[CentOS] centos 6.3 ipv6 default gateway

Thu Apr 11 19:38:07 UTC 2013
Michael H. Warfield <mhw at WittsEnd.com>


I may be totally off base here but...

On Thu, 2013-04-11 at 18:06 +0800, Jaze Lee wrote:
> hello,
>    i met a problem in configuratiion of ipv6 gw in my box
> i install centos 6.3 (64 bit) on my boxs, which have four netcard.
> i use a straight-through cable to connect centosv0:netcard-2 and
> centosv1:netcard2
> the topology is this:
> client c(windows xp) <-->centosv0:netcard-3 <--> centosv0:netcard-2 <--->
> centosv1:netcard-2 <---->centosv1:netcard-2  <---> client d (backtrack r2
> 32)
> 1:2:3:4::2/64          1:2:3:4::1/64                       1:2:3::4/64
>                   1:2:3::5/64             1:2:3:5::1/64
>  1:2:3:5::2/64

Surely, I hope you jest with those numbers.  You are not allowed to pick
numbers out of the air and just use them, even if it's for private use.
There are specific blocks of addresses for specific uses and assigned
"scopes" and all the "private use" addresses are in blocks very high up
in the address space beginning with fc or fd.  If those are literally
the addresses you used, they will not work and I would expect them to
give you all sorts of grief at some point or another.

> what i want to do is set default gw on centosv0 to centosv1

I take it "centosv0"  and "centosv1" are configured for ipv6 forwarding?
You didn't provide the information on that.  There are some gotcha's in
there with default routing on a router (basically there is no such
thing) and the router needs to be set up properly for both routing and
its routes.  But I don't think that's your problem you're describing
down below.

> i configure /etc/sysconfig/network-scripts/ifconfig-eth2  (centosv0) as this
> DEVICE="eth2"
> BOOTPROTO=static
> HWADDR="60:A4:4C:23:2F:6F"
> ONBOOT="yes"
> TYPE="Ethernet"
> #UUID="0ddcf499-878f-4ac7-9d1a-c27f85d2bccf"
> IPV6INIT=yes
> IPV6ADDR=1:2:3::4
 ^^^^^^^^^^^^^^^^^^ You didn't specify a netmask here (default /128).

> IPV6_DEFAULTGW=1:2:3::5
 ^^^^^^^^^^^^^^^^^^^^^^^^ Technically not on your interface's network (/128)

> and i also configure /etc/sysconfig/network to this:
> HOSTNAME=centosv0

For forwarding...

In that file you're also going to need:


You may also need to add lines to /etc/sysctl.conf (I've needed in the
past on Fedora):

net.ipv6.conf.all.forwarding = 1
net.ipv6.conf.default.forwarding = 1

But those aren't your problem with this...

> but i met an error:
> Bringing up interface eth2:  WARN     : [ipv6_add_route] Unknown error

I'm not totally sure if this is because you didn't specify a prefix
length on your IPV6ADDR line or the fact that it then conflicted with
your IPV6_DEFAULTGW which would not have been on 1:2:3::4/128 or if it
was because you choose and illegal IPv6 prefix or if it was a
combination of all of them.  The "WARN: [ipv6_add_route] Unknown error"
makes me suspicious because your default gatway conflicts with your
interface network definition (because you didn't specify the prefix size
and it defaulted to /128) and the kernel has no way to route it out any
interface.  IAC...  You won't be able to use a default route on a router
anyways (more below).

> i do not know how why,and can some one gives me some suggestion?
> thanks a lot.

If those were literally the addresses you used, It may be an address
that's in an illegal scope.  IPv6 does not behave quite like IPv4 does
and you need to know what some of these blocks of addresses do and what
their scope is.

"Local" IPv6 unicast addresses begin with the prefix fc00::/7 and there
are recommended procedures for assigning subnets out of them and
choosing network prefixes...


Those may be routed between your machines but may not be routed on the
global net either as a source or destination address.  Your machines
should also be given "link local" addresses which are valid only on that
network segment.  They're in the fe80::/64 prefix.

Global addresses are in the 2000::/3 block.  If you are using a Linux
system as an IPv6 router, the kernel is going to disable the default
route (::/0), preventing non-global addresses from routing.  You'll have
to add appropriate routes for all your "local" (fc00::/7) subnets and
also provide a global unicast default route using 2000::/3 on the

Don't try to do your setup above with the two routers pointing default
routes at each other.  Point specific static routes for each subnet
behind each respective opposite router.

Wikipedia has a rundown on the various address blocks and formats:


Local addresses in particular are described here:


Anything in 1::/16 (if that's what you're doing) is going to be illegal
afaik as it's not in an assigned block and scope.  It should reject it
as being unroutable or having a non-valid scope.

Certain addresses below 2000::/3 are used for compatibility purposes.

::a.b.c.d use to be an IPv4 compatibility address but is largely

::ffff:a.b.c.d are IPv4 / IPv6 transition addresses for applications
running in a dual stack environment where they see IPv4 addresses as
IPv6 addresses in the ::ffff:0:0/112 block.  All those addresses are for
internal use and are seriously hands off.

You can not treat IPv6 arbitrarily as if it were IPv4 with fat
addresses.  If you need to learn more about IPv6 and how it works, you
probably might want to start looking at Hurricane Electric aka
Tunnelbroker.net, http://www.tunnelbroker.net .  They have some very
good IPv6 interactive tutorials there for free and are very quick for
the basic stuff.  The first few exercise could be very helpful to you.
If you follow it all the way through, you will find yourself learning
how to set up DNS properly for IPv6 and registering your own IPv6 glue
records with your registrars.

Now, if I'm off base here and you were merely obfuscating your real
addresses, I would recommend obfuscating them with fc00: instead of 1:
and those would be valid example addresses.  You could use
fc00:1:1:1::/64 for one network and fc00:1:1:2::/64 for another and
fc00:1:1:3::/64 for yet another.  Read that RFC for recommendations on
what you really should chose (generally a random number for
fdxx:xxxx:xxxx::/48 before your SLA).  Since you've got 2 routers,
you'll need three network prefixes, which I see you have.  Generally,
you'll want to manipulate that fourth field as your SLA (Site Local
Address) which is IPv6 lingo for your subnet address.

Replace the leading "1:" in each of those nets with "fdxx:", add your
appropriate subnets, add your appropriate prefix lengths to those static
address, and add appropriate static routes, and you might get further
along the road.

Michael H. Warfield (AI4NB) | (770) 985-6132 |  mhw at WittsEnd.com
   /\/\|=mhw=|\/\/          | (678) 463-0932 |  http://www.wittsend.com/mhw/
   NIC whois: MHW9          | An optimist believes we live in the best of all
 PGP Key: 0x674627FF        | possible worlds.  A pessimist is sure of it!
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 482 bytes
Desc: This is a digitally signed message part
URL: <http://lists.centos.org/pipermail/centos/attachments/20130411/d394e961/attachment-0005.sig>