Merge branch 'tcp-change-pingpong-to-3-in-delayed-ack-logic'
authorDavid S. Miller <davem@davemloft.net>
Sun, 27 Jan 2019 21:29:43 +0000 (13:29 -0800)
committerDavid S. Miller <davem@davemloft.net>
Sun, 27 Jan 2019 21:29:43 +0000 (13:29 -0800)
commit889865cf54bddd8a4e258a11225bcbb646f90863
treef0afde54fb516f0293e849ade6c738ece48f16c1
parentfb1b69999130e9800c4e341a2b0f0991cdd8fcfc
parent4a41f453bedfd5e9cd040bad509d9da49feb3e2c
Merge branch 'tcp-change-pingpong-to-3-in-delayed-ack-logic'

Wei Wang says:

====================
tcp: change pingpong to 3 in delayed ack logic

TCP receiver today tries not to delay the ACKs to speed up the initial
slow start (a.k.a QUICK ACK mechanism). However the previous design
does not work well with modern TCP applications that starts with an
application-level handshake. For example, a HTTPs server often
receives the SSL hello and responds right away which triggers the TCP
stack to stop the quick ack and start delaying the ACKs based only one
instance of ping-pong. This patchset changes the threshold from 1 to 3
ping-pong transactions, so that we only start to delay the acks after
the receiver responds data quickly three times.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>