Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:2 as permitted sender) client-ip=2620:137:e000::3:2;
MIME-Version: 1.0
References: <20231016221900.4031141-1-roman.gushchin@linux.dev>
 <20231016221900.4031141-3-roman.gushchin@linux.dev> <d698b8d0-1697-e336-bccb-592e633e8b98@suse.cz>
 <ZTAUTWO2UfI0VoPL@P9FQF9L96D.corp.robot.car> <CALvZod6mb91o9pW57suovtW1UQ8G8j=2S3Tjoqzjh6L+jqz-EQ@mail.gmail.com>
 <ZTBeRu3iDu7nnPV8@P9FQF9L96D.corp.robot.car>
In-Reply-To: <ZTBeRu3iDu7nnPV8@P9FQF9L96D.corp.robot.car>
From:   Shakeel Butt <shakeelb@google.com>
Date:   Thu, 19 Oct 2023 09:36:44 -0700
Message-ID: <CALvZod5kXRY0LV6VOnctTYVhdHu+=yqzsQzKYa2_6_Jg+cOWfQ@mail.gmail.com>
Subject: Re: [PATCH v3 2/5] mm: kmem: add direct objcg pointer to task_struct
To:     Roman Gushchin <roman.gushchin@linux.dev>
Cc:     Vlastimil Babka <vbabka@suse.cz>,
        Andrew Morton <akpm@linux-foundation.org>,
        linux-kernel@vger.kernel.org, cgroups@vger.kernel.org,
        Johannes Weiner <hannes@cmpxchg.org>,
        Michal Hocko <mhocko@kernel.org>,
        Muchun Song <muchun.song@linux.dev>,
        Dennis Zhou <dennis@kernel.org>,
        David Rientjes <rientjes@google.com>,
        Naresh Kamboju <naresh.kamboju@linaro.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Wed, Oct 18, 2023 at 3:38=E2=80=AFPM Roman Gushchin <roman.gushchin@linu=
x.dev> wrote:
>
> On Wed, Oct 18, 2023 at 11:26:59AM -0700, Shakeel Butt wrote:
> > On Wed, Oct 18, 2023 at 10:22=E2=80=AFAM Roman Gushchin
> > <roman.gushchin@linux.dev> wrote:
> > >
> > [...]
> > > > >     struct mem_cgroup *memcg;
> > > > > @@ -3008,19 +3054,26 @@ __always_inline struct obj_cgroup *get_ob=
j_cgroup_from_current(void)
> > > > >
> > > > >     if (in_task()) {
> > > > >             memcg =3D current->active_memcg;
> > > > > +           if (unlikely(memcg))
> > > > > +                   goto from_memcg;
> > > > >
> > > > > -           /* Memcg to charge can't be determined. */
> > > > > -           if (likely(!memcg) && (!current->mm || (current->flag=
s & PF_KTHREAD)))
> > > >
> > > > The checks for current->mm and PF_KTHREAD seem to be gone completel=
y after
> > > > the patch, was that intended and why?
> > >
> > > There is no need for those anymore because it's as cheap or cheaper
> > > to check task->objcg for being NULL. Those were primarily used to rul=
e out
> > > kernel threads allocations early.
> > >
> >
> > I have the same understanding but please correct my suspicions here.
> > We can echo the kernel thread's pid to cgroup.procs which have
> > PF_NO_SETAFFINITY and thus this will cause the lower bit of the kernel
> > thread's task->objcg to be set. Please correct me if I am missing
> > something.
>
> Yes, you seem to be right. It's a gray zone because moving kernel threads=
 out of
> the root cgroup doesn't sound like a good idea, but I agree it's better t=
o keep
> the old behavior in place.
>
> Does this fixlet look good to you?
>

This looks fine. Another option is not to set the bit for such
task_structs in fork/attach.

> Thanks!
>
> --
>
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index 1a2835448028..0b0d2dc7a7d4 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -3021,6 +3021,10 @@ static struct obj_cgroup *current_objcg_update(voi=
d)
>                         old =3D NULL;
>                 }
>
> +               /* If new objcg is NULL, no reason for the second atomic =
update. */
> +               if (!current->mm || (current->flags & PF_KTHREAD))
> +                       return NULL;
> +
>                 /*
>                  * Release the objcg pointer from the previous iteration,
>                  * if try_cmpxcg() below fails.