[SOLVED] Weird DHCPv6 client breaking DHCPv4 when set to autostart

Software SW help
1

Hi,

I have DHCPv4 enabled on the LAN Interface handing out DHCPs with Options “6,192.168.1.36,192.168.1.1” (DNS) and as my uplink is fibre box on the LAN port I have to set the gateway to the fibre box in DHCP too with Option “3,192.168.1.3” (which is ofc the IP of the fibre box).

This works perfectly for IPV4 but it killed my IPV6 connectivity because I disabled IPV6 completely on the LAN Interface of turris hoping that the fibre box would make this available through SLAAC, but although it does hand out IPv6 Adresses and DHCPv6-IAID and DHCPv6-Client-DUID, to the machines they are not able to reach any IPV6 test site.

So I added “DHCPv6 Client” Interface on turris and set it try to ipv6 with an /48 adress range and it get’s IP Adresses too (like the machines) but on the turris in diagnostics I can reach out IPV6 hosts with PING6 in contrast to the machines.

But now comes ther very weird part. I noticed that after settting this up my client’s started to fail to detect the DHCP Server and don’t get an IP assigned even stopping the interface did not resolve the situation.

But: If I select eht DHCPv6 Client to not autostart at boot, everything is fine. I can start it manually and DHCPv4 keeps working, but whenever I set that checkbox to autostart DHCPv6 no matter it started or stopped I kills the DHCPv4 functionality. What the heck is going on here, how can the autostart on the DHCPv6 hamper the DHCPv4 from the lan interface?

I have no clue here how I could proceed with settting the proper Ipv6 settings for my machines without loosing DHCPv4 to set the IPV4 address.

Any insights are greatly appreciated

2

I didn’t investigate why, but I remember that I couldn’t make IPv6 work when I used a LAN port instead of WAN. (reassigned that LAN port in reForis to be in logical WAN, etc.)

3

Thanks @vcunat for your reply but I don’t understand yet. Did it work after you reassigned LAN port to WAN so there’s a solution or am I stuck with ipv4 as there’s no way to make it work in Turris/Openwrt?

I guess you understand that double NAT and limiting a 10GB link to 1GB can not be the solution, so connecting the fibre box to WAN port is actually out of the question at least as Gateway between WAN/LAN and having the Turris at hand sounded like a good idea, seems it’s more complicated then I thought though.

4

I was unable to make it work at all (for IPv6), but I didn’t try hard. In the end I just avoided using the physical LAN port for WAN purposes, as I need IPv6 (to allow connecting to various LAN machines from outside). It was years ago, too.

5

I don’t understand, why does the fibre box have to be on the LAN-Interface and not on the WAN?

Do you have two ISPs or WAN connections?

What happens when you connect the fibre box on the WAN interface?

I never understood why we need a physical wan for IPv4 and an additional virtual wan6 interface when IPv6 is used. But maybe that is the reason? Maybe try to create a virtual lan6 interface the same way as the wan6 to separate them somehow.

6

Because it’s a 10GBit Fibre connection, the network ports on the Turris are just 1GBit!

7

But the SFP is a WAN port, i.e. the same electric connection is switched either to the metallic WAN or SFP (unusable together).

EDIT: there’s a picture in docs:

8

There is some parts of your description of your setup that does not make sense.

You say you have dhcp enable on lan, but how come your ”fibre box” is also on lan, looks like it is the same subnet?

With that conf you will not use your turris router as a firewall, or router for that matter.

I think it will be hard so understand wherein your problem lies if you run your router as a switch with a dhcp server?

1 Like
9

It’s not connected to SFP+ port of turris but I’m forced to use some nasty Fibre Router from my ISP which has 10GB RJ45 Ports.

10

Then I’m more confused than ever. The RJ45 in Omnia are all 1G only.

11

Yes I see and you’re not the only one. I degraded the turris to be DHCP & DNS Server on the LAN side of the network including pi-hole and dnscrypt mask running on it as LXC containers and I disabled DHCP on that crippled router from ISP that I’m forced to use.

So the turris does not do the traditional job of the router. It’s just a DHCP & DNS Server nothing more as the one from the ISP won’t let me set my own DNS Server or forward some DHCP requests to my internal DNS for the Samba domain etc.and ofc I don’t want double NAT on IPv4, would be bad for port forwardings.

OFC I could have setup some other DHCPv4 & DHCPv6 server and pi-hole and dnscrypt and all that stuff on a raspberry maybe or some intel NUC or sth but why not just use turris where everything can be easily graphicly configred with luci apart from buying new hardware and finding proper software to do the job which openwrt is doing currently.

12

That is indeed a somewhat unorthodox use of your router :slight_smile:

I can not answer your questions but a way forward could be ro ask your isp to bridge the isp router so you can use your turris as it was intended.

You probably can get more help here that way as well.

13

There’s actually 2 problems with this:
1.) The box does not provide bridge/transparent mode from manufacturer side
2.) The Fibre connection is 10G whereas the turris interfaces are 1G

But I will close the thread now as I managed to get DHCPv6 working from the fibre box, I could hack there even that it uses my pihole as announcement in the RA Options.

1 Like
14

I think you can do ~2G by putting a suitable device into the SFP cage, but for speeds significantly over 1G you probably want a more powerful device than (original) Omnia anyway: