Problem with OF interrupt parsing code

[Benjamin Herrenschmidt]

On Mon, 2007-10-01 at 23:11 +0200, Gerhard Pircher wrote:
>         pci at 80000000 {
>                 device_type = "pci";
>                 compatible = "mai-logic,articia-s";
>                 bus-frequency = <01fca055>;             // 33.3MHz
>                 bus-range = <0 ff>;
>                 ranges = <01000000 0 00000000 fe000000 0
> 00c00000       // PCI I/O
>                           02000000 0 80000000 80000000 0
> 7d000000       // PCI memory
>                           02000000 0 fd000000 fd000000 0
> 01000000>;     // PCI alias memory
>                 8259-interrupt-acknowledge = <fef00000>;
>                 interrupt-parent = <&i8259>;
>                 #interrupt-cells = <1>;
>                 #address-cells = <3>;
>                 #size-cells = <2>;
> 

Part of your problem is that interrupt-parent property. You shouldn't
have such a property in a PCI host bridge. It's not technically illegal,
but it's triggering the "loop" you've been experiencing.

If you want the parsing to fail for PCI devices (to get the fallback to
config space values), you need to make sure it does fail. 

Another option is to put an empty interrupt-map in there. That will
guarantee failure.

But that's all very ugly. I don't understand why you don't setup a
proper map either from your bootloader, zImage wrapper or even prom_init
or platform code.

Ben.