Re: [PATCH v10 bpf-next/net 3/6] net-memcg: Introduce net.core.memcg_exclusive sysctl.

[Kuniyuki Iwashima <kuniyu@google.com>]

On Mon, Sep 22, 2025 at 6:03 PM Kuniyuki Iwashima <kuniyu@google.com> wrote:
>
> On Mon, Sep 22, 2025 at 5:54 PM Shakeel Butt <shakeel.butt@linux.dev> wrote:
> >
> > On Sat, Sep 20, 2025 at 12:07:17AM +0000, Kuniyuki Iwashima wrote:
> > > diff --git a/net/core/sock.c b/net/core/sock.c
> > > index 814966309b0e..348e599c3fbc 100644
> > > --- a/net/core/sock.c
> > > +++ b/net/core/sock.c
> > > @@ -2519,6 +2519,7 @@ struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
> > >  #ifdef CONFIG_MEMCG
> > >       /* sk->sk_memcg will be populated at accept() time */
> > >       newsk->sk_memcg = NULL;
> > > +     mem_cgroup_sk_set_flags(newsk, mem_cgroup_sk_get_flags(sk));
> >
> > Why do you need to set the flag here? Will doing in __inet_accept only
> > be too late i.e. protocol accounting would have happened?

Oh no, this is because I used newsk in __inet_accept(), and
we can use sock->sk in __inet_accept() and remove the inheritance
in sk_clone_lock() like this,

flags = mem_cgroup_sk_get_flags(sock->sk);

but I think using newsk reduces one cache miss.  Also, we touch the
parent sk_memcg during sock_copy() in sk_clone_lock().