M-L03: limit number of related ids per session key state#662
M-L03: limit number of related ids per session key state#662philanton merged 1 commit intofix/audit-findingsfrom
Conversation
📝 WalkthroughWalkthroughSession key state submission validation was refactored to check length limits after unmarshalling in the app session API, asset validation was added to the channel session API, comprehensive test suites were introduced for both handlers, and the default maximum session key ID limit was reduced from 256 to 10. Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Suggested reviewers
Poem
🚥 Pre-merge checks | ✅ 2 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
@coderabbitai review |
✅ Actions performedReview triggered.
|
|
@coderabbitai review |
✅ Actions performedReview triggered.
|
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (2)
clearnode/api/channel_v1/submit_session_key_state_test.go (1)
90-126: Strengthen failure-path assertions by checking store is never touched.For over-limit (and similar validation-failure tests), add explicit assertions that DB access methods were not called. That locks in the early-return guarantee.
♻️ Suggested test hardening
handler.SubmitSessionKeyState(ctx) require.NotNil(t, ctx.Response) respErr := ctx.Response.Error() require.NotNil(t, respErr) assert.Contains(t, respErr.Error(), "assets array exceeds maximum length of 2") + mockStore.AssertNotCalled(t, "GetLastChannelSessionKeyVersion", mock.Anything, mock.Anything) + mockStore.AssertNotCalled(t, "StoreChannelSessionKeyState", mock.Anything)🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@clearnode/api/channel_v1/submit_session_key_state_test.go` around lines 90 - 126, The test TestChannelSubmitSessionKeyState_AssetsExceedsMax should also assert that the store was never accessed to enforce the early-return validation: after calling handler.SubmitSessionKeyState(ctx) and verifying the validation error, add assertions on the MockStore (mockStore) to ensure its DB methods (the methods used by Handler via useStoreInTx/StoreTxHandler) were not invoked — e.g., verify no calls/expectations on mockStore occurred or that specific methods like Create/Update/Find were not called — so the test guarantees the handler returned before touching the store.clearnode/api/app_session_v1/submit_session_key_state_test.go (1)
104-151: Consider asserting no store interaction on validation failures.In max-length rejection tests, add
AssertNotCalledchecks for store methods to ensure invalid payloads never proceed into version lookup/persistence.♻️ Suggested test hardening
handler.SubmitSessionKeyState(ctx) require.NotNil(t, ctx.Response) respErr := ctx.Response.Error() require.NotNil(t, respErr) assert.Contains(t, respErr.Error(), "application_ids array exceeds maximum length of 2") + mockStore.AssertNotCalled(t, "GetLastAppSessionKeyVersion", mock.Anything, mock.Anything) + mockStore.AssertNotCalled(t, "StoreAppSessionKeyState", mock.Anything)🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@clearnode/api/app_session_v1/submit_session_key_state_test.go` around lines 104 - 151, The test TestSubmitSessionKeyState_ApplicationIDsExceedsMax should also assert that the store was never called when validation fails: after calling handler.SubmitSessionKeyState(ctx) and checking ctx.Response/Error, add AssertNotCalled (or equivalent) on the mockStore for any persistence/lookup methods used by the handler (e.g., the methods implemented on MockStore such as GetSessionKeyVersion, SaveSessionKeyState or other Store interface methods) to ensure Handler.useStoreInTx was not invoked and no version lookup or persistence occurred.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@clearnode/runtime.go`:
- Line 84: The change to MaxSessionKeyIDs (struct field MaxSessionKeyIDs with
env var CLEARNODE_MAX_SESSION_KEY_IDS) lowered the implicit default from 256 to
10 which can silently reject existing clients; either revert the env-default
back to "256" to preserve previous behavior and rely on new validation to bound
inputs, or keep "10" but add an explicit migration/rollout note in the
PR/CHANGELOG and documentation calling out the breaking default change and
required operator action to set CLEARNODE_MAX_SESSION_KEY_IDS if they need the
previous capacity.
---
Nitpick comments:
In `@clearnode/api/app_session_v1/submit_session_key_state_test.go`:
- Around line 104-151: The test
TestSubmitSessionKeyState_ApplicationIDsExceedsMax should also assert that the
store was never called when validation fails: after calling
handler.SubmitSessionKeyState(ctx) and checking ctx.Response/Error, add
AssertNotCalled (or equivalent) on the mockStore for any persistence/lookup
methods used by the handler (e.g., the methods implemented on MockStore such as
GetSessionKeyVersion, SaveSessionKeyState or other Store interface methods) to
ensure Handler.useStoreInTx was not invoked and no version lookup or persistence
occurred.
In `@clearnode/api/channel_v1/submit_session_key_state_test.go`:
- Around line 90-126: The test TestChannelSubmitSessionKeyState_AssetsExceedsMax
should also assert that the store was never accessed to enforce the early-return
validation: after calling handler.SubmitSessionKeyState(ctx) and verifying the
validation error, add assertions on the MockStore (mockStore) to ensure its DB
methods (the methods used by Handler via useStoreInTx/StoreTxHandler) were not
invoked — e.g., verify no calls/expectations on mockStore occurred or that
specific methods like Create/Update/Find were not called — so the test
guarantees the handler returned before touching the store.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 412e6ce8-0c32-4d22-a3dc-13fc64b03345
📒 Files selected for processing (5)
clearnode/api/app_session_v1/submit_session_key_state.goclearnode/api/app_session_v1/submit_session_key_state_test.goclearnode/api/channel_v1/submit_session_key_state.goclearnode/api/channel_v1/submit_session_key_state_test.goclearnode/runtime.go
H-H01: fix(contracts): disallow challenge with CLOSE or FIN_MIG intents (#631) H-M01: fix(contracts/ChannelHub): remove transferred amount check (#636) fix(contracts/ChannelHub): allow unblocking escrow ops after migration (#637) M-C01: feat(contracts): add token check between states (#639) H-L03: fix(contracts/ChannelHub): skip disputed escrow during purge (#640) M-H04: feat(contracts/ChannelHub): purge during escrow challenge finalization, count both skip and purge (#641) M-M02: fix(contracts): require zero allocs on close (#643) M-C02: fix(rpc/core): apply finalize escrow deposit correctly (#644) H-L02: fix(contracts/ChannelHub): revert on withdrawFromVault failure (#651) M-H03: fix(clearnode): log correct fields on failure (#647) M-C03: fix(pkg/core): disallow negative amount transitions (#645) H-L01: fix(contracts/ChannelHub): add CH address to prevent val addition replay (#650) M-H06(clearnode): restrict max channel challenge duration (#654) M-M03(clearnode): revert EscrowLock on insufficient home user balance (#655) M-M04(clearnode): support default signer even if it is not approved (#656) M-M05(clearnode): require correct intent on FinalizeEscrowWithdrawal (#657) M-M09: reject asset decimals exceeding its token's decimals (#659) M-L01: return correct errors during state advancement validation (#660) M-L03: limit number of related ids per session key state (#662) M-I02: normalize input hex addresses (#663) M-H01: feat(contracts/ChannelHub): restrict to one node (#649) M-H07: fix(contracts/ChannelHub): emit stored, not arbitrary candidate state (#664) M-I01: feat(contracts/ChannelHub): remove updateLastState flag (#665) H-L06: fix(contracts/ChannelHub): remove payable from methods, clarify in create (#666) H-L09: feat(contracts/ChannelEngine): add non-home migration version check (#669) YNU-839: fix blockchain listener lifecycle (#658) M-H09: use channel signer for all channel state node sigs (#667) H-L07: fix(contracts/ChannelHub): restrict createChannel to non-existing channels (#668) H-I02: docs(contracts): add a note that rebasing tokens are not supported (#670) H-I03: fix(contracts/ChannelHub): use CIE pattern in depositToHub (#671) M-H11: revert empty signatures on quorum verification (#672) M-H11: reject issuance of receiver state during escrow ops (#674) H-I01: docs(contracts): note that fee-on-transfer tokens are not supported (#675) M-L04: docs: mention liquidity monitoring (#677) fix(contracts/ChannelHub): fix initialize escrow deposit dos (#679) M-H08: enforce strict transition ordering after MutualLock and EscrowLock (#680) fix: run forge fmt (#681) M-L05: docs(contracts): document native token deposits (#685) fix(clearnode): resolve a set of audit findings (#686) M-H13: feat(contract): add validateChallengeSignature, revert in SK validator (#688)
Description
In
channel_v1,SubmitSessionKeyStateaccepts aChannelSessionKeyStateV1payload containing an unbounded assets array but never validates its length against the configuredmaxSessionKeyIDslimit. In contrast, the equivalentapp_session_v1.SubmitSessionKeyStatehandler correctly enforces this limit. As a result, a user request can supply an arbitrarily large asset list, which is then packed into the session key metadata hash and persisted as one database row per asset.Summary by CodeRabbit
Configuration Changes
Bug Fixes