码迷,mamicode.com
首页 > 移动开发 > 详细

Android 系统安装 apk 时解压 so 的逻辑问题

时间:2016-04-25 11:36:10      阅读:676      评论:0      收藏:0      [点我收藏+]

标签:

0X0 前言

在 Android 系统中,当我们安装apk文件的时候,lib 目录下的 so 文件会被解压到 app 的原生库目录,一般来说是放到 /data/data/<package-name>/lib 目录下,而根据系统和CPU架构的不同,其拷贝策略也是不一样的,在我们测试过程中发现不正确地配置了 so 文件,比如某些 app 使用第三方的 so 时,只配置了其中某一种 CPU 架构的 so,可能会造成 app 在某些机型上的适配问题。所以这篇文章主要介绍一下在不同版本的 Android 系统中,安装 apk 时,PackageManagerService 选择解压 so 库的策略,并给出一些 so 文件配置的建议。

0x1 Android4.0以前

当 apk 被安装时,执行路径虽然有差别,但最终要调用到的一个核心函数是 copyApk,负责拷贝 apk 中的资源。

参考2.3.6的 Android 源码,它的 copyApk 其内部函数一段选取原生库 so 逻辑:


public static int listPackageNativeBinariesLI(ZipFile zipFile, List> nativeFiles) throws ZipException, IOException {
    String cpuAbi = Build.CPU_ABI;
    int result = listPackageSharedLibsForAbiLI(zipFile, cpuAbi, nativeFiles);
    /*
     * Some architectures are capable of supporting several CPU ABIs
     * for example, ‘armeabi-v7a‘ also supports ‘armeabi‘ native code
     * this is indicated by the definition of the ro.product.cpu.abi2
     * system property.
     *
     * only scan the package twice in case of ABI mismatch
     */
    if (result == PACKAGE_INSTALL_NATIVE_ABI_MISMATCH) {
        final String cpuAbi2 = SystemProperties.get("ro.product.cpu.abi2", null);
        if (cpuAbi2 != null) {
            result = listPackageSharedLibsForAbiLI(zipFile, cpuAbi2, nativeFiles);
        }
        if (result == PACKAGE_INSTALL_NATIVE_ABI_MISMATCH) {
            Slog.w(TAG, "Native ABI mismatch from package file");
            return PackageManager.INSTALL_FAILED_INVALID_APK;
        }
        if (result == PACKAGE_INSTALL_NATIVE_FOUND_LIBRARIES) {
            cpuAbi = cpuAbi2;
        }
    }
    /*
     * Debuggable packages may have gdbserver embedded, so add it to
     * the list to the list of items to be extracted (as lib/gdbserver)
     * into the application‘s native library directory later.
     */
    if (result == PACKAGE_INSTALL_NATIVE_FOUND_LIBRARIES) {
        listPackageGdbServerLI(zipFile, cpuAbi, nativeFiles);
    }
    return PackageManager.INSTALL_SUCCEEDED;
}
                    

这段代码中的 Build.CPU_ABI 和 "ro.product.cpu.abi2" 分别为手机支持的主 abi 和次 abi 属性字符串,abi 为手机支持的指令集所代表的字符串,比如 armeabi-v7a、armeabi、x86、mips 等,而主 abi 和次 abi 分别表示手机支持的第一指令集和第二指令集。代码首先调用 listPackageSharedLibsForAbiLI 来遍历主 abi 目录。当主 abi 目录不存在时,才会接着调用 listPackageSharedLibsForAbiLI 遍历次 abi 目录。


private static int listPackageSharedLibsForAbiLI(ZipFile zipFile, String cpuAbi, List> libEntries) throws IOException, ZipException {
    final int cpuAbiLen = cpuAbi.length();
    boolean hasNativeLibraries = false;
    boolean installedNativeLibraries = false;
    if (DEBUG_NATIVE) {
        Slog.d(TAG, "Checking " + zipFile.getName() + " for shared libraries of CPU ABI type " + cpuAbi);
    }
    Enumeration entries = zipFile.entries();
    while (entries.hasMoreElements()) {
        ZipEntry entry = entries.nextElement();
        // skip directories
        if (entry.isDirectory()) {
            continue;
        }
        String entryName = entry.getName();
        /*
         * Check that the entry looks like lib//lib.so
         * here, but don‘t check the ABI just yet.
         *
         * - must be sufficiently long
         * - must end with LIB_SUFFIX, i.e. ".so"
         * - must start with APK_LIB, i.e. "lib/"
         */
        if (entryName.length() < MIN_ENTRY_LENGTH || !entryName.endsWith(LIB_SUFFIX) || !entryName.startsWith(APK_LIB)) {
            continue;
        }
        // file name must start with LIB_PREFIX, i.e. "lib"
        int lastSlash = entryName.lastIndexOf(‘/‘);
        if (lastSlash < 0 || !entryName.regionMatches(lastSlash + 1, LIB_PREFIX, 0, LIB_PREFIX_LENGTH)) {
            continue;
        }
        hasNativeLibraries = true;
        // check the cpuAbi now, between lib/ and /lib.so
        if (lastSlash != APK_LIB_LENGTH + cpuAbiLen || !entryName.regionMatches(APK_LIB_LENGTH, cpuAbi, 0, cpuAbiLen))
            continue;
        /*
         * Extract the library file name, ensure it doesn‘t contain
         * weird characters. we‘re guaranteed here that it doesn‘t contain
         * a directory separator though.
         */
        String libFileName = entryName.substring(lastSlash+1);
        if (!FileUtils.isFilenameSafe(new File(libFileName))) {
            continue;
        }
        installedNativeLibraries = true;
        if (DEBUG_NATIVE) {
            Log.d(TAG, "Caching shared lib " + entry.getName());
        }
        libEntries.add(Pair.create(entry, libFileName));
    }
    if (!hasNativeLibraries)
        return PACKAGE_INSTALL_NATIVE_NO_LIBRARIES;
    if (!installedNativeLibraries)
        return PACKAGE_INSTALL_NATIVE_ABI_MISMATCH;
    return PACKAGE_INSTALL_NATIVE_FOUND_LIBRARIES;
}
                    

listPackageSharedLibsForAbiLI 中判断当前遍历的 apk 中文件的 entry 名是否符合 so 命名的规范且包含相应 abi 字符串名。如果符合则规则则将 so 的 entry 名加入 list,如果遍历失败或者规则不匹配则返回相应错误码。

拷贝 so 策略:

遍历 apk 中文件,当 apk 中 lib 目录下主 abi 子目录中有 so 文件存在时,则全部拷贝主 abi 子目录下的 so;只有当主 abi 子目录下没有 so 文件的时候即 PACKAGE_INSTALL_NATIVE_ABI_MISMATCH 的情况,才会拷贝次 ABI 子目录下的 so 文件。

策略问题:

当 so 放置不当时,安装 apk 时会导致拷贝不全。比如 apk 的 lib 目录下存在 armeabi/libx.so , armeabi/liby.so , armeabi-v7a/libx.so 这3个 so 文件,那么在主 ABI 为 armeabi-v7a 且系统版本小于4.0的手机上, apk 安装后,按照拷贝策略,只会拷贝主 abi 目录下的文件即 armeabi-v7a/libx.so,当加载 liby.so 时就会报找不到 so 的异常。另外如果主 abi 目录不存在,这个策略会遍历2次 apk,效率偏低。

0x2 Android 4.0-Android 4.0.3

参考4.0.3的 Android 源码,同理,找到处理 so 拷贝的核心逻辑( native 层):


static install_status_t iterateOverNativeFiles(JNIEnv *env, jstring javaFilePath, jstring javaCpuAbi, jstring javaCpuAbi2, iterFunc callFunc, void* callArg) {
    ScopedUtfChars filePath(env, javaFilePath);
    ScopedUtfChars cpuAbi(env, javaCpuAbi);
    ScopedUtfChars cpuAbi2(env, javaCpuAbi2);
    ZipFileRO zipFile;
    if (zipFile.open(filePath.c_str()) != NO_ERROR) {
        LOGI("Couldn‘t open APK %s\n", filePath.c_str());
        return INSTALL_FAILED_INVALID_APK;
    }
    const int N = zipFile.getNumEntries();
    char fileName[PATH_MAX];
    for (int i = 0; i < N; i++) {
        const ZipEntryRO entry = zipFile.findEntryByIndex(i);
        if (entry == NULL) {
            continue;
        }
        // Make sure this entry has a filename.
        if (zipFile.getEntryFileName(entry, fileName, sizeof(fileName))) {
            continue;
        }
        // Make sure we‘re in the lib directory of the ZIP.
        if (strncmp(fileName, APK_LIB, APK_LIB_LEN)) {
            continue;
        }
        // Make sure the filename is at least to the minimum library name size.
        const size_t fileNameLen = strlen(fileName);
        static const size_t minLength = APK_LIB_LEN + 2 + LIB_PREFIX_LEN + 1 + LIB_SUFFIX_LEN;
        if (fileNameLen < minLength) {
            continue;
        }
        const char* lastSlash = strrchr(fileName, ‘/‘);
        LOG_ASSERT(lastSlash != NULL, "last slash was null somehow for %s\n", fileName);
        // Check to make sure the CPU ABI of this file is one we support.
        const char* cpuAbiOffset = fileName + APK_LIB_LEN;
        const size_t cpuAbiRegionSize = lastSlash - cpuAbiOffset;
        LOGV("Comparing ABIs %s and %s versus %s\n", cpuAbi.c_str(), cpuAbi2.c_str(), cpuAbiOffset);
        if (cpuAbi.size() == cpuAbiRegionSize
                && *(cpuAbiOffset + cpuAbi.size()) == ‘/‘
                && !strncmp(cpuAbiOffset, cpuAbi.c_str(), cpuAbiRegionSize)) {
            LOGV("Using ABI %s\n", cpuAbi.c_str());
        } else if (cpuAbi2.size() == cpuAbiRegionSize
                && *(cpuAbiOffset + cpuAbi2.size()) == ‘/‘
                && !strncmp(cpuAbiOffset, cpuAbi2.c_str(), cpuAbiRegionSize)) {
            LOGV("Using ABI %s\n", cpuAbi2.c_str());
        } else {
            LOGV("abi didn‘t match anything: %s (end at %zd)\n", cpuAbiOffset, cpuAbiRegionSize);
            continue;
        }
        // If this is a .so file, check to see if we need to copy it.
        if ((!strncmp(fileName + fileNameLen - LIB_SUFFIX_LEN, LIB_SUFFIX, LIB_SUFFIX_LEN)
                && !strncmp(lastSlash, LIB_PREFIX, LIB_PREFIX_LEN)
                && isFilenameSafe(lastSlash + 1))
                || !strncmp(lastSlash + 1, GDBSERVER, GDBSERVER_LEN)) {
            install_status_t ret = callFunc(env, callArg, &zipFile, entry, lastSlash + 1);
            if (ret != INSTALL_SUCCEEDED) {
                LOGV("Failure for entry %s", lastSlash + 1);
                return ret;
            }
        }
    }
    return INSTALL_SUCCEEDED;
}
                    

拷贝 so 策略:

遍历 apk 中所有文件,如果符合 so 文件的规则,且为主 ABI 目录或者次 ABI 目录下的 so,就解压拷贝到相应目录。

策略问题:

存在同名 so覆盖,比如一个 app 的 armeabi 和 armeabi-v7a 目录下都包含同名的 so,那么就会发生覆盖现象,覆盖的先后顺序根据 so 文件对应 ZipFileR0 中的 hash 值而定,考虑这样一个例子,假设一个 apk 同时有 armeabi/libx.so 和 armeabi-v7a/libx.so,安装到主 ABI 为 armeabi-v7a 的手机上,拷贝 so 时根据遍历顺序,存在一种可能即 armeab-v7a/libx.so 优先遍历并被拷贝,随后 armeabi/libx.so 被遍历拷贝,覆盖了前者。本来应该加载 armeabi-v7a 目录下的 so,结果按照这个策略拷贝了 armeabi 目录下的 so。

apk 中文件 entry 的散列计算函数如下:


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;
}
                    

0x3 Android 4.0.4以后

以4.1.2系统为例,遍历选择 so 逻辑如下:


static install_status_t iterateOverNativeFiles(JNIEnv *env, jstring javaFilePath, jstring javaCpuAbi, jstring javaCpuAbi2, iterFunc callFunc, void* callArg) {
    ScopedUtfChars filePath(env, javaFilePath);
    ScopedUtfChars cpuAbi(env, javaCpuAbi);
    ScopedUtfChars cpuAbi2(env, javaCpuAbi2);
    ZipFileRO zipFile;
    if (zipFile.open(filePath.c_str()) != NO_ERROR) {
        ALOGI("Couldn‘t open APK %s\n", filePath.c_str());
        return INSTALL_FAILED_INVALID_APK;
    }
    const int N = zipFile.getNumEntries();
    char fileName[PATH_MAX];
    bool hasPrimaryAbi = false;
    for (int i = 0; i < N; i++) {
        const ZipEntryRO entry = zipFile.findEntryByIndex(i);
        if (entry == NULL) {
            continue;
        }
        // Make sure this entry has a filename.
        if (zipFile.getEntryFileName(entry, fileName, sizeof(fileName))) {
            continue;
        }
        // Make sure we‘re in the lib directory of the ZIP.
        if (strncmp(fileName, APK_LIB, APK_LIB_LEN)) {
            continue;
        }
        // Make sure the filename is at least to the minimum library name size.
        const size_t fileNameLen = strlen(fileName);
        static const size_t minLength = APK_LIB_LEN + 2 + LIB_PREFIX_LEN + 1 + LIB_SUFFIX_LEN;
        if (fileNameLen < minLength) {
            continue;
        }
        const char* lastSlash = strrchr(fileName, ‘/‘);
        ALOG_ASSERT(lastSlash != NULL, "last slash was null somehow for %s\n", fileName);
        // Check to make sure the CPU ABI of this file is one we support.
        const char* cpuAbiOffset = fileName + APK_LIB_LEN;
        const size_t cpuAbiRegionSize = lastSlash - cpuAbiOffset;
        ALOGV("Comparing ABIs %s and %s versus %s\n", cpuAbi.c_str(), cpuAbi2.c_str(), cpuAbiOffset);
        if (cpuAbi.size() == cpuAbiRegionSize
                && *(cpuAbiOffset + cpuAbi.size()) == ‘/‘
                && !strncmp(cpuAbiOffset, cpuAbi.c_str(), cpuAbiRegionSize)) {
            ALOGV("Using primary ABI %s\n", cpuAbi.c_str());
            hasPrimaryAbi = true;
        } else if (cpuAbi2.size() == cpuAbiRegionSize
                && *(cpuAbiOffset + cpuAbi2.size()) == ‘/‘
                && !strncmp(cpuAbiOffset, cpuAbi2.c_str(), cpuAbiRegionSize)) {
        /*
         * If this library matches both the primary and secondary ABIs,
         * only use the primary ABI.
         */
            if (hasPrimaryAbi) {
                ALOGV("Already saw primary ABI, skipping secondary ABI %s\n", cpuAbi2.c_str());
                continue;
            } else {
                ALOGV("Using secondary ABI %s\n", cpuAbi2.c_str());
            }
        } else {
            ALOGV("abi didn‘t match anything: %s (end at %zd)\n", cpuAbiOffset, cpuAbiRegionSize);
            continue;
        }
        // If this is a .so file, check to see if we need to copy it.
        if ((!strncmp(fileName + fileNameLen - LIB_SUFFIX_LEN, LIB_SUFFIX, LIB_SUFFIX_LEN)
                && !strncmp(lastSlash, LIB_PREFIX, LIB_PREFIX_LEN)
                && isFilenameSafe(lastSlash + 1))
                || !strncmp(lastSlash + 1, GDBSERVER, GDBSERVER_LEN)) {
            install_status_t ret = callFunc(env, callArg, &zipFile, entry, lastSlash + 1);
            if (ret != INSTALL_SUCCEEDED) {
                ALOGV("Failure for entry %s", lastSlash + 1);
                return ret;
            }
        }
    }
    return INSTALL_SUCCEEDED;
}
                    

拷贝 so 策略:

遍历 apk 中文件,当遍历到有主 Abi 目录的 so 时,拷贝并设置标记 hasPrimaryAbi 为真,以后遍历则只拷贝主 Abi 目录下的 so,当标记为假的时候,如果遍历的 so 的 entry 名包含次 abi 字符串,则拷贝该 so。

策略问题:

经过实际测试, so 放置不当时,安装 apk 时存在 so 拷贝不全的情况。这个策略想解决的问题是在 4.0 ~ 4.0.3 系统中的 so 随意覆盖的问题,即如果有主 abi 目录的 so 则拷贝,如果主 abi 目录不存在这个 so 则拷贝次 abi 目录的 so,但代码逻辑是根据 ZipFileR0 的遍历顺序来决定是否拷贝 so,假设存在这样的 apk, lib 目录下存在 armeabi/libx.so , armeabi/liby.so , armeabi-v7a/libx.so 这三个 so 文件,且 hash 的顺序为 armeabi-v7a/libx.so 在 armeabi/liby.so 之前,则 apk 安装的时候 liby.so 根本不会被拷贝,因为按照拷贝策略, armeabi-v7a/libx.so 会优先遍历到,由于它是主 abi 目录的 so 文件,所以标记被设置了,当遍历到 armeabi/liby.so 时,由于标记被设置为真, liby.so 的拷贝就被忽略了,从而在加载 liby.so 的时候会报异常。

0x4 64位系统支持

Android 在5.0之后支持64位 ABI,以5.1.0系统为例:


public static int copyNativeBinariesWithOverride(Handle handle, File libraryRoot, String abiOverride) {
    try {
        if (handle.multiArch) {
            // Warn if we‘ve set an abiOverride for multi-lib packages..
            // By definition, we need to copy both 32 and 64 bit libraries for
            // such packages.
            if (abiOverride != null && !CLEAR_ABI_OVERRIDE.equals(abiOverride)) {
                Slog.w(TAG, "Ignoring abiOverride for multi arch application.");
            }
            int copyRet = PackageManager.NO_NATIVE_LIBRARIES;
            if (Build.SUPPORTED_32_BIT_ABIS.length > 0) {
                copyRet = copyNativeBinariesForSupportedAbi(handle, libraryRoot,
                        Build.SUPPORTED_32_BIT_ABIS, true /* use isa specific subdirs */);
                if (copyRet < 0 && copyRet != PackageManager.NO_NATIVE_LIBRARIES &&
                        copyRet != PackageManager.INSTALL_FAILED_NO_MATCHING_ABIS) {
                    Slog.w(TAG, "Failure copying 32 bit native libraries; copyRet=" +copyRet);
                    return copyRet;
                }
            }
            if (Build.SUPPORTED_64_BIT_ABIS.length > 0) {
                copyRet = copyNativeBinariesForSupportedAbi(handle, libraryRoot,
                        Build.SUPPORTED_64_BIT_ABIS, true /* use isa specific subdirs */);
                if (copyRet < 0 && copyRet != PackageManager.NO_NATIVE_LIBRARIES &&
                        copyRet != PackageManager.INSTALL_FAILED_NO_MATCHING_ABIS) {
                    Slog.w(TAG, "Failure copying 64 bit native libraries; copyRet=" +copyRet);
                    return copyRet;
                }
            }
        } else {
            String cpuAbiOverride = null;
            if (CLEAR_ABI_OVERRIDE.equals(abiOverride)) {
                cpuAbiOverride = null;
            } else if (abiOverride != null) {
                cpuAbiOverride = abiOverride;
            }
            String[] abiList = (cpuAbiOverride != null) ?
                    new String[] { cpuAbiOverride } : Build.SUPPORTED_ABIS;
            if (Build.SUPPORTED_64_BIT_ABIS.length > 0 && cpuAbiOverride == null &&
                    hasRenderscriptBitcode(handle)) {
                abiList = Build.SUPPORTED_32_BIT_ABIS;
            }
            int copyRet = copyNativeBinariesForSupportedAbi(handle, libraryRoot, abiList,
                    true /* use isa specific subdirs */);
            if (copyRet < 0 && copyRet != PackageManager.NO_NATIVE_LIBRARIES) {
                Slog.w(TAG, "Failure copying native libraries [errorCode=" + copyRet + "]");
                return copyRet;
            }
        }
        return PackageManager.INSTALL_SUCCEEDED;
    } catch (IOException e) {
        Slog.e(TAG, "Copying native libraries failed", e);
        return PackageManager.INSTALL_FAILED_INTERNAL_ERROR;
    }
}
                    

copyNativeBinariesWithOverride 分别处理32位和64位 so 的拷贝,内部函数 copyNativeBinariesForSupportedAbi 首先会根据 abilist 去找对应的 abi。


public static int copyNativeBinariesForSupportedAbi(Handle handle, File libraryRoot, String[] abiList, boolean useIsaSubdir) throws IOException {
    createNativeLibrarySubdir(libraryRoot);
    /*
     * If this is an internal application or our nativeLibraryPath points to
     * the app-lib directory, unpack the libraries if necessary.
     */
    int abi = findSupportedAbi(handle, abiList);
    if (abi >= 0) {
        /*
         * If we have a matching instruction set, construct a subdir under the native
         * library root that corresponds to this instruction set.
         */
        final String instructionSet = VMRuntime.getInstructionSet(abiList[abi]);
        final File subDir;
        if (useIsaSubdir) {
            final File isaSubdir = new File(libraryRoot, instructionSet);
            createNativeLibrarySubdir(isaSubdir);
            subDir = isaSubdir;
        } else {
            subDir = libraryRoot;
        }
        int copyRet = copyNativeBinaries(handle, subDir, abiList[abi]);
        if (copyRet != PackageManager.INSTALL_SUCCEEDED) {
            return copyRet;
        }
    }
    return abi;
}
                    

findSupportedAbi 内部实现是 native 函数,首先遍历 apk,如果 so 的全路径中包含 abilist 中的 abi 字符串,则记录该 abi 字符串的索引,最终返回所有记录索引中最靠前的,即排在 abilist 中最前面的索引。


static int findSupportedAbi(JNIEnv *env, jlong apkHandle, jobjectArray supportedAbisArray) {
    const int numAbis = env->GetArrayLength(supportedAbisArray);
    Vector supportedAbis;
    for (int i = 0; i < numAbis; ++i) {
        supportedAbis.add(new ScopedUtfChars(env,
                (jstring) env->GetObjectArrayElement(supportedAbisArray, i)));
    }
    ZipFileRO* zipFile = reinterpret_cast(apkHandle);
    if (zipFile == NULL) {
        return INSTALL_FAILED_INVALID_APK;
    }
    UniquePtr it(NativeLibrariesIterator::create(zipFile));
    if (it.get() == NULL) {
        return INSTALL_FAILED_INVALID_APK;
    }
    ZipEntryRO entry = NULL;
    char fileName[PATH_MAX];
    int status = NO_NATIVE_LIBRARIES;
    while ((entry = it->next()) != NULL) {
        // We‘re currently in the lib/ directory of the APK, so it does have some native
        // code. We should return INSTALL_FAILED_NO_MATCHING_ABIS if none of the
        // libraries match.
        if (status == NO_NATIVE_LIBRARIES) {
            status = INSTALL_FAILED_NO_MATCHING_ABIS;
        }
        const char* fileName = it->currentEntry();
        const char* lastSlash = it->lastSlash();
        // Check to see if this CPU ABI matches what we are looking for.
        const char* abiOffset = fileName + APK_LIB_LEN;
        const size_t abiSize = lastSlash - abiOffset;
        for (int i = 0; i < numAbis; i++) {
            const ScopedUtfChars* abi = supportedAbis[i];
            if (abi->size() == abiSize && !strncmp(abiOffset, abi->c_str(), abiSize)) {
                // The entry that comes in first (i.e. with a lower index) has the higher priority.
                if (((i < status) && (status >= 0)) || (status < 0) ) {
                    status = i;
                }
            }
        }
    }
    for (int i = 0; i < numAbis; ++i) {
        delete supportedAbis[i];
    }
    return status;
}
                    

举例说明,在某64位测试手机上的abi属性显示如下,它有2个 abilist,分别对应该手机支持的32位和64位 abi 的字符串组。

技术分享

当处理32位 so 拷贝时, findSupportedAbi 索引返回之后,若返回为0,则拷贝 armeabi-v7a 目录下的 so,如果为1,则拷贝 armeabi 目录下 so。

拷贝 so 策略:

分别处理32位和64位 abi 目录的 so 拷贝, abi 由遍历 apk 结果的所有 so 中符合 abilist 列表的最靠前的序号决定,然后拷贝该 abi 目录下的 so 文件。

策略问题:

策略假定每个 abi 目录下的 so 都放置完全的,这是和2.3.6一样的处理逻辑,存在遗漏拷贝 so 的可能。

0x5 建议

针对 Android 系统的这些拷贝策略的问题,我们给出了一些配置 so 的建议:

  1. 1)针对 armeabi 和 armeabi-v7a 两种 ABI

    方法1:由于 armeabi-v7a 指令集兼容 armeabi 指令集,所以如果损失一些应用的性能是可以接受的,同时不希望保留库的两份拷贝,可以移除 armeabi-v7a 目录和其下的库文件,只保留 armeabi 目录;比如 apk 使用第三方的 so 只有 armeabi 这一种 abi 时,可以考虑去掉 apk 中 lib 目录下 armeabi-v7a 目录。

    方法2:在 armeabi 和 armeabi-v7a 目录下各放入一份 so;

  2. 2)针对x86

    目前市面上的x86机型,为了兼容 arm 指令,基本都内置了 libhoudini 模块,即二进制转码支持,该模块负责把 ARM 指令转换为 X86 指令,所以如果是出于 apk 包大小的考虑,并且可以接受一些性能损失,可以选择删掉 x86 库目录, x86 下配置的 armeabi 目录的 so 库一样可以正常加载使用;

  3. 3)针对64位 ABI

    如果 app 开发者打算支持64位,那么64位的 so 要放全,否则可以选择不单独编译64位的 so,全部使用32位的 so,64位机型默认支持32位 so 的加载。比如 apk 使用第三方的 so 只有32位 abi 的 so,可以考虑去掉 apk 中 lib 目录下的64位 abi 子目录,保证 apk 安装后正常使用。

0x6 备注

其实本文是因为在 Android 的 so 加载上遇到很多坑,相信很多朋友都遇到过 UnsatisfiedLinkError 这个错误,反应在用户的机型上也是千差万别,但是有没有想过,可能不是 apk 逻辑的问题,而是 Android 系统在安装 APK 的时候,由于 PackageManager 的问题,并没有拷贝相应的 SO 呢?可以参考下面第4个链接,作者给出了解决方案,就是当出现 UnsatisfiedLinkError 错误时,手动拷贝 so 来解决的。

北京名片印刷:http://www.biyinjishi.com/index/bj/
http://www.biyinjishi.com/sellers/bj/
http://www.biyinjishi.com/bestes/
http://www.biyinjishi.com/nearby/bj/
名片印刷制作:http://www.biyinjishi.com/products/
印刷知识:http://www.biyinjishi.com/kdocs/

http://www.biyinjishi.com/products/a10/
http://www.biyinjishi.com/products/a20/
http://www.biyinjishi.com/products/a30/
http://www.biyinjishi.com/products/a35/
http://www.biyinjishi.com/products/a50/
http://www.biyinjishi.com/products/a60/
http://www.biyinjishi.com/products/a65/
http://www.biyinjishi.com/products/a70/
http://www.biyinjishi.com/products/a99/

 

Android 系统安装 apk 时解压 so 的逻辑问题

标签:

原文地址:http://www.cnblogs.com/SA-Jim/p/5430168.html

(0)
(0)
   
举报
评论 一句话评论(0
登录后才能评论!
© 2014 mamicode.com 版权所有  联系我们:gaon5@hotmail.com
迷上了代码!