[asterisk-dev] [Code Review] 4505: Asterisk 13 PJSIP sends RTP packets in wrong byte order on Intel platform when using slin codec

Matt Jordan reviewboard at asterisk.org
Mon Mar 23 09:56:49 CDT 2015



> On March 17, 2015, 11:57 a.m., Matt Jordan wrote:
> > /tags/13.2.0/include/asterisk/codec.h, lines 77-80
> > <https://reviewboard.asterisk.org/r/4505/diff/1/?file=72588#file72588line77>
> >
> >     I don't think you can trust that the codec will know its endianness. Looking at the resample code, I don't _think_ it actually determines the endianness of its encoding/decoding, and instead relies on the underlying machine to make that determination. As such, I don't think this should be a property on the codec structure.
> 
> Frankie Chin wrote:
>     Matt, I actually followed the implementation in Asterisk 12.8.1 where the AST_SMOOTHER_FLAG_BE was defined for all the SLIN codecs in main/format.c under the format_list_init() method. Do you mean this implementation back in 12.8.1 was inappropriate? FYI, slin codec used to work fine in Asterisk 12.8.1 for our application.
> 
> Matt Jordan wrote:
>     Apologies; you are absolutely correct about that.
>     
>     I'll need a day or two to think about whether or not this is the right way to handle passing smoother flags in. A part of me dislikes having an explicit BE byte order integer, but I'm also not a giant fan of storing a bunch of smoother properties directly on the codec. There may just not be a better place for it.

After looking at 12 some more, I think handling this in the same manner as Asterisk 12 is probably the best way forward. That would mean adding a 'smoother_flags' option 

struct ast_codec {
    ...

    unsigned int smooth;

    unsigned int smoother_flags;

}

That would change the API call from ast_format_smoothed_with_be to ast_format_get_smoother_flags (or something similar). That way, any additional smoother flags can be obtained without needing additional APIs added.


- Matt


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/4505/#review14719
-----------------------------------------------------------


On March 16, 2015, 10:36 p.m., Frankie Chin wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/4505/
> -----------------------------------------------------------
> 
> (Updated March 16, 2015, 10:36 p.m.)
> 
> 
> Review request for Asterisk Developers.
> 
> 
> Bugs: ASTERISK-24858
>     https://issues.asterisk.org/jira/browse/ASTERISK-24858
> 
> 
> Repository: Asterisk
> 
> 
> Description
> -------
> 
> In Asterisk 13.2.0 when SLIN codec is used in two Asterisk servers registered to one another via PJSIP, the RTP payload is sent in the wrong byte order. The patch addresses the following based on the correct behavior in Asterisk 12.8.1:
> 1) Save ptime = 20 as the framing in the ast_rtp_codecs structure when creating outgoing SDP packet (res_pjsip_sdp_rtp.c)
> 2) Do not copy the framing when copying the payload (rtp_engine.c)
> 3) Introduce the new "smoother_be" flagin the ast_codec structure. Set this flag = 1 for all the SLIN codecs (codec_builtin.c).
> 4) Check for this "smoother_be" flag before using the smoother on the data (res_rtp_asterisk.c)
> 
> 
> Diffs
> -----
> 
>   /tags/13.2.0/res/res_rtp_asterisk.c 433002 
>   /tags/13.2.0/res/res_pjsip_sdp_rtp.c 433002 
>   /tags/13.2.0/main/rtp_engine.c 433002 
>   /tags/13.2.0/main/format.c 433002 
>   /tags/13.2.0/main/codec_builtin.c 433002 
>   /tags/13.2.0/include/asterisk/format.h 433002 
>   /tags/13.2.0/include/asterisk/codec.h 433002 
> 
> Diff: https://reviewboard.asterisk.org/r/4505/diff/
> 
> 
> Testing
> -------
> 
> The patch was tested using the scenario described in ASTERISK-24858
> 
> 
> Thanks,
> 
> Frankie Chin
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20150323/25ddef65/attachment-0001.html>


More information about the asterisk-dev mailing list