Re: [PATCH RFC 1/2] drm/panel: Add new helpers for refcounted panel allocatons

[Luca Ceresoli <luca.ceresoli@bootlin.com>]

Hello Anusha,

On Thu, 13 Mar 2025 16:34:45 -0400
Anusha Srivatsa <asrivats@redhat.com> wrote:

> > > +void *__devm_drm_panel_alloc(struct device *dev, size_t size, size_t  
> > offset,  
> > > +                          const struct drm_panel_funcs *funcs)
> > > +{
> > > +     void *container;
> > > +     struct drm_panel *panel;
> > > +     int err;
> > > +
> > > +     if (!funcs) {
> > > +             dev_warn(dev, "Missing funcs pointer\n");
> > > +             return ERR_PTR(-EINVAL);
> > > +     }
> > > +
> > > +     container = kzalloc(size, GFP_KERNEL);
> > > +     if (!container)
> > > +             return ERR_PTR(-ENOMEM);
> > > +
> > > +     panel = container + offset;
> > > +     panel->container_offset = offset;
> > > +     panel->funcs = funcs;
> > > +     kref_init(&panel->refcount);
> > > +
> > > +     err = devm_add_action_or_reset(dev, drm_panel_put_void, panel);
> > > +     if (err)
> > > +             return ERR_PTR(err);
> > > +
> > > +     drm_panel_init(panel, dev, funcs, panel->connector_type);  
> >
> > panel->connector_type here is uninitialized. You are passing
> > panel->connector_type to drm_panel_init(), which will then copy it into
> > panel->connector_type itself.
> >
> > So you mean I pass connector_type from the driver calling the helper, so  
> there is access to the connector type here?

I'm not a panel expert, but I think it makes sense that to create the
panel you need to know the connection type, and that is what Maxime
suggested.

> > > +     /**
> > > +      * @container_offset: Offset of this struct within the container
> > > +      * struct embedding it. Used for refcounted panels to free the
> > > +      * embeddeing struct when the refcount drops to zero.
> > > +      */
> > > +     size_t container_offset;  
> >
> > While storing the offset obviously works, and that's what I had
> > implemented in my latest bridge refcounting series, after some
> > discussion with Maxime we agreed storing a container pointer instead of
> > the offset is cleaner. I think it would be good here as well.
> >
> > See:
> > https://lore.kernel.org/lkml/20250227-macho-convivial-tody-cea7dc@houat/
> >  
> 
> so just void *container instead of size_t container_offset.

Exactly. You can have a look at the patch I sent earlier today:
https://lore.kernel.org/lkml/20250314-drm-bridge-refcount-v7-2-152571f8c694@bootlin.com/

Luca

-- 
Luca Ceresoli, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com