Mailing-List: contact dev-help@httpd.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@httpd.apache.org
Message-ID: <00b201c1ed39$ed7ee640$50381b09@sashimi>
From: "Bill Stoddard" <bill@wstoddard.com>
To: <dev@httpd.apache.org>
References: <20020426051229.48294.qmail@icarus.apache.org>
 <00a201c1ed36$b7732c80$50381b09@sashimi>
Subject: Re: cvs commit: httpd-2.0 STATUS
Date: Fri, 26 Apr 2002 11:49:13 -0400
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit


> >   +    * The worker MPM should not accept more connections than it
> >   +      currently has available workers. Instead, the listener thread
> >   +      should block on a condition of the fdqueue such that it waits
> >   +      until there are at least one idle worker before continuing
> >   +      into the accept-loop.
>
> Here is an alternate idea... keep track of the number of available workers with a
counter.
> If the accept loop detects that the number of available workers has dropped to 0, the
> accept thread simply yields it's quanta. Very simple and no overhead associated with a
> condition variable. Something like this:
>
> if (!avail_workers) {
>    yield(); /* or what ever the Unix call is. I know there must be one */
>    continue;
> }
> else {
>    accept();
> }
>
> Bill

Forgot to mention what is probably obvious... this is not a perfect solution because of
race conditions between the threads incrementing/decrementing avail_workers and the
listener thread doing the checking. It narrows the exposure but does not eliminate it
entirely.

Bill