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:
1 | public final void takePicture(ShutterCallback shutter, PictureCallback raw, |
2 | PictureCallback postview, PictureCallback jpeg) { |
3 | mShutterCallback = shutter; |
4 | mRawImageCallback = raw; |
5 | mPostviewCallback = postview; |
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。
1 | class ICameraClient: public IInterface |
4 | DECLARE_META_INTERFACE(CameraClient); |
6 | virtual void notifyCallback(int32_t msgType, int32_t ext1, int32_t ext2) = 0; |
7 | virtual void dataCallback(int32_t msgType, const sp& data) = 0; |
8 | virtual void dataCallbackTimestamp(nsecs_t timestamp, int32_t msgType, const sp& 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:
01 | void CameraService::Client::dataCallback(int32_t msgType, |
02 | const sp& dataPtr, void* user) { |
03 | LOG2("dataCallback(%d)", msgType); |
05 | sp client = getClientFromCookie(user); |
06 | if (client == 0) return; |
07 | if (!client->lockIfMessageWanted(msgType)) return; |
10 | LOGE("Null data returned in data callback"); |
11 | client->handleGenericNotify(CAMERA_MSG_ERROR, UNKNOWN_ERROR, 0); |
16 | case CAMERA_MSG_PREVIEW_FRAME: |
17 | client->handlePreviewData(dataPtr); |
19 | case CAMERA_MSG_POSTVIEW_FRAME: |
20 | client->handlePostview(dataPtr); |
22 | case CAMERA_MSG_RAW_IMAGE: |
23 | client->handleRawPicture(dataPtr); |
25 | case CAMERA_MSG_COMPRESSED_IMAGE: |
26 | client->handleCompressedPicture(dataPtr); |
29 | client->handleGenericData(msgType, dataPtr); |
33 | // picture callback - compressed picture ready |
34 | void CameraService::Client::handleCompressedPicture(const sp& mem) { |
35 | int restPictures = mHardware->getPictureRestCount(); |
38 | disableMsgType(CAMERA_MSG_COMPRESSED_IMAGE); |
44 | c->dataCallback(CAMERA_MSG_COMPRESSED_IMAGE, mem); |
6. Camera.cpp会继续通知它的listener:
01 | // callback from camera service when frame or image is ready |
02 | void Camera::dataCallback(int32_t msgType, const sp& dataPtr) |
06 | Mutex::Autolock _l(mLock); |
09 | if (listener != NULL) { |
10 | listener->postData(msgType, dataPtr); |
7. 而这个listener就是我们的JNI层的JNICameraContext对象了:
01 | void JNICameraContext::postData(int32_t msgType, const sp& dataPtr) |
03 | // VM pointer will be NULL if object is released |
04 | Mutex::Autolock _l(mLock); |
05 | JNIEnv *env = AndroidRuntime::getJNIEnv(); |
06 | if (mCameraJObjectWeak == NULL) { |
07 | LOGW("callback on dead camera object"); |
11 | // return data based on callback type |
13 | case CAMERA_MSG_VIDEO_FRAME: |
14 | // should never happen |
16 | // don't return raw data to Java |
17 | case CAMERA_MSG_RAW_IMAGE: |
19 | env->CallStaticVoidMethod(mCameraJClass, fields.post_event, |
20 | mCameraJObjectWeak, msgType, 0, 0, NULL); |
23 | // TODO: Change to LOGV |
24 | LOGV("dataCallback(%d, %p)", msgType, dataPtr.get()); |
25 | copyAndPost(env, dataPtr, msgType); |
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。
阅读(3019) | 评论(0) | 转发(1) |