tcp: increase size at which tcp_bound_to_half_wnd bounds to > TCP_MSS_DEFAULT
authorSeymour, Shane M <shane.seymour@hpe.com>
Tue, 28 Jun 2016 23:06:48 +0000 (23:06 +0000)
committerDavid S. Miller <davem@davemloft.net>
Thu, 30 Jun 2016 12:17:21 +0000 (08:17 -0400)
commit2631b79f6cb8b634fe41b77de9c4add0ec6b3cae
treeaa9f2dde1f090f20fbe28e555c41d4f36d7409c4
parentb1ed4c4fa9a5ccf325184fd90edc50978ef6e33a
tcp: increase size at which tcp_bound_to_half_wnd bounds to > TCP_MSS_DEFAULT

In previous commit 01f83d69844d307be2aa6fea88b0e8fe5cbdb2f4
the following comments were added:

"When peer uses tiny windows, there is no use in packetizing to sub-MSS
pieces for the sake of SWS or making sure there are enough packets in
the pipe for fast recovery."

The test should be > TCP_MSS_DEFAULT not >= 512. This allows low end
devices that send an MSS of 536 (TCP_MSS_DEFAULT) to see better network
performance by sending it 536 bytes of data at a time instead of bounding
to half window size (268). Other network stacks work this way, e.g. HP-UX.

Signed-off-by: Shane Seymour <shane.seymour@hpe.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/net/tcp.h