DHCP client with static fallback IP weird connection issues

Installing and Using OpenWrt Network and Wireless Configuration
1

Hi y'all,

I am in the process of creating a preconfigured OpenWrt image to supply to a couple of sites.

To make life easy, I'd like to set them up as a DHCP-Client device to hook them into existing networks. But if the device is not hooked into an existing network, the device should automatically assume a static IP and start distributing DHCP-leases to clients.

In theory, this SHOULD work: Just add a 2nd Interface aliasing the 1st Interface with static IP and a non authorative dhcp-server.

But somehow I am always running into an issue.

  • 1st IF: DHCP on br-lan, 2nd IF static on br-lan

    Internet works, but no local network access and no DNS after reboot. Restarting the 2nd IF brings local addresses and dns back online.

  • 1st IF: DHCP on br-lan, 2nd IF static on alias @1st-lan
    DHCP Interface works, but static interface does only work when 1st interface has received a dhcp-adress, which is useless.

  • 1st IF: DHCP on alias @2nd-lan, 2nd IF static on br-lan
    Internet works, but no local network access and no DNS. restarting interfaces does not help.

Does anyone know about this and knows how to fix this?

For my testing I used a Raspberry Pi 4 + an amd64 VM in Proxmox, both Squashfs, compiled and heavily preconfigred by myself.

at the moment i think the best solution would be to just add an ifup lan2 to the startup routine, but this seems very hacky.

2

What are the devices in question?

What is the purpose of these devices on the networks when deployed? (Router, AP, some other purpose/service)?

3

Raspberry Pi 4 + an amd64 VM in Proxmox. Also Nanopi R3S LTS but I haven't tested these yet.

The use case is pretty wide but in a nutshell, it is a cloud solution, but at home: This is the software we're running on top of it: https://cosmos-cloud.io/

It's a web based server / app management including remote access, storage and backup, reverse proxy, auth via proxy or openid, monitoring and more.

The only things missing is routing/firewall and os package / update management, which is why I am using OpenWrt for this.

4

after a lot of testing and researching, I figured OpenWrt is simply not capable of doing this on its own. I created a script which does what I want. Two issues arised while trying to achieve this:

  • OpenWrt does NOT support having 2x IP's on the same NIC, even not with aliases due to the described issues.
  • Additionally, OpenWrt fails to detect a hotplug event when you plug in an ethernet-cable. It's simply ignored by hotplug. There are several threads about this as old as +10yrs out there.

My solution is a routine listening to logread about a cable plugged into eth0. The routine expects br-lan with eth0 as device and 2x Interfaces lan or dhcp and lan11 for static.

required packages: bash procps-ng-pkill

/etc/init.d/staticFallback

#!/bin/sh /etc/rc.common
# description: Fallback to static IP if dhcp-client does not get an IP after 30s
# processname: /bin/staticFallback

START=99
STOP=01
USE_PROCD=1

start_service() {
        procd_open_instance
        procd_set_param command /bin/bash "/bin/staticFallback"
        procd_append_param command daemon
        procd_set_param pidfile /var/run/staticFallback.pid
        procd_set_param term_timeout 60
        procd_set_param stdout 1
        procd_set_param stderr 1
        procd_set_param respawn 150 10 10
        procd_close_instance
}

stop_service() {
        /usr/bin/pkill -f --signal SIGTERM staticFallback
}

/bin/staticFallback

#!/bin/bash
# description: Fallback to static IP if dhcp-client does not get an IP after 30s
# processname: /bin/staticFallback

loglevel="DEBUG"         #INFO is default, DEBUG for more infos
nic="eth0"              #Device to listen to hotplug events (should be the physical nic, like eth0)
bridge="br-lan"         #bridge-device that is used as lan-bridge, default is br-lan
dhcp="lan"              #Interface with dhcp-client IP configuration
static="lan11"          #Interface with static IP configuration
delay=30                #how may seconds to wait for a dhcp-address before configuring the static IP

log() {
        [ "${loglevel}" == "DEBUG" ] && logger -t staticFallback "${1}"
}

while :; do

#wait until network device is up
(logread -f & ) | grep -q "Network device '${nic}' link is up"

#only do something if config is as expected (dhcp-nic is dhcp-client and static-nic is static)
if [ "$(uci get network.lan.proto)" == "dhcp" -a "$(uci get network.lan11.proto)" == "static" ]; then

        #make sure of nic state ( set dhcp-client up and static down )
        log "bring nics into default state (${dhcp} up and ${static} down)"
        ifup "${dhcp}" && ifdown "${static}"

        #check the local IP of the nic the script listens to
        log "check the local IP of the nic the script listens to"
        localIP="" && count=0
        while [[ -z "${localIP}" && ${count} -le ${delay} ]]; do
                localIP=$(ip -4 addr show "${bridge}" | grep -oP "(?<=inet ).*(?=/)")
                sleep 1
                ((count++))
                log "#${count}: localIP is: ${localIP}"
        done

        #check if the Interface did not receive an IP
        ( [ -n "${localIP}" ] && log "found dhcp-client IP: ${localIP}" ) || ( log "found no IP, start static nic" && ifdown "${dhcp}" && ifup "${static}" && service dnsmasq reload )        
        count=0
        while [[ -z "${localIP}" && ${count} -le ${delay} ]]; do
            localIP=$(ip -4 addr show "${bridge}" | grep -oP "(?<=inet ).*(?=/)")
            sleep 1
            ((count++))
        done
        logger -t staticFallback "staticFallback-routine completed, current LAN-IP is: ${localIP}"
else
        log "${dhcp} and ${static} are not setup as expected, do nothing"
fi

done