yamt@mwd.biglobe.ne.jp (YAMAMOTO Takashi) wrote:
> > Any better suggestions? Thoughts?
> 
> have you considered something more general like solaris's resource pools api
> or linux containerfs, on which pset_* api can be implemented if desired?

Not yet, that is a second stage :)

OK, preparing for that is reasonable. In such case we might do the same as
Solaris - use psetid_t as a system-wide ID, which might be assigned to
some analog of resource pool, or anything we want later. There would be an
array of pointers to the processor-set structures which would include the
bitmap of CPUs, reference counting and perhaps something more in the future.
The psetid_t would be an index of the array. All this would be kernel-only.
For the bitmap of CPUs, I would agree with David Holland to have it in the
same form as sigset_t.

Comments?

-- 
Best regards,
Mindaugas
www.NetBSD.org