Booting Imac G5 : success !

[Benjamin Herrenschmidt]

> As I said, we're lacking an IRQ pin in the PCI configuration space of
> the SATA controler, not in the OF tree (as you can see in the dump I
> previously sent).

(No, you wrote that we're lacking an IRQ _line_ which is different :)

Anyway, the correct fix is to do a quirk I suppose, that "corrects" the
IRQ pin register. Is any other device having the same bug ?

Note that we could also probably just ignore the IRQ PIN in the PCI
header, I have to check if it would have an impact on pSeries tho.
 
> So, pci_read_irq_line returns 0 at the first test:
> 	if (intpin == 0) {
> 		PPCDBG(PPCDBG_BUSWALK,"\tDevice: %s No Interrupt used by device.\n",
> 		       pci_name(pci_dev));
> 		return 0;	
> 	}
> 
> So, taking it from the OF tree instead of PCI space, the same way it's
> done for IDE, solves the problem.
> 
> > Once you have booted with your hack fix, can you send me a tarball
> > of /proc/device-tree please ?
> 
> Here you are... I also join a tarball of /proc/bus/pci so you can see
> all PCI devices configuration space dumps.
> 
-- 
Benjamin Herrenschmidt <benh at kernel.crashing.org>