Re: Problem with PHY offloc for Cadence ethernet

[Jared McNeill <jmcneill%invisible.ca@localhost>]

Hi Lloyd --

I think the right thing to do here is to have the bus glue pass the PHY 
ID to cemac_attach_common. For Zynq you can get that information from the 
device tree -- see 
https://nxr.netbsd.org/xref/src/sys/arch/arm/sunxi/sunxi_emac.c#1342 for 
an example of how other drivers do this.

HTH,
Jared

On Fri, 22 Mar 2024, Lloyd Parkes wrote:

> Kia ora koutou,
> I got NetBSD booting on my ALINX AX7021B Zynq system the other day and
> I've just now managed to get the ethernet online.
>
> The problem turned out to be that the PHY appears at offset location 0,
> but the Cadence ethernet driver calls mii_attach with a fixed offset
> location of 1 (i.e. "mii_attach(sc->sc_dev, mii, 0xffffffff,
> MII_PHY_ANY, 1, 0);").
>
> I changed "1" to "MII_OFFSET_ANY" made my ethernet come online.
>
> I went through the Mercurial logs to see if I can find out why that
> code is that way and I found this log message.
>
>  $ hg log -r 1021807 -v if_cemac.c
>  changeset:   1021807:df4107b887fe
>  branch:      trunk
>  user:        jmcneill <jmcneill%NetBSD.org@localhost>
>  date:        Sat Nov 12 16:54:36 2022 +0000
>  files:       sys/dev/cadence/if_cemac.c
>  description:
>  Skip the first PHY found to remove the shadow PHY that appears at
>  the broadcast address 0.
>
> Well, that makes things more difficult for me. My board doesn't have
> any extra PHYs that need to be skipped.
>
> Is this something that should be fixable with a suitable devicetree
> specification? I already have to write the devicetree source and I
> don't mind having a go at improving the Cadence driver's support for
> the FDT bindings if there is anything lacking there. I really just
> don't know what's right and what's feasible. Any advice will be
> appreciated.
>
> Ngā mihi,
> Lloyd