[Bug 350981] New: The system clock doesn't work correctly
https://bugzilla.novell.com/show_bug.cgi?id=350981 Summary: The system clock doesn't work correctly Product: openSUSE 10.3 Version: Final Platform: i686 OS/Version: openSUSE 10.3 Status: NEW Severity: Normal Priority: P5 - None Component: Kernel AssignedTo: kernel-maintainers@forge.provo.novell.com ReportedBy: robin.listas@telefonica.net QAContact: qa@suse.de Found By: --- This bug is closely related to Bug 344356 - please read it for more details. In a nutshell: This is what I see in the boot log: nimrodel:~ # grep -i "clock\|tsc\|acpi_pm" /var/log/boot.msg <6>Time: tsc clocksource has been installed. <6>Real Time Clock Driver v1.12ac <4>Marking TSC unstable due to: possible TSC halt in C2. <6>Time: acpi_pm clocksource has been installed. <6>intel8x0_measure_ac97_clock: measured 50864 usecs <6>intel8x0: clocking to 48000 doneSetting up the hardware clockdone The system thus defaults to acpi_pm clock, but this one doesn't keep correct time, and ntp ends by quitting: the error can be half an hour. Thus I forced the kernel to use "tsc" instead by doing: echo tsc > /sys/devices/system/clocksource/clocksource0/current_clocksource This does work and the clock keeps (I wrote it into /etc/init.d/boot.local), although I see this in the log, soon after boot: Dec 25 14:32:08 nimrodel kernel: Clocksource tsc unstable (delta = 9373788647 ns) I have also tried the "jiffies" clock, but this doesn't work: the clock simply stops ticking, and going back to "tsc" doesn't work, requiring a hard reboot. We have argued this in <http://lists.opensuse.org/opensuse/2007-12/msg00357.html >. At least one other person reports a similar issue (he switched to "jiffies"). I have been unable to find a document explaining what each type of clocksource is or does. I'm not a kernel expert. I have also tried to recompile the kernel in several small modifications, but so far, no luck. This thing started the next day I installed 10.3, but I found some symptoms in the 10.2 log, small to be noticed by me at the time. This bug might be related to Bug 350980 (lazy system) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=350981
Jeff Mahoney
https://bugzilla.novell.com/show_bug.cgi?id=350981
User ak@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c1
Andi Kleen
From the other bug
Nov 27 15:57:46 nimrodel kernel: set_rtc_mmss: can't update from 1 to 57 That's likely broken hardware of some sort. Note that hardware sometimes breaks so it might be possible that it wasn't broken already when you used the older version. Can you double check it doesn't happen in the older version? If you say it started with 10.3 -- with what version did it work? Can you add boot messages from both a broken and a working version? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=350981
User robin.listas@telefonica.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c2
--- Comment #2 from Carlos Robinson
Actually acpm_pm is just the offset timer -- since it has only 24bits precision and would wrap quiickly it used relative to some other; typically PIT. If it drifts the problem is likely there.
I'll try on another moment to force the PIT clock; in case the system crashes, I have to be in runlevel 3 when I do it.
TSC would be another method to be an offset timer, but it can't help if the base timer drifts.
But TSC doesn't drift. The kernel complains briefly, but the clock remains stable, it seems; I don't know of a method to output the current time difference between the OS clock and internet time as read by the NTP daemon, and thus check the possible drift.
From the other bug
Nov 27 15:57:46 nimrodel kernel: set_rtc_mmss: can't update from 1 to 57
That's likely broken hardware of some sort. Note that hardware sometimes breaks so it might be possible that it wasn't broken already when you used the older version. Can you double check it doesn't happen in the older version?
Ok, I still have a partition with 10.2 installed, I will try there.
If you say it started with 10.3 -- with what version did it work? Can you add boot messages from both a broken and a working version?
I noticed it the first day I upgraded to 10.3 because the time jumps were bigger than 10 minutes, the ntp daemon quited, and the error was very visible compared to the video machine clock: so I investigated. But looking in the old logs, knowing what to look for, I see some small time resets in 10.2. I think the log for 10.1 is lost, I had a bad HD crash at the time. These are messages (old) from the spare 10.2 partition: nimrodel:/otros/test_a/var/log # grep -i "clock\|tsc\|acpi_pm" boot.msg <6>Real Time Clock Driver v1.12ac <6>Time: tsc clocksource has been installed. <6>Time: acpi_pm clocksource has been installed. <6>intel8x0_measure_ac97_clock: measured 56454 usecs <6>intel8x0: clocking to 48000 <notice>boot.clock start Setting up the hardware clockdone <notice>'boot.clock start' exits with status 0 These are the messages from the last backup of 10.2; the file is dated November 1st, 2007 - on the 3nd I upgraded to 10.3 that same partition. nimrodel:/mnt/usb/usb_sg60/system/var/log # grep -i "clock\|tsc\|acpi_pm" boot.msg <6>Real Time Clock Driver v1.12ac <6>Time: tsc clocksource has been installed. <6>Time: acpi_pm clocksource has been installed. <6>intel8x0_measure_ac97_clock: measured 50928 usecs <6>intel8x0: clocking to 48000 <notice>boot.clock start Setting up the hardware clockdone <notice>'boot.clock start' exits with status 0 It did not complain of TSC being unstable. I will have to boot it to check the current state: the assumption would be that if it fails, it is hardware, if not, it is software; correct? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=350981
User ak@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c3
--- Comment #3 from Andi Kleen
https://bugzilla.novell.com/show_bug.cgi?id=350981
User robin.listas@telefonica.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c4
Carlos Robinson
https://bugzilla.novell.com/show_bug.cgi?id=350981
User robin.listas@telefonica.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c5
--- Comment #5 from Carlos Robinson
https://bugzilla.novell.com/show_bug.cgi?id=350981
User robin.listas@telefonica.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c6
--- Comment #6 from Carlos Robinson
https://bugzilla.novell.com/show_bug.cgi?id=350981
User robin.listas@telefonica.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c7
--- Comment #7 from Carlos Robinson
https://bugzilla.novell.com/show_bug.cgi?id=350981
User chrubis@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c8
--- Comment #8 from Cyril Hrubis
https://bugzilla.novell.com/show_bug.cgi?id=350981
User jeffm@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c10
Jeff Mahoney
https://bugzilla.novell.com/show_bug.cgi?id=350981
User robin.listas@telefonica.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c11
--- Comment #11 from Carlos Robinson
https://bugzilla.novell.com/show_bug.cgi?id=350981
User robin.listas@telefonica.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c12
--- Comment #12 from Carlos Robinson
https://bugzilla.novell.com/show_bug.cgi?id=350981
User robin.listas@telefonica.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c13
Carlos Robinson
https://bugzilla.novell.com/show_bug.cgi?id=350981
User robin.listas@telefonica.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c14
--- Comment #14 from Carlos Robinson
The clock shows no apparent problem here;meaning that it doesn't go slow. When I go to sleep I'll leave it on and check it on the morning.
No noticeable problem. ntpd was running, but has no peers - and the clock has no delay: minas-morgul:~ # rcntp status remote refid st t when poll reach delay offset jitter ============================================================================== *LOCAL(0) .LOCL. 10 l 23 64 377 0.000 0.000 0.001 Checking for network time protocol daemon (NTPD): running -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=350981
User robin.listas@telefonica.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c15
--- Comment #15 from Carlos Robinson
https://bugzilla.novell.com/show_bug.cgi?id=350981
User jeffm@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c16
--- Comment #16 from Jeff Mahoney
https://bugzilla.novell.com/show_bug.cgi?id=350981
User trenn@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c19
Thomas Renninger
https://bugzilla.novell.com/show_bug.cgi?id=350981
User robin.listas@telefonica.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c20
--- Comment #20 from Carlos Robinson
https://bugzilla.novell.com/show_bug.cgi?id=350981
User robin.listas@telefonica.net added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c21
Carlos Robinson
https://bugzilla.novell.com/show_bug.cgi?id=350981
User trenn@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c22
Thomas Renninger
https://bugzilla.novell.com/show_bug.cgi?id=350981
User nine@detonation.org added comment
https://bugzilla.novell.com/show_bug.cgi?id=350981#c23
Stefan Seifert
participants (1)
-
bugzilla_noreply@novell.com