| Message ID | 20190925203858.27870-9-michael@niedermayer.cc |
|---|---|
| State | New |
| Headers | show |
Why this does not set dirty for all decoding cases? On 9/25/19, Michael Niedermayer <michael@niedermayer.cc> wrote: > This avoids unneeded operations and makes the code faster. > > Fixes: Timeout > Fixes: > 15724/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_PAF_VIDEO_fuzzer-5750842205929472 > (12sec -> 9sec) > Fixes: > 17625/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_PAF_VIDEO_fuzzer-5640515311108096 > (16sec -> 4sec) > > Found-by: continuous fuzzing process > https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg > Signed-off-by: Michael Niedermayer <michael@niedermayer.cc> > --- > libavcodec/pafvideo.c | 10 ++++++++-- > 1 file changed, 8 insertions(+), 2 deletions(-) > > diff --git a/libavcodec/pafvideo.c b/libavcodec/pafvideo.c > index 6b4771cbce..07fa05caf8 100644 > --- a/libavcodec/pafvideo.c > +++ b/libavcodec/pafvideo.c > @@ -55,6 +55,7 @@ typedef struct PAFVideoDecContext { > > int current_frame; > uint8_t *frame[4]; > + int dirty[4]; > int frame_size; > int video_size; > > @@ -187,6 +188,7 @@ static int decode_0(PAFVideoDecContext *c, uint8_t *pkt, > uint8_t code) > j = bytestream2_get_le16(&c->gb) + offset; > if (bytestream2_get_bytes_left(&c->gb) < (j - offset) * 16) > return AVERROR_INVALIDDATA; > + c->dirty[page] = 1; > do { > offset++; > if (dst + 3 * c->width + 4 > dend) > @@ -329,9 +331,13 @@ static int paf_video_decode(AVCodecContext *avctx, void > *data, > c->pic->palette_has_changed = 1; > } > > + c->dirty[c->current_frame] = 1; > if (code & 0x20) > - for (i = 0; i < 4; i++) > - memset(c->frame[i], 0, c->frame_size); > + for (i = 0; i < 4; i++) { > + if (c->dirty[i]) > + memset(c->frame[i], 0, c->frame_size); > + c->dirty[i] = 0; > + } > > switch (code & 0x0F) { > case 0: > -- > 2.23.0 > > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-devel > > To unsubscribe, visit link above, or email > ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
On Thu, Sep 26, 2019 at 10:04:44AM +0200, Paul B Mahol wrote:
> Why this does not set dirty for all decoding cases?
dirty is set for c->current_frame, so cases which write only
into that do not need an explicit case.
But maybe i forgot something ?
Thanks
[...]
On Thu, Sep 26, 2019 at 11:18:37PM +0200, Michael Niedermayer wrote: > On Thu, Sep 26, 2019 at 10:04:44AM +0200, Paul B Mahol wrote: > > Why this does not set dirty for all decoding cases? > > dirty is set for c->current_frame, so cases which write only > into that do not need an explicit case. > > But maybe i forgot something ? do you see any remaining issues or can i apply it ? thx [...]
On 10/16/19, Michael Niedermayer <michael@niedermayer.cc> wrote: > On Thu, Sep 26, 2019 at 11:18:37PM +0200, Michael Niedermayer wrote: >> On Thu, Sep 26, 2019 at 10:04:44AM +0200, Paul B Mahol wrote: >> > Why this does not set dirty for all decoding cases? >> >> dirty is set for c->current_frame, so cases which write only >> into that do not need an explicit case. >> >> But maybe i forgot something ? > > do you see any remaining issues or can i apply it ? > Nope. > thx > > [...] > > -- > Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB > > If you fake or manipulate statistics in a paper in physics you will never > get a job again. > If you fake or manipulate statistics in a paper in medicin you will get > a job for life at the pharma industry. >
On Wed, Oct 16, 2019 at 06:27:10PM +0200, Paul B Mahol wrote: > On 10/16/19, Michael Niedermayer <michael@niedermayer.cc> wrote: > > On Thu, Sep 26, 2019 at 11:18:37PM +0200, Michael Niedermayer wrote: > >> On Thu, Sep 26, 2019 at 10:04:44AM +0200, Paul B Mahol wrote: > >> > Why this does not set dirty for all decoding cases? > >> > >> dirty is set for c->current_frame, so cases which write only > >> into that do not need an explicit case. > >> > >> But maybe i forgot something ? > > > > do you see any remaining issues or can i apply it ? > > > > Nope. ill assume that the "nope" referes to "remaining issues" and not the apply part and thus will apply in a few days thx [...]
diff --git a/libavcodec/pafvideo.c b/libavcodec/pafvideo.c index 6b4771cbce..07fa05caf8 100644 --- a/libavcodec/pafvideo.c +++ b/libavcodec/pafvideo.c @@ -55,6 +55,7 @@ typedef struct PAFVideoDecContext { int current_frame; uint8_t *frame[4]; + int dirty[4]; int frame_size; int video_size; @@ -187,6 +188,7 @@ static int decode_0(PAFVideoDecContext *c, uint8_t *pkt, uint8_t code) j = bytestream2_get_le16(&c->gb) + offset; if (bytestream2_get_bytes_left(&c->gb) < (j - offset) * 16) return AVERROR_INVALIDDATA; + c->dirty[page] = 1; do { offset++; if (dst + 3 * c->width + 4 > dend) @@ -329,9 +331,13 @@ static int paf_video_decode(AVCodecContext *avctx, void *data, c->pic->palette_has_changed = 1; } + c->dirty[c->current_frame] = 1; if (code & 0x20) - for (i = 0; i < 4; i++) - memset(c->frame[i], 0, c->frame_size); + for (i = 0; i < 4; i++) { + if (c->dirty[i]) + memset(c->frame[i], 0, c->frame_size); + c->dirty[i] = 0; + } switch (code & 0x0F) { case 0:
This avoids unneeded operations and makes the code faster. Fixes: Timeout Fixes: 15724/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_PAF_VIDEO_fuzzer-5750842205929472 (12sec -> 9sec) Fixes: 17625/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_PAF_VIDEO_fuzzer-5640515311108096 (16sec -> 4sec) Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc> --- libavcodec/pafvideo.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)