[FFmpeg-devel] [PATCH 1/2] fftools/ffmpeg_filter: fix null pointer dereference

[Anton Khirnov]

Quoting Zhao Zhili (2023-12-18 13:04:24)
> 
> 
> > On Dec 18, 2023, at 19:19, Anton Khirnov <anton at khirnov.net> wrote:
> > 
> > Quoting Zhao Zhili (2023-12-17 12:01:08)
> >> From: Zhao Zhili <zhilizhao at tencent.com>
> >> 
> >> A dummy frame is created with format NONE passed to enc_open(),
> >> which doesn't prepare for it. The null pointer dereference happened
> >> at av_pix_fmt_desc_get(enc_ctx->pix_fmt)->comp[0].depth.
> >> 
> >> frame #0: 0x0000005555bc34a4 ffmpeg_g`enc_open(opaque=0xb400007efe2db690, frame=0xb400007efe2d9f70) at ffmpeg_enc.c:235:44
> >> frame #1: 0x0000005555bef250 ffmpeg_g`enc_open(sch=0xb400007dde2d4090, enc=0xb400007e4e2daad0, frame=0xb400007efe2d9f70) at ffmpeg_sched.c:1462:11
> >> frame #2: 0x0000005555bee094 ffmpeg_g`send_to_enc(sch=0xb400007dde2d4090, enc=0xb400007e4e2daad0, frame=0xb400007efe2d9f70) at ffmpeg_sched.c:1571:19
> >> frame #3: 0x0000005555bee01c ffmpeg_g`sch_filter_send(sch=0xb400007dde2d4090, fg_idx=0, out_idx=0, frame=0xb400007efe2d9f70) at ffmpeg_sched.c:2154:12
> >> frame #4: 0x0000005555bcf124 ffmpeg_g`close_output(ofp=0xb400007e4e2d85b0, fgt=0x0000007d1790eb08) at ffmpeg_filter.c:2225:15
> >> frame #5: 0x0000005555bcb000 ffmpeg_g`fg_output_frame(ofp=0xb400007e4e2d85b0, fgt=0x0000007d1790eb08, frame=0x0000000000000000) at ffmpeg_filter.c:2317:16
> >> frame #6: 0x0000005555bc7e48 ffmpeg_g`filter_thread(arg=0xb400007eae2ce7a0) at ffmpeg_filter.c:2836:15
> >> frame #7: 0x0000005555bee568 ffmpeg_g`task_wrapper(arg=0xb400007d8e2db478) at ffmpeg_sched.c:2200:21
> >> 
> >> Signed-off-by: Zhao Zhili <zhilizhao at tencent.com>
> >> ---
> >> This bug is trigged by 10bit H.264 which doesn't supported by mediacodec.
> >> There is some misleading error messages from ffmpeg cli:
> >> 
> >> [h264_mediacodec @ 0xb400007513788010] Failed to dequeue output buffer (status=-10000)
> >> [vist#0:0/h264 @ 0xb4000074a3786b30] Error submitting packet to decoder: Generic error in an external library
> >> [h264_mediacodec @ 0xb400007513788010] Failed to dequeue output buffer (status=-10000)
> >> [vist#0:0/h264 @ 0xb4000074a3786b30] Error submitting packet to decoder: Generic error in an external library
> >> [vist#0:0/h264 @ 0xb4000074a3786b30] A decoder returned an unexpected error code. This is a bug, please report it.
> >> [vist#0:0/h264 @ 0xb4000074a3786b30] Error processing packet in decoder: Internal bug, should not have happened
> >> 
> >> fftools/ffmpeg_filter.c | 2 +-
> >> 1 file changed, 1 insertion(+), 1 deletion(-)
> >> 
> >> diff --git a/fftools/ffmpeg_filter.c b/fftools/ffmpeg_filter.c
> >> index bb755d7bb4..9dc47f9d90 100644
> >> --- a/fftools/ffmpeg_filter.c
> >> +++ b/fftools/ffmpeg_filter.c
> >> @@ -2192,7 +2192,7 @@ static int close_output(OutputFilterPriv *ofp, FilterGraphThread *fgt)
> >> 
> >>     // we are finished and no frames were ever seen at this output,
> >>     // at least initialize the encoder with a dummy frame
> >> -    if (!fgt->got_frame) {
> >> +    if (!fgt->got_frame && ofp->format != -1) {
> > 
> > I don't quite understand how precisely does this happen.
> > 
> > This code should only be reachable if the filtergraph was configured at
> > least once, then configure_filtergraph() should set ofp->format to the
> > format reported by the lavfi buffersink. Does this then mean that lavfi
> > is configured with AV_PIX_FMT_NONE? Or that this is somehow triggered
> > without the filtergraph being configured?
> 
> 
> Can reproduced with:
> 
> diff --git a/libavcodec/qsvdec.c b/libavcodec/qsvdec.c
> index 559f63698a..af4ea37812 100644
> --- a/libavcodec/qsvdec.c
> +++ b/libavcodec/qsvdec.c
> @@ -1050,6 +1050,7 @@ static int qsv_decode_frame(AVCodecContext *avctx, AVFrame *frame,
>      QSVDecContext *s = avctx->priv_data;
>      int ret;
>  
> +    return AVERROR_EXTERNAL;
>      /* buffer the input packet */
>      if (avpkt->size) {
>          AVPacket input_ref;
> 
> Then
> ./ffmpeg -c:v h264_qsv -xerror  -I input.mp4 -c:a copy  -c:v libx264  output.mp4 
> 
> Looks like filtergraph isn’t being configured.

I see, then seems to me it'd be better to run the final loop in
filter_thread() (the one calling fg_output_frame(, NULL)) only when
fgt->graph is non-NULL. That should fix this issue as well.

Thanks,
-- 
Anton Khirnov