Re: [PATCH v5 02/10] object: qom module support

From: "Daniel P. Berrangé" <berrange@redhat.com>
To: Christophe de Dinechin <dinechin@redhat.com>
Cc: "Eduardo Habkost" <ehabkost@redhat.com>,
	qemu-devel@nongnu.org, "Gerd Hoffmann" <kraxel@redhat.com>,
	"Paolo Bonzini" <pbonzini@redhat.com>,
	"Marc-André Lureau" <marcandre.lureau@redhat.com>,
	philmd@redhat.com
Subject: Re: [PATCH v5 02/10] object: qom module support
Date: Tue, 21 Jul 2020 15:35:22 +0100	[thread overview]
Message-ID: <20200721143522.GK843362@redhat.com> (raw)
In-Reply-To: <7himeimfnc.fsf@turbo.dinechin.lan>

On Mon, Jul 20, 2020 at 04:20:55PM +0200, Christophe de Dinechin wrote:
> 
> On 2020-06-24 at 15:10 CEST, Gerd Hoffmann wrote...
> > Little helper function to load modules on demand.  In most cases adding
> > module loading support for devices and other objects is just
> > s/object_class_by_name/module_object_class_by_name/ in the right spot.
> >
> > Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
> > ---
> >  include/qom/object.h | 12 ++++++++++++
> >  qom/object.c         | 14 ++++++++++++++
> >  2 files changed, 26 insertions(+)
> >
> > diff --git a/include/qom/object.h b/include/qom/object.h
> > index 94a61ccc3fe8..51f188137f1f 100644
> > --- a/include/qom/object.h
> > +++ b/include/qom/object.h
> > @@ -994,6 +994,18 @@ bool object_class_is_abstract(ObjectClass *klass);
> >   */
> >  ObjectClass *object_class_by_name(const char *typename);
> >
> > +/**
> > + * module_object_class_by_name:
> > + * @typename: The QOM typename to obtain the class for.
> > + *
> > + * For objects which might be provided by a module.  Behaves like
> > + * object_class_by_name, but additionally tries to load the module
> > + * needed in case the class is not available.
> > + *
> > + * Returns: The class for @typename or %NULL if not found.
> > + */
> > +ObjectClass *module_object_class_by_name(const char *typename);
> > +
> >  void object_class_foreach(void (*fn)(ObjectClass *klass, void *opaque),
> >                            const char *implements_type, bool include_abstract,
> >                            void *opaque);
> > diff --git a/qom/object.c b/qom/object.c
> > index 6ece96bc2bfc..34daaf1280f5 100644
> > --- a/qom/object.c
> > +++ b/qom/object.c
> > @@ -985,6 +985,20 @@ ObjectClass *object_class_by_name(const char *typename)
> >      return type->class;
> >  }
> >
> > +ObjectClass *module_object_class_by_name(const char *typename)
> > +{
> > +    ObjectClass *oc;
> > +
> > +    oc = object_class_by_name(typename);
> > +#ifdef CONFIG_MODULES
> > +    if (!oc) {
> > +        module_load_qom_one(typename);
> > +        oc = object_class_by_name(typename);
> > +    }
> > +#endif
> 
> I'm wondering if there is any reason to only trigger the module load when
> you don't find the object class. You could simply call module_load_qom_one
> under #ifdef CONFIG_MODULES.
> 
> Performance wise, I don't think this makes much of a difference, and it
> simplifies the logical flow IMO.

I wouldn't make that assumption about performance - module_load_qom_one()
does alot of work, and there are places where object / class creation is
performance critical. We might not happen to trigger them now with the
current set of modules, but we can easily do so in future.

Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|