Re: [PATCH v13 01/17] media: ti: j721e-csi2rx: Remove word size alignment on frame width

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Removing word-size alignment allows odd frame widths, violating V4L2 macro-pixel requirements for packed formats.
--

commit 1c8944cd1907204dbed865890aef72e4ff1103a8
Author: Rishikesh Donadkar <r-donadkar@ti.com>

media: ti: j721e-csi2rx: Remove word size alignment on frame width

This commit removes the limitation that the frame width must be a multiple
of the word size, as the hardware does not strictly enforce this constraint.
It updates the format clamping and stepwise frame sizes to a minimum and
step of 1.

Link: https://www.ti.com/lit/pdf/spruj16

> diff --git a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
> --- a/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
> +++ b/drivers/media/platform/ti/j721e-csi2rx/j721e-csi2rx.c
[ ... ]
> @@ -360,23 +352,15 @@ static int ti_csi2rx_enum_framesizes(struct file *file, void *fh,
>  				     struct v4l2_frmsizeenum *fsize)
>  {
[ ... ]
>  	fsize->type = V4L2_FRMSIZE_TYPE_STEPWISE;
> -	fsize->stepwise.min_width = pixels_in_word;
> -	fsize->stepwise.max_width = rounddown(MAX_WIDTH_BYTES * 8 / fmt->bpp,
> -					      pixels_in_word);
> -	fsize->stepwise.step_width = pixels_in_word;
> +	fsize->stepwise.min_width = 1;
> +	fsize->stepwise.max_width = MAX_WIDTH_BYTES * 8 / fmt->bpp;
> +	fsize->stepwise.step_width = 1;

[Severity: Medium]
Does removing this alignment entirely allow invalid widths for packed
macro-pixel formats?

For formats like YUYV (YUV 4:2:2), the V4L2 specification requires widths
to be an even number, because a single macro-pixel represents 2 pixels in
4 bytes.

By hardcoding step_width to 1 and removing the width rounddown in
ti_csi2rx_fill_fmt(), userspace could request an odd width like 3.
This results in a bytesperline calculation that physically cannot hold
full macro-pixels, which may lead to stream desynchronization or unread
data in the hardware FIFO.

Should we fall back to format-specific macro-pixel alignment using
something like v4l_bound_align_image() instead of removing all alignment
constraints?

>  	fsize->stepwise.min_height = 1;
>  	fsize->stepwise.max_height = MAX_HEIGHT_LINES;
>  	fsize->stepwise.step_height = 1;

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260520120022.539913-1-r-donadkar@ti.com?part=1