audio flinger流程分析

ToneGenerator.cpp

ToneGenerator中构造函数或者startTone会调用initAudioTrack：

mpAudioTrack = new AudioTrack(); 

mpAudioTrack->set(..., audioCallback, ...); 

创建了AudioTrack，并将audioCallback传入AudioTrack。

audioCallback调用ToneGenerator::WaveGenerator类的getSamples来获取tone sample，填入AudioTrack传过来的AudioTrack::Buffer结构的中：

ToneGenerator *lpToneGen = static_cast<ToneGenerator *>(user); 

short *lpOut = buffer->i16; 

lpWaveGen->getSamples(lpOut, lGenSmp, lWaveCmd); 

AudioTrack.cpp

set函数检查有没有传入callback，若有，则创建AudioTrackThread线程：

if (cbf != 0) {         

  mAudioTrackThread = new AudioTrackThread(*this, threadCanCallJava); 

AudioTrackThread线程调用AudioTrack::processAudioBuffer()，调用callback来获取frame：

obtainBuffer(&audioBuffer, 1); 

mCbf(EVENT_MORE_DATA, mUserData, &audioBuffer); releaseBuffer(&audioBuffer); 

obtainBuffer填充audioBuffer各域，最重要的是

audioBuffer->raw = (int8_t *)cblk->buffer(u); 

cblk指向mCblk。mCblk在createTrack时赋值，实际为AudioFlinger通过IPC传递的内存空间：

sp<IAudioTrack> track = audioFlinger->createTrack() 

sp<IMemory> cblk = track->getCblk(); 

mCblk = static_cast<audio_track_cblk_t*>(cblk->pointer()); 

AudioFinger.cpp

AudioFlinger::DirectOutputThread的函数threadLoop：

activeTrack->getNextBuffer(&buffer); 

mOutput->write(mMixBuffer, mixBufferSize); 

getNextBuffer从cblk中取得帧：

audio_track_cblk_t* cblk = this->cblk(); 

buffer->raw = getBuffer(s, framesReq); 

对于MixerThread，情况会复杂一些。创建MixerThread时，创建了AudioMixer:

mAudioMixer = new AudioMixer(mFrameCount, mSampleRate); 

MixerThread的threadLoop：

mAudioMixer->process(curBuf); 

mOutput->write(curBuf, mixBufferSize); 

进入AudioMixer::process

mState.hook(&mState, output); 

hook为AudioMixer被enable时设定，以process__nop为例：

t.bufferProvider->getNextBuffer(&t.buffer); 

t.bufferProvider->releaseBuffer(&t.buffer); 

hook调用AudioFinger的getNextBuffer来获取buffer

总体来说，应用程序使用tone generator来发送tone音，tone generator与AudioTrack打交道。AudioTrack将audio frame放入audio flinger通过IPC传递过来的一块内存中。AudioFlinger作为消费者从这一块内存中读取audio frame。