Re: 2.6.17-rc3-mm1

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

On Sat, 2006-05-06 at 10:30 +0000, Thomas Gleixner wrote:
> Benoit,
> 
> On Fri, 2006-05-05 at 17:05 +0200, Benoit Boissinot wrote:
> > @@ -437,6 +438,12 @@ hrtimer_start(struct hrtimer *timer, kti
> >  
> >  	if (mode == HRTIMER_REL) {
> - 		tim = ktime_add(tim, new_base->get_time());
> +		ktime_t curr = new_base->get_time();
> +		
> +		tim = ktime_add(tim, curr);
> +
> 
> Can you change the debug that way? So we have the values which are
> added. Please print out new_base->id too.
> 
> > and when urxvtd hanged I had the following in dmesg:
> > [  356.696000] urxvtd: empty nanosleep 356726124322 17948911854451
> > 
> > So I suppose something is wrong in ktime_add()
> 
> Well, ktime_add is adding two 64 bit values.
> 
> The delta between the two values is 0xFFFFFFB3451. That looks like the
> timekeeping on your box is screwed by 0x100000000000. 
> 
> John, any idea ?

I suspect the system is falling back to the PIT instead of the ACPI PM
due to mis-merged patch in 2.6.17-rc3-mm1. The PIT has had a few reports
of problems, and I've got a test patch which I'm waiting for some
results on from a tester before sending (I can't reproduce the issue
locally).

Applying the patch here should get the ACPI PM working again.
http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=f0ec5e39765cd254d436a6d86e211d81795952a4;hp=30d55280b867aa0cae99f836ad0181bb0bf8f9cb


thanks
-john