[PATCH v39 15/24] x86/sgx: Add SGX_IOC_ENCLAVE_PROVISION

[Jarkko Sakkinen]

On Sat, Oct 24, 2020 at 08:47:28AM -0700, Andy Lutomirski wrote:
> On Sat, Oct 24, 2020 at 4:34 AM Jarkko Sakkinen <kernel.org at kernel.org> wrote:
> >
> > On Fri, Oct 23, 2020 at 07:19:05AM -0700, Dave Hansen wrote:
> > > On 10/23/20 3:17 AM, Jarkko Sakkinen wrote:
> > > > On Tue, Oct 20, 2020 at 02:19:26PM -0700, Dave Hansen wrote:
> > > >> On 10/2/20 9:50 PM, Jarkko Sakkinen wrote:
> > > >>> + * Failure to explicitly request access to a restricted attribute will cause
> > > >>> + * sgx_ioc_enclave_init() to fail.  Currently, the only restricted attribute
> > > >>> + * is access to the PROVISION_KEY.
> > > >> Could we also justify why access is restricted, please?  Maybe:
> > > >>
> > > >>    Access is restricted because PROVISION_KEY is burned uniquely
> > > >>    into each each processor, making it a perfect unique identifier
> > > >>    with privacy and fingerprinting implications.
> > > >>
> > > >> Are there any other reasons for doing it this way?
> > > > AFAIK, if I interperet the SDM correctl, PROVISION_KEY and
> > > > PROVISION_SEALING_KEY also have random salt added, i.e. they change
> > > > every boot cycle.
> > > >
> > > > There is "RAND = yes" on those keys in Table 40-64 of Intel SDM volume
> > > > 3D :-)
> > >
> > > Does that mean there are no privacy implications from access to the
> > > provisioning keys?  If that's true, why do we need a separate permission
> > > framework for creating provisioning enclaves?
> >
> > As I've understood it, the key material for those keys is not even
> > required in the current SGX architecture, it was used in the legacy EPID
> > scheme, but the attribute itself is useful.
> >
> > Let's assume that we have some sort of quoting enclave Q, which guards a
> > public key pair, which signs quotes of other enclaves. Let's assume we
> > have an attestation server A, which will enable some capabilities [*],
> > if it receives a quote signed with that public key pair.
> >
> > 1. E gets the report key with EGETKEY.
> > 2. E constructs REPORTDATA (37.16) and TARGETINFO (37.17) structures.
> >    The former describes the enclaves contents and attributes and latter
> >    the target, i.e. Q in this artitificial example.
> > 3. E calls EREPORT to generate a structure called REPORT MAC'd with the
> >    *targets* report key. It knows, which key to usue from REPORTDATA.
> > 4. The runtime will then pass this to Q.
> > 5. Q will check if ATTRIBUTE.PROVISION_KEY is set. If it is, Q will
> >    know that the enclave is allowed to get attested. Then it will
> >    sign the report with the guarded public key pair and send it to
> >    the attestation server.
> 
> I think you have this a little bit off.  AIUI E won't have
> ATTRIBUTE.PROVISION_KEY set -- Q will.  Q uses the provisioning key to
> convince an Intel server that it's running on a genuine Intel CPU, and
> the Intel server will return a signed certificate that Q can chain off
> of to generate attestations for E.

Right, I was confused by that RAND column, until Jethro corrected me.

Actually, quoting enclave (QE) authorizes itself with a provisioning
certification enclave (PCE), which holds certificates and revocation
lists for provisioning secrets unique to a CPU. And the sequence that I
described happens between PCE and QE. It accepts requests from enclaves
with ATTRIBUTES.PROVISION key bits set to 1 according to:

  https://software.intel.com/content/dam/develop/external/us/en/documents/intel-sgx-support-for-third-party-attestation-801017.pdf

The source code for the reference  is available here:

  https://github.com/intel/SGXDataCenterAttestationPrimitives

And binaries are here:

  https://01.org/intel-softwareguard-extensions/downloads/intel-sgx-dcap-1.6-release

They are provided for the inevitable reason that, it is the way bind to
the hardware, i.e. proof that you are running on a genuine CPU.

The network part is that PCE and QE can certify to an application, if an
enclave running in a different computer is an enclave.

> Dave, I would rephrase what you're saying a bit.  The PROVISION_KEY
> attribute allows enclaves to access keys that are unique to a
> processor and unchangeable.  Unlike other SGX keys, these keys are not
> affected by OWNER_EPOCH changes and therefore cannot be reset.

/Jarkko