zeekayandHanzo Dev 4397552c91 fix(engine): stop steering the VM backwards, and stop calling a certified descendant a fork
All five devnet validators os.Exit(1)'d on v1.36.33, once each, at three distinct
height pairs (1258/1259, 1363/1364, 1364/1365) — always certified = head - 1, and
with every surviving node holding byte-identical blocks at every one of those
heights. There was no fork anywhere; in luxd-1's case the block it refused to
prefer was the block it had just built and the fleet had just accepted.

Two defects, one crash.

PRODUCER. acceptWithCertCore releases t.mu across every VM call-out, then steers
the VM with its STALE local blockID. A finalize that completed in that window has
already advanced the ledger AND the EVM to blockID+1, so the steer is BACKWARDS
and the EVM's accepted-irreversibility guard (evm/core/blockchain.go:1987,
commonBlock < lastAccepted) correctly refuses it. That guard is right and is left
alone. Fixed by steering at the LIVE build anchor — PreferredBuildTip, descended
from ledger.BuildAnchor, which the accept ordering (ApplyCert BEFORE VM.Accept)
keeps at or above the VM's own accepted head. It is the same value the build path
already steers with: one build target, computed one way, read in both places.

CLASSIFIER. reconcileVMToCertified asked only "is the head the ledger's certified
canonical at ITS OWN height?" — trivially true for every healthy node whose head
is certified — and called that a two-blocks-at-one-height double-finalization. It
never established that `certified` was at that same height. The ledger holds one
canonical per height along one contiguous chain (Finalize refuses equivocation
AND non-descendant branches), so when both blocks are certified at their own
heights they lie on that one chain and the head merely DESCENDS from the target:
nothing is orphaned. The fail-closed halt is NOT weakened — it now fires on the
state that is actually unsafe, steering off a certified head onto a block our own
ledger does not certify at its height, and TestReconcile_CertifiedHead_HaltsFail-
Closed proves it with both blocks readable so the LEDGER decides, not a missing id.

Direction chosen: NEITHER roll the head back nor certify forward. The VM head is
legitimately ahead and already CONTAINS the certified block. Rolling back is what
evm/core/blockchain.go:1987 exists to refuse, and FinalityLedger.BuildAnchor
already documents head > certified as the designed state. The correct action is
no action — plus not issuing the backwards steer in the first place.

Regression, failing before / passing after:
  TestReconcile_CertifiedDescendantHead_NoHalt   the incident, reduced
  TestReconcile_CertifiedAncestorHead_NoHalt     the mirror
  TestAcceptWithCert_StaleSteerIsNotIssued       the race window, injected
                                                 deterministically via Accept

Co-authored-by: Hanzo Dev <[email protected]>
2026-07-28 03:10:10 -07:00
2026-06-28 20:35:53 -07:00

consensus

Lux Consensus

Lux is not merely adding post-quantum signatures to a chain; it defines a hybrid finality architecture for DAG-native consensus, with protocol-agnostic threshold lifecycle, post-quantum threshold sealing, and cross-chain propagation of Horizon finality.

See LP-105 §Claims and evidence for the canonical claims/evidence table and the ten architectural commitments — single source of truth.

CI Status Go Version

Quasar Family of Consensus

Quasar is the consensus engine for the Lux network. It provides unified consensus for linear chains (P-Chain, C-Chain) and DAG chains (X-Chain) with post-quantum finality via three independent cryptographic signing paths.

Sub-Protocols

Protocol Role Package
Photon K-of-N committee selection via Fisher-Yates sampling with luminance-weighted reputation protocol/photon
Wave Per-round threshold voting with FPC (Fast Probabilistic Consensus) protocol/wave
Focus Confidence accumulation: beta consecutive successes = local finality protocol/focus
Nova Linear chain consensus mode (P-Chain, C-Chain) -- wraps Ray protocol/nova
Nebula DAG consensus mode (X-Chain) -- wraps Field protocol/nebula
Prism DAG geometry: frontiers, cuts, uniform peer sampling protocol/prism
Horizon DAG order-theory: reachability, LCA, transitive closure, skip lists protocol/horizon
Flare DAG certificate/skip detection via 2f+1 quorum protocol/flare
Ray Linear chain finality driver: Wave + Focus + Sink protocol/ray
Field DAG finality driver: Wave + safe-prefix commit protocol/field
Quasar BLS + Pulsar + ML-DSA threshold signing, epoch management protocol/quasar

Signing Modes

Each cryptographic layer is independently toggleable:

Mode Layers Use Case
BLS-only BLS12-381 threshold Fastest classical consensus
BLS + ML-DSA BLS + FIPS 204 ML-DSA-65 Dual with PQ identity proof
BLS + Pulsar BLS + Ring-LWE 2-round threshold Dual with PQ threshold proof
BLS + Pulsar + ML-DSA All three in parallel Full Quasar (triple mode)

IsTripleMode() returns true when all three paths are configured. TripleSignRound1 runs BLS + Pulsar + ML-DSA signing in parallel goroutines.

PQ Mode Selection (Configurable)

The active mode is selectable at runtime via config.PQMode or the CONSENSUS_PQ_MODE env var. Defined in config/pq_mode.go:

Mode Env value Maps to wire policy
BLSOnly bls PolicyQuorum
BLSPlusMLDSA bls-mldsa PolicyPQ
BLSPlusCorona bls-rt PolicyPZ
BLSPlusGroth16 bls-z PolicyQuantum (placeholder)
TripleQuantum triple PolicyQuantum

Run bench/pq_modes_bench_test.go for measured per-mode signing / aggregation / verify costs and cert sizes. Storage and cert size trade-offs:

Mode Sign Agg Verify Cert Storage 10K
bls 312 µs 8.6 ms 714 µs 123 B 1.17 MB
bls-mldsa 369 µs 8.5 ms 3.4 ms 69 KB 665 MB
bls-rt 39 ms 3.3 s 1.6 ms 33 KB 318 MB
triple 40 ms 3.3 s 4.3 ms 102 KB 981 MB

Pulsar / triple include the full 2-round Pulsar threshold protocol (github.com/luxfi/pulsar/threshold — Pulsar is Lux's variant with DKG2 and Pulsar-SHA3 hash suite). Production parameters M=8, N=7, LogN=8 (ring degree 256), Q=2^48-ish: classical 2^142, quantum 2^130. Cert-size floor of ~33 KB is set by the (C, Z, Delta) ring polynomials; protocol/quasar/cert_size_compare_test.go measures and verifies this.

Architecture

consensus/
  protocol/
    photon/       Committee selection (Emitter + Luminance)
    wave/         Threshold voting + FPC
      fpc/        Fast Probabilistic Consensus selector
    focus/        Confidence counter (beta tracker)
    prism/        DAG cuts, frontiers, uniform sampling
    horizon/      DAG reachability, LCA, skip lists
    flare/        DAG cert/skip classification (2f+1)
    ray/          Linear chain driver (Nova uses this)
    field/        DAG finality driver (Nebula uses this)
    nova/         Linear chain consensus mode
    nebula/       DAG consensus mode
    chain/        Block interface primitives
    quasar/       BLS + Pulsar + ML-DSA threshold signing
  engine/         Consensus engine (Chain, DAG, PQ wrappers)
  core/           Core types, DAG structures
  types/          Block, Vote, Config, Bag
  config/         Parameter presets (single, local, testnet, mainnet)
  runtime/        VM wiring (chain IDs, validators, logging)
  pkg/wire/       Wire protocol credentials (ML-DSA-44/65/87 + BLS + Ed25519)
  bench/          Benchmarks
  version/        Re-exports github.com/luxfi/version

Consensus Flow

Linear Chains (Nova)

Photon (select committee) -> Wave (threshold vote) -> Focus (count beta) -> Ray (decide) -> Sink

DAG Chains (Nebula)

Photon (select committee) -> Wave (threshold vote per frontier vertex)
  -> Flare (cert/skip) -> Horizon (safe prefix) -> Field (commit ordered prefix) -> Committer

Quasar PQ Finality

After consensus decision, the Quasar signing layer produces threshold certificates:

  1. BLS: single-round threshold share via crypto/threshold
  2. Pulsar: 2-round Ring-LWE threshold via luxfi/corona/threshold
  3. ML-DSA-65: single-round FIPS 204 identity signature

All three run in parallel. A block is quantum-final when all configured certificate layers are valid.

Key Hierarchy

Layer Key Type Purpose
Node-ID Ed25519 P2P transport auth
Validator-BLS BLS12-381 Classical finality votes
Validator-PQ Pulsar (Ring-LWE) PQ finality shares
Validator-ID ML-DSA-65 (FIPS 204) PQ identity attestation
Wallet (EVM) secp256k1 User transaction signatures
Wallet (X-Chain) secp256k1 UTXO locking

Quick Start

import "github.com/luxfi/consensus"

chain := consensus.NewChain(consensus.DefaultConfig())

ctx := context.Background()
if err := chain.Start(ctx); err != nil {
    log.Fatal(err)
}
defer chain.Stop()

block := &consensus.Block{
    ID:       consensus.NewID(),
    ParentID: consensus.GenesisID,
    Height:   1,
    Payload:  []byte("Hello, Lux!"),
}
if err := chain.Add(ctx, block); err != nil {
    log.Fatal(err)
}

Configuration

// Auto-configure based on validator count
params := consensus.GetConfig(21) // mainnet defaults

// Or use presets
consensus.SingleValidatorParams() // 1 node, dev
consensus.LocalParams()           // 5 nodes, local
consensus.TestnetParams()         // 11 nodes, testnet
consensus.MainnetParams()         // 21 nodes, production

Benchmarks

Measured on Apple M1 Max:

Component Operation Time/Op
Wave Vote round 3.38 us
Photon K-of-N selection 3.03 us
Luminance Brightness update 72 ns

ZAP wire protocol (measured, bench/):

Configuration Throughput
Single connection 114K TPS
20 parallel connections 376K TPS
50 conns + batch 1000 20.26M TPS

Testing

GOWORK=off go test ./...
GOWORK=off go test -bench=. ./bench/

Security Model

  1. Pre-quantum: Attacker must corrupt >= 1/3 stake to fork
  2. Q-day (BLS broken): Lattice certificates prevent unsafe finalization
  3. Post-quantum: Security rests on Module-LWE (Pulsar) + Module-LWE/SIS (ML-DSA)

License

BSD 3-Clause. See LICENSE.

S
Description
Quasar consensus engine: Go implementations of consensus protocols for the Lux Network.
Readme
251 MiB
Languages
Go 85.5%
Rust 7.6%
Python 3.1%
TeX 1.2%
C 0.9%
Other 1.7%