Clocks broken on Mac mini

[Hugh Blemings]

Hiya,

On 30/01/2017 19:56, Mark Atwood wrote:
> Hello Hugh,
>
> No need to be off list, we like to work in public as much as possible.

Cool cool, works for me :)

> Yes, please make those connections, and prod your friends on this issue.

No worries - by copy Hal - I apologise if I missed it but could you 
please remind me which version of Mac Mini this is ?  I think there was 
only the one CPU in the G4 based Minis - 7447A but other system 
components may have changed.

Please also confirm which Linux kernel version you're running (and 
FreeBSD for that matter though I lack any foo with that crowd :)

Any other info you think relevant would be grand - I'm still finding my 
way with ntpsec - Mark gave a particularly inspiring lightning talk at 
linux.conf.au earlier this month which prompted me to get involved so 
I'm very much a newbie :)

I've had a quick look at the ntp_loopfilter.c code and can see where the 
clock_step is occuring.  What I lack context on though is does it appear 
that on the Minis' the hardware clock/timebase is actually drifting too 
much (perhaps suggesting an issue with the underlying timebase init 
code) or is the kernel not allowing sufficiently large steps by calls to 
ntp_adjtime()

Or am I totally on the wrong track ?

In any case happy to gather the relevant data and summarise it into an 
email, CC the list to the kernel folk I know.

Cheers,
Hugh