[asterisk-dev] [Code Review] 4384: res_pjsip_outbound_publish: eventually crashes when no response is ever received

[Matt Jordan]

> On Jan. 29, 2015, 10:14 a.m., Joshua Colp wrote:
> > branches/13/res/res_pjsip_outbound_publish.c, line 927
> > <https://reviewboard.asterisk.org/r/4384/diff/2/?file=71172#file71172line927>
> >
> >     So...
> >     
> >     This may actually cause publish state to be reflected incorrectly on the remote side if you remove it from the head of the queue. Another publish for the same thing can be queued afterwards. If you move this to the tail of the queue then it will be out of order.
> >     
> >     As well - wouldn't this cause to the queue to potentially grow uncontrollably?
> 
> Kevin Harwell wrote:
>     Aaah yes things would be out of order. I see now and if I leave the current one in the queue and new ones are added then things would continue to grow.
>     
>     Hmm, what do you think about this: if the current publish receives no response check to see if it is the only one in the queue. If it is then leave it otherwise drop it allowing the the next one to try?
>     
>
> 
> Joshua Colp wrote:
>     Hum. Within the scope of the Asterisk publishing stuff that would mean that some updates may get lost. I don't think there's a really nice way to overcome that.

I think if we run into that problem, then there's probably bigger problems then merely reflecting device state. You'd have to have a very unresponsive system (it's down), at which point when it comes back, it'd want to resynch the state anyway.


- Matt


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/4384/#review14356
-----------------------------------------------------------


On Jan. 29, 2015, 12:52 p.m., Kevin Harwell wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/4384/
> -----------------------------------------------------------
> 
> (Updated Jan. 29, 2015, 12:52 p.m.)
> 
> 
> Review request for Asterisk Developers.
> 
> 
> Bugs: ASTERISK-24635
>     https://issues.asterisk.org/jira/browse/ASTERISK-24635
> 
> 
> Repository: Asterisk
> 
> 
> Description
> -------
> 
> When Asterisk attempts to send SIP outbound publish information and no response is ever received (no 200 okay, 412, 423) the system eventually crashes. A response is never received because the system Asterisk is attempting to send publish information to is not available. The underlying pjsip framework attempts to send publish information. After several attempts it calls back into the Asterisk outbound publish code. At this point if the "client->queue" is empty Asterisk attempts to schedule a refresh which utilizes "rdata" and since no response was received the given "rdata" struture is NULL. Attempting to dereference a NULL object of course results in a crash.
> 
> This patch re-queues the current message that has not received a response yet (has no "rdata"), thus removing the possibility of the queue being empty and having no "rdata" available. Consequently, in this scenario, the publish refresh is not called and the crash is avoided.
> 
> 
> Diffs
> -----
> 
>   branches/13/res/res_pjsip_outbound_publish.c 431402 
> 
> Diff: https://reviewboard.asterisk.org/r/4384/diff/
> 
> 
> Testing
> -------
> 
> First duplicated the problem by attempting to publish to a non existent system (after a bit Asterisk crashed). After applying the patch using the same setup Asterisk no longer crashed. Also ran the current set of outbound publish testsuite tests to make sure those still passed.
> 
> 
> Thanks,
> 
> Kevin Harwell
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20150129/562c6aba/attachment.html>