In <mtulngifu1m.fsf@contents-vnder-pressvre.mit.edu>
nathanw@MIT.EDU wrote:

> Ken Harrenstien <klh@us.oracle.com> writes:
> 
> > I recently picked up a Compaq Deskpro XL, which uses this chip, based
> > on the claims in the web sites of both FreeBSD and NetBSD that the
> > 53C794 was supported.
> > 
> > Foolish me -- this turns out not to be the case, and I finally realized
> > this is why I can't install either of them.  The FreeBSD people have
> > admitted that they dropped support without updating the documentation;
> > I'm not sure what the story is for NetBSD however.  (Look at
> > http://www.netbsd.org/Hardware/pci.html where it specifically lists the
> > AM53c974).
> 
> 	How does it not work? What did you try to boot, and what did
> you see? From the reference document you listed, it looks like a
> pretty straightforward two-pci-devices-on-one-chip widget. The
> Ethernet is likely supported by the le driver (PCI attachment), and
> the SCSI by the pcscp driver.

The 79c974 support should be easy, but currently it might not be
supported though i386 INSTALL doc mentioned that PCnet part of
the 79c974 would work fine. I think it would not be probed.

In /sys/dev/pci/if_le_pci.c:

---
le_pci_match(parent, match, aux)
  :
{
  :
	switch (PCI_PRODUCT(pa->pa_id)) {
	case PCI_PRODUCT_AMD_PCNET_PCI:
		return (1);
	}
---

and in /sys/dev/pci/pcscp.c:

---
pcscp_match(parent, match, aux)
   :
{
   :
	switch (PCI_PRODUCT(pa->pa_id)) {
	case PCI_PRODUCT_AMD_PCSCSI_PCI:
#if 0
	case PCI_PRODUCT_AMD_PCNETS_PCI:
#endif
		return 1;
	}
---

I have not got the 79c974 manual yet so I don't know how
should it be probed. I'm not familiar with PCI register
mappings...

> 	Admittedly, the pcscp driver only came into existence this
> January, so if you tried a 1.3 floppy it wouldn't work. But it sounds

I and Roger Brooks have made pcscp back-porting patches to 1.3.x
independently. I have not tested my version (because I'm using current),
but Roger is using his version on his 1.3.2 system.

BTW, Roger reported some problems with pcscp, and I have just
done send-pr to fix them. I hope someone takes a look
kern/7252 before 1.4 cut.
---
Izumi Tsutsui
tsutsui@ceres.dti.ne.jp