[lwip-users] Re: [lwip] Question and thoughts..

[Martin Glunz]

David Ryan wrote:
> 
> Hi..
> 
> I've recently downloaded lwip060rc1 and been working with that.  I
> finally got lwip so that I can *nearly* ping it from my pc, however
> ethereal is showing up that the ping response ICMP checksum is inccorect
> set as 0x335c ( should be 0x6e98 ).  Windows is throwing the response
> away as a bad packet.  I haven't started looking closely at the icmp or
> ip files yet..  but thought that this sort of error would not show up
> just in implementation?  Anyone have similar problem, or suggest how the
> environment might effect the outcome of the checksum?  The IP checksum
> is coming out correctly.
Just one try:
The ethernet packets must have a minimum length of 60 (or 64?) Bytes, 
otherwise they might not be received at all. Pad them with zeros if 
necessary, this must be done in the ethernet chip driver, if the chip
doesn't do that by itself.

> 
> I'd suggest that the whole sys.c module should be cleaned up to allow a
> seperate thread to handle timer code in a multithreaded implementation
> while still allowing a single threaded implementation to check timers
> another way.  Having sys_sem_wait have the side effects it does is very
> confusing.  Maybe a name change to sys_sem_wait_chk_timers?
> 
> Once I've got my implementation more stable I'll see if I can do some of
> these things.  Do people agree with where I'm going on this?
> 
I've implemented my own simple multithreading os (on a cooperative
basis, not preemptive). It seems to work ok at the moment, but
I'd like to have the timers separated from the semaphores, too.
In single thread implementations the semaphores and mailboxes
aren't used at all. They can be implemented as empty functions.


Martin Glunz
[This message was sent through the lwip discussion list.]