Re: [PATCH v3 5/6] arm64: dts: qcom: sdm845: Add OPP table support to UFSHC

[Manivannan Sadhasivam]

On Wed, Sep 27, 2023 at 04:45:46PM +0530, Viresh Kumar wrote:
> On 12-09-23, 12:29, Manivannan Sadhasivam wrote:
> > On Mon, Sep 11, 2023 at 04:15:10PM +0300, Dmitry Baryshkov wrote:
> > > I'd say, I'm still slightly unhappy about the 0 clock rates here.
> > 
> > Neither do I. But it is the only viable option I could found.
> > 
> > > We need only three clocks here: core, core_clk_unipro and optional
> > > ice_core_clk. Can we modify ufshcd_parse_operating_points() to pass only
> > > these two or three clock names to devm_pm_opp_set_config() ? The OPP core
> > > doesn't need to know about all the rest of the clocks.
> > > 
> > 
> > We need to enable/disable all of the clocks, but only need to control the rate
> > for these 3 clocks. So we cannot just use 3 clocks.
> > 
> > If the OPP table has only 3 entries (omitting the gate-only clocks), then we
> > need some hack in the driver to match the rates against the clock entries. Doing
> > so will result in hardcoding the clock info in the driver which I do not want to
> > do.
> > 
> > If we have something like "opp-hz-names" to relate the rates to clock-names, it
> > might do the job. But it needs some input from Viresh.
> 
> I have already given an option earlier about this [1]. You can change the order
> of clks in the "clock-names" field, so that the first three are the one with
> valid frequencies. You shouldn't need much of the hacks after that I guess.
> 

I do not remember all the details on the top of my head, but I did investigate on
that and concluded that it won't fit.

- Mani

> Or maybe I missed something else now, talked about this a long time ago :)
> 
> -- 
> viresh
> [1] https://lore.kernel.org/all/20230713040918.jnf5oqiwymrdnrmq@vireshk-i7/

-- 
மணிவண்ணன் சதாசிவம்