I have a VLAN reserved for guest users and my main VLAN. When I connect to either one I'm getting assigned correct IPv6 addresses using stateless configuration, it works as expected. However, after several minutes the interface clients get another set of IPv6 addresses belonging to the other VLAN.
So if I connect to the guest network I get a correct guest address (which works) and after a few minutes I getting an address from the main LAN (which does not work).
I've tried playing with delegated prefix sizes, with disabling built-in IPv6 and so on. No luck so far. Any ideas what I can do to debug this problem?
FYI limit is the pool size, not the last IP. To serve from 10 to 250, you need to change the limit to 151.
Remove them from dhcp and add them under wan interface. Although it would be best to separate the IPv6 part of wan interface with a separate wan6 interface.
You may want to change input and forward to REJECT or DROP. Otherwise your router is wide open.
Source is wan not lan.
You should not allow access to DNS from wan, so the wildcard * is wrong here.
Opening the uhttpd server on the wan is very bad practice, as it is not hardened to withstand attacks.
Same here, input and forward REJECT or DROP.
Wireguard running on OpenWrt listens to 51820, so nothing will be forwarded to lan. And WG uses only udp.
I won't go into more detail, I have spotted enough problems already to make me think it would be best to reset to defaults and start from scratch. Regarding your topic, I believe the managed you are using is not configured properly with the native vlan which is the lan interface and you have frames jumping from one vlan to the other. Assign a vlan on lan interface (it could be 10) and allow only tagged frames on the switch.
Lots of these notes are actually intended, like publiclly available DNS and permissive forwarding rules (OpenWRT actually serves as an authoritative DNS, with public recursion disabled). The management on the WAN interface is enabled for now to allow easier debugging from outside the network.
I've looked into the configuration several times and I don't see how RAs from one interface can leak into another.
I'm going to try to allow only tagged frames, let's see if it helps...
That was already turned off from what I see here. You assigned a /64. The delegate 0 is for DSlite/map/646xlate.
Could be another device that receives a delegated prefix from LAN and redistributes it in guest?
Indeed, I guess I left it on during one of my experiments.
I had the leakage even with my computer directly connected to the router and I can reproduce it again by using this config. The leakage disappeared after I tried a couple of small changes, like switching from a combined V4/V6 connection to V4 and a separate V6.
It definitely seems like a strange setting interaction somewhere in the stack. I'm going to try to debug this in a test environment.
I have still a problem with ipv6 RA in wrong/all vlans on E8450. Offload is not related as it is disabled.
A (DSA) bridge br-lan with lan1-4 and some vlans br-lan.X is configured.
Every bridge interface has a unique MAC set and got a public ipv6 (or ULA!) and then odhcp (or dnsmasq-v6) sends RAs for every bridge-interface.
Now RAs for all existings MACs/LL-IPv6 are send to every Vlan. To force: service odhcpd restart.
IPtable logs show that the RAs exit the device on the correct br-lan.X interface - but tcpdump on a host on the lan sees all RAs - and with SLAAC it get s many IPv6s of which only 1 works
I'm not sure if this is a problem of DSA or IPv6. As IPv6 also does not work with sw-offloading it could be IPv6
Its not a bug, its a feature ... of Windows!
All Windows versions strip the Vlan tags and used the RAs. So all router of all vlans are assigned and slaac then add many the ips. Even when a vlan-capable nertwork driver is installed and you set a Vlan you get all RAs.
Amusing: DHCP with IPv4 works well - maybe there are some workarounds in the OS