Re: [Qemu-devel] [PATCH 07/27] add memdev backend infrastructure

[Igor Mammedov]

On Wed, 27 Nov 2013 16:21:23 +0100
Paolo Bonzini <address@hidden> wrote:

> Il 27/11/2013 15:37, Igor Mammedov ha scritto:
> > It looks like "realize" for -object / object-add implemented via
> > an interface.
> 
> It does---but without unrealize and with the additional get_base_path.
> 
> > Maybe it should be renamed from QOMCommandLineIface to QOMRealizeIface
> > and s/complete/realize/ so anyone who knows about Device.realize would
> > get meaning without digging in complete() implementations.
> 
> There is an important difference; realize is an internal method in
> Device, the external interface is the property.  So perhaps it's the
> other way round; if Device implements QOMCommandLineIface you could
> start creating devices with -object.
> 
> > Alternative would be to behave just like Rng/Tpm do, i.e. use -object
> > to do late initialization in a backend user (DimmDevice.realize).
> > Draw back of it would be user won't get error during the first command
> > "object-add" and only will get error when creating DimmDevice calling
> > "device_add".
> 
> That's also a possibility.  But again, maybe it's the other way round
> and Rng/Tpm could enjoy better error handling if we add the interface.

Sure, I'll try to do as you described, it's much better to get error
earlier and from command/object that throws it than via proxy.

Thanks for suggestion, looking at netdev-add I even haven't thought
about using -object.

> Paolo