working ig

2026-07-16 00:54:33 -07:00
1 parent b834cd06f3
commit c14be296a8
@@ -32,6 +32,10 @@ RADIUS_SECRET=
 # Applied on FIRST start only; change later via the console itself.
 TECHNITIUM_ADMIN_PASSWORD=

+# ── DNS zone for VPN client registration (A + PTR via learn-address) ──
+# Leave empty to disable client DNS registration entirely.
+VPN_DNS_ZONE=roam.atlashorizon.net
+
 # ── Optional (defaults shown) ─────────────────────────────────────────
 # Duo Auth Proxy version; pin (e.g. 6.5.1) for deterministic updates
 #DUO_AUTHPROXY_VERSION=latest
@@ -34,6 +34,11 @@ radius_secret: ""
 # console (this var must still match for the init job's API login).
 technitium_admin_password: ""

+# DNS zone where VPN clients get registered as <username>.<zone> with a
+# matching PTR (Technitium primary zone, created by the init job). Set to
+# "" to disable registration.
+vpn_dns_zone: roam.atlashorizon.net
+
 # ── Optional ──
 # Pin for deterministic Duo updates (e.g. "6.5.1"); bumping it forces a
 # refetch of the tarball on the next rebuild.
@@ -23,6 +23,7 @@
    vpnhub_base_dir: /opt
    vpnhub_stack_dir: "{{ vpnhub_base_dir }}/vpnhub"
    vpnhub_force_rebuild: false
+    vpnhub_host_dns: [1.1.1.1, 9.9.9.9]
    public_iface: "{{ ansible_default_ipv4.interface }}"
    wg_peers_configured: >-
      {{ (troch_wg_public_key | default('')) != ''
@@ -30,9 +31,9 @@

  handlers:
    - name: reload firewall
+      # the reload flushes the Docker-managed chains, so restart docker after
      ansible.builtin.shell: |
        netfilter-persistent reload
-        # the reload flushes Docker's chains — put them back
        if systemctl is-active --quiet docker; then systemctl restart docker; fi

    - name: restart wireguard
@@ -50,17 +51,18 @@
        chdir: "{{ vpnhub_stack_dir }}"

    - name: recreate duo-authproxy
-      ansible.builtin.command: docker compose up -d --force-recreate duo-authproxy
+      # --build: entrypoint/cfg-template changes are baked into the image
+      ansible.builtin.command: docker compose up -d --build --force-recreate duo-authproxy
      args:
        chdir: "{{ vpnhub_stack_dir }}"

    - name: recreate openvpn
-      ansible.builtin.command: docker compose up -d --force-recreate openvpn
+      ansible.builtin.command: docker compose up -d --build --force-recreate openvpn
      args:
        chdir: "{{ vpnhub_stack_dir }}"

    - name: recreate stack
-      ansible.builtin.command: docker compose up -d --force-recreate
+      ansible.builtin.command: docker compose up -d --build --force-recreate
      args:
        chdir: "{{ vpnhub_stack_dir }}"

@@ -76,9 +78,12 @@
            update_cache: true
            cache_valid_time: 3600

+        # Distro-agnostic: ansible_distribution is Ubuntu or Debian, and
+        # Docker publishes separate repos per distro (a Debian codename
+        # like trixie does not exist under linux/ubuntu and vice versa).
        - name: Docker apt signing key
          ansible.builtin.get_url:
-            url: https://download.docker.com/linux/ubuntu/gpg
+            url: "https://download.docker.com/linux/{{ ansible_distribution | lower }}/gpg"
            dest: /etc/apt/keyrings/docker.asc
            mode: "0644"

@@ -87,7 +92,7 @@
            repo: >-
              deb [arch={{ 'arm64' if ansible_architecture == 'aarch64' else 'amd64' }}
              signed-by=/etc/apt/keyrings/docker.asc]
-              https://download.docker.com/linux/ubuntu
+              https://download.docker.com/linux/{{ ansible_distribution | lower }}
              {{ ansible_distribution_release }} stable
            filename: docker

@@ -124,10 +129,19 @@
        sysctl_file: /etc/sysctl.d/99-vpnhub.conf
        reload: true

-    # Technitium binds 0.0.0.0:53 — systemd-resolved's stub listener on
-    # 127.0.0.53:53 would collide with it.
+    # Technitium binds 0.0.0.0:53 — free the port while keeping host DNS
+    # working. Ubuntu runs systemd-resolved (stub on 127.0.0.53:53 must be
+    # disabled); plain Debian has no resolved and just needs a sane static
+    # /etc/resolv.conf.
+    - name: Check whether systemd-resolved is in use
+      tags: host
+      ansible.builtin.stat:
+        path: /run/systemd/resolve/resolv.conf
+      register: resolved_upstream
+
    - name: resolved drop-in directory
      tags: host
+      when: resolved_upstream.stat.exists
      ansible.builtin.file:
        path: /etc/systemd/resolved.conf.d
        state: directory
@@ -135,6 +149,7 @@

    - name: Disable systemd-resolved stub listener (frees :53 for Technitium)
      tags: host
+      when: resolved_upstream.stat.exists
      ansible.builtin.copy:
        dest: /etc/systemd/resolved.conf.d/99-vpnhub.conf
        mode: "0644"
@@ -143,14 +158,29 @@
          DNSStubListener=no
      notify: restart systemd-resolved

-    - name: Point /etc/resolv.conf at resolved's upstream file (not the dead stub)
+    - name: Point /etc/resolv.conf at resolved upstream file (not the dead stub)
      tags: host
+      when: resolved_upstream.stat.exists
      ansible.builtin.file:
        src: /run/systemd/resolve/resolv.conf
        dest: /etc/resolv.conf
        state: link
        force: true

+    - name: Static resolv.conf (no systemd-resolved on this host)
+      tags: host
+      when: not resolved_upstream.stat.exists
+      ansible.builtin.copy:
+        dest: /etc/resolv.conf
+        follow: false     # replace a dangling symlink rather than write through it
+        mode: "0644"
+        content: |
+          # Managed by Ansible (vpnhub). Public resolvers on purpose: the
+          # host must resolve registries/repos even when the stack is down.
+          {% for ns in vpnhub_host_dns %}
+          nameserver {{ ns }}
+          {% endfor %}
+
    - name: Firewall rules (rendered from host/rules.v4/v6.template)
      tags: [host, firewall]
      ansible.builtin.copy:
@@ -18,6 +18,8 @@ RADIUS_SECRET={{ radius_secret }}

 TECHNITIUM_ADMIN_PASSWORD={{ technitium_admin_password }}

+VPN_DNS_ZONE={{ vpn_dns_zone | default('roam.atlashorizon.net') }}
+
 DUO_AUTHPROXY_VERSION={{ duo_authproxy_version | default('latest') }}

 # /opt layout — consumed by compose ${*_DIR} interpolation
@@ -0,0 +1,62 @@
+client
+dev tun
+proto udp
+remote connect.atlashorizon.net 1194
+resolv-retry infinite
+nobind
+persist-key
+persist-tun
+dhcp-option DNS 172.31.8.1
+
+remote-cert-tls server
+tls-version-min 1.2
+data-ciphers AES-256-GCM:CHACHA20-POLY1305
+auth SHA256
+
+auth-user-pass
+auth-retry interact
+hand-window 120
+verb 3
+
+<ca>
+-----BEGIN CERTIFICATE-----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+-----END CERTIFICATE-----
+</ca>
+<tls-crypt>
+-----BEGIN OpenVPN Static key V1-----
+5fedf4b825a32eb88722a14f801ef416
+2d015c4b5c41654c556108fc2fa5006e
+f83c97ea210113935dbaf47469ba9cd4
+752f928caf48926b47b6da2b4573f6fa
+ed909502d027ba02e509dc48c6542d74
+53c3013ce00e443b43faac4f8f87717f
+1fe942bd1f1412fb38566a88ba7e0d5a
+b722ea79cb4a1b5e172185d2644db81b
+33fdc631815e47f6ae5ce099beee5c37
+b0880c7ff8f9c4ff7a8302e7eb692ce9
+eab50f743de8aa91bff266b2dc4f504c
+775137ae64d71c79e35a40e8bedb8f67
+65bfa02d05bfc2a5e8824eff9be34a26
+a8c8b151a215ae26c67c006e431c8a57
+8985d2feec6787b629b2e30dfdc5ffea
+a0d26b0b0f1aca83ccbfc77b660cceed
+-----END OpenVPN Static key V1-----
+</tls-crypt>
@@ -69,6 +69,7 @@ services:
    env_file: .env
    volumes:
      - ${OPENVPN_DIR:-./openvpn}/server.conf:/etc/openvpn/server.conf:ro
+      - ${OPENVPN_DIR:-./openvpn}/learn-address.sh:/etc/openvpn/learn-address.sh:ro
      - ${OPENVPN_DIR:-./openvpn}/pki:/etc/openvpn/pki:ro
      - openvpn-state:/var/lib/openvpn
    depends_on:
@@ -1,5 +1,11 @@
 # Duo publishes no official image; build from their source tarball.
 # Install flags match Duo's documented non-interactive install.
+#
+# Slimming option: switch BOTH stages to debian:bookworm-slim (~80MB less,
+# and the base layer is shared with the openvpn image). Both must move
+# together — the proxy links against the builder's glibc. Alpine/musl
+# will not work.
+#FROM debian:bookworm-slim AS build

 FROM ubuntu:24.04 AS build

@@ -20,6 +26,8 @@ RUN groupadd -r -g 35505 duo \
 && ./install --install-dir /opt/duoauthproxy --service-user duo \
      --log-group duo --create-init-script no --enable-selinux no

+# Slim variant (must match the build stage base — see note above):
+#FROM debian:bookworm-slim
 FROM ubuntu:24.04

 # gettext-base = envsubst for rendering authproxy.cfg from env vars
@@ -8,7 +8,7 @@ debug=false

 [ad_client]
 host=${TROCH_DC1_IP}
-host_2=${TROCH_DC2_IP}
+;host_2=${TROCH_DC2_IP}
 service_account_username=${DUO_SVC_USER}
 service_account_password=${DUO_SVC_PASSWORD}
 search_dn=${AD_SEARCH_DN}
@@ -22,7 +22,7 @@ ssl_ca_certs_file=/certs/ad-ldaps-ca.pem
 ssl_verify_hostname=false
 ; Restrict VPN access to one AD group (fill DUO_SECURITY_GROUP_DN in .env
 ; and uncomment):
-;security_group_dn=${DUO_SECURITY_GROUP_DN}
+security_group_dn=${DUO_SECURITY_GROUP_DN}

 [radius_server_auto]
 ikey=${DUO_IKEY}
@@ -13,7 +13,7 @@ if [ ! -f /certs/ad-ldaps-ca.pem ]; then
  exit 1
 fi

-envsubst '${TROCH_DC1_IP} ${TROCH_DC2_IP} ${DUO_IKEY} ${DUO_SKEY} ${DUO_API_HOST} ${DUO_SVC_USER} ${DUO_SVC_PASSWORD} ${AD_SEARCH_DN} ${RADIUS_SECRET}' \
+envsubst '${TROCH_DC1_IP} ${TROCH_DC2_IP} ${DUO_IKEY} ${DUO_SKEY} ${DUO_API_HOST} ${DUO_SVC_USER} ${DUO_SVC_PASSWORD} ${AD_SEARCH_DN} ${RADIUS_SECRET} ${DUO_SECURITY_GROUP_DN}' \
  < /opt/duoauthproxy/conf/authproxy.cfg.template \
  > /opt/duoauthproxy/conf/authproxy.cfg
 chmod 600 /opt/duoauthproxy/conf/authproxy.cfg
@@ -3,7 +3,7 @@ FROM debian:bookworm-slim
 # libpam-radius-auth: PAM -> RADIUS bridge to the Duo Auth Proxy on 127.0.0.1
 RUN apt-get update \
 && apt-get install -y --no-install-recommends \
-      openvpn libpam-radius-auth iproute2 procps ca-certificates \
+      openvpn libpam-radius-auth iproute2 procps ca-certificates socat curl jq \
 && rm -rf /var/lib/apt/lists/*

 COPY pam-openvpn /etc/pam.d/openvpn
@@ -0,0 +1,67 @@
+#!/bin/sh
+# OpenVPN learn-address hook: registers VPN clients in Technitium DNS.
+#   add/update <ip> <cn>  -> A <cn>.$VPN_DNS_ZONE + PTR, TTL 60
+#   delete <ip>           -> removes both records
+#
+# CRITICAL: this script must ALWAYS exit 0 — a non-zero exit makes
+# OpenVPN reject the client's address. DNS registration is best-effort;
+# it must never take the VPN down with it.
+
+OP="$1"; ADDR="$2"; CN="${3:-}"
+
+API="http://127.0.0.1:5380/api"
+[ -n "${TECHNITIUM_ADMIN_PASSWORD:-}" ] || exit 0
+[ -n "${VPN_DNS_ZONE:-}" ] || exit 0
+
+log() { echo "learn-address: $*"; }
+
+TOKEN=$(curl -sf -m 5 -G "$API/user/login" \
+  --data-urlencode "user=admin" \
+  --data-urlencode "pass=$TECHNITIUM_ADMIN_PASSWORD" 2>/dev/null \
+  | jq -r '.token // empty')
+[ -n "$TOKEN" ] || { log "Technitium API login failed; skipping $OP $ADDR"; exit 0; }
+AUTH="Authorization: Bearer $TOKEN"
+
+# 172.31.8.5 -> 5.8.31.172.in-addr.arpa
+PTR_HOST=$(printf '%s' "$ADDR" | awk -F. '{print $4"."$3"."$2"."$1".in-addr.arpa"}')
+
+deregister() {
+  # PTR tells us which A record this IP belonged to
+  TARGET=$(curl -sf -m 5 -H "$AUTH" -G "$API/zones/records/get" \
+    --data-urlencode "domain=$PTR_HOST" 2>/dev/null \
+    | jq -r '.response.records[]? | select(.type=="PTR") | .rData.ptrName' | head -n1)
+  [ -n "$TARGET" ] || return 0
+  curl -sf -m 5 -H "$AUTH" -G "$API/zones/records/delete" \
+    --data-urlencode "domain=$PTR_HOST" \
+    --data-urlencode "type=PTR" \
+    --data-urlencode "ptrName=$TARGET" >/dev/null 2>&1 || log "PTR delete failed for $ADDR"
+  curl -sf -m 5 -H "$AUTH" -G "$API/zones/records/delete" \
+    --data-urlencode "domain=$TARGET" \
+    --data-urlencode "zone=$VPN_DNS_ZONE" \
+    --data-urlencode "type=A" \
+    --data-urlencode "ipAddress=$ADDR" >/dev/null 2>&1 || log "A delete failed for $TARGET/$ADDR"
+}
+
+register() {
+  # sanitize CN (= AD username) into a DNS label
+  NAME=$(printf '%s' "$CN" | tr '[:upper:]' '[:lower:]' \
+    | tr -c 'a-z0-9-' '-' | sed 's/^-*//;s/-*$//' | cut -c1-63)
+  [ -n "$NAME" ] || return 0
+  curl -sf -m 5 -H "$AUTH" -G "$API/zones/records/add" \
+    --data-urlencode "domain=$NAME.$VPN_DNS_ZONE" \
+    --data-urlencode "zone=$VPN_DNS_ZONE" \
+    --data-urlencode "type=A" \
+    --data-urlencode "ttl=60" \
+    --data-urlencode "ipAddress=$ADDR" \
+    --data-urlencode "ptr=true" \
+    --data-urlencode "createPtrZone=true" >/dev/null 2>&1 \
+    || log "A/PTR add failed for $NAME/$ADDR"
+}
+
+case "$OP" in
+  add)    register ;;
+  update) deregister; register ;;   # IP reassigned to a different client
+  delete) deregister ;;
+esac
+
+exit 0
@@ -10,21 +10,28 @@ dev tun0
 topology subnet

 server 172.31.8.0 255.255.255.0
-ifconfig-pool-persist /var/lib/openvpn/ipp.txt
+
+# Identity = AD username (username-as-common-name), and one user WILL
+# connect from several devices (laptop + phone). duplicate-cn allows the
+# concurrent sessions and gives each its own pool IP. Sticky per-user IPs
+# (ifconfig-pool-persist) are incompatible with that — do not re-enable.
+duplicate-cn

 # Split tunnel: only the AD sites + transfer net ride the VPN.
 push "route 192.168.20.0 255.255.255.0"
 push "route 192.168.7.0 255.255.255.0"
 push "route 172.31.255.0 255.255.255.0"

-# All client DNS goes to Technitium on this box (block-outside-dns stops
-# Windows from leaking AD lookups to the local LAN resolver).
+# ALL client DNS goes to Technitium. Deliberately NO "dhcp-option DOMAIN"
+# or DOMAIN-SEARCH here: in OpenVPN3 clients (Connect on macOS/iOS/
+# Android/Windows) those options MEAN split-DNS — pushing them scopes the
+# VPN DNS to just those domains. With only DNS pushed, 172.31.8.1 becomes
+# the default resolver for every query. ADAPTER_DOMAIN_SUFFIX sets the
+# Windows connection-specific suffix (short names like "ahdc01" still
+# resolve) without triggering split-DNS. block-outside-dns additionally
+# stops Windows from leaking queries to the local LAN resolver.
 push "dhcp-option DNS 172.31.8.1"
-push "dhcp-option DOMAIN ad.atlashorizon.net"
-push "dhcp-option DOMAIN-SEARCH ad.atlashorizon.net"
-push "dhcp-option DOMAIN-SEARCH ad.calliecorner.net"
-push "dhcp-option DOMAIN-SEARCH ad.oswaldtheotter.net"
-push "dhcp-option DOMAIN-SEARCH ad.trochalakis.net"
+push "dhcp-option ADAPTER_DOMAIN_SUFFIX ad.atlashorizon.net"
 push "block-outside-dns"

 # PKI — server certificate only (offline CA, see pki/README.md);
@@ -44,8 +51,12 @@ plugin /usr/lib/openvpn/openvpn-plugin-auth-pam.so openvpn
 verify-client-cert none
 username-as-common-name
 hand-window 120        # allow time for the Duo push round-trip
-reneg-sec 0            # no mid-session TLS renegotiation (would re-prompt Duo)
-auth-gen-token 43200   # reconnects within 12h reuse a token instead of a new push
+# Hourly rekey + 12h auth token: renegotiations are satisfied silently by
+# the pushed token (no mid-session Duo prompts); a fresh password + push
+# is only required once the token expires. (auth-gen-token requires a
+# non-zero reneg-sec — do not set reneg-sec 0 here.)
+reneg-sec 3600
+auth-gen-token 43200

 keepalive 10 60
 explicit-exit-notify 1
@@ -60,6 +71,15 @@ group nogroup
 persist-key
 persist-tun

+# Register each client as <username>.$VPN_DNS_ZONE (A + PTR) in Technitium
+# on connect, clean up on disconnect. Invoked via sh so the mounted script
+# needs no exec bit; env (TECHNITIUM_ADMIN_PASSWORD, VPN_DNS_ZONE) comes
+# from the container environment.
+script-security 2
+learn-address "/bin/sh /etc/openvpn/learn-address.sh"
+
 status /var/lib/openvpn/status.log 30
-management 127.0.0.1 7505
+# Unix socket (root-only) instead of a passwordless TCP port:
+#   docker compose exec openvpn sh -c "socat - UNIX:/var/lib/openvpn/mgmt.sock"
+management /var/lib/openvpn/mgmt.sock unix
 verb 3
@@ -77,4 +77,23 @@ ensure_fwd_zone ad.oswaldtheotter.net "$OSWALD_DC_IP"
 ensure_fwd_zone 20.168.192.in-addr.arpa "$TROCH_DC1_IP" "$TROCH_DC2_IP"
 ensure_fwd_zone 7.168.192.in-addr.arpa  "$OSWALD_DC_IP"

+# Primary zones for VPN client registration (openvpn learn-address hook
+# writes <username>.$VPN_DNS_ZONE A records + PTRs for 172.31.8.0/24)
+ensure_primary_zone() {
+  zone=$1
+  if zone_exists "$zone"; then
+    echo "zone $zone already exists — leaving as-is"
+    return 0
+  fi
+  echo "creating primary zone $zone"
+  curl -sf -H "$AUTH" -G "$API/zones/create" \
+    --data-urlencode "zone=$zone" \
+    --data-urlencode "type=Primary" | api_ok
+}
+
+if [ -n "${VPN_DNS_ZONE:-}" ]; then
+  ensure_primary_zone "$VPN_DNS_ZONE"
+  ensure_primary_zone "8.31.172.in-addr.arpa"
+fi
+
 echo "Technitium AD forwarder zones OK."