media/formats/mp4/mp4_stream_parser.h

   1 // Copyright 2014 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #ifndef MEDIA_FORMATS_MP4_MP4_STREAM_PARSER_H_
   6 #define MEDIA_FORMATS_MP4_MP4_STREAM_PARSER_H_
   7
   8 #include <set>
   9 #include <vector>
  10
  11 #include "base/basictypes.h"
  12 #include "base/callback.h"
  13 #include "base/compiler_specific.h"
  14 #include "base/memory/scoped_ptr.h"
  15 #include "media/base/media_export.h"
  16 #include "media/base/stream_parser.h"
  17 #include "media/formats/common/offset_byte_queue.h"
  18 #include "media/formats/mp4/track_run_iterator.h"
  19
  20 namespace media {
  21 namespace mp4 {
  22
  23 struct Movie;
  24 class BoxReader;
  25
  26 class MEDIA_EXPORT MP4StreamParser : public StreamParser {
  27  public:
  28   MP4StreamParser(const std::set<int>& audio_object_types, bool has_sbr);
  29   ~MP4StreamParser() override;
  30
  31   void Init(const InitCB& init_cb,
  32             const NewConfigCB& config_cb,
  33             const NewBuffersCB& new_buffers_cb,
  34             bool ignore_text_tracks,
  35             const EncryptedMediaInitDataCB& encrypted_media_init_data_cb,
  36             const NewMediaSegmentCB& new_segment_cb,
  37             const base::Closure& end_of_segment_cb,
  38             const scoped_refptr<MediaLog>& media_log) override;
  39   void Flush() override;
  40   bool Parse(const uint8* buf, int size) override;
  41
  42  private:
  43   enum State {
  44     kWaitingForInit,
  45     kParsingBoxes,
  46     kWaitingForSampleData,
  47     kEmittingSamples,
  48     kError
  49   };
  50
  51   bool ParseBox(bool* err);
  52   bool ParseMoov(mp4::BoxReader* reader);
  53   bool ParseMoof(mp4::BoxReader* reader);
  54
  55   void OnEncryptedMediaInitData(
  56       const std::vector<ProtectionSystemSpecificHeader>& headers);
  57
  58   // To retain proper framing, each 'mdat' atom must be read; to limit memory
  59   // usage, the atom's data needs to be discarded incrementally as frames are
  60   // extracted from the stream. This function discards data from the stream up
  61   // to |max_clear_offset|, updating the |mdat_tail_| value so that framing can
  62   // be retained after all 'mdat' information has been read. |max_clear_offset|
  63   // is the upper bound on what can be removed from |queue_|. Anything below
  64   // this offset is no longer needed by the parser.
  65   // Returns 'true' on success, 'false' if there was an error.
  66   bool ReadAndDiscardMDATsUntil(int64 max_clear_offset);
  67
  68   void ChangeState(State new_state);
  69
  70   bool EmitConfigs();
  71   bool PrepareAVCBuffer(const AVCDecoderConfigurationRecord& avc_config,
  72                         std::vector<uint8>* frame_buf,
  73                         std::vector<SubsampleEntry>* subsamples) const;
  74   bool PrepareAACBuffer(const AAC& aac_config,
  75                         std::vector<uint8>* frame_buf,
  76                         std::vector<SubsampleEntry>* subsamples) const;
  77   bool EnqueueSample(BufferQueue* audio_buffers,
  78                      BufferQueue* video_buffers,
  79                      bool* err);
  80   bool SendAndFlushSamples(BufferQueue* audio_buffers,
  81                            BufferQueue* video_buffers);
  82
  83   void Reset();
  84
  85   // Checks to see if we have enough data in |queue_| to transition to
  86   // kEmittingSamples and start enqueuing samples.
  87   bool HaveEnoughDataToEnqueueSamples();
  88
  89   // Sets |highest_end_offset_| based on the data in |moov_|
  90   // and |moof|. Returns true if |highest_end_offset_| was successfully
  91   // computed.
  92   bool ComputeHighestEndOffset(const MovieFragment& moof);
  93
  94   State state_;
  95   InitCB init_cb_;
  96   NewConfigCB config_cb_;
  97   NewBuffersCB new_buffers_cb_;
  98   EncryptedMediaInitDataCB encrypted_media_init_data_cb_;
  99   NewMediaSegmentCB new_segment_cb_;
 100   base::Closure end_of_segment_cb_;
 101   scoped_refptr<MediaLog> media_log_;
 102
 103   OffsetByteQueue queue_;
 104
 105   // These two parameters are only valid in the |kEmittingSegments| state.
 106   //
 107   // |moof_head_| is the offset of the start of the most recently parsed moof
 108   // block. All byte offsets in sample information are relative to this offset,
 109   // as mandated by the Media Source spec.
 110   int64 moof_head_;
 111   // |mdat_tail_| is the stream offset of the end of the current 'mdat' box.
 112   // Valid iff it is greater than the head of the queue.
 113   int64 mdat_tail_;
 114
 115   // The highest end offset in the current moof. This offset is
 116   // relative to |moof_head_|. This value is used to make sure we have collected
 117   // enough bytes to parse all samples and aux_info in the current moof.
 118   int64 highest_end_offset_;
 119
 120   scoped_ptr<mp4::Movie> moov_;
 121   scoped_ptr<mp4::TrackRunIterator> runs_;
 122
 123   bool has_audio_;
 124   bool has_video_;
 125   uint32 audio_track_id_;
 126   uint32 video_track_id_;
 127   // The object types allowed for audio tracks.
 128   std::set<int> audio_object_types_;
 129   bool has_sbr_;
 130   bool is_audio_track_encrypted_;
 131   bool is_video_track_encrypted_;
 132
 133   // Tracks the number of MEDIA_LOGs for skipping top level boxes. Useful to
 134   // prevent log spam.
 135   int num_top_level_box_skipped_;
 136
 137   DISALLOW_COPY_AND_ASSIGN(MP4StreamParser);
 138 };
 139
 140 }  // namespace mp4
 141 }  // namespace media
 142
 143 #endif  // MEDIA_FORMATS_MP4_MP4_STREAM_PARSER_H_