On Thu, May 03, 2001 at 11:38:47AM -0700, Bill Studenmund wrote:

 > I think one problem we'll have here is that the pci code wants to attach a
 > different driver to each pci function of the device. So the pcib and
 > pciide each expect to be handed disjoint things.
 > 
 > What might work is for us to have an MI viapcib attach to the pci bus, and
 > have the (MD) pcib attach to it. All the viapcib would do is note what
 > revision it is in a place which the via-specific pciide code could find.

Er.. yuck.  What's the point?  It is easy (see my previous mail) for
the pciide code to get the information it wants without such a hack.

-- 
        -- Jason R. Thorpe <thorpej@zembu.com>