Message-ID: <alpine.DEB.2.00.0905201433060.1080@gandalf.stny.rr.com>
Date:	Wed, 20 May 2009 14:59:54 -0400 (EDT)
From:	Steven Rostedt <rostedt@...dmis.org>
To:	GeunSik Lim <leemgs1@...il.com>
cc:	Ingo Molnar <mingo@...e.hu>,
	Peter Zijlstra <a.p.zijlstra@...llo.nl>,
	tglx <tglx@...utronix.de>,
	linux-kernel <linux-kernel@...r.kernel.org>,
	linux-rt-users <linux-rt-users@...r.kernel.org>
Subject: Re: [PATCH 0/1] sched: fix typo in sched-rt-group.txt file


On Wed, 20 May 2009, GeunSik Lim wrote:

> 2009/5/20 GeunSik Lim <leemgs1@...il.com>:
> >
> >
> > Dear Ingo,
> >
> >
> > I made 2 patch files again about static prioriy levels according to ingo's advice
> > and explanation. (Reference: http://marc.info/?t=124211992800003&r=1&w=2 archives. )
> >
> > Fix static priority related range and chart map(userspace/kernelspace) of ftrace.
> >   - about chart map of static priority in  ftrace.txt file
> >   - about static priority levels(range) in sched-rt-group.txt file
> >
> >
> > commit d6af702008117ca489a2f5476239d1688a517e98
> > Author: GeunSik,Lim <leemgs1@...il.com>
> > Date:   Tue May 12 17:20:32 2009 +0900
> >
> >    sched: fix typo in sched-rt-group.txt file
> >
> >    Fix typo about static priority's range.
> >
> >    * Kernel Space priority 0(high) to 99(low)   --> User Space RT priority 99(high) to 1(low)
> >    * Kernel Space priority 100(high) to 139(low)--> User Space nice -20(high) to 19(low)
> >
> 
> Ingo,
> I want to  change some descriptions according to below discussion.
> 
> before) * Kernel Space priority 0(high) to 99(low)   --> User Space RT
> priority 99(high) to 1(low)
> after)    *  Kernel Space priority 0(high) to 98(low)   --> User Space
> RT priority 99(high) to 1(low)
> 
> I want to your advice and opinion about this changes.
> 
> If I mistook, Please correct me.
> 
> thks.
> 
> 
> 
> Steven Rostedt:
>   Shouldn't that be:
>   Kernel priority: 0(high) to 98(low)    ==> user RT priority 99(high) to 1(low)
> 
> GeunSik Lim :
>   Thank you for your opinion.  I think about corrected expression for
>   quick chart to map between kernel space and user space in ftrace .
> 
> Steven Rostedt:
>  100 items does not map to 99. Which begs the question, what can have
>  internal kernel priority 99?
> 
> GeunSik Lim :
>   ok.  Below example is about  kernel (static) priority 99 that you said.
>   please, refer
>   to http://www.kernel.org/doc/man-pages/online/pages/man2/sched_setscheduler.2.html
>  webpage.
> 
>   For processes scheduled under one of the normal scheduling policies
>   (SCHED_OTHER, SCHED_IDLE, SCHED_BATCH), sched_priority is not used in
>   scheduling decisions (it must be specified as 0).
> 
> int main (int argc, char **argv)
> {	int pidnumber, ret;
> 	struct sched_param p;
> 	p.sched_priority = 0; <------ Kernel (static) priority.
> 	if (argc != 2) {
> 		printf("usage: setbatch <pid>\n");
> 		exit(-1);
> 	}
> 	pidnumber = atol(argv[1]);
>          /* 3 = Number of SCHED_BATCH Constant */
> 	ret = sched_setscheduler(pidnumber, 3, &p);
> 	if (ret) {
> 		printf("could not set pid %d to SCHED_BATCH: err %d.\n", pid, ret);
> 		return -1;
> 	}
> 	printf("pid %d is SCHED_BATCH from now on.\n", pid);
> 	return 0;
> }
> 
> 
> >
> >    Processes scheduled with SCHED_OTHER or SCHED_BATCH must be assigned
> >    the static priority 0. Processes scheduled under SCHED_FIFO or SCHED_RR
> >    can have a static priority in the range 1 to 99.
> >    (reference: $> man 2 sched_setscheduler)
> >
> >    Signed-off-by: GeunSik Lim <geunsik.lim@...sung.com>
> >
> > diff --git a/Documentation/scheduler/sched-rt-group.txt b/Documentation/scheduler/sched-rt-group.txt
> > index 5ba4d3f..1537146 100644
> > --- a/Documentation/scheduler/sched-rt-group.txt
> > +++ b/Documentation/scheduler/sched-rt-group.txt
> > @@ -169,7 +169,7 @@ get their allocated time.
> >
> >  Implementing SCHED_EDF might take a while to complete. Priority Inheritance is
> >  the biggest challenge as the current linux PI infrastructure is geared towards
> > -the limited static priority levels 0-139. With deadline scheduling you need to
> > +the limited static priority levels 0-99. With deadline scheduling you need to
> >  do deadline inheritance (since priority is inversely proportional to the
> >  deadline delta (deadline - now).

Note, this document deals with internal kernel manipulations. The "prio" 
inside the kernel is indeed 0-139. Actually, if we count the idle task, it 
is 0-140. The PI code will also deal with non RT tasks so it is correct to 
say that the priority inheritance deals with 0-139 (I don't think the idle 
task ever touches anything needing PI).

The sched_setscheduler user API uses a different priority mapping. I was 
looking at the code inside sched.c and I believe this is the mapping:


0 to 98 - maps to RT tasks 99 to 1 (SCHED_RR or SCHED_FIFO)

99 - maps to internal kernel threads that want to be lower than RT tasks 
     but higher than SCHED_OTHER tasks. Although I'm not sure if any 
     kernel thread actually uses this. I'm not even sure how this can be
     set, because the internal sched_setscheduler function does not allow
     for it.

100 to 139 - maps nice levels -20 to 19. These are not set via 
    sched_setscheduler, but are set via the nice system call.

140 - reserved for idle tasks.

-- Steve

Powered by blists - more mailing lists