Noticed this while trying to respin my MST suspend/resume patch series. It's not technically possible (at least until someone moves amdgpu away from the deprecated drm_device->driver->{load,unload} hooks) for amdgpu to properly register all of it's encoders before registering with userspace. However, amdgpu also apparently adds and removes encoders along with MST connectors - which is a much bigger issue as userspace applications definitely do not expect this type of behavior.
So, let's fix it and add some WARNs() so new drivers don't accidentally make this mistake in the future.
Lyude Paul (6): drm/amdgpu/dm/mst: Don't create MST topology managers for eDP ports drm/amdgpu/dm/mst: Remove unnecessary NULL check drm/amdgpu/dm/mst: Use ->atomic_best_encoder drm/amdgpu/dm/mst: Make MST encoders per-CRTC and fix encoder usage drm/amdgpu/dm/mst: Report possible_crtcs incorrectly, for now drm/encoder: WARN() when adding/removing encoders after device registration
drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h | 3 ++ .../gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 15 ++++++ .../gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h | 1 - .../display/amdgpu_dm/amdgpu_dm_mst_types.c | 46 ++++++++++--------- .../display/amdgpu_dm/amdgpu_dm_mst_types.h | 3 ++ drivers/gpu/drm/drm_encoder.c | 31 ++++++++++--- 6 files changed, 70 insertions(+), 29 deletions(-)