Re: [lwip-users] Duplicate FINs and CLOSE_WAIT state

[Leon Woestenberg]

Hello all,

Jan Ulvesten wrote:

> Hi
>
> Thanks, it sounds like this is might be the solution. I will try it and
> come back with a status-report. Likewise with the new checksum routine
> (the old one never worked on Altera NIOS).
>  
Please heads up:

I am still discussing some rare cases with my collegea Christiaans 
Simons, also present here

Each pbuf can have a payload that is one of four cases:

aligned odd-length sequence
aligned even-length sequence
unaligned odd-length sequence
unaligned even-length sequence

A packet (in a generic case) is a chain of any mix of such pbufs.

The checksum routine must be aware of the MSb/LSb cadence of the byte 
sequence in the *packet* payload, i.e. *across* several pbufs.

Now, besides this, this must run correctly on little and big-endian 
systems alike, as well as some exoteric systems.

Although Christiaans' generic and Curt's optimized routines are very 
much an improvement, a code review is necessary to see if API changes 
are necessary to cope with all of this. This might incur simple changes 
of the API documentation, imposing restrictions so that the current 
stuff works as advertised, or a more thorough work-over off the 
checksumming routine and callers.

Any help is welcomed; we need insight and platform testers here.

Regards,

Leon.