19 Commits

Author SHA1 Message Date
Razvan Dimescu
82cc588c67 docs: explain the two DoT cert modes in README
Expands the DoT paragraph to make the trust model explicit. The
previous version said "self-signed or bring your own cert" without
explaining when to pick which or what the user experience looks like.

The two modes close numa's gap vs AdGuard Home: BYO cert mode is
functionally identical (Let's Encrypt via DNS-01 + cert_path/key_path),
and the self-signed mode is numa's advantage on LAN-only deploys.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
bc54ea930f docs: document DNS-over-TLS listener in README
Adds DoT to the four existing touchpoints in the README where the
feature naturally belongs:

- Hero paragraph: mentions DoT alongside DNSSEC as a headline feature
- Ad Blocking & Privacy section: dedicated paragraph with RFC 7858
  reference, config hint, and the ALPN strictness guarantee
- Comparison table: new "Encrypted clients (DoT listener)" row.
  Pi-hole "Needs stunnel sidecar" (verified — Pi-hole explicitly
  closed the native-DoT feature request as out of scope; community
  uses stunnel or AdGuard DNS Proxy as a TLS terminator)
- Roadmap: checks off "DNS-over-TLS listener" alongside the existing
  DoH entry

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
7001ba2e51 chore: bump version to 0.10.0
v0.10.0 ships DNS-over-TLS. Tagged release v0.10.0 on main after
merge will pick up this Cargo.toml version, keeping tag and manifest
aligned for release.yml.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
6887c8e02e refactor: move data_dir override from env var to [server] TOML field
Reverts the NUMA_DATA_DIR env var added in the previous commit and
replaces it with a [server] data_dir TOML field. Numa already has a
well-developed config system; adding a parallel env-var mechanism
for a single knob was wrong.

The principle: TOML is for application behavior configuration. Env
vars are for bootstrap values (HOME, SUDO_USER to discover paths
before config loads) and standard ecosystem conventions (RUST_LOG).
data_dir is neither — it's an app knob, so it belongs in the TOML.

Changes:
- lib.rs::data_dir() reverts to the platform-specific fallback only
- config.rs adds `data_dir: Option<PathBuf>` to ServerConfig
- main.rs resolves config.server.data_dir with fallback to
  numa::data_dir() and passes it to build_tls_config, then stores the
  resolved path on ctx.data_dir for downstream consumers
- tls.rs::build_tls_config takes `data_dir: &Path` as an explicit
  parameter instead of calling crate::data_dir() behind the caller's
  back. regenerate_tls and dot.rs self_signed_tls now pass
  &ctx.data_dir, honoring whatever path the config resolved to
- tests/integration.sh Suite 6 uses `data_dir = "$NUMA_DATA"` in its
  test TOML instead of the NUMA_DATA_DIR env var prefix
- numa.toml gains a commented-out data_dir example

No behavior change for existing production deployments (the default
path is unchanged). Test harness is now fully config-driven, and
containerized deploys can override data_dir via mount+config without
needing env var injection.

127/127 unit tests pass, Suite 6 passes end-to-end.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
7f52bd8a32 test: Suite 6 — proxy + DoT coexistence, NUMA_DATA_DIR override
Adds integration test coverage for the realistic production shape
where both the HTTPS proxy and DoT are enabled simultaneously. This
was previously untested — every existing suite had either one or the
other, so the interaction path was implicit.

What Suite 6 verifies:
- Both listeners bind without panic
- DoT still resolves queries with the proxy enabled
- Proxy HTTPS handshake still works with DoT enabled
- Both certs validate against the same shared CA

To run non-root, adds a NUMA_DATA_DIR env var override to data_dir()
that lets callers point the CA/cert storage at any writable path.
Useful beyond tests: containerized deployments, CI runners, dev
testing without sudo. The fallback is the existing platform-specific
path (unix: /usr/local/var/numa, windows: %PROGRAMDATA%\numa).

Suite 6 sets NUMA_DATA_DIR=/tmp/numa-integration-data before
starting numa, then trusts the generated CA at $NUMA_DATA_DIR/ca.pem
for both kdig (DoT query) and openssl s_client (HTTPS proxy
handshake) verification.

All 6 suites, 32 checks, run non-root and pass locally.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
c98e6c3ea9 fix: install rustls crypto provider when loading user DoT cert
Adds tests/integration.sh Suite 5 (DoT via kdig + openssl) and
fixes a startup panic caught by it.

Bug: when [dot] cert_path/key_path was set AND [proxy] was disabled,
numa panicked on the first DoT handshake with "Could not
automatically determine the process-level CryptoProvider from Rustls
crate features". In normal deployments the proxy's build_tls_config
installs the default provider as a side effect, masking the missing
call in dot.rs::load_tls_config. Disable the proxy and the panic
surfaces. Fix: call
rustls::crypto::ring::default_provider().install_default() at the
top of load_tls_config (no-op if already installed).

Suite 5 exercises:
- DoT listener binds on configured port
- Resolves a local zone A record over TLS (kdig +tls)
- Persistent connection reuse (kdig +keepopen, 3 queries, 1 handshake)
- ALPN "dot" negotiation (openssl s_client -alpn dot)
- ALPN mismatch rejected with no_application_protocol (openssl -alpn h2)

Uses a pre-generated cert at /tmp so the test runs non-root.
Skips gracefully if kdig or openssl aren't installed.

Also: Dockerfile now EXPOSE 853/tcp so docker run -p 853:853 works
out of the box when users enable DoT.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
186e709373 test: verify DoT server rejects mismatched ALPN
Adds dot_rejects_non_dot_alpn to assert the rustls server enforces
ALPN strictness rather than silently accepting a mismatched
negotiation. This is the load-bearing behavior behind the cross-
protocol confusion defense — without enforcement, the ALPN "dot"
advertisement is just a sign hung on an unlocked door.

Refactors test_tls_configs to return the leaf cert DER instead of a
prebuilt client config, and adds a dot_client(cert_der, alpn) helper
so each test can build a client config with the ALPN list it needs.
The five existing DoT tests gain one line each to call dot_client
with dot_alpn(); behavior unchanged.

127/127 tests pass.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
bacc49667a fix: DoT cert needs explicit {tld}.{tld} SAN, not just *.{tld} wildcard
self_signed_tls was passing an empty service_names list, so the
generated cert only had the *.numa wildcard SAN. Strict TLS clients
(browsers, possibly some iOS versions) reject wildcards under
single-label TLDs — see the existing comment in tls.rs explaining
why the proxy lists each service explicitly.

setup-phone's mobileconfig sends ServerName "numa.numa" as SNI, so
the DoT cert must have an explicit numa.numa SAN. Pass proxy_tld
itself as a service name, mirroring how main.rs already registers
"numa" as a service for the proxy's TLS cert.

Test fixture updated to mirror the production SAN shape (*.numa +
numa.numa) and switched the client to SNI "numa.numa", so the
existing DoT test suite implicitly exercises the SNI path used by
setup-phone clients.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
7d0fe19462 style: drop narrating comments on dot_alpn and ALPN test
Both were restating what the code already said — dot_alpn's doc
narrated the function name and the test comment restated the
assertion. RFC 7858 §3.2 is already cited on self_signed_tls and
build_tls_config where the "why" actually matters.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
1632fc36f2 feat: DoT write timeout and ALPN "dot" advertisement
Two DoS/interop hardening items:

1. Bound write_framed by WRITE_TIMEOUT (10s) so a slow-reader
   attacker can't indefinitely hold a worker task and its connection
   permit. Symmetric to the existing handshake timeout.

2. Advertise ALPN "dot" per RFC 7858 §3.2. Required by some strict
   DoT clients (newer Apple stacks, some Android versions). rustls
   ServerConfig exposes alpn_protocols as a pub field so we set it
   after with_single_cert:
   - load_tls_config (user-provided cert/key): set directly
   - self_signed_tls (new, replaces fallback_tls): builds a fresh
     DoT-specific TLS config via build_tls_config with the ALPN list

   build_tls_config now takes an `alpn: Vec<Vec<u8>>` parameter so
   DoT and the proxy can pass different ALPN lists while sharing the
   same CA. Proxy callers pass Vec::new() (unchanged behavior).

   Dropped the ctx.tls_config reuse branch: we can't mutate a shared
   Arc<ServerConfig> to add DoT-specific ALPN, and reusing the proxy
   config was already quietly broken re: SAN (proxy cert covers
   *.{tld}, not the DoT server's bind hostname/IP).

Added dot_negotiates_alpn test that asserts conn.alpn_protocol()
returns Some(b"dot") after handshake. 126/126 tests pass.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
0a73cdf4db docs: add commented-out [dot] example to numa.toml
Matches the style of the other opt-in sections (blocking, dnssec, lan).
Documents all five DotConfig fields with their defaults.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
2b0c4e3d5e refactor: trim DoT listener — let-else reads, drop MIN_MSG_LEN and redundant localhost test
- Collapse two 4-arm read/timeout matches to let-else (lose one
  defensive debug log on payload-read timeout; idle timeouts are
  routine on persistent DoT connections anyway)
- Drop MIN_MSG_LEN: DnsPacket::from_buffer rejects truncated input
  on its own, and BytePacketBuffer is zero-init so buf[0..2] for
  sub-2-byte messages just yields a harmless FORMERR with id=0
- Inline ACCEPT_ERROR_BACKOFF (single use site)
- Drop the partial cert/key warning: missing one of cert_path/
  key_path silently falls back to self-signed; users see the
  self-signed cert at startup and figure it out
- Drop dot_localhost_resolution test: RFC 6761 localhost is tested
  in ctx.rs; this test only verified DoT transport, which
  dot_resolves_local_zone already covers
- Drop self-documenting comment in dot_multiple_queries_on_persistent_connection

Net -32 lines, 125/125 tests pass, no behavior change users would notice.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
357c710ec4 style: rustfmt
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
7742858b7b refactor: simplify DoT cert/key match and extract send_response helper
- Flatten 4-arm cert/key match in start_dot to 2 arms with the
  partial-config warning hoisted into a one-liner above the match.
- Extract send_response() that serializes a DnsPacket and writes it
  framed, used by both the FORMERR-on-parse-error and SERVFAIL-on-
  resolve-error paths. Removes duplicated buffer/write/log boilerplate
  and unifies the rescode logging via {:?}.

No behavior change; 126/126 tests still pass.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
1239ed0e72 fix: parse DoT queries up-front and echo question in SERVFAIL
Address review findings on PR #25:

- Refactor resolve_query to take a pre-parsed DnsPacket. Parse-error
  handling moves to the UDP caller, eliminating the double warn! line
  on malformed UDP queries.
- Enforce MIN_MSG_LEN=12 (DNS header) in handle_dot_connection so
  query_id extraction is always reading client-sent bytes, not the
  zeroed buffer tail.
- Parse the DoT query before calling resolve_query and retain it, so
  SERVFAIL responses can echo the original question section via
  response_from(). Parse failures send FORMERR with the client id.
- Extract write_framed() helper for length-prefix + flush, reused by
  success, SERVFAIL, and FORMERR paths.
- Back off 100ms on listener.accept() errors to avoid tight-looping
  on fd exhaustion.
- Replace the hardcoded 127.0.0.1:53 upstream in dot_nxdomain_for_unknown
  with a bound-but-unresponsive UDP socket owned by the test, making it
  independent of the host's local resolver. Test now runs in ~220ms
  (timeout lowered to 200ms) instead of 3s and asserts the question is
  echoed in the SERVFAIL response.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
cb54ab3dfc fix: harden DoT listener against slowloris and stale handshakes
- Add 10s timeout on TLS handshake — prevents clients from holding a
  semaphore permit without completing the handshake
- Add IDLE_TIMEOUT on payload read_exact — prevents slowloris after
  sending a valid length prefix then trickling bytes
- Extract accept_loop() shared between start_dot and tests — eliminates
  duplicated accept logic that could drift
- Add 5s timeout on TCP reads in recursive test mock server

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
aa8923b2c6 fix: add debug logging for DoT SERVFAIL serialization failure, TC-bit TODO
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
14efc51340 fix: send SERVFAIL on DoT resolve errors, extract shared connection handler
- Send SERVFAIL response (with correct query ID) when resolve_query
  fails, preventing DoT clients from hanging until idle timeout
- Extract handle_dot_connection() so tests use the same logic as
  production, eliminating duplicated accept/read/resolve loop
- Replace magic 4096 with named MAX_MSG_LEN constant tied to BUF_SIZE
- Add flush() after each TLS write to prevent buffered responses
- Extract fallback_tls() helper, handle partial cert/key config,
  support IPv6 bind address, remove redundant crypto provider init

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
Razvan Dimescu
e4350ae81c feat: add DNS-over-TLS (DoT) listener (RFC 7858)
Refactor handle_query into transport-agnostic resolve_query that returns
a BytePacketBuffer, keeping the UDP path zero-alloc. Add a TLS listener
on port 853 with persistent connections, idle timeout, connection limits,
and coalesced writes. Supports user-provided certs or self-signed CA
fallback. Includes 5 integration tests.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-08 02:53:43 +03:00
13 changed files with 954 additions and 33 deletions

12
Cargo.lock generated
View File

@@ -1143,7 +1143,7 @@ dependencies = [
[[package]] [[package]]
name = "numa" name = "numa"
version = "0.9.1" version = "0.10.0"
dependencies = [ dependencies = [
"arc-swap", "arc-swap",
"axum", "axum",
@@ -1159,6 +1159,7 @@ dependencies = [
"reqwest", "reqwest",
"ring", "ring",
"rustls", "rustls",
"rustls-pemfile",
"serde", "serde",
"serde_json", "serde_json",
"socket2 0.5.10", "socket2 0.5.10",
@@ -1546,6 +1547,15 @@ dependencies = [
"zeroize", "zeroize",
] ]
[[package]]
name = "rustls-pemfile"
version = "2.2.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "dce314e5fee3f39953d46bb63bb8a46d40c2f8fb7cc5a3b6cab2bde9721d6e50"
dependencies = [
"rustls-pki-types",
]
[[package]] [[package]]
name = "rustls-pki-types" name = "rustls-pki-types"
version = "1.14.0" version = "1.14.0"

View File

@@ -1,6 +1,6 @@
[package] [package]
name = "numa" name = "numa"
version = "0.9.1" version = "0.10.0"
authors = ["razvandimescu <razvan@dimescu.com>"] authors = ["razvandimescu <razvan@dimescu.com>"]
edition = "2021" edition = "2021"
description = "Portable DNS resolver in Rust — .numa local domains, ad blocking, developer overrides, DNS-over-HTTPS" description = "Portable DNS resolver in Rust — .numa local domains, ad blocking, developer overrides, DNS-over-HTTPS"
@@ -29,6 +29,7 @@ rustls = "0.23"
tokio-rustls = "0.26" tokio-rustls = "0.26"
arc-swap = "1" arc-swap = "1"
ring = "0.17" ring = "0.17"
rustls-pemfile = "2.2.0"
[dev-dependencies] [dev-dependencies]
criterion = { version = "0.5", features = ["html_reports"] } criterion = { version = "0.5", features = ["html_reports"] }

View File

@@ -13,5 +13,5 @@ RUN cargo build --release
FROM alpine:3.20 FROM alpine:3.20
COPY --from=builder /app/target/release/numa /usr/local/bin/numa COPY --from=builder /app/target/release/numa /usr/local/bin/numa
EXPOSE 53/udp 80/tcp 443/tcp 5380/tcp EXPOSE 53/udp 80/tcp 443/tcp 853/tcp 5380/tcp
ENTRYPOINT ["numa"] ENTRYPOINT ["numa"]

View File

@@ -8,7 +8,7 @@
A portable DNS resolver in a single binary. Block ads on any network, name your local services (`frontend.numa`), and override any hostname with auto-revert — all from your laptop, no cloud account or Raspberry Pi required. A portable DNS resolver in a single binary. Block ads on any network, name your local services (`frontend.numa`), and override any hostname with auto-revert — all from your laptop, no cloud account or Raspberry Pi required.
Built from scratch in Rust. Zero DNS libraries. RFC 1035 wire protocol parsed by hand. Caching, ad blocking, and local service domains out of the box. Optional recursive resolution from root nameservers with full DNSSEC chain-of-trust validation. One ~8MB binary, everything embedded. Built from scratch in Rust. Zero DNS libraries. RFC 1035 wire protocol parsed by hand. Caching, ad blocking, and local service domains out of the box. Optional recursive resolution from root nameservers with full DNSSEC chain-of-trust validation, plus a DNS-over-TLS listener for encrypted client connections (iOS Private DNS, systemd-resolved, etc.). One ~8MB binary, everything embedded.
![Numa dashboard](assets/hero-demo.gif) ![Numa dashboard](assets/hero-demo.gif)
@@ -67,6 +67,13 @@ Three resolution modes:
DNSSEC validates the full chain of trust: RRSIG signatures, DNSKEY verification, DS delegation, NSEC/NSEC3 denial proofs. [Read how it works →](https://numa.rs/blog/posts/dnssec-from-scratch.html) DNSSEC validates the full chain of trust: RRSIG signatures, DNSKEY verification, DS delegation, NSEC/NSEC3 denial proofs. [Read how it works →](https://numa.rs/blog/posts/dnssec-from-scratch.html)
**DNS-over-TLS listener** (RFC 7858) — accept encrypted queries on port 853 from strict clients like iOS Private DNS, systemd-resolved, or stubby. Two modes:
- **Self-signed** (default) — numa generates a local CA automatically. Works on any network with zero DNS setup, but clients must manually trust the CA (on macOS/Linux add to the system trust store; on iOS install a `.mobileconfig`).
- **Bring-your-own cert** — point `[dot] cert_path` / `key_path` at a publicly-trusted cert (e.g., Let's Encrypt via DNS-01 challenge on a domain pointing at your numa instance). Clients connect without any trust-store setup — same UX as AdGuard Home or Cloudflare `1.1.1.1`.
ALPN `"dot"` is advertised and enforced in both modes; a handshake with mismatched ALPN is rejected as a cross-protocol confusion defense.
## LAN Discovery ## LAN Discovery
Run Numa on multiple machines. They find each other automatically via mDNS: Run Numa on multiple machines. They find each other automatically via mDNS:
@@ -96,6 +103,7 @@ From Machine B: `curl http://api.numa` → proxied to Machine A's port 8000. Ena
| Ad blocking | Yes | Yes | — | 385K+ domains | | Ad blocking | Yes | Yes | — | 385K+ domains |
| Web admin UI | Full | Full | — | Dashboard | | Web admin UI | Full | Full | — | Dashboard |
| Encrypted upstream (DoH) | Needs cloudflared | Yes | — | Native | | Encrypted upstream (DoH) | Needs cloudflared | Yes | — | Native |
| Encrypted clients (DoT listener) | Needs stunnel sidecar | Yes | Yes | Native (RFC 7858) |
| Portable (laptop) | No (appliance) | No (appliance) | Server | Single binary, macOS/Linux/Windows | | Portable (laptop) | No (appliance) | No (appliance) | Server | Single binary, macOS/Linux/Windows |
| Community maturity | 56K stars, 10 years | 33K stars | 20 years | New | | Community maturity | 56K stars, 10 years | 33K stars | 20 years | New |
@@ -116,6 +124,7 @@ From Machine B: `curl http://api.numa` → proxied to Machine A's port 8000. Ena
- [x] `.numa` local domains — auto TLS, path routing, WebSocket proxy - [x] `.numa` local domains — auto TLS, path routing, WebSocket proxy
- [x] LAN service discovery — mDNS, cross-machine DNS + proxy - [x] LAN service discovery — mDNS, cross-machine DNS + proxy
- [x] DNS-over-HTTPS — encrypted upstream - [x] DNS-over-HTTPS — encrypted upstream
- [x] DNS-over-TLS listener — encrypted client connections (RFC 7858, ALPN strict)
- [x] Recursive resolution + DNSSEC — chain-of-trust, NSEC/NSEC3 - [x] Recursive resolution + DNSSEC — chain-of-trust, NSEC/NSEC3
- [x] SRTT-based nameserver selection - [x] SRTT-based nameserver selection
- [ ] pkarr integration — self-sovereign DNS via Mainline DHT - [ ] pkarr integration — self-sovereign DNS via Mainline DHT

View File

@@ -2,6 +2,11 @@
bind_addr = "0.0.0.0:53" bind_addr = "0.0.0.0:53"
api_port = 5380 api_port = 5380
# api_bind_addr = "127.0.0.1" # default; set to "0.0.0.0" for LAN dashboard access # api_bind_addr = "127.0.0.1" # default; set to "0.0.0.0" for LAN dashboard access
# data_dir = "/usr/local/var/numa" # where numa stores TLS CA and cert material
# (default: /usr/local/var/numa on unix,
# %PROGRAMDATA%\numa on windows). Override for
# containerized deploys or tests that can't
# write to the system path.
# [upstream] # [upstream]
# mode = "forward" # "forward" (default) — relay to upstream # mode = "forward" # "forward" (default) — relay to upstream
@@ -83,6 +88,14 @@ tld = "numa"
# enabled = false # opt-in: verify chain of trust from root KSK # enabled = false # opt-in: verify chain of trust from root KSK
# strict = false # true = SERVFAIL on bogus signatures # strict = false # true = SERVFAIL on bogus signatures
# DNS-over-TLS listener (RFC 7858) — encrypted DNS on port 853
# [dot]
# enabled = false # opt-in: accept DoT queries
# port = 853 # standard DoT port
# bind_addr = "0.0.0.0" # IPv4 or IPv6; unspecified binds all interfaces
# cert_path = "/etc/numa/dot.crt" # PEM cert; omit to use self-signed (proxy CA if available)
# key_path = "/etc/numa/dot.key" # PEM private key; must be set together with cert_path
# LAN service discovery via mDNS (disabled by default — no network traffic unless enabled) # LAN service discovery via mDNS (disabled by default — no network traffic unless enabled)
# [lan] # [lan]
# enabled = true # discover other Numa instances via mDNS (_numa._tcp.local) # enabled = true # discover other Numa instances via mDNS (_numa._tcp.local)

View File

@@ -1,7 +1,7 @@
use std::collections::HashMap; use std::collections::HashMap;
use std::net::Ipv4Addr; use std::net::Ipv4Addr;
use std::net::Ipv6Addr; use std::net::Ipv6Addr;
use std::path::Path; use std::path::{Path, PathBuf};
use serde::Deserialize; use serde::Deserialize;
@@ -29,6 +29,8 @@ pub struct Config {
pub lan: LanConfig, pub lan: LanConfig,
#[serde(default)] #[serde(default)]
pub dnssec: DnssecConfig, pub dnssec: DnssecConfig,
#[serde(default)]
pub dot: DotConfig,
} }
#[derive(Deserialize)] #[derive(Deserialize)]
@@ -39,6 +41,10 @@ pub struct ServerConfig {
pub api_port: u16, pub api_port: u16,
#[serde(default = "default_api_bind_addr")] #[serde(default = "default_api_bind_addr")]
pub api_bind_addr: String, pub api_bind_addr: String,
/// Where numa writes TLS material (CA, leaf certs, regenerated state).
/// Defaults to `crate::data_dir()` (platform-specific system path) if unset.
#[serde(default)]
pub data_dir: Option<PathBuf>,
} }
impl Default for ServerConfig { impl Default for ServerConfig {
@@ -47,6 +53,7 @@ impl Default for ServerConfig {
bind_addr: default_bind_addr(), bind_addr: default_bind_addr(),
api_port: default_api_port(), api_port: default_api_port(),
api_bind_addr: default_api_bind_addr(), api_bind_addr: default_api_bind_addr(),
data_dir: None,
} }
} }
} }
@@ -370,6 +377,41 @@ pub struct DnssecConfig {
pub strict: bool, pub strict: bool,
} }
#[derive(Deserialize, Clone)]
pub struct DotConfig {
#[serde(default)]
pub enabled: bool,
#[serde(default = "default_dot_port")]
pub port: u16,
#[serde(default = "default_dot_bind_addr")]
pub bind_addr: String,
/// Path to TLS certificate (PEM). If None, uses self-signed CA.
#[serde(default)]
pub cert_path: Option<PathBuf>,
/// Path to TLS private key (PEM). If None, uses self-signed CA.
#[serde(default)]
pub key_path: Option<PathBuf>,
}
impl Default for DotConfig {
fn default() -> Self {
DotConfig {
enabled: false,
port: default_dot_port(),
bind_addr: default_dot_bind_addr(),
cert_path: None,
key_path: None,
}
}
}
fn default_dot_port() -> u16 {
853
}
fn default_dot_bind_addr() -> String {
"0.0.0.0".to_string()
}
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::*; use super::*;

View File

@@ -62,24 +62,21 @@ pub struct ServerCtx {
pub dnssec_strict: bool, pub dnssec_strict: bool,
} }
pub async fn handle_query( /// Transport-agnostic DNS resolution. Runs the full pipeline (overrides, blocklist,
mut buffer: BytePacketBuffer, /// cache, upstream, DNSSEC) and returns the serialized response in a buffer.
/// Callers use `.filled()` to get the response bytes without heap allocation.
/// Callers are responsible for parsing the incoming buffer into a `DnsPacket`
/// (and logging parse errors) before calling this function.
pub async fn resolve_query(
query: DnsPacket,
src_addr: SocketAddr, src_addr: SocketAddr,
ctx: &ServerCtx, ctx: &ServerCtx,
) -> crate::Result<()> { ) -> crate::Result<BytePacketBuffer> {
let start = Instant::now(); let start = Instant::now();
let query = match DnsPacket::from_buffer(&mut buffer) {
Ok(packet) => packet,
Err(e) => {
warn!("{} | PARSE ERROR | {}", src_addr, e);
return Ok(());
}
};
let (qname, qtype) = match query.questions.first() { let (qname, qtype) = match query.questions.first() {
Some(q) => (q.name.clone(), q.qtype), Some(q) => (q.name.clone(), q.qtype),
None => return Ok(()), None => return Err("empty question section".into()),
}; };
// Pipeline: overrides -> .tld interception -> blocklist -> local zones -> cache -> upstream // Pipeline: overrides -> .tld interception -> blocklist -> local zones -> cache -> upstream
@@ -306,17 +303,17 @@ pub async fn handle_query(
response.resources.len(), response.resources.len(),
); );
// Serialize response
// TODO: TC bit is UDP-specific; DoT connections could carry up to 65535 bytes.
// Once BytePacketBuffer supports larger buffers, skip truncation for TCP/TLS.
let mut resp_buffer = BytePacketBuffer::new(); let mut resp_buffer = BytePacketBuffer::new();
if response.write(&mut resp_buffer).is_err() { if response.write(&mut resp_buffer).is_err() {
// Response too large for UDP — set TC bit and send header + question only // Response too large — set TC bit and send header + question only
debug!("response too large, setting TC bit for {}", qname); debug!("response too large, setting TC bit for {}", qname);
let mut tc_response = DnsPacket::response_from(&query, response.header.rescode); let mut tc_response = DnsPacket::response_from(&query, response.header.rescode);
tc_response.header.truncated_message = true; tc_response.header.truncated_message = true;
let mut tc_buffer = BytePacketBuffer::new(); resp_buffer = BytePacketBuffer::new();
tc_response.write(&mut tc_buffer)?; tc_response.write(&mut resp_buffer)?;
ctx.socket.send_to(tc_buffer.filled(), src_addr).await?;
} else {
ctx.socket.send_to(resp_buffer.filled(), src_addr).await?;
} }
// Record stats and query log // Record stats and query log
@@ -339,6 +336,30 @@ pub async fn handle_query(
dnssec, dnssec,
}); });
Ok(resp_buffer)
}
/// Handle a DNS query received over UDP. Thin wrapper around resolve_query.
pub async fn handle_query(
mut buffer: BytePacketBuffer,
src_addr: SocketAddr,
ctx: &ServerCtx,
) -> crate::Result<()> {
let query = match DnsPacket::from_buffer(&mut buffer) {
Ok(packet) => packet,
Err(e) => {
warn!("{} | PARSE ERROR | {}", src_addr, e);
return Ok(());
}
};
match resolve_query(query, src_addr, ctx).await {
Ok(resp_buffer) => {
ctx.socket.send_to(resp_buffer.filled(), src_addr).await?;
}
Err(e) => {
warn!("{} | RESOLVE ERROR | {}", src_addr, e);
}
}
Ok(()) Ok(())
} }

542
src/dot.rs Normal file
View File

@@ -0,0 +1,542 @@
use std::net::{IpAddr, SocketAddr};
use std::path::Path;
use std::sync::Arc;
use std::time::Duration;
use log::{debug, error, info, warn};
use rustls::ServerConfig;
use tokio::io::{AsyncReadExt, AsyncWriteExt};
use tokio::net::TcpListener;
use tokio::sync::Semaphore;
use tokio_rustls::TlsAcceptor;
use crate::buffer::BytePacketBuffer;
use crate::config::DotConfig;
use crate::ctx::{resolve_query, ServerCtx};
use crate::header::ResultCode;
use crate::packet::DnsPacket;
const MAX_CONNECTIONS: usize = 512;
const IDLE_TIMEOUT: Duration = Duration::from_secs(30);
const HANDSHAKE_TIMEOUT: Duration = Duration::from_secs(10);
const WRITE_TIMEOUT: Duration = Duration::from_secs(10);
// Matches BytePacketBuffer::BUF_SIZE — RFC 7858 allows up to 65535 but our
// buffer would silently truncate anything larger.
const MAX_MSG_LEN: usize = 4096;
fn dot_alpn() -> Vec<Vec<u8>> {
vec![b"dot".to_vec()]
}
/// Build a TLS ServerConfig for DoT from user-provided cert/key PEM files.
fn load_tls_config(cert_path: &Path, key_path: &Path) -> crate::Result<Arc<ServerConfig>> {
// rustls needs a CryptoProvider installed before ServerConfig::builder().
// The proxy's build_tls_config also does this; we repeat it here because
// running DoT with user-provided certs while the proxy is disabled would
// otherwise panic on first handshake (no default provider).
let _ = rustls::crypto::ring::default_provider().install_default();
let cert_pem = std::fs::read(cert_path)?;
let key_pem = std::fs::read(key_path)?;
let certs: Vec<_> = rustls_pemfile::certs(&mut &cert_pem[..]).collect::<Result<_, _>>()?;
let key = rustls_pemfile::private_key(&mut &key_pem[..])?
.ok_or("no private key found in key file")?;
let mut config = ServerConfig::builder()
.with_no_client_auth()
.with_single_cert(certs, key)?;
config.alpn_protocols = dot_alpn();
Ok(Arc::new(config))
}
/// Build a self-signed DoT TLS config. Can't reuse `ctx.tls_config` (the
/// proxy's shared config) because DoT needs its own ALPN advertisement.
///
/// Pass `proxy_tld` itself as a service name so the cert gets an explicit
/// `{tld}.{tld}` SAN (e.g. "numa.numa") matching the ServerName that
/// setup-phone's mobileconfig sends as SNI. The `*.{tld}` wildcard alone
/// is rejected by strict TLS clients under single-label TLDs (per the
/// note in tls.rs::generate_service_cert).
fn self_signed_tls(ctx: &ServerCtx) -> Option<Arc<ServerConfig>> {
let service_names = [ctx.proxy_tld.clone()];
match crate::tls::build_tls_config(&ctx.proxy_tld, &service_names, dot_alpn(), &ctx.data_dir) {
Ok(cfg) => Some(cfg),
Err(e) => {
warn!(
"DoT: failed to generate self-signed TLS: {} — DoT disabled",
e
);
None
}
}
}
/// Start the DNS-over-TLS listener (RFC 7858).
pub async fn start_dot(ctx: Arc<ServerCtx>, config: &DotConfig) {
let tls_config = match (&config.cert_path, &config.key_path) {
(Some(cert), Some(key)) => match load_tls_config(cert, key) {
Ok(cfg) => cfg,
Err(e) => {
warn!("DoT: failed to load TLS cert/key: {} — DoT disabled", e);
return;
}
},
_ => match self_signed_tls(&ctx) {
Some(cfg) => cfg,
None => return,
},
};
let bind_addr: IpAddr = config
.bind_addr
.parse()
.unwrap_or(IpAddr::V4(std::net::Ipv4Addr::UNSPECIFIED));
let addr = SocketAddr::new(bind_addr, config.port);
let listener = match TcpListener::bind(addr).await {
Ok(l) => l,
Err(e) => {
warn!("DoT: could not bind {} ({}) — DoT disabled", addr, e);
return;
}
};
info!("DoT listening on {}", addr);
accept_loop(listener, TlsAcceptor::from(tls_config), ctx).await;
}
async fn accept_loop(listener: TcpListener, acceptor: TlsAcceptor, ctx: Arc<ServerCtx>) {
let semaphore = Arc::new(Semaphore::new(MAX_CONNECTIONS));
loop {
let (tcp_stream, remote_addr) = match listener.accept().await {
Ok(conn) => conn,
Err(e) => {
error!("DoT: TCP accept error: {}", e);
// Back off to avoid tight-looping on persistent failures (e.g. fd exhaustion).
tokio::time::sleep(Duration::from_millis(100)).await;
continue;
}
};
let permit = match semaphore.clone().try_acquire_owned() {
Ok(p) => p,
Err(_) => {
debug!("DoT: connection limit reached, rejecting {}", remote_addr);
continue;
}
};
let acceptor = acceptor.clone();
let ctx = Arc::clone(&ctx);
tokio::spawn(async move {
let _permit = permit; // held until task exits
let tls_stream =
match tokio::time::timeout(HANDSHAKE_TIMEOUT, acceptor.accept(tcp_stream)).await {
Ok(Ok(s)) => s,
Ok(Err(e)) => {
debug!("DoT: TLS handshake failed from {}: {}", remote_addr, e);
return;
}
Err(_) => {
debug!("DoT: TLS handshake timeout from {}", remote_addr);
return;
}
};
handle_dot_connection(tls_stream, remote_addr, &ctx).await;
});
}
}
/// Handle a single persistent DoT connection (RFC 7858).
/// Reads length-prefixed DNS queries until EOF, idle timeout, or error.
async fn handle_dot_connection<S>(mut stream: S, remote_addr: SocketAddr, ctx: &ServerCtx)
where
S: AsyncReadExt + AsyncWriteExt + Unpin,
{
loop {
// Read 2-byte length prefix (RFC 1035 §4.2.2) with idle timeout
let mut len_buf = [0u8; 2];
let Ok(Ok(_)) = tokio::time::timeout(IDLE_TIMEOUT, stream.read_exact(&mut len_buf)).await
else {
break;
};
let msg_len = u16::from_be_bytes(len_buf) as usize;
if msg_len > MAX_MSG_LEN {
debug!("DoT: oversized message {} from {}", msg_len, remote_addr);
break;
}
let mut buffer = BytePacketBuffer::new();
let Ok(Ok(_)) =
tokio::time::timeout(IDLE_TIMEOUT, stream.read_exact(&mut buffer.buf[..msg_len])).await
else {
break;
};
// Parse query up-front so we can echo its question section in SERVFAIL
// responses when resolve_query fails.
let query = match DnsPacket::from_buffer(&mut buffer) {
Ok(q) => q,
Err(e) => {
warn!("{} | PARSE ERROR | {}", remote_addr, e);
// BytePacketBuffer is zero-initialized, so buf[0..2] reads as 0x0000
// for sub-2-byte messages — harmless FORMERR with id=0.
let query_id = u16::from_be_bytes([buffer.buf[0], buffer.buf[1]]);
let mut resp = DnsPacket::new();
resp.header.id = query_id;
resp.header.response = true;
resp.header.rescode = ResultCode::FORMERR;
if send_response(&mut stream, &resp, remote_addr)
.await
.is_err()
{
break;
}
continue;
}
};
match resolve_query(query.clone(), remote_addr, ctx).await {
Ok(resp_buffer) => {
if write_framed(&mut stream, resp_buffer.filled())
.await
.is_err()
{
break;
}
}
Err(e) => {
warn!("{} | RESOLVE ERROR | {}", remote_addr, e);
// SERVFAIL that echoes the original question section.
let resp = DnsPacket::response_from(&query, ResultCode::SERVFAIL);
if send_response(&mut stream, &resp, remote_addr)
.await
.is_err()
{
break;
}
}
}
}
}
/// Serialize a DNS response and send it framed. Logs serialization failures
/// and returns Err so the caller can tear down the connection.
async fn send_response<S>(
stream: &mut S,
resp: &DnsPacket,
remote_addr: SocketAddr,
) -> std::io::Result<()>
where
S: AsyncWriteExt + Unpin,
{
let mut out_buf = BytePacketBuffer::new();
if resp.write(&mut out_buf).is_err() {
debug!(
"DoT: failed to serialize {:?} response for {}",
resp.header.rescode, remote_addr
);
return Err(std::io::Error::other("serialize failed"));
}
write_framed(stream, out_buf.filled()).await
}
/// Write a DNS message with its 2-byte length prefix, coalesced into one syscall.
/// Bounded by WRITE_TIMEOUT so a stalled reader can't indefinitely hold a worker.
async fn write_framed<S>(stream: &mut S, msg: &[u8]) -> std::io::Result<()>
where
S: AsyncWriteExt + Unpin,
{
let mut out = Vec::with_capacity(2 + msg.len());
out.extend_from_slice(&(msg.len() as u16).to_be_bytes());
out.extend_from_slice(msg);
match tokio::time::timeout(WRITE_TIMEOUT, async {
stream.write_all(&out).await?;
stream.flush().await
})
.await
{
Ok(result) => result,
Err(_) => Err(std::io::Error::other("write timeout")),
}
}
#[cfg(test)]
mod tests {
use super::*;
use std::collections::HashMap;
use std::sync::{Mutex, RwLock};
use rcgen::{CertificateParams, DnType, KeyPair};
use rustls::pki_types::{CertificateDer, PrivateKeyDer, PrivatePkcs8KeyDer, ServerName};
use tokio::io::{AsyncReadExt, AsyncWriteExt};
use crate::buffer::BytePacketBuffer;
use crate::header::ResultCode;
use crate::packet::DnsPacket;
use crate::question::QueryType;
use crate::record::DnsRecord;
/// Generate a self-signed DoT server config and return its leaf cert DER
/// so callers can build matching client configs with arbitrary ALPN.
fn test_tls_configs() -> (Arc<ServerConfig>, CertificateDer<'static>) {
let _ = rustls::crypto::ring::default_provider().install_default();
// Mirror production self_signed_tls SAN shape: *.numa wildcard plus
// explicit numa.numa apex (the ServerName setup-phone uses as SNI).
let key_pair = KeyPair::generate().unwrap();
let mut params = CertificateParams::default();
params
.distinguished_name
.push(DnType::CommonName, "Numa .numa services");
params.subject_alt_names = vec![
rcgen::SanType::DnsName("*.numa".try_into().unwrap()),
rcgen::SanType::DnsName("numa.numa".try_into().unwrap()),
];
let cert = params.self_signed(&key_pair).unwrap();
let cert_der = CertificateDer::from(cert.der().to_vec());
let key_der = PrivateKeyDer::Pkcs8(PrivatePkcs8KeyDer::from(key_pair.serialize_der()));
let mut server_config = ServerConfig::builder()
.with_no_client_auth()
.with_single_cert(vec![cert_der.clone()], key_der)
.unwrap();
server_config.alpn_protocols = dot_alpn();
(Arc::new(server_config), cert_der)
}
/// Build a TLS client config that trusts `cert_der` and advertises the
/// given ALPN protocols. Used by tests to vary ALPN per test case.
fn dot_client(
cert_der: &CertificateDer<'static>,
alpn: Vec<Vec<u8>>,
) -> Arc<rustls::ClientConfig> {
let mut root_store = rustls::RootCertStore::empty();
root_store.add(cert_der.clone()).unwrap();
let mut config = rustls::ClientConfig::builder()
.with_root_certificates(root_store)
.with_no_client_auth();
config.alpn_protocols = alpn;
Arc::new(config)
}
/// Spin up a DoT listener with a test TLS config. Returns the bind addr
/// and the leaf cert DER so callers can build clients with arbitrary ALPN.
/// The upstream is pointed at a bound-but-unresponsive UDP socket we own, so
/// any query that escapes to the upstream path times out deterministically
/// (SERVFAIL) regardless of what the host has running on port 53.
async fn spawn_dot_server() -> (SocketAddr, CertificateDer<'static>) {
let (server_tls, cert_der) = test_tls_configs();
let socket = tokio::net::UdpSocket::bind("127.0.0.1:0").await.unwrap();
// Bind an unresponsive upstream and leak it so it lives for the test duration.
let blackhole = Box::leak(Box::new(std::net::UdpSocket::bind("127.0.0.1:0").unwrap()));
let upstream_addr = blackhole.local_addr().unwrap();
let ctx = Arc::new(ServerCtx {
socket,
zone_map: {
let mut m = HashMap::new();
let mut inner = HashMap::new();
inner.insert(
QueryType::A,
vec![DnsRecord::A {
domain: "dot-test.example".to_string(),
addr: std::net::Ipv4Addr::new(10, 0, 0, 1),
ttl: 300,
}],
);
m.insert("dot-test.example".to_string(), inner);
m
},
cache: RwLock::new(crate::cache::DnsCache::new(100, 60, 86400)),
stats: Mutex::new(crate::stats::ServerStats::new()),
overrides: RwLock::new(crate::override_store::OverrideStore::new()),
blocklist: RwLock::new(crate::blocklist::BlocklistStore::new()),
query_log: Mutex::new(crate::query_log::QueryLog::new(100)),
services: Mutex::new(crate::service_store::ServiceStore::new()),
lan_peers: Mutex::new(crate::lan::PeerStore::new(90)),
forwarding_rules: Vec::new(),
upstream: Mutex::new(crate::forward::Upstream::Udp(upstream_addr)),
upstream_auto: false,
upstream_port: 53,
lan_ip: Mutex::new(std::net::Ipv4Addr::LOCALHOST),
timeout: Duration::from_millis(200),
proxy_tld: "numa".to_string(),
proxy_tld_suffix: ".numa".to_string(),
lan_enabled: false,
config_path: String::new(),
config_found: false,
config_dir: std::path::PathBuf::from("/tmp"),
data_dir: std::path::PathBuf::from("/tmp"),
tls_config: Some(arc_swap::ArcSwap::from(server_tls)),
upstream_mode: crate::config::UpstreamMode::Forward,
root_hints: Vec::new(),
srtt: RwLock::new(crate::srtt::SrttCache::new(true)),
inflight: Mutex::new(HashMap::new()),
dnssec_enabled: false,
dnssec_strict: false,
});
let listener = TcpListener::bind("127.0.0.1:0").await.unwrap();
let addr = listener.local_addr().unwrap();
let tls_config = Arc::clone(&*ctx.tls_config.as_ref().unwrap().load());
let acceptor = TlsAcceptor::from(tls_config);
tokio::spawn(accept_loop(listener, acceptor, ctx));
(addr, cert_der)
}
/// Open a TLS connection to the DoT server and return the stream.
/// Uses SNI "numa.numa" to mirror what setup-phone's mobileconfig sends.
async fn dot_connect(
addr: SocketAddr,
client_config: &Arc<rustls::ClientConfig>,
) -> tokio_rustls::client::TlsStream<tokio::net::TcpStream> {
let connector = tokio_rustls::TlsConnector::from(Arc::clone(client_config));
let tcp = tokio::net::TcpStream::connect(addr).await.unwrap();
connector
.connect(ServerName::try_from("numa.numa").unwrap(), tcp)
.await
.unwrap()
}
/// Send a DNS query over a DoT stream and read the response.
async fn dot_exchange(
stream: &mut tokio_rustls::client::TlsStream<tokio::net::TcpStream>,
query: &DnsPacket,
) -> DnsPacket {
let mut buf = BytePacketBuffer::new();
query.write(&mut buf).unwrap();
let msg = buf.filled();
let mut out = Vec::with_capacity(2 + msg.len());
out.extend_from_slice(&(msg.len() as u16).to_be_bytes());
out.extend_from_slice(msg);
stream.write_all(&out).await.unwrap();
let mut len_buf = [0u8; 2];
stream.read_exact(&mut len_buf).await.unwrap();
let resp_len = u16::from_be_bytes(len_buf) as usize;
let mut data = vec![0u8; resp_len];
stream.read_exact(&mut data).await.unwrap();
let mut resp_buf = BytePacketBuffer::from_bytes(&data);
DnsPacket::from_buffer(&mut resp_buf).unwrap()
}
#[tokio::test]
async fn dot_resolves_local_zone() {
let (addr, cert_der) = spawn_dot_server().await;
let client_config = dot_client(&cert_der, dot_alpn());
let mut stream = dot_connect(addr, &client_config).await;
let query = DnsPacket::query(0x1234, "dot-test.example", QueryType::A);
let resp = dot_exchange(&mut stream, &query).await;
assert_eq!(resp.header.id, 0x1234);
assert!(resp.header.response);
assert_eq!(resp.header.rescode, ResultCode::NOERROR);
assert_eq!(resp.answers.len(), 1);
match &resp.answers[0] {
DnsRecord::A { domain, addr, ttl } => {
assert_eq!(domain, "dot-test.example");
assert_eq!(*addr, std::net::Ipv4Addr::new(10, 0, 0, 1));
assert_eq!(*ttl, 300);
}
other => panic!("expected A record, got {:?}", other),
}
}
#[tokio::test]
async fn dot_multiple_queries_on_persistent_connection() {
let (addr, cert_der) = spawn_dot_server().await;
let client_config = dot_client(&cert_der, dot_alpn());
let mut stream = dot_connect(addr, &client_config).await;
for i in 0..3u16 {
let query = DnsPacket::query(0xA000 + i, "dot-test.example", QueryType::A);
let resp = dot_exchange(&mut stream, &query).await;
assert_eq!(resp.header.id, 0xA000 + i);
assert_eq!(resp.header.rescode, ResultCode::NOERROR);
assert_eq!(resp.answers.len(), 1);
}
}
#[tokio::test]
async fn dot_nxdomain_for_unknown() {
let (addr, cert_der) = spawn_dot_server().await;
let client_config = dot_client(&cert_der, dot_alpn());
let mut stream = dot_connect(addr, &client_config).await;
let query = DnsPacket::query(0xBEEF, "nonexistent.test", QueryType::A);
let resp = dot_exchange(&mut stream, &query).await;
assert_eq!(resp.header.id, 0xBEEF);
assert!(resp.header.response);
// Query goes to the blackhole upstream which never replies → SERVFAIL.
// The SERVFAIL response echoes the question section.
assert_eq!(resp.header.rescode, ResultCode::SERVFAIL);
assert_eq!(resp.questions.len(), 1);
assert_eq!(resp.questions[0].name, "nonexistent.test");
}
#[tokio::test]
async fn dot_negotiates_alpn() {
let (addr, cert_der) = spawn_dot_server().await;
let client_config = dot_client(&cert_der, dot_alpn());
let stream = dot_connect(addr, &client_config).await;
let (_io, conn) = stream.get_ref();
assert_eq!(conn.alpn_protocol(), Some(&b"dot"[..]));
}
#[tokio::test]
async fn dot_rejects_non_dot_alpn() {
// Cross-protocol confusion defense: a client that only offers "h2"
// (e.g. an HTTP/2 client mistakenly hitting :853) must not complete
// a TLS handshake with the DoT server. Verifies the rustls server
// sends `no_application_protocol` rather than silently negotiating.
let (addr, cert_der) = spawn_dot_server().await;
let client_config = dot_client(&cert_der, vec![b"h2".to_vec()]);
let connector = tokio_rustls::TlsConnector::from(client_config);
let tcp = tokio::net::TcpStream::connect(addr).await.unwrap();
let result = connector
.connect(ServerName::try_from("numa.numa").unwrap(), tcp)
.await;
assert!(
result.is_err(),
"DoT server must reject ALPN that doesn't include \"dot\""
);
}
#[tokio::test]
async fn dot_concurrent_connections() {
let (addr, cert_der) = spawn_dot_server().await;
let client_config = dot_client(&cert_der, dot_alpn());
let mut handles = Vec::new();
for i in 0..5u16 {
let cfg = Arc::clone(&client_config);
handles.push(tokio::spawn(async move {
let mut stream = dot_connect(addr, &cfg).await;
let query = DnsPacket::query(0xC000 + i, "dot-test.example", QueryType::A);
let resp = dot_exchange(&mut stream, &query).await;
assert_eq!(resp.header.id, 0xC000 + i);
assert_eq!(resp.header.rescode, ResultCode::NOERROR);
assert_eq!(resp.answers.len(), 1);
}));
}
for h in handles {
h.await.unwrap();
}
}
}

View File

@@ -5,6 +5,7 @@ pub mod cache;
pub mod config; pub mod config;
pub mod ctx; pub mod ctx;
pub mod dnssec; pub mod dnssec;
pub mod dot;
pub mod forward; pub mod forward;
pub mod header; pub mod header;
pub mod lan; pub mod lan;
@@ -65,7 +66,9 @@ fn config_dir_unix() -> std::path::PathBuf {
std::path::PathBuf::from("/usr/local/var/numa") std::path::PathBuf::from("/usr/local/var/numa")
} }
/// System-wide data directory for TLS certs. /// Default system-wide data directory for TLS certs. Overridable via
/// `[server] data_dir = "..."` in numa.toml — this function only provides
/// the fallback when the config doesn't set it.
/// Unix: /usr/local/var/numa /// Unix: /usr/local/var/numa
/// Windows: %PROGRAMDATA%\numa /// Windows: %PROGRAMDATA%\numa
pub fn data_dir() -> std::path::PathBuf { pub fn data_dir() -> std::path::PathBuf {

View File

@@ -204,10 +204,23 @@ async fn main() -> numa::Result<()> {
let forwarding_rules = system_dns.forwarding_rules; let forwarding_rules = system_dns.forwarding_rules;
// Resolve data_dir from config, falling back to the platform default.
// Used for TLS CA storage below and stored on ServerCtx for runtime use.
let resolved_data_dir = config
.server
.data_dir
.clone()
.unwrap_or_else(numa::data_dir);
// Build initial TLS config before ServerCtx (so ArcSwap is ready at construction) // Build initial TLS config before ServerCtx (so ArcSwap is ready at construction)
let initial_tls = if config.proxy.enabled && config.proxy.tls_port > 0 { let initial_tls = if config.proxy.enabled && config.proxy.tls_port > 0 {
let service_names = service_store.names(); let service_names = service_store.names();
match numa::tls::build_tls_config(&config.proxy.tld, &service_names) { match numa::tls::build_tls_config(
&config.proxy.tld,
&service_names,
Vec::new(),
&resolved_data_dir,
) {
Ok(tls_config) => Some(ArcSwap::from(tls_config)), Ok(tls_config) => Some(ArcSwap::from(tls_config)),
Err(e) => { Err(e) => {
log::warn!("TLS setup failed, HTTPS proxy disabled: {}", e); log::warn!("TLS setup failed, HTTPS proxy disabled: {}", e);
@@ -248,7 +261,7 @@ async fn main() -> numa::Result<()> {
config_path: resolved_config_path, config_path: resolved_config_path,
config_found, config_found,
config_dir: numa::config_dir(), config_dir: numa::config_dir(),
data_dir: numa::data_dir(), data_dir: resolved_data_dir,
tls_config: initial_tls, tls_config: initial_tls,
upstream_mode: resolved_mode, upstream_mode: resolved_mode,
root_hints, root_hints,
@@ -370,6 +383,9 @@ async fn main() -> numa::Result<()> {
); );
} }
} }
if config.dot.enabled {
row("DoT", g, &format!("tls://:{}", config.dot.port));
}
if config.lan.enabled { if config.lan.enabled {
row("LAN", g, "mDNS (_numa._tcp.local)"); row("LAN", g, "mDNS (_numa._tcp.local)");
} }
@@ -477,6 +493,15 @@ async fn main() -> numa::Result<()> {
}); });
} }
// Spawn DNS-over-TLS listener (RFC 7858)
if config.dot.enabled {
let dot_ctx = Arc::clone(&ctx);
let dot_config = config.dot.clone();
tokio::spawn(async move {
numa::dot::start_dot(dot_ctx, &dot_config).await;
});
}
// UDP DNS listener // UDP DNS listener
#[allow(clippy::infinite_loop)] #[allow(clippy::infinite_loop)]
loop { loop {

View File

@@ -870,14 +870,25 @@ mod tests {
}; };
let handler = handler.clone(); let handler = handler.clone();
tokio::spawn(async move { tokio::spawn(async move {
let timeout = std::time::Duration::from_secs(5);
// Read length-prefixed DNS query // Read length-prefixed DNS query
let mut len_buf = [0u8; 2]; let mut len_buf = [0u8; 2];
if stream.read_exact(&mut len_buf).await.is_err() { if tokio::time::timeout(timeout, stream.read_exact(&mut len_buf))
.await
.ok()
.and_then(|r| r.ok())
.is_none()
{
return; return;
} }
let len = u16::from_be_bytes(len_buf) as usize; let len = u16::from_be_bytes(len_buf) as usize;
let mut data = vec![0u8; len]; let mut data = vec![0u8; len];
if stream.read_exact(&mut data).await.is_err() { if tokio::time::timeout(timeout, stream.read_exact(&mut data))
.await
.ok()
.and_then(|r| r.ok())
.is_none()
{
return; return;
} }

View File

@@ -24,7 +24,7 @@ pub fn regenerate_tls(ctx: &ServerCtx) {
names.extend(ctx.lan_peers.lock().unwrap().names()); names.extend(ctx.lan_peers.lock().unwrap().names());
let names: Vec<String> = names.into_iter().collect(); let names: Vec<String> = names.into_iter().collect();
match build_tls_config(&ctx.proxy_tld, &names) { match build_tls_config(&ctx.proxy_tld, &names, Vec::new(), &ctx.data_dir) {
Ok(new_config) => { Ok(new_config) => {
tls.store(new_config); tls.store(new_config);
info!("TLS cert regenerated for {} services", names.len()); info!("TLS cert regenerated for {} services", names.len());
@@ -36,17 +36,26 @@ pub fn regenerate_tls(ctx: &ServerCtx) {
/// Build a TLS config with a cert covering all provided service names. /// Build a TLS config with a cert covering all provided service names.
/// Wildcards under single-label TLDs (*.numa) are rejected by browsers, /// Wildcards under single-label TLDs (*.numa) are rejected by browsers,
/// so we list each service explicitly as a SAN. /// so we list each service explicitly as a SAN.
pub fn build_tls_config(tld: &str, service_names: &[String]) -> crate::Result<Arc<ServerConfig>> { /// `alpn` is advertised in the TLS ServerHello — pass empty for the proxy
let dir = crate::data_dir(); /// (which accepts any ALPN), or `[b"dot"]` for DoT (RFC 7858 §3.2).
let (ca_cert, ca_key) = ensure_ca(&dir)?; /// `data_dir` is where the CA material is stored — taken from
/// `[server] data_dir` in numa.toml (defaults to `crate::data_dir()`).
pub fn build_tls_config(
tld: &str,
service_names: &[String],
alpn: Vec<Vec<u8>>,
data_dir: &Path,
) -> crate::Result<Arc<ServerConfig>> {
let (ca_cert, ca_key) = ensure_ca(data_dir)?;
let (cert_chain, key) = generate_service_cert(&ca_cert, &ca_key, tld, service_names)?; let (cert_chain, key) = generate_service_cert(&ca_cert, &ca_key, tld, service_names)?;
// Ensure a crypto provider is installed (rustls needs one) // Ensure a crypto provider is installed (rustls needs one)
let _ = rustls::crypto::ring::default_provider().install_default(); let _ = rustls::crypto::ring::default_provider().install_default();
let config = ServerConfig::builder() let mut config = ServerConfig::builder()
.with_no_client_auth() .with_no_client_auth()
.with_single_cert(cert_chain, key)?; .with_single_cert(cert_chain, key)?;
config.alpn_protocols = alpn;
info!( info!(
"TLS configured for {} .{} domains", "TLS configured for {} .{} domains",

View File

@@ -404,6 +404,241 @@ check "Cache flushed" \
kill "$NUMA_PID" 2>/dev/null || true kill "$NUMA_PID" 2>/dev/null || true
wait "$NUMA_PID" 2>/dev/null || true wait "$NUMA_PID" 2>/dev/null || true
sleep 1
# ---- Suite 5: DNS-over-TLS (RFC 7858) ----
echo ""
echo "╔══════════════════════════════════════════╗"
echo "║ Suite 5: DNS-over-TLS (RFC 7858) ║"
echo "╚══════════════════════════════════════════╝"
if ! command -v kdig >/dev/null 2>&1; then
printf " ${DIM}skipped — install 'knot' for kdig${RESET}\n"
elif ! command -v openssl >/dev/null 2>&1; then
printf " ${DIM}skipped — openssl not found${RESET}\n"
else
DOT_PORT=8853
DOT_CERT=/tmp/numa-integration-dot.crt
DOT_KEY=/tmp/numa-integration-dot.key
# Generate a test cert mirroring production self_signed_tls SAN shape
# (*.numa wildcard + explicit numa.numa apex).
openssl req -x509 -newkey rsa:2048 -nodes -days 1 \
-keyout "$DOT_KEY" -out "$DOT_CERT" \
-subj "/CN=Numa .numa services" \
-addext "subjectAltName=DNS:*.numa,DNS:numa.numa" \
>/dev/null 2>&1
# Suite 5 uses a local zone so it's upstream-independent — the point is
# to exercise the DoT transport layer (handshake, ALPN, framing,
# persistent connections), not re-test recursive resolution.
cat > "$CONFIG" << CONF
[server]
bind_addr = "127.0.0.1:$PORT"
api_port = $API_PORT
[upstream]
mode = "forward"
address = "127.0.0.1"
port = 65535
[cache]
max_entries = 10000
[blocking]
enabled = false
[proxy]
enabled = false
[dot]
enabled = true
port = $DOT_PORT
bind_addr = "127.0.0.1"
cert_path = "$DOT_CERT"
key_path = "$DOT_KEY"
[[zones]]
domain = "dot-test.example"
record_type = "A"
value = "10.0.0.1"
ttl = 60
CONF
RUST_LOG=info "$BINARY" "$CONFIG" > "$LOG" 2>&1 &
NUMA_PID=$!
sleep 4
if ! kill -0 "$NUMA_PID" 2>/dev/null; then
FAILED=$((FAILED + 1))
printf " ${RED}${RESET} DoT startup\n"
printf " ${DIM}%s${RESET}\n" "$(tail -5 "$LOG")"
else
echo ""
echo "=== Listener ==="
check "DoT bound on 127.0.0.1:$DOT_PORT" \
"DoT listening on 127.0.0.1:$DOT_PORT" \
"$(grep 'DoT listening' "$LOG")"
KDIG="kdig @127.0.0.1 -p $DOT_PORT +tls +tls-ca=$DOT_CERT +tls-hostname=numa.numa +time=5 +retry=0"
echo ""
echo "=== Queries over DoT ==="
check "DoT local zone A record" \
"10.0.0.1" \
"$($KDIG +short dot-test.example A 2>/dev/null)"
# +keepopen reuses one TLS connection for multiple queries — tests
# persistent connection handling. kdig applies options left-to-right,
# so +short and +keepopen must come before the query specs.
check "DoT persistent connection (3 queries, 1 handshake)" \
"10.0.0.1" \
"$($KDIG +keepopen +short dot-test.example A dot-test.example A dot-test.example A 2>/dev/null | head -1)"
echo ""
echo "=== ALPN ==="
# Positive case: client offers "dot", server picks it.
ALPN_OK=$(echo "" | openssl s_client -connect "127.0.0.1:$DOT_PORT" \
-servername numa.numa -alpn dot -CAfile "$DOT_CERT" 2>&1 </dev/null || true)
check "DoT negotiates ALPN \"dot\"" \
"ALPN protocol: dot" \
"$ALPN_OK"
# Negative case: client offers only "h2", server must reject the
# handshake with no_application_protocol alert (cross-protocol
# confusion defense, RFC 7858bis §3.2).
if echo "" | openssl s_client -connect "127.0.0.1:$DOT_PORT" \
-servername numa.numa -alpn h2 -CAfile "$DOT_CERT" \
</dev/null >/dev/null 2>&1; then
ALPN_MISMATCH="handshake unexpectedly succeeded"
else
ALPN_MISMATCH="rejected"
fi
check "DoT rejects non-dot ALPN" \
"rejected" \
"$ALPN_MISMATCH"
fi
kill "$NUMA_PID" 2>/dev/null || true
wait "$NUMA_PID" 2>/dev/null || true
rm -f "$DOT_CERT" "$DOT_KEY"
fi
sleep 1
# ---- Suite 6: Proxy + DoT coexistence ----
echo ""
echo "╔══════════════════════════════════════════╗"
echo "║ Suite 6: Proxy + DoT Coexistence ║"
echo "╚══════════════════════════════════════════╝"
if ! command -v kdig >/dev/null 2>&1 || ! command -v openssl >/dev/null 2>&1; then
printf " ${DIM}skipped — needs kdig + openssl${RESET}\n"
else
DOT_PORT=8853
PROXY_HTTP_PORT=8080
PROXY_HTTPS_PORT=8443
NUMA_DATA=/tmp/numa-integration-data
# Fresh data dir so we generate a fresh CA for this suite. Path is set
# via [server] data_dir in the TOML below, not an env var — numa treats
# its config file as the single source of truth for all knobs.
rm -rf "$NUMA_DATA"
mkdir -p "$NUMA_DATA"
cat > "$CONFIG" << CONF
[server]
bind_addr = "127.0.0.1:$PORT"
api_port = $API_PORT
data_dir = "$NUMA_DATA"
[upstream]
mode = "forward"
address = "127.0.0.1"
port = 65535
[cache]
max_entries = 10000
[blocking]
enabled = false
[proxy]
enabled = true
port = $PROXY_HTTP_PORT
tls_port = $PROXY_HTTPS_PORT
tld = "numa"
bind_addr = "127.0.0.1"
[dot]
enabled = true
port = $DOT_PORT
bind_addr = "127.0.0.1"
[[zones]]
domain = "dot-test.example"
record_type = "A"
value = "10.0.0.1"
ttl = 60
CONF
RUST_LOG=info "$BINARY" "$CONFIG" > "$LOG" 2>&1 &
NUMA_PID=$!
sleep 4
if ! kill -0 "$NUMA_PID" 2>/dev/null; then
FAILED=$((FAILED + 1))
printf " ${RED}${RESET} Startup with proxy + DoT\n"
printf " ${DIM}%s${RESET}\n" "$(tail -5 "$LOG")"
else
echo ""
echo "=== Both listeners ==="
check "DoT listener bound" \
"DoT listening on 127.0.0.1:$DOT_PORT" \
"$(grep 'DoT listening' "$LOG")"
check "HTTPS proxy listener bound" \
"HTTPS proxy listening on 127.0.0.1:$PROXY_HTTPS_PORT" \
"$(grep 'HTTPS proxy listening' "$LOG")"
PANIC_COUNT=$(grep -c 'panicked' "$LOG" 2>/dev/null || echo 0)
check "No startup panics in log" \
"^0$" \
"$PANIC_COUNT"
echo ""
echo "=== DoT works with proxy enabled ==="
# Proxy's build_tls_config runs first and creates the CA in
# $NUMA_DATA_DIR. DoT self_signed_tls then loads the same CA and
# issues its own leaf cert. One CA trusts both listeners.
CA="$NUMA_DATA/ca.pem"
KDIG="kdig @127.0.0.1 -p $DOT_PORT +tls +tls-ca=$CA +tls-hostname=numa.numa +time=5 +retry=0"
check "DoT local zone A (with proxy on)" \
"10.0.0.1" \
"$($KDIG +short dot-test.example A 2>/dev/null)"
echo ""
echo "=== Proxy TLS works with DoT enabled ==="
# Proxy cert has SAN numa.numa (auto-added "numa" service). A
# successful handshake validates that the proxy's separate
# ServerConfig wasn't disturbed by DoT's own cert generation.
PROXY_TLS=$(echo "" | openssl s_client -connect "127.0.0.1:$PROXY_HTTPS_PORT" \
-servername numa.numa -CAfile "$CA" 2>&1 </dev/null || true)
check "Proxy HTTPS TLS handshake succeeds" \
"Verify return code: 0 (ok)" \
"$PROXY_TLS"
fi
kill "$NUMA_PID" 2>/dev/null || true
wait "$NUMA_PID" 2>/dev/null || true
rm -rf "$NUMA_DATA"
fi
# Summary # Summary
echo "" echo ""