linux-kernel - Re: [PATCH 6/7] KVM-GST: adjust scheduler cpu power

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Tue, 14 Jun 2011 22:26:24 -0300
From:	Glauber Costa <glommer@...hat.com>
To:	Peter Zijlstra <peterz@...radead.org>
CC:	kvm@...r.kernel.org, linux-kernel@...r.kernel.org,
	Rik van Riel <riel@...hat.com>,
	Jeremy Fitzhardinge <jeremy.fitzhardinge@...rix.com>,
	Avi Kivity <avi@...hat.com>,
	Anthony Liguori <aliguori@...ibm.com>,
	Eric B Munson <emunson@...bm.net>
Subject: Re: [PATCH 6/7] KVM-GST: adjust scheduler cpu power

On 06/14/2011 07:42 AM, Peter Zijlstra wrote:
> On Mon, 2011-06-13 at 19:31 -0400, Glauber Costa wrote:
>> @@ -1981,12 +1987,29 @@ static void update_rq_clock_task(struct rq
>> *rq, s64 delta)
>>
>>          rq->prev_irq_time += irq_delta;
>>          delta -= irq_delta;
>> +#endif
>> +#ifdef CONFIG_PARAVIRT_TIME_ACCOUNTING
>> +       if (static_branch((&paravirt_steal_rq_enabled))) {
>
> Why is that a different variable from the touch_steal_time() one?

because they track different things, touch_steal_time() and 
update_rq_clock() are
called from different places at different situations.

If we advance prev_steal_time in touch_steal_time(), and later on call 
update_rq_clock_task(), we won't discount the time already flushed from 
the rq_clock. Conversely, if we call update_rq_clock_task(), and only 
then arrive at touch_steal_time, we won't account steal time properly.

update_rq_clock_task() is called whenever update_rq_clock() is called.
touch_steal_time is called every tick. If there is a causal relation 
between them that would allow us to track it in a single location, I 
fail to realize.

>> +
>> +               steal = paravirt_steal_clock(cpu_of(rq));
>> +               steal -= rq->prev_steal_time_acc;
>> +
>> +               rq->prev_steal_time_acc += steal;
>
> You have this addition in the wrong place, when you clip:

I begin by disagreeing
>> +               if (steal>  delta)
>> +                       steal = delta;
>
> you just lost your steal delta, so the addition to prev_steal_time_acc
> needs to be after the clip.
Unlike irq time, steal time can be extremely huge. Just think of a 
virtual machine that got interrupted for a very long time. We'd have 
steal >> delta, leading to steal == delta for a big number of iterations.
That would affect cpu power for an extended period of time, not
reflecting present situation, just the past. So I like to think of delta
as a hard cap for steal time.

Obviously, I am open to debate.
>
>> +               delta -= steal;
>> +       }
>> +#endif
>> +
>>          rq->clock_task += delta;
>>
>> -       if (irq_delta&&  sched_feat(NONIRQ_POWER))
>> -               sched_rt_avg_update(rq, irq_delta);
>> +       if ((irq_delta + steal)&&  sched_feat(NONTASK_POWER))
>> +               sched_rt_avg_update(rq, irq_delta + steal);
>>   }

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/