Re: [PATCH 1/2] KVM: nVMX: Always indicate HLT activity support in VMX_MISC MSR

[Jim Mattson <jmattson@google.com>]

On Mon, Aug 19, 2019 at 2:47 PM Nikita Leshenko
<nikita.leshchenko@oracle.com> wrote:
>
> Before this commit, userspace could disable the GUEST_ACTIVITY_HLT bit in
> VMX_MISC yet KVM would happily accept GUEST_ACTIVITY_HLT activity state in
> VMCS12. We can fix it by either failing VM entries with HLT activity state when
> it's not supported or by disallowing clearing this bit.
>
> The latter is preferable. If we go with the former, to disable
> GUEST_ACTIVITY_HLT userspace also has to make CPU_BASED_HLT_EXITING a "must be
> 1" control, otherwise KVM will be presenting a bogus model to L1.
>
> Don't fail writes that disable GUEST_ACTIVITY_HLT to maintain backwards
> compatibility.
>
> Reviewed-by: Liran Alon <liran.alon@oracle.com>
> Reviewed-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
> Signed-off-by: Nikita Leshenko <nikita.leshchenko@oracle.com>
> ---
>  arch/x86/kvm/vmx/nested.c | 8 ++++++++
>  1 file changed, 8 insertions(+)
>
> diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
> index 46af3a5e9209..24734946ec75 100644
> --- a/arch/x86/kvm/vmx/nested.c
> +++ b/arch/x86/kvm/vmx/nested.c
> @@ -1102,6 +1102,14 @@ static int vmx_restore_vmx_misc(struct vcpu_vmx *vmx, u64 data)
>         if (vmx_misc_mseg_revid(data) != vmx_misc_mseg_revid(vmx_misc))
>                 return -EINVAL;
>
> +       /*
> +        * We always support HLT activity state. In the past it was possible to
> +        * turn HLT bit off (without actually turning off HLT activity state
> +        * support) so we don't fail vmx_restore_vmx_misc if this bit is turned
> +        * off.
> +        */
> +       data |= VMX_MISC_ACTIVITY_HLT;
> +
>         vmx->nested.msrs.misc_low = data;
>         vmx->nested.msrs.misc_high = data >> 32;
>

This change breaks live migration to an upgraded kernel, since it
doesn't allow the IA32_VMX_MISC MSR to be restored to its original
value. I think this warrants a quirk.