Re: [PATCH v5 04/11] security: keys: trusted: Include TPM2 creation data

[James Bottomley]

On Mon, 2022-11-14 at 08:32 -0800, Evan Green wrote:
> On Sun, Nov 13, 2022 at 7:32 PM James Bottomley <jejb@xxxxxxxxxxxxx>
> wrote:
> > 
> > On Sun, 2022-11-13 at 13:20 -0800, Eric Biggers wrote:
> > > On Fri, Nov 11, 2022 at 03:16:29PM -0800, Evan Green wrote:
> > > > diff --git a/security/keys/trusted-keys/tpm2key.asn1
> > > > b/security/keys/trusted-keys/tpm2key.asn1
> > > > index f57f869ad60068..608f8d9ca95fa8 100644
> > > > --- a/security/keys/trusted-keys/tpm2key.asn1
> > > > +++ b/security/keys/trusted-keys/tpm2key.asn1
> > > > @@ -7,5 +7,18 @@ TPMKey ::= SEQUENCE {
> > > >         emptyAuth       [0] EXPLICIT BOOLEAN OPTIONAL,
> > > >         parent          INTEGER ({tpm2_key_parent}),
> > > >         pubkey          OCTET STRING ({tpm2_key_pub}),
> > > > -       privkey         OCTET STRING ({tpm2_key_priv})
> > > > +       privkey         OCTET STRING ({tpm2_key_priv}),
> > > > +       ---
> > > > +       --- A TPM2B_CREATION_DATA struct as returned from the
> > > > TPM2_Create command.
> > > > +       ---
> > > > +       creationData    [1] EXPLICIT OCTET STRING OPTIONAL
> > > > ({tpm2_key_creation_data}),
> > > > +       ---
> > > > +       --- A TPM2B_DIGEST of the creationHash as returned from
> > > > the
> > > > TPM2_Create
> > > > +       --- command.
> > > > +       ---
> > > > +       creationHash    [2] EXPLICIT OCTET STRING OPTIONAL
> > > > ({tpm2_key_creation_hash}),
> > > > +       ---
> > > > +       --- A TPMT_TK_CREATION ticket as returned from the
> > > > TPM2_Create command.
> > > > +       ---
> > > > +       creationTk      [3] EXPLICIT OCTET STRING OPTIONAL
> > > > ({tpm2_key_creation_tk})
> > > >         }
> > > 
> > > The commit that added this file claimed:
> > > 
> > >         "The benefit of the ASN.1 format is that it's a standard
> > > and thus the
> > >         exported key can be used by userspace tools
> > > (openssl_tpm2_engine,
> > >         openconnect and tpm2-tss-engine"
> > > 
> > > Are these new fields in compliance with whatever standard that
> > > was referring to?
> > 
> > Not really, no.  The current use case (and draft standard) is
> > already using [1] for policies and [2] for importable keys:
> > 
> > https://git.kernel.org/pub/scm/linux/kernel/git/jejb/openssl_tpm2_engine.git/tree/doc/draft-bottomley-tpm2-keys.xml
> > 
> > I'm actually planning to use [3] for signed policies.  There's no
> > reason why you can't use [4] though.  Since the creation data, hash
> > and ticket are likely used as a job lot, it strikes me they should
> > be a single numbered optional sequence instead of individually
> > numbered, since you're unlikely to have one without the others.
> 
> Thanks, I was hoping James might pipe up and tell me what to do.
> Grouping them as a single numbered optional sequence sounds
> reasonable to me. Is your draft too far along to squeeze this in?

Not at all.  The draft only becomes frozen once I submit it to the IETF
which, so far thanks to lack of any reviewers I haven't done (That's
why I was also thinking of adding signed policies).

>  If it is and I'm on my own to draft up and submit this, I would
> definitely appreciate any pointers on getting started you might have.
> 
> I notice the draft and the code seem to be out of alignment.

The kernel code is out of alignment just because development moves a
bit slowly.  Policy based keys were submitted a long time ago as part
of the original move to interoperable sealed keys based on ASN.1:

https://lore.kernel.org/all/20200616160229.8018-7-James.Bottomley@xxxxxxxxxxxxxxxxxxxxx/

But eventually the policy part was split out and forgotten about.  I
think the only complete implementation of the draft standard is the
openssl_tpm2_engine.

>  I'm unfamiliar with this process, is the idea to get through all the
> iterations and land the standard, then fix up the code? What happens
> to existing data handed out in the old format?

No, it doesn't matter at all.  That's the whole point of using ASN.1
explicit optionals: the ASN.1 is always backwards compatible.  If I
ever submit the draft, there'll have to be a new RFC to add new
explicit optionals, but keys conforming to the old RFC will still be
valid under the new one.

Of course, since openssl_tpm2_engine is the complete reference
implementation that means I'll have to add the creation PCRs
implementation to it ... unless you'd like to do it?

Regards,

James