PERFORCE change 75921 for review

[Marcel Moolenaar]

On Apr 24, 2005, at 11:37 PM, M. Warner Losh wrote:

> In message: <c228a787d0ab62d16e6c7457b9c5dba5 at xcllnt.net>
>             Marcel Moolenaar <marcel at xcllnt.net> writes:
> :
> : On Apr 24, 2005, at 10:49 PM, M. Warner Losh wrote:
> :
> : > In message: <200504250459.j3P4xWxs057683 at repoman.freebsd.org>
> : >             Marcel Moolenaar <marcel at freebsd.org> writes:
> : > : http://perforce.freebsd.org/chv.cgi?CH=75921
> : > :
> : > : Change 75921 by marcel at marcel_nfs on 2005/04/25 04:59:01
> : > :
> : > : 	Allow VGA resources across PCI-PCI bridges. Required to
> : > : 	support a PCI based VGA driver with the VGA hardware
> : > : 	behind a PCI-PCI bridge.
> : >
> : > WE need to make sure that the ISA bit and VGA bits are set 
> properly on
> : > those bridges with vga children and peers of bridges with vga
> : > children before allowing this...
> :
> : Ideally, yes. But since only one PCI device of class display and
> : subclass VGA can have the legacy VGA addresses enabled and we're
> : not mucking with the system configuration in that respect, we're
> : not really doing anything dangerous by allowing the VGA addresses
> : without such scrutiny provided we don't try to allocate those
> : addresses for any ol' VGA display device that we find on the PCI
> : busses in the system.
>
> Actually yes, really, we need to do this.  Most system BIOSes will get
> it right, but not all.

I think that means we already have problems with syscons(4) then on
those machines, right? I mean, syscons just peeks and pokes at the
addresses and if it gets something it expects, then there's a VGA
device. Allowing VGA addresses across PCI-PCI bridge devices yields
exactly the same: some newbusified driver starts to peek and poke
at the VGA addresses and if it gets something it expects, then the
device can attach.

-- 
  Marcel Moolenaar         USPA: A-39004          marcel at xcllnt.net