Quoting Greg Kroah-Hartman (2021-12-21 01:29:59)
> > @@ -471,9 +477,17 @@ int component_master_add_with_match(struct device
> > *parent,
> > if (!master)
> > return -ENOMEM;
> >
> > + id = ida_alloc(&aggregate_ida, GFP_KERNEL);
> > + if (id < 0) {
> > + kfree(master);
> > + return id;
> > + }
> > +
> > + master->id = id;
> > master->parent = parent;
> > master->ops = ops;
> > master->match = match;
> > + dev_set_name(&master->dev, "aggregate%d", id);
>
> You set the name yet the device is not "real"?
>
> I don't understand this patch at all, sorry.
The device is registered in the third patch. I'll squash this patch with
that one and move the rename patch to be the first patch in the series.
Then it should be clear why a struct device is introduced along with the
bus type it's associated with.
