[wt4hq] RX Buffer Overflow

Laurent HAAS - F6FVY f6fvy at free.fr
Wed Jul 9 20:18:13 CEST 2008


Hi Babsi

Dieter Kritzer a écrit :

> The overflow happens when the size of the already allocated RX Buffer and
> the 
> size of the new received data exceed 10000 characters.
> 
> I guess this happens when a lot of stations send lots of QSOs within short 
> time. I found the error in the logfile of the Windows server, but not yet on
> 
> the Linux server. Maybe my Linux server is faster in transmitting the data
> to 
> the other stations and deallocating memory, don't know yet. We'll be testing
> 
> the tunnel during the next two days.

Actually, from what I understand from the code (Olivier F5MZN wrote it) 
the RX buffer is only supposed to receive data and process them, ready 
to be sent to the other "peers". If you have overflows, it would mean 
that a bunch of data is received before *any* is sent back.

FYI, we currently have around 10 connected peers here in test, and 
haven't got this warning yet.

> Unfortunately my C knowledge is not good enough to be sure about what would 
> happen if I set the RX_BUFFER_SIZE_MAX to 20000. The server the tunnel is 
> running on has enough memory, I think. Memory allocation should not be the 
> problem. Should I give it a try if it happens on the Linux server too?

There is no risk at all trying to modify this value. You just need to 
recompile, of course. I'm not very familiar with Linux but a gcc command 
should do the trick (If you need help on this, I'm sure Mario DL5MLO or 
another Linux guru on this list should be able to help).

73

Larry - F6FVY


More information about the Wt4hq mailing list