On Fri, Feb 11, 2022, Sean Christopherson wrote: > On Wed, Feb 09, 2022, Paolo Bonzini wrote: > Hmm, so it doesn't matter (yet) because shadowing 4-level NPT with 5-level NPT > is fubar anyways[1], but this will prevent caching roots when shadowing 4-level > NPT with 5-level NPT (have I mentioned how much I love NPT?). This can be: > > if (VALID_PAGE(mmu->root.hpa) && mmu->root.hpa == __pa(mmu->pae_root)) Gah, that's wrong too, it will allow the pml4_root case. *sigh* I'm fine punting this until the special roots are less special, pml5_root is completely broken anyways.
