Re: How how latent should non-preemptive scheduling be?

[Ingo Molnar <mingo@elte.hu>]

* Sitsofe Wheeler <sitsofe@yahoo.com> wrote:

> Peter Zijlstra wrote:
>> What you can do to investigate this, is use the sched_wakeup tracer from
>> ftrace, that should give a function trace of the highest wakeup latency
>> showing what the kernel is doing.
>
> I struggled to find documentation of ftrace because it's quite new. I  
> have come across  
> http://www.redhat.com/docs/en-US/Red_Hat_Enterprise_MRG/1.0/html/Realtime_Tuning_Guide/sect-Realtime_Tuning_Guide-Realtime_Specific_Tuning-Using_the_ftrace_Utility_for_Tracing_Latencies.html 
> and  
> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=Documentation/ftrace.txt;h=d330fe3103da9c9a3cb8f888ac7255ce48e666d4;hb=45e9c0de2e86485f8b6633fd64ab19cfbff167f6 
>  .
>
> Thanks to those I started up the debugfs filesystem and went to the  
> trace directories but the on tracers in available_tracers are
> ftrace sched_switch none
>
> I can't see anything in the code that would disable wakeup... Any ideas  
> on what might be wrong? I'm using a 2.6.27rc6 kernel.

here's two quick howtos:

  http://redhat.com/~mingo/sched-devel.git/readme-tracer.txt
  http://redhat.com/~mingo/sched-devel.git/howto-trace-latencies.txt

you need to enable:

  CONFIG_SCHED_TRACER=y
  CONFIG_CONTEXT_SWITCH_TRACER=y

it's not particularly well named though. Why doesnt it say 
LATENCY_TRACER or something?

> Additionally I think I found a trigger - unplugging the power cable 
> from the EeePC and having it run on battery seems to then set off this 
> periodic stall every 30 seconds... There's no CPU frequency scaling 
> enabled either (Celeron M's seemingly don't have P states and support 
> for cpufreq is configured out).

sounds like potential SMM triggered latencies.

	Ingo