On Tue, Nov 24, 2020 at 04:21:48PM +0100, Linus Walleij wrote:

> > What people think they were sold was the idea that they shouldn't have
> > to write driver code or upstream things, something with more AML like
> > capabilities (not realising that AML works partly because ACPI hugely
> > constrains system design).

> This makes a lot of sense.

> I suppose what we need to think about is the bigger question of why
> people/companies/managers are so worried about working upstream
> that they will go to lengths to avoid it and jump at any chance of
> raising a wall of abstraction between their internal development and
> the in-kernel software development.

> I think of this as vendor/community couples therapy or something,
> there is some form of deep disconnect or mistrust going on at times
> and having worked on both ends myself I would think I could
> understand it but I can't.

In this case I think this is partly due to the way people were sold on
the DT conversion - part of the sales pitch was that you'd not need to
get board support upstream, which is a useful thing if you want to run
things like LTS or distro kernels on newer hardware.