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;
MIME-Version: 1.0
References: <20220728091341.20820-1-tujinjiang@bytedance.com> <YuJ2UeYVpSe+21Xs@dhcp22.suse.cz>
In-Reply-To: <YuJ2UeYVpSe+21Xs@dhcp22.suse.cz>
From:   Jinjiang Tu <tujinjiang@bytedance.com>
Date:   Fri, 29 Jul 2022 10:11:04 +0800
Message-ID: <CAF77dw_j+4ziV+wuXw0WWSTHqs0Usy-TDzC91skPxcPfKT66-Q@mail.gmail.com>
Subject: Re: Re: [PATCH v2] vmscan: Do not free nr_deferred for memcg aware shrinkers
To:     Michal Hocko <mhocko@suse.com>
Cc:     Yang Shi <shy828301@gmail.com>, akpm@linux-foundation.org,
        linux-mm@kvack.org, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, Jul 28, 2022 at 7:43 PM Michal Hocko <mhocko@suse.com> wrote:
>
> On Thu 28-07-22 17:13:41, tujinjiang@bytedance.com wrote:
> > From: Jinjiang Tu <tujinjiang@bytedance.com>
> >
> > When a memcg aware shrinker is registered by register_shrinker(),
> > shrinker->nr_deferred will not be initialized. But when the shrinker
> > is unregistered by unregister_shrinker(), shrinker->nr_deferred
> > will be freed.
> >
> > Luckily, the memcg aware shrinkers in the current kernel are pre-zeroed.
> > But a new memcg aware shrinker may be added in the future, and we should
> > not assume the shrinker is pre-zeroed.
> >
> > Another unregister API free_prealloced_shrinker() does not assume the
> > shrinker is pre-zered and free shrinker->nr_deferred only if it is
> > not memcg aware. So unregister_shrinker() should do like
> > free_prealloced_shrinker().
> >
> > Signed-off-by: Jinjiang Tu <tujinjiang@bytedance.com>
> > ---
> >  mm/vmscan.c | 5 ++++-
> >  1 file changed, 4 insertions(+), 1 deletion(-)
> >
> > diff --git a/mm/vmscan.c b/mm/vmscan.c
> > index f7d9a683e3a7..f8a9a5349b6e 100644
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
> > @@ -675,8 +675,11 @@ void unregister_shrinker(struct shrinker *shrinker)
> >       down_write(&shrinker_rwsem);
> >       list_del(&shrinker->list);
> >       shrinker->flags &= ~SHRINKER_REGISTERED;
> > -     if (shrinker->flags & SHRINKER_MEMCG_AWARE)
> > +     if (shrinker->flags & SHRINKER_MEMCG_AWARE) {
> >               unregister_memcg_shrinker(shrinker);
> > +             up_write(&shrinker_rwsem);
> > +             return;
> > +     }
> >       up_write(&shrinker_rwsem);
> >
> >       kfree(shrinker->nr_deferred);
>
> Can we get rid of the code duplication?
> ---
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index f7d9a683e3a7..308279414fe8 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -632,12 +632,10 @@ int prealloc_shrinker(struct shrinker *shrinker)
>         return 0;
>  }
>
> -void free_prealloced_shrinker(struct shrinker *shrinker)
> +static void __free_shrinker(struct shrinker *shrinker)
>  {
>         if (shrinker->flags & SHRINKER_MEMCG_AWARE) {
> -               down_write(&shrinker_rwsem);
>                 unregister_memcg_shrinker(shrinker);
> -               up_write(&shrinker_rwsem);
>                 return;
>         }
>
> @@ -645,6 +643,13 @@ void free_prealloced_shrinker(struct shrinker *shrinker)
>         shrinker->nr_deferred = NULL;
>  }
>
> +void free_prealloced_shrinker(struct shrinker *shrinker)
> +{
> +       down_write(&shrinker_rwsem);
> +       __free_shrinker(shrinker);
> +       up_write(&shrinker_rwsem);
> +}
> +
>  void register_shrinker_prepared(struct shrinker *shrinker)
>  {
>         down_write(&shrinker_rwsem);
> @@ -675,12 +680,9 @@ void unregister_shrinker(struct shrinker *shrinker)
>         down_write(&shrinker_rwsem);
>         list_del(&shrinker->list);
>         shrinker->flags &= ~SHRINKER_REGISTERED;
> -       if (shrinker->flags & SHRINKER_MEMCG_AWARE)
> -               unregister_memcg_shrinker(shrinker);
> -       up_write(&shrinker_rwsem);
>
> -       kfree(shrinker->nr_deferred);
> -       shrinker->nr_deferred = NULL;
> +       __free_shrinker(shrinker);
> +       up_write(&shrinker_rwsem);
>  }
>  EXPORT_SYMBOL(unregister_shrinker);
>
> --
> Michal Hocko
> SUSE Labs
Yes, the code is clearer.