Bug 1022434 - Extract crypto information from MP4 demuxer; r=cpearce

* * *
Bug 1022434 - "[EME] Expose decryption data in MP4Samples" []

media/libstagefright/binding/Adts.cpp

@@ -56,6 +56,14 @@ Adts::ConvertEsdsToAdts(uint16_t aChannelCount, int8_t aFrequencyIndex,
   header[6] = 0xfc;
 
   aSample->Prepend(&header[0], ArrayLength(header));
+  if (aSample->crypto.valid) {
+    if (aSample->crypto.plain_sizes.Length() == 0) {
+      aSample->crypto.plain_sizes.AppendElement(kADTSHeaderSize);
+      aSample->crypto.encrypted_sizes.AppendElement(aSample->size - kADTSHeaderSize);
+    } else {
+      aSample->crypto.plain_sizes[0] += kADTSHeaderSize;
+    }
+  }
 
   return true;
 }

media/libstagefright/binding/DecoderData.cpp

@@ -34,6 +34,94 @@ FindInt64(sp<MetaData>& mMetaData, uint32_t mKey)
   return value;
 }
 
+template <typename T, size_t N>
+static bool
+FindData(sp<MetaData>& aMetaData, uint32_t aKey, mozilla::Vector<T, N>* aDest)
+{
+  const void* data;
+  size_t size;
+  uint32_t type;
+
+  aDest->clear();
+  // There's no point in checking that the type matches anything because it
+  // isn't set consistently in the MPEG4Extractor.
+  if (!aMetaData->findData(aKey, &type, &data, &size) || size % sizeof(T)) {
+    return false;
+  }
+
+  aDest->append(reinterpret_cast<const T*>(data), size / sizeof(T));
+  return true;
+}
+
+template <typename T>
+static bool
+FindData(sp<MetaData>& aMetaData, uint32_t aKey, nsTArray<T>* aDest)
+{
+  const void* data;
+  size_t size;
+  uint32_t type;
+
+  aDest->Clear();
+  // There's no point in checking that the type matches anything because it
+  // isn't set consistently in the MPEG4Extractor.
+  if (!aMetaData->findData(aKey, &type, &data, &size) || size % sizeof(T)) {
+    return false;
+  }
+
+  aDest->AppendElements(reinterpret_cast<const T*>(data), size / sizeof(T));
+  return true;
+}
+
+bool
+CryptoFile::DoUpdate(sp<MetaData>& aMetaData)
+{
+  const void* data;
+  size_t size;
+  uint32_t type;
+
+  // There's no point in checking that the type matches anything because it
+  // isn't set consistently in the MPEG4Extractor.
+  if (!aMetaData->findData(kKeyPssh, &type, &data, &size)) {
+    return false;
+  }
+
+  ByteReader reader(reinterpret_cast<const uint8_t*>(data), size);
+  while (reader.Remaining()) {
+    PsshInfo psshInfo;
+    if (!reader.ReadArray(psshInfo.uuid, 16)) {
+      return false;
+    }
+
+    if (!reader.CanReadType<uint32_t>()) {
+      return false;
+    }
+    auto length = reader.ReadType<uint32_t>();
+
+    if (!reader.ReadArray(psshInfo.data, length)) {
+      return false;
+    }
+    pssh.append(Move(psshInfo));
+  }
+  return true;
+}
+
+void
+CryptoTrack::Update(sp<MetaData>& aMetaData)
+{
+  valid = aMetaData->findInt32(kKeyCryptoMode, &mode) &&
+          aMetaData->findInt32(kKeyCryptoDefaultIVSize, &iv_size) &&
+          FindData(aMetaData, kKeyCryptoKey, &key);
+}
+
+void
+CryptoSample::Update(sp<MetaData>& aMetaData)
+{
+  CryptoTrack::Update(aMetaData);
+  valid = valid && FindData(aMetaData, kKeyPlainSizes, &plain_sizes) &&
+          FindData(aMetaData, kKeyEncryptedSizes, &encrypted_sizes) &&
+          FindData(aMetaData, kKeyCryptoIV, &iv);
+}
+
 void
 AudioDecoderConfig::Update(sp<MetaData>& aMetaData, const char* aMimeType)
 {
@@ -46,20 +134,18 @@ AudioDecoderConfig::Update(sp<MetaData>& aMetaData, const char* aMimeType)
   frequency_index = Adts::GetFrequencyIndex(samples_per_second);
   aac_profile = FindInt32(aMetaData, kKeyAACProfile);
 
-  const void* data;
+  if (FindData(aMetaData, kKeyESDS, &extra_data)) {
-  size_t size;
-  uint32_t type;
-
-  if (aMetaData->findData(kKeyESDS, &type, &data, &size)) {
-    extra_data.clear();
-    extra_data.append(reinterpret_cast<const uint8_t*>(data), size);
-
     ESDS esds(&extra_data[0], extra_data.length());
+
+    const void* data;
+    size_t size;
     if (esds.getCodecSpecificInfo(&data, &size) == OK) {
       audio_specific_config.append(reinterpret_cast<const uint8_t*>(data),
                                    size);
     }
   }
+
+  crypto.Update(aMetaData);
 }
 
 bool
@@ -78,18 +164,14 @@ VideoDecoderConfig::Update(sp<MetaData>& aMetaData, const char* aMimeType)
   display_width = FindInt32(aMetaData, kKeyDisplayWidth);
   display_height = FindInt32(aMetaData, kKeyDisplayHeight);
 
-  const void* data;
+  if (FindData(aMetaData, kKeyAVCC, &extra_data) && extra_data.length() >= 7) {
-  size_t size;
-  uint32_t type;
-
-  if (aMetaData->findData(kKeyAVCC, &type, &data, &size) && size >= 7) {
-    extra_data.clear();
-    extra_data.append(reinterpret_cast<const uint8_t*>(data), size);
     // Set size of the NAL length to 4. The demuxer formats its output with
     // this NAL length size.
     extra_data[4] |= 3;
     annex_b = AnnexB::ConvertExtraDataToAnnexB(extra_data);
   }
+
+  crypto.Update(aMetaData);
 }
 
 bool
@@ -126,6 +208,8 @@ MP4Sample::Update()
   is_sync_point = FindInt32(m, kKeyIsSyncFrame);
   data = reinterpret_cast<uint8_t*>(mMediaBuffer->data());
   size = mMediaBuffer->range_length();
+
+  crypto.Update(m);
 }
 
 void

media/libstagefright/binding/include/mp4_demuxer/ByteReader.h

@@ -6,6 +6,7 @@
 #define BYTE_READER_H_
 
 #include "mozilla/Vector.h"
+#include "nsTArray.h"
 
 namespace mp4_demuxer
 {
@@ -13,7 +14,7 @@ namespace mp4_demuxer
 class ByteReader
 {
 public:
-  ByteReader(mozilla::Vector<uint8_t>& aData)
+  ByteReader(const mozilla::Vector<uint8_t>& aData)
     : mPtr(&aData[0]), mRemaining(aData.length())
   {
   }
@@ -72,6 +73,31 @@ public:
     return result;
   }
 
+  template <typename T> bool CanReadType() { return mRemaining >= sizeof(T); }
+
+  template <typename T> T ReadType()
+  {
+    auto ptr = Read(sizeof(T));
+    if (!ptr) {
+      MOZ_ASSERT(false);
+      return 0;
+    }
+    return *reinterpret_cast<const T*>(ptr);
+  }
+
+  template <typename T>
+  bool ReadArray(nsTArray<T>& aDest, size_t aLength)
+  {
+    auto ptr = Read(aLength * sizeof(T));
+    if (!ptr) {
+      return false;
+    }
+
+    aDest.Clear();
+    aDest.AppendElements(reinterpret_cast<const T*>(ptr), aLength);
+    return true;
+  }
+
 private:
   const uint8_t* mPtr;
   size_t mRemaining;

media/libstagefright/binding/include/mp4_demuxer/DecoderData.h

@@ -7,6 +7,7 @@
 
 #include "mozilla/Types.h"
 #include "mozilla/Vector.h"
+#include "nsTArray.h"
 #include "nsAutoPtr.h"
 
 namespace stagefright
@@ -21,6 +22,51 @@ namespace mp4_demuxer
 
 class MP4Demuxer;
 
+struct PsshInfo
+{
+  PsshInfo() {}
+  PsshInfo(PsshInfo&& aOther) : uuid(aOther.uuid), data(aOther.data) {}
+  nsTArray<uint8_t> uuid;
+  nsTArray<uint8_t> data;
+};
+
+class CryptoFile
+{
+public:
+  void Update(stagefright::sp<stagefright::MetaData>& aMetaData)
+  {
+    valid = DoUpdate(aMetaData);
+  }
+
+  bool valid;
+  mozilla::Vector<PsshInfo> pssh;
+
+private:
+  bool DoUpdate(stagefright::sp<stagefright::MetaData>& aMetaData);
+};
+
+class CryptoTrack
+{
+public:
+  CryptoTrack() : valid(false) {}
+  void Update(stagefright::sp<stagefright::MetaData>& aMetaData);
+
+  bool valid;
+  int32_t mode;
+  int32_t iv_size;
+  nsTArray<uint8_t> key;
+};
+
+class CryptoSample : public CryptoTrack
+{
+public:
+  void Update(stagefright::sp<stagefright::MetaData>& aMetaData);
+
+  nsTArray<uint16_t> plain_sizes;
+  nsTArray<uint32_t> encrypted_sizes;
+  nsTArray<uint8_t> iv;
+};
+
 class AudioDecoderConfig
 {
 public:
@@ -42,6 +88,7 @@ public:
   int8_t frequency_index;
   mozilla::Vector<uint8_t> extra_data;
   mozilla::Vector<uint8_t> audio_specific_config;
+  CryptoTrack crypto;
 
   void Update(stagefright::sp<stagefright::MetaData>& aMetaData, const char* aMimeType);
   bool IsValid();
@@ -69,6 +116,7 @@ public:
 
   mozilla::Vector<uint8_t> extra_data; // Unparsed AVCDecoderConfig payload.
   mozilla::Vector<uint8_t> annex_b;    // Parsed version for sample prepend.
+  CryptoTrack crypto;
 
   void Update(stagefright::sp<stagefright::MetaData>& aMetaData, const char* aMimeType);
   bool IsValid();
@@ -94,6 +142,8 @@ public:
   uint8_t* data;
   size_t size;
 
+  CryptoSample crypto;
+
   void Prepend(const uint8_t* aData, size_t aSize);
 
 private:

media/libstagefright/binding/include/mp4_demuxer/mp4_demuxer.h

@@ -52,6 +52,7 @@ public:
   MP4Sample* DemuxAudioSample();
   MP4Sample* DemuxVideoSample();
 
+  const CryptoFile& Crypto() { return mCrypto; }
   const AudioDecoderConfig& AudioConfig() { return mAudioConfig; }
   const VideoDecoderConfig& VideoConfig() { return mVideoConfig; }
 
@@ -62,6 +63,7 @@ public:
 private:
   AudioDecoderConfig mAudioConfig;
   VideoDecoderConfig mVideoConfig;
+  CryptoFile mCrypto;
 
   nsAutoPtr<StageFrightPrivate> mPrivate;
 };

media/libstagefright/binding/mp4_demuxer.cpp

@@ -111,6 +111,8 @@ MP4Demuxer::Init()
       mPrivate->mVideoIndex.Init(index);
     }
   }
+  sp<MetaData> metaData = e->getMetaData();
+  mCrypto.Update(metaData);
 
   return mPrivate->mAudio.get() || mPrivate->mVideo.get();
 }

media/libstagefright/frameworks/av/media/libstagefright/MPEG4Extractor.cpp

@@ -137,12 +137,12 @@ private:
 
     struct Sample {
         off64_t offset;
-        size_t size;
+        uint32_t size;
         uint32_t duration;
         uint32_t ctsOffset;
         uint8_t iv[16];
-        Vector<size_t> clearsizes;
+        Vector<uint16_t> clearsizes;
-        Vector<size_t> encryptedsizes;
+        Vector<uint32_t> encryptedsizes;
     };
     Vector<Sample> mCurrentSamples;
 
@@ -1522,7 +1522,8 @@ status_t MPEG4Extractor::parseChunk(off64_t *offset, int depth) {
                     kKeyESDS, kTypeESDS, &buffer[4], chunk_data_size - 4);
 
             if (mPath.size() >= 2
-                    && mPath[mPath.size() - 2] == FOURCC('m', 'p', '4', 'a')) {
+                    && (mPath[mPath.size() - 2] == FOURCC('m', 'p', '4', 'a') ||
+                       (mPath[mPath.size() - 2] == FOURCC('e', 'n', 'c', 'a')))) {
                 // Information from the ESDS must be relied on for proper
                 // setup of sample rate and channel count for MPEG4 Audio.
                 // The generic header appears to only contain generic