Re: [PATCH v2 1/3] media: mt9m111: add V4L2_CID_LINK_FREQ support

[Laurent Pinchart]

On Tue, Sep 20, 2022 at 11:43:37AM +0200, Jacopo Mondi wrote:
> On Fri, Sep 16, 2022 at 03:57:11PM +0200, Marco Felsch wrote:
> > Add support to report the link frequency.
> >
> > Signed-off-by: Marco Felsch <m.felsch@xxxxxxxxxxxxxx>
> > ---
> > The v1 of this small series can be found here:
> > https://lore.kernel.org/all/20220818144712.997477-1-m.felsch@xxxxxxxxxxxxxx/
> >
> > Thanks a lot to Jacopo for the review feedback on my v1.
> >
> > Changelog:
> >
> > v2:
> > - use V4L2_CID_LINK_FREQ instead of V4L2_CID_PIXEL_RATE
> > ---
> >  drivers/media/i2c/mt9m111.c | 21 ++++++++++++++++++++-
> >  1 file changed, 20 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/media/i2c/mt9m111.c b/drivers/media/i2c/mt9m111.c
> > index afc86efa9e3e..52be1c310455 100644
> > --- a/drivers/media/i2c/mt9m111.c
> > +++ b/drivers/media/i2c/mt9m111.c
> > @@ -1249,6 +1249,8 @@ static int mt9m111_probe(struct i2c_client *client)
> >  {
> >  	struct mt9m111 *mt9m111;
> >  	struct i2c_adapter *adapter = client->adapter;
> > +	static s64 extclk_rate;
> 
> Why static ?

I missed that one indeed. I assume it's static because the pointer is
stored in the v4l2_ctrl structure by v4l2_ctrl_new_int_menu(), but
that's wrong. The data should be in the mt9m111 structure instead,
otherwise it won't work right when using multiple sensors.

> Also clk_get_rate() returns an unsigned long

v4l2_ctrl_new_int_menu() requires an s64 pointer.

> > +	struct v4l2_ctrl *ctrl;
> >  	int ret;
> >
> >  	if (!i2c_check_functionality(adapter, I2C_FUNC_SMBUS_WORD_DATA)) {
> > @@ -1271,6 +1273,13 @@ static int mt9m111_probe(struct i2c_client *client)
> >  	if (IS_ERR(mt9m111->clk))
> >  		return PTR_ERR(mt9m111->clk);
> >
> > +	ret = clk_prepare_enable(mt9m111->clk);
> > +	if (ret < 0)
> > +		return ret;
> > +
> > +	extclk_rate = clk_get_rate(mt9m111->clk);
> > +	clk_disable_unprepare(mt9m111->clk);
> > +
> >  	mt9m111->regulator = devm_regulator_get(&client->dev, "vdd");
> >  	if (IS_ERR(mt9m111->regulator)) {
> >  		dev_err(&client->dev, "regulator not found: %ld\n",
> > @@ -1285,7 +1294,7 @@ static int mt9m111_probe(struct i2c_client *client)
> >  	mt9m111->subdev.flags |= V4L2_SUBDEV_FL_HAS_DEVNODE |
> >  				 V4L2_SUBDEV_FL_HAS_EVENTS;
> >
> > -	v4l2_ctrl_handler_init(&mt9m111->hdl, 7);
> > +	v4l2_ctrl_handler_init(&mt9m111->hdl, 8);
> >  	v4l2_ctrl_new_std(&mt9m111->hdl, &mt9m111_ctrl_ops,
> >  			V4L2_CID_VFLIP, 0, 1, 1, 0);
> >  	v4l2_ctrl_new_std(&mt9m111->hdl, &mt9m111_ctrl_ops,
> > @@ -1309,6 +1318,16 @@ static int mt9m111_probe(struct i2c_client *client)
> >  				BIT(V4L2_COLORFX_NEGATIVE) |
> >  				BIT(V4L2_COLORFX_SOLARIZATION)),
> >  			V4L2_COLORFX_NONE);
> 
> Empty line maybe ?
> 
> > +	/*
> > +	 * The extclk rate equals the link freq. if reg default values are used,
> > +	 * which is the case. This must be adapted as soon as we don't use the
> > +	 * default values anymore.
> > +	 */
> > +	ctrl = v4l2_ctrl_new_int_menu(&mt9m111->hdl, &mt9m111_ctrl_ops,
> > +				      V4L2_CID_LINK_FREQ, 0, 0, &extclk_rate);
> > +	if (ctrl)
> > +		ctrl->flags |= V4L2_CTRL_FLAG_READ_ONLY;
> > +
> 
> I'm sorry I have not replied to your previous email about using
> LINK_FREQ for parallel busses.. I see it mentioned in ext-ctrls-image-process.rst
> as you said:
> 
> ``V4L2_CID_LINK_FREQ (integer menu)``
>     The frequency of the data bus (e.g. parallel or CSI-2).
> 
> I still have a bit of troubles seeing it apply nicely on a parallel
> bus. Isn't PIXEL_RATE more appropriate ?

They are different. When transmitting YUYV_2X8 for instance, the link
frequency is twice the pixel clock rate.

> You said you need to know the
> overall bus bandwidth in bytes , and pixel_rate * bpp / 8 is equally
> valid and easy as link_freq / num_lanes, which requires the receiver
> to fetch the remote subdev media bus configuration instead of relying
> on the input format. Also LINK_FREQ is a menu control, something nasty
> already for CSI-2 busses, which requires to pre-calculate the link
> freqs based on the input mclk. It is also meant to be changed by
> userspace, while PIXEL_RATE is RO by default.
> 
> Sakari, Laurent, what's your take here ?

Ideally both should be implemented by the driver.

> >  	mt9m111->subdev.ctrl_handler = &mt9m111->hdl;
> >  	if (mt9m111->hdl.error) {
> >  		ret = mt9m111->hdl.error;

-- 
Regards,

Laurent Pinchart