feat: Sprint 1 — production hardening, TLS lifecycle, CI coverage, lint cleanup

- Fix 3 client panics: replace .unwrap()/.expect() with proper error
  handling in rpc.rs (AUTH_CONTEXT lock), repl.rs (pending_member),
  and retry.rs (last_err)
- Add --danger-accept-invalid-certs flag with InsecureServerCertVerifier
  for development TLS bypass, plus mdBook TLS documentation
- Add CI coverage job (cargo-tarpaulin) and Docker build validation
  to GitHub Actions workflow, plus README CI badge
- Add [workspace.lints] config, fix 46 clippy warnings across 8 crates,
  zero warnings on all buildable crates
- Update Dockerfile for all 11 workspace members
This commit is contained in:
2026-03-03 23:19:11 +01:00
parent dc4e4e49a0
commit 612b06aa8e
33 changed files with 388 additions and 67 deletions

View File

@@ -5,7 +5,7 @@ use std::sync::Arc;
use anyhow::Context;
use quinn::{ClientConfig, Endpoint};
use quinn_proto::crypto::rustls::QuicClientConfig;
use rustls::pki_types::CertificateDer;
use rustls::pki_types::{CertificateDer, ServerName, UnixTime};
use rustls::{ClientConfig as RustlsClientConfig, RootCertStore};
use tokio_util::compat::{TokioAsyncReadCompatExt, TokioAsyncWriteCompatExt};
use capnp_rpc::{rpc_twoparty_capnp::Side, twoparty, RpcSystem};
@@ -13,34 +13,101 @@ use capnp_rpc::{rpc_twoparty_capnp::Side, twoparty, RpcSystem};
use quicproquo_core::HybridPublicKey;
use quicproquo_proto::node_capnp::{auth, node_service};
use crate::AUTH_CONTEXT;
use crate::{AUTH_CONTEXT, INSECURE_SKIP_VERIFY};
use super::retry::{anyhow_is_retriable, retry_async, DEFAULT_BASE_DELAY_MS, DEFAULT_MAX_RETRIES};
/// Cap'n Proto traversal limit (words). 4 Mi words = 32 MiB; bounds DoS from deeply nested or large messages.
const CAPNP_TRAVERSAL_LIMIT_WORDS: usize = 4 * 1024 * 1024;
/// A [`rustls::client::danger::ServerCertVerifier`] that accepts any certificate.
///
/// **Development only.** Using this in production disables all TLS guarantees.
#[derive(Debug)]
struct InsecureServerCertVerifier;
impl rustls::client::danger::ServerCertVerifier for InsecureServerCertVerifier {
fn verify_server_cert(
&self,
_end_entity: &CertificateDer<'_>,
_intermediates: &[CertificateDer<'_>],
_server_name: &ServerName<'_>,
_ocsp_response: &[u8],
_now: UnixTime,
) -> Result<rustls::client::danger::ServerCertVerified, rustls::Error> {
Ok(rustls::client::danger::ServerCertVerified::assertion())
}
fn verify_tls12_signature(
&self,
_message: &[u8],
_cert: &CertificateDer<'_>,
_dss: &rustls::DigitallySignedStruct,
) -> Result<rustls::client::danger::HandshakeSignatureValid, rustls::Error> {
Ok(rustls::client::danger::HandshakeSignatureValid::assertion())
}
fn verify_tls13_signature(
&self,
_message: &[u8],
_cert: &CertificateDer<'_>,
_dss: &rustls::DigitallySignedStruct,
) -> Result<rustls::client::danger::HandshakeSignatureValid, rustls::Error> {
Ok(rustls::client::danger::HandshakeSignatureValid::assertion())
}
fn supported_verify_schemes(&self) -> Vec<rustls::SignatureScheme> {
rustls::crypto::ring::default_provider()
.signature_verification_algorithms
.supported_schemes()
}
}
/// Establish a QUIC/TLS connection and return a `NodeService` client.
///
/// Must be called from within a `LocalSet` because capnp-rpc is `!Send`.
///
/// Reads [`INSECURE_SKIP_VERIFY`] to decide whether to bypass certificate
/// verification (set once at startup via [`crate::set_insecure_skip_verify`]).
pub async fn connect_node(
server: &str,
ca_cert: &Path,
server_name: &str,
) -> anyhow::Result<node_service::Client> {
let skip = INSECURE_SKIP_VERIFY.load(std::sync::atomic::Ordering::Relaxed);
connect_node_opt(server, ca_cert, server_name, skip).await
}
/// Like [`connect_node`] but with an explicit `insecure_skip_verify` toggle.
///
/// When `insecure_skip_verify` is `true`, certificate verification is disabled entirely.
/// This is intended for development and testing only.
pub async fn connect_node_opt(
server: &str,
ca_cert: &Path,
server_name: &str,
insecure_skip_verify: bool,
) -> anyhow::Result<node_service::Client> {
let addr: SocketAddr = server
.parse()
.with_context(|| format!("server must be host:port, got {server}"))?;
let cert_bytes = std::fs::read(ca_cert).with_context(|| format!("read ca_cert {ca_cert:?}"))?;
let mut roots = RootCertStore::empty();
roots
.add(CertificateDer::from(cert_bytes))
.context("add root cert")?;
let mut tls = RustlsClientConfig::builder()
.with_root_certificates(roots)
.with_no_client_auth();
let mut tls = if insecure_skip_verify {
RustlsClientConfig::builder()
.dangerous()
.with_custom_certificate_verifier(Arc::new(InsecureServerCertVerifier))
.with_no_client_auth()
} else {
let cert_bytes =
std::fs::read(ca_cert).with_context(|| format!("read ca_cert {ca_cert:?}"))?;
let mut roots = RootCertStore::empty();
roots
.add(CertificateDer::from(cert_bytes))
.context("add root cert")?;
RustlsClientConfig::builder()
.with_root_certificates(roots)
.with_no_client_auth()
};
tls.alpn_protocols = vec![b"capnp".to_vec()];
let crypto = QuicClientConfig::try_from(tls)
@@ -76,7 +143,9 @@ pub async fn connect_node(
}
pub fn set_auth(auth: &mut auth::Builder<'_>) -> anyhow::Result<()> {
let guard = AUTH_CONTEXT.read().expect("AUTH_CONTEXT poisoned");
let guard = AUTH_CONTEXT
.read()
.map_err(|e| anyhow::anyhow!("AUTH_CONTEXT lock poisoned: {e}"))?;
let ctx = guard.as_ref().ok_or_else(|| {
anyhow::anyhow!(
"init_auth must be called before RPCs (use a bearer or session token for authenticated commands)"
@@ -257,7 +326,6 @@ pub async fn fetch_wait(
|| {
let client = client.clone();
let recipient_key = recipient_key.clone();
let timeout_ms = timeout_ms;
async move {
let mut req = client.fetch_wait_request();
{