Support for parsing encrypted WebM streams by src.

media/filters/ffmpeg_demuxer.cc

 // 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/filters/ffmpeg_demuxer.h"
 
 #include <algorithm>
 #include <string>
 
+#include "base/base64.h"
 #include "base/bind.h"
 #include "base/callback.h"
 #include "base/callback_helpers.h"
 #include "base/command_line.h"
 #include "base/memory/scoped_ptr.h"
 #include "base/message_loop.h"
 #include "base/stl_util.h"
 #include "base/string_util.h"
 #include "base/task_runner_util.h"
 #include "base/time.h"
 #include "media/base/audio_decoder_config.h"
 #include "media/base/bind_to_loop.h"
 #include "media/base/decoder_buffer.h"
+#include "media/base/decrypt_config.h"
 #include "media/base/limits.h"
 #include "media/base/media_switches.h"
 #include "media/base/video_decoder_config.h"
 #include "media/ffmpeg/ffmpeg_common.h"
 #include "media/filters/ffmpeg_glue.h"
 #include "media/filters/ffmpeg_h264_to_annex_b_bitstream_converter.h"
+#include "media/webm/webm_crypto_helpers.h"
 
 namespace media {
 
 //
 // FFmpegDemuxerStream
 //
 FFmpegDemuxerStream::FFmpegDemuxerStream(
     FFmpegDemuxer* demuxer,
     AVStream* stream)
     : demuxer_(demuxer),
       message_loop_(base::MessageLoopProxy::current()),
       stream_(stream),
       type_(UNKNOWN),
       stopped_(false),
       end_of_stream_(false),
       last_packet_timestamp_(kNoTimestamp()),
       bitstream_converter_enabled_(false) {
   DCHECK(demuxer_);
 
   // Determine our media format.
   switch (stream->codec->codec_type) {
     case AVMEDIA_TYPE_AUDIO:
       type_ = AUDIO;
-      AVCodecContextToAudioDecoderConfig(stream->codec, &audio_config_);
+      AVStreamToAudioDecoderConfig(stream, &audio_config_);
       break;
     case AVMEDIA_TYPE_VIDEO:
       type_ = VIDEO;
       AVStreamToVideoDecoderConfig(stream, &video_config_);
       break;
     default:
       NOTREACHED();
       break;
   }
 
   // Calculate the duration.
   duration_ = ConvertStreamTimestamp(stream->time_base, stream->duration);
 
   if (stream_->codec->codec_id == CODEC_ID_H264) {
     bitstream_converter_.reset(
         new FFmpegH264ToAnnexBBitstreamConverter(stream_->codec));
   }
+
+  if (audio_config_.is_encrypted() || video_config_.is_encrypted()) {

[ddorwin, 2013/03/10 21:29:39]

Shouldn't we only check the one for this stream type? Maybe set a local variable
in the switch statement at 52?

[fgalligan1, 2013/03/12 00:42:42]

Done. But I think having the two configs in demuxer stream is confusing. Should
I add a todo about generalizing the configs?

[ddorwin, 2013/03/12 04:40:06]

Agreed. Not sure what the best solution is. We can talk later.

+    AVDictionaryEntry *key = av_dict_get(stream->metadata, "enc_key_id", NULL,
+                                         0);
+    if (!key || !key->value)

[ddorwin, 2013/03/10 21:29:39]

DCHECK too?

[fgalligan1, 2013/03/12 00:42:42]

Done.

+      return;
+    base::StringPiece base64_key_id(key->value);
+    std::string enc_key_id;
+    base::Base64Decode(base64_key_id, &enc_key_id);
+    if (enc_key_id.empty())
+      return;
+
+    if (stream_->codec->codec_id == CODEC_ID_VORBIS) {

[ddorwin, 2013/03/10 21:29:39]

I don't think codec matters. If we read enc_key_id, it's WebM.
I think we can simplify this block to:
  encryption_key_id_.assign(enc_key_id);
  demuxer_->FireNeedKey(kWebMEncryptInitDataType, enc_key_id);

[fgalligan1, 2013/03/12 00:42:42]

The code I added to FFmpeg is not WebM specific. It will parse the EncKeyID
element and base64 encoded that data into enc_key_id. I removed the checks as I
was probably just being overly cautious. Done on the second part per the comment
from the h file.

[ddorwin, 2013/03/12 04:40:06]

In what cases would we do the wrong thing? Matroska files using enc_key_id? Some
other container that decides to use enc_key_id? For the first case, I think it's
probably fine since the goal is to tell the app how to interpret the data.
We don't really have a way to deterministically determine it here as far as I
can tell.

+      audio_encryption_key_id_.assign(enc_key_id);
+      demuxer_->FireNeedKey(kWebMEncryptInitDataType, enc_key_id);
+    } else if (stream_->codec->codec_id == CODEC_ID_VP8) {
+      video_encryption_key_id_.assign(enc_key_id);
+      demuxer_->FireNeedKey(kWebMEncryptInitDataType, enc_key_id);
+    }
+  }
 }
 
 void FFmpegDemuxerStream::EnqueuePacket(ScopedAVPacket packet) {
   DCHECK(message_loop_->BelongsToCurrentThread());
 
   if (stopped_ || end_of_stream_) {
     NOTREACHED() << "Attempted to enqueue packet on a stopped stream";
     return;
   }
 
   // Convert the packet if there is a bitstream filter.
   if (packet->data && bitstream_converter_enabled_ &&
       !bitstream_converter_->ConvertPacket(packet.get())) {
     LOG(ERROR) << "Format conversion failed.";
   }
 
   // If a packet is returned by FFmpeg's av_parser_parse2() the packet will
   // reference inner memory of FFmpeg.  As such we should transfer the packet
   // into memory we control.
   scoped_refptr<DecoderBuffer> buffer;
   buffer = DecoderBuffer::CopyFrom(packet->data, packet->size);
+
+  if (type() == DemuxerStream::AUDIO && audio_config_.is_encrypted() &&

[ddorwin, 2013/03/10 21:29:39]

I think we can collapse this to:
if (type() == DemuxerStream::AUDIO && audio_config_.is_encrypted() ||
    type() == DemuxerStream::VIDEO && video_config_.is_encrypted()) {
  buffer->SetDecryptConfig(create_decrtyp_config_cb_(...));
}

[fgalligan1, 2013/03/12 00:42:42]

Done.

+      stream_->codec->codec_id == CODEC_ID_VORBIS) {

[ddorwin, 2013/03/10 21:29:39]

Checking codecs doesn't scale. I think we may need to save the container type or
a callback to create the DecryptConfig in the constructor.

[fgalligan1, 2013/03/12 00:42:42]

I removed the checks.

+    buffer->SetDecryptConfig(WebMCreateDecryptConfig(
+        packet->data, packet->size,
+        reinterpret_cast<const uint8*>(audio_encryption_key_id_.data()),
+        audio_encryption_key_id_.size()));
+  } else if (type() == DemuxerStream::VIDEO && video_config_.is_encrypted() &&
+             stream_->codec->codec_id == CODEC_ID_VP8) {
+    buffer->SetDecryptConfig(WebMCreateDecryptConfig(
+         packet->data, packet->size,
+         reinterpret_cast<const uint8*>(video_encryption_key_id_.data()),
+         video_encryption_key_id_.size()));
+  }
+
   buffer->SetTimestamp(ConvertStreamTimestamp(
       stream_->time_base, packet->pts));
   buffer->SetDuration(ConvertStreamTimestamp(
       stream_->time_base, packet->duration));
   if (buffer->GetTimestamp() != kNoTimestamp() &&
       last_packet_timestamp_ != kNoTimestamp() &&
       last_packet_timestamp_ < buffer->GetTimestamp()) {
     buffered_ranges_.Add(last_packet_timestamp_, buffer->GetTimestamp());
     demuxer_->NotifyBufferingChanged();
   }
@@ skipping 123 matching lines @@
     return kNoTimestamp();
 
   return ConvertFromTimeBase(time_base, timestamp);
 }
 
 //
 // FFmpegDemuxer
 //
 FFmpegDemuxer::FFmpegDemuxer(
     const scoped_refptr<base::MessageLoopProxy>& message_loop,
-    const scoped_refptr<DataSource>& data_source)
+    const scoped_refptr<DataSource>& data_source,
+    const FFmpegNeedKeyCB& need_key_cb)
     : host_(NULL),
       message_loop_(message_loop),
       blocking_thread_("FFmpegDemuxer"),
       pending_read_(false),
       pending_seek_(false),
       data_source_(data_source),
       bitrate_(0),
       start_time_(kNoTimestamp()),
       audio_disabled_(false),
       duration_known_(false),
       url_protocol_(data_source, BindToLoop(message_loop_, base::Bind(
-          &FFmpegDemuxer::OnDataSourceError, base::Unretained(this)))) {
+          &FFmpegDemuxer::OnDataSourceError, base::Unretained(this)))),
+      need_key_cb_(need_key_cb) {
   DCHECK(message_loop_);
   DCHECK(data_source_);
 }
 
 FFmpegDemuxer::~FFmpegDemuxer() {}
 
 void FFmpegDemuxer::Stop(const base::Closure& callback) {
   DCHECK(message_loop_->BelongsToCurrentThread());
   url_protocol_.Abort();
   data_source_->Stop(BindToCurrentLoop(base::Bind(
@@ skipping 397 matching lines @@
   StreamVector::iterator iter;
   for (iter = streams_.begin(); iter != streams_.end(); ++iter) {
     if (!*iter ||
         (audio_disabled_ && (*iter)->type() == DemuxerStream::AUDIO)) {
       continue;
     }
     (*iter)->SetEndOfStream();
   }
 }
 
+void FFmpegDemuxer::FireNeedKey(const std::string& init_data_type,
+                                const std::string& encryption_key_id) {
+  int key_id_size = encryption_key_id.size();
+  scoped_array<uint8> key_id_local(new uint8[key_id_size]);
+  memcpy(key_id_local.get(), encryption_key_id.data(), key_id_size);
+  need_key_cb_.Run(init_data_type, key_id_local.Pass(), key_id_size);
+}
+
 void FFmpegDemuxer::NotifyCapacityAvailable() {
   DCHECK(message_loop_->BelongsToCurrentThread());
   ReadFrameIfNeeded();
 }
 
 void FFmpegDemuxer::NotifyBufferingChanged() {
   DCHECK(message_loop_->BelongsToCurrentThread());
   Ranges<base::TimeDelta> buffered;
   scoped_refptr<FFmpegDemuxerStream> audio =
       audio_disabled_ ? NULL : GetFFmpegStream(DemuxerStream::AUDIO);
   scoped_refptr<FFmpegDemuxerStream> video =
       GetFFmpegStream(DemuxerStream::VIDEO);
   if (audio && video) {
     buffered = audio->GetBufferedRanges().IntersectionWith(
         video->GetBufferedRanges());
   } else if (audio) {
     buffered = audio->GetBufferedRanges();
   } else if (video) {
     buffered = video->GetBufferedRanges();
   }
   for (size_t i = 0; i < buffered.size(); ++i)
     host_->AddBufferedTimeRange(buffered.start(i), buffered.end(i));
 }
 
 void FFmpegDemuxer::OnDataSourceError() {
   host_->OnDemuxerError(PIPELINE_ERROR_READ);
 }
 
 }  // namespace media