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>
34 #include "vport-internal_dev.h"
35 #include "vport-netdev.h"
37 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37) && \
38 !defined(HAVE_VLAN_BUG_WORKAROUND)
39 #include <linux/module.h>
41 static int vlan_tso __read_mostly;
42 module_param(vlan_tso, int, 0644);
43 MODULE_PARM_DESC(vlan_tso, "Enable TSO for VLAN packets");
48 static void netdev_port_receive(struct vport *vport, struct sk_buff *skb);
50 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,39)
51 /* Called with rcu_read_lock and bottom-halves disabled. */
52 static rx_handler_result_t netdev_frame_hook(struct sk_buff **pskb)
54 struct sk_buff *skb = *pskb;
57 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
58 return RX_HANDLER_PASS;
60 vport = ovs_netdev_get_vport(skb->dev);
62 netdev_port_receive(vport, skb);
64 return RX_HANDLER_CONSUMED;
66 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
67 defined HAVE_RHEL_OVS_HOOK
68 /* Called with rcu_read_lock and bottom-halves disabled. */
69 static struct sk_buff *netdev_frame_hook(struct sk_buff *skb)
73 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
76 vport = ovs_netdev_get_vport(skb->dev);
78 netdev_port_receive(vport, skb);
82 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,32)
84 * Used as br_handle_frame_hook. (Cannot run bridge at the same time, even on
85 * different set of devices!)
87 /* Called with rcu_read_lock and bottom-halves disabled. */
88 static struct sk_buff *netdev_frame_hook(struct net_bridge_port *p,
91 netdev_port_receive((struct vport *)p, skb);
98 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
99 defined HAVE_RHEL_OVS_HOOK
100 static int netdev_init(void) { return 0; }
101 static void netdev_exit(void) { }
103 static int port_count;
105 static void netdev_init(void)
111 /* Hook into callback used by the bridge to intercept packets.
112 * Parasites we are. */
113 br_handle_frame_hook = netdev_frame_hook;
118 static void netdev_exit(void)
124 br_handle_frame_hook = NULL;
128 static struct net_device *get_dpdev(struct datapath *dp)
132 local = ovs_vport_ovsl(dp, OVSP_LOCAL);
134 return netdev_vport_priv(local)->dev;
137 static struct vport *netdev_create(const struct vport_parms *parms)
140 struct netdev_vport *netdev_vport;
143 vport = ovs_vport_alloc(sizeof(struct netdev_vport),
144 &ovs_netdev_vport_ops, parms);
146 err = PTR_ERR(vport);
150 netdev_vport = netdev_vport_priv(vport);
152 netdev_vport->dev = dev_get_by_name(ovs_dp_get_net(vport->dp), parms->name);
153 if (!netdev_vport->dev) {
155 goto error_free_vport;
158 if (netdev_vport->dev->flags & IFF_LOOPBACK ||
159 netdev_vport->dev->type != ARPHRD_ETHER ||
160 ovs_is_internal_dev(netdev_vport->dev)) {
166 err = netdev_master_upper_dev_link(netdev_vport->dev,
167 get_dpdev(vport->dp));
171 err = netdev_rx_handler_register(netdev_vport->dev, netdev_frame_hook,
174 goto error_master_upper_dev_unlink;
176 dev_set_promiscuity(netdev_vport->dev, 1);
177 netdev_vport->dev->priv_flags |= IFF_OVS_DATAPATH;
183 error_master_upper_dev_unlink:
184 netdev_upper_dev_unlink(netdev_vport->dev, get_dpdev(vport->dp));
188 dev_put(netdev_vport->dev);
190 ovs_vport_free(vport);
195 static void free_port_rcu(struct rcu_head *rcu)
197 struct netdev_vport *netdev_vport = container_of(rcu,
198 struct netdev_vport, rcu);
200 dev_put(netdev_vport->dev);
201 ovs_vport_free(vport_from_priv(netdev_vport));
204 static void netdev_destroy(struct vport *vport)
206 struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
210 netdev_vport->dev->priv_flags &= ~IFF_OVS_DATAPATH;
211 netdev_rx_handler_unregister(netdev_vport->dev);
212 netdev_upper_dev_unlink(netdev_vport->dev, get_dpdev(vport->dp));
213 dev_set_promiscuity(netdev_vport->dev, -1);
216 call_rcu(&netdev_vport->rcu, free_port_rcu);
219 const char *ovs_netdev_get_name(const struct vport *vport)
221 const struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
222 return netdev_vport->dev->name;
225 /* Must be called with rcu_read_lock. */
226 static void netdev_port_receive(struct vport *vport, struct sk_buff *skb)
228 if (unlikely(!vport))
231 if (unlikely(skb_warn_if_lro(skb)))
234 /* Make our own copy of the packet. Otherwise we will mangle the
235 * packet for anyone who came before us (e.g. tcpdump via AF_PACKET).
236 * (No one comes after us, since we tell handle_bridge() that we took
238 skb = skb_share_check(skb, GFP_ATOMIC);
242 skb_push(skb, ETH_HLEN);
243 ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);
245 ovs_vport_receive(vport, skb, NULL);
252 static unsigned int packet_length(const struct sk_buff *skb)
254 unsigned int length = skb->len - ETH_HLEN;
256 if (skb->protocol == htons(ETH_P_8021Q))
262 static bool dev_supports_vlan_tx(struct net_device *dev)
264 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,37)
265 /* Software fallback means every device supports vlan_tci on TX. */
267 #elif defined(HAVE_VLAN_BUG_WORKAROUND)
268 return dev->features & NETIF_F_HW_VLAN_TX;
270 /* Assume that the driver is buggy. */
275 static int netdev_send(struct vport *vport, struct sk_buff *skb)
277 struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
278 int mtu = netdev_vport->dev->mtu;
281 if (unlikely(packet_length(skb) > mtu && !skb_is_gso(skb))) {
282 net_warn_ratelimited("%s: dropped over-mtu packet: %d > %d\n",
283 netdev_vport->dev->name,
284 packet_length(skb), mtu);
288 skb->dev = netdev_vport->dev;
290 if (vlan_tx_tag_present(skb) && !dev_supports_vlan_tx(skb->dev)) {
293 features = netif_skb_features(skb);
296 features &= ~(NETIF_F_TSO | NETIF_F_TSO6 |
297 NETIF_F_UFO | NETIF_F_FSO);
299 if (netif_needs_gso(skb, features)) {
300 struct sk_buff *nskb;
302 nskb = skb_gso_segment(skb, features);
304 if (unlikely(skb_cloned(skb) &&
305 pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
308 skb_shinfo(skb)->gso_type &= ~SKB_GSO_DODGY;
322 skb = __vlan_put_tag(skb, skb->vlan_proto, vlan_tx_tag_get(skb));
325 vlan_set_tci(skb, 0);
336 skb = __vlan_put_tag(skb, skb->vlan_proto, vlan_tx_tag_get(skb));
339 vlan_set_tci(skb, 0);
352 /* Returns null if this device is not attached to a datapath. */
353 struct vport *ovs_netdev_get_vport(struct net_device *dev)
355 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
356 defined HAVE_RHEL_OVS_HOOK
357 #if IFF_OVS_DATAPATH != 0
358 if (likely(dev->priv_flags & IFF_OVS_DATAPATH))
360 if (likely(rcu_access_pointer(dev->rx_handler) == netdev_frame_hook))
362 #ifdef HAVE_RHEL_OVS_HOOK
363 return (struct vport *)rcu_dereference_rtnl(dev->ax25_ptr);
365 return (struct vport *)rcu_dereference_rtnl(dev->rx_handler_data);
370 return (struct vport *)rcu_dereference_rtnl(dev->br_port);
374 const struct vport_ops ovs_netdev_vport_ops = {
375 .type = OVS_VPORT_TYPE_NETDEV,
376 .create = netdev_create,
377 .destroy = netdev_destroy,
378 .get_name = ovs_netdev_get_name,
382 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36) && \
383 !defined HAVE_RHEL_OVS_HOOK
385 * Enforces, mutual exclusion with the Linux bridge module, by declaring and
386 * exporting br_should_route_hook. Because the bridge module also exports the
387 * same symbol, the module loader will refuse to load both modules at the same
388 * time (e.g. "bridge: exports duplicate symbol br_should_route_hook (owned by
391 * Before Linux 2.6.36, Open vSwitch cannot safely coexist with the Linux
392 * bridge module, so openvswitch uses this macro in those versions. In
393 * Linux 2.6.36 and later, Open vSwitch can coexist with the bridge module.
395 * The use of "typeof" here avoids the need to track changes in the type of
396 * br_should_route_hook over various kernel versions.
398 typeof(br_should_route_hook) br_should_route_hook;
399 EXPORT_SYMBOL(br_should_route_hook);