fwp1@CC.MsState.Edu (Frank Peters) (07/20/90)
Hello, As part of my plan to begin synchronizing all of our UNIX systems with ntp I recently installed it on a Sun 4/390 running SunOS 4.0.3. I've had it running on our 4/490 running SunOS 4.1 without any problems. When I run it on the 390 however I get the following messages logged every few minutes: Jul 19 19:35:57 walt vmunix: resettodr: setting TOD chip to 08-20-22 00:35:57 Jul 19 19:35:57 walt vmunix: resettodr: TOD chip was 08-20-22 00:35:57 Jul 19 19:37:29 walt vmunix: resettodr: setting TOD chip to 08-20-22 00:37:29 Jul 19 19:37:29 walt vmunix: resettodr: TOD chip was 08-20-22 00:37:29 Jul 19 19:41:01 walt vmunix: resettodr: setting TOD chip to 08-20-22 00:41:01 Jul 19 19:41:01 walt vmunix: resettodr: TOD chip was 08-20-22 00:41:00 Jul 19 19:44:29 walt vmunix: resettodr: setting TOD chip to 08-20-22 00:44:29 Jul 19 19:44:29 walt vmunix: resettodr: TOD chip was 08-20-22 00:44:29 Jul 19 19:46:50 walt vmunix: resettodr: setting TOD chip to 08-20-22 00:46:50 Jul 19 19:46:50 walt vmunix: resettodr: TOD chip was 08-20-22 00:46:50 Jul 19 19:50:21 walt vmunix: resettodr: setting TOD chip to 08-20-22 00:50:21 Jul 19 19:50:21 walt vmunix: resettodr: TOD chip was 08-20-22 00:50:20 No such messages are logged on the 490. Is this some bug associated with 4.0.3 or the 390 that is not present in the 4.1/490 combination? The same code was compiled on both systems with the same Makefile. Here are some of the obvious things I've looked at: - I've tried it with and without tickadj set to 5. If I put tickadj 5 in the ntp.conf file or don't do it at all I see the errors as above. If I change the value of tickadj in the kernel and memory I get the above messages every few seconds instead of every minutes. I'm assuming from this that the tickadj in the ntp.conf file is ignored...but modifying the kernel seems to make the situation worse. The software was compiled with SETTICKADJ defined. - I've tried it with dosynctodr set to both 0 and 1. This seems to have no effect on the messages. - I've tried running the ntpd executable from the 390 on the 490 and didn't see the problem. The executable compiled on the 490 under 4.1 won't run on the 390 under 4.0.3. An executable compiled on a Sun 4/280 exhibits the same behavior on the 390 though it runs without a problem on the 4/280 (the 4/280 is running 4.0.3). Does anybody have any suggestions? Thanks --Frank -- Frank W. Peters Systems Programmer Computing Center & Services fwp1@CC.MsState.Edu Peters@MsState.Bitnet (601)325-2942 "I can't give you brains, but I can give you a diploma." -- The Wizard of OZ
dupuy@HUDSON.CS.COLUMBIA.EDU (Alexander Dupuy) (07/23/90)
We had this problem too, on a Sun-4/390 running 4.0.3.SPARCserver.390. The
problem is that the kernel is built with a bogus copy of clock.o which has all
the debugging statements turned on. Not only does this generate reams and
reams of stupid messages, but the kernel printfs block the timer interrupts,
causing the machine to keep truly terrible time - worse than a Sparcstation-1!
The fix is to take the Sun-4 clock.o module from a vanilla 4.0.3 release, or
the 4.0.3 patch tape, and make the following "well-known" adb patch to it:
# adb -w /sys/sun4/OBJ/clock.o
todget+0x1e0?X
<old value should be 80a3e006>
todget+0x1e0?W 80a3e007
$q
#
The rebuild your kernel and you're off and running! I did this a few months
back, and it turned my 4/390 from a lousy timekeeper which flooded my syslogs
into a happy (and quiet) chimer.
@alex