> If it were just for that purpose, it would be a hack.  FreeBSD uses it
> to decide whether to process a non-masked interrupt immediately, or to
> defer it until after the currently active ones have been processed.

NetBSD has other ways of doing this which don't depend on the number
of currently-running interrupt handlers.

it is unlikely that this feature will be necessary on all platforms.

I can see us adding support to allow this sort of thing to be used in
asserts in DEBUG/DIAGNOSTIC kernels (where performance is less of an
issue) but I don't see it showing up in production kernels -- why
should we make all interrupts pay for something which isn't needed by
most drivers?

						- Bill