To make a working kernel for my Toshiba 460CDX with a 3Com 3c589, I had to
make the following changes to the GENERIC config: 

add:

options	PCIC_ISA_INTR_ALLOC_MASK=0x1e08
options	PCIC_ISA_ALLOC_IOBASE=0x300
options PCIC_ISA_ALLOC_IOSIZE=0x060

remove:

wss0 at ... (the 460CDX has the Yamaha chipset but the driver doesn't have
             an ISA attachment yet)

optional:

sb0 at ... works but the quality isn't great.

scott


On Fri, 9 Oct 1998, Daniel Brown wrote:

> > Daniel Brown writes:
> > > Following myself up - this would appear to be something of a FAQ. Sorry.
> > > 
> > > I'm compiling a kernel with everything commented out (including lpt*), except
> > > all that's required for the job the machine's intended for. Hopefully this
> > > will correct things.
> > 
> > Actually, that will make things far worse.
> > 
> > Your problem is likely the PCMCIA subsystem's trouble with allocating
> > interrupts and I/O ports. It is absolutely critical that the kernel
> > probe and allocate all other devices on your machine before probing
> > the PCMCIA or it will attempt to assign interrupt and I/O ports in use 
> > by other devices. This would be BAD.
> 
> Perry,
> 
> I tried this, and with a similar lack of success, as you suggest.
> 
> In the end, I scavenged a Dayna Communicard E PCMCIA network card, and
> returned to a generic kernel (+pms0, -wss0 to avoid interrupt conflicts
> with sb0), with Havard's patch to provide Dayna Communicard E support
> (ne0 @ pcmcia0).
> 
> This works like a treat, which is a great relief.
> 
> It does suggest that the underlying problem is not with interrupt
> allocation, but with the driver itself. Assuming that if one PCMCIA
> network card works, they all should, of course...
> 
> Dan.
> 
> P.S. For the record, all this is under NetBSD-1.3.2/i386
>