I think evbarm is not a good reference point to start ARM port.  The
attempt is, in effect, to pretend a single port ( ~computer) by collecting
different designs and different configurations of hardwares.  There is no
such animal on this planet.  To achieve this goal, astonishing number
of files are made.  Let's take a look at how any single kernel is prepared
and configured to build, and recognize how bad to add "new evbarm" to
that.

The existent of .start section is another complication.  It was invented
to overcome an issue about how initial MMU arrangement is done prior
to CPU hand over its control to kernel_start.  Not all boot loader can
handle .start section as attemped.  It might not be a help but an obstacle
to start.

So, I would like to advise people who engage in potential ARM port in
future not to follow evbarm framework to begin.  Things can be done
simpler, I believe.

Toru Nishimura/ALKYL Technology