Fix muxed MP4 parsing so it won't crash on partial media segment appends.

Partial appends could cause the MP4StreamParser to emit buffers in
non-monotonically increasing timestamp order because of how we process
trun boxes. This patch makes sure that processing of trun boxes is
deferred until we have all the sample & aux_info data before emitting
samples. This prevents the parser from emitting samples in such a way
that will break downstream code. This is a minimal impact fix to
avoid bad behavior at the expense of buffering more data.

BUG=None
TEST=All MP4StreamParserTests that do partial appends now verify the fix.

Review URL: https://codereview.chromium.org/348623003

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@278650 0039d316-1c4b-4281-b951-d872f2087c98

media/formats/mp4/mp4_stream_parser.cc

@@ -92,18 +92,33 @@ bool MP4StreamParser::Parse(const uint8* buf, int size) {
   BufferQueue audio_buffers;
   BufferQueue video_buffers;
 
-  bool result, err = false;
+  bool result = false;
+  bool err = false;
 
   do {
-    if (state_ == kParsingBoxes) {
-      result = ParseBox(&err);
-    } else {
-      DCHECK_EQ(kEmittingSamples, state_);
-      result = EnqueueSample(&audio_buffers, &video_buffers, &err);
-      if (result) {
-        int64 max_clear = runs_->GetMaxClearOffset() + moof_head_;
-        err = !ReadAndDiscardMDATsUntil(max_clear);
-      }
+    switch (state_) {
+      case kWaitingForInit:
+      case kError:
+        NOTREACHED();
+        return false;
+
+      case kParsingBoxes:
+        result = ParseBox(&err);
+        break;
+
+      case kWaitingForSampleData:
+        result = HaveEnoughDataToEnqueueSamples();
+        if (result)
+          ChangeState(kEmittingSamples);
+        break;
+
+      case kEmittingSamples:
+        result = EnqueueSample(&audio_buffers, &video_buffers, &err);
+        if (result) {
+          int64 max_clear = runs_->GetMaxClearOffset() + moof_head_;
+          err = !ReadAndDiscardMDATsUntil(max_clear);
+        }
+        break;
     }
   } while (result && !err);
 
@@ -312,9 +327,10 @@ bool MP4StreamParser::ParseMoof(BoxReader* reader) {
   if (!runs_)
     runs_.reset(new TrackRunIterator(moov_.get(), log_cb_));
   RCHECK(runs_->Init(moof));
+  RCHECK(ComputeHighestEndOffset(moof));
   EmitNeedKeyIfNecessary(moof.pssh);
   new_segment_cb_.Run();
-  ChangeState(kEmittingSamples);
+  ChangeState(kWaitingForSampleData);
   return true;
 }
 
@@ -393,6 +409,8 @@ bool MP4StreamParser::PrepareAACBuffer(
 bool MP4StreamParser::EnqueueSample(BufferQueue* audio_buffers,
                                     BufferQueue* video_buffers,
                                     bool* err) {
+  DCHECK_EQ(state_, kEmittingSamples);
+
   if (!runs_->IsRunValid()) {
     // Flush any buffers we've gotten in this chunk so that buffers don't
     // cross NewSegment() calls
@@ -400,7 +418,7 @@ bool MP4StreamParser::EnqueueSample(BufferQueue* audio_buffers,
     if (*err)
       return false;
 
-    // Remain in kEnqueueingSamples state, discarding data, until the end of
+    // Remain in kEmittingSamples state, discarding data, until the end of
     // the current 'mdat' box has been appended to the queue.
     if (!queue_.Trim(mdat_tail_))
       return false;
@@ -426,8 +444,10 @@ bool MP4StreamParser::EnqueueSample(BufferQueue* audio_buffers,
   bool video = has_video_ && video_track_id_ == runs_->track_id();
 
   // Skip this entire track if it's not one we're interested in
-  if (!audio && !video)
+  if (!audio && !video) {
     runs_->AdvanceRun();
+    return true;
+  }
 
   // Attempt to cache the auxiliary information first. Aux info is usually
   // placed in a contiguous block before the sample data, rather than being
@@ -577,5 +597,40 @@ void MP4StreamParser::ChangeState(State new_state) {
   state_ = new_state;
 }
 
+bool MP4StreamParser::HaveEnoughDataToEnqueueSamples() {
+  DCHECK_EQ(state_, kWaitingForSampleData);
+  // For muxed content, make sure we have data up to |highest_end_offset_|
+  // so we can ensure proper enqueuing behavior. Otherwise assume we have enough
+  // data and allow per sample offset checks to meter sample enqueuing.
+  // TODO(acolwell): Fix trun box handling so we don't have to special case
+  // muxed content.
+  return !(has_audio_ && has_video_ &&
+           queue_.tail() < highest_end_offset_ + moof_head_);
+}
+
+bool MP4StreamParser::ComputeHighestEndOffset(const MovieFragment& moof) {
+  highest_end_offset_ = 0;
+
+  TrackRunIterator runs(moov_.get(), log_cb_);
+  RCHECK(runs.Init(moof));
+
+  while (runs.IsRunValid()) {
+    int64 aux_info_end_offset = runs.aux_info_offset() + runs.aux_info_size();
+    if (aux_info_end_offset > highest_end_offset_)
+      highest_end_offset_ = aux_info_end_offset;
+
+    while (runs.IsSampleValid()) {
+      int64 sample_end_offset = runs.sample_offset() + runs.sample_size();
+      if (sample_end_offset > highest_end_offset_)
+        highest_end_offset_ = sample_end_offset;
+
+      runs.AdvanceSample();
+    }
+    runs.AdvanceRun();
+  }
+
+  return true;
+}
+
 }  // namespace mp4
 }  // namespace media

media/formats/mp4/mp4_stream_parser.h

@@ -42,6 +42,7 @@ class MEDIA_EXPORT MP4StreamParser : public StreamParser {
   enum State {
     kWaitingForInit,
     kParsingBoxes,
+    kWaitingForSampleData,
     kEmittingSamples,
     kError
   };
@@ -78,6 +79,15 @@ class MEDIA_EXPORT MP4StreamParser : public StreamParser {
 
   void Reset();
 
+  // Checks to see if we have enough data in |queue_| to transition to
+  // kEmittingSamples and start enqueuing samples.
+  bool HaveEnoughDataToEnqueueSamples();
+
+  // Sets |highest_end_offset_| based on the data in |moov_|
+  // and |moof|. Returns true if |highest_end_offset_| was successfully
+  // computed.
+  bool ComputeHighestEndOffset(const MovieFragment& moof);
+
   State state_;
   InitCB init_cb_;
   NewConfigCB config_cb_;
@@ -99,6 +109,11 @@ class MEDIA_EXPORT MP4StreamParser : public StreamParser {
   // Valid iff it is greater than the head of the queue.
   int64 mdat_tail_;
 
+  // The highest end offset in the current moof. This offset is
+  // relative to |moof_head_|. This value is used to make sure we have collected
+  // enough bytes to parse all samples and aux_info in the current moof.
+  int64 highest_end_offset_;
+
   scoped_ptr<mp4::Movie> moov_;
   scoped_ptr<mp4::TrackRunIterator> runs_;
 

media/formats/mp4/mp4_stream_parser_unittest.cc

@@ -31,7 +31,8 @@ static const char kMp4InitDataType[] = "video/mp4";
 class MP4StreamParserTest : public testing::Test {
  public:
   MP4StreamParserTest()
-      : configs_received_(false) {
+      : configs_received_(false),
+        lower_bound_(base::TimeDelta::Max()) {
     std::set<int> audio_object_types;
     audio_object_types.insert(kISO_14496_3);
     parser_.reset(new MP4StreamParser(audio_object_types, false));
@@ -40,6 +41,7 @@ class MP4StreamParserTest : public testing::Test {
  protected:
   scoped_ptr<MP4StreamParser> parser_;
   bool configs_received_;
+  base::TimeDelta lower_bound_;
 
   bool AppendData(const uint8* data, size_t length) {
     return parser_->Parse(data, length);
@@ -73,7 +75,6 @@ class MP4StreamParserTest : public testing::Test {
     return true;
   }
 
-
   void DumpBuffers(const std::string& label,
                    const StreamParser::BufferQueue& buffers) {
     DVLOG(2) << "DumpBuffers: " << label << " size " << buffers.size();
@@ -96,6 +97,24 @@ class MP4StreamParserTest : public testing::Test {
     if (!text_map.empty())
       return false;
 
+    // Find the second highest timestamp so that we know what the
+    // timestamps on the next set of buffers must be >= than.
+    base::TimeDelta audio = !audio_buffers.empty() ?
+        audio_buffers.back()->GetDecodeTimestamp() : kNoTimestamp();
+    base::TimeDelta video = !video_buffers.empty() ?
+        video_buffers.back()->GetDecodeTimestamp() : kNoTimestamp();
+    base::TimeDelta second_highest_timestamp =
+        (audio == kNoTimestamp() ||
+         (video != kNoTimestamp() && audio > video)) ? video : audio;
+
+    DCHECK(second_highest_timestamp != kNoTimestamp());
+
+    if (lower_bound_ != kNoTimestamp() &&
+        second_highest_timestamp < lower_bound_) {
+      return false;
+    }
+
+    lower_bound_ = second_highest_timestamp;
     return true;
   }
 
@@ -108,10 +127,12 @@ class MP4StreamParserTest : public testing::Test {
 
   void NewSegmentF() {
     DVLOG(1) << "NewSegmentF";
+    lower_bound_ = kNoTimestamp();
   }
 
   void EndOfSegmentF() {
     DVLOG(1) << "EndOfSegmentF()";
+    lower_bound_ = base::TimeDelta::Max();
   }
 
   void InitializeParser() {