On 08/12/2011 06:23 AM, Peter Maydell wrote:
> static int realview_mpcore_init(SysBusDevice *dev) > { > mpcore_rirq_state *s = FROM_SYSBUS(mpcore_rirq_state, dev); > @@ -79,7 +84,7 @@ static int realview_mpcore_init(SysBusDevice *dev) > } > } > qdev_init_gpio_in(&dev->qdev, mpcore_rirq_set_irq, 64); > - sysbus_init_mmio_cb(dev, 0x2000, mpcore_rirq_map); > + sysbus_init_mmio_cb2(dev, mpcore_rirq_map, mpcore_rirq_unmap); > return 0; > }Since the mpcore code here is only using sysbus mmio callbacks as a way to pass things through to a sysbus mmio provided by another device, I think it would be cleaner to just have a way to say "my mmio is actually provided by this other device". For instance you could have MemoryRegion* sysbus_mmio_get_region(SysBusDevice *dev, int n)) { return dev->mmio[n].memory; } and then realview_mpcore_init() can just sysbus_init_mmio_region(dev, sysbus_mmio_get_region(s->priv, 0)); Or you could have a sysbus_pass_mmio() which worked like sysbus_pass_irq() and just said "my mmios are all this other device's mmios"; that's less flexible though.
Yes. Yet another option is to create a container MemoryRegion and pass it upwards, then allow the device that actually implements that mmio to register its regions with the container.
I'll leave this to the device maintainer though. -- I have a truly marvellous patch that fixes the bug which this signature is too narrow to contain.
