[Xen-devel] RE: rdtsc strangeness on upstream kernel

[Dan Magenheimer <dan.magenheimer@xxxxxxxxxx>]

> > The number of rdtsc/second goes up dramatically when
> > there is CPU-intensive load on an upstream kernel!
> 
> Are you looking at rdtsc emulation traps when running a PV domain?

Yes, with "for i in {0..99}; do xm debug-key s; sleep 1; done".
 
> > I know we both observed some cases where rdtsc/sec
> > was very high, but I don't think we ever were able to
> > reproduce this consistently.
> 
> It would be interesting to compare that to the context switch rate (cs
> column in vmstat output) to see if they correlate.

Doesn't appear to be related.  Vmstat continues to show
cs at about 40/sec while the rdtsc/sec is very high.

> Also, how does it relate to timer interrupts?

With vcpus=2, timer0 goes up to about 2000 irq/sec on CPU0
and timer1 on CPU1 is the same while rdtsc/sec is about 20000.
Is it possible that the timer interrupt is using rdtsc
five times each interrupt?

> > First, this is a single vcpu, 64-bit 2.6.32 (RHEL6b2)
> > kernel.
> >
> > I am observing ~300 rdtsc/sec on an idle VM.  When
> > I run a load of:
> >
> > main() {while(1);}
> >
> > I am observing about 10000 rdtsc/sec!!
> >
> > This is a CONFIG_HZ_1000=y kernel, so I would
> > expect 1000 rdtsc/sec, or maybe 2000 rdtsc/sec,
> > but 10000 makes me wonder if there is some hidden
> > bug.
> 
> Do you have preemption running?  And why HZ=1000?

CONFIG_PREEMPT is off, though there are some other
PREEMPT-related configs that are on.  Are any others
relevant?

HZ=1000 is because this is a straight RHEL6b2 kernel,
but running as a PV guest.

Dan

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel