[asterisk-dev] [Code Review] 3982: res_rtp_asterisk: Fix a slew of TURN issues.

[Matt Jordan]

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/3982/#review13299
-----------------------------------------------------------

Ship it!



/branches/13/res/res_rtp_asterisk.c
<https://reviewboard.asterisk.org/r/3982/#comment23775>

    Blob cleanup



/branches/13/res/res_rtp_asterisk.c
<https://reviewboard.asterisk.org/r/3982/#comment23776>

    Slap an assert here, as something has gone wrong if we get a transport type that we don't support passed to this function



/branches/13/res/res_rtp_asterisk.c
<https://reviewboard.asterisk.org/r/3982/#comment23777>

    This probably needs a WARNING message, as PJPROJECT not creating something won't get caught by our allocation routines


- Matt Jordan


On Sept. 10, 2014, 6:24 a.m., Joshua Colp wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/3982/
> -----------------------------------------------------------
> 
> (Updated Sept. 10, 2014, 6:24 a.m.)
> 
> 
> Review request for Asterisk Developers.
> 
> 
> Bugs: ASTERISK-23577 and ASTERISK-23634
>     https://issues.asterisk.org/jira/browse/ASTERISK-23577
>     https://issues.asterisk.org/jira/browse/ASTERISK-23634
> 
> 
> Repository: Asterisk
> 
> 
> Description
> -------
> 
> The TURN support in res_rtp_asterisk has been exercised by only a few, which has uncovered a slew of issues.
> 
> 1. The number of file descriptors that an ioqueue instance in pjlib can support is a fixed number. Exceeding this causes an assertion. The code will now dynamically create/terminate threads as needed to ensure that this limit is not exceeded on ioqueue instances.
> 2. The API did not allow users of the TURN code to explicitly request a TURN session with details. This has been added.
> 3. The ICE code has a fixed size array of 4 for transports. As our transport identifiers started at 1 we were exceeding this causing an assertion. Our identifiers now start at 0.
> 4. The TURN client did not set up client binding causing needless bandwidth usage. Upon ICE completion if TURN is used channel binding is now established.
> 5. The code will no longer update address information on every sent packet. The remote address is now updated only once upon ICE completion to the target address.
> 6. When relaying was in use STUN traffic was getting looped back to res_rtp_asterisk due to it being handled on the normal socket. It is now handled in the TURN session callback instead.
> 7. Logging when a TURN relay is in use now uses the IP address that the TURN relay is sending/receiving to/from on our behalf.
> 8. Synchronization between the TURN session and the RTP instance now ensures that the session has been fully created or fully destroyed before continuing.
> 9. Some cleanup has occurred when tearing down the pj environment.
> 
> 
> Diffs
> -----
> 
>   /branches/13/res/res_rtp_asterisk.c 422898 
>   /branches/13/include/asterisk/rtp_engine.h 422898 
> 
> Diff: https://reviewboard.asterisk.org/r/3982/diff/
> 
> 
> Testing
> -------
> 
> Sabotaged the code so the only candidates I sent were of the relay type. Confirmed bidirectional media flow using the TURN relay.
> 
> 
> Thanks,
> 
> Joshua Colp
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20140915/4a20b851/attachment.html>