On Tue, May 18, 2021, Reiji Watanabe wrote: > > > > + if (kvm_cr0_mmu_role_changed(old_cr0, kvm_read_cr0(vcpu)) || > > > > + kvm_cr4_mmu_role_changed(old_cr4, kvm_read_cr4(vcpu))) > > > > + kvm_mmu_reset_context(vcpu); > > > > } > > > > > > I'm wondering if kvm_vcpu_reset() should call kvm_mmu_reset_context() > > > for a change in EFER.NX as well. > > > > Oooh. So there _should_ be no need. Paging has to be enabled for EFER.NX to > > be relevant, and INIT toggles CR0.PG 1=>0 if paging was enabled and so is > > guaranteed to trigger a context reset. And we do want to skip the context reset, > > e.g. INIT-SIPI-SIPI when the vCPU has paging disabled should continue using the > > same MMU. > > > > But, kvm_calc_mmu_role_common() neglects to ignore NX if CR0.PG=0, and so the > > MMU role will be stale if INIT clears EFER.NX without forcing a context reset. > > However, that's benign from a functionality perspective because the context > > itself correctly incorporates CR0.PG, it's only the role that's borked. I.e. > > KVM will fail to reuse a page/context due to the spurious role.nxe, but the > > permission checks are always be correct. > > > > I'll add a comment here and send a patch to fix the role calculation. > > Thank you so much for the explanation ! > I understand your intention and why it would be benign. > > Then, I'm wondering if kvm_cr4_mmu_role_changed() needs to be > called here. Looking at the Intel SDM, in my understanding, > all the bits kvm_cr4_mmu_role_changed() checks are relevant > only if paging is enabled. (Or is my understanding incorrect ??) Duh, yes. And it goes even beyond that, CR0.WP is only relevant if CR0.PG=1, i.e. INIT with CR0.PG=0 and CR0.WP=1 will incorrectly trigger a MMU reset with the current logic. Sadly, simply omitting the CR4 check puts us in an awkward situation where, due to the MMU role CR4 calculations not accounting for CR0.PG=0, KVM will run with a stale role. The other consideration is that kvm_post_set_cr4() and kvm_post_set_cr0() should also skip kvm_mmu_reset_context() if CR0.PG=0, but again that requires fixing the role calculations first (or at the same time). I think I'll throw in those cleanups to the beginning of this series. The result is going to be disgustingly long, but I really don't want to introduce code that knowingly leaves KVM in an inconsistent state, nor do I want to add useless checks on CR4 and EFER.
