TCP/IP problem : 'Still waiting for mbuf cluster'

Sylvain Prestavoine sylvain.prestavoine at memobox.fr
Wed Sep 13 10:16:34 UTC 2006


----- Original Message ----- 
From: "Ian Caddy" <ianc at goanna.iinet.net.au>

> This is just a guess, but what have you set your mbuf and mbuf cluster
> allocations up as?
>

At present, we use :
rtems_bsdnet_config.mbuf_bytecount=(64*1024) ;
rtems_bsdnet_config.mbuf_cluster_bytecount = (128 * 1024);
because ours board has only 1Mo of SRAM.
In my case, the performance is not important.
But, even slowly, it must working really.

> It looks like you are running out of room, when you increase your socket
> send buffer size.  Is there any scope for increasing your mbuf and mbuf
> cluster sizes?

I don't know, I'll see it.
But why the problem appear when the TCP client send one message window
update or
window is null ?
Why that disrupt the RTEMS stack IP ?
The flow control shouldn't have cause this effect, no ?

I don't understand it.

regards,

sorry for my verry bad english ;-)

Sylvain




More information about the users mailing list