5.1 RELEASE: clock running wild?

[Kevin Oberman]

> Date: Thu, 16 Oct 2003 08:27:02 -0500
> From: "G. Panula" <greg.panula at lexisnexis.com>
> Sender: owner-freebsd-stable at freebsd.org
> 
> 
> Alex Povolotsky wrote:
> > Hello!
> > 
> > Not sure if I'm posting into right list.
> 
> Problems with the 5.x branch should be directed to the freebsd-current 
> mail-list.  Rumor has it, 5.3 will be deemed stable.
> 
> > 
> > After upgrading to 5.1-Release I've mentioned that clock on server is running wild (about twice faster than all other world). I am not sure how it worked before upgrade, but... hmm... I can't recall receiving daily output twice a day.
> > 
> > What can I do besides downgrading back?
> > 
> 
> Run ntpd.  Pretty much just need two files; /etc/ntp.conf and 
> /etc/ntp.drift.
> 
> ntp.conf lists the driftfile and the upstream time source.
> example ntp.conf
> server some.ntpserver.com
> driftfile /etc/ntp.drift
> 
> Your local ISP should have an ntp server that you can get the correct 
> time from.  Otherwise there is a list of public stratum 2 ntp servers 
> available at: http://www.eecis.udel.edu/~mills/ntp/clock2a.html
> 
> Might want to use ntpdate to get your time synched up before running ntpd.
> 
> And of course you'll want to read the manpages. :)

As stated, V5.x questions should go to freebsd-current@, not stable.

The problem you are seeing has shown up on a few mobos. The most
common is the ASUS P5A, but others have also reported the issue. It
appears to be a bug in the board ACPI firmware. The clock runs at
exactly double speed. The problem is bad enough that simply using ntp
is inadequate, but it is easily worked around.

Edit /etc/sysctl.conf to contain
"kern.timecounter.hardware="TSC". This will make the clock run
correctly after a reboot.

Enter the command:
sysctl kern.timecounter.hardware="TSC"
to slow down the clock until you reboot. 

Under V4, the TSC was used for clocking, but the ACPI-fast clock is a
better choice when ACPI is available and working.
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman at es.net			Phone: +1 510 486-8634