Merge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6

Documentation/networking/ip-sysctl.txt

@@ -78,6 +78,11 @@ inet_peer_gc_maxtime - INTEGER
 
 TCP variables: 
 
+tcp_abc - INTEGER
+	Controls Appropriate Byte Count defined in RFC3465. If set to
+	0 then does congestion avoid once per ack. 1 is conservative
+	value, and 2 is more agressive.
+
 tcp_syn_retries - INTEGER
 	Number of times initial SYNs for an active TCP connection attempt
 	will be retransmitted. Should not be higher than 255. Default value

include/linux/sysctl.h

@@ -390,6 +390,7 @@ enum
 	NET_TCP_BIC_BETA=108,
 	NET_IPV4_ICMP_ERRORS_USE_INBOUND_IFADDR=109,
 	NET_TCP_CONG_CONTROL=110,
+	NET_TCP_ABC=111,
 };
 
 enum {

include/linux/tcp.h

@@ -307,6 +307,21 @@ struct tcp_sock {
 	struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
 	struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
 
+	struct tcp_sack_block recv_sack_cache[4];
+
+	/* from STCP, retrans queue hinting */
+	struct sk_buff* lost_skb_hint;
+
+	struct sk_buff *scoreboard_skb_hint;
+	struct sk_buff *retransmit_skb_hint;
+	struct sk_buff *forward_skb_hint;
+	struct sk_buff *fastpath_skb_hint;
+
+	int     fastpath_cnt_hint;
+	int     lost_cnt_hint;
+	int     retransmit_cnt_hint;
+	int     forward_cnt_hint;
+
 	__u16	advmss;		/* Advertised MSS			*/
 	__u16	prior_ssthresh; /* ssthresh saved at recovery start	*/
 	__u32	lost_out;	/* Lost packets			*/
@@ -326,6 +341,7 @@ struct tcp_sock {
 	__u32	snd_up;		/* Urgent pointer		*/
 
 	__u32	total_retrans;	/* Total retransmits for entire connection */
+	__u32	bytes_acked;	/* Appropriate Byte Counting - RFC3465 */
 
 	unsigned int		keepalive_time;	  /* time before keep alive takes place */
 	unsigned int		keepalive_intvl;  /* time interval between keep alive probes */

include/net/sock.h

@@ -1247,6 +1247,12 @@ static inline struct page *sk_stream_alloc_page(struct sock *sk)
 		     (skb != (struct sk_buff *)&(sk)->sk_write_queue);	\
 		     skb = skb->next)
 
+/*from STCP for fast SACK Process*/
+#define sk_stream_for_retrans_queue_from(skb, sk)			\
+		for (; (skb != (sk)->sk_send_head) &&                   \
+		     (skb != (struct sk_buff *)&(sk)->sk_write_queue);	\
+		     skb = skb->next)
+
 /*
  *	Default write policy as shown to user space via poll/select/SIGIO
  */

include/net/tcp.h

@@ -89,10 +89,10 @@ extern void tcp_time_wait(struct sock *sk, int state, int timeo);
 				 */
 
 #define TCP_SYN_RETRIES	 5	/* number of times to retry active opening a
-				 * connection: ~180sec is RFC minumum	*/
+				 * connection: ~180sec is RFC minimum	*/
 
 #define TCP_SYNACK_RETRIES 5	/* number of times to retry passive opening a
-				 * connection: ~180sec is RFC minumum	*/
+				 * connection: ~180sec is RFC minimum	*/
 
 
 #define TCP_ORPHAN_RETRIES 7	/* number of times to retry on an orphaned
@@ -180,7 +180,7 @@ extern void tcp_time_wait(struct sock *sk, int state, int timeo);
 /* Flags in tp->nonagle */
 #define TCP_NAGLE_OFF		1	/* Nagle's algo is disabled */
 #define TCP_NAGLE_CORK		2	/* Socket is corked	    */
-#define TCP_NAGLE_PUSH		4	/* Cork is overriden for already queued data */
+#define TCP_NAGLE_PUSH		4	/* Cork is overridden for already queued data */
 
 extern struct inet_timewait_death_row tcp_death_row;
 
@@ -218,6 +218,7 @@ extern int sysctl_tcp_low_latency;
 extern int sysctl_tcp_nometrics_save;
 extern int sysctl_tcp_moderate_rcvbuf;
 extern int sysctl_tcp_tso_win_divisor;
+extern int sysctl_tcp_abc;
 
 extern atomic_t tcp_memory_allocated;
 extern atomic_t tcp_sockets_allocated;
@@ -551,13 +552,13 @@ extern u32	__tcp_select_window(struct sock *sk);
 
 /* TCP timestamps are only 32-bits, this causes a slight
  * complication on 64-bit systems since we store a snapshot
- * of jiffies in the buffer control blocks below.  We decidely
+ * of jiffies in the buffer control blocks below.  We decidedly
  * only use of the low 32-bits of jiffies and hide the ugly
  * casts with the following macro.
  */
 #define tcp_time_stamp		((__u32)(jiffies))
 
-/* This is what the send packet queueing engine uses to pass
+/* This is what the send packet queuing engine uses to pass
  * TCP per-packet control information to the transmission
  * code.  We also store the host-order sequence numbers in
  * here too.  This is 36 bytes on 32-bit architectures,
@@ -597,7 +598,7 @@ struct tcp_skb_cb {
 #define TCPCB_EVER_RETRANS	0x80	/* Ever retransmitted frame	*/
 #define TCPCB_RETRANS		(TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
 
-#define TCPCB_URG		0x20	/* Urgent pointer advenced here	*/
+#define TCPCB_URG		0x20	/* Urgent pointer advanced here	*/
 
 #define TCPCB_AT_TAIL		(TCPCB_URG)
 
@@ -765,6 +766,33 @@ static inline __u32 tcp_current_ssthresh(const struct sock *sk)
 			    (tp->snd_cwnd >> 2)));
 }
 
+/*
+ * Linear increase during slow start
+ */
+static inline void tcp_slow_start(struct tcp_sock *tp)
+{
+	if (sysctl_tcp_abc) {
+		/* RFC3465: Slow Start
+		 * TCP sender SHOULD increase cwnd by the number of
+		 * previously unacknowledged bytes ACKed by each incoming
+		 * acknowledgment, provided the increase is not more than L
+		 */
+		if (tp->bytes_acked < tp->mss_cache)
+			return;
+
+		/* We MAY increase by 2 if discovered delayed ack */
+		if (sysctl_tcp_abc > 1 && tp->bytes_acked > 2*tp->mss_cache) {
+			if (tp->snd_cwnd < tp->snd_cwnd_clamp)
+				tp->snd_cwnd++;
+		}
+	}
+	tp->bytes_acked = 0;
+
+	if (tp->snd_cwnd < tp->snd_cwnd_clamp)
+		tp->snd_cwnd++;
+}
+
+
 static inline void tcp_sync_left_out(struct tcp_sock *tp)
 {
 	if (tp->rx_opt.sack_ok &&
@@ -794,6 +822,7 @@ static inline void tcp_enter_cwr(struct sock *sk)
 	struct tcp_sock *tp = tcp_sk(sk);
 
 	tp->prior_ssthresh = 0;
+	tp->bytes_acked = 0;
 	if (inet_csk(sk)->icsk_ca_state < TCP_CA_CWR) {
 		__tcp_enter_cwr(sk);
 		tcp_set_ca_state(sk, TCP_CA_CWR);
@@ -810,6 +839,27 @@ static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
 	return 3;
 }
 
+/* RFC2861 Check whether we are limited by application or congestion window
+ * This is the inverse of cwnd check in tcp_tso_should_defer
+ */
+static inline int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight)
+{
+	const struct tcp_sock *tp = tcp_sk(sk);
+	u32 left;
+
+	if (in_flight >= tp->snd_cwnd)
+		return 1;
+
+	if (!(sk->sk_route_caps & NETIF_F_TSO))
+		return 0;
+
+	left = tp->snd_cwnd - in_flight;
+	if (sysctl_tcp_tso_win_divisor)
+		return left * sysctl_tcp_tso_win_divisor < tp->snd_cwnd;
+	else
+		return left <= tcp_max_burst(tp);
+}
+
 static __inline__ void tcp_minshall_update(struct tcp_sock *tp, int mss, 
 					   const struct sk_buff *skb)
 {
@@ -1157,6 +1207,15 @@ static inline void tcp_mib_init(void)
 	TCP_ADD_STATS_USER(TCP_MIB_MAXCONN, -1);
 }
 
+/*from STCP */
+static inline void clear_all_retrans_hints(struct tcp_sock *tp){
+	tp->lost_skb_hint = NULL;
+	tp->scoreboard_skb_hint = NULL;
+	tp->retransmit_skb_hint = NULL;
+	tp->forward_skb_hint = NULL;
+	tp->fastpath_skb_hint = NULL;
+}
+
 /* /proc */
 enum tcp_seq_states {
 	TCP_SEQ_STATE_LISTENING,

net/ipv4/sysctl_net_ipv4.c

@@ -645,6 +645,14 @@ ctl_table ipv4_table[] = {
 		.proc_handler	= &proc_tcp_congestion_control,
 		.strategy	= &sysctl_tcp_congestion_control,
 	},
+	{
+		.ctl_name	= NET_TCP_ABC,
+		.procname	= "tcp_abc",
+		.data		= &sysctl_tcp_abc,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec,
+	},
 
 	{ .ctl_name = 0 }
 };

net/ipv4/tcp.c

@@ -1640,7 +1640,7 @@ int tcp_disconnect(struct sock *sk, int flags)
 	} else if (tcp_need_reset(old_state) ||
 		   (tp->snd_nxt != tp->write_seq &&
 		    (1 << old_state) & (TCPF_CLOSING | TCPF_LAST_ACK))) {
-		/* The last check adjusts for discrepance of Linux wrt. RFC
+		/* The last check adjusts for discrepancy of Linux wrt. RFC
 		 * states
 		 */
 		tcp_send_active_reset(sk, gfp_any());
@@ -1669,6 +1669,7 @@ int tcp_disconnect(struct sock *sk, int flags)
 	tp->packets_out = 0;
 	tp->snd_ssthresh = 0x7fffffff;
 	tp->snd_cwnd_cnt = 0;
+	tp->bytes_acked = 0;
 	tcp_set_ca_state(sk, TCP_CA_Open);
 	tcp_clear_retrans(tp);
 	inet_csk_delack_init(sk);

net/ipv4/tcp_bic.c

@@ -217,14 +217,12 @@ static void bictcp_cong_avoid(struct sock *sk, u32 ack,
 
 	bictcp_low_utilization(sk, data_acked);
 
-	if (in_flight < tp->snd_cwnd)
+	if (!tcp_is_cwnd_limited(sk, in_flight))
 		return;
 
-	if (tp->snd_cwnd <= tp->snd_ssthresh) {
-		/* In "safe" area, increase. */
-		if (tp->snd_cwnd < tp->snd_cwnd_clamp)
-			tp->snd_cwnd++;
-	} else {
+	if (tp->snd_cwnd <= tp->snd_ssthresh)
+		tcp_slow_start(tp);
+	else {
 		bictcp_update(ca, tp->snd_cwnd);
 
 		/* In dangerous area, increase slowly.

net/ipv4/tcp_cong.c

@@ -186,17 +186,25 @@ void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 rtt, u32 in_flight,
 {
 	struct tcp_sock *tp = tcp_sk(sk);
 
-	if (in_flight < tp->snd_cwnd)
+	if (!tcp_is_cwnd_limited(sk, in_flight))
 		return;
 
-        if (tp->snd_cwnd <= tp->snd_ssthresh) {
 	/* In "safe" area, increase. */
+        if (tp->snd_cwnd <= tp->snd_ssthresh)
+		tcp_slow_start(tp);
+
+ 	/* In dangerous area, increase slowly. */
+	else if (sysctl_tcp_abc) {
+ 		/* RFC3465: Apppriate Byte Count
+ 		 * increase once for each full cwnd acked
+ 		 */
+ 		if (tp->bytes_acked >= tp->snd_cwnd*tp->mss_cache) {
+ 			tp->bytes_acked -= tp->snd_cwnd*tp->mss_cache;
  			if (tp->snd_cwnd < tp->snd_cwnd_clamp)
  				tp->snd_cwnd++;
+ 		}
  	} else {
-                /* In dangerous area, increase slowly.
-		 * In theory this is tp->snd_cwnd += 1 / tp->snd_cwnd
-		 */
+ 		/* In theory this is tp->snd_cwnd += 1 / tp->snd_cwnd */
  		if (tp->snd_cwnd_cnt >= tp->snd_cwnd) {
  			if (tp->snd_cwnd < tp->snd_cwnd_clamp)
  				tp->snd_cwnd++;

net/ipv4/tcp_highspeed.c

@@ -111,18 +111,17 @@ static void hstcp_init(struct sock *sk)
 }
 
 static void hstcp_cong_avoid(struct sock *sk, u32 adk, u32 rtt,
-			     u32 in_flight, int good)
+			     u32 in_flight, u32 pkts_acked)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
 	struct hstcp *ca = inet_csk_ca(sk);
 
-	if (in_flight < tp->snd_cwnd)
+	if (!tcp_is_cwnd_limited(sk, in_flight))
 		return;
 
-	if (tp->snd_cwnd <= tp->snd_ssthresh) {
-		if (tp->snd_cwnd < tp->snd_cwnd_clamp)
-			tp->snd_cwnd++;
-	} else {
+	if (tp->snd_cwnd <= tp->snd_ssthresh)
+		tcp_slow_start(tp);
+	else {
 		/* Update AIMD parameters */
 		if (tp->snd_cwnd > hstcp_aimd_vals[ca->ai].cwnd) {
 			while (tp->snd_cwnd > hstcp_aimd_vals[ca->ai].cwnd &&

net/ipv4/tcp_htcp.c

@@ -207,14 +207,13 @@ static void htcp_cong_avoid(struct sock *sk, u32 ack, u32 rtt,
 	struct tcp_sock *tp = tcp_sk(sk);
 	struct htcp *ca = inet_csk_ca(sk);
 
-	if (in_flight < tp->snd_cwnd)
+	if (!tcp_is_cwnd_limited(sk, in_flight))
 		return;
 
-        if (tp->snd_cwnd <= tp->snd_ssthresh) {
-                /* In "safe" area, increase. */
-		if (tp->snd_cwnd < tp->snd_cwnd_clamp)
-			tp->snd_cwnd++;
-	} else {
+        if (tp->snd_cwnd <= tp->snd_ssthresh)
+		tcp_slow_start(tp);
+	else {
+
 		measure_rtt(sk);
 
 		/* keep track of number of round-trip times since last backoff event */

net/ipv4/tcp_hybla.c

@@ -100,12 +100,12 @@ static void hybla_cong_avoid(struct sock *sk, u32 ack, u32 rtt,
 		ca->minrtt = tp->srtt;
 	}
 
+	if (!tcp_is_cwnd_limited(sk, in_flight))
+		return;
+
 	if (!ca->hybla_en)
 		return tcp_reno_cong_avoid(sk, ack, rtt, in_flight, flag);
 
-	if (in_flight < tp->snd_cwnd)
-		return;
-
 	if (ca->rho == 0)
 		hybla_recalc_param(sk);
 

net/ipv4/tcp_ipv4.c

@@ -39,7 +39,7 @@
  *					request_sock handling and moved
  *					most of it into the af independent code.
  *					Added tail drop and some other bugfixes.
- *					Added new listen sematics.
+ *					Added new listen semantics.
  *		Mike McLagan	:	Routing by source
  *	Juan Jose Ciarlante:		ip_dynaddr bits
  *		Andi Kleen:		various fixes.
@@ -1210,7 +1210,7 @@ int tcp_v4_rcv(struct sk_buff *skb)
 
 	/* An explanation is required here, I think.
 	 * Packet length and doff are validated by header prediction,
-	 * provided case of th->doff==0 is elimineted.
+	 * provided case of th->doff==0 is eliminated.
 	 * So, we defer the checks. */
 	if ((skb->ip_summed != CHECKSUM_UNNECESSARY &&
 	     tcp_v4_checksum_init(skb)))

net/ipv4/tcp_minisocks.c

@@ -158,7 +158,7 @@ tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb,
 		/* I am shamed, but failed to make it more elegant.
 		 * Yes, it is direct reference to IP, which is impossible
 		 * to generalize to IPv6. Taking into account that IPv6
-		 * do not undertsnad recycling in any case, it not
+		 * do not understand recycling in any case, it not
 		 * a big problem in practice. --ANK */
 		if (tw->tw_family == AF_INET &&
 		    tcp_death_row.sysctl_tw_recycle && tcptw->tw_ts_recent_stamp &&
@@ -194,7 +194,7 @@ tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb,
 		/* In window segment, it may be only reset or bare ack. */
 
 		if (th->rst) {
-			/* This is TIME_WAIT assasination, in two flavors.
+			/* This is TIME_WAIT assassination, in two flavors.
 			 * Oh well... nobody has a sufficient solution to this
 			 * protocol bug yet.
 			 */
@@ -380,6 +380,7 @@ struct sock *tcp_create_openreq_child(struct sock *sk, struct request_sock *req,
 		 */
 		newtp->snd_cwnd = 2;
 		newtp->snd_cwnd_cnt = 0;
+		newtp->bytes_acked = 0;
 
 		newtp->frto_counter = 0;
 		newtp->frto_highmark = 0;
@@ -550,7 +551,7 @@ struct sock *tcp_check_req(struct sock *sk,struct sk_buff *skb,
 
 	/* RFC793 page 36: "If the connection is in any non-synchronized state ...
 	 *                  and the incoming segment acknowledges something not yet
-	 *                  sent (the segment carries an unaccaptable ACK) ...
+	 *                  sent (the segment carries an unacceptable ACK) ...
 	 *                  a reset is sent."
 	 *
 	 * Invalid ACK: reset will be sent by listening socket

net/ipv4/tcp_output.c

@@ -436,6 +436,8 @@ int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len, unsigned int mss
 	u16 flags;
 
 	BUG_ON(len > skb->len);
+
+ 	clear_all_retrans_hints(tp);
 	nsize = skb_headlen(skb) - len;
 	if (nsize < 0)
 		nsize = 0;
@@ -599,7 +601,7 @@ int tcp_trim_head(struct sock *sk, struct sk_buff *skb, u32 len)
    for TCP options, but includes only bare TCP header.
 
    tp->rx_opt.mss_clamp is mss negotiated at connection setup.
-   It is minumum of user_mss and mss received with SYN.
+   It is minimum of user_mss and mss received with SYN.
    It also does not include TCP options.
 
    tp->pmtu_cookie is last pmtu, seen by this function.
@@ -1171,7 +1173,7 @@ u32 __tcp_select_window(struct sock *sk)
 {
 	struct inet_connection_sock *icsk = inet_csk(sk);
 	struct tcp_sock *tp = tcp_sk(sk);
-	/* MSS for the peer's data.  Previous verions used mss_clamp
+	/* MSS for the peer's data.  Previous versions used mss_clamp
 	 * here.  I don't know if the value based on our guesses
 	 * of peer's MSS is better for the performance.  It's more correct
 	 * but may be worse for the performance because of rcv_mss
@@ -1260,6 +1262,9 @@ static void tcp_retrans_try_collapse(struct sock *sk, struct sk_buff *skb, int m
 		BUG_ON(tcp_skb_pcount(skb) != 1 ||
 		       tcp_skb_pcount(next_skb) != 1);
 
+		/* changing transmit queue under us so clear hints */
+		clear_all_retrans_hints(tp);
+
 		/* Ok.	We will be able to collapse the packet. */
 		__skb_unlink(next_skb, &sk->sk_write_queue);
 
@@ -1330,6 +1335,8 @@ void tcp_simple_retransmit(struct sock *sk)
 		}
 	}
 
+	clear_all_retrans_hints(tp);
+
 	if (!lost)
 		return;
 
@@ -1361,7 +1368,7 @@ int tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb)
 	int err;
 
 	/* Do not sent more than we queued. 1/4 is reserved for possible
-	 * copying overhead: frgagmentation, tunneling, mangling etc.
+	 * copying overhead: fragmentation, tunneling, mangling etc.
 	 */
 	if (atomic_read(&sk->sk_wmem_alloc) >
 	    min(sk->sk_wmem_queued + (sk->sk_wmem_queued >> 2), sk->sk_sndbuf))
@@ -1468,13 +1475,25 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
 	const struct inet_connection_sock *icsk = inet_csk(sk);
 	struct tcp_sock *tp = tcp_sk(sk);
 	struct sk_buff *skb;
-	int packet_cnt = tp->lost_out;
+	int packet_cnt;
+
+	if (tp->retransmit_skb_hint) {
+		skb = tp->retransmit_skb_hint;
+		packet_cnt = tp->retransmit_cnt_hint;
+	}else{
+		skb = sk->sk_write_queue.next;
+		packet_cnt = 0;
+	}
 
 	/* First pass: retransmit lost packets. */
-	if (packet_cnt) {
-		sk_stream_for_retrans_queue(skb, sk) {
+	if (tp->lost_out) {
+		sk_stream_for_retrans_queue_from(skb, sk) {
 			__u8 sacked = TCP_SKB_CB(skb)->sacked;
 
+			/* we could do better than to assign each time */
+			tp->retransmit_skb_hint = skb;
+			tp->retransmit_cnt_hint = packet_cnt;
+
 			/* Assume this retransmit will generate
 			 * only one packet for congestion window
 			 * calculation purposes.  This works because
@@ -1485,10 +1504,12 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
 			if (tcp_packets_in_flight(tp) >= tp->snd_cwnd)
 				return;
 
-			if (sacked&TCPCB_LOST) {
+			if (sacked & TCPCB_LOST) {
 				if (!(sacked&(TCPCB_SACKED_ACKED|TCPCB_SACKED_RETRANS))) {
-					if (tcp_retransmit_skb(sk, skb))
+					if (tcp_retransmit_skb(sk, skb)) {
+						tp->retransmit_skb_hint = NULL;
 						return;
+					}
 					if (icsk->icsk_ca_state != TCP_CA_Loss)
 						NET_INC_STATS_BH(LINUX_MIB_TCPFASTRETRANS);
 					else
@@ -1501,8 +1522,8 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
 									  TCP_RTO_MAX);
 				}
 
-				packet_cnt -= tcp_skb_pcount(skb);
-				if (packet_cnt <= 0)
+				packet_cnt += tcp_skb_pcount(skb);
+				if (packet_cnt >= tp->lost_out)
 					break;
 			}
 		}
@@ -1528,9 +1549,18 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
 	if (tcp_may_send_now(sk, tp))
 		return;
 
+	if (tp->forward_skb_hint) {
+		skb = tp->forward_skb_hint;
+		packet_cnt = tp->forward_cnt_hint;
+	} else{
+		skb = sk->sk_write_queue.next;
 		packet_cnt = 0;
+	}
+
+	sk_stream_for_retrans_queue_from(skb, sk) {
+		tp->forward_cnt_hint = packet_cnt;
+		tp->forward_skb_hint = skb;
 
-	sk_stream_for_retrans_queue(skb, sk) {
 		/* Similar to the retransmit loop above we
 		 * can pretend that the retransmitted SKB
 		 * we send out here will be composed of one
@@ -1547,8 +1577,10 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
 			continue;
 
 		/* Ok, retransmit it. */
-		if (tcp_retransmit_skb(sk, skb))
+		if (tcp_retransmit_skb(sk, skb)) {
+			tp->forward_skb_hint = NULL;
 			break;
+		}
 
 		if (skb == skb_peek(&sk->sk_write_queue))
 			inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
@@ -2058,3 +2090,4 @@ EXPORT_SYMBOL(tcp_connect);
 EXPORT_SYMBOL(tcp_make_synack);
 EXPORT_SYMBOL(tcp_simple_retransmit);
 EXPORT_SYMBOL(tcp_sync_mss);
+EXPORT_SYMBOL(sysctl_tcp_tso_win_divisor);

net/ipv4/tcp_scalable.c

@@ -20,20 +20,20 @@ static void tcp_scalable_cong_avoid(struct sock *sk, u32 ack, u32 rtt,
 				    u32 in_flight, int flag)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
-	if (in_flight < tp->snd_cwnd)
+
+	if (!tcp_is_cwnd_limited(sk, in_flight))
 		return;
 
-	if (tp->snd_cwnd <= tp->snd_ssthresh) {
-		tp->snd_cwnd++;
-	} else {
+	if (tp->snd_cwnd <= tp->snd_ssthresh)
+		tcp_slow_start(tp);
+	else {
 		tp->snd_cwnd_cnt++;
 		if (tp->snd_cwnd_cnt > min(tp->snd_cwnd, TCP_SCALABLE_AI_CNT)){
+			if (tp->snd_cwnd < tp->snd_cwnd_clamp)
 				tp->snd_cwnd++;
 			tp->snd_cwnd_cnt = 0;
 		}
 	}
-	tp->snd_cwnd = min_t(u32, tp->snd_cwnd, tp->snd_cwnd_clamp);
-	tp->snd_cwnd_stamp = tcp_time_stamp;
 }
 
 static u32 tcp_scalable_ssthresh(struct sock *sk)

net/ipv4/tcp_timer.c

@@ -58,7 +58,7 @@ static void tcp_write_err(struct sock *sk)
  * to prevent DoS attacks. It is called when a retransmission timeout
  * or zero probe timeout occurs on orphaned socket.
  *
- * Criterium is still not confirmed experimentally and may change.
+ * Criteria is still not confirmed experimentally and may change.
  * We kill the socket, if:
  * 1. If number of orphaned sockets exceeds an administratively configured
  *    limit.
@@ -132,7 +132,7 @@ static int tcp_write_timeout(struct sock *sk)
 			   hole detection. :-(
 
 			   It is place to make it. It is not made. I do not want
-			   to make it. It is disguisting. It does not work in any
+			   to make it. It is disgusting. It does not work in any
 			   case. Let me to cite the same draft, which requires for
 			   us to implement this:
 

net/ipv4/tcp_vegas.c

@@ -236,8 +236,7 @@ static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack,
 			/* We don't have enough RTT samples to do the Vegas
 			 * calculation, so we'll behave like Reno.
 			 */
-			if (tp->snd_cwnd > tp->snd_ssthresh)
-				tp->snd_cwnd++;
+			tcp_reno_cong_avoid(sk, ack, seq_rtt, in_flight, cnt);
 		} else {
 			u32 rtt, target_cwnd, diff;
 
@@ -275,7 +274,7 @@ static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack,
 			 */
 			diff = (old_wnd << V_PARAM_SHIFT) - target_cwnd;
 
-			if (tp->snd_cwnd < tp->snd_ssthresh) {
+			if (tp->snd_cwnd <= tp->snd_ssthresh) {
 				/* Slow start.  */
 				if (diff > gamma) {
 					/* Going too fast. Time to slow down
@@ -295,6 +294,7 @@ static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack,
 							    V_PARAM_SHIFT)+1);
 
 				}
+				tcp_slow_start(tp);
 			} else {
 				/* Congestion avoidance. */
 				u32 next_snd_cwnd;
@@ -327,37 +327,17 @@ static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack,
 				else if (next_snd_cwnd < tp->snd_cwnd)
 					tp->snd_cwnd--;
 			}
+
+			if (tp->snd_cwnd < 2)
+				tp->snd_cwnd = 2;
+			else if (tp->snd_cwnd > tp->snd_cwnd_clamp)
+				tp->snd_cwnd = tp->snd_cwnd_clamp;
+		}
 	}
 
 	/* Wipe the slate clean for the next RTT. */
 	vegas->cntRTT = 0;
 	vegas->minRTT = 0x7fffffff;
-	}
-
-	/* The following code is executed for every ack we receive,
-	 * except for conditions checked in should_advance_cwnd()
-	 * before the call to tcp_cong_avoid(). Mainly this means that
-	 * we only execute this code if the ack actually acked some
-	 * data.
-	 */
-
-	/* If we are in slow start, increase our cwnd in response to this ACK.
-	 * (If we are not in slow start then we are in congestion avoidance,
-	 * and adjust our congestion window only once per RTT. See the code
-	 * above.)
-	 */
-	if (tp->snd_cwnd <= tp->snd_ssthresh)
-		tp->snd_cwnd++;
-
-	/* to keep cwnd from growing without bound */
-	tp->snd_cwnd = min_t(u32, tp->snd_cwnd, tp->snd_cwnd_clamp);
-
-	/* Make sure that we are never so timid as to reduce our cwnd below
-	 * 2 MSS.
-	 *
-	 * Going below 2 MSS would risk huge delayed ACKs from our receiver.
-	 */
-	tp->snd_cwnd = max(tp->snd_cwnd, 2U);
 }
 
 /* Extract info for Tcp socket info provided via netlink. */