beaglebone, take two

[Taylor R Campbell <campbell+netbsd-port-arm%mumble.net@localhost>]

I seem to have completely botched my last attempt to send this
message, so let's try that again, with an out-of-line patch, and
apologies to the mail archives for that...

(I am not subscribed to this list, so please cc me in replies.)

The patch at

http://www.NetBSD.org/~riastradh/tmp/beaglebone-20121209.patch

gets my beaglebone booted to the point that it asks for a root device.
It adds a driver to handle both versions of the OMAP-style dual-mode
timer (v1 is found on the OMAP35x; v1 and v2 are both found on the
AM335x as in the beaglebone), and some preliminary infrastructure for
enabling and disabling OMAP-style module clocks with the PRCM, which
is currently done in a totally ad-hoc way throughout arm/omap.

This is nowhere near complete, but since I am new to this part of the
world, I didn't want to start committing without some review.

Eventually I'd like to disentangle the #ifdefs throughout arm/omap so
that it's easier to navigate the common OMAP-style parts (e.g.,
omap_dmtimer.c), the SoC-specific layouts (e.g., am335x_dmtimer.c),
and the board configurations (e.g., conf/BEAGLEBONE).  I think it will
also be useful to separate the autoconf tree (obio0/obio1/obio2) from
the power and clock domain structure, because there is no good order
for autoconf to configure the devices as it is currently structured.

Comments?  OK to commit?