Re: [PATCH v5 2/6] media: imx335: Parse fwnode properties

[Umang Jain]

Hi Sakari,

On 12/04/24 6:33 pm, Sakari Ailus wrote:
> Hi Umang,
>
> On Fri, Apr 12, 2024 at 05:58:38PM +0530, Umang Jain wrote:
> > From: Kieran Bingham <kieran.bingham@xxxxxxxxxxxxxxxx>
> >
> > Call the V4L2 fwnode device parser to handle controls that are
> > standardised by the framework.
> >
> > Signed-off-by: Kieran Bingham <kieran.bingham@xxxxxxxxxxxxxxxx>
> > Signed-off-by: Umang Jain <umang.jain@xxxxxxxxxxxxxxxx>
> > ---
> >   drivers/media/i2c/imx335.c | 16 ++++++++++++----
> >   1 file changed, 12 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/media/i2c/imx335.c b/drivers/media/i2c/imx335.c
> > index c633ea1380e7..3ea9c0ebe278 100644
> > --- a/drivers/media/i2c/imx335.c
> > +++ b/drivers/media/i2c/imx335.c
> > @@ -1227,10 +1227,12 @@ static int imx335_init_controls(struct imx335 *imx335)
> >   {
> >   	struct v4l2_ctrl_handler *ctrl_hdlr = &imx335->ctrl_handler;
> >   	const struct imx335_mode *mode = imx335->cur_mode;
> > +	struct v4l2_fwnode_device_properties props;
> >   	u32 lpfr;
> >   	int ret;
> >   
> > -	ret = v4l2_ctrl_handler_init(ctrl_hdlr, 7);
> > +	/* v4l2_fwnode_device_properties can add two more controls */
> > +	ret = v4l2_ctrl_handler_init(ctrl_hdlr, 9);
> >   	if (ret)
> >   		return ret;
> >   
> > @@ -1295,9 +1297,15 @@ static int imx335_init_controls(struct imx335 *imx335)
> >   	if (imx335->hblank_ctrl)
> >   		imx335->hblank_ctrl->flags |= V4L2_CTRL_FLAG_READ_ONLY;
> >   
> > -	if (ctrl_hdlr->error) {
> > -		dev_err(imx335->dev, "control init failed: %d\n",
> > -			ctrl_hdlr->error);
> > +	ret = v4l2_fwnode_device_parse(imx335->dev, &props);
> > +	if (!ret) {
> > +		/* Failure sets ctrl_hdlr->error, which we check afterwards anyway */
> > +		v4l2_ctrl_new_fwnode_properties(ctrl_hdlr, &imx335_ctrl_ops,
> > +						&props);
> > +	}
> > +
> > +	if (ctrl_hdlr->error || ret) {
> > +		dev_err(imx335->dev, "control init failed: %d\n", ctrl_hdlr->error);
> Too long line.
>
> >   		v4l2_ctrl_handler_free(ctrl_hdlr);
> >   		return ctrl_hdlr->error;
> The handler may not be in error state if only v4l2_fwnode_device_parse()
> failed.

I read some more drivers and it seems v4l2_fwnode_device_parse() can 
probably be checked at start of init_controls() and return early on 
non-zero return value.

Is that something that should be done here as well ?
> Should that be something that should prevent probing a driver though, or
> could it just be ignored? I.e. in that case I'd only check for handler's
> error, not ret.

So we should probably check for ret and ctrl_hdlr->error separately ?
> >   	}