[Olsr-users] OLSR 0.6.8 on OpenWrt with ipv6 and kernel 3.18.x

[Henning Rogge]

Wait,

I remember that thing from Debian...

Most likely the IPv6 socket also receives the IPv4-datagrams because
of "legacy IPv4 support".

can you look for files named "bindv6only" in the /proc/sys/net/ipv6
subdirectories (most likely in conf/all and conf/<if>). If you set it
to "true" the problem should disappear (and we know that my guess was
right).

In Olsrd2 I explicitly mark all sockets as "bindv6only" with a socket
flag, but there might be a reason (beyond "someone forgot") to do this
in olsrd.

I will have a look at the code and will prepare a patch to test while
you test the procfile entries.

Henning

On Mon, Mar 16, 2015 at 8:37 AM, Russell Senior
<(spam-protected)> wrote:
>>>>>> "Henning" == Henning Rogge <(spam-protected)> writes:
>
> Henning> Hi, typically "detected duplicate packet with seqnr" happens if
> Henning> you either have a problem with your linklayer driver OR a bad
> Henning> bridge configuration.
>
> Henning> Are "br-pub" and "ptp" somehow connected on one side or the
> Henning> other side with a bridge?
>
> I ran strace on olsr with ipv6 on OpenWrt r44764 on an alix2 with kernel
> version 3.18.8, as well as on kernel version 3.14.34.  It appear that
> with both kernels, olsrd opens two sockets on each interface.  However,
> with kernel 3.14 only one of the sockets call recvfrom to get incoming
> data, while on kernel 3.18 both of the sockets do.
>
> Apparently, since the same UDP datagram is received on both sockets, it
> sees the same seqno twice, hence the syslog message.
>
>
> --
> Russell Senior, President
> (spam-protected)