> > PCIBus *ppce500_pci_init(qemu_irq pci_irqs[4], target_phys_addr_t
> > registers) {
> > - PPCE500PCIState *controller;
> > + DeviceState *dev;
> > + PCIBus *b;
> > + PCIHostState *h;
> > + PPCE500PCIState *s;
> >
> > PCIDevice *d;
> >
> > - int index;
> >
> > static int ppce500_pci_id;
> >
> > + SysBusDevice *sb;
> > +
> > + dev = qdev_create(NULL, "e500-pcihost");
> > + sb = sysbus_from_qdev(dev);
> > + h = FROM_SYSBUS(PCIHostState, sb);
> > + s = DO_UPCAST(PPCE500PCIState, pci_state, h);
> > +
> > + b = pci_register_bus(&s->pci_state.busdev.qdev, NULL,
>
> No. This function should not exist. All this should be done in
> e500_pcihost_initfn. Please do the qdev conversion properly.
Or more precicely it should not depend on the internals of ppce500_pci.c.
In principle the only public entry point in that file should be
device_init(...). If used by multiple boards a simple helper function may be
appropriate (e.g. smc91c111_init). Note that this helper function has no ties
to the rest of that file, and could be trivially moved into a different file
or replaced with a macro/inline funciton.
ppce500_pci_init definitely should not be creating the PCI bus.
Paul
