NetBSD thread model: a few questions

[Jean-Yves Migeon <jeanyves.migeon%free.fr@localhost>]

Hi list,

I would like to have a few clarification about NetBSD's thread model, 
concerning kernel preemption (when enabled) and threads scheduling, 
under particular situations.

These questions relates to my work under Xen, especially suspension of 
the LWP and CPUs. I know that we are not supporting SMP under Xen 
currently, but I am preparing for it.

- is it valid for a LWP to call kpreempt_disable() but still use 
synchronisation primitives in the non-preemptible part, thus resulting 
in voluntary context switches? KASSERTs in the mutex code in this regard 
are not particularly clear (yes, I know, the question is somewhat 
stupid, but I would like to know if such a situation may ever happen)

- when a LWP is bound to a CPU and that CPU is put offline, bound 
threads may still execute on it. However, Xen only saves the context of 
CPU0, the rest being at the responsibility of the domain to save/restart 
the CPU on suspend/resume.

Is there a way to correctly assert that bound threads running on a CPU 
do not hold any locks when we are about to save the CPU context, to 
avoid any possible deadlock between bound LWP and LWP waiting for locked 
ressources? Or is it acceptable to alter the affinity of bound threads 
arbitrarily and restore it later?

Many thanks in advance for your answers.

--
Jean-Yves Migeon
jeanyves.migeon%free.fr@localhost