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

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (47 commits)
  [SCTP]: Fix local_addr deletions during list traversals.
  net: fix build with CONFIG_NET=n
  [TCP]: Prevent sending past receiver window with TSO (at last skb)
  rt2x00: Add new D-Link USB ID
  rt2x00: never disable multicast because it disables broadcast too
  libertas: fix the 'compare command with itself' properly
  drivers/net/Kconfig: fix whitespace for GELIC_WIRELESS entry
  [NETFILTER]: nf_queue: don't return error when unregistering a non-existant handler
  [NETFILTER]: nfnetlink_queue: fix EPERM when binding/unbinding and instance 0 exists
  [NETFILTER]: nfnetlink_log: fix EPERM when binding/unbinding and instance 0 exists
  [NETFILTER]: nf_conntrack: replace horrible hack with ksize()
  [NETFILTER]: nf_conntrack: add \n to "expectation table full" message
  [NETFILTER]: xt_time: fix failure to match on Sundays
  [NETFILTER]: nfnetlink_log: fix computation of netlink skb size
  [NETFILTER]: nfnetlink_queue: fix computation of allocated size for netlink skb.
  [NETFILTER]: nfnetlink: fix ifdef in nfnetlink_compat.h
  [NET]: include <linux/types.h> into linux/ethtool.h for __u* typedef
  [NET]: Make /proc/net a symlink on /proc/self/net (v3)
  RxRPC: fix rxrpc_recvmsg()'s returning of msg_name
  net/enc28j60: oops fix
  ...

1 files changed, 10 insertions, 2 deletions

diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index ed750f9ceb0..01578f544ad 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -1035,6 +1035,13 @@ static void tcp_cwnd_validate(struct sock *sk)
  * introducing MSS oddities to segment boundaries. In rare cases where
  * mss_now != mss_cache, we will request caller to create a small skb
  * per input skb which could be mostly avoided here (if desired).
+ *
+ * We explicitly want to create a request for splitting write queue tail
+ * to a small skb for Nagle purposes while avoiding unnecessary modulos,
+ * thus all the complexity (cwnd_len is always MSS multiple which we
+ * return whenever allowed by the other factors). Basically we need the
+ * modulo only when the receiver window alone is the limiting factor or
+ * when we would be allowed to send the split-due-to-Nagle skb fully.
  */
 static unsigned int tcp_mss_split_point(struct sock *sk, struct sk_buff *skb,
 					unsigned int mss_now, unsigned int cwnd)
@@ -1048,10 +1055,11 @@ static unsigned int tcp_mss_split_point(struct sock *sk, struct sk_buff *skb,
 	if (likely(cwnd_len <= window && skb != tcp_write_queue_tail(sk)))
 		return cwnd_len;
 
-	if (skb == tcp_write_queue_tail(sk) && cwnd_len <= skb->len)
+	needed = min(skb->len, window);
+
+	if (skb == tcp_write_queue_tail(sk) && cwnd_len <= needed)
 		return cwnd_len;
 
-	needed = min(skb->len, window);
 	return needed - needed % mss_now;
 }