[lwip-users] Re: [lwip] lwIP 0.5.3 tcp Checksum issues?

[Jani Monoses]

Hi
as Florian said it could be an alignment issue.
I had bad checksums because the data in httpd was not 4byte aligned.
In the file where those binary structs live (fsdata.c) make the structs 
aligned with the gcc attribute feature.I used an ARM processor.
Jani.

> Hi all.
> 
>   Are there any known issues with 0.5.3 with regards to checksum calculation?
> 
>   Specifically, I am testing my own httpd server (heavily modified from the 
> original using the raw api) and pages don't get displayed. I am tracking this 
> down to the fact that the responding packet of data is sent, but packet 
> sniffing software says it has an invalid tcp checksum and my PC isn't ACKing. 
> The data just keeps getting resent with the same incorrect checksum.
> 
>   The main index web page supplied in the lwIP archive works ok, others 
> don't, so it seems to depend on the data content.
> 
>   What's strange is I had tested this code and it works under the win32 port 
> (and still does work) - although due to different timings the exact sequence 
> of events are different.
> 
>   What's even more annoying is that I have compared the ROM pbuf checksum 
> total with the win32 port's ROM pbuf checksum total - they are equal. I have 
> calculated manually the TCP pbuf checksum and added this all up with the 
> pseudo header and get the same checksum as that which was sent with the 
> packet !! Surely the packet sniffing software and windows 2000 checksum 
> implementation is correct ? ! I can assume that can't I ? ! :-)
> 
> Regards,
> Michael Portmann
> 
> HYDRA Electronic Design Solutions Pty Ltd
> 140 Ashley Street, Underdale, SA 5032.
> Ph. +61 8 8234-0477
> Fx. +61 8 8234-1840
> 
> [This message was sent through the lwip discussion list.]
[This message was sent through the lwip discussion list.]