Clock running fast on Mac Mini?

[Benjamin Herrenschmidt]

On Mon, 2006-04-24 at 23:37 +0200, Mich Lanners wrote:
> Hello all,
> 
> I have a problem on my Mac Mini with the kernel clock, which is running
> fast.
> 
> In fact, it is gaining 10 ms every 5 seconds, as can be seen here, with
> ntpdate called every 5 seconds:
> 
> owl:~# while true; do ntpdate -u be.pool.ntp.org; sleep 5; done
> 24 Apr 23:14:17 ntpdate[2601]: adjust time server 195.47.215.226 offset 0.151197 sec
> 24 Apr 23:14:22 ntpdate[2607]: adjust time server 195.47.215.226 offset 0.160348 sec
> 24 Apr 23:14:28 ntpdate[2613]: adjust time server 195.47.215.226 offset 0.170311 sec
> [...]
> 24 Apr 23:17:15 ntpdate[2813]: adjust time server 195.47.215.226 offset 0.489635 sec
> 24 Apr 23:17:20 ntpdate[2819]: adjust time server 195.47.215.226 offset 0.499386 sec
> 24 Apr 23:17:25 ntpdate[2825]: step time server 195.47.215.226 offset 0.509342 sec
> 24 Apr 23:17:31 ntpdate[2831]: adjust time server 195.47.215.226 offset 0.013693 sec
> 24 Apr 23:17:36 ntpdate[2837]: adjust time server 195.47.215.226 offset 0.021668 sec
> 
> and so on.
> 
> This is with ntpd stopped, after a fresh boot with /etc/adjtime removed
> to make sure hwclock is not falsly tuning the kernel timekeeping (I did
> this because I once had a 2.6.16 kernel running, where clock problems
> have been reported. I thought that may have falsified ntp's drift file
> and/or hwclock's /etc/adjtime).
> 
> All this is with a 2.6.15 kernel.
> 
> My Alu powerbook with a 2.6.15 kernel does not show this problem.

Could be several things... the clock calibration data from the
device-tree incorrect, or a rouding bug we fixed recently... I'd expect
2.6.16 to be good.

Ben