On Thu, Dec 12, 2024 at 12:10:58PM +0200, Jani Nikula wrote:
> [...]
> On Thu, 12 Dec 2024, Imre Deak <imre.deak@xxxxxxxxx> wrote:
> > @@ -863,9 +866,14 @@ EXPORT_SYMBOL(drm_connector_dynamic_register);
> > * drm_connector_unregister - unregister a connector
> > * @connector: the connector to unregister
> > *
> > - * Unregister userspace interfaces for a connector. Only call this for
> > - * connectors which have been registered explicitly by calling
> > - * drm_connector_register().
> > + * Unregister userspace interfaces for a connector. Drivers should call this
> > + * for dynamic connectors (MST) only, which were registered explicitly by
> > + * calling drm_connector_dynamic_register(). All other - static - connectors
> > + * will be unregistered automatically by DRM core and drivers shouldn't call
> > + * this function for those.
>
> This kind of supports my point about a single
> drm_connector_register(). There's no
> drm_connector_dynamic_unregister(). After all the
> drm_connector_register() calls have been removed, we're left with the
> asymmetric pair:
>
> - drm_connector_dynamic_register()
> - drm_connector_unregister()
>
> Then again, all of these should become internal and not for drivers?
Yes, drm_connector_register() - after this patchset - could be removed
from drivers. The use of drm_connector_unregister() in drivers for
static connectors should be also removed, left it for DRM core internal
use only and exporting drm_connector_dynamic_unregister() for driver
use (atm only MST).
> > + *
> > + * Note: Existing uses of this function in drivers for static connectors
> > + * should be a nop already and are scheduled to be removed.
> > */
> > void drm_connector_unregister(struct drm_connector *connector)
> > {
>
> --
> Jani Nikula, Intel
