On 16/06/2016 08:05, Haozhong Zhang wrote:
> From: Ashok Raj <ashok....@intel.com>
>
> On Intel platforms, this patch adds LMCE to KVM MCE supported
> capabilities and handles guest access to LMCE related MSRs.
>
> Signed-off-by: Ashok Raj <ashok....@intel.com>
> [Haozhong: macro KVM_MCE_CAP_SUPPORTED => variable kvm_mce_cap_supported
> Only enable LMCE on Intel platform
> Check MSR_IA32_FEATURE_CONTROL when handling guest
> access to MSR_IA32_MCG_EXT_CTL]
> Signed-off-by: Haozhong Zhang <haozhong.zh...@intel.com>
> ---
> arch/x86/include/asm/kvm_host.h | 5 +++++
> arch/x86/kvm/vmx.c | 36 +++++++++++++++++++++++++++++++++++-
> arch/x86/kvm/x86.c | 15 +++++++++------
> 3 files changed, 49 insertions(+), 7 deletions(-)
>
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index e0fbe7e..75defa6 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -598,6 +598,7 @@ struct kvm_vcpu_arch {
> u64 mcg_cap;
> u64 mcg_status;
> u64 mcg_ctl;
> + u64 mcg_ext_ctl;
> u64 *mce_banks;
>
> /* Cache MMIO info */
> @@ -1005,6 +1006,8 @@ struct kvm_x86_ops {
> int (*update_pi_irte)(struct kvm *kvm, unsigned int host_irq,
> uint32_t guest_irq, bool set);
> void (*apicv_post_state_restore)(struct kvm_vcpu *vcpu);
> +
> + void (*setup_mce)(struct kvm_vcpu *vcpu);
> };
>
> struct kvm_arch_async_pf {
> @@ -1077,6 +1080,8 @@ extern u8 kvm_tsc_scaling_ratio_frac_bits;
> /* maximum allowed value of TSC scaling ratio */
> extern u64 kvm_max_tsc_scaling_ratio;
>
> +extern u64 kvm_mce_cap_supported;
> +
> enum emulation_result {
> EMULATE_DONE, /* no further processing */
> EMULATE_USER_EXIT, /* kvm_run ready for userspace exit */
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index 1dc89c5..42db42e 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -638,7 +638,7 @@ static struct pi_desc *vcpu_to_pi_desc(struct kvm_vcpu
> *vcpu)
> * feature_control_valid_bits_add/del(), so it's not included here.
> */
> #define FEATURE_CONTROL_MAX_VALID_BITS \
> - FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX
> + (FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX | FEATURE_CONTROL_LMCE)
>
> static void feature_control_valid_bits_add(struct kvm_vcpu *vcpu, uint64_t
> bits)
> {
> @@ -2905,6 +2905,15 @@ static inline bool
> vmx_feature_control_msr_valid(struct kvm_vcpu *vcpu,
> return valid_bits && !(val & ~valid_bits);
> }
>
> +static inline bool vmx_mcg_ext_ctrl_msr_present(struct kvm_vcpu *vcpu,
> + bool host_initiated)
> +{
> + return (vcpu->arch.mcg_cap & MCG_LMCE_P) &&
Checking MCG_LMCE_P is unnecessary, because you cannot set
FEATURE_CONTROL_LMCE unless MCG_LMCE_P is present.
You can just inline this function in the callers, it's simpler.
> + (host_initiated ||
> + (to_vmx(vcpu)->msr_ia32_feature_control &
> + FEATURE_CONTROL_LMCE));
> +}
> +
> /*
> * Reads an msr value (of 'msr_index') into 'pdata'.
> * Returns 0 on success, non-0 otherwise.
> @@ -2946,6 +2955,12 @@ static int vmx_get_msr(struct kvm_vcpu *vcpu, struct
> msr_data *msr_info)
> return 1;
> msr_info->data = vmcs_read64(GUEST_BNDCFGS);
> break;
> + case MSR_IA32_MCG_EXT_CTL:
> + if (!vmx_mcg_ext_ctrl_msr_present(vcpu,
> + msr_info->host_initiated))
> + return 1;
> + msr_info->data = vcpu->arch.mcg_ext_ctl;
> + break;
> case MSR_IA32_FEATURE_CONTROL:
> if (!vmx_feature_control_msr_valid(vcpu, 0))
> return 1;
> @@ -3039,6 +3054,13 @@ static int vmx_set_msr(struct kvm_vcpu *vcpu, struct
> msr_data *msr_info)
> case MSR_IA32_TSC_ADJUST:
> ret = kvm_set_msr_common(vcpu, msr_info);
> break;
> + case MSR_IA32_MCG_EXT_CTL:
> + if (!vmx_mcg_ext_ctrl_msr_present(vcpu,
> + msr_info->host_initiated) ||
> + (data & ~MCG_EXT_CTL_LMCE_EN))
> + return 1;
> + vcpu->arch.mcg_ext_ctl = data;
> + break;
> case MSR_IA32_FEATURE_CONTROL:
> if (!vmx_feature_control_msr_valid(vcpu, data) ||
> (to_vmx(vcpu)->msr_ia32_feature_control &
> @@ -6433,6 +6455,8 @@ static __init int hardware_setup(void)
>
> kvm_set_posted_intr_wakeup_handler(wakeup_handler);
>
> + kvm_mce_cap_supported |= MCG_LMCE_P;
Ah, so virtual LMCE is available on all processors! This is
interesting, but it also makes it more complicated to handle in QEMU; a
new QEMU generally doesn't require a new kernel.
Eduardo, any ideas?
Thanks,
Paolo
> return alloc_kvm_area();
>
> out8:
> @@ -10950,6 +10974,14 @@ out:
> return ret;
> }
>
> +static void vmx_setup_mce(struct kvm_vcpu *vcpu)
> +{
> + if (vcpu->arch.mcg_cap & MCG_LMCE_P)
> + feature_control_valid_bits_add(vcpu, FEATURE_CONTROL_LMCE);
> + else
> + feature_control_valid_bits_del(vcpu, FEATURE_CONTROL_LMCE);
> +}
> +
> static struct kvm_x86_ops vmx_x86_ops = {
> .cpu_has_kvm_support = cpu_has_kvm_support,
> .disabled_by_bios = vmx_disabled_by_bios,
> @@ -11074,6 +11106,8 @@ static struct kvm_x86_ops vmx_x86_ops = {
> .pmu_ops = &intel_pmu_ops,
>
> .update_pi_irte = vmx_update_pi_irte,
> +
> + .setup_mce = vmx_setup_mce,
> };
>
> static int __init vmx_init(void)
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index bf22721..5bf76ab 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -70,7 +70,8 @@
>
> #define MAX_IO_MSRS 256
> #define KVM_MAX_MCE_BANKS 32
> -#define KVM_MCE_CAP_SUPPORTED (MCG_CTL_P | MCG_SER_P)
> +u64 __read_mostly kvm_mce_cap_supported = MCG_CTL_P | MCG_SER_P;
> +EXPORT_SYMBOL_GPL(kvm_mce_cap_supported);
>
> #define emul_to_vcpu(ctxt) \
> container_of(ctxt, struct kvm_vcpu, arch.emulate_ctxt)
> @@ -983,6 +984,7 @@ static u32 emulated_msrs[] = {
> MSR_IA32_MISC_ENABLE,
> MSR_IA32_MCG_STATUS,
> MSR_IA32_MCG_CTL,
> + MSR_IA32_MCG_EXT_CTL,
> MSR_IA32_SMBASE,
> };
>
> @@ -2684,11 +2686,9 @@ long kvm_arch_dev_ioctl(struct file *filp,
> break;
> }
> case KVM_X86_GET_MCE_CAP_SUPPORTED: {
> - u64 mce_cap;
> -
> - mce_cap = KVM_MCE_CAP_SUPPORTED;
> r = -EFAULT;
> - if (copy_to_user(argp, &mce_cap, sizeof mce_cap))
> + if (copy_to_user(argp, &kvm_mce_cap_supported,
> + sizeof(kvm_mce_cap_supported)))
> goto out;
> r = 0;
> break;
> @@ -2866,7 +2866,7 @@ static int kvm_vcpu_ioctl_x86_setup_mce(struct kvm_vcpu
> *vcpu,
> r = -EINVAL;
> if (!bank_num || bank_num >= KVM_MAX_MCE_BANKS)
> goto out;
> - if (mcg_cap & ~(KVM_MCE_CAP_SUPPORTED | 0xff | 0xff0000))
> + if (mcg_cap & ~(kvm_mce_cap_supported | 0xff | 0xff0000))
> goto out;
> r = 0;
> vcpu->arch.mcg_cap = mcg_cap;
> @@ -2876,6 +2876,9 @@ static int kvm_vcpu_ioctl_x86_setup_mce(struct kvm_vcpu
> *vcpu,
> /* Init IA32_MCi_CTL to all 1s */
> for (bank = 0; bank < bank_num; bank++)
> vcpu->arch.mce_banks[bank*4] = ~(u64)0;
> +
> + if (kvm_x86_ops->setup_mce)
> + kvm_x86_ops->setup_mce(vcpu);
> out:
> return r;
> }
>
