In message <mtuu0as4x1z.fsf@contents-vnder-pressvre.mit.edu>, "Nathan J. Willia
ms" writes:
>Note the "extent_flags" hack in powerpc/powerpc/bus_space.c. It
>doesn't bother to allocate the extent for bootstrap uses... but
>there's no corresponding check in the unmap code. My interpretation of
>this is that bootstrap code should feel free to not bother to unmap
>the region; adding the matching extent_flags hack to the unmap code
>would be okay, too.

Hmm.  My main concern about not unmapping is just that, according to the
bus_space man page, it's perfectly reasonable to refuse to map something
that's already mapped.  So, if I've mapped the registers for this device,
a later bus_space_map for these registers could legitimately fail.

Of course, I don't care, because I can't even find a possible code path
that would lead to accessing this device.

-s