Re: NetBSD 11.0 on Netgate SG-1000 AM335x cpsw, KSZ9031 and i2c notes

[Try The Post Office <trythepostoffice%gmail.com@localhost>]

Hi Yuri,

Thanks for going through all of it.

> Before committing the DT, I'd prefer if you could fix the pinmux
> configuration rather than just dropping it.

Okay. ufw.dts has the RGMII pinmux for both ports so I'll start from
that and check it against what U-Boot leaves in the control module so
the kernel and U-Boot set the pins the same way.

> However, I'm wondering if disabling hardware is the right approach.
> You aren't building a DT for a BBB, so you can just strip BBB
> hardware not present on your board out.

Makes sense. The patched BBB tree was just to get it booting. The real
one will drop the PMIC, HDMI, LCDC, audio, cape EEPROMs and the rest
instead of disabling them.

> It is normal that disabled nodes get enumerated. They shouldn't
> attach though.

I think there's still a bug then. With only status "disabled" on
tps@24 and the compatible left as "ti,tps65217" tps65217pmic0 did
attach and kept logging "cannot execute operation", that's
logs/i2c-tps65217.txt in the repo. I'll file the PR as the disabled
node attaching not as it being enumerated. I'll capture a clean dmesg
of that case with the attach line for the PR.

> In addition to sys/external/gpl2/dts, there is also sys/arch/arm/dts.
> This is where NetBSD-only patches to the DTs go.

Thanks, I'll put the SG-1000 tree there following your commit and send
a patch once it boots with the pinmux fixed and the BBB stuff stripped.
For the new file, is am335x-netgate-sg1000.dts with compatible
"netgate,sg1000" okay, or is there a naming convention you'd rather
I follow?

On the driver side I'll probably start with cpsw_mii_statchg() and
MACCONTROL since that one is small and I can test it on the board.

Elias Howell