2 * Copyright (c) 2010 Nicira Networks.
3 * Distributed under the terms of the GNU GPL version 2.
5 * Significant portions of this file may be copied from parts of the Linux
6 * kernel, by Linus Torvalds and others.
9 #include <linux/if_arp.h>
10 #include <linux/if_bridge.h>
11 #include <linux/if_vlan.h>
12 #include <linux/kernel.h>
13 #include <linux/llc.h>
14 #include <linux/rtnetlink.h>
15 #include <linux/skbuff.h>
20 #include "vport-internal_dev.h"
21 #include "vport-netdev.h"
25 /* If the native device stats aren't 64 bit use the vport stats tracking instead. */
26 #define USE_VPORT_STATS (sizeof(((struct net_device_stats *)0)->rx_bytes) < sizeof(u64))
28 static void netdev_port_receive(struct vport
*vport
, struct sk_buff
*skb
);
30 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36)
31 /* Called with rcu_read_lock and bottom-halves disabled. */
32 static struct sk_buff
*netdev_frame_hook(struct sk_buff
*skb
)
36 if (unlikely(skb
->pkt_type
== PACKET_LOOPBACK
))
39 vport
= netdev_get_vport(skb
->dev
);
41 netdev_port_receive(vport
, skb
);
45 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
47 * Used as br_handle_frame_hook. (Cannot run bridge at the same time, even on
48 * different set of devices!)
50 /* Called with rcu_read_lock and bottom-halves disabled. */
51 static struct sk_buff
*netdev_frame_hook(struct net_bridge_port
*p
,
54 netdev_port_receive((struct vport
*)p
, skb
);
57 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
59 * Used as br_handle_frame_hook. (Cannot run bridge at the same time, even on
60 * different set of devices!)
62 /* Called with rcu_read_lock and bottom-halves disabled. */
63 static int netdev_frame_hook(struct net_bridge_port
*p
, struct sk_buff
**pskb
)
65 netdev_port_receive((struct vport
*)p
, *pskb
);
72 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36)
73 static int netdev_init(void) { return 0; }
74 static void netdev_exit(void) { }
76 static int netdev_init(void)
78 /* Hook into callback used by the bridge to intercept packets.
79 * Parasites we are. */
80 br_handle_frame_hook
= netdev_frame_hook
;
85 static void netdev_exit(void)
87 br_handle_frame_hook
= NULL
;
91 static struct vport
*netdev_create(const char *name
, const void __user
*config
)
94 struct netdev_vport
*netdev_vport
;
97 vport
= vport_alloc(sizeof(struct netdev_vport
), &netdev_vport_ops
);
103 netdev_vport
= netdev_vport_priv(vport
);
105 netdev_vport
->dev
= dev_get_by_name(&init_net
, name
);
106 if (!netdev_vport
->dev
) {
108 goto error_free_vport
;
111 if (netdev_vport
->dev
->flags
& IFF_LOOPBACK
||
112 netdev_vport
->dev
->type
!= ARPHRD_ETHER
||
113 is_internal_dev(netdev_vport
->dev
)) {
118 /* If we are using the vport stats layer initialize it to the current
119 * values so we are roughly consistent with the device stats. */
120 if (USE_VPORT_STATS
) {
121 struct rtnl_link_stats64 stats
;
123 err
= netdev_get_stats(vport
, &stats
);
125 vport_set_stats(vport
, &stats
);
131 dev_put(netdev_vport
->dev
);
138 static int netdev_destroy(struct vport
*vport
)
140 struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
142 dev_put(netdev_vport
->dev
);
148 static int netdev_attach(struct vport
*vport
)
150 struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
153 err
= netdev_rx_handler_register(netdev_vport
->dev
, netdev_frame_hook
,
158 dev_set_promiscuity(netdev_vport
->dev
, 1);
159 dev_disable_lro(netdev_vport
->dev
);
160 netdev_vport
->dev
->priv_flags
|= IFF_OVS_DATAPATH
;
165 static int netdev_detach(struct vport
*vport
)
167 struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
169 netdev_vport
->dev
->priv_flags
&= ~IFF_OVS_DATAPATH
;
170 netdev_rx_handler_unregister(netdev_vport
->dev
);
171 dev_set_promiscuity(netdev_vport
->dev
, -1);
176 int netdev_set_mtu(struct vport
*vport
, int mtu
)
178 struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
179 return dev_set_mtu(netdev_vport
->dev
, mtu
);
182 int netdev_set_addr(struct vport
*vport
, const unsigned char *addr
)
184 struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
187 sa
.sa_family
= ARPHRD_ETHER
;
188 memcpy(sa
.sa_data
, addr
, ETH_ALEN
);
190 return dev_set_mac_address(netdev_vport
->dev
, &sa
);
193 const char *netdev_get_name(const struct vport
*vport
)
195 const struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
196 return netdev_vport
->dev
->name
;
199 const unsigned char *netdev_get_addr(const struct vport
*vport
)
201 const struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
202 return netdev_vport
->dev
->dev_addr
;
205 struct kobject
*netdev_get_kobj(const struct vport
*vport
)
207 const struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
208 return &netdev_vport
->dev
->NETDEV_DEV_MEMBER
.kobj
;
211 int netdev_get_stats(const struct vport
*vport
, struct rtnl_link_stats64
*stats
)
213 const struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
214 dev_get_stats(netdev_vport
->dev
, stats
);
218 unsigned netdev_get_dev_flags(const struct vport
*vport
)
220 const struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
221 return dev_get_flags(netdev_vport
->dev
);
224 int netdev_is_running(const struct vport
*vport
)
226 const struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
227 return netif_running(netdev_vport
->dev
);
230 unsigned char netdev_get_operstate(const struct vport
*vport
)
232 const struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
233 return netdev_vport
->dev
->operstate
;
236 int netdev_get_ifindex(const struct vport
*vport
)
238 const struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
239 return netdev_vport
->dev
->ifindex
;
242 int netdev_get_iflink(const struct vport
*vport
)
244 const struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
245 return netdev_vport
->dev
->iflink
;
248 int netdev_get_mtu(const struct vport
*vport
)
250 const struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
251 return netdev_vport
->dev
->mtu
;
254 /* Must be called with rcu_read_lock. */
255 static void netdev_port_receive(struct vport
*vport
, struct sk_buff
*skb
)
257 /* Make our own copy of the packet. Otherwise we will mangle the
258 * packet for anyone who came before us (e.g. tcpdump via AF_PACKET).
259 * (No one comes after us, since we tell handle_bridge() that we took
261 skb
= skb_share_check(skb
, GFP_ATOMIC
);
265 skb_warn_if_lro(skb
);
267 /* Push the Ethernet header back on. */
268 skb_push(skb
, ETH_HLEN
);
269 skb_reset_mac_header(skb
);
270 compute_ip_summed(skb
, false);
272 vport_receive(vport
, skb
);
275 static int netdev_send(struct vport
*vport
, struct sk_buff
*skb
)
277 struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
280 skb
->dev
= netdev_vport
->dev
;
281 forward_ip_summed(skb
);
287 /* Returns null if this device is not attached to a datapath. */
288 struct vport
*netdev_get_vport(struct net_device
*dev
)
290 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36)
291 /* XXX: The bridge code may have registered the data.
292 * So check that the handler pointer is the datapath's.
293 * Once the merge is done and IFF_OVS_DATAPATH stops
294 * being the same value as IFF_BRIDGE_PORT the check can
295 * simply be netdev_vport->dev->priv_flags & IFF_OVS_DATAPATH. */
296 if (rcu_dereference(dev
->rx_handler
) != netdev_frame_hook
)
298 return (struct vport
*)rcu_dereference(dev
->rx_handler_data
);
300 return (struct vport
*)rcu_dereference(dev
->br_port
);
304 struct vport_ops netdev_vport_ops
= {
306 .flags
= (VPORT_F_REQUIRED
|
307 (USE_VPORT_STATS
? VPORT_F_GEN_STATS
: 0)),
310 .create
= netdev_create
,
311 .destroy
= netdev_destroy
,
312 .attach
= netdev_attach
,
313 .detach
= netdev_detach
,
314 .set_mtu
= netdev_set_mtu
,
315 .set_addr
= netdev_set_addr
,
316 .get_name
= netdev_get_name
,
317 .get_addr
= netdev_get_addr
,
318 .get_kobj
= netdev_get_kobj
,
319 .get_stats
= netdev_get_stats
,
320 .get_dev_flags
= netdev_get_dev_flags
,
321 .is_running
= netdev_is_running
,
322 .get_operstate
= netdev_get_operstate
,
323 .get_ifindex
= netdev_get_ifindex
,
324 .get_iflink
= netdev_get_iflink
,
325 .get_mtu
= netdev_get_mtu
,
329 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36)
331 * In kernels earlier than 2.6.36, Open vSwitch cannot safely coexist with
332 * the Linux bridge module on any released version of Linux, because there
333 * is only a single bridge hook function and only a single br_port member
334 * in struct net_device.
336 * Declaring and exporting this symbol enforces mutual exclusion. The bridge
337 * module also exports the same symbol, so the module loader will refuse to
338 * load both modules at the same time (e.g. "bridge: exports duplicate symbol
339 * br_should_route_hook (owned by openvswitch_mod)").
341 * The use of "typeof" here avoids the need to track changes in the type of
342 * br_should_route_hook over various kernel versions.
344 typeof(br_should_route_hook
) br_should_route_hook
;
345 EXPORT_SYMBOL(br_should_route_hook
);