Validating Lightning Signer

Please see the VLS Project Overview for more information. Our web site.

Limitations

The following remain to be implemented:

• vlsd2 --recover-to can only handle a simple force-close by us. It cannot sweep a force-close or a breach by the peer. It also cannot sweep HTLC outputs.
• there is no facility to recover from loss of signer state.
• on-chain tracking is not fully implemented, so a malicious node can steal funds by failing to remedy a breach (for example)

Additional Crates

• a no_std VLS wire protocol encoder/decoder - in ./vls-protocol
• a no_std protocol handler for VLS - in ./vls-protocol-signer
• a replacement for the UNIX CLN hsmd binary, implemented in Rust in ./vls-proxy.

Development Information

Additional HOWTO Documentation

Formatting Code

Enable formatting precommit hooks:

./scripts/enable-githooks

For some reason, the ignore configuration for rustfmt is only available on the nightly channel, even though it's documented as stable.

rustup install nightly

cargo +nightly fmt

Building Validating Lightning Signer

Build VLS and related crates:

cargo build

Running Unit Tests

cargo test

To enable logging for a failing test (adjust log level to preference):

RUST_LOG=trace cargo test

Using llvm-cov for Code Coverage

Dependencies:

cargo +stable install cargo-llvm-cov --locked

Run coverage:

./scripts/run-llvm-cov

Changing linker to mold instead of ld:

cp .cargo/config.sample.toml .cargo/config.toml

Benchmarks

Running Benchmarks

cargo bench -p vls-core --bench secp_bench

Note that you might need to add --features=test_utils if you want to run all benches in vls-core.

Without optimizations:

cargo bench -p vls-core --bench secp_bench --profile=dev

Expect something like:

    test fib1_bench        ... bench:           1 ns/iter (+/- 0)
    test fib_bench         ... bench:      17,247 ns/iter (+/- 198)
    test hash_bench        ... bench:         258 ns/iter (+/- 2)
    test secp_create_bench ... bench:      49,981 ns/iter (+/- 642)
    test sign_bench        ... bench:      25,692 ns/iter (+/- 391)
    test verify_bench      ... bench:      31,705 ns/iter (+/- 1,445)

i.e. around 30 microseconds per secp256k1 crypto operation. We also see that creating a secp context is expensive, but not prohibitively so.