avformat/matroskadec: Don't skip too much when unseekable

The Matroska (and WebM) file format achieves forward-compability by
insisting that demuxers ignore and skip elements they don't know about.
Unfortunately, this complicates the detection of errors as errors
resulting from loosing sync can't be reliably distinguished from
unknown elements that are part of a future version of the standard.

Up until now, the strategy to deal with this situation was to skip all
unknown elements that are not obviously erroneous; if an error happened,
it was tried to seek to the last known good position to resync from (and
resync to level 1 elements). This is working fine if the input is
seekable, but if it is not, then the skipped data can usually not be
rechecked lateron. This is particularly acute if unknown-length clusters
are in use, as the check for whether a child element exceeds the
containing master element is ineffective in this situation.

To remedy this, a new heuristic has been introduced: If an unknown
element is encountered in non-seekable mode, an error is presumed to
have happened based upon a combination of the length of the row of the
already encountered unknown elements and of how far away skipping this
element would take us.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>

1 files changed, 52 insertions, 2 deletions

diff --git a/libavformat/matroskadec.c b/libavformat/matroskadec.c
index 2a1ad485d2..2485d8edf4 100644
--- a/libavformat/matroskadec.c
+++ b/libavformat/matroskadec.c
@@ -73,6 +73,12 @@
                                          * still need to be performed */
 #define LEVEL_ENDED                   3 /* return value of ebml_parse when the
                                          * syntax level used for parsing ended. */
+#define SKIP_THRESHOLD      1024 * 1024 /* In non-seekable mode, if more than SKIP_THRESHOLD
+                                         * of unkown, potentially damaged data is encountered,
+                                         * it is considered an error. */
+#define UNKNOWN_EQUIV         50 * 1024 /* An unknown element is considered equivalent
+                                         * to this many bytes of unknown data for the
+                                         * SKIP_THRESHOLD check. */
 
 typedef enum {
     EBML_NONE,
@@ -338,6 +344,7 @@ typedef struct MatroskaDemuxContext {
     int      num_levels;
     uint32_t current_id;
     int64_t  resync_pos;
+    int      unknown_count;
 
     uint64_t time_scale;
     double   duration;
@@ -763,8 +770,9 @@ static int matroska_reset_status(MatroskaDemuxContext *matroska,
             return err;
     }
 
-    matroska->current_id = id;
-    matroska->num_levels = 1;
+    matroska->current_id    = id;
+    matroska->num_levels    = 1;
+    matroska->unknown_count = 0;
     matroska->resync_pos = avio_tell(matroska->ctx->pb);
     if (id)
         matroska->resync_pos -= (av_log2(id) + 7) / 8;
@@ -1289,6 +1297,48 @@ static int ebml_parse(MatroskaDemuxContext *matroska,
         } else
             level_check = 0;
 
+        if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
+            // Loosing sync will likely manifest itself as encountering unknown
+            // elements which are not reliably distinguishable from elements
+            // belonging to future extensions of the format.
+            // We use a heuristic to detect such situations: If the current
+            // element is not expected at the current syntax level and there
+            // were only a few unknown elements in a row, then the element is
+            // skipped or considered defective based upon the length of the
+            // current element (i.e. how much would be skipped); if there were
+            // more than a few skipped elements in a row and skipping the current
+            // element would lead us more than SKIP_THRESHOLD away from the last
+            // known good position, then it is inferred that an error occured.
+            // The dependency on the number of unknown elements in a row exists
+            // because the distance to the last known good position is
+            // automatically big if the last parsed element was big.
+            // In both cases, each unknown element is considered equivalent to
+            // UNKNOWN_EQUIV of skipped bytes for the check.
+            // The whole check is only done for non-seekable output, because
+            // in this situation skipped data can't simply be rechecked later.
+            // This is especially important when using unkown length elements
+            // as the check for whether a child exceeds its containing master
+            // element is not effective in this situation.
+            if (update_pos) {
+                matroska->unknown_count = 0;
+            } else {
+                int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
+
+                if (matroska->unknown_count > 3)
+                    dist += pos_alt - matroska->resync_pos;
+
+                if (dist > SKIP_THRESHOLD) {
+                    av_log(matroska->ctx, AV_LOG_ERROR,
+                           "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
+                           "length 0x%"PRIx64" considered as invalid data. Last "
+                           "known good position 0x%"PRIx64", %d unknown elements"
+                           " in a row\n", id, pos, length, matroska->resync_pos,
+                           matroska->unknown_count);
+                    return AVERROR_INVALIDDATA;
+                }
+            }
+        }
+
         if (update_pos) {
             // We have found an element that is allowed at this place
             // in the hierarchy and it passed all checks, so treat the beginning