Re: [PATCH 06/17] dt-bindings: net: mscc-miim: Add resets property

[Andrew Lunn <andrew@lunn.ch>]

On Tue, Apr 30, 2024 at 10:37:15AM +0200, Herve Codina wrote:
> Add the (optional) resets property.
> The mscc-miim device is impacted by the switch reset especially when the
> mscc-miim device is used as part of the LAN966x PCI device.
> 
> Signed-off-by: Herve Codina <herve.codina@bootlin.com>
> ---
>  Documentation/devicetree/bindings/net/mscc,miim.yaml | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git a/Documentation/devicetree/bindings/net/mscc,miim.yaml b/Documentation/devicetree/bindings/net/mscc,miim.yaml
> index 5b292e7c9e46..a8c92cec85a6 100644
> --- a/Documentation/devicetree/bindings/net/mscc,miim.yaml
> +++ b/Documentation/devicetree/bindings/net/mscc,miim.yaml
> @@ -38,6 +38,14 @@ properties:
>  
>    clock-frequency: true
>  
> +  resets:
> +    items:
> +      - description: Reset controller used for switch core reset (soft reset)

A follow up to the comment on the next patch. I think it should be
made clear in the patch and the binding, the aim is to reset the MDIO
bus master, not the switch. It just happens that the MDIO bus master
is within the domain of the switch core, and so the switch core reset
also resets the MDIO bus master.

Architecturally, this is important. I would not expect the MDIO driver
to be resetting the switch, the switch driver should be doing
that. But we have seen some odd Qualcomm patches where the MDIO driver
has been doing things outside the scope of MDIO, playing with resets
and clocks which are not directly related to the MDIO bus master. I
want to avoid any confusion here, especially when Qualcomm tries
again, and maybe points at this code.

	Andrew