diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index ad2021ccc55a..9d5078bd23a3 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -248,6 +248,7 @@ struct tcp_sock {
 	/* inet_connection_sock has to be the first member of tcp_sock */
 	struct inet_connection_sock	inet_conn;
 	u16	tcp_header_len;	/* Bytes of tcp header to send		*/
+	u16	xmit_size_goal_segs; /* Goal for segmenting output packets */
 
 /*
  *	Header prediction flags
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 886596ff0aae..0db9f3b984f7 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -665,7 +665,7 @@ static unsigned int tcp_xmit_size_goal(struct sock *sk, u32 mss_now,
 				       int large_allowed)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
-	u32 xmit_size_goal;
+	u32 xmit_size_goal, old_size_goal;
 
 	xmit_size_goal = mss_now;
 
@@ -676,7 +676,17 @@ static unsigned int tcp_xmit_size_goal(struct sock *sk, u32 mss_now,
 				  tp->tcp_header_len);
 
 		xmit_size_goal = tcp_bound_to_half_wnd(tp, xmit_size_goal);
-		xmit_size_goal -= (xmit_size_goal % mss_now);
+
+		/* We try hard to avoid divides here */
+		old_size_goal = tp->xmit_size_goal_segs * mss_now;
+
+		if (likely(old_size_goal <= xmit_size_goal &&
+			   old_size_goal + mss_now > xmit_size_goal)) {
+			xmit_size_goal = old_size_goal;
+		} else {
+			tp->xmit_size_goal_segs = xmit_size_goal / mss_now;
+			xmit_size_goal = tp->xmit_size_goal_segs * mss_now;
+		}
 	}
 
 	return xmit_size_goal;