lkml.org 
[lkml]   [2006]   [May]   [15]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: rt20 scheduling latency testcase and failure data
Date
On Monday 15 May 2006 04:20, Sébastien Dugué wrote:
> On Sat, 2006-05-13 at 11:06 -0700, Darren Hart wrote:
> > These tests are running on a 4 way opteron at 2 GHz with 4 GB RAM. I
> > have attached the .config and a listing of all the rt tasks running on
> > the system (which addresses the questions regarding priority setup, IRQ
> > handlers, and softirqs - all default). I am running with the futex
> > priority based wakeup patches from Sebastien Duque, but I don't think
> > this test excercises those paths.
>
> Which watchdog are you using here? Have you tried without the
> watchdog?

Those are the softlockup watchdog threads (kernel/softlockup.c). They run
once a second and reports a bug if the watchdog failed to run in 10 seconds.
It is difficult to reproduce but at run 126 (1,260,000 iterations) it finally
failed. Note that I am only counting runs that completely miss an entire
period as a failure for the purposes of this test. I want to knock out the
10+ms latencies before I concern myself too much with the >100us failures :-)

--------------------
ITERATION 26
--------------------
-------------------------------
Scheduling Latency
-------------------------------

Running 10000 iterations with a period of 5 ms
Expected running time: 50 s

ITERATION DELAY(US) MAX_DELAY(US) FAILURES
--------- --------- ------------- --------


PERIOD MISSED!
scheduled delta: 4076 us
actual delta: 14892 us
latency: 10815 us
---------------------------------------
previous start: 18365818 us
now: 18366739 us
scheduled start: 18360000 us
next scheduled start is in the past!


Start Latency: 106 us: FAIL
Min Latency: 8 us: PASS
Avg Latency: 4 us: PASS
Max Latency: 10818 us: FAIL
Failed Iterations: 1


It's interesting, this 10ms latency seems to be the most common result. I'm
going to take a look at ingo's tracing script now, more info a bit later...

--
Darren Hart
IBM Linux Technology Center
Realtime Linux Team

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2006-05-15 23:51    [W:0.120 / U:1.816 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site