[vlc-devel] [PATCH] vout: spu: fix data-race

[Rémi Denis-Courmont]

Le tiistaina 7. toukokuuta 2019, 15.17.44 EEST Thomas Guillem a écrit :
> The clock pointer, used by the vout thread was already protected by this
> lock.

Adding the lock like that still leaves the possibility that a timestamp is 
computed with one clock, and then used after the clock is changed. How does it 
work then?

> ---
>  src/video_output/vout_subpictures.c | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/src/video_output/vout_subpictures.c
> b/src/video_output/vout_subpictures.c index 6b106534d3..ae242dda11 100644
> --- a/src/video_output/vout_subpictures.c
> +++ b/src/video_output/vout_subpictures.c
> @@ -1489,19 +1489,25 @@ void spu_Detach(spu_t *spu)
> 
>  void spu_clock_Set(spu_t *spu, vlc_clock_t *clock)
>  {
> +    vlc_mutex_lock(&spu->p->lock);
>      spu->p->clock = clock;
> +    vlc_mutex_unlock(&spu->p->lock);
>  }
> 
>  void spu_clock_Reset(spu_t *spu)
>  {
> +    vlc_mutex_lock(&spu->p->lock);
>      if (spu->p->clock)
>          vlc_clock_Reset(spu->p->clock);
> +    vlc_mutex_unlock(&spu->p->lock);
>  }
> 
>  void spu_clock_SetDelay(spu_t *spu, vlc_tick_t delay)
>  {
> +    vlc_mutex_lock(&spu->p->lock);
>      if (spu->p->clock)
>          vlc_clock_SetDelay(spu->p->clock, delay);
> +    vlc_mutex_unlock(&spu->p->lock);
>  }
> 
>  /**


-- 
Rémi Denis-Courmont
http://www.remlab.net/