[PATCH v3 2/3] dt-bindings: i2c-ast2600: Add bindings for AST2600 i2C driver

[Andrew Jeffery]

On Tue, 2 Aug 2022, at 18:34, Ryan Chen wrote:
> Hello,
>
>> -----Original Message-----
>> From: Andrew Jeffery <andrew at aj.id.au>
>> Sent: Friday, July 29, 2022 11:13 AM
>> To: Ryan Chen <ryan_chen at aspeedtech.com>; Joel Stanley <joel at jms.id.au>;
>> Philipp Zabel <p.zabel at pengutronix.de>; linux-arm-kernel at lists.infradead.org;
>> linux-aspeed at lists.ozlabs.org; linux-kernel at vger.kernel.org;
>> openbmc at lists.ozlabs.org
>> Cc: BMC-SW <BMC-SW at aspeedtech.com>
>> Subject: Re: [PATCH v3 2/3] dt-bindings: i2c-ast2600: Add bindings for AST2600
>> i2C driver
>> 
>> 
>> 
>> On Fri, 29 Jul 2022, at 12:33, Ryan Chen wrote:
>> > Hello Andrew,
>> >
>> >> -----Original Message-----
>> >> From: Andrew Jeffery <andrew at aj.id.au>
>> >> Sent: Friday, July 29, 2022 10:29 AM
>> >> To: Ryan Chen <ryan_chen at aspeedtech.com>; Joel Stanley
>> >> <joel at jms.id.au>; Philipp Zabel <p.zabel at pengutronix.de>;
>> >> linux-arm-kernel at lists.infradead.org;
>> >> linux-aspeed at lists.ozlabs.org; linux-kernel at vger.kernel.org;
>> >> openbmc at lists.ozlabs.org
>> >> Cc: BMC-SW <BMC-SW at aspeedtech.com>
>> >> Subject: Re: [PATCH v3 2/3] dt-bindings: i2c-ast2600: Add bindings
>> >> for AST2600 i2C driver
>> >>
>> >> Hi Ryan,
>> >>
>> >> On Mon, 16 May 2022, at 16:18, ryan_chen wrote:
>> >> > +    i2c0: i2c-bus at 80 {
>> >> > +      #address-cells = <1>;
>> >> > +      #size-cells = <0>;
>> >> > +      #interrupt-cells = <1>;
>> >> > +      compatible = "aspeed,ast2600-i2c-bus";
>> >>
>> >> This isn't quite right with respect to your binding description above
>> >> :)
>> > Yes, the compatible need to be " aspeed,ast2600-i2c" is that your point ?
>> 
>> Yes, but only if we agree that we should have different compatibles for the
>> different drivers. I'm not convinced about that yet.
>> 
>> I think it's enough to have different Kconfig symbols, and select the old driver
>> in aspeed_g4_defconfig, and the new driver in aspeed_g5_defconfig. Won't
>> that gives us the right outcome without requiring a new set of compatibles?
>> 
> The new driver in aspeed_g5_defconfig.

Right, behind a new Kconfig option.

> And different compatible string 
> claim will
> Load the new or legacy driver,

I don't think we need this. It's enough to enable the new driver in the 
defconfig (and possibly disable the config option for the old driver).

> it should ok like the different 
> generation SOC. Have 
> different design.
> Am I right?

We have SoC-specific compatibles already, so the new driver can just 
bind on the compatibles for the SoC revisions that have the new 
register interface. The old driver just binds to the device in the SoCs 
that have the old register interface.

There's an overlap in support between the two drivers, but for people 
who care about which implementation they use they can choose to exclude 
that driver from their kernel config.

None of this requires more compatibles be added.

Does that help?

Andrew