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: <20221024051744.GA48642@debian> <20221101085153.12ccae1c@kernel.org>
 <CAJLv34RKj6u_7EZwYWiNujC-R4nxKHJ24DVYqydgHPy88NqMPA@mail.gmail.com> <CANn89iJeg+wQUdi7i=EbSS7Z__j+LEakPab3oKD7_Rr4hmV_xg@mail.gmail.com>
In-Reply-To: <CANn89iJeg+wQUdi7i=EbSS7Z__j+LEakPab3oKD7_Rr4hmV_xg@mail.gmail.com>
From:   Eric Dumazet <edumazet@google.com>
Date:   Wed, 2 Nov 2022 11:25:54 -0700
Message-ID: <CANn89iKKFABZxAv5PkNmCVZTyHxqghC9zoRSvbfDo+04qrHH9w@mail.gmail.com>
Subject: Re: [PATCH net-next] gro: avoid checking for a failed search
To:     Richard Gobert <richardbgobert@gmail.com>
Cc:     Jakub Kicinski <kuba@kernel.org>, davem@davemloft.net,
        pabeni@redhat.com, lixiaoyan@google.com, alexanderduyck@fb.com,
        steffen.klassert@secunet.com, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, Nov 2, 2022 at 11:20 AM Eric Dumazet <edumazet@google.com> wrote:
>
> On Wed, Nov 2, 2022 at 9:46 AM Richard Gobert <richardbgobert@gmail.com> wrote:
> >
> > > Why does it matter? You see a measurable perf win?
> >
> > In the common case, we will exit the loop with a break,
> > so this patch eliminates an unnecessary check.
> >
> > On some architectures this optimization might be done
> > automatically by the compiler, but I think it will be better
> > to make it explicit here. Although on x86 this optimization
> > happens automatically, I noticed that on my build target
> > (ARM/GCC) this does change the binary.
>
> What about taking this as an opportunity to reduce the indentation
> level by one tab ?
>
> Untested patch:
>
> diff --git a/net/core/gro.c b/net/core/gro.c
> index bc9451743307bc380cca96ae6995aa0a3b83d185..ddfe92c9a5e869d241931b72d6b3426a0e858468
> 100644
> --- a/net/core/gro.c
> +++ b/net/core/gro.c
> @@ -491,43 +491,44 @@ static enum gro_result dev_gro_receive(struct
> napi_struct *napi, struct sk_buff
>         list_for_each_entry_rcu(ptype, head, list) {
>                 if (ptype->type != type || !ptype->callbacks.gro_receive)
>                         continue;
> +               goto found_ptype;
> +       }
> +       rcu_read_unlock();
> +       goto normal;


Or even better:

        list_for_each_entry_rcu(ptype, head, list) {
               if (ptype->type == type && ptype->callbacks.gro_receive)
                       goto found_ptype;
       }
       rcu_read_unlock();
       goto normal;