Received-SPF: pass (google.com: domain of linux-kernel+bounces-181252-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.48.161 as permitted sender) client-ip=147.75.48.161;
Date: Thu, 16 May 2024 17:32:56 +0200
From: Frederic Weisbecker <frederic@kernel.org>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Yun Levi <ppbuk5246@gmail.com>, Joel Fernandes <joel@joelfernandes.org>,
	Vineeth Pillai <vineeth@bitbyteword.org>,
	Vincent Guittot <vincent.guittot@linaro.org>,
	Dietmar Eggemann <dietmar.eggemann@arm.com>,
	anna-maria@linutronix.de, mingo@kernel.org, tglx@linutronix.de,
	Markus.Elfring@web.de, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v4] time/tick-sched: idle load balancing when nohz_full
 cpu becomes idle.
Message-ID: <ZkYnKAd1Qy+yvjDY@lothringen>
References: <20240516075628.GC22557@noisy.programming.kicks-ass.net>
 <CAM7-yPRHp3tiZjuBTesdRQoU8WJNg1scon_txS_6R-pZq9MXHw@mail.gmail.com>
 <20240516084911.GF22557@noisy.programming.kicks-ass.net>
 <ZkXtHv+fHUD2+lFJ@lothringen>
 <CAM7-yPTSq0CSmRsTpeXwzhFk77gfwUK_LZKnbgo4NPk5zPCaAg@mail.gmail.com>
 <20240516140003.GJ22557@noisy.programming.kicks-ass.net>
 <ZkYW48dTX2FH5NaD@lothringen>
 <20240516144504.GL22557@noisy.programming.kicks-ass.net>
 <ZkYgG9KYMpUPeJsM@lothringen>
 <20240516151953.GM22557@noisy.programming.kicks-ass.net>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20240516151953.GM22557@noisy.programming.kicks-ass.net>

On Thu, May 16, 2024 at 05:19:53PM +0200, Peter Zijlstra wrote:
> On Thu, May 16, 2024 at 05:02:51PM +0200, Frederic Weisbecker wrote:
> 
> > > I'm confused, none of that makes sense. If you're part of a
> > > load-balancer, you're part of a load-balancer, no ifs buts or other
> > > nonsense.
> > > 
> > > idle load balancer is no different from regular load balancing.
> > > 
> > > Fundamentally, you can't disable the tick if you're part of a
> > > load-balance group, the load-balancer needs the tick.
> > > 
> > > The only possible way to use nohz_full is to not be part of a
> > > load-balancer, and the only way that is so is by having (lots of) single
> > > CPU partitions.
> > 
> > So you're suggesting that nohz_full should just be part of the whole
> > ilb machinery by default (that is, not fiddle with ilb internals) and
> > then it's up to CPU partitioning (through cpuset or isolcpus) to disable
> > ilb naturally. Right?
> 
> Yes, but stronger, as long as the CPU is part of a load-balance domain,
> it must not disable the tick while running anything.
> 
> that is, NOHZ_FULL must not become active unless it's running on a
> single CPU partition.

I like the idea but I'm afraid to introduce regressions while doing so,
with people currently using nohz_full without proper partionning...