On Wed, Feb 06, 2019 at 04:33:54PM -0800, Dave Hansen wrote: > I wonder if the patches that you bisected to just changed the flushing > from being CR3-based (and not taking an address) to being INVPCID-based, > and taking an address that is sensitive to canonicality.
That is indeed one of the things that patch series does; before this it would always flush world for ARRAY interfaces. I'll have a look at the code; I seem to remember there being test for canonical addresses, maybe they're not in the right place.