Re: [lwip-users] Case Of The Missing ACK

[davidm]

address@hidden wrote:
> Kieran Mansley wrote:
> > On Mon, 2009-09-07 at 14:58 +1000, address@hidden wrote:
> >  
> > > Perhaps others have encountered this and would offer comments and 
> > > perhaps comments could be offered in general.
> > >     
> >
> > That is weird.  It's hard to tell which end is lwIP and which is WinXP
> > in your packet capture, but at the point at which things go wrong I
> > notice the direction of the traffic changes.  Until then, 192.168.2.2 is
> > sending to 192.168.2.3, at that point ...2.3 starts sending to ...2.2
> >
> > Is there anything in the lwIP stats that might explain why packets are
> > dropped?  Could it be that your board just gets very busy (perhaps with
> > something like debug output?) and takes a very long time to respond?
> >
> > Kieran
> >
> >
> >
> > _______________________________________________
> > lwip-users mailing list
> > address@hidden
> > http://lists.nongnu.org/mailman/listinfo/lwip-users
> >
> >   
> I should have explained - 192.168.2.2 is lwIP. At line 32 of 
> Wireshark, a concatenated number of requests are sent which the MCU 
> responds individually, echoing the request prior to sending the 
> response data. This explains the multiple packets from lwIP to the 
> client (192.168.2.3).
>
> At line 103 the request to enter binary mode is sent with the request 
> echoed at line 105 and the acceptance response at line 106.
>
> I have attached a file which shows the stats just prior to entering 
> binary mode and just after returning to text mode. All seems to be ok 
> there.
>
> There should be no delay due to the MCU being busy (other than task 
> switching) as all that is done is to partially load and internal 
> buffer which once filled, is written to an SD card Flash sector.
>
> A question: When the debug output "tcp_recved: recveived 46 bytes, wnd 
> 2048 (0)" is delivered, should an ACK already have been sent by lwIP, 
> or does that occur later?
>
> At a later time I should put a debug breakpoint on the receipt of the 
> 46 byte packet and trace through the internals of lwIP.
>
> And by the way, thank you for your input. If nothing else I'm honing 
> my skills in the finer art of lwIP debugging.
>
> Regards - davidm
>
> .
> ------------------------------------------------------------------------
>
> _______________________________________________
> lwip-users mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/lwip-users
Quick note to put matters at rest. After debugging with time stamped 
host strings on my Red Suite/Eclipse IDE I am getting a fuller picture 
of the sequence of events which lay fault within my app, not lwIP. 
Simply put, in binary mode the necessary things for TCP maintenance such 
as fast/slow timer attendance, interrupt servicing was not being done!

davidm