media/formats/mpeg/mpeg_audio_stream_parser_base.cc

   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 #include "media/formats/mpeg/mpeg_audio_stream_parser_base.h"
   6
   7 #include "base/bind.h"
   8 #include "base/callback_helpers.h"
   9 #include "base/message_loop/message_loop.h"
  10 #include "media/base/stream_parser_buffer.h"
  11 #include "media/base/text_track_config.h"
  12 #include "media/base/timestamp_constants.h"
  13 #include "media/base/video_decoder_config.h"
  14
  15 namespace media {
  16
  17 static const uint32 kICYStartCode = 0x49435920; // 'ICY '
  18
  19 // Arbitrary upper bound on the size of an IceCast header before it
  20 // triggers an error.
  21 static const int kMaxIcecastHeaderSize = 4096;
  22
  23 static const uint32 kID3StartCodeMask = 0xffffff00;
  24 static const uint32 kID3v1StartCode = 0x54414700; // 'TAG\0'
  25 static const int kID3v1Size = 128;
  26 static const int kID3v1ExtendedSize = 227;
  27 static const uint32 kID3v2StartCode = 0x49443300; // 'ID3\0'
  28
  29 static int LocateEndOfHeaders(const uint8_t* buf, int buf_len, int i) {
  30   bool was_lf = false;
  31   char last_c = '\0';
  32   for (; i < buf_len; ++i) {
  33     char c = buf[i];
  34     if (c == '\n') {
  35       if (was_lf)
  36         return i + 1;
  37       was_lf = true;
  38     } else if (c != '\r' || last_c != '\n') {
  39       was_lf = false;
  40     }
  41     last_c = c;
  42   }
  43   return -1;
  44 }
  45
  46 MPEGAudioStreamParserBase::MPEGAudioStreamParserBase(uint32 start_code_mask,
  47                                                      AudioCodec audio_codec,
  48                                                      int codec_delay)
  49     : state_(UNINITIALIZED),
  50       in_media_segment_(false),
  51       start_code_mask_(start_code_mask),
  52       audio_codec_(audio_codec),
  53       codec_delay_(codec_delay) {}
  54
  55 MPEGAudioStreamParserBase::~MPEGAudioStreamParserBase() {}
  56
  57 void MPEGAudioStreamParserBase::Init(
  58     const InitCB& init_cb,
  59     const NewConfigCB& config_cb,
  60     const NewBuffersCB& new_buffers_cb,
  61     bool ignore_text_tracks,
  62     const EncryptedMediaInitDataCB& encrypted_media_init_data_cb,
  63     const NewMediaSegmentCB& new_segment_cb,
  64     const base::Closure& end_of_segment_cb,
  65     const scoped_refptr<MediaLog>& media_log) {
  66   DVLOG(1) << __FUNCTION__;
  67   DCHECK_EQ(state_, UNINITIALIZED);
  68   init_cb_ = init_cb;
  69   config_cb_ = config_cb;
  70   new_buffers_cb_ = new_buffers_cb;
  71   new_segment_cb_ = new_segment_cb;
  72   end_of_segment_cb_ = end_of_segment_cb;
  73   media_log_ = media_log;
  74
  75   ChangeState(INITIALIZED);
  76 }
  77
  78 void MPEGAudioStreamParserBase::Flush() {
  79   DVLOG(1) << __FUNCTION__;
  80   DCHECK_NE(state_, UNINITIALIZED);
  81   queue_.Reset();
  82   timestamp_helper_->SetBaseTimestamp(base::TimeDelta());
  83   in_media_segment_ = false;
  84 }
  85
  86 bool MPEGAudioStreamParserBase::Parse(const uint8* buf, int size) {
  87   DVLOG(1) << __FUNCTION__ << "(" << size << ")";
  88   DCHECK(buf);
  89   DCHECK_GT(size, 0);
  90   DCHECK_NE(state_, UNINITIALIZED);
  91
  92   if (state_ == PARSE_ERROR)
  93     return false;
  94
  95   DCHECK_EQ(state_, INITIALIZED);
  96
  97   queue_.Push(buf, size);
  98
  99   bool end_of_segment = true;
 100   BufferQueue buffers;
 101   for (;;) {
 102     const uint8* data;
 103     int data_size;
 104     queue_.Peek(&data, &data_size);
 105
 106     if (data_size < 4)
 107       break;
 108
 109     uint32 start_code = data[0] << 24 | data[1] << 16 | data[2] << 8 | data[3];
 110     int bytes_read = 0;
 111     bool parsed_metadata = true;
 112     if ((start_code & start_code_mask_) == start_code_mask_) {
 113       bytes_read = ParseFrame(data, data_size, &buffers);
 114
 115       // Only allow the current segment to end if a full frame has been parsed.
 116       end_of_segment = bytes_read > 0;
 117       parsed_metadata = false;
 118     } else if (start_code == kICYStartCode) {
 119       bytes_read = ParseIcecastHeader(data, data_size);
 120     } else if ((start_code & kID3StartCodeMask) == kID3v1StartCode) {
 121       bytes_read = ParseID3v1(data, data_size);
 122     } else if ((start_code & kID3StartCodeMask) == kID3v2StartCode) {
 123       bytes_read = ParseID3v2(data, data_size);
 124     } else {
 125       bytes_read = FindNextValidStartCode(data, data_size);
 126
 127       if (bytes_read > 0) {
 128         DVLOG(1) << "Unexpected start code 0x" << std::hex << start_code;
 129         DVLOG(1) << "SKIPPING " << bytes_read << " bytes of garbage.";
 130       }
 131     }
 132
 133     CHECK_LE(bytes_read, data_size);
 134
 135     if (bytes_read < 0) {
 136       ChangeState(PARSE_ERROR);
 137       return false;
 138     } else if (bytes_read == 0) {
 139       // Need more data.
 140       break;
 141     }
 142
 143     // Send pending buffers if we have encountered metadata.
 144     if (parsed_metadata && !buffers.empty() && !SendBuffers(&buffers, true))
 145       return false;
 146
 147     queue_.Pop(bytes_read);
 148     end_of_segment = true;
 149   }
 150
 151   if (buffers.empty())
 152     return true;
 153
 154   // Send buffers collected in this append that haven't been sent yet.
 155   return SendBuffers(&buffers, end_of_segment);
 156 }
 157
 158 void MPEGAudioStreamParserBase::ChangeState(State state) {
 159   DVLOG(1) << __FUNCTION__ << "() : " << state_ << " -> " << state;
 160   state_ = state;
 161 }
 162
 163 int MPEGAudioStreamParserBase::ParseFrame(const uint8* data,
 164                                           int size,
 165                                           BufferQueue* buffers) {
 166   DVLOG(2) << __FUNCTION__ << "(" << size << ")";
 167
 168   int sample_rate;
 169   ChannelLayout channel_layout;
 170   int frame_size;
 171   int sample_count;
 172   bool metadata_frame = false;
 173   int bytes_read = ParseFrameHeader(data,
 174                                     size,
 175                                     &frame_size,
 176                                     &sample_rate,
 177                                     &channel_layout,
 178                                     &sample_count,
 179                                     &metadata_frame);
 180
 181   if (bytes_read <= 0)
 182     return bytes_read;
 183
 184   // Make sure data contains the entire frame.
 185   if (size < frame_size)
 186     return 0;
 187
 188   DVLOG(2) << " sample_rate " << sample_rate
 189            << " channel_layout " << channel_layout
 190            << " frame_size " << frame_size
 191            << " sample_count " << sample_count;
 192
 193   if (config_.IsValidConfig() &&
 194       (config_.samples_per_second() != sample_rate ||
 195        config_.channel_layout() != channel_layout)) {
 196     // Clear config data so that a config change is initiated.
 197     config_ = AudioDecoderConfig();
 198
 199     // Send all buffers associated with the previous config.
 200     if (!buffers->empty() && !SendBuffers(buffers, true))
 201       return -1;
 202   }
 203
 204   if (!config_.IsValidConfig()) {
 205     config_.Initialize(audio_codec_,
 206                        kSampleFormatF32,
 207                        channel_layout,
 208                        sample_rate,
 209                        NULL,
 210                        0,
 211                        false,
 212                        base::TimeDelta(),
 213                        codec_delay_);
 214
 215     base::TimeDelta base_timestamp;
 216     if (timestamp_helper_)
 217       base_timestamp = timestamp_helper_->GetTimestamp();
 218
 219     timestamp_helper_.reset(new AudioTimestampHelper(sample_rate));
 220     timestamp_helper_->SetBaseTimestamp(base_timestamp);
 221
 222     VideoDecoderConfig video_config;
 223     if (!config_cb_.Run(config_, video_config, TextTrackConfigMap()))
 224       return -1;
 225
 226     if (!init_cb_.is_null()) {
 227       InitParameters params(kInfiniteDuration());
 228       params.auto_update_timestamp_offset = true;
 229       base::ResetAndReturn(&init_cb_).Run(params);
 230     }
 231   }
 232
 233   if (metadata_frame)
 234     return frame_size;
 235
 236   // TODO(wolenetz/acolwell): Validate and use a common cross-parser TrackId
 237   // type and allow multiple audio tracks, if applicable. See
 238   // https://crbug.com/341581.
 239   scoped_refptr<StreamParserBuffer> buffer =
 240       StreamParserBuffer::CopyFrom(data, frame_size, true,
 241                                    DemuxerStream::AUDIO, 0);
 242   buffer->set_timestamp(timestamp_helper_->GetTimestamp());
 243   buffer->set_duration(timestamp_helper_->GetFrameDuration(sample_count));
 244   buffers->push_back(buffer);
 245
 246   timestamp_helper_->AddFrames(sample_count);
 247
 248   return frame_size;
 249 }
 250
 251 int MPEGAudioStreamParserBase::ParseIcecastHeader(const uint8* data, int size) {
 252   DVLOG(1) << __FUNCTION__ << "(" << size << ")";
 253
 254   if (size < 4)
 255     return 0;
 256
 257   if (memcmp("ICY ", data, 4))
 258     return -1;
 259
 260   int locate_size = std::min(size, kMaxIcecastHeaderSize);
 261   int offset = LocateEndOfHeaders(data, locate_size, 4);
 262   if (offset < 0) {
 263     if (locate_size == kMaxIcecastHeaderSize) {
 264       MEDIA_LOG(ERROR, media_log_) << "Icecast header is too large.";
 265       return -1;
 266     }
 267
 268     return 0;
 269   }
 270
 271   return offset;
 272 }
 273
 274 int MPEGAudioStreamParserBase::ParseID3v1(const uint8* data, int size) {
 275   DVLOG(1) << __FUNCTION__ << "(" << size << ")";
 276
 277   if (size < kID3v1Size)
 278     return 0;
 279
 280   // TODO(acolwell): Add code to actually validate ID3v1 data and
 281   // expose it as a metadata text track.
 282   return !memcmp(data, "TAG+", 4) ? kID3v1ExtendedSize : kID3v1Size;
 283 }
 284
 285 int MPEGAudioStreamParserBase::ParseID3v2(const uint8* data, int size) {
 286   DVLOG(1) << __FUNCTION__ << "(" << size << ")";
 287
 288   if (size < 10)
 289     return 0;
 290
 291   BitReader reader(data, size);
 292   int32 id;
 293   int version;
 294   uint8 flags;
 295   int32 id3_size;
 296
 297   if (!reader.ReadBits(24, &id) ||
 298       !reader.ReadBits(16, &version) ||
 299       !reader.ReadBits(8, &flags) ||
 300       !ParseSyncSafeInt(&reader, &id3_size)) {
 301     return -1;
 302   }
 303
 304   int32 actual_tag_size = 10 + id3_size;
 305
 306   // Increment size if 'Footer present' flag is set.
 307   if (flags & 0x10)
 308     actual_tag_size += 10;
 309
 310   // Make sure we have the entire tag.
 311   if (size < actual_tag_size)
 312     return 0;
 313
 314   // TODO(acolwell): Add code to actually validate ID3v2 data and
 315   // expose it as a metadata text track.
 316   return actual_tag_size;
 317 }
 318
 319 bool MPEGAudioStreamParserBase::ParseSyncSafeInt(BitReader* reader,
 320                                                  int32* value) {
 321   *value = 0;
 322   for (int i = 0; i < 4; ++i) {
 323     uint8 tmp;
 324     if (!reader->ReadBits(1, &tmp) || tmp != 0) {
 325       MEDIA_LOG(ERROR, media_log_) << "ID3 syncsafe integer byte MSb is not 0!";
 326       return false;
 327     }
 328
 329     if (!reader->ReadBits(7, &tmp))
 330       return false;
 331
 332     *value <<= 7;
 333     *value += tmp;
 334   }
 335
 336   return true;
 337 }
 338
 339 int MPEGAudioStreamParserBase::FindNextValidStartCode(const uint8* data,
 340                                                       int size) const {
 341   const uint8* start = data;
 342   const uint8* end = data + size;
 343
 344   while (start < end) {
 345     int bytes_left = end - start;
 346     const uint8* candidate_start_code =
 347         static_cast<const uint8*>(memchr(start, 0xff, bytes_left));
 348
 349     if (!candidate_start_code)
 350       return 0;
 351
 352     bool parse_header_failed = false;
 353     const uint8* sync = candidate_start_code;
 354     // Try to find 3 valid frames in a row. 3 was selected to decrease
 355     // the probability of false positives.
 356     for (int i = 0; i < 3; ++i) {
 357       int sync_size = end - sync;
 358       int frame_size;
 359       int sync_bytes = ParseFrameHeader(
 360           sync, sync_size, &frame_size, NULL, NULL, NULL, NULL);
 361
 362       if (sync_bytes == 0)
 363         return 0;
 364
 365       if (sync_bytes > 0) {
 366         DCHECK_LT(sync_bytes, sync_size);
 367
 368         // Skip over this frame so we can check the next one.
 369         sync += frame_size;
 370
 371         // Make sure the next frame starts inside the buffer.
 372         if (sync >= end)
 373           return 0;
 374       } else {
 375         DVLOG(1) << "ParseFrameHeader() " << i << " failed @" << (sync - data);
 376         parse_header_failed = true;
 377         break;
 378       }
 379     }
 380
 381     if (parse_header_failed) {
 382       // One of the frame header parses failed so |candidate_start_code|
 383       // did not point to the start of a real frame. Move |start| forward
 384       // so we can find the next candidate.
 385       start = candidate_start_code + 1;
 386       continue;
 387     }
 388
 389     return candidate_start_code - data;
 390   }
 391
 392   return 0;
 393 }
 394
 395 bool MPEGAudioStreamParserBase::SendBuffers(BufferQueue* buffers,
 396                                             bool end_of_segment) {
 397   DCHECK(!buffers->empty());
 398
 399   if (!in_media_segment_) {
 400     in_media_segment_ = true;
 401     new_segment_cb_.Run();
 402   }
 403
 404   BufferQueue empty_video_buffers;
 405   TextBufferQueueMap empty_text_map;
 406   if (!new_buffers_cb_.Run(*buffers, empty_video_buffers, empty_text_map))
 407     return false;
 408   buffers->clear();
 409
 410   if (end_of_segment) {
 411     in_media_segment_ = false;
 412     end_of_segment_cb_.Run();
 413   }
 414
 415   timestamp_helper_->SetBaseTimestamp(base::TimeDelta());
 416   return true;
 417 }
 418
 419 }  // namespace media