Am 30.09.2010 um 20:46 schrieb Blue Swirl:
On Thu, Sep 30, 2010 at 5:02 PM, Andreas Färber <andreas.faerber@web.de
wrote: Not having (low-level) access to an IBM Power System / System p / pSeries, the best source I could find was a JS20 device tree [1]. (Some older RS/6000 ones for comparison are available at [2].)
If you don't have to implement JS20, are there any other machines where all devices are located inside the first 4G?
It doesn't need to be a JS20; any POWER4, PPC970, POWER5, POWER6 or POWER7 CHRP system from IBM should do. Maybe Alex has access to one for KVM?
Finding such a chrp system doesn't solve the mac99 issue though. Haiku is happy that on OpenBIOS/ppc64 everything is like on 32-bit since it hadn't been ported to ppc64 yet, so I'd rather have qemu-system-ppc64 for testing the actual differences than hiding them. As you confirmed, it's not a question of cell size but of cell count. I guess I'll have to try using a new type like phys_addr_t to replace ucell in the ofmem backend.
There are more trees here: http://penguinppc.org/historical/dev-trees-html/trees-index.html
Yeah, those were linked from tibit.org, but they're all Mac trees and equally old.
Andreas