Add support for XING header

1 files changed, 182 insertions, 11 deletions

diff --git a/media/libstagefright/MP3Extractor.cpp b/media/libstagefright/MP3Extractor.cpp
index 39d4c0b..14e6177 100644
--- a/media/libstagefright/MP3Extractor.cpp
+++ b/media/libstagefright/MP3Extractor.cpp
@@ -173,6 +173,133 @@ static bool get_mp3_frame_size(
     return true;
 }
 
+static bool parse_xing_header(
+        const sp<DataSource> &source, off_t first_frame_pos,
+        int32_t *frame_number = NULL, int32_t *byte_number = NULL,
+        char *table_of_contents = NULL, int32_t *quality_indicator = NULL,
+        int64_t *duration = NULL) {
+
+    if (frame_number) {
+        *frame_number = 0;
+    }
+    if (byte_number) {
+        *byte_number = 0;
+    }
+    if (table_of_contents) {
+        table_of_contents[0] = 0;
+    }
+    if (quality_indicator) {
+        *quality_indicator = 0;
+    }
+    if (duration) {
+        *duration = 0;
+    }
+
+    uint8_t buffer[4];
+    int offset = first_frame_pos;
+    if (source->readAt(offset, &buffer, 4) < 4) { // get header
+        return false;
+    }
+    offset += 4;
+
+    uint8_t id, layer, sr_index, mode;
+    layer = (buffer[1] >> 1) & 3;
+    id = (buffer[1] >> 3) & 3;
+    sr_index = (buffer[2] >> 2) & 3;
+    mode = (buffer[3] >> 6) & 3;
+    if (layer == 0) {
+        return false;
+    }
+    if (id == 1) {
+        return false;
+    }
+    if (sr_index == 3) {
+        return false;
+    }
+    // determine offset of XING header
+    if(id&1) { // mpeg1
+        if (mode != 3) offset += 32;
+        else offset += 17;
+    } else { // mpeg2
+        if (mode != 3) offset += 17;
+        else offset += 9;
+    }
+
+    if (source->readAt(offset, &buffer, 4) < 4) { // XING header ID
+        return false;
+    }
+    offset += 4;
+    // Check XING ID
+    if ((buffer[0] != 'X') || (buffer[1] != 'i')
+                || (buffer[2] != 'n') || (buffer[3] != 'g')) {
+        if ((buffer[0] != 'I') || (buffer[1] != 'n')
+                    || (buffer[2] != 'f') || (buffer[3] != 'o')) {
+            return false;
+        }
+    }
+
+    if (source->readAt(offset, &buffer, 4) < 4) { // flags
+        return false;
+    }
+    offset += 4;
+    uint32_t flags = U32_AT(buffer);
+
+    if (flags & 0x0001) {  // Frames field is present
+        if (source->readAt(offset, buffer, 4) < 4) {
+             return false;
+        }
+        if (frame_number) {
+           *frame_number = U32_AT(buffer);
+        }
+        int32_t frame = U32_AT(buffer);
+        // Samples per Frame: 1. index = MPEG Version ID, 2. index = Layer
+        const int samplesPerFrames[2][3] =
+        {
+            { 384, 1152, 576  }, // MPEG 2, 2.5: layer1, layer2, layer3
+            { 384, 1152, 1152 }, // MPEG 1: layer1, layer2, layer3
+        };
+        // sampling rates in hertz: 1. index = MPEG Version ID, 2. index = sampling rate index
+        const int samplingRates[4][3] =
+        {
+            { 11025, 12000, 8000,  },    // MPEG 2.5
+            { 0,     0,     0,     },    // reserved
+            { 22050, 24000, 16000, },    // MPEG 2
+            { 44100, 48000, 32000, }     // MPEG 1
+        };
+        if (duration) {
+            *duration = (int64_t)frame * samplesPerFrames[id&1][3-layer] * 1000000LL
+                / samplingRates[id][sr_index];
+        }
+        offset += 4;
+    }
+    if (flags & 0x0002) {  // Bytes field is present
+        if (byte_number) {
+            if (source->readAt(offset, buffer, 4) < 4) {
+                return false;
+            }
+            *byte_number = U32_AT(buffer);
+        }
+        offset += 4;
+    }
+    if (flags & 0x0004) {  // TOC field is present
+       if (table_of_contents) {
+            if (source->readAt(offset + 1, table_of_contents, 99) < 99) {
+                return false;
+            }
+        }
+        offset += 100;
+    }
+    if (flags & 0x0008) {  // Quality indicator field is present
+        if (quality_indicator) {
+            if (source->readAt(offset, buffer, 4) < 4) {
+                return false;
+            }
+            *quality_indicator = U32_AT(buffer);
+        }
+    }
+    return true;
+}
+
 static bool Resync(
         const sp<DataSource> &source, uint32_t match_header,
         off_t *inout_pos, uint32_t *out_header) {
@@ -314,7 +441,8 @@ class MP3Source : public MediaSource {
 public:
     MP3Source(
             const sp<MetaData> &meta, const sp<DataSource> &source,
-            off_t first_frame_pos, uint32_t fixed_header);
+            off_t first_frame_pos, uint32_t fixed_header,
+            int32_t byte_number, const char *table_of_contents);
 
     virtual status_t start(MetaData *params = NULL);
     virtual status_t stop();
@@ -335,7 +463,9 @@ private:
     off_t mCurrentPos;
     int64_t mCurrentTimeUs;
     bool mStarted;
-
+    int32_t mByteNumber; // total number of bytes in this MP3
+    // TOC entries in XING header. Skip the first one since it's always 0.
+    char mTableOfContents[99];
     MediaBufferGroup *mGroup;
 
     MP3Source(const MP3Source &);
@@ -345,7 +475,8 @@ private:
 MP3Extractor::MP3Extractor(const sp<DataSource> &source)
     : mDataSource(source),
       mFirstFramePos(-1),
-      mFixedHeader(0) {
+      mFixedHeader(0),
+      mByteNumber(0) {
     off_t pos = 0;
     uint32_t header;
     bool success = Resync(mDataSource, 0, &pos, &header);
@@ -369,11 +500,19 @@ MP3Extractor::MP3Extractor(const sp<DataSource> &source)
         mMeta->setInt32(kKeyBitRate, bitrate * 1000);
         mMeta->setInt32(kKeyChannelCount, num_channels);
 
-        off_t fileSize;
-        if (mDataSource->getSize(&fileSize) == OK) {
-            mMeta->setInt64(
-                    kKeyDuration,
-                    8000LL * (fileSize - mFirstFramePos) / bitrate);
+        int64_t duration;
+        parse_xing_header(
+                mDataSource, mFirstFramePos, NULL, &mByteNumber,
+                mTableOfContents, NULL, &duration);
+        if (duration > 0) {
+            mMeta->setInt64(kKeyDuration, duration);
+        } else {
+            off_t fileSize;
+            if (mDataSource->getSize(&fileSize) == OK) {
+                mMeta->setInt64(
+                        kKeyDuration,
+                        8000LL * (fileSize - mFirstFramePos) / bitrate);
+            }
         }
     }
 }
@@ -391,7 +530,8 @@ sp<MediaSource> MP3Extractor::getTrack(size_t index) {
     }
 
     return new MP3Source(
-            mMeta, mDataSource, mFirstFramePos, mFixedHeader);
+            mMeta, mDataSource, mFirstFramePos, mFixedHeader,
+            mByteNumber, mTableOfContents);
 }
 
 sp<MetaData> MP3Extractor::getTrackMetaData(size_t index, uint32_t flags) {
@@ -406,7 +546,8 @@ sp<MetaData> MP3Extractor::getTrackMetaData(size_t index, uint32_t flags) {
 
 MP3Source::MP3Source(
         const sp<MetaData> &meta, const sp<DataSource> &source,
-        off_t first_frame_pos, uint32_t fixed_header)
+        off_t first_frame_pos, uint32_t fixed_header,
+        int32_t byte_number, const char *table_of_contents)
     : mMeta(meta),
       mDataSource(source),
       mFirstFramePos(first_frame_pos),
@@ -414,7 +555,9 @@ MP3Source::MP3Source(
       mCurrentPos(0),
       mCurrentTimeUs(0),
       mStarted(false),
+      mByteNumber(byte_number),
       mGroup(NULL) {
+    memcpy (mTableOfContents, table_of_contents, strlen(table_of_contents));
 }
 
 MP3Source::~MP3Source() {
@@ -469,7 +612,35 @@ status_t MP3Source::read(
         }
 
         mCurrentTimeUs = seekTimeUs;
-        mCurrentPos = mFirstFramePos + seekTimeUs * bitrate / 8000000;
+        // interpolate in TOC to get file seek point in bytes
+        int64_t duration;
+        if ((mByteNumber > 0) && (mTableOfContents[0] > 0)
+            && mMeta->findInt64(kKeyDuration, &duration)) {
+            float percent = (float)seekTimeUs * 100 / duration;
+            float fx;
+            if( percent <= 0.0f ) {
+                fx = 0.0f;
+            } else if( percent >= 100.0f ) {
+                fx = 256.0f;
+            } else {
+                int a = (int)percent;
+                float fa, fb;
+                if ( a == 0 ) {
+                    fa = 0.0f;
+                } else {
+                    fa = (float)mTableOfContents[a-1];
+                }
+                if ( a < 99 ) {
+                    fb = (float)mTableOfContents[a];
+                } else {
+                    fb = 256.0f;
+                }
+                fx = fa + (fb-fa)*(percent-a);
+            }
+            mCurrentPos = mFirstFramePos + (int)((1.0f/256.0f)*fx*mByteNumber);
+        } else {
+            mCurrentPos = mFirstFramePos + seekTimeUs * bitrate / 8000000;
+        }
     }
 
     MediaBuffer *buffer;