Date: Thu, 19 Nov 1998 18:39:33 -0800 (PST) From: John Polstra <jdp@polstra.com> To: garman@earthling.net Cc: current@FreeBSD.ORG Subject: Re: more dying daemons Message-ID: <199811200239.SAA17040@vashon.polstra.com> In-Reply-To: <199811190247.SAA25805@hub.freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
In article <199811190247.SAA25805@hub.freebsd.org>,
<garman@earthling.net> wrote:
> while sitting here trying to find a common thread through all this
> mess...
>
> it appears to happen only with servers that fork() copies of themselves
> on incoming requests.
>
> inetd does this, samba does this, and sshd does this. consequently,
> inetd dies, samba dies, and sshd dies :)
>
> however, servers that *don't* fork off a new connection on incoming
> requests don't die.
This comment nudged my feeble memory and caused me to notice that
there are some definite similarities between this problem and the
"can't build Modula-3 on an SMP system" bug which I watched Tor Egge
fix a few months ago. I won't recap all the details here, but it had
to do with forking and corrupted memory and SMP. (Somebody remarked
that the dying daemon problem went away for him when he switched to
SMP.)
Someone who is experiencing this problem please try a very simple
experiment for me. On a freshly rebooted system (to ensure a clean
slate), do this:
sysctl -w kern.fast_vfork=0
Better yet, add it to your /etc/rc.local file too, so that you won't
forget to do it the next time you reboot.
Then see if the dying daemon problem still happens, and let me know.
This is just a hunch, but it's an easy experiment to try.
John
--
John Polstra jdp@polstra.com
John D. Polstra & Co., Inc. Seattle, Washington USA
"Nobody ever went broke underestimating the taste of the American public."
-- H. L. Mencken
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199811200239.SAA17040>