[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lwip-users] How to force dup ACK

From: yueyue papa
Subject: Re: [lwip-users] How to force dup ACK
Date: Thu, 15 Apr 2010 18:06:10 +0800

>From the wireshark,

item 13, is a 1460 data resend,

but in terminal log, we confirmed the data is get, and ACK is also
send to remote.

So I think we just lost ACK.

The problem is easy to meet in  larger TSS 1460, but with 512 TSS , I
could see the send/receive is very stable.

But smaller TSS give a lower performance.

If I increase the TSS, the resend cause the complete flow is low and not stable.

I tried to make duplicate ACK sent to remote, so as to solve my current problem.

I make resend ACK in different code, but still could not see the
duplicate ACK in the server, so I think my changing is not correct.

Kieran, could you guild me to add code for force send duplicate ACK


On Thu, Apr 15, 2010 at 5:49 PM, yueyue papa <address@hidden> wrote:
> Kieran,
> thank you to review.
> On Thu, Apr 15, 2010 at 5:44 PM, Kieran Mansley <address@hidden> wrote:
>> On Thu, 2010-04-15 at 15:32 +0800, yueyue papa wrote:
>>> Hi All,
>>> I use lwIP in GPRS environment.  Wireshark is setup in the server.  I
>>> find these information:
>>> lwip  device  <--------------------> server (Wireshark running there)
>>> if  TSS = 512, the TCP stream is working very stable.
>>> if TSS = 1024, 1460, there is always TCP ACK lost in server. The
>>> resend is always happen.
>>> The resend cause the performance down.
>> Are you sure it is ACKs that are being lost rather than packets with
>> payload?  Could you post an example wireshark capture to show this?
>>> Is there a solution I could configure lwIP make duplicate ACK send, so
>>> as to decrease the ACK lost possibility?
>> Not that I know of.
>> Kieran
>> _______________________________________________
>> lwip-users mailing list
>> address@hidden
>> http://lists.nongnu.org/mailman/listinfo/lwip-users

reply via email to

[Prev in Thread] Current Thread [Next in Thread]