Re: SCHED_RR preempted by SCHED_OTHER

["Luis Claudio R. Goncalves" <lclaudio@uudg.org>]

On Mon, Jul 27, 2009 at 02:36:09PM +0200, Uwe Kleine-König wrote:
| On Mon, Jul 27, 2009 at 01:57:31PM +0200, Denis Richard wrote:
| > Hi all,
| > We work on powerpc board with kernel 2.6.29.6-rt23.
| > The kernel is configured with  "Complete Preemption (Real-Time)".
| > Working on our software, we have a strange behavior.
| > A real time thread (SCHED_RR) can be preempted by non real time one  
| > (SCHED_OTHER).
| >
| > We wrote a little test program (at the end of  this mail).
| > This program creates 2 threads, a SCHED_RR one which processes during  
| > about 12s
| > (long for a real time, but just for test), and a SCHED_OTHER one,  
| > started after the first one,
| > which just modifies a global variable. The second one should not run  
| > during the execution of
| > the first one. But it appends, the global variable is modified.
| >
| > > rr_other
| > Create thread
| > OTHER ran during RR
| > The end
| > >
| >
| > Is it normal, a non real time thread preempts a real time one ? (we  
| > think, it is not)
| > Or is there something wrong in the test program, the kernel  
| > configuration, ... ?
| Does
| 
| 	# echo -1 > /proc/sys/kernel/sched_rt_runtime_us
| 
| help?

Also be prepared to find a WARNING on the kernel logs as the kernel gets
worried everytime a CPU runs the same process for more than 10s.

If Uwe's suggestion doesn't solve the issue, you may want to run the
sched_switch tracer from ftrace. Steven Rostedt recently posted a userspace
application to make this task easier.

Have fun,
Luis
-- 
[ Luis Claudio R. Goncalves                    Bass - Gospel - RT ]
[ Fingerprint: 4FDD B8C4 3C59 34BD 8BE9  2696 7203 D980 A448 C8F8 ]

--
To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html