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;
Date:   Thu, 8 Jun 2023 17:38:22 +0000
In-Reply-To: <20230608171256.17827-1-mkoutny@suse.com>
Mime-Version: 1.0
References: <20230608171256.17827-1-mkoutny@suse.com>
Message-ID: <20230608173822.zezggshxhh62mxoi@google.com>
Subject: Re: [RFC PATCH] mm: Sync percpu mm RSS counters before querying
From:   Shakeel Butt <shakeelb@google.com>
To:     "Michal =?utf-8?Q?Koutn=C3=BD?=" <mkoutny@suse.com>
Cc:     linux-mm@kvack.org, linux-kernel@vger.kernel.org,
        Andrew Morton <akpm@linux-foundation.org>,
        Christian Brauner <brauner@kernel.org>,
        "Liam R . Howlett" <Liam.Howlett@oracle.com>,
        Suren Baghdasaryan <surenb@google.com>,
        "Michael S . Tsirkin" <mst@redhat.com>,
        Mike Christie <michael.christie@oracle.com>,
        Andrei Vagin <avagin@gmail.com>,
        Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
        Nicholas Piggin <npiggin@gmail.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Adam Majer <amajer@suse.com>, Jan Kara <jack@suse.cz>,
        Michal Hocko <mhocko@kernel.org>
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Jun 08, 2023 at 07:12:56PM +0200, Michal Koutn=C3=BD wrote:
> An issue was observed with stats collected in struct rusage on ppc64le
> with 64kB pages. The percpu counters use batching with
> 	percpu_counter_batch =3D max(32, nr*2) # in PAGE_SIZE
> i.e. with larger pages but similar RSS consumption (bytes), there'll be
> less flushes and error more noticeable.
>=20
> In this given case (getting consumption of exited child), we can request
> percpu counter's flush without worrying about contention with updaters.
>=20
> Fortunately, the commit f1a7941243c1 ("mm: convert mm's rss stats into
> percpu_counter") didn't eradicate all traces of SPLIT_RSS_COUNTING and
> this mechanism already provided some synchronization points before
> reading stats.
> Therefore, use sync_mm_rss as carrier for percpu counters refreshes and
> forget SPLIT_RSS_COUNTING macro for good.
>=20
> Fixes: f1a7941243c1 ("mm: convert mm's rss stats into percpu_counter")
> Reported-by: Adam Majer <amajer@suse.com>
> Signed-off-by: Michal Koutn=C3=BD <mkoutny@suse.com>

The patch seems reasonable to me. Are any of the callsites of
sync_mm_rss performance sensitive?

> ---
>  include/linux/mm.h | 6 ++----
>  kernel/fork.c      | 4 ----
>  2 files changed, 2 insertions(+), 8 deletions(-)
>=20
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 27ce77080c79..30cfde88d5b2 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -2547,13 +2547,11 @@ static inline void setmax_mm_hiwater_rss(unsigned=
 long *maxrss,
>  		*maxrss =3D hiwater_rss;
>  }
> =20
> -#if defined(SPLIT_RSS_COUNTING)
> -void sync_mm_rss(struct mm_struct *mm);
> -#else
>  static inline void sync_mm_rss(struct mm_struct *mm)
>  {
> +	for (int i =3D 0; i < NR_MM_COUNTERS; ++i)
> +		percpu_counter_sum(&mm->rss_stat[i]);
>  }
> -#endif
> =20
>  #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
>  static inline int pte_special(pte_t pte)
> diff --git a/kernel/fork.c b/kernel/fork.c
> index 81cba91f30bb..e030eb902e4b 100644
> --- a/kernel/fork.c
> +++ b/kernel/fork.c
> @@ -2412,10 +2412,6 @@ __latent_entropy struct task_struct *copy_process(
>  	p->io_uring =3D NULL;
>  #endif
> =20
> -#if defined(SPLIT_RSS_COUNTING)
> -	memset(&p->rss_stat, 0, sizeof(p->rss_stat));
> -#endif
> -
>  	p->default_timer_slack_ns =3D current->timer_slack_ns;
> =20
>  #ifdef CONFIG_PSI
> --=20
> 2.40.1
>=20