| Message ID | 20201203231120.27307-1-rick.p.edgecombe@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | kvm: x86/mmu: Use cpuid to determine max gfn | expand |
On Thu, Dec 03, 2020, Rick Edgecombe wrote: > In the TDP MMU, use shadow_phys_bits to dermine the maximum possible GFN > mapped in the guest for zapping operations. boot_cpu_data.x86_phys_bits > may be reduced in the case of HW features that steal HPA bits for other > purposes. However, this doesn't necessarily reduce GPA space that can be > accessed via TDP. So zap based on a maximum gfn calculated with MAXPHYADDR > retrieved from CPUID. This is already stored in shadow_phys_bits, so use > it instead of x86_phys_bits. > > Fixes: faaf05b00aec ("kvm: x86/mmu: Support zapping SPTEs in the TDP MMU") > Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com> Dang, in hindsight it'd be nice if KVM_CAP_SMALLER_MAXPHYADDR allowed explicitly setting the max MAXPHYADDR for an entire VM instead of being a simple toggle. E.g. TDX and SEV-ES likely could also make use of "what's this VM's max GPA?". Reviewed-by: Sean Christopherson <seanjc@google.com>
On 04/12/20 01:43, Sean Christopherson wrote: > On Thu, Dec 03, 2020, Rick Edgecombe wrote: >> In the TDP MMU, use shadow_phys_bits to dermine the maximum possible GFN >> mapped in the guest for zapping operations. boot_cpu_data.x86_phys_bits >> may be reduced in the case of HW features that steal HPA bits for other >> purposes. However, this doesn't necessarily reduce GPA space that can be >> accessed via TDP. So zap based on a maximum gfn calculated with MAXPHYADDR >> retrieved from CPUID. This is already stored in shadow_phys_bits, so use >> it instead of x86_phys_bits. >> >> Fixes: faaf05b00aec ("kvm: x86/mmu: Support zapping SPTEs in the TDP MMU") >> Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com> > > Dang, in hindsight it'd be nice if KVM_CAP_SMALLER_MAXPHYADDR allowed explicitly > setting the max MAXPHYADDR for an entire VM instead of being a simple toggle. > E.g. TDX and SEV-ES likely could also make use of "what's this VM's max GPA?". > > Reviewed-by: Sean Christopherson <seanjc@google.com> > Queued, thanks. Paolo
diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c index ff28a5c6abd6..84c8f06bec26 100644 --- a/arch/x86/kvm/mmu/tdp_mmu.c +++ b/arch/x86/kvm/mmu/tdp_mmu.c @@ -66,7 +66,7 @@ static bool zap_gfn_range(struct kvm *kvm, struct kvm_mmu_page *root, void kvm_tdp_mmu_free_root(struct kvm *kvm, struct kvm_mmu_page *root) { - gfn_t max_gfn = 1ULL << (boot_cpu_data.x86_phys_bits - PAGE_SHIFT); + gfn_t max_gfn = 1ULL << (shadow_phys_bits - PAGE_SHIFT); lockdep_assert_held(&kvm->mmu_lock); @@ -456,7 +456,7 @@ bool kvm_tdp_mmu_zap_gfn_range(struct kvm *kvm, gfn_t start, gfn_t end) void kvm_tdp_mmu_zap_all(struct kvm *kvm) { - gfn_t max_gfn = 1ULL << (boot_cpu_data.x86_phys_bits - PAGE_SHIFT); + gfn_t max_gfn = 1ULL << (shadow_phys_bits - PAGE_SHIFT); bool flush; flush = kvm_tdp_mmu_zap_gfn_range(kvm, 0, max_gfn);
In the TDP MMU, use shadow_phys_bits to dermine the maximum possible GFN mapped in the guest for zapping operations. boot_cpu_data.x86_phys_bits may be reduced in the case of HW features that steal HPA bits for other purposes. However, this doesn't necessarily reduce GPA space that can be accessed via TDP. So zap based on a maximum gfn calculated with MAXPHYADDR retrieved from CPUID. This is already stored in shadow_phys_bits, so use it instead of x86_phys_bits. Fixes: faaf05b00aec ("kvm: x86/mmu: Support zapping SPTEs in the TDP MMU") Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com> --- arch/x86/kvm/mmu/tdp_mmu.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)