KVM: VMX: VMXON/VMXOFF usage changes
authorDongxiao Xu <dongxiao.xu@intel.com>
Tue, 11 May 2010 10:29:48 +0000 (18:29 +0800)
committerAvi Kivity <avi@redhat.com>
Sun, 1 Aug 2010 07:35:43 +0000 (10:35 +0300)
SDM suggests VMXON should be called before VMPTRLD, and VMXOFF
should be called after doing VMCLEAR.

Therefore in vmm coexistence case, we should firstly call VMXON
before any VMCS operation, and then call VMXOFF after the
operation is done.

Signed-off-by: Dongxiao Xu <dongxiao.xu@intel.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
arch/x86/kvm/vmx.c

index 9529bff042626f0eeb5e640a3a85413fa1199641..b8aac4e989036772015339d8e29a9ee4f0736d72 100644 (file)
@@ -176,6 +176,8 @@ static inline struct vcpu_vmx *to_vmx(struct kvm_vcpu *vcpu)
 
 static int init_rmode(struct kvm *kvm);
 static u64 construct_eptp(unsigned long root_hpa);
+static void kvm_cpu_vmxon(u64 addr);
+static void kvm_cpu_vmxoff(void);
 
 static DEFINE_PER_CPU(struct vmcs *, vmxarea);
 static DEFINE_PER_CPU(struct vmcs *, current_vmcs);
@@ -847,8 +849,11 @@ static void vmx_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 {
        struct vcpu_vmx *vmx = to_vmx(vcpu);
        u64 tsc_this, delta, new_offset;
+       u64 phys_addr = __pa(per_cpu(vmxarea, cpu));
 
-       if (vmm_exclusive && vcpu->cpu != cpu)
+       if (!vmm_exclusive)
+               kvm_cpu_vmxon(phys_addr);
+       else if (vcpu->cpu != cpu)
                vcpu_clear(vmx);
 
        if (per_cpu(current_vmcs, cpu) != vmx->vmcs) {
@@ -894,8 +899,10 @@ static void vmx_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 static void vmx_vcpu_put(struct kvm_vcpu *vcpu)
 {
        __vmx_load_host_state(to_vmx(vcpu));
-       if (!vmm_exclusive)
+       if (!vmm_exclusive) {
                __vcpu_clear(to_vmx(vcpu));
+               kvm_cpu_vmxoff();
+       }
 }
 
 static void vmx_fpu_activate(struct kvm_vcpu *vcpu)
@@ -1327,9 +1334,11 @@ static int hardware_enable(void *garbage)
                wrmsrl(MSR_IA32_FEATURE_CONTROL, old | test_bits);
        }
        write_cr4(read_cr4() | X86_CR4_VMXE); /* FIXME: not cpu hotplug safe */
-       kvm_cpu_vmxon(phys_addr);
 
-       ept_sync_global();
+       if (vmm_exclusive) {
+               kvm_cpu_vmxon(phys_addr);
+               ept_sync_global();
+       }
 
        return 0;
 }
@@ -1355,8 +1364,10 @@ static void kvm_cpu_vmxoff(void)
 
 static void hardware_disable(void *garbage)
 {
-       vmclear_local_vcpus();
-       kvm_cpu_vmxoff();
+       if (vmm_exclusive) {
+               vmclear_local_vcpus();
+               kvm_cpu_vmxoff();
+       }
        write_cr4(read_cr4() & ~X86_CR4_VMXE);
 }
 
@@ -3991,6 +4002,19 @@ static void vmx_free_vcpu(struct kvm_vcpu *vcpu)
        kmem_cache_free(kvm_vcpu_cache, vmx);
 }
 
+static inline void vmcs_init(struct vmcs *vmcs)
+{
+       u64 phys_addr = __pa(per_cpu(vmxarea, raw_smp_processor_id()));
+
+       if (!vmm_exclusive)
+               kvm_cpu_vmxon(phys_addr);
+
+       vmcs_clear(vmcs);
+
+       if (!vmm_exclusive)
+               kvm_cpu_vmxoff();
+}
+
 static struct kvm_vcpu *vmx_create_vcpu(struct kvm *kvm, unsigned int id)
 {
        int err;
@@ -4016,7 +4040,7 @@ static struct kvm_vcpu *vmx_create_vcpu(struct kvm *kvm, unsigned int id)
        if (!vmx->vmcs)
                goto free_msrs;
 
-       vmcs_clear(vmx->vmcs);
+       vmcs_init(vmx->vmcs);
 
        cpu = get_cpu();
        vmx_vcpu_load(&vmx->vcpu, cpu);