[FFmpeg-devel] [PATCH] hwcontext_d3d11va: properly reset values after release/close

Hendrik Leppkes h.leppkes at gmail.com
Fri Nov 24 11:00:18 EET 2017


On Fri, Nov 24, 2017 at 2:01 AM, Jan Ekström <jeebjp at gmail.com> wrote:
> Makes the uninit function re-entrable, which can be a common case
> when an API user first tries to initialize its context, fails, and
> then finally unrefs the AVHWDevice.
>
> Fixes a crash reported by sm2345 on IRC.
> ---
>  libavutil/hwcontext_d3d11va.c | 21 ++++++++++++++++-----
>  1 file changed, 16 insertions(+), 5 deletions(-)
>
> diff --git a/libavutil/hwcontext_d3d11va.c b/libavutil/hwcontext_d3d11va.c
> index 65dd6651fc..845a4a45fe 100644
> --- a/libavutil/hwcontext_d3d11va.c
> +++ b/libavutil/hwcontext_d3d11va.c
> @@ -458,20 +458,31 @@ static void d3d11va_device_uninit(AVHWDeviceContext *hwdev)
>  {
>      AVD3D11VADeviceContext *device_hwctx = hwdev->hwctx;
>
> -    if (device_hwctx->device)
> +    if (device_hwctx->device) {
>          ID3D11Device_Release(device_hwctx->device);
> +        device_hwctx->device = NULL;
> +    }
>
> -    if (device_hwctx->device_context)
> +    if (device_hwctx->device_context) {
>          ID3D11DeviceContext_Release(device_hwctx->device_context);
> +        device_hwctx->device_context = NULL;
> +    }
>
> -    if (device_hwctx->video_device)
> +    if (device_hwctx->video_device) {
>          ID3D11VideoDevice_Release(device_hwctx->video_device);
> +        device_hwctx->video_device = NULL;
> +    }
>
> -    if (device_hwctx->video_context)
> +    if (device_hwctx->video_context) {
>          ID3D11VideoContext_Release(device_hwctx->video_context);
> +        device_hwctx->video_context = NULL;
> +    }
>
> -    if (device_hwctx->lock == d3d11va_default_lock)
> +    if (device_hwctx->lock == d3d11va_default_lock) {
>          CloseHandle(device_hwctx->lock_ctx);
> +        device_hwctx->lock_ctx = INVALID_HANDLE_VALUE;
> +        device_hwctx->lock = NULL;
> +    }
>  }
>

LGTM. In some cases the uninit can be called twice, which seems like a
bit of an odd design, but we better don't leak any pointers out then.
Should probably go through the other hwcontext implementation to check
for similar problems.

- Hendrik


More information about the ffmpeg-devel mailing list