R: [Serusers] SER -> PSTN Gateway+NAT: BYE handling problem

Fabio Macchi f.macchi at keeptelecom.com
Fri Mar 16 12:56:06 CET 2007


First, thanks for answer.

I've tryed your trik and in effect this solve the problem of the '200 ok'
forwarded to the UA, but my problem still remain alive: when BYE is sent
from Gateway, it reaches correctly SER, but it still forward it to the
private UA address. I was wondering about the nat_uac_test in this case, as
the source of the BYE message is the gateway ( not natted ) and not the UA.

Have any idea about this ?

Fabio

-----Messaggio originale-----
Da: Kostas Marneris [mailto:K.Marneris at otenet.gr] 
Inviato: giovedì 15 marzo 2007 20.39
A: Fabio Macchi
Cc: serusers at lists.iptel.org
Oggetto: Re: [Serusers] SER -> PSTN Gateway+NAT: BYE handling problem

Hello, 
I was working on about the same problem today either with 'Mediaproxy
solution' 
or with 'SER's Nathelper only solution' .

The NAT issue is a nightmare, not because of SER but because of 
different implementations on NAT boxes.

Actually my problem was :
if the NATed UA send a BYE to SER, SER forward it to PSTN-GW,
then the '200 Ok' Response from PSTN-GW is forwarded by SER to UA
to the wrong port (Contact or Via header port).

I used the following block on Loose Route section, 
(because BYE is loose_routed if you use Record-Route),
and it seems to work.

        # ---------------------------------------
        # Loose Route Section
        # ---------------------------------------
        if (loose_route()) {
                # mark routing logic in request
                if (method == "BYE") {
                        if (nat_uac_test("22")) {
                                xlog("L_NOTICE", "*** LR -> NATed BYE - Use
force_rport()");
                                force_rport();
                        };
                };
                route(1);
                break;
        };




I faced up your second problem too. 
The solution was to move the NAT handling block before proxy_authorize
block.

I think that the different behaviour does not come with the 'standard
RFC1918 addresses',
but with the different NAT type.

I realize that the provisional mesgs '100 Trying' and '407 Proxy
Authentication Required' 
are relayed back to the real IP addr of NATed UA (this is correct),
but to the WRONG port (that of Contact/Via header and not the signalling
received port).
It seems that these mesgs use the IP address part of 'Received' field of
Location DB
but not the port. 

It happens to work if NAT box use the SAME port (eg. 5060) on NAT
translation
(10.10.10.1:5060  --> Real_IP:5060) (eg. with a SAGEM1500 Router)
But it does not work if NAT box doesn't use the same port
(10.10.10.1:5060  --> Real_IP:38181)


I think that this has to be verified by SER developers or SER experts. 



Kostas

---
K.Marneris at otenet.gr

----- Original Message ----- 
From: "Fabio Macchi" <f.macchi at keeptelecom.com>
To: <serusers at lists.iptel.org>
Sent: 15 March 2007 19:34
Subject: [Serusers] SER -> PSTN Gateway+NAT: BYE handling problem


> Hi all,
> 
> 
> 
> I'm running the following schema:
> 
> 
> 
> UA ( possibly natted ) -> SER -> PSTN Gateway 
> 
> 
> 
> I have a problem with UA belonging to a particular network with private
> address not RFC1918 compliant ( class 1.x.x.x ), SER and PSTN Gateway have
> pubblic address.
> 
> 
> 
> The problem is that, after a succesfull call, if the PSTN gateway send a
BYE
> to SER, then SER forward BYE to the private address of UA instead of
pubblic
> one.
> 
> 
> 
> I don't understand which is the section that handle BYE messages and how
can
> I solve this problem: anyone may help ?
> 
> 
> 
> 
> 
> 
> 
> Second, another question: with this particular network I had problem with
> INVITE too, because SER was sending "proxy authorization request" to the
> wrong TCP port. To solve this, I've moved the nat handling ( with
> force_rport ) before the proxy_authorize block and it's working, but why
> this is not necessary on standard RFC1918 compliant natted address ? 
> 
> 
> 
> Many thanks for any explanation
> 
> 
> 
> Fabio
> 
> 




More information about the sr-users mailing list