Re: [PATCH v4 2/2] drm/rocket: Keep scheduler allocation in rocket_file_priv

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Critical] Deterministic double free in rocket_job_close() for multi-core devices.
--

--- Patch [2]: [PATCH v4 2/2] drm/rocket: Keep scheduler allocation in rocket_file_priv ---
commit 3f1ce60677689fe14af070b96e041b01b9fbc718
Author: Triet Hoang <[email protected]>

drm/rocket: Keep scheduler allocation in rocket_file_priv

Keep the scheduler allocation in rocket_file_priv and free it
unconditionally in rocket_job_close().

Link: https://lore.kernel.org/all/[email protected]/#t

[ ... ]
> diff --git a/drivers/accel/rocket/rocket_job.c b/drivers/accel/rocket/rocket_job.c
> index adcc792541ec1..ff1d9e802024d 100644
> --- a/drivers/accel/rocket/rocket_job.c
> +++ b/drivers/accel/rocket/rocket_job.c
[ ... ]
> @@ -525,6 +528,7 @@ void rocket_job_close(struct rocket_file_priv *rocket_priv)
>  	struct drm_sched_entity *entity = &rocket_priv->sched_entity;
>  
>  	kfree(entity->sched_list);
> +	kfree(rocket_priv->scheds);

[Severity: Critical]
Does this cause a double free of the scheds array on multi-core devices?

When the device has multiple cores (rdev->num_cores > 1),
drm_sched_entity_init() internally assigns the provided scheds array to
entity->sched_list. Since the patch additionally saves this same pointer
to rocket_priv->scheds, both entity->sched_list and rocket_priv->scheds
will point to the exact same heap allocation. 

By calling kfree() on both sequentially, are we freeing the same memory twice?

>  	drm_sched_entity_destroy(entity);
>  }
>

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=2