| Message ID | 1553160583-14557-1-git-send-email-decai.lin@intel.com |
|---|---|
| State | New |
| Headers | show |
> -----Original Message----- > From: ffmpeg-devel [mailto:ffmpeg-devel-bounces@ffmpeg.org] On Behalf > Of Decai Lin > Sent: Thursday, March 21, 2019 17:30 > To: ffmpeg-devel@ffmpeg.org > Cc: Yan, FengX <fengx.yan@intel.com>; Lin, Decai <decai.lin@intel.com> > Subject: [FFmpeg-devel] [PATCH v1 1/1] vaapi_hevc: Fix double-free issue. > > From: "Yan, FengX" <fengx.yan@intel.com> > > Signed-off-by: Yan, FengX <fengx.yan@intel.com> > Signed-off-by: Decai Lin <decai.lin@intel.com> > --- > libavcodec/vaapi_hevc.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/libavcodec/vaapi_hevc.c b/libavcodec/vaapi_hevc.c > index 19aabcd..373ffc4 100644 > --- a/libavcodec/vaapi_hevc.c > +++ b/libavcodec/vaapi_hevc.c > @@ -271,7 +271,7 @@ static int vaapi_hevc_end_frame(AVCodecContext > *avctx) > > ret = ff_vaapi_decode_issue(avctx, &pic->pic); > if (ret < 0) > - goto fail; > + return ret; > > return 0; > fail: I believe this should match the behavior in vaapi_h264/mpeg4/.../vc1.c, and remove the redundant ff_vaapi_decode_cancel in vaapi_hevc_end_frame(). --Linjie
> -----Original Message----- > From: Fu, Linjie > Sent: 2019年3月21日 20:20 > To: FFmpeg development discussions and patches > <ffmpeg-devel@ffmpeg.org> > Cc: Yan, FengX <fengx.yan@intel.com>; Lin, Decai <decai.lin@intel.com> > Subject: RE: [FFmpeg-devel] [PATCH v1 1/1] vaapi_hevc: Fix double-free > issue. > > > -----Original Message----- > > From: ffmpeg-devel [mailto:ffmpeg-devel-bounces@ffmpeg.org] On Behalf > > Of Decai Lin > > Sent: Thursday, March 21, 2019 17:30 > > To: ffmpeg-devel@ffmpeg.org > > Cc: Yan, FengX <fengx.yan@intel.com>; Lin, Decai <decai.lin@intel.com> > > Subject: [FFmpeg-devel] [PATCH v1 1/1] vaapi_hevc: Fix double-free issue. > > > > From: "Yan, FengX" <fengx.yan@intel.com> > > > > Signed-off-by: Yan, FengX <fengx.yan@intel.com> > > Signed-off-by: Decai Lin <decai.lin@intel.com> > > --- > > libavcodec/vaapi_hevc.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git a/libavcodec/vaapi_hevc.c b/libavcodec/vaapi_hevc.c index > > 19aabcd..373ffc4 100644 > > --- a/libavcodec/vaapi_hevc.c > > +++ b/libavcodec/vaapi_hevc.c > > @@ -271,7 +271,7 @@ static int vaapi_hevc_end_frame(AVCodecContext > > *avctx) > > > > ret = ff_vaapi_decode_issue(avctx, &pic->pic); > > if (ret < 0) > > - goto fail; > > + return ret; > > > > return 0; > > fail: > > I believe this should match the behavior in vaapi_h264/mpeg4/.../vc1.c, and > remove the redundant ff_vaapi_decode_cancel in vaapi_hevc_end_frame(). > > --Linjie Hello Linjie, there is little different between vaapi_hevc and other vaapi codecs. vaapi_hevc has an extra code block to handle slice buffer before ff_vaapi_decode_issue() , and if it's failed to handle, the buffers have to be destroyed separately. So the calling of ff_vaapi_decode_cancel() here is a must for hevc.
diff --git a/libavcodec/vaapi_hevc.c b/libavcodec/vaapi_hevc.c index 19aabcd..373ffc4 100644 --- a/libavcodec/vaapi_hevc.c +++ b/libavcodec/vaapi_hevc.c @@ -271,7 +271,7 @@ static int vaapi_hevc_end_frame(AVCodecContext *avctx) ret = ff_vaapi_decode_issue(avctx, &pic->pic); if (ret < 0) - goto fail; + return ret; return 0; fail: