[PATCH V3] drm/sched: Fix fence reference count leak

[Philipp Stanner]

On Wed, 2025-02-26 at 17:05 +0800, Qianyi Liu wrote:
> From: qianyi liu <liuqianyi125 at gmail.com>
> 
> The last_scheduled fence leaked when an entity was being killed and
> adding its callback failed.
> 
> Decrement the reference count of prev when dma_fence_add_callback()
> fails, ensuring proper balance.
> 
> Cc: stable at vger.kernel.org
> Fixes: 2fdb8a8f07c2 ("drm/scheduler: rework entity flush, kill and
> fini")
> Signed-off-by: qianyi liu <liuqianyi125 at gmail.com>

@Matt: since you in principle agreed with this patch, could you give it
an official RB?

I could then take it [but will probably rephrase some nits in the
commit message]


P.

> ---
> v2 -> v3: Rework commit message (Markus)
> v1 -> v2: Added 'Fixes:' tag and clarified commit message (Philipp
> and Matthew)
> ---
>  drivers/gpu/drm/scheduler/sched_entity.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/scheduler/sched_entity.c
> b/drivers/gpu/drm/scheduler/sched_entity.c
> index 69bcf0e99d57..1c0c14bcf726 100644
> --- a/drivers/gpu/drm/scheduler/sched_entity.c
> +++ b/drivers/gpu/drm/scheduler/sched_entity.c
> @@ -259,9 +259,12 @@ static void drm_sched_entity_kill(struct
> drm_sched_entity *entity)
>  		struct drm_sched_fence *s_fence = job->s_fence;
>  
>  		dma_fence_get(&s_fence->finished);
> -		if (!prev || dma_fence_add_callback(prev, &job-
> >finish_cb,
> -					  
> drm_sched_entity_kill_jobs_cb))
> +		if (!prev ||
> +		    dma_fence_add_callback(prev, &job->finish_cb,
> +					  
> drm_sched_entity_kill_jobs_cb)) {
> +			dma_fence_put(prev);
>  			drm_sched_entity_kill_jobs_cb(NULL, &job-
> >finish_cb);
> +		}
>  
>  		prev = &s_fence->finished;
>  	}