[TCP]: Add two new spurious RTO responses to FRTO

New sysctl tcp_frto_response is added to select amongst these
responses:
	- Rate halving based; reuses CA_CWR state (default)
	- Very conservative; used to be the only one available (=1)
	- Undo cwr; undoes ssthresh and cwnd reductions (=2)

The response with rate halving requires a new parameter to
tcp_enter_cwr because FRTO has already reduced ssthresh and
doing a second reduction there has to be prevented. In addition,
to keep things nice on 80 cols screen, a local variable was
added.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 98e0fd241a255063a54e5eda7620278d16f4f26d..c9ccb550206fb0accb948bd42ab655b4f3fefcb1 100644 (file)
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -439,6 +439,7 @@ enum
        NET_TCP_AVAIL_CONG_CONTROL=122,
        NET_TCP_ALLOWED_CONG_CONTROL=123,
        NET_TCP_MAX_SSTHRESH=124,
+       NET_TCP_FRTO_RESPONSE=125,
 };
 
 enum {

diff --git a/include/net/tcp.h b/include/net/tcp.h
index 6d09f5085f6a9ef4c6a46648b2e79a4cc912cfa2..f0c9e3400a090491598e438367e13d688829e9e0 100644 (file)
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -220,6 +220,7 @@ extern int sysctl_tcp_app_win;
 extern int sysctl_tcp_adv_win_scale;
 extern int sysctl_tcp_tw_reuse;
 extern int sysctl_tcp_frto;
+extern int sysctl_tcp_frto_response;
 extern int sysctl_tcp_low_latency;
 extern int sysctl_tcp_dma_copybreak;
 extern int sysctl_tcp_nometrics_save;
@@ -738,7 +739,7 @@ static inline void tcp_sync_left_out(struct tcp_sock *tp)
        tp->left_out = tp->sacked_out + tp->lost_out;
 }
 
-extern void tcp_enter_cwr(struct sock *sk);
+extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
 extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
 
 /* Slow start with delack produces 3 packets of burst, so that

diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c
index d68effe98e8db7d6d994651c5bd1ce028115f6de..6817d6485df5c195430ab808c2c345e8d6e15ba4 100644 (file)
--- a/net/ipv4/sysctl_net_ipv4.c
+++ b/net/ipv4/sysctl_net_ipv4.c
@@ -646,6 +646,14 @@ ctl_table ipv4_table[] = {
                .mode           = 0644,
                .proc_handler   = &proc_dointvec
        },
+       {
+               .ctl_name       = NET_TCP_FRTO_RESPONSE,
+               .procname       = "tcp_frto_response",
+               .data           = &sysctl_tcp_frto_response,
+               .maxlen         = sizeof(int),
+               .mode           = 0644,
+               .proc_handler   = &proc_dointvec
+       },
        {
                .ctl_name       = NET_TCP_LOW_LATENCY,
                .procname       = "tcp_low_latency",

diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index f6ba07f0d81688ed81ed962bd1425f5548d962fe..322e43c56461a950c14ee369d6ccf2842e015157 100644 (file)
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -86,6 +86,7 @@ int sysctl_tcp_stdurg __read_mostly;
 int sysctl_tcp_rfc1337 __read_mostly;
 int sysctl_tcp_max_orphans __read_mostly = NR_FILE;
 int sysctl_tcp_frto __read_mostly;
+int sysctl_tcp_frto_response __read_mostly;
 int sysctl_tcp_nometrics_save __read_mostly;
 
 int sysctl_tcp_moderate_rcvbuf __read_mostly = 1;
@@ -762,15 +763,17 @@ __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst)
 }
 
 /* Set slow start threshold and cwnd not falling to slow start */
-void tcp_enter_cwr(struct sock *sk)
+void tcp_enter_cwr(struct sock *sk, const int set_ssthresh)
 {
        struct tcp_sock *tp = tcp_sk(sk);
+       const struct inet_connection_sock *icsk = inet_csk(sk);
 
        tp->prior_ssthresh = 0;
        tp->bytes_acked = 0;
        if (inet_csk(sk)->icsk_ca_state < TCP_CA_CWR) {
                tp->undo_marker = 0;
-               tp->snd_ssthresh = inet_csk(sk)->icsk_ca_ops->ssthresh(sk);
+               if (set_ssthresh)
+                       tp->snd_ssthresh = icsk->icsk_ca_ops->ssthresh(sk);
                tp->snd_cwnd = min(tp->snd_cwnd,
                                   tcp_packets_in_flight(tp) + 1U);
                tp->snd_cwnd_cnt = 0;
@@ -2003,7 +2006,7 @@ static void tcp_try_to_open(struct sock *sk, struct tcp_sock *tp, int flag)
                tp->retrans_stamp = 0;
 
        if (flag&FLAG_ECE)
-               tcp_enter_cwr(sk);
+               tcp_enter_cwr(sk, 1);
 
        if (inet_csk(sk)->icsk_ca_state != TCP_CA_CWR) {
                int state = TCP_CA_Open;
@@ -2579,6 +2582,21 @@ static void tcp_conservative_spur_to_response(struct tcp_sock *tp)
        tcp_moderate_cwnd(tp);
 }
 
+/* A conservative spurious RTO response algorithm: reduce cwnd using
+ * rate halving and continue in congestion avoidance.
+ */
+static void tcp_ratehalving_spur_to_response(struct sock *sk)
+{
+       struct tcp_sock *tp = tcp_sk(sk);
+       tcp_enter_cwr(sk, 0);
+       tp->high_seq = tp->frto_highmark;       /* Smoother w/o this? - ij */
+}
+
+static void tcp_undo_spur_to_response(struct sock *sk)
+{
+       tcp_undo_cwr(sk, 1);
+}
+
 /* F-RTO spurious RTO detection algorithm (RFC4138)
  *
  * F-RTO affects during two new ACKs following RTO (well, almost, see inline
@@ -2661,7 +2679,17 @@ static int tcp_process_frto(struct sock *sk, u32 prior_snd_una, int flag)
                tp->frto_counter = 2;
                return 1;
        } else /* frto_counter == 2 */ {
-               tcp_conservative_spur_to_response(tp);
+               switch (sysctl_tcp_frto_response) {
+               case 2:
+                       tcp_undo_spur_to_response(sk);
+                       break;
+               case 1:
+                       tcp_conservative_spur_to_response(tp);
+                       break;
+               default:
+                       tcp_ratehalving_spur_to_response(sk);
+                       break;
+               };
                tp->frto_counter = 0;
        }
        return 0;

diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 3c24881f2a6598941fbc5663d8bc6c2263e0ee21..d19b2f3b70fd0a53af71a36688e17d7d4b320222 100644 (file)
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -545,7 +545,7 @@ static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, int clone_it,
        if (likely(err <= 0))
                return err;
 
-       tcp_enter_cwr(sk);
+       tcp_enter_cwr(sk, 1);
 
        return net_xmit_eval(err);