Re: [PATCH v2 5/6] media: verisilicon: Reinitialize the IOMMU after resetting an encoder core

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [Medium] media: verisilicon: missing v4l2_ctrl_request_complete() in
  device_run()
--

commit 9b8b7ddcd6a2f08763d5a6b0f18fb2a01381b678
Author: Arun Iyengar <aiyengar1990@gmail.com>

media: verisilicon: Reinitialize the IOMMU after resetting an encoder core

This commit allocates an empty IOMMU paging domain at probe and attaches
and detaches it after resetting a Rockchip encoder core so the IOMMU is
reinitialized, refusing further jobs if the reattachment fails.

> diff --git a/drivers/media/platform/verisilicon/hantro_drv.c b/drivers/media/platform/verisilicon/hantro_drv.c
> index 4542322dba71a..6d027c2770387 100644
> --- a/drivers/media/platform/verisilicon/hantro_drv.c
> +++ b/drivers/media/platform/verisilicon/hantro_drv.c

[ ... ]

> @@ -174,6 +204,11 @@ static void device_run(void *priv)
>  	src = hantro_get_src_buf(ctx);
>  	dst = hantro_get_dst_buf(ctx);
>  
> +	if (unlikely(vpu->iommu_lost)) {
> +		hantro_job_finish_no_pm(ctx, VB2_BUF_STATE_ERROR);
> +		return;
> +	}

[Severity: Medium]
Does this leave the control handler's media_request_object incomplete when
src was queued with a media_request containing controls?

Normal runs complete the control request object in
hantro_end_prepare_run() via v4l2_ctrl_request_complete(
src_buf->vb2_buf.req_obj.req, &ctx->ctrl_handler).  Returning early via
hantro_job_finish_no_pm() calls v4l2_m2m_buf_done_and_job_finish() ->
vb2_buffer_done(), which removes src from the m2m queue and clears
src->vb2_buf.req_obj.req without completing the control object (as on the
adjacent pm_runtime_resume_and_get() error path).  The media_request then
stays stuck in MEDIA_REQUEST_STATE_QUEUED until the file descriptor is
closed.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261112-hantro-robust-v2-0-18561fd48838@gmail.com?part=5