media/webm/webm_cluster_parser.cc

   1 // Copyright (c) 2012 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/webm/webm_cluster_parser.h"
   6
   7 #include <vector>
   8
   9 #include "base/logging.h"
  10 #include "base/sys_byteorder.h"
  11 #include "media/base/buffers.h"
  12 #include "media/base/decrypt_config.h"
  13 #include "media/webm/webm_constants.h"
  14 #include "media/webm/webm_crypto_helpers.h"
  15
  16 namespace media {
  17
  18 WebMClusterParser::TextTrackIterator::TextTrackIterator(
  19     const TextTrackMap& text_track_map) :
  20     iterator_(text_track_map.begin()),
  21     iterator_end_(text_track_map.end()) {
  22 }
  23
  24 WebMClusterParser::TextTrackIterator::TextTrackIterator(
  25     const TextTrackIterator& rhs) :
  26     iterator_(rhs.iterator_),
  27     iterator_end_(rhs.iterator_end_) {
  28 }
  29
  30 WebMClusterParser::TextTrackIterator::~TextTrackIterator() {
  31 }
  32
  33 bool WebMClusterParser::TextTrackIterator::operator()(
  34   int* track_num,
  35   const BufferQueue** buffers) {
  36   if (iterator_ == iterator_end_) {
  37     *track_num = 0;
  38     *buffers = NULL;
  39
  40     return false;
  41   }
  42
  43   *track_num = iterator_->first;
  44   *buffers = &iterator_->second.buffers();
  45
  46   ++iterator_;
  47   return true;
  48 }
  49
  50 WebMClusterParser::WebMClusterParser(
  51     int64 timecode_scale, int audio_track_num, int video_track_num,
  52     const WebMTracksParser::TextTracks& text_tracks,
  53     const std::set<int64>& ignored_tracks,
  54     const std::string& audio_encryption_key_id,
  55     const std::string& video_encryption_key_id,
  56     const LogCB& log_cb)
  57     : timecode_multiplier_(timecode_scale / 1000.0),
  58       ignored_tracks_(ignored_tracks),
  59       audio_encryption_key_id_(audio_encryption_key_id),
  60       video_encryption_key_id_(video_encryption_key_id),
  61       parser_(kWebMIdCluster, this),
  62       last_block_timecode_(-1),
  63       block_data_size_(-1),
  64       block_duration_(-1),
  65       block_add_id_(-1),
  66       block_additional_data_size_(-1),
  67       discard_padding_(-1),
  68       cluster_timecode_(-1),
  69       cluster_start_time_(kNoTimestamp()),
  70       cluster_ended_(false),
  71       audio_(audio_track_num, false),
  72       video_(video_track_num, true),
  73       log_cb_(log_cb) {
  74   for (WebMTracksParser::TextTracks::const_iterator it = text_tracks.begin();
  75        it != text_tracks.end();
  76        ++it) {
  77     text_track_map_.insert(std::make_pair(it->first, Track(it->first, false)));
  78   }
  79 }
  80
  81 WebMClusterParser::~WebMClusterParser() {}
  82
  83 void WebMClusterParser::Reset() {
  84   last_block_timecode_ = -1;
  85   cluster_timecode_ = -1;
  86   cluster_start_time_ = kNoTimestamp();
  87   cluster_ended_ = false;
  88   parser_.Reset();
  89   audio_.Reset();
  90   video_.Reset();
  91   ResetTextTracks();
  92 }
  93
  94 int WebMClusterParser::Parse(const uint8* buf, int size) {
  95   audio_.Reset();
  96   video_.Reset();
  97   ResetTextTracks();
  98
  99   int result = parser_.Parse(buf, size);
 100
 101   if (result < 0) {
 102     cluster_ended_ = false;
 103     return result;
 104   }
 105
 106   cluster_ended_ = parser_.IsParsingComplete();
 107   if (cluster_ended_) {
 108     // If there were no buffers in this cluster, set the cluster start time to
 109     // be the |cluster_timecode_|.
 110     if (cluster_start_time_ == kNoTimestamp()) {
 111       DCHECK_GT(cluster_timecode_, -1);
 112       cluster_start_time_ = base::TimeDelta::FromMicroseconds(
 113           cluster_timecode_ * timecode_multiplier_);
 114     }
 115
 116     // Reset the parser if we're done parsing so that
 117     // it is ready to accept another cluster on the next
 118     // call.
 119     parser_.Reset();
 120
 121     last_block_timecode_ = -1;
 122     cluster_timecode_ = -1;
 123   }
 124
 125   return result;
 126 }
 127
 128 WebMClusterParser::TextTrackIterator
 129 WebMClusterParser::CreateTextTrackIterator() const {
 130   return TextTrackIterator(text_track_map_);
 131 }
 132
 133 WebMParserClient* WebMClusterParser::OnListStart(int id) {
 134   if (id == kWebMIdCluster) {
 135     cluster_timecode_ = -1;
 136     cluster_start_time_ = kNoTimestamp();
 137   } else if (id == kWebMIdBlockGroup) {
 138     block_data_.reset();
 139     block_data_size_ = -1;
 140     block_duration_ = -1;
 141     discard_padding_ = -1;
 142     discard_padding_set_ = false;
 143   } else if (id == kWebMIdBlockAdditions) {
 144     block_add_id_ = -1;
 145     block_additional_data_.reset();
 146     block_additional_data_size_ = -1;
 147   }
 148
 149   return this;
 150 }
 151
 152 bool WebMClusterParser::OnListEnd(int id) {
 153   if (id != kWebMIdBlockGroup)
 154     return true;
 155
 156   // Make sure the BlockGroup actually had a Block.
 157   if (block_data_size_ == -1) {
 158     MEDIA_LOG(log_cb_) << "Block missing from BlockGroup.";
 159     return false;
 160   }
 161
 162   bool result = ParseBlock(false, block_data_.get(), block_data_size_,
 163                            block_additional_data_.get(),
 164                            block_additional_data_size_, block_duration_,
 165                            discard_padding_set_ ? discard_padding_ : 0);
 166   block_data_.reset();
 167   block_data_size_ = -1;
 168   block_duration_ = -1;
 169   block_add_id_ = -1;
 170   block_additional_data_.reset();
 171   block_additional_data_size_ = -1;
 172   discard_padding_ = -1;
 173   discard_padding_set_ = false;
 174   return result;
 175 }
 176
 177 bool WebMClusterParser::OnUInt(int id, int64 val) {
 178   int64* dst;
 179   switch (id) {
 180     case kWebMIdTimecode:
 181       dst = &cluster_timecode_;
 182       break;
 183     case kWebMIdBlockDuration:
 184       dst = &block_duration_;
 185       break;
 186     case kWebMIdBlockAddID:
 187       dst = &block_add_id_;
 188       break;
 189     case kWebMIdDiscardPadding:
 190       if (discard_padding_set_)
 191         return false;
 192       discard_padding_set_ = true;
 193       discard_padding_ = val;
 194       return true;
 195     default:
 196       return true;
 197   }
 198   if (*dst != -1)
 199     return false;
 200   *dst = val;
 201   return true;
 202 }
 203
 204 bool WebMClusterParser::ParseBlock(bool is_simple_block, const uint8* buf,
 205                                    int size, const uint8* additional,
 206                                    int additional_size, int duration,
 207                                    int64 discard_padding) {
 208   if (size < 4)
 209     return false;
 210
 211   // Return an error if the trackNum > 127. We just aren't
 212   // going to support large track numbers right now.
 213   if (!(buf[0] & 0x80)) {
 214     MEDIA_LOG(log_cb_) << "TrackNumber over 127 not supported";
 215     return false;
 216   }
 217
 218   int track_num = buf[0] & 0x7f;
 219   int timecode = buf[1] << 8 | buf[2];
 220   int flags = buf[3] & 0xff;
 221   int lacing = (flags >> 1) & 0x3;
 222
 223   if (lacing) {
 224     MEDIA_LOG(log_cb_) << "Lacing " << lacing << " is not supported yet.";
 225     return false;
 226   }
 227
 228   // Sign extend negative timecode offsets.
 229   if (timecode & 0x8000)
 230     timecode |= ~0xffff;
 231
 232   const uint8* frame_data = buf + 4;
 233   int frame_size = size - (frame_data - buf);
 234   return OnBlock(is_simple_block, track_num, timecode, duration, flags,
 235                  frame_data, frame_size, additional, additional_size,
 236                  discard_padding);
 237 }
 238
 239 bool WebMClusterParser::OnBinary(int id, const uint8* data, int size) {
 240   switch (id) {
 241     case kWebMIdSimpleBlock:
 242       return ParseBlock(true, data, size, NULL, -1, -1, 0);
 243
 244     case kWebMIdBlock:
 245       if (block_data_) {
 246         MEDIA_LOG(log_cb_) << "More than 1 Block in a BlockGroup is not "
 247                               "supported.";
 248         return false;
 249       }
 250       block_data_.reset(new uint8[size]);
 251       memcpy(block_data_.get(), data, size);
 252       block_data_size_ = size;
 253       return true;
 254
 255     case kWebMIdBlockAdditional: {
 256       uint64 block_add_id = base::HostToNet64(block_add_id_);
 257       if (block_additional_data_) {
 258         // TODO(vigneshv): Technically, more than 1 BlockAdditional is allowed
 259         // as per matroska spec. But for now we don't have a use case to
 260         // support parsing of such files. Take a look at this again when such a
 261         // case arises.
 262         MEDIA_LOG(log_cb_) << "More than 1 BlockAdditional in a BlockGroup is "
 263                               "not supported.";
 264         return false;
 265       }
 266       // First 8 bytes of side_data in DecoderBuffer is the BlockAddID
 267       // element's value in Big Endian format. This is done to mimic ffmpeg
 268       // demuxer's behavior.
 269       block_additional_data_size_ = size + sizeof(block_add_id);
 270       block_additional_data_.reset(new uint8[block_additional_data_size_]);
 271       memcpy(block_additional_data_.get(), &block_add_id,
 272              sizeof(block_add_id));
 273       memcpy(block_additional_data_.get() + 8, data, size);
 274       return true;
 275     }
 276
 277     default:
 278       return true;
 279   }
 280 }
 281
 282 bool WebMClusterParser::OnBlock(bool is_simple_block, int track_num,
 283                                 int timecode,
 284                                 int  block_duration,
 285                                 int flags,
 286                                 const uint8* data, int size,
 287                                 const uint8* additional, int additional_size,
 288                                 int64 discard_padding) {
 289   DCHECK_GE(size, 0);
 290   if (cluster_timecode_ == -1) {
 291     MEDIA_LOG(log_cb_) << "Got a block before cluster timecode.";
 292     return false;
 293   }
 294
 295   // TODO(acolwell): Should relative negative timecode offsets be rejected?  Or
 296   // only when the absolute timecode is negative?  See http://crbug.com/271794
 297   if (timecode < 0) {
 298     MEDIA_LOG(log_cb_) << "Got a block with negative timecode offset "
 299                        << timecode;
 300     return false;
 301   }
 302
 303   if (last_block_timecode_ != -1 && timecode < last_block_timecode_) {
 304     MEDIA_LOG(log_cb_)
 305         << "Got a block with a timecode before the previous block.";
 306     return false;
 307   }
 308
 309   Track* track = NULL;
 310   std::string encryption_key_id;
 311   if (track_num == audio_.track_num()) {
 312     track = &audio_;
 313     encryption_key_id = audio_encryption_key_id_;
 314   } else if (track_num == video_.track_num()) {
 315     track = &video_;
 316     encryption_key_id = video_encryption_key_id_;
 317   } else if (ignored_tracks_.find(track_num) != ignored_tracks_.end()) {
 318     return true;
 319   } else if (Track* const text_track = FindTextTrack(track_num)) {
 320     if (is_simple_block)  // BlockGroup is required for WebVTT cues
 321       return false;
 322     if (block_duration < 0)  // not specified
 323       return false;
 324     track = text_track;
 325   } else {
 326     MEDIA_LOG(log_cb_) << "Unexpected track number " << track_num;
 327     return false;
 328   }
 329
 330   last_block_timecode_ = timecode;
 331
 332   base::TimeDelta timestamp = base::TimeDelta::FromMicroseconds(
 333       (cluster_timecode_ + timecode) * timecode_multiplier_);
 334
 335   // The first bit of the flags is set when a SimpleBlock contains only
 336   // keyframes. If this is a Block, then inspection of the payload is
 337   // necessary to determine whether it contains a keyframe or not.
 338   // http://www.matroska.org/technical/specs/index.html
 339   bool is_keyframe =
 340       is_simple_block ? (flags & 0x80) != 0 : track->IsKeyframe(data, size);
 341
 342   scoped_refptr<StreamParserBuffer> buffer =
 343       StreamParserBuffer::CopyFrom(data, size, additional, additional_size,
 344                                    is_keyframe);
 345
 346   // Every encrypted Block has a signal byte and IV prepended to it. Current
 347   // encrypted WebM request for comments specification is here
 348   // http://wiki.webmproject.org/encryption/webm-encryption-rfc
 349   if (!encryption_key_id.empty()) {
 350     scoped_ptr<DecryptConfig> config(WebMCreateDecryptConfig(
 351         data, size,
 352         reinterpret_cast<const uint8*>(encryption_key_id.data()),
 353         encryption_key_id.size()));
 354     if (!config)
 355       return false;
 356     buffer->set_decrypt_config(config.Pass());
 357   }
 358
 359   buffer->set_timestamp(timestamp);
 360   if (cluster_start_time_ == kNoTimestamp())
 361     cluster_start_time_ = timestamp;
 362
 363   if (block_duration >= 0) {
 364     buffer->set_duration(base::TimeDelta::FromMicroseconds(
 365         block_duration * timecode_multiplier_));
 366   }
 367
 368   if (discard_padding != 0) {
 369     buffer->set_discard_padding(base::TimeDelta::FromMicroseconds(
 370                                     discard_padding / 1000));
 371   }
 372
 373   return track->AddBuffer(buffer);
 374 }
 375
 376 WebMClusterParser::Track::Track(int track_num, bool is_video)
 377     : track_num_(track_num),
 378       is_video_(is_video) {
 379 }
 380
 381 WebMClusterParser::Track::~Track() {}
 382
 383 bool WebMClusterParser::Track::AddBuffer(
 384     const scoped_refptr<StreamParserBuffer>& buffer) {
 385   DVLOG(2) << "AddBuffer() : " << track_num_
 386            << " ts " << buffer->timestamp().InSecondsF()
 387            << " dur " << buffer->duration().InSecondsF()
 388            << " kf " << buffer->IsKeyframe()
 389            << " size " << buffer->data_size();
 390
 391   buffers_.push_back(buffer);
 392   return true;
 393 }
 394
 395 void WebMClusterParser::Track::Reset() {
 396   buffers_.clear();
 397 }
 398
 399 bool WebMClusterParser::Track::IsKeyframe(const uint8* data, int size) const {
 400   // For now, assume that all blocks are keyframes for datatypes other than
 401   // video. This is a valid assumption for Vorbis, WebVTT, & Opus.
 402   if (!is_video_)
 403     return true;
 404
 405   // Make sure the block is big enough for the minimal keyframe header size.
 406   if (size < 7)
 407     return false;
 408
 409   // The LSb of the first byte must be a 0 for a keyframe.
 410   // http://tools.ietf.org/html/rfc6386 Section 19.1
 411   if ((data[0] & 0x01) != 0)
 412     return false;
 413
 414   // Verify VP8 keyframe startcode.
 415   // http://tools.ietf.org/html/rfc6386 Section 19.1
 416   if (data[3] != 0x9d || data[4] != 0x01 || data[5] != 0x2a)
 417     return false;
 418
 419   return true;
 420 }
 421
 422 void WebMClusterParser::ResetTextTracks() {
 423   for (TextTrackMap::iterator it = text_track_map_.begin();
 424        it != text_track_map_.end();
 425        ++it) {
 426     it->second.Reset();
 427   }
 428 }
 429
 430 WebMClusterParser::Track*
 431 WebMClusterParser::FindTextTrack(int track_num) {
 432   const TextTrackMap::iterator it = text_track_map_.find(track_num);
 433
 434   if (it == text_track_map_.end())
 435     return NULL;
 436
 437   return &it->second;
 438 }
 439
 440 }  // namespace media