Re: [PATCH 2/9] media: iris: account for UBWC extradata in the QC08C buffer size

On 7/10/2026 1:06 AM, Dmitry Baryshkov wrote:
> iris_yuv_buffer_size_qc08c() sums only the luma/chroma data and metadata
> planes. The UBWC engine additionally writes a trailing extradata/scratch
> region below the frame, so the computed size is too small for the
> reconstruction (DPB) buffer. When the firmware writes that trailing region
> it overruns the buffer and triggers an IOMMU fault; this is hit when
> decoding interlaced content, where the reconstruction buffers are packed
> with a gap after each buffer.
>
> Add the trailing region to the buffer size, sized as the venus driver does
> (max of 16K and 48 luma lines).
>
> Fixes: cfd71b14b0d6 ("media: iris: Add support for QC08C format for decoder")
> Assisted-by: Claude:claude-opus-4-8
> Signed-off-by: Dmitry Baryshkov <[email protected]>
> ---
>  drivers/media/platform/qcom/iris/iris_buffer.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/media/platform/qcom/iris/iris_buffer.c b/drivers/media/platform/qcom/iris/iris_buffer.c
> index 246ad0abbac3..f1de4442be1f 100644
> --- a/drivers/media/platform/qcom/iris/iris_buffer.c
> +++ b/drivers/media/platform/qcom/iris/iris_buffer.c
> @@ -261,7 +261,8 @@ static u32 iris_yuv_buffer_size_qc08c(struct iris_inst *inst)
>  	uv_plane = ALIGN(uv_stride * ALIGN(f->fmt.pix_mp.height / 2, UV_SCANLINE_ALIGN_QC08C),
>  			 PIXELS_4K);
>  
> -	return ALIGN(y_meta_plane + y_plane + uv_meta_plane + uv_plane, PIXELS_4K);
> +	return ALIGN(y_meta_plane + y_plane + uv_meta_plane + uv_plane +
> +		     max_t(u32, SZ_16K, y_stride * 48), PIXELS_4K);

Can max_t(u32, SZ_16K, y_stride * 48) be stored in a variable such as
trailing_region or extradata and used here?

>  }
>  
>  /*
>