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

[Jacopo Mondi]

Hello

On Tue, Sep 20, 2022 at 10:56:17AM +0200, Marco Felsch wrote:
> Hi Sakari,
>
> On 22-09-19, Sakari Ailus wrote:
>
> ...
>
> > > > > +	ret = clk_prepare_enable(mt9m111->clk);
> > > > > +	if (ret < 0)
> > > > > +		return ret;
> > > > > +
> > > > > +	extclk_rate = clk_get_rate(mt9m111->clk);
> > > > > +	clk_disable_unprepare(mt9m111->clk);
> > > >
> > > > I don't think you'll need to enable a clock to just get its frequency.
> > >
> > > The official API states that you need to turn on the clk before
> > > requesting it and it makes sense. Also there is a new helper
> > > devm_clk_get_enabled() which addresses simple clk usage since most of
> > > drivers don't enable it before requesting the rate.

Had the same question on v1 and Marco pointed me to the clk_get_rate()
documentation
https://elixir.bootlin.com/linux/v6.0-rc1/source/include/linux/clk.h#L682

which indeed specifies
"This is only valid once the clock source has been enabled."

However none (or very few) of the linux-media i2c drivers actually do
that.

I have added in cc the clk framework maintainer to see if he can help
shed some light on this


> >
> > I guess the rate could change in the meantime, unless exclusive access is
> > requested.
>
> Not only that, there are a bunch of clk provider hw around which may
> need to turned on first. Anyway, I really don't care on this topic. As
> I said I wanted to fullfil the API and if drop clk_prepare_enable() I
> don't. So if this okay for you I will go that way.
>
> > The clock framework currently doesn't offer a way to set the assigned
> > rate and prevent changing it. But above, couldn't the clock frequency
> > be changed again once the clock has been disabled?
>
> Yes it could.
>
> Regards,
>   Marco