#!/bin/sh # Runs ON THE ROUTER (OpenWrt has no bash, this must stay POSIX sh). # Invoked as: TS= sh skyg-apply.sh # Expects the new files staged at /tmp/skyg_firewall.batch, /tmp/dnsmasq.new.conf # and /tmp/ethers.new. Backs up the configs it overrides, applies the new # firewall/dnsmasq/ethers, ensures bridge-netfilter sysctls are enabled, # validates, and auto-restores on failure. set -euo pipefail BK=/etc/skyg-backups mkdir -p "$BK" cp /etc/config/firewall "$BK/firewall.$TS" cp /etc/dnsmasq.conf "$BK/dnsmasq.conf.$TS" cp /etc/ethers "$BK/ethers.$TS" 2>/dev/null || true cp /etc/sysctl.conf "$BK/sysctl.conf.$TS" 2>/dev/null || true # --- sysctl: enable bridge netfilter --- # All of this router's "networks" (apl, cam, lab, ...) are just address # ranges on one flat L2 bridge (br-lan, single /16 across the whole LAN) -- # there's no real VLAN separation. Without bridge-nf-call-iptables, unicast # traffic between two hosts on the same bridge is switched at L2 and never # reaches the FORWARD chain, so the skyg_* DROP rules below silently never # fire. /etc/sysctl.d/*.conf ships bridge-nf-call-* disabled and is reset on # firmware upgrades, but /etc/sysctl.conf is applied last and persists. if ! grep -q '^# BEGIN skyg-managed$' /etc/sysctl.conf 2>/dev/null; then { echo '' echo '# BEGIN skyg-managed' echo 'net.bridge.bridge-nf-call-iptables=1' echo 'net.bridge.bridge-nf-call-ip6tables=1' echo '# END skyg-managed' } >> /etc/sysctl.conf fi sysctl -e -p /etc/sysctl.conf >/dev/null # --- Firewall: replace the skyg-managed rules --- for ref in $(uci show firewall | grep -E '^firewall\.skyg_[^.=]+=' | cut -d= -f1); do uci -q del "$ref" || true done if ! uci -m import firewall < /tmp/skyg_firewall.batch; then echo 'FIREWALL: uci import failed, restoring' cp "$BK/firewall.$TS" /etc/config/firewall exit 1 fi if ! uci commit firewall; then echo 'FIREWALL: uci commit failed, restoring' cp "$BK/firewall.$TS" /etc/config/firewall exit 1 fi /etc/init.d/firewall reload # --- dnsmasq --- cp /tmp/dnsmasq.new.conf /etc/dnsmasq.conf if dnsmasq --test; then /etc/init.d/dnsmasq restart else echo 'dnsmasq config invalid, restoring' cp "$BK/dnsmasq.conf.$TS" /etc/dnsmasq.conf exit 1 fi # --- ethers --- cp /tmp/ethers.new /etc/ethers echo 'Done'