Re: [lwip-users] lwip_send hang (probably due to my allocator)

[Kieran Mansley]

On Fri, 2011-09-23 at 13:41 +0000, Stephen Cleary wrote:
> 
> One solution would be to limit the amount of data ftp passes to
> lwip_send. My TCP_SND_BUF is TCP_WND (== 4 * TCP_MSS, where TCP_MSS ==
> 1460), so it's only 5840 bytes, but I'm pretty sure lwip_send is
> allowing me to queue more data than that. Is there a procedure for
> detecting "how full" the send buffer is on a socket so that I can
> throttle in my FTP code? 

This would be my preferred solution.  lwIP will allow you to have
TCP_SND_BUF bytes in the send queue (i.e. not yet sent on the wire)
*and* TCP_WND bytes already sent waiting for an ACK from the other end.
There is another variable to control the length of the send buffer in
number of packets, but I think that is less relevant to your problem.

There isn't an official API to access the send queue length, but you
could hack something in for debugging.

A packet capture would show if you're right about the problem being that
the stack can't allocate packets to send an ACK.  I'm a bit dubious
about this.  Could it be that it couldn't allocate memory to post an
mbox message to the sockets API instead?  The LWIP_STATS code could also
throw light on where the problem is and so help focus your efforts.

Kieran