Re: [REGRESSION] sudden cpu hogging in kernel 3.2 rc-series

[Mike Galbraith <efault@gmx.de>]

On Tue, 2011-11-22 at 09:19 +0100, Nicolas Kalkhof wrote:
> Hi Mike,
> 
> thanks for your reply! I didn't set any polling explicitly.
> 
> My boot params are:
> intel_iommu=off pcie_aspm=force i915.powersave=1 i915.i915_enable_rc6=1
> 
> and in the cpufrequtils startup daemon the only parameter set is:
> # Options when starting cpufreq (given to the `cpufreq-set` program)
> START_OPTS="--governor ondemand"
> 
> egrep "IDLE|GOV" .config yields:
> 
> CONFIG_ARCH_HAS_CPU_IDLE_WAIT=y
> CONFIG_ARCH_HAS_DEFAULT_IDLE=y
> # CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE is not set
> # CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE is not set
> CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND=y
> # CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE is not set
> CONFIG_CPU_FREQ_GOV_PERFORMANCE=y
> CONFIG_CPU_FREQ_GOV_POWERSAVE=y
> CONFIG_CPU_FREQ_GOV_USERSPACE=y
> CONFIG_CPU_FREQ_GOV_ONDEMAND=y
> CONFIG_CPU_FREQ_GOV_CONSERVATIVE=y
> CONFIG_CPU_IDLE=y
> CONFIG_CPU_IDLE_GOV_LADDER=y
> CONFIG_INTEL_IDLE=y
> # CONFIG_I7300_IDLE is not set
> 
> Any advice?

Not really.  If you apply the below, then..

make -C tools/perf prefix=/usr/local install
perf top -U

..does perf top agree with oprofile?

diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index c9cdedb..28ef86d 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -636,7 +636,6 @@ static const char *skip_symbols[] = {
 	"exit_idle",
 	"mwait_idle",
 	"mwait_idle_with_hints",
-	"poll_idle",
 	"ppc64_runlatch_off",
 	"pseries_dedicated_idle_sleep",
 	NULL