[kernel] bug#110: kernel time faster than RTC


X-PA-RISC Linux-PR-Message: report 110
X-PA-RISC Linux-PR-Package: kernel
X-Loop: daniel_frazier@hp.com
Received: via spool by bugs@bugs.parisc-linux.org id=B.98771097612402
          (code B ref -1); Thu, 19 Apr 2001 20:18:02 GMT
Date: Thu, 19 Apr 2001 14:03:19 -0600
From: Grant Grundler <grundler@puffin.external.hp.com>
Message-Id: <200104192003.OAA25531@puffin.external.hp.com>
To: submit@bugs.parisc-linux.org

Package: kernel
Version: 20010418
Severity: normal

Not sure how much drift between RTC and kernel is acceptable but we are
going to have some.  But 1 second per 24h seems a bit high to me. That's
30sec per month and even my cheapo casio does better than that.

Perhaps the kernel needs to note the drift and on occasion tune the
"fudge factor" used in the itimer path. Then assume ntpdate is available
to update the RTC. Other thoughts?

Workaround is to add a cron job which runs ntpdate and then hwclock daily.

To: Andreas Thienemann <andreas@thienemann.net>
Subject: Re: [parisc-linux] Clock skew problems 
Date: Thu, 19 Apr 2001 12:02:54 -0600
From: Grant Grundler <grundler@puffin.external.hp.com>

Andreas Thienemann wrote:
> I'm getting the following message from make:
> make: *** Warning: File `libdb.a' has modification time in the future
> (2001-04-19 15:15:33 > 2001-04-19 15:15:04)
> And I have absolutely no Idea why this is happening.

Everytime one updates the kernel time and it gets set *back*,
make will notice the derived objects (ie libdb.a) are newer than
current time and complain.

However, I suspect the kernel isn't keeping time perfectly and updating
from the HW clock (YMMV: historically, some servers had better RTC crystals)
is causing the problem. Having a bad RTC will also look like
the kernel isn't keeping time when in fact it is.

a500# date
Thu Apr 19 11:11:02 MDT 2001
a500# hwclock
Thu Apr 19 11:11:01 2001  -1.003308 seconds
a500# uptime
 11:11:36 up 21:25,  1 user,  load average: 0.00, 0.04, 0.04

So not doing too badly - but it needs to be fixed.
I'm assuming the A500 has a good RTC.
I'll submit a bug for you.


Andreas Thienemann <andreas@thienemann.net> replied later with:

| > I'm assuming the A500 has a good RTC.
| I've got a 821/D250 and I don't know anything about the accuracy of the
| RTC clock. But hwclock tells me the following:
| [root@hp9000 root]# hwclock
| Thu Apr 19 20:17:25 2001  -1.063290 seconds