Subject: Re: [accounting regression since rc1]  scheduler updates
From: Martin Schwidefsky <schwidefsky@de.ibm.com>
Reply-To: schwidefsky@de.ibm.com
To: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Ingo Molnar <mingo@elte.hu>,
       Linus Torvalds <torvalds@linux-foundation.org>,
       Andrew Morton <akpm@linux-foundation.org>, linux-kernel@vger.kernel.org,
       Jan Glauber <jang@linux.vnet.ibm.com>, heiko.carstens@de.ibm.com,
       Paul Mackerras <paulus@samba.org>
In-Reply-To: <200708210900.54549.borntraeger@de.ibm.com>
References: <20070812163225.GA11996@elte.hu>
	 <20070820180810.GA25160@elte.hu> <1187634782.4028.10.camel@localhost>
	 <200708210900.54549.borntraeger@de.ibm.com>
Content-Type: text/plain
Organization: IBM Corporation
Date: Tue, 21 Aug 2007 11:18:46 +0200
Message-Id: <1187687926.7623.10.camel@localhost>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1917
Lines: 41

On Tue, 2007-08-21 at 09:00 +0200, Christian Borntraeger wrote:
> Am Montag, 20. August 2007 schrieb Martin Schwidefsky:
> > On Mon, 2007-08-20 at 20:08 +0200, Ingo Molnar wrote:
> > Ok, that would mean that sched_clock can just return the virtual cpu
> > time and the two hooks starts and stops the idle periods as far as the
> > scheduler is concerned. In this case we can use the patch from Jan with
> > the new implementation for sched_clock and add the two hooks to the
> > places where the cpu-idle notifiers are done (do_monitor_call and
> > default_idle). In fact this could be an idle-notifier. Hmm, I take a
> > closer look tomorrow when I'm back at the office.
> > 
> > > If a virtual CPU is idle then i think the "real = steal, virtual = 0" 
> > > way of thinking about idle looks a bit unnatural to me - wouldnt it be 
> > > better to think in terms of "steal = 0, virtual = real" ? Basically a 
> > > virtual CPU can idle at "perfect speed", without the host "stealing" any 
> > > cycles from it. And with that way of thinking, if s390 passed in the 
> > > real-idle-time value to the new callbacks below it would all fall into 
> > > place. Hm?
> 
> Martin,
> 
> I think we already do something like this. If you look at cpustat in 2.6.22 
> and earlier we already have steal increase = 0, idle increase = 100 % on an 
> idle cpu, even on s390. So while from the hardware perspective steal is 
> growing, we do the right thing in Linux, no?

This is done in kernel/sched.c:account_steal_time(). If the architecture
backend reports steal time for idle it is accounted as idle time.

-- 
blue skies,
  Martin.

"Reality continues to ruin my life." - Calvin.


-
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/