Skip to content

WIP: PCR Event combination logic #45

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
wants to merge 20 commits into
base: main
Choose a base branch
from
Open

WIP: PCR Event combination logic #45

wants to merge 20 commits into from

Conversation

@bgartzi
Copy link
Contributor

@bgartzi bgartzi commented Oct 29, 2025 •
edited
Loading

The main goal of this PR is to support combining events for PCR${N} generated for image versions A and B.

The logic must return all possible resulting PCR value/struct combinations. To do so, it must know which events change together, and how to combine those with the events from the other images. In other words, it's not just about combining all events from A and B in all combinations that are possible without any context, but about returning the minimum amount of combinations that are actually logically possible on a system going through an upgrade.

To do so, this PR also goes through a bunch of other tasks meanwhile, including just some housekeeping, refactoring, and actual functionality:

  • Refactors Pcr and Part implementations.
  • Makes hash-related fields byte vectors in memory while keeping them hex strings when serialized.
  • Unit tests to make sure it was what I expected.
  • Adds TPMEvent abstractions.
  • Adds TPMEvent related constants.
  • &TPMEvent -> Part cast implementation (there is some expected information loss).
  • Unit tests for the cast implementation
  • &Vec -> Pcr computation logic (refactored from every single compute_pcr function).
  • Unit tests for the Pcr from TPMEvent vector computation logic.
  • Adds the TPMEvent combination logic
    • base logic
    • Group-based event selection.
    • Group conflict detection (can't select a event)
    • Get rid of the root event
    • compile Vec from heterogeneous Vecs
  • Adds docs explaining the design behind the logic
  • Unit tests for the TPMEvent combination logic
  • Integration tests for the TPMEvent combination logic
  • Removes some dead code along the way
  • Fix CI :)
  • Adds a function that calculates all PCR combinations of all possible 2-permutations of a list of TPMEvents coming from n different images.

And probably some more things still missing that I will need to implement and figure out.
For now, I'm posting this here just for awareness while I finish the rest.

@bgartzi bgartzi force-pushed the preds-split_reusability branch from 3c3e3cb to 779d16e Compare October 29, 2025 17:12
@bgartzi
Copy link
Contributor Author

bgartzi commented Oct 29, 2025 •
edited
Loading

Still would need to fix the minimum supported toolchain tests caused by the bitshifting function calls...

P.S. I underestimated clippy

@bgartzi bgartzi force-pushed the preds-split_reusability branch 3 times, most recently from 3583ac8 to d38e77e Compare November 5, 2025 08:18
@bgartzi bgartzi force-pushed the preds-split_reusability branch 4 times, most recently from c1883b2 to b833548 Compare November 11, 2025 16:32
bgartzi added 17 commits November 17, 2025 17:13
@bgartzi
lib: Keep hashes as Vec<u8> in memory
34dd81f 
Part and PCR hashes were modelled as Strings. This made it a bit more
difficult than it should to handle them in memory (or reuse them later).

This commit uses the serde_as library to represent hash values as
Vec<u8> in memory in favour of comfort but still serialize them into
hex-encoded Strings for ease of comprehension.

Note this is a breaking API change that could affect the way the
cocl-operator consumes the library.

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
pcrs: Move Pcr and Part type def to separate mod
c10974c 
And add some unit tests to check that Vec<u8> -> hex-str
(de)serialization works as expected.

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
tpmevents: Constants and basic typedef
4c5bd18 
TPMEvent will be the struct that will hold everything needed to support
the calculation of possible PCR combinations based on different image
events.

The commit also add a bunch of constants that define each of the events
that are modelled now, and the input-change group they belong to.

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
pcrs: Cast TPMEvents into Part and Pcr structs
aa91f51 
Implement the from TPMEvent traits for Part and Pcr.

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
PCR4: Break down pcr computation into event + hash steps
b977199 
2 steps are needed to compute PCRs: computing the TPM events, and
combining all of them together into the resulting PCR hashes.

Right now, we were taking both steps, all together, inside the same
function.

This commit breaks down those functions into event computation + hash
computation for PCR4.

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
PCR7: Break down pcr computation into event + hash steps
a4fe7c8 
2 steps are needed to compute PCRs: computing the TPM events, and
combining all of them together into the resulting PCR hashes.

Right now, we were taking both steps, all together, inside the same
function.

This commit breaks down those functions into event computation + hash
computation for PCR7.

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
tpmevents: PCR11 event modelling
6557c5c 
Adds consts and enums modelling expected PCR11 events.

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
PCR11: Break down pcr computation into event + hash steps
b78664f 
2 steps are needed to compute PCRs: computing the TPM events, and
combining all of them together into the resulting PCR hashes.

Right now, we were taking both steps, all together, inside the same
function.

This commit breaks down those functions into event computation + hash
computation for PCR11.

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
PCR14: Break down pcr computation into event + hash steps
b37aaee 
2 steps are needed to compute PCRs: computing the TPM events, and
combining all of them together into the resulting PCR hashes.

Right now, we were taking both steps, all together, inside the same
function.

This commit breaks down those functions into event computation + hash
computation for PCR14.

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
uefi::secureboot: Delete module, it's not needed
42c31ef 
collect_secure_boot_hashes is not longer used. That makes all the fancy
stuff written around it as Traits and such not longer needed too. Which
is a good thing as, honestly, it added unneeded complexity to the
source.

To remove it cleanly, this commit moves the secureboot_db implementation
for EFIVarsLoader from the SecureBootdbLoader (now removed) trait into
the actual context for method implementations of EFIVarsLoader.

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
TPMEvents: Remove MixModel
afaa7da 
Knowing the group models outside of the execution is not needed. It
doesn't add any value to the TPMEvents. In other words, the tpm event
group-id relationships are static. They do not change over time. If they
do, it will be a bug, which will be easier to fix by just changing a
line of code than updating many json annotations and files wherever
they are used.

This commit removes the TPMEventMixModel struct from there, adds a
function to set the tpm-id->group relationships, and makes the TPMEvent
struct to only hold the Event ID. It also renames the field from mix, to
id.

Also runs some updates to the code where "mix" or anything related was
referenced instead of the brand new "id".

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
tpmevents: Simple event tree implementation
9d46986 
Now simple trees can be build, nodes can checked for being leafs or root
nodes and tree branches can be computed.

All of this is tested.

Children capacity of each node is set to 2 for now.

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
tpmevents: Derive PartialEq and Eq for TPMEventID
7b9986b 
Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
tpmevents: Derive PartialEq and Debug for TPMEvent
df6929a 
Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
tpmevents: Introduce PcrRootNodeEvent
71b127c 
PcrRootNodeEvent will be used as the root node of the event tree built
during the TPM event combination logic.

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
lib: Add strum to Cargo.toml
883d25f 
This will let us implement something that is close to an iterator.

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
tpmevents: Implement next method for TPMEventID
fb49262 
This will let us know which is the next expected TPMEventID and have
something similar to an iterator, but without having to manage one.

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi bgartzi force-pushed the preds-split_reusability branch 4 times, most recently from f3008db to 6c8998c Compare November 19, 2025 13:45
@bgartzi
pcrs: Support heterogeneous pcr# vector compilation
eca0669 
Until now, compiling TPMEvents into PCRs assumed that all TPMEvents were
from the same PCR.

This patch adds a check to the Pcr::compile_from method so it checks
that all events processed belong to the same PCR.

Based on that, it adds another function that computes vectors of Pcrs
from an heterogeneous TPMEvent vector.

Finally it adds a few slight tests to cover the new functionality and
updates some old ones that for some reason assumed that having different
pcr events compiled under the same compile_from call was right.

Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi
WIP: tpmevents.combine: Combine Vec<TPMEvent>s
81fa7db 
Combine Vec<TPMEvent> from image "this" and image "that" into a vector
of all possible solutions.

There's still some work to do, such as:
- Logic for error recovery
@bgartzi
combine: Support TPMEvent combinations for N images
964001e 
Signed-off-by: Beñat Gartzia Arruabarrena <[email protected]>
@bgartzi bgartzi force-pushed the preds-split_reusability branch from 6c8998c to 964001e Compare November 20, 2025 12:00
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@bgartzi