Subject: Re: PCI mode probe [was Re: INSTALL kernel support ]
To: Manuel BOUYER <bouyer@antioche.lip6.fr>
From: Matthias Drochner <drochner@zelux6.zel.kfa-juelich.de>
List: port-i386
Date: 12/11/1997 12:56:53
Excerpts from netbsd: 10-Dec-97 Re: PCI mode probe [was Re:.. Manuel
BOUYER@antioche.l (407*)

> Good news: it works on my compaq deskpro 4000, my old pci-mode-2 motherboard,
> my asus/cyrix motherboard and my compaq professional workstation.

That's encouraging.

But it's not yet perfect.
Jarkko Torppa reported that a PCNET-PCI network interface
stopped to work on a Compaq after he applied the patch. The
lance initialization failed, which is in my experience caused by the
master enable bit not set in the PCI command register. This bit is
explicitely set in if_le_pci.c.
My assumption is that the probe for mode 1 (which pedantically
follows the PCI spec) confuses the Compac chipset in a way
that subsequent PCI config write accesses either do nothing or
write junk.
A workaround should be to check for Compaq chips before the
normal mode-1 probe.

Does anybody have a Compaq and hardware which allows
to check if config write accesses do the right thing?

> Perhaps someone should test this on a 486 pci too ?

I did, with a CDC (Saturn?) chipset, and it correctly recognized
mode 2.

best regards
Matthias