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/buffers.h"
  11 #include "media/base/stream_parser_buffer.h"
  12 #include "media/base/text_track_config.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(const InitCB& init_cb,
  58                                      const NewConfigCB& config_cb,
  59                                      const NewBuffersCB& new_buffers_cb,
  60                                      bool ignore_text_tracks,
  61                                      const NeedKeyCB& need_key_cb,
  62                                      const NewMediaSegmentCB& new_segment_cb,
  63                                      const base::Closure& end_of_segment_cb,
  64                                      const LogCB& log_cb) {
  65   DVLOG(1) << __FUNCTION__;
  66   DCHECK_EQ(state_, UNINITIALIZED);
  67   init_cb_ = init_cb;
  68   config_cb_ = config_cb;
  69   new_buffers_cb_ = new_buffers_cb;
  70   new_segment_cb_ = new_segment_cb;
  71   end_of_segment_cb_ = end_of_segment_cb;
  72   log_cb_ = log_cb;
  73
  74   ChangeState(INITIALIZED);
  75 }
  76
  77 void MPEGAudioStreamParserBase::Flush() {
  78   DVLOG(1) << __FUNCTION__;
  79   DCHECK_NE(state_, UNINITIALIZED);
  80   queue_.Reset();
  81   timestamp_helper_->SetBaseTimestamp(base::TimeDelta());
  82   in_media_segment_ = false;
  83 }
  84
  85 bool MPEGAudioStreamParserBase::Parse(const uint8* buf, int size) {
  86   DVLOG(1) << __FUNCTION__ << "(" << size << ")";
  87   DCHECK(buf);
  88   DCHECK_GT(size, 0);
  89   DCHECK_NE(state_, UNINITIALIZED);
  90
  91   if (state_ == PARSE_ERROR)
  92     return false;
  93
  94   DCHECK_EQ(state_, INITIALIZED);
  95
  96   queue_.Push(buf, size);
  97
  98   bool end_of_segment = true;
  99   BufferQueue buffers;
 100   for (;;) {
 101     const uint8* data;
 102     int data_size;
 103     queue_.Peek(&data, &data_size);
 104
 105     if (data_size < 4)
 106       break;
 107
 108     uint32 start_code = data[0] << 24 | data[1] << 16 | data[2] << 8 | data[3];
 109     int bytes_read = 0;
 110     bool parsed_metadata = true;
 111     if ((start_code & start_code_mask_) == start_code_mask_) {
 112       bytes_read = ParseFrame(data, data_size, &buffers);
 113
 114       // Only allow the current segment to end if a full frame has been parsed.
 115       end_of_segment = bytes_read > 0;
 116       parsed_metadata = false;
 117     } else if (start_code == kICYStartCode) {
 118       bytes_read = ParseIcecastHeader(data, data_size);
 119     } else if ((start_code & kID3StartCodeMask) == kID3v1StartCode) {
 120       bytes_read = ParseID3v1(data, data_size);
 121     } else if ((start_code & kID3StartCodeMask) == kID3v2StartCode) {
 122       bytes_read = ParseID3v2(data, data_size);
 123     } else {
 124       bytes_read = FindNextValidStartCode(data, data_size);
 125
 126       if (bytes_read > 0) {
 127         DVLOG(1) << "Unexpected start code 0x" << std::hex << start_code;
 128         DVLOG(1) << "SKIPPING " << bytes_read << " bytes of garbage.";
 129       }
 130     }
 131
 132     CHECK_LE(bytes_read, data_size);
 133
 134     if (bytes_read < 0) {
 135       ChangeState(PARSE_ERROR);
 136       return false;
 137     } else if (bytes_read == 0) {
 138       // Need more data.
 139       break;
 140     }
 141
 142     // Send pending buffers if we have encountered metadata.
 143     if (parsed_metadata && !buffers.empty() && !SendBuffers(&buffers, true))
 144       return false;
 145
 146     queue_.Pop(bytes_read);
 147     end_of_segment = true;
 148   }
 149
 150   if (buffers.empty())
 151     return true;
 152
 153   // Send buffers collected in this append that haven't been sent yet.
 154   return SendBuffers(&buffers, end_of_segment);
 155 }
 156
 157 void MPEGAudioStreamParserBase::ChangeState(State state) {
 158   DVLOG(1) << __FUNCTION__ << "() : " << state_ << " -> " << state;
 159   state_ = state;
 160 }
 161
 162 int MPEGAudioStreamParserBase::ParseFrame(const uint8* data,
 163                                           int size,
 164                                           BufferQueue* buffers) {
 165   DVLOG(2) << __FUNCTION__ << "(" << size << ")";
 166
 167   int sample_rate;
 168   ChannelLayout channel_layout;
 169   int frame_size;
 170   int sample_count;
 171   bool metadata_frame = false;
 172   int bytes_read = ParseFrameHeader(data,
 173                                     size,
 174                                     &frame_size,
 175                                     &sample_rate,
 176                                     &channel_layout,
 177                                     &sample_count,
 178                                     &metadata_frame);
 179
 180   if (bytes_read <= 0)
 181     return bytes_read;
 182
 183   // Make sure data contains the entire frame.
 184   if (size < frame_size)
 185     return 0;
 186
 187   DVLOG(2) << " sample_rate " << sample_rate
 188            << " channel_layout " << channel_layout
 189            << " frame_size " << frame_size
 190            << " sample_count " << sample_count;
 191
 192   if (config_.IsValidConfig() &&
 193       (config_.samples_per_second() != sample_rate ||
 194        config_.channel_layout() != channel_layout)) {
 195     // Clear config data so that a config change is initiated.
 196     config_ = AudioDecoderConfig();
 197
 198     // Send all buffers associated with the previous config.
 199     if (!buffers->empty() && !SendBuffers(buffers, true))
 200       return -1;
 201   }
 202
 203   if (!config_.IsValidConfig()) {
 204     config_.Initialize(audio_codec_,
 205                        kSampleFormatF32,
 206                        channel_layout,
 207                        sample_rate,
 208                        NULL,
 209                        0,
 210                        false,
 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     bool success = config_cb_.Run(config_, video_config, TextTrackConfigMap());
 224
 225     if (!init_cb_.is_null()) {
 226       InitParameters params(kInfiniteDuration());
 227       params.auto_update_timestamp_offset = true;
 228       base::ResetAndReturn(&init_cb_).Run(success, params);
 229     }
 230
 231     if (!success)
 232       return -1;
 233   }
 234
 235   if (metadata_frame)
 236     return frame_size;
 237
 238   // TODO(wolenetz/acolwell): Validate and use a common cross-parser TrackId
 239   // type and allow multiple audio tracks, if applicable. See
 240   // https://crbug.com/341581.
 241   scoped_refptr<StreamParserBuffer> buffer =
 242       StreamParserBuffer::CopyFrom(data, frame_size, true,
 243                                    DemuxerStream::AUDIO, 0);
 244   buffer->set_timestamp(timestamp_helper_->GetTimestamp());
 245   buffer->set_duration(timestamp_helper_->GetFrameDuration(sample_count));
 246   buffers->push_back(buffer);
 247
 248   timestamp_helper_->AddFrames(sample_count);
 249
 250   return frame_size;
 251 }
 252
 253 int MPEGAudioStreamParserBase::ParseIcecastHeader(const uint8* data, int size) {
 254   DVLOG(1) << __FUNCTION__ << "(" << size << ")";
 255
 256   if (size < 4)
 257     return 0;
 258
 259   if (memcmp("ICY ", data, 4))
 260     return -1;
 261
 262   int locate_size = std::min(size, kMaxIcecastHeaderSize);
 263   int offset = LocateEndOfHeaders(data, locate_size, 4);
 264   if (offset < 0) {
 265     if (locate_size == kMaxIcecastHeaderSize) {
 266       MEDIA_LOG(log_cb_) << "Icecast header is too large.";
 267       return -1;
 268     }
 269
 270     return 0;
 271   }
 272
 273   return offset;
 274 }
 275
 276 int MPEGAudioStreamParserBase::ParseID3v1(const uint8* data, int size) {
 277   DVLOG(1) << __FUNCTION__ << "(" << size << ")";
 278
 279   if (size < kID3v1Size)
 280     return 0;
 281
 282   // TODO(acolwell): Add code to actually validate ID3v1 data and
 283   // expose it as a metadata text track.
 284   return !memcmp(data, "TAG+", 4) ? kID3v1ExtendedSize : kID3v1Size;
 285 }
 286
 287 int MPEGAudioStreamParserBase::ParseID3v2(const uint8* data, int size) {
 288   DVLOG(1) << __FUNCTION__ << "(" << size << ")";
 289
 290   if (size < 10)
 291     return 0;
 292
 293   BitReader reader(data, size);
 294   int32 id;
 295   int version;
 296   uint8 flags;
 297   int32 id3_size;
 298
 299   if (!reader.ReadBits(24, &id) ||
 300       !reader.ReadBits(16, &version) ||
 301       !reader.ReadBits(8, &flags) ||
 302       !ParseSyncSafeInt(&reader, &id3_size)) {
 303     return -1;
 304   }
 305
 306   int32 actual_tag_size = 10 + id3_size;
 307
 308   // Increment size if 'Footer present' flag is set.
 309   if (flags & 0x10)
 310     actual_tag_size += 10;
 311
 312   // Make sure we have the entire tag.
 313   if (size < actual_tag_size)
 314     return 0;
 315
 316   // TODO(acolwell): Add code to actually validate ID3v2 data and
 317   // expose it as a metadata text track.
 318   return actual_tag_size;
 319 }
 320
 321 bool MPEGAudioStreamParserBase::ParseSyncSafeInt(BitReader* reader,
 322                                                  int32* value) {
 323   *value = 0;
 324   for (int i = 0; i < 4; ++i) {
 325     uint8 tmp;
 326     if (!reader->ReadBits(1, &tmp) || tmp != 0) {
 327       MEDIA_LOG(log_cb_) << "ID3 syncsafe integer byte MSb is not 0!";
 328       return false;
 329     }
 330
 331     if (!reader->ReadBits(7, &tmp))
 332       return false;
 333
 334     *value <<= 7;
 335     *value += tmp;
 336   }
 337
 338   return true;
 339 }
 340
 341 int MPEGAudioStreamParserBase::FindNextValidStartCode(const uint8* data,
 342                                                       int size) const {
 343   const uint8* start = data;
 344   const uint8* end = data + size;
 345
 346   while (start < end) {
 347     int bytes_left = end - start;
 348     const uint8* candidate_start_code =
 349         static_cast<const uint8*>(memchr(start, 0xff, bytes_left));
 350
 351     if (!candidate_start_code)
 352       return 0;
 353
 354     bool parse_header_failed = false;
 355     const uint8* sync = candidate_start_code;
 356     // Try to find 3 valid frames in a row. 3 was selected to decrease
 357     // the probability of false positives.
 358     for (int i = 0; i < 3; ++i) {
 359       int sync_size = end - sync;
 360       int frame_size;
 361       int sync_bytes = ParseFrameHeader(
 362           sync, sync_size, &frame_size, NULL, NULL, NULL, NULL);
 363
 364       if (sync_bytes == 0)
 365         return 0;
 366
 367       if (sync_bytes > 0) {
 368         DCHECK_LT(sync_bytes, sync_size);
 369
 370         // Skip over this frame so we can check the next one.
 371         sync += frame_size;
 372
 373         // Make sure the next frame starts inside the buffer.
 374         if (sync >= end)
 375           return 0;
 376       } else {
 377         DVLOG(1) << "ParseFrameHeader() " << i << " failed @" << (sync - data);
 378         parse_header_failed = true;
 379         break;
 380       }
 381     }
 382
 383     if (parse_header_failed) {
 384       // One of the frame header parses failed so |candidate_start_code|
 385       // did not point to the start of a real frame. Move |start| forward
 386       // so we can find the next candidate.
 387       start = candidate_start_code + 1;
 388       continue;
 389     }
 390
 391     return candidate_start_code - data;
 392   }
 393
 394   return 0;
 395 }
 396
 397 bool MPEGAudioStreamParserBase::SendBuffers(BufferQueue* buffers,
 398                                             bool end_of_segment) {
 399   DCHECK(!buffers->empty());
 400
 401   if (!in_media_segment_) {
 402     in_media_segment_ = true;
 403     new_segment_cb_.Run();
 404   }
 405
 406   BufferQueue empty_video_buffers;
 407   TextBufferQueueMap empty_text_map;
 408   if (!new_buffers_cb_.Run(*buffers, empty_video_buffers, empty_text_map))
 409     return false;
 410   buffers->clear();
 411
 412   if (end_of_segment) {
 413     in_media_segment_ = false;
 414     end_of_segment_cb_.Run();
 415   }
 416
 417   timestamp_helper_->SetBaseTimestamp(base::TimeDelta());
 418   return true;
 419 }
 420
 421 }  // namespace media