[TCP]: skb pcount with MTU discovery

The problem is that when doing MTU discovery, the too-large segments in
the write queue will be calculated as having a pcount of >1.  When
tcp_write_xmit() is trying to send, tcp_snd_test() fails the cwnd test
when pcount > cwnd.

The segments are eventually transmitted one at a time by keepalive, but
this can take a long time.

This patch checks if TSO is enabled when setting pcount.

Signed-off-by: John Heffner <jheffner@psc.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>

2 files changed, 20 insertions, 16 deletions

diff --git a/include/net/tcp.h b/include/net/tcp.h
index 503810a70e21..9355ae5b1d75 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -1417,19 +1417,20 @@ tcp_nagle_check(const struct tcp_sock *tp, const struct sk_buff *skb,
 		  tcp_minshall_check(tp))));
 }
 
-extern void tcp_set_skb_tso_segs(struct sk_buff *, unsigned int);
+extern void tcp_set_skb_tso_segs(struct sock *, struct sk_buff *);
 
 /* This checks if the data bearing packet SKB (usually sk->sk_send_head)
  * should be put on the wire right now.
  */
-static __inline__ int tcp_snd_test(const struct tcp_sock *tp, 
+static __inline__ int tcp_snd_test(struct sock *sk,
 				   struct sk_buff *skb,
 				   unsigned cur_mss, int nonagle)
 {
+	struct tcp_sock *tp = tcp_sk(sk);
 	int pkts = tcp_skb_pcount(skb);
 
 	if (!pkts) {
-		tcp_set_skb_tso_segs(skb, tp->mss_cache_std);
+		tcp_set_skb_tso_segs(sk, skb);
 		pkts = tcp_skb_pcount(skb);
 	}
 
@@ -1490,7 +1491,7 @@ static __inline__ void __tcp_push_pending_frames(struct sock *sk,
 	if (skb) {
 		if (!tcp_skb_is_last(sk, skb))
 			nonagle = TCP_NAGLE_PUSH;
-		if (!tcp_snd_test(tp, skb, cur_mss, nonagle) ||
+		if (!tcp_snd_test(sk, skb, cur_mss, nonagle) ||
 		    tcp_write_xmit(sk, nonagle))
 			tcp_check_probe_timer(sk, tp);
 	}
@@ -1508,7 +1509,7 @@ static __inline__ int tcp_may_send_now(struct sock *sk, struct tcp_sock *tp)
 	struct sk_buff *skb = sk->sk_send_head;
 
 	return (skb &&
-		tcp_snd_test(tp, skb, tcp_current_mss(sk, 1),
+		tcp_snd_test(sk, skb, tcp_current_mss(sk, 1),
 			     tcp_skb_is_last(sk, skb) ? TCP_NAGLE_PUSH : tp->nonagle));
 }
 
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 13c14cb6dee4..a12df6979ffd 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -427,7 +427,7 @@ void tcp_push_one(struct sock *sk, unsigned cur_mss)
 	struct tcp_sock *tp = tcp_sk(sk);
 	struct sk_buff *skb = sk->sk_send_head;
 
-	if (tcp_snd_test(tp, skb, cur_mss, TCP_NAGLE_PUSH)) {
+	if (tcp_snd_test(sk, skb, cur_mss, TCP_NAGLE_PUSH)) {
 		/* Send it out now. */
 		TCP_SKB_CB(skb)->when = tcp_time_stamp;
 		tcp_tso_set_push(skb);
@@ -440,9 +440,12 @@ void tcp_push_one(struct sock *sk, unsigned cur_mss)
 	}
 }
 
-void tcp_set_skb_tso_segs(struct sk_buff *skb, unsigned int mss_std)
+void tcp_set_skb_tso_segs(struct sock *sk, struct sk_buff *skb)
 {
-	if (skb->len <= mss_std) {
+	struct tcp_sock *tp = tcp_sk(sk);
+
+	if (skb->len <= tp->mss_cache_std ||
+	    !(sk->sk_route_caps & NETIF_F_TSO)) {
 		/* Avoid the costly divide in the normal
 		 * non-TSO case.
 		 */
@@ -451,10 +454,10 @@ void tcp_set_skb_tso_segs(struct sk_buff *skb, unsigned int mss_std)
 	} else {
 		unsigned int factor;
 
-		factor = skb->len + (mss_std - 1);
-		factor /= mss_std;
+		factor = skb->len + (tp->mss_cache_std - 1);
+		factor /= tp->mss_cache_std;
 		skb_shinfo(skb)->tso_segs = factor;
-		skb_shinfo(skb)->tso_size = mss_std;
+		skb_shinfo(skb)->tso_size = tp->mss_cache_std;
 	}
 }
 
@@ -525,8 +528,8 @@ static int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len)
 	}
 
 	/* Fix up tso_factor for both original and new SKB.  */
-	tcp_set_skb_tso_segs(skb, tp->mss_cache_std);
-	tcp_set_skb_tso_segs(buff, tp->mss_cache_std);
+	tcp_set_skb_tso_segs(sk, skb);
+	tcp_set_skb_tso_segs(sk, buff);
 
 	if (TCP_SKB_CB(skb)->sacked & TCPCB_LOST) {
 		tp->lost_out += tcp_skb_pcount(skb);
@@ -601,7 +604,7 @@ int tcp_trim_head(struct sock *sk, struct sk_buff *skb, u32 len)
 	 * factor and mss.
 	 */
 	if (tcp_skb_pcount(skb) > 1)
-		tcp_set_skb_tso_segs(skb, tcp_skb_mss(skb));
+		tcp_set_skb_tso_segs(sk, skb);
 
 	return 0;
 }
@@ -752,7 +755,7 @@ int tcp_write_xmit(struct sock *sk, int nonagle)
 		mss_now = tcp_current_mss(sk, 1);
 
 		while ((skb = sk->sk_send_head) &&
-		       tcp_snd_test(tp, skb, mss_now,
+		       tcp_snd_test(sk, skb, mss_now,
 			       	    tcp_skb_is_last(sk, skb) ? nonagle :
 				    			       TCP_NAGLE_PUSH)) {
 			if (skb->len > mss_now) {
@@ -1676,7 +1679,7 @@ int tcp_write_wakeup(struct sock *sk)
 					tp->mss_cache = tp->mss_cache_std;
 				}
 			} else if (!tcp_skb_pcount(skb))
-				tcp_set_skb_tso_segs(skb, tp->mss_cache_std);
+				tcp_set_skb_tso_segs(sk, skb);
 
 			TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH;
 			TCP_SKB_CB(skb)->when = tcp_time_stamp;