Re: [lwip-users] tcp_enqueue fails at more than two tcp_enqueue's

[Jimi Simpson]

Hi Kieran,

It's getting a little late and after a full day off debugging with this it 
has taken its toll :-( I did however find the problem - finally, and it's in 
tcp_receive. I initially hard coded the following line (immediate below), 
then debugged a little to establish that the following is necessary (one but 
next) - it's all preprocessor stuff:

   } else if (((ackno - pcb->lastack+1) >= 0) && ((ackno - pcb->snd_nxt) <= 
0)) {
//    } else if (TCP_SEQ_BETWEEN(ackno, pcb->lastack+1, pcb->snd_nxt)){

inside tcp.h I amended as follows:

#define TCP_SEQ_LT(a,b)     ((s32_t)((a)-(b)) < 0) -- no change
#define TCP_SEQ_LEQ(a,b)    ((s32_t)((a)-(b)) <= 0) -- no change
#define TCP_SEQ_GT(a,b)     ((s32_t)((a)-(b)) > 0) -- no change
#define TCP_SEQ_GEQ(a,b)    (((s32_t)(a)-(b)) >= 0) -- this line, I have 
shifted the cast and this works now

I will endeavor to check a little further tomorrow, but this is now working 
well :-) - not sure if this is a core bug but at least I hope this lwip-user 
thread may be informative to others...

Best regards, Jimi

----- Original Message ----- 
From: "Kieran Mansley" <address@hidden>
To: "Mailing list for lwIP users" <address@hidden>
Sent: Tuesday, May 04, 2010 2:51 PM
Subject: Re: [lwip-users] tcp_enqueue fails at more than two tcp_enqueue's


> On Tue, 2010-05-04 at 14:15 +0100, Jimi Simpson wrote:
> > Hi Kieran,
> >
> > Thank you for the response.
> >
> > I have 4 threads running:
>
> That sounds fine, and your lwipopts settings look OK too.
>
> > Sorry for the lengthy email and lots of information, but your help is
> > greatly appreciated.
>
> It looks from the debug output and packet capture as though lwIP is
> sending some data, the other side is sending an ACK, but the stack is
> not receiving or processing the ACK.  It therefore retransmits the old
> packet, which the other side again ACKs, but again lwIP doesn't receive
> or process the ACK and so it gets stuck in this cycle.
>
> It would be worth turning on the LWIP_STATS code to see if there is some
> resource that is running dry that would account for the dropped ACKs.
> The debug log does suggest that tcp_input() is being called, but the
> received packets aren't (after the first couple) recognised as
> acknowledging data we have outstanding.  I would look through the code
> and add more debugging to work out why these packets get ignored.
>
> Kieran
>
>
>
> _______________________________________________
> lwip-users mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/lwip-users