Re: [PATCH v2 3/3] arm64: dts: rockchip: Add Luckfox Omni3576 Board support

["Diederik de Haas" <didi.debian@cknow.org>]

[-- Attachment #1: Type: text/plain, Size: 2652 bytes --]

Hi,

On Mon May 5, 2025 at 2:52 AM CEST, John Clark wrote:
> On 5/4/25 8:45 PM, Andrew Lunn wrote:
>>>> What PHY is it? Are you using the correct PHY driver for it, or
>>>> genphy?
>>>>
>>> MAE0621A-Q3C
>>> http://www.maxio-tech.com/product/12928/12929/12930/12931.html
>> 
>> Mainline does not have a PHY driver for this. So nothing is
>> controlling the delays in the PHY. So what you have above works by
>> luck, and is likely to break once there is a PHY driver. So i suggest
>> you drop the Ethernet nodes for the moment.
>> 
> The chip claims to be a pin-for-pin clone of the rtl8211f. Empirical 
> testing has demonstrated it to be extremely stable. Without networking 
> IO the board is very difficult to develop against. I can disable 
> networking if that is the consensus.
>
>> There does appear to be a PHY driver here:
>> 
>> https://github.com/CoreELEC/linux-amlogic/blob/5.15.153_202501/drivers/net/phy/maxio.c
>> 
>> but it has a number of things wrong with it. You might want to search
>> around and see if there are any cleaner versions around, or if anybody
>> is working on upstreaming a driver for this PHY.

It may be nothing, but to me this is getting too much of a coincidence.

I have a NanoPi R5S and I want(ed) to look at its gmac node as well,
because it (also) has some deprecated properties ... and
``phy-mode = "rgmii"`` which was done deliberately in commit
31425b1fadb2 ("arm64: dts: rockchip: fix gmac support for NanoPi R5S")
and the change was away from "rgmii-id" ... and it has a RTL8211F-CG.
Unfortunately the commit message doesn't say *why* it changed.

I also have a Quartz64-B, which also has a RTL8211F-CG and it has
``phy-mode = "rgmii"``, changed from "rgmii-id" in commit
16bc4d196b2a ("arm64: dts: rockchip: Fix ethernet on production Quartz64-B")

And maybe related, but a bit different is the Quartz64-A which I also
have, which also has ``phy-mode = "rgmii"`` which I'm reasonably sure
was done deliberately (by Peter Geis, who also made the previously
mentioned commit for Q64-B). Q64-A's PHY is YT-8511C 'though'.

And then there's an 'issue' with Q64-A and Q64-B when U-Boot enabled the
ethernet driver causing massive packet loss: 
https://lore.kernel.org/u-boot/2086393.9F9pDXStbY@bagend/
"configs: rockchip: Enable ethernet driver on RK356x boards"
(packet loss on Q64-B was significantly higher then on Q64-A)

As I said, it could be a coincidence and if it is, just tell me and I'll
stay out of this thread going further.
In any case, thanks for "Add informative text about RGMII delays", that
is quite useful :-)

Cheers,
  Diederik

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]