ipv4: allow local fragmentation in ip_finish_output_gso()
authorLance Richardson <lrichard@redhat.com>
Wed, 2 Nov 2016 20:36:17 +0000 (16:36 -0400)
committerDavid S. Miller <davem@davemloft.net>
Thu, 3 Nov 2016 20:10:26 +0000 (16:10 -0400)
commit9ee6c5dc816aa8256257f2cd4008a9291ec7e985
treeb52ae6938861f8a9ce99c46db958b28a7ac321fd
parentda96786e26c3ae47316db2b92046b11268c4379c
ipv4: allow local fragmentation in ip_finish_output_gso()

Some configurations (e.g. geneve interface with default
MTU of 1500 over an ethernet interface with 1500 MTU) result
in the transmission of packets that exceed the configured MTU.
While this should be considered to be a "bad" configuration,
it is still allowed and should not result in the sending
of packets that exceed the configured MTU.

Fix by dropping the assumption in ip_finish_output_gso() that
locally originated gso packets will never need fragmentation.
Basic testing using iperf (observing CPU usage and bandwidth)
have shown no measurable performance impact for traffic not
requiring fragmentation.

Fixes: c7ba65d7b649 ("net: ip: push gso skb forwarding handling down the stack")
Reported-by: Jan Tluka <jtluka@redhat.com>
Signed-off-by: Lance Richardson <lrichard@redhat.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/net/ip.h
net/ipv4/ip_forward.c
net/ipv4/ip_output.c
net/ipv4/ip_tunnel_core.c
net/ipv4/ipmr.c