lkml.org 
[lkml]   [2008]   [Apr]   [23]   [last100]   RSS Feed
Views: [more markup]  [less markup]  [headers]  [forward] 
 
Messages in this thread
/
DateWed, 23 Apr 2008 11:46:44 +0200
FromIngo Molnar <>
SubjectRe: [git pull] scheduler changes for v2.6.26
* Frans Pop <elendil@planet.nl> wrote:

> If I enable tracing with these settings, I get data in both trace 
> *and* in latency_trace. Is the last correct? From 
> http://lkml.org/lkml/2008/2/10/43 I got the impression that file 
> should only be used if the "wakeup" tracer is active.

yeah, that's the intended behavior. 'trace' and 'latency_trace' are two 
kinds of 'views' on the same set of tracer data (with a different output 
format). They are both non-destructive - i.e. they dont deplete the 
trace buffers.

there's even a third view: 'trace_pipe' - which depletes the trace 
buffers and can be used to pipe trace data into a larger file:

    cat trace_pipe > /tmp/large-trace.txt
i guess it would be less confusing if 'trace' and 'latency_trace' was 
just a single 'view' and its behavior depended on iter_ctl settings. (In 
fact, we could probably just get rid of latency_trace - it is a leftover 
from -rt.)

> From the last run I've got three traces with 50000 entries (about a 
> minute worth each). Traces 1 and 2 should each have two skips and 
> trace 3 should have four. I've saved both the trace and latency_trace 
> (ltrace) files.
> 
> They are available at: 
> http://people.debian.org/~fjp/tmp/kernel/sched/.

thanks, we'll have a look.

> BTW, did either of you actually look at the traces I sent for .24? I 
> never got any feedback on those.

hm, i looked at them - i thought i mailed you about that but indeed i 
didnt (sorry about that) - the traces were too short in scope - they 
just covered a few dozen milliseconds in scope, while your description 
said that the problem occured seconds into the trace.

> P.S. I've got group scheduling active in this config as my tests with 
> .24 showed that did not make any difference. Can rerun without if 
> needed.

no need to rerun - but if you do the next test it might make sense to 
disable it, just to reduce the number of variables.

another thing - Peter observed skipping on NOHZ. Does your skipping go 
away if you boot with 'nohz' and/or with CONFIG_NOHZ=y disabled in the 
.config?

	Ingo


\
 
 \ /
  Last update: 2008-04-23 11:49    [from the cache]
©2003-2008