[PATCH] drm/sched: Don't crash kernel on wrong params

[PATCH] drm/sched: Don't crash kernel on wrong params

[Philipp Stanner]

drm_sched_job_arm() just panics the kernel with BUG_ON() in case of an
entity being NULL. While drm_sched_job_arm() crashing or not effectively
arming jobs is certainly a huge issue that needs to be noticed,
completely shooting down the kernel reduces the probability of reaching
and debugging a system to 0.

Moreover, the checkpatch script by now strongly discourages all new uses
of BUG_ON() for this reason.

Replace the BUG_ON() in drm_sched_job_arm() with a WARN_ON().

Signed-off-by: Philipp Stanner <phasta@kernel.org>
---
 drivers/gpu/drm/scheduler/sched_main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
index 1d4f1b822e7b..3bf4ae0ca4bc 100644
--- a/drivers/gpu/drm/scheduler/sched_main.c
+++ b/drivers/gpu/drm/scheduler/sched_main.c
@@ -857,7 +857,7 @@ void drm_sched_job_arm(struct drm_sched_job *job)
 	struct drm_gpu_scheduler *sched;
 	struct drm_sched_entity *entity = job->entity;
 
-	BUG_ON(!entity);
+	WARN_ON(!entity);
 	drm_sched_entity_select_rq(entity);
 	sched = entity->rq->sched;
 
-- 
2.49.0

Re: [PATCH] drm/sched: Don't crash kernel on wrong params

[Alex Deucher]

On Fri, Nov 7, 2025 at 8:01 AM Philipp Stanner <phasta@kernel.org> wrote:
>
> drm_sched_job_arm() just panics the kernel with BUG_ON() in case of an
> entity being NULL. While drm_sched_job_arm() crashing or not effectively
> arming jobs is certainly a huge issue that needs to be noticed,
> completely shooting down the kernel reduces the probability of reaching
> and debugging a system to 0.
>
> Moreover, the checkpatch script by now strongly discourages all new uses
> of BUG_ON() for this reason.
>
> Replace the BUG_ON() in drm_sched_job_arm() with a WARN_ON().
>
> Signed-off-by: Philipp Stanner <phasta@kernel.org>

Acked-by: Alex Deucher <alexander.deucher@amd.com>

> ---
>  drivers/gpu/drm/scheduler/sched_main.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
> index 1d4f1b822e7b..3bf4ae0ca4bc 100644
> --- a/drivers/gpu/drm/scheduler/sched_main.c
> +++ b/drivers/gpu/drm/scheduler/sched_main.c
> @@ -857,7 +857,7 @@ void drm_sched_job_arm(struct drm_sched_job *job)
>         struct drm_gpu_scheduler *sched;
>         struct drm_sched_entity *entity = job->entity;
>
> -       BUG_ON(!entity);
> +       WARN_ON(!entity);
>         drm_sched_entity_select_rq(entity);
>         sched = entity->rq->sched;
>
> --
> 2.49.0
>

Re: [PATCH] drm/sched: Don't crash kernel on wrong params

[Tvrtko Ursulin]

On 07/11/2025 12:55, Philipp Stanner wrote:
> drm_sched_job_arm() just panics the kernel with BUG_ON() in case of an
> entity being NULL. While drm_sched_job_arm() crashing or not effectively
> arming jobs is certainly a huge issue that needs to be noticed,
> completely shooting down the kernel reduces the probability of reaching
> and debugging a system to 0.
> 
> Moreover, the checkpatch script by now strongly discourages all new uses
> of BUG_ON() for this reason.
> 
> Replace the BUG_ON() in drm_sched_job_arm() with a WARN_ON().
> 
> Signed-off-by: Philipp Stanner <phasta@kernel.org>
> ---
>   drivers/gpu/drm/scheduler/sched_main.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
> index 1d4f1b822e7b..3bf4ae0ca4bc 100644
> --- a/drivers/gpu/drm/scheduler/sched_main.c
> +++ b/drivers/gpu/drm/scheduler/sched_main.c
> @@ -857,7 +857,7 @@ void drm_sched_job_arm(struct drm_sched_job *job)
>   	struct drm_gpu_scheduler *sched;
>   	struct drm_sched_entity *entity = job->entity;
>   
> -	BUG_ON(!entity);
> +	WARN_ON(!entity);
>   	drm_sched_entity_select_rq(entity);

void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
{
	struct dma_fence *fence;
	struct drm_gpu_scheduler *sched;
	struct drm_sched_rq *rq;

	/* single possible engine and already selected */
	if (!entity->sched_list)

Still the same end result.

I believe best practice is to simply not have BUG_ON's followed by null 
pointer dereferences since they add no value. Ie. it should just be 
removed and not replaced.

Regards,

Tvrtko

Regards,

Tvrtko

>   	sched = entity->rq->sched;
>   

Re: [PATCH] drm/sched: Don't crash kernel on wrong params

[Tvrtko Ursulin]

On 07/11/2025 14:42, Tvrtko Ursulin wrote:
> 
> On 07/11/2025 12:55, Philipp Stanner wrote:
>> drm_sched_job_arm() just panics the kernel with BUG_ON() in case of an
>> entity being NULL. While drm_sched_job_arm() crashing or not effectively
>> arming jobs is certainly a huge issue that needs to be noticed,
>> completely shooting down the kernel reduces the probability of reaching
>> and debugging a system to 0.
>>
>> Moreover, the checkpatch script by now strongly discourages all new uses
>> of BUG_ON() for this reason.
>>
>> Replace the BUG_ON() in drm_sched_job_arm() with a WARN_ON().
>>
>> Signed-off-by: Philipp Stanner <phasta@kernel.org>
>> ---
>>   drivers/gpu/drm/scheduler/sched_main.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/ 
>> scheduler/sched_main.c
>> index 1d4f1b822e7b..3bf4ae0ca4bc 100644
>> --- a/drivers/gpu/drm/scheduler/sched_main.c
>> +++ b/drivers/gpu/drm/scheduler/sched_main.c
>> @@ -857,7 +857,7 @@ void drm_sched_job_arm(struct drm_sched_job *job)
>>       struct drm_gpu_scheduler *sched;
>>       struct drm_sched_entity *entity = job->entity;
>> -    BUG_ON(!entity);
>> +    WARN_ON(!entity);
>>       drm_sched_entity_select_rq(entity);
> 
> void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
> {
>      struct dma_fence *fence;
>      struct drm_gpu_scheduler *sched;
>      struct drm_sched_rq *rq;
> 
>      /* single possible engine and already selected */
>      if (!entity->sched_list)
> 
> Still the same end result.
> 
> I believe best practice is to simply not have BUG_ON's followed by null 

Sorry I meant WARN_ONs. If we think there is scope for bad things to 
happen after letting the execution proceed then I would add an early 
return to the WARN_ON (_ONCE?). That probably means to push job as well.

Regards,

Tvrtko

> pointer dereferences since they add no value. Ie. it should just be 
> removed and not replaced.
> 
> Regards,
> 
> Tvrtko
> 
> Regards,
> 
> Tvrtko
> 
>>       sched = entity->rq->sched;
> 

Re: [PATCH] drm/sched: Don't crash kernel on wrong params

[Philipp Stanner]

On Fri, 2025-11-07 at 14:42 +0000, Tvrtko Ursulin wrote:
> 
> On 07/11/2025 12:55, Philipp Stanner wrote:
> > drm_sched_job_arm() just panics the kernel with BUG_ON() in case of an
> > entity being NULL. While drm_sched_job_arm() crashing or not effectively
> > arming jobs is certainly a huge issue that needs to be noticed,
> > completely shooting down the kernel reduces the probability of reaching
> > and debugging a system to 0.
> > 
> > Moreover, the checkpatch script by now strongly discourages all new uses
> > of BUG_ON() for this reason.
> > 
> > Replace the BUG_ON() in drm_sched_job_arm() with a WARN_ON().
> > 
> > Signed-off-by: Philipp Stanner <phasta@kernel.org>
> > ---
> >   drivers/gpu/drm/scheduler/sched_main.c | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
> > index 1d4f1b822e7b..3bf4ae0ca4bc 100644
> > --- a/drivers/gpu/drm/scheduler/sched_main.c
> > +++ b/drivers/gpu/drm/scheduler/sched_main.c
> > @@ -857,7 +857,7 @@ void drm_sched_job_arm(struct drm_sched_job *job)
> >   	struct drm_gpu_scheduler *sched;
> >   	struct drm_sched_entity *entity = job->entity;
> >   
> > -	BUG_ON(!entity);
> > +	WARN_ON(!entity);
> >   	drm_sched_entity_select_rq(entity);
> 
> void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
> {
> 	struct dma_fence *fence;
> 	struct drm_gpu_scheduler *sched;
> 	struct drm_sched_rq *rq;
> 
> 	/* single possible engine and already selected */
> 	if (!entity->sched_list)
> 
> Still the same end result.
> 
> I believe best practice is to simply not have BUG_ON's followed by null 
> pointer dereferences since they add no value. Ie. it should just be 
> removed and not replaced.

Hm, right. I agree. Both jobs and entities are initialized to 0
anyways. Existing users work, and new users would notice a null ptr
crash.

Let's remove it altogether.


P.