Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [bpf-next 2/2] bpf: add BPF_SOCK_OPS_STATS_CB for tcp_stats()
To:     Yafang Shao <laoar.shao@gmail.com>, daniel@iogearbox.net,
        ast@kernel.org
Cc:     yhs@fb.com, brakmo@fb.com, edumazet@google.com,
        davem@davemloft.net, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org, shaoyafang@didiglobal.com
References: <1549971097-12627-1-git-send-email-laoar.shao@gmail.com>
 <1549971097-12627-3-git-send-email-laoar.shao@gmail.com>
From:   Eric Dumazet <eric.dumazet@gmail.com>
Message-ID: <e5fabb00-4fad-183a-5c2d-e22ed58006c7@gmail.com>
Date:   Tue, 12 Feb 2019 07:11:20 -0800
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
 Thunderbird/52.9.1
MIME-Version: 1.0
In-Reply-To: <1549971097-12627-3-git-send-email-laoar.shao@gmail.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


On 02/12/2019 03:31 AM, Yafang Shao wrote:
> Introuce this new op BPF_SOCK_OPS_STATS_CB for tcp_stats() such that it
> can be traced via BPF on a per socket basis.
> There's one argument in BPF_SOCK_OPS_STATS_CB, which is Linux MIB index
> LINUX_MIB_* to indicate the TCP event.
> All these Linux MIBs are defined in include/uapi/linux/snmp.h.
> 
> Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
> ---
>  include/uapi/linux/bpf.h | 5 +++++
>  net/ipv4/tcp_input.c     | 1 +
>  2 files changed, 6 insertions(+)
> 
> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> index 1777fa0..0314ddd 100644
> --- a/include/uapi/linux/bpf.h
> +++ b/include/uapi/linux/bpf.h
> @@ -2894,6 +2894,11 @@ enum {
>  	BPF_SOCK_OPS_TCP_LISTEN_CB,	/* Called on listen(2), right after
>  					 * socket transition to LISTEN state.
>  					 */
> +	BPF_SOCK_OPS_STATS_CB,		/*
> +					 * Called on tcp_stats().
> +					 * Arg1: Linux MIB index
> +					 * 	 LINUX_MIB_*
> +					 */
>  };
>  
>  /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
> diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
> index 88deb1f..4acf458 100644
> --- a/net/ipv4/tcp_input.c
> +++ b/net/ipv4/tcp_input.c
> @@ -3557,6 +3557,7 @@ static u32 tcp_newly_delivered(struct sock *sk, u32 prior_delivered, int flag)
>  static void tcp_stats(struct sock *sk, int mib_idx)
>  {
>  	NET_INC_STATS(sock_net(sk), mib_idx);
> +	tcp_call_bpf(sk, BPF_SOCK_OPS_STATS_CB, 1, &mib_idx);
>  }
>  
>  /* This routine deals with incoming acks, but not outgoing ones. */
> 

If the plan is to add to all NET_INC_STATS() calls in TCP an additional tcp_call_bpf()
I will say no.

So far, tcp_call_bpf() has not been used in fast path.