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

Message-ID: <CANn89iJSQh-5DAhEL4Fh5ZDrtY47y0Mo9YJbG-rnj17pdXqoXA@mail.gmail.com>
Date:   Tue, 12 Jul 2022 15:20:37 +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 Sun, Jul 10, 2022 at 12:20 AM Cong Wang <xiyou.wangcong@...il.com> wrote:
>
> From: Cong Wang <cong.wang@...edance.com>
>
> Before commit 965b57b469a5 ("net: Introduce a new proto_ops
> ->read_skb()"), skb was not dequeued from receive queue hence
> when we close TCP socket skb can be just flushed synchronously.
>
> After this commit, we have to uncharge skb immediately after being
> dequeued, otherwise it is still charged in the original sock. And we
> still need to retain skb->sk, as eBPF programs may extract sock
> information from skb->sk. Therefore, we have to call
> skb_set_owner_sk_safe() here.
>
> Fixes: 965b57b469a5 ("net: Introduce a new proto_ops ->read_skb()")
> Reported-and-tested-by: syzbot+a0e6f8738b58f7654417@...kaller.appspotmail.com
> Tested-by: Stanislav Fomichev <sdf@...gle.com>
> Cc: Eric Dumazet <edumazet@...gle.com>
> Cc: John Fastabend <john.fastabend@...il.com>
> Signed-off-by: Cong Wang <cong.wang@...edance.com>
> ---
>  net/ipv4/tcp.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
> index 9d2fd3ced21b..c6b1effb2afd 100644
> --- a/net/ipv4/tcp.c
> +++ b/net/ipv4/tcp.c
> @@ -1749,6 +1749,7 @@ int tcp_read_skb(struct sock *sk, skb_read_actor_t recv_actor)
>                 int used;
>
>                 __skb_unlink(skb, &sk->sk_receive_queue);
> +               WARN_ON(!skb_set_owner_sk_safe(skb, sk));
>                 used = recv_actor(sk, skb);
>                 if (used <= 0) {
>                         if (!copied)
> --
> 2.34.1
>

I am reading tcp_read_skb(),it seems to have other bugs.
I wonder why syzbot has not caught up yet.

It ignores the offset value from tcp_recv_skb(), this looks wrong to me.
The reason tcp_read_sock() passes a @len parameter is that is it not
skb->len, but (skb->len - offset)

Also if recv_actor(sk, skb) returns 0, we probably still need to
advance tp->copied_seq,
for instance if skb had a pure FIN (and thus skb->len == 0), since you
removed the skb from sk_receive_queue ?