KVM: nVMX: Synchronize VMCS12 content with the shadow vmcs
authorAbel Gordon <abelg@il.ibm.com>
Thu, 18 Apr 2013 11:39:25 +0000 (14:39 +0300)
committerGleb Natapov <gleb@redhat.com>
Mon, 22 Apr 2013 07:52:45 +0000 (10:52 +0300)
Synchronize between the VMCS12 software controlled structure and the
processor-specific shadow vmcs

Signed-off-by: Abel Gordon <abelg@il.ibm.com>
Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
arch/x86/kvm/vmx.c

index 8dc59aa..c5baecc 100644 (file)
@@ -356,6 +356,11 @@ struct nested_vmx {
        struct page *current_vmcs12_page;
        struct vmcs12 *current_vmcs12;
        struct vmcs *current_shadow_vmcs;
+       /*
+        * Indicates if the shadow vmcs must be updated with the
+        * data hold by vmcs12
+        */
+       bool sync_shadow_vmcs;
 
        /* vmcs02_list cache of VMCSs recently used to run L2 guests */
        struct list_head vmcs02_pool;
@@ -5611,6 +5616,14 @@ static int nested_vmx_check_permission(struct kvm_vcpu *vcpu)
 
 static inline void nested_release_vmcs12(struct vcpu_vmx *vmx)
 {
+       if (enable_shadow_vmcs) {
+               if (vmx->nested.current_vmcs12 != NULL) {
+                       /* copy to memory all shadowed fields in case
+                          they were modified */
+                       copy_shadow_to_vmcs12(vmx);
+                       vmx->nested.sync_shadow_vmcs = false;
+               }
+       }
        kunmap(vmx->nested.current_vmcs12_page);
        nested_release_page(vmx->nested.current_vmcs12_page);
 }
@@ -5739,6 +5752,10 @@ static void nested_vmx_failValid(struct kvm_vcpu *vcpu,
                            X86_EFLAGS_SF | X86_EFLAGS_OF))
                        | X86_EFLAGS_ZF);
        get_vmcs12(vcpu)->vm_instruction_error = vm_instruction_error;
+       /*
+        * We don't need to force a shadow sync because
+        * VM_INSTRUCTION_ERROR is not shadowed
+        */
 }
 
 /* Emulate the VMCLEAR instruction */
@@ -6137,6 +6154,9 @@ static int handle_vmptrld(struct kvm_vcpu *vcpu)
                vmx->nested.current_vmptr = vmptr;
                vmx->nested.current_vmcs12 = new_vmcs12;
                vmx->nested.current_vmcs12_page = page;
+               if (enable_shadow_vmcs) {
+                       vmx->nested.sync_shadow_vmcs = true;
+               }
        }
 
        nested_vmx_succeed(vcpu);
@@ -6895,6 +6915,11 @@ static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu)
        if (vmx->emulation_required)
                return;
 
+       if (vmx->nested.sync_shadow_vmcs) {
+               copy_vmcs12_to_shadow(vmx);
+               vmx->nested.sync_shadow_vmcs = false;
+       }
+
        if (test_bit(VCPU_REGS_RSP, (unsigned long *)&vcpu->arch.regs_dirty))
                vmcs_writel(GUEST_RSP, vcpu->arch.regs[VCPU_REGS_RSP]);
        if (test_bit(VCPU_REGS_RIP, (unsigned long *)&vcpu->arch.regs_dirty))
@@ -7504,6 +7529,9 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
        skip_emulated_instruction(vcpu);
        vmcs12 = get_vmcs12(vcpu);
 
+       if (enable_shadow_vmcs)
+               copy_shadow_to_vmcs12(vmx);
+
        /*
         * The nested entry process starts with enforcing various prerequisites
         * on vmcs12 as required by the Intel SDM, and act appropriately when
@@ -7950,6 +7978,8 @@ static void nested_vmx_vmexit(struct kvm_vcpu *vcpu)
                nested_vmx_failValid(vcpu, vmcs_read32(VM_INSTRUCTION_ERROR));
        } else
                nested_vmx_succeed(vcpu);
+       if (enable_shadow_vmcs)
+               vmx->nested.sync_shadow_vmcs = true;
 }
 
 /*
@@ -7967,6 +7997,8 @@ static void nested_vmx_entry_failure(struct kvm_vcpu *vcpu,
        vmcs12->vm_exit_reason = reason | VMX_EXIT_REASONS_FAILED_VMENTRY;
        vmcs12->exit_qualification = qualification;
        nested_vmx_succeed(vcpu);
+       if (enable_shadow_vmcs)
+               to_vmx(vcpu)->nested.sync_shadow_vmcs = true;
 }
 
 static int vmx_check_intercept(struct kvm_vcpu *vcpu,