tcp: fix SNMP TCP timeout under-estimation
authorYuchung Cheng <ycheng@google.com>
Thu, 29 Nov 2018 00:06:45 +0000 (16:06 -0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 21 Dec 2019 09:41:17 +0000 (10:41 +0100)
[ Upstream commit e1561fe2dd69dc5dddd69bd73aa65355bdfb048b ]

Previously the SNMP TCPTIMEOUTS counter has inconsistent accounting:
1. It counts all SYN and SYN-ACK timeouts
2. It counts timeouts in other states except recurring timeouts and
   timeouts after fast recovery or disorder state.

Such selective accounting makes analysis difficult and complicated. For
example the monitoring system needs to collect many other SNMP counters
to infer the total amount of timeout events. This patch makes TCPTIMEOUTS
counter simply counts all the retransmit timeout (SYN or data or FIN).

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
net/ipv4/tcp_timer.c

index ad0083f7b5dd30e4764fde503475f511dac57ec1..761a198ed5f307dec45b81e0aa7c1bc0a82fb6fe 100644 (file)
@@ -478,11 +478,12 @@ void tcp_retransmit_timer(struct sock *sk)
                goto out_reset_timer;
        }
 
+       __NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPTIMEOUTS);
        if (tcp_write_timeout(sk))
                goto out;
 
        if (icsk->icsk_retransmits == 0) {
-               int mib_idx;
+               int mib_idx = 0;
 
                if (icsk->icsk_ca_state == TCP_CA_Recovery) {
                        if (tcp_is_sack(tp))
@@ -497,10 +498,9 @@ void tcp_retransmit_timer(struct sock *sk)
                                mib_idx = LINUX_MIB_TCPSACKFAILURES;
                        else
                                mib_idx = LINUX_MIB_TCPRENOFAILURES;
-               } else {
-                       mib_idx = LINUX_MIB_TCPTIMEOUTS;
                }
-               __NET_INC_STATS(sock_net(sk), mib_idx);
+               if (mib_idx)
+                       __NET_INC_STATS(sock_net(sk), mib_idx);
        }
 
        tcp_enter_loss(sk);