On Thu, May 27, 2010 at 12:24 AM, Igor Kovalenko igor.v.kovalenko@gmail.com wrote:
On Wed, May 26, 2010 at 11:02 PM, Blue Swirl blauwirbel@gmail.com wrote:
On Tue, May 25, 2010 at 12:38 PM, Igor V. Kovalenko igor.v.kovalenko@gmail.com wrote:
From: Igor V. Kovalenko igor.v.kovalenko@gmail.com
- refactor scan procedure to start with PCI host controller
- initiate scan of subordinate PCI bus from PCI host and PCI-to-PCI bridges
- find out PCI subordinate bus numbers and write them to bride devices
- automated assignment of "reg", "ranges", and "bus-range" properties
Unfortunately this breaks the PPC Mac99 machine.
To test, simply run 'qemu-system-ppc -M mac99', the screen remains black. 5/10 still works.
The expectation to have host PCI bridge at pci0:0:0 is invalid, mac99 has it at device 0xb The following change attached hereto fixes that by adding a scan for host bridge.
It would be great if you could include it into the set after currently failing one, thanks!
Unfortunately there is a typo in ob_pci_read_identification, it must check pointers before assignment, not data pointed to. It does not produce errors in my -M mac99 test but breaks sparc64 test.
Please check the updated version attached. I now tested qemu-system-sparc64 and qemu-system-ppc -M mac99 with updated version.