Re: ntp problems

[john stultz <johnstul@us.ibm.com>]

On Mon, 2005-12-05 at 21:07 -0500, Gene Heskett wrote:
> On Monday 05 December 2005 19:14, john stultz wrote:
> >On Mon, 2005-12-05 at 18:33 -0500, Gene Heskett wrote:
> >> On Monday 05 December 2005 16:39, john stultz wrote:
> >> >On Mon, 2005-12-05 at 00:31 -0500, Gene Heskett wrote:
> >> >> Greetings everybody;
> >> >>
> >> >> I seem to have an ntp problem.  I noticed a few minutes ago that
> >> >> if my watch was anywhere near correct, then the computer was about
> >> >> 6 minutes fast.  Doing a service ntpd restart crash set it back
> >> >> nearly 6 minutes.
> >> >
> >> >Not sure exactly what is going on, but you might want to try
> >> > dropping the LOCAL server reference in your ntp.conf. It could be
> >> > you're just syncing w/ yourself.
> >>
> >> Joanne, bless her, pointed out that I had probably turned the ACPI
> >> stuff in my kernel back on.  She was of course correct, shut it off &
> >> ntpd works just fine.
> >
> >Err. ACPI stuff? Could you elaborate? Sounds like you have some sort of
> >bug hiding there.
> 
> This has been a relatively long standing problem, John.  I think its 
> possibly related to some access path in the nforce2 chipset as it seems 
> to plague that chipset worse than others.  But its long been, and I had 
> forgotten, that if ntpd didn't work, turning off the ACPI stuff was the 
> fix.

Hey Gene,

	I know we've spoken before about a few timekeeping problems, but
sometimes I have a hard time keeping all the issues straight. Have you
filed a bug on this issue? I queried for your email in
bugzilla.kernel.org and I couldn't find anything. If not, would you mind
opening a bug describing the behavior you are seeing with the different
boot options (possibly also dmesg output for both configurations)? It
will greatly help make sure this doesn't slip by without notice.


> It had worked for a few kernel.org kernels and I had become complacent.  
> My mistake.
> 
> OTOH, calling it a local bug, no, I certainly wouldn't call it a local to 
> my machine bug.  Jdow OTOH, running an FC4 box, has it enabled, and hers 
> is working just fine.  She is I believe, running the FC4's latest kernel 
> too, so maybe the redhat people have massaged it.  However, at one time 
> several months ago I believe she also had to have a grub argument 
> turning acpi=no.

Hmmm. Indeed the nforce2 has had a number of problems, but I'm not sure
why it would have changed recently. Can you bound at all the kernel
versions where it worked and where it broke? Additionally, do be sure
you have the most recent BIOS, I've seen a number of nforce2 issues be
resolved with a BIOS update.


> There was a bunch of ntp related patches submitted recently, and I have 
> no idea which of them may have restored the broken acpi vs ntp scenario 
> to its formerly broken status, again.

I don't think the NTP changes would have triggered this, but I want to
be sure (its more likely something else in the timekeeping code is
causing problems).

> Should it be looked at?  Certainly, but I don't have the knowledge to do 
> so.  So I build kernels, and report problems areas.  The canary in the 
> coal mine so to speak. :-)

The testing and problem report is very much appreciated! Start with
filing a bugzilla bug and then I'll point you at a few more mines to
checkout :)

thanks
-john


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/