Issues getting IO times using Schedtimes.stp example script

Issues getting IO times using Schedtimes.stp example script

[Sunny Fugate]

I'm having some difficulties in getting one of the examples scripts to 
run.  According to the documentation, the kernel.trace("sched_switch") 
probe should expose $prev, $next, and $rq.  I've played around with the 
stp script and on my system the $rq variable is never defined.  As such, 
the script never reports io scheduling times.  Other results from the same script seem ok. 

Anyone have an idea what would cause my issues with the schedtimes.stp 
example script?  

I'm running on a recent Ubuntu 11.04. 

Thanks,

Sunny

Re: Issues getting IO times using Schedtimes.stp example script

[Josh Stone]

On 01/30/2012 03:13 PM, Sunny Fugate wrote:
> I'm having some difficulties in getting one of the examples scripts to 
> run.  According to the documentation, the kernel.trace("sched_switch") 
> probe should expose $prev, $next, and $rq.  I've played around with the 
> stp script and on my system the $rq variable is never defined.

There used to be a rq parameter, but it was removed from the kernels
2.6.35+ by commit 27a9da65:

http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commitdiff;h=27a9da6538ee18046d7bff8e36a9f783542c54c3

I don't know any other way off-hand to get at the runqueue -- it is held
very secretively in sched internals.

Josh

Re: Issues getting IO times using Schedtimes.stp example script

[Sunny Fugate]

Thanks for the help....

The justification for the kernel change is odd....apparently internal kernel features disappear if nobody is using them in an obvious manner. 


Cheers, 

-Sunny

On Jan 30, 2012, at 4:48 PM, Josh Stone wrote:

> On 01/30/2012 03:13 PM, Sunny Fugate wrote:
>> I'm having some difficulties in getting one of the examples scripts to 
>> run.  According to the documentation, the kernel.trace("sched_switch") 
>> probe should expose $prev, $next, and $rq.  I've played around with the 
>> stp script and on my system the $rq variable is never defined.
> 
> There used to be a rq parameter, but it was removed from the kernels
> 2.6.35+ by commit 27a9da65:
> 
> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commitdiff;h=27a9da6538ee18046d7bff8e36a9f783542c54c3
> 
> I don't know any other way off-hand to get at the runqueue -- it is held
> very secretively in sched internals.
> 
> Josh

Re: Issues getting IO times using Schedtimes.stp example script

[Timo Juhani Lindfors]

Sunny Fugate <fugate@unm.edu> writes:
> I'm running on a recent Ubuntu 11.04. 

I'm aware of this, the issue is tracked as

"examples/profiling/sched_switch.stp: semantic error: not accessible at
this address (0xffffffff812f9f86): identifier '$prev' .."

in http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=604761

Re: Issues getting IO times using Schedtimes.stp example script

[David Smith]

On 01/30/2012 05:48 PM, Josh Stone wrote:

> On 01/30/2012 03:13 PM, Sunny Fugate wrote:
>> I'm having some difficulties in getting one of the examples scripts to 
>> run.  According to the documentation, the kernel.trace("sched_switch") 
>> probe should expose $prev, $next, and $rq.  I've played around with the 
>> stp script and on my system the $rq variable is never defined.
> 
> There used to be a rq parameter, but it was removed from the kernels
> 2.6.35+ by commit 27a9da65:
> 
> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commitdiff;h=27a9da6538ee18046d7bff8e36a9f783542c54c3
> 
> I don't know any other way off-hand to get at the runqueue -- it is held
> very secretively in sched internals.


I wonder if we couldn't set a kprobe on finish_task_switch() (which gets
an rq parameter).  We'd have to match the calls up, but it might be
possible.

-- 
David Smith
dsmith@redhat.com
Red Hat
http://www.redhat.com
256.217.0141 (direct)
256.837.0057 (fax)