Re: [PATCH net-next 01/12] dt-bindings: ptp: add bindings for NETC Timer

[Krzysztof Kozlowski <krzk@kernel.org>]

On 14/07/2025 12:09, Krzysztof Kozlowski wrote:
> On 14/07/2025 11:56, Wei Fang wrote:
>>
>>>
>>> How does the other consumer - ethernet - reference this one here? Paste
>>> complete DTS of this and users, otherwise it is just ping-pong
>>> discussion where you put just a little effort to bounce back my question.
>>
>> Below is the DTS node of enetc (ethernet device) and timer node.
>>
>> enetc_port0: ethernet@0,0 {
>> 	compatible = "pci1131,e101";
>> 	reg = <0x000000 0 0 0 0>;
>> 	pinctrl-names = "default";
>> 	pinctrl-0 = <&pinctrl_enetc0>;
>> 	phy-handle = <&ethphy0>;
>> 	phy-mode = "rgmii-id";
>> 	status = "okay";
> 
> How do you use netc_timer in such case?
> 
>> };
>>
>> netc_timer: ethernet@18,0 {
>> 	compatible = "pci1131,ee02";
>> 	reg = <0x00c000 0 0 0 0>;
>> 	clocks = <&netc_system333m>;
>> 	clock-names = "system";
>> };
>>
>> Currently, the enetc driver uses the PCIe device number and function number
>> of the Timer to obtain the Timer device, so there is no related binding in DTS.
> 
> So you just tightly coupled these devices. Looks poor design for me, but
> your choice. Anyway, then use that channel as information to pass the
> pin/timer/channel number. You do not get a new property for that.
> 
>> In the future, we plan to add phandle to the enetc document to bind enetc
>> and Timer, because there will be multiple Timer instances on subsequent
>> platforms.
> 
> Bindings must be complete, not "in the future" but now. Start sending
> complete work, so we won't have to guess it.

BTW, above DTS is not a patch. Post complete upstream DTS as patch for
review, evaluation and merging. Otherwise you will be posting
incomplete, unvalidated code which in fact does not work and shortly
after you will be bringing fix-up patches for bindings.

No. This is not how the process is working for upstreaming SoC
components from major vendor.

Best regards,
Krzysztof