Mike Galbraith wrote:

On Sat, 2011-08-13 at 09:27 -0700, Paul E. McKenney wrote:

quoted

Strange.  By the time it got around to printing the stall, no one was
stalling:

[  100.409012] INFO: rcu_preempt_state detected stalls on CPUs/tasks: {} (detected by 21, t=60002 jiffies)

Now it -is- possible for the stall to end just as we get ready to detect
it, but that window is really really small.  The most recent occurrence
of this sort of thing was due misconfigured timekeeping, but I don't see
any sign of that in the trace.

This happens repeatedly?

Yeah.  When I put -rt9 on the UV box, it was doing the just idling, and
boom thing I saw with earlier releases on both my Q6600 and x3550 M3
boxes.  -rt10 seems to have just changed the timing.

I'll update the x3550 M3, and see if it still gripes.  UV isn't virgin
source, has to be twiddled a little. (symptom was identical though)

Mike,

Which test were you running? Because I'm not seeing it
on our x3550 M3..

thanks,
Nivedita