blob: 2067d88740c9f08f9a47eaf1ece4892f64fa0ab1 [file] [log] [blame]
* Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
#include <memory>
#include "webrtc/sdk/objc/Framework/Headers/WebRTC/RTCMacros.h"
#include "webrtc/modules/audio_device/audio_device_generic.h"
#include "webrtc/modules/audio_device/ios/audio_session_observer.h"
#include "webrtc/modules/audio_device/ios/voice_processing_audio_unit.h"
#include "webrtc/rtc_base/buffer.h"
#include "webrtc/rtc_base/gtest_prod_util.h"
#include "webrtc/rtc_base/thread.h"
#include "webrtc/rtc_base/thread_annotations.h"
#include "webrtc/rtc_base/thread_checker.h"
namespace webrtc {
class FineAudioBuffer;
// Implements full duplex 16-bit mono PCM audio support for iOS using a
// Voice-Processing (VP) I/O audio unit in Core Audio. The VP I/O audio unit
// supports audio echo cancellation. It also adds automatic gain control,
// adjustment of voice-processing quality and muting.
// An instance must be created and destroyed on one and the same thread.
// All supported public methods must also be called on the same thread.
// A thread checker will RTC_DCHECK if any supported method is called on an
// invalid thread.
// Recorded audio will be delivered on a real-time internal I/O thread in the
// audio unit. The audio unit will also ask for audio data to play out on this
// same thread.
class AudioDeviceIOS : public AudioDeviceGeneric,
public AudioSessionObserver,
public VoiceProcessingAudioUnitObserver,
public rtc::MessageHandler {
void AttachAudioBuffer(AudioDeviceBuffer* audioBuffer) override;
InitStatus Init() override;
int32_t Terminate() override;
bool Initialized() const override;
int32_t InitPlayout() override;
bool PlayoutIsInitialized() const override;
int32_t InitRecording() override;
bool RecordingIsInitialized() const override;
int32_t StartPlayout() override;
int32_t StopPlayout() override;
bool Playing() const override { return playing_; }
int32_t StartRecording() override;
int32_t StopRecording() override;
bool Recording() const override { return recording_; }
int32_t SetLoudspeakerStatus(bool enable) override;
int32_t GetLoudspeakerStatus(bool& enabled) const override;
// These methods returns hard-coded delay values and not dynamic delay
// estimates. The reason is that iOS supports a built-in AEC and the WebRTC
// AEC will always be disabled in the Libjingle layer to avoid running two
// AEC implementations at the same time. And, it saves resources to avoid
// updating these delay values continuously.
// TODO(henrika): it would be possible to mark these two methods as not
// implemented since they are only called for A/V-sync purposes today and
// A/V-sync is not supported on iOS. However, we avoid adding error messages
// the log by using these dummy implementations instead.
int32_t PlayoutDelay(uint16_t& delayMS) const override;
int32_t RecordingDelay(uint16_t& delayMS) const override;
// Native audio parameters stored during construction.
// These methods are unique for the iOS implementation.
int GetPlayoutAudioParameters(AudioParameters* params) const override;
int GetRecordAudioParameters(AudioParameters* params) const override;
// These methods are currently not fully implemented on iOS:
// See for trivial implementations.
int32_t ActiveAudioLayer(
AudioDeviceModule::AudioLayer& audioLayer) const override;
int32_t PlayoutIsAvailable(bool& available) override;
int32_t RecordingIsAvailable(bool& available) override;
int32_t SetAGC(bool enable) override;
bool AGC() const override;
int16_t PlayoutDevices() override;
int16_t RecordingDevices() override;
int32_t PlayoutDeviceName(uint16_t index,
char name[kAdmMaxDeviceNameSize],
char guid[kAdmMaxGuidSize]) override;
int32_t RecordingDeviceName(uint16_t index,
char name[kAdmMaxDeviceNameSize],
char guid[kAdmMaxGuidSize]) override;
int32_t SetPlayoutDevice(uint16_t index) override;
int32_t SetPlayoutDevice(
AudioDeviceModule::WindowsDeviceType device) override;
int32_t SetRecordingDevice(uint16_t index) override;
int32_t SetRecordingDevice(
AudioDeviceModule::WindowsDeviceType device) override;
int32_t InitSpeaker() override;
bool SpeakerIsInitialized() const override;
int32_t InitMicrophone() override;
bool MicrophoneIsInitialized() const override;
int32_t SpeakerVolumeIsAvailable(bool& available) override;
int32_t SetSpeakerVolume(uint32_t volume) override;
int32_t SpeakerVolume(uint32_t& volume) const override;
int32_t MaxSpeakerVolume(uint32_t& maxVolume) const override;
int32_t MinSpeakerVolume(uint32_t& minVolume) const override;
int32_t MicrophoneVolumeIsAvailable(bool& available) override;
int32_t SetMicrophoneVolume(uint32_t volume) override;
int32_t MicrophoneVolume(uint32_t& volume) const override;
int32_t MaxMicrophoneVolume(uint32_t& maxVolume) const override;
int32_t MinMicrophoneVolume(uint32_t& minVolume) const override;
int32_t MicrophoneMuteIsAvailable(bool& available) override;
int32_t SetMicrophoneMute(bool enable) override;
int32_t MicrophoneMute(bool& enabled) const override;
int32_t SpeakerMuteIsAvailable(bool& available) override;
int32_t SetSpeakerMute(bool enable) override;
int32_t SpeakerMute(bool& enabled) const override;
int32_t StereoPlayoutIsAvailable(bool& available) override;
int32_t SetStereoPlayout(bool enable) override;
int32_t StereoPlayout(bool& enabled) const override;
int32_t StereoRecordingIsAvailable(bool& available) override;
int32_t SetStereoRecording(bool enable) override;
int32_t StereoRecording(bool& enabled) const override;
bool PlayoutWarning() const override;
bool PlayoutError() const override;
bool RecordingWarning() const override;
bool RecordingError() const override;
void ClearPlayoutWarning() override {}
void ClearPlayoutError() override {}
void ClearRecordingWarning() override {}
void ClearRecordingError() override {}
// AudioSessionObserver methods. May be called from any thread.
void OnInterruptionBegin() override;
void OnInterruptionEnd() override;
void OnValidRouteChange() override;
void OnCanPlayOrRecordChange(bool can_play_or_record) override;
void OnChangedOutputVolume() override;
// VoiceProcessingAudioUnitObserver methods.
OSStatus OnDeliverRecordedData(AudioUnitRenderActionFlags* flags,
const AudioTimeStamp* time_stamp,
UInt32 bus_number,
UInt32 num_frames,
AudioBufferList* io_data) override;
OSStatus OnGetPlayoutData(AudioUnitRenderActionFlags* flags,
const AudioTimeStamp* time_stamp,
UInt32 bus_number,
UInt32 num_frames,
AudioBufferList* io_data) override;
// Handles messages from posts.
void OnMessage(rtc::Message *msg) override;
// Called by the relevant AudioSessionObserver methods on |thread_|.
void HandleInterruptionBegin();
void HandleInterruptionEnd();
void HandleValidRouteChange();
void HandleCanPlayOrRecordChange(bool can_play_or_record);
void HandleSampleRateChange(float sample_rate);
void HandlePlayoutGlitchDetected();
void HandleOutputVolumeChange();
// Uses current |playout_parameters_| and |record_parameters_| to inform the
// audio device buffer (ADB) about our internal audio parameters.
void UpdateAudioDeviceBuffer();
// Since the preferred audio parameters are only hints to the OS, the actual
// values may be different once the AVAudioSession has been activated.
// This method asks for the current hardware parameters and takes actions
// if they should differ from what we have asked for initially. It also
// defines |playout_parameters_| and |record_parameters_|.
void SetupAudioBuffersForActiveAudioSession();
// Creates the audio unit.
bool CreateAudioUnit();
// Updates the audio unit state based on current state.
void UpdateAudioUnit(bool can_play_or_record);
// Configures the audio session for WebRTC.
bool ConfigureAudioSession();
// Unconfigures the audio session.
void UnconfigureAudioSession();
// Activates our audio session, creates and initializes the voice-processing
// audio unit and verifies that we got the preferred native audio parameters.
bool InitPlayOrRecord();
// Closes and deletes the voice-processing I/O unit.
void ShutdownPlayOrRecord();
// Ensures that methods are called from the same thread as this object is
// created on.
rtc::ThreadChecker thread_checker_;
// Native I/O audio thread checker.
rtc::ThreadChecker io_thread_checker_;
// Thread that this object is created on.
rtc::Thread* thread_;
// Raw pointer handle provided to us in AttachAudioBuffer(). Owned by the
// AudioDeviceModuleImpl class and called by AudioDeviceModule::Create().
// The AudioDeviceBuffer is a member of the AudioDeviceModuleImpl instance
// and therefore outlives this object.
AudioDeviceBuffer* audio_device_buffer_;
// Contains audio parameters (sample rate, #channels, buffer size etc.) for
// the playout and recording sides. These structure is set in two steps:
// first, native sample rate and #channels are defined in Init(). Next, the
// audio session is activated and we verify that the preferred parameters
// were granted by the OS. At this stage it is also possible to add a third
// component to the parameters; the native I/O buffer duration.
// A RTC_CHECK will be hit if we for some reason fail to open an audio session
// using the specified parameters.
AudioParameters playout_parameters_;
AudioParameters record_parameters_;
// The AudioUnit used to play and record audio.
std::unique_ptr<VoiceProcessingAudioUnit> audio_unit_;
// FineAudioBuffer takes an AudioDeviceBuffer which delivers audio data
// in chunks of 10ms. It then allows for this data to be pulled in
// a finer or coarser granularity. I.e. interacting with this class instead
// of directly with the AudioDeviceBuffer one can ask for any number of
// audio data samples. Is also supports a similar scheme for the recording
// side.
// Example: native buffer size can be 128 audio frames at 16kHz sample rate.
// WebRTC will provide 480 audio frames per 10ms but iOS asks for 128
// in each callback (one every 8ms). This class can then ask for 128 and the
// FineAudioBuffer will ask WebRTC for new data only when needed and also
// cache non-utilized audio between callbacks. On the recording side, iOS
// can provide audio data frames of size 128 and these are accumulated until
// enough data to supply one 10ms call exists. This 10ms chunk is then sent
// to WebRTC and the remaining part is stored.
std::unique_ptr<FineAudioBuffer> fine_audio_buffer_;
// Temporary storage for recorded data. AudioUnitRender() renders into this
// array as soon as a frame of the desired buffer size has been recorded.
// On real iOS devices, the size will be fixed and set once. For iOS
// simulators, the size can vary from callback to callback and the size
// will be changed dynamically to account for this behavior.
rtc::BufferT<int8_t> record_audio_buffer_;
// Set to 1 when recording is active and 0 otherwise.
volatile int recording_;
// Set to 1 when playout is active and 0 otherwise.
volatile int playing_;
// Set to true after successful call to Init(), false otherwise.
bool initialized_ RTC_ACCESS_ON(thread_checker_);
// Set to true after successful call to InitRecording() or InitPlayout(),
// false otherwise.
bool audio_is_initialized_;
// Set to true if audio session is interrupted, false otherwise.
bool is_interrupted_;
// Audio interruption observer instance.
RTCAudioSessionDelegateAdapter* audio_session_observer_
// Set to true if we've activated the audio session.
bool has_configured_session_ RTC_ACCESS_ON(thread_checker_);
// Counts number of detected audio glitches on the playout side.
int64_t num_detected_playout_glitches_ RTC_ACCESS_ON(thread_checker_);
int64_t last_playout_time_ RTC_ACCESS_ON(io_thread_checker_);
// Counts number of playout callbacks per call.
// The value isupdated on the native I/O thread and later read on the
// creating thread (see thread_checker_) but at this stage no audio is
// active. Hence, it is a "thread safe" design and no lock is needed.
int64_t num_playout_callbacks_;
// Contains the time for when the last output volume change was detected.
int64_t last_output_volume_change_time_ RTC_ACCESS_ON(thread_checker_);
// Exposes private members for testing purposes only.
FRIEND_TEST_ALL_PREFIXES(AudioDeviceTest, testInterruptedAudioSession);
} // namespace webrtc