Sender: Tejun Heo <htejun@gmail.com>
Date:   Wed, 12 Jan 2022 11:21:00 -1000
From:   Tejun Heo <tj@kernel.org>
To:     Waiman Long <longman@redhat.com>
Cc:     Michal =?iso-8859-1?Q?Koutn=FD?= <mkoutny@suse.com>,
        Zefan Li <lizefan.x@bytedance.com>,
        Johannes Weiner <hannes@cmpxchg.org>,
        Jonathan Corbet <corbet@lwn.net>,
        Shuah Khan <shuah@kernel.org>, cgroups@vger.kernel.org,
        linux-kernel@vger.kernel.org, linux-doc@vger.kernel.org,
        linux-kselftest@vger.kernel.org,
        Andrew Morton <akpm@linux-foundation.org>,
        Roman Gushchin <guro@fb.com>, Phil Auld <pauld@redhat.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Juri Lelli <juri.lelli@redhat.com>,
        Frederic Weisbecker <frederic@kernel.org>,
        Marcelo Tosatti <mtosatti@redhat.com>
Subject: Re: [PATCH v9 6/7] cgroup/cpuset: Update description of
 cpuset.cpus.partition in cgroup-v2.rst
Message-ID: <Yd9GPLytGQ6XpYK6@slm.duckdns.org>
References: <20211205183220.818872-1-longman@redhat.com>
 <20211205183220.818872-7-longman@redhat.com>
 <Ybe0YWEo7Wp7wib9@slm.duckdns.org>
 <20211215144450.GC25459@blackbody.suse.cz>
 <96018978-6b7f-1e7f-1012-9df7f7996ec5@redhat.com>
 <Ybo1jmNvM6sblcJq@slm.duckdns.org>
 <58c06961-ffc4-27d7-01d2-4c91b0c9161d@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <58c06961-ffc4-27d7-01d2-4c91b0c9161d@redhat.com>
Precedence: bulk

Hello,

On Wed, Dec 15, 2021 at 01:55:05PM -0500, Waiman Long wrote:
> How about we allow transition to an invalid partition state but still return
> an error?

Like -EAGAIN as Michal suggested? I don't know. I understand the motivation
but one problem is that error return usually means that the operation didn't
change the state of the system and that holds even for -EAGAIN. So, we'd be
trading one locally jarring thing (this thing is asynchrnous and the actual
state transitions should be monitored separately) to another one which is
jarring in a wider context (this thing returned error but the system state
changed anyway). To me, the latter seems noticeably worse given how common
the assumption that an error return indicate that nothing actually happened.

We have other cases where we split operation submissions and completions
(aios being the obvious one) but I don't think we have any where -EAGAIN
indicates successful initiation of an operation. At least I hope not.

Thanks.

-- 
tejun