[FFmpeg-devel] [PATCH 1/3] avcodec/mv30: Fix multiple integer overflows in idct_1d()

Michael Niedermayer michael at niedermayer.cc
Mon May 25 01:38:41 EEST 2020


Fixes: signed integer overflow: -4869937 * 473 cannot be represented in type 'int'
Fixes: 21934/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_MV30_fuzzer-5667289925156864

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael at niedermayer.cc>
---
 libavcodec/mv30.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/libavcodec/mv30.c b/libavcodec/mv30.c
index 658f32c6ff..013a5753fe 100644
--- a/libavcodec/mv30.c
+++ b/libavcodec/mv30.c
@@ -107,7 +107,7 @@ static inline void idct_1d(int *blk, int step)
     const int t0 = blk[0 * step] + blk[4 * step];
     const int t1 = blk[0 * step] - blk[4 * step];
     const int t2 = blk[2 * step] + blk[6 * step];
-    const int t3 = (((blk[2 * step] - blk[6 * step]) * 362) >> 8) - t2;
+    const int t3 = ((int)((blk[2 * step] - blk[6 * step]) * 362U) >> 8) - t2;
     const int t4 = t0 + t2;
     const int t5 = t0 - t2;
     const int t6 = t1 + t3;
@@ -117,10 +117,10 @@ static inline void idct_1d(int *blk, int step)
     const int tA = blk[1 * step] + blk[7 * step];
     const int tB = blk[1 * step] - blk[7 * step];
     const int tC = t8 + tA;
-    const int tD = (tB + t9) * 473 >> 8;
-    const int tE = ((t9 * -669 >> 8) - tC) + tD;
-    const int tF = ((tA - t8) * 362 >> 8) - tE;
-    const int t10 = ((tB * 277 >> 8) - tD) + tF;
+    const int tD = (int)((tB + t9) * 473U) >> 8;
+    const int tE = (((int)(t9 * -669U) >> 8) - tC) + tD;
+    const int tF = ((int)((tA - t8) * 362U) >> 8) - tE;
+    const int t10 = (((int)(tB * 277U) >> 8) - tD) + tF;
 
     blk[0 * step] = t4 + tC;
     blk[1 * step] = t6 + tE;
-- 
2.17.1



More information about the ffmpeg-devel mailing list