KVM: nVMX: Enable VMFUNC for the L1 hypervisor
authorBandan Das <bsd@redhat.com>
Thu, 3 Aug 2017 19:54:42 +0000 (15:54 -0400)
committerPaolo Bonzini <pbonzini@redhat.com>
Mon, 7 Aug 2017 13:29:21 +0000 (15:29 +0200)
Expose VMFUNC in MSRs and VMCS fields. No actual VMFUNCs are enabled.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Bandan Das <bsd@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
arch/x86/kvm/vmx.c

index f2063c9ee275d0d0fafebae806087af984857ced..e04b2a8dd9b0066468f753f4f0cfafdd6827fa17 100644 (file)
@@ -243,6 +243,7 @@ struct __packed vmcs12 {
        u64 virtual_apic_page_addr;
        u64 apic_access_addr;
        u64 posted_intr_desc_addr;
+       u64 vm_function_control;
        u64 ept_pointer;
        u64 eoi_exit_bitmap0;
        u64 eoi_exit_bitmap1;
@@ -481,6 +482,7 @@ struct nested_vmx {
        u64 nested_vmx_cr4_fixed0;
        u64 nested_vmx_cr4_fixed1;
        u64 nested_vmx_vmcs_enum;
+       u64 nested_vmx_vmfunc_controls;
 };
 
 #define POSTED_INTR_ON  0
@@ -763,6 +765,7 @@ static const unsigned short vmcs_field_to_offset_table[] = {
        FIELD64(VIRTUAL_APIC_PAGE_ADDR, virtual_apic_page_addr),
        FIELD64(APIC_ACCESS_ADDR, apic_access_addr),
        FIELD64(POSTED_INTR_DESC_ADDR, posted_intr_desc_addr),
+       FIELD64(VM_FUNCTION_CONTROL, vm_function_control),
        FIELD64(EPT_POINTER, ept_pointer),
        FIELD64(EOI_EXIT_BITMAP0, eoi_exit_bitmap0),
        FIELD64(EOI_EXIT_BITMAP1, eoi_exit_bitmap1),
@@ -1380,6 +1383,11 @@ static inline bool nested_cpu_has_posted_intr(struct vmcs12 *vmcs12)
        return vmcs12->pin_based_vm_exec_control & PIN_BASED_POSTED_INTR;
 }
 
+static inline bool nested_cpu_has_vmfunc(struct vmcs12 *vmcs12)
+{
+       return nested_cpu_has2(vmcs12, SECONDARY_EXEC_ENABLE_VMFUNC);
+}
+
 static inline bool is_nmi(u32 intr_info)
 {
        return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VALID_MASK))
@@ -2828,6 +2836,12 @@ static void nested_vmx_setup_ctls_msrs(struct vcpu_vmx *vmx)
        } else
                vmx->nested.nested_vmx_ept_caps = 0;
 
+       if (cpu_has_vmx_vmfunc()) {
+               vmx->nested.nested_vmx_secondary_ctls_high |=
+                       SECONDARY_EXEC_ENABLE_VMFUNC;
+               vmx->nested.nested_vmx_vmfunc_controls = 0;
+       }
+
        /*
         * Old versions of KVM use the single-context version without
         * checking for support, so declare that it is supported even
@@ -3197,6 +3211,9 @@ static int vmx_get_vmx_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata)
                *pdata = vmx->nested.nested_vmx_ept_caps |
                        ((u64)vmx->nested.nested_vmx_vpid_caps << 32);
                break;
+       case MSR_IA32_VMX_VMFUNC:
+               *pdata = vmx->nested.nested_vmx_vmfunc_controls;
+               break;
        default:
                return 1;
        }
@@ -7847,7 +7864,29 @@ static int handle_preemption_timer(struct kvm_vcpu *vcpu)
 
 static int handle_vmfunc(struct kvm_vcpu *vcpu)
 {
-       kvm_queue_exception(vcpu, UD_VECTOR);
+       struct vcpu_vmx *vmx = to_vmx(vcpu);
+       struct vmcs12 *vmcs12;
+       u32 function = vcpu->arch.regs[VCPU_REGS_RAX];
+
+       /*
+        * VMFUNC is only supported for nested guests, but we always enable the
+        * secondary control for simplicity; for non-nested mode, fake that we
+        * didn't by injecting #UD.
+        */
+       if (!is_guest_mode(vcpu)) {
+               kvm_queue_exception(vcpu, UD_VECTOR);
+               return 1;
+       }
+
+       vmcs12 = get_vmcs12(vcpu);
+       if ((vmcs12->vm_function_control & (1 << function)) == 0)
+               goto fail;
+       WARN_ONCE(1, "VMCS12 VM function control should have been zero");
+
+fail:
+       nested_vmx_vmexit(vcpu, vmx->exit_reason,
+                         vmcs_read32(VM_EXIT_INTR_INFO),
+                         vmcs_readl(EXIT_QUALIFICATION));
        return 1;
 }
 
@@ -10220,7 +10259,8 @@ static int prepare_vmcs02(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
                                  SECONDARY_EXEC_ENABLE_INVPCID |
                                  SECONDARY_EXEC_RDTSCP |
                                  SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY |
-                                 SECONDARY_EXEC_APIC_REGISTER_VIRT);
+                                 SECONDARY_EXEC_APIC_REGISTER_VIRT |
+                                 SECONDARY_EXEC_ENABLE_VMFUNC);
                if (nested_cpu_has(vmcs12,
                                   CPU_BASED_ACTIVATE_SECONDARY_CONTROLS)) {
                        vmcs12_exec_ctrl = vmcs12->secondary_vm_exec_control &
@@ -10228,6 +10268,10 @@ static int prepare_vmcs02(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
                        exec_control |= vmcs12_exec_ctrl;
                }
 
+               /* All VMFUNCs are currently emulated through L0 vmexits.  */
+               if (exec_control & SECONDARY_EXEC_ENABLE_VMFUNC)
+                       vmcs_write64(VM_FUNCTION_CONTROL, 0);
+
                if (exec_control & SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY) {
                        vmcs_write64(EOI_EXIT_BITMAP0,
                                vmcs12->eoi_exit_bitmap0);
@@ -10480,6 +10524,11 @@ static int check_vmentry_prereqs(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12)
                                vmx->nested.nested_vmx_entry_ctls_high))
                return VMXERR_ENTRY_INVALID_CONTROL_FIELD;
 
+       if (nested_cpu_has_vmfunc(vmcs12) &&
+           (vmcs12->vm_function_control &
+            ~vmx->nested.nested_vmx_vmfunc_controls))
+               return VMXERR_ENTRY_INVALID_CONTROL_FIELD;
+
        if (vmcs12->cr3_target_count > nested_cpu_vmx_misc_cr3_count(vcpu))
                return VMXERR_ENTRY_INVALID_CONTROL_FIELD;