Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
Date:   Fri, 27 May 2022 00:51:41 +0200
From:   Frederic Weisbecker <frederic@kernel.org>
To:     Tejun Heo <tj@kernel.org>
Cc:     LKML <linux-kernel@vger.kernel.org>,
        Peter Zijlstra <peterz@infradead.org>,
        "Paul E . McKenney" <paulmck@kernel.org>,
        Paul Gortmaker <paul.gortmaker@windriver.com>,
        Johannes Weiner <hannes@cmpxchg.org>,
        Marcelo Tosatti <mtosatti@redhat.com>,
        Phil Auld <pauld@redhat.com>,
        Zefan Li <lizefan.x@bytedance.com>,
        Waiman Long <longman@redhat.com>,
        Daniel Bristot de Oliveira <bristot@kernel.org>,
        Nicolas Saenz Julienne <nsaenz@kernel.org>,
        rcu@vger.kernel.org
Subject: Re: [RFC PATCH 4/4] cpuset: Support RCU-NOCB toggle on v2 root
 partitions
Message-ID: <20220526225141.GA1214445@lothringen>
References: <20220525221055.1152307-1-frederic@kernel.org>
 <20220525221055.1152307-5-frederic@kernel.org>
 <Yo/FGcG+uiBh88sT@slm.duckdns.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Yo/FGcG+uiBh88sT@slm.duckdns.org>
Precedence: bulk

On Thu, May 26, 2022 at 08:21:13AM -1000, Tejun Heo wrote:
> On Thu, May 26, 2022 at 12:10:55AM +0200, Frederic Weisbecker wrote:
> > Introduce a new "isolation.rcu_nocb" file within a cgroup2/cpuset
> > directory which provides support for a set of CPUs to either enable ("1")
> > or disable ("0") RCU callbacks offloading (aka. RCU NOCB). This can
> > overwrite previous boot settings towards "rcu_nocbs=" kernel parameter.
> > 
> > The file is only writeable on "root" type partitions to exclude any
> > overlap. The deepest root type partition has the highest priority.
> > This means that given the following setting:
> > 
> >                     Top cpuset (CPUs: 0-7)
> >                     cpuset.isolation.rcu_nocb = 0
> >                               |
> >                               |
> >                     Subdirectory A (CPUs: 5-7)
> >                     cpuset.cpus.partition = root
> >                     cpuset.isolation.rcu_nocb = 0
> >                               |
> >                               |
> >                     Subdirectory B (CPUs: 7)
> >                     cpuset.cpus.partition = root
> >                     cpuset.isolation.rcu_nocb = 1
> > 
> > the result is that only CPU 7 is in rcu_nocb mode.
> > 
> > Note that "rcu_nocbs" kernel parameter must be passed on boot, even
> > without a cpulist, so that nocb support is enabled.
> 
> Does it even make sense to make this hierarchical? What's wrong with a
> cpumask under sys/ or proc/?

I'm usually told that cpusets is the current place where CPU attributes are
supposed to go. I personally don't mind much /sys either even though cpusets
looks like a more flexible way to partition CPUs with properties and tasks
placement altogether...