On 2 December 2011 14:51, Benoît Canet <benoit.ca...@gmail.com> wrote:
> @@ -320,10 +321,15 @@ static void mpc8544ds_init(ram_addr_t ram_size,
> sysbus_create_simple("mpc8544-guts", MPC8544_UTIL_BASE, NULL);
>
> /* PCI */
> - dev = sysbus_create_varargs("e500-pcihost", MPC8544_PCI_REGS_BASE,
> - mpic[pci_irq_nrs[0]], mpic[pci_irq_nrs[1]],
> - mpic[pci_irq_nrs[2]], mpic[pci_irq_nrs[3]],
> - NULL);
> + dev = qdev_create(NULL, "e500-pcihost");
> + busdev = sysbus_from_qdev(dev);
> + qdev_init_nofail(dev);
> + sysbus_mmio_map(busdev, 0, MPC8544_PCI_REGS_BASE);
> + sysbus_connect_irq(busdev, 0, mpic[pci_irq_nrs[0]]);
> + sysbus_connect_irq(busdev, 1, mpic[pci_irq_nrs[1]]);
> + sysbus_connect_irq(busdev, 2, mpic[pci_irq_nrs[2]]);
> + sysbus_connect_irq(busdev, 3, mpic[pci_irq_nrs[3]]);
> +
Is this change actually necessary? If we only have one MMIO region
to map then sysbus_create_varargs() suffices.
How much did you test this patch, by the way? I've been running into
some odd behaviour with sub-page-sized memory regions which I'm
wondering if you hit here...
-- PMM
