[ntp:questions] Hopelessly broken clock?

Hal Murray hal-usenet at ip-64-139-1-69.sjc.megapath.net
Sun Nov 16 23:14:34 UTC 2008


>I'm running CentOS 5 on my system and the only clock source I have available
>is jiffies. Newer kernels should give you options like: acpi_pm jiffies hpet
>tsc pit.

I've been happier since I added
  clocksource=acpi_pm
to the boot command line.


TSC is the default, at least on most of the systems I've used.

There is a bug in the TSC calibration routine.  It gets close, but
it doesn't get the same answer if you run it several times.
If you scan your syslog, you will see things like this:

/var/log/messages.8:Mar 31 00:07:37 shuksan kernel: Detected 2793.147 MHz process
/var/log/messages.9:Feb 11 02:00:18 shuksan kernel: Detected 2793.060 MHz process
/var/log/messages.9:Feb 11 02:19:04 shuksan kernel: Detected 2793.109 MHz process
/var/log/messages.9:Feb 29 00:28:05 shuksan kernel: Detected 2793.156 MHz process

That's enough to add a significant offset to the drift.  So it
might take several hours to recover after a reboot.

I'd try removing the drift file and rebooting.

-- 
These are my opinions, not necessarily my employer's.  I hate spam.




More information about the questions mailing list