Re: [PATCH] tpm: enable HMAC encryption for only x86-64 and aarch64

["Jarkko Sakkinen" <jarkko@kernel.org>]

On Tue May 21, 2024 at 5:56 PM EEST, James Bottomley wrote:
> On Tue, 2024-05-21 at 17:35 +0300, Jarkko Sakkinen wrote:
> > On Tue May 21, 2024 at 5:26 PM EEST, Jarkko Sakkinen wrote:
> > > On Tue May 21, 2024 at 5:13 PM EEST, James Bottomley wrote:
> > > > On Tue, 2024-05-21 at 17:02 +0300, Jarkko Sakkinen wrote:
> > > > > Secondly, it also roots to the null key if a parent is not
> > > > > given. So it covers all the basic features of the HMAC patch
> > > > > set.
> > > > 
> > > > I don't think that can work.  The key file would be wrapped to
> > > > the parent and the null seed (and hence the wrapping) changes
> > > > with every reboot.  If you want a permanent key, it has to be in
> > > > one of the accessible permanent hierarchies (storage ideally or
> > > > endorsement).
> > > 
> > > I'm fully aware that null seed is randomized per power cycle.
>
> OK, as long as this gets documented, I'm OK with it
>
> > > The fallback was inherited from James Prestwood's original code and
> > > I decided to keep it as a testing feature, and also to test HMAC
> > > changes.
> > > 
> > > If you look at the testing transcript in the cover letter, it
> > > should beobvious that a primary key is created in my basic test.
> > 
> > I think what could be done to it in v3 would be to return -EOPNOTSUPP
> > if parent is not defined. I.e. rationale here is that this way the
> > empty option is still usable for something in future kernel releases.
>
> You can absolutely have null derived parent keys (I use them for
> testing as well).  However, the spec says the parent handle in that
> case should be TPM_RH_NULL (i.e. 0x40000007) not zero:
>
> https://www.hansenpartnership.com/draft-bottomley-tpm2-keys.html#name-parent

Yep. I somehow recalled that it replaced 0x0 with RH_NULL but it
actually checked whether the handle is RH_NULL and then loaded
the null key if that was the case.

BR, Jarkko