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

[Jani Monoses]

> >>
> >What platform do you run lwip on?
 
> Attempting :)  to run it on a playstation2 on the IOP(IO Processor) 
> which is mips3000.  Playstation 2 has two processors..  one handling IO 
> (which is actually the old playstation one chip) and the other (EE) 
> handling the main programs.  I'm using a NetGear EA101 (Kawasaki 
> chipset) over its USB connections.  This isn't related to playstaion2 
> linux either.. this is direct access to the system.
Then the ICMP problem is probably related to your netif driver or misalignment
have you tried sending raw ethernet frames form your app (using linkoutput )
and see if they get through correctly?
What it your MEM_ALIGMENT?What is required for MIPS?

> On a different note.. is *anyone* using lwip in a multithreaded 
> environment?  I'm currently looking closely at the mbox code used in the 

Yes I do and I used sys.c but recently tried without it and did sys_sem_wait
and mbox_fetch in sys_arch too.And sys_timeout as well.This part of lwip
need a bit of rethinking especially because of threaded implemenetations.

> unix arch.. I've used it as the basis for the ps2.  One thing I noticed 
> is that the mbox->mail semaphore gets posted when the first message is 
> sent to the mbox.. but is never signalled when a message is removed. 

This is in the current code...what do you mean by never signalled?
  mbox->first++;

  if(mbox->wait_send) {
    sys_sem_signal(mbox->mail);
  }


>  I've added the following just after mbox->first++ in mbox_fetch.  This 
> just resets the semaphore ready for next time when we've emptied the queue.
> 
> if ( mbox->first == mbox-> last )
>    sys_sem_signal( mbox->mail );
> 
> I've still got a race condition which is causing the main thread to 
> stall.  It could be in my implementation of sys_arch_sem_wait.  I'll 
> keep searching.
[This message was sent through the lwip discussion list.]