[PATCH] Separate pci bits out of struct device_node

[Nathan Lynch]

Hi Paul,

No objection to the intent, but some concerns below.

Paul Mackerras wrote:
> @@ -40,16 +42,26 @@
>  	struct pci_controller *phb = data;
>  	int *type = (int *)get_property(dn, "ibm,pci-config-space-type", NULL);
>  	u32 *regs;
> +	struct pci_dn *pdn;
>  
> -	dn->phb = phb;
> +	if (phb->is_dynamic)
> +		pdn = kmalloc(sizeof(*pdn), GFP_KERNEL);
> +	else
> +		pdn = alloc_bootmem(sizeof(*pdn));

I don't see in the patch where the pci_dn is freed when its
device_node goes away.

Should the lifetime of a pci_dn be tied to its corresponding
device_node?

Should code which accesses a pci_dn hold a reference to its
corresponding device_node?