Skip to content
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.

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

Non-digested containers-storage references should resolve to the manifest that exists #1929

Open
mtrmac opened this issue Apr 24, 2023 · 1 comment
Labels
kind/bug A defect in an existing functionality (or a PR fixing it)

Comments

@mtrmac
Copy link
Collaborator

mtrmac commented Apr 24, 2023

cri-o/cri-o#6457 (comment)

@mtrmac mtrmac added the kind/bug A defect in an existing functionality (or a PR fixing it) label Apr 24, 2023
@mtrmac mtrmac changed the title containers-storage @id references should resolve to the manifest that exists Non-digested containers-storage references should resolve to the manifest that exists Nov 28, 2023
@mtrmac
Copy link
Collaborator Author

mtrmac commented Nov 28, 2023

This affects not just @id references, but also all tagged references (where a manifest digest is not provided by the caller).

The original report is about a multi-arch manifest where we have only pulled a non-default architecture and consuming the multi-arch manifest the usual way would look for a manifest which we don’t have.

There’s another problematic case: if we are validating signatures for a specific name, we need to find the manifest that is signed under that name (if multiple manifests, some signed, some not, end up being stored in the same deduplicated image). We don’t currently, by default validate signatures in c/storage during ordinary Podman operation, but CRI-O wants to introduce that around cri-o/cri-o#7046 .

In that case we probably need to maintain native RepoTags records (name:tag → digest, not just name@digest) to be able to find “the right” manifest without having to parse signatures. But that would also make handling the non-default-architecture case even harder, because we would explicitly record that we want to refer to the multi-arch manifest. ?!

I.e. users who pull an image with a non-default architecture option would have to also later consume it with a non-default architecture option. That’s somewhat plausible for a single-image operation, but it completely breaks down for podman images and similar operations which want to work with all images.

So, I guess, instead of “the obvious RepoTags” we would need to record “resolved RepoTags”, where we don’t point at the multi-arch manifest list, but only at the per-arch manifest. Or, possibly, record both, one for reporting to users in inspect, and one for actual image lookups?!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
kind/bug A defect in an existing functionality (or a PR fixing it)
Projects
None yet
Development

No branches or pull requests

1 participant