Re: [RFC PATCH 08/35] KVM: SVM: Prevent debugging under SEV-ES

From: Sean Christopherson
Date: Wed Sep 16 2020 - 18:50:32 EST


On Wed, Sep 16, 2020 at 03:27:13PM -0500, Tom Lendacky wrote:
> On 9/16/20 11:49 AM, Sean Christopherson wrote:
> > On Wed, Sep 16, 2020 at 11:38:38AM -0500, Tom Lendacky wrote:
> >>
> >>
> >> On 9/16/20 11:02 AM, Sean Christopherson wrote:
> >>> On Wed, Sep 16, 2020 at 10:11:10AM -0500, Tom Lendacky wrote:
> >>>> On 9/15/20 3:13 PM, Tom Lendacky wrote:
> >>>>> On 9/15/20 11:30 AM, Sean Christopherson wrote:
> >>>>>> I don't quite follow the "doesn't mean debugging can't be done in the future".
> >>>>>> Does that imply that debugging could be supported for SEV-ES guests, even if
> >>>>>> they have an encrypted VMSA?
> >>>>>
> >>>>> Almost anything can be done with software. It would require a lot of
> >>>>> hypervisor and guest code and changes to the GHCB spec, etc. So given
> >>>>> that, probably just the check for arch.guest_state_protected is enough for
> >>>>> now. I'll just need to be sure none of the debugging paths can be taken
> >>>>> before the VMSA is encrypted.
> >>>>
> >>>> So I don't think there's any guarantee that the KVM_SET_GUEST_DEBUG ioctl
> >>>> couldn't be called before the VMSA is encrypted, meaning I can't check the
> >>>> arch.guest_state_protected bit for that call. So if we really want to get
> >>>> rid of the allow_debug() op, I'd need some other way to indicate that this
> >>>> is an SEV-ES / protected state guest.
> >>>
> >>> Would anything break if KVM "speculatively" set guest_state_protected before
> >>> LAUNCH_UPDATE_VMSA? E.g. does KVM need to emulate before LAUNCH_UPDATE_VMSA?
> >>
> >> Yes, the way the code is set up, the guest state (VMSA) is initialized in
> >> the same way it is today (mostly) and that state is encrypted by the
> >> LAUNCH_UPDATE_VMSA call. I check the guest_state_protected bit to decide
> >> on whether to direct the updates to the real VMSA (before it's encrypted)
> >> or the GHCB (that's the get_vmsa() function from patch #5).
> >
> > Ah, gotcha. Would it work to set guest_state_protected[*] from time zero,
> > and move vmsa_encrypted to struct vcpu_svm? I.e. keep vmsa_encrypted, but
> > use it only for guiding get_vmsa() and related behavior.
>
> It is mainly __set_sregs() that needs to know when to allow the register
> writes and when not to. During guest initialization, __set_sregs is how
> some of the VMSA is initialized by Qemu.

Hmm. I assume that also means KVM_SET_REGS and KVM_GET_XCRS are also legal
before the VMSA is encrypted? If so, then the current behavior of setting
vmsa_encrypted "late" make sense. KVM_SET_FPU/XSAVE can be handled by not
allocating guest_fpu, i.e. they can be disallowed from time zero without
adding an SEV-ES specific check.

Which brings us back to KVM_SET_GUEST_DEBUG. What would happen if that were
allowed prior to VMSA encryption? If LAUNCH_UPDATE_VMSA acts as a sort of
reset, one thought would be to allow KVM_SET_GUEST_DEBUG and then sanitize
KVM's state during LAUNCH_UPDATE_VMSA. Or perhaps even better, disallow
LAUNCH_UPDATE_VMSA if vcpu->guest_debug!=0. That would allow using debug
capabilities up until LAUNCH_UPDATE_VMSA without adding much burden to KVM.