split_huge_pmd() for file mappings (and DAX too) is implemented by just clearing pmd entry as we can re-fill this area from page cache on pte level la
↧