tcp: reset reordering est. selectively on timeout
authorYuchung Cheng <ycheng@google.com>
Mon, 12 Aug 2013 23:41:25 +0000 (16:41 -0700)
committerDavid S. Miller <davem@davemloft.net>
Tue, 13 Aug 2013 23:08:33 +0000 (16:08 -0700)
commit74c181d528bd8b5989f424a489262d0742ca31ae
treedb72e02385e949e1b3a9c305e661db65a27f639b
parent98f1b7f3820a50a42e51f9bd3e7014cf9b2688a8
tcp: reset reordering est. selectively on timeout

On timeout the TCP sender unconditionally resets the estimated degree
of network reordering (tp->reordering). The idea behind this is that
the estimate is too large to trigger fast recovery (e.g., due to a IP
path change).

But for example if the sender only had 2 packets outstanding, then a
timeout doesn't tell much about reordering. A sender that learns about
reordering on big writes and loses packets on small writes will end up
falsely retransmitting again and again, especially when reordering is
more likely on big writes.

Therefore the sender should only suspect that tp->reordering is too
high if it could have gone into fast recovery with the (lower) default
estimate.

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/tcp_input.c