Yes, I agree that X may be in the way.  It is also possible, if you are
logging in via telnet or some other remote service that the I/O service
tasks don't have the needed priority.

My solution for X is to bump its priority too.

Attached, find rt, a program to allow you to change the priority of any
task, and getrt a simple reporting program to get the priority of any
task.

By the way, if you have a relatively secure system, and you want to use
telnet, you can set the priority of inetd.  This way all the I/O
services and the resulting session are boosted in priority.  You must do
this ahead of time, after your task goes into a loop, its too late to
boost anything, but once you do this, you should be able to telnet into
the machine even if your task is "lost" in some loop.

-g

Nicolas Bonnefon wrote:
> 
> > My application uses the possibility to change scheduling mechanism. When
> > changing from SCHED_OTHER to SCHED_FIFO or SCHED_RR, my application causes
> > the whole Linux PC to "hang". Prior to launching my own application, I
> > launch a "superbash" application with higher priority than my own
> > application. But I am not able to switch to the window where this
> > "superbash" application is running to kill my own application. The PC
> > simply does not respond to any input from the keyboard.
> 
> Have you tried without running X ?
> A SCHED_FIFO or SCHED_RR task has ALWAYS priority over any SCHED_OTHER
> task, so if your application sits in an infinite loop without blocking, the
> window manager cannot preempt it and you simply cannot switch x-term.
> I think you should carefully debug your app in SCHED_OTHER mode first, and
> then switch to some real-time policy.
> 
> >
> > I have used this mechanism before, and I am quite confident that this
> > worked under kernel 2.4.2. Are you aware of anything that could explain
> > this behaviour?
> 
> It would be surprising, but if you were sure of that, it would deserve some
> deeper investigations !
> 
> Regards
> --
> Nicolas Bonnefon
> Radar Development/Digital Processing Engineer
> Thales Air Defence - 7-9 rue des Mathurins
> 92223 Bagneux - France
> -
> 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/

-- 
George Anzinger   george@mvista.com
High-res-timers:  http://sourceforge.net/projects/high-res-timers/
Real time sched:  http://sourceforge.net/projects/rtsched/
Preemption patch: http://www.kernel.org/pub/linux/kernel/people/rml