tcp_output.c, tcp.c, tcp.h:

  reconciling TCP_CORK and TCP_NODELAY

include/net/tcp.h

@@ -450,6 +450,11 @@ static __inline__ int tcp_sk_listen_hashfn(struct sock *sk)
 #define TCP_TIME_PROBE0		3	/* Zero window probe timer */
 #define TCP_TIME_KEEPOPEN	4	/* Keepalive timer */
 
+/* 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 */
+
 /* sysctl variables for tcp */
 extern int sysctl_max_syn_backlog;
 extern int sysctl_tcp_timestamps;
@@ -1216,7 +1221,7 @@ tcp_nagle_check(struct tcp_opt *tp, struct sk_buff *skb, unsigned mss_now, int n
 {
 	return (skb->len < mss_now &&
 		!(TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN) &&
-		(nonagle == 2 ||
+		((nonagle&TCP_NAGLE_CORK) ||
 		 (!nonagle &&
 		  tp->packets_out &&
 		  tcp_minshall_check(tp))));
@@ -1252,7 +1257,7 @@ static __inline__ int tcp_snd_test(struct tcp_opt *tp, struct sk_buff *skb,
 	/* Don't be strict about the congestion window for the
 	 * final FIN frame.  -DaveM
 	 */
-	return ((nonagle==1 || tp->urg_mode
+	return (((nonagle&TCP_NAGLE_PUSH) || tp->urg_mode
 		 || !tcp_nagle_check(tp, skb, cur_mss, nonagle)) &&
 		((tcp_packets_in_flight(tp) < tp->snd_cwnd) ||
 		 (TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN)) &&
@@ -1283,7 +1288,7 @@ static __inline__ void __tcp_push_pending_frames(struct sock *sk,
 
 	if (skb) {
 		if (!tcp_skb_is_last(sk, skb))
-			nonagle = 1;
+			nonagle = TCP_NAGLE_PUSH;
 		if (!tcp_snd_test(tp, skb, cur_mss, nonagle) ||
 		    tcp_write_xmit(sk, nonagle))
 			tcp_check_probe_timer(sk, tp);
@@ -1303,7 +1308,7 @@ static __inline__ int tcp_may_send_now(struct sock *sk, struct tcp_opt *tp)
 
 	return (skb &&
 		tcp_snd_test(tp, skb, tcp_current_mss(sk, 1),
-			     tcp_skb_is_last(sk, skb) ? 1 : tp->nonagle));
+			     tcp_skb_is_last(sk, skb) ? TCP_NAGLE_PUSH : tp->nonagle));
 }
 
 static __inline__ void tcp_init_wl(struct tcp_opt *tp, u32 ack, u32 seq)

net/ipv4/tcp.c

@@ -800,6 +800,8 @@ static inline void skb_entail(struct sock *sk, struct tcp_opt *tp,
 	tcp_charge_skb(sk, skb);
 	if (!tp->send_head)
 		tp->send_head = skb;
+	else if (tp->nonagle&TCP_NAGLE_PUSH)
+		tp->nonagle &= ~TCP_NAGLE_PUSH; 
 }
 
 static inline void tcp_mark_urg(struct tcp_opt *tp, int flags,
@@ -821,7 +823,7 @@ static inline void tcp_push(struct sock *sk, struct tcp_opt *tp, int flags,
 			tcp_mark_push(tp, skb);
 		tcp_mark_urg(tp, flags, skb);
 		__tcp_push_pending_frames(sk, tp, mss_now,
-					  (flags & MSG_MORE) ? 2 : nonagle);
+					  (flags & MSG_MORE) ? TCP_NAGLE_CORK : nonagle);
 	}
 }
 
@@ -911,7 +913,7 @@ ssize_t do_tcp_sendpages(struct sock *sk, struct page **pages, int poffset,
 
 		if (forced_push(tp)) {
 			tcp_mark_push(tp, skb);
-			__tcp_push_pending_frames(sk, tp, mss_now, 1);
+			__tcp_push_pending_frames(sk, tp, mss_now, TCP_NAGLE_PUSH);
 		} else if (skb == tp->send_head)
 			tcp_push_one(sk, mss_now);
 		continue;
@@ -920,7 +922,7 @@ ssize_t do_tcp_sendpages(struct sock *sk, struct page **pages, int poffset,
 		set_bit(SOCK_NOSPACE, &sk->socket->flags);
 wait_for_memory:
 		if (copied)
-			tcp_push(sk, tp, flags & ~MSG_MORE, mss_now, 1);
+			tcp_push(sk, tp, flags & ~MSG_MORE, mss_now, TCP_NAGLE_PUSH);
 
 		if ((err = wait_for_tcp_memory(sk, &timeo)) != 0)
 			goto do_error;
@@ -1199,7 +1201,7 @@ int tcp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 
 			if (forced_push(tp)) {
 				tcp_mark_push(tp, skb);
-				__tcp_push_pending_frames(sk, tp, mss_now, 1);
+				__tcp_push_pending_frames(sk, tp, mss_now, TCP_NAGLE_PUSH);
 			} else if (skb == tp->send_head)
 				tcp_push_one(sk, mss_now);
 			continue;
@@ -1208,7 +1210,7 @@ int tcp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 			set_bit(SOCK_NOSPACE, &sk->socket->flags);
 wait_for_memory:
 			if (copied)
-				tcp_push(sk, tp, flags & ~MSG_MORE, mss_now, 1);
+				tcp_push(sk, tp, flags & ~MSG_MORE, mss_now, TCP_NAGLE_PUSH);
 
 			if ((err = wait_for_tcp_memory(sk, &timeo)) != 0)
 				goto do_error;
@@ -2300,16 +2302,20 @@ int tcp_setsockopt(struct sock *sk, int level, int optname, char *optval,
 		break;
 
 	case TCP_NODELAY:
-		/* You cannot try to use this and TCP_CORK in
-		 * tandem, so let the user know.
-		 */
-		if (tp->nonagle == 2) {
-			err = -EINVAL;
-			break;
-		}
-		tp->nonagle = !val ? 0 : 1;
-		if (val)
+		if (val) {
+			/* TCP_NODELAY is weaker than TCP_CORK, so that
+			 * this option on corked socket is remembered, but
+			 * it is not activated until cork is cleared.
+			 *
+			 * However, when TCP_NODELAY is set we make
+			 * an explicit push, which overrides even TCP_CORK
+			 * for currently queued segments.
+			 */
+			tp->nonagle |= TCP_NAGLE_OFF|TCP_NAGLE_PUSH;
 			tcp_push_pending_frames(sk, tp);
+		} else {
+			tp->nonagle &= ~TCP_NAGLE_OFF;
+		}
 		break;
 
 	case TCP_CORK:
@@ -2321,18 +2327,15 @@ int tcp_setsockopt(struct sock *sk, int level, int optname, char *optval,
 		 * out headers with a write() call first and then use
 		 * sendfile to send out the data parts.
 		 *
-		 * You cannot try to use TCP_NODELAY and this mechanism
-		 * at the same time, so let the user know.
+		 * TCP_CORK can be set together with TCP_NODELAY and it is
+		 * stronger than TCP_NODELAY.
 		 */
-		if (tp->nonagle == 1) {
-			err = -EINVAL;
-			break;
-		}
-		if (val != 0) {
-			tp->nonagle = 2;
+		if (val) {
+			tp->nonagle |= TCP_NAGLE_CORK;
 		} else {
-			tp->nonagle = 0;
-
+			tp->nonagle &= ~TCP_NAGLE_CORK;
+			if (tp->nonagle&TCP_NAGLE_OFF)
+				tp->nonagle |= TCP_NAGLE_PUSH;
 			tcp_push_pending_frames(sk, tp);
 		}
 		break;
@@ -2455,10 +2458,10 @@ int tcp_getsockopt(struct sock *sk, int level, int optname, char *optval,
 			val = tp->user_mss;
 		break;
 	case TCP_NODELAY:
-		val = (tp->nonagle == 1);
+		val = !!(tp->nonagle&TCP_NAGLE_OFF);
 		break;
 	case TCP_CORK:
-		val = (tp->nonagle == 2);
+		val = !!(tp->nonagle&TCP_NAGLE_CORK);
 		break;
 	case TCP_KEEPIDLE:
 		val = (tp->keepalive_time ? : sysctl_tcp_keepalive_time) / HZ;

net/ipv4/tcp_output.c

@@ -336,7 +336,7 @@ void tcp_push_one(struct sock *sk, unsigned cur_mss)
 	struct tcp_opt *tp = tcp_sk(sk);
 	struct sk_buff *skb = tp->send_head;
 
-	if (tcp_snd_test(tp, skb, cur_mss, 1)) {
+	if (tcp_snd_test(tp, skb, cur_mss, TCP_NAGLE_PUSH)) {
 		/* Send it out now. */
 		TCP_SKB_CB(skb)->when = tcp_time_stamp;
 		if (tcp_transmit_skb(sk, skb_clone(skb, sk->allocation)) == 0) {
@@ -632,7 +632,7 @@ int tcp_write_xmit(struct sock *sk, int nonagle)
 		mss_now = tcp_current_mss(sk, 1);
 
 		while((skb = tp->send_head) &&
-		      tcp_snd_test(tp, skb, mss_now, tcp_skb_is_last(sk, skb) ? nonagle : 1)) {
+		      tcp_snd_test(tp, skb, mss_now, tcp_skb_is_last(sk, skb) ? nonagle : TCP_NAGLE_PUSH)) {
 			if (skb->len > mss_now) {
 				if (tcp_fragment(sk, skb, mss_now))
 					break;
@@ -1106,7 +1106,7 @@ void tcp_send_fin(struct sock *sk)
 		TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq + 1;
 		tcp_send_skb(sk, skb, 1, mss_now);
 	}
-	__tcp_push_pending_frames(sk, tp, mss_now, 1);
+	__tcp_push_pending_frames(sk, tp, mss_now, TCP_NAGLE_OFF);
 }
 
 /* We get here when a process closes a file descriptor (either due to