Re: [PATCH v7 3/3] drm/imagination: Release pm references in case of error

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] drm/imagination: misplaced code adds dead block and fails to fix PM leaks
--

commit ee55e37dd6c9d6ec50a0931a6efbed821fcc945d
Author: Alexandru Dadu <alexandru.dadu@imgtec.com>
drm/imagination: Release pm references in case of error

This commit attempts to fix power management reference leaks by adding early
releases on error.

> diff --git a/drivers/gpu/drm/imagination/pvr_queue.c b/drivers/gpu/drm/imagination/pvr_queue.c
> index b58c0887cafe5..c72f1d0b8bc00 100644
> --- a/drivers/gpu/drm/imagination/pvr_queue.c
> +++ b/drivers/gpu/drm/imagination/pvr_queue.c
> @@ -820,6 +820,19 @@ static struct dma_fence *pvr_queue_run_job(struct drm_sched_job *sched_job)
>  	if (WARN_ON(err))
>  		goto err_release;
>  
> +	err = pvr_job_get_pm_ref(job);
> +
> +	if (WARN_ON(err))
> +		goto err_release;
> +
> +	if (job->paired_job) {
> +		err = pvr_job_get_pm_ref(job->paired_job);
> +		if (WARN_ON(err)) {
> +			pvr_job_release_pm_ref(job);
> +			goto err_release;
> +		}
> +	}
> +

[Severity: Medium]
Does this block act as dead code instead of fixing the error paths?

This code is placed in the success path of pvr_queue_run_job() where the PM
references are already successfully held from earlier in the function. Because
job->has_pm_ref is already true when execution reaches here,
pvr_job_get_pm_ref() simply returns 0 and does nothing.

>  	return dma_fence_get(job->done_fence);
>  
>  err_release:
>  	pvr_kccb_release_slot(pvr_dev);
>  	return ERR_PTR(err);
>  }

[Severity: Medium]
Does the targeted PM leak remain unfixed in this error path?

When an internal job submission fails (for example, if
pvr_cccb_send_kccb_kick() fails) and execution jumps to err_release, the PM
references acquired earlier in the function are not released, which would leak
the references and prevent the GPU from entering runtime suspend.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260924-b4-upstream-propagate-all-errors-from-kccb-cmd-submission-code-v7-0-584796cb599d@imgtec.com?part=3