MIME-Version: 1.0
References: <20230220230624.lkobqeagycx7bi7p@google.com> <6563189C-7765-4FFA-A8F2-A5CC4860A1EF@linux.dev>
In-Reply-To: <6563189C-7765-4FFA-A8F2-A5CC4860A1EF@linux.dev>
From:   Shakeel Butt <shakeelb@google.com>
Date:   Mon, 20 Feb 2023 22:52:10 -0800
Message-ID: <CALvZod55K5zbbVYptq8ud=nKVyU1xceGVf6UcambBZ3BA2TZqA@mail.gmail.com>
Subject: Re: [PATCH] mm: change memcg->oom_group access with atomic operations
To:     Roman Gushchin <roman.gushchin@linux.dev>
Cc:     Yue Zhao <findns94@gmail.com>, linux-mm@kvack.org,
        akpm@linux-foundation.org, hannes@cmpxchg.org, mhocko@kernel.org,
        muchun.song@linux.dev, cgroups@vger.kernel.org,
        linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Mon, Feb 20, 2023 at 9:17 PM Roman Gushchin <roman.gushchin@linux.dev> w=
rote:
>
> > On Feb 20, 2023, at 3:06 PM, Shakeel Butt <shakeelb@google.com> wrote:
> >
> > =EF=BB=BFOn Mon, Feb 20, 2023 at 01:09:44PM -0800, Roman Gushchin wrote=
:
> >>> On Mon, Feb 20, 2023 at 11:16:38PM +0800, Yue Zhao wrote:
> >>> The knob for cgroup v2 memory controller: memory.oom.group
> >>> will be read and written simultaneously by user space
> >>> programs, thus we'd better change memcg->oom_group access
> >>> with atomic operations to avoid concurrency problems.
> >>>
> >>> Signed-off-by: Yue Zhao <findns94@gmail.com>
> >>
> >> Hi Yue!
> >>
> >> I'm curious, have any seen any real issues which your patch is solving=
?
> >> Can you, please, provide a bit more details.
> >>
> >
> > IMHO such details are not needed. oom_group is being accessed
> > concurrently and one of them can be a write access. At least
> > READ_ONCE/WRITE_ONCE is needed here.
>
> Needed for what?

For this particular case, documenting such an access. Though I don't
think there are any architectures which may tear a one byte read/write
and merging/refetching is not an issue for this.

>
> I mean it=E2=80=99s obviously not a big deal to put READ_ONCE()/WRITE_ONC=
E() here, but I struggle to imagine a scenario when it will make any differ=
ence. IMHO it=E2=80=99s easier to justify a proper atomic operation here, e=
ven if it=E2=80=99s most likely an overkill.
>
> My question is very simple: the commit log mentions =E2=80=9C=E2=80=A6 to=
 avoid concurrency problems=E2=80=9D, so I wonder what problems are these.
>
> Also there are other similar cgroup interfaces without READ_ONCE()/WRITE_=
ONCE()

Yeah and those are v1 interfaces e.g. oom_kill_disable, swappiness,
soft_limit. These definitely need [READ|WRITE]_ONCE primitive.

Yue, can you update your patch and convert all accesses to these
fields through [READ|WRITE]_ONCE ?