Do any of your local patches touch radeon_evict_flags or
radeon_ttm_placement_from_domain? I don't see why it would recurse so deeply
otherwise.

A full public git tree to reproduce the problem and seeing what patches are
applied would also be nice.