[PATCH net-next] net: sock: do not set sk_cookie in sk_clone_lock()

[PATCH net-next] net: sock: do not set sk_cookie in sk_clone_lock()

[Yafang Shao]

The only call site of sk_clone_lock is in inet_csk_clone_lock,
and sk_cookie will be set there.
So we don't need to set sk_cookie in sk_clone_lock().
That can save an atomic operation.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
---
 net/core/sock.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/net/core/sock.c b/net/core/sock.c
index f00902c..21e2a84 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -1726,7 +1726,6 @@ struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
 		newsk->sk_err_soft = 0;
 		newsk->sk_priority = 0;
 		newsk->sk_incoming_cpu = raw_smp_processor_id();
-		atomic64_set(&newsk->sk_cookie, 0);
 		if (likely(newsk->sk_net_refcnt))
 			sock_inuse_add(sock_net(newsk), 1);
 
-- 
1.8.3.1

Re: [PATCH net-next] net: sock: do not set sk_cookie in sk_clone_lock()

[Eric Dumazet]

On Thu, Jan 17, 2019 at 2:02 AM Yafang Shao <laoar.shao@gmail.com> wrote:
>
> The only call site of sk_clone_lock is in inet_csk_clone_lock,
> and sk_cookie will be set there.
> So we don't need to set sk_cookie in sk_clone_lock().
> That can save an atomic operation.
>

Patch is fine, although the wording of ' atomic operation'  is a bit misleading.

atomic_set or atomic_read are plain memory writes and reads.

Real ' atomic and expensive'  operations are the ones doing RMW
operations (with lock semantic on SMP)

Reviewed-by: Eric Dumazet <edumazet@google.com>


> Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
> ---
>  net/core/sock.c | 1 -
>  1 file changed, 1 deletion(-)
>
> diff --git a/net/core/sock.c b/net/core/sock.c
> index f00902c..21e2a84 100644
> --- a/net/core/sock.c
> +++ b/net/core/sock.c
> @@ -1726,7 +1726,6 @@ struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
>                 newsk->sk_err_soft = 0;
>                 newsk->sk_priority = 0;
>                 newsk->sk_incoming_cpu = raw_smp_processor_id();
> -               atomic64_set(&newsk->sk_cookie, 0);
>                 if (likely(newsk->sk_net_refcnt))
>                         sock_inuse_add(sock_net(newsk), 1);
>
> --
> 1.8.3.1
>

Re: [PATCH net-next] net: sock: do not set sk_cookie in sk_clone_lock()

[Yafang Shao]

On Fri, Jan 18, 2019 at 12:40 AM Eric Dumazet <edumazet@google.com> wrote:
>
> On Thu, Jan 17, 2019 at 2:02 AM Yafang Shao <laoar.shao@gmail.com> wrote:
> >
> > The only call site of sk_clone_lock is in inet_csk_clone_lock,
> > and sk_cookie will be set there.
> > So we don't need to set sk_cookie in sk_clone_lock().
> > That can save an atomic operation.
> >
>
> Patch is fine, although the wording of ' atomic operation'  is a bit misleading.
>
> atomic_set or atomic_read are plain memory writes and reads.
>
> Real ' atomic and expensive'  operations are the ones doing RMW
> operations (with lock semantic on SMP)
>
> Reviewed-by: Eric Dumazet <edumazet@google.com>
>

Thanks for your correction.
Will change it and send v2.

>
> > Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
> > ---
> >  net/core/sock.c | 1 -
> >  1 file changed, 1 deletion(-)
> >
> > diff --git a/net/core/sock.c b/net/core/sock.c
> > index f00902c..21e2a84 100644
> > --- a/net/core/sock.c
> > +++ b/net/core/sock.c
> > @@ -1726,7 +1726,6 @@ struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
> >                 newsk->sk_err_soft = 0;
> >                 newsk->sk_priority = 0;
> >                 newsk->sk_incoming_cpu = raw_smp_processor_id();
> > -               atomic64_set(&newsk->sk_cookie, 0);
> >                 if (likely(newsk->sk_net_refcnt))
> >                         sock_inuse_add(sock_net(newsk), 1);
> >
> > --
> > 1.8.3.1
> >