Leap v5.0.0-rc3

Leap 5.0.0-rc3 builds upon the first Leap 5.0 release candidate with important updates to the BLS host functions, as well as bug fixes, and improvements to performance, efficiency, and stability of the Leap software. It also includes a mitigation for a security issue.

📝	This release includes consensus changes from prior 5.0 releases. Node Operators running prior 5.0 release candidates must upgrade.

See the Upgrade Guide for details on required/recommended configuration changes, deprecations and new nodeos options.

Expand the release notes below for an in depth look at the Leap v5.0.0-rc3 release including a summary of important changes, and a complete change log.

Leap v5.0.0-rc3 Release Notes

Security

Mitigation of Potential Denial of Service for Snapshots & SHiP

A security vulnerability has been mitigated, preventing errors in state_history_plugin and snapshot functionality in certain edge case conditions (PR 1964).

Node Operators that depend on snapshots or state_history_plugin with chain-state-history enabled should upgrade. Users running state_history_plugin with only trace-history are not affected.

Compatibility

basee64 encoding compatibility

This release includes a change to base64 encoding that is incompatible with versions of leap prior to 3.2.5.

For more details, see PR 1888.

Enhancements

BLS Updates

Changed host functions and added benchmarking for BLS. (PRs #1882, #1884, #1904, #1938)

BLS Host Function Updates

Added, removed, and changed host functions to require affine and non-Montgomery form and enable additional BLS mathematical operations to be feasibly implemented in WASM such as compressed signature conversion.

Changes to BLS host functions

• Replaced these
  • bls_g1_mul
  • bls_g2_mul
  • bls_g1_exp
  • bls_g2_exp
• With these
  • bls_g1_weighted_sum
  • bls_g2_weighted_sum
• Added these (which make it feasible to implement more operations in WASM such as converting from compressed encodings of public keys and signatures to the affine form required for use with the host functions)
  • bls_fp_mul
  • bls_fp_exp
• BLS host functions originally expected and returned points in Jacobian and Mongtomery form. They now expect and return points in affine and non-Montgomery form.

BLS Benchmarking

Add benchmarking of BLS host functions into Leap Benchmark suites. Use benchmark/benchmark -f bls to benchmark.

A sample result looks like

function                                runs        average       minimum       maximum

bls:
bls_g1_add                              5,000      76,699 ns      75,622 ns     123,329 ns
bls_g2_add                              5,000      82,521 ns      81,196 ns     131,229 ns
bls_pairing 1 pair                      5,000   1,638,763 ns   1,557,463 ns   3,569,653 ns
bls_pairing 3 pairs                     5,000   2,321,344 ns   2,232,847 ns   3,740,820 ns
bls_g1_weighted_sum 1 point             5,000     260,683 ns     258,196 ns     328,138 ns
bls_g1_weighted_sum 3 points            5,000     670,546 ns     661,700 ns   2,038,831 ns
bls_g1_weighted_sum 5 points            5,000     817,884 ns     808,866 ns   2,200,022 ns
bls_g2_weighted_sum 1 point             5,000     818,862 ns     794,358 ns   2,368,434 ns
bls_g2_weighted_sum 3 points            5,000   2,338,746 ns   2,285,345 ns   4,072,288 ns
bls_g2_weighted_sum 5 points            5,000   2,924,081 ns   2,845,713 ns  22,833,113 ns
bls_g1_map                              5,000     379,593 ns     372,223 ns     566,554 ns
bls_g2_map                              5,000     531,899 ns     522,480 ns     722,943 ns
bls_fp_mod                              5,000         794 ns         703 ns       9,219 ns
bls_fp_mul                              5,000         692 ns         621 ns       7,913 ns
bls_fp_exp                              5,000      32,397 ns      31,702 ns      43,478 ns

Updated Protocol Feature

Renamed BLS_PRIMITIVES to BLS_PRIMITIVES2 to make explicit that there is a breaking change in the interface and behavior of the BLS host functions enabled by the BLS_PRIMITIVES2 protocol feature compared to the BLS_PRIMITIVES protocol feature that was added in v5.0.0-rc2.

Replace cpu-effort-percent with produce-block-offset-ms

5.0.0-rc3 builds upon changes we made in rc2 to optimize block start time but replaces the existing config parameter expressed in miliseconds instead of a percentage.

Specifically, we have replaced the cpu-effort-percent configuration parameter with produce-block-offset-ms, which now specifies the amount of time to leave at the end of the 12-block production round for the last block to reach the next block producer. This value can be configured to be larger than 500 milliseconds if necessary.

See PR 1800 for complete details on the change. BP Documentation has been updated.

Only return wasm config settings if configurable wasm limits enabled

Before this change, these wasm config settings were included in the response regardless of the protocol features in use. However, it has been recognized that these settings are only valid and relevant when the CONFIGURABLE_WASM_LIMITS2 protocol feature is active.

With this update, a change has been made to the behavior of the /v1/chain/get_consensus_parameters endpoint. The wasm config settings, which include various parameters related to WebAssembly execution, will now only be returned if the protocol feature CONFIGURABLE_WASM_LIMITS2 is enabled.

This modification ensures that clients requesting consensus parameters receive only the information that is applicable to the current protocol configuration, making the responses more accurate and relevant.

For complete details on this change see PR 1770.

Prometheus Enhancements

Added stable identifiers for P2P connections and ensured valid unique_conn_node_id. (PRs #1750, #1879).

Stable Identifiers

Improved identification in prometheus_plugin by using node_id as a stable identifier for P2P connections, addressing the issue of changing identifiers with each disconnect and reconnect. This enhancement, implemented in PR #1750, resolves the challenge outlined in Issue #1683 of tracking consistent metrics in nodeos_p2p_connections.

Ensure Valid Connection ID

Ensured the presence of a valid and unique connection ID for Prometheus metrics, as detailed in PR #1879. This update addresses Issue #1871, which highlighted the inefficiency of stats displayed without connection IDs, and now temporarily assigns an ID until an established connection is confirmed.

Performance & Efficiency Improvements

• VM and ABI Serializer Performance: Various improvements to ABI serializer performance. (PRs #1770, #1773, #1792, #1922)
• Chainbase Performance Improvement: Notable enhancements in performance. (PR #1772)
• Snapshot Scheduler Test: Refactored threading for better efficiency. (PR #1821)
• P2P Network Improvements: Several updates including fixes for throttling issues and improvements in sync window management. (PRs #1791, #1811)

Stability Improvements

• Resource Monitor Plugin: Hardened shutdown handling and made tests deterministic. (PRs #1774, #1783, #1826)

Bug Fixes

• Compiler Warnings and Errors: Addressed various compiler warnings and errors to ensure a clean build. (PRs #1752, #1836)
• Test Fixes: Various fixes in tests to avoid deadlocks and improve reliability. (PRs #1766, #1818, #1905)
• Chainbase and EOS VM OC Fixes: Resolved issues with Chainbase database size and EOS VM OC's subjective compilation limits. (PRs #1827, #1843, #1877)
• PH Reliability and Error Handling: Improved error handling and reliability in PH. (PRs #1814, #1866)
• Miscellaneous Fixes: Various other fixes including those for base64 encoding and contract exceptions. (PRs #1886, #1888, #1889, #1901, #1903)

Operational Improvements

• CICD: Enhanced branch protection and integration with CI Build & Test workflow for more stable builds. (PRs #1753, #1710, #1797, #1867)

Deprecations & Removals

Deferred Transactions Removed

Deferred Transactions were previously deprecated and have been removed. No new deferred transactions may be created. Existing deferred transactions, will be blocked from executing.

Get Block Header State Deprecated

As of Leap v5.0.0 v1/chain/get_block_header_state is deprecated. It will be removed in Leap v6.0.0.

Documentation Updates

• Updated tutorial readme to reflect the latest changes. (PR #1810)

Change Log

Pull Requests

• [5.0] fix "All Required Tests Passed" CI Branch Protection by @spoonincode in #1753
• [5.0] Fix compiler warning by @heifner in #1752
• [5.0] Prometheus: Add stable identifier for P2P connections by @heifner in #1750
• [5.0] Test: read-only trxs should only be posted to read_exclusive queue by @heifner in #1766
• [4.0 -> 5.0] Only return wasm config settings if configurable wasm limits enabled by @heifner in #1773
• [4.0 -> 5.0] Signal accepted_block after it is marked valid by @heifner in #1777
• [4.0 -> 5.0] hardening resource monitor manager plugin shutdown handling by @linh2931 in #1783
• [5.0] Chainbase performance improvement by @greg7mdp in #1772
• [5.0] P2P: Fix: Throttling of last block of request caused lost block by @heifner in #1791
• [5.0] only use --eos-vm-oc-enable option in tests when OC is available, and tweak/fix location of its usage by @spoonincode in #1792
• [4.0 -> 5.0] SHiP: Fixes: Stack overflow, invalid index, split file access by @heifner in #1798
• [5.0] Replaced cpu-effort-percent with produce-block-offset-ms by @heifner in #1800
• [5.0] automatically create experimental-binaries package on release by @spoonincode in #1797
• [5.0] P2P: Throttle over sync window by @heifner in #1811
• [5.0] integrate reproducible build with CI Build & Test workflow by @spoonincode in #1710
• [5.0] Test Fix: Avoid deadlock on app_thread by @heifner in #1818
• [5.0] refactor threading of snapshot_scheduler_test by @spoonincode in #1821
• [5.0] make resource monitor plugin tests deterministic by using fixed fc::temp_directory instead of dynamic /tmp by @linh2931 in #1826
• [5.0] don't run EOS VM OC's monitor compile task callback when socket being dtored by @spoonincode in #1827
• [5.0] Increase max-transaction-time for read_only_trx unit tests by @linh2931 in #1819
• [5.0] Undo the temporary switch to mapped mode when loading a snapshot in mapped_private mode. by @greg7mdp in #1825
• [5.0] Update Tutorial Readme by @ericpassmore in #1810
• [5.0] fix new signedness warnings to have a clean build for rc3 by @linh2931 in #1836
• [5.0] Normalize speculative & producer block intervals by @heifner in #1840
• [5.0] Minimize abort/start block for speculative blocks by @heifner in #1845
• [5.0] revert usage of zstd for .deb packages by @spoonincode in #1831
• [5.0] PH: Reliability improvements by @heifner in #1814
• [5.0] backport: upload core dumps from failed tests in CI by @spoonincode in #1853
• [5.0] Recover keys in chain thread pool by @heifner in #1846
• [5.0] fix release workflow's artifact download to wait-for-exact-target not wait-for-exact-target-workflow by @spoonincode in #1867
• [5.0] PH: Improve error handling and use one strand by @heifner in #1866
• [5.0] Disable EOS VM OC's subjective compilation limits in unit tests by @linh2931 in #1843
• [5.0] Prometheus: Ensure valid unique_conn_node_id by @heifner in #1879
• [5.0] Fix chainbase issue when shrinking the database size. by @greg7mdp in #1877
• [5.0] Add additional tests for mapped_private by @heifner in #1887
• [5.0] Revert "Fix base64 encoding" by @heifner in #1886
• [5.0] Fix base64 encoding - take 2 by @heifner in #1888
• [5.0] Add dynamic check for pagemap support by @greg7mdp in #1895
• [5.0] Benchmark BLS host functions by @linh2931 in #1884
• [4.0 -> 5.0] Do not require trailing = for base64 encoded strings by @heifner in #1894
• [5.0] Change remaining references of mandel to leap by @heifner in #1901
• [5.0] Add context to contract exceptions by @YaroShkvorets in #1903
• [5.0] Test: net_plugin test cleanup by @heifner in #1905
• BLS: add benchmarking for Affine form host functions by @linh2931 in #1904
• [4.0 -> 5.0] Fix non-working cleos set code and set abi commands, and add tests by @linh2931 in #1909
• [5.0] Fix two small performance issues in abi_serializer. by @greg7mdp in #1922
• [5.0] Adapt to changes in bls12-381 lib and add tests by @yarkinwho in #1882
• [5.0] Update BLS host function doxygen comments by @heifner in #1938
• [5.0] Update bls12-381 lib with affine non-montgomery public key proof of possession by @heifner in #1948
• [5.0] Return error for num == 0 for bls_g1_weighted_sum, bls_g2_weighted_sum and bls_pairing by @heifner in #1954
• [3.2] consolidated security fixes for 3.2.5 by @spoonincode in #1961
• [3.2] bump version to 3.2.5 by @spoonincode in #1962
• [3.2 -> 4.0] consolidated security fixes for 4.0.5 by @spoonincode in #1964
• [4.0] bump version to 4.0.5 by @spoonincode in #1965
• [4.0 -> 5.0] consolidated security fixes for 5.0.0-rc3 by @spoonincode in #1966
• [5.0] bump version to 5.0.0-rc3 by @spoonincode in #1967

Contributors

• @spoonincode
• @heifner
• @linh2931
• @greg7mdp
• @ericpassmore
• @YaroShkvorets
• @yarkinwho

Full Changelog: v5.0.0-rc2...v5.0.0-rc3