Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210916122943.19849-1-yajun.deng@linux.dev> <20210917183311.2db5f332@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com>
 <87275ec67ed69d077e0265bc01acd8a2@linux.dev> <CAM_iQpXcqpEFpnyX=wLQFTWJBjWiAMofighQkpnrV2a0Fh83AQ@mail.gmail.com>
 <202109202028152977817@linux.dev>
In-Reply-To: <202109202028152977817@linux.dev>
From:   Cong Wang <xiyou.wangcong@gmail.com>
Date:   Mon, 20 Sep 2021 09:33:09 -0700
Message-ID: <CAM_iQpV4eaiD6msaiYNAOYE9Eoy2AjnGrSSwihhCO-yPb-61ww@mail.gmail.com>
Subject: Re: Re: [PATCH net-next] net: socket: add the case sock_no_xxx support
To:     "yajun.deng@linux.dev" <yajun.deng@linux.dev>
Cc:     kuba <kuba@kernel.org>, davem <davem@davemloft.net>,
        netdev <netdev@vger.kernel.org>,
        linux-kernel <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Mon, Sep 20, 2021 at 5:28 AM yajun.deng@linux.dev
<yajun.deng@linux.dev> wrote:
>
> From: Cong Wang
> Date: 2021-09-20 07:52
> To: Yajun Deng
> CC: Jakub Kicinski; David Miller; Linux Kernel Network Developers; LKML
> Subject: Re: [PATCH net-next] net: socket: add the case sock_no_xxx suppo=
rt
> On Sat, Sep 18, 2021 at 5:11 AM <yajun.deng@linux.dev> wrote:
> >
> > September 18, 2021 9:33 AM, "Jakub Kicinski" <kuba@kernel.org> wrote:
> >
> > > On Thu, 16 Sep 2021 20:29:43 +0800 Yajun Deng wrote:
> > >
> > >> Those sock_no_{mmap, socketpair, listen, accept, connect, shutdown,
> > >> sendpage} functions are used many times in struct proto_ops, but the=
y are
> > >> meaningless. So we can add them support in socket and delete them in=
 struct
> > >> proto_ops.
> > >
> > > So the reason to do this is.. what exactly?
> > >
> > > Removing a couple empty helpers (which is not even part of this patch=
)?
> > >
> > > I'm not sold, sorry.
> >
> > When we define a struct proto_ops xxx, we only need to assign meaningfu=
l member variables that we need.
> > Those {mmap, socketpair, listen, accept, connect, shutdown, sendpage} m=
embers we don't need assign
> > it if we don't need. We just need do once in socket, not in every struc=
t proto_ops.
> >
> > These members are assigned meaningless values far more often than meani=
ngful ones, so this patch I used likely(!!sock->ops->xxx) for this case. Th=
is is the reason why I send this patch.
>
> But you end up adding more code:
>
> 1 file changed, 58 insertions(+), 13 deletions(-)
>
> Yes=EF=BC=8CThis would add more code, but this is at the cost of reducing=
 other codes. At the same time, the code will only run  likely(!sock->ops->=
xxx) in most cases.  Don=E2=80=99t you think that this kind of meaningless =
thing shouldn=E2=80=99t be done by socket?

I have no idea why you call it reducing code while adding 45 lines
of code. So this does not make sense to me.

Thanks.