Re: [PATCH 2/2] ARM: dts: BCM5301X: Set fixed-link for extra Netgear R8000 CPU ports

["Rafał Miłecki" <rafal@milecki.pl>]

On 2023-10-18 20:10, Florian Fainelli wrote:
> On 10/16/23 08:45, Andrew Lunn wrote:
>> On Mon, Oct 16, 2023 at 05:36:24PM +0200, Rafał Miłecki wrote:
>>> On 2023-10-14 18:50, Andrew Lunn wrote:
>>>> On Fri, Oct 13, 2023 at 12:33:14PM +0200, Rafał Miłecki wrote:
>>>>> From: Rafał Miłecki <rafal@milecki.pl>
>>>>> 
>>>>> While switch ports 5 and 7 are disabled (vendor designed port 8 to 
>>>>> be
>>>>> used for CPU traffic) they could be used strictly technically. For
>>>>> some
>>>>> reason however both those ports need forcing link to be usable.
>>>> 
>>>> This explanation is not making much sense to me.
>>>> 
>>>> I assume this board does not have an RJ45 for these two ports? But
>>>> does it have a header so you can access the MII interface?
>>> 
>>> This PATCH as it is requires a basic familiarity with Northstar 
>>> platform
>>> or checking bcm-ns.dtsi.
>>> 
>>> All Northstar (BCM5301X) devices have 3 Ethernet interfaces. 99% of 
>>> them
>>> have:
>>> 1. gmac0 connected to port 5
>>> 2. gmac1 connected to port 7
>>> 3. gmac2 connected to port 8
>>> (it's described in bcm-ns.dtsi).
>>> 
>>> 
>>> Some vendors decide to use gmac0 and switch port 5. They fill NVRAM 
>>> with
>>> MAC for gmac0.
>>> 
>>> Some vendors decide to use gmac2 & port 8. They set MAC for gmac2.
>>> 
>>> 
>>> Netgear decided to use gmac2 & port 8 for R8000. They fill NVRAM with
>>> MAC for gmac2.
>>> 
>>> If you however insist on using gmac0 you could do that. That just
>>> requires setting up gmac0 with a custom/random MAC and forcing link 
>>> for
>>> switch ports as described in this PATCH.
>> 
>> If the ports are not used, you have them set to disabled, why do they
>> need a fixed-link? That is what i don't understand yet.
> 
> It seems to me like the commit message could be reworded such that:
> 
> Even though ports 5 and 7 are disabled and the system is intended to
> use port 8, make it possible for users to experiment with using ports
> 5 and/or 7 if they desire so by ensuring that they have the necessary
> 'fixed-link' properties to describe the internal connection within the
> SoC between the switch ports and the two Ethernet controllers.
> 
> Rafal, does that capture the intent? If so I can amend the commit
> message while applying.

I believe so. You're correct that in practice it's for experimenting
mainly. Formally it also describes hardware which DT is for.

I'm sure we can find a lot of *disabled* hardware blocks in Linux's DTS
files that are still described for the sake of documenting it.

-- 
Rafał Miłecki