Re: Seagate Dockstar, current and clock

["Jonathan A. Kollasch" <jakllsch%kollasch.net@localhost>]

On Thu, Jun 09, 2011 at 02:13:45PM +0200, Stefano Marinelli wrote:
> Hello everybody,
> yesterday I got a Seagate Dockstar. I compiled a evbarm system from -current, 
> adapted the SHEEVAPLUG kernel according to some information I got from this 
> mailing list, installed and everything is working properly.
> What I've seen is that even if I let ntpd daemon running, if there's a huge 
> I/O (like now, compiling some pkgsrc packages and the system is swapping on 
> the disk), the clock starts to lose time. It lost many seconds in 5 minutes.
> 
> More, here's the output of "ntpq -p":
> 
>      remote           refid      st t when poll reach   delay   offset  jitter
> ==============================================================================
>  tucano.isti.cnr 193.204.114.232  2 u   11   64  377   24.536  4526.44 3506.33
>  gamma.rueckgr.a 213.239.239.165  3 u    4   64  377   43.405  2328.81 4766.73
>  sd-21102.dedibo 2.128.0.0        6 u   58   64  377   55.609  4616.32 2987.79
>  hermes.develer. 193.204.114.233  2 u   50   64  377   26.046  4989.88 3070.85
> 
> The jitter gets very high and, more, I can't see the "*" and the "+" near the 
> server names, so I think it's not getting in sync anymore.
> 
> Any idea?

The timecounter implementation for marvell is faulty.
I should really commit my changes.

        Jonathan Kollasch