如何阅读Android framework层源代码

阅读Android Framework层的源码可能是Android 开发者深入学习的必经之路。但在我学习的过程中看到最多的是各路大神的源码分析，而很少有展示如何一步一步找到相关代码的。直到前不久看到老罗的视频《Android源代码情景分析法》，很有启发。但是老罗也只是讲到Java层的分析追踪，在我的日常工作中，经常涉及从Java 层到JNI 层，再到Native层，经常追着追着就丢了。

之后我仿照老罗的方法，尝试追踪MediaRecorder的setVideoEncoder方法是怎么实现的，现在总结如下。

准备：

Source Insight 3

Android Framework层源码

分析开始

首先看MediaRecorder.java内的setVideoEncoder方法：

/** * Sets the video encoder to be used for recording. If this method is not * called, the output file will not contain an video track. Call this after * setOutputFormat() and before prepare(). * * @param video_encoder the video encoder to use. * @throws IllegalStateException if it is called before * setOutputFormat() or after prepare() * @see android.media.MediaRecorder.VideoEncoder */ public native void setVideoEncoder(int video_encoder) throws IllegalStateException;

这是一个native方法，根据JNI的规则，我们应该去android_media_MediaRecorder.cpp里看这个方法的实现：

static void android_media_MediaRecorder_setVideoEncoder(JNIEnv *env, jobject thiz, jint ve) { ALOGV("setVideoEncoder(%d)", ve); if (ve < VIDEO_ENCODER_DEFAULT || ve >= VIDEO_ENCODER_LIST_END) { jniThrowException(env, "java/lang/IllegalArgumentException", "Invalid video encoder"); return; } sp<MediaRecorder> mr = getMediaRecorder(env, thiz); process_media_recorder_call(env, mr->setVideoEncoder(ve), "java/lang/RuntimeException", "setVideoEncoder failed."); }

关键是最后的两句，sp是Android 里的一个指针，就当没看见~ 所以最后两句的意思是调用Native层的MediaRecorder的setVideoEncoder方法。

继续去MediaRecorder.cpp里看setVideoEncoder方法：

status_t MediaRecorder::setVideoEncoder(int ve) { ALOGV("setVideoEncoder(%d)", ve); if (mMediaRecorder == NULL) { ALOGE("media recorder is not initialized yet"); return INVALID_OPERATION; } if (!mIsVideoSourceSet) { ALOGE("try to set the video encoder without setting the video source first"); return INVALID_OPERATION; } if (mIsVideoEncoderSet) { ALOGE("video encoder has already been set"); return INVALID_OPERATION; } if (!(mCurrentState & MEDIA_RECORDER_DATASOURCE_CONFIGURED)) { ALOGE("setVideoEncoder called in an invalid state(%d)", mCurrentState); return INVALID_OPERATION; } status_t ret = mMediaRecorder->setVideoEncoder(ve); if (OK != ret) { ALOGV("setVideoEncoder failed: %d", ret); mCurrentState = MEDIA_RECORDER_ERROR; return ret; } mIsVideoEncoderSet = true; return ret; }

这里的mMediaReocorder定义在MediaRecorder.h里：

sp<IMediaRecorder> mMediaRecorder;

可见其是一个IMediaRecorder类型的变量，这是一个接口类型，看一下其接口定义，在IMediaRecorder.h中：

class IMediaRecorder: public IInterface { public: DECLARE_META_INTERFACE(MediaRecorder); virtual status_t setCamera(const sp<hardware::ICamera>& camera, const sp<ICameraRecordingProxy>& proxy) = 0; virtual status_t setPreviewSurface(const sp<IGraphicBufferProducer>& surface) = 0; virtual status_t setVideoSource(int vs) = 0; virtual status_t setAudioSource(int as) = 0; virtual status_t setOutputFormat(int of) = 0; virtual status_t setVideoEncoder(int ve) = 0; virtual status_t setAudioEncoder(int ae) = 0; virtual status_t setOutputFile(int fd, int64_t offset, int64_t length) = 0; virtual status_t setVideoSize(int width, int height) = 0; virtual status_t setVideoFrameRate(int frames_per_second) = 0; virtual status_t setParameters(const String8& params) = 0; virtual status_t setListener(const sp<IMediaRecorderClient>& listener) = 0; virtual status_t setClientName(const String16& clientName) = 0; virtual status_t prepare() = 0; virtual status_t getMaxAmplitude(int* max) = 0; virtual status_t start() = 0; virtual status_t stop() = 0; virtual status_t reset() = 0; virtual status_t pause() = 0; virtual status_t resume() = 0; virtual status_t init() = 0; virtual status_t close() = 0; virtual status_t release() = 0; virtual status_t setInputSurface(const sp<IGraphicBufferConsumer>& surface) = 0; virtual sp<IGraphicBufferProducer> querySurfaceMediaSource() = 0; }; // ---------------------------------------------------------------------------- class BnMediaRecorder: public BnInterface<IMediaRecorder> { public: virtual status_t onTransact( uint32_t code, const Parcel& data, Parcel* reply, uint32_t flags = 0); };

呐~关键的步骤来了，在以前我基本就追到这，就算完蛋了，不知道怎么继续往下追了，后来看Binder相关的内容，知道了BnInterface的作用，这是Binder中关键的一环，我们这就不展开讲了，只要记得，接下来我们要追的是BnMediaRecorder这个类。

全局搜索一下“: public BnMediaRecorder”,也就是去找它的实现类，发现在MediaRecorderClient.h当中有如下的定义：

class MediaRecorderClient : public BnMediaRecorder { class ServiceDeathNotifier: public IBinder::DeathRecipient { public: ServiceDeathNotifier( const sp<IBinder>& service, const sp<IMediaRecorderClient>& listener, int which); virtual ~ServiceDeathNotifier(); virtual void binderDied(const wp<IBinder>& who); private: int mWhich; sp<IBinder> mService; wp<IMediaRecorderClient> mListener; }; public: virtual status_t setCamera(const sp<hardware::ICamera>& camera, const sp<ICameraRecordingProxy>& proxy); virtual status_t setPreviewSurface(const sp<IGraphicBufferProducer>& surface); virtual status_t setVideoSource(int vs); virtual status_t setAudioSource(int as); virtual status_t setOutputFormat(int of); virtual status_t setVideoEncoder(int ve); virtual status_t setAudioEncoder(int ae); virtual status_t setOutputFile(int fd, int64_t offset, int64_t length); virtual status_t setVideoSize(int width, int height); virtual status_t setVideoFrameRate(int frames_per_second); virtual status_t setParameters(const String8& params); virtual status_t setListener( const sp<IMediaRecorderClient>& listener); virtual status_t setClientName(const String16& clientName); virtual status_t prepare(); virtual status_t getMaxAmplitude(int* max); virtual status_t start(); virtual status_t stop(); virtual status_t reset(); virtual status_t pause(); virtual status_t resume(); virtual status_t init(); virtual status_t close(); virtual status_t release(); virtual status_t dump(int fd, const Vector<String16>& args); virtual status_t setInputSurface(const sp<IGraphicBufferConsumer>& surface); virtual sp<IGraphicBufferProducer> querySurfaceMediaSource(); private: friend class MediaPlayerService; // for accessing private constructor MediaRecorderClient( const sp<MediaPlayerService>& service, pid_t pid, const String16& opPackageName); virtual ~MediaRecorderClient(); sp<IBinder::DeathRecipient> mCameraDeathListener; sp<IBinder::DeathRecipient> mCodecDeathListener; pid_t mPid; Mutex mLock; MediaRecorderBase *mRecorder; sp<MediaPlayerService> mMediaPlayerService; };

那么我们去MediaRecorderClient.cpp里看看：

status_t MediaRecorderClient::setVideoEncoder(int ve) { ALOGV("setVideoEncoder(%d)", ve); Mutex::Autolock lock(mLock); if (mRecorder == NULL) { ALOGE("recorder is not initialized"); return NO_INIT; } return mRecorder->setVideoEncoder((video_encoder)ve); }

这里的mRecorder 定义在MediaRecorderClient.h当中：

MediaRecorderBase *mRecorder;

来，我们继续搜索“: public MediaRecorderBase”，在StagefrightRecorder.h中找到了，那么我们去StagefrightRecorder.cpp当中看看~

status_t StagefrightRecorder::setVideoEncoder(video_encoder ve) { ALOGV("setVideoEncoder: %d", ve); if (ve < VIDEO_ENCODER_DEFAULT || ve >= VIDEO_ENCODER_LIST_END) { ALOGE("Invalid video encoder: %d", ve); return BAD_VALUE; } mVideoEncoder = ve; return OK; }

呐，这就算完了~最后我们会发现MediaRecorder中的接口实现基本到最后都会落在这个StagefrightRecorder.cpp当中。

总结

在Android源码阅读当中，由于大多数人都没接触过大型的C++工程，所以在Native层的追踪上一开始会有较大困难，但只要静下心来，会发现其实只要记住C++的继承、实现、接口的相关概念，和一般的读代码没啥区别。