[Devel] handle_publish: Bad body format

[Bogdan-Andrei Iancu]

Hi Klaus,

yes, doing debug via debug logs is not all the time the best options. In 
this case, probably the best way will be to attach with GDB to a running 
process (after the leak occurred) and inspect the shm memory - but is 
difficult to do this over email :(

as I guess is not possible to get access to tour test system, could send 
me a description of how to reproduce your test scenario (sipp scripts & 
commands, relevant part of scripts). I will ask Anca to reproduce it on 
our systems....

I really want to solve this till the 1.2.1 release, otherwise I see no 
option than to delay the release until we figure out the problem.

Regards,
Bogdan

Klaus Darilion wrote:
> Hi Bogdan!
>
> When running in debug=4 mode, the problem is gone. Maybe the problem 
> happens only during high traffic, but in debug mode openser is really 
> slow.
>
> regards
> klaus
>
> Bogdan-Andrei Iancu wrote:
>> Klaus,
>>
>> can you get a post the entire log (including all the mem logs from 
>> runtime) in debug mode? It is important to see in what context the 
>> transactions were allocated.
>>
>> thanks,
>> bogdan
>>
>> Klaus Darilion wrote:
>>> Maybe this is memory related too?
>>>
>>> I've made a memdump (after 20 minutes idle) of this strange situation.
>>>
>>> is02:/home/klaus3000# openserctl fifo ps
>>> 200 OK
>>> Process::  ID=0 PID=6700 Type=attendant
>>> Process::  ID=1 PID=6705 Type=receiver child=0 sock= 1.2.3.4:6060
>>> Process::  ID=2 PID=6708 Type=timer
>>> Process::  ID=3 PID=6710 Type=tcp receiver
>>> Process::  ID=4 PID=6712 Type=tcp main process
>>>
>>>
>>> http://pernau.at/kd/openser/otherPCmemleak-sigusr1.openser-SVN.log
>>>
>>> regards
>>> klaus
>>>
>>> Klaus Darilion wrote:
>>>> Hi!
>>>>
>>>> I've tried with todays SVN 1.2. In the beginning was everything 
>>>> fine - but suddenly after ~150000 messages presence module reports:
>>>>
>>>> May 11 15:14:00 is02 /usr/sbin/openser[6705]: PRESENCE: 
>>>> generate_ETag: etag= a.1178878809.6705.311731 / 24
>>>> May 11 15:14:00 is02 /usr/sbin/openser[6705]: PRESENCE: 
>>>> handle_publish: Bad body format
>>>> May 11 15:14:00 is02 /usr/sbin/openser[6705]: PRESENCE: 
>>>> handle_publish: ERROR occured
>>>>
>>>> Following is the PUBLISH - which is generated by SIPP. Any hints?
>>>>
>>>> U 11.22.33.123:5061 -> 11.22.33.123:6060
>>>> PUBLISH sip:klaus at pernau.at SIP/2.0.
>>>> Via: SIP/2.0/UDP 11.22.33.123:5061;branch=z9hG4bK-7275-39-0.
>>>> Max-Forwards: 70.
>>>> Contact: sip:ua1 at 11.22.33.123:5061.
>>>> To: "klaus"<sip:klaus at pernau.at>.
>>>> From: "klaus"<sip:klaus at pernau.at>;tag=39.
>>>> Call-ID: 39-7275 at 11.22.33.123.
>>>> CSeq: 1 PUBLISH.
>>>> Expires: 60.
>>>> Allow: INVITE, ACK, CANCEL, OPTIONS, BYE, REFER, NOTIFY, MESSAGE, 
>>>> SUBSCRIBE, INFO.
>>>> Content-Type: application/pidf+xml.
>>>> User-Agent: sipp.
>>>> Event: presence.
>>>> Content-Length:  420.
>>>> .
>>>> <?xml version='1.0' encoding='UTF-8'?><presence 
>>>> xmlns='urn:ietf:params:xml:ns:pidf' 
>>>> xmlns:dm='urn:ietf:params:xml:ns:pidf:data-model' 
>>>> xmlns:rpid='urn:ietf:params:xml:ns:pidf:rpid' 
>>>> xmlns:c='urn:ietf:params:xml:ns:pidf:cipid' 
>>>> entity='sip:klaus at pernau.at'><tuple 
>>>> id='t532d494f'><status><basic>open</basic></status></tuple><dm:person 
>>>> id='p98169736'><rpid:activities><rpid:unknown/></rpid:activities></dm:person></presence>. 
>>>>
>>>>
>>>> #
>>>> U 11.22.33.123:6060 -> 11.22.33.123:5061
>>>> SIP/2.0 415 Unsupported media type.
>>>> Via: SIP/2.0/UDP 11.22.33.123:5061;branch=z9hG4bK-7275-39-0.
>>>> To: 
>>>> "klaus"<sip:klaus at pernau.at>;tag=0847ae4039e784441c60c0ce3b39ea30.8a59. 
>>>>
>>>> From: "klaus"<sip:klaus at pernau.at>;tag=39.
>>>> Call-ID: 39-7275 at 11.22.33.123.
>>>> CSeq: 1 PUBLISH.
>>>> Server: OpenSER (1.2.0-tls (i386/linux)).
>>>> Content-Length: 0.
>>>> .
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> Bogdan-Andrei Iancu wrote:
>>>>> Klaus,
>>>>>
>>>>> I'm not 100% sure if the mem leak can be just avoided only with 
>>>>> t_release(). Theoretically, it should ; but practically, there is 
>>>>> a difference between practice and theory :D....
>>>>> Just be sure, try to get the SVN version of 1.2 and run the tests 
>>>>> on it. This way we can be 100% sure !
>>>>>
>>>>> regards,
>>>>> bogdan
>>>>>
>>>>> Klaus Darilion wrote:
>>>>>>
>>>>>>
>>>>>> Bogdan-Andrei Iancu wrote:
>>>>>>> Hi Klaus,
>>>>>>>
>>>>>>> the fix was also backported to 1.2.
>>>>>>
>>>>>> Uups. I'missed that. But i still use my old binaries with 
>>>>>> t_release(). This should be fine too - isn't it?
>>>>>>
>>>>>> regards
>>>>>> klaus
>>>>>>
>>>>>>>
>>>>>>> regards,
>>>>>>> bogdan
>>>>>>>
>>>>>>> Klaus Darilion wrote:
>>>>>>>>
>>>>>>>>
>>>>>>>> Juha Heinanen wrote:
>>>>>>>>> Klaus Darilion writes:
>>>>>>>>>
>>>>>>>>>  > I still use openser 1.2, thus I have to use t_release()
>>>>>>>>>
>>>>>>>>> i have missed this: how do 1.2 and trunk differ regarding 
>>>>>>>>> t_release?
>>>>>>>>
>>>>>>>> With openser 1.2, if you use t_newtran() before 
>>>>>>>> handle_publish() you have to explicitely free the transaction 
>>>>>>>> with t_release().
>>>>>>>>
>>>>>>>> In trunk it is fixed and the transaction gets freed also if you 
>>>>>>>> omit t_release().
>>>>>>>>
>>>>>>>> regards
>>>>>>>> klaus