> Would grabbing
> struct drm_plane_state from the atomic branch fix this for you?

That looks like it's meant to house the user requested coordinates
rather than the clipped ones. What you could do is just shovel the
drm_rects we use during the clipping into a new i915 specific plane
config struct and pass that to both check and commit. We can later
look into moving that stuff into some core struct if seems like a win
for more than one driver.