Whoa Turbo!

While I fully agree that pmap_kenter() is an abortion (and I didn't
want it added in the first place), I think people are a little
confused.

The entire *point* of pmap_kenter() was to optimize one particular
case -- that being where R/M information is (in theory) genuinely not
needed.  This is the case for pages that are exclusively mapped into
the kernel (e.g. with uvm_km_*()).  Similarly, the VAC aliasing issues
are not relevant in this case.

If pmap_kenter() is being used for other purposes, then something is
seriously broken.  But I scanned a bit and didn't find anything that
looked suspicious.