Daniel Vetter <daniel@ffwll.ch> writes:

> Hm, I can't see v1 anywhere, but I think it'd be better. You can't store
> any transient state related to the current update in struct drm_plane. In
> this case the cleanup_buffers from a previous update might overlap (for
> nonblocking atomic commits) with the prepare_planes for the next one.
> Either we need special cleanup vs. error-path code, or some flag somewhere
> in the drm_plane_state.

Ok, here's pretty much the previous version, which works now that I've
fixed the intel driver. Instead of just comparing the fb's, I'm using
the framebuffer_changed function, which seems like a nice bit of
documentation if nothing else.