115

Unix/Linux 网络时间协议版本3 Network Time Protocol Version 3 (NTP) distribution for Unix systems

Received: from snow-white.ee.udel.edu by whimsy.udel.edu id aa10420;
          28 Mar 97 5:00 GMT
Received: from fang.dsto.defence.gov.au (fang.dsto.defence.gov.au [131.185.2.5]) by digger1.defence.gov.au (8.7.5/8.7.3) with ESMTP id PAA27717; Fri, 28 Mar 1997 15:30:38 +1030 (CST)
Received: from sibyl.chez-dall.org.au (dall-i.remote.dsto.defence.gov.au [131.185.44.51]) by fang.dsto.defence.gov.au (8.7.6/8.7.6) with ESMTP id PAA10167; Fri, 28 Mar 1997 15:30:14 +1030 (CST)
Received: (from ian@localhost)
	by sibyl.chez-dall.org.au (8.8.5/8.8.5) id PAA20983;
	Fri, 28 Mar 1997 15:30:11 +1030 (CST)
Date: Fri, 28 Mar 1997 15:30:11 +1030 (CST)
Message-Id: <[e-mail]>
From: Ian Dall <[e-mail]>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
To: [e-mail], [e-mail]
Subject: Re:  xntpd gets confused by unreachable networks
In-Reply-To: <[e-mail]>
References: <[e-mail]>
X-Mailer: VM 6.22 under Emacs 19.34.2

OK. I have solved the problem. It is not, as I thought, packets not
being delivered, but rather, packets not being received. This is fixed
by configuring with --enable-udp-wildcard=yes. I note from the
ChangeLog that this became the default for freebsd recently. Perhaps
it should be the default for netbsd as well.  In fact, from the long
list of systems in configure.in for this condition, I wonder if it
should be the default and only list the systems where it *shouldn't*
be enabled.

On inspection, the code in ntp_io does use sendto. There is some code
to make sure you don't get swamped by messages. I did add the following
information message:

--- ntp_io.c~   Thu Mar 13 17:56:27 1997
+++ ntp_io.c    Thu Mar 27 23:03:33 1997
@@ -1277,8 +1277,10 @@
       /*
        * He's not bad any more
        */
-      if (slot >= 0)
+      if (slot >= 0) {
+       msyslog(LOG_INFO, "Connection to %s", ntoa(dest));
        badaddrs[slot].port = 0;
+      }
     }
 

The idea is if you get a message saying there is "no route to xxx"
it would be nice to get a message when there it becomes reachable
again. But this is still assymetric the no route to host is
at LOG_ERR.

This all arises because if ppp is not up when xntpd is started, it
doesn't recognize the interface, and it sends on the "default"
interface. If udp-wildcard is not enabled, data arriving on 
the default interface is dropped.

This interface bit is all pretty contorted and non portable, I would
have thought it would be easier to specify interfaces in the config
file (only relevant for broadcast anyway I think).

Ian