[lwip-users] is this a reentrancy problem with LWIP ver 1.3.1?

[shogun]

Can someone tell me if this is a reentrancy issue?  If I send data too fast
on a socket and the stack is still sending the data and I make another call
to send more data, will that cause a problem?

I am using a Stellaris 9b92 CPU running 80 MHz using lwip to send TCPIP data
on a server socket.  The data I am sending  is a block of data 2894 bytes
long.  I am using Code Composer Studio without an OS.  I have Nagel disabled
on both  ends  and I am sending the data to a unloaded PC that is running
wireshark and is just receiving the data and throwing it away.  When I look
at the wireshark data I can see the data is sent in three packets 1260, 1260
and then 324 bytes.  Wireshart shows the following:

Time = 0  first packet is sent 1260 bytes long
Time 64uS PC ACK's
Time 110uS, second 1260 packet sent
Time 122uS PC ACK's
Time 177uS,  send 374 bytes
Time 187uS, PC ACK's

The software then will send the data again 200mS later so there is quite a
bit of dead time before the next send.  The total time it takes to send the
data is ~ 200uS every 200mS.

This works well till I try to send the data with less delay between the
sends of the data.  With the delay at about 100mS between the sends, it will
still work but I think it may be marginal.  When I try to go much faster, I
see the sends going out about every 90mS then after about 7 sends of the
data at the rate of a send every 90mS, I start getting the out of memory
error code on the send.  I know it takes time for the send to process that
data and break it into packets and calculate a checksum etc and at some
point things will back up.  The data rates where I am seeing problems seems
low where I think something may be wrong in what I am doing.  Does anyone
have advice and is this a reentrancy problem?  I could provide the wireshark
trace is that would help.

Thanks,
db