Re: [PATCH] PM / AVS: rockchip-io: add GRF and PMUGRF types to distinguish

[David.Wu]

Hi Heiko,

å 2016/2/2 5:17, Heiko StÃbner åé:
> Hi David,
>
> Am Montag, 1. Februar 2016, 16:54:38 schrieb David.Wu:
> > å 2016/1/30 20:39, Heiko Stuebner åé:
> > > Am Samstag, 30. Januar 2016, 20:01:45 schrieb David Wu:
> > > > As rk3368 contained two separated iodomain areas, this was
> > > > determined to use which regmap base address.
> > > >
> > > > Signed-off-by: David Wu <david.wu@xxxxxxxxxxxxxx>
> > > I don't think we need to specify this on a driver level. Both GRF areas
> > > are
> > > "General register files" only located in two separate power-domains.
> > > So the rockchip,grf property should work for both. Especially as nothing
> > > keeps designers from introducing yet another GRF-area somewhere else ;-)
> > >
> > > >From when I started working on the rk3368, I still have a preliminary
> > >
> > > patches for that sitting here, so I've attached on how I envisoned that to
> > > work.
> > Okay, i agree to you, but it make someone a little confused just from
> > the drive code,
> > not DT file,  about  pmugrf regmap base address.:-)
> >
> > How do you feel about intergating GRF and PMU drivers in one driver?
> > Thanks!
> I will very strongly disagree here ;-) .
> Similar to the power-domains being part of the pmu, the io-domains are
> part of their individual GRFs. So if you want it really clean and tidy the way
> to go foward will be the attached patches. Compile-tested only.

Thanks for your reply, the patchs look better than mine.
I have tested them on sdk board and i found something may be wrong.
"parent->of_node" instead of "parent", as the parent is not null if 
parent-node not used.
        if (parent->of_node) {
            iod->grf = syscon_node_to_regmap(parent->of_node);
        } else {
            dev_dbg(&pdev->dev, "falling back to old binding\n");
            iod->grf = syscon_regmap_lookup_by_phandle(np, "rockchip,grf");
        }

> Other things like the usbphy control should move there as well in the long
> run. But that's not immediate necessary.
>
>
> Heiko