From: Chuck Murcko <chuck@n2k.com>
Message-Id: <199701141805.NAA16721@spudboy.telebase.com>
Subject: Re: lingering_close
To: new-httpd@hyperreal.com
Date: Tue, 14 Jan 1997 13:05:44 -0500 (EST)
In-Reply-To: <Pine.BSI.3.95.970114105005.24103A-100000@valis.worldgate.com>
 from "Marc Slemko" at Jan 14, 97 10:51:15 am
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: new-httpd-owner@apache.org
Precedence: bulk
Reply-To: new-httpd@hyperreal.com

Well, that would then not do the lingering close for timeouts, right?

BTW, where do we call lingering_close in timeout? I don't see that. The
only place I see it getting called is in child_main().

FYI, SGI has an IP patch that sets a configurable timeout for sockets with
SO_LINGER set, so that might be helping on those platforms.

Marc Slemko liltingly intones:
> 
> Except that on a number of platforms, people are saying that NO_LINGCLOSE
> _is_ helping, ie. making the difference between their machines crashing
> and them staying  alive with no problems.
> 
> Details on what various people are seeing under what conditions to come
> later...
> 
> On Tue, 14 Jan 1997, Randy Terbush wrote:
> 
> > Actually, as I think about this more, I believe that the reason
> > we are seeing more FIN_WAIT_2 in 1.2 is because we are slamming
> > the door on timeouts. The whole purpose of lingering_close() was
> > to work around some of the buggier TCP stacks. This would also
> > explain why defining NO_LINGCLOSE is not helping the situation.
> > 
> > 
> > > Then I think we need to be setting timeout_req = NULL. It seems 
> > > that the frequency of calls to lingering_close() with Invalid
> > > sd are directly related to timeouts. We're longjumping back
> > > to child_main() and calling lingering_close() there if(r).
> > > 
> > > 
> > > > Randy Terbush wrote:
> > > > > 
> > > > > Should we be calling lingering_close() in timeout()?
> > > > 
> > > > Naah - if we've timed out we don't want to waste even more time trying to
> > > > deliver stuff to them.
> > > > 
> > > > Cheers,
> > > > 
> > > > Ben.
> > > > 
> > > 
> > 
> 

chuck
Chuck Murcko	N2K Inc.	Wayne PA	chuck@telebase.com
And now, on a lighter note:
I can read your mind, and you should be ashamed of yourself.