> > can anyone, esp. non-x86 folks, review these api?  thanks.
> 
> [cpu_switchto]
> 
> "cpu_switchto() doesn't switch address spaces.". I know sparc currently
> needs to play with address space identifiers on switch. Could lazy switching
> be done for that instead?

i guess it's more natural to do it in pmap_activate/deactivate,
which are explicitly called by mi_switch on the branch.

> [cpu_need_resched]
> 
> I noticed that you implemented monitor/mwait, which was on my list of SMP
> tasks (thanks!). It might be useful have a "bool immediate" argument. If a
> high priority LWP has been released (e.g. realtime) it would probably make
> sense to force the resched to happen immediately.

i think it makes sense.

YAMAMOTO Takashi