netdev - 答复: [PATCH][RFC] udp: cache sock to avoid searching it twice

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <2AD939572F25A448A3AE3CAEA61328C237AE9FFD@BC-MAIL-M30.internal.baidu.com>
Date:   Mon, 12 Nov 2018 05:39:27 +0000
From:   "Li,Rongqing" <lirongqing@...du.com>
To:     Eric Dumazet <eric.dumazet@...il.com>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: 答复: [PATCH][RFC] udp: cache sock to avoid searching it twice


On Sat, Nov 10, 2018 at 1:29 AM Eric Dumazet <eric.dumazet@...il.com> wrote:
>
>
>
> On 11/08/2018 10:21 PM, Li RongQing wrote:
> > GRO for UDP needs to lookup socket twice, first is in gro receive,
> > second is gro complete, so if store sock to skb to avoid looking up
> > twice, this can give small performance boost
> >
> > netperf -t UDP_RR -l 10
> >
> > Before:
> >       Rate per sec: 28746.01
> > After:
> >       Rate per sec: 29401.67
> >
> > Signed-off-by: Li RongQing <lirongqing@...du.com>
> > ---
> >  net/ipv4/udp_offload.c | 18 +++++++++++++++++-
> >  1 file changed, 17 insertions(+), 1 deletion(-)
> >
> > diff --git a/net/ipv4/udp_offload.c b/net/ipv4/udp_offload.c
> > index 0646d61f4fa8..429570112a33 100644
> > --- a/net/ipv4/udp_offload.c
> > +++ b/net/ipv4/udp_offload.c
> > @@ -408,6 +408,11 @@ struct sk_buff *udp_gro_receive(struct list_head *head, struct sk_buff *skb,
> >
> >       if (udp_sk(sk)->gro_enabled) {
> >               pp = call_gro_receive(udp_gro_receive_segment, head, skb);
> > +
> > +             if (!IS_ERR(pp) && NAPI_GRO_CB(pp)->count > 1) {
> > +                     sock_hold(sk);
> > +                     pp->sk = sk;
>
>
> You also have to set pp->destructor to sock_edemux
>
> flush_gro_hash -> kfree_skb()
>
> If there is no destructor, the reference on pp->sk will never be released.
>
>

Ok, thanks,

does it need to reset sk in udp_gro_complete,  ip early demuxing will lookup udp socket again, if we can keep it, we can avoid to lookup socket again


-RongQing
>
>
> > +             }
> >               rcu_read_unlock();
> >               return pp;
> >       }
> > @@ -444,6 +449,10 @@ struct sk_buff *udp_gro_receive(struct list_head *head, struct sk_buff *skb,
> >       skb_gro_postpull_rcsum(skb, uh, sizeof(struct udphdr));
> >       pp = call_gro_receive_sk(udp_sk(sk)->gro_receive, sk, head, skb);
> >
> > +     if (!IS_ERR(pp) && NAPI_GRO_CB(pp)->count > 1) {
> > +             sock_hold(sk);
> > +             pp->sk = sk;
> > +     }
> >  out_unlock:
> >       rcu_read_unlock();
> >       skb_gro_flush_final(skb, pp, flush);
> > @@ -502,7 +511,9 @@ int udp_gro_complete(struct sk_buff *skb, int nhoff,
> >       uh->len = newlen;
> >
> >       rcu_read_lock();
> > -     sk = (*lookup)(skb, uh->source, uh->dest);
> > +     sk = skb->sk;
> > +     if (!sk)
> > +             sk = (*lookup)(skb, uh->source, uh->dest);
> >       if (sk && udp_sk(sk)->gro_enabled) {
> >               err = udp_gro_complete_segment(skb);
> >       } else if (sk && udp_sk(sk)->gro_complete) {
> > @@ -516,6 +527,11 @@ int udp_gro_complete(struct sk_buff *skb, int nhoff,
> >               err = udp_sk(sk)->gro_complete(sk, skb,
> >                               nhoff + sizeof(struct udphdr));
> >       }
> > +
> > +     if (skb->sk) {
> > +             sock_put(skb->sk);
> > +             skb->sk = NULL;
> > +     }
> >       rcu_read_unlock();
> >
> >       if (skb->remcsum_offload)
> >