I am having problems getting IPv6 to work under OpenWRT, I want to run it DHCPv6 stateful (like DHCPv4), but for some reason setting it to that (or stateless) in the LAN interface page will stop clients having internet access, bar having a valid (as far as I can see) IPv6 address and their DNS requests being seen and responded to by the router. The only way I have got it to work is with the stateless+stateful setting, but that seems to just decide to stop working after an hour or so (and I have to disconnect the client to get it working again).
I know that some fool at google has decided that DHCPv6 is the work of the devil and that Android users should suffer, but the problems I am having are with a Windows 10 client!
I'm ok with the default stateless+stateful setting so I don't comment on that.
But for the problem with losing the IPv6 connectivity after a certain time, I had a similar issue activating my IPv6.
Seems there's a bug somewhere that RA updates from your ISP is not updating route validity time.
Can check with ifstatus wan6 whether validity timer just expires and will not get updated after ISP RA announcements;
Once the last "valid" field reach 0 the IPv6 connectivity is gone. My ISP is announcing RA every 3sec with 1800sec lifetime so I had a drops after every 30min. Restart of the box restored IPv6 connectivity for that 30min.
In some other thread here I found the workaround for this by setting wan interface to allmulti.
So now I've added below command in my rc.local and all works ok.
ifconfig eth0.2 allmulti
Please don't ask me any cons of this setup as I just followed instructions and findings by others.
What happens when it stops working is very odd.
I can ping the LAN IPv6 address from the PC, I can ping the PC's publicly routerable address from the router, stubby (which is what I use for DNS over TLS) seems to be able to make a connection to cloudflare's IPv6 DNS servers, but dnsmasq does not get any IPv6 addresses returned to it and I can not ping public IPv6 addresses from the PC.
I'm using also stubby for DNS over TLS and Cloudflare DNS server. Only oddity (most likely misconfiguration somewhere) I have is that Client IPv6 DNS requests are using IPv4 DoT toward Cloudflare servers unless I monitor with stubby -l. Client IPv6 connectivity works ok though and nothing is sent out from router using port 53.
I am using 18.06.02 and IPv6 is coming from my ISP
Edit: After changing the LAN assignment length to /64 instead of /56 and removing always announce default router, I now have connection again, but only for a little while before it stops.
To get the connection back I have tried restarting odhcpd, refreshing IPv6 leases, shutting down and restarting the WAN connection, disconnecting and reconnecting the client to the router and even using other computers with different WiFI adaptors The only way I have found to get connection again is to reboot the router!
I also do not think it is a routing table issue, I have looked at it both when it was working and when it was not and I see no differences.
I have now set the LAN interface's Router Advertisement-Service and DHCPv6-Service to "Hybrid" and that does give me IPv6 for a bit longer, but after a while the PC loses it's IPv6 gateway (which is fixed with a reconnection)