Re: lost /dev/agpgart , X11 confused.

On Wednesday 04 January 2006 04:26 pm, Ted Faber wrote:
> On Wed, Jan 04, 2006 at 02:58:22PM -0500, John Baldwin wrote:
> > On Wednesday 04 January 2006 02:20 pm, Ted Faber wrote:
> > > On Wed, Jan 04, 2006 at 02:08:27PM -0500, John Baldwin wrote:
> > > > Odd, it seems you don't have any AGP devices.  The 865G isn't a PCI-e
> > > > chipset, so you should have them.
> > >
> > > Yeah, my X server seems upset about it, all right.
> > >
> > > Am I not loading a module or device in my kernel config?  (attached)
> > > The config that's attached has been working for a while, and I didn't
> > > see anything in /usr/src/UPDATING that told me to change anything.  Of
> > > course, if I've been doing something wrong and getting away with it for
> > > a while, that's to be expected.
> > >
> > > The change between 1.33 and 1.34 to /sys/pci/agp_i810.c seems to delete
> > > some code that added a device for the drm, but that's just me reading a
> > > comment from
> > > http://www.freebsd.org/cgi/cvsweb.cgi/src/sys/pci/agp_i810.c.diff?r1=1.
> > >33&r 2=1.34&f=h I haven't really gone digging.
> > >
> > > Any suggestions?
> >
> > For i915 systems some people have hacked vga_pci.c to always add an "agp"
> > child device rather than doing it conditionally, but that I wouldn't
> > expect non-PCI-e chipsets to have the same issue.  Do you have AGP
> > enabled in your BIOS?
>
> As far as I can tell.  The BIOS setup didn't have a particular AGP
> on/off entry, but all the entries seemed reasonable to me and are the
> factory defaults.  The box is a Dell Dimension 3000 if that helps.
>
> I'm currently running a -CURRENT kernel from Mon Dec 19 11:21:13 PST and
> it properly detects the AGP, the /dev/agppart device appears, and the X
> server is fine.  I'm assuming that means AGP is enabled in my BIOS.  If
> that's a bad assumption, do you know how I can test it?
>
> In case it helps, I've included a dmesg from the Dec 19 kernel.

Ok, use the workaround other folks have posted for now.  Hack vga_pci.c to 
always add the "agp" device in its attach routine instead of checking for 
PCIY_AGP.

-- 
John Baldwin <jhb_at_FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve"  =  http://www.FreeBSD.org