Upstreaming MMP clock devicetree support

[Daniel Drake]

On Thu, Jun 13, 2013 at 11:56 AM, Paul Fox <pgf at laptop.org> wrote:
> does this imply a future flag day, at which time new firmware will be
> incompatible with old kernels, and vice-versa?

Unfortunately even without the DT changes described here, old firmware
versions already will not be able to boot new/upstream kernels, once
we get to that point. The DT is missing regulator information so the
MMC controller doesn't get powered up. This is the "old firmware, new
kernel" problem.

The DT clock changes described here would also mean that new firmwares
become incompatible with released kernels. This is the "new firmware,
old kernel" problem. I think this will also be unavoidable.

So yes, we are looking at both of those problems.

There are some thoughts and possibilities here:

http://wiki.laptop.org/go/Device_tree_upgrade_considerations
https://lists.ozlabs.org/pipermail/devicetree-discuss/2012-August/019024.html

but no definite, attractive solution. I suspect it is something we are
just going to have to live with.

I think the "old firmware, new kernel" is the most painful of the two
problems. Maybe we can fix that up with a bootwrapper type solution
mentioned in the mail linked to above (and maybe I already half-wrote
the bootwrapper in my previous mail).

There is also the possibility that the kernel will accept bindings for
old/imperfect DTs once good bindings are put in place - or
alternatively we could patch them in. I'm working on the "good
bindings" aspect at the moment.

Daniel