Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <44843f08-b7c2-42e3-1951-a0c5688190a1@redhat.com>
Date:   Mon, 3 Apr 2023 13:29:57 -0400
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.7.1
Subject: Re: [PATCH 1/3] cgroup/cpuset: Make cpuset_fork() handle
 CLONE_INTO_CGROUP properly
Content-Language: en-US
From:   Waiman Long <longman@redhat.com>
To:     =?UTF-8?Q?Michal_Koutn=c3=bd?= <mkoutny@suse.com>
Cc:     Tejun Heo <tj@kernel.org>, Zefan Li <lizefan.x@bytedance.com>,
        Johannes Weiner <hannes@cmpxchg.org>,
        Christian Brauner <brauner@kernel.org>,
        cgroups@vger.kernel.org, linux-kernel@vger.kernel.org,
        Juri Lelli <juri.lelli@redhat.com>,
        Dietmar Eggemann <dietmar.eggemann@arm.com>,
        gscrivan@redhat.com
References: <20230331145045.2251683-1-longman@redhat.com>
 <20230331145045.2251683-2-longman@redhat.com>
 <20230403165523.aphsec2epqi72k27@blackpad>
 <d9f0005c-6825-b2a0-eac3-fcbad6e32b2f@redhat.com>
In-Reply-To: <d9f0005c-6825-b2a0-eac3-fcbad6e32b2f@redhat.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk


On 4/3/23 13:18, Waiman Long wrote:
>
> On 4/3/23 12:55, Michal Koutný wrote:
>> On Fri, Mar 31, 2023 at 10:50:43AM -0400, Waiman Long 
>> <longman@redhat.com> wrote:
>>> By default, the clone(2) syscall spawn a child process into the same
>>> cgroup as its parent. With the use of the CLONE_INTO_CGROUP flag
>>> introduced by commit ef2c41cf38a7 ("clone3: allow spawning processes
>>> into cgroups"), the child will be spawned into a different cgroup which
>>> is somewhat similar to writing the child's tid into "cgroup.threads".
>>>
>>> The current cpuset_fork() method does not properly handle the
>>> CLONE_INTO_CGROUP case where the cpuset of the child may be different
>>> from that of its parent.  Update the cpuset_fork() method to treat the
>>> CLONE_INTO_CGROUP case similar to cpuset_attach().
>> Should .can_fork=cpuset_can_fork in analogy to cpuset_can_attach be also
>> devised? (Sorry if I missed that in the previous discussion.)
>
> I have thought about that too.
>
> However, the can_attach method checks only a couple of things:
>
> 1) PF_NO_SETAFFINITY flag - which won't be set in the case of fork() 
> as it is for kthread only.
> 2) DL bandwidth - Juri has a cpuset outstanding to modify the way this 
> check is being done. I want to wait until it is settled before 
> tackling this, if necessary.

BTW, even cloning without CLONE_INTO_CGROUP can be an issue WRT DL 
bandwidth. So I will let the scheduler folk figure out what to do with that.

Cheers,
Longman