fix(net): gate peer traffic on hello validation and enforce timeout

[317787106]

What does this PR do?

Require a validated application-layer HELLO before processing peer traffic, enforce a handshake deadline, and keep invalid HELLO diagnostic logs bounded.

• Before HELLO validation succeeds, allow only HELLO and DISCONNECT. Reject other message types with BAD_PROTOCOL before parsing or PBFT dispatch, and reject null or empty frames safely.
• Add a 10-second HELLO timeout based on the local channel start time, independent of synchronization flags or recent peer activity. Use the existing status-check cycle and TIME_OUT disconnect, while retaining synchronization and request timeout checks. Make helloMessageReceive volatile for visibility to the status-check thread.
• For invalid HELLO messages, log the three block-hash lengths instead of full hashes. Read diagnostic lengths with ByteString.size() to avoid copying untrusted fields for logging.

Why are these changes required?

Peers could reach business handlers, including PBFT, before completing the application handshake, while handshake timeout depended indirectly on synchronization state. Oversized hashes could also produce excessive warning output.

Handshake admission retains the existing version, genesis, and solid-block compatibility checks without adding a main-chain membership requirement for the peer's unfinalized head. This avoids rejecting otherwise compatible peers during a temporary fork, including reconnection after a disconnect or restart. Subsequent fork handling remains with the existing synchronization and broadcast logic.

This PR has been tested by:

• Focused unit tests and regressions on ARM64 / JDK 17 covering message gating, bounded logging, handshake compatibility, peer connections, and timeouts. The latest run passed HelloMessageTest, HandShakeServiceTest, HandshakeHeadCheckTest, P2pHelloGateTest, and PeerStatusCheckMockTest after removing the endpoint format checks.
• Logging scenarios are consolidated into HandShakeServiceTest; HELLO timeout scenarios are consolidated into PeerStatusCheckMockTest.
• HelloMessageTest.testValidAddressLogFormat was verified on an actual Java 8 runtime after fixing its JDK-dependent IPv6 formatting expectation.
• ./gradlew :framework:checkstyleMain :framework:checkstyleTest :plugins:checkstyleMain, bash .github/scripts/check_math_usage.sh, and git diff --check passed locally.

The full repository test suite and manual multi-node network testing were not run locally.

Follow up

The planned libp2p v2.3.0 fixes will be integrated into java-tron separately. This PR covers java-tron's application-layer HELLO handling and does not update the libp2p dependency.

Extra details

• Ten seconds is the timeout threshold; disconnection occurs on the existing periodic status-check cycle.
• No changes to the wire format, configuration, or fork-choice logic are introduced. The implementation remains Java 8 compatible.

[3for]

LGTM now.

[3for]

[SHOULD] Duplicate-connection regression test does not establish the intended connection state.

HandshakeHeadCheckTest.java (line 282) still uses clear() and put() on ChannelManager.getChannels(). However, after the merge, ChannelManager.java (line 110) returns an independent copy.

As a result, these operations do not populate the actual connection registry, so the assertion that other connections remain open does not exercise the intended scenario. The merge also removed the original assertions verifying that the test setup was valid.

Please set up and clean up the test connections through the actual registry, and first assert that both connections are registered before testing duplicate-connection handling.

[317787106]

To avoid conflicts with other changes to HelloMessage, PR #6993 drops its newly added Endpoint validation. The remaining handshake gating, HELLO timeout, and bounded invalid-HELLO logging changes are retained.