Re: [V1 PATCH 1/6] KVM: x86: Add support for testing private memory

[Sean Christopherson <seanjc@google.com>]

On Wed, Nov 23, 2022, Marc Orr wrote:
> On Tue, Nov 22, 2022 at 12:06 PM Sean Christopherson <seanjc@google.com> wrote:
> > > > @@ -221,6 +220,9 @@ struct kvm_page_fault {
> > > >     /* The memslot containing gfn. May be NULL. */
> > > >     struct kvm_memory_slot *slot;
> > > >
> > > > +   /* Derived from encryption bits of the faulting GPA for CVMs. */
> > > > +   bool is_private;
> > >
> > > Either we can wrap it with the CONFIG_HAVE_KVM_PRIVATE_MEM_TESTING or if
> > > it looks ugly I can remove the "const" in my code.
> >
> > Hmm, I think we can keep the const.  Similar to the bug in kvm_faultin_pfn()[*],
> > the kvm_slot_can_be_private() is bogus.  A fault should be considered private if
> > it's marked as private, whether or not userspace has configured the slot to be
> > private is irrelevant.  I.e. the xarray is the single source of truth, memslots
> > are just plumbing.
> 
> If we incorporate Sean's suggestion and use xarray as the single
> source of truth, then can we get rid of the
> HAVE_KVM_PRIVATE_MEM_TESTING config?

No, we still want the opt-in config.  

> Specifically, the self test can call the KVM_MEMORY_ENCRYPT_REG_REGION
> ioctl which will set the bits for the private FD within KVM's xarray.

Yes, but that should be disallowed for regular VMs without HAVE_KVM_PRIVATE_MEM_TESTING=y.

> (Maybe this was part of the point that Sean was making; but his
> feedback seemed focused on the discussion about keeping `is_private`
> const, whereas I've been staring at this trying to figure out if we
> can run the UPM selftests on a non-TDX/SNP VM WITHOUT a special
> test-only config. And Sean's idea seems to eliminate the need for the
> awkward CONFIG.)

"need" was always relative.  It's obviously possible to enable any code without a
Kconfig, the question is whether or not it's a good idea to do so.  In this case,
the answer is "no", because allowing private memory opens up a number a of code
paths and thus potential bugs.  And we need something for kvm_arch_has_private_mem()
because returning "true" unconditionally is not correct for regular VMs.