Repository navigation
test(store): check that what ck writes stays readable under the frozen format (LAB-10651) - #25
Merged
Merged
Conversation
…ader an_older_entry_still_decrypts checks that ck reads old entries. Nothing checked the other direction: a writer change that the same release can still read (a prepended header byte, say) passed every test while an older ck sharing the backend would miss on every entry. The new test stores a value through Store::set and opens the exact bytes handed to the backend with HKDF-SHA256 and AES-256-GCM from RustCrypto, rebuilding the key, the v0x03 AAD and the nonce/ciphertext/tag split from the protocol spec. The reader is pinned to the protocol's basic_bytes vector and to the cachekit-rs 0.9.0 ciphertext, so it reads the frozen format rather than echoing today's writer. aes-gcm, hkdf and sha2 become dev-dependencies at the versions already locked through cachekit-core: no package or version changes in Cargo.lock. Also corrects the old-entry test's doc comment: its ciphertext comes from the EncryptionLayer call ck makes, and its plaintext is not an Envelope.
Code Review Completed! 🔥The code review was successfully completed based on your current configurations. Kody Guide: Usage and ConfigurationInteracting with Kody
Providing Context (Files & MCPs)Add these hints in your PR description (or a comment) to unlock deeper checks:
Current Kody ConfigurationReview OptionsThe following review options are enabled or disabled:
|
![kodus-27b[bot]](https://avatars.githubusercontent.com/u/223823690?s=60&v=4){kind=small}
The writer check now also stores the saas capture cap's worth of zeros, so a writer branch gated on size (compression past a threshold, say) cannot pass unseen. Its doc comment claims only what it shows: the bytes put hands to the backend open under the frozen encryption format. A failure now names both lengths and the leading stored bytes. The reader's docs say it covers the subset ck writes. The key-fingerprint check is dropped, since the basic_bytes open already pins the derived key. OLD_KEY becomes KNOWN_ANSWER_KEY, and the old ciphertext's master key is a named constant.
Code Review Completed! 🔥The code review was successfully completed based on your current configurations. Kody Guide: Usage and ConfigurationInteracting with Kody
Providing Context (Files & MCPs)Add these hints in your PR description (or a comment) to unlock deeper checks:
Current Kody ConfigurationReview OptionsThe following review options are enabled or disabled:
|
The file store keeps up to OUTPUT_CAP, not the saas cap, so a 1 MiB write let a writer branch gated between the two pass unseen. The large case now writes store.max_output() zeros.
Code Review Completed! 🔥The code review was successfully completed based on your current configurations. Kody Guide: Usage and ConfigurationInteracting with Kody
Providing Context (Files & MCPs)Add these hints in your PR description (or a comment) to unlock deeper checks:
Current Kody ConfigurationReview OptionsThe following review options are enabled or disabled:
|
Every key the format checks used repeated one byte, so it read the same reversed and could not show a byte-order mistake in the key's path into HKDF, in the writer or in the reference reader. The writer check now uses the protocol's master_key_every_hex_digit key (32 distinct bytes), and the reader is pinned to that vector's sealed entry. Two sizes could not catch a writer branch gated on a range between them. The writer check now also stores zeros at every power of 4 from 1 byte to 16 MiB, and its doc says what that covers.
Contributor
Author
|
@kody start-review |
Code Review Completed! 🔥The code review was successfully completed based on your current configurations. Kody Guide: Usage and ConfigurationInteracting with Kody
Providing Context (Files & MCPs)Add these hints in your PR description (or a comment) to unlock deeper checks:
Current Kody ConfigurationReview OptionsThe following review options are enabled or disabled:
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
an_older_entry_still_decryptschecks that ck can read entries an older release wrote. Nothing checked the other direction. A cachekit-rs or cachekit-core writer change that the same release can still read, such as a prepended header byte, would pass every ck test, while an older ck sharing the backend would miss on every new entry and re-run the command.What changed
what_ck_writes_the_frozen_format_readswrites values throughStore::seton the file backend, takes the exact bytes ck handed to the backend, and opens them with an independent reader. It writes a short text value, an empty value, and zeros at every power of 4 from 1 byte to 16 MiB and atstore.max_output()(20 MiB on the file store). A writer branch gated on size therefore passes unseen only if its whole range falls between two neighbouring sizes. The master key has 32 distinct bytes, so a byte-order mistake in the key's path into HKDF changes the key.store::tests::frozen) rebuilds the subset of the format ck writes from the protocol spec, on RustCrypto alone: the HKDF-SHA256 key with thecachekit_v1_length-prefixed salt and theencryptiondomain, the four-part v0x03 AAD (tenant, cache key,msgpack,False), andnonce(12) ‖ ciphertext ‖ tag(16)with nothing around it. Neither cachekit-rs nor cachekit-core derives the key or decrypts.the_frozen_reader_reads_the_protocol_vectorspins that reader to known answers:test-vectors/encryption.json1.5.0, thebasic_bytesAAD and ciphertext;master_key_every_hex_digitentry (32 distinct key bytes, tenantdefault), sealed outside the Rust SDKs;OLD_CIPHERTEXT).aes-gcm0.10,hkdf0.12 andsha20.10 become dev-dependencies. All three were already locked through cachekit-core (0.10.3, 0.12.4, 0.10.9). TheCargo.lockdiff only lists them undercachekit-cli: it adds no package and changes no version.an_older_entry_still_decryptsnow says what it checks: a ciphertext from the sameEncryptionLayercall ck makes, on cachekit-rs 0.9.0 and cachekit-core 0.6.0, whose plaintext is a fixed string rather than anEnvelope.Proof it catches the gap
Each change below went into a local copy of cachekit-rs 0.11.1 or cachekit-core 0.7.0 through
[patch.crates-io], which is not committed. In each case the same release still reads its own output:encryptprepends a0xC4byte;decryptstrips itwhat_ck_writes_the_frozen_format_readsfails"True"for plaintexts above 64 KiB"True"above 2 MiB"True"for 64 KiB to 1 MiB onlyA reference reader that reverses its key fails
the_frozen_reader_reads_the_protocol_vectorson themaster_key_every_hex_digitentry. Unmodified, every test passes.Checks run locally
cargo fmt --all -- --check,cargo clippy --locked --all-targets -- -D warnings,cargo test --locked(21 unit, 45 run, 17 saas, 1 readme),cargo +1.85 check --locked --all-targetsandcargo test --locked --release --test bench -- --ignoredpass.cargo deny checkfails on RUSTSEC-2026-0386 (tokio 1.53.1, also locked onmain). This PR's lock changes no version. #26 fixes it, and it should land first.No docs needed: this is test-only and changes no behaviour or documented surface.
Closes LAB-10651