Re: [net-next: PATCH v3 6/8] net: core: switch to fwnode_find_net_device_by_node()

[Marcin Wojtas]

czw., 28 lip 2022 o 22:18 Andrew Lunn <andrew@xxxxxxx> napisał(a):
>
> > The 'label' thing is actually one of the things that I'm seriously
> > considering skipping parsing if this is an ACPI system, simply because
> > best practices are different today than they were when the OF bindings
> > were created.
>
> Agreed. We want the ACPI binding to learn from what has worked and not
> worked in DT. We should clean up some of the historical mess. And
> enforce things we don't in DT simply because there is too much
> history.
>
> So a straight one to one conversion is not going to happen.

I understand your standpoint - there is a long history, possible
clean-ups, backward compatibility considerations, etc. that should not
be zero-day baggage of ACPI. Otoh, we don't need to be worried about
the ACPI binding too much now - as agreed it was removed from this
series, beginning from v2. IMO it may be better to return to that once
the ACPI Spec is updated with the MDIOSerialBus and the patches are
resubmitted on whatever shape of the DSA subsystem is established
within the next weeks/months from now.

In v1 we discussed also the resubmission of the non-ACPI-related
patches, which would pave the way to dropping the explicit OF_
dependency in the DSA and moving to a generic hardware description
kernel API - without any functional change. Modifying DT bindings and
clean-ups could be done on top this patchset as well. Of course, it is
the subsystems' Maintainers call and I'll adjust accordingly - if you
wish me to wait and rebase after the 'validation patch' lands in
net-next, I'll do that.

A side note: I was of course aware that making it for the v5.20 would
be extremely hard, but I decided to give it a try anyway - I had to
wait for some time, as this series was gated by fate of the eventually
abandoned phylink-related changes.

Thanks,
Marcin