* Mark Brown <broo...@opensource.wolfsonmicro.com> [111123 10:34]: > On Wed, Nov 23, 2011 at 10:55:19AM -0800, Tony Lindgren wrote: > > * Russell King - ARM Linux <li...@arm.linux.org.uk> [111123 09:31]: > > > > Keep the clk API as a fundamental thing which should be initialized early > > > so we don't have to invent new clk APIs to get around its unavailability. > > > What else are you aware of that is really needed early for clocks other > > than clockevent? > > If nothing else we'd have to resolve all the device probe ordering > issues (Grant's patches seem a bit stalled) and convert all the systems > using the API to handle deferring probes until their clocks appear. > Using an early initcall of some kind would help with that but it does > seem like a lot of trouble and I'd expect it'll end up getting forced in > on most systems anyway.
Yes the ordering depends on the system. In any case, initcalls are not super early compared to setup_timer. Tony _______________________________________________ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev