Merge branch 'tcp-more-reliable-window-probes'

Eric Dumazet says:

====================
tcp: more reliable window probes

This series address a problem caused by small rto_min timers in DC,
leading to either timer storms or early flow terminations.

We also add two new SNMP counters for proper monitoring :
TCPWinProbe and TCPKeepAlive

v2: added TCPKeepAlive counter, as suggested by Yuchung & Neal
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

4 files changed, 12 insertions, 9 deletions

diff --git a/net/ipv4/proc.c b/net/ipv4/proc.c
index e1f3b911dd1e..da5d483e236a 100644
--- a/net/ipv4/proc.c
+++ b/net/ipv4/proc.c
@@ -298,6 +298,8 @@ static const struct snmp_mib snmp4_net_list[] = {
         SNMP_MIB_ITEM("TCPACKSkippedFinWait2", LINUX_MIB_TCPACKSKIPPEDFINWAIT2),
         SNMP_MIB_ITEM("TCPACKSkippedTimeWait", LINUX_MIB_TCPACKSKIPPEDTIMEWAIT),
         SNMP_MIB_ITEM("TCPACKSkippedChallenge", LINUX_MIB_TCPACKSKIPPEDCHALLENGE),
+        SNMP_MIB_ITEM("TCPWinProbe", LINUX_MIB_TCPWINPROBE),
+        SNMP_MIB_ITEM("TCPKeepAlive", LINUX_MIB_TCPKEEPALIVE),
         SNMP_MIB_SENTINEL
 };
 
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index df2ca615cd0c..cf8b20ff6658 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -3233,7 +3233,7 @@ static void tcp_ack_probe(struct sock *sk)
                  * This function is not for random using!
                  */
         } else {
-                unsigned long when = inet_csk_rto_backoff(icsk, TCP_RTO_MAX);
+                unsigned long when = tcp_probe0_when(sk, TCP_RTO_MAX);
 
                 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
                                           when, TCP_RTO_MAX);
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index a369e8a70b2c..7386d32cd670 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -3382,7 +3382,7 @@ EXPORT_SYMBOL_GPL(tcp_send_ack);
  * one is with SEG.SEQ=SND.UNA to deliver urgent pointer, another is
  * out-of-date with SND.UNA-1 to probe window.
  */
-static int tcp_xmit_probe_skb(struct sock *sk, int urgent)
+static int tcp_xmit_probe_skb(struct sock *sk, int urgent, int mib)
 {
         struct tcp_sock *tp = tcp_sk(sk);
         struct sk_buff *skb;
@@ -3400,6 +3400,7 @@ static int tcp_xmit_probe_skb(struct sock *sk, int urgent)
          */
         tcp_init_nondata_skb(skb, tp->snd_una - !urgent, TCPHDR_ACK);
         skb_mstamp_get(&skb->skb_mstamp);
+        NET_INC_STATS_BH(sock_net(sk), mib);
         return tcp_transmit_skb(sk, skb, 0, GFP_ATOMIC);
 }
 
@@ -3407,12 +3408,12 @@ void tcp_send_window_probe(struct sock *sk)
 {
         if (sk->sk_state == TCP_ESTABLISHED) {
                 tcp_sk(sk)->snd_wl1 = tcp_sk(sk)->rcv_nxt - 1;
-                tcp_xmit_probe_skb(sk, 0);
+                tcp_xmit_probe_skb(sk, 0, LINUX_MIB_TCPWINPROBE);
         }
 }
 
 /* Initiate keepalive or window probe from timer. */
-int tcp_write_wakeup(struct sock *sk)
+int tcp_write_wakeup(struct sock *sk, int mib)
 {
         struct tcp_sock *tp = tcp_sk(sk);
         struct sk_buff *skb;
@@ -3449,8 +3450,8 @@ int tcp_write_wakeup(struct sock *sk)
                 return err;
         } else {
                 if (between(tp->snd_up, tp->snd_una + 1, tp->snd_una + 0xFFFF))
-                        tcp_xmit_probe_skb(sk, 1);
-                return tcp_xmit_probe_skb(sk, 0);
+                        tcp_xmit_probe_skb(sk, 1, mib);
+                return tcp_xmit_probe_skb(sk, 0, mib);
         }
 }
 
@@ -3464,7 +3465,7 @@ void tcp_send_probe0(struct sock *sk)
         unsigned long probe_max;
         int err;
 
-        err = tcp_write_wakeup(sk);
+        err = tcp_write_wakeup(sk, LINUX_MIB_TCPWINPROBE);
 
         if (tp->packets_out || !tcp_send_head(sk)) {
                 /* Cancel probe timer, if it is not required. */
@@ -3490,7 +3491,7 @@ void tcp_send_probe0(struct sock *sk)
                 probe_max = TCP_RESOURCE_PROBE_INTERVAL;
         }
         inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
-                                  inet_csk_rto_backoff(icsk, probe_max),
+                                  tcp_probe0_when(sk, probe_max),
                                   TCP_RTO_MAX);
 }
 
diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c
index 8c65dc147d8b..65bf670e8714 100644
--- a/net/ipv4/tcp_timer.c
+++ b/net/ipv4/tcp_timer.c
@@ -616,7 +616,7 @@ static void tcp_keepalive_timer (unsigned long data)
                         tcp_write_err(sk);
                         goto out;
                 }
-                if (tcp_write_wakeup(sk) <= 0) {
+                if (tcp_write_wakeup(sk, LINUX_MIB_TCPKEEPALIVE) <= 0) {
                         icsk->icsk_probes_out++;
                         elapsed = keepalive_intvl_when(tp);
                 } else {