Re: sched_setscheduler() vs idle_balance() race

[Peter Zijlstra]

On Fri, May 29, 2015 at 08:48:56PM +0200, Mike Galbraith wrote:
> 
> P.S. intel_idle is not all that wonderful on this box.
> 
> -   78.31%  [kernel]       [k] _raw_spin_lock                                                                                                                                                                                              â
>    - _raw_spin_lock                                                                                                                                                                                                                        â
>       - 94.91% tick_broadcast_oneshot_control                                                                                                                                                                                              â

Your DL980 G7 has E7-4800 parts in, right? Which if Wikipedia is
correct, resolves to a Nehalem-EX.

Now the NHM-EX has a fun 'feature' that for (some?) idle states the
local timer stops, so we have to revert back to a global broadcast
timer.

Now go count the number of cpus on your box and then imagine a global
spinlock, oh wait, you already found it ^ :-)
--
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/