The simplistic clts implementation has a couple of flaws: - kvm_read_cr0_bits() is temporarily unsynchronized when vcpu->arch.cr0 changes - if the fpu is active, we need to clear GUEST_CR0.TS, not just CR_READ_SHADOW.TS, so that we don't send the guest an unexpected #NM. Fix by replacing custom logic with a call to vmx_set_cr0(), which does the right thing, albeit less efficiently. Signed-off-by: Avi Kivity <avi@xxxxxxxxxx> --- arch/x86/kvm/vmx.c | 3 +-- 1 files changed, 1 insertions(+), 2 deletions(-) diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c index 9f56110..d27ff6a 100644 --- a/arch/x86/kvm/vmx.c +++ b/arch/x86/kvm/vmx.c @@ -3007,8 +3007,7 @@ static int handle_cr(struct kvm_vcpu *vcpu) }; break; case 2: /* clts */ - vcpu->arch.cr0 &= ~X86_CR0_TS; - vmcs_writel(CR0_READ_SHADOW, kvm_read_cr0(vcpu)); + vmx_set_cr0(vcpu, kvm_read_cr0_bits(vcpu, ~X86_CR0_TS)); trace_kvm_cr_write(0, kvm_read_cr0(vcpu)); skip_emulated_instruction(vcpu); return 1; -- 1.6.5.3 -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
