From b790e01aee74c23a5d92576177934f13aa51f718 Mon Sep 17 00:00:00 2001
From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Mon, 27 Sep 2010 23:05:47 +0000
Subject: ip_gre: lockless xmit

GRE tunnels can benefit from lockless xmits, using NETIF_F_LLTX

Note: If tunnels are created with the "oseq" option, LLTX is not
enabled :

Even using an atomic_t o_seq, we would increase chance for packets being
out of order at receiver.

Bench on a 16 cpus machine (dual E5540 cpus), 16 threads sending
10000000 UDP frames via one gre tunnel (size:200 bytes per frame)

Before patch :
real	3m0.094s
user	0m9.365s
sys	47m50.103s

After patch:
real	0m29.756s
user	0m11.097s
sys	7m33.012s

Last problem to solve is the contention on dst :

38660.00 21.4% __ip_route_output_key          vmlinux
20786.00 11.5% dst_release                    vmlinux
14191.00  7.8% __xfrm_lookup                  vmlinux
12410.00  6.9% ip_finish_output               vmlinux
 4540.00  2.5% ip_push_pending_frames         vmlinux
 4427.00  2.4% ip_append_data                 vmlinux
 4265.00  2.4% __alloc_skb                    vmlinux
 4140.00  2.3% __ip_local_out                 vmlinux
 3991.00  2.2% dev_queue_xmit                 vmlinux

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/net/ipv4/ip_gre.c b/net/ipv4/ip_gre.c
index a1b5d5e..035db63 100644
--- a/net/ipv4/ip_gre.c
+++ b/net/ipv4/ip_gre.c
@@ -1557,6 +1557,10 @@ static int ipgre_newlink(struct net *src_net, struct net_device *dev, struct nla
 	if (!tb[IFLA_MTU])
 		dev->mtu = mtu;
 
+	/* Can use a lockless transmit, unless we generate output sequences */
+	if (!(nt->parms.o_flags & GRE_SEQ))
+		dev->features |= NETIF_F_LLTX;
+
 	err = register_netdevice(dev);
 	if (err)
 		goto out;
-- 
cgit v0.10.2