Re: [PATCH] ARM: dts: aspeed: Add dts for ASUS Kommando IPMI card

Hi Andrew

On Mon, Mar 23, 2026 at 6:42 PM Andrew Jeffery
<[email protected]> wrote:
>
> Hi Anirudh,
>
> Sorry for the delay in reply to this thread.
>
> On Sat, 2026-03-21 at 17:24 -0500, Anirudh Srinivasan wrote:
> > Helo Andrew,
> >
> > After doing some more reading and experimentation, I think I might
> > have a clearer understanding of how things work.
> >
> > Andrew Lunn's suggestion in the patch series from Aspeed was to patch
> > u-boot to not insert clock delays anymore. And then use rgmii-id as
> > the phy mode in u-boot and linux for all new device trees.
> >
> > I tested this by setting the mac2 clk delays to 0 in the u-boot device
> > tree like this
> >
> > &scu {
> >     mac2-clk-delay = <0 0
> >               0 0
> >               0 0>;
> > };
> >
> > Setting the phy-mode to rgmii-id now gives us working networking in
> > u-boot and linux.
> >
> > So how do you suggest we proceed here?
> > 1. Add this DT to u-boot now, set phy mode to rgmii-id and set
> > mac2-clk-delay values to 0. Networking will work in u-boot and linux.
> > Update the DT to remove the clk delay values once Aspeed updates
> > u-boot
> > 2. Add this DT to u-boot now. set phy mode to rgmii-id. This means
> > networking won't work in u-boot, but it works once you're in linux
> > 3. Add this DT to u-boot now without networking. Add networking later
> > when Aspeed updates u-boot based on the discussion in that thread.
> > 3. Wait for aspeed to update u-boot, then add this DT to u-boot.
> >
> > In any of these cases, we can also now add networking to the linux DT
> > for this card (it's missing currently) with phy-mode set to rgmii-id.
> > No need to wait for Aspeed to update the linux driver (or I think
> > they're not gonna make any updates at all if it's just u-boot that
> > needs fixing).
>
> Working backwards here, yes, I think we should add the correct (but not
> necessarily yet functional) description to the linux devicetree
> (probably best to point out the lack of functionality in a comment in
> the relevant devicetree node).
>
> Then, do your first option 3 above, except add the description of
> option 1 in the commit message so people understand how to make
> networking function for both environments. That way it's apparent that
> anyone who has enabled networking in u-boot is also aware of the
> (future) problems of doing so, as they likely will have read that
> description. Please add links to the relevant discussions between
> Andrew Lunn and Jacky to the commit message as well.

Sounds good. I'll do things in this manner:
1. I'll send v2 of this u-boot patch without networking support, with
a description of how to enable networking in the commit message.
2. I'll send a kernel patch to enable networking with rgmii-id as the
phy mode. The commit message will mention that this requires patched
u-boot without the clock delays (with a link to the above u-boot patch
series).

I'll add links to the discussion between Andrew and Jacky in both.
Once aspeed updates u-boot, we can add networking in u-boot.

>
> Any concerns with that approach?
>
> >
> > Let me know what you think (or if I've gotten any info incorrect here).
>
> That all sounds about right to me.
>
> Andrew



-- 
Regards
Anirudh Srinivasan