Wireshark-dev: Re: [Wireshark-dev] G.722 and G.726 decoders for Wireshark
Date Index · Thread Index · Other Months · All Mailing Lists
Date Prev · Date Next · Thread Prev · Thread Next
From: Jaap Keuter <jaap.keuter@xxxxxxxxx>
Date: Wed, 26 Jan 2011 17:28:41 +0100
Hi,

Well, Wireshark should pick up on SDP negotiations for your RTP stream,
and set the correct payload type interpretation. I would have to look at 
the specific capture file to see what's happening.

The fact that "The G.726 decoder function returns twice the number of
bytes fed to it as output buffer size" is indeed to cause of your problems. 
You have decoded from 4 bit (ADPCM) samples to 8 bit (companded PCM)
samples, while the player expects 16 bit (linear PCM) samples.
So your output size should be four times your input size.

Thanks,
Jaap

On Wed, 26 Jan 2011 09:23:58 +0100, Dietfrid Mali wrote:
Yeah, I wasn't expressing this quite right. What I basically wanted to 
say is that Wireshark
doesn't assign a payload type using the SDP but just forwards the
payload type
given in the RTP packets (102, which Wireshark wraps into a define that 
basically says
"undefined packet type").


Date: Wed, 26 Jan 2011 00:09:50 +0100
From: jaap.keuter@xxxxxxxxx
To: wireshark-dev@xxxxxxxxxxxxx
Subject: Re: [Wireshark-dev] G.722 and G.726 decoders for Wireshark

On 01/25/2011 05:48 PM, Dietfrid Mali wrote:
> The problem with e.g. G.726 is that Wireshark gives those packets

RTP

> type 102 which afaik is an error code ("unknown encoding").

No, that's your RTP endpoint configured to label these as such. RFC

3550 says:
"A profile MAY specify a default static mapping of payload type codes 
to payload

formats. Additional payload type codes MAY be defined dynamically

through

non-RTP means (see Section 3)."
RFC 1890/RFC 3551 defines the "RTP Profile for Audio and Video

Conferences with

Minimal Control", which lists several static payload types. The old

RFC lists

G.721 (aka G.726-32), while the new one dropped that one and added

references to

G.726 at various bit rate, with a dynamic payload type.
RFC 3550 says in Section 3: "Non-RTP means: Protocols and mechanisms

that may be
needed in addition to RTP to provide a usable service. In particular, 
..., and

define dynamic mappings between RTP payload type values and the

payload formats
they represent for formats that do not have a predefined payload type 
value."

with reference to Session Description Protocol (SDP)

So, payload type 102 is a dynamic payload type which has to be given

meaning
(through SDP for instance) within the session. In your case Wireshark 
didn't

pick that up from the trace, hence cannot give you the proper

interpretation of

that payload type within that session.

> I would need to know where and how Wireshark maps dynamic payload

types

> (negotiated via SDP) to internal static ones. Above that RFC3551

notes that
> static G.726 payload types are obsolete, and afaik there aren't even > (obsolete) static payload types for all G.726 variants, so Wireshark 
> would need to
> take care of that by using some (more or less arbitrary) internal

static

> type numbers.
Yep, that is done by the SDP dissector. It tries to interpret the SDP 
offer

(should be the answer, but that a whole other story) and create

conversations
(see doc/README.developer, section 2.2) for the RTP dissector, feeding 
it

dynamic payload type information it has learned from the media

attributes.


The RTP dissector does then the heavy lifting on the RTP packets,

based on the

information feed in by the SDP dissector.

> I will do my best to provide a patch once I have fully integrated

all

> codecs (currently only G.726-32 has been implemented as proof of
> concept, but since
> this is working adding more is no big deal).

Just one to get started is fine. Does it integrate into codecs/

directory

besides G711a and G711u (and G729 and G723, if you have them)?

> Getting G.726 to work was a bit of a pain btw because of the weird

frame

> sync calculation in rtp_player.c::play_channels() as this function

seems to

> assume 1:1 relationships of decoder input and output stream sizes

and

> thus simply halves the decoder output batch sizes to determine

whether

> frames
> are properly sync'd. This doesn't work for compressed audio. To
> compensate, my decodeG726_32() function doubles the number of bytes 
returned

> (as it has a 1:2 relationship of input and output buffer sizes).

Before

> it did that, lots of silence frames were inserted and half of the

audio

> data was
> dropped by the player.
The G.726 decoder function returns twice the number of bytes fed to it 
as
output buffer size. Just returning that number leads to incorrect audio 
playback. I had to double that number once more to make it work. This
has something
to do with how rtp_player() determines whether it has received enough
packets
for the given time frame and that it inserts silence frames when it
thinks it hasn't.
There is no (inline) documentation of what rtp_player() expects. Maybe 
this is
described in some Wireshark programming or API documentation?


I'm no sure if I understand you correctly. Working with these decode

functions

there is an input buffer with its length as input, and two output

parameters,
being the output buffer and it a pointer to store its size. This size 
of the
output buffer has to be set, by the decoder, to the number of samples 
in output

buffer. That should be enough, see for instance

rtp_player.c:decode_rtp_packet()

the handling of G.279 and G.723.
Be aware that you have to store 16 bit linear samples in the output

buffer,

maybe that's your factor 2?

Thanks,
Jaap


>
> Dietfrid
>
> > From: jaap.keuter@xxxxxxxxx
> > Date: Tue, 25 Jan 2011 16:54:25 +0100
> > To: wireshark-dev@xxxxxxxxxxxxx
> > Subject: Re: [Wireshark-dev] G.722 and G.726 decoders for

Wireshark

> >
> > Hi,
> >
> > That would be interesting. Can you put the code in a patch on

bugzilla?

> >
> > Can't work on it right now, but would be nice to have.
> >
> > btw: their are already static RTP types assigned for both codecs. 
The

> dynamic types should come in through protocols like SDP, or a

dissector

> preference.
> >
> > Thanks,
> > Jaap
> >
> > Send from my iPhone
> >
> > On 25 jan. 2011, at 16:07, Dietfrid Mali wrote:
> >
> > > Hi,
> > >
> > > using spandsp, I have added G.722 and G.726 decoders to

Wireshark.

> > >
> > > Currently this is a bit of a hack job, particularly regarding
> inclusion of the spandsp lib, and I could need a bit help to

properly

> integrate it into Wireshark's automake hell (configure.in).
> > >
> > > There also isn't a proper Wireshark signature for that RTP type 
(I

> am simply reacting to RTP type 102, which actually is an error

code), so

> some help getting this straight and introducing proper codec types

would

> be appreciated, too.
> > >