On Sat, Mar 23, 2002 at 09:31:55PM -0800, Chuck Silvers wrote:
> one change that might improve things would be to add a new PMAP_LOAN
> flag for pmap_enter(), ...
> there are probably additional subtleties here.

... and it didn't take too long for me to think of one.
the result of the alias-detection code noticing a bad alias
for a page is that it would make all the mappings of that page
uncachable, which would be unacceptably slow.  so to make this work
we'd have to arrange for loan-to-kernel mappings to have
good cache alignment in the common case, even for anon pages.
that would involve yet more pmap interface changes since PMAP_PREFER()
only really makes sense for object mappings.

if people want to try implementing various ideas and benchmarking them,
that would be great, but we need to be extremely careful with this stuff.

-Chuck