> Index: svn/src/mainboard/gigabyte/ga_2761gxdk/Config.lb > =================================================================== > --- svn.orig/src/mainboard/gigabyte/ga_2761gxdk/Config.lb > +++ svn/src/mainboard/gigabyte/ga_2761gxdk/Config.lb > @@ -178,9 +178,7 @@ chip northbridge/amd/amdk8/root_complex > chip southbridge/sis/sis966 > device pci 0.0 on end # Northbridge > device pci 1.0 on # AGP bridge > - chip drivers/pci/onboard # Integrated VGA > device pci 0.0 on end > This looks fishy, but then again, I never understood the v2 device tree syntax completely.I'm not sure what looks fishy here. This is the way I understand it:
- Devices between the 'on' and 'end' tokens are children (or children
of children) of the device. 2. The 'chip' token assigns the driver for devices inside it.
So in this case, device 1.0 has an AGP bus with 0.0 hanging off of it.
In which case the inner device pci 0.0 on end should be removed. Otherwise it gets assigned an on_mainboard = 1, and be treated as an on-mainboard graphics card even though it's an AGP plugin card.
Yeah. I thought maybe it was on-mainboard AGP.
Generally, the static tree should only contain devices that can not be removed
Agreed. That gets a little problematic with some of the "stuff options" where some of the boards have SCSI or an extra nic or...
But, to be sure we might want to trust the original authors of the
device trees that they knew what they were doing unless someone reports a problem or fixes one.
Sounds good.
Thanks, Myles