Re: [Qemu-devel] [PATCH 1/1] Support default interface types per QEMUMachine

[Anthony Liguori]

On 04/26/2012 01:17 PM, Peter Maydell wrote:
> On 26 April 2012 19:10, Anthony Liguori<address@hidden>  wrote:
> > QOM represents logic objects.  It's a data model and when we model hardware,
> > we can model something concrete (because hardware for the most part has a
> > physical implementation).
> >
> > Machines are entirely a QEMU concept.
>
> Not this one again? The real physical beagle board on my desk doesn't
> look much like a QEMU concept to me...

So where's your beagle board's "serial_hds[]" array inputs?  Where can I put a 
DriveInfo into it?

Not everything needs to be modeled as a device.  The QEMUMachine for a beagle 
board ought to have an init function that looks like:

void machine_init()
{
   BeagleBoard *bb = BEAGLE_BOARD(object_new(TYPE_BEAGLE_BOARD));

   for (i = 0; serial_hds[i]; i++) {
       // add a UART connected to GPIO pins off of the beagle board
   }

   for (i = 0; net_drives[i]; i++) {
       // add a USB-net connected to the USB controller on bb
   }
}

But I think modeling a BeagleBoard as a device with an array of CharDriverState 
links with logic in realize to do the above is a bit silly.  Just because we 
could do that doesn't mean we should.

Regards,

Anthony Liguori

> >   There's logic in the machines that
> > will never be able to be expressed in a config file.
>
> So? There's logic in a UART model that won't be able to be expressed
> in a config file, that doesn't constitute an argument for it not
> being a QOM object.
>
> -- PMM