Implement ISO BMFF support in Media Source.

media/mp4/mp4_stream_parser.cc

Index: media/mp4/mp4_stream_parser.cc
diff --git a/media/mp4/mp4_stream_parser.cc b/media/mp4/mp4_stream_parser.cc
new file mode 100644
index 0000000000000000000000000000000000000000..541cb35ae83c1cf71fa556340d2f9d4b226ffc36
--- /dev/null
+++ b/media/mp4/mp4_stream_parser.cc
@@ -0,0 +1,340 @@
+// Copyright (c) 2012 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "media/mp4/mp4_stream_parser.h"
+
+#include "base/callback.h"
+#include "base/logging.h"
+#include "base/time.h"
+#include "media/base/audio_decoder_config.h"
+#include "media/base/stream_parser_buffer.h"
+#include "media/base/video_decoder_config.h"
+#include "media/filters/ffmpeg_glue.h"
+#include "media/mp4/box_definitions.h"
+#include "media/mp4/box_reader.h"
+#include "media/mp4/rcheck.h"
+
+namespace media {
+namespace mp4 {
+
+MP4StreamParser::MP4StreamParser()
+    : state_(kWaitingForInit),
+      moof_head_(0),
+      mdat_tail_(0),
+      has_audio_(false),
+      has_video_(false) {
+}
+
+MP4StreamParser::~MP4StreamParser() {}
+
+void MP4StreamParser::Init(const InitCB& init_cb,
+                           const NewConfigCB& config_cb,
+                           const NewBuffersCB& audio_cb,
+                           const NewBuffersCB& video_cb,
+                           const KeyNeededCB& key_needed_cb) {
+  DCHECK_EQ(state_, kWaitingForInit);
+  DCHECK(init_cb_.is_null());
+  DCHECK(!init_cb.is_null());
+  DCHECK(!config_cb.is_null());
+  DCHECK(!audio_cb.is_null() || !video_cb.is_null());
+  DCHECK(!key_needed_cb.is_null());
+
+  // TODO(strobe): Codec searches fail unless the FFmpegGlue singleton is
+  // initialized. Find the appropriate place to put this initialization.
+  FFmpegGlue::GetInstance();
+
+  ChangeState(kParsingBoxes);
+  init_cb_ = init_cb;
+  config_cb_ = config_cb;
+  audio_cb_ = audio_cb;
+  video_cb_ = video_cb;
+  key_needed_cb_ = key_needed_cb;
+}
+
+void MP4StreamParser::Flush() {
+  DCHECK_NE(state_, kWaitingForInit);
+
+  queue_.Reset();
+  moof_head_ = 0;
+  mdat_tail_ = 0;
+}
+
+bool MP4StreamParser::Parse(const uint8* buf, int size) {
+  DCHECK_NE(state_, kWaitingForInit);
+
+  if (state_ == kError)
+    return false;
+
+  queue_.Push(buf, size);
+
+  BufferQueue audio_buffers;
+  BufferQueue video_buffers;
+
+  bool result, err = false;
+
+  do {
+    if (state_ == kParsingBoxes) {
+      if (mdat_tail_ > queue_.head()) {
+        result = queue_.Trim(mdat_tail_);
+      } else {
+        result = ParseBox(&err);
+      }
+    } else {
+      DCHECK_EQ(kEmittingSamples, state_);
+      result = EnqueueSample(&audio_buffers, &video_buffers, &err);
+      if (result) {
+        size_t max_clear = runs_.GetMaxClearOffset() + moof_head_;
+        DCHECK(max_clear <= queue_.tail());
+        err = !(ReadMDATsUntil(max_clear) && queue_.Trim(max_clear));
+      }
+    }
+  } while (result && !err);
+
+  if (err) {
+    DLOG(ERROR) << "Unknown error while parsing MP4";
+    queue_.Reset();
+    moov_.reset();
+    ChangeState(kError);
+    return false;
+  }
+
+  if (!audio_buffers.empty() &&
+      (audio_cb_.is_null() || !audio_cb_.Run(audio_buffers)))
+    return false;
+  if (!video_buffers.empty() &&
+      (video_cb_.is_null() || !video_cb_.Run(video_buffers)))
+    return false;
+
+  return true;
+}
+
+bool MP4StreamParser::ParseBox(bool* err) {
+  const uint8* buf;
+  size_t size;
+  queue_.Peek(&buf, &size);
+  if (!size) return false;
+
+  scoped_ptr<BoxReader> reader(BoxReader::ReadTopLevelBox(buf, size, err));
+  if (reader.get() == NULL) return false;
+
+  if (reader->type() == FOURCC_MOOV) {
+    *err = !ParseMoov(reader.get());
+  } else if (reader->type() == FOURCC_MOOF) {
+    moof_head_ = queue_.head();
+    *err = !ParseMoof(reader.get());
+
+    // Set up first mdat offset for ParseMDATsUntil()
+    mdat_tail_ = queue_.head() + reader->size();
+  } else {
+    DVLOG(2) << "Skipping unrecognized top-level box: "
+             << FourCCToString(reader->type());
+  }
+
+  queue_.Pop(reader->size());
+  return !(*err);
+}
+
+
+bool MP4StreamParser::ParseMoov(BoxReader* reader) {
+  // TODO(strobe): respect edit lists
+  if (moov_.get()) {
+    DLOG(FATAL) << "TODO: Reinitialization not yet supported.";
+    return false;
+  }
+
+  moov_.reset(new Movie);
+
+  RCHECK(mp4::Parse(reader, moov_.get()));
+
+  has_audio_ = false;
+  has_video_ = false;
+
+  // Select first video and audio track
+  AudioDecoderConfig audio_config;
+  VideoDecoderConfig video_config;
+
+  for (auto track = moov_->tracks.begin();
+       track != moov_->tracks.end(); ++track) {
+    // TODO(strobe): We only take the first of each type of track for now
+    const SampleDescription& samp_descr =
+        track->media.information.sample_table.description;
+    if (track->media.handler.type == kAudio && !audio_config.IsValidConfig()) {
+      RCHECK(!samp_descr.audio_entries.empty());
+      const AudioSampleEntry& entry = samp_descr.audio_entries[0];
+
+      // TODO(strobe): this does not conform to specs for protected audio in the
+      // test files provided by Widevine. Need clarification on the scheme and
+      // whether we expect to deviate from it.
+      //RCHECK(entry.format == FOURCC_MP4A ||
+      //       (entry.format == FOURCC_ENCA &&
+      //        entry.sinf.format.format == FOURCC_MP4A));
+
+      const ChannelLayout layout =
+          ConvertAACChannelCountToChannelLayout(entry.channelcount);
+      audio_config.Initialize(kCodecAAC, entry.samplesize, layout,
+                              entry.samplerate, NULL, 0, false);
+      has_audio_ = true;
+      audio_track_id_ = track->header.track_id;
+    }
+    if (track->media.handler.type == kVideo && !video_config.IsValidConfig()) {
+      RCHECK(!samp_descr.video_entries.empty());
+      const VideoSampleEntry& entry = samp_descr.video_entries[0];
+
+      //RCHECK(entry.format == FOURCC_AVC1 ||
+      //       (entry.format == FOURCC_ENCV &&
+      //        entry.sinf.format.format == FOURCC_AVC1));
+
+      // TODO(strobe): Profile level recovery
+      // TODO(strobe): Crop box (if needed)
+      // TODO(strobe): Rectangle semantics
+      // TODO(strobe): PAR (if needed)
+      // TODO(strobe): Ensure framerate is not used
+      video_config.Initialize(kCodecH264, H264PROFILE_MAIN,  VideoFrame::YV12,
+                              gfx::Size(entry.width, entry.height),
+                              gfx::Rect(0, 0, entry.width, entry.height),
+                              1000, track->media.header.timescale,
+                              1, 1,
+                              NULL, 0, false);
+      has_video_ = true;
+      video_track_id_ = track->header.track_id;
+
+      size_of_nalu_length_ = entry.avcc.length_size;
+    }
+  }
+
+  RCHECK(config_cb_.Run(audio_config, video_config));
+
+  base::TimeDelta duration;
+  if (moov_->extends.header.fragment_duration > 0) {
+    duration = TimeDeltaFromFrac(moov_->extends.header.fragment_duration,
+                                      moov_->header.timescale);

[acolwell GONE FROM CHROMIUM, 2012/06/07 19:44:02]

nit:indent

[strobe_, 2012/06/07 22:04:22]

Done.

+  } else if (moov_->header.duration > 0) {
+    duration = TimeDeltaFromFrac(moov_->header.duration,
+                                      moov_->header.timescale);

[acolwell GONE FROM CHROMIUM, 2012/06/07 19:44:02]

nit: indent

[strobe_, 2012/06/07 22:04:22]

Done.

+  } else {
+    duration = kInfiniteDuration();
+  }
+
+  init_cb_.Run(true, duration);
+  return true;
+}
+
+bool MP4StreamParser::ParseMoof(BoxReader* reader) {
+  RCHECK(moov_.get());  // Must already have initialization segment
+  MovieFragment moof;
+  RCHECK(mp4::Parse(reader, &moof));
+  RCHECK(runs_.Init(*moov_, moof));
+  ChangeState(kEmittingSamples);
+  return true;
+}
+
+bool MP4StreamParser::EnqueueSample(BufferQueue* audio_buffers,
+                                    BufferQueue* video_buffers,
+                                    bool* err) {
+  if (!runs_.RunValid()) {
+    ChangeState(kParsingBoxes);
+    return true;
+  }
+
+  if (!runs_.SampleValid()) {
+    runs_.AdvanceRun();
+    return true;
+  }
+
+  DCHECK(!(*err));
+
+  const uint8* buf;
+  size_t size;
+  queue_.Peek(&buf, &size);
+  if (!size) return false;
+
+  bool audio = has_audio_ && audio_track_id_ == runs_.track_id();
+  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) runs_.AdvanceRun();
+
+  // Attempt to cache the auxiliary information first. Aux info is usually
+  // placed in a contiguous block before the sample data, rather than being
+  // interleaved. If we didn't cache it, this would require that we retain the
+  // start of the segment buffer while reading samples. Aux info is typically
+  // quite small compared to sample data, so this pattern is useful on
+  // memory-constrained devices where the source buffer consumes a substantial
+  // portion of the total system memory.
+  if (runs_.NeedsCENC()) {
+    queue_.PeekAt(runs_.cenc_offset() + moof_head_, &buf, &size);
+    return runs_.CacheCENC(buf, size);
+  }
+
+  queue_.PeekAt(runs_.offset() + moof_head_, &buf, &size);
+  if (size < runs_.size()) return false;
+
+  std::vector<uint8> frame_buf(buf, buf + runs_.size());
+  if (video) {
+    // TODO(strobe): reinject SPS/PPS at every keyframe (or whatever the
+    // spec says to do).
+    RCHECK(ConvertAVCCToAnnexB(size_of_nalu_length_, &frame_buf));
+  }
+
+  scoped_refptr<StreamParserBuffer> stream_buf =
+    StreamParserBuffer::CopyFrom(&frame_buf[0], frame_buf.size(),
+                                 runs_.is_keyframe());
+
+  stream_buf->SetDuration(runs_.duration());
+  stream_buf->SetTimestamp(runs_.dts());
+  // TODO(strobe): verify that Chrome's various media backends perform
+  // B-frame reordering by zipping the un-reordered frame timestamp stream
+  // with the reordered frame data stream, such that we only need to pass
+  // DTS to the source buffer. (Doing so will always carry a risk for
+  // bizarre videos that do things like have a DTS stream that isn't a
+  // strict reordering of the CTS stream, but I'm in favor of ignoring that
+  // case.)
+
+  DVLOG(3) << "Pushing frame: aud=" << audio
+           << ", dur=" << runs_.duration().InMilliseconds()
+           << ", dts=" << runs_.dts().InMilliseconds()
+           << ", size=" << runs_.size();
+
+  if (audio) {
+    audio_buffers->push_back(stream_buf);
+  } else {
+    video_buffers->push_back(stream_buf);
+  }
+
+  runs_.AdvanceSample();
+  return true;
+}
+
+bool MP4StreamParser::ReadMDATsUntil(const size_t tgt_offset) {
+  DCHECK(tgt_offset <= queue_.tail());
+
+  while (mdat_tail_ < tgt_offset) {
+    const uint8* buf;
+    size_t size;
+    queue_.PeekAt(mdat_tail_, &buf, &size);
+
+    FourCC type;
+    size_t box_sz;
+    bool err;
+    if (!BoxReader::StartTopLevelBox(buf, size, &type, &box_sz, &err))
+      return false;
+
+    if (type != FOURCC_MDAT) {
+      DLOG(WARNING) << "Unexpected type while parsing MDATs: "
+                    << FourCCToString(type);
+    }
+
+    mdat_tail_ += box_sz;
+  }
+
+  return true;
+}
+
+void MP4StreamParser::ChangeState(State new_state) {
+  DVLOG(2) << "Changing state: " << new_state;
+  state_ = new_state;
+}
+
+}  // namespace mp4
+}  // namespace media