Re: SVSM draft specification

Gerd Hoffmann <[email protected]> Wed, 2 Jul 2025 15:51:10 +0200
Newsgroups dev.linux.lists.coconut-svsm
Message-ID <gs37rpeemichi3wsaujsutjclml5vykenia7iytljiujumdfny@zknvphjyax6w>
On Wed, Jul 02, 2025 at 12:40:55PM +0100, Geoffrey Ndu wrote:
> Comments and Suggestions:
> 
> * Bit 2 of Flags (Table 23):
> 
> The description for bit 2 is, "Secure boot databases are updatable by
> the guest (db/dbx)."
> 
> Does this also apply to other updatable Secure Boot databases like
> Platform Key (PK), Key Exchange Key (KEK), and Timestamp Signature
> Database (dbt)? If not, perhaps allocating one bit per UEFI database
> could enhance flexibility.

One bit per database is not useful I think.  They all should be treated
the same.

What possibly makes sense is have separate bits for enrollment and
update.  "enrollment" means the guest can set those variables if the
system is in setup mode (i.e. no keys present).  "update" means the
guest can update these variables with proper pkcs7 signatures.  The
later depends on crypto support for the signature verification, the
former does not.

> * Confusing Statement:
> 
> The following statement is unclear:
> 
> "If a database does not contain any data or is not present, the offset
> and size for that database must be 0."

I think this is a leftover from an older version where 'db' + 'dbx'
where added to the manifest, which is not the case any more because the
uefi firmware measures them into tpm pcr registers anyway.

take care,
  Gerd