IPv6 /127

RouterOS General
1

Not sure RouterOS likes /127’s or at least it doesn’t seem to like /127’s where the gateway is ::

For instance
2607:XXXX:XXa::1

with a gateway of 2607:XXXX:XXa::

RouterOS would ping 2607:XXXX:XXa:: but it would show unreachable in ipv6 routes and wouldn’t use it as the default gateway

I swapped to a /126 and made the gateway 2607:XXXX:XXa::2 instead and routerOS was happy with that.

CHR v6.35.2

The previous VyOS VM didn’t seem to have any trouble with that config (I swapped it as I wanted to use Winbox + Ipsec was rather temperamental)

2

do not use /127 network mask in RouterOS.
https://tools.ietf.org/html/rfc3627 AFAIK, no points in this has been addressed. Should use /128 instead.

3

https://tools.ietf.org/html/rfc6164
https://tools.ietf.org/html/rfc6547

4

here is an updated answer. Yes, /127 was reconsidered after one company that used /31 addressing extensively and pushed for /127 to keep same functionality when converting to IPv6. Currently, such a change is not considered and /127 should not be used in RouterOS configurations.

5

Perhaps you should consider using only link-local addresses on your attachment circuits. Routers don’t really need a WAN address when using IPv6. If your router needs to generate ICMPv6 messages to other global IPv6 addresses on the internet, then it will just use one of its LAN interfaces’ routable addresses for that purpose. (e.g. it will use the LAN address when replying to traceroutes, for instance).

I’ve been considering using link-local addressing for all router-to-router links - there are a few drawbacks to this method, though. For instance if there are multiple links between a pair of routers, you couldn’t use traceroute to see which one was used because the loopback IP of the router is the only address that would show up as each hop…

6

In this case the other “Router” is actually a proxmox host, due to it being a rented dedicated server and me not having a seperate router onsite.
Essentially the Host has an IPv6 address (now a /126) and next-hops the CHR virtual machine.

Same for Ipv4 execpt in that case it’s RFC1918 space and only a /30 of public IP space routed.

Edit:

/126 even, I wasn’t stupid enough to try /128 on the router just wasn’t paying enough attention to what I’d posted.

7

It goes against 20 years of deeply-ingrained IP miserliness in me, but apparently the correct thing to do in IPv6 is to assign a /64 to this interface.
In v6, a /64 is roughly analogous to a /32 in v4.

(I’m still going to use a single /64 for all router loopback interfaces system-wide and assign /128s from it - I just can’t make myself burn space so extravagantly that I’d put a /64 on every router’s loopback interface. That’s just obscene.)

8

I thought the advise had since changed regarding the use of such subnet sizes on router2router links.

9

Yes, router-to-router links are an exception to the /64 rule. Sitting here, I can’t recall if /127 or /126 is the recommended point-to-point netmask.
If it’s /127, then Mikrotik needs to fix this limitation in ROS. (shouldn’t be terrible since there’s no such thing as “broadcast address” in IPv6)

10

for point to point links it is /128 in RouterOS and AFAIK elsewhere also, same as with IPv4 it is /32

11

You are confusing PPP with Ethernet links that only have 2 devices on (which can also be refereed to as a point2point link as there are only 2 possible destinations) in which case a /31 is acceptable to use providing both devices know how to support it.

12

Both PPP and Ethernet with only two devices are point to point and in both cases /128 is applicable.

13

A /128 contains 1 IP address what do you propose to put on the other end of the Link?

14

Any other /128 address and route it over link-local. I know it is not as easy as simply adding /127, but you must understand that there will be no /127 support in RouterOS v6.

15

I already worked around that issue by making it a /126 :wink:

16

/128 contains 1 address, on the other end you set up any other /128. Set up routes for this to work. For IPv4 setting /32 network is set to other ends /32 address.

Also, /127 will work in RouterOS 7.x

17

Without a corresponding interface route it won’t work as it won’t know which interface to send the packet out of to reach that /32 or /128

/32 /128 should really only be used on PPP links where you can just next hop the interface and in loopbacks for IGP’s.etc

I have done things with /32’s on providers like OVH as that’s the way they assign additional IP’s so yes out of subnet routes are possible but as to how well it will work will depend on the platform you are using.

18

RouterOS v7 (aka ROS Longhorn :smiley:) /127 teaser:

[admin@rack2_b6] /ipv6 address> print   
 7  G 2001:1::1/127                                         ether6      no 
[admin@rack2_b6] /ipv6 address> /ipv6/route/  
[admin@rack2_b6] /ipv6 route> print 
   DAC>' 2001:1::/127             ether6                          0
[admin@rack2_b6] /ipv6 route> /ping 2001:1:: 
  SEQ HOST                                     SIZE TTL TIME  STATUS             
    0 2001:1::                                   56  64 53ms  echo reply         
    1 2001:1::                                   56  64 0ms   echo reply
[admin@rack2_b6] /routing bgp peer> print status 
Flags: D - dynamic, F - stopping, X - disabled, I - inactive, E - established 
 0     ;;;     RemoteAddress=2001:1::
       ;;;     sk=Socket{ 28[1] cb } [afi IP vrf main fastNotReady=0 slowNotReady=0 ] seenEOR=0 sendEOR=2 state=established
       name="bgp-peer-1" remote.address=2001:1:: .as=65530 multihop=no remote.id=10.5.130.150 .hold-time=3m 
       used.keepalive-time=1m .hold-time=3m uptime=24s110ms prefix-count=0 sent.updates=0 .withdrawn=0 .bytes=0 
       received.updates=0 .withdrawn=0 .bytes=68 refresh-capability=yes as4-capability=yes state=established 
       poll-in restart-timer=2m35s870ms
19

Every time there’s a MUM, I check to see if Longhorn beta release has been announced.
Luckily for me, I only started doing this in late 2015 so I haven’t been hungry for it as long as some others here. :slight_smile: