Android中Camera的调用流程可分为以下几个层次：
Package->Framework->JNI->Camera(cpp)--(binder)-->CameraService->Camera HAL->Camera Driver
以拍照流程为例：
1. 各个参数设置完成，对焦完成后，位于Package的Camera.java会调用Framework中Camera.java的takePicture函数，如下：
public final void takePicture(ShutterCallback shutter, PictureCallback raw,
PictureCallback postview, PictureCallback jpeg) {
mShutterCallback = shutter;
mRawImageCallback = raw;
mPostviewCallback = postview;
mJpegCallback = jpeg;
native_takePicture();
}
此函数保存Package层传下的callback函数，同时调用JNI层的native_takePicture
2. JNI层的native_takePicture自己并没有做太多事情，只是简单地调用cpp的Camera中的takePicture函数。此前已经把JNI中的一个对象注册成了Camera.cpp的listener
3. 位于frameworks/base/libs/camera是向CameraService请求服务的客户端，但它本身也继承了一个BnCameraClient类，用于CameraService回调自己。
class ICameraClient: public IInterface
{
public:
DECLARE_META_INTERFACE(CameraClient);
virtual void notifyCallback(int32_t msgType, int32_t ext1, int32_t ext2) = 0;
virtual void dataCallback(int32_t msgType, const sp<IMemory>& data) = 0;
virtual void dataCallbackTimestamp(nsecs_t timestamp, int32_t msgType, const sp<IMemory>& data) = 0;
};
从上面的接口定义可以看到，这个类就是用于回调。
Camera.cpp的takePicture函数是利用open Camera时得到的ICamera对象来继续调用takePicture
4. 接下来通过binder转到另一个进程CameraService中的处理。CameraService中之前已经实例化了一个HAL层的CameraHardware，并把自己的data callback传递给了CameraHardware，这些工作都是由CameraService的内部类Client来完成的，这个Client类继承自BnCamera，是真正提供Camera操作API的类
5. 然后自然是调用HAL层CameraHardware的takePicture函数。从HAL层向下就不是Android的标准代码了，各个厂商有自己不同的实现。但思路应该都是相同的：Camera遵循V4L2架构，利用ioctl发送VIDIOC_DQBUF命令得到有效的图像数据，接着回调HAL层的data callback接口以通知CameraService，CameraService会通过binder通知Camera.cpp，如下：
void CameraService::Client::dataCallback(int32_t msgType,
const sp<IMemory>& dataPtr, void* user) {
LOG2("dataCallback(%d)", msgType);
sp<Client> client = getClientFromCookie(user);
if (client == 0) return;
if (!client->lockIfMessageWanted(msgType)) return;
if (dataPtr == 0) {
LOGE("Null data returned in data callback");
client->handleGenericNotify(CAMERA_MSG_ERROR, UNKNOWN_ERROR, 0);
return;
}
switch (msgType) {
case CAMERA_MSG_PREVIEW_FRAME:
client->handlePreviewData(dataPtr);
break;
case CAMERA_MSG_POSTVIEW_FRAME:
client->handlePostview(dataPtr);
break;
case CAMERA_MSG_RAW_IMAGE:
client->handleRawPicture(dataPtr);
break;
case CAMERA_MSG_COMPRESSED_IMAGE:
client->handleCompressedPicture(dataPtr);
break;
default:
client->handleGenericData(msgType, dataPtr);
break;
}
}
// picture callback - compressed picture ready
void CameraService::Client::handleCompressedPicture(const sp<IMemory>& mem) {
int restPictures = mHardware->getPictureRestCount();
if (!restPictures)
{
disableMsgType(CAMERA_MSG_COMPRESSED_IMAGE);
}
sp<ICameraClient> c = mCameraClient;
mLock.unlock();
if (c != 0) {
c->dataCallback(CAMERA_MSG_COMPRESSED_IMAGE, mem);
}
}
6. Camera.cpp会继续通知它的listener：
// callback from camera service when frame or image is ready
void Camera::dataCallback(int32_t msgType, const sp<IMemory>& dataPtr)
{
sp<CameraListener> listener;
{
Mutex::Autolock _l(mLock);
listener = mListener;
}
if (listener != NULL) {
listener->postData(msgType, dataPtr);
}
}
7. 而这个listener就是我们的JNI层的JNICameraContext对象了：
void JNICameraContext::postData(int32_t msgType, const sp<IMemory>& dataPtr)
{
// VM pointer will be NULL if object is released
Mutex::Autolock _l(mLock);
JNIEnv *env = AndroidRuntime::getJNIEnv();
if (mCameraJObjectWeak == NULL) {
LOGW("callback on dead camera object");
return;
}
// return data based on callback type
switch(msgType) {
case CAMERA_MSG_VIDEO_FRAME:
// should never happen
break;
// don't return raw data to Java
case CAMERA_MSG_RAW_IMAGE:
LOGV("rawCallback");
env->CallStaticVoidMethod(mCameraJClass, fields.post_event,
mCameraJObjectWeak, msgType, 0, 0, NULL);
break;
default:
// TODO: Change to LOGV
LOGV("dataCallback(%d, %p)", msgType, dataPtr.get());
copyAndPost(env, dataPtr, msgType);
break;
}
}
8. 可以看到JNI层最终都会调用来自java层的函数postEventFromNative，这个函数会发送对应的消息给自己的eventhandler，收到消息后就会根据消息的类型回调Package层Camera.java最初传下来的callback函数。至此，我们就在最上层拿到了图像数据。