[asterisk-dev] IAX hardphone

[Bill Shaw]

On 2/15/2011 12:11 PM, Kevin P. Fleming wrote:
> On 02/15/2011 10:58 AM, Mihai Balea wrote:
>>
>> On Feb 15, 2011, at 11:21 AM, Bill Shaw wrote:
>>
>>> Just thought I'd confirm that the UDP checksum being incorrect was 
>>> my problem (thank you Mihai),  and also that Wireshark incorrectly 
>>> handles the UDP checksum.  I've corrected my checksum routine and 
>>> deleted Wireshark - fallen back to Ethereal.
>>
>> Glad I could help.
>> Weird that your copy of Wireshark did not flag this - I'm using 
>> Wireshark on Mac OSX (v 1.4.3) and it pointed me to it. And, after 
>> all, Wireshark is the new Ethereal, they renamed the project a while 
>> back.
>> Maybe it was just a buggy version or maybe just some settings...
>
> In addition, there's really no such thing as a 'disabled' UDP 
> checksum; UDP checksums are not optional in the protocol, although 
> some endpoints can be configured to not compute/check them. Even in 
> that case, the checksum field will still be present in the UDP header, 
> and will usually have a constant value.
>
> It is possible that Wireshark senses a particularly common constant 
> value (like all zero) and reports that the endpoint must be configured 
> with UDP checksum generation/checking 'disabled'... which I suppose 
> would be better than it reporting every single packet with a UDP 
> checksum error for such an endpoint.
>
Actually,  Wireshark reported checksums disabled when I had values in 
the field.  The values changed depending on the data so they were not 
some constant value.  When I hard coded 0s in the checksum field for 
testing,  wireshark, ethereal and the linux box all accepted the packets 
as valid,  checksums were in fact disabled.

I think I'd buy Jeff's explanation - Wireshark disables checksums for 
some odd reason.  I don't buy the entire thing though,  since it was 
disabling checksums on incoming packets where the checksums should have 
been present and valid,  not just outgoing packets where some external 
processor might be computing the checksum.

Regardless,  I'll just write it off to some networking weirdness and 
avoid Wireshark in the future.

BTW - Wireshark 1.4.1 on Win XP fwiw

Thanks again

Bill