[Qemu-devel] [Patch V0] This patch adds some support required for KVM in order to support LMCE.

[Ashok Raj]

- Add support for MSR_IA32_MCG_EXT_CTL
- Add MCG_LMCE_P to KVM_MCE_CAP_SUPPORTED
- Changes to IA32_FEATURE_CONTROL, allow this MSR to be defined just not for
  nested VMM, but now its required for Local MCE.

Reviewed-by: Andi Kleen <address@hidden>
Reviewed-by: Tony Luck <address@hidden>
Tested-by: Gong Chen <address@hidden>
Signed-off-by: Ashok Raj <address@hidden>
---
 arch/x86/include/asm/kvm_host.h |  1 +
 arch/x86/kvm/vmx.c              | 26 +++++++++++++++++++++-----
 arch/x86/kvm/x86.c              | 17 ++++++++++++++++-
 3 files changed, 38 insertions(+), 6 deletions(-)

diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 30cfd64..6940141 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -525,6 +525,7 @@ struct kvm_vcpu_arch {
        u64 mcg_cap;
        u64 mcg_status;
        u64 mcg_ctl;
+       u64 mcg_ext_ctl;
        u64 *mce_banks;
 
        /* Cache MMIO info */
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index 87acc52..c2ce9f4 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -2747,6 +2747,20 @@ static int vmx_get_vmx_msr(struct kvm_vcpu *vcpu, u32 
msr_index, u64 *pdata)
        return 0;
 }
 
+bool can_feature_control_exist(struct kvm_vcpu *vcpu)
+{
+       /*
+        * There are some features that require BIOS enabling.
+        * In such cases BIOS is supposed to set this bit and indicate
+        * the feature is enabled and available to the OS.
+        * Local Machine Check Exception (LMCE) is one such feature.
+        */
+       if (vcpu->arch.mcg_cap & MCG_LMCE_P)
+               return true;
+
+       return (nested_vmx_allowed(vcpu));
+}
+
 /*
  * Reads an msr value (of 'msr_index') into 'pdata'.
  * Returns 0 on success, non-0 otherwise.
@@ -2789,9 +2803,11 @@ static int vmx_get_msr(struct kvm_vcpu *vcpu, struct 
msr_data *msr_info)
                msr_info->data = vmcs_read64(GUEST_BNDCFGS);
                break;
        case MSR_IA32_FEATURE_CONTROL:
-               if (!nested_vmx_allowed(vcpu))
+               if (can_feature_control_exist(vcpu))
+                       msr_info->data =
+                       to_vmx(vcpu)->nested.msr_ia32_feature_control;
+               else
                        return 1;
-               msr_info->data = to_vmx(vcpu)->nested.msr_ia32_feature_control;
                break;
        case MSR_IA32_VMX_BASIC ... MSR_IA32_VMX_VMFUNC:
                if (!nested_vmx_allowed(vcpu))
@@ -2882,9 +2898,9 @@ static int vmx_set_msr(struct kvm_vcpu *vcpu, struct 
msr_data *msr_info)
                ret = kvm_set_msr_common(vcpu, msr_info);
                break;
        case MSR_IA32_FEATURE_CONTROL:
-               if (!nested_vmx_allowed(vcpu) ||
-                   (to_vmx(vcpu)->nested.msr_ia32_feature_control &
-                    FEATURE_CONTROL_LOCKED && !msr_info->host_initiated))
+               if ((can_feature_control_exist(vcpu) == false) ||
+                   ((to_vmx(vcpu)->nested.msr_ia32_feature_control &
+                    FEATURE_CONTROL_LOCKED) && !msr_info->host_initiated))
                        return 1;
                vmx->nested.msr_ia32_feature_control = data;
                if (msr_info->host_initiated && data == 0)
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 00462bd..0da3871 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -70,7 +70,7 @@
 
 #define MAX_IO_MSRS 256
 #define KVM_MAX_MCE_BANKS 32
-#define KVM_MCE_CAP_SUPPORTED (MCG_CTL_P | MCG_SER_P)
+#define KVM_MCE_CAP_SUPPORTED (MCG_CTL_P | MCG_SER_P | MCG_LMCE_P)
 
 #define emul_to_vcpu(ctxt) \
        container_of(ctxt, struct kvm_vcpu, arch.emulate_ctxt)
@@ -974,6 +974,7 @@ static u32 emulated_msrs[] = {
        MSR_IA32_MISC_ENABLE,
        MSR_IA32_MCG_STATUS,
        MSR_IA32_MCG_CTL,
+       MSR_IA32_MCG_EXT_CTL,
        MSR_IA32_SMBASE,
 };
 
@@ -1913,6 +1914,13 @@ static int set_msr_mce(struct kvm_vcpu *vcpu, u32 msr, 
u64 data)
                        return -1;
                vcpu->arch.mcg_ctl = data;
                break;
+       case MSR_IA32_MCG_EXT_CTL:
+               if (!(mcg_cap & MCG_LMCE_P))
+                       return 1;
+               if (data != 0 && data != 0x1)
+                       return -1;
+               vcpu->arch.mcg_ext_ctl = data;
+               break;
        default:
                if (msr >= MSR_IA32_MC0_CTL &&
                    msr < MSR_IA32_MCx_CTL(bank_num)) {
@@ -2170,6 +2178,7 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct 
msr_data *msr_info)
 
        case MSR_IA32_MCG_CTL:
        case MSR_IA32_MCG_STATUS:
+       case MSR_IA32_MCG_EXT_CTL:
        case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1:
                return set_msr_mce(vcpu, msr, data);
 
@@ -2266,6 +2275,11 @@ static int get_msr_mce(struct kvm_vcpu *vcpu, u32 msr, 
u64 *pdata)
                        return 1;
                data = vcpu->arch.mcg_ctl;
                break;
+       case MSR_IA32_MCG_EXT_CTL:
+               if (!(mcg_cap & MCG_LMCE_P))
+                       return 1;
+               data = vcpu->arch.mcg_ext_ctl;
+               break;
        case MSR_IA32_MCG_STATUS:
                data = vcpu->arch.mcg_status;
                break;
@@ -2384,6 +2398,7 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct 
msr_data *msr_info)
        case MSR_IA32_P5_MC_TYPE:
        case MSR_IA32_MCG_CAP:
        case MSR_IA32_MCG_CTL:
+       case MSR_IA32_MCG_EXT_CTL:
        case MSR_IA32_MCG_STATUS:
        case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1:
                return get_msr_mce(vcpu, msr_info->index, &msr_info->data);
-- 
2.4.3