[TCP]: Add vegas congestion avoidance support.

A forward port of an old 2.3.x kernel hack done
years ago.  I (DaveM) did the first rough port,
Stephen Hemminger actually cleaned it up and
made it usable.

3 files changed, 68 insertions, 0 deletions

diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index f32c83568201..46c9268c9079 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -327,6 +327,10 @@ enum
 	NET_TCP_WESTWOOD=95,
 	NET_IPV4_IGMP_MAX_MSF=96,
 	NET_TCP_NO_METRICS_SAVE=97,
+	NET_TCP_VEGAS=98,
+	NET_TCP_VEGAS_ALPHA=99,
+	NET_TCP_VEGAS_BETA=100,
+	NET_TCP_VEGAS_GAMMA=101,
 };
 
 enum {
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index f8c91e734737..00bece841380 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -388,6 +388,18 @@ struct tcp_opt {
                 __u32    rtt;
                 __u32    rtt_min;          /* minimum observed RTT */
         } westwood;
+
+/* Vegas variables */
+	struct {
+		__u32	beg_snd_nxt;	/* right edge during last RTT */
+		__u32	beg_snd_una;	/* left edge  during last RTT */
+		__u32	beg_snd_cwnd;	/* saves the size of the cwnd */
+		__u8	do_vegas;	/* do vegas for this connection */
+		__u8	doing_vegas_now;/* if true, do vegas for this RTT */
+		__u16	cntRTT;		/* # of RTTs measured within last RTT */
+		__u32	minRTT;		/* min of RTTs measured within last RTT (in usec) */
+		__u32	baseRTT;	/* the min of all Vegas RTT measurements seen (in usec) */
+	} vegas;
 };
 
 /* WARNING: don't change the layout of the members in tcp_sock! */
diff --git a/include/net/tcp.h b/include/net/tcp.h
index af427a279038..076811e43c2c 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -583,6 +583,10 @@ extern int sysctl_tcp_tw_reuse;
 extern int sysctl_tcp_frto;
 extern int sysctl_tcp_low_latency;
 extern int sysctl_tcp_westwood;
+extern int sysctl_tcp_vegas_cong_avoid;
+extern int sysctl_tcp_vegas_alpha;
+extern int sysctl_tcp_vegas_beta;
+extern int sysctl_tcp_vegas_gamma;
 extern int sysctl_tcp_nometrics_save;
 
 extern atomic_t tcp_memory_allocated;
@@ -1212,8 +1216,56 @@ static inline __u32 tcp_recalc_ssthresh(struct tcp_opt *tp)
 	return max(tp->snd_cwnd >> 1U, 2U);
 }
 
+/* Stop taking Vegas samples for now. */
+#define tcp_vegas_disable(__tp)	((__tp)->vegas.doing_vegas_now = 0)
+
+/* Is this TCP connection using Vegas (regardless of whether it is taking
+ * Vegas measurements at the current time)?
+ */
+#define tcp_is_vegas(__tp)	((__tp)->vegas.do_vegas)
+    
+static inline void tcp_vegas_enable(struct tcp_opt *tp)
+{
+	/* There are several situations when we must "re-start" Vegas:
+	 *
+	 *  o when a connection is established
+	 *  o after an RTO
+	 *  o after fast recovery
+	 *  o when we send a packet and there is no outstanding
+	 *    unacknowledged data (restarting an idle connection)
+	 *
+	 * In these circumstances we cannot do a Vegas calculation at the
+	 * end of the first RTT, because any calculation we do is using
+	 * stale info -- both the saved cwnd and congestion feedback are
+	 * stale.
+	 *
+	 * Instead we must wait until the completion of an RTT during
+	 * which we actually receive ACKs.
+	 */
+    
+	/* Begin taking Vegas samples next time we send something. */
+	tp->vegas.doing_vegas_now = 1;
+     
+	/* Set the beginning of the next send window. */
+	tp->vegas.beg_snd_nxt = tp->snd_nxt;
+
+	tp->vegas.cntRTT = 0;
+	tp->vegas.minRTT = 0x7fffffff;
+}
+
+/* Should we be taking Vegas samples right now? */
+#define tcp_vegas_enabled(__tp)	((__tp)->vegas.doing_vegas_now)
+
+extern void tcp_vegas_init(struct tcp_opt *tp);
+
 static inline void tcp_set_ca_state(struct tcp_opt *tp, u8 ca_state)
 {
+	if (tcp_is_vegas(tp)) {
+		if (ca_state == TCP_CA_Open) 
+			tcp_vegas_enable(tp);
+		else
+			tcp_vegas_disable(tp);
+	}
 	tp->ca_state = ca_state;
 }