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 static void netdev_port_receive(struct vport
*vport
, struct sk_buff
*skb
);
39 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,39)
40 /* Called with rcu_read_lock and bottom-halves disabled. */
41 static rx_handler_result_t
netdev_frame_hook(struct sk_buff
**pskb
)
43 struct sk_buff
*skb
= *pskb
;
46 if (unlikely(skb
->pkt_type
== PACKET_LOOPBACK
))
47 return RX_HANDLER_PASS
;
49 vport
= ovs_netdev_get_vport(skb
->dev
);
51 netdev_port_receive(vport
, skb
);
53 return RX_HANDLER_CONSUMED
;
55 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
56 defined HAVE_RHEL_OVS_HOOK
57 /* Called with rcu_read_lock and bottom-halves disabled. */
58 static struct sk_buff
*netdev_frame_hook(struct sk_buff
*skb
)
62 if (unlikely(skb
->pkt_type
== PACKET_LOOPBACK
))
65 vport
= ovs_netdev_get_vport(skb
->dev
);
67 netdev_port_receive(vport
, skb
);
71 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,32)
73 * Used as br_handle_frame_hook. (Cannot run bridge at the same time, even on
74 * different set of devices!)
76 /* Called with rcu_read_lock and bottom-halves disabled. */
77 static struct sk_buff
*netdev_frame_hook(struct net_bridge_port
*p
,
80 netdev_port_receive((struct vport
*)p
, skb
);
87 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
88 defined HAVE_RHEL_OVS_HOOK
89 static int netdev_init(void) { return 0; }
90 static void netdev_exit(void) { }
92 static int port_count
;
94 static void netdev_init(void)
100 /* Hook into callback used by the bridge to intercept packets.
101 * Parasites we are. */
102 br_handle_frame_hook
= netdev_frame_hook
;
107 static void netdev_exit(void)
113 br_handle_frame_hook
= NULL
;
117 static struct net_device
*get_dpdev(struct datapath
*dp
)
121 local
= ovs_vport_ovsl(dp
, OVSP_LOCAL
);
123 return netdev_vport_priv(local
)->dev
;
126 static struct vport
*netdev_create(const struct vport_parms
*parms
)
129 struct netdev_vport
*netdev_vport
;
132 vport
= ovs_vport_alloc(sizeof(struct netdev_vport
),
133 &ovs_netdev_vport_ops
, parms
);
135 err
= PTR_ERR(vport
);
139 netdev_vport
= netdev_vport_priv(vport
);
141 netdev_vport
->dev
= dev_get_by_name(ovs_dp_get_net(vport
->dp
), parms
->name
);
142 if (!netdev_vport
->dev
) {
144 goto error_free_vport
;
147 if (netdev_vport
->dev
->flags
& IFF_LOOPBACK
||
148 netdev_vport
->dev
->type
!= ARPHRD_ETHER
||
149 ovs_is_internal_dev(netdev_vport
->dev
)) {
155 err
= netdev_master_upper_dev_link(netdev_vport
->dev
,
156 get_dpdev(vport
->dp
));
160 err
= netdev_rx_handler_register(netdev_vport
->dev
, netdev_frame_hook
,
163 goto error_master_upper_dev_unlink
;
165 dev_set_promiscuity(netdev_vport
->dev
, 1);
166 netdev_vport
->dev
->priv_flags
|= IFF_OVS_DATAPATH
;
172 error_master_upper_dev_unlink
:
173 netdev_upper_dev_unlink(netdev_vport
->dev
, get_dpdev(vport
->dp
));
177 dev_put(netdev_vport
->dev
);
179 ovs_vport_free(vport
);
184 static void free_port_rcu(struct rcu_head
*rcu
)
186 struct netdev_vport
*netdev_vport
= container_of(rcu
,
187 struct netdev_vport
, rcu
);
189 dev_put(netdev_vport
->dev
);
190 ovs_vport_free(vport_from_priv(netdev_vport
));
193 static void netdev_destroy(struct vport
*vport
)
195 struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
199 netdev_vport
->dev
->priv_flags
&= ~IFF_OVS_DATAPATH
;
200 netdev_rx_handler_unregister(netdev_vport
->dev
);
201 netdev_upper_dev_unlink(netdev_vport
->dev
, get_dpdev(vport
->dp
));
202 dev_set_promiscuity(netdev_vport
->dev
, -1);
205 call_rcu(&netdev_vport
->rcu
, free_port_rcu
);
208 const char *ovs_netdev_get_name(const struct vport
*vport
)
210 const struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
211 return netdev_vport
->dev
->name
;
214 /* Must be called with rcu_read_lock. */
215 static void netdev_port_receive(struct vport
*vport
, struct sk_buff
*skb
)
217 if (unlikely(!vport
))
220 if (unlikely(skb_warn_if_lro(skb
)))
223 /* Make our own copy of the packet. Otherwise we will mangle the
224 * packet for anyone who came before us (e.g. tcpdump via AF_PACKET).
225 * (No one comes after us, since we tell handle_bridge() that we took
227 skb
= skb_share_check(skb
, GFP_ATOMIC
);
231 skb_push(skb
, ETH_HLEN
);
232 ovs_skb_postpush_rcsum(skb
, skb
->data
, ETH_HLEN
);
234 ovs_vport_receive(vport
, skb
, NULL
);
241 static unsigned int packet_length(const struct sk_buff
*skb
)
243 unsigned int length
= skb
->len
- ETH_HLEN
;
245 if (skb
->protocol
== htons(ETH_P_8021Q
))
251 static int netdev_send(struct vport
*vport
, struct sk_buff
*skb
)
253 struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
254 int mtu
= netdev_vport
->dev
->mtu
;
257 if (unlikely(packet_length(skb
) > mtu
&& !skb_is_gso(skb
))) {
258 net_warn_ratelimited("%s: dropped over-mtu packet: %d > %d\n",
259 netdev_vport
->dev
->name
,
260 packet_length(skb
), mtu
);
264 skb
->dev
= netdev_vport
->dev
;
275 /* Returns null if this device is not attached to a datapath. */
276 struct vport
*ovs_netdev_get_vport(struct net_device
*dev
)
278 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
279 defined HAVE_RHEL_OVS_HOOK
280 #if IFF_OVS_DATAPATH != 0
281 if (likely(dev
->priv_flags
& IFF_OVS_DATAPATH
))
283 if (likely(rcu_access_pointer(dev
->rx_handler
) == netdev_frame_hook
))
285 #ifdef HAVE_RHEL_OVS_HOOK
286 return (struct vport
*)rcu_dereference_rtnl(dev
->ax25_ptr
);
288 return (struct vport
*)rcu_dereference_rtnl(dev
->rx_handler_data
);
293 return (struct vport
*)rcu_dereference_rtnl(dev
->br_port
);
297 const struct vport_ops ovs_netdev_vport_ops
= {
298 .type
= OVS_VPORT_TYPE_NETDEV
,
299 .create
= netdev_create
,
300 .destroy
= netdev_destroy
,
301 .get_name
= ovs_netdev_get_name
,
305 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36) && \
306 !defined HAVE_RHEL_OVS_HOOK
308 * Enforces, mutual exclusion with the Linux bridge module, by declaring and
309 * exporting br_should_route_hook. Because the bridge module also exports the
310 * same symbol, the module loader will refuse to load both modules at the same
311 * time (e.g. "bridge: exports duplicate symbol br_should_route_hook (owned by
314 * Before Linux 2.6.36, Open vSwitch cannot safely coexist with the Linux
315 * bridge module, so openvswitch uses this macro in those versions. In
316 * Linux 2.6.36 and later, Open vSwitch can coexist with the bridge module.
318 * The use of "typeof" here avoids the need to track changes in the type of
319 * br_should_route_hook over various kernel versions.
321 typeof(br_should_route_hook
) br_should_route_hook
;
322 EXPORT_SYMBOL(br_should_route_hook
);