On Thursday 26 April 2012, Ryan Mallon wrote:

quoted

See Shawn's "[PATCH 00/17] arch/arm/mach-* late_initcall cleanup".

I haven't looked closely at this yet. I think there may be some platforms
where the late_initcall's might cause a problem when making an ARM
unified kernel.

Hmm, I see. It's a bit annoying that it makes the code less clean, but I
can't really see any other nice way of doing it :-/.

Well, we could treat crunch as a special case and not call it through
the init_late callback but instead add a check in crunch_init() detect
the presence of crunch hardware. We don't have to replace every
late_initcall with the new method, although it only makes sense to
apply the series if we do most of them.

	Arnd