[PATCH net-next] tcp: make TCP_INFO more consistent

[PATCH net-next] tcp: make TCP_INFO more consistent

[Eric Dumazet]

From: Eric Dumazet <edumazet@google.com>

tcp_get_info() has to lock the socket, so lets lock it
for an extended critical section, so that various fields
have consistent values.

This solves an annoying issue that some applications
reported when multiple counters are updated during one
particular rx/rx event, and TCP_INFO was called from
another cpu.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
---
 net/ipv4/tcp.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index ec97e4b4a62f44decdcb559991500a737c7d6379..c8d46c140b4a47ee7c954317977393664e295abf 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2766,6 +2766,9 @@ void tcp_get_info(struct sock *sk, struct tcp_info *info)
 		info->tcpi_sacked = sk->sk_max_ack_backlog;
 		return;
 	}
+
+	slow = lock_sock_fast(sk);
+
 	info->tcpi_ca_state = icsk->icsk_ca_state;
 	info->tcpi_retransmits = icsk->icsk_retransmits;
 	info->tcpi_probes = icsk->icsk_probes_out;
@@ -2816,15 +2819,11 @@ void tcp_get_info(struct sock *sk, struct tcp_info *info)
 
 	info->tcpi_total_retrans = tp->total_retrans;
 
-	slow = lock_sock_fast(sk);
-
 	info->tcpi_bytes_acked = tp->bytes_acked;
 	info->tcpi_bytes_received = tp->bytes_received;
 	info->tcpi_notsent_bytes = max_t(int, 0, tp->write_seq - tp->snd_nxt);
 	tcp_get_info_chrono_stats(tp, info);
 
-	unlock_sock_fast(sk, slow);
-
 	info->tcpi_segs_out = tp->segs_out;
 	info->tcpi_segs_in = tp->segs_in;
 
@@ -2840,6 +2839,7 @@ void tcp_get_info(struct sock *sk, struct tcp_info *info)
 		do_div(rate64, intv);
 		info->tcpi_delivery_rate = rate64;
 	}
+	unlock_sock_fast(sk, slow);
 }
 EXPORT_SYMBOL_GPL(tcp_get_info);
 

Re: [PATCH net-next] tcp: make TCP_INFO more consistent

[David Miller]

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Mon, 09 Jan 2017 10:29:27 -0800

> From: Eric Dumazet <edumazet@google.com>
> 
> tcp_get_info() has to lock the socket, so lets lock it
> for an extended critical section, so that various fields
> have consistent values.
> 
> This solves an annoying issue that some applications
> reported when multiple counters are updated during one
> particular rx/rx event, and TCP_INFO was called from
> another cpu.
> 
> Signed-off-by: Eric Dumazet <edumazet@google.com>

Applied.