contrib/sni-router: use host networking for HAProxy to preserve client IPs

Move HAProxy into the host network namespace so it sees the real
client source IP on inbound connections.  With bridge networking +
published ports the source IP is rewritten to the bridge gateway by
the runtime (Docker's userland-proxy, rootless Podman's slirp4netns
or pasta), and the PROXY v2 header HAProxy then sends to mtg and
Caddy carries that useless address.

mtg and Caddy stay on the compose bridge and publish their ports on
host loopback; the host-mode HAProxy dials them at 127.0.0.1.  Caddy's
proxy_protocol allow list is tightened to loopback only.

The 'sysctls: net.ipv4.ip_unprivileged_port_start=80' line is removed
because Docker refuses to apply namespaced sysctls when the netns is
shared with the host.  Rootless Podman users binding the privileged
ports need the equivalent host-side sysctl once; this is documented
in README.md.

Fixes #498.

contrib/sni-router/Caddyfile

 	# to Caddy's access log.  The `tls` wrapper must follow so that TLS
 	# is terminated on the unwrapped connection.
 	#
-	# `allow` lists the networks permitted to send PROXY headers.  These
-	# ranges cover docker compose's default bridge networks; tighten
-	# them if you pin a specific subnet in docker-compose.yml.
+	# `allow` lists the networks permitted to send PROXY headers.
+	# HAProxy runs in the host netns and reaches Caddy via host loopback
+	# (see docker-compose.yml), so the only legitimate peer is loopback.
 	servers :8443 {
 		listener_wrappers {
 			proxy_protocol {
 				timeout 5s
-				allow 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
+				allow 127.0.0.0/8 ::1/128
 			}
 			tls
 		}

contrib/sni-router/README.md

 If you disable one, disable all four, otherwise the backend will fail
 to parse the connection.
 
+## Why host networking for HAProxy
+
+HAProxy runs in the host network namespace (`network_mode: host` in
+`docker-compose.yml`) so it sees the real client source IP on every
+inbound connection.  With the default bridge networking + published
+ports the source IP is rewritten to the bridge gateway — by Docker's
+userland proxy (`docker-proxy`), by rootless Podman's `slirp4netns`
+or `pasta`, or by NAT on the Docker host — and the PROXY v2 header
+HAProxy then sends to mtg and Caddy carries that useless address.
+Host networking lifts HAProxy out of the rewrite path; mtg and Caddy
+stay on the compose bridge and HAProxy dials them via host loopback
+(`127.0.0.1`).
+
+Trade-off: HAProxy occupies the host's `:443` and `:80` directly, so
+nothing else on the host may listen on those ports.  For a dedicated
+mtg/SNI-router host that is the intended layout.
+
+Rootless Podman users binding the privileged ports `:80`/`:443` need
+the host-side sysctl once (rootful Docker handles this implicitly):
+
+```bash
+sudo sysctl -w net.ipv4.ip_unprivileged_port_start=80
+```
+
 ## Fronting loop (why `[domain-fronting]` is set explicitly)
 
 When mtg sees TLS that isn't valid Telegram (a probe or a browser

contrib/sni-router/docker-compose.yml

 services:
   haproxy:
     image: haproxy:lts-alpine
-    ports:
-      - "443:443"
-      - "80:80"
+    # Host networking so HAProxy sees the real client source IP on
+    # inbound (bridge networking with published ports rewrites it to
+    # the bridge gateway under both Docker's userland-proxy and
+    # rootless Podman's slirp4netns/pasta).  See "Why host networking"
+    # in README.md.
+    network_mode: host
     volumes:
       - ./haproxy.cfg:/usr/local/etc/haproxy/haproxy.cfg:ro,Z
     environment:
       - mtg
       - web
     restart: unless-stopped
-    sysctls:
-      - net.ipv4.ip_unprivileged_port_start=80
 
   mtg:
     image: nineseconds/mtg:2
     volumes:
       - ./mtg-config.toml:/config/config.toml:ro,Z
-    expose:
-      - "3128"
+    # Publish on host loopback so the host-mode HAProxy can reach it.
+    ports:
+      - "127.0.0.1:3128:3128"
     restart: unless-stopped
     extra_hosts:
       - "host.containers.internal:host-gateway"
       - ./Caddyfile:/etc/caddy/Caddyfile:ro,Z
       - caddy_data:/data
       - ./www:/srv:ro,Z
-    expose:
-      - "80"
-      - "8443"
+    # Publish on host loopback so the host-mode HAProxy can reach it.
+    # Caddy's HTTP listener is mapped off :80 (occupied by HAProxy)
+    # to :8080; haproxy.cfg dials it on 127.0.0.1:8080 for ACME.
+    ports:
+      - "127.0.0.1:8080:80"
+      - "127.0.0.1:8443:8443"
     environment:
       <<: *domain-env
     restart: unless-stopped

contrib/sni-router/haproxy.cfg

     # send-proxy-v2 prepends a PROXY protocol v2 header so mtg sees the
     # real client IP instead of HAProxy's.  mtg must have
     # `proxy-protocol-listener = true` in its config.
-    server mtg mtg:3128 send-proxy-v2
+    #
+    # HAProxy runs in the host network namespace (see docker-compose.yml)
+    # and mtg publishes :3128 on host loopback, so we dial 127.0.0.1.
+    server mtg 127.0.0.1:3128 send-proxy-v2
 
 backend web
     # send-proxy-v2 prepends a PROXY protocol v2 header so Caddy logs the
     # real client IP instead of HAProxy's.  Caddy must enable the
     # proxy_protocol listener wrapper on :8443 (see Caddyfile).
-    server web web:8443 send-proxy-v2
+    server web 127.0.0.1:8443 send-proxy-v2
 
 backend web_acme
     mode http
-    server web web:80
+    # Caddy's :80 is published on host 127.0.0.1:8080 (HAProxy occupies
+    # the host's :80 itself).
+    server web 127.0.0.1:8080