Implement SampleEncryption box parsing and generation

Issue #41

Change-Id: Ic57436b6e6c8b58fc264037fd81c98627c525932

packager/app/test/testdata/bear-640x360-a-cenc-golden.mp4.media_info

@@ -1,4 +1,4 @@
-bandwidth: 128635
+bandwidth: 128728
 audio_info {
   codec: "mp4a.40.2"
   sampling_frequency: 44100

packager/app/test/testdata/bear-640x360-av-cenc-golden.mpd

@@ -2,7 +2,7 @@
 <MPD xmlns="urn:mpeg:DASH:schema:MPD:2011" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xlink="http://www.w3.org/1999/xlink" xsi:schemaLocation="urn:mpeg:DASH:schema:MPD:2011 DASH-MPD.xsd" xmlns:cenc="urn:mpeg:cenc:2013" minBufferTime="PT2S" type="static" profiles="urn:mpeg:dash:profile:isoff-on-demand:2011" mediaPresentationDuration="PT2.763174533843994S">
   <Period>
     <AdaptationSet id="0" contentType="video" width="640" height="360" frameRate="30000/1001" subSegmentAlignment="true" par="16:9">
-      <Representation id="0" bandwidth="885058" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360" frameRate="30000/1001">
+      <Representation id="0" bandwidth="885151" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360" frameRate="30000/1001">
         <ContentProtection value="cenc" schemeIdUri="urn:mpeg:dash:mp4protection:2011" cenc:default_KID="31323334-3536-3738-3930-313233343536"/>
         <ContentProtection schemeIdUri="urn:uuid:edef8ba9-79d6-4ace-a3c8-27dcd51d21ed">
           <cenc:pssh>AAAAMHBzc2gAAAAA7e+LqXnWSs6jyCfc1R0h7QAAABAxMjM0NTY3ODkwMTIzNDU2</cenc:pssh>
@@ -14,7 +14,7 @@
       </Representation>
     </AdaptationSet>
     <AdaptationSet id="1" contentType="audio" subSegmentAlignment="true">
-      <Representation id="1" bandwidth="128635" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
+      <Representation id="1" bandwidth="128728" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
         <AudioChannelConfiguration schemeIdUri="urn:mpeg:dash:23003:3:audio_channel_configuration:2011" value="2"/>
         <ContentProtection value="cenc" schemeIdUri="urn:mpeg:dash:mp4protection:2011" cenc:default_KID="31323334-3536-3738-3930-313233343536"/>
         <ContentProtection schemeIdUri="urn:uuid:edef8ba9-79d6-4ace-a3c8-27dcd51d21ed">

packager/app/test/testdata/bear-640x360-av-cenc-iop-golden.mpd

@@ -6,7 +6,7 @@
       <ContentProtection schemeIdUri="urn:uuid:edef8ba9-79d6-4ace-a3c8-27dcd51d21ed">
         <cenc:pssh>AAAAMHBzc2gAAAAA7e+LqXnWSs6jyCfc1R0h7QAAABAxMjM0NTY3ODkwMTIzNDU2</cenc:pssh>
       </ContentProtection>
-      <Representation id="0" bandwidth="885058" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360" frameRate="30000/1001">
+      <Representation id="0" bandwidth="885151" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360" frameRate="30000/1001">
         <BaseURL>output_video.mp4</BaseURL>
         <SegmentBase indexRange="941-1008" timescale="30000">
           <Initialization range="0-940"/>
@@ -18,7 +18,7 @@
       <ContentProtection schemeIdUri="urn:uuid:edef8ba9-79d6-4ace-a3c8-27dcd51d21ed">
         <cenc:pssh>AAAAMHBzc2gAAAAA7e+LqXnWSs6jyCfc1R0h7QAAABAxMjM0NTY3ODkwMTIzNDU2</cenc:pssh>
       </ContentProtection>
-      <Representation id="1" bandwidth="128635" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
+      <Representation id="1" bandwidth="128728" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
         <AudioChannelConfiguration schemeIdUri="urn:mpeg:dash:23003:3:audio_channel_configuration:2011" value="2"/>
         <BaseURL>output_audio.mp4</BaseURL>
         <SegmentBase indexRange="817-884" timescale="44100">

packager/app/test/testdata/bear-640x360-av-live-cenc-golden.mpd

@@ -2,7 +2,7 @@
 <MPD xmlns="urn:mpeg:DASH:schema:MPD:2011" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xlink="http://www.w3.org/1999/xlink" xsi:schemaLocation="urn:mpeg:DASH:schema:MPD:2011 DASH-MPD.xsd" xmlns:cenc="urn:mpeg:cenc:2013" minBufferTime="PT2S" type="dynamic" profiles="urn:mpeg:dash:profile:isoff-live:2011" availabilityStartTime="place_holder" minimumUpdatePeriod="PT5S" timeShiftBufferDepth="PT1800S">
   <Period start="PT0S">
     <AdaptationSet id="0" contentType="video" width="640" height="360" frameRate="30000/1001" segmentAlignment="true" par="16:9">
-      <Representation id="0" bandwidth="875528" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360" frameRate="30000/1001">
+      <Representation id="0" bandwidth="875620" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360" frameRate="30000/1001">
         <ContentProtection value="cenc" schemeIdUri="urn:mpeg:dash:mp4protection:2011" cenc:default_KID="31323334-3536-3738-3930-313233343536"/>
         <ContentProtection schemeIdUri="urn:uuid:edef8ba9-79d6-4ace-a3c8-27dcd51d21ed">
           <cenc:pssh>AAAAMHBzc2gAAAAA7e+LqXnWSs6jyCfc1R0h7QAAABAxMjM0NTY3ODkwMTIzNDU2</cenc:pssh>
@@ -16,7 +16,7 @@
       </Representation>
     </AdaptationSet>
     <AdaptationSet id="1" contentType="audio" segmentAlignment="true">
-      <Representation id="1" bandwidth="124085" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
+      <Representation id="1" bandwidth="124195" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
         <AudioChannelConfiguration schemeIdUri="urn:mpeg:dash:23003:3:audio_channel_configuration:2011" value="2"/>
         <ContentProtection value="cenc" schemeIdUri="urn:mpeg:dash:mp4protection:2011" cenc:default_KID="31323334-3536-3738-3930-313233343536"/>
         <ContentProtection schemeIdUri="urn:uuid:edef8ba9-79d6-4ace-a3c8-27dcd51d21ed">

packager/app/test/testdata/bear-640x360-av-live-cenc-iop-golden.mpd

@@ -6,7 +6,7 @@
       <ContentProtection schemeIdUri="urn:uuid:edef8ba9-79d6-4ace-a3c8-27dcd51d21ed">
         <cenc:pssh>AAAAMHBzc2gAAAAA7e+LqXnWSs6jyCfc1R0h7QAAABAxMjM0NTY3ODkwMTIzNDU2</cenc:pssh>
       </ContentProtection>
-      <Representation id="0" bandwidth="875528" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360" frameRate="30000/1001">
+      <Representation id="0" bandwidth="875620" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360" frameRate="30000/1001">
         <SegmentTemplate timescale="30000" initialization="output_video-init.mp4" media="output_video-$Number$.m4s" startNumber="1">
           <SegmentTimeline>
             <S t="2002" d="30030" r="1"/>
@@ -20,7 +20,7 @@
       <ContentProtection schemeIdUri="urn:uuid:edef8ba9-79d6-4ace-a3c8-27dcd51d21ed">
         <cenc:pssh>AAAAMHBzc2gAAAAA7e+LqXnWSs6jyCfc1R0h7QAAABAxMjM0NTY3ODkwMTIzNDU2</cenc:pssh>
       </ContentProtection>
-      <Representation id="1" bandwidth="124085" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
+      <Representation id="1" bandwidth="124195" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
         <AudioChannelConfiguration schemeIdUri="urn:mpeg:dash:23003:3:audio_channel_configuration:2011" value="2"/>
         <SegmentTemplate timescale="44100" initialization="output_audio-init.mp4" media="output_audio-$Number$.m4s" startNumber="1">
           <SegmentTimeline>

packager/app/test/testdata/bear-640x360-av-live-cenc-rotation-golden.mpd

@@ -2,7 +2,7 @@
 <MPD xmlns="urn:mpeg:DASH:schema:MPD:2011" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xlink="http://www.w3.org/1999/xlink" xsi:schemaLocation="urn:mpeg:DASH:schema:MPD:2011 DASH-MPD.xsd" xmlns:cenc="urn:mpeg:cenc:2013" minBufferTime="PT2S" type="dynamic" profiles="urn:mpeg:dash:profile:isoff-live:2011" availabilityStartTime="place_holder" minimumUpdatePeriod="PT5S" timeShiftBufferDepth="PT1800S">
   <Period start="PT0S">
     <AdaptationSet id="0" contentType="video" width="640" height="360" frameRate="30000/1001" segmentAlignment="true" par="16:9">
-      <Representation id="0" bandwidth="876377" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360" frameRate="30000/1001">
+      <Representation id="0" bandwidth="876470" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360" frameRate="30000/1001">
         <ContentProtection value="cenc" schemeIdUri="urn:mpeg:dash:mp4protection:2011" cenc:default_KID="00000000-0000-0000-0000-000000000000"/>
         <ContentProtection schemeIdUri="urn:uuid:edef8ba9-79d6-4ace-a3c8-27dcd51d21ed"/>
         <SegmentTemplate timescale="30000" initialization="output_video-init.mp4" media="output_video-$Number$.m4s" startNumber="1">
@@ -14,7 +14,7 @@
       </Representation>
     </AdaptationSet>
     <AdaptationSet id="1" contentType="audio" segmentAlignment="true">
-      <Representation id="1" bandwidth="125028" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
+      <Representation id="1" bandwidth="125138" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
         <AudioChannelConfiguration schemeIdUri="urn:mpeg:dash:23003:3:audio_channel_configuration:2011" value="2"/>
         <ContentProtection value="cenc" schemeIdUri="urn:mpeg:dash:mp4protection:2011" cenc:default_KID="00000000-0000-0000-0000-000000000000"/>
         <ContentProtection schemeIdUri="urn:uuid:edef8ba9-79d6-4ace-a3c8-27dcd51d21ed"/>

packager/app/test/testdata/bear-640x360-av-live-cenc-rotation-iop-golden.mpd

@@ -4,7 +4,7 @@
     <AdaptationSet id="0" contentType="video" width="640" height="360" frameRate="30000/1001" segmentAlignment="true" par="16:9">
       <ContentProtection value="cenc" schemeIdUri="urn:mpeg:dash:mp4protection:2011" cenc:default_KID="00000000-0000-0000-0000-000000000000"/>
       <ContentProtection schemeIdUri="urn:uuid:edef8ba9-79d6-4ace-a3c8-27dcd51d21ed"/>
-      <Representation id="0" bandwidth="876377" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360" frameRate="30000/1001">
+      <Representation id="0" bandwidth="876470" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360" frameRate="30000/1001">
         <SegmentTemplate timescale="30000" initialization="output_video-init.mp4" media="output_video-$Number$.m4s" startNumber="1">
           <SegmentTimeline>
             <S t="2002" d="30030" r="1"/>
@@ -16,7 +16,7 @@
     <AdaptationSet id="1" contentType="audio" segmentAlignment="true">
       <ContentProtection value="cenc" schemeIdUri="urn:mpeg:dash:mp4protection:2011" cenc:default_KID="00000000-0000-0000-0000-000000000000"/>
       <ContentProtection schemeIdUri="urn:uuid:edef8ba9-79d6-4ace-a3c8-27dcd51d21ed"/>
-      <Representation id="1" bandwidth="125028" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
+      <Representation id="1" bandwidth="125138" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
         <AudioChannelConfiguration schemeIdUri="urn:mpeg:dash:23003:3:audio_channel_configuration:2011" value="2"/>
         <SegmentTemplate timescale="44100" initialization="output_audio-init.mp4" media="output_audio-$Number$.m4s" startNumber="1">
           <SegmentTimeline>

packager/app/test/testdata/bear-640x360-v-cenc-golden.mp4.media_info

@@ -1,4 +1,4 @@
-bandwidth: 885058
+bandwidth: 885151
 video_info {
   codec: "avc1.64001e"
   width: 640

packager/media/formats/mp4/cenc.cc

@@ -1,93 +0,0 @@
-// 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 "packager/media/formats/mp4/cenc.h"
-
-#include <cstring>
-
-#include "packager/media/base/buffer_reader.h"
-#include "packager/media/base/buffer_writer.h"
-#include "packager/media/formats/mp4/rcheck.h"
-
-namespace {
-// According to ISO/IEC FDIS 23001-7: CENC spec, IV should be either
-// 64-bit (8-byte) or 128-bit (16-byte).
-bool IsIvSizeValid(size_t iv_size) { return iv_size == 8 || iv_size == 16; }
-
-// 16-bit |clear_bytes| and 32-bit |cipher_bytes|.
-const size_t kSubsampleEntrySize = sizeof(uint16_t) + sizeof(uint32_t);
-}  // namespace
-
-namespace edash_packager {
-namespace media {
-namespace mp4 {
-
-FrameCENCInfo::FrameCENCInfo() {}
-FrameCENCInfo::FrameCENCInfo(const std::vector<uint8_t>& iv) : iv_(iv) {
-}
-FrameCENCInfo::~FrameCENCInfo() {}
-
-bool FrameCENCInfo::Parse(uint8_t iv_size, BufferReader* reader) {
-  DCHECK(reader);
-  // Mandated by CENC spec.
-  RCHECK(IsIvSizeValid(iv_size));
-
-  iv_.resize(iv_size);
-  RCHECK(reader->ReadToVector(&iv_, iv_size));
-
-  if (!reader->HasBytes(1))
-    return true;
-
-  uint16_t subsample_count;
-  RCHECK(reader->Read2(&subsample_count) &&
-         reader->HasBytes(subsample_count * kSubsampleEntrySize));
-
-  subsamples_.resize(subsample_count);
-  for (uint16_t i = 0; i < subsample_count; ++i) {
-    uint16_t clear_bytes;
-    uint32_t cipher_bytes;
-    RCHECK(reader->Read2(&clear_bytes) &&
-           reader->Read4(&cipher_bytes));
-    subsamples_[i].clear_bytes = clear_bytes;
-    subsamples_[i].cipher_bytes = cipher_bytes;
-  }
-  return true;
-}
-
-void FrameCENCInfo::Write(BufferWriter* writer) const {
-  DCHECK(writer);
-  DCHECK(IsIvSizeValid(iv_.size()));
-  writer->AppendVector(iv_);
-
-  uint16_t subsample_count = subsamples_.size();
-  if (subsample_count == 0)
-    return;
-  writer->AppendInt(subsample_count);
-
-  for (uint16_t i = 0; i < subsample_count; ++i) {
-    writer->AppendInt(subsamples_[i].clear_bytes);
-    writer->AppendInt(subsamples_[i].cipher_bytes);
-  }
-}
-
-size_t FrameCENCInfo::ComputeSize() const {
-  uint16_t subsample_count = subsamples_.size();
-  if (subsample_count == 0)
-    return iv_.size();
-
-  return iv_.size() + sizeof(subsample_count) +
-         subsample_count * kSubsampleEntrySize;
-}
-
-size_t FrameCENCInfo::GetTotalSizeOfSubsamples() const {
-  size_t size = 0;
-  for (size_t i = 0; i < subsamples_.size(); ++i) {
-    size += subsamples_[i].clear_bytes + subsamples_[i].cipher_bytes;
-  }
-  return size;
-}
-
-}  // namespace mp4
-}  // namespace media
-}  // namespace edash_packager

packager/media/formats/mp4/cenc.h

@@ -1,52 +0,0 @@
-// 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.
-
-#ifndef MEDIA_FORMATS_MP4_CENC_H_
-#define MEDIA_FORMATS_MP4_CENC_H_
-
-#include <stdint.h>
-
-#include <vector>
-
-#include "packager/media/base/decrypt_config.h"
-
-namespace edash_packager {
-namespace media {
-
-class BufferReader;
-class BufferWriter;
-
-namespace mp4 {
-
-class FrameCENCInfo {
- public:
-  FrameCENCInfo();
-  explicit FrameCENCInfo(const std::vector<uint8_t>& iv);
-  ~FrameCENCInfo();
-
-  bool Parse(uint8_t iv_size, BufferReader* reader);
-  void Write(BufferWriter* writer) const;
-  size_t ComputeSize() const;
-  size_t GetTotalSizeOfSubsamples() const;
-
-  void AddSubsample(const SubsampleEntry& subsample) {
-    subsamples_.push_back(subsample);
-  }
-
-  const std::vector<uint8_t>& iv() const { return iv_; }
-  const std::vector<SubsampleEntry>& subsamples() const { return subsamples_; }
-
- private:
-  std::vector<uint8_t> iv_;
-  std::vector<SubsampleEntry> subsamples_;
-
-  // Not using DISALLOW_COPY_AND_ASSIGN here intentionally to allow the compiler
-  // generated copy constructor and assignment operator.
-};
-
-}  // namespace mp4
-}  // namespace media
-}  // namespace edash_packager
-
-#endif  // MEDIA_FORMATS_MP4_CENC_H_

packager/media/formats/mp4/encrypting_fragmenter.cc

@@ -13,7 +13,6 @@
 #include "packager/media/filters/vp8_parser.h"
 #include "packager/media/filters/vp9_parser.h"
 #include "packager/media/formats/mp4/box_definitions.h"
-#include "packager/media/formats/mp4/cenc.h"
 
 namespace {
 // Generate 64bit IV by default.
@@ -67,6 +66,11 @@ Status EncryptingFragmenter::InitializeFragment(int64_t first_sample_dts) {
 
   traf()->auxiliary_size.sample_info_sizes.clear();
   traf()->auxiliary_offset.offsets.clear();
+  if (IsSubsampleEncryptionRequired()) {
+    traf()->sample_encryption.flags |=
+        SampleEncryption::kUseSubsampleEncryption;
+  }
+  traf()->sample_encryption.sample_encryption_entries.clear();
 
   const bool enable_encryption = clear_time_ <= 0;
   if (!enable_encryption) {
@@ -116,6 +120,7 @@ void EncryptingFragmenter::FinalizeFragmentForEncryption() {
     DCHECK(!IsSubsampleEncryptionRequired());
     saiz.default_sample_info_size = encryptor_->iv().size();
   }
+  traf()->sample_encryption.iv_size = encryptor_->iv().size();
 }
 
 Status EncryptingFragmenter::CreateEncryptor() {
@@ -141,7 +146,8 @@ void EncryptingFragmenter::EncryptBytes(uint8_t* data, uint32_t size) {
 Status EncryptingFragmenter::EncryptSample(scoped_refptr<MediaSample> sample) {
   DCHECK(encryptor_);
 
-  FrameCENCInfo cenc_info(encryptor_->iv());
+  SampleEncryptionEntry sample_encryption_entry;
+  sample_encryption_entry.initialization_vector = encryptor_->iv();
   uint8_t* data = sample->writable_data();
   if (IsSubsampleEncryptionRequired()) {
     if (vpx_parser_) {
@@ -155,7 +161,7 @@ Status EncryptingFragmenter::EncryptSample(scoped_refptr<MediaSample> sample) {
         subsample.clear_bytes = frame.uncompressed_header_size;
         subsample.cipher_bytes =
             frame.frame_size - frame.uncompressed_header_size;
-        cenc_info.AddSubsample(subsample);
+        sample_encryption_entry.subsamples.push_back(subsample);
         if (subsample.cipher_bytes > 0)
           EncryptBytes(data + subsample.clear_bytes, subsample.cipher_bytes);
         data += frame.frame_size;
@@ -167,27 +173,30 @@ Status EncryptingFragmenter::EncryptSample(scoped_refptr<MediaSample> sample) {
         if (!reader.ReadNBytesInto8(&nalu_length, nalu_length_size_))
           return Status(error::MUXER_FAILURE, "Fail to read nalu_length.");
 
-        SubsampleEntry subsample;
-        subsample.clear_bytes = nalu_length_size_ + 1;
-        subsample.cipher_bytes = nalu_length - 1;
         if (!reader.SkipBytes(nalu_length)) {
           return Status(error::MUXER_FAILURE,
                         "Sample size does not match nalu_length.");
         }
 
+        SubsampleEntry subsample;
+        subsample.clear_bytes = nalu_length_size_ + 1;
+        subsample.cipher_bytes = nalu_length - 1;
+        sample_encryption_entry.subsamples.push_back(subsample);
+
         EncryptBytes(data + subsample.clear_bytes, subsample.cipher_bytes);
-        cenc_info.AddSubsample(subsample);
         data += nalu_length_size_ + nalu_length;
       }
     }
 
     // The length of per-sample auxiliary datum, defined in CENC ch. 7.
-    traf()->auxiliary_size.sample_info_sizes.push_back(cenc_info.ComputeSize());
+    traf()->auxiliary_size.sample_info_sizes.push_back(
+        sample_encryption_entry.ComputeSize());
   } else {
     EncryptBytes(data, sample->data_size());
   }
 
-  cenc_info.Write(aux_data());
+  traf()->sample_encryption.sample_encryption_entries.push_back(
+      sample_encryption_entry);
   encryptor_->UpdateIv();
   return Status::OK;
 }

packager/media/formats/mp4/fragmenter.cc

@@ -84,7 +84,6 @@ Status Fragmenter::InitializeFragment(int64_t first_sample_dts) {
   earliest_presentation_time_ = kInvalidTime;
   first_sap_time_ = kInvalidTime;
   data_.reset(new BufferWriter());
-  aux_data_.reset(new BufferWriter());
   return Status::OK;
 }
 

packager/media/formats/mp4/fragmenter.h

@@ -59,7 +59,6 @@ class Fragmenter {
   bool fragment_initialized() const { return fragment_initialized_; }
   bool fragment_finalized() const { return fragment_finalized_; }
   BufferWriter* data() { return data_.get(); }
-  BufferWriter* aux_data() { return aux_data_.get(); }
 
  protected:
   TrackFragment* traf() { return traf_; }
@@ -82,7 +81,6 @@ class Fragmenter {
   int64_t earliest_presentation_time_;
   int64_t first_sap_time_;
   scoped_ptr<BufferWriter> data_;
-  scoped_ptr<BufferWriter> aux_data_;
 
   DISALLOW_COPY_AND_ASSIGN(Fragmenter);
 };

packager/media/formats/mp4/mp4.gyp

@@ -22,8 +22,6 @@
         'box_definitions.h',
         'box_reader.cc',
         'box_reader.h',
-        'cenc.cc',
-        'cenc.h',
         'chunk_info_iterator.cc',
         'chunk_info_iterator.h',
         'composition_offset_iterator.cc',

packager/media/formats/mp4/mp4_media_parser_unittest.cc

@@ -25,6 +25,10 @@ namespace edash_packager {
 namespace media {
 
 namespace {
+const char kKey[] =
+    "\xeb\xdd\x62\xf1\x68\x14\xd2\x7b\x68\xef\x12\x2a\xfc\xe4\xae\x3c";
+const char kKeyId[] = "0123456789012345";
+
 class MockKeySource : public KeySource {
  public:
   MOCK_METHOD1(FetchKeys, Status(const std::vector<uint8_t>& pssh_data));
@@ -225,7 +229,7 @@ TEST_F(MP4MediaParserTest, NON_FRAGMENTED_MP4) {
 
 TEST_F(MP4MediaParserTest, CencWithoutDecryptionSource) {
   // Parsing should fail but it will get the streams successfully.
-  EXPECT_FALSE(ParseMP4File("bear-640x360-v_frag-cenc.mp4", 512));
+  EXPECT_FALSE(ParseMP4File("bear-640x360-v_frag-cenc-aux.mp4", 512));
   EXPECT_EQ(1u, num_streams_);
 }
 
@@ -233,20 +237,16 @@ TEST_F(MP4MediaParserTest, CencInitWithoutDecryptionSource) {
   InitializeParser(NULL);
 
   std::vector<uint8_t> buffer =
-      ReadTestDataFile("bear-640x360-v_frag-cenc.mp4");
+      ReadTestDataFile("bear-640x360-v_frag-cenc-aux.mp4");
   const int kFirstMoofOffset = 1646;
   EXPECT_TRUE(AppendDataInPieces(buffer.data(), kFirstMoofOffset, 512));
   EXPECT_EQ(1u, num_streams_);
 }
 
-TEST_F(MP4MediaParserTest, CencWithDecryptionSource) {
+TEST_F(MP4MediaParserTest, CencWithDecryptionSourceAndAuxInMdat) {
   MockKeySource mock_key_source;
   EXPECT_CALL(mock_key_source, FetchKeys(_)).WillOnce(Return(Status::OK));
 
-  const char kKey[] =
-      "\xeb\xdd\x62\xf1\x68\x14\xd2\x7b\x68\xef\x12\x2a\xfc\xe4\xae\x3c";
-  const char kKeyId[] = "0123456789012345";
-
   EncryptionKey encryption_key;
   encryption_key.key.assign(kKey, kKey + strlen(kKey));
   EXPECT_CALL(mock_key_source,
@@ -256,7 +256,26 @@ TEST_F(MP4MediaParserTest, CencWithDecryptionSource) {
   InitializeParser(&mock_key_source);
 
   std::vector<uint8_t> buffer =
-      ReadTestDataFile("bear-640x360-v_frag-cenc.mp4");
+      ReadTestDataFile("bear-640x360-v_frag-cenc-aux.mp4");
+  EXPECT_TRUE(AppendDataInPieces(buffer.data(), buffer.size(), 512));
+  EXPECT_EQ(1u, num_streams_);
+  EXPECT_EQ(82u, num_samples_);
+}
+
+TEST_F(MP4MediaParserTest, CencWithDecryptionSourceAndSenc) {
+  MockKeySource mock_key_source;
+  EXPECT_CALL(mock_key_source, FetchKeys(_)).WillOnce(Return(Status::OK));
+
+  EncryptionKey encryption_key;
+  encryption_key.key.assign(kKey, kKey + strlen(kKey));
+  EXPECT_CALL(mock_key_source,
+              GetKey(std::vector<uint8_t>(kKeyId, kKeyId + strlen(kKeyId)), _))
+      .WillOnce(DoAll(SetArgPointee<1>(encryption_key), Return(Status::OK)));
+
+  InitializeParser(&mock_key_source);
+
+  std::vector<uint8_t> buffer =
+      ReadTestDataFile("bear-640x360-v_frag-cenc-senc.mp4");
   EXPECT_TRUE(AppendDataInPieces(buffer.data(), buffer.size(), 512));
   EXPECT_EQ(1u, num_streams_);
   EXPECT_EQ(82u, num_samples_);

packager/media/formats/mp4/segmenter.cc

@@ -385,39 +385,41 @@ Status Segmenter::FinalizeFragment(bool finalize_segment,
   }
 
   MediaData mdat;
-  // Fill in data offsets. Data offset base is moof size + mdat box size.
+  // Data offset relative to 'moof': moof size + mdat header size.
-  // (mdat is still empty, mdat size is the same as mdat box size).
+  // The code will also update box sizes for moof_ and its child boxes.
-  uint64_t base = moof_->ComputeSize() + mdat.ComputeSize();
+  uint64_t data_offset = moof_->ComputeSize() + mdat.HeaderSize();
+  // 'traf' should follow 'mfhd' moof header box.
+  uint64_t next_traf_position = moof_->HeaderSize() + moof_->header.box_size();
   for (size_t i = 0; i < moof_->tracks.size(); ++i) {
     TrackFragment& traf = moof_->tracks[i];
-    Fragmenter* fragmenter = fragmenters_[i];
+    if (traf.auxiliary_offset.offsets.size() > 0) {
-    if (fragmenter->aux_data()->Size() > 0) {
+      DCHECK_EQ(traf.auxiliary_offset.offsets.size(), 1u);
-      traf.auxiliary_offset.offsets[0] += base;
+      DCHECK(!traf.sample_encryption.sample_encryption_entries.empty());
-      base += fragmenter->aux_data()->Size();
+
+      next_traf_position += traf.box_size();
+      // SampleEncryption 'senc' box should be the last box in 'traf'.
+      // |auxiliary_offset| should point to the data of SampleEncryption.
+      traf.auxiliary_offset.offsets[0] =
+          next_traf_position - traf.sample_encryption.box_size() +
+          traf.sample_encryption.HeaderSize() +
+          sizeof(uint32_t);  // for sample count field in 'senc'
     }
-    traf.runs[0].data_offset += base;
+    traf.runs[0].data_offset = data_offset + mdat.data_size;
-    base += fragmenter->data()->Size();
+    mdat.data_size += fragmenters_[i]->data()->Size();
   }
 
   // Generate segment reference.
   sidx_->references.resize(sidx_->references.size() + 1);
   fragmenters_[GetReferenceStreamId()]->GenerateSegmentReference(
       &sidx_->references[sidx_->references.size() - 1]);
-  sidx_->references[sidx_->references.size() - 1].referenced_size = base;
+  sidx_->references[sidx_->references.size() - 1].referenced_size =
+      data_offset + mdat.data_size;
 
   // Write the fragment to buffer.
   moof_->Write(fragment_buffer_.get());
-
+  mdat.WriteHeader(fragment_buffer_.get());
-  for (size_t i = 0; i < moof_->tracks.size(); ++i) {
+  for (Fragmenter* fragmenter : fragmenters_)
-    Fragmenter* fragmenter = fragmenters_[i];
-    mdat.data_size =
-        fragmenter->aux_data()->Size() + fragmenter->data()->Size();
-    mdat.WriteHeader(fragment_buffer_.get());
-    if (fragmenter->aux_data()->Size()) {
-      fragment_buffer_->AppendBuffer(*fragmenter->aux_data());
-    }
     fragment_buffer_->AppendBuffer(*fragmenter->data());
-  }
 
   // Increase sequence_number for next fragment.
   ++moof_->header.sequence_number;

packager/media/formats/mp4/track_run_iterator.cc

@@ -40,6 +40,12 @@ struct TrackRunInfo {
   const AudioSampleEntry* audio_description;
   const VideoSampleEntry* video_description;
 
+  // Stores sample encryption entries, which is populated from 'senc' box if it
+  // is available, otherwise will try to load from cenc auxiliary information.
+  std::vector<SampleEncryptionEntry> sample_encryption_entries;
+
+  // These variables are useful to load |sample_encryption_entries| from cenc
+  // auxiliary information when 'senc' box is not available.
   int64_t aux_info_start_offset;  // Only valid if aux_info_total_size > 0.
   int aux_info_default_size;
   std::vector<uint8_t> aux_info_sizes;  // Populated if default_size == 0.
@@ -300,6 +306,40 @@ bool TrackRunIterator::Init(const MovieFragment& moof) {
     RCHECK(desc_idx > 0);  // Descriptions are one-indexed in the file
     desc_idx -= 1;
 
+    const AudioSampleEntry* audio_sample_entry = NULL;
+    const VideoSampleEntry* video_sample_entry = NULL;
+    switch (stsd.type) {
+      case kAudio:
+        RCHECK(!stsd.audio_entries.empty());
+        if (desc_idx > stsd.audio_entries.size())
+          desc_idx = 0;
+        audio_sample_entry = &stsd.audio_entries[desc_idx];
+        break;
+      case kVideo:
+        RCHECK(!stsd.video_entries.empty());
+        if (desc_idx > stsd.video_entries.size())
+          desc_idx = 0;
+        video_sample_entry = &stsd.video_entries[desc_idx];
+        break;
+      default:
+        NOTREACHED();
+        break;
+    }
+
+    // SampleEncryptionEntries should not have been parsed, without having
+    // iv_size. Parse the box now.
+    DCHECK(traf.sample_encryption.sample_encryption_entries.empty());
+    std::vector<SampleEncryptionEntry> sample_encryption_entries;
+    if (!traf.sample_encryption.sample_encryption_data.empty()) {
+      RCHECK(audio_sample_entry || video_sample_entry);
+      const uint8_t default_iv_size =
+          audio_sample_entry
+              ? audio_sample_entry->sinf.info.track_encryption.default_iv_size
+              : video_sample_entry->sinf.info.track_encryption.default_iv_size;
+      RCHECK(traf.sample_encryption.ParseFromSampleEncryptionData(
+          default_iv_size, &sample_encryption_entries));
+    }
+
     int64_t run_start_dts = traf.decode_time_absent
                                 ? next_fragment_start_dts_[i]
                                 : traf.decode_time.decode_time;
@@ -314,27 +354,30 @@ bool TrackRunIterator::Init(const MovieFragment& moof) {
       tri.sample_start_offset = trun.data_offset;
 
       tri.track_type = stsd.type;
-      if (tri.track_type == kAudio) {
+      tri.audio_description = audio_sample_entry;
-        RCHECK(!stsd.audio_entries.empty());
+      tri.video_description = video_sample_entry;
-        if (desc_idx > stsd.audio_entries.size())
-          desc_idx = 0;
-        tri.audio_description = &stsd.audio_entries[desc_idx];
-      } else if (tri.track_type == kVideo) {
-        RCHECK(!stsd.video_entries.empty());
-        if (desc_idx > stsd.video_entries.size())
-          desc_idx = 0;
-        tri.video_description = &stsd.video_entries[desc_idx];
-      }
 
-      // Collect information from the auxiliary_offset entry with the same index
+      tri.aux_info_start_offset = -1;
-      // in the 'saiz' container as the current run's index in the 'trun'
+      tri.aux_info_total_size = 0;
-      // container, if it is present.
+      // Populate sample encryption entries from SampleEncryption 'senc' box if
-      if (traf.auxiliary_offset.offsets.size() > j) {
+      // it is available; otherwise initialize aux_info variables, which will
+      // be used to populate sample encryption entries later in CacheAuxInfo.
+      if (!sample_encryption_entries.empty()) {
+        RCHECK(sample_encryption_entries.size() >=
+               sample_count_sum + trun.sample_count);
+        for (size_t k = 0; k < trun.sample_count; ++k) {
+          tri.sample_encryption_entries.push_back(
+              sample_encryption_entries[sample_count_sum + k]);
+        }
+      } else if (traf.auxiliary_offset.offsets.size() > j) {
+        // Collect information from the auxiliary_offset entry with the same
+        // index in the 'saiz' container as the current run's index in the
+        // 'trun' container, if it is present.
+        tri.aux_info_start_offset = traf.auxiliary_offset.offsets[j];
         // There should be an auxiliary info entry corresponding to each sample
         // in the auxiliary offset entry's corresponding track run.
         RCHECK(traf.auxiliary_size.sample_count >=
                sample_count_sum + trun.sample_count);
-        tri.aux_info_start_offset = traf.auxiliary_offset.offsets[j];
         tri.aux_info_default_size =
             traf.auxiliary_size.default_sample_info_size;
         if (tri.aux_info_default_size == 0) {
@@ -358,9 +401,6 @@ bool TrackRunIterator::Init(const MovieFragment& moof) {
             tri.aux_info_total_size += tri.aux_info_sizes[k];
           }
         }
-      } else {
-        tri.aux_info_start_offset = -1;
-        tri.aux_info_total_size = 0;
       }
 
       tri.samples.resize(trun.sample_count);
@@ -391,7 +431,6 @@ void TrackRunIterator::ResetRun() {
   sample_dts_ = run_itr_->start_dts;
   sample_offset_ = run_itr_->sample_start_offset;
   sample_itr_ = run_itr_->samples.begin();
-  cenc_info_.clear();
 }
 
 void TrackRunIterator::AdvanceSample() {
@@ -405,14 +444,17 @@ void TrackRunIterator::AdvanceSample() {
 // info is available in the stream.
 bool TrackRunIterator::AuxInfoNeedsToBeCached() {
   DCHECK(IsRunValid());
-  return is_encrypted() && aux_info_size() > 0 && cenc_info_.size() == 0;
+  return is_encrypted() && aux_info_size() > 0 &&
+         run_itr_->sample_encryption_entries.size() == 0;
 }
 
 // This implementation currently only caches CENC auxiliary info.
 bool TrackRunIterator::CacheAuxInfo(const uint8_t* buf, int buf_size) {
   RCHECK(AuxInfoNeedsToBeCached() && buf_size >= aux_info_size());
 
-  cenc_info_.resize(run_itr_->samples.size());
+  std::vector<SampleEncryptionEntry>& sample_encryption_entries =
+      runs_[run_itr_ - runs_.begin()].sample_encryption_entries;
+  sample_encryption_entries.resize(run_itr_->samples.size());
   int64_t pos = 0;
   for (size_t i = 0; i < run_itr_->samples.size(); i++) {
     int info_size = run_itr_->aux_info_default_size;
@@ -420,7 +462,9 @@ bool TrackRunIterator::CacheAuxInfo(const uint8_t* buf, int buf_size) {
       info_size = run_itr_->aux_info_sizes[i];
 
     BufferReader reader(buf + pos, info_size);
-    RCHECK(cenc_info_[i].Parse(track_encryption().default_iv_size, &reader));
+    const bool has_subsamples = info_size > track_encryption().default_iv_size;
+    RCHECK(sample_encryption_entries[i].ParseFromBuffer(
+        track_encryption().default_iv_size, has_subsamples, &reader));
     pos += info_size;
   }
 
@@ -539,12 +583,14 @@ const TrackEncryption& TrackRunIterator::track_encryption() const {
 
 scoped_ptr<DecryptConfig> TrackRunIterator::GetDecryptConfig() {
   size_t sample_idx = sample_itr_ - run_itr_->samples.begin();
-  DCHECK_LT(sample_idx, cenc_info_.size());
+  DCHECK_LT(sample_idx, run_itr_->sample_encryption_entries.size());
-  const FrameCENCInfo& cenc_info = cenc_info_[sample_idx];
+  const SampleEncryptionEntry& sample_encryption_entry =
+      run_itr_->sample_encryption_entries[sample_idx];
   DCHECK(is_encrypted());
   DCHECK(!AuxInfoNeedsToBeCached());
 
-  const size_t total_size_of_subsamples = cenc_info.GetTotalSizeOfSubsamples();
+  const size_t total_size_of_subsamples =
+      sample_encryption_entry.GetTotalSizeOfSubsamples();
   if (total_size_of_subsamples != 0 &&
       total_size_of_subsamples != static_cast<size_t>(sample_size())) {
     LOG(ERROR) << "Incorrect CENC subsample size.";
@@ -553,9 +599,9 @@ scoped_ptr<DecryptConfig> TrackRunIterator::GetDecryptConfig() {
 
   return scoped_ptr<DecryptConfig>(new DecryptConfig(
       track_encryption().default_kid,
-      cenc_info.iv(),
+      sample_encryption_entry.initialization_vector,
       0,  // No offset to start of media data in MP4 using CENC.
-      cenc_info.subsamples()));
+      sample_encryption_entry.subsamples));
 }
 
 }  // namespace mp4

packager/media/formats/mp4/track_run_iterator.h

@@ -9,7 +9,6 @@
 
 #include "packager/base/memory/scoped_ptr.h"
 #include "packager/media/formats/mp4/box_definitions.h"
-#include "packager/media/formats/mp4/cenc.h"
 
 namespace edash_packager {
 namespace media {
@@ -111,7 +110,6 @@ class TrackRunIterator {
   std::vector<TrackRunInfo>::const_iterator run_itr_;
   std::vector<SampleInfo>::const_iterator sample_itr_;
 
-  std::vector<FrameCENCInfo> cenc_info_;
   // Track the start dts of the next segment, only useful if decode_time box is
   // absent.
   std::vector<int64_t> next_fragment_start_dts_;

packager/media/formats/mp4/track_run_iterator_unittest.cc

@@ -11,14 +11,16 @@
 #include "packager/media/formats/mp4/rcheck.h"
 #include "packager/media/formats/mp4/track_run_iterator.h"
 
+namespace {
+
 // The sum of the elements in a vector initialized with SumAscending,
 // less the value of the last element.
-static const int kSumAscending1 = 45;
+const int kSumAscending1 = 45;
 
-static const int kAudioScale = 48000;
+const int kAudioScale = 48000;
-static const int kVideoScale = 25;
+const int kVideoScale = 25;
 
-static const uint8_t kAuxInfo[] = {
+const uint8_t kAuxInfo[] = {
     // Sample 1: IV (no subsumples).
     0x41, 0x54, 0x65, 0x73, 0x74, 0x49, 0x76, 0x31,
     // Sample 2: IV.
@@ -30,11 +32,39 @@ static const uint8_t kAuxInfo[] = {
     // Sample 2: Subsample 2.
     0x00, 0x03, 0x00, 0x00, 0x00, 0x04};
 
-static const char kIv1[] = {0x41, 0x54, 0x65, 0x73, 0x74, 0x49, 0x76, 0x31, };
+const uint8_t kSampleEncryptionDataWithSubsamples[] = {
+    // Sample count.
+    0x00, 0x00, 0x00, 0x02,
+    // Sample 1: IV.
+    0x41, 0x54, 0x65, 0x73, 0x74, 0x49, 0x76, 0x31,
+    // Sample 1: Subsample count.
+    0x00, 0x01,
+    // Sample 1: Subsample 1.
+    0x00, 0x01, 0x00, 0x00, 0x00, 0x02,
+    // Sample 2: IV.
+    0x41, 0x54, 0x65, 0x73, 0x74, 0x49, 0x76, 0x32,
+    // Sample 2: Subsample count.
+    0x00, 0x02,
+    // Sample 2: Subsample 1.
+    0x00, 0x01, 0x00, 0x00, 0x00, 0x02,
+    // Sample 2: Subsample 2.
+    0x00, 0x03, 0x00, 0x00, 0x00, 0x04};
 
-static const uint8_t kKeyId[] = {0x41, 0x47, 0x6f, 0x6f, 0x67, 0x6c,
+const uint8_t kSampleEncryptionDataWithoutSubsamples[] = {
-                                 0x65, 0x54, 0x65, 0x73, 0x74, 0x4b,
+    // Sample count.
-                                 0x65, 0x79, 0x49, 0x44};
+    0x00, 0x00, 0x00, 0x02,
+    // Sample 1: IV.
+    0x41, 0x54, 0x65, 0x73, 0x74, 0x49, 0x76, 0x31,
+    // Sample 2: IV.
+    0x41, 0x54, 0x65, 0x73, 0x74, 0x49, 0x76, 0x32};
+
+const char kIv1[] = {0x41, 0x54, 0x65, 0x73, 0x74, 0x49, 0x76, 0x31};
+const char kIv2[] = {0x41, 0x54, 0x65, 0x73, 0x74, 0x49, 0x76, 0x32};
+
+const uint8_t kKeyId[] = {0x41, 0x47, 0x6f, 0x6f, 0x67, 0x6c, 0x65, 0x54,
+                          0x65, 0x73, 0x74, 0x4b, 0x65, 0x79, 0x49, 0x44};
+
+}  // namespace
 
 namespace edash_packager {
 namespace media {
@@ -143,6 +173,38 @@ class TrackRunIteratorTest : public testing::Test {
     frag->runs[0].sample_sizes[1] = 10;
   }
 
+  void AddSampleEncryption(uint8_t use_subsample_flag, TrackFragment* frag) {
+    frag->sample_encryption.iv_size = 8;
+    frag->sample_encryption.flags = use_subsample_flag;
+    if (use_subsample_flag) {
+      frag->sample_encryption.sample_encryption_data.assign(
+          kSampleEncryptionDataWithSubsamples,
+          kSampleEncryptionDataWithSubsamples +
+              arraysize(kSampleEncryptionDataWithSubsamples));
+    } else {
+      frag->sample_encryption.sample_encryption_data.assign(
+          kSampleEncryptionDataWithoutSubsamples,
+          kSampleEncryptionDataWithoutSubsamples +
+              arraysize(kSampleEncryptionDataWithoutSubsamples));
+    }
+
+    // Update sample sizes and aux info header.
+    frag->runs.resize(1);
+    frag->runs[0].sample_count = 2;
+    frag->auxiliary_offset.offsets.push_back(0);
+    frag->auxiliary_size.sample_count = 2;
+    if (use_subsample_flag) {
+      // Update sample sizes to match with subsample entries above.
+      frag->runs[0].sample_sizes[0] = 3;
+      frag->runs[0].sample_sizes[1] = 10;
+      // Set aux info header.
+      frag->auxiliary_size.sample_info_sizes.push_back(16);
+      frag->auxiliary_size.sample_info_sizes.push_back(22);
+    } else {
+      frag->auxiliary_size.default_sample_info_size = 8;
+    }
+  }
+
   void SetAscending(std::vector<uint32_t>* vec) {
     vec->resize(10);
     for (size_t i = 0; i < vec->size(); i++)
@@ -303,7 +365,77 @@ TEST_F(TrackRunIteratorTest, IgnoreUnknownAuxInfoTest) {
   EXPECT_FALSE(iter_->AuxInfoNeedsToBeCached());
 }
 
-TEST_F(TrackRunIteratorTest, DecryptConfigTest) {
+TEST_F(TrackRunIteratorTest,
+       DecryptConfigTestWithSampleEncryptionAndSubsample) {
+  AddEncryption(&moov_.tracks[1]);
+  iter_.reset(new TrackRunIterator(&moov_));
+
+  MovieFragment moof = CreateFragment();
+  AddSampleEncryption(SampleEncryption::kUseSubsampleEncryption,
+                      &moof.tracks[1]);
+
+  ASSERT_TRUE(iter_->Init(moof));
+  // The run for track 2 will be the second, which is parsed according to
+  // data_offset.
+  iter_->AdvanceRun();
+  EXPECT_EQ(iter_->track_id(), 2u);
+
+  EXPECT_TRUE(iter_->is_encrypted());
+  // No need to cache aux info as it is already available in SampleEncryption.
+  EXPECT_FALSE(iter_->AuxInfoNeedsToBeCached());
+  EXPECT_EQ(iter_->aux_info_size(), 0);
+  EXPECT_EQ(iter_->sample_offset(), 200);
+  EXPECT_EQ(iter_->GetMaxClearOffset(), moof.tracks[1].runs[0].data_offset);
+  scoped_ptr<DecryptConfig> config = iter_->GetDecryptConfig();
+  EXPECT_EQ(std::vector<uint8_t>(kKeyId, kKeyId + arraysize(kKeyId)),
+            config->key_id());
+  EXPECT_EQ(std::vector<uint8_t>(kIv1, kIv1 + arraysize(kIv1)), config->iv());
+  EXPECT_EQ(config->subsamples().size(), 1u);
+  EXPECT_EQ(config->subsamples()[0].clear_bytes, 1u);
+  EXPECT_EQ(config->subsamples()[0].cipher_bytes, 2u);
+  iter_->AdvanceSample();
+  config = iter_->GetDecryptConfig();
+  EXPECT_EQ(std::vector<uint8_t>(kIv2, kIv2 + arraysize(kIv2)), config->iv());
+  EXPECT_EQ(config->subsamples().size(), 2u);
+  EXPECT_EQ(config->subsamples()[0].clear_bytes, 1u);
+  EXPECT_EQ(config->subsamples()[0].cipher_bytes, 2u);
+  EXPECT_EQ(config->subsamples()[1].clear_bytes, 3u);
+  EXPECT_EQ(config->subsamples()[1].cipher_bytes, 4u);
+}
+
+TEST_F(TrackRunIteratorTest,
+       DecryptConfigTestWithSampleEncryptionAndNoSubsample) {
+  AddEncryption(&moov_.tracks[1]);
+  iter_.reset(new TrackRunIterator(&moov_));
+
+  MovieFragment moof = CreateFragment();
+  AddSampleEncryption(!SampleEncryption::kUseSubsampleEncryption,
+                      &moof.tracks[1]);
+
+  ASSERT_TRUE(iter_->Init(moof));
+  // The run for track 2 will be the second, which is parsed according to
+  // data_offset.
+  iter_->AdvanceRun();
+  EXPECT_EQ(iter_->track_id(), 2u);
+
+  EXPECT_TRUE(iter_->is_encrypted());
+  // No need to cache aux info as it is already available in SampleEncryption.
+  EXPECT_FALSE(iter_->AuxInfoNeedsToBeCached());
+  EXPECT_EQ(iter_->aux_info_size(), 0);
+  EXPECT_EQ(iter_->sample_offset(), 200);
+  EXPECT_EQ(iter_->GetMaxClearOffset(), moof.tracks[1].runs[0].data_offset);
+  scoped_ptr<DecryptConfig> config = iter_->GetDecryptConfig();
+  EXPECT_EQ(std::vector<uint8_t>(kKeyId, kKeyId + arraysize(kKeyId)),
+            config->key_id());
+  EXPECT_EQ(std::vector<uint8_t>(kIv1, kIv1 + arraysize(kIv1)), config->iv());
+  EXPECT_EQ(config->subsamples().size(), 0u);
+  iter_->AdvanceSample();
+  config = iter_->GetDecryptConfig();
+  EXPECT_EQ(std::vector<uint8_t>(kIv2, kIv2 + arraysize(kIv2)), config->iv());
+  EXPECT_EQ(config->subsamples().size(), 0u);
+}
+
+TEST_F(TrackRunIteratorTest, DecryptConfigTestWithAuxInfo) {
   AddEncryption(&moov_.tracks[1]);
   iter_.reset(new TrackRunIterator(&moov_));
 
@@ -316,7 +448,7 @@ TEST_F(TrackRunIteratorTest, DecryptConfigTest) {
   // element in the file.
   EXPECT_EQ(iter_->track_id(), 2u);
   EXPECT_TRUE(iter_->is_encrypted());
-  EXPECT_TRUE(iter_->AuxInfoNeedsToBeCached());
+  ASSERT_TRUE(iter_->AuxInfoNeedsToBeCached());
   EXPECT_EQ(static_cast<uint32_t>(iter_->aux_info_size()), arraysize(kAuxInfo));
   EXPECT_EQ(iter_->aux_info_offset(), 50);
   EXPECT_EQ(iter_->GetMaxClearOffset(), 50);
@@ -328,11 +460,9 @@ TEST_F(TrackRunIteratorTest, DecryptConfigTest) {
   EXPECT_EQ(iter_->sample_offset(), 200);
   EXPECT_EQ(iter_->GetMaxClearOffset(), moof.tracks[0].runs[0].data_offset);
   scoped_ptr<DecryptConfig> config = iter_->GetDecryptConfig();
-  ASSERT_EQ(arraysize(kKeyId), config->key_id().size());
+  EXPECT_EQ(std::vector<uint8_t>(kKeyId, kKeyId + arraysize(kKeyId)),
-  EXPECT_TRUE(
+            config->key_id());
-      !memcmp(kKeyId, config->key_id().data(), config->key_id().size()));
+  EXPECT_EQ(std::vector<uint8_t>(kIv1, kIv1 + arraysize(kIv1)), config->iv());
-  ASSERT_EQ(arraysize(kIv1), config->iv().size());
-  EXPECT_TRUE(!memcmp(kIv1, config->iv().data(), config->iv().size()));
   EXPECT_TRUE(config->subsamples().empty());
   iter_->AdvanceSample();
   config = iter_->GetDecryptConfig();

packager/media/test/data/README

@@ -3,9 +3,9 @@
 // found in the LICENSE file.
 
 bear-320x240.webm - WebM encode of bear.1280x720.mp4 resized to 320x240.
-no_streams.webm - Header, Info, & Tracks element from bear-320x240.webm slightly corrupted so it looks 
+no_streams.webm - Header, Info, & Tracks element from bear-320x240.webm slightly corrupted so it looks
                   like there are no tracks.
-nonzero-start-time.webm - Has the same headers as bear-320x240.webm but the first cluster of this file 
+nonzero-start-time.webm - Has the same headers as bear-320x240.webm but the first cluster of this file
                           is the second cluster of bear-320x240.webm. This creates the situation where
                           the media data doesn't start at time 0.
 bear-320x240-live.webm - bear-320x240.webm remuxed w/o a duration and using clusters with unknown sizes.
@@ -41,7 +41,11 @@ bear-640x360-non_square_pixel-with_pasp.mp4 - A non-square pixel version of the
 bear-640x360-non_square_pixel-without_pasp.mp4 - A non-square pixel version of the video track of bear-640x360.mp4 without PixelAspectRatio box.
 
 // Encrypted Files.
-bear-640x360-v_frag-cenc.mp4 - A fragmented MP4 version of the video track of bear-640x360.mp4 encrypted (ISO CENC) using key ID [1] and key [2].
+bear-640x360-v_frag-cenc-aux.mp4 - A fragmented MP4 version of the video track of bear-640x360.mp4
+                                   encrypted (ISO CENC) using key ID [1] and key [2] and with sample
+                                   encryption auxiliary information in the beginning of mdat box.
+bear-640x360-v_frag-cenc-senc.mp4 - Same as above, but with sample encryption information stored in
+                                    senc box.
 
 [1] 30313233343536373839303132333435
 [2] ebdd62f16814d27b68ef122afce4ae3c