Re: [Accel-ppp-users] discarding message with invalid tid 0

[Guillaume N. <g....@al...>]

On Fri, Oct 05, 2018 at 02:21:16PM +0200, Alarig Le Lay wrote:
> On ven.  5 oct. 12:53:58 2018, Guillaume Nault wrote:
> > Well, the tunnel is not established yet, but it has been allocated and
> > associated with the 5-tuple of the original SCCRQ. Given that a SCCRP
> > message has been set, accel-ppp now expects a SCCCN message from the
> > client and that message must contain a valid Tunnel ID.
> 
> From what I see in the capture, I don’t see any SCCCN from the peer.
> https://bulbizarre.swordarmor.fr/garbage/documents/l2tp-lns.pcap
> 
Given that the peer sends a StopCCN with a null Tunnel ID, it is very
likely that it never got your SCCRP.

> > The "discarding message with invalid tid 0" may comes from a
> > retransmission of the original SCCRQ, which is rightfully ignored.
> > If it comes from a SCCCN message then you will probably need to enable
> > the "dir300_quirk" l2tp option, but I would be really surprised if that
> > was the issue.
> > 
> > Most likely, either the LAC does not receive the SCCRP, or the LNS
> > (accel-ppp) does not receive the SCCCN. Do you have some filtering
> > involved?
> 
> I don’t have any filtering on my side. And if I do a tcpdump on the
> interface where the BGP is done with the peer, I see the exact same
> packets, just two of them are inverted.
> https://bulbizarre.swordarmor.fr/garbage/documents/l2tp-bgp.pcap
> 
Thanks for the trace. Do BGP and L2TP run on the same machine? Or is
the second "BGP" trace taken from a different machine? I ask because
I'm surprised to see incorrect checksums on your second trace. If this
machine only acts as a router between your L2TP server and your
provider we should have correct checksums, regardless of offloading
options.

Also, I think there was an issue a few years ago that prevented the
kernel l2tp module from working properly when checksum offloads were
enabled. That's old memory, so I'm not so sure about it. But if your
kernel is fairly old, I would be worth trying to disable checksum
offloads (ethtool -K ethX rx off tx off).