Android的Surface的创建

ViewRootImpl管理着整个view tree。 对于ViewRootImpl.setView()，我们可以简单的把它当做一个UI渲染操作的入口。

http://androidxref.com/6.0.1_r10/xref/frameworks/base/core/java/android/view/WindowManagerImpl.java

public void setView(View view, WindowManager.LayoutParams attrs, View panelParentView) {
    ...
    //mWindowSession是一个aidl，ViewRootImpl利用它来和WindowManagerService交互
    //mWindow是一个aidl，WindowManagerService可以利用这个对象与服务端交互
    //mAttachInfo可以理解为是一个data bean，可以跨进程传递
    res = mWindowSession.addToDisplay(mWindow, mSeq, mWindowAttributes,
            getHostVisibility(), mDisplay.getDisplayId(), mWinFrame,
            mAttachInfo.mContentInsets, mAttachInfo.mStableInsets,
            mAttachInfo.mOutsets, mAttachInfo.mDisplayCutout, mInputChannel);
    ...
}

ViewRootImpl.setView()方法会向WindowManagerService请求添加一个Window,mWindowSession.addToDisplay()跨进程最终调用到了WindowManagerService.addWindow():

http://androidxref.com/6.0.1_r10/xref/frameworks/base/services/core/java/com/android/server/wm/WindowManagerService.java

public int addWindow(Session session, IWindow client...) {
    ...
    //WindowState用来描述一个Window
    final WindowState win = new WindowState(this, session, client, token, parentWindow,
                appOp[0], seq, attrs, viewVisibility, session.mUid,
                session.mCanAddInternalSystemWindow);
    ...
    win.attach();  //会创建一个SurfaceSession

    mWindowMap.put(client.asBinder(), win); //mWindowMap是WindowManagerService用来保存当前所有Window新的的集合
    ...
    win.mToken.addWindow(win); //一个token下会有多个win state。 其实token与PhoneWindow是一一对应的。
    ...
}

http://androidxref.com/6.0.1_r10/xref/frameworks/base/services/core/java/com/android/server/wm/WindowState.java

 void attach() {
       if (WindowManagerService.localLOGV) Slog.v(
            TAG, "Attaching " + this + " token=" + mToken
            + ", list=" + mToken.windows);
        mSession.windowAddedLocked();
   }

WindowState是WindowManagerService用来描述应用程序的一个Window的对象。上面注释我标注了win.attach(),这个方法可以说是Window与SurfaceFlinger链接的起点,它最终会调用到Session.windowAddedLocked():

http://androidxref.com/6.0.1_r10/xref/frameworks/base/services/core/java/com/android/server/wm/Session.java

void windowAddedLocked(String packageName) {
    ...
    if (mSurfaceSession == null) {
        ...
        mSurfaceSession = new SurfaceSession();
        ...
    }
}

http://androidxref.com/6.0.1_r10/xref/frameworks/base/core/java/android/view/SurfaceSession.java

//SurfaceSession类的构造方法
public final class SurfaceSession {
    private long mNativeClient; // SurfaceComposerClient*

    public SurfaceSession() {
        mNativeClient = nativeCreate();
}

这里调用了native方法nativeCreate()，这个方法其实是返回了一个SurfaceComposerClient指针。那这个对象是怎么创建的呢？

SurfaceComposerClient的创建

http://androidxref.com/6.0.1_r10/xref/frameworks/base/core/jni/android_view_SurfaceSession.cpp

static jlong nativeCreate(JNIEnv* env, jclass clazz) {
    SurfaceComposerClient* client = new SurfaceComposerClient(); //构造函数其实并没有做什么
    client->incStrong((void*)nativeCreate);
    return reinterpret_cast<jlong>(client);
}

即构造了一个SurfaceComposerClient对象。并返回它的指针。这个对象一个应用程序就有一个，它是应用程序与SurfaceFlinger沟通的桥梁，为什么这么说呢？在SurfaceComposerClient指针第一次使用时会调用下面这个方法:

//这个方法在第一次使用SurfaceComposerClient的指针的时候会调用
void SurfaceComposerClient::onFirstRef() {
    ....
    sp<ISurfaceComposerClient> conn;
    //sf 就是SurfaceFlinger
    conn = (rootProducer != nullptr) ? sf->createScopedConnection(rootProducer) :
            sf->createConnection();
    ...
}

即通过SurfaceFlinger(它本身具有跨进程通信的能力)创建了一个ISurfaceComposerClient对象:

http://androidxref.com/6.0.1_r10/xref/frameworks/native/services/surfaceflinger/SurfaceFlinger.cpp

sp<ISurfaceComposerClient> SurfaceFlinger::createConnection() {
    return initClient(new Client(this)); //initClient这个方法其实并没有做什么，
}

即构造了一个Client对象，Client实现了ISurfaceComposerClient接口。是一个可以跨进程通信的aidl对象。即SurfaceComposerClient可以通过它来和SurfaceFlinger通信。除此之外它还可以创建Surface，并且维护一个应用程序的所有Layer(下文会分析到它是什么)。它是一个十分重要的对象，我们先来看一下它的组成,它所涉及的其他东西在下文分析中都会讲到:

http://androidxref.com/6.0.1_r10/xref/frameworks/native/services/surfaceflinger/Client.h

class Client : public BnSurfaceComposerClient
{
public:
    ...
    void attachLayer(const sp<IBinder>& handle, const sp<Layer>& layer);
    void detachLayer(const Layer* layer);
    ...
private:
    // ISurfaceComposerClient interface。   gbp很重要，它维护这一个应用程序的渲染 Buffer队列
    virtual status_t createSurface(...sp<IBinder>* handle, sp<IGraphicBufferProducer>* gbp);

    virtual status_t destroySurface(const sp<IBinder>& handle); 

    //跨进程通信方法
    virtual status_t onTransact(uint32_t code, const Parcel& data, Parcel* reply, uint32_t flags);
    ...

    // constant
    sp<SurfaceFlinger> mFlinger;

    // protected by mLock
    DefaultKeyedVector< wp<IBinder>, wp<Layer> > mLayers; // 一个应用程序的所有Layer
    ...
};

经过上面这一顿源码分析，我们大概知道了ViewRootImpl.setView()所引发的主要操作:

1. WindowManagerService创建了一个WindowState。用来表示客户端的一个Window
2. WindowManagerService创建了一个SurfaceSession,SurfaceSession会与SurfaceFlinger构建链接，创建了一个SurfaceComposerClient对象，一个应用程序只具有一个这个对象。
3. SurfaceComposerClient创建了一个Client, 这个对象十分重要，它维护这应用程序的渲染核心数据，并负责与SurfaceFlinger通信。

经过上面的步骤，应用程序的ViewRootImpl已经被WindowManagerService识别，并且应用程序已经与SurfaceFlinger建立连接。即创建了SurfaceComposerClient和Client对象

文章开始就已经说了Surface是Window(ViewRootImpl)的UI载体，那Surface是在哪里创建的呢？

Surface的创建

其实一个ViewRootImpl就对应一个Surface。

http://androidxref.com/6.0.1_r10/xref/frameworks/base/core/java/android/view/Surface.java

这点可以通过ViewRootImpl的源码看出:

即ViewRootImpl在构造的时候就new 了一个 Surface。但其实这个新new的Surface并没有什么逻辑，它的构造函数是空的。

http://androidxref.com/6.0.1_r10/xref/frameworks/base/core/java/android/view/ViewRootImpl.java

public final class ViewRootImpl implements ViewParent,
        View.AttachInfo.Callbacks, HardwareRenderer.HardwareDrawCallbacks {
//...
public final Surface mSurface = new Surface();
//...
public void setView(View view, WindowManager.LayoutParams attrs, View panelParentView) {
    ...
    requestLayout(); //susion 请求layout。先添加到待渲染队列中
    ...
    res = mWindowSession.addToDisplay(mWindow, ...); //WindowManagerService会创建mWindow对应的WindowState
    ...
}
 @Override
 public void requestLayout() {
        if (!mHandlingLayoutInLayoutRequest) {
            checkThread();
            mLayoutRequested = true;
            scheduleTraversals();
        }
    }

}

即在向WindowManagerService请求创建WindowState之前，调用了requestLayout(),这个方法会引起ViewRootImpl所管理的整个view tree的重新渲染。它最终会调用到scheduleTraversals():

http://androidxref.com/6.0.1_r10/xref/frameworks/base/core/java/android/view/ViewRootImpl.java

void scheduleTraversals() {
    ...
    mChoreographer.postCallback(Choreographer.CALLBACK_TRAVERSAL, mTraversalRunnable, null);
    ...
}

scheduleTraversals()会通过Choreographer来post一个mTraversalRunnable，Choreographer接收显示系统的时间脉冲(垂直同步信号-VSync信号，16ms发出一次)，在下一个frame渲染时控制执行这个mTraversalRunnable。

但是mTraversalRunnable的执行至少要在应用程序与SurfaceFlinger建立连接之后。这是因为渲染操作是由SurfaceFlinger负责调度了，如果应用程序还没有与SurfaceFlinger创建连接，那SurfaceFlinger当然不会渲染这个应用程序。所以在执行完mWindowSession.addToDisplay(mWindow, ...)之后，才会执行mTraversalRunnable:

final class TraversalRunnable implements Runnable {
    @Override
    public void run() {
        doTraversal();
    }
}

doTraversal()会调用到ViewRootImpl.performTraversals(),大部分同学可能知道这个方法是一个view tree的measure/layout/draw的控制方法:

private void performTraversals() {
    finalView host = mView; //mView是一个Window的根View，对于Activity来说就是DecorView
    ...
    relayoutWindow(params, viewVisibility, insetsPending);
    ...
    performMeasure(childWidthMeasureSpec, childHeightMeasureSpec);
    ...
    performLayout(lp, mWidth, mHeight);
    ...
    performDraw();
    ...
}

Surface的具体创建就由relayoutWindow(params, viewVisibility, insetsPending)这个方法来完成的。这个方法会通过IPC调用到WindowManagerService.relayoutWindow():

http://androidxref.com/6.0.1_r10/xref/frameworks/base/core/java/android/view/ViewRootImpl.java

private int relayoutWindow(WindowManager.LayoutParams params, ...) throws RemoteException {
    ...
    int relayoutResult = mWindowSession.relayout(mWindow,..., mSurface);
    ...
}

mWindowSession.relayout()方法的很多参数，不过有一个十分重要的参数我没有省略，就是mSurface。前面已经分析了它就是一个空的Surface对象。其实:

真正的Surface创建是由SurfaceControl完成的，应用程序ViewRootImpl的Surface只是一个指针，指向这个Surface

下面就来看一下SurfaceControl是如何创建Surface的:

mWindowSession.relayout()会调用到WindowManagerService.relayoutWindow():

http://androidxref.com/6.0.1_r10/xref/frameworks/base/services/core/java/com/android/server/wm/WindowManagerService.java

 //这里的outSurface其实就是ViewRootImpl中的那个Surface
public int relayoutWindow(Session session, IWindow client....Surface outSurface){
    ...
    result = createSurfaceControl(outSurface, result, win, winAnimator);
    ...
}

private int createSurfaceControl(Surface outSurface, int result, WindowState win,WindowStateAnimator winAnimator) {
    ...
    surfaceController = winAnimator.createSurfaceLocked(win.mAttrs.type, win.mOwnerUid);
    ...
    surfaceController.getSurface(outSurface);
}

winAnimator.createSurfaceLocked实际上是创建了一个SurfaceControl。即上面是先构造SurfaceControl，然后在构造Surface。

http://androidxref.com/6.0.1_r10/xref/frameworks/base/services/core/java/com/android/server/wm/WindowStateAnimator.java

 SurfaceControl createSurfaceLocked() {
        final WindowState w = mWin;
        if (mSurfaceControl == null) {
           //...
             mSurfaceFormat = format;
                if (DEBUG_SURFACE_TRACE) {
                    mSurfaceControl = new SurfaceTrace(
                            mSession.mSurfaceSession,
                            attrs.getTitle().toString(),
                            width, height, format, flags);
                } else {
                    mSurfaceControl = new SurfaceControl(
                        mSession.mSurfaceSession,
                        attrs.getTitle().toString(),
                        width, height, format, flags);
                }
             //...
        }
        return mSurfaceControl;
    }        

SurfaceControl的创建

winAnimator.createSurfaceLocked其实是通过SurfaceControl的构造函数创建了一个SurfaceControl对象,这个对象的作用其实就是负责维护Surface,Surface其实也是由这个对象负责创建的，我们看一下这个对象的构造方法:

http://androidxref.com/6.0.1_r10/xref/frameworks/base/core/java/android/view/SurfaceControl.java

long mNativeObject; //成员指针变量，指向native创建的SurfaceControl

private SurfaceControl(SurfaceSession session, String name, int w, int h, int format, int flags,
            SurfaceControl parent, int windowType, int ownerUid){
    ...
    mNativeObject = nativeCreate(session, name, w, h, format, flags,
        parent != null ? parent.mNativeObject : 0, windowType, ownerUid);
    ...
}

即调用了nativeCreate()并返回一个SurfaceControl指针:

http://androidxref.com/6.0.1_r10/xref/frameworks/native/libs/gui/SurfaceControl.cpp

static jlong nativeCreate(JNIEnv* env, ...) {
    //这个client其实就是前面创建的SurfaceComposerClinent
    sp<SurfaceComposerClient> client(android_view_SurfaceSession_getClient(env, sessionObj));
    sp<SurfaceControl> surface; //创建成功之后，这个指针会指向新创建的SurfaceControl
    status_t err = client->createSurfaceChecked(String8(name.c_str()), w, h, format, &surface, flags, parent, windowType, ownerUid);
    ...
    return reinterpret_cast<jlong>(surface.get()); //返回这个SurfaceControl的地址
}

即调用到SurfaceComposerClient.createSurfaceChecked():

http://androidxref.com/6.0.1_r10/xref/frameworks/native/libs/gui/SurfaceComposerClient.cpp

//outSurface会指向新创建的SurfaceControl
status_t SurfaceComposerClient::createSurfaceChecked(...sp<SurfaceControl>* outSurface..)
{
    sp<IGraphicBufferProducer> gbp; //这个对象很重要
    ...
    err = mClient->createSurface(name, w, h, format, flags, parentHandle, windowType, ownerUid, &handle, &gbp);
    if (err == NO_ERROR) {
        //SurfaceControl创建成功, 指针赋值
        *outSurface = new SurfaceControl(this, handle, gbp, true);
    }
    return err;
}

上面这个方法实际上是调用Client.createSurface()来创建一个Surface。在创建时有一个很重要的参数sp<IGraphicBufferProducer> gbp,在下面源码分析中我们也要重点注意它。这是因为应用所渲染的每一帧，实际上都会添加到IGraphicBufferProducer中，来等待SurfaceFlinger的渲染。

http://androidxref.com/6.0.1_r10/xref/frameworks/native/services/surfaceflinger/Client.cpp

status_t Client::createSurface(...)
{
    ...
    //gbp 直接透传到了SurfaceFlinger
    return mFlinger->createLayer(name, this, w, h, format, flags, windowType, ownerUid, handle, gbp, &parent);
}

Surface在SurfaceFlinger中对应的实体其实是Layer

我们继续看一下mFlinger->createLayer()

http://androidxref.com/6.0.1_r10/xref/frameworks/native/services/surfaceflinger/SurfaceFlinger.cpp

status_t SurfaceFlinger::createLayer(const String8& name,const sp<Client>& client...)
{
    status_t result = NO_ERROR;
    sp<Layer> layer; //将要创建的layer
    switch (flags & ISurfaceComposerClient::eFXSurfaceMask) {
        case ISurfaceComposerClient::eFXSurfaceNormal:
            result = createBufferLayer(client,
                    uniqueName, w, h, flags, format,
                    handle, gbp, &layer); // 注意gbp，这时候还没有构造呢！
            break;
            ... //Layer 分为好几种，这里不全部列出
    }
    ...
    result = addClientLayer(client, *handle, *gbp, layer, *parent);  //这个layer和client相关联, 添加到Client的mLayers集合中。
    ...
    return result;
}

从SurfaceFlinger.createLayer()方法可以看出Layer分为好几种。我们这里只对普通的BufferLayer的创建做一下分析，看createBufferLayer():

http://androidxref.com/6.0.1_r10/xref/frameworks/native/services/surfaceflinger/SurfaceFlinger.cpp

status_t SurfaceFlinger::createBufferLayer(const sp<Client>& client... sp<Layer>* outLayer)
{
    ...
    sp<BufferLayer> layer = new BufferLayer(this, client, name, w, h, flags);
    status_t err = layer->setBuffers(w, h, format, flags);  //设置layer的宽高
    if (err == NO_ERROR) {
        *handle = layer->getHandle(); //创建handle
        *gbp = layer->getProducer(); //创建 gbp IGraphicBufferProducer
        *outLayer = layer; //把新建的layer的指针拷贝给outLayer,这样outLayer就指向了新建的BufferLayer
    }
    return err;
}

前面我说过IGraphicBufferProducer(gbp)是一个很重要的对象，它涉及到SurfaceFlinger的渲染逻辑，下面我们就看一下这个对象的创建逻辑:

IGraphicBufferProducer(gbp)的创建

sp<IGraphicBufferProducer> BufferLayer::getProducer() const {
    return mProducer;
}

即mProducer其实是Layer的成员变量，它的创建时机是Layer第一次被使用时:

void BufferLayer::onFirstRef() {
    ...
    BufferQueue::createBufferQueue(&producer, &consumer, true);
    mProducer = new MonitoredProducer(producer, mFlinger, this);
    ...
}

所以mProducer的实例是MonitoredProducer,但其实它只是一个装饰类，它实际功能都委托给构造它的参数producer:

BufferQueue.cpp

void BufferQueue::createBufferQueue(sp<IGraphicBufferProducer>* outProducer,
    ...
    sp<IGraphicBufferProducer> producer(new BufferQueueProducer(core, consumerIsSurfaceFlinger));
    sp<IGraphicBufferConsumer> consumer(new BufferQueueConsumer(core)); //注意这个consumer
    ...
    *outProducer = producer;
    *outConsumer = consumer;
}

所以实际实现mProducer的工作的queueProducer是BufferQueueProducer。

所以构造一个SurfaceControl所做的工作就是创建了一个SurfaceControl,并让SurfaceFlinger创建了一个对应的Layer，Layer中有一个IGraphicBufferProducer,它的实例是BufferQueueProducer。

可以用下面这个图来描述SurfaceControl的创建过程:

从SurfaceControl中获取Surface

我们回看WindowManagerService.createSurfaceControl()， 来看一下java层的Surface对象到底是个什么:

http://androidxref.com/6.0.1_r10/xref/frameworks/base/services/core/java/com/android/server/wm/WindowManagerService.java

private int createSurfaceControl(Surface outSurface, int result, WindowState win,WindowStateAnimator winAnimator) {
    ...
    surfaceController = winAnimator.createSurfaceLocked(win.mAttrs.type, win.mOwnerUid);
    ...
    surfaceController.getSurface(outSurface);
}

上面我们已经了解了winAnimator.createSurfaceLocked的整个过程，我们看一下surfaceController.getSurface(outSurface), surfaceController是WindowSurfaceController的实例:

//WindowSurfaceController.java
void getSurface(Surface outSurface) {
    outSurface.copyFrom(mSurfaceControl);
}

//Surface.java
public void copyFrom(SurfaceControl other) {
    ...
    long surfaceControlPtr = other.mNativeObject;
    ...
    long newNativeObject = nativeGetFromSurfaceControl(surfaceControlPtr);
    ...
    mNativeObject = ptr; // mNativeObject指向native创建的Surface
}

即Surface.copyFrom()方法调用nativeGetFromSurfaceControl()来获取一个指针，这个指针是根据前面创建的SurfaceControl的指针来寻找的，即传入的参数surfaceControlPtr:

android_view_Surface.cpp

static jlong nativeGetFromSurfaceControl(JNIEnv* env, jclass clazz, jlong surfaceControlNativeObj) {
    sp<SurfaceControl> ctrl(reinterpret_cast<SurfaceControl *>(surfaceControlNativeObj)); //把java指针转化内native指针
    sp<Surface> surface(ctrl->getSurface()); //直接构造一个Surface，指向 ctrl->getSurface()
    if (surface != NULL) {
        surface->incStrong(&sRefBaseOwner); //强引用
    }
    return reinterpret_cast<jlong>(surface.get());
}

这里的ctrl指向前面创建的SurfaceControl，继续追溯ctrl->getSurface():

sp<Surface> SurfaceControl::getSurface() const
{
    Mutex::Autolock _l(mLock);
    if (mSurfaceData == 0) {
        return generateSurfaceLocked();
    }
    return mSurfaceData;
}

sp<Surface> SurfaceControl::generateSurfaceLocked() const
{
    //这个mGraphicBufferProducer其实就是上面分析的BufferQueueProducer
    mSurfaceData = new Surface(mGraphicBufferProducer, false);
    return mSurfaceData;
}

即直接new了一个nativie的Surface返回给java层，java层的Surface指向的就是native层的Surface。

所以Surface的实际创建可以用下图表示:

经过上面这个图，也可以理解SurfaceControl为什么叫SurfaceControl了。