On 19.07.2018 13:45, Cornelia Huck wrote:
> On Thu, 19 Jul 2018 13:06:59 +0200
> Thomas Huth <th...@redhat.com> wrote:
>
>> On 17.07.2018 19:00, Juan Quintela wrote:
>
>>> So far so good, but look at virtio-pci.c:
>>>
>>> static void virtio_rng_pci_realize(VirtIOPCIProxy *vpci_dev, Error **errp)
>>> {
>>> ...
>>> }
>>>
>>> static void virtio_rng_pci_class_init(ObjectClass *klass, void *data)
>>> {
>>> ....
>>> }
>>>
>>> static void virtio_rng_initfn(Object *obj)
>>> {
>>> ...
>>> }
>>>
>>> static const TypeInfo virtio_rng_pci_info = {
>>> .name = TYPE_VIRTIO_RNG_PCI,
>>> .parent = TYPE_VIRTIO_PCI,
>>> .instance_size = sizeof(VirtIORngPCI),
>>> .instance_init = virtio_rng_initfn,
>>> .class_init = virtio_rng_pci_class_init,
>>> };
>>>
>>> static void virtio_pci_register_types(void)
>>> {
>>> type_register_static(&virtio_rng_pci_info);
>>> ...
>>> }
>>>
>>> See, we have defined the device "virtio-rng-pci", but there is no
>>> implementation. WHen I run device-intronspection-test on that qemu with
>>> CONFIG_VIRTIO_RNG, it fails to run. If we can agree that something is
>>> wrong, then we can search for a solution.
>>
>> I agree with you that the current situation with virtio-pci. c is bad. I
>> think we should split it up into individual files instead
>> (virtio-pci-rng.c etc.).
>
> We should then do the same thing for virtio-ccw as well.
Yes. I can do that if you like.
Thomas
