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: <20230619082547.73929-1-wuyun.abel@bytedance.com>
In-Reply-To: <20230619082547.73929-1-wuyun.abel@bytedance.com>
From:   Eric Dumazet <edumazet@google.com>
Date:   Mon, 19 Jun 2023 12:08:07 +0200
Message-ID: <CANn89i+deprQWB0dmsUD1sRmy1VQCQwKnZUkLu_AEGV=ow=PKQ@mail.gmail.com>
Subject: Re: [PATCH net-next] inet: Save one atomic op if no memcg to charge
To:     Abel Wu <wuyun.abel@bytedance.com>
Cc:     "David S. Miller" <davem@davemloft.net>,
        David Ahern <dsahern@kernel.org>,
        Jakub Kicinski <kuba@kernel.org>,
        Paolo Abeni <pabeni@redhat.com>,
        "open list:NETWORKING [IPv4/IPv6]" <netdev@vger.kernel.org>,
        open list <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Mon, Jun 19, 2023 at 10:26=E2=80=AFAM Abel Wu <wuyun.abel@bytedance.com>=
 wrote:
>
> If there is no net-memcg associated with the sock, don't bother
> calculating its memory usage for charge.
>
> Signed-off-by: Abel Wu <wuyun.abel@bytedance.com>
> ---
>  net/ipv4/inet_connection_sock.c | 18 +++++++++++-------
>  1 file changed, 11 insertions(+), 7 deletions(-)
>
> diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_s=
ock.c
> index 65ad4251f6fd..73798282c1ef 100644
> --- a/net/ipv4/inet_connection_sock.c
> +++ b/net/ipv4/inet_connection_sock.c
> @@ -706,20 +706,24 @@ struct sock *inet_csk_accept(struct sock *sk, int f=
lags, int *err, bool kern)
>  out:
>         release_sock(sk);
>         if (newsk && mem_cgroup_sockets_enabled) {
> -               int amt;
> +               int amt =3D 0;
>
>                 /* atomically get the memory usage, set and charge the
>                  * newsk->sk_memcg.
>                  */
>                 lock_sock(newsk);
>
> -               /* The socket has not been accepted yet, no need to look =
at
> -                * newsk->sk_wmem_queued.
> -                */
> -               amt =3D sk_mem_pages(newsk->sk_forward_alloc +
> -                                  atomic_read(&newsk->sk_rmem_alloc));
>                 mem_cgroup_sk_alloc(newsk);
> -               if (newsk->sk_memcg && amt)
> +               if (newsk->sk_memcg) {
> +                       /* The socket has not been accepted yet, no need
> +                        * to look at newsk->sk_wmem_queued.
> +                        */
> +                       amt =3D sk_mem_pages(newsk->sk_forward_alloc +
> +                                          atomic_read(&newsk->sk_rmem_al=
loc));
> +
> +               }
> +
> +               if (amt)
>                         mem_cgroup_charge_skmem(newsk->sk_memcg, amt,
>                                                 GFP_KERNEL | __GFP_NOFAIL=
);

This looks correct, but claiming reading an atomic_t is an 'atomic op'
is a bit exaggerated.