[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CANn89iLe9spogp7eaXPziA0L-FqJ0w=6VxdWDL6NKGobTyuQRw@mail.gmail.com>
Date: Sat, 3 Sep 2022 09:42:43 -0700
From: Eric Dumazet <edumazet@...gle.com>
To: Breno Leitao <leitao@...ian.org>
Cc: David Miller <davem@...emloft.net>,
Jakub Kicinski <kuba@...nel.org>,
netdev <netdev@...r.kernel.org>, leit@...com,
Hideaki YOSHIFUJI <yoshfuji@...ux-ipv6.org>,
Paolo Abeni <pabeni@...hat.com>,
David Ahern <dsahern@...nel.org>,
LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH RESEND net-next] tcp: socket-specific version of WARN_ON_ONCE()
On Wed, Aug 31, 2022 at 6:38 AM Breno Leitao <leitao@...ian.org> wrote:
>
> There are cases where we need information about the socket during a
> warning, so, it could help us to find bugs that happens that do not have
> a easily repro.
>
> BPF congestion control algorithms can change socket state in unexpected
> ways, leading to WARNings. Additional information about the socket state
> is useful to identify the culprit.
Well, this suggests we need to fix BPF side ?
Not sure how this can happen, because TCP_BPF_IW has
if (val <= 0 || tp->data_segs_out > tp->syn_data)
ret = -EINVAL;
else
tcp_snd_cwnd_set(tp, val);
It seems you already found the issue in an eBPF CC, can you share the details ?
>
> This diff creates a TCP socket-specific version of WARN_ON_ONCE(), and
> attaches it to tcp_snd_cwnd_set().
Well, I feel this will need constant additions... the state of a
custom BPF CC is opaque to core TCP stack anyway ?
>
> Signed-off-by: Breno Leitao <leitao@...ian.org>
> ---
> include/net/tcp.h | 3 ++-
> include/net/tcp_debug.h | 10 ++++++++++
> net/ipv4/tcp.c | 30 ++++++++++++++++++++++++++++++
> 3 files changed, 42 insertions(+), 1 deletion(-)
> create mode 100644 include/net/tcp_debug.h
>
> diff --git a/include/net/tcp.h b/include/net/tcp.h
> index d10962b9f0d0..73c3970d8839 100644
> --- a/include/net/tcp.h
> +++ b/include/net/tcp.h
> @@ -40,6 +40,7 @@
> #include <net/inet_ecn.h>
> #include <net/dst.h>
> #include <net/mptcp.h>
> +#include <net/tcp_debug.h>
>
> #include <linux/seq_file.h>
> #include <linux/memcontrol.h>
> @@ -1222,7 +1223,7 @@ static inline u32 tcp_snd_cwnd(const struct tcp_sock *tp)
>
> static inline void tcp_snd_cwnd_set(struct tcp_sock *tp, u32 val)
> {
> - WARN_ON_ONCE((int)val <= 0);
> + TCP_SOCK_WARN_ON_ONCE(tp, (int)val <= 0);
> tp->snd_cwnd = val;
> }
>
> diff --git a/include/net/tcp_debug.h b/include/net/tcp_debug.h
> new file mode 100644
> index 000000000000..50e96d87d335
> --- /dev/null
> +++ b/include/net/tcp_debug.h
> @@ -0,0 +1,10 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +#ifndef _LINUX_TCP_DEBUG_H
> +#define _LINUX_TCP_DEBUG_H
> +
> +void tcp_sock_warn(const struct tcp_sock *tp);
> +
> +#define TCP_SOCK_WARN_ON_ONCE(tcp_sock, condition) \
> + DO_ONCE_LITE_IF(condition, tcp_sock_warn, tcp_sock)
> +
> +#endif /* _LINUX_TCP_DEBUG_H */
> diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
> index bbe218753662..71771fee72f7 100644
> --- a/net/ipv4/tcp.c
> +++ b/net/ipv4/tcp.c
> @@ -4684,6 +4684,36 @@ int tcp_abort(struct sock *sk, int err)
> }
> EXPORT_SYMBOL_GPL(tcp_abort);
>
> +void tcp_sock_warn(const struct tcp_sock *tp)
> +{
> + const struct sock *sk = (const struct sock *)tp;
> + struct inet_sock *inet = inet_sk(sk);
> + struct inet_connection_sock *icsk = inet_csk(sk);
> +
> + WARN_ON(1);
> +
> + if (!tp)
> + return;
> +
> + pr_warn("Socket Info: family=%u state=%d sport=%u dport=%u ccname=%s cwnd=%u",
> + sk->sk_family, sk->sk_state, ntohs(inet->inet_sport),
> + ntohs(inet->inet_dport), icsk->icsk_ca_ops->name, tcp_snd_cwnd(tp));
> +
> + switch (sk->sk_family) {
> + case AF_INET:
> + pr_warn("saddr=%pI4 daddr=%pI4", &inet->inet_saddr,
> + &inet->inet_daddr);
> + break;
> +#if IS_ENABLED(CONFIG_IPV6)
> + case AF_INET6:
> + pr_warn("saddr=%pI6 daddr=%pI6", &sk->sk_v6_rcv_saddr,
> + &sk->sk_v6_daddr);
> + break;
> +#endif
> + }
> +}
> +EXPORT_SYMBOL_GPL(tcp_sock_warn);
> +
> extern struct tcp_congestion_ops tcp_reno;
>
> static __initdata unsigned long thash_entries;
> --
> 2.30.2
>
Powered by blists - more mailing lists