[ntp:questions] Large and apparently permanent step in PPM error - confused?

Hal Murray hal-usenet at ip-64-139-1-69.sjc.megapath.net
Sat Nov 1 22:39:50 UTC 2008


In article <490BA73F.7070500 at dcureton.com>,
 david.cureton at dcureton.com (David Cureton) writes:
>Hi,
>    I am syncing a Linux 2.6.26 kernel to a Serial DCD PPS GPS source.
>The kernel time  has been running for quite some time with a ~160ppm 
>frequency error however inexplicably the ppm error over night simply 
>jumped to ~105ppm. The system  has for the last couple of day tracked in 
>a stable manner at that ~105ppm rather than the ~160ppm offset it used 
>to be.
>
>There has been no change to the system that I am aware of. The machine 
>was not rebooted. There has been no dramatic change in temperature 
>although I think temperature change would be very dramatic cause 55ppm jump.
>(In fact you can see the daily temperature variation in the ppm error 
>which is in the order of a couple of ppm over time)

50 ppm is a huge change.  That's either something quite rare in
the hardware or an interesting software quirk.

Is there anything interesting in syslog at about that time?
Did it get warm and shift to a slower CPU frequency?

There is a bug in the Linux 2.6 kernels.  The TSC calibration
routine doesn't get the same answer each time.  So you could get
something like that if you rebooted.  But you said it wasn't
rebooted so this seems unlikely.

How long has the system been up?  I wonder if some parameter
finally overflowed N bits and is now taking a slightly different
path through some code.

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




More information about the questions mailing list