[PATCH v2 1/6] riscv: dts: sophgo: Put sdhci compatible in dt of specific SoC

[Inochi Amaoto]

On Mon, Jun 17, 2024 at 10:57:54AM GMT, Samuel Holland wrote:
> Hi Jisheng, Thomas,
> 
> On 2024-06-17 10:40 AM, Conor Dooley wrote:
> > On Mon, Jun 17, 2024 at 09:16:43PM +0800, Jisheng Zhang wrote:
> >> On Mon, Jun 17, 2024 at 11:16:32AM +0200, Thomas Bonnefille wrote:
> >>> On 6/17/24 1:58 AM, Yixun Lan wrote:
> >>>> On 18:47 Wed 12 Jun     , Inochi Amaoto wrote:
> > 
> >>>>> Is this change necessary? IIRC, the sdhci is the same across
> >>>>> the whole series.
> > 
> >> sorry for being late, I was busy in the past 2.5 month. Per my
> >> understanding, the sdhci in cv1800b is the same as the one in
> >> sg200x. Maybe I'm wrong, but this was my impression when I cooked
> >> the sdhci driver patch for these SoCs.
> >>
> >>>> I tend to agree with Inochi here, if it's same across all SoC, then no bother to
> >>>> split, it will cause more trouble to maintain..
> >>>>
> >>>
> >>> To be honest, I agree with this to, but as a specific compatible for the
> >>> SG2002 was created in commit 849e81817b9b, I thought that the best practice
> >>> was to use it.
> >>
> >> I'd like to take this chance to query DT maintainers: FWICT, in the past
> >> even if the PLIC is the same between SoCs, adding a new compatible for
> >> them seems a must. So when time goes on, the compatbile list would be
> >> longer and longer, is it really necessary? Can we just use the existing
> >> compatible string?
> >> DT maintainers may answered the query in the past, if so, sorry for
> >> querying again.
> > 
> > For new integrations of an IP, yes, new specific compatibles please. New
> > integrations may have different bugs etc, even if the IP itself is the
> > same. If there's different SoCs that are the same die, but with elements
> > fused off, then sure, use the same compatible.
> > 
> > I expect the list of compatibles in the binding to grow rather large, but
> > that is fine. No one SoC is going to do anything other than something like
> > compatible = "renesas,$soc-plic", "andestech,corecomplex-plic", "riscv,plic";
> > which I think is perfectly fine.
> 
> And you can do the same thing here for the SDHCI controller: if you think sg200x
> has the same controller (and integration! e.g. number of clocks/resets) as
> cv1800b, then you should keep sophgo,cv1800b-dwcmshc as a fallback compatible
> string. Then the driver doesn't need any changes until/unless you eventually
> find some reason they are not compatible.
> 
> It's better to have a SoC-specific compatible string in the DT and not need it,
> than find out later you need one and not have it. :)
> 
> Regards,
> Samuel
> 

This is excellect and reasonable. I will take your advice for the DT
change. With your suggetion, I think it may be acceptable to mark the
low-profile SoC as the default value and let other override it.

Let take the clk as the example:

// in the base file cv18xx.dtsi
clk: clock-controller at 3002000 {
	// set the "sophgo,cv1800-clk" as the fallback.
	compatible = "sophgo,cv1800-clk";
	[...]
};

// in the cv1800b.dtsi
// now no need for the clk override since we have the compatible
// default.

// in the cv1812h.dtsi
// we still need this override as it is not compatible
&clk {
	compatible = "sophgo,cv1810-clk";
};

// in the sg2002.dtsi of the patch
// we need this override as it is also not compatible
&clk {
	compatible = "sophgo,sg2000-clk";
};

Do I understand correctly? Or we still need to duplicate these node
for SoCs with different profiles.

Regards,
Inochi