[深入理解Android卷一全文-第八章]深入理解Surface系统

说明：为书写方便起见，后文将SurfaceFlinger简写为SF。

private final voidhandleLaunchActivity(ActivityRecord r, Intent customIntent) {

//①performLaunchActivity返回一个Activity

Activitya = performLaunchActivity(r, customIntent);

if(a != null) {

r.createdConfig = new Configuration(mConfiguration);

Bundle oldState = r.state;

//②调用handleResumeActivity

handleResumeActivity(r.token, false, r.isForward);

}

......

}

private final ActivityperformLaunchActivity(ActivityRecord r,

Intent customIntent) {

ActivityInfo aInfo = r.activityInfo;

......//完毕一些准备工作

//Activity定义在Activity.java中

Activity activity = null;

try {

java.lang.ClassLoader cl = r.packageInfo.getClassLoader();

/*

mInstrumentation为Instrumentation类型，源文件为Instrumentation.java。

它在newActivity函数中依据Activity的类名通过Java反射机制来创建相应的Activity，

这个函数比較复杂。待会我们再分析它。

*/

activity = mInstrumentation.newActivity(

cl,component.getClassName(), r.intent);

r.intent.setExtrasClassLoader(cl);

if (r.state != null) {

r.state.setClassLoader(cl);

}

}catch (Exception e) {

......

}

try {

Application app =

r.packageInfo.makeApplication(false,mInstrumentation);

if (activity != null) {

//在Activity中getContext函数返回的就是这个ContextImpl类型的对象

ContextImpl appContext = new ContextImpl();

......

//以下这个函数会调用Activity的onCreate函数

mInstrumentation.callActivityOnCreate(activity, r.state);

......

return activity;

}

final void handleResumeActivity(IBinder token,boolean clearHide,

boolean isForward) {

boolean willBeVisible = !a.mStartedActivity;

if (r.window == null && !a.mFinished&& willBeVisible) {

r.window= r.activity.getWindow();

//①获得一个View对象

Viewdecor = r.window.getDecorView();

decor.setVisibility(View.INVISIBLE);

//②获得ViewManager对象

ViewManagerwm = a.getWindowManager();

......

//③把刚才的decor对象增加到ViewManager中

wm.addView(decor,l);

}

......//其它处理

}

public void setContentView(View view) {

//getWindow返回的是什么呢？一起来看看。

getWindow().setContentView(view);

}

public Window getWindow() {

returnmWindow; //返回一个类型为Window的mWindow。它是什么？

}

activity = mInstrumentation.newActivity(

cl,component.getClassName(), r.intent);

public Activity newActivity(Class<?

>clazz, Context context,

IBinder token, Application application, Intent intent,

ActivityInfo info, CharSequencetitle, Activity parent,

String id,Object lastNonConfigurationInstance)

throws InstantiationException, IllegalAccessException{

Activity activity = (Activity)clazz.newInstance();

ActivityThread aThread = null;

//关键函数attach!!

activity.attach(context, aThread, this, token, application, intent,

info, title,parent, id, lastNonConfigurationInstance,

new Configuration());

return activity;

}

final void attach(Context context,ActivityThread aThread,

Instrumentation instr, IBinder token, int ident,

Application application, Intent intent, ActivityInfo info,

CharSequence title, Activity parent, String id,

Object lastNonConfigurationInstance,

HashMap<String,Object> lastNonConfigurationChildInstances,

Configuration config) {

......

//利用PolicyManager来创建Window对象

mWindow = PolicyManager.makeNewWindow(this);

mWindow.setCallback(this);

......

//创建WindowManager对象

mWindow.setWindowManager(null, mToken, mComponent.flattenToString());

if(mParent != null) {

mWindow.setContainer(mParent.getWindow());

}

//保存这个WindowManager对象

mWindowManager = mWindow.getWindowManager();

mCurrentConfig = config;

}

注意，上面路径中的灰色文件夹phone是针对智能手机这样的小屏幕的；另外另一个平级的文件夹叫mid，是针对Mid设备的。

mid文件夹的代码比較少，可能眼下还没有开发完毕。

public final class PolicyManager {

private static final String POLICY_IMPL_CLASS_NAME =

"com.android.internal.policy.impl.Policy";

private static final IPolicy sPolicy;

static{

//

try {

Class policyClass = Class.forName(POLICY_IMPL_CLASS_NAME);

//创建Policy对象

sPolicy = (IPolicy)policyClass.newInstance();

}catch (ClassNotFoundException ex) {

......

}

private PolicyManager() {}

//通过Policy对象的makeNewWindow创建一个Window

publicstatic Window makeNewWindow(Context context) {

return sPolicy.makeNewWindow(context);

}

......

}

public class Policy implements IPolicy {

private static final String TAG = "PhonePolicy";

private static final String[] preload_classes = {

"com.android.internal.policy.impl.PhoneLayoutInflater",

"com.android.internal.policy.impl.PhoneWindow",

"com.android.internal.policy.impl.PhoneWindow$1",

"com.android.internal.policy.impl.PhoneWindow$ContextMenuCallback",

"com.android.internal.policy.impl.PhoneWindow$DecorView",

"com.android.internal.policy.impl.PhoneWindow$PanelFeatureState",

"com.android.internal.policy.impl.PhoneWindow$PanelFeatureState$SavedState",

};

static{

//载入全部的类

for (String s : preload_classes) {

try {

Class.forName(s);

} catch (ClassNotFoundException ex) {

......

}

}

}

public PhoneWindow makeNewWindow(Contextcontext) {

//makeNewWindow返回的是PhoneWindow对象

return new PhoneWindow(context);

}

......

}

mWindow = PolicyManager.makeNewWindow(this);

......//创建mWindow对象

//调用mWindow的setWindowManager函数

mWindow.setWindowManager(null, mToken,mComponent.flattenToString());

.....

public class PhoneWindow extends Windowimplements MenuBuilder.Callback

public void setWindowManager(WindowManagerwm,IBinder appToken, String appName) {     //注意。传入的wm值为null

mAppToken = appToken;

mAppName = appName;

if(wm == null) {

//假设wm为空的话，则创建WindowManagerImpl对象

wm = WindowManagerImpl.getDefault();

}

//mWindowManager是一个LocalWindowManager

mWindowManager = new LocalWindowManager(wm);

}

private class LocalWindowManager implementsWindowManager {

LocalWindowManager(WindowManager wm) {

mWindowManager = wm;//还好。仅仅是简单地保存了传入的wm參数

mDefaultDisplay = mContext.getResources().getDefaultDisplay(

mWindowManager.getDefaultDisplay());

}

......

public class WindowManagerImpl implementsWindowManager {

......

public static WindowManagerImpl getDefault()

{

return mWindowManager; //返回的就是WindowManagerImpl对象

}

private static WindowManagerImpl mWindowManager= new WindowManagerImpl();

}

public void setContentView(View view) {

getWindow().setContentView(view);//getWindow返回的是PhoneWindow

}

public void setContentView(View view) {

//调用另一个setContentView

setContentView(view,

new ViewGroup.LayoutParams(MATCH_PARENT,MATCH_PARENT));

}

public void setContentView(View view,ViewGroup.LayoutParams params) {

//mContentParent为ViewGroup类型，它的初值为null

if(mContentParent == null) {

installDecor();

}else {

mContentParent.removeAllViews();

}

//把view增加到ViewGroup中

mContentParent.addView(view, params);

......

}

这里採用的是23种设计模式中的Composite模式，它是UI编程中经常使用的模式之中的一个。

private void installDecor() {

if (mDecor == null) {

//创建mDecor。它为DecorView类型，从FrameLayout派生

mDecor= generateDecor();

......

}

if(mContentParent == null) {

//得到这个mContentParent

mContentParent = generateLayout(mDecor);

//创建标题栏

mTitleView= (TextView)findViewById(com.android.internal.R.id.title);

......

}

protected ViewGroup generateLayout(DecorViewdecor){

......

intlayoutResource;

intfeatures = getLocalFeatures();

if((features & ((1 << FEATURE_LEFT_ICON) |(1 <<FEATURE_RIGHT_ICON))) != 0) {

if(mIsFloating) {

//依据情况取得相应标题栏的资源id

layoutResource =  com.android.internal.R.layout.dialog_title_icons;

}

......

}

mDecor.startChanging();

View in =mLayoutInflater.inflate(layoutResource, null);

//增加标题栏

decor.addView(in,new ViewGroup.LayoutParams(MATCH_PARENT, MATCH_PARENT));

/*

ID_ANDROID_CONTENT的值为”com.android.internal.R.id.content”

这个contentParent由findViewById返回。实际上就是mDecorView的一部分。

*/

ViewGroupcontentParent = (ViewGroup)findViewById(ID_ANDROID_CONTENT);

......

mDecor.finishChanging();

returncontentParent;

}

public View findViewById(int id) {

//getDecorView将返回mDecorView,所以contentParent确实是DecorView的一部分

returngetDecorView().findViewById(id);

}

注意，这里使用了设计模式中的Decorator（装饰）模式，它也是UI编程中经常使用的模式之中的一个。

final void handleResumeActivity(IBinder token,boolean clearHide,

boolean isForward) {

booleanwillBeVisible = !a.mStartedActivity;

......

if (r.window == null && !a.mFinished&& willBeVisible) {

r.window= r.activity.getWindow();

//①获得一个View对象。如今知道这个view就是DecorView

Viewdecor = r.window.getDecorView();

decor.setVisibility(View.INVISIBLE);

//②获得ViewManager对象,这个wm就是LocalWindowManager

ViewManagerwm = a.getWindowManager();

WindowManager.LayoutParamsl = r.window.getAttributes();

a.mDecor= decor;

l.type =WindowManager.LayoutParams.TYPE_BASE_APPLICATION;

if(a.mVisibleFromClient) {

a.mWindowAdded= true;

//③把刚才的decor对象增加到ViewManager中

wm.addView(decor,l);

}

......//其它处理

}

由于代码比較长，跳转关系也非常多，在分析代码时。请读者把握流程。在大脑中建立一个代码分析的堆栈。

wm.addView(decor, l);//wm类型实际是LocalWindowManager

public final void addView(View view,ViewGroup.LayoutParams params) {

WindowManager.LayoutParams wp =(WindowManager.LayoutParams)params;

CharSequence curTitle = wp.getTitle();

...... //做一些操作，能够无论它

//还记得前面提到过的Proxy模式吗？mWindowManager对象实际上是WindowManagerImpl类型

mWindowManager.addView(view, params);

}

private void addView(View view,ViewGroup.LayoutParams params, boolean nest)

{

ViewRootroot; //ViewRoot。幕后的主角终于登场了！

synchronized(this) {

//①创建ViewRoot

root =new ViewRoot(view.getContext());

root.mAddNesting = 1;

view.setLayoutParams(wparams);

if(mViews == null) {

index = 1;

mViews = new View[1];

mRoots= new ViewRoot[1];

mParams = new WindowManager.LayoutParams[1];

} else{

......

}

index--;

mViews[index]= view;

mRoots[index]= root;//保存这个root

mParams[index]= wparams;

//②setView,当中view是刚才我们介绍的DecorView

root.setView(view,wparams, panelParentView);//

}

public final class ViewRoot extends Handlerimplements ViewParent,

View.AttachInfo.Callbacks //从Handler类派生

{

private final Surface mSurface = new Surface();//这里创建了一个Surface对象

final W mWindow; //这个是什么？

View mView;

}

static class W extends IWindow.Stub

public ViewRoot(Context context) {

super();

....

// getWindowSession？我们进去看看

getWindowSession(context.getMainLooper());

......//ViewRoot的mWindow是一个W类型，注意它不是Window类型，而是IWindow类型

mWindow= new W(this, context);

}

ublic static IWindowSessiongetWindowSession(Looper mainLooper) {

synchronized (mStaticInit) {

if(!mInitialized) {

try {

InputMethodManagerimm =

InputMethodManager.getInstance(mainLooper);

//以下这个函数先得到WindowManagerService的Binder代理，然后调用它的openSession

sWindowSession = IWindowManager.Stub.asInterface(

ServiceManager.getService("window"))

.openSession(imm.getClient(), imm.getInputContext());

mInitialized = true;

} catch (RemoteException e) {

}

}

return sWindowSession;

}

}

openSession的操作是一个使用Binder通信的跨进程调用，暂且记住这个函数。在精简流程之后再来分析。

public void setView(View view, WindowManager.LayoutParamsattrs,

View panelParentView){//第一个參数view是DecorView

......

mView= view;//保存这个view

synchronized (this) {

requestLayout(); //待会先看看这个。

try {

//调用IWindowSession的add函数，第一个參数是mWindow

res =sWindowSession.add(mWindow, mWindowAttributes,

getHostVisibility(), mAttachInfo.mContentInsets);

}

......

}

public void requestLayout() {

checkThread();

mLayoutRequested = true;

scheduleTraversals();

}

public void scheduleTraversals() {

if(!mTraversalScheduled) {

mTraversalScheduled = true;

sendEmptyMessage(DO_TRAVERSAL); //发送DO_TRAVERSAL消息

}

}

public IWindowSessionopenSession(IInputMethodClient client,

IInputContextinputContext) {

......

return new Session(client, inputContext);

}

public int add(IWindow window,WindowManager.LayoutParams attrs,

int viewVisibility, Rect outContentInsets) {

//调用外部类对象的addWindow,也就是WMS的addWindow

returnaddWindow(this, window, attrs, viewVisibility,

outContentInsets);

}

public int addWindow(Session session, IWindowclient,

WindowManager.LayoutParams attrs, int viewVisibility,

Rect outContentInsets) {

......

//创建一个WindowState

win = new WindowState(session, client, token,

attachedWindow, attrs,viewVisibility);

......

//调用attach函数

win.attach();

......

return res;

}

void attach() {

//mSession就是Session对象。调用它的windowAddedLocked函数

mSession.windowAddedLocked();

}

void windowAddedLocked() {

if(mSurfaceSession == null) {

......

//创建一个SurfaceSession对象

mSurfaceSession= new SurfaceSession();

......

}

mNumWindow++;

}

void dispatchKey(in KeyEvent event);

void dispatchPointer(in MotionEvent event, longeventTime,

boolean callWhenDone);

void dispatchTrackball(in MotionEvent event,long eventTime,

boolean callWhenDone);

注意：上面的描写叙述实际上大大简化了真实的处理流程，读者可在了解大体知识后进行更深入的研究。

public void handleMessage(Message msg) {

switch (msg.what) {

......

case DO_TRAVERSAL:

......

performTraversals();//调用performTraversals函数

......

break;

......

}

}

private void performTraversals() {

finalView host = mView;//还记得这mView吗？它就是DecorView喔

booleaninitialized = false;

booleancontentInsetsChanged = false;

booleanvisibleInsetsChanged;

try {

relayoutResult= //①关键函数relayoutWindow

relayoutWindow(params, viewVisibility,insetsPending);

}

......

draw(fullRedrawNeeded);// ②開始绘制

......

}

private intrelayoutWindow(WindowManager.LayoutParams params,

int viewVisibility, boolean insetsPending)throws RemoteException {

//原来是调用IWindowSession的relayOut，暂且记住这个调用

int relayoutResult = sWindowSession.relayout(

mWindow, params,

(int) (mView.mMeasuredWidth * appScale + 0.5f),

(int) (mView.mMeasuredHeight * appScale + 0.5f),

viewVisibility, insetsPending, mWinFrame,

mPendingContentInsets, mPendingVisibleInsets,

mPendingConfiguration, mSurface); mSurface做为參数传进去了。

}

......

}

private void draw(boolean fullRedrawNeeded) {

Surface surface = mSurface;//mSurface是ViewRoot的成员变量

......

Canvascanvas;

try {

int left = dirty.left;

int top = dirty.top;

int right = dirty.right;

int bottom = dirty.bottom;

//从mSurface中lock一块Canvas

canvas = surface.lockCanvas(dirty);

......

mView.draw(canvas);//调用DecorView的draw函数，canvas就是画布的意思啦！

......

//unlock画布，屏幕上立即就会见到美丽宝贝的长相了。

surface.unlockCanvasAndPost(canvas);

}

......

}

private final Surface mSurface = new Surface();

void windowAddedLocked() {

if(mSurfaceSession == null) {

mSurfaceSession = new SurfaceSession();

mNumWindow++;

}

}

private intrelayoutWindow(WindowManager.LayoutParams params,

int viewVisibility, boolean insetsPending)

throws RemoteException {

int relayoutResult = sWindowSession.relayout(

mWindow, params,

(int) (mView.mMeasuredWidth * appScale + 0.5f),

(int) (mView.mMeasuredHeight * appScale + 0.5f),

viewVisibility, insetsPending, mWinFrame,

mPendingContentInsets, mPendingVisibleInsets,

mPendingConfiguration, mSurface);//mSurface传了进去

......

return relayoutResult;

}

public int relayout(IWindow window,WindowManager.LayoutParams attrs,

int requestedWidth, int requestedHeight, int viewFlags,

boolean insetsPending, Rect outFrame, Rect outContentInsets,

Rect outVisibleInsets, Configuration outConfig,

Surface outSurface) {

//注意最后这个參数的名字，叫outSurface

//调用外部类对象的relayoutWindow

returnrelayoutWindow(this, window, attrs,

requestedWidth,requestedHeight, viewFlags, insetsPending,

outFrame, outContentInsets,outVisibleInsets, outConfig,

outSurface);

}

public int relayoutWindow(Session session,IWindow client,

WindowManager.LayoutParams attrs, int requestedWidth,

int requestedHeight, int viewVisibility, boolean insetsPending,

Rect outFrame, Rect outContentInsets, Rect outVisibleInsets,

Configuration outConfig, SurfaceoutSurface){

.....

try {

//win就是WinState，这里将创建一个本地的Surface对象

Surfacesurface = win.createSurfaceLocked();

if(surface != null) {

//先创建一个本地surface，然后在outSurface的对象上调用copyFrom

//将本地Surface的信息复制到outSurface中，为什么要这么麻烦呢？

outSurface.copyFrom(surface);

......

}

Surface createSurfaceLocked() {

......

try {

//mSurfaceSession就是在Session上创建的SurfaceSession对象

//这里，以它为參数。构造一个新的Surface对象

mSurface = new Surface(

mSession.mSurfaceSession, mSession.mPid,

mAttrs.getTitle().toString(),

0, w, h, mAttrs.format, flags);

}

Surface.openTransaction();//打开一个事务处理

......

Surface.closeTransaction();//关闭一个事务处理。关于事务处理以后再分析

......

}

interface IWindowSession {

......

intrelayout(IWindow window, in WindowManager.LayoutParams attrs,

int requestedWidth, int requestedHeight, int viewVisibility,

boolean insetsPending, out Rect outFrame, out Rect outContentInsets,

out Rect outVisibleInsets, out Configuration outConfig,

out Surface outSurface);

在命令行下能够输入：

aidl –Ie:\froyo\source\frameworks\base\core\java\ -Ie:\froyo\source\frameworks\base\Graphics\java e:\froyo\source\frameworks\base\core\java\android\view\IWindowSession.aidltest.java

新生成的Java文件叫test.java。当中，-I參数指定include文件夹。比如aidl文件里使用了别的Java文件里的类。所以须要指定这些Java文件所在的文件夹。

public int relayout(android.view.IWindow window,

android.view.WindowManager.LayoutParams attrs,

int requestedWidth, intrequestedHeight,

int viewVisibility, boolean insetsPending,

android.graphics.Rect outFrame,

android.graphics.Rect outContentInsets,

android.graphics.Rect outVisibleInsets,

android.content.res.Configuration outConfig,

android.view.Surface outSurface)//outSurface是第11个參数

throwsandroid.os.RemoteException

{

android.os.Parcel_data = android.os.Parcel.obtain();

android.os.Parcel_reply = android.os.Parcel.obtain();

int_result;

try {

_data.writeInterfaceToken(DESCRIPTOR);

_data.writeStrongBinder((((window!=null))?(window.asBinder()):(null)));

if((attrs!=null)) {

_data.writeInt(1);

attrs.writeToParcel(_data,0);

}

else {

_data.writeInt(0);

}

_data.writeInt(requestedWidth);

_data.writeInt(requestedHeight);

_data.writeInt(viewVisibility);

_data.writeInt(((insetsPending)?(1):(0)));

//奇怪，outSurface的信息没有写到请求包_data中，就直接发送请求消息了

mRemote.transact(Stub.TRANSACTION_relayout,_data, _reply, 0);

_reply.readException();

_result= _reply.readInt();

if((0!=_reply.readInt())) {

outFrame.readFromParcel(_reply);

}

....

if((0!=_reply.readInt())) {

outSurface.readFromParcel(_reply);//从Parcel中读取信息来填充outSurface

}

}

......

return_result;

}

public boolean onTransact(int code,android.os.Parcel data,

android.os.Parcelreply, int flags)

throwsandroid.os.RemoteException

{

switch(code)

{

caseTRANSACTION_relayout:

{

data.enforceInterface(DESCRIPTOR);

android.view.IWindow_arg0;

android.view.Surface_arg10;

//刚才讲了，Surface信息并没有传过来。那么在relayOut中看到的outSurface是怎么

//出来的呢？看以下这句可知。原来在服务端这边居然new了一个新的Surface!!!

_arg10= new android.view.Surface();

int_result = this.relayout(_arg0, _arg1, _arg2, _arg3, _arg4,

_arg5,_arg6, _arg7, _arg8, _arg9, _arg10);

reply.writeNoException();

reply.writeInt(_result);

//_arg10就是调用copyFrom的那个outSurface，那怎么传到客户端呢？

if((_arg10!=null)) {

reply.writeInt(1);

//调用Surface的writeToParcel，把信息写到reply包中。

//注意最后一个參数为PARCELABLE_WRITE_RETURN_VALUE

_arg10.writeToParcel(reply,

android.os.Parcelable.PARCELABLE_WRITE_RETURN_VALUE);

}

}

......

returntrue;

}

注意。这里，将BpWindowSession作为了IWindowSessionBinder在客户端的代表。

public Surface() {

......

//CompatibleCanvas从Canvas类派生

mCanvas = new CompatibleCanvas();

}

public SurfaceSession() {

init();//这是一个native函数

}

static void SurfaceSession_init(JNIEnv* env,jobject clazz)

{

//创建一个SurfaceComposerClient对象

sp<SurfaceComposerClient> client = new SurfaceComposerClient;

client->incStrong(clazz);

//在Java对象中保存这个client对象的指针，类型为SurfaceComposerClient

env->SetIntField(clazz, sso.client, (int)client.get());

}

publicSurface(SurfaceSession s,//传入一个SurfaceSession对象

int pid, String name, int display, int w, int h, int format, int flags)

throws OutOfResourcesException {

......

mCanvas = new CompatibleCanvas();

//又一个native函数，注意传递的參数：display以后再说，w,h代表画图区域的宽高值

init(s,pid,name,display,w,h,format,flags);

mName = name;

}

static void Surface_init(

JNIEnv*env, jobject clazz,

jobject session,

jint pid, jstring jname, jint dpy, jint w, jint h, jint format, jintflags)

{

//从SurfaceSession对象中取出之前创建的那个SurfaceComposerClient对象

SurfaceComposerClient* client =

(SurfaceComposerClient*)env->GetIntField(session, sso.client);

sp<SurfaceControl> surface;//注意它的类型是SurfaceControl

if (jname == NULL) {

/*

调用SurfaceComposerClient的createSurface函数，返回的surface是一个

SurfaceControl类型。

*/

surface = client->createSurface(pid, dpy, w, h, format, flags);

} else{

......

}

//把这个surfaceControl对象设置到Java层的Surface对象中，对这个函数就不再分析了

setSurfaceControl(env, clazz, surface);

}

static void Surface_copyFrom(JNIEnv* env,jobject clazz, jobject other)

{

//依据JNI函数的规则，clazz是copyFrom的调用对象，而other是copyFrom的參数。

//目标对象此时还没有设置SurfaceControl，而源对象在前面已经创建了SurfaceControl

constsp<SurfaceControl>& surface = getSurfaceControl(env, clazz);

constsp<SurfaceControl>& rhs = getSurfaceControl(env, other);

if (!SurfaceControl::isSameSurface(surface, rhs)) {

//把源SurfaceControl对象设置到目标Surface中。

setSurfaceControl(env, clazz, rhs);

}

}

static void Surface_writeToParcel(JNIEnv* env,jobject clazz,

jobject argParcel, jint flags)

{

Parcel* parcel = (Parcel*)env->GetIntField(argParcel, no.native_parcel);

//clazz就是Surface对象，从这个Surface对象中取出保存的SurfaceControl对象

const sp<SurfaceControl>&control(getSurfaceControl(env, clazz));

/*

把SurfaceControl中的信息写到Parcel包中。然后利用Binder通信传递到对端。

对端通过readFromParcel来处理Parcel包。

*/

SurfaceControl::writeSurfaceToParcel(control, parcel);

if (flags & PARCELABLE_WRITE_RETURN_VALUE) {

//还记得PARCELABLE_WRITE_RETURN_VALUE吗？flags的值就等于它

//所以本地Surface对象的SurfaceControl值被置空了

setSurfaceControl(env, clazz, 0);

}

}

static void Surface_readFromParcel(

JNIEnv* env, jobject clazz, jobject argParcel)

{

Parcel* parcel = (Parcel*)env->GetIntField( argParcel,no.native_parcel);

//注意以下定义的变量类型是Surface，而不是SurfaceControl

const sp<Surface>&control(getSurface(env, clazz));

//依据服务端传递的Parcel包来构造一个新的surface。

sp<Surface> rhs = new Surface(*parcel);

if (!Surface::isSameSurface(control, rhs)) {

//把这个新surface赋给ViewRoot中的mSurface对象。

setSurface(env,clazz, rhs);

}

}

精简流程后。寥寥数语就可把过程说清楚。以后我们在研究代码时。也能够採取这样的方式。

public Canvas lockCanvas(Rect dirty)

throws OutOfResourcesException,IllegalArgumentException

{

return lockCanvasNative(dirty);//调用native的lockCanvasNative函数。

}

static jobject Surface_lockCanvas(JNIEnv* env,jobject clazz, jobject dirtyRect)

{

//从Java中的Surface对象中，取出费尽千辛万苦得到的Native的Surface对象

constsp<Surface>& surface(getSurface(env, clazz));

......

// dirtyRect表示须要重绘的矩形块，以下依据这个dirtyRect设置dirtyRegion

RegiondirtyRegion;

if(dirtyRect) {

Rect dirty;

dirty.left  =env->GetIntField(dirtyRect, ro.l);

dirty.top   =env->GetIntField(dirtyRect, ro.t);

dirty.right = env->GetIntField(dirtyRect, ro.r);

dirty.bottom=env->GetIntField(dirtyRect, ro.b);

if(!dirty.isEmpty()) {

dirtyRegion.set(dirty);

}

} else{

dirtyRegion.set(Rect(0x3FFF,0x3FFF));

}

//调用NativeSurface对象的lock函数。

//传入了一个參数Surface::SurfaceInfo info和一块表示脏区域的dirtyRegion

Surface::SurfaceInfo info;

status_t err = surface->lock(&info, &dirtyRegion);

......

//Java的Surface对象构造的时候会创建一个CompatibleCanvas。

//这里就取出这个CompatibleCanvas对象

jobject canvas = env->GetObjectField(clazz, so.canvas);

env->SetIntField(canvas, co.surfaceFormat, info.format);

//从Canvas对象中取出SkCanvas对象

SkCanvas* nativeCanvas =(SkCanvas*)env->GetIntField(

canvas, no.native_canvas);

SkBitmap bitmap;

ssize_t bpr = info.s *bytesPerPixel(info.format);

bitmap.setConfig(convertPixelFormat(info.format), info.w, info.h, bpr);

......

if (info.w > 0 && info.h > 0) {

//info.bits指向一块存储区域。

bitmap.setPixels(info.bits);

} else{

bitmap.setPixels(NULL);

}

//给这个SkCanvas设置一个Bitmap。还记得前面说的，画图须要的四大金刚吗？

//这里将Bitmap设置到这个Canvas中，这样进UI绘画时就有画布了。

nativeCanvas->setBitmapDevice(bitmap);

......

returncanvas;

}

注意。本书不拟讨论Android系统上Skia和OpenGL方面的知识，有兴趣的读者可自行研究。

static void Surface_unlockCanvasAndPost(JNIEnv*env, jobject clazz,

jobject argCanvas)

{

jobjectcanvas = env->GetObjectField(clazz, so.canvas);

//取出Native的Surface对象

const sp<Surface>& surface(getSurface(env,clazz));

//以下这些内容。不拟讨论。读者若有兴趣。可结合Skia库，自行研究。

SkCanvas* nativeCanvas =(SkCanvas*)env->GetIntField(canvas,

no.native_canvas);

intsaveCount = env->GetIntField(clazz, so.saveCount);

nativeCanvas->restoreToCount(saveCount);

nativeCanvas->setBitmapDevice(SkBitmap());

env->SetIntField(clazz, so.saveCount, 0);

//调用Surface对象的unlockAndPost函数。

status_t err = surface->unlockAndPost();

......

}

注意，Surface系统中定义了一个名为Layer类型的类，为了区分广义概念上的Layer和代码中的Layer，这里称广义层的Layer为显示层。以免混淆。

注意，关于Surface系统的显示层属性定义，读者可參考ISurfaceComposer.h。

注意：上面所说的框架将引出另外一个概念Linux FrameBuffer（简称LFB）。LFB是Linux平台提供的一种可直接操作FB的机制，依托这个机制，应用层通过标准的系统调用。就能够操作显示设备了。从使用的角度来看，它和Linux Audio中的OSS有些相似。

struct fbinfo {//定义一个结构体

unsigned int version;

unsigned int bpp;

unsigned int size;

unsigned int width;

unsigned int height;

unsigned int red_offset;

unsigned int red_length;

unsigned int blue_offset;

unsigned int blue_length;

unsigned int green_offset;

unsigned int green_length;

unsigned int alpha_offset;

unsigned int alpha_length;

} __attribute__((packed));

//fd是一个文件的描写叙述符，这个函数的目的，是把当前屏幕的内容写到一个文件里

void framebuffer_service(int fd, void *cookie)

{

structfb_var_screeninfo vinfo;

intfb, offset;

charx[256];

structfbinfo fbinfo;

unsigned i, bytespp;

//Android系统上的fb设备路径在/dev/graphics文件夹下

fb =open("/dev/graphics/fb0", O_RDONLY);

if(fb< 0) goto done;

//取出屏幕的属性

if(ioctl(fb, FBIOGET_VSCREENINFO, &vinfo) < 0) goto done;

fcntl(fb, F_SETFD, FD_CLOEXEC);

bytespp = vinfo.bits_per_pixel / 8;

//依据屏幕的属性填充fbinfo结构。这个结构要写到输出文件的头部

fbinfo.version = DDMS_RAWIMAGE_VERSION;

fbinfo.bpp = vinfo.bits_per_pixel;

fbinfo.size = vinfo.xres * vinfo.yres * bytespp;

fbinfo.width = vinfo.xres;

fbinfo.height = vinfo.yres;

/*

以下几个变量和颜色格式有关，以RGB565为例，简介一下。

RGB565表示一个像素点中R分量为5位。G分量为6位，B分量为5位，而且没有Alpha分量。

这样一个像素点的大小为16位，占两个字节，比RGB888格式的一个像素少一个字节（它一个像素是三个字节）。

x_length的值为x分量的位数。比如，RGB565中R分量就是5位。

x_offset的值代表x分量在内存中的位置。如RGB565一个像素占两个字节。那么x_offeset

表示x分量在这两个字节内存区域中的起始位置，但这个顺序是反的。也就是B分量在前，

R在最后。

所以red_offset的值就是11。而blue_offset的值是0,green_offset的值是6。

这些信息在做格式转换时（比如从RGB565转到RGB888的时候）实用。

*/

fbinfo.red_offset = vinfo.red.offset;

fbinfo.red_length = vinfo.red.length;

fbinfo.green_offset = vinfo.green.offset;

fbinfo.green_length = vinfo.green.length;

fbinfo.blue_offset = vinfo.blue.offset;

fbinfo.blue_length = vinfo.blue.length;

fbinfo.alpha_offset = vinfo.transp.offset;

fbinfo.alpha_length = vinfo.transp.length;

offset= vinfo.xoffset * bytespp;

offset+= vinfo.xres * vinfo.yoffset * bytespp;

//将fb信息写到文件头部

if(writex(fd, &fbinfo, sizeof(fbinfo))) goto done;

lseek(fb, offset, SEEK_SET);

for(i= 0; i < fbinfo.size; i += 256) {

if(readx(fb, &x, 256)) goto done;//读取FBD中的数据

if(writex(fd, &x, 256)) goto done;//将数据写到文件

}

if(readx(fb, &x, fbinfo.size % 256)) goto done;

if(writex(fd, &x, fbinfo.size % 256)) goto done;

done:

if(fb>= 0) close(fb);

close(fd);

}

注意：我们可依据这段代码，写一个简单的Native可运行程序，然后adb push到设备上运行。

注意上面写到文件里的是RGB565格式的原始数据。如想在台式机上看到这幅图片。可将它转换成BMP格式。我的个人博客上提供一个RGB565转BMP的程序。读者能够下载或自己另写一个，这样也许有助于更深入理解图形/图像方面的知识。

说白了，PageFlipping事实上就是使用了一个仅仅有两个成员的帧缓冲队列，以后在分析传输数据的时候还会见到诸如dequeue和queue的操作。

static void SurfaceSession_init(JNIEnv* env,jobject clazz)

{

//new 一个SurfaceComposerClient对象

sp<SurfaceComposerClient> client = newSurfaceComposerClient;

//sp的使用也有让人烦恼的地方。有时须要显式地增加强弱引用计数，要是忘记，可就麻烦了

client->incStrong(clazz);

env->SetIntField(clazz, sso.client,(int)client.get());

}

SurfaceComposerClient::SurfaceComposerClient()

{

//getComposerService()将返回SF的Binder代理端的BpSurfaceFlinger对象

sp<ISurfaceComposer> sm(getComposerService());

//先调用SF的createConnection。再调用_init

_init(sm, sm->createConnection());

if(mClient != 0) {

Mutex::Autolock _l(gLock);

//gActiveConnections是全局变量，把刚才创建的client保存到这个map中去

gActiveConnections.add(mClient->asBinder(), this);

}

}

sp<ISurfaceFlingerClient>SurfaceFlinger::createConnection()

{

Mutex::Autolock _l(mStateLock);

uint32_t token = mTokens.acquire();

//先创建一个Client。

sp<Client> client = new Client(token, this);

//把这个Client对象保存到mClientsMap中，token是它的标识。

status_t err = mClientsMap.add(token, client);

/*

创建一个用于Binder通信的BClient，BClient派生于ISurfaceFlingerClient，

它的作用是接受客户端的请求，然后把处理提交给SF。注意，并非提交给Client。

Client会创建一块共享内存，该内存由getControlBlockMemory函数返回

*/

sp<BClient> bclient =

new BClient(this, token,client->getControlBlockMemory());

returnbclient;

}

Client::Client(ClientID clientID, constsp<SurfaceFlinger>& flinger)

:ctrlblk(0), cid(clientID), mPid(0), mBitmap(0), mFlinger(flinger)

{

const int pgsize = getpagesize();

//以下这个操作会使cblksize为页的大小，眼下是4096字节。

constint cblksize = ((sizeof(SharedClient)+(pgsize-1))&~(pgsize-1));

//MemoryHeapBase是我们的老朋友了。不熟悉的读者能够回想Audio系统中所介绍的内容

mCblkHeap = new MemoryHeapBase(cblksize, 0,

"SurfaceFlinger Clientcontrol-block");

ctrlblk = static_cast<SharedClient *>(mCblkHeap->getBase());

if(ctrlblk) {

new(ctrlblk) SharedClient; //再一次觉得眼熟吧？使用了placement new

}

}

class SharedClient

{

public:

SharedClient();

~SharedClient();

status_t validate(size_t token) const;

uint32_t getIdentity(size_t token) const;//取出标识本Client的token

private:

Mutexlock;

Condition cv; //支持跨进程的同步对象

//NUM_LAYERS_MAX为31，SharedBufferStack是什么？

SharedBufferStack surfaces[ NUM_LAYERS_MAX ];

};

//SharedClient的构造函数，没什么新意。不如Audio的CB对象复杂

SharedClient::SharedClient()

:lock(Mutex::SHARED), cv(Condition::SHARED)

{

}

class  SharedBufferStack{

......

//Buffer是按块使用的，每一个Buffer都有自己的编号，事实上就是数组中的索引號。

volatile int32_t head;     //FrontBuffer的编号

volatile int32_t available; //空暇Buffer的个数

volatile int32_t queued;  //脏Buffer的个数。脏Buffer表示有新数据的Buffer

volatile int32_t inUse; //SF当前正在使用的Buffer的编号

volatilestatus_t status; //状态码

......

}

注意。这里的显示层指的是Normal类型的显示层。

SurfaceComposerClient::SurfaceComposerClient()

{

......

_init(sm, sm->createConnection());

......

}

void SurfaceComposerClient::_init(

const sp<ISurfaceComposer>& sm, constsp<ISurfaceFlingerClient>& conn)

{

mPrebuiltLayerState = 0;

mTransactionOpen = 0;

mStatus = NO_ERROR;

mControl = 0;

mClient = conn;//mClient就是BClient的客户端

mControlMemory =mClient->getControlBlock();

mSignalServer = sm;// mSignalServer就是BpSurfaceFlinger

//mControl就是那个创建于共享内存之中的SharedClient

mControl = static_cast<SharedClient*>(mControlMemory->getBase());

}

static void Surface_init(JNIEnv* env, jobjectclazz, jobject session,

jint pid, jstring jname, jint dpy, jint w, jint h, jint format, jintflags)

{

SurfaceComposerClient* client =

(SurfaceComposerClient*)env->GetIntField(session, sso.client);

//注意这个变量。类型是SurfaceControl。名字却叫surface。稍不留神就出错了。

sp<SurfaceControl>surface;

if (jname == NULL) {

//调用Client的createSurface函数。得到一个SurfaceControl对象。

surface= client->createSurface(pid, dpy, w, h, format, flags);

}

......

//将这个SurfaceControl对象设置到Java层的对象中保存。

setSurfaceControl(env, clazz, surface);

}

此时，读者也许会被代码中任意起的变量名搞糊涂。由于我的处理方法碰到了easy混淆的地方，尽量以对象类型来表示这个对象。

sp<SurfaceControl>SurfaceComposerClient::createSurface(

int pid,

DisplayID display,//DisplayID是什么意思？

uint32_t w,

uint32_t h,

PixelFormat format,

uint32_t flags)

{

String8 name;

constsize_t SIZE = 128;

charbuffer[SIZE];

snprintf(buffer, SIZE, "<pid_%d>", getpid());

name.append(buffer);

//调用另外一个createSurface。多一个name參数

returnSurfaceComposerClient::createSurface(pid, name, display,

w, h, format, flags);

}

typedef int32_t    DisplayID;

sp<SurfaceControl>SurfaceComposerClient::createSurface(

int pid,const String8& name,DisplayID display,uint32_t w,

uint32_t h,PixelFormat format,uint32_t flags)

{

sp<SurfaceControl> result;

if(mStatus == NO_ERROR) {

ISurfaceFlingerClient::surface_data_t data;

//调用BpSurfaceFlingerClient的createSurface函数

sp<ISurface> surface = mClient->createSurface(&data, pid,name,

display, w, h,format, flags);

if(surface != 0) {

if (uint32_t(data.token) < NUM_LAYERS_MAX) {

//以返回的ISurface对象创建一个SurfaceControl对象

result = new SurfaceControl(this, surface, data, w, h,

format, flags);

}

}

}

returnresult;//返回的是SurfaceControl对象

}

在Surface系统定义了非常多类型，咱们也中途歇息一下，最好还是来看看和字符串“Surface”有关的有多少个类。权当其为小小的娱乐：

Native层有Surface、ISurface、SurfaceControl、SurfaceComposerClient。

Java层有Surface、SurfaceSession。

上面列出的还仅仅是一部分。后面还有呢！

*&@&*%￥*

sp<ISurface> BClient::createSurface(

ISurfaceFlingerClient::surface_data_t* params, int pid,

const String8& name,

DisplayID display, uint32_t w, uint32_t h, PixelFormat format,

uint32_t flags)

{

//果然是交给SF处理，以后我们将跳过BClient这个代理。

return mFlinger->createSurface(mId, pid,name, params, display, w, h,

format, flags);

}

sp<ISurface>SurfaceFlinger::createSurface(ClientID clientId, int pid,

const String8& name, ISurfaceFlingerClient::surface_data_t* params,

DisplayID d, uint32_t w, uint32_t h, PixelFormat format,

uint32_t flags)

{

sp<LayerBaseClient> layer;//LayerBaseClient是Layer家族的基类

//这里又冒出一个LayerBaseClient的内部类，它也叫Surface，是不是有点头晕了？

sp<LayerBaseClient::Surface> surfaceHandle;

Mutex::Autolock _l(mStateLock);

//依据clientId找到createConnection时增加的那个Client对象

sp<Client> client = mClientsMap.valueFor(clientId);

......

//注意这个id。它的值表示Client创建的是第几个显示层，依据图8-14能够看出，这个id

//同一时候也表示将使用SharedBufferStatck数组的第id个元素。

int32_t id = client->generateId(pid);

//一个Client不能创建多于NUM_LAYERS_MAX个的Layer。

if(uint32_t(id) >= NUM_LAYERS_MAX) {

return surfaceHandle;

}

//依据flags參数来创建不同类型的显示层，我们在8.4.1节介绍过相关知识

switch(flags & eFXSurfaceMask) {

case eFXSurfaceNormal:

if (UNLIKELY(flags & ePushBuffers)) {

//创建PushBuffer类型的显示层，我们将在拓展思考部分分析它

layer = createPushBuffersSurfaceLocked(client, d, id,

w, h, flags);

} else {

//①创建Normal类型的显示层，我们分析待会这个

layer = createNormalSurfaceLocked(client, d, id,

w, h, flags, format);

}

break;

case eFXSurfaceBlur:

//创建Blur类型的显示层

layer = createBlurSurfaceLocked(client, d, id, w, h, flags);

break;

case eFXSurfaceDim:

//创建Dim类型的显示层

layer = createDimSurfaceLocked(client, d, id, w, h, flags);

break;

}

if(layer != 0) {

layer->setName(name);

setTransactionFlags(eTransactionNeeded);

//从显示层对象中取出一个ISurface对象赋值给SurfaceHandle

surfaceHandle = layer->getSurface();

if(surfaceHandle != 0) {

params->token = surfaceHandle->getToken();

params->identity = surfaceHandle->getIdentity();

params->width = w;

params->height = h;

params->format = format;

}

}

returnsurfaceHandle;//ISurface的Bn端就是这个对象。

}

sp<LayerBaseClient>SurfaceFlinger::createNormalSurfaceLocked(

const sp<Client>& client, DisplayID display,

int32_t id, uint32_t w, uint32_t h, uint32_t flags,

PixelFormat& format)

{

switch(format) { //一些图像方面的參数设置。能够不去管它。

casePIXEL_FORMAT_TRANSPARENT:

casePIXEL_FORMAT_TRANSLUCENT:

format = PIXEL_FORMAT_RGBA_8888;

break;

casePIXEL_FORMAT_OPAQUE:

format = PIXEL_FORMAT_RGB_565;

break;

}

//①创建一个Layer类型的对象

sp<Layer> layer = new Layer(this, display,client, id);

//②设置Buffer

status_t err = layer->setBuffers(w, h, format, flags);

if (LIKELY(err == NO_ERROR)) {

//初始化这个新layer的一些状态

layer->initStates(w, h, flags);

//③ 还记得在图8-10中提到的Z轴吗？以下这个函数把这个layer增加到Z轴大军中。

addLayer_l(layer);

}

......

returnlayer;

}

result = new SurfaceControl(this, surface, data,w, h,format, flags);

SurfaceControl::SurfaceControl(

const sp<SurfaceComposerClient>& client,

const sp<ISurface>& surface,

const ISurfaceFlingerClient::surface_data_t& data,

uint32_t w, uint32_t h, PixelFormat format, uint32_t flags)

//mClient为SurfaceComposerClient。而mSurface指向跨进程createSurface调用

//返回的ISurface对象。

:mClient(client), mSurface(surface),

mToken(data.token), mIdentity(data.identity),

mWidth(data.width), mHeight(data.height), mFormat(data.format),

mFlags(flags)

{

}

Layer::Layer(SurfaceFlinger* flinger, DisplayIDdisplay,

const sp<Client>& c, int32_t i)//这个i表示SharedBufferStack数组的索引

:   LayerBaseClient(flinger, display, c, i),//先调用基类构造函数

mSecure(false),

mNoEGLImageForSwBuffers(false),

mNeedsBlending(true),

mNeedsDithering(false)

{

//getFrontBuffer实际取出的是FrontBuffer的位置

mFrontBufferIndex = lcblk->getFrontBuffer();

}

LayerBaseClient::LayerBaseClient(SurfaceFlinger*flinger, DisplayID display,

const sp<Client>& client, int32_t i)

:LayerBase(flinger, display), lcblk(NULL), client(client), mIndex(i),

mIdentity(uint32_t(android_atomic_inc(&sIdentity)))

{

/*

创建一个SharedBufferServer对象，注意它使用了SharedClient对象，

而且传入了表示SharedBufferStack数组索引的i和一个常量NUM_BUFFERS

*/

lcblk = new SharedBufferServer(

client->ctrlblk, i, NUM_BUFFERS,//该值为常量2，在Layer.h中定义

mIdentity);

}

注意，在拓展思考部分。会有单独章节来分析生产/消费过程中的读写控制。

void LayerBaseClient::onFirstRef()

{

sp<Client> client(this->client.promote());

if (client != 0) {

//把自己增加client对象的mLayers数组中,这部分内容比較简单，读者能够自行研究

client->bindLayer(this, mIndex);

}

}

status_t Layer::setBuffers( uint32_t w, uint32_th,

PixelFormat format,uint32_t flags)

{

PixelFormatInfo info;

status_t err = getPixelFormatInfo(format, &info);

if(err) return err;

//DisplayHardware是代表显示设备的HAL对象，0代表第一块屏幕的显示设备。

//这里将从HAL中取出一些和显示相关的信息。

constDisplayHardware& hw(graphicPlane(0).displayHardware());

uint32_t const maxSurfaceDims = min(

hw.getMaxTextureSize(), hw.getMaxViewportDims());

PixelFormatInfo displayInfo;

getPixelFormatInfo(hw.getFormat(),&displayInfo);

constuint32_t hwFlags = hw.getFlags();

......

/*

创建Buffer，这里将创建两个GraphicBuffer。这两个GraphicBuffer就是我们前面

所说的FrontBuffer和BackBuffer。

*/

for (size_t i=0 ; i<NUM_BUFFERS ; i++) {

//注意，这里调用的是GraphicBuffer的无參构造函数,mBuffers是一个二元数组。

mBuffers[i] = new GraphicBuffer();

}

//又冒出来一个SurfaceLayer类型，#￥%……&*！@

mSurface = new SurfaceLayer(mFlinger, clientIndex(), this);

returnNO_ERROR;

}

GraphicBuffer是Android提供的显示内存管理类。关于它的故事，将在8.4.7节中介绍。我们暂把它当做普通的Buffer就可以。

status_t SurfaceFlinger::addLayer_l(constsp<LayerBase>& layer)

{

/*

mCurrentState是SurfaceFlinger定义的一个结构，它有一个成员变量叫

layersSortedByZ，事实上就是一个排序数组。

以下这个add函数将把这个新的layer依照

它在Z轴的位置增加到排序数组中。mCurrentState保存了全部的显示层。

*/

ssize_t i = mCurrentState.layersSortedByZ.add(

layer,&LayerBase::compareCurrentStateZ);

sp<LayerBaseClient> lbc =

LayerBase::dynamicCast< LayerBaseClient*>(layer.get());

if(lbc != 0) {

mLayerMap.add(lbc->serverIndex(), lbc);

}

returnNO_ERROR;

}

强烈建议Google把Surface相关代码好好整理一下，至少让类型名取得更直观些，如今这样确实有点让人头晕。

好。来小小娱乐一下。看之前介绍的和“Surface”有关的名字：

Native层有Surface、ISurface、SurfaceControl、SurfaceComposerClient。

Java层有Surface、SurfaceSession。

在介绍完Layer家族后，与它相关的名字又多了几个。它们是

LayerBaseClient::Surface、Layer::SurfaceLayer、LayerBuffer::SurfaceLayerBuffer。

注意。mOwner变量由SurfaceLayer的基类Surface（LayBaseClient的内部类）定义。

后文将Activity所在的进程简称为Activity端。

status_t SurfaceControl::writeSurfaceToParcel(

const sp<SurfaceControl>& control, Parcel* parcel)

{

uint32_t flags = 0;

uint32_t format = 0;

SurfaceID token = -1;

uint32_t identity = 0;

uint32_t width = 0;

uint32_t height = 0;

sp<SurfaceComposerClient> client;

sp<ISurface> sur;

if(SurfaceControl::isValid(control)) {

token    = control->mToken;

identity= control->mIdentity;

client   = control->mClient;

sur      = control->mSurface;

width    = control->mWidth;

height   = control->mHeight;

format   = control->mFormat;

flags    = control->mFlags;

}

//SurfaceComposerClient的信息须要传递到Activity端。这样客户端那边会构造一个

//SurfaceComposerClient对象

parcel->writeStrongBinder(client!=0  ?

client->connection() : NULL);

//把ISurface对象信息也写到Parcel中，这样Activity端那边也会构造一个ISurface对象

parcel->writeStrongBinder(sur!=0?

sur->asBinder(): NULL);

parcel->writeInt32(token);

parcel->writeInt32(identity);

parcel->writeInt32(width);

parcel->writeInt32(height);

parcel->writeInt32(format);

parcel->writeInt32(flags);

returnNO_ERROR;

}

static void Surface_readFromParcel(

JNIEnv* env, jobject clazz, jobject argParcel)

{

Parcel* parcel = (Parcel*)env->GetIntField( argParcel, no.native_parcel);

const sp<Surface>& control(getSurface(env,clazz));

//依据服务端的parcel信息来构造客户端的Surface

sp<Surface> rhs = new Surface(*parcel);

if(!Surface::isSameSurface(control, rhs)) {

setSurface(env, clazz, rhs);

}

}

Surface::Surface(const Parcel& parcel)

:mBufferMapper(GraphicBufferMapper::get()),

mSharedBufferClient(NULL)

{

/*

Surface定义了一个mBuffers变量。它是一个sp<GraphicBuffer>的二元数组。也就是说Surface也存在二个GraphicBuffer。而之前在创建Layer的时候也有两个GraphicBuffer，难道一共同拥有四个GraphicBuffer？这个问题。后面再解答。

*/

sp<IBinder> clientBinder =parcel.readStrongBinder();

//得到ISurface的Bp端BpSurface。

mSurface    =interface_cast<ISurface>(parcel.readStrongBinder());

mToken      = parcel.readInt32();

mIdentity   = parcel.readInt32();

mWidth      = parcel.readInt32();

mHeight     = parcel.readInt32();

mFormat     = parcel.readInt32();

mFlags      = parcel.readInt32();

if (clientBinder != NULL) {

/*

依据ISurfaceFlingerClient对象构造一个SurfaceComposerClient对象，注意我们

如今位于Activity端。这里还没有创建SurfaceComposerClient对象，所以须要创建一个

*/

mClient = SurfaceComposerClient::clientForConnection(clientBinder);

//SharedBuffer家族的最后一员ShardBufferClient终于出现了。

mSharedBufferClient = new SharedBufferClient(

mClient->mControl, mToken, 2,mIdentity);

}

init();//做一些初始化工作。

}

SharedBufferBase::SharedBufferBase(SharedClient*sharedClient,

int surface, int num, int32_t identity)

: mSharedClient(sharedClient),

mSharedStack(sharedClient->surfaces+ surface),

mNumBuffers(num), //依据前面PageFlipping的知识可知。num值为2

mIdentity(identity)

{

/*

上面的赋值语句中最重要的是第二句：

mSharedStack(sharedClient->surfaces +surface)

这条语句使得这个SharedBufferXXX对象，和SharedClient中SharedBufferStack数组

的第surface个元素建立了关系

*/

}

status_t Surface::lock(SurfaceInfo* other,Region* dirtyIn, bool blocking)

{

//传入的參数中。other用来接收一些返回信息，dirtyIn表示须要重绘的区域

......

if (mApiLock.tryLock() != NO_ERROR) {//多线程的情况下要锁住

......

returnWOULD_BLOCK;

}

//设置usage标志。这个标志在GraphicBuffer分配缓冲时有指导作用

setUsage(GRALLOC_USAGE_SW_READ_OFTEN | GRALLOC_USAGE_SW_WRITE_OFTEN);

//定义一个GraphicBuffer，名字就叫backBuffer。

sp<GraphicBuffer>backBuffer;

//①还记得我们说的2个元素的缓冲队列吗？以下的dequeueBuffer将取出一个空暇缓冲

status_terr = dequeueBuffer(&backBuffer);

if (err== NO_ERROR) {

//② 锁住这块buffer

err = lockBuffer(backBuffer.get());

if(err == NO_ERROR) {

const Rect bounds(backBuffer->width, backBuffer->height);

Region scratch(bounds);

Region& newDirtyRegion(dirtyIn ? *dirtyIn : scratch);

......

//mPostedBuffer是上一次绘画时使用的Buffer，也就是如今的frontBuffer

const sp<GraphicBuffer>& frontBuffer(mPostedBuffer);

if (frontBuffer !=0 &&

backBuffer->width  ==frontBuffer->width &&

backBuffer->height == frontBuffer->height &&

!(mFlags & ISurfaceComposer::eDestroyBackbuffer))

{

const Region copyback(mOldDirtyRegion.subtract(newDirtyRegion));

if (!copyback.isEmpty() && frontBuffer!=0) {

/③把frontBuffer中的数据复制到BackBuffer中。这是为什么？

copyBlt(backBuffer,frontBuffer, copyback);

}

}

mDirtyRegion = newDirtyRegion;

mOldDirtyRegion = newDirtyRegion;

void* vaddr;

//调用GraphicBuffer的lock得到一块内存，内存地址被赋值给了vaddr,

//兴许的作画将在这块内存上展开

status_t res = backBuffer->lock(

GRALLOC_USAGE_SW_READ_OFTEN |GRALLOC_USAGE_SW_WRITE_OFTEN,

newDirtyRegion.bounds(),&vaddr);

mLockedBuffer = backBuffer;

//other用来接收一些信息。

other->w      =backBuffer->width;  //宽度信息

other->h      =backBuffer->height;

other->s      =backBuffer->stride;

other->usage  =backBuffer->usage;

other->format = backBuffer->format;

other->bits   = vaddr; //最重要的是这个内存地址

}

}

mApiLock.unlock();

returnerr;

}

status_tSurface::dequeueBuffer(sp<GraphicBuffer>* buffer) {

android_native_buffer_t* out;

status_t err = dequeueBuffer(&out);//调用另外一个dequeueBuffer

if(err == NO_ERROR) {

*buffer = GraphicBuffer::getSelf(out);

}

returnerr;

}

intSurface::dequeueBuffer(android_native_buffer_t** buffer)

{

sp<SurfaceComposerClient> client(getClient());

//①调用SharedBufferClient的dequeue函数，它返回当前空暇的缓冲号

ssize_tbufIdx = mSharedBufferClient->dequeue();

const uint32_t usage(getUsage());

/*

mBuffers就是我们前面在Surface创建中介绍的那个二元sp<GraphicBuffer>数组。

这里定义的backBuffer是一个引用类型，也就是说假设改动backBuffer的信息，

就相当于改动了mBuffers[bufIdx]

*/

const sp<GraphicBuffer>&backBuffer(mBuffers[bufIdx]);

//mBuffers定义的GraphicBuffer使用的也是无參构造函数，所以此时还没有真实的存储被创建

if(backBuffer == 0 || //第一次进来满足backBuffer为空这个条件

((uint32_t(backBuffer->usage) & usage) != usage) ||

mSharedBufferClient->needNewBuffer(bufIdx))

{

//调用getBufferLocked，须要进去看看。

err = getBufferLocked(bufIdx, usage);

if(err == NO_ERROR) {

mWidth  =uint32_t(backBuffer->width);

mHeight = uint32_t(backBuffer->height);

}

}

......

}

tatus_t Surface::getBufferLocked(int index, intusage)

{

sp<ISurface> s(mSurface);

status_t err = NO_MEMORY;

//注意这个currentBuffer也被定义为引用类型

sp<GraphicBuffer>&currentBuffer(mBuffers[index]);

//终于用上了ISurface对象,调用它的requestBuffer得到指定索引index的Buffer

sp<GraphicBuffer> buffer =s->requestBuffer(index, usage);

if (buffer != 0) {

err = mSharedBufferClient->getStatus();

if(!err && buffer->handle != NULL) {

//getBufferMapper返回GraphicBufferMapper对象

//调用它的registerBuffer干什么？这个问题我们在8.4.7节回答

err = getBufferMapper().registerBuffer(buffer->handle);

if (err == NO_ERROR) {

//把requestBuffer得到的值赋给currentBuffer，由于currentBuffer是引用类型，

//实际上相当于mBuffers[index]=buffer

currentBuffer = buffer;

//设置currentBuffer的编号

currentBuffer->setIndex(index);

mNeedFullUpdate = true;

}

}else {

err = err<0 ?

err : NO_MEMORY;

}

return err;

}

sp<GraphicBuffer> Layer::requestBuffer(intindex, int usage)

{

sp<GraphicBuffer> buffer;

sp<Client> ourClient(client.promote());

//lcblk就是那个SharedBufferServer对象，以下这个调用确保index号GraphicBuffer

//没有被SF当做FrontBuffer使用。

status_t err = lcblk->assertReallocate(index);

......

if(err != NO_ERROR) {

return buffer;

}

uint32_t w, h;

{

Mutex::Autolock _l(mLock);

w= mWidth;

h= mHeight;

/*

mBuffers是SF端创建的一个二元数组，这里取出第index个元素。之前说过，

mBuffers使用的也是GraphicBuffer的无參构造函数，所以此时也没有真实存储被创建。

*/

buffer = mBuffers[index];

mBuffers[index].clear();

}

constuint32_t effectiveUsage = getEffectiveUsage(usage);

if(buffer!=0 && buffer->getStrongCount() == 1) {

//①分配物理存储，后面会分析这个。

err = buffer->reallocate(w, h, mFormat, effectiveUsage);

} else{

buffer.clear();

//使用GraphicBuffer的有參构造，这也使得物理存储被分配

buffer = new GraphicBuffer(w, h, mFormat, effectiveUsage);

err = buffer->initCheck();

}

......

if(err == NO_ERROR && buffer->handle != 0) {

Mutex::Autolock _l(mLock);

if(mWidth && mHeight) {

mBuffers[index] = buffer;

mTextures[index].dirty = true;

}else {

buffer.clear();

}

}

returnbuffer;//返回

}

int Surface::lockBuffer(android_native_buffer_t*buffer)

{

sp<SurfaceComposerClient> client(getClient());

status_t err = validate();

int32_t bufIdx = GraphicBuffer::getSelf(buffer)->getIndex();

err =mSharedBufferClient->lock(bufIdx); //调用SharedBufferClient的lock

return err;

}

status_t SharedBufferClient::lock(int buf)

{

LockCondition condition(this, buf);//这个buf是BackBuffer的索引號

status_t err = waitForCondition(condition);

returnerr;

}

template <typename T> //这是一个模板函数

status_t SharedBufferBase::waitForCondition(Tcondition)

{

constSharedBufferStack& stack( *mSharedStack );

SharedClient& client( *mSharedClient );

constnsecs_t TIMEOUT = s2ns(1);

Mutex::Autolock _l(client.lock);

while((condition()==false) && //注意这个condition()的使用方法

(stack.identity == mIdentity) &&

(stack.status == NO_ERROR))

{

status_t err = client.cv.waitRelative(client.lock, TIMEOUT);

if(CC_UNLIKELY(err != NO_ERROR)) {

if (err == TIMED_OUT) {

if (condition()) {//注意这个：condition()。condition是一个对象

break;

} else {

}

} else {

return err;

}

}

}

return(stack.identity != mIdentity) ? status_t(BAD_INDEX) : stack.status;

}

这就是Funcition Object（函数对象）的概念。函数对象的本质是一个对象，只是是重载了操作符()。这和重载操作符+、-等没什么差别。能够把它当作是一个函数来看待。

为什么须要函数对象呢？由于对象能够保存信息。所以调用这个对象的()函数就能够利用这个对象的信息了。

boolSharedBufferClient::LockCondition::operator()() {

//stack、buf等都是这个对象的内部成员，这个对象的目的就是依据读写位置推断这个buffer是

//否空暇。

return(buf != stack.head ||

(stack.queued > 0 && stack.inUse != buf));

}

status_t Surface::lock(SurfaceInfo* other,Region* dirtyIn, bool blocking)

{

......

const sp<GraphicBuffer>& frontBuffer(mPostedBuffer);

if (frontBuffer !=0 &&

backBuffer->width  ==frontBuffer->width &&

backBuffer->height == frontBuffer->height &&

!(mFlags & ISurfaceComposer::eDestroyBackbuffer))

{

const Region copyback(mOldDirtyRegion.subtract(newDirtyRegion));

if (!copyback.isEmpty() && frontBuffer!=0) {

//③把frontBuffer中的数据复制到BackBuffer中，这是为什么？

copyBlt(backBuffer,frontBuffer, copyback);

}

}

......

}

status_t Surface::unlockAndPost()

{

//调用GraphicBuffer的unlock函数

status_t err = mLockedBuffer->unlock();

//get返回这个GraphicBuffer的编号，queueBuffer将含有新数据的缓冲增加队中。

err =queueBuffer(mLockedBuffer.get());

mPostedBuffer = mLockedBuffer; //保存这个BackBuffer为mPostedBuffer

mLockedBuffer = 0;

returnerr;

}

intSurface::queueBuffer(android_native_buffer_t* buffer)

{

sp<SurfaceComposerClient> client(getClient());

int32_t bufIdx =GraphicBuffer::getSelf(buffer)->getIndex();

//设置脏Region

mSharedBufferClient->setDirtyRegion(bufIdx,mDirtyRegion);

//更新写位置。

err =mSharedBufferClient->queue(bufIdx);

if (err== NO_ERROR) {

//client是BpSurfaceFlinger，调用它的signalServer，这样SF就知道新数据准备好了

client->signalServer();

}

returnerr;

}

status_t SharedBufferClient::queue(int buf)

{

//QueueUpdate也是一个函数对象

QueueUpdate update(this);

//调用updateCondition函数。

status_t err = updateCondition( update );

SharedBufferStack& stack( *mSharedStack );

constnsecs_t now = systemTime(SYSTEM_TIME_THREAD);

stack.stats.totalTime = ns2us(now - mDequeueTime[buf]);

returnerr;

}

template <typename T>

status_t SharedBufferBase::updateCondition(Tupdate) {

SharedClient& client( *mSharedClient );

Mutex::Autolock _l(client.lock);

ssize_t result = update();//调用update对象的()函数

client.cv.broadcast(); //触发同步对象

returnresult;

}

class GraphicBuffer

:public EGLNativeBase<android_native_buffer_t,

GraphicBuffer,LightRefBase<GraphicBuffer>>,

public Flattenable

template <typename NATIVE_TYPE, typenameTYPE, typename REF>

class EGLNativeBase : public NATIVE_TYPE, publicREF

typedef struct android_native_buffer_t

{

#ifdef __cplusplus

android_native_buffer_t() {

common.magic = ANDROID_NATIVE_BUFFER_MAGIC;

common.version = sizeof(android_native_buffer_t);

memset(common.reserved, 0, sizeof(common.reserved));

}

#endif

//这个android_native_base_t是struct的第一个成员，依据C/C++编译的特性。这个成员

//在它的派生类对象所占有的内存中也是排第一个。

structandroid_native_base_t common;

intwidth;

intheight;

intstride;

intformat;

intusage;

void* reserved[2];

//这是一个关键成员，保存一些和显示内存分配/管理相关的内容

buffer_handle_t handle;

void*reserved_proc[8];

} android_native_buffer_t;

typedef const native_handle* buffer_handle_t;

typedef struct

{

intversion;        /* version值为sizeof(native_handle_t) */

intnumFds;

intnumInts;

intdata[0];        /* data是数据存储空间的首地址 */

} native_handle_t;

typedef native_handle_t native_handle;

GraphicBuffer::GraphicBuffer()

:BASE(), mOwner(ownData), mBufferMapper(GraphicBufferMapper::get()),

mInitCheck(NO_ERROR), mVStride(0), mIndex(-1)

{

/*

当中mBufferMapper为GraphicBufferMapper类型，它的创建採用的是单例模式，也就是每一个

进程仅仅有一个GraphicBufferMapper对象，读者能够去看看get的实现。

*/

width  =

height=

stride=

format=

usage  = 0;

handle= NULL; //handle为空

}

status_t GraphicBuffer::reallocate(uint32_t w,uint32_t h, PixelFormat f,

uint32_t reqUsage)

{

if(mOwner != ownData)

return INVALID_OPERATION;

if(handle) {//handle值在无參构造函数中初始化为空，所以不满足if的条件

GraphicBufferAllocator& allocator(GraphicBufferAllocator::get());

allocator.free(handle);

handle = 0;

}

returninitSize(w, h, f, reqUsage);//调用initSize函数

}

status_t GraphicBuffer::initSize(uint32_t w,uint32_t h, PixelFormat format,

uint32_t reqUsage)

{

if(format == PIXEL_FORMAT_RGBX_8888)

format = PIXEL_FORMAT_RGBA_8888;

/*

GraphicBufferAllocator才是真正的存储分配的管理类，它的创建也是採用的单例模式，

也就是每一个进程仅仅有一个GraphicBufferAllocator对象

*/

GraphicBufferAllocator& allocator =GraphicBufferAllocator::get();

//调用GraphicBufferAllocator的alloc来分配存储，注意handle作为指针

//被传了进去。看来handle的值会被改动

status_t err = allocator.alloc(w, h, format, reqUsage, &handle,&stride);

if(err == NO_ERROR) {

this->width  = w;

this->height = h;

this->format = format;

this->usage  = reqUsage;

mVStride = 0;

}

returnerr;

}

GraphicBufferAllocator::GraphicBufferAllocator()

:mAllocDev(0)

{

hw_module_t const* module;

//调用hw_get_module。得到hw_module_t

interr = hw_get_module(GRALLOC_HARDWARE_MODULE_ID, &module);

if (err == 0) {

//调用gralloc_open函数，注意我们把module參数传了进去。

gralloc_open(module, &mAllocDev);

}

}

读者可通过运行adb getprop ro.board.platform命令。得到具体手机上硬件平台的名字。图8-22总结了GraphicBufferAllocator分配内存的途径。

这部分代码，读者可參考hardware/libhardware/hardware.c和hardware/msm7k/libgralloc-qsd8k/gralloc.cpp。后文将不再深入探讨和硬件平台有关的知识。

注意。这里是以G7的libgralloc.qsk-8k.so为演示样例的。

当中pmem设备用来创建一块连续的内存。由于有些硬件设备（比如Camera）工作时须要使用一块连续的内存，对于这样的情况。一般就会使用pmem设备来分配内存。

status_t GraphicBufferAllocator::alloc(uint32_tw, uint32_t h, PixelFormat format,int usage, buffer_handle_t* handle, int32_t*stride)

{

//依据前面的定义可知buffer_handle_t为native_handle_t*类型

status_t err;

if (usage & GRALLOC_USAGE_HW_MASK) {

err =mAllocDev->alloc(mAllocDev, w, h, format, usage, handle, stride);

} else {

//SW分配，能够做到和HW无关了。

err = sw_gralloc_handle_t::alloc(w, h, format, usage, handle, stride);

}

......

returnerr;

}

status_t sw_gralloc_handle_t::alloc(uint32_t w,uint32_t h, int format,

int usage, buffer_handle_t* pHandle, int32_t*pStride)

{

intalign = 4;

intbpp = 0;

......//格式转换

size_tbpr = (w*bpp + (align-1)) & ~(align-1);

size_tsize = bpr * h;

size_tstride = bpr / bpp;

size =(size + (PAGE_SIZE-1)) & ~(PAGE_SIZE-1);

//直接使用了ashmem创建共享内存

int fd= ashmem_create_region("sw-gralloc-buffer", size);

......

//进行内存映射。得到共享内存起始地址

void*base = mmap(0, size, prot, MAP_SHARED, fd, 0);

sw_gralloc_handle_t* hnd = new sw_gralloc_handle_t();

hnd->fd = fd;//保存文件描写叙述符

hnd->size = size;//保存共享内存的大小

hnd->base = intptr_t(base);//intptr_t将void*类型转换成int*类型

hnd->prot = prot;//保存属性

*pStride = stride;

*pHandle = hnd; //pHandle就是传入的那个handle变量的指针。这里对它进行赋值

returnNO_ERROR;

}

//requestBuffer的响应端

status_t BnSurface::onTransact(

uint32_t code, const Parcel& data, Parcel* reply, uint32_t flags)

{

switch(code) {

case REQUEST_BUFFER: {

CHECK_INTERFACE(ISurface, data, reply);

int bufferIdx = data.readInt32();

int usage = data.readInt32();

sp<GraphicBuffer> buffer(requestBuffer(bufferIdx, usage));

......

/*

requestBuffer的返回值被写到Parcel包中，由于GraphicBuffer从

Flattenable类派生，这将导致它的flatten函数被调用

*/

return reply->write(*buffer);

}

.......

}

//再来看请求端的处理。在BpSurface中

virtual sp<GraphicBuffer> requestBuffer(intbufferIdx, int usage)

{

Parcel data, reply;

data.writeInterfaceToken(ISurface::getInterfaceDescriptor());

data.writeInt32(bufferIdx);

data.writeInt32(usage);

remote()->transact(REQUEST_BUFFER, data, &reply);

sp<GraphicBuffer> buffer = new GraphicBuffer();

reply.read(*buffer);//Parcel调用unflatten函数把信息反序列化到这个buffer中。

return buffer;//requestBuffer实际上返回的是本地new出来的这个GraphicBuffer

}

status_t GraphicBuffer::flatten(void* buffer,size_t size,

int fds[], size_t count) const

{

//buffer是装载数据的缓冲区，由Parcel提供

......

if(handle) {

buf[6] = handle->numFds;

buf[7] = handle->numInts;

native_handle_t const* const h = handle;

//把handle的信息也写到buffer中

memcpy(fds,     h->data,             h->numFds*sizeof(int));

memcpy(&buf[8], h->data + h->numFds,h->numInts*sizeof(int));

}

returnNO_ERROR;

}

status_t GraphicBuffer::unflatten(void const*buffer, size_t size,

int fds[], size_t count)

{

......

if(numFds || numInts) {

width  = buf[1];

height = buf[2];

stride = buf[3];

format = buf[4];

usage  = buf[5];

native_handle* h =native_handle_create(numFds, numInts);

memcpy(h->data,         fds,     numFds*sizeof(int));

memcpy(h->data + numFds, &buf[8],numInts*sizeof(int));

handle = h;//依据Parcel包中的数据还原一个handle

} else{

width = height = stride = format = usage = 0;

handle = NULL;

}

mOwner= ownHandle;

returnNO_ERROR;

}

status_tsw_gralloc_handle_t::registerBuffer(sw_gralloc_handle_t* hnd)

{

if (hnd->pid != getpid()) {

//原来是做一次内存映射操作

void* base = mmap(0, hnd->size, hnd->prot, MAP_SHARED, hnd->fd,0);

......

//base保存着共享内存的起始地址

hnd->base = intptr_t(base);

}

returnNO_ERROR;

}

//lock操作

int sw_gralloc_handle_t::lock(sw_gralloc_handle_t*hnd, int usage,

int l, int t, int w, int h, void** vaddr)

{

*vaddr= (void*)hnd->base;//得到共享内存的起始地址，兴许作画就使用这块内存了。

returnNO_ERROR;

}

//unlock操作

status_tsw_gralloc_handle_t::unlock(sw_gralloc_handle_t* hnd)

{

returnNO_ERROR;//没有不论什么操作

}

void SurfaceFlinger::instantiate() {

defaultServiceManager()->addService(

String16("SurfaceFlinger"), new SurfaceFlinger());

}

class SurfaceFlinger : public BnSurfaceComposer,protected Thread

SurfaceFlinger::SurfaceFlinger()

:   BnSurfaceComposer(), Thread(false),

mTransactionFlags(0),

mTransactionCount(0),

mResizeTransationPending(false),

mLayersRemoved(false),

mBootTime(systemTime()),

mHardwareTest("android.permission.HARDWARE_TEST"),

mAccessSurfaceFlinger("android.permission.ACCESS_SURFACE_FLINGER"),

mDump("android.permission.DUMP"),

mVisibleRegionsDirty(false),

mDeferReleaseConsole(false),

mFreezeDisplay(false),

mFreezeCount(0),

mFreezeDisplayTime(0),

mDebugRegion(0),

mDebugBackground(0),

mDebugInSwapBuffers(0),

mLastSwapBufferTime(0),

mDebugInTransaction(0),

mLastTransactionTime(0),

mBootFinished(false),

mConsoleSignals(0),

mSecureFrameBuffer(0)

{

init();//上面没有调用run。

必须到init去检查一番。

}

//init函数更简单了。

void SurfaceFlinger::init()

{

charvalue[PROPERTY_VALUE_MAX];

property_get("debug.sf.showupdates", value, "0");

mDebugRegion = atoi(value);

property_get("debug.sf.showbackground", value, "0");

mDebugBackground = atoi(value);

}

void SurfaceFlinger::onFirstRef()

{

//真是梦里寻他千百度，果然是在onFirstRef中创建了工作线程

run("SurfaceFlinger",PRIORITY_URGENT_DISPLAY);

/*

mReadyToRunBarrier类型为Barrier，这个类就是封装了一个Mutex对象和一个Condition

对象。

假设读者还记得第5章有关同步类的介绍，理解这个Barrier就非常easy了。

以下调用的

wait函数表示要等待一个同步条件的满足。

*/

mReadyToRunBarrier.wait();

}

不清楚Thread类的读者能够复习一下与第5章有关的Thread类的知识。

status_t SurfaceFlinger::readyToRun()

{

intdpy = 0;

{

//①GraphicPlane是什么？

GraphicPlane& plane(graphicPlane(dpy));

//②为这个GraphicPlane设置一个HAL对象——DisplayHardware

DisplayHardware* const hw = new DisplayHardware(this, dpy);

plane.setDisplayHardware(hw);

}

//创建Surface系统中的“CB”对象。依照老规矩。应该先创建一块共享内存，然后使用placment new

mServerHeap = new MemoryHeapBase(4096,

MemoryHeapBase::READ_ONLY,

"SurfaceFlingerread-only heap");

/*

注意这个“CB“对象的类型是surface_flinger_cblk_t。为什么在CB上打引號呢？由于这个对象

谈不上什么控制，仅仅只是被用来存储一些信息罢了。

其控制作用全然达不到audio_track_cblk_t

的程度。基于这样的事实，我们把前面提到的SharedBuffer家族称之为CB对象。

*/

mServerCblk=

static_cast<surface_flinger_cblk_t*>(mServerHeap->getBase());

//placementnew创建surface_flinger_cblk_t

new(mServerCblk) surface_flinger_cblk_t;

constGraphicPlane& plane(graphicPlane(dpy));

constDisplayHardware& hw = plane.displayHardware();

constuint32_t w = hw.getWidth();

constuint32_t h = hw.getHeight();

constuint32_t f = hw.getFormat();

hw.makeCurrent();

//当前仅仅有一块屏

mServerCblk->connected|= 1<<dpy;

//屏幕在“CB”对象中的代表是display_cblk_t

display_cblk_t* dcblk = mServerCblk->displays + dpy;

memset(dcblk, 0, sizeof(display_cblk_t));

dcblk->w            =plane.getWidth();

dcblk->h            =plane.getHeight();

......//获取屏幕信息

//还用上了内联汇编语句。

asmvolatile ("":::"memory");

/*

以下是一些和OpenGL相关的函数调用。读者如感兴趣，能够研究一下。

至少SurfaceFlinger.cpp中所涉及的相关代码还不算难懂

*/

glActiveTexture(GL_TEXTURE0);

glBindTexture(GL_TEXTURE_2D, 0);

......

glOrthof(0, w, h, 0, 0, 1);

//LayerDim是Dim类型的Layer

LayerDim::initDimmer(this, w, h);

//还记得在onFirstRef函数中的wait吗？以下的open将触发这个同步条件

mReadyToRunBarrier.open();

//资源准备好后。init将启动bootanim程序。这样就见到开机动画了。

property_set("ctl.start", "bootanim");

returnNO_ERROR;

}

GraphicPlane     mGraphicPlanes[1];

DisplayHardware::DisplayHardware(

const sp<SurfaceFlinger>& flinger,

uint32_t dpy)

:DisplayHardwareBase(flinger, dpy)

{

init(dpy); //最重要的是这个init函数。

}

void DisplayHardware::init(uint32_t dpy)

{

//FrameBufferNativeWindow实现了对FrameBuffer的管理和操作，该类中创建了两个

//FrameBuffer，分别起到FrontBuffer和BackBuffer的作用。

mNativeWindow = new FramebufferNativeWindow();

framebuffer_device_t const * fbDev = mNativeWindow->getDevice();

mOverlayEngine = NULL;

hw_module_t const* module;//Overlay相关

if(hw_get_module(OVERLAY_HARDWARE_MODULE_ID, &module) == 0) {

overlay_control_open(module, &mOverlayEngine);

}

......

EGLint w, h, dummy;

EGLintnumConfigs=0;

EGLSurface surface;

EGLContext context;

mFlags= CACHED_BUFFERS;

//EGLDisplay在EGL中代表屏幕

EGLDisplay display = eglGetDisplay(EGL_DEFAULT_DISPLAY);

......

/*

surface是EGLSurface类型，以下这个函数会将EGL和Android中的Display系统绑定起来，

兴许就能够利用OpenGL在这个Surface上绘画。然后通过eglSwappBuffers输出图像了。

*/

surface= eglCreateWindowSurface(display, config,

mNativeWindow.get(),NULL);

......

mDisplay = display;

mConfig  = config;

mSurface = surface;

mContext = context;

mFormat  = fbDev->format;

mPageFlipCount = 0;

}

本节的内容，实际上涉及另外一个比Surface更复杂的Display系统。出于篇幅和精力的原因，本书眼下不打算讨论它。

bool SurfaceFlinger::threadLoop()

{

waitForEvent();//① 等待什么事件呢？

if (UNLIKELY(mConsoleSignals)) {

handleConsoleEvents();

}

if(LIKELY(mTransactionCount == 0)) {

const uint32_t mask = eTransactionNeeded | eTraversalNeeded;

uint32_t transactionFlags = getTransactionFlags(mask);

if(LIKELY(transactionFlags)) {

//Transaction（事务）处理。放到本节最后来讨论

handleTransaction(transactionFlags);

}

}

//②处理PageFlipping工作

handlePageFlip();

constDisplayHardware& hw(graphicPlane(0).displayHardware());

if (LIKELY(hw.canDraw() && !isFrozen())) {

//③处理重绘

handleRepaint();

hw.compositionComplete();

//④投递BackBuffer

unlockClients();

postFramebuffer();

} else{

unlockClients();

usleep(16667);

}

returntrue;

}

void SurfaceFlinger::waitForEvent()

{

while(true) {

nsecs_t timeout = -1;

const nsecs_t freezeDisplayTimeout = ms2ns(5000);

......

MessageList::value_type msg = mEventQueue.waitMessage(timeout);

......//另一些和冻屏相关的内容

if(msg != 0) {

switch (msg->what) {

//千辛万苦就等这一个重绘消息

case MessageQueue::INVALIDATE:

return;

}

}

}

}

void SurfaceFlinger::signal() const {

const_cast<SurfaceFlinger*>(this)->signalEvent();

}

void SurfaceFlinger::signalEvent() {

mEventQueue.invalidate(); //往消息队列中增加INVALIDATE消息

}

注意：事务处理将在8.5.3节中介绍。

void SurfaceFlinger::handlePageFlip()

{

bool visibleRegions = mVisibleRegionsDirty;

/*

还记得前面所说的mCurrentState吗？它保存了全部显示层的信息，而绘制的时候使用的

mDrawingState则保存了当前须要显示的显示层信息。

*/

LayerVector& currentLayers =

const_cast<LayerVector&>(mDrawingState.layersSortedByZ);

//①调用lockPageFlip

visibleRegions |= lockPageFlip(currentLayers);

const DisplayHardware& hw =graphicPlane(0).displayHardware();

//取得屏幕的区域

const Region screenRegion(hw.bounds());

if (visibleRegions) {

Region opaqueRegion;

computeVisibleRegions(currentLayers, mDirtyRegion,opaqueRegion);

mWormholeRegion = screenRegion.subtract(opaqueRegion);

mVisibleRegionsDirty = false;

}

//② 调用unlockPageFlip

unlockPageFlip(currentLayers);

mDirtyRegion.andSelf(screenRegion);

}

bool SurfaceFlinger::lockPageFlip(constLayerVector& currentLayers)

{

boolrecomputeVisibleRegions = false;

size_tcount = currentLayers.size();

sp<LayerBase> const* layers = currentLayers.array();

for(size_t i=0 ; i<count ; i++) {

const sp<LayerBase>& layer = layers[i];

//调用每一个显示层的lockPageFlip

layer->lockPageFlip(recomputeVisibleRegions);

}

returnrecomputeVisibleRegions;

}

void Layer::lockPageFlip(bool&recomputeVisibleRegions)

{

//lcblk是SharedBufferServer类型，调用retireAndLock函数将返回FrontBuffer的

//索引號

ssize_tbuf = lcblk->retireAndLock();

......

mFrontBufferIndex = buf;

//得到FrontBuffer相应的GraphicBuffer

sp<GraphicBuffer> newFrontBuffer(getBuffer(buf));

if (newFrontBuffer != NULL) {

//取出脏区域

const Region dirty(lcblk->getDirtyRegion(buf));

//和GraphicBuffer所表示的区域进行裁剪，得到一个脏区域

mPostedDirtyRegion = dirty.intersect( newFrontBuffer->getBounds() );

const Layer::State& front(drawingState());

if(newFrontBuffer->getWidth()  ==front.requested_w &&

newFrontBuffer->getHeight() ==front.requested_h)

{

if ((front.w != front.requested_w) ||

(front.h != front.requested_h))

{

...... //须要又一次计算可见区域

recomputeVisibleRegions = true;

}

mFreezeLock.clear();

}

} else{

mPostedDirtyRegion.clear();

}

if(lcblk->getQueuedCount()) {

mFlinger->signalEvent();

}

/*

假设脏区域不为空，则须要绘制成纹理，reloadTexture将绘制一张纹理保存在

mTextures数组中，里边涉及非常多OpenGL的操作，读者有兴趣能够自己研究。

*/

if(!mPostedDirtyRegion.isEmpty()) {

reloadTexture( mPostedDirtyRegion );

}

}

void SurfaceFlinger::unlockPageFlip(constLayerVector& currentLayers)

{

constGraphicPlane& plane(graphicPlane(0));

constTransform& planeTransform(plane.transform());

size_tcount = currentLayers.size();

sp<LayerBase> const* layers = currentLayers.array();

for(size_t i=0 ; i<count ; i++) {

const sp<LayerBase>& layer = layers[i];

//调用每一个显示层的unlockPageFlip，Layer的unlockPageFlip主要做一些

//区域的清理工作。读者能够自己看看。

layer->unlockPageFlip(planeTransform, mDirtyRegion);

}

}

void SurfaceFlinger::handleRepaint()

{

mInvalidRegion.orSelf(mDirtyRegion);

if(mInvalidRegion.isEmpty()) {

return;

}

......

constDisplayHardware& hw(graphicPlane(0).displayHardware());

glMatrixMode(GL_MODELVIEW);

glLoadIdentity();

uint32_t flags = hw.getFlags();

if((flags & DisplayHardware::SWAP_RECTANGLE) ||

(flags & DisplayHardware::BUFFER_PRESERVED))

{

......//计算mDirtyRegion

}

// 在脏区域上进行绘制

composeSurfaces(mDirtyRegion);

mDirtyRegion.clear();

}

void SurfaceFlinger::composeSurfaces(constRegion& dirty)

{

constSurfaceFlinger& flinger(*this);

constLayerVector& drawingLayers(mDrawingState.layersSortedByZ);

constsize_t count = drawingLayers.size();

sp<LayerBase> const* const layers = drawingLayers.array();

for(size_t i=0 ; i<count ; ++i) {

const sp<LayerBase>& layer = layers[i];

const Region&visibleRegion(layer->visibleRegionScreen);

if(!visibleRegion.isEmpty())  {

const Region clip(dirty.intersect(visibleRegion));

if (!clip.isEmpty()) {

layer->draw(clip); //调用各个显示层的layer函数

}

}

}

}

void LayerBase::draw(const Region& inClip)const

{

......

glEnable(GL_SCISSOR_TEST);

onDraw(clip);//调用子类的onDraw函数

}

void Layer::onDraw(const Region& clip) const

{

intindex = mFrontBufferIndex;

if(mTextures[index].image == EGL_NO_IMAGE_KHR)

index = 0;

GLuint textureName = mTextures[index].name;

....

Region holes(clip.subtract(under));

if(!holes.isEmpty()) {

clearWithOpenGL(holes);

}

return;

}

//index是FrontBuffer相应生成的纹理，在lockPageFlip函数中就已经生成了。

drawWithOpenGL(clip,mTextures[index]);//将纹理画上去，里面有非常多和OpenGL相关内容

}

void LayerBase::drawWithOpenGL(const Region&clip, const Texture& texture) const

{

constDisplayHardware& hw(graphicPlane(0).displayHardware());

constuint32_t fbHeight = hw.getHeight();

constState& s(drawingState());

//validateTexture函数内部将绑定指定的纹理

validateTexture(texture.name);

//以下就是OpenGL操作函数了

glEnable(GL_TEXTURE_2D);

......

glMatrixMode(GL_TEXTURE);

glLoadIdentity();

//坐标旋转

switch(texture.transform) {

case HAL_TRANSFORM_ROT_90:

glTranslatef(0, 1, 0);

glRotatef(-90, 0, 0, 1);

break;

case HAL_TRANSFORM_ROT_180:

glTranslatef(1, 1, 0);

glRotatef(-180, 0, 0, 1);

break;

case HAL_TRANSFORM_ROT_270:

glTranslatef(1, 0, 0);

glRotatef(-270, 0, 0, 1);

break;

}

if (texture.NPOTAdjust) {

//缩放处理

glScalef(texture.wScale, texture.hScale, 1.0f);

}

//使能纹理坐标

glEnableClientState(GL_TEXTURE_COORD_ARRAY);

//设置顶点坐标

glVertexPointer(2, GL_FIXED, 0, mVertices);

//设置纹理坐标

glTexCoordPointer(2, GL_FIXED, 0, texCoords);

while(it != end) {

const Rect& r = *it++;

const GLint sy = fbHeight - (r.top + r.height());

//裁剪

glScissor(r.left, sy, r.width(), r.height());

//画矩形

glDrawArrays(GL_TRIANGLE_FAN, 0, 4);

}

//禁止纹理坐标

glDisableClientState(GL_TEXTURE_COORD_ARRAY);

}

void LayerBase::validateTexture(GLinttextureName) const

{

//以下这个函数将绑定纹理

glBindTexture(GL_TEXTURE_2D, textureName);

......//其它一些设置

}

我在Android平台上有几个月的OpenGL开发经历，还谈不上非常深刻，当中的一些资料，希望能够给感兴趣的读者提供參考。

1）OpenGL的入门教材当选NeHe的资料。大略看前几章就可以。

2） Android平台上关于OpenGL ES的开发，有一篇非常具体的Word文档叫《OpenGL ESTutorial for Android》。

该文具体描写叙述了在Android平台上进行OpenGL开发的流程。大家可跟着这篇教材，在模拟器上做一些练习。那里面涉及到的一些基础知识，从前面介绍的入门教材中能够学到。

3）有了前面两点的基础后，就须要对整个OpenGL有比較完整深入的了解了。我在那时所看的书是《OpenGL Programming Guide (7th Edition)》。

该书非常厚，有1000多页。里面有一些内容可能与工作无涉。仅仅要大概知道有那回事就可以了，临时不必深入学习，等须要时再进一步学习并运用。我在开发的项目中曾用到的光照、雾化等效果，都是之前先知道有这个东西。后来在项目中才逐渐学习运用的。

4）嵌入式平台上用的事实上是OpenGL ES。这里。另一本书叫《OpenGL ES 2.0 Programming Guide》，它介绍了OpenGL ES的开发，读者可认真修习。

5）在Android SDK文档中。对OpenGL API的描写叙述仅仅寥寥数语。怎么办？由于它使用了J2ME中的javax.microedition.khronos.opengles包。所以J2ME的SDK文档中对OpenGL的API有着非常具体的描写叙述。读者手头应该要有一个J2ME的文档。

6）假设想做深入开发，就不得不学习计算机图形学了。我后来买了书。可惜没时间学了。

void SurfaceFlinger::unlockClients()

{

constLayerVector& drawingLayers(mDrawingState.layersSortedByZ);

constsize_t count = drawingLayers.size();

sp<LayerBase> const* const layers = drawingLayers.array();

for (size_t i=0 ; i<count ; ++i) {

const sp<LayerBase>& layer = layers[i];

layer->finishPageFlip();

}

}

void Layer::finishPageFlip()

{

//释放FrontBufferIndex

status_t err = lcblk->unlock( mFrontBufferIndex );

}

void SurfaceFlinger::postFramebuffer()

{

if(!mInvalidRegion.isEmpty()) {

const DisplayHardware& hw(graphicPlane(0).displayHardware());

const nsecs_t now = systemTime();

mDebugInSwapBuffers = now;

//调用这个函数后。混合后的图像就会传递到屏幕中显示了

hw.flip(mInvalidRegion);

mLastSwapBufferTime = systemTime() - now;

mDebugInSwapBuffers = 0;

mInvalidRegion.clear();

}

}

void DisplayHardware::flip(const Region&dirty) const

{

checkGLErrors();

EGLDisplay dpy = mDisplay;

EGLSurface surface = mSurface;

......

if(mFlags & PARTIAL_UPDATES) {

mNativeWindow->setUpdateRectangle(dirty.getBounds());

}

mPageFlipCount++;

eglSwapBuffers(dpy, surface);//PageFlipping，此后图像终于显示在屏幕上了！

}

Surface createSurfaceLocked() {

Surface.openTransaction(); //開始一次transaction

try {

try {

mSurfaceX = mFrame.left + mXOffset;

mSurfaceY = mFrame.top + mYOffset;

//设置Surface的位置

mSurface.setPosition(mSurfaceX, mSurfaceY);

......

}

}finally {

Surface.closeTransaction(); //关闭这次事务

}

static void Surface_openTransaction(JNIEnv* env,jobject clazz)

{

//调用SurfaceComposerClient的openGlobalTransaction函数

SurfaceComposerClient::openGlobalTransaction();

}

voidSurfaceComposerClient::openGlobalTransaction()

{

Mutex::Autolock _l(gLock);

......

constsize_t N = gActiveConnections.size();

for(size_t i=0; i<N; i++) {

sp<SurfaceComposerClient>client(gActiveConnections.valueAt(i).promote());

//gOpenTransactions存储当前提交事务请求的Client

if(client != 0 && gOpenTransactions.indexOf(client) < 0) {

//Client是保存在全局变量gActiveConnections中的SurfaceComposerClient

//对象。调用它的openTransaction。

if (client->openTransaction() == NO_ERROR) {

if (gOpenTransactions.add(client) < 0) {

client->closeTransaction();

}

}

......

}

}

}

status_tSurfaceComposerClient::openTransaction()

{

if(mStatus != NO_ERROR)

return mStatus;

Mutex::Autolock _l(mLock);

mTransactionOpen++; //一个计数值，用来控制事务的提交。

if(mPrebuiltLayerState == 0) {

mPrebuiltLayerState = new layer_state_t;

}

returnNO_ERROR;

}

static void Surface_setPosition(JNIEnv* env,jobject clazz, jint x, jint y)

{

constsp<SurfaceControl>& surface(getSurfaceControl(env, clazz));

if(surface == 0) return;

status_t err = surface->setPosition(x, y);

}

status_t SurfaceControl::setPosition(int32_t x,int32_t y) {

constsp<SurfaceComposerClient>& client(mClient);

status_t err = validate();

if (err < 0) return err;

//调用SurfaceComposerClient的setPosition函数

returnclient->setPosition(mToken, x, y);

}

status_tSurfaceComposerClient::setPosition(SurfaceID id, int32_t x, int32_t y)

{

layer_state_t* s = _lockLayerState(id); //找到相应的layer_state_t

if(!s) return BAD_INDEX;

s->what |= ISurfaceComposer::ePositionChanged;

s->x = x;

s->y = y;  //上面几句改动了这块layer的參数

_unlockLayerState(); //该函数将unlock一个同步对象，其它没有做什么工作

returnNO_ERROR;

}

static void Surface_closeTransaction(JNIEnv*env, jobject clazz)

{

SurfaceComposerClient::closeGlobalTransaction();

}

voidSurfaceComposerClient::closeGlobalTransaction()

{

......

const size_t N = clients.size();

sp<ISurfaceComposer>sm(getComposerService());

//①先调用SF的openGlobalTransaction

sm->openGlobalTransaction();

for (size_t i=0; i<N; i++) {

//②然后调用每一个SurfaceComposerClient对象的closeTransaction

clients[i]->closeTransaction();

}

//③最后调用SF的closeGlobalTransaction

sm->closeGlobalTransaction();

}

void SurfaceFlinger::openGlobalTransaction()

{

android_atomic_inc(&mTransactionCount);//又是一个计数控制

}

status_tSurfaceComposerClient::closeTransaction()

{

if(mStatus != NO_ERROR)

return mStatus;

Mutex::Autolock _l(mLock);

......

constssize_t count = mStates.size();

if (count) {

//mStates是这个SurfaceComposerClient中保存的全部layer_state_t数组。也就是

//每一个Surface一个。

然后调用跨进程的setState

mClient->setState(count, mStates.array());

mStates.clear();

}

returnNO_ERROR;

}

status_t SurfaceFlinger::setClientState(ClientIDcid, int32_t count,

const layer_state_t*states)

{

Mutex::Autolock _l(mStateLock);

uint32_t flags = 0;

cid<<= 16;

for(int i=0 ; i<count ; i++) {

const layer_state_t& s = states[i];

sp<LayerBaseClient> layer(getLayerUser_l(s.surface | cid));

if(layer != 0) {

const uint32_t what = s.what;

if (what & ePositionChanged) {

if (layer->setPosition(s.x, s.y))

//eTraversalNeeded表示须要遍历全部显示层

flags |= eTraversalNeeded;

}

....

if(flags) {

setTransactionFlags(flags);//这里将会触发threadLoop的事件。

}

returnNO_ERROR;

}

uint32_tSurfaceFlinger::setTransactionFlags(uint32_t flags, nsecs_t delay)

{

uint32_t old = android_atomic_or(flags, &mTransactionFlags);

if((old & flags)==0) {

if(delay > 0) {

signalDelayedEvent(delay);

}else {

signalEvent();  //设置完mTransactionFlags后，触发事件。

}

}

returnold;

}

void SurfaceFlinger::closeGlobalTransaction()

{

if (android_atomic_dec(&mTransactionCount) ==1) {

//注意以下语句的运行条件，当mTransactionCount变为零时才运行，这意味着

//openGlobalTransaction两次的话。仅仅有最后一个closeGlobalTransaction调用

//才会真正地提交事务

signalEvent();

Mutex::Autolock _l(mStateLock);

//假设这次事务涉及尺寸调整，则须要等一段时间

while (mResizeTransationPending) {

status_t err = mTransactionCV.waitRelative(mStateLock, s2ns(5));

if (CC_UNLIKELY(err != NO_ERROR)) {

mResizeTransationPending = false;

break;

}

}

}

}

bool SurfaceFlinger::threadLoop()

{

waitForEvent();

if(LIKELY(mTransactionCount == 0)) {

const uint32_t mask = eTransactionNeeded | eTraversalNeeded;

uint32_ttransactionFlags = getTransactionFlags(mask);

if(LIKELY(transactionFlags)) {

handleTransaction(transactionFlags);

}

}

...

}

uint32_t SurfaceFlinger::getTransactionFlags(uint32_tflags)

{

//先通过原子操作去掉mTransactionFlags中相应的位。

//而后原子操作返回的旧值和flags进行与操作

return android_atomic_and(~flags,&mTransactionFlags) & flags;

}

void SurfaceFlinger::handleTransaction(uint32_ttransactionFlags)

{

Vector< sp<LayerBase> > ditchedLayers;

{

Mutex::Autolock _l(mStateLock);

//调用handleTransactionLocked函数处理

handleTransactionLocked(transactionFlags, ditchedLayers);

}

constsize_t count = ditchedLayers.size();

for(size_t i=0 ; i<count ; i++) {

if(ditchedLayers[i] != 0) {

//ditch是丢弃的意思，有些显示层可能被hide了。所以这里做些收尾的工作

ditchedLayers[i]->ditch();

}

}

}

void SurfaceFlinger::handleTransactionLocked(

uint32_t transactionFlags, Vector< sp<LayerBase> >&ditchedLayers)

{

//这里使用了mCurrentState，它的layersSortedByZ数组存储了SF中全部的显示层

constLayerVector& currentLayers(mCurrentState.layersSortedByZ);

constsize_t count = currentLayers.size();

constbool layersNeedTransaction = transactionFlags & eTraversalNeeded;

//假设须要遍历全部显示的话。

if(layersNeedTransaction) {

for (size_t i=0 ; i<count ; i++) {

const sp<LayerBase>& layer = currentLayers[i];

uint32_t trFlags = layer->getTransactionFlags(eTransactionNeeded);

if (!trFlags) continue;

//调用各个显示层的doTransaction函数。

constuint32_t flags = layer->doTransaction(0);

if (flags & Layer::eVisibleRegion)

mVisibleRegionsDirty = true;

}

}

if(transactionFlags & eTransactionNeeded) {

if(mCurrentState.orientation != mDrawingState.orientation) {

//横竖屏假设发生切换。须要相应变换设置。

const int dpy = 0;

const int orientation = mCurrentState.orientation;

const uint32_t type = mCurrentState.orientationType;

GraphicPlane& plane(graphicPlane(dpy));

plane.setOrientation(orientation);

......

}

/*

mLayersRemoved变量在显示层被移除的时候设置。比如removeLayer函数。这些函数

也会触发handleTranscation函数的运行

*/

if(mLayersRemoved) {

mLayersRemoved = false;

mVisibleRegionsDirty = true;

const LayerVector& previousLayers(mDrawingState.layersSortedByZ);

const size_t count = previousLayers.size();

for (size_t i=0 ; i<count ; i++) {

const sp<LayerBase>& layer(previousLayers[i]);

if (currentLayers.indexOf( layer ) < 0) {

ditchedLayers.add(layer);

mDirtyRegionRemovedLayer.orSelf(layer->visibleRegionScreen);

}

}

}

free_resources_l();

}

//提交事务处理。有必要进去看看。

commitTransaction();

}

void SurfaceFlinger::commitTransaction()

{

//mDrawingState将使用更新后的mCurrentState

mDrawingState = mCurrentState;

mResizeTransationPending = false;

//触发一个条件变量。这样等待在closeGlobalTransaction函数中的线程能够放心地返回了。

mTransactionCV.broadcast();

}

class SharedBufferStack{

......

/*

尽管PageFlipping使用Front和Back两个Buffer就能够了，可是SBT的结构和相关算法

是支持多个缓冲的。

另外，缓冲是依照块来获取的，也就是一次获得一块缓冲，每块缓冲用

一个编号表示（这一点在之前的分析已经介绍过了）。

*/

int32_t head;

int32_tavailable; //当前可用的空暇缓冲个数

int32_t queued;    //SBC投递的脏缓冲个数

int32_tinUse;  //SBS当前正在使用的缓冲编号

......//上面这几个參数联合SBC中的tail。我称之为控制參数。

}

SharedBufferServer::SharedBufferServer(SharedClient*sharedClient,

int surface, int num, int32_t identity)

:SharedBufferBase(sharedClient, surface, num, identity)

{

mSharedStack->init(identity);//这个函数将设置inUse为-1

//以下设置SBT中的參数，我们关注前三个

mSharedStack->head = num-1;

mSharedStack->available = num;

mSharedStack->queued = 0;

//设置完后，head=2-1=1,available=2,queued=0，inUse=-1

mSharedStack->reallocMask = 0;

memset(mSharedStack->dirtyRegion, 0,sizeof(mSharedStack->dirtyRegion));

}

SharedBufferClient::SharedBufferClient(SharedClient*sharedClient,

int surface, int num, int32_t identity)

:SharedBufferBase(sharedClient, surface, num, identity), tail(0)

{

tail =computeTail(); //tail是SBC定义的变量。注意它不是SBT定义的。

}

int32_t SharedBufferClient::computeTail() const

{

SharedBufferStack& stack( *mSharedStack );

int32_t newTail;

int32_t avail;

int32_t head;

do {

avail = stack.available; //available=2,head=1

head = stack.head;

}while (stack.available != avail);

newTail = head - avail + 1;//newTail=1-2+1=0

if(newTail < 0) {

newTail += mNumBuffers;

} elseif (newTail >= mNumBuffers) {

newTail -= mNumBuffers;

}

return newTail;//计算得到newTail=0

}

ssize_t SharedBufferClient::dequeue()

{

SharedBufferStack& stack( *mSharedStack );

......

//DequeueCondition函数对象

DequeueCondition condition(this);

status_t err = waitForCondition(condition);

//成功以后，available减1，表示当前可用的空暇buffer仅仅有1个

if (android_atomic_dec(&stack.available) == 0) {

......

}

int dequeued = tail; //tail值为0。所以dequeued的值为0。

//tail加1。假设超过2，则又一次置为0，这表明tail的值在0,1间循环。

tail =((tail+1 >= mNumBuffers) ? 0 : tail+1);

......

//返回的这个dequeued值为零，也就是tail加1操作前的旧值。这一点请读者务必注意。

returndequeued;

}

bool SharedBufferClient::DequeueCondition::operator()(){

returnstack.available > 0;//仅仅要available大于0就算满足条件。第一次进来肯定满足

}

status_t SharedBufferClient::lock(int buf)

{

LockCondition condition(this, buf);

status_terr = waitForCondition(condition);

returnerr;

}

boolSharedBufferClient::LockCondition::operator()() {

/*

这个条件事实上就是推断编号为buf的Buffer是不是被使用了。

buf值为0，head值为1。queued为0，inUse为-1

*/

return(buf != stack.head ||

(stack.queued > 0 && stack.inUse!= buf));

}

status_t SharedBufferClient::queue(int buf)

{

QueueUpdate update(this);

status_t err = updateCondition( update );

......

returnerr;

}

//直接看这个QueueUpdate函数对象

ssize_tSharedBufferClient::QueueUpdate::operator()() {

android_atomic_inc(&stack.queued);//queued增加1。如今该值由零变为1

returnNO_ERROR;

}

ssize_t SharedBufferServer::retireAndLock()

{

RetireUpdate update(this, mNumBuffers);

ssize_t buf = updateCondition( update );

returnbuf;

}

ssize_tSharedBufferServer::RetireUpdate::operator()() {

//先取得head值，为1

int32_t head = stack.head;

//inUse被设置为1。表明要使用1吗？眼下的脏缓冲应该是0才对

android_atomic_write(head, &stack.inUse);

int32_tqueued;

do {

queued = stack.queued; //queued眼下为1

if(queued == 0) {

return NOT_ENOUGH_DATA;

}

//以下这个原子操作使得stack.queued减1.

}while (android_atomic_cmpxchg(queued, queued-1, &stack.queued));

//while循环退出后，queued减1，又变为0。

//head值也在0,1间循环。如今head值变为0了

head =((head+1 >= numBuffers) ? 0 : head+1);

//inUse被设置为0

android_atomic_write(head, &stack.inUse);

// head值被设为0

android_atomic_write(head, &stack.head);

// available加1，变成2.

android_atomic_inc(&stack.available);

returnhead;//返回0

}

ssize_tSharedBufferServer::UnlockUpdate::operator()() {

......

android_atomic_write(-1, &stack.inUse);//inUse被设置为-1

returnNO_ERROR;

}

像以下这样的代码，假设有锁控制的话根本用不着一个while循环，由于有锁的保护。没有其它线程

能够改动stack.queued的值，所以用while来循环推断android_atomic_cmpxchg没有什么意义。

int32_tqueued;

do {

queued = stack.queued;

if(queued == 0) {

return NOT_ENOUGH_DATA;

}

}while (android_atomic_cmpxchg(queued, queued-1, &stack.queued));

为什么我对生产/消费的同步控制如此感兴趣呢？这和自己工作的经历有些关系。

由于之前曾做过一个单写多读的跨进程缓冲类，也就是一个生产者。多个消费者。为了保证正确性和一定的效率，我们在算法上曾做了非常多改进，但还是大量使用了锁。所以我非常好奇Google是怎么做到的，这也体现了一个高手的内功修养。要是由读者自己来实现，结果会如何呢？

sp<LayerBaseClient> SurfaceFlinger::createPushBuffersSurfaceLocked(

const sp<Client>& client, DisplayID display,

int32_t id, uint32_t w, uint32_t h, uint32_t flags)

{

sp<LayerBuffer> layer = new LayerBuffer(this, display, client,id);

layer->initStates(w, h, flags);

addLayer_l(layer);

returnlayer;

}

status_tCameraService::Client::registerPreviewBuffers()

{

int w, h;

CameraParameters params(mHardware->getParameters());

params.getPreviewSize(&w, &h);

/*

①mHardware代表Camera设备的HAL对象。本书讨论CameraHardwareStub设备。它事实上是

一个虚拟的设备。只是其代码却具有參考价值。

BufferHeap定义为ISurface的内部类，事实上就是对IMemoryHeap的封装

*/

ISurface::BufferHeapbuffers(w, h, w, h,

HAL_PIXEL_FORMAT_YCrCb_420_SP,

mOrientation,

0,

mHardware->getPreviewHeap());

//②调用SurfaceLayerBuffer的registerBuffers函数。

status_t ret = mSurface->registerBuffers(buffers);

returnret;

}

classBufferHeap {

public:

......

//使用这个构造函数

BufferHeap(uint32_t w, uint32_t h,

int32_t hor_stride, int32_t ver_stride,

PixelFormat format, const sp<IMemoryHeap>& heap);

......

~BufferHeap();

uint32_t w;

uint32_t h;

int32_t hor_stride;

int32_t ver_stride;

PixelFormat format;

uint32_t transform;

uint32_t flags;

sp<IMemoryHeap> heap; //heap指向真实的存储对象

};

sp<IMemoryHeap>CameraHardwareStub::getPreviewHeap() const

{

returnmPreviewHeap;//返回一个成员变量，它又是在哪创建的呢？

}

//上面的mPreivewHeap对象由initHeapLocked函数创建。该函数在HAL对象创建的时候被调用

void CameraHardwareStub::initHeapLocked()

{

......

/*

创建一个MemoryHeapBase对象，大小是mPreviewFrameSize * kBufferCount,当中

kBufferCount为4。注意这是一段连续的缓冲。

*/

mPreviewHeap= new MemoryHeapBase(mPreviewFrameSize * kBufferCount);

//mBuffer为MemoryBase数组。元素为4

for (inti = 0; i < kBufferCount; i++) {

mBuffers[i] = new MemoryBase(mPreviewHeap,

i * mPreviewFrameSize, mPreviewFrameSize);

}

}

status_t LayerBuffer::SurfaceLayerBuffer::registerBuffers(

const ISurface::BufferHeap& buffers)

{

sp<LayerBuffer> owner(getOwner());

if (owner != 0)

//调用外部类对象的registerBuffers，所以SurfaceLayerBuffer也是一个Proxy哦。

return owner->registerBuffers(buffers);

returnNO_INIT;

}

//外部类是LayerBuffer。调用它的registerBuffers函数

status_t LayerBuffer::registerBuffers(constISurface::BufferHeap& buffers)

{

Mutex::Autolock _l(mLock);

//创建数据的来源BufferSource，注意我们事实上把MemoryHeap设置上去了

sp<BufferSource> source = new BufferSource(*this, buffers);

status_t result = source->getStatus();

if(result == NO_ERROR) {

mSource = source;//保存这个数据源为mSource。

}

returnresult;

}

//preview线程从Thread类派生。以下这个函数在threadLoop中循环调用

int CameraHardwareStub::previewThread()

{

mLock.lock();

//每次进来mCurrentPreviewFrame都会加1

ssize_t offset = mCurrentPreviewFrame * mPreviewFrameSize;

sp<MemoryHeapBase> heap = mPreviewHeap;

FakeCamera* fakeCamera = mFakeCamera;//虚拟的摄像机设备

//从mBuffers中取一块内存，用于接收来自硬件的数据

sp<MemoryBase>buffer = mBuffers[mCurrentPreviewFrame];

mLock.unlock();

if(buffer != 0) {

intdelay = (int)(1000000.0f / float(previewFrameRate));

void *base = heap->base();//base是mPreviewHeap的起始位置

//以下这个frame代表buffer在mPreviewHeap中的起始位置，还记得图8-31吗？

//四块MemoryBase的起始位置由以下这个代码计算得来

uint8_t *frame = ((uint8_t *)base) + offset;

//取出一帧数据，放到相应的MemoryBase中

fakeCamera->getNextFrameAsYuv422(frame);

//①把含有帧数据的buffer传递到上层

if(mMsgEnabled & CAMERA_MSG_PREVIEW_FRAME)

mDataCb(CAMERA_MSG_PREVIEW_FRAME, buffer, mCallbackCookie);

//mCurrentPreviewFrame 递增，在0到3之间循环

mCurrentPreviewFrame = (mCurrentPreviewFrame + 1) % kBufferCount;

usleep(delay);//模拟真实硬件的延时

}

returnNO_ERROR;

}

voidCameraService::Client::handlePreviewData(const sp<IMemory>& mem)

{

ssize_t offset;

size_t size;

//注意传入的mem參数，它实际上是Camera HAL创建的mBuffers数组中的一个

//offset返回的是这个数组在mPreviewHeap中的偏移量

sp<IMemoryHeap> heap = mem->getMemory(&offset, &size);

if (!mUseOverlay)

{

Mutex::Autolock surfaceLock(mSurfaceLock);

if(mSurface != NULL) {

//调用ISurface的postBuffer，注意我们传入的參数是offset。

mSurface->postBuffer(offset);

}

}

......

}

void LayerBuffer::postBuffer(ssize_t offset)

{

sp<Source> source(getSource());//getSource返回mSource，为BufferSource类型

if(source != 0)

source->postBuffer(offset);//调用BufferSource的postBuffer函数。

}

voidLayerBuffer::BufferSource::postBuffer(ssize_t offset)

{

ISurface::BufferHeap buffers;

{

Mutex::Autolock _l(mBufferSourceLock);

buffers = mBufferHeap;//还记得图8-32吗？

if(buffers.heap != 0) {

//BufferHeap的heap变量指向MemoryHeap,以下取出它的大小

const size_t memorySize = buffers.heap->getSize();

//做一下检查，推断这个offset是不是有问题

if ((size_t(offset) + mBufferSize) > memorySize) {

LOGE("LayerBuffer::BufferSource::postBuffer() "

"invalid buffer(offset=%d, size=%d, heap-size=%d",

int(offset),int(mBufferSize), int(memorySize));

return;

}

}

}

sp<Buffer> buffer;

if (buffers.heap != 0) {

//创建一个LayerBuffer::Buffer

buffer = new LayerBuffer::Buffer(buffers, offset, mBufferSize);

if(buffer->getStatus() != NO_ERROR)

buffer.clear();

setBuffer(buffer);//setBuffer？我们要看看

//mLayer就是外部类LayerBuffer。调用它的invalidate函数将触发SF的重绘

mLayer.invalidate();

}

}

void LayerBuffer::BufferSource::setBuffer(

const sp<LayerBuffer::Buffer>& buffer)

{

//setBuffer函数就是简单地将new出来的Buffer设置给成员变量mBuffer，这么做会有问题吗？Mutex::Autolock_l(mBufferSourceLock);

mBuffer = buffer; //将新的buffer设置为mBuffer，mBuffer原来指向的那个被delete

}

LayerBuffer::Buffer::Buffer(constISurface::BufferHeap& buffers,

ssize_t offset, size_t bufferSize)

:mBufferHeap(buffers), mSupportsCopybit(false)

{

//注意，这个src被定义为引用。所以改动src的信息相当于改动mNativeBuffer的信息

NativeBuffer& src(mNativeBuffer);

src.crop.l = 0;

src.crop.t = 0;

src.crop.r = buffers.w;

src.crop.b = buffers.h;

src.img.w       =buffers.hor_stride ?

: buffers.w;

src.img.h       =buffers.ver_stride ?: buffers.h;

src.img.format  = buffers.format;

//这个base将指向相应的内存起始地址

src.img.base    =(void*)(intptr_t(buffers.heap->base()) + offset);

src.img.handle  = 0;

gralloc_module_tconst * module = LayerBuffer::getGrallocModule();

//做一些处理，有兴趣的读者能够去看看。

if(module && module->perform) {

int err = module->perform(module,

GRALLOC_MODULE_PERFORM_CREATE_HANDLE_FROM_BUFFER,

buffers.heap->heapID(), bufferSize,

offset, buffers.heap->base(),

&src.img.handle);

mSupportsCopybit = (err == NO_ERROR);

}

}

void LayerBuffer::onDraw(const Region& clip)const

{

sp<Source> source(getSource());

if(LIKELY(source != 0)) {

source->onDraw(clip);//source实际类型是BufferSource，我们去看看。

} else{

clearWithOpenGL(clip);

}

}

void LayerBuffer::BufferSource::onDraw(constRegion& clip) const

{

sp<Buffer> ourBuffer(getBuffer());

......//使用这个Buffer，注意使用的时候没有锁控制

mLayer.drawWithOpenGL(clip, mTexture);//生成一个贴图，然后绘制它

}

sp<LayerBuffer::Buffer>LayerBuffer::BufferSource::getBuffer() const

{

Mutex::Autolock_l(mBufferSourceLock);

returnmBuffer;

}

从代码的分析来看，这样的方式确实有问题。我在真实设备上測试的结果，也在一定程度上验证了这一点。通过改动LayerBuffer来解决这问题的难度比較大。是否可在读写具体缓存时加上同步控制呢（比如使用mBuffers[0]的时候调用一下lock。用完后调用unlock）？这样就不用改动LayerBuffer了。读者可再深入研究这个问题。