netdev - Re: [PATCH net-next 10/10] udp: use skb_attempt_defer

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: Windows password security audit tool. GUI, reports in PDF.

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

Message-ID: <CANn89iJLC=sAS_9=dOaRv0P69+8cG8ZEW5boq9f4JxeXYDeBzQ@mail.gmail.com>
Date: Wed, 17 Sep 2025 09:32:05 -0700
From: Eric Dumazet <edumazet@...gle.com>
To: Paolo Abeni <pabeni@...hat.com>
Cc: Simon Horman <horms@...nel.org>, Willem de Bruijn <willemb@...gle.com>, 
	Kuniyuki Iwashima <kuniyu@...gle.com>, David Ahern <dsahern@...nel.org>, netdev@...r.kernel.org, 
	eric.dumazet@...il.com, "David S . Miller" <davem@...emloft.net>, 
	Jakub Kicinski <kuba@...nel.org>
Subject: Re: [PATCH net-next 10/10] udp: use skb_attempt_defer_free()

On Wed, Sep 17, 2025 at 9:15 AM Paolo Abeni <pabeni@...hat.com> wrote:
>
>
>
> On 9/16/25 6:09 PM, Eric Dumazet wrote:
> > Move skb freeing from udp recvmsg() path to the cpu
> > which allocated/received it, as TCP did in linux-5.17.
> >
> > This increases max thoughput by 20% to 30%, depending
> > on number of BH producers.
> >
> > Signed-off-by: Eric Dumazet <edumazet@...gle.com>
> > ---
> >  net/ipv4/udp.c | 7 +++++++
> >  1 file changed, 7 insertions(+)
> >
> > diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
> > index 7d1444821ee51a19cd5fd0dd5b8d096104c9283c..0c40426628eb2306b609881341a51307c4993871 100644
> > --- a/net/ipv4/udp.c
> > +++ b/net/ipv4/udp.c
> > @@ -1825,6 +1825,13 @@ void skb_consume_udp(struct sock *sk, struct sk_buff *skb, int len)
> >       if (unlikely(READ_ONCE(udp_sk(sk)->peeking_with_offset)))
> >               sk_peek_offset_bwd(sk, len);
> >
> > +     if (!skb_shared(skb)) {
> > +             if (unlikely(udp_skb_has_head_state(skb)))
> > +                     skb_release_head_state(skb);
> > +             skb_attempt_defer_free(skb);
> > +             return;
> > +     }
> > +
> >       if (!skb_unref(skb))
> >               return;
>
> What about consolidating the release path with something alternative
> like the following, does the skb_unref()/additional smp_rmb() affects
> performances badly?
>
> Thanks,
>
> Paolo
> ---
> diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
> index cc3ce0f762ec..ed2e370ad4de 100644
> --- a/net/ipv4/udp.c
> +++ b/net/ipv4/udp.c
> @@ -1836,7 +1836,7 @@ void skb_consume_udp(struct sock *sk, struct
> sk_buff *skb, int len)
>          */
>         if (unlikely(udp_skb_has_head_state(skb)))
>                 skb_release_head_state(skb);
> -       __consume_stateless_skb(skb);
> +       skb_attempt_defer_free(skb);

This will not work, skb_attempt_defer_free(skb) wants to perform an skb_unref(),
from skb_defer_free_flush()

while (skb != NULL) {
     next = skb->next;
     napi_consume_skb(skb, 1);
     skb = next;
}

MSG_PEEK support for UDP adds this complexity, unfortunately.