[OpenSIPS-Users] Dialogs with fix_nated_contact() have wrong RURI domain on sequential requests

[Jeff Pyle]

Hi Johan,

There typically isn't loose_route() in my script because there
is topology_hiding_match() instead.  But, I've tested without topology
hiding (using loose_route for sequential requests) and there is no
difference.

The docs for fix_route_dialog()
<https://opensips.org/html/docs/modules/3.1.x/dialog.html#func_fix_route_dialog>
say that it "forces an in dialog SIP message to contain the ruri, route
headers and dst_uri, as specified by the internal data of the dialog it
belongs to."  That's not a problem here; the in-dialog request already has
the same values as the internal data of the dialog it belongs to.  This
function looks more to prevent bad actors from doing nasty things in
in-dialog requests.  In my case everyone is playing by the rules.

The caller_contact and callee_contact from the "internal data of the
dialog" (as viewed with the dlg_list MI command) contain the
public/received IP and port rather than the internal/private IP and port
each UA provided.  That occurs because of the fix_nated_contact() function
in the script prior to dialog creation.  In other words, by the time the
dialog is created, the internal IP:port is lost.

My questions are:
 - how to preserve the private/internal Contact info in the dialog, and
 - use it for signaling in the RURI but continue to use the received/public
info for routing for in-dialog requests


- Jeff

On Tue, Jan 26, 2021 at 11:04 AM Johan De Clercq <Johan at democon.be> wrote:

> did you change the loose route part to fix route dialog ?
>
> Op di 26 jan. 2021 om 16:39 schreef Jeff Pyle <jeff at ugnd.org>:
>
>> Hello,
>>
>> This is on OpenSIPS nightly 3.1.1~20210125~8bab0da7b-1.
>>
>> I have a registrar configured with basic call routing between the
>> registered AORs.  I use topology_hiding("D") to create the dialog on calls
>> and normal stuff like has_totag() and topology_hiding_match() for
>> sequential request handling.  All this seems fine.
>>
>> This appears high in the main route and appears to do exactly what it
>> should:
>>
>>         if (has_body("application/sdp")) {
>>                 if (nat_uac_test(14)) {
>>                         setflag("NAT_FLAG");
>>                 }
>>         } else {
>>                 if (nat_uac_test(6)) {
>>                         setflag("NAT_FLAG");
>>                 }
>>         }
>>
>>         if (isflagset("NAT_FLAG")) {
>>                 force_rport();
>>                 if ($rm == "REGISTER") {
>>                         fix_nated_register();
>>                 } else {
>>                         fix_nated_contact();
>>                 }
>>         }
>>
>> And, for replies:
>>
>> onreply_route [handle_rtprelay_onreply] {
>>         # rtpengine and such, omitted for brevity
>>         if (isbflagset("NAT_BFLAG")) {
>>                 fix_nated_contact();
>>         }
>>
>>         exit;
>> }
>>
>> When one client calls another, everything works fine.  lookup("location")
>> works to update $rd with the original (private) Contact provided upon
>> registration, and $du contains the actual received source IP:port to get to
>> the device.  Excellent.  The INVITE goes out accordingly, and all is well.
>>
>> My problem occurs with sequential requests, say, re-INVITEs from on-hold
>> events.  The dialogs themselves save the received IP:port values as the
>> caller_contact and callee_contact values (from fix_nated_contact() above),
>> so when the requests pass through the sequential handling section of the
>> script and topology_hiding_match() does its fixups, the request URI domain
>> of the relayed request has the received IP:port values of the target UA
>> rather than the private IP:port values the UA provided during the initial
>> request that established the dialog.
>>
>> I can't wrap my head around how to fix this.  The initial requests work
>> because lookup() has the intelligence to distinguish the UAC's Contact from
>> the received IP:port at REGISTER-time, but I can't see how to achieve this
>> at dialog-creation time so sequential requests have the right RURI domain.
>> Force the caller_contact and callee_contact to the private values somehow,
>> and manage the route_set to point to the appropriate received IP:port?  I'm
>> not sure how to configure that if it is the solution.
>>
>> Any direction would be appreciated!
>>
>>
>> Regards,
>> Jeff
>>
>> _______________________________________________
>> Users mailing list
>> Users at lists.opensips.org
>> http://lists.opensips.org/cgi-bin/mailman/listinfo/users
>>
> _______________________________________________
> Users mailing list
> Users at lists.opensips.org
> http://lists.opensips.org/cgi-bin/mailman/listinfo/users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opensips.org/pipermail/users/attachments/20210126/ddba2e3c/attachment.html>