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>
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 /* Called with rcu_read_lock and bottom-halves disabled. */
68 static struct sk_buff *netdev_frame_hook(struct sk_buff *skb)
72 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
75 vport = ovs_netdev_get_vport(skb->dev);
77 netdev_port_receive(vport, skb);
81 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
83 * Used as br_handle_frame_hook. (Cannot run bridge at the same time, even on
84 * different set of devices!)
86 /* Called with rcu_read_lock and bottom-halves disabled. */
87 static struct sk_buff *netdev_frame_hook(struct net_bridge_port *p,
90 netdev_port_receive((struct vport *)p, skb);
93 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
95 * Used as br_handle_frame_hook. (Cannot run bridge at the same time, even on
96 * different set of devices!)
98 /* Called with rcu_read_lock and bottom-halves disabled. */
99 static int netdev_frame_hook(struct net_bridge_port *p, struct sk_buff **pskb)
101 netdev_port_receive((struct vport *)p, *pskb);
108 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36)
109 static int netdev_init(void) { return 0; }
110 static void netdev_exit(void) { }
112 static int netdev_init(void)
114 /* Hook into callback used by the bridge to intercept packets.
115 * Parasites we are. */
116 br_handle_frame_hook = netdev_frame_hook;
121 static void netdev_exit(void)
123 br_handle_frame_hook = NULL;
127 static struct vport *netdev_create(const struct vport_parms *parms)
130 struct netdev_vport *netdev_vport;
133 vport = ovs_vport_alloc(sizeof(struct netdev_vport),
134 &ovs_netdev_vport_ops, parms);
136 err = PTR_ERR(vport);
140 netdev_vport = netdev_vport_priv(vport);
142 netdev_vport->dev = dev_get_by_name(ovs_dp_get_net(vport->dp), parms->name);
143 if (!netdev_vport->dev) {
145 goto error_free_vport;
148 if (netdev_vport->dev->flags & IFF_LOOPBACK ||
149 netdev_vport->dev->type != ARPHRD_ETHER ||
150 ovs_is_internal_dev(netdev_vport->dev)) {
155 err = netdev_rx_handler_register(netdev_vport->dev, netdev_frame_hook,
160 dev_set_promiscuity(netdev_vport->dev, 1);
161 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,24)
162 dev_disable_lro(netdev_vport->dev);
164 netdev_vport->dev->priv_flags |= IFF_OVS_DATAPATH;
169 dev_put(netdev_vport->dev);
171 ovs_vport_free(vport);
176 static void free_port_rcu(struct rcu_head *rcu)
178 struct netdev_vport *netdev_vport = container_of(rcu,
179 struct netdev_vport, rcu);
181 dev_put(netdev_vport->dev);
182 ovs_vport_free(vport_from_priv(netdev_vport));
185 static void netdev_destroy(struct vport *vport)
187 struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
189 netdev_vport->dev->priv_flags &= ~IFF_OVS_DATAPATH;
190 netdev_rx_handler_unregister(netdev_vport->dev);
191 dev_set_promiscuity(netdev_vport->dev, -1);
193 call_rcu(&netdev_vport->rcu, free_port_rcu);
196 const char *ovs_netdev_get_name(const struct vport *vport)
198 const struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
199 return netdev_vport->dev->name;
202 int ovs_netdev_get_ifindex(const struct vport *vport)
204 const struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
205 return netdev_vport->dev->ifindex;
208 /* Must be called with rcu_read_lock. */
209 static void netdev_port_receive(struct vport *vport, struct sk_buff *skb)
211 if (unlikely(!vport))
214 if (unlikely(skb_warn_if_lro(skb)))
217 /* Make our own copy of the packet. Otherwise we will mangle the
218 * packet for anyone who came before us (e.g. tcpdump via AF_PACKET).
219 * (No one comes after us, since we tell handle_bridge() that we took
221 skb = skb_share_check(skb, GFP_ATOMIC);
225 skb_push(skb, ETH_HLEN);
227 if (unlikely(compute_ip_summed(skb, false)))
230 vlan_copy_skb_tci(skb);
232 ovs_vport_receive(vport, skb);
239 static unsigned int packet_length(const struct sk_buff *skb)
241 unsigned int length = skb->len - ETH_HLEN;
243 if (skb->protocol == htons(ETH_P_8021Q))
249 static bool dev_supports_vlan_tx(struct net_device *dev)
251 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,37)
252 /* Software fallback means every device supports vlan_tci on TX. */
254 #elif defined(HAVE_VLAN_BUG_WORKAROUND)
255 return dev->features & NETIF_F_HW_VLAN_TX;
257 /* Assume that the driver is buggy. */
262 static int netdev_send(struct vport *vport, struct sk_buff *skb)
264 struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
265 int mtu = netdev_vport->dev->mtu;
268 if (unlikely(packet_length(skb) > mtu && !skb_is_gso(skb))) {
269 net_warn_ratelimited("%s: dropped over-mtu packet: %d > %d\n",
270 netdev_vport->dev->name,
271 packet_length(skb), mtu);
275 skb->dev = netdev_vport->dev;
276 forward_ip_summed(skb, true);
278 if (vlan_tx_tag_present(skb) && !dev_supports_vlan_tx(skb->dev)) {
281 features = netif_skb_features(skb);
284 features &= ~(NETIF_F_TSO | NETIF_F_TSO6 |
285 NETIF_F_UFO | NETIF_F_FSO);
287 if (netif_needs_gso(skb, features)) {
288 struct sk_buff *nskb;
290 nskb = skb_gso_segment(skb, features);
292 if (unlikely(skb_cloned(skb) &&
293 pskb_expand_head(skb, 0, 0, GFP_ATOMIC))) {
298 skb_shinfo(skb)->gso_type &= ~SKB_GSO_DODGY;
314 skb = __vlan_put_tag(skb, vlan_tx_tag_get(skb));
317 vlan_set_tci(skb, 0);
328 skb = __vlan_put_tag(skb, vlan_tx_tag_get(skb));
331 vlan_set_tci(skb, 0);
341 ovs_vport_record_error(vport, VPORT_E_TX_DROPPED);
345 /* Returns null if this device is not attached to a datapath. */
346 struct vport *ovs_netdev_get_vport(struct net_device *dev)
348 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36)
349 #if IFF_OVS_DATAPATH != 0
350 if (likely(dev->priv_flags & IFF_OVS_DATAPATH))
352 if (likely(rcu_access_pointer(dev->rx_handler) == netdev_frame_hook))
354 return (struct vport *)rcu_dereference_rtnl(dev->rx_handler_data);
358 return (struct vport *)rcu_dereference_rtnl(dev->br_port);
362 const struct vport_ops ovs_netdev_vport_ops = {
363 .type = OVS_VPORT_TYPE_NETDEV,
364 .flags = VPORT_F_REQUIRED,
367 .create = netdev_create,
368 .destroy = netdev_destroy,
369 .get_name = ovs_netdev_get_name,
370 .get_ifindex = ovs_netdev_get_ifindex,
374 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36)
376 * Enforces, mutual exclusion with the Linux bridge module, by declaring and
377 * exporting br_should_route_hook. Because the bridge module also exports the
378 * same symbol, the module loader will refuse to load both modules at the same
379 * time (e.g. "bridge: exports duplicate symbol br_should_route_hook (owned by
382 * Before Linux 2.6.36, Open vSwitch cannot safely coexist with the Linux
383 * bridge module, so openvswitch uses this macro in those versions. In
384 * Linux 2.6.36 and later, Open vSwitch can coexist with the bridge module.
386 * The use of "typeof" here avoids the need to track changes in the type of
387 * br_should_route_hook over various kernel versions.
389 typeof(br_should_route_hook) br_should_route_hook;
390 EXPORT_SYMBOL(br_should_route_hook);