>  * Force the controller to interrupt.
> 
> I know that that was the way in 4.xBSD. According to driver(9) NetBSD
> runs autoconf(9) when interrupts are disabled. So I designed a rfc_match
> 
If the docs say that, it's wrong. I assume that it is arch dependent; the
Unibus needs an interrupt to be generated to find the interrupt vector
and interrupt level (and sometimes to verify the existance of a device).

> that does not generate an interrupt. 
> Now I get from sys/dev/qbus/uba.c:ubasearch()
> rfc0 at uba0 csr 177170 didn't interrupt
> I had a look at this fuction and it seams that it needs to get an interrupt 
> from the child device to catch the interrupt vector. Is that correct?
YES!

-- Ragge