Jason Thorpe <thorpej@nas.nasa.gov> writes:
> We could invent a new call: {isa,pci}_intr_pending().  That could tell us if
> an ISA interrupt is pending or not.  It could take the appropriate interrupt
> handle (IRQ for ISA, PCI interrupt handle for PCI, etc.) and return TRUE
> or FALSE corresponding to whether or not that interrupt is pending delivery
> when interrupts are enabled again.

The question then becomes, what do you do to increase the confidence
value of the result gained from using this function?

You'd really like something to (try) to clear 'interrupt pending'
status, i.e. so you could do:

	clear pending interrupt (or try to, if you can)
	generate interrupt
	is interrupt pending?

which should give you the right value more often than just:

	generate interrupt
	is interrupt pending?



cgd
-- 
Chris Demetriou - cgd@netbsd.org - http://www.netbsd.org/People/Pages/cgd.html
Disclaimer: Not speaking for NetBSD, just expressing my own opinion.