Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Thu, 5 Sep 2019 13:47:09 +0200
From:   Peter Zijlstra <peterz@infradead.org>
To:     Patrick Bellasi <patrick.bellasi@arm.com>
Cc:     Qais Yousef <qais.yousef@arm.com>,
        Subhra Mazumdar <subhra.mazumdar@oracle.com>,
        linux-kernel@vger.kernel.org, mingo@redhat.com, tglx@linutronix.de,
        steven.sistare@oracle.com, dhaval.giani@oracle.com,
        daniel.lezcano@linaro.org, vincent.guittot@linaro.org,
        viresh.kumar@linaro.org, tim.c.chen@linux.intel.com,
        mgorman@techsingularity.net, parth@linux.ibm.com
Subject: Re: [RFC PATCH 1/9] sched,cgroup: Add interface for latency-nice
Message-ID: <20190905114709.GM2349@hirez.programming.kicks-ass.net>
References: <20190830174944.21741-1-subhra.mazumdar@oracle.com>
 <20190830174944.21741-2-subhra.mazumdar@oracle.com>
 <20190905083127.GA2332@hirez.programming.kicks-ass.net>
 <87r24v2i14.fsf@arm.com>
 <20190905104616.GD2332@hirez.programming.kicks-ass.net>
 <20190905111346.2w6kuqrdvaqvgilu@e107158-lin.cambridge.arm.com>
 <87h85r2d5f.fsf@arm.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <87h85r2d5f.fsf@arm.com>
User-Agent: Mutt/1.10.1 (2018-07-13)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, Sep 05, 2019 at 12:30:52PM +0100, Patrick Bellasi wrote:

> I see this concept possibly evolving into something more then just a
> binary switch. Not yet convinced if it make sense and/or it's possible
> but, in principle, I was thinking about these possible usages for CFS
> tasks:
> 
>  - dynamically tune the policy of a task among SCHED_{OTHER,BATCH,IDLE}
>    depending on crossing certain pre-configured threshold of latency
>    niceness.

A big part of BATCH is wakeup preemption (batch doesn't preempt itself),
and wakeup preemption is a task-task propery and can thus be completely
relative.

>  - dynamically bias the vruntime updates we do in place_entity()
>    depending on the actual latency niceness of a task.

That is dangerous; theory says we should keep track of the 0-lag point
and place it back where we found it. BFQ does this correctly IIRC, but
for CFS I've never done that because 'expensive'.

But yes, we could (carefully) fumble a bit there.

>  - bias the decisions we take in check_preempt_tick() still depending
>    on a relative comparison of the current and wakeup task latency
>    niceness values.

Ack.

Placing relative and absolute behaviour on the same scale is going to be
'fun' :-)