Description of problem:

ntpd on kernel 2.6.22.9-61.fc6 on an intel xserve x345 (dual hyperthreaded xeon)
can't keep in sync. Booting with clocksource=acpi_pm made kernel-2.6.22.7-57.fc6
happy and keep time, but that workaround fails on 2.6.22.9-61.

messages show [ when booted with clocksource=acpi_pm] :
Oct 15 22:14:32 db01 ntpd[2367]: synchronized to 10.0.5.254, stratum 3
Oct 15 22:14:35 db01 ntpd[2367]: time reset +2.423105 s
Oct 15 22:14:35 db01 ntpd[2367]: kernel time sync enabled 0001
Oct 15 22:17:57 db01 ntpd[2367]: synchronized to LOCAL(0), stratum 10
Oct 16 02:59:54 db01 kernel: set_rtc_mmss: can't update from 0 to 59
Oct 16 02:59:59 db01 last message repeated 5 times


And then time begins to get lost -- more than an hour over 12 hours.

When booting w/o clocksource=acpi_pm, we also loose lots of time -- on the order
of 4 seconds per minute.

Version-Release number of selected component (if applicable):
kernel 2.6.22.9-61.fc6

How reproducible:
Very.


Steps to Reproduce:
Boot kernel on intel x345 server box, run ntpd, watch time get lost.
  
Actual results:
can't keep time, even with local NTPD server, whereas kernel
kernel-2.6.22.7-57.fc6 w/ boot option clocksource=acpi_pm kept us in sync with
stratum 3 NTPD on the local net.

Expected results:
keep time, at least within tolerance for NTPD to help us out.

Additional info: