On Thu, Jul 09, 2026 at 03:49:41PM +0200, Thomas Zimmermann wrote: > Hi Maxime > > Am 09.07.26 um 13:50 schrieb Maxime Ripard: > > The simple-kms helper provides a reset_plane hook to let drivers > > override the default plane reset. However, the reset hook is > > overloaded: it is used to create the initial software state at probe > > time, but also to reset both the hardware and software state during > > suspend/resume. These two roles have different expectations, and > > reset is not fallible which makes error handling difficult. > > > > The atomic_create_state pattern was introduced to untangle this by > > providing a hook that only allocates and initializes a pristine state > > without any side effect. > > > > Add a create_plane_state hook to struct > > drm_simple_display_pipe_funcs. When provided, > > drm_simple_kms_plane_reset() will use it to allocate a fresh state, > > falling back to reset_plane and then to the default > > drm_atomic_helper_plane_reset(). > > > > This will allow simple-kms drivers to be converted to the > > atomic_create_state pattern. > > Have you seen > https://lore.kernel.org/dri-devel/20260705-drm-simple-kms-removal-v1-0-b4e1ca053...@oss.qualcomm.com/ > ? Once this lands, simple-kms will have no users left and be gone soon > after.
Oh, yeah, absolutely, if these get removed it makes much more sense to just remove the code instead of converting to it. Maxime
signature.asc
Description: PGP signature
