Re: 2.6.17-mm2 hrtimer code wedges at boot?

[john stultz]

On Sun, 2006-07-02 at 22:20 -0400, Valdis.Kletnieks@xxxxxx wrote:
> On Sun, 02 Jul 2006 18:56:22 PDT, Daniel Walker said:
> 
> > I was reviewing these new ntp adjustment functions, and it seems like it
> > would be a lot easier to just switch to a better clocksource. These new
> > functions seems to compensate for a clock that has a high rating but is
> > actually quite poor..
> 
> It is indeed poor - a few -mm's ago I tripped over a bug that kept it from
> recognizing the PM timer clocksource, and it refused to NTP sync because
> the clock drift was well outside the 500 ppm that NTP wants.  All the same,
> it's *one* thing for a clock to be drifting 10 seconds per hour.  It's
> something else to totally explode when handed a drifting clock.

Yes, the TSC is a very poor clocksource. Although its high resolution
and very quick access time makes it hard to quit. 

> Currently, the kernel is build with CONFIG_RTC=m, and the clock starts
> behaving as soom as rc.sysinit modprobes it.  Questions this raises:
> 
> 1) What's up with *that*?

Hmmm. Thats an interesting observation. Let me look to see if maybe
something is going oddly in the settimeofday() path.

> 2) Anybody want to place bets that building with CONFIG_RTC=y will make
> the clock work right off the bat?

No bets here, but please let us know if you see any change in behavior.

thanks
-john

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