Kernel time problem 2.2.15/2.2.16

From: Richard B. Johnson (root@chaos.analogic.com)
Date: Thu Aug 31 2000 - 08:59:27 EDT

  • Next message: almesber@lrc.di.epfl.ch: "Re: [PATCH] 2.2: /proc/config.gz"

    Every morning at 00:00:00, I have cron set my system clock to
    NIST (time-a.timefreq.bldrdoc.gov). Then I use hwclock to set
    the hardware clock to the new system time.

    # Min Hour Day Month Day_of_week Command
    #
    0 0 * * * /usr/local/sbin/nettime time-A.timefreq.bldrdoc.gov >>/var/adm/set_time.log;/sbin/hwclock --systohc >>/var/adm/set_time.log

    `nettime` is `rdate` that shows the time before and after it's set.
     
    This has worked very well with a relatively old version of Linux,
    linux-2.2.6. The time of my server has always been within about 10
    seconds of NIST.

    I upgraded my server to:
    Linux boneserver 2.2.15 #22 SMP Fri Aug 25 17:31:25 EDT 2000 i686

    Now the time is diverging. Every night, the time is further and
    further from NIST, just before it is set.

    The current contents of /etc/adjtime shows that the last time hwclock
    set the time, it was 394 seconds off:

            -394.634766 967724179 0.000000
            967724179

    This is a bit much. Water-clocks are more accurate than this.
    This may be a "set-back-time" problem because the kernel thinks
    that it's midnight, the time gets set back roughly 6 minutes,
    then there may be a problem when it's midnight again, 6 minutes
    later. Just a WAG.

    Is this a known problem?
    Should I just zero out or remove /etc/adjtime and see if it starts
    to converge?
    Should we always zero out or remove /etc/adjtime during an upgrade to handle
    possible code changes?

    FYI, I just deleted /etc/adjtime and created a new one with hwclock.
    I will see what it does tonight.

    Cheers,
    Dick Johnson

    Penguin : Linux version 2.2.15 on an i686 machine (797.90 BogoMips).

    "Memory is like gasoline. You use it up when you are running. Of
    course you get it all back when you reboot..."; Actual explanation
    obtained from the Micro$oft help desk.

    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.kernel.org
    Please read the FAQ at http://www.tux.org/lkml/



    This archive was generated by hypermail 2b29 : Thu Aug 31 2000 - 08:54:28 EDT