ipv6 default - ::/3 ?

[Marcus Keane]

Hi,

Just FYI, we saw something similar internally. We introduced ::/1 and 8000::/1 to get around some problems(don't ask :-)), and from that point, we could never get rid of ::/1. It was with 6500s and 12.2(33)SXI*. We could see with debug that the 6500s were definitely receiving ::/0 from the upstream AS, but they interpreted it as ::/1 and they also passed on ::/1 to the route-reflectors. The only way we could get rid it was to take an outage and completely remove the default for a few seconds. We had a TAC case, but in discussing with colleagues, we never got a bug ID. However, they did say the fix was being ported to the SXI branch. We didn't pursue it because we just wanted to put in the workaround immediately. I haven't tested, but apparently IOS-XR will reject these routes as bogons.
I know this probably won't be very helpful in your case, but it might be interesting.
Marcus

-----Original Message-----
From: ipv6-ops-bounces+mkeane=microsoft.com at lists.cluenet.de [mailto:ipv6-ops-bounces+mkeane=microsoft.com at lists.cluenet.de] On Behalf Of Garry Peirce
Sent: Wednesday, 14 November 2012 12:24 AM
To: 'Sander Steffann'
Cc: ipv6-ops at lists.cluenet.de
Subject: RE: ipv6 default - ::/3 ?

Thanks Sander - that's very interesting.

However, I've never adjusted the timers and would characterize my config as being quite vanilla.
Surely there seems something amiss within IOS involving this mutation from
::/0 to ::/3.
Perhaps there is some other function , beyond just timer adjustment, that causes this.
Also to note, enabling soft inbound also shows the ::/3 entry as being RX'd although a bit of debugging shows the ::/0 is coming in.
I do not know the specific platforms/images running on the peers (yet), but the above seems to tell me it's on my side. 

Sounds like best to open a case - I'll send an update when/if something is found.



> -----Original Message-----
> From: ipv6-ops-bounces+peirce=maine.edu at lists.cluenet.de [mailto:ipv6-
> ops-bounces+peirce=maine.edu at lists.cluenet.de] On Behalf Of Sander
> Steffann
> Sent: Monday, November 12, 2012 7:20 AM
> To: peirce at maine.edu
> Cc: ipv6-ops at lists.cluenet.de
> Subject: Re: ipv6 default - ::/3 ?
> 
> Hi Garry,
> 
> > I've been seeing an odd v6 situation and wondered if any may be 
> > seeing
> the same.
> >
> > I have two different v6 providers originating a default v6 (expected 
> > to
be
> ::/0) into two different routers of mine.
> > I'm not seeing it from either ;  however my routers tell me I am 
> > RX'ing
a ::/3
> from each which I believe is a bogon.
> > They claim they are default-originating a ::/0  , are not 
> > advertising a
::/3   ,
> nor do they have this prefix in their tables.
> >
> > As it looks oddly similar to 2000::/3, I'm curious if it's being 
> > mucked
with
> within IOS when received.
> > I'm seeing this on Cisco76xx under SRE6,7 & 7a.
> > I've tried removing filtering/routemaps with no change in behavior.
I've
> not had an opportunity to sniff it.
> >
> > Outside of the aspect of accepting/using a default, anyone else 
> > seeing
this
> or know why it may be?
> 
> This is a bug that I ran into earlier this year. It happens when you 
> set
the BGP
> timers to a low value on the receiving router. Here is an example. 
> This is before setting the timers:
> 
> > router1#sh bgp ipv6 unicast neighbors 2001:DB8:810:1::1 
> > received-routes BGP table version is 1, local router ID is 10.0.56.5 
> > Status codes: s suppressed, d damped, h history, * valid, > best, i 
> > -
internal,
> >              r RIB-failure, S Stale, m multipath, b backup-path, x 
> > best-external, f RT-Filter Origin codes: i - IGP, e - EGP, ? - 
> > incomplete
> >
> >   Network          Next Hop            Metric LocPrf Weight Path
> > *  ::/0             2001:DB8:810:1::1
> >                                             0             0 65531 i
> >
> > Total number of prefixes 1
> 
> Then I change the timers:
> 
> > router1#conf t
> > Enter configuration commands, one per line.  End with CNTL/Z.
> > router1(config)#router bgp 65530
> > router1(config-router)#neighbor 2001:DB8:810:1::1 timers 10 30
> 
> Now clear the session and look again:
> 
> > router1(config-router)#do clear bgp ipv6 unicast 2001:DB8:810:1::1 
> > router1(config-router)#do sh bgp ipv6 unicast neighbors
> > 2001:DB8:810:1::1 received-routes BGP table version is 54339, local 
> > router ID is 10.0.56.5 Status codes: s suppressed, d damped, h 
> > history,
*
> valid, > best, i - internal,
> >              r RIB-failure, S Stale, m multipath, b backup-path, x 
> > best-external, f RT-Filter Origin codes: i - IGP, e - EGP, ? - 
> > incomplete
> >
> >   Network          Next Hop            Metric LocPrf Weight Path
> > *  ::/3             2001:DB8:810:1::1
> >                                             0             0 65531 i
> >
> > Total number of prefixes 1
> > router1(config-router)#
> 
> And suddenly we have the ::/3 route. The very annoying thing is that 
> it doesn't disappear when removing the timers:
> 
> > router1(config-router)#no neighbor 2001:DB8:810:1::1 timers 10 30 
> > router1(config-router)#do clear bgp ipv6 unicast 2001:DB8:810:1::1 
> > router1(config-router)#do sh bgp ipv6 unicast neighbors
> > 2001:DB8:810:1::1 received-routes BGP table version is 54339, local 
> > router ID is 10.0.56.5 Status codes: s suppressed, d damped, h 
> > history,
*
> valid, > best, i - internal,
> >             r RIB-failure, S Stale, m multipath, b backup-path, x 
> > best-external, f RT-Filter Origin codes: i - IGP, e - EGP, ? - 
> > incomplete
> >
> >  Network          Next Hop            Metric LocPrf Weight Path
> > *  ::/3             2001:DB8:810:1::1
> >                                            0             0 65531 i
> >
> > Total number of prefixes 1
> > router1(config-router)#
> 
> Only reloading the box after removing the timers fixes this...
> 
> This was tested on a test-box without a active support contact so I
couldn't
> create a TAC case for it, but I did report it to people inside Cisco. 
> The firmware was an 15.1XB* release from lisp.cisco.com from around 
> April. I don't know the exact XB release anymore though :-(
> 
> - Sander