在camera service这端的结构还是很容易让人迷惑的，我就是看了好久，才缕清楚关系。
service这部分包括以下几个头文件：ICamera.h, ICameraService.h, CameraService.h，对应的实现ICamera.cpp, ICameraService.cpp, CameraService.cpp。
CameraService中包含了一个内部类CameraService::Client，这个CameraService::Client是对ICamera的实现，camera client得到的ICamera就是这个CameraService::Client。
也就是说与camera client真正通信的，是这个CameraService::Client。下面我们来逐步分析service这块的实现。
1.ICameraService
（1）ICameraService.h
其中只定义了3个方法：
virtual int32_t getNumberOfCameras() = 0;
virtual status_t getCameraInfo(int cameraId, struct CameraInfo* cameraInfo) = 0;
virtual sp<ICamera> connect(const sp<ICameraClient>& cameraClient, int cameraId) = 0;
（2） ICameraService.cpp
看代码：
status_t BnCameraService::onTransact( uint32_t code, const Parcel& data, Parcel* reply, uint32_t flags)
{
switch(code) {
case GET_NUMBER_OF_CAMERAS: {
CHECK_INTERFACE(ICameraService, data, reply);
reply->writeInt32(getNumberOfCameras());
return NO_ERROR;
} break;
case GET_CAMERA_INFO: {
CHECK_INTERFACE(ICameraService, data, reply);
CameraInfo cameraInfo;
memset(&cameraInfo, 0, sizeof(cameraInfo));
status_t result = getCameraInfo(data.readInt32(), &cameraInfo);
reply->writeInt32(cameraInfo.facing);
reply->writeInt32(cameraInfo.orientation);
reply->writeInt32(result);
return NO_ERROR;
} break;
case CONNECT: {
CHECK_INTERFACE(ICameraService, data, reply);
sp<ICameraClient> cameraClient = interface_cast<ICameraClient>(data.readStrongBinder());
sp<ICamera> camera = connect(cameraClient, data.readInt32());
reply->writeStrongBinder(camera->asBinder());
return NO_ERROR;
} break;
default:
return BBinder::onTransact(code, data, reply, flags);
}
}
这3个函数的真正实现在CameraService.cpp中。
sp<ICameraClient> cameraClient = interface_cast<ICameraClient>(data.readStrongBinder());中得到此次连接的client，并且传递到connect函数中。保存client信息的目的是保存在CameraService::Client中，在适当的时机回调client。
2.CameraService
（1）CameraService.h
class CameraService : public BinderService<CameraService>, public BnCameraService
这个 BinderService<CameraService>是什么东西呢？看看它的定义，在frameworks/base/include/binder下BinderService.h。
template<typename SERVICE>
class BinderService
{
public:
static status_t publish() {
sp<IServiceManager> sm(defaultServiceManager());
return sm->addService(String16(SERVICE::getServiceName()), new SERVICE());
}
static void publishAndJoinThreadPool() {
sp<ProcessState> proc(ProcessState::self());
sp<IServiceManager> sm(defaultServiceManager());
sm->addService(String16(SERVICE::getServiceName()), new SERVICE());
ProcessState::self()->startThreadPool();
IPCThreadState::self()->joinThreadPool();
}
static void instantiate() { publish(); }
static status_t shutdown() {
return NO_ERROR;
}
};
return sm->addService(String16(SERVICE::getServiceName()), new SERVICE());从这句代码，可以看出，这是native service在向SM注册。
其实这个模板类，就是为native service提供了一向service注册的统一方式。
publish和publishAndJoinThreadPool的区别就是在于，注册之后是否开启线程池来监听。
因为之前看过media server的代码，在media server的main函数里，调用的就是CameraService.instantiate来注册的，因为camera service是跑在media server进程里的，所以camera service不需要自己来开启线程来循环监听。
所以，我觉得调用publish和publishAndJoinThreadPool的不同场景是：如果这个service是跑在别的进程里的，就调用publish，如果是自己单独一个进程，就调用publishAndJoinThreadPool。
（2）CameraService.cpp
实现在ICameraService定义的3个函数。
getNumberOfCameras（）和getCameraInfo（）的实现很简单。
int32_t CameraService::getNumberOfCameras() {
return mNumberOfCameras; //mNumberOfCameras是在构造函数中初始化的，mNumberOfCameras = HAL_getNumberOfCameras();
}
status_t CameraService::getCameraInfo(int cameraId,
struct CameraInfo* cameraInfo) {
if (cameraId < 0 || cameraId >= mNumberOfCameras) {
return BAD_VALUE;
}
HAL_getCameraInfo(cameraId, cameraInfo);
return OK;
}
主要看connect函数：
sp<ICamera> CameraService::connect( const sp<ICameraClient>& cameraClient, int cameraId) {
sp<Client> client; //CameraService::Client
…………
if (mClient[cameraId] != 0) {
client = mClient[cameraId].promote();
if (client != 0) {
if (cameraClient->asBinder() == client->getCameraClient()->asBinder()) {
LOG1("CameraService::connect X (pid %d) (the same client)",
callingPid);
return client;
}
}
}
…………
sp<CameraHardwareInterface> hardware = HAL_openCameraHardware(cameraId); //获取CameraHardwareInterface接口，在下边的代码构造Client时传递进入
…………
CameraInfo info;
HAL_getCameraInfo(cameraId, &info);
client = new Client(this, cameraClient, hardware, cameraId, info.facing, callingPid); //这里就是之前说过的，在onTransact保存的client
mClient[cameraId] = client;
return client;
}
3.CameraService::Client
class Client : public BnCamera， Client是对ICamera的实现。
构造函数：
CameraService::Client::Client(const sp<CameraService>& cameraService,
const sp<ICameraClient>& cameraClient,
const sp<CameraHardwareInterface>& hardware,
int cameraId, int cameraFacing, int clientPid) {
int callingPid = getCallingPid();
LOG1("Client::Client E (pid %d)", callingPid);
mCameraService = cameraService;
mCameraClient = cameraClient;
mHardware = hardware;
mCameraId = cameraId;
mCameraFacing = cameraFacing;
mClientPid = clientPid;
mUseOverlay = mHardware->useOverlay();
mMsgEnabled = 0;
mHardware->setCallbacks(notifyCallback,
dataCallback,
dataCallbackTimestamp,
(void *)cameraId);
// Enable zoom, error, and focus messages by default
enableMsgType(CAMERA_MSG_ERROR |
CAMERA_MSG_ZOOM |
CAMERA_MSG_FOCUS);
mOverlayW = 0;
mOverlayH = 0;
// Callback is disabled by default
mPreviewCallbackFlag = FRAME_CALLBACK_FLAG_NOOP;
mOrientation = getOrientation(0, mCameraFacing == CAMERA_FACING_FRONT);
mOrientationChanged = false;
cameraService->setCameraBusy(cameraId);
cameraService->loadSound();
LOG1("Client::Client X (pid %d)", callingPid);
}
主要进行一些成员变量的初始化，主要有mCameraClient， mHardware等。都是在CameraService::connect时进行的。
Client的各个函数可以自己看一下，主要就是对mHardware的调用，这是一个CameraHardwareInterface接口，也就是对HAL层的一个封装。比如stopRecording函数：
// stop recording mode
void CameraService::Client::stopRecording() {
……
mCameraService->playSound(SOUND_RECORDING);
disableMsgType(CAMERA_MSG_VIDEO_FRAME);
mHardware->stopRecording();
…………
}
4.ICamera
最后再来说一下ICamera。
这个ICamera定义的其实就是camera client与camera service的接口。
当camera client连接camera service时，也就是调用CameraService::connect时，返回一个ICamera接口给camera client调用，这个ICamera接口的真正实现是CameraService::Client。
