Media: harden TrackID handling.

9 files changed, 183 insertions, 37 deletions

diff --git a/dom/html/HTMLMediaElement.cpp b/dom/html/HTMLMediaElement.cpp
index 1f1a545faf..0b9f606f19 100644
--- a/dom/html/HTMLMediaElement.cpp
+++ b/dom/html/HTMLMediaElement.cpp
@@ -817,6 +817,7 @@ NS_IMPL_CYCLE_COLLECTION_TRAVERSE_BEGIN_INHERITED(HTMLMediaElement, nsGenericHTM
   NS_IMPL_CYCLE_COLLECTION_TRAVERSE(mErrorSink->mError)
   for (uint32_t i = 0; i < tmp->mOutputStreams.Length(); ++i) {
     NS_IMPL_CYCLE_COLLECTION_TRAVERSE(mOutputStreams[i].mStream);
+    NS_IMPL_CYCLE_COLLECTION_TRAVERSE(mOutputStreams[i].mPreCreatedTracks);
   }
   NS_IMPL_CYCLE_COLLECTION_TRAVERSE(mPlayed);
   NS_IMPL_CYCLE_COLLECTION_TRAVERSE(mTextTrackManager)
@@ -1020,6 +1021,17 @@ void HTMLMediaElement::ShutdownDecoder()
   RemoveMediaElementFromURITable();
   NS_ASSERTION(mDecoder, "Must have decoder to shut down");
   mWaitingForKeyListener.DisconnectIfExists();
+  for (OutputMediaStream& out : mOutputStreams) {
+    if (!out.mCapturingDecoder) {
+      continue;
+    }
+    if (!out.mStream) {
+      continue;
+    }
+    out.mNextAvailableTrackID = std::max<TrackID>(
+      mDecoder->NextAvailableTrackIDFor(out.mStream->GetInputStream()),
+      out.mNextAvailableTrackID);
+  }
   mDecoder->Shutdown();
   mDecoder = nullptr;
 }
@@ -2730,6 +2742,7 @@ HTMLMediaElement::CaptureStreamInternal(bool aFinishWhenEnded,
   if (mDecoder) {
     out->mCapturingDecoder = true;
     mDecoder->AddOutputStream(out->mStream->GetInputStream()->AsProcessedStream(),
+                              out->mNextAvailableTrackID,
                               aFinishWhenEnded);
   } else if (mSrcStream) {
     out->mCapturingMediaStream = true;
@@ -2743,23 +2756,26 @@ HTMLMediaElement::CaptureStreamInternal(bool aFinishWhenEnded,
 
   if (mDecoder) {
     if (HasAudio()) {
-      TrackID audioTrackId = mMediaInfo.mAudio.mTrackId;
+      TrackID audioTrackId = out->mNextAvailableTrackID++;
       RefPtr<MediaStreamTrackSource> trackSource =
         getter->GetMediaStreamTrackSource(audioTrackId);
       RefPtr<MediaStreamTrack> track =
-        out->mStream->CreateDOMTrack(audioTrackId, MediaSegment::AUDIO,
+        out->mStream->CreateDOMTrack(audioTrackId,
+                                     MediaSegment::AUDIO,
                                      trackSource);
+      out->mPreCreatedTracks.AppendElement(track);
       out->mStream->AddTrackInternal(track);
       LOG(LogLevel::Debug,
           ("Created audio track %d for captured decoder", audioTrackId));
     }
     if (IsVideo() && HasVideo() && !out->mCapturingAudioOnly) {
-      TrackID videoTrackId = mMediaInfo.mVideo.mTrackId;
+      TrackID videoTrackId = out->mNextAvailableTrackID++;
       RefPtr<MediaStreamTrackSource> trackSource =
         getter->GetMediaStreamTrackSource(videoTrackId);
       RefPtr<MediaStreamTrack> track =
         out->mStream->CreateDOMTrack(videoTrackId, MediaSegment::VIDEO,
                                      trackSource);
+      out->mPreCreatedTracks.AppendElement(track);
       out->mStream->AddTrackInternal(track);
       LOG(LogLevel::Debug,
           ("Created video track %d for captured decoder", videoTrackId));
@@ -2848,6 +2864,25 @@ NS_IMETHODIMP HTMLMediaElement::GetMozAudioCaptured(bool* aCaptured)
   return NS_OK;
 }
 
+void
+HTMLMediaElement::EndPreCreatedCapturedDecoderTracks()
+{
+  MOZ_ASSERT(NS_IsMainThread());
+  for (OutputMediaStream& ms : mOutputStreams) {
+    if (!ms.mCapturingDecoder) {
+      continue;
+    }
+    for (RefPtr<MediaStreamTrack>& t : ms.mPreCreatedTracks) {
+      if (t->Ended()) {
+        continue;
+      }
+      NS_DispatchToMainThread(NewRunnableMethod(
+        t, &MediaStreamTrack::OverrideEnded));
+    }
+    ms.mPreCreatedTracks.Clear();
+  }
+}
+
 class MediaElementSetForURI : public nsURIHashKey {
 public:
   explicit MediaElementSetForURI(const nsIURI* aKey) : nsURIHashKey(aKey) {}
@@ -3380,11 +3415,12 @@ HTMLMediaElement::WakeLockRelease()
 }
 
 HTMLMediaElement::OutputMediaStream::OutputMediaStream()
-  : mFinishWhenEnded(false)
+  : mNextAvailableTrackID(1)
+  , mFinishWhenEnded(false)
   , mCapturingAudioOnly(false)
   , mCapturingDecoder(false)
   , mCapturingMediaStream(false)
-  , mNextAvailableTrackID(1) {}
+{}
 
 HTMLMediaElement::OutputMediaStream::~OutputMediaStream()
 {
@@ -4008,6 +4044,7 @@ nsresult HTMLMediaElement::FinishDecoderSetup(MediaDecoder* aDecoder,
 
     ms.mCapturingDecoder = true;
     aDecoder->AddOutputStream(ms.mStream->GetInputStream()->AsProcessedStream(),
+                              ms.mNextAvailableTrackID,
                               ms.mFinishWhenEnded);
   }
 
diff --git a/dom/html/HTMLMediaElement.h b/dom/html/HTMLMediaElement.h
index b65049206d..af944a3187 100644
--- a/dom/html/HTMLMediaElement.h
+++ b/dom/html/HTMLMediaElement.h
@@ -675,6 +675,11 @@ public:
     return mAudioCaptured;
   }
 
+  /**
+   * Ensures any MediaStreamTracks captured from a MediaDecoder are ended.
+   */
+  void EndPreCreatedCapturedDecoderTracks();
+
   void MozGetMetadata(JSContext* aCx, JS::MutableHandle<JSObject*> aResult,
                       ErrorResult& aRv);
 
@@ -815,13 +820,18 @@ protected:
     ~OutputMediaStream();
 
     RefPtr<DOMMediaStream> mStream;
+    TrackID mNextAvailableTrackID;
     bool mFinishWhenEnded;
     bool mCapturingAudioOnly;
     bool mCapturingDecoder;
     bool mCapturingMediaStream;
 
+    // The following members are keeping state for a captured MediaDecoder.
+    // Tracks that were created on main thread before MediaDecoder fed them
+    // to the MediaStreamGraph.
+    nsTArray<RefPtr<MediaStreamTrack>> mPreCreatedTracks;
+
     // The following members are keeping state for a captured MediaStream.
-    TrackID mNextAvailableTrackID;
     nsTArray<Pair<nsString, RefPtr<MediaInputPort>>> mTrackPorts;
   };
 
diff --git a/dom/media/MediaDecoder.cpp b/dom/media/MediaDecoder.cpp
index ab39886ba7..9334d1bcb9 100644
--- a/dom/media/MediaDecoder.cpp
+++ b/dom/media/MediaDecoder.cpp
@@ -327,11 +327,13 @@ MediaDecoder::SetVolume(double aVolume)
 
 void
 MediaDecoder::AddOutputStream(ProcessedMediaStream* aStream,
+                              TrackID aNextAvailableTrackID,
                               bool aFinishWhenEnded)
 {
   MOZ_ASSERT(NS_IsMainThread());
   MOZ_ASSERT(mDecoderStateMachine, "Must be called after Load().");
-  mDecoderStateMachine->AddOutputStream(aStream, aFinishWhenEnded);
+  mDecoderStateMachine->AddOutputStream(
+    aStream, aNextAvailableTrackID, aFinishWhenEnded);
 }
 
 void
@@ -342,6 +344,14 @@ MediaDecoder::RemoveOutputStream(MediaStream* aStream)
   mDecoderStateMachine->RemoveOutputStream(aStream);
 }
 
+TrackID
+MediaDecoder::NextAvailableTrackIDFor(MediaStream* aOutputStream) const
+{
+  MOZ_ASSERT(NS_IsMainThread());
+  MOZ_ASSERT(mDecoderStateMachine, "Must be called after Load().");
+  return mDecoderStateMachine->NextAvailableTrackIDFor(aOutputStream);
+}
+
 double
 MediaDecoder::GetDuration()
 {
@@ -1736,6 +1746,8 @@ MediaDecoder::RemoveMediaTracks()
     videoList->RemoveTracks();
   }
 
+  element->EndPreCreatedCapturedDecoderTracks();
+
   mMediaTracksConstructed = false;
 }
 
diff --git a/dom/media/MediaDecoder.h b/dom/media/MediaDecoder.h
index 825bbbd2c0..a4edcbe727 100644
--- a/dom/media/MediaDecoder.h
+++ b/dom/media/MediaDecoder.h
@@ -210,9 +210,13 @@ public:
   // Add an output stream. All decoder output will be sent to the stream.
   // The stream is initially blocked. The decoder is responsible for unblocking
   // it while it is playing back.
-  virtual void AddOutputStream(ProcessedMediaStream* aStream, bool aFinishWhenEnded);
+  virtual void AddOutputStream(ProcessedMediaStream* aStream,
+                               TrackID aNextAvailableTrackID,
+                               bool aFinishWhenEnded);
   // Remove an output stream added with AddOutputStream.
   virtual void RemoveOutputStream(MediaStream* aStream);
+  // The next TrackID that can be used without risk of a collision.
+  virtual TrackID NextAvailableTrackIDFor(MediaStream* aOutputStream) const;
 
   // Return the duration of the video in seconds.
   virtual double GetDuration();
diff --git a/dom/media/MediaDecoderStateMachine.cpp b/dom/media/MediaDecoderStateMachine.cpp
index f13e59b6cf..c586139ad9 100644
--- a/dom/media/MediaDecoderStateMachine.cpp
+++ b/dom/media/MediaDecoderStateMachine.cpp
@@ -2835,6 +2835,10 @@ MediaDecoderStateMachine::FinishDecodeFirstFrame()
 RefPtr<ShutdownPromise>
 MediaDecoderStateMachine::BeginShutdown()
 {
+  MOZ_ASSERT(NS_IsMainThread());
+  if (mOutputStreamManager) {
+    mOutputStreamManager->Clear();
+  }
   return InvokeAsync(OwnerThread(), this, __func__,
                      &MediaDecoderStateMachine::Shutdown);
 }
@@ -3250,11 +3254,12 @@ MediaDecoderStateMachine::DumpDebugInfo()
 }
 
 void MediaDecoderStateMachine::AddOutputStream(ProcessedMediaStream* aStream,
+                                               TrackID aNextAvailableTrackID,
                                                bool aFinishWhenEnded)
 {
   MOZ_ASSERT(NS_IsMainThread());
   DECODER_LOG("AddOutputStream aStream=%p!", aStream);
-  mOutputStreamManager->Add(aStream, aFinishWhenEnded);
+  mOutputStreamManager->Add(aStream, aNextAvailableTrackID, aFinishWhenEnded);
   nsCOMPtr<nsIRunnable> r = NewRunnableMethod<bool>(
     this, &MediaDecoderStateMachine::SetAudioCaptured, true);
   OwnerThread()->Dispatch(r.forget());
@@ -3269,9 +3274,17 @@ void MediaDecoderStateMachine::RemoveOutputStream(MediaStream* aStream)
     nsCOMPtr<nsIRunnable> r = NewRunnableMethod<bool>(
       this, &MediaDecoderStateMachine::SetAudioCaptured, false);
     OwnerThread()->Dispatch(r.forget());
+
   }
 }
 
+TrackID
+MediaDecoderStateMachine::NextAvailableTrackIDFor(MediaStream* aOutputStream) const
+{
+  MOZ_ASSERT(NS_IsMainThread());
+  return mOutputStreamManager->NextAvailableTrackIDFor(aOutputStream);
+}
+
 size_t
 MediaDecoderStateMachine::SizeOfVideoQueue() const
 {
diff --git a/dom/media/MediaDecoderStateMachine.h b/dom/media/MediaDecoderStateMachine.h
index a61fe13d2a..ff3258ff1f 100644
--- a/dom/media/MediaDecoderStateMachine.h
+++ b/dom/media/MediaDecoderStateMachine.h
@@ -168,9 +168,12 @@ public:
 
   void DumpDebugInfo();
 
-  void AddOutputStream(ProcessedMediaStream* aStream, bool aFinishWhenEnded);
+  void AddOutputStream(ProcessedMediaStream* aStream,
+                       TrackID aNextAvailableTrackID,
+                       bool aFinishWhenEnded);
   // Remove an output stream added with AddOutputStream.
   void RemoveOutputStream(MediaStream* aStream);
+  TrackID NextAvailableTrackIDFor(MediaStream* aOutputStream) const;
 
   // Seeks to the decoder to aTarget asynchronously.
   RefPtr<MediaDecoder::SeekPromise> InvokeSeek(SeekTarget aTarget);
diff --git a/dom/media/mediasink/DecodedStream.cpp b/dom/media/mediasink/DecodedStream.cpp
index 9501a6cde9..00bc5ea493 100644
--- a/dom/media/mediasink/DecodedStream.cpp
+++ b/dom/media/mediasink/DecodedStream.cpp
@@ -181,17 +181,22 @@ DecodedStreamData::DecodedStreamData(OutputStreamManager* aOutputStreamManager,
   , mOutputStreamManager(aOutputStreamManager)
 {
   mStream->AddListener(mListener);
-  mOutputStreamManager->Connect(mStream);
+  TrackID audioTrack = TRACK_NONE;
+  TrackID videoTrack = TRACK_NONE;
 
   // Initialize tracks.
   if (aInit.mInfo.HasAudio()) {
-    mStream->AddAudioTrack(aInit.mInfo.mAudio.mTrackId,
+    audioTrack = aInit.mInfo.mAudio.mTrackId;
+    mStream->AddAudioTrack(audioTrack,
                            aInit.mInfo.mAudio.mRate,
                            0, new AudioSegment());
   }
   if (aInit.mInfo.HasVideo()) {
-    mStream->AddTrack(aInit.mInfo.mVideo.mTrackId, 0, new VideoSegment());
+    videoTrack = aInit.mInfo.mVideo.mTrackId;
+    mStream->AddTrack(videoTrack, 0, new VideoSegment());
   }
+
+  mOutputStreamManager->Connect(mStream, audioTrack, videoTrack);
 }
 
 DecodedStreamData::~DecodedStreamData()
diff --git a/dom/media/mediasink/OutputStreamManager.cpp b/dom/media/mediasink/OutputStreamManager.cpp
index d5685837a6..7ecc203ed8 100644
--- a/dom/media/mediasink/OutputStreamManager.cpp
+++ b/dom/media/mediasink/OutputStreamManager.cpp
@@ -13,29 +13,41 @@ OutputStreamData::~OutputStreamData()
 {
   MOZ_ASSERT(NS_IsMainThread());
   // Break the connection to the input stream if necessary.
-  if (mPort) {
-    mPort->Destroy();
+  for (RefPtr<MediaInputPort>& port : mPorts) {
+    port->Destroy();
   }
 }
 
 void
-OutputStreamData::Init(OutputStreamManager* aOwner, ProcessedMediaStream* aStream)
+OutputStreamData::Init(OutputStreamManager* aOwner,
+                       ProcessedMediaStream* aStream,
+                       TrackID aNextAvailableTrackID)
 {
   mOwner = aOwner;
   mStream = aStream;
+  mNextAvailableTrackID = aNextAvailableTrackID;
 }
 
 bool
-OutputStreamData::Connect(MediaStream* aStream)
+OutputStreamData::Connect(MediaStream* aStream,
+                          TrackID aInputAudioTrackID,
+                          TrackID aInputVideoTrackID)
 {
   MOZ_ASSERT(NS_IsMainThread());
-  MOZ_ASSERT(!mPort, "Already connected?");
+  MOZ_ASSERT(mPorts.IsEmpty(), "Already connected?");
 
   if (mStream->IsDestroyed()) {
     return false;
   }
 
-  mPort = mStream->AllocateInputPort(aStream);
+  for (TrackID tid : {aInputAudioTrackID, aInputVideoTrackID}) {
+    if (tid == TRACK_NONE) {
+      continue;
+    }
+    MOZ_ASSERT(IsTrackIDExplicit(tid));
+    mPorts.AppendElement(mStream->AllocateInputPort(
+        aStream, tid, mNextAvailableTrackID++));
+  }
   return true;
 }
 
@@ -51,11 +63,11 @@ OutputStreamData::Disconnect()
     return false;
   }
 
-  // Disconnect the existing port if necessary.
-  if (mPort) {
-    mPort->Destroy();
-    mPort = nullptr;
+  // Disconnect any existing port.
+  for (RefPtr<MediaInputPort>& port : mPorts) {
+    port->Destroy();
   }
+  mPorts.Clear();
   return true;
 }
 
@@ -71,8 +83,16 @@ OutputStreamData::Graph() const
   return mStream->Graph();
 }
 
+TrackID
+OutputStreamData::NextAvailableTrackID() const
+{
+  return mNextAvailableTrackID;
+}
+
 void
-OutputStreamManager::Add(ProcessedMediaStream* aStream, bool aFinishWhenEnded)
+OutputStreamManager::Add(ProcessedMediaStream* aStream,
+                         TrackID aNextAvailableTrackID,
+                         bool aFinishWhenEnded)
 {
   MOZ_ASSERT(NS_IsMainThread());
   // All streams must belong to the same graph.
@@ -84,12 +104,12 @@ OutputStreamManager::Add(ProcessedMediaStream* aStream, bool aFinishWhenEnded)
   }
 
   OutputStreamData* p = mStreams.AppendElement();
-  p->Init(this, aStream);
+  p->Init(this, aStream, aNextAvailableTrackID);
 
   // Connect to the input stream if we have one. Otherwise the output stream
   // will be connected in Connect().
   if (mInputStream) {
-    p->Connect(mInputStream);
+    p->Connect(mInputStream, mInputAudioTrackID, mInputVideoTrackID);
   }
 }
 
@@ -106,12 +126,35 @@ OutputStreamManager::Remove(MediaStream* aStream)
 }
 
 void
-OutputStreamManager::Connect(MediaStream* aStream)
+OutputStreamManager::Clear()
+{
+  MOZ_ASSERT(NS_IsMainThread());
+  mStreams.Clear();
+}
+
+TrackID
+OutputStreamManager::NextAvailableTrackIDFor(MediaStream* aOutputStream) const
+{
+  MOZ_ASSERT(NS_IsMainThread());
+  for (const OutputStreamData& out : mStreams) {
+    if (out.Equals(aOutputStream)) {
+      return out.NextAvailableTrackID();
+    }
+  }
+  return TRACK_INVALID;
+}
+
+void
+OutputStreamManager::Connect(MediaStream* aStream,
+                             TrackID aAudioTrackID,
+                             TrackID aVideoTrackID)
 {
   MOZ_ASSERT(NS_IsMainThread());
   mInputStream = aStream;
+  mInputAudioTrackID = aAudioTrackID;
+  mInputVideoTrackID = aVideoTrackID;
   for (int32_t i = mStreams.Length() - 1; i >= 0; --i) {
-    if (!mStreams[i].Connect(aStream)) {
+    if (!mStreams[i].Connect(aStream, mInputAudioTrackID, mInputVideoTrackID)) {
       // Probably the DOMMediaStream was GCed. Clean up.
       mStreams.RemoveElementAt(i);
     }
@@ -123,6 +166,8 @@ OutputStreamManager::Disconnect()
 {
   MOZ_ASSERT(NS_IsMainThread());
   mInputStream = nullptr;
+  mInputAudioTrackID = TRACK_INVALID;
+  mInputVideoTrackID = TRACK_INVALID;
   for (int32_t i = mStreams.Length() - 1; i >= 0; --i) {
     if (!mStreams[i].Disconnect()) {
       // Probably the DOMMediaStream was GCed. Clean up.
diff --git a/dom/media/mediasink/OutputStreamManager.h b/dom/media/mediasink/OutputStreamManager.h
index 7f91a60c1e..941a86cf09 100644
--- a/dom/media/mediasink/OutputStreamManager.h
+++ b/dom/media/mediasink/OutputStreamManager.h
@@ -9,6 +9,7 @@
 
 #include "mozilla/RefPtr.h"
 #include "nsTArray.h"
+#include "MediaSegment.h"
 
 namespace mozilla {
 
@@ -21,11 +22,13 @@ class ProcessedMediaStream;
 class OutputStreamData {
 public:
   ~OutputStreamData();
-  void Init(OutputStreamManager* aOwner, ProcessedMediaStream* aStream);
+  void Init(OutputStreamManager* aOwner,
+            ProcessedMediaStream* aStream,
+            TrackID aNextAvailableTrackID);
 
-  // Connect mStream to the input stream.
+  // Connect the given input stream's audio and video tracks to mStream.
   // Return false is mStream is already destroyed, otherwise true.
-  bool Connect(MediaStream* aStream);
+  bool Connect(MediaStream* aStream, TrackID aAudioTrackID, TrackID aVideoTrackID);
   // Disconnect mStream from its input stream.
   // Return false is mStream is already destroyed, otherwise true.
   bool Disconnect();
@@ -34,12 +37,16 @@ public:
   bool Equals(MediaStream* aStream) const;
   // Return the graph mStream belongs to.
   MediaStreamGraph* Graph() const;
+  // The next TrackID that will not cause a collision in mStream.
+  TrackID NextAvailableTrackID() const;
 
 private:
   OutputStreamManager* mOwner;
   RefPtr<ProcessedMediaStream> mStream;
-  // mPort connects our mStream to an input stream.
-  RefPtr<MediaInputPort> mPort;
+  // mPort connects an input stream to our mStream.
+  nsTArray<RefPtr<MediaInputPort>> mPorts;
+  // For guaranteeing TrackID uniqueness in our mStream.
+  TrackID mNextAvailableTrackID = TRACK_INVALID;
 };
 
 class OutputStreamManager {
@@ -47,18 +54,26 @@ class OutputStreamManager {
 
 public:
   // Add the output stream to the collection.
-  void Add(ProcessedMediaStream* aStream, bool aFinishWhenEnded);
+  void Add(ProcessedMediaStream* aStream,
+           TrackID aNextAvailableTrackID,
+           bool aFinishWhenEnded);
   // Remove the output stream from the collection.
   void Remove(MediaStream* aStream);
+  // Clear all output streams from the collection.
+  void Clear();
+  // The next TrackID that will not cause a collision in aOutputStream.
+  TrackID NextAvailableTrackIDFor(MediaStream* aOutputStream) const;
   // Return true if the collection empty.
   bool IsEmpty() const
   {
     MOZ_ASSERT(NS_IsMainThread());
     return mStreams.IsEmpty();
   }
-  // Connect all output streams in the collection to the input stream.
-  void Connect(MediaStream* aStream);
-  // Disconnect all output streams from the input stream.
+  // Connect the given input stream's tracks to all output streams.
+  void Connect(MediaStream* aStream,
+               TrackID aAudioTrackID,
+               TrackID aVideoTrackID);
+  // Disconnect the input stream to all output streams.
   void Disconnect();
   // Return the graph these streams belong to or null if empty.
   MediaStreamGraph* Graph() const
@@ -72,6 +87,8 @@ private:
   // Keep the input stream so we can connect the output streams that
   // are added after Connect().
   RefPtr<MediaStream> mInputStream;
+  TrackID mInputAudioTrackID = TRACK_INVALID;
+  TrackID mInputVideoTrackID = TRACK_INVALID;
   nsTArray<OutputStreamData> mStreams;
 };