Re: [PATCH v4 1/4] dt-bindings: net: phy: Add support for NXP TJA11xx

From: Florian Fainelli <f.fainelli@gmail.com>
To: Oleksij Rempel <o.rempel@pengutronix.de>,
	Mark Rutland <mark.rutland@arm.com>,
	devicetree@vger.kernel.org, Rob Herring <robh+dt@kernel.org>
Cc: Andrew Lunn <andrew@lunn.ch>, Marek Vasut <marex@denx.de>,
	netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
	Pengutronix Kernel Team <kernel@pengutronix.de>,
	David Jander <david@protonic.nl>,
	"David S. Miller" <davem@davemloft.net>,
	Heiner Kallweit <hkallweit1@gmail.com>
Subject: Re: [PATCH v4 1/4] dt-bindings: net: phy: Add support for NXP TJA11xx
Date: Tue, 17 Mar 2020 12:48:47 -0700	[thread overview]
Message-ID: <137a6dd3-c5ba-25b1-67ff-f0112afd7f34@gmail.com> (raw)
In-Reply-To: <20200317115626.4ncavxdcw4wu5zgc@pengutronix.de>


On 3/17/2020 4:56 AM, Oleksij Rempel wrote:
> On Fri, Mar 13, 2020 at 07:53:27PM +0100, Oleksij Rempel wrote:
>> On Fri, Mar 13, 2020 at 11:20:35AM -0700, Florian Fainelli wrote:
>>>
>>>
>>> On 3/13/2020 11:16 AM, Oleksij Rempel wrote:
>>>> On Fri, Mar 13, 2020 at 07:10:56PM +0100, Andrew Lunn wrote:
>>>>>>> diff --git a/Documentation/devicetree/bindings/net/nxp,tja11xx.yaml b/Documentation/devicetree/bindings/net/nxp,tja11xx.yaml
>>>>>>> new file mode 100644
>>>>>>> index 000000000000..42be0255512b
>>>>>>> --- /dev/null
>>>>>>> +++ b/Documentation/devicetree/bindings/net/nxp,tja11xx.yaml
>>>>>>> @@ -0,0 +1,61 @@
>>>>>>> +# SPDX-License-Identifier: GPL-2.0+
>>>>>>> +%YAML 1.2
>>>>>>> +---
>>>>>>> +$id: http://devicetree.org/schemas/net/nxp,tja11xx.yaml#
>>>>>>> +$schema: http://devicetree.org/meta-schemas/core.yaml#
>>>>>>> +
>>>>>>> +title: NXP TJA11xx PHY
>>>>>>> +
>>>>>>> +maintainers:
>>>>>>> +  - Andrew Lunn <andrew@lunn.ch>
>>>>>>> +  - Florian Fainelli <f.fainelli@gmail.com>
>>>>>>> +  - Heiner Kallweit <hkallweit1@gmail.com>
>>>>>>> +
>>>>>>> +description:
>>>>>>> +  Bindings for NXP TJA11xx automotive PHYs
>>>>>>> +
>>>>>>> +allOf:
>>>>>>> +  - $ref: ethernet-phy.yaml#
>>>>>>> +
>>>>>>> +patternProperties:
>>>>>>> +  "^ethernet-phy@[0-9a-f]+$":
>>>>>>> +    type: object
>>>>>>> +    description: |
>>>>>>> +      Some packages have multiple PHYs. Secondary PHY should be defines as
>>>>>>> +      subnode of the first (parent) PHY.
>>>>>>
>>>>>>
>>>>>> There are QSGMII PHYs which have 4 PHYs embedded and AFAICT they are
>>>>>> defined as 4 separate Ethernet PHY nodes and this would not be quite a
>>>>>> big stretch to represent them that way compared to how they are.
>>>>>>
>>>>>> I would recommend doing the same thing and not bend the MDIO framework
>>>>>> to support the registration of "nested" Ethernet PHY nodes.
>>>>>
>>>>> Hi Florian
>>>>>
>>>>> The issue here is the missing PHY ID in the secondary PHY. Because of
>>>>> that, the secondary does not probe in the normal way. We need the
>>>>> primary to be involved to some degree. It needs to register it. What
>>>>> i'm not so clear on is if it just needs to register it, or if these
>>>>> sub nodes are actually needed, given the current code.
>>>>
>>>> There are a bit more dependencies:
>>>> - PHY0 is responsible for health monitoring. If some thing wrong, it may
>>>>   shut down complete chip.
>>>> - We have shared reset. It make no sense to probe PHY1 before PHY0 with
>>>>   more controlling options will be probed
>>>> - It is possible bat dangerous to use PHY1 without PHY0.
>>>
>>> probing is a software problem though. If we want to describe the PHY
>>> package more correctly, we should be using a container node, something
>>> like this maybe:
>>>
>>> phy-package {
>>> 	compatible = "nxp,tja1102";
>>>
>>> 	ethernet-phy@4 {
>>> 		reg = <4>;
>>> 	};
>>>
>>> 	ethernet-phy@5 {
>>> 		reg = <5>;
>>> 	};
>>> };
>>
>> Yes, this is almost the same as it is currently done:
>>
>> phy-package {
>> 	reg = <4>;
>>  
>>  	ethernet-phy@5 {
>>  		reg = <5>;
>>  	};
>> };
>>
>> Because the primary PHY0 can be autodetected by the bus scan.
>> But I have nothing against your suggestions. Please, some one should say the
>> last word here, how exactly it should be implemented?

It's not for me to decide, I was hoping the Device Tree maintainers
could chime in, your current approach would certainly work although it
feels visually awkward.
-- 
Florian