Since PCIDs will (probably?) not be available, I wonder whether it would make sense to unmap the kernel memory from the page tables (minus a stub for...