From 163d5a9fa742e6a39f058fb9ed726fc31f4ecb28 Mon Sep 17 00:00:00 2001 From: Colin Cross Date: Thu, 22 Mar 2012 18:46:44 -0700 Subject: [PATCH] frameworks/native: move Zip* from libandroidfw to libutils ZipUtils is needed by build/tools, move it from libandroidfw (frameworks/base) to libutils (frameworks/native). Change-Id: Ib8c41134ccdec6d6aa8d49860f8dddef49a24783 --- include/utils/ZipFileCRO.h | 61 ++ include/utils/ZipFileRO.h | 262 ++++++++ include/utils/ZipUtils.h | 67 ++ libs/utils/Android.mk | 9 +- libs/utils/ZipFileCRO.cpp | 54 ++ libs/utils/ZipFileRO.cpp | 931 ++++++++++++++++++++++++++++ libs/utils/ZipUtils.cpp | 343 ++++++++++ libs/utils/tests/Android.mk | 1 + libs/utils/tests/ZipFileRO_test.cpp | 64 ++ 9 files changed, 1790 insertions(+), 2 deletions(-) create mode 100644 include/utils/ZipFileCRO.h create mode 100644 include/utils/ZipFileRO.h create mode 100644 include/utils/ZipUtils.h create mode 100644 libs/utils/ZipFileCRO.cpp create mode 100644 libs/utils/ZipFileRO.cpp create mode 100644 libs/utils/ZipUtils.cpp create mode 100644 libs/utils/tests/ZipFileRO_test.cpp diff --git a/include/utils/ZipFileCRO.h b/include/utils/ZipFileCRO.h new file mode 100644 index 000000000..3e42a958b --- /dev/null +++ b/include/utils/ZipFileCRO.h @@ -0,0 +1,61 @@ +/* + * Copyright (C) 2008 The Android Open Source Project + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +// +// C API for ead-only access to Zip archives, with minimal heap allocation. +// +#ifndef __LIBS_ZIPFILECRO_H +#define __LIBS_ZIPFILECRO_H + +#include +#include +#include + +#include + +#ifdef __cplusplus +extern "C" { +#endif + +/* + * Trivial typedef to ensure that ZipFileCRO is not treated as a simple integer. + */ +typedef void* ZipFileCRO; + +/* + * Trivial typedef to ensure that ZipEntryCRO is not treated as a simple + * integer. We use NULL to indicate an invalid value. + */ +typedef void* ZipEntryCRO; + +extern ZipFileCRO ZipFileXRO_open(const char* path); + +extern void ZipFileCRO_destroy(ZipFileCRO zip); + +extern ZipEntryCRO ZipFileCRO_findEntryByName(ZipFileCRO zip, + const char* fileName); + +extern bool ZipFileCRO_getEntryInfo(ZipFileCRO zip, ZipEntryCRO entry, + int* pMethod, size_t* pUncompLen, + size_t* pCompLen, off64_t* pOffset, long* pModWhen, long* pCrc32); + +extern bool ZipFileCRO_uncompressEntry(ZipFileCRO zip, ZipEntryCRO entry, int fd); + +#ifdef __cplusplus +} +#endif + +#endif /*__LIBS_ZIPFILECRO_H*/ diff --git a/include/utils/ZipFileRO.h b/include/utils/ZipFileRO.h new file mode 100644 index 000000000..547e36a09 --- /dev/null +++ b/include/utils/ZipFileRO.h @@ -0,0 +1,262 @@ +/* + * Copyright (C) 2007 The Android Open Source Project + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* + * Read-only access to Zip archives, with minimal heap allocation. + * + * This is similar to the more-complete ZipFile class, but no attempt + * has been made to make them interchangeable. This class operates under + * a very different set of assumptions and constraints. + * + * One such assumption is that if you're getting file descriptors for + * use with this class as a child of a fork() operation, you must be on + * a pread() to guarantee correct operation. This is because pread() can + * atomically read at a file offset without worrying about a lock around an + * lseek() + read() pair. + */ +#ifndef __LIBS_ZIPFILERO_H +#define __LIBS_ZIPFILERO_H + +#include +#include +#include +#include + +#include +#include +#include +#include + +namespace android { + +/* + * Trivial typedef to ensure that ZipEntryRO is not treated as a simple + * integer. We use NULL to indicate an invalid value. + */ +typedef void* ZipEntryRO; + +/* + * Open a Zip archive for reading. + * + * We want "open" and "find entry by name" to be fast operations, and we + * want to use as little memory as possible. We memory-map the file, + * and load a hash table with pointers to the filenames (which aren't + * null-terminated). The other fields are at a fixed offset from the + * filename, so we don't need to extract those (but we do need to byte-read + * and endian-swap them every time we want them). + * + * To speed comparisons when doing a lookup by name, we could make the mapping + * "private" (copy-on-write) and null-terminate the filenames after verifying + * the record structure. However, this requires a private mapping of + * every page that the Central Directory touches. Easier to tuck a copy + * of the string length into the hash table entry. + * + * NOTE: If this is used on file descriptors inherited from a fork() operation, + * you must be on a platform that implements pread() to guarantee correctness + * on the shared file descriptors. + */ +class ZipFileRO { +public: + ZipFileRO() + : mFd(-1), mFileName(NULL), mFileLength(-1), + mDirectoryMap(NULL), + mNumEntries(-1), mDirectoryOffset(-1), + mHashTableSize(-1), mHashTable(NULL) + {} + + ~ZipFileRO(); + + /* + * Open an archive. + */ + status_t open(const char* zipFileName); + + /* + * Find an entry, by name. Returns the entry identifier, or NULL if + * not found. + * + * If two entries have the same name, one will be chosen at semi-random. + */ + ZipEntryRO findEntryByName(const char* fileName) const; + + /* + * Return the #of entries in the Zip archive. + */ + int getNumEntries(void) const { + return mNumEntries; + } + + /* + * Return the Nth entry. Zip file entries are not stored in sorted + * order, and updated entries may appear at the end, so anyone walking + * the archive needs to avoid making ordering assumptions. We take + * that further by returning the Nth non-empty entry in the hash table + * rather than the Nth entry in the archive. + * + * Valid values are [0..numEntries). + * + * [This is currently O(n). If it needs to be fast we can allocate an + * additional data structure or provide an iterator interface.] + */ + ZipEntryRO findEntryByIndex(int idx) const; + + /* + * Copy the filename into the supplied buffer. Returns 0 on success, + * -1 if "entry" is invalid, or the filename length if it didn't fit. The + * length, and the returned string, include the null-termination. + */ + int getEntryFileName(ZipEntryRO entry, char* buffer, int bufLen) const; + + /* + * Get the vital stats for an entry. Pass in NULL pointers for anything + * you don't need. + * + * "*pOffset" holds the Zip file offset of the entry's data. + * + * Returns "false" if "entry" is bogus or if the data in the Zip file + * appears to be bad. + */ + bool getEntryInfo(ZipEntryRO entry, int* pMethod, size_t* pUncompLen, + size_t* pCompLen, off64_t* pOffset, long* pModWhen, long* pCrc32) const; + + /* + * Create a new FileMap object that maps a subset of the archive. For + * an uncompressed entry this effectively provides a pointer to the + * actual data, for a compressed entry this provides the input buffer + * for inflate(). + */ + FileMap* createEntryFileMap(ZipEntryRO entry) const; + + /* + * Uncompress the data into a buffer. Depending on the compression + * format, this is either an "inflate" operation or a memcpy. + * + * Use "uncompLen" from getEntryInfo() to determine the required + * buffer size. + * + * Returns "true" on success. + */ + bool uncompressEntry(ZipEntryRO entry, void* buffer) const; + + /* + * Uncompress the data to an open file descriptor. + */ + bool uncompressEntry(ZipEntryRO entry, int fd) const; + + /* Zip compression methods we support */ + enum { + kCompressStored = 0, // no compression + kCompressDeflated = 8, // standard deflate + }; + + /* + * Utility function: uncompress deflated data, buffer to buffer. + */ + static bool inflateBuffer(void* outBuf, const void* inBuf, + size_t uncompLen, size_t compLen); + + /* + * Utility function: uncompress deflated data, buffer to fd. + */ + static bool inflateBuffer(int fd, const void* inBuf, + size_t uncompLen, size_t compLen); + + /* + * Utility function to convert ZIP's time format to a timespec struct. + */ + static inline void zipTimeToTimespec(long when, struct tm* timespec) { + const long date = when >> 16; + timespec->tm_year = ((date >> 9) & 0x7F) + 80; // Zip is years since 1980 + timespec->tm_mon = (date >> 5) & 0x0F; + timespec->tm_mday = date & 0x1F; + + timespec->tm_hour = (when >> 11) & 0x1F; + timespec->tm_min = (when >> 5) & 0x3F; + timespec->tm_sec = (when & 0x1F) << 1; + } + + /* + * Some basic functions for raw data manipulation. "LE" means + * Little Endian. + */ + static inline unsigned short get2LE(const unsigned char* buf) { + return buf[0] | (buf[1] << 8); + } + static inline unsigned long get4LE(const unsigned char* buf) { + return buf[0] | (buf[1] << 8) | (buf[2] << 16) | (buf[3] << 24); + } + +private: + /* these are private and not defined */ + ZipFileRO(const ZipFileRO& src); + ZipFileRO& operator=(const ZipFileRO& src); + + /* locate and parse the central directory */ + bool mapCentralDirectory(void); + + /* parse the archive, prepping internal structures */ + bool parseZipArchive(void); + + /* add a new entry to the hash table */ + void addToHash(const char* str, int strLen, unsigned int hash); + + /* compute string hash code */ + static unsigned int computeHash(const char* str, int len); + + /* convert a ZipEntryRO back to a hash table index */ + int entryToIndex(const ZipEntryRO entry) const; + + /* + * One entry in the hash table. + */ + typedef struct HashEntry { + const char* name; + unsigned short nameLen; + //unsigned int hash; + } HashEntry; + + /* open Zip archive */ + int mFd; + + /* Lock for handling the file descriptor (seeks, etc) */ + mutable Mutex mFdLock; + + /* zip file name */ + char* mFileName; + + /* length of file */ + size_t mFileLength; + + /* mapped file */ + FileMap* mDirectoryMap; + + /* number of entries in the Zip archive */ + int mNumEntries; + + /* CD directory offset in the Zip archive */ + off64_t mDirectoryOffset; + + /* + * We know how many entries are in the Zip archive, so we have a + * fixed-size hash table. We probe for an empty slot. + */ + int mHashTableSize; + HashEntry* mHashTable; +}; + +}; // namespace android + +#endif /*__LIBS_ZIPFILERO_H*/ diff --git a/include/utils/ZipUtils.h b/include/utils/ZipUtils.h new file mode 100644 index 000000000..42c42b6c0 --- /dev/null +++ b/include/utils/ZipUtils.h @@ -0,0 +1,67 @@ +/* + * Copyright (C) 2007 The Android Open Source Project + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +// +// Miscellaneous zip/gzip utility functions. +// +#ifndef __LIBS_ZIPUTILS_H +#define __LIBS_ZIPUTILS_H + +#include + +namespace android { + +/* + * Container class for utility functions, primarily for namespace reasons. + */ +class ZipUtils { +public: + /* + * General utility function for uncompressing "deflate" data from a file + * to a buffer. + */ + static bool inflateToBuffer(int fd, void* buf, long uncompressedLen, + long compressedLen); + static bool inflateToBuffer(FILE* fp, void* buf, long uncompressedLen, + long compressedLen); + + /* + * Someday we might want to make this generic and handle bzip2 ".bz2" + * files too. + * + * We could declare gzip to be a sub-class of zip that has exactly + * one always-compressed entry, but we currently want to treat Zip + * and gzip as distinct, so there's no value. + * + * The zlib library has some gzip utilities, but it has no interface + * for extracting the uncompressed length of the file (you do *not* + * want to gzseek to the end). + * + * Pass in a seeked file pointer for the gzip file. If this is a gzip + * file, we set our return values appropriately and return "true" with + * the file seeked to the start of the compressed data. + */ + static bool examineGzip(FILE* fp, int* pCompressionMethod, + long* pUncompressedLen, long* pCompressedLen, unsigned long* pCRC32); + +private: + ZipUtils() {} + ~ZipUtils() {} +}; + +}; // namespace android + +#endif /*__LIBS_ZIPUTILS_H*/ diff --git a/libs/utils/Android.mk b/libs/utils/Android.mk index b4fc994c0..a5a59d8e3 100644 --- a/libs/utils/Android.mk +++ b/libs/utils/Android.mk @@ -42,6 +42,9 @@ commonSources:= \ Unicode.cpp \ VectorImpl.cpp \ WorkQueue.cpp \ + ZipFileCRO.cpp \ + ZipFileRO.cpp \ + ZipUtils.cpp \ misc.cpp host_commonCflags := -DLIBUTILS_NATIVE=1 $(TOOL_CFLAGS) @@ -96,7 +99,8 @@ LOCAL_LDLIBS += -lrt -ldl endif LOCAL_C_INCLUDES += \ - bionic/libc/private + bionic/libc/private \ + external/zlib LOCAL_LDLIBS += -lpthread @@ -104,7 +108,8 @@ LOCAL_SHARED_LIBRARIES := \ liblog \ libcutils \ libdl \ - libcorkscrew + libcorkscrew \ + libz LOCAL_MODULE:= libutils include $(BUILD_SHARED_LIBRARY) diff --git a/libs/utils/ZipFileCRO.cpp b/libs/utils/ZipFileCRO.cpp new file mode 100644 index 000000000..55dfd9f87 --- /dev/null +++ b/libs/utils/ZipFileCRO.cpp @@ -0,0 +1,54 @@ +/* + * Copyright (C) 2008 The Android Open Source Project + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include +#include + +using namespace android; + +ZipFileCRO ZipFileXRO_open(const char* path) { + ZipFileRO* zip = new ZipFileRO(); + if (zip->open(path) == NO_ERROR) { + return (ZipFileCRO)zip; + } + return NULL; +} + +void ZipFileCRO_destroy(ZipFileCRO zipToken) { + ZipFileRO* zip = (ZipFileRO*)zipToken; + delete zip; +} + +ZipEntryCRO ZipFileCRO_findEntryByName(ZipFileCRO zipToken, + const char* fileName) { + ZipFileRO* zip = (ZipFileRO*)zipToken; + return (ZipEntryCRO)zip->findEntryByName(fileName); +} + +bool ZipFileCRO_getEntryInfo(ZipFileCRO zipToken, ZipEntryRO entryToken, + int* pMethod, size_t* pUncompLen, + size_t* pCompLen, off64_t* pOffset, long* pModWhen, long* pCrc32) { + ZipFileRO* zip = (ZipFileRO*)zipToken; + ZipEntryRO entry = (ZipEntryRO)entryToken; + return zip->getEntryInfo(entry, pMethod, pUncompLen, pCompLen, pOffset, + pModWhen, pCrc32); +} + +bool ZipFileCRO_uncompressEntry(ZipFileCRO zipToken, ZipEntryRO entryToken, int fd) { + ZipFileRO* zip = (ZipFileRO*)zipToken; + ZipEntryRO entry = (ZipEntryRO)entryToken; + return zip->uncompressEntry(entry, fd); +} diff --git a/libs/utils/ZipFileRO.cpp b/libs/utils/ZipFileRO.cpp new file mode 100644 index 000000000..cad772075 --- /dev/null +++ b/libs/utils/ZipFileRO.cpp @@ -0,0 +1,931 @@ +/* + * Copyright (C) 2007 The Android Open Source Project + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +// +// Read-only access to Zip archives, with minimal heap allocation. +// +#define LOG_TAG "zipro" +//#define LOG_NDEBUG 0 +#include +#include +#include +#include + +#include + +#include +#include +#include +#include +#include + +#if HAVE_PRINTF_ZD +# define ZD "%zd" +# define ZD_TYPE ssize_t +#else +# define ZD "%ld" +# define ZD_TYPE long +#endif + +/* + * We must open binary files using open(path, ... | O_BINARY) under Windows. + * Otherwise strange read errors will happen. + */ +#ifndef O_BINARY +# define O_BINARY 0 +#endif + +/* + * TEMP_FAILURE_RETRY is defined by some, but not all, versions of + * . (Alas, it is not as standard as we'd hoped!) So, if it's + * not already defined, then define it here. + */ +#ifndef TEMP_FAILURE_RETRY +/* Used to retry syscalls that can return EINTR. */ +#define TEMP_FAILURE_RETRY(exp) ({ \ + typeof (exp) _rc; \ + do { \ + _rc = (exp); \ + } while (_rc == -1 && errno == EINTR); \ + _rc; }) +#endif + +using namespace android; + +/* + * Zip file constants. + */ +#define kEOCDSignature 0x06054b50 +#define kEOCDLen 22 +#define kEOCDNumEntries 8 // offset to #of entries in file +#define kEOCDSize 12 // size of the central directory +#define kEOCDFileOffset 16 // offset to central directory + +#define kMaxCommentLen 65535 // longest possible in ushort +#define kMaxEOCDSearch (kMaxCommentLen + kEOCDLen) + +#define kLFHSignature 0x04034b50 +#define kLFHLen 30 // excluding variable-len fields +#define kLFHNameLen 26 // offset to filename length +#define kLFHExtraLen 28 // offset to extra length + +#define kCDESignature 0x02014b50 +#define kCDELen 46 // excluding variable-len fields +#define kCDEMethod 10 // offset to compression method +#define kCDEModWhen 12 // offset to modification timestamp +#define kCDECRC 16 // offset to entry CRC +#define kCDECompLen 20 // offset to compressed length +#define kCDEUncompLen 24 // offset to uncompressed length +#define kCDENameLen 28 // offset to filename length +#define kCDEExtraLen 30 // offset to extra length +#define kCDECommentLen 32 // offset to comment length +#define kCDELocalOffset 42 // offset to local hdr + +/* + * The values we return for ZipEntryRO use 0 as an invalid value, so we + * want to adjust the hash table index by a fixed amount. Using a large + * value helps insure that people don't mix & match arguments, e.g. to + * findEntryByIndex(). + */ +#define kZipEntryAdj 10000 + +ZipFileRO::~ZipFileRO() { + free(mHashTable); + if (mDirectoryMap) + mDirectoryMap->release(); + if (mFd >= 0) + TEMP_FAILURE_RETRY(close(mFd)); + if (mFileName) + free(mFileName); +} + +/* + * Convert a ZipEntryRO to a hash table index, verifying that it's in a + * valid range. + */ +int ZipFileRO::entryToIndex(const ZipEntryRO entry) const +{ + long ent = ((long) entry) - kZipEntryAdj; + if (ent < 0 || ent >= mHashTableSize || mHashTable[ent].name == NULL) { + ALOGW("Invalid ZipEntryRO %p (%ld)\n", entry, ent); + return -1; + } + return ent; +} + + +/* + * Open the specified file read-only. We memory-map the entire thing and + * close the file before returning. + */ +status_t ZipFileRO::open(const char* zipFileName) +{ + int fd = -1; + + assert(mDirectoryMap == NULL); + + /* + * Open and map the specified file. + */ + fd = ::open(zipFileName, O_RDONLY | O_BINARY); + if (fd < 0) { + ALOGW("Unable to open zip '%s': %s\n", zipFileName, strerror(errno)); + return NAME_NOT_FOUND; + } + + mFileLength = lseek64(fd, 0, SEEK_END); + if (mFileLength < kEOCDLen) { + TEMP_FAILURE_RETRY(close(fd)); + return UNKNOWN_ERROR; + } + + if (mFileName != NULL) { + free(mFileName); + } + mFileName = strdup(zipFileName); + + mFd = fd; + + /* + * Find the Central Directory and store its size and number of entries. + */ + if (!mapCentralDirectory()) { + goto bail; + } + + /* + * Verify Central Directory and create data structures for fast access. + */ + if (!parseZipArchive()) { + goto bail; + } + + return OK; + +bail: + free(mFileName); + mFileName = NULL; + TEMP_FAILURE_RETRY(close(fd)); + return UNKNOWN_ERROR; +} + +/* + * Parse the Zip archive, verifying its contents and initializing internal + * data structures. + */ +bool ZipFileRO::mapCentralDirectory(void) +{ + ssize_t readAmount = kMaxEOCDSearch; + if (readAmount > (ssize_t) mFileLength) + readAmount = mFileLength; + + unsigned char* scanBuf = (unsigned char*) malloc(readAmount); + if (scanBuf == NULL) { + ALOGW("couldn't allocate scanBuf: %s", strerror(errno)); + free(scanBuf); + return false; + } + + /* + * Make sure this is a Zip archive. + */ + if (lseek64(mFd, 0, SEEK_SET) != 0) { + ALOGW("seek to start failed: %s", strerror(errno)); + free(scanBuf); + return false; + } + + ssize_t actual = TEMP_FAILURE_RETRY(read(mFd, scanBuf, sizeof(int32_t))); + if (actual != (ssize_t) sizeof(int32_t)) { + ALOGI("couldn't read first signature from zip archive: %s", strerror(errno)); + free(scanBuf); + return false; + } + + { + unsigned int header = get4LE(scanBuf); + if (header == kEOCDSignature) { + ALOGI("Found Zip archive, but it looks empty\n"); + free(scanBuf); + return false; + } else if (header != kLFHSignature) { + ALOGV("Not a Zip archive (found 0x%08x)\n", header); + free(scanBuf); + return false; + } + } + + /* + * Perform the traditional EOCD snipe hunt. + * + * We're searching for the End of Central Directory magic number, + * which appears at the start of the EOCD block. It's followed by + * 18 bytes of EOCD stuff and up to 64KB of archive comment. We + * need to read the last part of the file into a buffer, dig through + * it to find the magic number, parse some values out, and use those + * to determine the extent of the CD. + * + * We start by pulling in the last part of the file. + */ + off64_t searchStart = mFileLength - readAmount; + + if (lseek64(mFd, searchStart, SEEK_SET) != searchStart) { + ALOGW("seek %ld failed: %s\n", (long) searchStart, strerror(errno)); + free(scanBuf); + return false; + } + actual = TEMP_FAILURE_RETRY(read(mFd, scanBuf, readAmount)); + if (actual != (ssize_t) readAmount) { + ALOGW("Zip: read " ZD ", expected " ZD ". Failed: %s\n", + (ZD_TYPE) actual, (ZD_TYPE) readAmount, strerror(errno)); + free(scanBuf); + return false; + } + + /* + * Scan backward for the EOCD magic. In an archive without a trailing + * comment, we'll find it on the first try. (We may want to consider + * doing an initial minimal read; if we don't find it, retry with a + * second read as above.) + */ + int i; + for (i = readAmount - kEOCDLen; i >= 0; i--) { + if (scanBuf[i] == 0x50 && get4LE(&scanBuf[i]) == kEOCDSignature) { + ALOGV("+++ Found EOCD at buf+%d\n", i); + break; + } + } + if (i < 0) { + ALOGD("Zip: EOCD not found, %s is not zip\n", mFileName); + free(scanBuf); + return false; + } + + off64_t eocdOffset = searchStart + i; + const unsigned char* eocdPtr = scanBuf + i; + + assert(eocdOffset < mFileLength); + + /* + * Grab the CD offset and size, and the number of entries in the + * archive. After that, we can release our EOCD hunt buffer. + */ + unsigned int numEntries = get2LE(eocdPtr + kEOCDNumEntries); + unsigned int dirSize = get4LE(eocdPtr + kEOCDSize); + unsigned int dirOffset = get4LE(eocdPtr + kEOCDFileOffset); + free(scanBuf); + + // Verify that they look reasonable. + if ((long long) dirOffset + (long long) dirSize > (long long) eocdOffset) { + ALOGW("bad offsets (dir %ld, size %u, eocd %ld)\n", + (long) dirOffset, dirSize, (long) eocdOffset); + return false; + } + if (numEntries == 0) { + ALOGW("empty archive?\n"); + return false; + } + + ALOGV("+++ numEntries=%d dirSize=%d dirOffset=%d\n", + numEntries, dirSize, dirOffset); + + mDirectoryMap = new FileMap(); + if (mDirectoryMap == NULL) { + ALOGW("Unable to create directory map: %s", strerror(errno)); + return false; + } + + if (!mDirectoryMap->create(mFileName, mFd, dirOffset, dirSize, true)) { + ALOGW("Unable to map '%s' (" ZD " to " ZD "): %s\n", mFileName, + (ZD_TYPE) dirOffset, (ZD_TYPE) (dirOffset + dirSize), strerror(errno)); + return false; + } + + mNumEntries = numEntries; + mDirectoryOffset = dirOffset; + + return true; +} + +bool ZipFileRO::parseZipArchive(void) +{ + bool result = false; + const unsigned char* cdPtr = (const unsigned char*) mDirectoryMap->getDataPtr(); + size_t cdLength = mDirectoryMap->getDataLength(); + int numEntries = mNumEntries; + + /* + * Create hash table. We have a minimum 75% load factor, possibly as + * low as 50% after we round off to a power of 2. + */ + mHashTableSize = roundUpPower2(1 + (numEntries * 4) / 3); + mHashTable = (HashEntry*) calloc(mHashTableSize, sizeof(HashEntry)); + + /* + * Walk through the central directory, adding entries to the hash + * table. + */ + const unsigned char* ptr = cdPtr; + for (int i = 0; i < numEntries; i++) { + if (get4LE(ptr) != kCDESignature) { + ALOGW("Missed a central dir sig (at %d)\n", i); + goto bail; + } + if (ptr + kCDELen > cdPtr + cdLength) { + ALOGW("Ran off the end (at %d)\n", i); + goto bail; + } + + long localHdrOffset = (long) get4LE(ptr + kCDELocalOffset); + if (localHdrOffset >= mDirectoryOffset) { + ALOGW("bad LFH offset %ld at entry %d\n", localHdrOffset, i); + goto bail; + } + + unsigned int fileNameLen, extraLen, commentLen, hash; + + fileNameLen = get2LE(ptr + kCDENameLen); + extraLen = get2LE(ptr + kCDEExtraLen); + commentLen = get2LE(ptr + kCDECommentLen); + + /* add the CDE filename to the hash table */ + hash = computeHash((const char*)ptr + kCDELen, fileNameLen); + addToHash((const char*)ptr + kCDELen, fileNameLen, hash); + + ptr += kCDELen + fileNameLen + extraLen + commentLen; + if ((size_t)(ptr - cdPtr) > cdLength) { + ALOGW("bad CD advance (%d vs " ZD ") at entry %d\n", + (int) (ptr - cdPtr), (ZD_TYPE) cdLength, i); + goto bail; + } + } + ALOGV("+++ zip good scan %d entries\n", numEntries); + result = true; + +bail: + return result; +} + +/* + * Simple string hash function for non-null-terminated strings. + */ +/*static*/ unsigned int ZipFileRO::computeHash(const char* str, int len) +{ + unsigned int hash = 0; + + while (len--) + hash = hash * 31 + *str++; + + return hash; +} + +/* + * Add a new entry to the hash table. + */ +void ZipFileRO::addToHash(const char* str, int strLen, unsigned int hash) +{ + int ent = hash & (mHashTableSize-1); + + /* + * We over-allocate the table, so we're guaranteed to find an empty slot. + */ + while (mHashTable[ent].name != NULL) + ent = (ent + 1) & (mHashTableSize-1); + + mHashTable[ent].name = str; + mHashTable[ent].nameLen = strLen; +} + +/* + * Find a matching entry. + * + * Returns NULL if not found. + */ +ZipEntryRO ZipFileRO::findEntryByName(const char* fileName) const +{ + /* + * If the ZipFileRO instance is not initialized, the entry number will + * end up being garbage since mHashTableSize is -1. + */ + if (mHashTableSize <= 0) { + return NULL; + } + + int nameLen = strlen(fileName); + unsigned int hash = computeHash(fileName, nameLen); + int ent = hash & (mHashTableSize-1); + + while (mHashTable[ent].name != NULL) { + if (mHashTable[ent].nameLen == nameLen && + memcmp(mHashTable[ent].name, fileName, nameLen) == 0) + { + /* match */ + return (ZipEntryRO)(long)(ent + kZipEntryAdj); + } + + ent = (ent + 1) & (mHashTableSize-1); + } + + return NULL; +} + +/* + * Find the Nth entry. + * + * This currently involves walking through the sparse hash table, counting + * non-empty entries. If we need to speed this up we can either allocate + * a parallel lookup table or (perhaps better) provide an iterator interface. + */ +ZipEntryRO ZipFileRO::findEntryByIndex(int idx) const +{ + if (idx < 0 || idx >= mNumEntries) { + ALOGW("Invalid index %d\n", idx); + return NULL; + } + + for (int ent = 0; ent < mHashTableSize; ent++) { + if (mHashTable[ent].name != NULL) { + if (idx-- == 0) + return (ZipEntryRO) (ent + kZipEntryAdj); + } + } + + return NULL; +} + +/* + * Get the useful fields from the zip entry. + * + * Returns "false" if the offsets to the fields or the contents of the fields + * appear to be bogus. + */ +bool ZipFileRO::getEntryInfo(ZipEntryRO entry, int* pMethod, size_t* pUncompLen, + size_t* pCompLen, off64_t* pOffset, long* pModWhen, long* pCrc32) const +{ + bool ret = false; + + const int ent = entryToIndex(entry); + if (ent < 0) + return false; + + HashEntry hashEntry = mHashTable[ent]; + + /* + * Recover the start of the central directory entry from the filename + * pointer. The filename is the first entry past the fixed-size data, + * so we can just subtract back from that. + */ + const unsigned char* ptr = (const unsigned char*) hashEntry.name; + off64_t cdOffset = mDirectoryOffset; + + ptr -= kCDELen; + + int method = get2LE(ptr + kCDEMethod); + if (pMethod != NULL) + *pMethod = method; + + if (pModWhen != NULL) + *pModWhen = get4LE(ptr + kCDEModWhen); + if (pCrc32 != NULL) + *pCrc32 = get4LE(ptr + kCDECRC); + + size_t compLen = get4LE(ptr + kCDECompLen); + if (pCompLen != NULL) + *pCompLen = compLen; + size_t uncompLen = get4LE(ptr + kCDEUncompLen); + if (pUncompLen != NULL) + *pUncompLen = uncompLen; + + /* + * If requested, determine the offset of the start of the data. All we + * have is the offset to the Local File Header, which is variable size, + * so we have to read the contents of the struct to figure out where + * the actual data starts. + * + * We also need to make sure that the lengths are not so large that + * somebody trying to map the compressed or uncompressed data runs + * off the end of the mapped region. + * + * Note we don't verify compLen/uncompLen if they don't request the + * dataOffset, because dataOffset is expensive to determine. However, + * if they don't have the file offset, they're not likely to be doing + * anything with the contents. + */ + if (pOffset != NULL) { + long localHdrOffset = get4LE(ptr + kCDELocalOffset); + if (localHdrOffset + kLFHLen >= cdOffset) { + ALOGE("ERROR: bad local hdr offset in zip\n"); + return false; + } + + unsigned char lfhBuf[kLFHLen]; + +#ifdef HAVE_PREAD + /* + * This file descriptor might be from zygote's preloaded assets, + * so we need to do an pread64() instead of a lseek64() + read() to + * guarantee atomicity across the processes with the shared file + * descriptors. + */ + ssize_t actual = + TEMP_FAILURE_RETRY(pread64(mFd, lfhBuf, sizeof(lfhBuf), localHdrOffset)); + + if (actual != sizeof(lfhBuf)) { + ALOGW("failed reading lfh from offset %ld\n", localHdrOffset); + return false; + } + + if (get4LE(lfhBuf) != kLFHSignature) { + ALOGW("didn't find signature at start of lfh; wanted: offset=%ld data=0x%08x; " + "got: data=0x%08lx\n", + localHdrOffset, kLFHSignature, get4LE(lfhBuf)); + return false; + } +#else /* HAVE_PREAD */ + /* + * For hosts don't have pread64() we cannot guarantee atomic reads from + * an offset in a file. Android should never run on those platforms. + * File descriptors inherited from a fork() share file offsets and + * there would be nothing to protect from two different processes + * calling lseek64() concurrently. + */ + + { + AutoMutex _l(mFdLock); + + if (lseek64(mFd, localHdrOffset, SEEK_SET) != localHdrOffset) { + ALOGW("failed seeking to lfh at offset %ld\n", localHdrOffset); + return false; + } + + ssize_t actual = + TEMP_FAILURE_RETRY(read(mFd, lfhBuf, sizeof(lfhBuf))); + if (actual != sizeof(lfhBuf)) { + ALOGW("failed reading lfh from offset %ld\n", localHdrOffset); + return false; + } + + if (get4LE(lfhBuf) != kLFHSignature) { + off64_t actualOffset = lseek64(mFd, 0, SEEK_CUR); + ALOGW("didn't find signature at start of lfh; wanted: offset=%ld data=0x%08x; " + "got: offset=" ZD " data=0x%08lx\n", + localHdrOffset, kLFHSignature, (ZD_TYPE) actualOffset, get4LE(lfhBuf)); + return false; + } + } +#endif /* HAVE_PREAD */ + + off64_t dataOffset = localHdrOffset + kLFHLen + + get2LE(lfhBuf + kLFHNameLen) + get2LE(lfhBuf + kLFHExtraLen); + if (dataOffset >= cdOffset) { + ALOGW("bad data offset %ld in zip\n", (long) dataOffset); + return false; + } + + /* check lengths */ + if ((off64_t)(dataOffset + compLen) > cdOffset) { + ALOGW("bad compressed length in zip (%ld + " ZD " > %ld)\n", + (long) dataOffset, (ZD_TYPE) compLen, (long) cdOffset); + return false; + } + + if (method == kCompressStored && + (off64_t)(dataOffset + uncompLen) > cdOffset) + { + ALOGE("ERROR: bad uncompressed length in zip (%ld + " ZD " > %ld)\n", + (long) dataOffset, (ZD_TYPE) uncompLen, (long) cdOffset); + return false; + } + + *pOffset = dataOffset; + } + + return true; +} + +/* + * Copy the entry's filename to the buffer. + */ +int ZipFileRO::getEntryFileName(ZipEntryRO entry, char* buffer, int bufLen) + const +{ + int ent = entryToIndex(entry); + if (ent < 0) + return -1; + + int nameLen = mHashTable[ent].nameLen; + if (bufLen < nameLen+1) + return nameLen+1; + + memcpy(buffer, mHashTable[ent].name, nameLen); + buffer[nameLen] = '\0'; + return 0; +} + +/* + * Create a new FileMap object that spans the data in "entry". + */ +FileMap* ZipFileRO::createEntryFileMap(ZipEntryRO entry) const +{ + /* + * TODO: the efficient way to do this is to modify FileMap to allow + * sub-regions of a file to be mapped. A reference-counting scheme + * can manage the base memory mapping. For now, we just create a brand + * new mapping off of the Zip archive file descriptor. + */ + + FileMap* newMap; + size_t compLen; + off64_t offset; + + if (!getEntryInfo(entry, NULL, NULL, &compLen, &offset, NULL, NULL)) + return NULL; + + newMap = new FileMap(); + if (!newMap->create(mFileName, mFd, offset, compLen, true)) { + newMap->release(); + return NULL; + } + + return newMap; +} + +/* + * Uncompress an entry, in its entirety, into the provided output buffer. + * + * This doesn't verify the data's CRC, which might be useful for + * uncompressed data. The caller should be able to manage it. + */ +bool ZipFileRO::uncompressEntry(ZipEntryRO entry, void* buffer) const +{ + const size_t kSequentialMin = 32768; + bool result = false; + int ent = entryToIndex(entry); + if (ent < 0) + return -1; + + int method; + size_t uncompLen, compLen; + off64_t offset; + const unsigned char* ptr; + + getEntryInfo(entry, &method, &uncompLen, &compLen, &offset, NULL, NULL); + + FileMap* file = createEntryFileMap(entry); + if (file == NULL) { + goto bail; + } + + ptr = (const unsigned char*) file->getDataPtr(); + + /* + * Experiment with madvise hint. When we want to uncompress a file, + * we pull some stuff out of the central dir entry and then hit a + * bunch of compressed or uncompressed data sequentially. The CDE + * visit will cause a limited amount of read-ahead because it's at + * the end of the file. We could end up doing lots of extra disk + * access if the file we're prying open is small. Bottom line is we + * probably don't want to turn MADV_SEQUENTIAL on and leave it on. + * + * So, if the compressed size of the file is above a certain minimum + * size, temporarily boost the read-ahead in the hope that the extra + * pair of system calls are negated by a reduction in page faults. + */ + if (compLen > kSequentialMin) + file->advise(FileMap::SEQUENTIAL); + + if (method == kCompressStored) { + memcpy(buffer, ptr, uncompLen); + } else { + if (!inflateBuffer(buffer, ptr, uncompLen, compLen)) + goto unmap; + } + + if (compLen > kSequentialMin) + file->advise(FileMap::NORMAL); + + result = true; + +unmap: + file->release(); +bail: + return result; +} + +/* + * Uncompress an entry, in its entirety, to an open file descriptor. + * + * This doesn't verify the data's CRC, but probably should. + */ +bool ZipFileRO::uncompressEntry(ZipEntryRO entry, int fd) const +{ + bool result = false; + int ent = entryToIndex(entry); + if (ent < 0) + return -1; + + int method; + size_t uncompLen, compLen; + off64_t offset; + const unsigned char* ptr; + + getEntryInfo(entry, &method, &uncompLen, &compLen, &offset, NULL, NULL); + + FileMap* file = createEntryFileMap(entry); + if (file == NULL) { + goto bail; + } + + ptr = (const unsigned char*) file->getDataPtr(); + + if (method == kCompressStored) { + ssize_t actual = write(fd, ptr, uncompLen); + if (actual < 0) { + ALOGE("Write failed: %s\n", strerror(errno)); + goto unmap; + } else if ((size_t) actual != uncompLen) { + ALOGE("Partial write during uncompress (" ZD " of " ZD ")\n", + (ZD_TYPE) actual, (ZD_TYPE) uncompLen); + goto unmap; + } else { + ALOGI("+++ successful write\n"); + } + } else { + if (!inflateBuffer(fd, ptr, uncompLen, compLen)) + goto unmap; + } + + result = true; + +unmap: + file->release(); +bail: + return result; +} + +/* + * Uncompress "deflate" data from one buffer to another. + */ +/*static*/ bool ZipFileRO::inflateBuffer(void* outBuf, const void* inBuf, + size_t uncompLen, size_t compLen) +{ + bool result = false; + z_stream zstream; + int zerr; + + /* + * Initialize the zlib stream struct. + */ + memset(&zstream, 0, sizeof(zstream)); + zstream.zalloc = Z_NULL; + zstream.zfree = Z_NULL; + zstream.opaque = Z_NULL; + zstream.next_in = (Bytef*)inBuf; + zstream.avail_in = compLen; + zstream.next_out = (Bytef*) outBuf; + zstream.avail_out = uncompLen; + zstream.data_type = Z_UNKNOWN; + + /* + * Use the undocumented "negative window bits" feature to tell zlib + * that there's no zlib header waiting for it. + */ + zerr = inflateInit2(&zstream, -MAX_WBITS); + if (zerr != Z_OK) { + if (zerr == Z_VERSION_ERROR) { + ALOGE("Installed zlib is not compatible with linked version (%s)\n", + ZLIB_VERSION); + } else { + ALOGE("Call to inflateInit2 failed (zerr=%d)\n", zerr); + } + goto bail; + } + + /* + * Expand data. + */ + zerr = inflate(&zstream, Z_FINISH); + if (zerr != Z_STREAM_END) { + ALOGW("Zip inflate failed, zerr=%d (nIn=%p aIn=%u nOut=%p aOut=%u)\n", + zerr, zstream.next_in, zstream.avail_in, + zstream.next_out, zstream.avail_out); + goto z_bail; + } + + /* paranoia */ + if (zstream.total_out != uncompLen) { + ALOGW("Size mismatch on inflated file (%ld vs " ZD ")\n", + zstream.total_out, (ZD_TYPE) uncompLen); + goto z_bail; + } + + result = true; + +z_bail: + inflateEnd(&zstream); /* free up any allocated structures */ + +bail: + return result; +} + +/* + * Uncompress "deflate" data from one buffer to an open file descriptor. + */ +/*static*/ bool ZipFileRO::inflateBuffer(int fd, const void* inBuf, + size_t uncompLen, size_t compLen) +{ + bool result = false; + const size_t kWriteBufSize = 32768; + unsigned char writeBuf[kWriteBufSize]; + z_stream zstream; + int zerr; + + /* + * Initialize the zlib stream struct. + */ + memset(&zstream, 0, sizeof(zstream)); + zstream.zalloc = Z_NULL; + zstream.zfree = Z_NULL; + zstream.opaque = Z_NULL; + zstream.next_in = (Bytef*)inBuf; + zstream.avail_in = compLen; + zstream.next_out = (Bytef*) writeBuf; + zstream.avail_out = sizeof(writeBuf); + zstream.data_type = Z_UNKNOWN; + + /* + * Use the undocumented "negative window bits" feature to tell zlib + * that there's no zlib header waiting for it. + */ + zerr = inflateInit2(&zstream, -MAX_WBITS); + if (zerr != Z_OK) { + if (zerr == Z_VERSION_ERROR) { + ALOGE("Installed zlib is not compatible with linked version (%s)\n", + ZLIB_VERSION); + } else { + ALOGE("Call to inflateInit2 failed (zerr=%d)\n", zerr); + } + goto bail; + } + + /* + * Loop while we have more to do. + */ + do { + /* + * Expand data. + */ + zerr = inflate(&zstream, Z_NO_FLUSH); + if (zerr != Z_OK && zerr != Z_STREAM_END) { + ALOGW("zlib inflate: zerr=%d (nIn=%p aIn=%u nOut=%p aOut=%u)\n", + zerr, zstream.next_in, zstream.avail_in, + zstream.next_out, zstream.avail_out); + goto z_bail; + } + + /* write when we're full or when we're done */ + if (zstream.avail_out == 0 || + (zerr == Z_STREAM_END && zstream.avail_out != sizeof(writeBuf))) + { + long writeSize = zstream.next_out - writeBuf; + int cc = write(fd, writeBuf, writeSize); + if (cc != (int) writeSize) { + ALOGW("write failed in inflate (%d vs %ld)\n", cc, writeSize); + goto z_bail; + } + + zstream.next_out = writeBuf; + zstream.avail_out = sizeof(writeBuf); + } + } while (zerr == Z_OK); + + assert(zerr == Z_STREAM_END); /* other errors should've been caught */ + + /* paranoia */ + if (zstream.total_out != uncompLen) { + ALOGW("Size mismatch on inflated file (%ld vs " ZD ")\n", + zstream.total_out, (ZD_TYPE) uncompLen); + goto z_bail; + } + + result = true; + +z_bail: + inflateEnd(&zstream); /* free up any allocated structures */ + +bail: + return result; +} diff --git a/libs/utils/ZipUtils.cpp b/libs/utils/ZipUtils.cpp new file mode 100644 index 000000000..cf5467ba5 --- /dev/null +++ b/libs/utils/ZipUtils.cpp @@ -0,0 +1,343 @@ +/* + * Copyright (C) 2007 The Android Open Source Project + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +// +// Misc zip/gzip utility functions. +// + +#define LOG_TAG "ziputil" + +#include +#include +#include + +#include +#include +#include + +#include + +using namespace android; + +/* + * Utility function that expands zip/gzip "deflate" compressed data + * into a buffer. + * + * "fd" is an open file positioned at the start of the "deflate" data + * "buf" must hold at least "uncompressedLen" bytes. + */ +/*static*/ bool ZipUtils::inflateToBuffer(int fd, void* buf, + long uncompressedLen, long compressedLen) +{ + bool result = false; + const unsigned long kReadBufSize = 32768; + unsigned char* readBuf = NULL; + z_stream zstream; + int zerr; + unsigned long compRemaining; + + assert(uncompressedLen >= 0); + assert(compressedLen >= 0); + + readBuf = new unsigned char[kReadBufSize]; + if (readBuf == NULL) + goto bail; + compRemaining = compressedLen; + + /* + * Initialize the zlib stream. + */ + memset(&zstream, 0, sizeof(zstream)); + zstream.zalloc = Z_NULL; + zstream.zfree = Z_NULL; + zstream.opaque = Z_NULL; + zstream.next_in = NULL; + zstream.avail_in = 0; + zstream.next_out = (Bytef*) buf; + zstream.avail_out = uncompressedLen; + zstream.data_type = Z_UNKNOWN; + + /* + * Use the undocumented "negative window bits" feature to tell zlib + * that there's no zlib header waiting for it. + */ + zerr = inflateInit2(&zstream, -MAX_WBITS); + if (zerr != Z_OK) { + if (zerr == Z_VERSION_ERROR) { + ALOGE("Installed zlib is not compatible with linked version (%s)\n", + ZLIB_VERSION); + } else { + ALOGE("Call to inflateInit2 failed (zerr=%d)\n", zerr); + } + goto bail; + } + + /* + * Loop while we have data. + */ + do { + unsigned long getSize; + + /* read as much as we can */ + if (zstream.avail_in == 0) { + getSize = (compRemaining > kReadBufSize) ? + kReadBufSize : compRemaining; + ALOGV("+++ reading %ld bytes (%ld left)\n", + getSize, compRemaining); + + int cc = read(fd, readBuf, getSize); + if (cc != (int) getSize) { + ALOGD("inflate read failed (%d vs %ld)\n", + cc, getSize); + goto z_bail; + } + + compRemaining -= getSize; + + zstream.next_in = readBuf; + zstream.avail_in = getSize; + } + + /* uncompress the data */ + zerr = inflate(&zstream, Z_NO_FLUSH); + if (zerr != Z_OK && zerr != Z_STREAM_END) { + ALOGD("zlib inflate call failed (zerr=%d)\n", zerr); + goto z_bail; + } + + /* output buffer holds all, so no need to write the output */ + } while (zerr == Z_OK); + + assert(zerr == Z_STREAM_END); /* other errors should've been caught */ + + if ((long) zstream.total_out != uncompressedLen) { + ALOGW("Size mismatch on inflated file (%ld vs %ld)\n", + zstream.total_out, uncompressedLen); + goto z_bail; + } + + // success! + result = true; + +z_bail: + inflateEnd(&zstream); /* free up any allocated structures */ + +bail: + delete[] readBuf; + return result; +} + +/* + * Utility function that expands zip/gzip "deflate" compressed data + * into a buffer. + * + * (This is a clone of the previous function, but it takes a FILE* instead + * of an fd. We could pass fileno(fd) to the above, but we can run into + * trouble when "fp" has a different notion of what fd's file position is.) + * + * "fp" is an open file positioned at the start of the "deflate" data + * "buf" must hold at least "uncompressedLen" bytes. + */ +/*static*/ bool ZipUtils::inflateToBuffer(FILE* fp, void* buf, + long uncompressedLen, long compressedLen) +{ + bool result = false; + const unsigned long kReadBufSize = 32768; + unsigned char* readBuf = NULL; + z_stream zstream; + int zerr; + unsigned long compRemaining; + + assert(uncompressedLen >= 0); + assert(compressedLen >= 0); + + readBuf = new unsigned char[kReadBufSize]; + if (readBuf == NULL) + goto bail; + compRemaining = compressedLen; + + /* + * Initialize the zlib stream. + */ + memset(&zstream, 0, sizeof(zstream)); + zstream.zalloc = Z_NULL; + zstream.zfree = Z_NULL; + zstream.opaque = Z_NULL; + zstream.next_in = NULL; + zstream.avail_in = 0; + zstream.next_out = (Bytef*) buf; + zstream.avail_out = uncompressedLen; + zstream.data_type = Z_UNKNOWN; + + /* + * Use the undocumented "negative window bits" feature to tell zlib + * that there's no zlib header waiting for it. + */ + zerr = inflateInit2(&zstream, -MAX_WBITS); + if (zerr != Z_OK) { + if (zerr == Z_VERSION_ERROR) { + ALOGE("Installed zlib is not compatible with linked version (%s)\n", + ZLIB_VERSION); + } else { + ALOGE("Call to inflateInit2 failed (zerr=%d)\n", zerr); + } + goto bail; + } + + /* + * Loop while we have data. + */ + do { + unsigned long getSize; + + /* read as much as we can */ + if (zstream.avail_in == 0) { + getSize = (compRemaining > kReadBufSize) ? + kReadBufSize : compRemaining; + ALOGV("+++ reading %ld bytes (%ld left)\n", + getSize, compRemaining); + + int cc = fread(readBuf, 1, getSize, fp); + if (cc != (int) getSize) { + ALOGD("inflate read failed (%d vs %ld)\n", + cc, getSize); + goto z_bail; + } + + compRemaining -= getSize; + + zstream.next_in = readBuf; + zstream.avail_in = getSize; + } + + /* uncompress the data */ + zerr = inflate(&zstream, Z_NO_FLUSH); + if (zerr != Z_OK && zerr != Z_STREAM_END) { + ALOGD("zlib inflate call failed (zerr=%d)\n", zerr); + goto z_bail; + } + + /* output buffer holds all, so no need to write the output */ + } while (zerr == Z_OK); + + assert(zerr == Z_STREAM_END); /* other errors should've been caught */ + + if ((long) zstream.total_out != uncompressedLen) { + ALOGW("Size mismatch on inflated file (%ld vs %ld)\n", + zstream.total_out, uncompressedLen); + goto z_bail; + } + + // success! + result = true; + +z_bail: + inflateEnd(&zstream); /* free up any allocated structures */ + +bail: + delete[] readBuf; + return result; +} + +/* + * Look at the contents of a gzip archive. We want to know where the + * data starts, and how long it will be after it is uncompressed. + * + * We expect to find the CRC and length as the last 8 bytes on the file. + * This is a pretty reasonable thing to expect for locally-compressed + * files, but there's a small chance that some extra padding got thrown + * on (the man page talks about compressed data written to tape). We + * don't currently deal with that here. If "gzip -l" whines, we're going + * to fail too. + * + * On exit, "fp" is pointing at the start of the compressed data. + */ +/*static*/ bool ZipUtils::examineGzip(FILE* fp, int* pCompressionMethod, + long* pUncompressedLen, long* pCompressedLen, unsigned long* pCRC32) +{ + enum { // flags + FTEXT = 0x01, + FHCRC = 0x02, + FEXTRA = 0x04, + FNAME = 0x08, + FCOMMENT = 0x10, + }; + int ic; + int method, flags; + int i; + + ic = getc(fp); + if (ic != 0x1f || getc(fp) != 0x8b) + return false; // not gzip + method = getc(fp); + flags = getc(fp); + + /* quick sanity checks */ + if (method == EOF || flags == EOF) + return false; + if (method != ZipFileRO::kCompressDeflated) + return false; + + /* skip over 4 bytes of mod time, 1 byte XFL, 1 byte OS */ + for (i = 0; i < 6; i++) + (void) getc(fp); + /* consume "extra" field, if present */ + if ((flags & FEXTRA) != 0) { + int len; + + len = getc(fp); + len |= getc(fp) << 8; + while (len-- && getc(fp) != EOF) + ; + } + /* consume filename, if present */ + if ((flags & FNAME) != 0) { + do { + ic = getc(fp); + } while (ic != 0 && ic != EOF); + } + /* consume comment, if present */ + if ((flags & FCOMMENT) != 0) { + do { + ic = getc(fp); + } while (ic != 0 && ic != EOF); + } + /* consume 16-bit header CRC, if present */ + if ((flags & FHCRC) != 0) { + (void) getc(fp); + (void) getc(fp); + } + + if (feof(fp) || ferror(fp)) + return false; + + /* seek to the end; CRC and length are in the last 8 bytes */ + long curPosn = ftell(fp); + unsigned char buf[8]; + fseek(fp, -8, SEEK_END); + *pCompressedLen = ftell(fp) - curPosn; + + if (fread(buf, 1, 8, fp) != 8) + return false; + /* seek back to start of compressed data */ + fseek(fp, curPosn, SEEK_SET); + + *pCompressionMethod = method; + *pCRC32 = ZipFileRO::get4LE(&buf[0]); + *pUncompressedLen = ZipFileRO::get4LE(&buf[4]); + + return true; +} diff --git a/libs/utils/tests/Android.mk b/libs/utils/tests/Android.mk index a6811fcf8..98e4936e8 100644 --- a/libs/utils/tests/Android.mk +++ b/libs/utils/tests/Android.mk @@ -9,6 +9,7 @@ test_src_files := \ Looper_test.cpp \ String8_test.cpp \ Unicode_test.cpp \ + ZipFileRO_test.cpp shared_libraries := \ libz \ diff --git a/libs/utils/tests/ZipFileRO_test.cpp b/libs/utils/tests/ZipFileRO_test.cpp new file mode 100644 index 000000000..7a1d0bd95 --- /dev/null +++ b/libs/utils/tests/ZipFileRO_test.cpp @@ -0,0 +1,64 @@ +/* + * Copyright (C) 2011 The Android Open Source Project + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#define LOG_TAG "ZipFileRO_test" +#include +#include + +#include + +#include +#include + +namespace android { + +class ZipFileROTest : public testing::Test { +protected: + virtual void SetUp() { + } + + virtual void TearDown() { + } +}; + +TEST_F(ZipFileROTest, ZipTimeConvertSuccess) { + struct tm t; + + // 2011-06-29 14:40:40 + long when = 0x3EDD7514; + + ZipFileRO::zipTimeToTimespec(when, &t); + + EXPECT_EQ(2011, t.tm_year + 1900) + << "Year was improperly converted."; + + EXPECT_EQ(6, t.tm_mon) + << "Month was improperly converted."; + + EXPECT_EQ(29, t.tm_mday) + << "Day was improperly converted."; + + EXPECT_EQ(14, t.tm_hour) + << "Hour was improperly converted."; + + EXPECT_EQ(40, t.tm_min) + << "Minute was improperly converted."; + + EXPECT_EQ(40, t.tm_sec) + << "Second was improperly converted."; +} + +}