KVM: x86: Add emulation for MSR_IA32_MCx_CTL2 MSRs.

This patch adds the emulation of IA32_MCi_CTL2 registers to KVM. A
separate mci_ctl2_banks array is used to keep the existing mce_banks
register layout intact.

In Machine Check Architecture, in addition to MCG_CMCI_P, bit 30 of
the per-bank register IA32_MCi_CTL2 controls whether Corrected Machine
Check error reporting is enabled.

Signed-off-by: Jue Wang <juew@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220610171134.772566-7-juew@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

arch/x86/include/asm/kvm_host.h

@@ -826,6 +826,7 @@ struct kvm_vcpu_arch {
 	u64 mcg_ctl;
 	u64 mcg_ext_ctl;
 	u64 *mce_banks;
+	u64 *mci_ctl2_banks;
 
 	/* Cache MMIO info */
 	u64 mmio_gva;

arch/x86/kvm/x86.c

@@ -3191,6 +3191,16 @@ static void kvmclock_sync_fn(struct work_struct *work)
 					KVMCLOCK_SYNC_PERIOD);
 }
 
+/* These helpers are safe iff @msr is known to be an MCx bank MSR. */
+static bool is_mci_control_msr(u32 msr)
+{
+	return (msr & 3) == 0;
+}
+static bool is_mci_status_msr(u32 msr)
+{
+	return (msr & 3) == 1;
+}
+
 /*
  * On AMD, HWCR[McStatusWrEn] controls whether setting MCi_STATUS results in #GP.
  */
@@ -3209,6 +3219,7 @@ static int set_msr_mce(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 	unsigned bank_num = mcg_cap & 0xff;
 	u32 msr = msr_info->index;
 	u64 data = msr_info->data;
+	u32 offset, last_msr;
 
 	switch (msr) {
 	case MSR_IA32_MCG_STATUS:
@@ -3222,35 +3233,53 @@ static int set_msr_mce(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 			return 1;
 		vcpu->arch.mcg_ctl = data;
 		break;
-	default:
+	case MSR_IA32_MC0_CTL2 ... MSR_IA32_MCx_CTL2(KVM_MAX_MCE_BANKS) - 1:
-		if (msr >= MSR_IA32_MC0_CTL &&
+		last_msr = MSR_IA32_MCx_CTL2(bank_num) - 1;
-		    msr < MSR_IA32_MCx_CTL(bank_num)) {
+		if (msr > last_msr)
-			u32 offset = array_index_nospec(
+			return 1;
-				msr - MSR_IA32_MC0_CTL,
-				MSR_IA32_MCx_CTL(bank_num) - MSR_IA32_MC0_CTL);
 
-			/* only 0 or all 1s can be written to IA32_MCi_CTL
+		if (!(mcg_cap & MCG_CMCI_P) && (data || !msr_info->host_initiated))
-			 * some Linux kernels though clear bit 10 in bank 4 to
+			return 1;
-			 * workaround a BIOS/GART TBL issue on AMD K8s, ignore
+		/* An attempt to write a 1 to a reserved bit raises #GP */
-			 * this to avoid an uncatched #GP in the guest.
+		if (data & ~(MCI_CTL2_CMCI_EN | MCI_CTL2_CMCI_THRESHOLD_MASK))
+			return 1;
+		offset = array_index_nospec(msr - MSR_IA32_MC0_CTL2,
+					    last_msr + 1 - MSR_IA32_MC0_CTL2);
+		vcpu->arch.mci_ctl2_banks[offset] = data;
+		break;
+	case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1:
+		last_msr = MSR_IA32_MCx_CTL(bank_num) - 1;
+		if (msr > last_msr)
+			return 1;
+
+		/*
+		 * Only 0 or all 1s can be written to IA32_MCi_CTL, all other
+		 * values are architecturally undefined.  But, some Linux
+		 * kernels clear bit 10 in bank 4 to workaround a BIOS/GART TLB
+		 * issue on AMD K8s, allow bit 10 to be clear when setting all
+		 * other bits in order to avoid an uncaught #GP in the guest.
 			 *
 			 * UNIXWARE clears bit 0 of MC1_CTL to ignore
 			 * correctable, single-bit ECC data errors.
-			 */
+		 */
-			if ((offset & 0x3) == 0 &&
+		if (is_mci_control_msr(msr) &&
-			    data != 0 && (data | (1 << 10) | 1) != ~(u64)0)
+		    data != 0 && (data | (1 << 10) | 1) != ~(u64)0)
-				return -1;
+			return 1;
 
-			/* MCi_STATUS */
+		/*
-			if (!msr_info->host_initiated &&
+		 * All CPUs allow writing 0 to MCi_STATUS MSRs to clear the MSR.
-			    (offset & 0x3) == 1 && data != 0) {
+		 * AMD-based CPUs allow non-zero values, but if and only if
-				if (!can_set_mci_status(vcpu))
+		 * HWCR[McStatusWrEn] is set.
-					return -1;
+		 */
-			}
+		if (!msr_info->host_initiated && is_mci_status_msr(msr) &&
+		    data != 0 && !can_set_mci_status(vcpu))
+			return 1;
 
-			vcpu->arch.mce_banks[offset] = data;
+		offset = array_index_nospec(msr - MSR_IA32_MC0_CTL,
-			break;
+					    last_msr + 1 - MSR_IA32_MC0_CTL);
-		}
+		vcpu->arch.mce_banks[offset] = data;
+		break;
+	default:
 		return 1;
 	}
 	return 0;
@@ -3534,7 +3563,8 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 			return 1;
 		}
 		break;
-	case 0x200 ... 0x2ff:
+	case 0x200 ... MSR_IA32_MC0_CTL2 - 1:
+	case MSR_IA32_MCx_CTL2(KVM_MAX_MCE_BANKS) ... 0x2ff:
 		return kvm_mtrr_set_msr(vcpu, msr, data);
 	case MSR_IA32_APICBASE:
 		return kvm_set_apic_base(vcpu, msr_info);
@@ -3704,6 +3734,7 @@ 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_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1:
+	case MSR_IA32_MC0_CTL2 ... MSR_IA32_MCx_CTL2(KVM_MAX_MCE_BANKS) - 1:
 		return set_msr_mce(vcpu, msr_info);
 
 	case MSR_K7_PERFCTR0 ... MSR_K7_PERFCTR3:
@@ -3819,6 +3850,7 @@ static int get_msr_mce(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata, bool host)
 	u64 data;
 	u64 mcg_cap = vcpu->arch.mcg_cap;
 	unsigned bank_num = mcg_cap & 0xff;
+	u32 offset, last_msr;
 
 	switch (msr) {
 	case MSR_IA32_P5_MC_ADDR:
@@ -3836,16 +3868,27 @@ static int get_msr_mce(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata, bool host)
 	case MSR_IA32_MCG_STATUS:
 		data = vcpu->arch.mcg_status;
 		break;
-	default:
+	case MSR_IA32_MC0_CTL2 ... MSR_IA32_MCx_CTL2(KVM_MAX_MCE_BANKS) - 1:
-		if (msr >= MSR_IA32_MC0_CTL &&
+		last_msr = MSR_IA32_MCx_CTL2(bank_num) - 1;
-		    msr < MSR_IA32_MCx_CTL(bank_num)) {
+		if (msr > last_msr)
-			u32 offset = array_index_nospec(
+			return 1;
-				msr - MSR_IA32_MC0_CTL,
-				MSR_IA32_MCx_CTL(bank_num) - MSR_IA32_MC0_CTL);
 
-			data = vcpu->arch.mce_banks[offset];
+		if (!(mcg_cap & MCG_CMCI_P) && !host)
-			break;
+			return 1;
-		}
+		offset = array_index_nospec(msr - MSR_IA32_MC0_CTL2,
+					    last_msr + 1 - MSR_IA32_MC0_CTL2);
+		data = vcpu->arch.mci_ctl2_banks[offset];
+		break;
+	case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1:
+		last_msr = MSR_IA32_MCx_CTL(bank_num) - 1;
+		if (msr > last_msr)
+			return 1;
+
+		offset = array_index_nospec(msr - MSR_IA32_MC0_CTL,
+					    last_msr + 1 - MSR_IA32_MC0_CTL);
+		data = vcpu->arch.mce_banks[offset];
+		break;
+	default:
 		return 1;
 	}
 	*pdata = data;
@@ -3949,7 +3992,8 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 		break;
 	}
 	case MSR_MTRRcap:
-	case 0x200 ... 0x2ff:
+	case 0x200 ... MSR_IA32_MC0_CTL2 - 1:
+	case MSR_IA32_MCx_CTL2(KVM_MAX_MCE_BANKS) ... 0x2ff:
 		return kvm_mtrr_get_msr(vcpu, msr_info->index, &msr_info->data);
 	case 0xcd: /* fsb frequency */
 		msr_info->data = 3;
@@ -4065,6 +4109,7 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 	case MSR_IA32_MCG_CTL:
 	case MSR_IA32_MCG_STATUS:
 	case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1:
+	case MSR_IA32_MC0_CTL2 ... MSR_IA32_MCx_CTL2(KVM_MAX_MCE_BANKS) - 1:
 		return get_msr_mce(vcpu, msr_info->index, &msr_info->data,
 				   msr_info->host_initiated);
 	case MSR_IA32_XSS:
@@ -4842,9 +4887,12 @@ static int kvm_vcpu_ioctl_x86_setup_mce(struct kvm_vcpu *vcpu,
 	/* Init IA32_MCG_CTL to all 1s */
 	if (mcg_cap & MCG_CTL_P)
 		vcpu->arch.mcg_ctl = ~(u64)0;
-	/* Init IA32_MCi_CTL to all 1s */
+	/* Init IA32_MCi_CTL to all 1s, IA32_MCi_CTL2 to all 0s */
-	for (bank = 0; bank < bank_num; bank++)
+	for (bank = 0; bank < bank_num; bank++) {
 		vcpu->arch.mce_banks[bank*4] = ~(u64)0;
+		if (mcg_cap & MCG_CMCI_P)
+			vcpu->arch.mci_ctl2_banks[bank] = 0;
+	}
 	vcpu->arch.apic->nr_lvt_entries =
 		KVM_APIC_MAX_NR_LVT_ENTRIES - !(mcg_cap & MCG_CMCI_P);
 
@@ -11449,7 +11497,9 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu)
 
 	vcpu->arch.mce_banks = kcalloc(KVM_MAX_MCE_BANKS * 4, sizeof(u64),
 				       GFP_KERNEL_ACCOUNT);
-	if (!vcpu->arch.mce_banks)
+	vcpu->arch.mci_ctl2_banks = kcalloc(KVM_MAX_MCE_BANKS, sizeof(u64),
+					    GFP_KERNEL_ACCOUNT);
+	if (!vcpu->arch.mce_banks || !vcpu->arch.mci_ctl2_banks)
 		goto fail_free_pio_data;
 	vcpu->arch.mcg_cap = KVM_MAX_MCE_BANKS;
 
@@ -11503,6 +11553,7 @@ free_wbinvd_dirty_mask:
 	free_cpumask_var(vcpu->arch.wbinvd_dirty_mask);
 fail_free_mce_banks:
 	kfree(vcpu->arch.mce_banks);
+	kfree(vcpu->arch.mci_ctl2_banks);
 fail_free_pio_data:
 	free_page((unsigned long)vcpu->arch.pio_data);
 fail_free_lapic:
@@ -11548,6 +11599,7 @@ void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
 	kvm_hv_vcpu_uninit(vcpu);
 	kvm_pmu_destroy(vcpu);
 	kfree(vcpu->arch.mce_banks);
+	kfree(vcpu->arch.mci_ctl2_banks);
 	kvm_free_lapic(vcpu);
 	idx = srcu_read_lock(&vcpu->kvm->srcu);
 	kvm_mmu_destroy(vcpu);