On Wed, Nov 2, 2011 at 2:04 PM, Timo Reimann
<sr(a)foo-lounge.de> wrote:
Hey,
Am 02.11.2011 um 08:24 schrieb Brandon Armstead:
Some interesting results indeed. If I place
the same block of code with or without dlg_get first then dlg_bye AFTER loose_route --
kamailio ignores the packet.
If I place it before loose_route WITH dlg_get FIRST - it works as expected.
What do you mean by "kamailio ignores the packet"? Is this equivalent to what
you described initially regarding dialogs that were terminated single-sided only?
Cheers,
--Timo
On Tue, Nov 1, 2011 at 4:42 PM, Timo Reimann
<sr(a)foo-lounge.de> wrote:
Hey Brandon,
Am 02.11.2011 um 00:32 schrieb Brandon Armstead:
Thats correct - however I am calling this
before loose_route, perhaps that is my problem and need for calling dlg_get? Let me give
this a go - and I will respond back with my findings, thanks!
Yeah, that should be it -- linking to the currently active dialog is being implemented as
a callback to record-routed messages which takes place after a call to loose_route().
Looking forward to hearing your findings. :)
Cheers,
--Timo
On Tue, Nov 1, 2011 at 4:21 PM, Timo Reimann
<sr(a)foo-lounge.de> wrote:
Hey Brandon,
Am 01.11.2011 um 23:44 schrieb Brandon Armstead:
Thank you for the input - I actually just
figured out a resolution.
Apparently dlg_bye("all") must be called after dlg_get - so what I'm doing
now is simply checking for request with a special request uri with the callid / from tag /
to tag in the request that correlates to the call I desire to terminate. I am then
feeding $ci, $tt, and $ft into dlg_get and then calling dlg_bye("all") - in
which Kamailio then proceeds to send BYE out to both legs of the call.
I think we can mark this as resolved :).
OK, very good. :)
Let me make sure that this isn't a dialog module bug: You are now calling dlg_get()
followed by dlg_bye() with proper parameters from within the Kamailio configuration
script? if so, you shouldn't need to call dlg_get() in the first place, at least not
if you place dlg_bye() after the call to loose_route(). Contrary, using certain dialog
features (including dlg_bye()) from reply routes isn't completely possible yet which
could explain your findings.
Cheers,
--Timo
On Tue, Nov 1, 2011 at 3:21 PM, Timo Reimann
<sr(a)foo-lounge.de> wrote:
Hey Brandon,
Am 01.11.2011 um 22:48 schrieb Brandon Armstead:
> I am attempting to tear down a call with a BYE packet generated externally (kind
of similar to Kamailio fifo dlg_end_dlg).
>
> Let me describe what I am trying to do in more depth and then I will continue to
tell you the problem I think I am experiencing.
>
> [PSTN SIP Proxy] -> [CORE SIP Proxy] -> [REGISTRAR] -> [UAC]
>
> So the above layout is the normal call flow / structure of calls (incoming when
originating from pstn) (outgoing when originating from uac).
>
> I then have an "external" host - I am attempting to generate a BYE to
[CORE SIP Proxy] and have it go both directions [PSTN] + [UAC].
>
> So far I am able to get the call to tear down in only a single direction (only kill
call with PSTN) or (only kill call with UAC).
>
> I have not been able to kill both legs of the call.
[snip]
> My Question - what am I doing wrong - or what is the best method to tackle this
task?
If I get you right you are trying to emulate Kamailio's dlg_end_dlg functionality
without involving Kamailio at all (except for passing forward the BYE requests). This
makes your question sort of off-topic as processing BYE requests is a matter of RFC 3261
alone.
Apart from that, what you could do is take a look at the dialog module code and check how
it implements dlg_end_dlg. An idea I have what you could possibly be doing wrong is not
using the CSeq numbers that each party expects. I don't have the details at hand but
if either CSeq number is off your UAs won't accept the BYE request.
Cheers,
--Timo