Re: 2.6.1 "clock preempt"?

[timothy parkinson <t@timothyparkinson.com>]

On Fri, Jan 23, 2004 at 12:06:45PM -0800, john stultz wrote:
> On Fri, 2004-01-23 at 11:36, timothy parkinson wrote:
> > On Fri, Jan 23, 2004 at 11:17:29AM -0800, john stultz wrote:
> > > Well, lost ticks can be caused by many things, but your point is valid,
> > > the message could be a bit more elightening. 
> > 
> > googling for this issue turns up quite a few questions about it - there's
> > already one possible answer in the source, couldn't hurt to stick in a few
> > more:
> > 
> > 
> >       if (lost_count++ > 100) {
> >               printk(KERN_WARNING "Losing too many ticks!\n");
> >               printk(KERN_WARNING "TSC cannot be used as a timesource.\n"
> >                     "Are you running with SpeedStep?\n"
> > +                   "Perhaps you should enable DMA using \"hdparm\"?\n"
> > +                   "etc..........)\n");
> >               printk(KERN_WARNING "Falling back to a sane timesource.\n");
> >               clock_fallback();
> >       }
> > 
> > not that you have to actually listen to me or anything...  :-)
> 
> Looks good by me. Would you mind sending such a patch to Andrew?
> 
> thanks
> -john
> 

no problem, i'll take it from here.

* Running with SpeedStep (this is a cpu thing i assume?) could cause this.
* Not having DMA enabled on your hard disk(s) could cause this.  See the hdparm
  utility to enable it.
* Incorrect TSC synchronization on SMP systems could cause this.
* Anything else?

thanks,

timothy