From owner-freebsd-hackers  Tue Nov  4 03:17:48 1997
Return-Path: <owner-freebsd-hackers>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.7/8.8.7) id DAA04709
          for hackers-outgoing; Tue, 4 Nov 1997 03:17:48 -0800 (PST)
          (envelope-from owner-freebsd-hackers)
Received: from ren.dtir.qld.gov.au (firewall-user@ns.dtir.qld.gov.au [203.108.138.66])
          by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id DAA04700
          for <freebsd-hackers@freebsd.org>; Tue, 4 Nov 1997 03:17:43 -0800 (PST)
          (envelope-from syssgm@dtir.qld.gov.au)
Received: by ren.dtir.qld.gov.au; id VAA22054; Tue, 4 Nov 1997 21:33:09 +1000 (EST)
Received: from ogre.dtir.qld.gov.au(167.123.8.3) by ren.dtir.qld.gov.au via smap (3.2)
	id xma022049; Tue, 4 Nov 97 21:32:46 +1000
Received: from localhost.dtir.qld.gov.au (localhost.dtir.qld.gov.au [127.0.0.1])
	by ogre.dtir.qld.gov.au (8.8.7/8.8.7) with SMTP id VAA14283;
	Tue, 4 Nov 1997 21:17:16 +1000 (EST)
Message-Id: <199711041117.VAA14283@ogre.dtir.qld.gov.au>
X-Authentication-Warning: ogre.dtir.qld.gov.au: localhost.dtir.qld.gov.au [127.0.0.1] didn't use HELO protocol
To: Charles Mott <cmott@srv.net>
cc: freebsd-hackers@freebsd.org, syssgm@dtir.qld.gov.au
Subject: Re: Maximum number of forked processes 
References: <Pine.BSF.3.96.971103192451.29009E-100000@darkstar.home>
In-Reply-To: <Pine.BSF.3.96.971103192451.29009E-100000@darkstar.home>
    from Charles Mott at "Tue, 04 Nov 1997 02:27:17 +0000"
Date: Tue, 04 Nov 1997 21:17:15 +1000
From: Stephen McKay <syssgm@dtir.qld.gov.au>
Sender: owner-freebsd-hackers@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

On Tuesday, 4th November 1997, Charles Mott wrote:

>On Tue, 4 Nov 1997, Greg Lehey wrote:
>> On Mon, Nov 03, 1997 at 04:16:36PM -0700, Charles Mott wrote:
>> > How does one control the maximum number of forked processes from within
>> > the parent process?
>> 
>> setrlimit(2) allows you to limit
>> the  number of processes (=forks)    per user, but   I  don't know  of
>> anything that would limit the number of children of a process.

>Tentatively, I am thinking of incrementing a counter when a process is
>forked and decrementing it when a SIGCLD is received.  However, I don't
>know what standard procedure is in this matter.

This is often tried.  Some people even try to defend it.  But the truth
of the matter is that SIGCHLD may be delivered more or less times than
the number of exiting processes since it is delivered for stopped children
and since only one signal is delivered even if multiple children exit
before the signal is processed.

The correct way to spot child deaths is with wait() or any of the wait
variants we have available nowadays.  Of course, the SIGCHLD can give
you a hint that wait() is necessary. :-)

Of the variants available,
    wait() is old and portable to the early days of Unix, 
    wait3() is traditional BSD but I can only remember back to 4.2 :-)
    waitpid() is POSIX, so probably blessed by most people
    wait4() is BSD 4.4 or maybe Sun derived (Help me out here!)
    waitid() is SysV (and we don't have it)

You probably want waitpid() in a loop in the SIGCHLD handler.  Or better
yet, if you have some select() loop you can detect EINTR, check the flag
set by your SIGCHLD handler, and apply waitpid() as appropriate.

Stephen.