Subject: Re: what does "DAD detected duplicate IPv6 address" mean?
To: Robert Elz <kre@munnari.OZ.AU>
From: Klaus Klein <kleink@reziprozitaet.de>
List: netbsd-help
Date: 12/17/2003 13:58:57
On Wednesday 17 December 2003 12:15, Robert Elz wrote:

>     Date:        Wed, 17 Dec 2003 14:22:08 +0900
>     From:        henry nelson <netb@irm.nara.kindai.ac.jp>
>     Message-ID:  <20031217052208.GA28790@irm.nara.kindai.ac.jp>
>
>   | Adding interface aliases:
>   | ep0: DAD detected duplicate IPv6 address
>   | fe80:0001::0260:97ff:fe13:39e7: 20 NS, 0 NA ep0: DAD complete for
>   | fe80:0001::0260:97ff:fe13:39e7 - duplicate found ep0: manual
>   | intervention required
>   |
>   | I have absolutely no idea what "DAD" is or where to begin to research
>   | this problem.  What kind of "manual intervention [is] required"? 
>   | Should I just turn IPv6 off (How?) since I'm not using it?
>
> DAD is Duplicate Address Detection
>
> You're being told that two different systems have the same IPv6 address
> on the LAN - this is not good at all.
>
> Since that address is usually automatically generated from the MAC
> address, unless some other system has been manually set up to defeat you,
> it is likely to mean duplicate MAC addresses.
>
> The manual intervention is to trash the card that has a duplicate
> address.
>
> Or, it may be that something on the LAN is deliberately faking duplicate
> addresses, in which case the intervention is to find whoever installed
> that, and have them drawn and quartered - or if the culprit can't be
> located, at least remove the offending equipment.
>
> It is also possible this can happen if you are using a Sun (or other
> system that uses the same MAC address on multiple interfaces) and you
> have more than one interface connected to the same LAN - the intervention
> required there is to manually assign a different Link Local address to
> one (or really, all but one) of the affected interfaces.   (But given
> "ep0" that's not likely).

It is also possible that this is an issue with the ep(4) driver
itself; see PR kern/17523 for a discussion of such an issue with hme(4).
For debugging purposes, putting another system on the same wire and
running tcpdump on it might be useful.


- Klaus