Re: [Qemu-devel] [PATCH v8 2/3] qom: pass original path to unparent method

["Michael S. Tsirkin" <mst@redhat.com>]

On Mon, Mar 18, 2013 at 09:24:16AM -0500, Anthony Liguori wrote:
> "Michael S. Tsirkin" <mst@redhat.com> writes:
> 
> > We need to know the original path since unparenting loses this state.
> >
> > Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> > ---
> >  hw/qdev.c            | 4 ++--
> >  include/qom/object.h | 3 ++-
> >  qom/object.c         | 4 +++-
> >  3 files changed, 7 insertions(+), 4 deletions(-)
> >
> > diff --git a/hw/qdev.c b/hw/qdev.c
> > index 741af96..64546cf 100644
> > --- a/hw/qdev.c
> > +++ b/hw/qdev.c
> > @@ -436,7 +436,7 @@ static void qbus_realize(BusState *bus, DeviceState *parent, const char *name)
> >      }
> >  }
> >  
> > -static void bus_unparent(Object *obj)
> > +static void bus_unparent(Object *obj, const char *path)
> >  {
> >      BusState *bus = BUS(obj);
> >      BusChild *kid;
> > @@ -756,7 +756,7 @@ static void device_class_base_init(ObjectClass *class, void *data)
> >      klass->props = NULL;
> >  }
> >  
> > -static void device_unparent(Object *obj)
> > +static void device_unparent(Object *obj, const char *path)
> >  {
> >      DeviceState *dev = DEVICE(obj);
> >      DeviceClass *dc = DEVICE_GET_CLASS(dev);
> > diff --git a/include/qom/object.h b/include/qom/object.h
> > index cf094e7..f0790d4 100644
> > --- a/include/qom/object.h
> > +++ b/include/qom/object.h
> > @@ -330,11 +330,12 @@ typedef struct ObjectProperty
> >  /**
> >   * ObjectUnparent:
> >   * @obj: the object that is being removed from the composition tree
> > + * @path: canonical path that object had if any
> >   *
> >   * Called when an object is being removed from the QOM composition tree.
> >   * The function should remove any backlinks from children objects to @obj.
> >   */
> > -typedef void (ObjectUnparent)(Object *obj);
> > +typedef void (ObjectUnparent)(Object *obj, const char *path);
> >  
> >  /**
> >   * ObjectFree:
> > diff --git a/qom/object.c b/qom/object.c
> > index 3d638ff..21c9da4 100644
> > --- a/qom/object.c
> > +++ b/qom/object.c
> > @@ -362,14 +362,16 @@ static void object_property_del_child(Object *obj, Object *child, Error **errp)
> >  
> >  void object_unparent(Object *obj)
> >  {
> > +    gchar *path = object_get_canonical_path(obj);
> >      object_ref(obj);
> >      if (obj->parent) {
> >          object_property_del_child(obj->parent, obj, NULL);
> >      }
> >      if (obj->class->unparent) {
> > -        (obj->class->unparent)(obj);
> > +        (obj->class->unparent)(obj, path);
> >      }
> 
> I think you should actually just move this call above
> if (obj->parent) { object_parent_del_child(...); }.
> 
> There's no harm AFAICT in doing this and it seems more logical to me to
> have destruction flow start with the subclass and move up to the base
> class.

At Paolo's request children are intentionally reported before parents,
shouldn't this apply?

> 
> This avoids needing a hack like this because the object is still in a
> reasonable state when unparent is called.
> 
> Paolo, do you see anything wrong with this?  I looked at the commit you
> added this in and it doesn't look like it would be a problem.
> 
> Regards,
> 
> Anthony Liguori

Hmm I already put this on my branch (and sent a pull request).
I guess I could back it out, though it will create minor problems if
someone is basing on my tree.

Cleanup in a separate patch?

> >      object_unref(obj);
> > +    g_free(path);
> >  }
> >  
> >  static void object_deinit(Object *obj, TypeImpl *type)
> > -- 
> > MST