Sometimes ago, I wrote:

> Hello,
>
> I have an embedded application running under Linux 2.6.22 #1 PREEMPT
> with 1000 ticks per second (PC architecture, AMD Geode Processor).
> Since precise timing is needed, the devices are synchronized to a time
> server via ntpd 4.2.2p4.
>
> After running 48 days and 13 hours without problems, things got crazy.
> The system time slowed down for a period of time, so that the system
> time lagged about three minutes to the real time. Some time later (an
> hour or so), ntp corrected the problem with a time jump. The device
> worked well again until 48 days and 13 hours after the first occurence
> of the problem the same thing happenend again. The system clock slowed
> down some 23 percent for a period of 13 minutes and 32 seconds (real
> time). This behavier was the same for the four devices I run, although
> the time when the problem began varied some 100 seconds between the
> devices and seemingly the subjektive time too. Again ntp corrected the
> problem after some time. The first time the problem occurred, the
> systems claimed clocksource tsc unstable and switched to clocksource
> pit. Didn't seem to make a difference.
>
> I have no idea, what the problem is. Does anybody know, what's going
> on and how to cure it?
>
> Regards, Volker Meyer

So here's the solution for the curious. You won't believe it. With the
Geode Processor
the bios can be active without the operating system knowing anything
about that. I
don't know how this works. Anyway, using a bios version without this
kind of
background activity solved the problem.

Greetings, Volker Meyer

_______________________________________________
questions mailing list
[email protected]
https://lists.ntp.org/mailman/listinfo/questions

Reply via email to