[PATCH net-next v3 06/10] dt-bindings: net: Add Synopsys DW xPCS bindings

[Serge Semin]

On Fri, Jun 28, 2024 at 04:12:46PM -0600, Rob Herring wrote:
> On Thu, Jun 27, 2024 at 08:10:48PM +0300, Serge Semin wrote:
> > On Thu, Jun 27, 2024 at 04:51:22PM +0100, Conor Dooley wrote:
> > > On Thu, Jun 27, 2024 at 03:41:26AM +0300, Serge Semin wrote:
> > > > +  clocks:
> > > > +    description:
> > > > +      Both MCI and APB3 interfaces are supposed to be equipped with a clock
> > > > +      source connected via the clk_csr_i line.
> > > > +
> > > > +      PCS/PMA layer can be clocked by an internal reference clock source
> > > > +      (phyN_core_refclk) or by an externally connected (phyN_pad_refclk) clock
> > > > +      generator. Both clocks can be supplied at a time.
> > > > +    minItems: 1
> > > > +    maxItems: 3
> > > > +
> > > > +  clock-names:
> > > > +    oneOf:
> > > > +      - minItems: 1
> > > > +        items:
> > > > +          - enum: [core, pad]
> > > > +          - const: pad
> > > > +      - minItems: 1
> > > > +        items:
> > > > +          - const: pclk
> > > > +          - enum: [core, pad]
> > > > +          - const: pad
> > > 
> > 
> > > While reading this, I'm kinda struggling to map "clk_csr_i" to a clock
> > > name. Is that pclk? And why pclk if it is connected to "clk_csr_i"?
> > 
> > Right. It's "pclk". The reason of using the "pclk" name is that it has
> > turned to be a de-facto standard name in the DT-bindings for the
> > peripheral bus clock sources utilized for the CSR-space IO buses.
> > Moreover the STMMAC driver responsible for the parental DW *MAC
> > devices handling also has the "pclk" name utilized for the clk_csr_i
> > signal. So using the "pclk" name in the tightly coupled devices (MAC
> > and PCS) for the same signal seemed a good idea.
> 

> It is? That's really just the name of the bus clock for APB (Arm 
> Peripheral Bus). If there's a name that matches the docs, use that. 
> Though I'd drop 'clk_' part.

Yes, it's normally should have been utilized for APB, but as I see it
the name utilization has gone wider than to just the ARM Peripheral
bus clock. The DW MAC clock-names DT-property bindings is just one
example of that.

Anyway. Ok. I'll convert the name to "csr". (I'll drop the _i suffix
too since it's obvious that the clock signal is the connected to the
device input pin.)

-Serge(y)

> 
> Rob