2 * Copyright (c) 2007-2012 Nicira, Inc.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
21 #include <linux/if_arp.h>
22 #include <linux/if_bridge.h>
23 #include <linux/if_vlan.h>
24 #include <linux/kernel.h>
25 #include <linux/llc.h>
26 #include <linux/rtnetlink.h>
27 #include <linux/skbuff.h>
28 #include <linux/openvswitch.h>
29 #include <linux/netdevice.h>
35 #include "vport-internal_dev.h"
36 #include "vport-netdev.h"
38 static struct vport_ops ovs_netdev_vport_ops;
39 static void netdev_port_receive(struct vport *vport, struct sk_buff *skb);
41 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,39)
42 /* Called with rcu_read_lock and bottom-halves disabled. */
43 static rx_handler_result_t netdev_frame_hook(struct sk_buff **pskb)
45 struct sk_buff *skb = *pskb;
48 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
49 return RX_HANDLER_PASS;
51 vport = ovs_netdev_get_vport(skb->dev);
53 netdev_port_receive(vport, skb);
55 return RX_HANDLER_CONSUMED;
57 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
58 defined HAVE_RHEL_OVS_HOOK
59 /* Called with rcu_read_lock and bottom-halves disabled. */
60 static struct sk_buff *netdev_frame_hook(struct sk_buff *skb)
64 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
67 vport = ovs_netdev_get_vport(skb->dev);
69 netdev_port_receive(vport, skb);
73 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,32)
75 * Used as br_handle_frame_hook. (Cannot run bridge at the same time, even on
76 * different set of devices!)
78 /* Called with rcu_read_lock and bottom-halves disabled. */
79 static struct sk_buff *netdev_frame_hook(struct net_bridge_port *p,
82 netdev_port_receive((struct vport *)p, skb);
89 static struct net_device *get_dpdev(const struct datapath *dp)
93 local = ovs_vport_ovsl(dp, OVSP_LOCAL);
95 return netdev_vport_priv(local)->dev;
98 static struct vport *netdev_create(const struct vport_parms *parms)
101 struct netdev_vport *netdev_vport;
104 vport = ovs_vport_alloc(sizeof(struct netdev_vport),
105 &ovs_netdev_vport_ops, parms);
107 err = PTR_ERR(vport);
111 netdev_vport = netdev_vport_priv(vport);
113 netdev_vport->dev = dev_get_by_name(ovs_dp_get_net(vport->dp), parms->name);
114 if (!netdev_vport->dev) {
116 goto error_free_vport;
119 if (netdev_vport->dev->flags & IFF_LOOPBACK ||
120 netdev_vport->dev->type != ARPHRD_ETHER ||
121 ovs_is_internal_dev(netdev_vport->dev)) {
127 err = netdev_master_upper_dev_link(netdev_vport->dev,
128 get_dpdev(vport->dp));
132 err = netdev_rx_handler_register(netdev_vport->dev, netdev_frame_hook,
135 goto error_master_upper_dev_unlink;
137 dev_disable_lro(netdev_vport->dev);
138 dev_set_promiscuity(netdev_vport->dev, 1);
139 netdev_vport->dev->priv_flags |= IFF_OVS_DATAPATH;
144 error_master_upper_dev_unlink:
145 netdev_upper_dev_unlink(netdev_vport->dev, get_dpdev(vport->dp));
149 dev_put(netdev_vport->dev);
151 ovs_vport_free(vport);
156 static void free_port_rcu(struct rcu_head *rcu)
158 struct netdev_vport *netdev_vport = container_of(rcu,
159 struct netdev_vport, rcu);
161 dev_put(netdev_vport->dev);
162 ovs_vport_free(vport_from_priv(netdev_vport));
165 void ovs_netdev_detach_dev(struct vport *vport)
167 struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
170 netdev_vport->dev->priv_flags &= ~IFF_OVS_DATAPATH;
171 netdev_rx_handler_unregister(netdev_vport->dev);
172 netdev_upper_dev_unlink(netdev_vport->dev,
173 netdev_master_upper_dev_get(netdev_vport->dev));
174 dev_set_promiscuity(netdev_vport->dev, -1);
177 static void netdev_destroy(struct vport *vport)
179 struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
182 if (ovs_netdev_get_vport(netdev_vport->dev))
183 ovs_netdev_detach_dev(vport);
186 call_rcu(&netdev_vport->rcu, free_port_rcu);
189 const char *ovs_netdev_get_name(const struct vport *vport)
191 const struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
192 return netdev_vport->dev->name;
195 /* Must be called with rcu_read_lock. */
196 static void netdev_port_receive(struct vport *vport, struct sk_buff *skb)
198 if (unlikely(!vport))
201 if (unlikely(skb_warn_if_lro(skb)))
204 /* Make our own copy of the packet. Otherwise we will mangle the
205 * packet for anyone who came before us (e.g. tcpdump via AF_PACKET).
206 * (No one comes after us, since we tell handle_bridge() that we took
209 skb = skb_share_check(skb, GFP_ATOMIC);
213 skb_push(skb, ETH_HLEN);
214 ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);
216 ovs_vport_receive(vport, skb, NULL);
223 static unsigned int packet_length(const struct sk_buff *skb)
225 unsigned int length = skb->len - ETH_HLEN;
227 if (skb->protocol == htons(ETH_P_8021Q))
233 static int netdev_send(struct vport *vport, struct sk_buff *skb)
235 struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
236 int mtu = netdev_vport->dev->mtu;
239 if (unlikely(packet_length(skb) > mtu && !skb_is_gso(skb))) {
240 net_warn_ratelimited("%s: dropped over-mtu packet: %d > %d\n",
241 netdev_vport->dev->name,
242 packet_length(skb), mtu);
246 skb->dev = netdev_vport->dev;
257 /* Returns null if this device is not attached to a datapath. */
258 struct vport *ovs_netdev_get_vport(struct net_device *dev)
260 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
261 defined HAVE_RHEL_OVS_HOOK
262 #ifdef HAVE_OVS_DATAPATH
263 if (likely(dev->priv_flags & IFF_OVS_DATAPATH))
265 if (likely(rcu_access_pointer(dev->rx_handler) == netdev_frame_hook))
267 #ifdef HAVE_RHEL_OVS_HOOK
268 return (struct vport *)rcu_dereference_rtnl(dev->ax25_ptr);
270 return (struct vport *)rcu_dereference_rtnl(dev->rx_handler_data);
275 return (struct vport *)rcu_dereference_rtnl(dev->br_port);
279 static struct vport_ops ovs_netdev_vport_ops = {
280 .type = OVS_VPORT_TYPE_NETDEV,
281 .create = netdev_create,
282 .destroy = netdev_destroy,
283 .get_name = ovs_netdev_get_name,
287 int __init ovs_netdev_init(void)
289 return ovs_vport_ops_register(&ovs_netdev_vport_ops);
292 void ovs_netdev_exit(void)
294 ovs_vport_ops_unregister(&ovs_netdev_vport_ops);
297 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36) && \
298 !defined HAVE_RHEL_OVS_HOOK
300 * Enforces, mutual exclusion with the Linux bridge module, by declaring and
301 * exporting br_should_route_hook. Because the bridge module also exports the
302 * same symbol, the module loader will refuse to load both modules at the same
303 * time (e.g. "bridge: exports duplicate symbol br_should_route_hook (owned by
306 * Before Linux 2.6.36, Open vSwitch cannot safely coexist with the Linux
307 * bridge module, so openvswitch uses this macro in those versions. In
308 * Linux 2.6.36 and later, Open vSwitch can coexist with the bridge module.
310 * The use of "typeof" here avoids the need to track changes in the type of
311 * br_should_route_hook over various kernel versions.
313 typeof(br_should_route_hook) br_should_route_hook;
314 EXPORT_SYMBOL(br_should_route_hook);