> > > > > bits for the supported features. > > > - Add kvm_get_active_cr3_ctrl_bits() to get the active control bits to form > > > a new guest CR3 (in vmx_load_mmu_pgd()). > > KVM handles #PF for shadow MMU, and for TDP (EPT) there's also a case that KVM > > will trap the #PF (see allow_smaller_maxphyaddr). Do we need any handling to > > the linear address in the #PF, i.e. stripping metadata off while walking page > > table? > > > > I guess it's already done automatically? Anyway, IMO it would be better if you > > can add one or two sentences in the changelog to clarify whether such handling > > is needed, and if not, why. > > LAM masking applies before paging, so the faulting linear address > doesn't contain the metadata. > It has been mentioned in cover letter, but to be clear, I will add the > clarification in the changelog > of patch 4. Yes I think this will be helpful. Thanks.
