[lwip-users] Re: RTO issues

[Chen]

Hi Andrew,

I believe the bottleneck is in lwip. As I stated in my previous post,
this is NOT a bug, but a room for performance improvement for
lwip.

Please see

http://lists.nongnu.org/archive/html/lwip-users/2010-12/msg00059.html
for my first post, and you can find out the rest of them in

http://lists.nongnu.org/archive/html/lwip-users/2010-12/index.html
under the subject "Re: lwip-users Digest, Vol
88, Issue 17" (My bad for not editing the title :(

I think the best way to rule out my platform or bad coding is someone
write a quick telnet program on another platform(it is very very simple
and won't take long at all, mine is similar to

http://www.ultimaserial.com/avr_lwip_tcp.html) and it simply sends
out 1200 byte packets continously once TCP connection is made (I used
lwip_send)

Configure the network like this:

lwip->hub->PC1, and connect PC2 to the hub

Run WireShark on PC2 to capture packets between lwip and PC1

Once lwip begins to send out packets to PC1, pull the cable between PC1
and hub.

My capture shows lwip will not tranfer more than two packets, it will
wait for the ACK. (non-lwip's RTO capture can be found in my first
post)

If you can capture a different result, than the problem is on my
end.

At 05:09 PM 12/15/2010, you wrote:

> You may want to check Atmel’s
> website for the latest errata. They have
> an upgraded version of your framework that already incorporates
> 1.3.2.
> Even if you don’t want to rework your entire project, it would be
> worth
> looking into the low_level_input functions to ensure there weren’t
> any
> changes.
>
> Andrew