Bug 1195073: [MSE] P5. Detect out of order appends and recreate demuxer. r=gerald

The webm demuxer will only handle data where frames's a monotonically increasing.

dom/media/mediasource/TrackBuffersManager.cpp

@@ -97,6 +97,7 @@ TrackBuffersManager::TrackBuffersManager(dom::SourceBufferAttributes* aAttribute
   , mAppendState(AppendState::WAITING_FOR_SEGMENT)
   , mBufferFull(false)
   , mFirstInitializationSegmentReceived(false)
+  , mNewSegmentStarted(false)
   , mActiveTrack(false)
   , mType(aType)
   , mParser(ContainerParser::CreateForMIMEType(aType))
@@ -655,10 +656,12 @@ TrackBuffersManager::SegmentParserLoop()
           // This is a new initialization segment. Obsolete the old one.
           RecreateParser(false);
         }
+        mNewSegmentStarted = true;
         continue;
       }
       if (mParser->IsMediaSegmentPresent(mInputBuffer)) {
         SetAppendState(AppendState::PARSING_MEDIA_SEGMENT);
+        mNewSegmentStarted = true;
         continue;
       }
       // We have neither an init segment nor a media segment, this is either
@@ -669,7 +672,7 @@ TrackBuffersManager::SegmentParserLoop()
     }
 
     int64_t start, end;
-    mParser->ParseStartAndEndTimestamps(mInputBuffer, start, end);
+    bool newData = mParser->ParseStartAndEndTimestamps(mInputBuffer, start, end);
     mProcessedInput += mInputBuffer->Length();
 
     // 5. If the append state equals PARSING_INIT_SEGMENT, then run the
@@ -696,6 +699,22 @@ TrackBuffersManager::SegmentParserLoop()
         NeedMoreData();
         return;
       }
+
+      // We can't feed some demuxers (WebMDemuxer) with data that do not have
+      // monotonizally increasing timestamps. So we check if we have a
+      // discontinuity from the previous segment parsed.
+      // If so, recreate a new demuxer to ensure that the demuxer is only fed
+      // monotonically increasing data.
+      if (newData) {
+        if (mNewSegmentStarted && mLastParsedEndTime.isSome() &&
+            start < mLastParsedEndTime.ref().ToMicroseconds()) {
+          ResetDemuxingState();
+          return;
+        }
+        mNewSegmentStarted = false;
+        mLastParsedEndTime = Some(TimeUnit::FromMicroseconds(end));
+      }
+
       // 3. If the input buffer contains one or more complete coded frames, then run the coded frame processing algorithm.
       nsRefPtr<TrackBuffersManager> self = this;
       mProcessingRequest.Begin(CodedFrameProcessing()
@@ -756,6 +775,7 @@ TrackBuffersManager::ShutdownDemuxers()
     mAudioTracks.mDemuxer = nullptr;
   }
   mInputDemuxer = nullptr;
+  mLastParsedEndTime.reset();
 }
 
 void
@@ -780,6 +800,58 @@ TrackBuffersManager::CreateDemuxerforMIMEType()
   return;
 }
 
+// We reset the demuxer by creating a new one and initializing it.
+void
+TrackBuffersManager::ResetDemuxingState()
+{
+  MOZ_ASSERT(mParser && mParser->HasInitData());
+  RecreateParser(true);
+  mCurrentInputBuffer = new SourceBufferResource(mType);
+  // The demuxer isn't initialized yet ; we don't want to notify it
+  // that data has been appended yet ; so we simply append the init segment
+  // to the resource.
+  mCurrentInputBuffer->AppendData(mParser->InitData());
+  CreateDemuxerforMIMEType();
+  if (!mInputDemuxer) {
+    RejectAppend(NS_ERROR_FAILURE, __func__);
+    return;
+  }
+  mDemuxerInitRequest.Begin(mInputDemuxer->Init()
+                      ->Then(GetTaskQueue(), __func__,
+                             this,
+                             &TrackBuffersManager::OnDemuxerResetDone,
+                             &TrackBuffersManager::OnDemuxerInitFailed));
+}
+
+void
+TrackBuffersManager::OnDemuxerResetDone(nsresult)
+{
+  MOZ_ASSERT(OnTaskQueue());
+  MSE_DEBUG("mAbort:%d", static_cast<bool>(mAbort));
+  mDemuxerInitRequest.Complete();
+  if (mAbort) {
+    RejectAppend(NS_ERROR_ABORT, __func__);
+    return;
+  }
+
+  // Recreate track demuxers.
+  uint32_t numVideos = mInputDemuxer->GetNumberTracks(TrackInfo::kVideoTrack);
+  if (numVideos) {
+    // We currently only handle the first video track.
+    mVideoTracks.mDemuxer = mInputDemuxer->GetTrackDemuxer(TrackInfo::kVideoTrack, 0);
+    MOZ_ASSERT(mVideoTracks.mDemuxer);
+  }
+
+  uint32_t numAudios = mInputDemuxer->GetNumberTracks(TrackInfo::kAudioTrack);
+  if (numAudios) {
+    // We currently only handle the first audio track.
+    mAudioTracks.mDemuxer = mInputDemuxer->GetTrackDemuxer(TrackInfo::kAudioTrack, 0);
+    MOZ_ASSERT(mAudioTracks.mDemuxer);
+  }
+
+  SegmentParserLoop();
+}
+
 void
 TrackBuffersManager::AppendDataToCurrentInputBuffer(MediaByteBuffer* aData)
 {
@@ -1056,6 +1128,14 @@ TrackBuffersManager::CodedFrameProcessing()
     // The mediaRange is offset by the init segment position previously added.
     uint32_t length =
       mediaRange.mEnd - (mProcessedInput - mInputBuffer->Length());
+    if (!length) {
+      // We've completed our earlier media segment and no new data is to be
+      // processed. This happens with some containers that can't detect that a
+      // media segment is ending until a new one starts.
+      nsRefPtr<CodedFrameProcessingPromise> p = mProcessingPromise.Ensure(__func__);
+      CompleteCodedFrameProcessing();
+      return p;
+    }
     nsRefPtr<MediaByteBuffer> segment = new MediaByteBuffer;
     if (!segment->AppendElements(mInputBuffer->Elements(), length, fallible)) {
       return CodedFrameProcessingPromise::CreateAndReject(NS_ERROR_OUT_OF_MEMORY, __func__);

dom/media/mediasource/TrackBuffersManager.h

@@ -113,6 +113,7 @@ private:
   void InitializationSegmentReceived();
   void ShutdownDemuxers();
   void CreateDemuxerforMIMEType();
+  void ResetDemuxingState();
   void NeedMoreData();
   void RejectAppend(nsresult aRejectValue, const char* aName);
   // Will return a promise that will be resolved once all frames of the current
@@ -151,6 +152,8 @@ private:
   // TODO: Unused for now.
   Atomic<bool> mBufferFull;
   bool mFirstInitializationSegmentReceived;
+  // Set to true once a new segment is started.
+  bool mNewSegmentStarted;
   bool mActiveTrack;
   Maybe<media::TimeUnit> mGroupStartTimestamp;
   media::TimeUnit mGroupEndTimestamp;
@@ -171,9 +174,11 @@ private:
   nsRefPtr<MediaDataDemuxer> mInputDemuxer;
   // Length already processed in current media segment.
   uint32_t mProcessedInput;
+  Maybe<media::TimeUnit> mLastParsedEndTime;
 
   void OnDemuxerInitDone(nsresult);
   void OnDemuxerInitFailed(DemuxerFailureReason aFailure);
+  void OnDemuxerResetDone(nsresult);
   MozPromiseRequestHolder<MediaDataDemuxer::InitPromise> mDemuxerInitRequest;
   bool mEncrypted;
 

dom/media/webm/WebMBufferedParser.cpp

@@ -208,6 +208,7 @@ void WebMBufferedParser::Append(const unsigned char* aBuffer, uint32_t aLength,
       if (!mSkipBytes) {
         if (mInitEndOffset < 0) {
           mInitEndOffset = mCurrentOffset + (p - aBuffer);
+          mBlockEndOffset = mCurrentOffset + (p - aBuffer);
         }
         mState = READ_ELEMENT_ID;
       }