Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.31 as permitted sender) client-ip=23.128.96.31;
Sender: Tejun Heo <htejun@gmail.com>
Date:   Wed, 18 Oct 2023 08:08:51 -1000
From:   Tejun Heo <tj@kernel.org>
To:     Waiman Long <longman@redhat.com>
Cc:     Zefan Li <lizefan.x@bytedance.com>,
        Johannes Weiner <hannes@cmpxchg.org>,
        Jonathan Corbet <corbet@lwn.net>,
        Lai Jiangshan <jiangshanlai@gmail.com>,
        Shuah Khan <shuah@kernel.org>, cgroups@vger.kernel.org,
        linux-doc@vger.kernel.org, linux-kernel@vger.kernel.org,
        linux-kselftest@vger.kernel.org
Subject: Re: [PATCH-cgroup 3/4] cgroup/cpuset: Keep track of CPUs in isolated
 partitions
Message-ID: <ZTAfM0msp8Cg-qLy@slm.duckdns.org>
References: <20231013181122.3518610-1-longman@redhat.com>
 <20231013181122.3518610-4-longman@redhat.com>
 <ZS-kt6X5Dd1lktAw@slm.duckdns.org>
 <9e2772e3-f615-5e80-6922-5a2dd06a8b07@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <9e2772e3-f615-5e80-6922-5a2dd06a8b07@redhat.com>
Precedence: bulk

On Wed, Oct 18, 2023 at 09:30:04AM -0400, Waiman Long wrote:
> On 10/18/23 05:26, Tejun Heo wrote:
> > On Fri, Oct 13, 2023 at 02:11:21PM -0400, Waiman Long wrote:
> > ...
> > > @@ -3875,6 +3931,13 @@ static struct cftype dfl_files[] = {
> > >   		.flags = CFTYPE_ONLY_ON_ROOT | CFTYPE_DEBUG,
> > >   	},
> > > +	{
> > > +		.name = "cpus.isolated",
> > > +		.seq_show = cpuset_common_seq_show,
> > > +		.private = FILE_ISOLATED_CPULIST,
> > > +		.flags = CFTYPE_ONLY_ON_ROOT | CFTYPE_DEBUG,
> > > +	},
> > I'd much rather show this in a wq sysfs file along with other related masks,
> > and not in a DEBUG file.
> 
> It can certainly be exposed as a permanent addition to the cgroup control
> files instead of a debug only file. However this set of isolated CPUs may be
> used by others not just by workqueue. So I doubt if it should be a sysfs
> file in the workqueue directory. I can see if it is possible to put a
> symlink there point back to the cgroupfs.

I don't know whether it will happen but let's say there will be three
subsystems which call into workqueue for this. Wouldn't it be better to have
all of them in workqueue sysfs using a consistent naming scheme? What does
putting it in cgroupfs buy us?

Thanks.

-- 
tejun