blob: c627f334ebba0940ceb84b825af6e8200d3c1366 [file] [log] [blame]
/*
* Copyright (C) 2010 Google Inc. All rights reserved.
* Copyright (C) 2016 Apple Inc. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY APPLE INC. AND ITS CONTRIBUTORS ``AS IS'' AND ANY
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
* DISCLAIMED. IN NO EVENT SHALL APPLE INC. OR ITS CONTRIBUTORS BE LIABLE FOR ANY
* DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "config.h"
#if ENABLE(WEB_AUDIO)
#include "AudioContext.h"
#include "AnalyserNode.h"
#include "AsyncAudioDecoder.h"
#include "AudioBuffer.h"
#include "AudioBufferCallback.h"
#include "AudioBufferSourceNode.h"
#include "AudioListener.h"
#include "AudioNodeInput.h"
#include "AudioNodeOutput.h"
#include "AudioSession.h"
#include "BiquadFilterNode.h"
#include "ChannelMergerNode.h"
#include "ChannelSplitterNode.h"
#include "ConvolverNode.h"
#include "DefaultAudioDestinationNode.h"
#include "DelayNode.h"
#include "Document.h"
#include "DynamicsCompressorNode.h"
#include "EventNames.h"
#include "FFTFrame.h"
#include "Frame.h"
#include "FrameLoader.h"
#include "GainNode.h"
#include "GenericEventQueue.h"
#include "HRTFDatabaseLoader.h"
#include "HRTFPanner.h"
#include "JSDOMPromiseDeferred.h"
#include "Logging.h"
#include "NetworkingContext.h"
#include "OfflineAudioCompletionEvent.h"
#include "OfflineAudioDestinationNode.h"
#include "OscillatorNode.h"
#include "Page.h"
#include "PannerNode.h"
#include "PeriodicWave.h"
#include "ScriptController.h"
#include "ScriptProcessorNode.h"
#include "WaveShaperNode.h"
#include <JavaScriptCore/ScriptCallStack.h>
#if ENABLE(MEDIA_STREAM)
#include "MediaStream.h"
#include "MediaStreamAudioDestinationNode.h"
#include "MediaStreamAudioSource.h"
#include "MediaStreamAudioSourceNode.h"
#endif
#if ENABLE(VIDEO)
#include "HTMLMediaElement.h"
#include "MediaElementAudioSourceNode.h"
#endif
#if DEBUG_AUDIONODE_REFERENCES
#include <stdio.h>
#endif
#if USE(GSTREAMER)
#include "GStreamerCommon.h"
#endif
#if PLATFORM(IOS_FAMILY)
#include "ScriptController.h"
#include "Settings.h"
#endif
#include <JavaScriptCore/ArrayBuffer.h>
#include <wtf/Atomics.h>
#include <wtf/IsoMallocInlines.h>
#include <wtf/MainThread.h>
#include <wtf/Ref.h>
#include <wtf/RefCounted.h>
#include <wtf/Scope.h>
#include <wtf/text/WTFString.h>
const unsigned MaxPeriodicWaveLength = 4096;
namespace WebCore {
WTF_MAKE_ISO_ALLOCATED_IMPL(AudioContext);
#define RELEASE_LOG_IF_ALLOWED(fmt, ...) RELEASE_LOG_IF(document() && document()->page() && document()->page()->isAlwaysOnLoggingAllowed(), Media, "%p - AudioContext::" fmt, this, ##__VA_ARGS__)
bool AudioContext::isSampleRateRangeGood(float sampleRate)
{
// FIXME: It would be nice if the minimum sample-rate could be less than 44.1KHz,
// but that will require some fixes in HRTFPanner::fftSizeForSampleRate(), and some testing there.
return sampleRate >= 44100 && sampleRate <= 96000;
}
// Don't allow more than this number of simultaneous AudioContexts talking to hardware.
const unsigned MaxHardwareContexts = 4;
unsigned AudioContext::s_hardwareContextCount = 0;
RefPtr<AudioContext> AudioContext::create(Document& document)
{
ASSERT(isMainThread());
if (s_hardwareContextCount >= MaxHardwareContexts)
return nullptr;
RefPtr<AudioContext> audioContext(adoptRef(new AudioContext(document)));
audioContext->suspendIfNeeded();
return audioContext;
}
// Constructor for rendering to the audio hardware.
AudioContext::AudioContext(Document& document)
: ActiveDOMObject(document)
#if !RELEASE_LOG_DISABLED
, m_logger(document.logger())
, m_logIdentifier(uniqueLogIdentifier())
#endif
, m_mediaSession(PlatformMediaSession::create(*this))
, m_eventQueue(MainThreadGenericEventQueue::create(*this))
{
// According to spec AudioContext must die only after page navigate.
// Lets mark it as ActiveDOMObject with pending activity and unmark it in clear method.
makePendingActivity();
constructCommon();
m_destinationNode = DefaultAudioDestinationNode::create(*this);
// Initialize the destination node's muted state to match the page's current muted state.
pageMutedStateDidChange();
document.addAudioProducer(*this);
document.registerForVisibilityStateChangedCallbacks(*this);
}
// Constructor for offline (non-realtime) rendering.
AudioContext::AudioContext(Document& document, unsigned numberOfChannels, size_t numberOfFrames, float sampleRate)
: ActiveDOMObject(document)
#if !RELEASE_LOG_DISABLED
, m_logger(document.logger())
, m_logIdentifier(uniqueLogIdentifier())
#endif
, m_isOfflineContext(true)
, m_mediaSession(PlatformMediaSession::create(*this))
, m_eventQueue(MainThreadGenericEventQueue::create(*this))
{
constructCommon();
// Create a new destination for offline rendering.
m_renderTarget = AudioBuffer::create(numberOfChannels, numberOfFrames, sampleRate);
m_destinationNode = OfflineAudioDestinationNode::create(*this, m_renderTarget.get());
}
void AudioContext::constructCommon()
{
FFTFrame::initialize();
m_listener = AudioListener::create();
ASSERT(document());
if (document()->audioPlaybackRequiresUserGesture())
addBehaviorRestriction(RequireUserGestureForAudioStartRestriction);
else
m_restrictions = NoRestrictions;
#if PLATFORM(COCOA)
addBehaviorRestriction(RequirePageConsentForAudioStartRestriction);
#endif
}
AudioContext::~AudioContext()
{
#if DEBUG_AUDIONODE_REFERENCES
fprintf(stderr, "%p: AudioContext::~AudioContext()\n", this);
#endif
ASSERT(!m_isInitialized);
ASSERT(m_isStopScheduled);
ASSERT(m_nodesToDelete.isEmpty());
ASSERT(m_referencedNodes.isEmpty());
ASSERT(m_finishedNodes.isEmpty()); // FIXME (bug 105870): This assertion fails on tests sometimes.
ASSERT(m_automaticPullNodes.isEmpty());
if (m_automaticPullNodesNeedUpdating)
m_renderingAutomaticPullNodes.resize(m_automaticPullNodes.size());
ASSERT(m_renderingAutomaticPullNodes.isEmpty());
// FIXME: Can we assert that m_deferredFinishDerefList is empty?
if (!isOfflineContext() && scriptExecutionContext()) {
document()->removeAudioProducer(*this);
document()->unregisterForVisibilityStateChangedCallbacks(*this);
}
}
void AudioContext::lazyInitialize()
{
ASSERT(!m_isStopScheduled);
if (m_isInitialized)
return;
// Don't allow the context to initialize a second time after it's already been explicitly uninitialized.
ASSERT(!m_isAudioThreadFinished);
if (m_isAudioThreadFinished)
return;
if (m_destinationNode) {
m_destinationNode->initialize();
if (!isOfflineContext()) {
// This starts the audio thread. The destination node's provideInput() method will now be called repeatedly to render audio.
// Each time provideInput() is called, a portion of the audio stream is rendered. Let's call this time period a "render quantum".
// NOTE: for now default AudioContext does not need an explicit startRendering() call from JavaScript.
// We may want to consider requiring it for symmetry with OfflineAudioContext.
startRendering();
++s_hardwareContextCount;
}
}
m_isInitialized = true;
}
void AudioContext::clear()
{
Ref<AudioContext> protectedThis(*this);
// We have to release our reference to the destination node before the context will ever be deleted since the destination node holds a reference to the context.
if (m_destinationNode)
m_destinationNode = nullptr;
// Audio thread is dead. Nobody will schedule node deletion action. Let's do it ourselves.
do {
deleteMarkedNodes();
m_nodesToDelete.appendVector(m_nodesMarkedForDeletion);
m_nodesMarkedForDeletion.clear();
} while (m_nodesToDelete.size());
clearPendingActivity();
}
void AudioContext::uninitialize()
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (!m_isInitialized)
return;
// This stops the audio thread and all audio rendering.
if (m_destinationNode)
m_destinationNode->uninitialize();
// Don't allow the context to initialize a second time after it's already been explicitly uninitialized.
m_isAudioThreadFinished = true;
if (!isOfflineContext()) {
ASSERT(s_hardwareContextCount);
--s_hardwareContextCount;
// Offline contexts move to 'Closed' state when dispatching the completion event.
setState(State::Closed);
}
// Get rid of the sources which may still be playing.
derefUnfinishedSourceNodes();
m_isInitialized = false;
}
bool AudioContext::isInitialized() const
{
return m_isInitialized;
}
void AudioContext::addReaction(State state, DOMPromiseDeferred<void>&& promise)
{
size_t stateIndex = static_cast<size_t>(state);
if (stateIndex >= m_stateReactions.size())
m_stateReactions.grow(stateIndex + 1);
m_stateReactions[stateIndex].append(WTFMove(promise));
}
void AudioContext::setState(State state)
{
if (m_state == state)
return;
m_state = state;
m_eventQueue->enqueueEvent(Event::create(eventNames().statechangeEvent, Event::CanBubble::Yes, Event::IsCancelable::No));
size_t stateIndex = static_cast<size_t>(state);
if (stateIndex >= m_stateReactions.size())
return;
Vector<DOMPromiseDeferred<void>> reactions;
m_stateReactions[stateIndex].swap(reactions);
for (auto& promise : reactions)
promise.resolve();
}
void AudioContext::stop()
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
// Usually ScriptExecutionContext calls stop twice.
if (m_isStopScheduled)
return;
m_isStopScheduled = true;
ASSERT(document());
document()->updateIsPlayingMedia();
uninitialize();
clear();
}
bool AudioContext::canSuspendForDocumentSuspension() const
{
// FIXME: We should be able to suspend while rendering as well with some more code.
return m_state == State::Suspended || m_state == State::Closed;
}
const char* AudioContext::activeDOMObjectName() const
{
return "AudioContext";
}
Document* AudioContext::document() const
{
return downcast<Document>(m_scriptExecutionContext);
}
Document* AudioContext::hostingDocument() const
{
return downcast<Document>(m_scriptExecutionContext);
}
String AudioContext::sourceApplicationIdentifier() const
{
Document* document = this->document();
if (Frame* frame = document ? document->frame() : nullptr) {
if (NetworkingContext* networkingContext = frame->loader().networkingContext())
return networkingContext->sourceApplicationIdentifier();
}
return emptyString();
}
bool AudioContext::processingUserGestureForMedia() const
{
return document() ? document()->processingUserGestureForMedia() : false;
}
bool AudioContext::isSuspended() const
{
return !document() || document()->activeDOMObjectsAreSuspended() || document()->activeDOMObjectsAreStopped();
}
void AudioContext::visibilityStateChanged()
{
// Do not suspend if audio is audible.
if (!document() || mediaState() == MediaProducer::IsPlayingAudio || m_isStopScheduled)
return;
if (document()->hidden()) {
if (state() == State::Running) {
RELEASE_LOG_IF_ALLOWED("visibilityStateChanged() Suspending playback after going to the background");
m_mediaSession->beginInterruption(PlatformMediaSession::EnteringBackground);
}
} else {
if (state() == State::Interrupted) {
RELEASE_LOG_IF_ALLOWED("visibilityStateChanged() Resuming playback after entering foreground");
m_mediaSession->endInterruption(PlatformMediaSession::MayResumePlaying);
}
}
}
bool AudioContext::wouldTaintOrigin(const URL& url) const
{
if (url.protocolIsData())
return false;
if (auto* document = this->document())
return !document->securityOrigin().canRequest(url);
return false;
}
ExceptionOr<Ref<AudioBuffer>> AudioContext::createBuffer(unsigned numberOfChannels, size_t numberOfFrames, float sampleRate)
{
auto audioBuffer = AudioBuffer::create(numberOfChannels, numberOfFrames, sampleRate);
if (!audioBuffer)
return Exception { NotSupportedError };
return audioBuffer.releaseNonNull();
}
ExceptionOr<Ref<AudioBuffer>> AudioContext::createBuffer(ArrayBuffer& arrayBuffer, bool mixToMono)
{
auto audioBuffer = AudioBuffer::createFromAudioFileData(arrayBuffer.data(), arrayBuffer.byteLength(), mixToMono, sampleRate());
if (!audioBuffer)
return Exception { SyntaxError };
return audioBuffer.releaseNonNull();
}
void AudioContext::decodeAudioData(Ref<ArrayBuffer>&& audioData, RefPtr<AudioBufferCallback>&& successCallback, RefPtr<AudioBufferCallback>&& errorCallback)
{
if (!m_audioDecoder)
m_audioDecoder = makeUnique<AsyncAudioDecoder>();
m_audioDecoder->decodeAsync(WTFMove(audioData), sampleRate(), WTFMove(successCallback), WTFMove(errorCallback));
}
ExceptionOr<Ref<AudioBufferSourceNode>> AudioContext::createBufferSource()
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled)
return Exception { InvalidStateError };
lazyInitialize();
Ref<AudioBufferSourceNode> node = AudioBufferSourceNode::create(*this, sampleRate());
// Because this is an AudioScheduledSourceNode, the context keeps a reference until it has finished playing.
// When this happens, AudioScheduledSourceNode::finish() calls AudioContext::notifyNodeFinishedProcessing().
refNode(node);
return node;
}
#if ENABLE(VIDEO)
ExceptionOr<Ref<MediaElementAudioSourceNode>> AudioContext::createMediaElementSource(HTMLMediaElement& mediaElement)
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled || mediaElement.audioSourceNode())
return Exception { InvalidStateError };
lazyInitialize();
auto node = MediaElementAudioSourceNode::create(*this, mediaElement);
mediaElement.setAudioSourceNode(node.ptr());
refNode(node.get()); // context keeps reference until node is disconnected
return node;
}
#endif
#if ENABLE(MEDIA_STREAM)
ExceptionOr<Ref<MediaStreamAudioSourceNode>> AudioContext::createMediaStreamSource(MediaStream& mediaStream)
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled)
return Exception { InvalidStateError };
auto audioTracks = mediaStream.getAudioTracks();
if (audioTracks.isEmpty())
return Exception { InvalidStateError };
MediaStreamTrack* providerTrack = nullptr;
for (auto& track : audioTracks) {
if (track->audioSourceProvider()) {
providerTrack = track.get();
break;
}
}
if (!providerTrack)
return Exception { InvalidStateError };
lazyInitialize();
auto node = MediaStreamAudioSourceNode::create(*this, mediaStream, *providerTrack);
node->setFormat(2, sampleRate());
refNode(node); // context keeps reference until node is disconnected
return node;
}
ExceptionOr<Ref<MediaStreamAudioDestinationNode>> AudioContext::createMediaStreamDestination()
{
if (m_isStopScheduled)
return Exception { InvalidStateError };
// FIXME: Add support for an optional argument which specifies the number of channels.
// FIXME: The default should probably be stereo instead of mono.
return MediaStreamAudioDestinationNode::create(*this, 1);
}
#endif
ExceptionOr<Ref<ScriptProcessorNode>> AudioContext::createScriptProcessor(size_t bufferSize, size_t numberOfInputChannels, size_t numberOfOutputChannels)
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled)
return Exception { InvalidStateError };
lazyInitialize();
// W3C Editor's Draft 06 June 2017
// https://webaudio.github.io/web-audio-api/#widl-BaseAudioContext-createScriptProcessor-ScriptProcessorNode-unsigned-long-bufferSize-unsigned-long-numberOfInputChannels-unsigned-long-numberOfOutputChannels
// The bufferSize parameter determines the buffer size in units of sample-frames. If it's not passed in,
// or if the value is 0, then the implementation will choose the best buffer size for the given environment,
// which will be constant power of 2 throughout the lifetime of the node. ... If the value of this parameter
// is not one of the allowed power-of-2 values listed above, an IndexSizeError must be thrown.
switch (bufferSize) {
case 0:
#if USE(AUDIO_SESSION)
// Pick a value between 256 (2^8) and 16384 (2^14), based on the buffer size of the current AudioSession:
bufferSize = 1 << std::max<size_t>(8, std::min<size_t>(14, std::log2(AudioSession::sharedSession().bufferSize())));
#else
bufferSize = 2048;
#endif
break;
case 256:
case 512:
case 1024:
case 2048:
case 4096:
case 8192:
case 16384:
break;
default:
return Exception { IndexSizeError };
}
// An IndexSizeError exception must be thrown if bufferSize or numberOfInputChannels or numberOfOutputChannels
// are outside the valid range. It is invalid for both numberOfInputChannels and numberOfOutputChannels to be zero.
// In this case an IndexSizeError must be thrown.
if (!numberOfInputChannels && !numberOfOutputChannels)
return Exception { NotSupportedError };
// This parameter [numberOfInputChannels] determines the number of channels for this node's input. Values of
// up to 32 must be supported. A NotSupportedError must be thrown if the number of channels is not supported.
if (numberOfInputChannels > maxNumberOfChannels())
return Exception { NotSupportedError };
// This parameter [numberOfOutputChannels] determines the number of channels for this node's output. Values of
// up to 32 must be supported. A NotSupportedError must be thrown if the number of channels is not supported.
if (numberOfOutputChannels > maxNumberOfChannels())
return Exception { NotSupportedError };
auto node = ScriptProcessorNode::create(*this, sampleRate(), bufferSize, numberOfInputChannels, numberOfOutputChannels);
refNode(node); // context keeps reference until we stop making javascript rendering callbacks
return node;
}
ExceptionOr<Ref<BiquadFilterNode>> AudioContext::createBiquadFilter()
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled)
return Exception { InvalidStateError };
lazyInitialize();
return BiquadFilterNode::create(*this, sampleRate());
}
ExceptionOr<Ref<WaveShaperNode>> AudioContext::createWaveShaper()
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled)
return Exception { InvalidStateError };
lazyInitialize();
return WaveShaperNode::create(*this);
}
ExceptionOr<Ref<PannerNode>> AudioContext::createPanner()
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled)
return Exception { InvalidStateError };
lazyInitialize();
return PannerNode::create(*this, sampleRate());
}
ExceptionOr<Ref<ConvolverNode>> AudioContext::createConvolver()
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled)
return Exception { InvalidStateError };
lazyInitialize();
return ConvolverNode::create(*this, sampleRate());
}
ExceptionOr<Ref<DynamicsCompressorNode>> AudioContext::createDynamicsCompressor()
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled)
return Exception { InvalidStateError };
lazyInitialize();
return DynamicsCompressorNode::create(*this, sampleRate());
}
ExceptionOr<Ref<AnalyserNode>> AudioContext::createAnalyser()
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled)
return Exception { InvalidStateError };
lazyInitialize();
return AnalyserNode::create(*this, sampleRate());
}
ExceptionOr<Ref<GainNode>> AudioContext::createGain()
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled)
return Exception { InvalidStateError };
lazyInitialize();
return GainNode::create(*this, sampleRate());
}
ExceptionOr<Ref<DelayNode>> AudioContext::createDelay(double maxDelayTime)
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled)
return Exception { InvalidStateError };
lazyInitialize();
return DelayNode::create(*this, sampleRate(), maxDelayTime);
}
ExceptionOr<Ref<ChannelSplitterNode>> AudioContext::createChannelSplitter(size_t numberOfOutputs)
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled)
return Exception { InvalidStateError };
lazyInitialize();
auto node = ChannelSplitterNode::create(*this, sampleRate(), numberOfOutputs);
if (!node)
return Exception { IndexSizeError };
return node.releaseNonNull();
}
ExceptionOr<Ref<ChannelMergerNode>> AudioContext::createChannelMerger(size_t numberOfInputs)
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled)
return Exception { InvalidStateError };
lazyInitialize();
auto node = ChannelMergerNode::create(*this, sampleRate(), numberOfInputs);
if (!node)
return Exception { IndexSizeError };
return node.releaseNonNull();
}
ExceptionOr<Ref<OscillatorNode>> AudioContext::createOscillator()
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled)
return Exception { InvalidStateError };
lazyInitialize();
Ref<OscillatorNode> node = OscillatorNode::create(*this, sampleRate());
// Because this is an AudioScheduledSourceNode, the context keeps a reference until it has finished playing.
// When this happens, AudioScheduledSourceNode::finish() calls AudioContext::notifyNodeFinishedProcessing().
refNode(node);
return node;
}
ExceptionOr<Ref<PeriodicWave>> AudioContext::createPeriodicWave(Float32Array& real, Float32Array& imaginary)
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
if (m_isStopScheduled)
return Exception { InvalidStateError };
if (real.length() != imaginary.length() || (real.length() > MaxPeriodicWaveLength) || !real.length())
return Exception { IndexSizeError };
lazyInitialize();
return PeriodicWave::create(sampleRate(), real, imaginary);
}
void AudioContext::notifyNodeFinishedProcessing(AudioNode* node)
{
ASSERT(isAudioThread());
m_finishedNodes.append(node);
}
void AudioContext::derefFinishedSourceNodes()
{
ASSERT(isGraphOwner());
ASSERT(isAudioThread() || isAudioThreadFinished());
for (auto& node : m_finishedNodes)
derefNode(*node);
m_finishedNodes.clear();
}
void AudioContext::refNode(AudioNode& node)
{
ASSERT(isMainThread());
AutoLocker locker(*this);
node.ref(AudioNode::RefTypeConnection);
m_referencedNodes.append(&node);
}
void AudioContext::derefNode(AudioNode& node)
{
ASSERT(isGraphOwner());
node.deref(AudioNode::RefTypeConnection);
ASSERT(m_referencedNodes.contains(&node));
m_referencedNodes.removeFirst(&node);
}
void AudioContext::derefUnfinishedSourceNodes()
{
ASSERT(isMainThread() && isAudioThreadFinished());
for (auto& node : m_referencedNodes)
node->deref(AudioNode::RefTypeConnection);
m_referencedNodes.clear();
}
void AudioContext::lock(bool& mustReleaseLock)
{
// Don't allow regular lock in real-time audio thread.
ASSERT(isMainThread());
Thread& thisThread = Thread::current();
if (&thisThread == m_graphOwnerThread) {
// We already have the lock.
mustReleaseLock = false;
} else {
// Acquire the lock.
m_contextGraphMutex.lock();
m_graphOwnerThread = &thisThread;
mustReleaseLock = true;
}
}
bool AudioContext::tryLock(bool& mustReleaseLock)
{
Thread& thisThread = Thread::current();
bool isAudioThread = &thisThread == audioThread();
// Try to catch cases of using try lock on main thread - it should use regular lock.
ASSERT(isAudioThread || isAudioThreadFinished());
if (!isAudioThread) {
// In release build treat tryLock() as lock() (since above ASSERT(isAudioThread) never fires) - this is the best we can do.
lock(mustReleaseLock);
return true;
}
bool hasLock;
if (&thisThread == m_graphOwnerThread) {
// Thread already has the lock.
hasLock = true;
mustReleaseLock = false;
} else {
// Don't already have the lock - try to acquire it.
hasLock = m_contextGraphMutex.tryLock();
if (hasLock)
m_graphOwnerThread = &thisThread;
mustReleaseLock = hasLock;
}
return hasLock;
}
void AudioContext::unlock()
{
ASSERT(m_graphOwnerThread == &Thread::current());
m_graphOwnerThread = nullptr;
m_contextGraphMutex.unlock();
}
bool AudioContext::isAudioThread() const
{
return m_audioThread == &Thread::current();
}
bool AudioContext::isGraphOwner() const
{
return m_graphOwnerThread == &Thread::current();
}
void AudioContext::addDeferredFinishDeref(AudioNode* node)
{
ASSERT(isAudioThread());
m_deferredFinishDerefList.append(node);
}
void AudioContext::handlePreRenderTasks()
{
ASSERT(isAudioThread());
// At the beginning of every render quantum, try to update the internal rendering graph state (from main thread changes).
// It's OK if the tryLock() fails, we'll just take slightly longer to pick up the changes.
bool mustReleaseLock;
if (tryLock(mustReleaseLock)) {
// Fixup the state of any dirty AudioSummingJunctions and AudioNodeOutputs.
handleDirtyAudioSummingJunctions();
handleDirtyAudioNodeOutputs();
updateAutomaticPullNodes();
if (mustReleaseLock)
unlock();
}
}
void AudioContext::handlePostRenderTasks()
{
ASSERT(isAudioThread());
// Must use a tryLock() here too. Don't worry, the lock will very rarely be contended and this method is called frequently.
// The worst that can happen is that there will be some nodes which will take slightly longer than usual to be deleted or removed
// from the render graph (in which case they'll render silence).
bool mustReleaseLock;
if (tryLock(mustReleaseLock)) {
// Take care of finishing any derefs where the tryLock() failed previously.
handleDeferredFinishDerefs();
// Dynamically clean up nodes which are no longer needed.
derefFinishedSourceNodes();
// Don't delete in the real-time thread. Let the main thread do it.
// Ref-counted objects held by certain AudioNodes may not be thread-safe.
scheduleNodeDeletion();
// Fixup the state of any dirty AudioSummingJunctions and AudioNodeOutputs.
handleDirtyAudioSummingJunctions();
handleDirtyAudioNodeOutputs();
updateAutomaticPullNodes();
if (mustReleaseLock)
unlock();
}
}
void AudioContext::handleDeferredFinishDerefs()
{
ASSERT(isAudioThread() && isGraphOwner());
for (auto& node : m_deferredFinishDerefList)
node->finishDeref(AudioNode::RefTypeConnection);
m_deferredFinishDerefList.clear();
}
void AudioContext::markForDeletion(AudioNode& node)
{
ASSERT(isGraphOwner());
if (isAudioThreadFinished())
m_nodesToDelete.append(&node);
else
m_nodesMarkedForDeletion.append(&node);
// This is probably the best time for us to remove the node from automatic pull list,
// since all connections are gone and we hold the graph lock. Then when handlePostRenderTasks()
// gets a chance to schedule the deletion work, updateAutomaticPullNodes() also gets a chance to
// modify m_renderingAutomaticPullNodes.
removeAutomaticPullNode(node);
}
void AudioContext::scheduleNodeDeletion()
{
bool isGood = m_isInitialized && isGraphOwner();
ASSERT(isGood);
if (!isGood)
return;
// Make sure to call deleteMarkedNodes() on main thread.
if (m_nodesMarkedForDeletion.size() && !m_isDeletionScheduled) {
m_nodesToDelete.appendVector(m_nodesMarkedForDeletion);
m_nodesMarkedForDeletion.clear();
m_isDeletionScheduled = true;
callOnMainThread([protectedThis = makeRef(*this)]() mutable {
protectedThis->deleteMarkedNodes();
});
}
}
void AudioContext::deleteMarkedNodes()
{
ASSERT(isMainThread());
// Protect this object from being deleted before we release the mutex locked by AutoLocker.
Ref<AudioContext> protectedThis(*this);
{
AutoLocker locker(*this);
while (m_nodesToDelete.size()) {
AudioNode* node = m_nodesToDelete.takeLast();
// Before deleting the node, clear out any AudioNodeInputs from m_dirtySummingJunctions.
unsigned numberOfInputs = node->numberOfInputs();
for (unsigned i = 0; i < numberOfInputs; ++i)
m_dirtySummingJunctions.remove(node->input(i));
// Before deleting the node, clear out any AudioNodeOutputs from m_dirtyAudioNodeOutputs.
unsigned numberOfOutputs = node->numberOfOutputs();
for (unsigned i = 0; i < numberOfOutputs; ++i)
m_dirtyAudioNodeOutputs.remove(node->output(i));
// Finally, delete it.
delete node;
}
m_isDeletionScheduled = false;
}
}
void AudioContext::markSummingJunctionDirty(AudioSummingJunction* summingJunction)
{
ASSERT(isGraphOwner());
m_dirtySummingJunctions.add(summingJunction);
}
void AudioContext::removeMarkedSummingJunction(AudioSummingJunction* summingJunction)
{
ASSERT(isMainThread());
AutoLocker locker(*this);
m_dirtySummingJunctions.remove(summingJunction);
}
void AudioContext::markAudioNodeOutputDirty(AudioNodeOutput* output)
{
ASSERT(isGraphOwner());
m_dirtyAudioNodeOutputs.add(output);
}
void AudioContext::handleDirtyAudioSummingJunctions()
{
ASSERT(isGraphOwner());
for (auto& junction : m_dirtySummingJunctions)
junction->updateRenderingState();
m_dirtySummingJunctions.clear();
}
void AudioContext::handleDirtyAudioNodeOutputs()
{
ASSERT(isGraphOwner());
for (auto& output : m_dirtyAudioNodeOutputs)
output->updateRenderingState();
m_dirtyAudioNodeOutputs.clear();
}
void AudioContext::addAutomaticPullNode(AudioNode& node)
{
ASSERT(isGraphOwner());
if (m_automaticPullNodes.add(&node).isNewEntry)
m_automaticPullNodesNeedUpdating = true;
}
void AudioContext::removeAutomaticPullNode(AudioNode& node)
{
ASSERT(isGraphOwner());
if (m_automaticPullNodes.remove(&node))
m_automaticPullNodesNeedUpdating = true;
}
void AudioContext::updateAutomaticPullNodes()
{
ASSERT(isGraphOwner());
if (m_automaticPullNodesNeedUpdating) {
// Copy from m_automaticPullNodes to m_renderingAutomaticPullNodes.
m_renderingAutomaticPullNodes.resize(m_automaticPullNodes.size());
unsigned i = 0;
for (auto& output : m_automaticPullNodes)
m_renderingAutomaticPullNodes[i++] = output;
m_automaticPullNodesNeedUpdating = false;
}
}
void AudioContext::processAutomaticPullNodes(size_t framesToProcess)
{
ASSERT(isAudioThread());
for (auto& node : m_renderingAutomaticPullNodes)
node->processIfNecessary(framesToProcess);
}
ScriptExecutionContext* AudioContext::scriptExecutionContext() const
{
return ActiveDOMObject::scriptExecutionContext();
}
void AudioContext::nodeWillBeginPlayback()
{
// Called by scheduled AudioNodes when clients schedule their start times.
// Prior to the introduction of suspend(), resume(), and stop(), starting
// a scheduled AudioNode would remove the user-gesture restriction, if present,
// and would thus unmute the context. Now that AudioContext stays in the
// "suspended" state if a user-gesture restriction is present, starting a
// schedule AudioNode should set the state to "running", but only if the
// user-gesture restriction is set.
if (userGestureRequiredForAudioStart())
startRendering();
}
bool AudioContext::willBeginPlayback()
{
if (!document())
return false;
if (userGestureRequiredForAudioStart()) {
if (!processingUserGestureForMedia() && !document()->isCapturing()) {
ALWAYS_LOG(LOGIDENTIFIER, "returning false, not processing user gesture or capturing");
return false;
}
removeBehaviorRestriction(AudioContext::RequireUserGestureForAudioStartRestriction);
}
if (pageConsentRequiredForAudioStart()) {
Page* page = document()->page();
if (page && !page->canStartMedia()) {
document()->addMediaCanStartListener(*this);
ALWAYS_LOG(LOGIDENTIFIER, "returning false, page doesn't allow media to start");
return false;
}
removeBehaviorRestriction(AudioContext::RequirePageConsentForAudioStartRestriction);
}
auto willBegin = m_mediaSession->clientWillBeginPlayback();
ALWAYS_LOG(LOGIDENTIFIER, "returning ", willBegin);
return willBegin;
}
bool AudioContext::willPausePlayback()
{
if (!document())
return false;
if (userGestureRequiredForAudioStart()) {
if (!processingUserGestureForMedia())
return false;
removeBehaviorRestriction(AudioContext::RequireUserGestureForAudioStartRestriction);
}
if (pageConsentRequiredForAudioStart()) {
Page* page = document()->page();
if (page && !page->canStartMedia()) {
document()->addMediaCanStartListener(*this);
return false;
}
removeBehaviorRestriction(AudioContext::RequirePageConsentForAudioStartRestriction);
}
return m_mediaSession->clientWillPausePlayback();
}
void AudioContext::startRendering()
{
ALWAYS_LOG(LOGIDENTIFIER);
if (m_isStopScheduled || !willBeginPlayback())
return;
makePendingActivity();
destination()->startRendering();
setState(State::Running);
}
void AudioContext::mediaCanStart(Document& document)
{
ASSERT_UNUSED(document, &document == this->document());
removeBehaviorRestriction(AudioContext::RequirePageConsentForAudioStartRestriction);
mayResumePlayback(true);
}
MediaProducer::MediaStateFlags AudioContext::mediaState() const
{
if (!m_isStopScheduled && m_destinationNode && m_destinationNode->isPlayingAudio())
return MediaProducer::IsPlayingAudio;
return MediaProducer::IsNotPlaying;
}
void AudioContext::pageMutedStateDidChange()
{
if (m_destinationNode && document() && document()->page())
m_destinationNode->setMuted(document()->page()->isAudioMuted());
}
void AudioContext::isPlayingAudioDidChange()
{
// Make sure to call Document::updateIsPlayingMedia() on the main thread, since
// we could be on the audio I/O thread here and the call into WebCore could block.
callOnMainThread([protectedThis = makeRef(*this)] {
if (protectedThis->document())
protectedThis->document()->updateIsPlayingMedia();
});
}
void AudioContext::finishedRendering(bool didRendering)
{
ASSERT(isOfflineContext());
ASSERT(isMainThread());
if (!isMainThread())
return;
auto clearPendingActivityIfExitEarly = WTF::makeScopeExit([this] {
clearPendingActivity();
});
ALWAYS_LOG(LOGIDENTIFIER);
if (!didRendering)
return;
AudioBuffer* renderedBuffer = m_renderTarget.get();
setState(State::Closed);
ASSERT(renderedBuffer);
if (!renderedBuffer)
return;
// Avoid firing the event if the document has already gone away.
if (m_isStopScheduled)
return;
clearPendingActivityIfExitEarly.release();
m_eventQueue->enqueueEvent(OfflineAudioCompletionEvent::create(renderedBuffer));
}
void AudioContext::dispatchEvent(Event& event)
{
EventTarget::dispatchEvent(event);
if (event.eventInterface() == OfflineAudioCompletionEventInterfaceType)
clearPendingActivity();
}
void AudioContext::incrementActiveSourceCount()
{
++m_activeSourceCount;
}
void AudioContext::decrementActiveSourceCount()
{
--m_activeSourceCount;
}
void AudioContext::suspend(DOMPromiseDeferred<void>&& promise)
{
if (isOfflineContext() || m_isStopScheduled) {
promise.reject(InvalidStateError);
return;
}
if (m_state == State::Suspended) {
promise.resolve();
return;
}
if (m_state == State::Closed || m_state == State::Interrupted || !m_destinationNode) {
promise.reject();
return;
}
addReaction(State::Suspended, WTFMove(promise));
if (!willPausePlayback())
return;
lazyInitialize();
m_destinationNode->suspend([this, protectedThis = makeRef(*this)] {
setState(State::Suspended);
});
}
void AudioContext::resume(DOMPromiseDeferred<void>&& promise)
{
if (isOfflineContext() || m_isStopScheduled) {
promise.reject(InvalidStateError);
return;
}
if (m_state == State::Running) {
promise.resolve();
return;
}
if (m_state == State::Closed || !m_destinationNode) {
promise.reject();
return;
}
addReaction(State::Running, WTFMove(promise));
if (!willBeginPlayback())
return;
lazyInitialize();
m_destinationNode->resume([this, protectedThis = makeRef(*this)] {
setState(State::Running);
});
}
void AudioContext::close(DOMPromiseDeferred<void>&& promise)
{
if (isOfflineContext() || m_isStopScheduled) {
promise.reject(InvalidStateError);
return;
}
if (m_state == State::Closed || !m_destinationNode) {
promise.resolve();
return;
}
addReaction(State::Closed, WTFMove(promise));
lazyInitialize();
m_destinationNode->close([this, protectedThis = makeRef(*this)] {
setState(State::Closed);
uninitialize();
});
}
void AudioContext::suspendPlayback()
{
if (!m_destinationNode || m_state == State::Closed)
return;
if (m_state == State::Suspended) {
if (m_mediaSession->state() == PlatformMediaSession::Interrupted)
setState(State::Interrupted);
return;
}
lazyInitialize();
m_destinationNode->suspend([this, protectedThis = makeRef(*this)] {
bool interrupted = m_mediaSession->state() == PlatformMediaSession::Interrupted;
setState(interrupted ? State::Interrupted : State::Suspended);
});
}
void AudioContext::mayResumePlayback(bool shouldResume)
{
if (!m_destinationNode || m_state == State::Closed || m_state == State::Running)
return;
if (!shouldResume) {
setState(State::Suspended);
return;
}
if (!willBeginPlayback())
return;
lazyInitialize();
m_destinationNode->resume([this, protectedThis = makeRef(*this)] {
setState(State::Running);
});
}
void AudioContext::postTask(WTF::Function<void()>&& task)
{
if (m_isStopScheduled)
return;
m_scriptExecutionContext->postTask(WTFMove(task));
}
const SecurityOrigin* AudioContext::origin() const
{
return m_scriptExecutionContext ? m_scriptExecutionContext->securityOrigin() : nullptr;
}
void AudioContext::addConsoleMessage(MessageSource source, MessageLevel level, const String& message)
{
if (m_scriptExecutionContext)
m_scriptExecutionContext->addConsoleMessage(source, level, message);
}
void AudioContext::clearPendingActivity()
{
if (!m_pendingActivity)
return;
m_pendingActivity = nullptr;
// FIXME: Remove this specific deref() and ref() call in makePendingActivity().
deref();
}
void AudioContext::makePendingActivity()
{
if (m_pendingActivity)
return;
m_pendingActivity = ActiveDOMObject::makePendingActivity(*this);
ref();
}
#if !RELEASE_LOG_DISABLED
WTFLogChannel& AudioContext::logChannel() const
{
return LogMedia;
}
#endif
} // namespace WebCore
#endif // ENABLE(WEB_AUDIO)