Re: [ECOS] assert: "no sleep slots"

[Christoph Csebits <christoph dot csebits at frequentis dot com>]

On Thu, Jul 22, 2004 at 06:04:04AM -0600, Gary Thomas wrote:
> On Thu, 2004-07-22 at 05:35, Christoph Csebits wrote:
> > hi,
> > 
> > in net/bsd_tcpip/current/src/ecos/synch.c
> > 
> > there are the following code lines:
> > 
> >  CYG_ASSERT( i <  CYGPKG_NET_NUM_WAKEUP_EVENTS, "no sleep slots" );
> >  
> >  // Defensive:
> >  if ( i >= CYGPKG_NET_NUM_WAKEUP_EVENTS ) {
> >      cyg_scheduler_unlock();
> >      return ETIMEDOUT;
> >  }
> > 
> > The things on this:
> > 
> > 1) The if clause is redundant, because the assert will stop execution anyway.
> 
> Only if you have asserts enabled!

Ok, i forgot about that.

> 
> > 2) I didn't try to figure out how this works, so my question:
> >    Is the assert there because, this must not happen (because of a
> >    bug/failure) or is it just here to say "no resources, please adjust
> >    your CDL parameters"?
> > 
> > So, would it be possible to remove the assert and let the if clause 
> > return ETIMEDOUT instead?
> 
> This does just what it should - if you have asserts on, it will tell you
> when you've run out of these resources which is arguably considered an 
> error.  If you have no asserts, it at least tries to keep going.
> 
> We put the assert in there to trap this condition which we never expect
> to see.  In fact, to date it has never happened for anyone AFAIK, so it
> is all merely defensive programming (as the comment states)

Unfortunately we had it at least two times. I don't know the real circumstances,
but it was associated with real heavy network traffic, because of an
errorneous loop in our ethernet switching environment. We can't reproduce it,
but seemingly we had to deal with this problem, and an assert isn't acceptable.

Thanks for clarification,
Christoph

-- 

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss