No traffic gets past HE ipv6 tunnel

toskium

That's the current ipv6 routing table:

Destination	Gateway	Flags	Use	Mtu	Netif	Expire
default	2001:470:6c:aaaa::1	UGS	151	1500	gif0	
::1	link#19	UH	1747	16384	lo0	
2001:470:20::2	2001:470:6c:aaaa::1	UGHS	0	1500	gif0	
2001:470:6c:aaaa::1	link#43	UH	23260	1280	gif0	
2001:470:6c:aaaa::2	link#43	UHS	6	16384	lo0	
2001:470:6d:aaaa::/64	link#32	U	6249	1500	lagg0.4088	
2001:470:6d:aaaa::1	link#32	UHS	0	16384	lo0	
2001:4860:4860::8888	2001:470:6c:aaaa::1	UGHS	0	1500	gif0	
fe80::%igb0/64	link#1	U	0	1500	igb0	
fe80::8261:5fff:fe04:ea3f%igb0	link#1	UHS	0	16384	lo0	
fe80::%lo0/64	link#19	U	0	16384	lo0	
fe80::1%lo0	link#19	UHS	0	16384	lo0	
fe80::%lagg0/64	link#23	U	0	1500	lagg0	
fe80::208:a2ff:fe11:5f66%lagg0	link#23	UHS	0	16384	lo0	
fe80::%igb0.7/64	link#24	U	0	1500	igb0.7	
fe80::8261:5fff:fe04:ea3f%igb0.7	link#24	UHS	0	16384	lo0	
fe80::%lagg0.4081/64	link#25	U	0	1500	lagg0.4081	
fe80::208:a2ff:fe11:5f66%lagg0.4081	link#25	UHS	0	16384	lo0	
fe80::%lagg0.4082/64	link#26	U	21	1500	lagg0.4082	
fe80::208:a2ff:fe11:5f66%lagg0.4082	link#26	UHS	0	16384	lo0	
fe80::%lagg0.4083/64	link#27	U	0	1500	lagg0.4083	
fe80::208:a2ff:fe11:5f66%lagg0.4083	link#27	UHS	0	16384	lo0	
fe80::%lagg0.4084/64	link#28	U	0	1500	lagg0.4084	
fe80::208:a2ff:fe11:5f66%lagg0.4084	link#28	UHS	0	16384	lo0	
fe80::%lagg0.4085/64	link#29	U	0	1500	lagg0.4085	
fe80::208:a2ff:fe11:5f66%lagg0.4085	link#29	UHS	0	16384	lo0	
fe80::%lagg0.4086/64	link#30	U	0	1500	lagg0.4086	
fe80::208:a2ff:fe11:5f66%lagg0.4086	link#30	UHS	0	16384	lo0	
fe80::%lagg0.4087/64	link#31	U	0	1500	lagg0.4087	
fe80::208:a2ff:fe11:5f66%lagg0.4087	link#31	UHS	0	16384	lo0	
fe80::%lagg0.4088/64	link#32	U	619	1500	lagg0.4088	
fe80::208:a2ff:fe11:5f66%lagg0.4088	link#32	UHS	0	16384	lo0	
fe80::%lagg0.4000/64	link#33	U	0	1500	lagg0.4000	
fe80::208:a2ff:fe11:5f66%lagg0.4000	link#33	UHS	0	16384	lo0	
fe80::%lagg0.20/64	link#34	U	0	1500	lagg0.20	
fe80::208:a2ff:fe11:5f66%lagg0.20	link#34	UHS	0	16384	lo0	
fe80::%lagg0.30/64	link#35	U	0	1500	lagg0.30	
fe80::208:a2ff:fe11:5f66%lagg0.30	link#35	UHS	0	16384	lo0	
fe80::%lagg0.40/64	link#36	U	0	1500	lagg0.40	
fe80::208:a2ff:fe11:5f66%lagg0.40	link#36	UHS	0	16384	lo0	
fe80::%lagg0.50/64	link#37	U	0	1500	lagg0.50	
fe80::208:a2ff:fe11:5f66%lagg0.50	link#37	UHS	0	16384	lo0	
fe80::%lagg0.60/64	link#38	U	0	1500	lagg0.60	
fe80::208:a2ff:fe11:5f66%lagg0.60	link#38	UHS	0	16384	lo0	
fe80::%lagg0.70/64	link#39	U	0	1500	lagg0.70	
fe80::208:a2ff:fe11:5f66%lagg0.70	link#39	UHS	0	16384	lo0	
fe80::%lagg0.80/64	link#40	U	0	1500	lagg0.80	
fe80::208:a2ff:fe11:5f66%lagg0.80	link#40	UHS	0	16384	lo0	
fe80::%lagg0.90/64	link#41	U	0	1500	lagg0.90	
fe80::208:a2ff:fe11:5f66%lagg0.90	link#41	UHS	0	16384	lo0	
fe80::%pppoe0/64	link#42	U	0	1492	pppoe0	
fe80::208:a2ff:fe11:5f66%pppoe0	link#42	UHS	0	16384	lo0	
fe80::%gif0/64	link#43	U	0	1500	gif0	
fe80::8261:5fff:fe04:ea3f%gif0	link#43	UHS	0	16384	lo0

for better readability:

kiokoman

@toskium said in No traffic gets past HE ipv6 tunnel:

2001:470:20::2

what is it? ^
and why do you have google dns there ?

ah i see it's he net dns

toskium

@kiokoman this comes from my general DNS settings, the howto on docs.netgate.com stated to add google DNS servers in System > General Setup like so:

kiokoman

but i don't have any dns server on my routing table (i'm also using the /48 from he net)

Internet6:
Destination                       Gateway                       Flags     Netif Expire
default                           2001:470:25:xxx::1            UGS        gif0
::1                               link#4                        UH          lo0
2001:470:25:xxx::1                link#9                        UH         gif0
2001:470:25:xxx::2                link#9                        UHS         lo0
2001:470:26:xxx::/64              link#2                        U           em1
2001:470:26:xxx::1                link#2                        UHS         lo0
2001:470:b4e1:xxx::/64           link#3                        U           em2
2001:470:b4e1:xxx::1             link#3                        UHS         lo0
fe80::%em0/64                     link#1                        U           em0
fe80::5054:ff:fe3d:64cc%em0       link#1                        UHS         lo0
fe80::%em1/64                     link#2                        U           em1
fe80::5054:ff:fe91:db46%em1       link#2                        UHS         lo0
fe80::%em2/64                     link#3                        U           em2
fe80::5054:ff:fe27:556a%em2       link#3                        UHS         lo0
fe80::%lo0/64                     link#4                        U           lo0
fe80::1%lo0                       link#4                        UHS         lo0
fe80::%em1.10/64                  link#8                        U        em1.10
fe80::5054:ff:fe91:db46%em1.10    link#8                        UHS         lo0
fe80::%gif0/64                    link#9                        U          gif0
fe80::6097:dd62:2e35:991d%gif0    link#9                        UHS         lo0
fe80::6097:dd62:2e35:991d%ovpnc1  link#10                       UHS         lo0

toskium

@kiokoman fair enough, but how did they end up there? (I guess that's a rhetorical question...)
Removing them from System > General Setup does not purge them from the routing table.

Edit:
okay, restarting the gif0 interface purges them. It seems like they are added to the routing table when being entered in System > General Setup as a DNS server.

Now that I am able to ping the ipv6 address of the tunnel server over at HE (2001:470:....::1) using:

ping6 -I gif0 2001...

I should also be able to ping other ipv6 hosts, but I can't. For instance ipv6.google.com

ping6 -I gif0 2a00:1450:4005:803::200e

leads to 100% package loss

kiokoman

manually delete it
route -6 del 2001:470:20::2 2001:470:6c:aaaa::1
route -6 del 2001:4860:4860::8888 2001:470:6c:aaaa::1
ok sorry i'm at work, i was too late on answering

i think you have discovered a bug there ^ ...
i have one of my pfsense with a route that appear at boot out of nowhere, i have setup an earlyshellscript to remove everytime that offending route, since 2.4.4-p3
https://forum.netgate.com/topic/147254/lost-ipv6-connectivity-from-one-interface

toskium

Discovering bugs is fine :-) where can I report that properly so it has a chance of being fixed?

kiokoman

https://redmine.pfsense.org

Gertjan

@toskium said in No traffic gets past HE ipv6 tunnel:

@kiokoman this comes from my general DNS settings, the howto on docs.netgate.com stated to add google DNS servers in System > General Setup like so:

A bug, maybe -I'll add some @home and see what happens.

Why did you add all these DNS servers ?
You are aware that you don't need them ?? The resolver, out of the box is close to perfect. [ and then people start forwarding because ... / [ we never know why ] /..... and things go downhill ]

edit :
When I add these :

...the IPv6 of the DNS of he.net, I wind up seeing this :

in the routing table.
Which doesn't look 'wrong' to me, as 2001:470:20::2 should be reached over the interface gif0 = he.net = my (their) 2001:470:1f12:5xx::1

edit : and my IPv6 still works ....

kiokoman

because i use bind on another server and not unbound nor forwarder for example ^^

toskium

@Gertjan I added the DNS servers because the howto says so.

Gertjan

@toskium said in No traffic gets past HE ipv6 tunnel:

the howto says so

Source ?

Read again the initial setup instruction : you'll find https://docs.netgate.com/pfsense/en/latest/config/general.html where it says :

Note
The DNS Resolver is active by default and uses resolver mode (DNS Resolver). When set this way the DNS Resolver does not need forwarding DNS servers as it will communicate directly with root DNS servers and other authoritative DNS servers.

I don't want to say that forwarding - using tiers DNS servers, is bad.
It was somewhat mandatory in the early ages. But not anymore.
You use DNS : use the source aka "Internet itself".

toskium

@Gertjan according to the recipe for setting up a ipv6 tunnel. You will need to go to docs.netgate.com and search vor Hurricane Electric. Askismet doesn't allow me to post the direct link. It's the first result, if you look in the DNS chapter of that recipe you will find the reference.

Gertjan

You mean :

Or https: slash slash docs dot netgate dot com slash /pfsense/en/latest/recipes slash ipv6-tunnel-broker.html#setup-ipv6-dns ?
(Askismet can be circumvented so easily)

Because the doc is somewhat old.

This :

If the DNS Resolver is used in non-forwarding mode, it will talk to IPv6 root servers automatically once IPv6 connectivity is functional.

is not an "if" any more.
The DNS Resolver is used out of the box.
pfSense used a forwarder in the past, it's still there : the lightweight forwarder (dnsmasq), mutual exclusive with the functionality of the resolver (unbound).

Way back, code had to be mean, lean and small, as devices had limited resources.
The Internet start with these, Internet IS https://en.wikipedia.org/wiki/Root_name_server.
So : why take your info from "some one" if you can tap into the source ?

But ...... the damage has been done.
people like to use VPN 'for protection' and '8.8.8.8' for their DNS, and antivirals for their safety. They haven't figured out yet that it's all - and only - a "€/$" thing.

To gain some milli seconds, it could be useful to use a close by DNS server. he.net has a (their) own DNS servers close at every POP.
You don't need them, they are just optional, as all the others.

toskium

@gertjan @kiokoman
Thank you all for providing support but in the end it turns out it's a routing issue on Hurricane Electrics end.

What did I do?
I first registered a tunnel on the nearest server here in Germany (Berlin), I followed the setup but it didn't work.

After doing some debugging with the packet filter it became clear that absolutely no traffic was routed from HEs end to my router.

Since desperate times call for desperate measures I deleted the endpoint and reregistered a new one but in Frankfurt instead of Berlin.

The Frankfurt tunnel works as expected, unfortunately the latency is significantly higher.
I'll write an email to HEs support since the not working Berlin tunnel had been the second attempt, so I guess there is something up with the Berlin endpoint.

JKnott

@toskium said in No traffic gets past HE ipv6 tunnel:

The Frankfurt tunnel works as expected, unfortunately the latency is significantly higher.

You should be able to use any tunnel site. Try another to see if it works any better.

Bob.Dig

Berlin is working fine here. But it was down some days ago.

Gertjan

@toskium said in No traffic gets past HE ipv6 tunnel:

I'll write an email to HEs support since the not working Berlin tunnel had been the second attempt, so I guess there is something up with the Berlin endpoint.

Use their forum. They are reactive. https://forums.he.net/
Do the save approach : if thousands are connected to "Berlin / Düsseldorf or Frankfurt" but you can't, then where would lie the issue ? ;)
If a tunnel goes bad, you will find notification about it on the forum after a minute or so, as it impact many users.

https://www.tunnelbroker.net/usage/tunnels_by_country.php : 5000 (german) users !?

toskium

@gertjan I have read in the internets that this sometimes happens and it is advisable to just delete the tunnel and request a new prefix.

I tried that, but then I realized that I got the same tunnel prefix assigned again.

For the life of me, I did not change anything other than the GIF interface and the ipv6 net on my LAN interface and ping from the router to ipv6.google.com instantly worked.

Yes, I know how that sounds

Gertjan

@toskium said in No traffic gets past HE ipv6 tunnel:

I have read in the internets that this sometimes happens and it is advisable to just delete the tunnel and request a new prefix.

Read that where ? ? ?
Deleting your tunnel will discard your IPv6 /64 and IPv6/48 and you start with new ones.
I'm using tunnel.he.net nearly 10 years now, and never had to do that.
The POP in Paris (France) goes down ones in a while : not an issue as IPv4 and IPv6 are complementary for now : one can replace the other.
If my IPv6 network prefix is not working well, then throwing away a prefix and taking another one on the same POP won't help much : the issue stays.
If the issue happens often, the I tend to say : it's a local issue.