On Mon, 4 Jan 2016, Mark Cave-Ayland wrote:
On 04/01/16 08:20, Alexander Graf wrote:
On 04.01.16 02:57, BALATON Zoltan wrote:
On Sun, 3 Jan 2016, Mark Cave-Ayland wrote:
Actually this may be a red herring, since the AAPL,address properties are present in the G4 AGP device trees but not the G4 PCI device trees. So unless G4 AGP are old world then this patch might be a red herring. Alex?
I think all G4 Macs are New World as it started somewhere around G3 and all later models had New World ROM (that is OpenFirmware without Toolbox). So I'm not sure why you don't see these in G4 PCI or if you do it's not because of Old World ROM.
The model id we have now (PowerMac3,1) is corresponding to the G4 AGP and I've picked this because it is the minimum required by MorphOS. Previously it was probably some PCI model but the set of devices currently emulated by QEMU may not really correspond to any of these real machines.
Yeah, my guess is that they left the legacy AAPL, bits in for Mac OS 9 compatibility in systems where they cared about it. Since it is present in real world DTs, I think it's perfectly valid to expose it in OpenBIOS as well.
In the examples I've looked at, AAPL,address isn't included in all devices on the G4 PCI so I suspect it may be generated by older FCode ROMs for cards that can work on both old world and new world Macs.
The device tree I've seen here: http://web.archive.org/web/20090107145016/http://penguinppc.org/historical/d....
only contains APPL,address properties for i2c nodes that lack the assigned-addresses property so I think it may have been added as an extension or were used before assigned-addresses was standardised and only retained for those devices where no standard property exists so having both of these on the same node may not make sense.
There's also AAPL,address-step that seems to appear together but I'm not sure what would be the standard property for that. Maybe #address-cells?
Regards, BALATON Zoltan