Re: [Qemu-devel] [RFC] QEMU Object Model status/merge plan

[Anthony Liguori]

On 12/13/2011 09:05 AM, Kevin Wolf wrote:
> Am 13.12.2011 14:43, schrieb Anthony Liguori:
> > On 12/13/2011 05:35 AM, Stefan Hajnoczi wrote:
> > > On Mon, Dec 12, 2011 at 7:36 PM, Anthony Liguori<address@hidden>   wrote:
> > > > I choose the serial device to showcase what we'll eventually be able to do.
> > > >    The three relevant files are:
> > > >
> > > > https://github.com/aliguori/qemu/blob/qom-next/hw/isa-serial.c
> > > >
> > > > https://github.com/aliguori/qemu/blob/qom-next/hw/mm-serial.c
> > > >
> > > > https://github.com/aliguori/qemu/blob/qom-next/hw/serial.c
> > >
> > > I'm not sure I understand how init functions are called for derived
> > > classes.
> >
> > There are three types of init functions:
> >
> > class_init
> > ==========
> >
> > This lives in (TypeInit) and is called when a class is first created for a type.
> >    It is only ever called once.  Within this function, you should override any
> > methods in your base classes and set default implementations for any methods you
> > implement.
>
> I guess in most cases this could be replaced by a static table and the
> function could be made optional? (That is, there could be a default
> implementation for the NULL case)

As it turns out, you can pass an opaque to class_init so you could have 
something like:

typedef struct PCIDeviceOps {
   void (*foo)(PCIDevice *dev, ...);
   ...
};

And then you could write a generic:

void pci_generic_class_init(ObjectClass *klass, void *data)
{
    PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
    PCIDeviceOps *ops = data;

    if (ops->foo) {
        k->foo = ops->foo;
    }
}

Which would then let you do:

static PCIDeviceOps e1000_device_ops = {
    .foo = e1000_foo,
    ...
};

static TypeInfo e1000_device_info = {
    .name = TYPE_E1000,
    .parent = TYPE_PCI_DEVICE,
    .instance_size = sizeof(E1000State),
    .class_init = pci_generic_class_init,
    .class_data = &e1000_device_ops,
};

I didn't really plan on this, but it looks like it would work pretty well.  It 
might be reasonable to do for really common devices.  I don't think it's all 
that nice to do for everything though because the *_generic_class_init() 
functions get really ugly and makes allowing subclassing pretty hard.

> > instance_init
> > =============
> >
> > This is the constructor for a type.  It is called when an object is created and
> > chained such that the base class constructors are called first to initialize the
> > object.
>
> Same for this one, in your serial code it looks like this doesn't do
> anything interesting in the common case and could be made optional (it
> adds an UART child device, but this is static property and should be
> moved anyway)

It could potentially, yes.  instance_init functions will often times not be 
needed.  It's really meant to do things like initialize lists and stuff like 
that that property accessors may need to work with.

> I think even in the future the really interesting work will be done in
> realize.

Yes.  The various TypeInfo methods can all be omitted if they don't do any work.

Regards,

Anthony Liguori

> Kevin