
169        size_t uncompLen, size_t compLen);
170
171    /*
172     * Utility function: uncompress deflated data, buffer to fd.
173     */
174    static bool inflateBuffer(int fd, const void* inBuf,
175        size_t uncompLen, size_t compLen);
176
177    /*
178     * Utility function to convert ZIP's time format to a timespec struct.
179     */
180    static inline void zipTimeToTimespec(long when, struct tm* timespec) {
181        const long date = when >> 16;
182        timespec->tm_year = ((date >> 9) & 0x7F) + 80; // Zip is years since 1980
183        timespec->tm_mon = (date >> 5) & 0x0F;
184        timespec->tm_mday = date & 0x1F;
185
186        timespec->tm_hour = (when >> 11) & 0x1F;
187        timespec->tm_min = (when >> 5) & 0x3F;
188        timespec->tm_sec = (when & 0x1F) << 1;
189    }
190
191    /*
192     * Some basic functions for raw data manipulation.  "LE" means
193     * Little Endian.
194     */
195    static inline unsigned short get2LE(const unsigned char* buf) {
196        return buf[0] | (buf[1] << 8);
197    }
198    static inline unsigned long get4LE(const unsigned char* buf) {
199        return buf[0] | (buf[1] << 8) | (buf[2] << 16) | (buf[3] << 24);
200    }
201
202private:
203    /* these are private and not defined */
204    ZipFileRO(const ZipFileRO& src);
205    ZipFileRO& operator=(const ZipFileRO& src);
206
207    /* locate and parse the central directory */
208    bool mapCentralDirectory(void);
209
210    /* parse the archive, prepping internal structures */
211    bool parseZipArchive(void);
212
213    /* add a new entry to the hash table */
214    void addToHash(const char* str, int strLen, unsigned int hash);
215
216    /* compute string hash code */
217    static unsigned int computeHash(const char* str, int len);
218
219    /* convert a ZipEntryRO back to a hash table index */
220    int entryToIndex(const ZipEntryRO entry) const;
221
222    /*
223     * One entry in the hash table.
224     */
225    typedef struct HashEntry {
226        const char*     name;
227        unsigned short  nameLen;
228        //unsigned int    hash;
229    } HashEntry;
230
231    /* open Zip archive */
232    int         mFd;
233
234    /* Lock for handling the file descriptor (seeks, etc) */
235    mutable Mutex mFdLock;
236
237    /* zip file name */
238    char*       mFileName;
239
240    /* length of file */
241    size_t      mFileLength;
242
243    /* mapped file */
244    FileMap*    mDirectoryMap;
245
246    /* number of entries in the Zip archive */
247    int         mNumEntries;
248
249    /* CD directory offset in the Zip archive */
250    off64_t     mDirectoryOffset;
251
252    /*
253     * We know how many entries are in the Zip archive, so we have a
254     * fixed-size hash table.  We probe for an empty slot.
255     */
256    int         mHashTableSize;
257    HashEntry*  mHashTable;
258};
259
260}; // namespace android
261
262#endif /*__LIBS_ZIPFILERO_H*/
263