Subject: Re: -current (1.2A-960626) snapshot available
To: Webmaster Jim <jspath@mail.bcpl.lib.md.us>
From: Michael Graff <explorer@flame.org>
List: port-i386
Date: 07/12/1996 15:31:08
The card isn't broken, it's just a bastard card.  (btw, I'm the one who
originally sent out the patches.  The files were since split into ahc_BUSTYPE
files, and no ahc_isa.c file exists, and I believe it should.)

When adaptec decided to make a VLB card, they simply took the EISA chipset,
put some glue between it and the VLB bus, and dubbed it done.  This was the
simple thing to do, but here's the problems with it.

Since some machines have both VLB and EISA busses, they couldn't simply
make it map into the same IO address space as the EISA cards do (why?
I dunno really...) so they made it a requirement to ``prime'' the address
before you could read the card ident.

Normal EISA cards ``just work'' by reading the slot's address where the
identification information is stored.  This VLB card sits in those addresses
as well, but the EISA routines ignore them, since EISA doesn't prime the
addresses.

Also, NetBSD will only scan the EISA bus if an EISA BIOS is installed, I
believe, so even if NetBSD _did_ prime the ports, you'd still loose.

This is why I think thre needs to be an ahc_isa.c written, since the VLB
card really is an ISA-like device, not EISA, or PCI.

--Michael