netdev - Re: [Patch bpf-next] tcp: fix sock skb accounting in tcp_read

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] [day] [month] [year] [list]

Message-ID: <CANn89iJ5yTX7f-Pv-yQs_kz+Xg7Ar7V_eF4cdeu6GimdVg-j9Q@mail.gmail.com>
Date:   Tue, 26 Jul 2022 18:53:48 +0200
From:   Eric Dumazet <edumazet@...gle.com>
To:     Cong Wang <xiyou.wangcong@...il.com>
Cc:     netdev <netdev@...r.kernel.org>, bpf <bpf@...r.kernel.org>,
        Cong Wang <cong.wang@...edance.com>,
        syzbot <syzbot+a0e6f8738b58f7654417@...kaller.appspotmail.com>,
        Stanislav Fomichev <sdf@...gle.com>,
        John Fastabend <john.fastabend@...il.com>
Subject: Re: [Patch bpf-next] tcp: fix sock skb accounting in tcp_read_skb()

On Tue, Jul 26, 2022 at 6:49 PM Eric Dumazet <edumazet@...gle.com> wrote:
>
> On Tue, Jul 26, 2022 at 6:47 PM Eric Dumazet <edumazet@...gle.com> wrote:
> >
> > On Tue, Jul 26, 2022 at 6:14 PM Cong Wang <xiyou.wangcong@...il.com> wrote:
> >
> > > If TCP really wants to queue a FIN with skb->len==0, then we have to
> > > adjust the return value for recv_actor(), because we currently use 0 as
> > > an error too (meaning no data is consumed):
> > >
> > >         if (sk_psock_verdict_apply(psock, skb, ret) < 0)
> > >                 len = 0;  // here!
> > > out:
> > >         rcu_read_unlock();
> > >         return len;
> > >
> > >
> > > BTW, what is wrong if we simply drop it before queueing to
> > > sk_receive_queue in TCP? Is it there just for collapse?
> >
> > Because an incoming segment can have payload and FIN.
> >
> > The consumer will need to consume the payload before FIN is considered/consumed,
> > with the complication of MSG_PEEK ...
> >
> > Right after tcp_read_skb() removes the skb from sk_receive_queue,
> > we need to update TCP state, regardless of recv_actor().
> >
> > Maybe like that:
> >
> > diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
> > index ba2bdc81137490bd1748cde95789f8d2bff3ab0f..6e2c11cd921872e406baffc475c9870e147578a1
> > 100644
> > --- a/net/ipv4/tcp.c
> > +++ b/net/ipv4/tcp.c
> > @@ -1759,20 +1759,15 @@ int tcp_read_skb(struct sock *sk,
> > skb_read_actor_t recv_actor)
> >                 int used;
> >
> >                 __skb_unlink(skb, &sk->sk_receive_queue);
> > +               seq = TCP_SKB_CB(skb)->end_seq;
> >                 used = recv_actor(sk, skb);
> >                 if (used <= 0) {
> >                         if (!copied)
> >                                 copied = used;
> >                         break;
> >                 }
> > -               seq += used;
> >                 copied += used;
> >
> > -               if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN) {
> > -                       consume_skb(skb);
> > -                       ++seq;
> > -                       break;
> > -               }
> >                 consume_skb(skb);
> >                 break;
> >         }
>
> Or even better:
>
> diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
> index ba2bdc81137490bd1748cde95789f8d2bff3ab0f..66c187a2592c042565211565adb3f40a811dfd7d
> 100644
> --- a/net/ipv4/tcp.c
> +++ b/net/ipv4/tcp.c
> @@ -1759,21 +1759,15 @@ int tcp_read_skb(struct sock *sk,
> skb_read_actor_t recv_actor)
>                 int used;
>
>                 __skb_unlink(skb, &sk->sk_receive_queue);
> +               seq = TCP_SKB_CB(skb)->end_seq;
>                 used = recv_actor(sk, skb);
> +               consume_skb(skb);
>                 if (used <= 0) {
>                         if (!copied)
>                                 copied = used;
>                         break;
>                 }
> -               seq += used;
>                 copied += used;
> -
> -               if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN) {
> -                       consume_skb(skb);
> -                       ++seq;
> -                       break;
> -               }
> -               consume_skb(skb);
>                 break;
>         }
>         WRITE_ONCE(tp->copied_seq, seq);

Note that this code will still not behave properly if we have in
receive queues two skbs of 1000 bytes of payload like:

seq 1:1001
seq 501:1501

tcp_recvmsg() would copy 1000 bytes from the first skb, then 500 bytes
from second skb.