tcp: do not set queue_mapping on SYNACK
authorEric Dumazet <edumazet@google.com>
Fri, 16 Oct 2015 20:00:01 +0000 (13:00 -0700)
committerDavid S. Miller <davem@davemloft.net>
Mon, 19 Oct 2015 05:26:02 +0000 (22:26 -0700)
commitdc6ef6be52154490c5c03f742e28bc781cc751b2
tree8e297dd00c9cef73f67424b72a5669d6a97d13f7
parent951b5d959f1da4bae8910085a2d8d6a3d374c72d
tcp: do not set queue_mapping on SYNACK

At the time of commit fff326990789 ("tcp: reflect SYN queue_mapping into
SYNACK packets") we had little ways to cope with SYN floods.

We no longer need to reflect incoming skb queue mappings, and instead
can pick a TX queue based on cpu cooking the SYNACK, with normal XPS
affinities.

Note that all SYNACK retransmits were picking TX queue 0, this no longer
is a win given that SYNACK rtx are now distributed on all cpus.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/net/tcp.h
net/ipv4/ip_output.c
net/ipv4/tcp_input.c
net/ipv4/tcp_ipv4.c
net/ipv4/tcp_output.c
net/ipv6/tcp_ipv6.c