[Serusers] Race Condition on CANCEL?
Greger V. Teigre
greger at teigre.com
Wed May 16 10:03:42 CEST 2007
Gustavo Passos Tourinho wrote:
> Last question, I promise :)
say, for now, and I may believe you ;-)
>
> A CANCEL request is hop-by-hop. If the proxy receives a cancel (using
> his server transaction) it should response and issue a CANCEL request
> (using his client transaction) to next hop.
Yes and no. There's a difference between a locally generated CANCEL and
an externally generated one. SER will generate CANCEL where required in
the RFC, but CANCELs coming from downstream needs to be handled by
ser.cfg and relayed. The most robust method in 0.9 and 2.0 (not future
2.1) is to let CANCELs with Route be handled by loose_route and then let
any CANCELs without Route be handled like INVITE (i.e. rewrite request uri)
> So, if it receives a 200 OK after sent a CANCEL, shouldnt it send a
> BYE they self? Insted of relying it to caller?
No, SER will NEVER generate a BYE. This can only be done by a UAC.
According to the RFC, the 200 OK is most likely due to the callee UA not
being able to cancel the call before it was picked up, thus a dialog is
created. In order to tear down the dialog, the caller has to send a BYE
in response even though it CANCELed the call previsouly. How you account
is another story, the same happens in PSTN. I believe many telcos bill
those calls.
>
> As I promise, last question and thank you very much for your help.
Sure, no probs.
g-)
>
> Regards,
>
> Greger V. Teigre escreveu:
>> Inline.
>>
>> Gustavo Passos Tourinho wrote:
>>> Hello All,
>>>
>>> I Guess that the Proxy relays 200 OK (INVITE) because it didnt
>>> destroyed the transaction correctely affter it was canceled, right?
>>>
>> No, the proxy MUST relay the 200 OK. The caller UA (UAC) should then
>> send an immediate BYE. The callee UA (UAS) wrongly(?) sends a 200 OK
>> after it claims to cancel the call. However, this may happen due to a
>> race condition that is unavoidable (RFC3261)
>>> I was looking for a possible configuration bug (ie. relaying 200 OK
>>> statelessly), but there is no such think. All relays are done by
>>> t_relay () function.
>>>
>> Your problem is that the caller UA does not immediately send a BYE.
>>> Another question. If a cancel request is not replyed with a response
>>> after some time, the UA should try to retransmit it? After reading
>>> the RFC 3 times I just so that the CANCEL cannot be challenge.
>>>
>> That's correct, a CANCEL cannot be challenged (as in authentication).
>> g-)
>>> Well, thank you very much for your help.
>>>
>>> Regards,
>>> Gustavo
>>>
>>> Greger V. Teigre escreveu:
>>> > As I said, I would start finding out why the UA sends 200 OK after
>>> it has sent > canceling.
>>> > There is no sign of any bugs in CANCEL transaction. As I
>>> explained, the CVS > trunk changes are related to how CANCELs are
>>> handled. Your CANCEL reaches the > UA ok.
>>> > g-)
>>> >
>>> > Gustavo Passos Tourinho wrote:
>>> >> Hi,
>>> >>
>>> >> So, is there a BUG in CANCEL transaction?
>>> >>
>>> >> How can I confirm this information?
>>> >>
>>> >> Thanks
>>> >>
>>> >> Atle Samuelsen escreveu:
>>> >>> Hi,
>>> >>>
>>> >>> I might be wrong, but I think Andrei added some code in CVS head
>>> a few
>>> >>> days back that addresses this issue. >>>
>>> >>> -A
>>> >>>
>>> >>> * Gustavo Passos Tourinho <gustavo.passos.tourinho at gmail.com>
>>> [070511 13:57]:
>>> >>> >>>> Thanks for your reply.
>>> >>>>
>>> >>>> Yes, I have this problem right now. The problem is that when
>>> the proxy receives 200 OK (for INVITE, confirmed by CSeq), insted of
>>> 200 Cancelling, it issue >>>> an RADIUS request for billing.
>>> >>>>
>>> >>>> So, I will have an "Start-Invite" (200 OK), but will have not a
>>> "Stop-Bye" because the BYE message will not be generated.
>>> >>>>
>>> >>>> How can I ensure that the proxy will not forward 200 OK for
>>> INVITE? I mean, if it is a transaction statefull and the transaction
>>> doesnt existis, why it is >>>> stills forwarding the message? Is
>>> there any thing that I can do to prevent this kind of situation to
>>> occour?
>>> >>>>
>>> >>>> Thanks again.
>>> >>>>
>>> >>>> Regards,
>>> >>>> Gustavo
>>> >>>>
>>> >>>> Greger V. Teigre escreveu:
>>> >>>> >>>>> No, the UAS (U2) shall answer with 200 OK to confirm
>>> that the call has been canceled and yes, it should be sent to U1.
>>> >>>>> Do you have an actual experienced problem or was the 200 OK
>>> the problem?
>>> >>>>> g-)
>>> >>>>>
>>> >>>>> Gustavo Passos Tourinho wrote:
>>> >>>>> >>>>>> Hello,
>>> >>>>>>
>>> >>>>>> Im having some problems with cancelled calls. This is the
>>> scenario:
>>> >>>>>>
>>> >>>>>> U1
>>> Proxy U2
>>> >>>>>>
>>> >>>>>> INVITE -->>>
>>> <<--- 100 Trying
>>> >>>>>> INVITE -->>>
>>> >>>>>>
>>> <<--- 100 Trying
>>> >>>>>> <<--- 100 Trying CANCEL
>>> ->>> <<-- 200 Cancelling
>>> >>>>>> CANCEL
>>> ->> <<-- 180 Ringing
>>> >>>>>> <<-- 487 Cancelled
>>> >>>>>> <<-- 180 Ringing
>>> >>>>>>
>>> <<-- 200 OK
>>> >>>>>> (Wrong??)
>>> >>>>>> <<-- 200 OK
>>> >>>>>>
>>> >>>>>>
>>> >>>>>> My problem is that after some time waiting for "ringing", the
>>> user cancel the call. Even that proxy responses "487" it still
>>> forward the late 200 OK.
>>> >>>>>>
>>> >>>>>> Should it forward? I guess not because the transaction was
>>> destroyed, right?
>>> >>>>>>
>>> >>>>>> Can it be a configuration problem on my ser.cfg ou it can be
>>> in t_relay implementation?
>>> >>>>>>
>>> >>>>>> Thanks in advanced.
>>> >>>>>> Regards,
>>> >>>>>>
>>> >>>>>> Gustavo
>>> _______________________________________________
>>> >>>>>> Serusers mailing list
>>> >>>>>> Serusers at lists.iptel.org
>>> >>>>>> http://lists.iptel.org/mailman/listinfo/serusers
>>> >>>>>>
>>> >>>>>>
>>> >>>>>> >>>> _______________________________________________
>>> >>>> Serusers mailing list
>>> >>>> Serusers at lists.iptel.org
>>> >>>> http://lists.iptel.org/mailman/listinfo/serusers
>>> >>>> >>>
>>> >>> >>
>>>
>>>
>>
>
>
More information about the sr-users
mailing list