Reject streaming .mp4 files over http that are not muxed reasonably.

Audio and video samples must be reasonably close by at any given timestamp,
for now we define "reasonably" as "their offsets must not be more than 1MB apart".

related-to-bug: 3509977
Change-Id: I4c3a2a239bf911306833f3b17b73c9e2f76ab901

1 files changed, 105 insertions, 7 deletions

diff --git a/media/libstagefright/MPEG4Extractor.cpp b/media/libstagefright/MPEG4Extractor.cpp
index 7b96d01..1ca2d6d 100644
--- a/media/libstagefright/MPEG4Extractor.cpp
+++ b/media/libstagefright/MPEG4Extractor.cpp
@@ -262,7 +262,7 @@ static const char *FourCC2MIME(uint32_t fourcc) {
 
 MPEG4Extractor::MPEG4Extractor(const sp<DataSource> &source)
     : mDataSource(source),
-      mHaveMetadata(false),
+      mInitCheck(NO_INIT),
       mHasVideo(false),
       mFirstTrack(NULL),
       mLastTrack(NULL),
@@ -361,8 +361,8 @@ sp<MetaData> MPEG4Extractor::getTrackMetaData(
 }
 
 status_t MPEG4Extractor::readMetaData() {
-    if (mHaveMetadata) {
-        return OK;
+    if (mInitCheck != NO_INIT) {
+        return mInitCheck;
     }
 
     off64_t offset = 0;
@@ -370,17 +370,20 @@ status_t MPEG4Extractor::readMetaData() {
     while ((err = parseChunk(&offset, 0)) == OK) {
     }
 
-    if (mHaveMetadata) {
+    if (mInitCheck == OK) {
         if (mHasVideo) {
             mFileMetaData->setCString(kKeyMIMEType, "video/mp4");
         } else {
             mFileMetaData->setCString(kKeyMIMEType, "audio/mp4");
         }
 
-        return OK;
+        mInitCheck = verifyIfStreamable();
+    } else {
+        mInitCheck = err;
     }
 
-    return err;
+    CHECK_NE(err, (status_t)NO_INIT);
+    return mInitCheck;
 }
 
 void MPEG4Extractor::setDrmFlag(bool flag) {
@@ -755,7 +758,7 @@ status_t MPEG4Extractor::parseChunk(off64_t *offset, int depth) {
                     return err;
                 }
             } else if (chunk_type == FOURCC('m', 'o', 'o', 'v')) {
-                mHaveMetadata = true;
+                mInitCheck = OK;
 
                 if (!mIsDrm) {
                     return UNKNOWN_ERROR;  // Return a dummy error.
@@ -2077,6 +2080,101 @@ status_t MPEG4Source::read(
     }
 }
 
+MPEG4Extractor::Track *MPEG4Extractor::findTrackByMimePrefix(
+        const char *mimePrefix) {
+    for (Track *track = mFirstTrack; track != NULL; track = track->next) {
+        const char *mime;
+        if (track->meta != NULL
+                && track->meta->findCString(kKeyMIMEType, &mime)
+                && !strncasecmp(mime, mimePrefix, strlen(mimePrefix))) {
+            return track;
+        }
+    }
+
+    return NULL;
+}
+
+status_t MPEG4Extractor::verifyIfStreamable() {
+    if (!(mDataSource->flags() & DataSource::kIsCachingDataSource)) {
+        return OK;
+    }
+
+    Track *audio = findTrackByMimePrefix("audio/");
+    Track *video = findTrackByMimePrefix("video/");
+
+    if (audio == NULL || video == NULL) {
+        return OK;
+    }
+
+    sp<SampleTable> audioSamples = audio->sampleTable;
+    sp<SampleTable> videoSamples = video->sampleTable;
+
+    off64_t maxOffsetDiff = 0;
+    int64_t maxOffsetTimeUs = -1;
+
+    for (uint32_t i = 0; i < videoSamples->countSamples(); ++i) {
+        off64_t videoOffset;
+        uint32_t videoTime;
+        bool isSync;
+        CHECK_EQ((status_t)OK, videoSamples->getMetaDataForSample(
+                    i, &videoOffset, NULL, &videoTime, &isSync));
+
+        int64_t videoTimeUs = (int64_t)(videoTime * 1E6 / video->timescale);
+
+        uint32_t reqAudioTime = (videoTimeUs * audio->timescale) / 1000000;
+        uint32_t j;
+        if (audioSamples->findSampleAtTime(
+            reqAudioTime, &j, SampleTable::kFlagClosest) != OK) {
+            continue;
+        }
+
+        off64_t audioOffset;
+        uint32_t audioTime;
+        CHECK_EQ((status_t)OK, audioSamples->getMetaDataForSample(
+                    j, &audioOffset, NULL, &audioTime));
+
+        int64_t audioTimeUs = (int64_t)(audioTime * 1E6 / audio->timescale);
+
+        off64_t offsetDiff = videoOffset - audioOffset;
+        if (offsetDiff < 0) {
+            offsetDiff = -offsetDiff;
+        }
+
+#if 0
+        printf("%s%d/%d videoTime %.2f secs audioTime %.2f secs "
+               "videoOffset %lld audioOffset %lld offsetDiff %lld\n",
+               isSync ? "*" : " ",
+               i,
+               j,
+               videoTimeUs / 1E6,
+               audioTimeUs / 1E6,
+               videoOffset,
+               audioOffset,
+               offsetDiff);
+#endif
+
+        if (offsetDiff > maxOffsetDiff) {
+            maxOffsetDiff = offsetDiff;
+            maxOffsetTimeUs = videoTimeUs;
+        }
+    }
+
+#if 0
+    printf("max offset diff: %lld at video time: %.2f secs\n",
+           maxOffsetDiff, maxOffsetTimeUs / 1E6);
+#endif
+
+    if (maxOffsetDiff < 1024 * 1024) {
+        return OK;
+    }
+
+    LOGE("This content is not streamable, "
+         "max offset diff: %lld at video time: %.2f secs",
+         maxOffsetDiff, maxOffsetTimeUs / 1E6);
+
+    return ERROR_UNSUPPORTED;
+}
+
 static bool LegacySniffMPEG4(
         const sp<DataSource> &source, String8 *mimeType, float *confidence) {
     uint8_t header[8];