Re: rcu_preempt running flat out on idle desktop.

[Dave Jones]

On Wed, May 08, 2013 at 02:16:35PM -0700, Paul E. McKenney wrote:
 > On Wed, May 08, 2013 at 04:56:33PM -0400, Dave Jones wrote:
 > > On Wed, May 08, 2013 at 01:52:58PM -0700, Paul E. McKenney wrote:
 > >  > On Wed, May 08, 2013 at 04:30:42PM -0400, Dave Jones wrote:
 > > 
 > >  > >   PID USER      PR  NI  VIRT  RES  SHR S  %CPU %MEM    TIME+  COMMAND 
 > >  > >    10 root      20   0     0    0    0 S 200.0  0.0 185301:36 rcu_preempt
 > >  > >   553 root      20   0  268m  76m 6764 S 200.0  2.6 144579:53 Xorg
 > >  > >  1199 root      20   0     0    0    0 S 200.0  0.0 306:17.85 kworker/1:0
 > >  > >   501 root      20   0     0    0    0 S 200.0  0.0   4471:03 kworker/0:2
 > >  > >    12 root      20   0     0    0    0 S 200.0  0.0  67277:16 rcuop/1
 > >  > >  1237 davej     20   0  535m  15m 8484 S 200.0  0.5   3645:16 Terminal
 > >  > >   859 davej     20   0  117m 3036 1336 S 200.0  0.1   1986:55 htop
 > >  > > 
 > >  > > There are a lot of processes allegedly using "200%" of CPU time, a handful
 > >  > > in the "196%" range, and then a bunch at 0.
 > >  > 
 > >  > Yow!!!  185301 minutes is 128 days, which is a truly impressive amount of
 > >  > CPU time to accumulate in a few short hours.
 > > 
 > > perf top doesn't show anything out of the ordinary. Just as an idle desktop
 > > should behave, it's spending a bunch of time in delay_tsc.
 > > 
 > >  > This is 3.9, or Linus's current tree?  I am guessing the latter, but
 > >  > figured I should ask.
 > > 
 > > Yeah, the latter. (v3.9-11572-g5af43c2)
 > 
 > Do you have CONFIG_NO_HZ_FULL=y?  We just might have a bug in tickless
 > CPU-time accounting...

I never saw a new config option I didn't like.

CONFIG_NO_HZ_FULL=y
CONFIG_NO_HZ_FULL_ALL=y

	Dave

--
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/