Windows 7 IPv6 fail

[Arifumi Matsumoto]

Martin,

I can understand that can happen in the situation where you mentioned.
The source address selection algorithm uses address status, that is
deprecated or perfered, before looking at the outgoing interface selected.

But, I'm wondering why such a situation happens. The router lifetime
is usually very shorter than that of address preferred lifetime.
The former is 1800 seconds, and the latter is 7 days by default.

Best regards,

On 2010/10/30, at 4:51, Martin Millnert wrote:

> On Fri, October 29, 2010 9:36 pm, Brian E Carpenter wrote:
>>> But Teredo was used, but not through the tunnel. It took the wrong way
>>> out
>>> (directly on my ethernet).
>> 
>> Do you mean that it generated IPv6 native packets, rather than
>> proto-41 IPv4 packets? That would certainly be weird. If it
>> happened to me, I'd write direct to Dave Thaler <dthaler at microsoft.com>.
>> 
> 
> That's precisely what happened.  The return packets from the IPv6
> resolvers were routed via Teredo boxes, and presumably came over the
> Teredo tunnel interface back (the tcpdump agrees with this, since it sits
> as the only v6 router involved in routing traffic back and forth between
> the host computer's native v6 connectivity and the IPv6 resolvers).
> 
> This is also consistent with traffic I have been observing for some period
> here on that v6 router, which also serves other customers, so my host is
> certainly not the only host doing it.
> 
> I'll write to Dave then.
> 
> Thanks Brian,
> Regards,
> Martin
> 
>> Regards
>>   Brian
>> 
>> On 2010-10-30 05:13, Martin Millnert wrote:
>>> On Fri, October 29, 2010 12:57 pm, Sean Siler wrote:
>>>> Here is my translation:
>>>> 
>>>> You were trying to reach an IPv6 DST (resolvers) with no valid native
>>>> v6
>>>> SRC, therefore Teredo was used to connect instead. Additionally, the
>>>> RAs
>>>> (I assume) were still present and configured your host for those v6
>>>> resolvers. Since they were still present, so were the resolvers.
>>>> 
>>>> Sean
>>>> 
>>> Hi Sean,
>>> 
>>> thanks for your reply.
>>> 
>>> See the attached file for more detail (if you didn't before).
>>> Specifically, the addresses configured from the prefix carried in the RA
>>> (2a02:9a0:4:104::/64) were deprecated, which you note.
>>> 
>>> But Teredo was used, but not through the tunnel. It took the wrong way
>>> out
>>> (directly on my ethernet).
>>> 
>>> It really does seem to me that this situation can be avoided somehow. :)
>>> 
>>> Cheers,
>>> Martin
>>> 
>>>> -----Original Message-----
>>>> From: ipv6-ops-bounces+sean.siler=microsoft.com at lists.cluenet.de
>>>> [mailto:ipv6-ops-bounces+sean.siler=microsoft.com at lists.cluenet.de] On
>>>> Behalf Of Martin Millnert
>>>> Sent: Friday, October 29, 2010 5:39 AM
>>>> To: ipv6-ops at lists.cluenet.de
>>>> Subject: Windows 7 IPv6 fail
>>>> 
>>>> Hi list,
>>>> 
>>>> I'm running Windows 7 on my home box dual-booted, so I had the luck to
>>>> find a quite severe error (IMHO) in its IPv6 stack today:
>>>>   When (all?) IPv6 addresses on its wired interface becomes deprecated
>>>> for whatever reason, it will still send out packets to the (wired)
>>>> on-link router with Teredo source address.  (uhm, yeah, we were missing
>>>> BCP38-style filters on the v6 router)
>>>> 
>>>> My Windows 7 network configuration is entirely vanilla, everything
>>>> automatic from the network.
>>>> 
>>>> For a better description on what I found, please see the attached
>>>> text-file.
>>>> 
>>>> 
>>>> So this is incorrect behaviour, for various reasons. In my case, our v6
>>>> resolvers that we hand out via stateless DHCPv6 (resolvers only), are
>>>> being used for domain name look-up with a Teredo source address.
>>>> Reasonably, this is unsuccessful, since the source address isn't within
>>>> the ISP allocated space, and the resolvers won't answer that.
>>>> 
>>>> Most web browsing worked fine, since there are 3 IPv4 resolvers to
>>>> fall-back to.  But I stumbled on one name which the system failed
>>>> entirely
>>>> to look-up: sr6.se. I'm not sure how this came about; the v4 resolvers
>>>> would answer this name, yet the system wouldn't give me an answer.
>>>> ping/browsing to sr6.se failed.
>>>> 
>>>> A myriad of things could be 'fixed' / looked in to, IMO. For example:
>>>> - Why query resolvers configured on the wired interface using an
>>>> address
>>>> configured on another interface?
>>>> - Why send packets on the wired interface with a source address taken
>>>> from another interface, in this case Teredo?
>>>> - Why keep/use the v6 resolvers on the wired interface when addresses
>>>> becomes deprecated, and wired IPv6 essentially drops? At least when
>>>> resolvers aren't using link-local addresses?
>>>> 
>>>> 
>>>> MS clue greatly appreciated.
>>>> 
>>>> Cheers,
>>>> --
>>>> Martin Millnert <martin at millnert.se>
>>>> 
>>>> 
>>>> 
>>> 
>>> 
>> 
> 
> 
> -- 
> Martin Millnert <martin at millnert.se>
>