Re: [PATCH v7 4/7] KVM: VMX: Load Guest CET via VMCS when CET is enabled in Guest

[Jim Mattson]

On Wed, Oct 9, 2019 at 6:28 PM Yang Weijiang <weijiang.yang@xxxxxxxxx> wrote:
>
> On Wed, Oct 09, 2019 at 04:08:50PM -0700, Jim Mattson wrote:
> > On Tue, Oct 8, 2019 at 11:41 PM Yang Weijiang <weijiang.yang@xxxxxxxxx> wrote:
> > >
> > > On Wed, Oct 02, 2019 at 11:54:26AM -0700, Jim Mattson wrote:
> > > > On Thu, Sep 26, 2019 at 7:17 PM Yang Weijiang <weijiang.yang@xxxxxxxxx> wrote:
> > > > > +       if (cet_on)
> > > > > +               vmcs_set_bits(VM_ENTRY_CONTROLS,
> > > > > +                             VM_ENTRY_LOAD_GUEST_CET_STATE);
> > > >
> > > > Have we ensured that this VM-entry control is supported on the platform?
> > > >
> > > If all the checks pass, is it enought to ensure the control bit supported?
> >
> > I don't think so. The only way to check to see if a VM-entry control
> > is supported is to check the relevant VMX capability MSR.
> >
> It's a bit odd, there's no relevant CET bit in VMX cap. MSR, so I have
> to check like this.

Bit 52 of the IA32_VMX_ENTRY_CTLS MSR (index 484H) [and bit 52 of the
IA32_VMX_TRUE_ENTRY_CTLS MSR (index 490H), on hardware that supports
the "true" VMX capability MSRs] will be 1 if it is legal to set bit 20
of the VM-entry controls field to 1.

> > BTW, what about the corresponding VM-exit control?
> The kernel supervisor mode CET is not implemented yet, so I don't load host CET
> states on VM-exit, in future, I'll add it.

If you don't clear the supervisor mode CET state on VM-exit and the
guest has set IA32_S_CET.SH_STK_EN, doesn't that mean that
supervisor-mode shadow stacks will then be enabled on the host after
VM-exit?