Re: [RFC net-next 1/2] dt-bindings: net: dsa: add bindings for GSW Series switches

[Vladimir Oltean]

On Thu, Oct 27, 2022 at 12:08:06PM -0400, Krzysztof Kozlowski wrote:
> On 27/10/2022 09:57, Vladimir Oltean wrote:
> > Hi Camel,
> > 
> > On Thu, Oct 27, 2022 at 08:46:27AM -0400, Krzysztof Kozlowski wrote:
> >>>  >> +      - enum:
> >>>  >> +          - mxl,gsw145-mdio
> >>>  >
> >>>  > Why "mdio" suffix?
> >>>
> >>> Inspired by others dsa chips.
> >>> lan9303.txt:  - "smsc,lan9303-mdio" for mdio managed mode
> >>> lantiq-gswip.txt:- compatible   : "lantiq,xrx200-mdio" for the MDIO bus
> >>> inside the GSWIP
> >>> nxp,sja1105.yaml:                  - nxp,sja1110-base-t1-mdio
> >>
> >> As I replied to Andrew, this is discouraged.
> > 
> > Let's compare apples to apples, shall we?
> > "nxp,sja1110-base-t1-mdio" is the 100Base-T1 MDIO controller of the
> > NXP SJA1110 switch, hence the name. It is not a SJA1110 switch connected
> > over MDIO.
> 
> Thanks for clarifying. Then this could be fine. Let me then explain what
> is discouraged:
> 1. Adding bus suffixes to the compatible, so for example foo,bar LED
> controller is on I2C bus, so you call it "foo,bar-i2c".
> 
> 2. Adding device types to the compatible, if this is the only
> function/variant of the device, so for example calling foo,bar LED
> controller "foo,bar-led". This makes sense in case of multi functional
> devices (PMICs, SoCs), but not standalone ones.
> 
> So what do we have here? Is it one of the cases above?

We are in agreement about SJA1110 (it's in case 2), this is what I said
about Camel's comparison not being apples to apples. "mxl,gsw145-mdio"
is in case 1, and I've also been recommending people to not add such
suffixes to compatible strings (also see the discussion with Colin
Foster about "mscc,vsc7512-switch" vs "mscc,vsc7512-ext-switch" to
denote an "external" switch which is otherwise the exact same hw but on
a different bus).