fix(server): web UI listens on IPv6 — the mesh could never load it

The mesh is IPv6-only; the main web listener bound 0.0.0.0 only, so a
phone reaching a node over its fips0 ULA got RST on :80
(ERR_CONNECTION_ABORTED) — UI over mesh was structurally impossible.
Confirmed 2026-07-26 on framework-pt: v4:80 = 200, v6:80 = refused.
Mirror an IPv4-any main listener with a V6ONLY [::] socket on the same
port; V6ONLY so it coexists with the v4 listener regardless of
net.ipv6.bindv6only.

Also: fips.yaml generator carries the fast-reconnect profile (validated
live on framework-pt since 2026-07-24; snapshot tests updated).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
Dorian 2026-07-26 21:48:38 +01:00
parent a78aa02890
commit 4db1b85fc5
2 changed files with 118 additions and 0 deletions

View File

@ -48,6 +48,30 @@ pub struct FipsConfig {
pub struct NodeSection {
pub identity: IdentitySection,
pub discovery: DiscoverySection,
pub retry: RetrySection,
pub rate_limit: RateLimitSection,
}
/// Fast-reconnect profile (`node.retry.*`). Upstream defaults (5s base
/// doubling to 300s) are tuned for stable always-on links; a node redialing
/// a recycled anchor sat off-mesh for ~90s. Field names verified live on
/// 2026-07-24: a daemon restarted with these keys in fips.yaml and peered.
#[derive(Debug, Clone, PartialEq, Serialize)]
pub struct RetrySection {
pub base_interval_secs: u64,
pub max_backoff_secs: u64,
pub max_retries: u32,
}
/// Session-handshake resend pacing (`node.rate_limit.*`). Stock 1s x2.0
/// gaps out to 8-16s between resends exactly when a route has just
/// appeared; 400ms x1.5 keeps continuous coverage through the connect
/// window (phone measured session-after-route: 225ms).
#[derive(Debug, Clone, PartialEq, Serialize)]
pub struct RateLimitSection {
pub handshake_resend_interval_ms: u64,
pub handshake_resend_backoff: f64,
pub handshake_max_resends: u32,
}
#[derive(Debug, Clone, PartialEq, Serialize)]
@ -60,6 +84,14 @@ pub struct IdentitySection {
#[derive(Debug, Clone, PartialEq, Serialize)]
pub struct DiscoverySection {
/// Lookup completion timeout. Lookups fired while the tree position is
/// still settling are doomed; failing them fast (5s, not 10s) lets the
/// 1s-backoff retry find the route the moment it exists.
pub timeout_secs: u64,
pub backoff_base_secs: u64,
pub backoff_max_secs: u64,
pub retry_interval_secs: u64,
pub max_attempts: u8,
pub lan: LanDiscoverySection,
}
@ -123,8 +155,23 @@ impl Default for FipsConfig {
node: NodeSection {
identity: IdentitySection { persistent: true },
discovery: DiscoverySection {
timeout_secs: 5,
backoff_base_secs: 1,
backoff_max_secs: 30,
retry_interval_secs: 2,
max_attempts: 3,
lan: LanDiscoverySection { enabled: true },
},
retry: RetrySection {
base_interval_secs: 1,
max_backoff_secs: 30,
max_retries: 30,
},
rate_limit: RateLimitSection {
handshake_resend_interval_ms: 400,
handshake_resend_backoff: 1.5,
handshake_max_resends: 10,
},
},
tun: TunSection {
enabled: true,
@ -318,8 +365,21 @@ node:
identity:
persistent: true
discovery:
timeout_secs: 5
backoff_base_secs: 1
backoff_max_secs: 30
retry_interval_secs: 2
max_attempts: 3
lan:
enabled: true
retry:
base_interval_secs: 1
max_backoff_secs: 30
max_retries: 30
rate_limit:
handshake_resend_interval_ms: 400
handshake_resend_backoff: 1.5
handshake_max_resends: 10
tun:
enabled: true
name: fips0

View File

@ -988,6 +988,43 @@ impl Server {
main_addr,
));
// The mesh is IPv6-only: a phone reaching the node over its fips0
// ULA lands on port 80 over v6, where a 0.0.0.0 listener never
// answers — the UI was structurally unreachable over the mesh
// (RST -> ERR_CONNECTION_ABORTED, confirmed 2026-07-26: v4:80 = 200,
// v6:80 = refused). Mirror an IPv4-any main listener with a
// V6ONLY [::] socket on the same port — v6-only so it coexists
// with the v4 listener regardless of net.ipv6.bindv6only.
let v4_any_port = match main_addr {
SocketAddr::V4(v4) if v4.ip().is_unspecified() => Some(v4.port()),
_ => None,
};
let v6_task = if let Some(port) = v4_any_port {
let v6_addr = SocketAddr::new(
std::net::IpAddr::V6(std::net::Ipv6Addr::UNSPECIFIED),
port,
);
match bind_v6_only(v6_addr) {
Ok(listener) => {
info!("IPv6 web listener bound {} (mesh ULA reachable)", v6_addr);
Some(tokio::spawn(accept_loop(
self.api_handler.clone(),
listener,
active_connections.clone(),
false, // same semantics as the main listener
tx.subscribe(),
v6_addr,
)))
}
Err(e) => {
warn!("IPv6 web listener bind {} failed: {} — UI stays v4-only", v6_addr, e);
None
}
}
} else {
None
};
// Peer listener: late-binding so we don't need an archipelago
// restart when fips0 comes up after onboarding.
let peer_task = tokio::spawn(peer_late_bind_loop(
@ -1012,6 +1049,9 @@ impl Server {
}
let _ = main_task.await;
if let Some(t) = v6_task {
let _ = t.await;
}
let _ = peer_task.await;
info!("Shutdown complete");
@ -1019,6 +1059,24 @@ impl Server {
}
}
/// Bind a V6ONLY `[::]` TCP listener. V6ONLY is set explicitly so the
/// socket never claims the IPv4 side (which the main listener owns) —
/// without it, Linux hosts with `net.ipv6.bindv6only=0` would fail with
/// EADDRINUSE.
fn bind_v6_only(addr: SocketAddr) -> std::io::Result<tokio::net::TcpListener> {
let socket = socket2::Socket::new(
socket2::Domain::IPV6,
socket2::Type::STREAM,
Some(socket2::Protocol::TCP),
)?;
socket.set_only_v6(true)?;
socket.set_reuse_address(true)?;
socket.set_nonblocking(true)?;
socket.bind(&addr.into())?;
socket.listen(1024)?;
tokio::net::TcpListener::from_std(socket.into())
}
/// Poll every 30s for `fips0`'s ULA; when it appears, bind the peer
/// listener and run the normal accept loop. If the bind fails (port
/// already taken, permissions), log and keep retrying. Returns on