理解 Audio 音频系统三 之 [2] AudioFlinger --- AudioThread
- 三、AudioFlinger
- 6. AudioFlinger 目录介绍
- 7. Audio Thread 音效线程管理分析
- 7.1 Threads.h 代码分析
- 7.1.1 定义一系列类Class ThreadBase 的内部类
- 7.1.2 类Class Threadbase 方法 public 成员
- 7.1.3 类Class Threadbase 方法 protected 成员
- 7.1.4 内部类Class ActiveTracks
- 7.1.5 定义类Class VolumeInterface 类 用于音量相关操作方法
- 7.1.6 定义类Class PlaybackThread (继承自 ThreadBase, VolumeInterface ) 用于对播放线程的管理
- 7.1.7 定义类Class MixerThread (继承自 PlaybackThread ) 用于 Mixer
- 7.1.8 定义类Class DirectOutputThread (继承自 PlaybackThread ) 用于播放不需混音的音乐
- 7.1.9 定义类Class OffloadThread (继承自 DirectOutputThread ) 用于播放不需混音的音乐
- 7.1.10 定义类Class AsyncCallbackThread (继承自 Thread ) 用于回调相关
- 7.1.11 定义类Class DuplicatingThread (继承自 MixerThread ) 复制线程
- 7.1.12 定义类Class RecordThread (继承自 ThreadBase ) 录音线程
- 7.1.13 定义类Class MmapThread (继承自 ThreadBase ) 录音线程
- 7.1.14 定义类Class MmapPlaybackThread (继承自 MmapThread，VolumeInterface )
- 7.1.15 定义类Class MmapCaptureThread (继承自 MmapThread )
- 7.1.16 ThreadBase 及其内部类关系总结
- 7.2 Threads.cpp 具体实现分析
- 8. Audio EffectModule 音效模块分析
本文接着前面的《理解 Audio 音频系统三 之 [1] AudioFlinger 启动流程 及 Audio PatchPanel初始化》 继续写。
在前面的文章中，我们主要是根据AudioFlinger 启动流程，重点分析了下 Audio PatchPanel 的初始化，接下来，我们主要介绍下 AudioFlinger 中的其他对象，比如，AudioTrack 和 AudioThread 等 。
三、AudioFlinger
6. AudioFlinger 目录介绍
在 AudioFlinger 目录下：
@ \frameworks\av\services\audioflinger
Android.mk
AudioFlinger.cpp / AudioFlinger.h // AudioFlinger 主代码，Class AudioFlinger 实现，Class SyncEvent 实现，
AudioHwDevice.cpp / AudioHwDevice.h // Class AudioHwDevice 实现
AudioStreamOut.cpp / AudioStreamOut.h // Class AudioStreamOut 实现
AudioWatchdog.cpp / AudioWatchdog.h // 监控CPU 及 记录其相关信息
AutoPark.h
BufLog.cpp / BufLog.h // Audio log 相关实现，log位于/data/misc/audioserver
Configuration.h // 对整个 AudioFlinger 功能宏控的单独配置
Effects.cpp / Effects.h // Class EffectModule 、 EffectHandle 、 EffectChain 类实现
FastCapture.cpp / FastCapture.h // Class FastCapture 类实现
FastCaptureDumpState.cpp / FastCaptureDumpState.h // 打印 FastCapture 的相关信息
FastCaptureState.cpp / FastCaptureState.h
FastMixer.cpp / FastMixer.h // Class FastMixer 类实现
FastMixerDumpState.cpp / FastMixerDumpState.h
FastMixerState.cpp / FastMixerState.h
FastThread.cpp / FastThread.h // class FastThread 是 FastMixer 和 FastCapture 的基类
FastThreadDumpState.cpp / FastThreadDumpState.h
FastThreadState.cpp / FastThreadState.h // FastThread 状态管理
MmapTracks.h
PatchPanel.cpp / PatchPanel.h // Class PatchPanel 类实现
PlaybackTracks.h // Class Track, 及其子类 OutputTrack、PatchTrack 的类定义
RecordTracks.h // Class RecordTrack，及其子类 PatchRecord 的类定义
ServiceUtilities.cpp / ServiceUtilities.h // Services 权限检查
SpdifStreamOut.cpp / SpdifStreamOut.h // Class SpdifStreamOut 实现：对HAL层的PCM数据进行封装
StateQueue.cpp / StateQueue.h
StateQueueInstantiations.cpp
Threads.cpp / Threads.h
// class ThreadBase 类
// 子类 Class PlaybackThread, MixerThread, DirectOutputThread, OffloadThread, AsyncCallbackThread
// 子类 Class RecordThread, MmapThread , MmapPlaybackThread , MmapCaptureThread 等类定义
TrackBase.h // Class TrackBase 定义， 及 Class PatchProxyBufferProvider 定义
Tracks.cpp // Class TrackBase 的方法实现，及其他 AudioFlinger 内 Track 相关实现
TypedLogger.cpp
TypedLogger.h
从上面的目录，可以看出，除了前面写的 Class PatchPanel 外，
AudioFlinger 中还包含了如下几个大模块：
- Audio Device 音频设备管理
- Audio EffectModule 音效模块
- Audio Thread 音频线程管理
- Audio Track 音频流管理
- Audio PatchPanel 音频路径管理
接下来，我们对其先各个突破，一个个地来详细学习下：
7. Audio Thread 音效线程管理分析
AudioThread 是AudioFlinger 中的一重要模块。
我们先来看下它的头文件。
7.1 Threads.h 代码分析
可以看出，Threads.h 是只为 AudioFlinger 服务的。
其主要工作如下：
-
定义Class ThreadBase，共继承自 Thread ，拥有 Thread 线程的一切方法
Thread 源码位于 #include <utils/threads.h> 中 -
定义了 Audio 中的六种线程，分别用于不同的音频场景
MIXER：用于混音输出音频的线程
DIRECT：用于直接输出音频的线程 (不需要混音)
DUPLICATING：用于拷贝的线程
RECORD：用于录音的线程
OFFLOAD：用于硬解码的线程（需有硬解码硬件支持）
MMAP：用于MMAP 的线程 -
定义录音 和 播放时的事件类型
CFG_EVENT_IO, // IO 访问操作
CFG_EVENT_PRIO, // 申请优先级
CFG_EVENT_SET_PARAMETER, // 配置参数
CFG_EVENT_CREATE_AUDIO_PATCH, // 创建 Audio patch
CFG_EVENT_RELEASE_AUDIO_PATCH, // 释放 Audio patch
以上代码如下：
@ \src\frameworks\av\services\audioflinger\Threads.h
#ifndef INCLUDING_FROM_AUDIOFLINGER_H
#error This header file should only be included from AudioFlinger.h
#endif
// ThreadBase 继承自 Thread
class ThreadBase : public Thread {
public:
#include "TrackBase.h"
// 定义了 Audio 中的六种线程
enum type_t {
MIXER, // Thread class is MixerThread 用于混音输出音频的线程
DIRECT, // Thread class is DirectOutputThread 用于直接输出音频的线程 (不需要混音)
DUPLICATING, // Thread class is DuplicatingThread 用于拷贝的线程
RECORD, // Thread class is RecordThread 用于录音的线程
OFFLOAD, // Thread class is OffloadThread 用于硬解码的线程（需有硬解码硬件支持）
MMAP // control thread for MMAP stream 用于MMAP 的线程
// If you add any values here, also update ThreadBase::threadTypeToString()
};
// 获取线程类型的字符串描述
static const char *threadTypeToString(type_t type);
// 构造函数
ThreadBase(const sp<AudioFlinger>& audioFlinger, audio_io_handle_t id,
audio_devices_t outDevice, audio_devices_t inDevice, type_t type, bool systemReady);
virtual ~ThreadBase(); // 析构函数
// 检查当前线程 是否可以运行
virtual status_t readyToRun();
// 打印当前线程 及 对应 Audio 的所有信息
void dumpBase(int fd, const Vector<String16>& args);
// 打印当前线程的 音效链表信息
void dumpEffectChains(int fd, const Vector<String16>& args);
void clearPowerManager();
// 定义 录音 和 播放时的事件类型
// base for record and playback
enum {
CFG_EVENT_IO, // IO 访问操作
CFG_EVENT_PRIO, // 申请优先级
CFG_EVENT_SET_PARAMETER, // 配置参数
CFG_EVENT_CREATE_AUDIO_PATCH, // 创建 Audio patch
CFG_EVENT_RELEASE_AUDIO_PATCH, // 释放 Audio patch
};
7.1.1 定义一系列类Class ThreadBase 的内部类
- ConfigEventData ： 用于打印当前 Event 事件的数据
- ConfigEvent ： 基类，用于 Event 事件的管理
对前边的 六种 Event 事件分别定义了其 ConfigEventData 和 ConfigEvent 内部类。
-
IoConfigEventData ：继承自ConfigEventData ， 用于 IO 事件
-
IoConfigEvent ：继承自 ConfigEvent，IO事件管理
-
PrioConfigEventData ：继承自ConfigEventData ，申请优先级
-
PrioConfigEvent ：继承自 ConfigEvent
-
SetParameterConfigEventData ：继承自ConfigEventData ，配置参数
-
SetParameterConfigEvent ：继承自 ConfigEvent
-
CreateAudioPatchConfigEventData ：继承自ConfigEventData ，创建 Audio Patch
-
CreateAudioPatchConfigEvent ：继承自 ConfigEvent
-
ReleaseAudioPatchConfigEventData ：继承自ConfigEventData ，释放Audio Patch
-
ReleaseAudioPatchConfigEvent ：继承自 ConfigEvent
@ \src\frameworks\av\services\audioflinger\Threads.h
class ThreadBase : public Thread {
public:
// 用于打印当前 Event 事件的数据
class ConfigEventData: public RefBase {
public:
virtual ~ConfigEventData() {}
virtual void dump(char *buffer, size_t size) = 0;
protected:
ConfigEventData() {}
};
// 用于 Event 事件的管理
class ConfigEvent: public RefBase {
public:
virtual ~ConfigEvent() {}
void dump(char *buffer, size_t size) { mData->dump(buffer, size); }
const int mType; // event type e.g. CFG_EVENT_IO
Mutex mLock; // mutex associated with mCond
Condition mCond; // condition for status return
status_t mStatus; // status communicated to sender
bool mWaitStatus; // true if sender is waiting for status
bool mRequiresSystemReady; // true if must wait for system ready to enter event queue
sp<ConfigEventData> mData; // event specific parameter data
protected:
explicit ConfigEvent(int type, bool requiresSystemReady = false) :
mType(type), mStatus(NO_ERROR), mWaitStatus(false),
mRequiresSystemReady(requiresSystemReady), mData(NULL) {}
};
// 用于 IO Event 事件管理
class IoConfigEventData : public ConfigEventData {
public:
IoConfigEventData(audio_io_config_event event, pid_t pid) :
mEvent(event), mPid(pid) {}
virtual void dump(char *buffer, size_t size) {
snprintf(buffer, size, "IO event: event %d\n", mEvent);
}
const audio_io_config_event mEvent;
const pid_t mPid;
};
class IoConfigEvent : public ConfigEvent {
public:
IoConfigEvent(audio_io_config_event event, pid_t pid) :
ConfigEvent(CFG_EVENT_IO) {
mData = new IoConfigEventData(event, pid);
}
virtual ~IoConfigEvent() {}
};
class PrioConfigEventData : public ConfigEventData {
public:
PrioConfigEventData(pid_t pid, pid_t tid, int32_t prio, bool forApp) :
mPid(pid), mTid(tid), mPrio(prio), mForApp(forApp) {}
virtual void dump(char *buffer, size_t size) {
snprintf(buffer, size, "Prio event: pid %d, tid %d, prio %d, for app? %d\n",
mPid, mTid, mPrio, mForApp);
}
const pid_t mPid;
const pid_t mTid;
const int32_t mPrio;
const bool mForApp;
};
class PrioConfigEvent : public ConfigEvent {
public:
PrioConfigEvent(pid_t pid, pid_t tid, int32_t prio, bool forApp) :
ConfigEvent(CFG_EVENT_PRIO, true) {
mData = new PrioConfigEventData(pid, tid, prio, forApp);
}
virtual ~PrioConfigEvent() {}
};
class SetParameterConfigEventData : public ConfigEventData {
public:
explicit SetParameterConfigEventData(String8 keyValuePairs) :
mKeyValuePairs(keyValuePairs) {}
virtual void dump(char *buffer, size_t size) {
snprintf(buffer, size, "KeyValue: %s\n", mKeyValuePairs.string());
}
const String8 mKeyValuePairs;
};
class SetParameterConfigEvent : public ConfigEvent {
public:
explicit SetParameterConfigEvent(String8 keyValuePairs) :
ConfigEvent(CFG_EVENT_SET_PARAMETER) {
mData = new SetParameterConfigEventData(keyValuePairs);
mWaitStatus = true;
}
virtual ~SetParameterConfigEvent() {}
};
class CreateAudioPatchConfigEventData : public ConfigEventData {
public:
CreateAudioPatchConfigEventData(const struct audio_patch patch,
audio_patch_handle_t handle) :
mPatch(patch), mHandle(handle) {}
virtual void dump(char *buffer, size_t size) {
snprintf(buffer, size, "Patch handle: %u\n", mHandle);
}
const struct audio_patch mPatch;
audio_patch_handle_t mHandle;
};
class CreateAudioPatchConfigEvent : public ConfigEvent {
public:
CreateAudioPatchConfigEvent(const struct audio_patch patch,
audio_patch_handle_t handle) :
ConfigEvent(CFG_EVENT_CREATE_AUDIO_PATCH) {
mData = new CreateAudioPatchConfigEventData(patch, handle);
mWaitStatus = true;
}
virtual ~CreateAudioPatchConfigEvent() {}
};
class ReleaseAudioPatchConfigEventData : public ConfigEventData {
public:
explicit ReleaseAudioPatchConfigEventData(const audio_patch_handle_t handle) :
mHandle(handle) {}
virtual void dump(char *buffer, size_t size) {
snprintf(buffer, size, "Patch handle: %u\n", mHandle);
}
audio_patch_handle_t mHandle;
};
class ReleaseAudioPatchConfigEvent : public ConfigEvent {
public:
explicit ReleaseAudioPatchConfigEvent(const audio_patch_handle_t handle) :
ConfigEvent(CFG_EVENT_RELEASE_AUDIO_PATCH) {
mData = new ReleaseAudioPatchConfigEventData(handle);
mWaitStatus = true;
}
virtual ~ReleaseAudioPatchConfigEvent() {}
};
}
7.1.2 类Class Threadbase 方法 public 成员
在 public 成员中包括：
- 参数获取方法
- 事件处理函数
- 创建 AudioTrack 方法
- 音效相关接口
@ \src\frameworks\av\services\audioflinger\Threads.h
class ThreadBase : public Thread {
public:
virtual status_t initCheck() const = 0; //检查是否已经初始化，判断 mOutput == NULL
// static externally-visible
type_t type() const { return mType; }
bool isDuplicating() const { return (mType == DUPLICATING); }
audio_io_handle_t id() const { return mId;}
// 参数获取方法
// dynamic externally-visible
uint32_t sampleRate() const { return mSampleRate; } // 返回采样速率
audio_channel_mask_t channelMask() const { return mChannelMask; } // 反回通道掩码
audio_format_t format() const { return mHALFormat; } // 返回格式
uint32_t channelCount() const { return mChannelCount; }
// Called by AudioFlinger::frameCount(audio_io_handle_t output) and effects,
// and returns the [normal mix] buffer's frame count.
virtual size_t frameCount() const = 0;
// Return's the HAL's frame count i.e. fast mixer buffer size.
size_t frameCountHAL() const { return mFrameCount; }
size_t frameSize() const { return mFrameSize; }
// Should be "virtual status_t requestExitAndWait()" and override same
// method in Thread, but Thread::requestExitAndWait() is not yet virtual.
void exit();
virtual bool checkForNewParameter_l(const String8& keyValuePair,
status_t& status) = 0;
virtual status_t setParameters(const String8& keyValuePairs);
virtual String8 getParameters(const String8& keys) = 0;
virtual void ioConfigChanged(audio_io_config_event event, pid_t pid = 0) = 0;
// 事件处理函数
// sendConfigEvent_l() must be called with ThreadBase::mLock held
// Can temporarily release the lock if waiting for a reply from
// processConfigEvents_l().
status_t sendConfigEvent_l(sp<ConfigEvent>& event);
void sendIoConfigEvent(audio_io_config_event event, pid_t pid = 0);
void sendIoConfigEvent_l(audio_io_config_event event, pid_t pid = 0);
void sendPrioConfigEvent(pid_t pid, pid_t tid, int32_t prio, bool forApp);
void sendPrioConfigEvent_l(pid_t pid, pid_t tid, int32_t prio, bool forApp);
status_t sendSetParameterConfigEvent_l(const String8& keyValuePair);
status_t sendCreateAudioPatchConfigEvent(const struct audio_patch *patch, audio_patch_handle_t *handle);
status_t sendReleaseAudioPatchConfigEvent(audio_patch_handle_t handle);
void processConfigEvents_l();
// 创建 AudioTrack 方法
virtual void cacheParameters_l() = 0;
virtual status_t createAudioPatch_l(const struct audio_patch *patch,audio_patch_handle_t *handle) = 0;
virtual status_t releaseAudioPatch_l(const audio_patch_handle_t handle) = 0;
virtual void getAudioPortConfig(struct audio_port_config *config) = 0;
// see note at declaration of mStandby, mOutDevice and mInDevice
bool standby() const { return mStandby; }
audio_devices_t outDevice() const { return mOutDevice; }
audio_devices_t inDevice() const { return mInDevice; }
audio_devices_t getDevice() const { return isOutput() ? mOutDevice : mInDevice; }
virtual bool isOutput() const = 0;
virtual sp<StreamHalInterface> stream() const = 0;
// 音效相关接口
sp<EffectHandle> createEffect_l( const sp<AudioFlinger::Client>& client,
const sp<IEffectClient>& effectClient, int32_t priority,
audio_session_t sessionId, effect_descriptor_t *desc, int *enabled,
status_t *status /*non-NULL*/, bool pinned);
// return values for hasAudioSession (bit field)
enum effect_state {
EFFECT_SESSION = 0x1, // the audio session corresponds to at least one
// effect
TRACK_SESSION = 0x2, // the audio session corresponds to at least one
// track
FAST_SESSION = 0x4 // the audio session corresponds to at least one
// fast track
};
// get effect chain corresponding to session Id.
sp<EffectChain> getEffectChain(audio_session_t sessionId);
// same as getEffectChain() but must be called with ThreadBase mutex locked
sp<EffectChain> getEffectChain_l(audio_session_t sessionId) const;
// add an effect chain to the chain list (mEffectChains)
virtual status_t addEffectChain_l(const sp<EffectChain>& chain) = 0;
// remove an effect chain from the chain list (mEffectChains)
virtual size_t removeEffectChain_l(const sp<EffectChain>& chain) = 0;
// lock all effect chains Mutexes. Must be called before releasing the
// ThreadBase mutex before processing the mixer and effects. This guarantees the
// integrity of the chains during the process.
// Also sets the parameter 'effectChains' to current value of mEffectChains.
void lockEffectChains_l(Vector< sp<EffectChain> >& effectChains);
// unlock effect chains after process
void unlockEffectChains(const Vector< sp<EffectChain> >& effectChains);
// get a copy of mEffectChains vector
Vector< sp<EffectChain> > getEffectChains_l() const { return mEffectChains; };
// set audio mode to all effect chains
void setMode(audio_mode_t mode);
// get effect module with corresponding ID on specified audio session
sp<AudioFlinger::EffectModule> getEffect(audio_session_t sessionId, int effectId);
sp<AudioFlinger::EffectModule> getEffect_l(audio_session_t sessionId, int effectId);
// add and effect module. Also creates the effect chain is none exists for
// the effects audio session
status_t addEffect_l(const sp< EffectModule>& effect);
// remove and effect module. Also removes the effect chain is this was the last
// effect
void removeEffect_l(const sp< EffectModule>& effect, bool release = false);
// disconnect an effect handle from module and destroy module if last handle
void disconnectEffectHandle(EffectHandle *handle, bool unpinIfLast);
// detach all tracks connected to an auxiliary effect
virtual void detachAuxEffect_l(int effectId __unused) {}
// returns a combination of:
// - EFFECT_SESSION if effects on this audio session exist in one chain
// - TRACK_SESSION if tracks on this audio session exist
// - FAST_SESSION if fast tracks on this audio session exist
virtual uint32_t hasAudioSession_l(audio_session_t sessionId) const = 0;
uint32_t hasAudioSession(audio_session_t sessionId) const {
Mutex::Autolock _l(mLock);
return hasAudioSession_l(sessionId);
}
// the value returned by default implementation is not important as the
// strategy is only meaningful for PlaybackThread which implements this method
virtual uint32_t getStrategyForSession_l(audio_session_t sessionId __unused)
{ return 0; }
// check if some effects must be suspended/restored when an effect is enabled or disabled
void checkSuspendOnEffectEnabled(const sp<EffectModule>& effect, bool enabled,
audio_session_t sessionId = AUDIO_SESSION_OUTPUT_MIX);
void checkSuspendOnEffectEnabled_l(const sp<EffectModule>& effect, bool enabled,
audio_session_t sessionId = AUDIO_SESSION_OUTPUT_MIX);
virtual status_t setSyncEvent(const sp<SyncEvent>& event) = 0;
virtual bool isValidSyncEvent(const sp<SyncEvent>& event) const = 0;
// Return a reference to a per-thread heap which can be used to allocate IMemory
// objects that will be read-only to client processes, read/write to mediaserver,
// and shared by all client processes of the thread.
// The heap is per-thread rather than common across all threads, because
// clients can't be trusted not to modify the offset of the IMemory they receive.
// If a thread does not have such a heap, this method returns 0.
virtual sp<MemoryDealer> readOnlyHeap() const { return 0; }
virtual sp<IMemory> pipeMemory() const { return 0; }
void systemReady();
// checkEffectCompatibility_l() must be called with ThreadBase::mLock held
virtual status_t checkEffectCompatibility_l(const effect_descriptor_t *desc, audio_session_t sessionId) = 0;
void broadcast_l();
mutable Mutex mLock;
｝
7.1.3 类Class Threadbase 方法 protected 成员
包含了一系列的 Audio 参数
protected:
const sp<AudioFlinger> mAudioFlinger;
// updated by PlaybackThread::readOutputParameters_l() or RecordThread::readInputParameters_l()
uint32_t mSampleRate;
size_t mFrameCount; // output HAL, direct output, record
audio_channel_mask_t mChannelMask;
uint32_t mChannelCount;
size_t mFrameSize;
// not HAL frame size, this is for output sink (to pipe to fast mixer)
audio_format_t mFormat; // Source format for Recording and
// Sink format for Playback.
// Sink format may be different than
// HAL format if Fastmixer is used.
audio_format_t mHALFormat;
size_t mBufferSize; // HAL buffer size for read() or write()
Vector< sp<ConfigEvent> > mConfigEvents;
Vector< sp<ConfigEvent> > mPendingConfigEvents; // events awaiting system ready
// These fields are written and read by thread itself without lock or barrier,
// and read by other threads without lock or barrier via standby(), outDevice()
// and inDevice().
// Because of the absence of a lock or barrier, any other thread that reads
// these fields must use the information in isolation, or be prepared to deal
// with possibility that it might be inconsistent with other information.
bool mStandby; // Whether thread is currently in standby.
audio_devices_t mOutDevice; // output device
audio_devices_t mInDevice; // input device
audio_devices_t mPrevOutDevice; // previous output device
audio_devices_t mPrevInDevice; // previous input device
struct audio_patch mPatch;
audio_source_t mAudioSource;
const audio_io_handle_t mId;
Vector< sp<EffectChain> > mEffectChains;
static const int kThreadNameLength = 16; // prctl(PR_SET_NAME) limit
char mThreadName[kThreadNameLength]; // guaranteed NUL-terminated
sp<IPowerManager> mPowerManager;
sp<IBinder> mWakeLockToken;
const sp<PMDeathRecipient> mDeathRecipient;
7.1.4 内部类Class ActiveTracks
定义了一个内部类， 用于对 ActiveTracks 的管理
@ \src\frameworks\av\services\audioflinger\Threads.h
class ThreadBase : public Thread {
template <typename T>
class ActiveTracks {
public:
explicit ActiveTracks(SimpleLog *localLog = nullptr)
: mActiveTracksGeneration(0)
, mLastActiveTracksGeneration(0)
, mLocalLog(localLog)
{ }
~ActiveTracks() {
ALOGW_IF(!mActiveTracks.isEmpty(),
"ActiveTracks should be empty in destructor");
}
// returns the last track added (even though it may have been
// subsequently removed from ActiveTracks).
//
// Used for DirectOutputThread to ensure a flush is called when transitioning
// to a new track (even though it may be on the same session).
// Used for OffloadThread to ensure that volume and mixer state is
// taken from the latest track added.
//
// The latest track is saved with a weak pointer to prevent keeping an
// otherwise useless track alive. Thus the function will return nullptr
// if the latest track has subsequently been removed and destroyed.
sp<T> getLatest() {
return mLatestActiveTrack.promote();
}
// SortedVector methods
ssize_t add(const sp<T> &track);
ssize_t remove(const sp<T> &track);
size_t size() const {
return mActiveTracks.size();
}
ssize_t indexOf(const sp<T>& item) {
return mActiveTracks.indexOf(item);
}
sp<T> operator[](size_t index) const {
return mActiveTracks[index];
}
typename SortedVector<sp<T>>::iterator begin() {
return mActiveTracks.begin();
}
typename SortedVector<sp<T>>::iterator end() {
return mActiveTracks.end();
}
// Due to Binder recursion optimization, clear() and updatePowerState()
// cannot be called from a Binder thread because they may call back into
// the original calling process (system server) for BatteryNotifier
// (which requires a Java environment that may not be present).
// Hence, call clear() and updatePowerState() only from the
// ThreadBase thread.
void clear();
// periodically called in the threadLoop() to update power state uids.
void updatePowerState(sp<ThreadBase> thread, bool force = false);
private:
void logTrack(const char *funcName, const sp<T> &track) const;
SortedVector<uid_t> getWakeLockUids() {
SortedVector<uid_t> wakeLockUids;
for (const sp<T> &track : mActiveTracks) {
wakeLockUids.add(track->uid());
}
return wakeLockUids; // moved by underlying SharedBuffer
}
std::map<uid_t, std::pair<ssize_t /* previous */, ssize_t /* current */>>
mBatteryCounter;
SortedVector<sp<T>> mActiveTracks;
int mActiveTracksGeneration;
int mLastActiveTracksGeneration;
wp<T> mLatestActiveTrack; // latest track added to ActiveTracks
SimpleLog * const mLocalLog;
};
｝
7.1.5 定义类Class VolumeInterface 类 用于音量相关操作方法
在 Threads.h 文件中，同样也定义了一个 VolumeInterface 结构体，用于控制音量。
@ \src\frameworks\av\services\audioflinger\Threads.h
class VolumeInterface {
public:
virtual ~VolumeInterface() {}
virtual void setMasterVolume(float value) = 0;
virtual void setMasterMute(bool muted) = 0;
virtual void setStreamVolume(audio_stream_type_t stream, float value) = 0;
virtual void setStreamMute(audio_stream_type_t stream, bool muted) = 0;
virtual float streamVolume(audio_stream_type_t stream) const = 0;
};
7.1.6 定义类Class PlaybackThread (继承自 ThreadBase, VolumeInterface ) 用于对播放线程的管理
@ \src\frameworks\av\services\audioflinger\Threads.h
// --- PlaybackThread ---
class PlaybackThread : public ThreadBase, public StreamOutHalInterfaceCallback, public VolumeInterface
{
public:
#include "PlaybackTracks.h"
// 定义了 Mixer 的几种状态
enum mixer_state {
MIXER_IDLE, // no active tracks
MIXER_TRACKS_ENABLED, // at least one active track, but no track has any data ready
MIXER_TRACKS_READY, // at least one active track, and at least one track has data
MIXER_DRAIN_TRACK, // drain currently playing track
MIXER_DRAIN_ALL, // fully drain the hardware
// standby mode does not have an enum value
// suspend by audio policy manager is orthogonal to mixer state
};
// retry count before removing active track in case of underrun on offloaded thread:
// we need to make sure that AudioTrack client has enough time to send large buffers
static const int8_t kMaxTrackRetriesOffload = 20;
static const int8_t kMaxTrackStartupRetriesOffload = 100;
static const int8_t kMaxTrackStopRetriesOffload = 2;
// 每个 UID client 最大可以支持 14 个线程
// 14 tracks max per client allows for 2 misbehaving application leaving 4 available tracks.
static const uint32_t kMaxTracksPerUid = 14;
// Maximum delay (in nanoseconds) for upcoming buffers in suspend mode, otherwise
// if delay is greater, the estimated time for timeLoopNextNs is reset.
// This allows for catch-up to be done for small delays, while resetting the estimate
// for initial conditions or large delays.
static const nsecs_t kMaxNextBufferDelayNs = 100000000;
PlaybackThread(const sp<AudioFlinger>& audioFlinger, AudioStreamOut* output, audio_io_handle_t id, audio_devices_t device, type_t type, bool systemReady);
virtual ~PlaybackThread();
void dump(int fd, const Vector<String16>& args);
// Thread virtuals
virtual bool threadLoop();
// RefBase
virtual void onFirstRef();
virtual status_t checkEffectCompatibility_l(const effect_descriptor_t *desc, audio_session_t sessionId);
protected:
// Code snippets that were lifted up out of threadLoop()
virtual void threadLoop_mix() = 0;
virtual void threadLoop_sleepTime() = 0;
virtual ssize_t threadLoop_write();
virtual void threadLoop_drain();
virtual void threadLoop_standby();
virtual void threadLoop_exit();
virtual void threadLoop_removeTracks(const Vector< sp<Track> >& tracksToRemove);
// prepareTracks_l reads and writes mActiveTracks, and returns
// the pending set of tracks to remove via Vector 'tracksToRemove'. The caller
// is responsible for clearing or destroying this Vector later on, when it
// is safe to do so. That will drop the final ref count and destroy the tracks.
virtual mixer_state prepareTracks_l(Vector< sp<Track> > *tracksToRemove) = 0;
void removeTracks_l(const Vector< sp<Track> >& tracksToRemove);
// StreamOutHalInterfaceCallback implementation
virtual void onWriteReady();
virtual void onDrainReady();
virtual void onError();
void resetWriteBlocked(uint32_t sequence);
void resetDraining(uint32_t sequence);
virtual bool waitingAsyncCallback();
virtual bool waitingAsyncCallback_l();
virtual bool shouldStandby_l();
virtual void onAddNewTrack_l();
void onAsyncError(); // error reported by AsyncCallbackThread
// ThreadBase virtuals
virtual void preExit();
virtual void onIdleMixer();
virtual bool keepWakeLock() const { return true; }
virtual void acquireWakeLock_l() {
ThreadBase::acquireWakeLock_l();
mActiveTracks.updatePowerState(this, true /* force */); }
public:
virtual status_t initCheck() const { return (mOutput == NULL) ? NO_INIT : NO_ERROR; }
// return estimated latency in milliseconds, as reported by HAL
uint32_t latency() const;
// same, but lock must already be held
uint32_t latency_l() const;
// VolumeInterface
virtual void setMasterVolume(float value);
virtual void setMasterMute(bool muted);
virtual void setStreamVolume(audio_stream_type_t stream, float value);
virtual void setStreamMute(audio_stream_type_t stream, bool muted);
virtual float streamVolume(audio_stream_type_t stream) const;
sp<Track> createTrack_l(
const sp<AudioFlinger::Client>& client,
audio_stream_type_t streamType,
uint32_t sampleRate,
audio_format_t format,
audio_channel_mask_t channelMask,
size_t *pFrameCount,
const sp<IMemory>& sharedBuffer,
audio_session_t sessionId,
audio_output_flags_t *flags,
pid_t tid,
uid_t uid,
status_t *status /*non-NULL*/,
audio_port_handle_t portId);
AudioStreamOut* getOutput() const;
AudioStreamOut* clearOutput();
virtual sp<StreamHalInterface> stream() const;
// a very large number of suspend() will eventually wraparound, but unlikely
void suspend() { (void) android_atomic_inc(&mSuspended); }
void restore() {
// if restore() is done without suspend(), get back into
// range so that the next suspend() will operate correctly
if (android_atomic_dec(&mSuspended) <= 0) {
android_atomic_release_store(0, &mSuspended);
}
}
bool isSuspended() const { return android_atomic_acquire_load(&mSuspended) > 0; }
virtual String8 getParameters(const String8& keys);
virtual void ioConfigChanged(audio_io_config_event event, pid_t pid = 0);
status_t getRenderPosition(uint32_t *halFrames, uint32_t *dspFrames);
// FIXME rename mixBuffer() to sinkBuffer() and remove int16_t* dependency.
// Consider also removing and passing an explicit mMainBuffer initialization
// parameter to AF::PlaybackThread::Track::Track().
int16_t *mixBuffer() const { return reinterpret_cast<int16_t *>(mSinkBuffer); };
virtual void detachAuxEffect_l(int effectId);
status_t attachAuxEffect(const sp<AudioFlinger::PlaybackThread::Track>& track,int EffectId);
status_t attachAuxEffect_l(const sp<AudioFlinger::PlaybackThread::Track>& track, int EffectId);
virtual status_t addEffectChain_l(const sp<EffectChain>& chain);
virtual size_t removeEffectChain_l(const sp<EffectChain>& chain);
virtual uint32_t hasAudioSession_l(audio_session_t sessionId) const;
virtual uint32_t getStrategyForSession_l(audio_session_t sessionId);
virtual status_t setSyncEvent(const sp<SyncEvent>& event);
virtual bool isValidSyncEvent(const sp<SyncEvent>& event) const;
// called with AudioFlinger lock held
bool invalidateTracks_l(audio_stream_type_t streamType);
virtual void invalidateTracks(audio_stream_type_t streamType);
virtual size_t frameCount() const { return mNormalFrameCount; }
virtual status_t getTimestamp_l(AudioTimestamp& timestamp);
void addPatchTrack(const sp<PatchTrack>& track);
void deletePatchTrack(const sp<PatchTrack>& track);
virtual void getAudioPortConfig(struct audio_port_config *config);
// Return the asynchronous signal wait time.
virtual int64_t computeWaitTimeNs_l() const { return INT64_MAX; }
virtual bool isOutput() const override { return true; }
}
7.1.7 定义类Class MixerThread (继承自 PlaybackThread ) 用于 Mixer
MixerThread Class 继承自 PlaybackThread。
@ /frameworks/av/services/audioflinger/Threads.h
class MixerThread : public PlaybackThread {
public:
MixerThread(const sp<AudioFlinger>& audioFlinger, AudioStreamOut* output, audio_io_handle_t id,
audio_devices_t device, bool systemReady, type_t type = MIXER);
virtual ~MixerThread();
// Thread virtuals
virtual bool checkForNewParameter_l(const String8& keyValuePair, status_t& status);
virtual void dumpInternals(int fd, const Vector<String16>& args);
protected:
virtual mixer_state prepareTracks_l(Vector< sp<Track> > *tracksToRemove);
virtual int getTrackName_l(audio_channel_mask_t channelMask, audio_format_t format,
audio_session_t sessionId, uid_t uid);
virtual void deleteTrackName_l(int name);
virtual uint32_t idleSleepTimeUs() const;
virtual uint32_t suspendSleepTimeUs() const;
virtual void cacheParameters_l();
// threadLoop snippets
virtual ssize_t threadLoop_write();
virtual void threadLoop_standby();
virtual void threadLoop_mix();
virtual void threadLoop_sleepTime();
virtual void threadLoop_removeTracks(const Vector< sp<Track> >& tracksToRemove);
virtual void onIdleMixer();
virtual uint32_t correctLatency_l(uint32_t latency) const;
virtual status_t createAudioPatch_l(const struct audio_patch *patch, audio_patch_handle_t *handle);
virtual status_t releaseAudioPatch_l(const audio_patch_handle_t handle);
AudioMixer* mAudioMixer; // normal mixer
private:
// one-time initialization, no locks required
sp<FastMixer> mFastMixer; // non-0 if there is also a fast mixer
sp<AudioWatchdog> mAudioWatchdog; // non-0 if there is an audio watchdog thread
// contents are not guaranteed to be consistent, no locks required
FastMixerDumpState mFastMixerDumpState;
#ifdef STATE_QUEUE_DUMP
StateQueueObserverDump mStateQueueObserverDump;
StateQueueMutatorDump mStateQueueMutatorDump;
#endif
AudioWatchdogDump mAudioWatchdogDump;
};
7.1.8 定义类Class DirectOutputThread (继承自 PlaybackThread ) 用于播放不需混音的音乐
class DirectOutputThread : public PlaybackThread {
public:
DirectOutputThread(const sp<AudioFlinger>& audioFlinger, AudioStreamOut* output,
audio_io_handle_t id, audio_devices_t device, bool systemReady);
virtual ~DirectOutputThread();
// Thread virtuals
virtual bool checkForNewParameter_l(const String8& keyValuePair,
status_t& status);
virtual void flushHw_l();
protected:
virtual int getTrackName_l(audio_channel_mask_t channelMask, audio_format_t format,
audio_session_t sessionId, uid_t uid);
virtual void deleteTrackName_l(int name);
virtual uint32_t activeSleepTimeUs() const;
virtual uint32_t idleSleepTimeUs() const;
virtual uint32_t suspendSleepTimeUs() const;
virtual void cacheParameters_l();
// threadLoop snippets
virtual mixer_state prepareTracks_l(Vector< sp<Track> > *tracksToRemove);
virtual void threadLoop_mix();
virtual void threadLoop_sleepTime();
virtual void threadLoop_exit();
virtual bool shouldStandby_l();
virtual void onAddNewTrack_l();
bool mVolumeShaperActive;
DirectOutputThread(const sp<AudioFlinger>& audioFlinger, AudioStreamOut* output,
audio_io_handle_t id, uint32_t device, ThreadBase::type_t type,
bool systemReady);
void processVolume_l(Track *track, bool lastTrack);
// prepareTracks_l() tells threadLoop_mix() the name of the single active track
sp<Track> mActiveTrack;
wp<Track> mPreviousTrack; // used to detect track switch
uint64_t mFramesWrittenAtStandby;// used to reset frames on track reset
uint64_t mFramesWrittenForSleep; // used to reset frames on track removal
// or underrun before entering standby
public:
virtual bool hasFastMixer() const { return false; }
virtual int64_t computeWaitTimeNs_l() const override;
virtual status_t getTimestamp_l(AudioTimestamp& timestamp) override;
};
7.1.9 定义类Class OffloadThread (继承自 DirectOutputThread ) 用于播放不需混音的音乐
class OffloadThread : public DirectOutputThread {
public:
OffloadThread(const sp<AudioFlinger>& audioFlinger, AudioStreamOut* output,
audio_io_handle_t id, uint32_t device, bool systemReady);
virtual ~OffloadThread() {};
virtual void flushHw_l();
protected:
// threadLoop snippets
virtual mixer_state prepareTracks_l(Vector< sp<Track> > *tracksToRemove);
virtual void threadLoop_exit();
virtual bool waitingAsyncCallback();
virtual bool waitingAsyncCallback_l();
virtual void invalidateTracks(audio_stream_type_t streamType);
virtual bool keepWakeLock() const { return (mKeepWakeLock || (mDrainSequence & 1)); }
private:
size_t mPausedWriteLength; // length in bytes of write interrupted by pause
size_t mPausedBytesRemaining; // bytes still waiting in mixbuffer after resume
bool mKeepWakeLock; // keep wake lock while waiting for write callback
uint64_t mOffloadUnderrunPosition; // Current frame position for offloaded playback
// used and valid only during underrun. ~0 if
// no underrun has occurred during playback and
// is not reset on standby.
};
7.1.10 定义类Class AsyncCallbackThread (继承自 Thread ) 用于回调相关
class AsyncCallbackThread : public Thread {
public:
explicit AsyncCallbackThread(const wp<PlaybackThread>& playbackThread);
virtual ~AsyncCallbackThread();
// Thread virtuals
virtual bool threadLoop();
// RefBase
virtual void onFirstRef();
void exit();
void setWriteBlocked(uint32_t sequence);
void resetWriteBlocked();
void setDraining(uint32_t sequence);
void resetDraining();
void setAsyncError();
private:
const wp<PlaybackThread> mPlaybackThread;
// mWriteAckSequence corresponds to the last write sequence passed by the offload thread via
// setWriteBlocked(). The sequence is shifted one bit to the left and the lsb is used
// to indicate that the callback has been received via resetWriteBlocked()
uint32_t mWriteAckSequence;
// mDrainSequence corresponds to the last drain sequence passed by the offload thread via
// setDraining(). The sequence is shifted one bit to the left and the lsb is used
// to indicate that the callback has been received via resetDraining()
uint32_t mDrainSequence;
Condition mWaitWorkCV;
Mutex mLock;
bool mAsyncError;
};
7.1.11 定义类Class DuplicatingThread (继承自 MixerThread ) 复制线程
class DuplicatingThread : public MixerThread {
public:
DuplicatingThread(const sp<AudioFlinger>& audioFlinger, MixerThread* mainThread,
audio_io_handle_t id, bool systemReady);
virtual ~DuplicatingThread();
// Thread virtuals
void addOutputTrack(MixerThread* thread);
void removeOutputTrack(MixerThread* thread);
uint32_t waitTimeMs() const { return mWaitTimeMs; }
protected:
virtual uint32_t activeSleepTimeUs() const;
private:
bool outputsReady(const SortedVector< sp<OutputTrack> > &outputTracks);
protected:
// threadLoop snippets
virtual void threadLoop_mix();
virtual void threadLoop_sleepTime();
virtual ssize_t threadLoop_write();
virtual void threadLoop_standby();
virtual void cacheParameters_l();
private:
// called from threadLoop, addOutputTrack, removeOutputTrack
virtual void updateWaitTime_l();
protected:
virtual void saveOutputTracks();
virtual void clearOutputTracks();
private:
uint32_t mWaitTimeMs;
SortedVector < sp<OutputTrack> > outputTracks;
SortedVector < sp<OutputTrack> > mOutputTracks;
public:
virtual bool hasFastMixer() const { return false; }
};
7.1.12 定义类Class RecordThread (继承自 ThreadBase ) 录音线程
// record thread
class RecordThread : public ThreadBase
{
public:
class RecordTrack;
/* The ResamplerBufferProvider is used to retrieve recorded input data from the
* RecordThread. It maintains local state on the relative position of the read
* position of the RecordTrack compared with the RecordThread. */
class ResamplerBufferProvider : public AudioBufferProvider
{
public:
explicit ResamplerBufferProvider(RecordTrack* recordTrack) :
mRecordTrack(recordTrack), mRsmpInUnrel(0), mRsmpInFront(0) { }
virtual ~ResamplerBufferProvider() { }
// called to set the ResamplerBufferProvider to head of the RecordThread data buffer,
// skipping any previous data read from the hal.
virtual void reset();
/* Synchronizes RecordTrack position with the RecordThread.
* Calculates available frames and handle overruns if the RecordThread
* has advanced faster than the ResamplerBufferProvider has retrieved data.
* TODO: why not do this for every getNextBuffer?
*
* Parameters
* framesAvailable: pointer to optional output size_t to store record track
* frames available.
* hasOverrun: pointer to optional boolean, returns true if track has overrun.
*/
virtual void sync(size_t *framesAvailable = NULL, bool *hasOverrun = NULL);
// AudioBufferProvider interface
virtual status_t getNextBuffer(AudioBufferProvider::Buffer* buffer);
virtual void releaseBuffer(AudioBufferProvider::Buffer* buffer);
private:
RecordTrack * const mRecordTrack;
size_t mRsmpInUnrel; // unreleased frames remaining from
// most recent getNextBuffer
// for debug only
int32_t mRsmpInFront; // next available frame
// rolling counter that is never cleared
};
#include "RecordTracks.h"
RecordThread(const sp<AudioFlinger>& audioFlinger,
AudioStreamIn *input,
audio_io_handle_t id,
audio_devices_t outDevice,
audio_devices_t inDevice,
bool systemReady
#ifdef TEE_SINK
, const sp<NBAIO_Sink>& teeSink
#endif
);
virtual ~RecordThread();
// no addTrack_l ?
void destroyTrack_l(const sp<RecordTrack>& track);
void removeTrack_l(const sp<RecordTrack>& track);
void dumpInternals(int fd, const Vector<String16>& args);
void dumpTracks(int fd, const Vector<String16>& args);
// Thread virtuals
virtual bool threadLoop();
virtual void preExit();
// RefBase
virtual void onFirstRef();
virtual status_t initCheck() const { return (mInput == NULL) ? NO_INIT : NO_ERROR; }
virtual sp<MemoryDealer> readOnlyHeap() const { return mReadOnlyHeap; }
virtual sp<IMemory> pipeMemory() const { return mPipeMemory; }
sp<AudioFlinger::RecordThread::RecordTrack> createRecordTrack_l(
const sp<AudioFlinger::Client>& client,
uint32_t sampleRate,
audio_format_t format,
audio_channel_mask_t channelMask,
size_t *pFrameCount,
audio_session_t sessionId,
size_t *notificationFrames,
uid_t uid,
audio_input_flags_t *flags,
pid_t tid,
status_t *status /*non-NULL*/,
audio_port_handle_t portId);
status_t start(RecordTrack* recordTrack,
AudioSystem::sync_event_t event,
audio_session_t triggerSession);
// ask the thread to stop the specified track, and
// return true if the caller should then do it's part of the stopping process
bool stop(RecordTrack* recordTrack);
void dump(int fd, const Vector<String16>& args);
AudioStreamIn* clearInput();
virtual sp<StreamHalInterface> stream() const;
virtual bool checkForNewParameter_l(const String8& keyValuePair,status_t& status);
virtual void cacheParameters_l() {}
virtual String8 getParameters(const String8& keys);
virtual void ioConfigChanged(audio_io_config_event event, pid_t pid = 0);
virtual status_t createAudioPatch_l(const struct audio_patch *patch,audio_patch_handle_t *handle);
virtual status_t releaseAudioPatch_l(const audio_patch_handle_t handle);
void addPatchRecord(const sp<PatchRecord>& record);
void deletePatchRecord(const sp<PatchRecord>& record);
void readInputParameters_l();
virtual uint32_t getInputFramesLost();
virtual status_t addEffectChain_l(const sp<EffectChain>& chain);
virtual size_t removeEffectChain_l(const sp<EffectChain>& chain);
virtual uint32_t hasAudioSession_l(audio_session_t sessionId) const;
// Return the set of unique session IDs across all tracks.
// The keys are the session IDs, and the associated values are meaningless.
// FIXME replace by Set [and implement Bag/Multiset for other uses].
KeyedVector<audio_session_t, bool> sessionIds() const;
virtual status_t setSyncEvent(const sp<SyncEvent>& event);
virtual bool isValidSyncEvent(const sp<SyncEvent>& event) const;
static void syncStartEventCallback(const wp<SyncEvent>& event);
virtual size_t frameCount() const { return mFrameCount; }
bool hasFastCapture() const { return mFastCapture != 0; }
virtual void getAudioPortConfig(struct audio_port_config *config);
virtual status_t checkEffectCompatibility_l(const effect_descriptor_t *desc, audio_session_t sessionId);
virtual void acquireWakeLock_l() {
ThreadBase::acquireWakeLock_l();
mActiveTracks.updatePowerState(this, true /* force */); }
virtual bool isOutput() const override { return false; }
void checkBtNrec();
private:
// Enter standby if not already in standby, and set mStandby flag
void standbyIfNotAlreadyInStandby();
// Call the HAL standby method unconditionally, and don't change mStandby flag
void inputStandBy();
void checkBtNrec_l();
AudioStreamIn *mInput;
SortedVector < sp<RecordTrack> > mTracks;
// mActiveTracks has dual roles: it indicates the current active track(s), and
// is used together with mStartStopCond to indicate start()/stop() progress
ActiveTracks<RecordTrack> mActiveTracks;
Condition mStartStopCond;
// resampler converts input at HAL Hz to output at AudioRecord client Hz
void *mRsmpInBuffer; // size = mRsmpInFramesOA
size_t mRsmpInFrames; // size of resampler input in frames
size_t mRsmpInFramesP2;// size rounded up to a power-of-2
size_t mRsmpInFramesOA;// mRsmpInFramesP2 + over-allocation
// rolling index that is never cleared
int32_t mRsmpInRear; // last filled frame + 1
// For dumpsys
const sp<NBAIO_Sink> mTeeSink;
const sp<MemoryDealer> mReadOnlyHeap;
// one-time initialization, no locks required
sp<FastCapture> mFastCapture; // non-0 if there is also
// a fast capture
// FIXME audio watchdog thread
// contents are not guaranteed to be consistent, no locks required
FastCaptureDumpState mFastCaptureDumpState;
#ifdef STATE_QUEUE_DUMP
// FIXME StateQueue observer and mutator dump fields
#endif
// FIXME audio watchdog dump
// accessible only within the threadLoop(), no locks required
// mFastCapture->sq() // for mutating and pushing state
int32_t mFastCaptureFutex; // for cold idle
// The HAL input source is treated as non-blocking,
// but current implementation is blocking
sp<NBAIO_Source> mInputSource;
// The source for the normal capture thread to read from: mInputSource or mPipeSource
sp<NBAIO_Source> mNormalSource;
// If a fast capture is present, the non-blocking pipe sink written to by fast capture,
// otherwise clear
sp<NBAIO_Sink> mPipeSink;
// If a fast capture is present, the non-blocking pipe source read by normal thread,
// otherwise clear
sp<NBAIO_Source> mPipeSource;
// Depth of pipe from fast capture to normal thread and fast clients, always power of 2
size_t mPipeFramesP2;
// If a fast capture is present, the Pipe as IMemory, otherwise clear
sp<IMemory> mPipeMemory;
static const size_t kFastCaptureLogSize = 4 * 1024;
sp<NBLog::Writer> mFastCaptureNBLogWriter;
bool mFastTrackAvail; // true if fast track available
// common state to all record threads
std::atomic_bool mBtNrecSuspended;
};
7.1.13 定义类Class MmapThread (继承自 ThreadBase ) 录音线程
class MmapThread : public ThreadBase
{
public:
#include "MmapTracks.h"
MmapThread(const sp<AudioFlinger>& audioFlinger, audio_io_handle_t id,
AudioHwDevice *hwDev, sp<StreamHalInterface> stream,
audio_devices_t outDevice, audio_devices_t inDevice, bool systemReady);
virtual ~MmapThread();
virtual void configure(const audio_attributes_t *attr,
audio_stream_type_t streamType,
audio_session_t sessionId,
const sp<MmapStreamCallback>& callback,
audio_port_handle_t deviceId,
audio_port_handle_t portId);
void disconnect();
// MmapStreamInterface
status_t createMmapBuffer(int32_t minSizeFrames, struct audio_mmap_buffer_info *info);
status_t getMmapPosition(struct audio_mmap_position *position);
status_t start(const AudioClient& client, audio_port_handle_t *handle);
status_t stop(audio_port_handle_t handle);
status_t standby();
// RefBase
virtual void onFirstRef();
// Thread virtual
virtual bool threadLoop();
virtual void threadLoop_exit();
virtual void threadLoop_standby();
virtual bool shouldStandby_l() { return false; }
virtual status_t initCheck() const { return (mHalStream == 0) ? NO_INIT : NO_ERROR; }
virtual size_t frameCount() const { return mFrameCount; }
virtual bool checkForNewParameter_l(const String8& keyValuePair, status_t& status);
virtual String8 getParameters(const String8& keys);
virtual void ioConfigChanged(audio_io_config_event event, pid_t pid = 0);
void readHalParameters_l();
virtual void cacheParameters_l() {}
virtual status_t createAudioPatch_l(const struct audio_patch *patch, audio_patch_handle_t *handle);
virtual status_t releaseAudioPatch_l(const audio_patch_handle_t handle);
virtual void getAudioPortConfig(struct audio_port_config *config);
virtual sp<StreamHalInterface> stream() const { return mHalStream; }
virtual status_t addEffectChain_l(const sp<EffectChain>& chain);
virtual size_t removeEffectChain_l(const sp<EffectChain>& chain);
virtual status_t checkEffectCompatibility_l(const effect_descriptor_t *desc, audio_session_t sessionId);
virtual uint32_t hasAudioSession_l(audio_session_t sessionId) const;
virtual status_t setSyncEvent(const sp<SyncEvent>& event);
virtual bool isValidSyncEvent(const sp<SyncEvent>& event) const;
virtual void checkSilentMode_l() {}
virtual void processVolume_l() {}
void checkInvalidTracks_l();
virtual audio_stream_type_t streamType() { return AUDIO_STREAM_DEFAULT; }
virtual void invalidateTracks(audio_stream_type_t streamType __unused) {}
void dump(int fd, const Vector<String16>& args);
virtual void dumpInternals(int fd, const Vector<String16>& args);
void dumpTracks(int fd, const Vector<String16>& args);
protected:
audio_attributes_t mAttr;
audio_session_t mSessionId;
audio_port_handle_t mDeviceId;
audio_port_handle_t mPortId;
wp<MmapStreamCallback> mCallback;
sp<StreamHalInterface> mHalStream;
sp<DeviceHalInterface> mHalDevice;
AudioHwDevice* const mAudioHwDev;
ActiveTracks<MmapTrack> mActiveTracks;
};
7.1.14 定义类Class MmapPlaybackThread (继承自 MmapThread，VolumeInterface )
class MmapPlaybackThread : public MmapThread, public VolumeInterface
{
public:
MmapPlaybackThread(const sp<AudioFlinger>& audioFlinger, audio_io_handle_t id,
AudioHwDevice *hwDev, AudioStreamOut *output,
audio_devices_t outDevice, audio_devices_t inDevice, bool systemReady);
virtual ~MmapPlaybackThread() {}
virtual void configure(const audio_attributes_t *attr,
audio_stream_type_t streamType,
audio_session_t sessionId,
const sp<MmapStreamCallback>& callback,
audio_port_handle_t deviceId,
audio_port_handle_t portId);
AudioStreamOut* clearOutput();
// VolumeInterface
virtual void setMasterVolume(float value);
virtual void setMasterMute(bool muted);
virtual void setStreamVolume(audio_stream_type_t stream, float value);
virtual void setStreamMute(audio_stream_type_t stream, bool muted);
virtual float streamVolume(audio_stream_type_t stream) const;
void setMasterMute_l(bool muted) { mMasterMute = muted; }
virtual void invalidateTracks(audio_stream_type_t streamType);
virtual audio_stream_type_t streamType() { return mStreamType; }
virtual void checkSilentMode_l();
virtual void processVolume_l();
virtual void dumpInternals(int fd, const Vector<String16>& args);
virtual bool isOutput() const override { return true; }
protected:
audio_stream_type_t mStreamType;
float mMasterVolume;
float mStreamVolume;
bool mMasterMute;
bool mStreamMute;
float mHalVolFloat;
AudioStreamOut* mOutput;
};
7.1.15 定义类Class MmapCaptureThread (继承自 MmapThread )
class MmapCaptureThread : public MmapThread
{
public:
MmapCaptureThread(const sp<AudioFlinger>& audioFlinger, audio_io_handle_t id, AudioHwDevice *hwDev,
AudioStreamIn *input, audio_devices_t outDevice, audio_devices_t inDevice, bool systemReady);
virtual ~MmapCaptureThread() {}
AudioStreamIn* clearInput();
virtual bool isOutput() const override { return false; }
protected:
AudioStreamIn* mInput;
};
7.1.16 ThreadBase 及其内部类关系总结
ThreadBase {
ConfigEventData
------>
+ IoConfigEventData
+ PrioConfigEventData
+ SetParameterConfigEventData
+ CreateAudioPatchConfigEventData
+ ReleaseAudioPatchConfigEventData
<------
IoConfigEvent
------>
+ IoConfigEvent
+ PrioConfigEvent
+ SetParameterConfigEvent
+ CreateAudioPatchConfigEvent
+ ReleaseAudioPatchConfigEvent
<------
ActiveTracks
VolumeInterface
------>
+ PlaybackThread
+ ------>
+ + MixerThread
+ + ------> DuplicatingThread
+ +
+ + DirectOutputThread
+ + ------> OffloadThread
+ <------
+ AsyncCallbackThread
+ RecordThread
+ MmapThread
+ ------>
+ + MmapPlaybackThread
+ + MmapCaptureThread
+ <------
<------
}
7.2 Threads.cpp 具体实现分析
7.2.1 ThreadBase 构造方法实现
@ /frameworks/av/services/audioflinger/Threads.cpp
// 在构造函数中，主要是对一些参数的初始化
AudioFlinger::ThreadBase::ThreadBase(const sp<AudioFlinger>& audioFlinger, audio_io_handle_t id,
audio_devices_t outDevice, audio_devices_t inDevice, type_t type, bool systemReady)
: Thread(false /*canCallJava*/),
mType(type),
mAudioFlinger(audioFlinger),
// mSampleRate, mFrameCount, mChannelMask, mChannelCount, mFrameSize, mFormat, mBufferSize
// are set by PlaybackThread::readOutputParameters_l() or
// RecordThread::readInputParameters_l()
//FIXME: mStandby should be true here. Is this some kind of hack?
mStandby(false), mOutDevice(outDevice), mInDevice(inDevice),
mPrevOutDevice(AUDIO_DEVICE_NONE), mPrevInDevice(AUDIO_DEVICE_NONE),
mAudioSource(AUDIO_SOURCE_DEFAULT), mId(id),
// mName will be set by concrete (non-virtual) subclass
mDeathRecipient(new PMDeathRecipient(this)),
mSystemReady(systemReady),
mSignalPending(false)
{
memset(&mPatch, 0, sizeof(struct audio_patch));
}
}
7.2.2 设置参数函数 ThreadBase::setParameters( )
// 设置参数，其实就是发送事件
status_t AudioFlinger::ThreadBase::setParameters(const String8& keyValuePairs)
{
ALOGV("ThreadBase::setParameters() %s", keyValuePairs.string());
return sendSetParameterConfigEvent_l(keyValuePairs);
------>
param.remove(String8(AudioParameter::keyMonoOutput));
configEvent = new SetParameterConfigEvent(param.toString());
sendConfigEvent_l(configEvent);
<------
// sendConfigEvent_l() must be called with ThreadBase::mLock held
// Can temporarily release the lock if waiting for a reply from processConfigEvents_l().
status_t AudioFlinger::ThreadBase::sendConfigEvent_l(sp<ConfigEvent>& event)
{
status_t status = NO_ERROR;
if (event->mRequiresSystemReady && !mSystemReady) {
event->mWaitStatus = false;
mPendingConfigEvents.add(event);
return status;
}
mConfigEvents.add(event);
ALOGV("sendConfigEvent_l() num events %zu event %d", mConfigEvents.size(), event->mType);
mWaitWorkCV.signal();
return status;
}
sendIoConfigEvent(event, pid);
------> sendIoConfigEvent_l(event, pid);
------>
sp<ConfigEvent> configEvent = (ConfigEvent *)new IoConfigEvent(event, pid);
sendConfigEvent_l(configEvent);
<------
<------
sendPrioConfigEvent(pid, tid, prio, forApp)
------> sendPrioConfigEvent_l(pid, tid, prio, forApp);
------>
sp<ConfigEvent> configEvent = (ConfigEvent *)new PrioConfigEvent(pid, tid, prio, forApp);
sendConfigEvent_l(configEvent);
<------
<------
7.2.3 事件处理函数实现 ThreadBase::processConfigEvents_l( )
在 processConfigEvents_l 函数中，主要是对 如下几种事件的处理
CFG_EVENT_IO, // 调用 ioConfigChanged 函数
CFG_EVENT_PRIO, // 调用 requestPriority 函数 申请优先级
CFG_EVENT_SET_PARAMETER, // 调用 checkForNewParameter_l 函数
CFG_EVENT_CREATE_AUDIO_PATCH, // 调用 createAudioPatch_l 函数，创建 Audio patch
CFG_EVENT_RELEASE_AUDIO_PATCH, // 调用 releaseAudioPatch_l 函数，释放 Audio patch
// post condition: mConfigEvents.isEmpty()
void AudioFlinger::ThreadBase::processConfigEvents_l()
{
bool configChanged = false;
while (!mConfigEvents.isEmpty()) {
ALOGV("processConfigEvents_l() remaining events %zu", mConfigEvents.size());
sp<ConfigEvent> event = mConfigEvents[0];
mConfigEvents.removeAt(0);
switch (event->mType) {
case CFG_EVENT_PRIO: {
PrioConfigEventData *data = (PrioConfigEventData *)event->mData.get();
// FIXME Need to understand why this has to be done asynchronously
int err = requestPriority(data->mPid, data->mTid, data->mPrio, data->mForApp, true /*asynchronous*/);
} break;
case CFG_EVENT_IO: {
IoConfigEventData *data = (IoConfigEventData *)event->mData.get();
ioConfigChanged(data->mEvent, data->mPid);
} break;
case CFG_EVENT_SET_PARAMETER: {
SetParameterConfigEventData *data = (SetParameterConfigEventData *)event->mData.get();
if (checkForNewParameter_l(data->mKeyValuePairs, event->mStatus)) {
configChanged = true;
mLocalLog.log("CFG_EVENT_SET_PARAMETER: (%s) configuration changed",
data->mKeyValuePairs.string());
}
} break;
case CFG_EVENT_CREATE_AUDIO_PATCH: {
const audio_devices_t oldDevice = getDevice();
CreateAudioPatchConfigEventData *data = (CreateAudioPatchConfigEventData *)event->mData.get();
event->mStatus = createAudioPatch_l(&data->mPatch, &data->mHandle);
const audio_devices_t newDevice = getDevice();
mLocalLog.log("CFG_EVENT_CREATE_AUDIO_PATCH: old device %#x (%s) new device %#x (%s)",
(unsigned)oldDevice, devicesToString(oldDevice).c_str(),
(unsigned)newDevice, devicesToString(newDevice).c_str());
} break;
case CFG_EVENT_RELEASE_AUDIO_PATCH: {
const audio_devices_t oldDevice = getDevice();
ReleaseAudioPatchConfigEventData *data = (ReleaseAudioPatchConfigEventData *)event->mData.get();
event->mStatus = releaseAudioPatch_l(data->mHandle);
const audio_devices_t newDevice = getDevice();
mLocalLog.log("CFG_EVENT_RELEASE_AUDIO_PATCH: old device %#x (%s) new device %#x (%s)",
(unsigned)oldDevice, devicesToString(oldDevice).c_str(),
(unsigned)newDevice, devicesToString(newDevice).c_str());
} break;
default:
ALOG_ASSERT(false, "processConfigEvents_l() unknown event type %d", event->mType);
break;
}
{
Mutex::Autolock _l(event->mLock);
if (event->mWaitStatus) {
event->mWaitStatus = false;
event->mCond.signal();
}
}
ALOGV_IF(mConfigEvents.isEmpty(), "processConfigEvents_l() DONE thread %p", this);
}
if (configChanged) {
cacheParameters_l();
}
}
7.2.4 创建音效 ThreadBase::createEffect_l( )
- 判断音效是否已经初始化过了
- 通过线程类型来 判断 加载哪一种音效
- 获得音效
- 初始化 并配置 音效策略
- 如果已经初始化过了，则直接获取即可
- 注册音效，如果没注册过。则分配一个 effectID 重新注册，并创建 音效模块
- 配置音效的 输入、输出 设备
- 添加到 handle 中
// ThreadBase::createEffect_l() must be called with AudioFlinger::mLock held
sp<AudioFlinger::EffectHandle> AudioFlinger::ThreadBase::createEffect_l(
const sp<AudioFlinger::Client>& client,
const sp<IEffectClient>& effectClient,
int32_t priority,
audio_session_t sessionId,
effect_descriptor_t *desc,
int *enabled,
status_t *status,
bool pinned)
{
lStatus = initCheck(); // 1. 判断音效是否已经初始化过了
ALOGV("createEffect_l() thread %p effect %s on session %d", this, desc->name, sessionId);
{ // scope for mLock
// 2. 通过线程类型来 判断 加载哪一种音效
lStatus = checkEffectCompatibility_l(desc, sessionId);
// 3. 获得音效
// check for existing effect chain with the requested audio session
chain = getEffectChain_l(sessionId);
if (chain == 0) {
// create a new chain for this session
ALOGV("createEffect_l() new effect chain for session %d", sessionId);
// 4. 初始化 并配置 音效策略
chain = new EffectChain(this, sessionId);
addEffectChain_l(chain);
chain->setStrategy(getStrategyForSession_l(sessionId));
chainCreated = true;
} else {
// 5. 如果已经初始化过了，则直接获取即可
effect = chain->getEffectFromDesc_l(desc);
}
ALOGV("createEffect_l() got effect %p on chain %p", effect.get(), chain.get());
// 6. 注册音效，如果没注册过。则分配一个 effectID 重新注册，并创建 音效模块
if (effect == 0) {
effectId = mAudioFlinger->nextUniqueId(AUDIO_UNIQUE_ID_USE_EFFECT);
// Check CPU and memory usage
lStatus = AudioSystem::registerEffect(desc, mId, chain->strategy(), sessionId, effectId);
effectRegistered = true;
// create a new effect module if none present in the chain
lStatus = chain->createEffect_l(effect, this, desc, effectId, sessionId, pinned);
effectCreated = true;
// 配置音效的 输入、输出 设备，及模式
effect->setDevice(mOutDevice);
effect->setDevice(mInDevice);
effect->setMode(mAudioFlinger->getMode());
effect->setAudioSource(mAudioSource);
}
// 7. 创建 EffectHandle 用于绑定 effect module
// create effect handle and connect it to effect module
handle = new EffectHandle(effect, client, effectClient, priority);
lStatus = handle->initCheck();
// 8. 添加到 handle 中。
if (lStatus == OK) {
lStatus = effect->addHandle(handle.get());
}
// 9. 初始化音效
if (enabled != NULL) {
*enabled = (int)effect->isEnabled();
}
}
*status = lStatus;
return handle;
}
7.2.5 ActiveTracks 操作方法
-
添加 track
mActiveTracks.add(track); -
移除 track
mActiveTracks.remove(track); -
清楚 track
mActiveTracks.clear();
7.2.6 PlaybackTread 类方法实现
7.2.6.1 PlaybackTread 初始化
// ----------------------------------------------------------------------------
// Playback
// ----------------------------------------------------------------------------
AudioFlinger::PlaybackThread::PlaybackThread(const sp<AudioFlinger>& audioFlinger,
AudioStreamOut* output,
audio_io_handle_t id,
audio_devices_t device,
type_t type,
bool systemReady)
: ThreadBase(audioFlinger, id, device, AUDIO_DEVICE_NONE, type, systemReady),
mNormalFrameCount(0), mSinkBuffer(NULL),
mMixerBufferEnabled(AudioFlinger::kEnableExtendedPrecision),
mMixerBuffer(NULL),
mMixerBufferSize(0),
mMixerBufferFormat(AUDIO_FORMAT_INVALID),
mMixerBufferValid(false),
mEffectBufferEnabled(AudioFlinger::kEnableExtendedPrecision),
mEffectBuffer(NULL),
mEffectBufferSize(0),
mEffectBufferFormat(AUDIO_FORMAT_INVALID),
mEffectBufferValid(false),
mSuspended(0), mBytesWritten(0),
mFramesWritten(0),
mSuspendedFrames(0),
mActiveTracks(&this->mLocalLog),
// mStreamTypes[] initialized in constructor body
mOutput(output),
mLastWriteTime(-1), mNumWrites(0), mNumDelayedWrites(0), mInWrite(false),
mMixerStatus(MIXER_IDLE),
mMixerStatusIgnoringFastTracks(MIXER_IDLE),
mStandbyDelayNs(AudioFlinger::mStandbyTimeInNsecs),
mBytesRemaining(0),
mCurrentWriteLength(0),
mUseAsyncWrite(false),
mWriteAckSequence(0),
mDrainSequence(0),
mScreenState(AudioFlinger::mScreenState),
// index 0 is reserved for normal mixer's submix
mFastTrackAvailMask(((1 << FastMixerState::sMaxFastTracks) - 1) & ~1),
mHwSupportsPause(false), mHwPaused(false), mFlushPending(false),
mLeftVolFloat(-1.0), mRightVolFloat(-1.0), mHwSupportsSuspend(false)
{
snprintf(mThreadName, kThreadNameLength, "AudioOut_%X", id);
mNBLogWriter = audioFlinger->newWriter_l(kLogSize, mThreadName);
// Assumes constructor is called by AudioFlinger with it's mLock held, but
// it would be safer to explicitly pass initial masterVolume/masterMute as
// parameter.
//
// If the HAL we are using has support for master volume or master mute,
// then do not attenuate or mute during mixing (just leave the volume at 1.0
// and the mute set to false).
mMasterVolume = audioFlinger->masterVolume_l();
mMasterMute = audioFlinger->masterMute_l();
if (mOutput && mOutput->audioHwDev) {
if (mOutput->audioHwDev->canSetMasterVolume()) {
mMasterVolume = 1.0;
}
if (mOutput->audioHwDev->canSetMasterMute()) {
mMasterMute = false;
}
}
readOutputParameters_l();
// ++ operator does not compile
for (audio_stream_type_t stream = AUDIO_STREAM_MIN; stream < AUDIO_STREAM_CNT;
stream = (audio_stream_type_t) (stream + 1)) {
mStreamTypes[stream].volume = mAudioFlinger->streamVolume_l(stream);
mStreamTypes[stream].mute = mAudioFlinger->streamMute_l(stream);
}
}
// Thread virtuals
void AudioFlinger::PlaybackThread::onFirstRef()
{
run(mThreadName, ANDROID_PRIORITY_URGENT_AUDIO);
}
在实例化 PlaybackThread 后，在构造函数中初始化相关的量后，就会调用 onFirstRef 函数，运行线程。
线程的名字为 AudioOut_%x。
snprintf(mThreadName, kThreadNameLength, “AudioOut_%X”, id);
7.2.6.2 PlaybackThread::createTrack_l( )
waitting update…
@ \frameworks\av\services\audioflinger\Threads.cpp
// PlaybackThread::createTrack_l() must be called with AudioFlinger::mLock held
sp<AudioFlinger::PlaybackThread::Track> AudioFlinger::PlaybackThread::createTrack_l()
{
// special case for FAST flag considered OK if fast mixer is present
if (hasFastMixer()) {
outputFlags = (audio_output_flags_t)(outputFlags | AUDIO_OUTPUT_FLAG_FAST);
}
// Check if requested flags are compatible with output stream flags
if ((*flags & outputFlags) != *flags) {
ALOGW("createTrack_l(): mismatch between requested flags (%08x) and output flags (%08x)", *flags, outputFlags);
*flags = (audio_output_flags_t)(*flags & outputFlags);
}
// client expresses a preference for FAST, but we get the final say
if (*flags & AUDIO_OUTPUT_FLAG_FAST) {
if (
// PCM data
audio_is_linear_pcm(format) &&
// TODO: extract as a data library function that checks that a computationally
// expensive downmixer is not required: isFastOutputChannelConversion()
(channelMask == mChannelMask ||
mChannelMask != AUDIO_CHANNEL_OUT_STEREO ||
(channelMask == AUDIO_CHANNEL_OUT_MONO
/* && mChannelMask == AUDIO_CHANNEL_OUT_STEREO */)) &&
// hardware sample rate
(sampleRate == mSampleRate) &&
// normal mixer has an associated fast mixer
hasFastMixer() &&
// there are sufficient fast track slots available
(mFastTrackAvailMask != 0)
// FIXME test that MixerThread for this fast track has a capable output HAL
// FIXME add a permission test also?
) {
// static tracks can have any nonzero framecount, streaming tracks check against minimum.
if (sharedBuffer == 0) {
// read the fast track multiplier property the first time it is needed
int ok = pthread_once(&sFastTrackMultiplierOnce, sFastTrackMultiplierInit);
if (ok != 0) {
ALOGE("%s pthread_once failed: %d", __func__, ok);
}
frameCount = max(frameCount, mFrameCount * sFastTrackMultiplier); // incl framecount 0
}
// check compatibility with audio effects.
{ // scope for mLock
Mutex::Autolock _l(mLock);
for (audio_session_t session : {
AUDIO_SESSION_OUTPUT_STAGE,
AUDIO_SESSION_OUTPUT_MIX,
sessionId,
}) {
sp<EffectChain> chain = getEffectChain_l(session);
if (chain.get() != nullptr) {
audio_output_flags_t old = *flags;
chain->checkOutputFlagCompatibility(flags);
if (old != *flags) {
ALOGV("AUDIO_OUTPUT_FLAGS denied by effect, session=%d old=%#x new=%#x",
(int)session, (int)old, (int)*flags);
}
}
}
}
ALOGV_IF((*flags & AUDIO_OUTPUT_FLAG_FAST) != 0,
"AUDIO_OUTPUT_FLAG_FAST accepted: frameCount=%zu mFrameCount=%zu",
frameCount, mFrameCount);
} else {
ALOGV("AUDIO_OUTPUT_FLAG_FAST denied: sharedBuffer=%p frameCount=%zu "
"mFrameCount=%zu format=%#x mFormat=%#x isLinear=%d channelMask=%#x "
"sampleRate=%u mSampleRate=%u "
"hasFastMixer=%d tid=%d fastTrackAvailMask=%#x",
sharedBuffer.get(), frameCount, mFrameCount, format, mFormat,
audio_is_linear_pcm(format),
channelMask, sampleRate, mSampleRate, hasFastMixer(), tid, mFastTrackAvailMask);
*flags = (audio_output_flags_t)(*flags & ~AUDIO_OUTPUT_FLAG_FAST);
}
}
// For normal PCM streaming tracks, update minimum frame count.
// For compatibility with AudioTrack calculation, buffer depth is forced
// to be at least 2 x the normal mixer frame count and cover audio hardware latency.
// This is probably too conservative, but legacy application code may depend on it.
// If you change this calculation, also review the start threshold which is related.
if (!(*flags & AUDIO_OUTPUT_FLAG_FAST)
&& audio_has_proportional_frames(format) && sharedBuffer == 0) {
// this must match AudioTrack.cpp calculateMinFrameCount().
// TODO: Move to a common library
uint32_t latencyMs = 0;
lStatus = mOutput->stream->getLatency(&latencyMs);
if (lStatus != OK) {
ALOGE("Error when retrieving output stream latency: %d", lStatus);
goto Exit;
}
uint32_t minBufCount = latencyMs / ((1000 * mNormalFrameCount) / mSampleRate);
if (minBufCount < 2) {
minBufCount = 2;
}
// For normal mixing tracks, if speed is > 1.0f (normal), AudioTrack
// or the client should compute and pass in a larger buffer request.
size_t minFrameCount =
minBufCount * sourceFramesNeededWithTimestretch(
sampleRate, mNormalFrameCount,
mSampleRate, AUDIO_TIMESTRETCH_SPEED_NORMAL /*speed*/);
if (frameCount < minFrameCount) { // including frameCount == 0
frameCount = minFrameCount;
}
}
*pFrameCount = frameCount;
switch (mType) {
case DIRECT:
if (audio_is_linear_pcm(format)) { // TODO maybe use audio_has_proportional_frames()?
if (sampleRate != mSampleRate || format != mFormat || channelMask != mChannelMask) {
ALOGE("createTrack_l() Bad parameter: sampleRate %u format %#x, channelMask 0x%08x "
"for output %p with format %#x",
sampleRate, format, channelMask, mOutput, mFormat);
lStatus = BAD_VALUE;
goto Exit;
}
}
break;
case OFFLOAD:
if (sampleRate != mSampleRate || format != mFormat || channelMask != mChannelMask) {
ALOGE("createTrack_l() Bad parameter: sampleRate %d format %#x, channelMask 0x%08x \""
"for output %p with format %#x",
sampleRate, format, channelMask, mOutput, mFormat);
lStatus = BAD_VALUE;
goto Exit;
}
break;
default:
if (!audio_is_linear_pcm(format)) {
ALOGE("createTrack_l() Bad parameter: format %#x \""
"for output %p with format %#x",
format, mOutput, mFormat);
lStatus = BAD_VALUE;
goto Exit;
}
if (sampleRate > mSampleRate * AUDIO_RESAMPLER_DOWN_RATIO_MAX) {
ALOGE("Sample rate out of range: %u mSampleRate %u", sampleRate, mSampleRate);
lStatus = BAD_VALUE;
goto Exit;
}
break;
}
lStatus = initCheck();
if (lStatus != NO_ERROR) {
ALOGE("createTrack_l() audio driver not initialized");
goto Exit;
}
{ // scope for mLock
Mutex::Autolock _l(mLock);
// all tracks in same audio session must share the same routing strategy otherwise
// conflicts will happen when tracks are moved from one output to another by audio policy
// manager
uint32_t strategy = AudioSystem::getStrategyForStream(streamType);
for (size_t i = 0; i < mTracks.size(); ++i) {
sp<Track> t = mTracks[i];
if (t != 0 && t->isExternalTrack()) {
uint32_t actual = AudioSystem::getStrategyForStream(t->streamType());
if (sessionId == t->sessionId() && strategy != actual) {
ALOGE("createTrack_l() mismatched strategy; expected %u but found %u",
strategy, actual);
lStatus = BAD_VALUE;
goto Exit;
}
}
}
track = new Track(this, client, streamType, sampleRate, format,
channelMask, frameCount,
nullptr /* buffer */, (size_t)0 /* bufferSize */, sharedBuffer,
sessionId, uid, *flags, TrackBase::TYPE_DEFAULT, portId);
lStatus = track != 0 ? track->initCheck() : (status_t) NO_MEMORY;
if (lStatus != NO_ERROR) {
ALOGE("createTrack_l() initCheck failed %d; no control block?", lStatus);
// track must be cleared from the caller as the caller has the AF lock
goto Exit;
}
mTracks.add(track);
sp<EffectChain> chain = getEffectChain_l(sessionId);
if (chain != 0) {
ALOGV("createTrack_l() setting main buffer %p", chain->inBuffer());
track->setMainBuffer(chain->inBuffer());
chain->setStrategy(AudioSystem::getStrategyForStream(track->streamType()));
chain->incTrackCnt();
}
if ((*flags & AUDIO_OUTPUT_FLAG_FAST) && (tid != -1)) {
pid_t callingPid = IPCThreadState::self()->getCallingPid();
// we don't have CAP_SYS_NICE, nor do we want to have it as it's too powerful,
// so ask activity manager to do this on our behalf
sendPrioConfigEvent_l(callingPid, tid, kPriorityAudioApp, true /*forApp*/);
}
}
lStatus = NO_ERROR;
线程的类型 ，保存在 mType 私有变量中：
@ /frameworks/av/services/audioflinger/Threads.cpp
// 获得当前线程的名字
const char *AudioFlinger::ThreadBase::threadTypeToString(AudioFlinger::ThreadBase::type_t type)
{
switch (type) {
case MIXER: return "MIXER"; ---> MixerThread
case DIRECT: return "DIRECT"; ---> DirectOutputThread
case DUPLICATING: return "DUPLICATING"; ---> DuplicatingThread
case RECORD: return "RECORD"; ---> RecordThread
case OFFLOAD: return "OFFLOAD"; ---> OffloadThread
case MMAP: return "MMAP"; ---> MmapThread
default: return "unknown";
}
}
@ /frameworks/av/services/audioflinger/Threads.cpp
// 获得当前输入类型 的名字
const char *sourceToString(audio_source_t source)
{
switch (source) {
case AUDIO_SOURCE_DEFAULT: return "default";
case AUDIO_SOURCE_MIC: return "mic";
case AUDIO_SOURCE_VOICE_UPLINK: return "voice uplink";
case AUDIO_SOURCE_VOICE_DOWNLINK: return "voice downlink";
case AUDIO_SOURCE_VOICE_CALL: return "voice call";
case AUDIO_SOURCE_CAMCORDER: return "camcorder";
case AUDIO_SOURCE_VOICE_RECOGNITION: return "voice recognition";
case AUDIO_SOURCE_VOICE_COMMUNICATION: return "voice communication";
case AUDIO_SOURCE_REMOTE_SUBMIX: return "remote submix";
case AUDIO_SOURCE_UNPROCESSED: return "unprocessed";
case AUDIO_SOURCE_FM_TUNER: return "FM tuner";
case AUDIO_SOURCE_HOTWORD: return "hotword";
default: return "unknown";
}
}
8. Audio EffectModule 音效模块分析
Audio EffectModule 主要是对 Effect 库的方法封装。
给不同的 thread client 提供 process() 和 command() 方法实现。
8.1 EffectModule 类定义
先来看下类定义：
@ \frameworks\av\services\audioflinger\Effects.h
class EffectModule : public RefBase {
public:
// 构造函数
EffectModule(ThreadBase *thread, const wp<AudioFlinger::EffectChain>& chain,
effect_descriptor_t *desc, int id, audio_session_t sessionId, bool pinned);
virtual ~EffectModule(); // 析构函数
enum effect_state { // 音效的状态 enum
IDLE, // 0
RESTART, // 1
STARTING, // 2
ACTIVE, // 3
STOPPING, // 4
STOPPED, // 5
DESTROYED // 6
};
int id() const { return mId; } // 返回当前入口的 特殊ID (unique ID)
void process(); // 音效处理函数
bool updateState(); // 更新音效状态
// 音效命令处理
status_t command(uint32_t cmdCode,uint32_t cmdSize,void *pCmdData, uint32_t *replySize,void *pReplyData);
void reset_l(); // 重置音效 发送 command( EFFECT_CMD_RESET )
status_t configure(); // 音效参数配置，发送 command( EFFECT_CMD_SET_PARAM )
status_t init(); // 音效初始化，发送 command( EFFECT_CMD_INIT )
effect_state state() const { return mState; } // 返回当前的状态
uint32_t status() { return mStatus; }
audio_session_t sessionId() const { return mSessionId; }
status_t setEnabled(bool enabled); // 使能 EffectModule
status_t setEnabled_l(bool enabled);// 使能 EffectModule
bool isEnabled() const; // 返回当前 EffectModule 是否已使能
bool isProcessEnabled() const;
void setInBuffer(const sp<EffectBufferHalInterface>& buffer); // 配置 buff 为输入 buff
int16_t *inBuffer() const { // 返回输入 buff
return mInBuffer != 0 ? reinterpret_cast<int16_t*>(mInBuffer->ptr()) : NULL;
}
void setOutBuffer(const sp<EffectBufferHalInterface>& buffer); // 配置 buff 为输出 buff
int16_t *outBuffer() const { // 返回输出 buff
return mOutBuffer != 0 ? reinterpret_cast<int16_t*>(mOutBuffer->ptr()) : NULL;
}
void setChain(const wp<EffectChain>& chain) { mChain = chain; } // 配置通道
void setThread(const wp<ThreadBase>& thread) { mThread = thread; } // 配置线程
const wp<ThreadBase>& thread() { return mThread; } // 返回当前全局线程
status_t addHandle(EffectHandle *handle); // 将当前 handle 插入到全局 mHandles 链表中
ssize_t disconnectHandle(EffectHandle *handle, bool unpinIfLast); //
ssize_t removeHandle(EffectHandle *handle); // 多除 handle
ssize_t removeHandle_l(EffectHandle *handle);
const effect_descriptor_t& desc() const { return mDescriptor; }
wp<EffectChain>& chain() { return mChain; }
status_t setDevice(audio_devices_t device);
status_t setVolume(uint32_t *left, uint32_t *right, bool controller);
status_t setMode(audio_mode_t mode);
status_t setAudioSource(audio_source_t source);
status_t start();
status_t stop();
void setSuspended(bool suspended);
bool suspended() const;
EffectHandle* controlHandle_l();
bool isPinned() const { return mPinned; }
void unPin() { mPinned = false; }
bool purgeHandles();
void lock() { mLock.lock(); }
void unlock() { mLock.unlock(); }
bool isOffloadable() const
{ return (mDescriptor.flags & EFFECT_FLAG_OFFLOAD_SUPPORTED) != 0; }
bool isImplementationSoftware() const
{ return (mDescriptor.flags & EFFECT_FLAG_HW_ACC_MASK) == 0; }
bool isProcessImplemented() const
{ return (mDescriptor.flags & EFFECT_FLAG_NO_PROCESS) == 0; }
status_t setOffloaded(bool offloaded, audio_io_handle_t io);
bool isOffloaded() const;
void addEffectToHal_l();
void release_l();
void dump(int fd, const Vector<String16>& args);
private:
friend class AudioFlinger; // for mHandles
bool mPinned;
// Maximum time allocated to effect engines to complete the turn off sequence
static const uint32_t MAX_DISABLE_TIME_MS = 10000;
DISALLOW_COPY_AND_ASSIGN(EffectModule);
status_t start_l();
status_t stop_l();
status_t remove_effect_from_hal_l();
mutable Mutex mLock; // mutex for process, commands and handles list protection
wp<ThreadBase> mThread; // parent thread
wp<EffectChain> mChain; // parent effect chain
const int mId; // this instance unique ID
const audio_session_t mSessionId; // audio session ID
const effect_descriptor_t mDescriptor;// effect descriptor received from effect engine
effect_config_t mConfig; // input and output audio configuration
sp<EffectHalInterface> mEffectInterface; // Effect module HAL
sp<EffectBufferHalInterface> mInBuffer; // Buffers for interacting with HAL
sp<EffectBufferHalInterface> mOutBuffer;
status_t mStatus; // initialization status
effect_state mState; // current activation state
Vector<EffectHandle *> mHandles; // list of client handles
// First handle in mHandles has highest priority and controls the effect module
uint32_t mMaxDisableWaitCnt; // maximum grace period before forcing an effect off after
// sending disable command.
uint32_t mDisableWaitCnt; // current process() calls count during disable period.
bool mSuspended; // effect is suspended: temporarily disabled by framework
bool mOffloaded; // effect is currently offloaded to the audio DSP
wp<AudioFlinger> mAudioFlinger;
};
