lwip

Hi all,     I have a problem that I have seen lots or users straggling with, but without any real solution.   I am trying to send data in a loop. I have triad closing NAGLE  as follows:     // this should shut down the NAGLE algorithm   pcb->flags |= TF_NODELAY | TF_ACK_NOW;     I am calling tcp_output() on every tcp_write() but this does not help as well. I got ERR_MEM after the 20th something call to tcp_write()     I managed to find in one of the answers here that I should use a smaller window, meaning change the settings in lwipopts.h… So I did, and it was a bit better.   /* TCP Maximum segment size. */ //#define TCP_MSS                         1460 #define TCP_MSS                         512   /* TCP sender buffer space (bytes). */ //#define TCP_SND_BUF                     (3*TCP_MSS) #define TCP_SND_BUF                     (8*TCP_MSS)     Well the above helped and I was able to send more small packets but after 32 instead of 20 it again It stopped sending and I got ERR_MEM.   The most important thing is when I check wireshark I see that the stack is sending all my data in one Frame ???     Now can someone explain what is going on ?   For every call to tcp_write the enqueue mechanism is advancing  snd_queulen by one !   I see two problems here. If the stack eventually sends all these in one TCP packet why is snd_queulen incremented at all ??   Secondly if I use  pcb->flags |= TF_NODELAY  why is the stack adding all the data into one packet ??     Your input is more than welcomed.     BR, Noam.   ************************************************************************************ This footnote confirms that this email message has been scanned by PineApp Mail-SeCure for the presence of malicious code, vandals & computer viruses. ************************************************************************************ _______________________________________________ lwip-users mailing list [hidden email] http://lists./mailman/listinfo/lwip-users Remove Ads

Simon Goldschmidt

Reply | Threaded | More  

Mar 27, 2011; 9:22pm

RE: enqueing problem

1951 posts

On Sun, 2011-03-27 at 15:40 +0200, Noam weissman wrote: > > I do understand that on every call to tcp_write the snd_queuelen is > incremented but I do not understand why it is not > decremented when I call  tcp_output. Also take into consideration that > I have closed NGLE so it should have worked ?? I think there's a little confusion here.  Simon said "The only thing changed by disabling nagle is that tcp_output always sends everything that is enqueued." but this isn't quite right.  Disabling Nagle's algorithm stops us delaying transmission when there are small writes and packets are already in flight.  It doesn't stop us delaying transmission for other reasons, so tcp_output() won't always send everything that is enqueued.  For example, if we have run out of congestion window or receive window we will be unable to send immediately.  I think this will explain what you see, particularly as the initial value for the congestion window is quite small. A packet capture would be useful to see what is happening on the wire. Detail about how large your sends are would also help. Kieran _______________________________________________ lwip-users mailing list [hidden email] http://lists./mailman/listinfo/lwip-users Remove Ads