Excerpts from netbsd: 23-May-97 Re: Son of diskless boot maximum
entropy@zippy.be (977)

> Yes, I did see that netboot.com passes in a struct in one of the
> arguments, which contains "ep@isa,0x200" (e.g. device type, bus, and
> iobase for the adapter) and I couldn't find anything in the kernel
> that tried to use this information.  I assume this is the information
> your startup patches uses to determine the correct device name, based
> on the devices probed at boot time.

Yes, this is what the patches contain.
This is not very elaborated yet, I'm still playing with different
variants. The scheme breaks eg with ISA plug-'n-pray devices
because the NetBSD implementation overwrites the configuration
information. That's why I only check with the device type, which
works as long as there are not 2 cards of 1 type.
Perhaps the hardware address should be passed too, but this
is not accessible via the "device" structure...
I'll work on this.
Simply overloading the "cyloffset" boot argument is not a good
idea either. The whole argument passing needs to be extended
(in a compatible way).
However, if it helps you, I'll put patches to
zelof1.zel.kfa-juelich.de:/pub/NetBSD/i386_bootdev-*.tar.gz
later today.

best regards
Matthias