http://bugzilla.novell.com/show_bug.cgi?id=566319 http://bugzilla.novell.com/show_bug.cgi?id=566319#c32 --- Comment #32 from mail ignored <0.bugs.only.0@gmail.com> 2010-01-18 22:19:09 UTC --- the issue is real, fully reproducible, and, is, apparently, a result of the default setting of Xen's default Platform Timer == HPET @ boot to default xen settings, xm dmesg | egrep -i "timer|clock" (XEN) ACPI: PM-Timer IO Port: 0x808 (XEN) ..TIMER: vector=0xF0 apic1=0 pin1=2 apic2=-1 pin2=-1 (XEN) Platform timer is 14.284MHz HPET (XEN) HPET: 4 timers in total, 3 timers will be used for broadcast (XEN) mcheck_poll: Machine check polling timer started. results in the above documented noisy ntp behavior. otoh, adding a default-override spec'n of "clocksource=" on the Xen cmd line, NOT the widely used spec'n ('xen' vs 'jiffies') on the kernel line, fixes the problem. afaict, assigning clocksource= on the Xen cmd line is completely undocumented @ either the Xen wiki or in the Xen user's guide. e.g., booting to, /xen.gz ... clocksource=acpi ... then, xm dmesg | grep -i timer (XEN) ACPI: PM-Timer IO Port: 0x808 (XEN) ..TIMER: vector=0xF0 apic1=0 pin1=2 apic2=-1 pin2=-1 (XEN) Platform timer is 3.579MHz ACPI PM Timer (XEN) HPET: 4 timers in total, 3 timers will be used for broadcast (XEN) mcheck_poll: Machine check polling timer started. ~ 15 minutes, quiet/stable operation @ stratum=2, full reach. assID=0 status=0684 leap_none, sync_ntp, 8 events, event_peer/strat_chg, version="ntpd 4.2.4p8@1.1612 Mon Dec 28 17:26:30 UTC 2009 (1)", processor="x86_64", system="Linux/2.6.31.8-0.1-xen", leap=00, stratum=2, precision=-20, rootdelay=13.970, rootdispersion=26.079, peer=33500, refid=216.218.254.202, reftime=ceff1d2c.ddbe13c0 Mon, Jan 18 2010 9:47:24.866, poll=6, clock=ceff1e07.464f2053 Mon, Jan 18 2010 9:51:03.274, state=4, offset=15.245, frequency=24.962, jitter=10.571, noise=17.706, stability=7.306 remote refid st t when poll reach delay offset jitter ============================================================================== -clock.fmt.he.ne .PPS. 1 u 41 64 377 16.047 20.346 10.336 -otc2.psu.edu 130.207.244.240 2 u 39 64 377 100.626 14.538 5.387 +clock.isc.org .GPS. 1 u 54 64 377 15.174 -0.843 10.578 *clock.sjc.he.ne .CDMA. 1 u 26 64 377 13.970 12.255 7.051 zorro.sf-bay.or 216.218.254.202 2 u 31 64 377 16.489 13.730 5.781 +nist1.aol-ca.tr .ACTS. 1 u 15 64 377 14.891 5.947 14.596 booting to, /xen.gz ... clocksource=pit ... xm dmesg | grep -i timer (XEN) ACPI: PM-Timer IO Port: 0x808 (XEN) ..TIMER: vector=0xF0 apic1=0 pin1=2 apic2=-1 pin2=-1 (XEN) Platform timer is 1.193MHz PIT (XEN) mcheck_poll: Machine check polling timer started. ~ 20 minutes, quiet/stable operation @ stratum=2, full reach. assID=0 status=0654 leap_none, sync_ntp, 5 events, event_peer/strat_chg, version="ntpd 4.2.4p8@1.1612 Mon Dec 28 17:26:30 UTC 2009 (1)", processor="x86_64", system="Linux/2.6.31.8-0.1-xen", leap=00, stratum=2, precision=-20, rootdelay=16.981, rootdispersion=31.958, peer=63536, refid=207.200.81.113, reftime=ceff2377.503b4ceb Mon, Jan 18 2010 10:14:15.313, poll=6, clock=ceff240e.fa89c747 Mon, Jan 18 2010 10:16:46.978, state=4, offset=9.641, frequency=4.070, jitter=12.958, noise=6.244, stability=0.834 remote refid st t when poll reach delay offset jitter ============================================================================== +clock.fmt.he.ne .PPS. 1 u 30 64 177 15.250 8.422 9.657 +otc2.psu.edu 147.84.59.145 2 u 33 64 177 99.375 10.359 8.906 +clock.isc.org .GPS. 1 u 25 64 177 13.060 2.679 9.501 +clock.sjc.he.ne .CDMA. 1 u 22 64 177 15.097 10.608 7.866 +zorro.sf-bay.or 216.218.254.202 2 u 27 64 177 15.082 9.639 8.443 *nist1.aol-ca.tr .ACTS. 1 u 23 64 177 16.981 15.780 7.743 wrt usgin clocksource=acpi vs clocksource=pit, reading @ http://the-b.org/Linux_timers, it seems (?) acpi may be, in general, the better options, as pit " ... doesn't gain lost time because it doesn't have lost tick correction code ..." iiuc, that references those timers @kernel -- not clear if it is/isn't relevant to Xen ... whether the noisy performance with clocksource=HPET is considered a bug or not is beyond me. as it's, apparently, the DEFAULT that's installed in/with opensuse 11.2, i'd suggest it is. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.