3

来自「Unix/Linux 网络时间协议版本3 Network Time Proto」· 代码 · 共 40 行

Replied: Wed, 06 Nov 1996 19:25:05 -0500
Replied: "Jeffrey C Honig <[e-mail]> Dave Mills <[e-mail]>"
Received: from snow-white.ee.udel.edu by whimsy.udel.edu id aa03421;
          7 Nov 96 0:14 GMT
Received: from kismet.bsdi.com ([e-mail] [207.16.111.10]) by external.BSDI.COM (8.8.2/8.8.2) with ESMTP id RAA28328; Wed, 6 Nov 1996 17:14:21 -0700 (MST)
Received: from kismet.bsdi.com ([e-mail] [127.0.0.1]) by kismet.bsdi.com (8.7.6/8.7.3) with ESMTP id TAA10572; Wed, 6 Nov 1996 19:14:20 -0500 (EST)
Message-Id: <[e-mail]>
To: Harlan Stenn <[e-mail]>
cc: Dave Mills <[e-mail]>
Subject: Re: BSD/OS 3.0 
In-reply-to: Message from Harlan Stenn <[e-mail]> on
             Wed, 06 Nov 1996 17:48:44 -0500.<[e-mail]> 
References: <[e-mail]>
Organization: Berkeley Software Design, Inc.; Ithaca, NY  USA
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <[e-mail]>
Date: Wed, 06 Nov 1996 19:14:04 -0500
From: Jeffrey C Honig <[e-mail]>

> Just a stupid question before I go ahead and see what I can do...
> 
> If BSDI-3.x doesn't let the time go backwards if the security
> level is greater than zero, then would it be better to try, notice a
> failure, and then slew the time with adjtime?  (I hope I said that right
> - I still get confused by "step" and "slew"; Dave, can you set me
> straight on these terms?)

Yes, that would work.  It would allow xntpd() to step the clock
forward quickly.  I suspect that operation would be much more common
than having to set the time back a significant amount.  Especially
since we try to run ntpdate at boot time to get the clock.

It would also be good not to log syslog messages on an EPERM error
message from settimeofday(), as I suspect this could happen often.

Thanks.

Jeff