On Wed, Mar 23, 2011 at 11:12:20AM +0000, Mark Brown wrote: > On Wed, Mar 23, 2011 at 07:23:48AM +0900, Paul Mundt wrote: > > On Tue, Mar 22, 2011 at 11:00:56PM +0100, Kay Sievers wrote: > > > > Which is what we need to get rid of. It does not make any sense on the > > > global picture to have anything like that exported to userspace. > > > So far I haven't heard any rationale for why it doesn't. Exporting CPU > > state to userspace certainly makes sense, and the sysdev model has worked > > reasonably for CPUs, memory nodes, etc. > > FWIW it'd be really helpful to have CPUs (or at least SoCs) be regular > struct devices for integration with the regulator API so we can have all > things that might use a regulator (like DVFS) be struct devices but... > Sure, that makes sense. The easiest would probably be to just replace the struct cpu sysdev with a struct device pointer and fix up drivers/base/cpu.c accordingly. The linux/cpu.h API is unfortunately rather coupled to the idea of having a sysdev, but this is purely for attributes and attribute groups and primarily impacts powerpc, so the conversion shouldn't be too painful. For simple topology registration the bulk of the architectures ultimately don't care what's backing the struct cpu within the sysfs context. _______________________________________________ linux-pm mailing list linux-pm@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/linux-pm
