PCR7 change in 15.7+ due to the switch to using MokListRT instead of MokList

[vittyvk]

Shim 15.7 switched to checking MokListRT instead of MokList to support cases when the former is extended by someone else, e.g. grub:

commit 092c2b2bbed950727e41cf450b61c794881c33e7
Author: Eric Snowberg <[email protected]>
Date:   Fri Jun 17 12:37:28 2022 -0400

    Reference MokListRT instead of MokList

Unfortunately, this commit also changed what's measured in PCR7 for shim builds using 'vendor_cert' (and not 'vendor_db'). Pre-change:

- EventNum: 28
  PCRIndex: 7
  EventType: EV_EFI_VARIABLE_AUTHORITY
  ...
  Event:
    VariableName: 605dab50-e046-4300-abb6-3dd810dd8b23
    UnicodeNameLength: 4
    VariableDataLength: 1119
    UnicodeName: Shim
    VariableData: ...

Post-change:

- EventNum: 28
  PCRIndex: 7
  EventType: EV_EFI_VARIABLE_AUTHORITY
  ...
  Event:
    VariableName: 605dab50-e046-4300-abb6-3dd810dd8b23
    UnicodeNameLength: 9
    VariableDataLength: 1135
    UnicodeName: MokListRT
    VariableData: ...

The reason for the change is that MokListRT gets 'vendor_cert'/'vendor_db' mirrored in it, however, shim checks MokListRT before it gets to checking 'vendor_cert' (but after 'vendor_db' so there's no change for these builds, see verify_one_signature()/check_allowlist()).

Now, the question is what can (or should) be done to this. I see the following options:

• We change the order of the checks, namely move SHIM_CERT/VENDOR_CERT check to check_allowlist() before it gets to checking MokListRT. Unfortunately, this will result in PCR7 change again.
• We do nothing and stick to the new behavior. It would probably be nice to have this documented somewhere as the discrepancy between 'vendor_db' and 'vendor_cert' is quite un-obvious.
• We eradicate 'vendor_cert' completely and make everyone use 'vendor_db' even for a single cert.

[vittyvk]

Cc @esnowberg

[chrisccoulson]

Huh, I thought I'd already opened an issue for this because this is something we had to add a workaround for in Ubuntu Core some time ago, but it seems like I didn't.

[vittyvk]

It was brought to my attention that there's #616 already which basically describes the same issue but as it only mentions 'dead code' and not PCR7 measurements change, this went unnoticed.

[trungams]

We also noticed this PCR7 value change in Azure Linux after switching from shim 15.3 to shim 15.8. We depend on PCR7 for our confidential VM image so it would be nice to have the behaviour stable.