commit 37486135d3a7b03acc7755b63627a130437f066a upstream.

In 5.4.y only, vcpu->arch.host_pkru is being set on every run thru of vcpu_enter_guest, when it really only needs to be set on load. As a result, we're doing a rdpkru on supported CPUs on every iteration of vcpu_enter_guest even though the value never changes.

Mainline and 5.10.y already has host_pkru being initialized in kvm_arch_vcpu_load. This change is 5.4.y specific and moves host_pkru save to kvm_arch_vcpu_load.

Fixes: 99e392a4979b ("KVM: x86: Fix pkru save/restore when guest CR4.PKE=0, move it to x86.c") Cc: stable@vger.kernel.org # 5.4.y Cc: Babu Moger babu.moger@amd.com Signed-off-by: Jon Kohler jon@nutanix.com --- arch/x86/kvm/x86.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 153659e8f403..1f7521752a94 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -3507,6 +3507,9 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)

kvm_x86_ops->vcpu_load(vcpu, cpu);

+ /* Save host pkru register if supported */ + vcpu->arch.host_pkru = read_pkru(); + /* Apply any externally detected TSC adjustments (due to suspend) */ if (unlikely(vcpu->arch.tsc_offset_adjustment)) { adjust_tsc_offset_host(vcpu, vcpu->arch.tsc_offset_adjustment); @@ -8253,9 +8256,6 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu) trace_kvm_entry(vcpu->vcpu_id); guest_enter_irqoff();

- /* Save host pkru register if supported */ - vcpu->arch.host_pkru = read_pkru(); - fpregs_assert_state_consistent(); if (test_thread_flag(TIF_NEED_FPU_LOAD)) switch_fpu_return();