2.6.8.1 slews system clock

2.6.8.1 slews system clock

[Jan Spitalnik]

Hi,

after updating kernel to 2.6.8.1 the system clock slews by 1 second every 10
seconds into future. I tried turning off ACPI, but that had no effect.

root@largo:~# ntpdate tik.cesnet.cz;sleep 10;ntpdate tik.cesnet.cz
20 Aug 13:55:01 ntpdate[5315]: step time server 195.113.144.201 offset
-24.488611 sec
20 Aug 13:55:13 ntpdate[5321]: step time server 195.113.144.201 offset
-1.042110 sec

on second machine the effect is opposite, ie the clock slews backwards.
Any ideas?

Thanks,

--
Jan Spitalnik
jan@spitalnik.net

Re: 2.6.8.1 slews system clock

[Tim Schmielau]

On Fri, 20 Aug 2004, Jan Spitalnik wrote:

> after updating kernel to 2.6.8.1 the system clock slews by 1 second every 10
> seconds into future. I tried turning off ACPI, but that had no effect.
> 
> root@largo:~# ntpdate tik.cesnet.cz;sleep 10;ntpdate tik.cesnet.cz
> 20 Aug 13:55:01 ntpdate[5315]: step time server 195.113.144.201 offset
> -24.488611 sec
> 20 Aug 13:55:13 ntpdate[5321]: step time server 195.113.144.201 offset
> -1.042110 sec
> 
> on second machine the effect is opposite, ie the clock slews backwards.
> Any ideas?

Which was the last kernel that worked?

Re: 2.6.8.1 slews system clock

[Jan Spitalnik]

Dne pá 20. srpna 2004 16:02 Tim Schmielau napsal(a):
> On Fri, 20 Aug 2004, Jan Spitalnik wrote:
> > after updating kernel to 2.6.8.1 the system clock slews by 1 second every
> > 10 seconds into future. I tried turning off ACPI, but that had no effect.
> >
> > root@largo:~# ntpdate tik.cesnet.cz;sleep 10;ntpdate tik.cesnet.cz
> > 20 Aug 13:55:01 ntpdate[5315]: step time server 195.113.144.201 offset
> > -24.488611 sec
> > 20 Aug 13:55:13 ntpdate[5321]: step time server 195.113.144.201 offset
> > -1.042110 sec
> >
> > on second machine the effect is opposite, ie the clock slews backwards.
> > Any ideas?
>
> Which was the last kernel that worked?

2.6.7 didn't exhibit this problem. I will test 2.6.8-rc's to find which one 
caused this regression. 

-- 
Jan Spitalnik
jan@spitalnik.net

Re: 2.6.8.1 slews system clock

[Jan Spitalnik]

Dne pá 20. srpna 2004 17:02 Jan Spitalnik napsal(a):
> > Which was the last kernel that worked?
>
> 2.6.7 didn't exhibit this problem. I will test 2.6.8-rc's to find which one
> caused this regression.

Hi,

I've did further testing and I've found out that fortunately it's not a kernel 
problem (sorry for the buzz :-/ ). adjtimex was the culprit. 
adjtimexconfigure said that the CMOS clock is faster by 970secs a day to 
system clock. When I disabled adjtimex and rebooted, the system clock is now 
rock solid and doesn't slew. Now I wonder what's wrong with adjtimex or my 
system. I've tried on several other systems and the it was all the same.
Thanks,
  jan
-- 
Jan Spitalnik
jan@spitalnik.net