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 #ifdef HAVE_RHEL_OVS_HOOK
49 static atomic_t nr_bridges = ATOMIC_INIT(0);
51 extern struct sk_buff *(*openvswitch_handle_frame_hook)(struct sk_buff *skb);
54 static void netdev_port_receive(struct vport *vport, struct sk_buff *skb);
56 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,39)
57 /* Called with rcu_read_lock and bottom-halves disabled. */
58 static rx_handler_result_t netdev_frame_hook(struct sk_buff **pskb)
60 struct sk_buff *skb = *pskb;
63 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
64 return RX_HANDLER_PASS;
66 vport = ovs_netdev_get_vport(skb->dev);
68 netdev_port_receive(vport, skb);
70 return RX_HANDLER_CONSUMED;
72 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
73 defined HAVE_RHEL_OVS_HOOK
74 /* Called with rcu_read_lock and bottom-halves disabled. */
75 static struct sk_buff *netdev_frame_hook(struct sk_buff *skb)
79 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
82 vport = ovs_netdev_get_vport(skb->dev);
84 netdev_port_receive(vport, skb);
88 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
90 * Used as br_handle_frame_hook. (Cannot run bridge at the same time, even on
91 * different set of devices!)
93 /* Called with rcu_read_lock and bottom-halves disabled. */
94 static struct sk_buff *netdev_frame_hook(struct net_bridge_port *p,
97 netdev_port_receive((struct vport *)p, skb);
100 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
102 * Used as br_handle_frame_hook. (Cannot run bridge at the same time, even on
103 * different set of devices!)
105 /* Called with rcu_read_lock and bottom-halves disabled. */
106 static int netdev_frame_hook(struct net_bridge_port *p, struct sk_buff **pskb)
108 netdev_port_receive((struct vport *)p, *pskb);
115 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
116 defined HAVE_RHEL_OVS_HOOK
117 static int netdev_init(void) { return 0; }
118 static void netdev_exit(void) { }
120 static int netdev_init(void)
122 /* Hook into callback used by the bridge to intercept packets.
123 * Parasites we are. */
124 br_handle_frame_hook = netdev_frame_hook;
129 static void netdev_exit(void)
131 br_handle_frame_hook = NULL;
135 static struct vport *netdev_create(const struct vport_parms *parms)
138 struct netdev_vport *netdev_vport;
141 vport = ovs_vport_alloc(sizeof(struct netdev_vport),
142 &ovs_netdev_vport_ops, parms);
144 err = PTR_ERR(vport);
148 netdev_vport = netdev_vport_priv(vport);
150 netdev_vport->dev = dev_get_by_name(ovs_dp_get_net(vport->dp), parms->name);
151 if (!netdev_vport->dev) {
153 goto error_free_vport;
156 if (netdev_vport->dev->flags & IFF_LOOPBACK ||
157 netdev_vport->dev->type != ARPHRD_ETHER ||
158 ovs_is_internal_dev(netdev_vport->dev)) {
163 #ifdef HAVE_RHEL_OVS_HOOK
164 rcu_assign_pointer(netdev_vport->dev->ax25_ptr, vport);
165 atomic_inc(&nr_bridges);
166 rcu_assign_pointer(openvswitch_handle_frame_hook, netdev_frame_hook);
168 err = netdev_rx_handler_register(netdev_vport->dev, netdev_frame_hook,
174 dev_set_promiscuity(netdev_vport->dev, 1);
175 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,24)
176 dev_disable_lro(netdev_vport->dev);
178 netdev_vport->dev->priv_flags |= IFF_OVS_DATAPATH;
183 dev_put(netdev_vport->dev);
185 ovs_vport_free(vport);
190 static void free_port_rcu(struct rcu_head *rcu)
192 struct netdev_vport *netdev_vport = container_of(rcu,
193 struct netdev_vport, rcu);
195 #ifdef HAVE_RHEL_OVS_HOOK
196 rcu_assign_pointer(netdev_vport->dev->ax25_ptr, NULL);
198 if (atomic_dec_and_test(&nr_bridges))
199 rcu_assign_pointer(openvswitch_handle_frame_hook, NULL);
201 dev_put(netdev_vport->dev);
202 ovs_vport_free(vport_from_priv(netdev_vport));
205 static void netdev_destroy(struct vport *vport)
207 struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
209 netdev_vport->dev->priv_flags &= ~IFF_OVS_DATAPATH;
210 netdev_rx_handler_unregister(netdev_vport->dev);
211 dev_set_promiscuity(netdev_vport->dev, -1);
213 call_rcu(&netdev_vport->rcu, free_port_rcu);
216 int ovs_netdev_set_addr(struct vport *vport, const unsigned char *addr)
218 struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
221 sa.sa_family = ARPHRD_ETHER;
222 memcpy(sa.sa_data, addr, ETH_ALEN);
224 return dev_set_mac_address(netdev_vport->dev, &sa);
227 const char *ovs_netdev_get_name(const struct vport *vport)
229 const struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
230 return netdev_vport->dev->name;
233 const unsigned char *ovs_netdev_get_addr(const struct vport *vport)
235 const struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
236 return netdev_vport->dev->dev_addr;
239 struct kobject *ovs_netdev_get_kobj(const struct vport *vport)
241 const struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
242 return &netdev_vport->dev->NETDEV_DEV_MEMBER.kobj;
245 unsigned ovs_netdev_get_dev_flags(const struct vport *vport)
247 const struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
248 return dev_get_flags(netdev_vport->dev);
251 int ovs_netdev_is_running(const struct vport *vport)
253 const struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
254 return netif_running(netdev_vport->dev);
257 unsigned char ovs_netdev_get_operstate(const struct vport *vport)
259 const struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
260 return netdev_vport->dev->operstate;
263 int ovs_netdev_get_ifindex(const struct vport *vport)
265 const struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
266 return netdev_vport->dev->ifindex;
269 int ovs_netdev_get_mtu(const struct vport *vport)
271 const struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
272 return netdev_vport->dev->mtu;
275 /* Must be called with rcu_read_lock. */
276 static void netdev_port_receive(struct vport *vport, struct sk_buff *skb)
278 if (unlikely(!vport))
281 if (unlikely(skb_warn_if_lro(skb)))
284 /* Make our own copy of the packet. Otherwise we will mangle the
285 * packet for anyone who came before us (e.g. tcpdump via AF_PACKET).
286 * (No one comes after us, since we tell handle_bridge() that we took
288 skb = skb_share_check(skb, GFP_ATOMIC);
292 skb_push(skb, ETH_HLEN);
294 if (unlikely(compute_ip_summed(skb, false)))
297 vlan_copy_skb_tci(skb);
299 ovs_vport_receive(vport, skb);
306 static unsigned int packet_length(const struct sk_buff *skb)
308 unsigned int length = skb->len - ETH_HLEN;
310 if (skb->protocol == htons(ETH_P_8021Q))
316 static bool dev_supports_vlan_tx(struct net_device *dev)
318 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,37)
319 /* Software fallback means every device supports vlan_tci on TX. */
321 #elif defined(HAVE_VLAN_BUG_WORKAROUND)
322 return dev->features & NETIF_F_HW_VLAN_TX;
324 /* Assume that the driver is buggy. */
329 static int netdev_send(struct vport *vport, struct sk_buff *skb)
331 struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
332 int mtu = netdev_vport->dev->mtu;
335 if (unlikely(packet_length(skb) > mtu && !skb_is_gso(skb))) {
336 net_warn_ratelimited("%s: dropped over-mtu packet: %d > %d\n",
337 netdev_vport->dev->name,
338 packet_length(skb), mtu);
342 skb->dev = netdev_vport->dev;
343 forward_ip_summed(skb, true);
345 if (vlan_tx_tag_present(skb) && !dev_supports_vlan_tx(skb->dev)) {
348 features = netif_skb_features(skb);
351 features &= ~(NETIF_F_TSO | NETIF_F_TSO6 |
352 NETIF_F_UFO | NETIF_F_FSO);
354 if (netif_needs_gso(skb, features)) {
355 struct sk_buff *nskb;
357 nskb = skb_gso_segment(skb, features);
359 if (unlikely(skb_cloned(skb) &&
360 pskb_expand_head(skb, 0, 0, GFP_ATOMIC))) {
365 skb_shinfo(skb)->gso_type &= ~SKB_GSO_DODGY;
381 skb = __vlan_put_tag(skb, vlan_tx_tag_get(skb));
384 vlan_set_tci(skb, 0);
395 skb = __vlan_put_tag(skb, vlan_tx_tag_get(skb));
398 vlan_set_tci(skb, 0);
408 ovs_vport_record_error(vport, VPORT_E_TX_DROPPED);
412 /* Returns null if this device is not attached to a datapath. */
413 struct vport *ovs_netdev_get_vport(struct net_device *dev)
415 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
416 defined HAVE_RHEL_OVS_HOOK
417 #if IFF_OVS_DATAPATH != 0
418 if (likely(dev->priv_flags & IFF_OVS_DATAPATH))
420 if (likely(rcu_access_pointer(dev->rx_handler) == netdev_frame_hook))
422 #ifdef HAVE_RHEL_OVS_HOOK
423 return (struct vport *)rcu_dereference_rtnl(dev->ax25_ptr);
425 return (struct vport *)rcu_dereference_rtnl(dev->rx_handler_data);
430 return (struct vport *)rcu_dereference_rtnl(dev->br_port);
434 const struct vport_ops ovs_netdev_vport_ops = {
435 .type = OVS_VPORT_TYPE_NETDEV,
436 .flags = VPORT_F_REQUIRED,
439 .create = netdev_create,
440 .destroy = netdev_destroy,
441 .set_addr = ovs_netdev_set_addr,
442 .get_name = ovs_netdev_get_name,
443 .get_addr = ovs_netdev_get_addr,
444 .get_kobj = ovs_netdev_get_kobj,
445 .get_dev_flags = ovs_netdev_get_dev_flags,
446 .is_running = ovs_netdev_is_running,
447 .get_operstate = ovs_netdev_get_operstate,
448 .get_ifindex = ovs_netdev_get_ifindex,
449 .get_mtu = ovs_netdev_get_mtu,
453 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36) && \
454 !defined HAVE_RHEL_OVS_HOOK
456 * In kernels earlier than 2.6.36, Open vSwitch cannot safely coexist with the
457 * Linux bridge module, because there is only a single bridge hook function and
458 * only a single br_port member in struct net_device, so this prevents loading
459 * both bridge and openvswitch at the same time.
461 BRIDGE_MUTUAL_EXCLUSION;