netdev - Re: [PATCH v5] sock: add tracepoint for send recv length

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: <20230110104419.67294691@gandalf.local.home>
Date:   Tue, 10 Jan 2023 10:44:19 -0500
From:   Steven Rostedt <rostedt@...dmis.org>
To:     Eric Dumazet <edumazet@...gle.com>
Cc:     Yunhui Cui <cuiyunhui@...edance.com>, mhiramat@...nel.org,
        davem@...emloft.net, kuba@...nel.org, pabeni@...hat.com,
        kuniyu@...zon.com, xiyou.wangcong@...il.com,
        duanxiongchun@...edance.com, linux-kernel@...r.kernel.org,
        linux-trace-kernel@...r.kernel.org, netdev@...r.kernel.org,
        dust.li@...ux.alibaba.com
Subject: Re: [PATCH v5] sock: add tracepoint for send recv length

On Tue, 10 Jan 2023 12:49:30 +0100
Eric Dumazet <edumazet@...gle.com> wrote:

> > +static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags)
> > +{
> > +       trace_sock_recv_length(sk, !(flags & MSG_PEEK) ? ret :
> > +                              (ret < 0 ? ret : 0), flags);  
> 
> Maybe we should only 'fast assign' the two fields (ret and flags),
> and let this logic happen later at 'print' time ?
> 
> This would reduce storage by one integer, and make fast path really fast.
> 
> This also could potentially remove the need for the peculiar construct with
> these noinline helpers.

I noticed that the trace_sock_send_length() doesn't have this logic, and
they are both DEFINE_EVENT() of the same DECLARE_EVENT_CLASS(). But we
could change this too, by the following:

/*
 * sock send/recv msg length
 */
DECLARE_EVENT_CLASS(sock_msg_length,

	TP_PROTO(struct sock *sk, int ret, int flags),

	TP_ARGS(sk, ret, flags),

	TP_STRUCT__entry(
		__field(void *, sk)
		__field(__u16, family)
		__field(__u16, protocol)
		__field(int, ret)
		__field(int, flags)
	),

	TP_fast_assign(
		__entry->sk = sk;
		__entry->family = sk->sk_family;
		__entry->protocol = sk->sk_protocol;
		__entry->length = ret > 0 ? ret : 0;
		__entry->error = ret < 0 ? ret : 0;
		__entry->flags = flags;
	),

	TP_printk("sk address = %p, family = %s protocol = %s, length = %d, error = %d, flags = 0x%x",
		  __entry->sk, show_family_name(__entry->family),
		  show_inet_protocol_name(__entry->protocol),
		  __entry->ret > 0 ? ret : 0, __entry->ret < 0 ? ret : 0,
		  __entry->flags)
);

DEFINE_EVENT(sock_msg_length, sock_send_length,
	TP_PROTO(struct sock *sk, int ret, int flags),

	TP_ARGS(sk, ret, flags)
);

DEFINE_EVENT_PRINT(sock_msg_length, sock_recv_length,
	TP_PROTO(struct sock *sk, int ret, int flags),

	TP_ARGS(sk, ret, flags)

	TP_printk("sk address = %p, family = %s protocol = %s, length = %d, error = %d, flags = 0x%x",
		  __entry->sk, show_family_name(__entry->family),
		  show_inet_protocol_name(__entry->protocol),
		  !(__entry->flags & MSG_PEEK) ? __entry->ret : __entry->ret > 0 ? ret : 0,
		  __entry->ret < 0 ? ret : 0,
		  __entry->flags)
);
#endif /* _TRACE_SOCK_H */

As DEFINE_EVENT_PRINT() uses the class template, but overrides the
TP_printk() portion (still saving memory).

And then both calls can just do:

	trace_sock_send_length(sk, ret, 0);

	trace_sock_recv_length(sock->sk, ret, flags);

And I bet that will also solve all the gcc being smart waste.

-- Steve