[Serdev] STUN support in SER

[Vladimir Marek]

Alfred E. Heggestad wrote:
> hi
>
> I see that there is now support for STUN in SER (v0.10) - well done guys,
> that is a very useful feature..
>
>
> I have a couple of comments/questions:
>
>
> * I did a quick test running a STUN client against a running SER with 
> STUN
>   enabled, and I got back:
>
> 00000000: 0111 007c 2112 a442 bdbe 2838 6245 ed66  ...|!..B..(8bE.f
> 00000010: 89f7 90ee 0009 0076 0000 0400 5468 6520  .......v....The
> 00000020: 7265 7175 6573 7420 7761 7320 6d61 6c66  request was malf
> 00000030: 6f72 6d65 642e 2054 6865 2063 6c69 656e  ormed. The clien
> 00000040: 7420 7368 6f75 6c64 206e 6f74 2072 6574  t should not ret
> 00000050: 7279 2074 6865 2072 6571 7565 7374 2077  ry the request w
> 00000060: 6974 686f 7574 206d 6f64 6966 6963 6174  ithout modificat
> 00000070: 696f 6e20 6672 6f6d 2074 6865 2070 7265  ion from the pre
> 00000080: 7669 6f75 7320 6174 7465 6d70 742e 0000  vious attempt...
> 00000090: 8022 003d 5365 7276 6572 3a20 5369 7020  .".=Server: Sip
> 000000a0: 4558 7072 6573 7320 726f 7574 6572 2028  EXpress router (
> 000000b0: 302e 3130 2e39 392d 6465 7636 332d 746c  0.10.99-dev63-tl
> 000000c0: 7320 2878 3836 5f36 342f 6c69 6e75 7829  s (x86_64/linux)
> 000000d0: 2900 0000 0023 0014 91ad 2a16 387e 3e6c  )....#....*.8~>l
> 000000e0: 9e22 0c86 eed4 4f92 2ae6 0180            ."....O.*...
>
>
> according to this packet the length of the STUN message (excluding
> header) should be 124 bytes, however the length is actually 216 bytes.
> Is there perhaps a calculation error of the length field?
>

Yes, it's a bug. The problem is that it was not clear to me if 
fingerprint should be counted in message length or not. The draft 04 say 
nothing about it and when I asked in a behave mailing list, nobody gave 
me the answer ... I decide it not to include it but If I'm thinking 
about it right know I would say, that it should be a part of the message 
length. Finger print is a 24 bytes. The rest 68 missing bytes is the 
attribute SERVER that I forgot to count (the total counting of body 
length is before adding SERVER attribute ...). So we have 124B of 
message + 24B of finger print + 68B for attribute SERVER ... that's the 
216B. I'll fix it ASAP.

Vladimir

>
> * I see that SER's STUN implementation is expecting a mandatory 
> FINGERPRINT
>   headerfield (SHA1) as of rfc3489-bis04 - this has now changed to 
> optional
>   and CRC32 in rfc3489-bis05 - do you have any plans of upgrading your 
> code
>   to support this?
>
>
> * Do you also support STUN over TCP ?
>
>
> thanks.
>
>
> /alfred
>
> _______________________________________________
> Serdev mailing list
> Serdev at lists.iptel.org
> http://lists.iptel.org/mailman/listinfo/serdev