> > i think you need to change the api or make everything which can
> > use vmem interrupt threads.
> 
> We would need an additional flag to vmem_create, or an ipl argument. I
> have a couple of questions:
> 
> - Is it useful to call vmem_create from an interrupt handler, or for it
>   not to sleep (VM_NOSLEEP)? I can't see how but there are undoubtedly
>   other uses I have not thought of.

i personally don't think it's useful.
people familiar with solaris might have a better answer, i guess.

> - Is it intended that the quantum cache could in the future be used from
>   an interrupt handler?

it's already used from interrupt handlers.  (arch/xen/i386/xen_shm_machdep.c)

YAMAMOTO Takashi