From owner-freebsd-stable Fri May 17 2:43:55 2002 Delivered-To: freebsd-stable@freebsd.org Received: from iguana.icir.org (iguana.icir.org [192.150.187.36]) by hub.freebsd.org (Postfix) with ESMTP id 70FA537B406 for ; Fri, 17 May 2002 02:43:46 -0700 (PDT) Received: (from rizzo@localhost) by iguana.icir.org (8.11.6/8.11.3) id g4H9hj176564; Fri, 17 May 2002 02:43:45 -0700 (PDT) (envelope-from rizzo) Date: Fri, 17 May 2002 02:43:45 -0700 From: Luigi Rizzo To: stable@freebsd.org Subject: syslogd and inetd problems Message-ID: <20020517024345.A75815@iguana.icir.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.3.23i Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG Hi, recently I have seen the following problems with inetd and syslogd, so was wondering if my diagnose is correct, other have seen them, and how urgent is a fix. syslogd: + i am using syslogd to log to a remote system. From time to time, the syslog daemon stops sending to the remote system, and I need to restart it. Haven't traced the problem in detail, but it usually happens across some network outage between the two. A high level inspection of the code in syslogd.c suggests that the daemon closes the connection on a sendto failure, which is probably not the smartest thing to do. At the very least, the code should implement some recovery mechanism. inetd + i have a soekris box running inetd and bootpd and a few other simple services. The box is connected to a switch powered by the same power strip. I noticed that upon a recent power outage, inetd came up "deaf", i.e. it did not seem to receive bootp requests, even though I can see them running a tcpdump on the box (and I can ssh into the box, of course). The fix was to issue a 'kill -HUP' to the inetd process. Here, i suspect that the soekris box is somewhat faster than the switch to come up, for some reason inetd fails to do some operation on the socket, and this prevents further activity. Now, the inetd code seems to use a call to alarm() to deal with bind() failures, but i am not sure how effective is the process. cheers luigi To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message