Re: [Qemu-devel] [PATCH v3 2/6] hw/arm/dyn_sysbus_devtree: helpers for sysbus device dynamic dt node generation

[Paolo Bonzini <pbonzini@redhat.com>]

Il 09/09/2014 17:56, Eric Auger ha scritto:
>>> >> +    if (!sbdev) {
>>> >> +        /* Container, traverse it for children */
>>> >> +        return object_child_foreach(obj,
>>> >> +                                    arm_sysbus_device_create_devtree, data);
>>> >> +    }
> When we add support for a dynamically instantiable device we add
> something like
> 
>     if (object_dynamic_cast(obj, TYPE_ETSEC_COMMON)) {
>         create_devtree_etsec(ETSEC_COMMON(dev), data);
>         matched = true;
>     }
>>> >> +
>>> >> +    if (!matched) {
>> > 
>> > Who is going to set "matched", since it doesn't escape?
> 
> 

That's not part of this patch though, right?  So this code for now is
dead.  Please remove the dead code if it is not used in this series.

We really should make that an interface, so that the code can do just

    if (object_dynamic_cast(obj, TYPE_FDT_BUILDER)) {
        fdt_builder_create_fdt(FDT_BUILDER(dev), data);
    } else {
        ...
    }

(and so can the generic virt.c code) but that can come later.

>> Why does arm_sysbus_device_create_devtree need intc and irq_start?
> 
> irq_start: needed because when the "interrupts" property is set for the
> leaf component the irq number is irq_start +
> object_property_get_int(obj, "irq[i]", NULL)
> irq[i] being in [0, params->platform_bus_num_irqs]

Ah, it's passed to the not-yet-existing create_* functions.

> intc: this was in case the leaf component would use "interrupt-parent"
> prop. I miss experience on device trees and I don't know if it make
> sense the leaf component uses a different interrupt controller than the
> parent platform bus or if such property is mandatory in some cases.
> Maybe not needed indeed.

No idea, sorry.

Paolo