[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [lwip-users] Retransmission timeout
From: |
Kieran Mansley |
Subject: |
Re: [lwip-users] Retransmission timeout |
Date: |
Mon, 23 Jun 2003 11:56:48 +0100 (BST) |
On Fri, 20 Jun 2003, Zschocke, Florian wrote:
> This maybe a stupid question, but I fail to answer it myself.
>
> I understand that lwIP is using van Jacobson's algorithm for RTO
> determination. But what is the minimum timeout value? The minimum timeout
> before a retransmission is started? What I was able to deduce from the code
> reads as if it would be 3000*TCP_SLOW_INTERVAL, i.e. 3000*500ms, i.e. 25min.
> That seems a bit long thus I assume I'm missing something important here.
Where do you get the 3000 from? From taking a quick look at the code a
retransmission will occur when the number of coarse timer ticks since
the last segment was output (pcb->rtime) is greater than (the
round trip estimator) pcb->rto.
pcb->rto seems to be interpreted (when it is used) as being in units of
TCP_SLOW_INTERVAL, but I haven't checked to make sure that things that set
it also interpret it in this way.
The thing that makes me suspicious is I see retransmissions begin much
sooner than 25 min. I am however using an older version of the source, so
maybe we've introduced a bug since then. Are you seeing an actual problem
or just thought it was wrong when reading the source?
Thanks!
Kieran