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 port_count
;
122 static void netdev_init(void)
128 /* Hook into callback used by the bridge to intercept packets.
129 * Parasites we are. */
130 br_handle_frame_hook
= netdev_frame_hook
;
135 static void netdev_exit(void)
141 br_handle_frame_hook
= NULL
;
145 static struct vport
*netdev_create(const struct vport_parms
*parms
)
148 struct netdev_vport
*netdev_vport
;
151 vport
= ovs_vport_alloc(sizeof(struct netdev_vport
),
152 &ovs_netdev_vport_ops
, parms
);
154 err
= PTR_ERR(vport
);
158 netdev_vport
= netdev_vport_priv(vport
);
160 netdev_vport
->dev
= dev_get_by_name(ovs_dp_get_net(vport
->dp
), parms
->name
);
161 if (!netdev_vport
->dev
) {
163 goto error_free_vport
;
166 if (netdev_vport
->dev
->flags
& IFF_LOOPBACK
||
167 netdev_vport
->dev
->type
!= ARPHRD_ETHER
||
168 ovs_is_internal_dev(netdev_vport
->dev
)) {
174 #ifdef HAVE_RHEL_OVS_HOOK
175 rcu_assign_pointer(netdev_vport
->dev
->ax25_ptr
, vport
);
176 atomic_inc(&nr_bridges
);
177 rcu_assign_pointer(openvswitch_handle_frame_hook
, netdev_frame_hook
);
179 err
= netdev_rx_handler_register(netdev_vport
->dev
, netdev_frame_hook
,
185 dev_set_promiscuity(netdev_vport
->dev
, 1);
186 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,24)
187 dev_disable_lro(netdev_vport
->dev
);
189 netdev_vport
->dev
->priv_flags
|= IFF_OVS_DATAPATH
;
195 #ifndef HAVE_RHEL_OVS_HOOK
200 dev_put(netdev_vport
->dev
);
202 ovs_vport_free(vport
);
207 static void free_port_rcu(struct rcu_head
*rcu
)
209 struct netdev_vport
*netdev_vport
= container_of(rcu
,
210 struct netdev_vport
, rcu
);
212 #ifdef HAVE_RHEL_OVS_HOOK
213 rcu_assign_pointer(netdev_vport
->dev
->ax25_ptr
, NULL
);
215 if (atomic_dec_and_test(&nr_bridges
))
216 rcu_assign_pointer(openvswitch_handle_frame_hook
, NULL
);
218 dev_put(netdev_vport
->dev
);
219 ovs_vport_free(vport_from_priv(netdev_vport
));
222 static void netdev_destroy(struct vport
*vport
)
224 struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
228 netdev_vport
->dev
->priv_flags
&= ~IFF_OVS_DATAPATH
;
229 netdev_rx_handler_unregister(netdev_vport
->dev
);
230 dev_set_promiscuity(netdev_vport
->dev
, -1);
233 call_rcu(&netdev_vport
->rcu
, free_port_rcu
);
236 const char *ovs_netdev_get_name(const struct vport
*vport
)
238 const struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
239 return netdev_vport
->dev
->name
;
242 /* Must be called with rcu_read_lock. */
243 static void netdev_port_receive(struct vport
*vport
, struct sk_buff
*skb
)
245 if (unlikely(!vport
))
248 if (unlikely(skb_warn_if_lro(skb
)))
251 /* Make our own copy of the packet. Otherwise we will mangle the
252 * packet for anyone who came before us (e.g. tcpdump via AF_PACKET).
253 * (No one comes after us, since we tell handle_bridge() that we took
255 skb
= skb_share_check(skb
, GFP_ATOMIC
);
259 if (unlikely(compute_ip_summed(skb
, false)))
262 skb_push(skb
, ETH_HLEN
);
263 ovs_skb_postpush_rcsum(skb
, skb
->data
, ETH_HLEN
);
265 vlan_copy_skb_tci(skb
);
267 ovs_vport_receive(vport
, skb
, NULL
);
274 static unsigned int packet_length(const struct sk_buff
*skb
)
276 unsigned int length
= skb
->len
- ETH_HLEN
;
278 if (skb
->protocol
== htons(ETH_P_8021Q
))
284 static bool dev_supports_vlan_tx(struct net_device
*dev
)
286 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,37)
287 /* Software fallback means every device supports vlan_tci on TX. */
289 #elif defined(HAVE_VLAN_BUG_WORKAROUND)
290 return dev
->features
& NETIF_F_HW_VLAN_TX
;
292 /* Assume that the driver is buggy. */
297 static int netdev_send(struct vport
*vport
, struct sk_buff
*skb
)
299 struct netdev_vport
*netdev_vport
= netdev_vport_priv(vport
);
300 int mtu
= netdev_vport
->dev
->mtu
;
303 if (unlikely(packet_length(skb
) > mtu
&& !skb_is_gso(skb
))) {
304 net_warn_ratelimited("%s: dropped over-mtu packet: %d > %d\n",
305 netdev_vport
->dev
->name
,
306 packet_length(skb
), mtu
);
310 skb
->dev
= netdev_vport
->dev
;
311 forward_ip_summed(skb
, true);
313 if (vlan_tx_tag_present(skb
) && !dev_supports_vlan_tx(skb
->dev
)) {
316 features
= netif_skb_features(skb
);
319 features
&= ~(NETIF_F_TSO
| NETIF_F_TSO6
|
320 NETIF_F_UFO
| NETIF_F_FSO
);
322 if (netif_needs_gso(skb
, features
)) {
323 struct sk_buff
*nskb
;
325 nskb
= skb_gso_segment(skb
, features
);
327 if (unlikely(skb_cloned(skb
) &&
328 pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
)))
331 skb_shinfo(skb
)->gso_type
&= ~SKB_GSO_DODGY
;
345 skb
= __vlan_put_tag(skb
, vlan_tx_tag_get(skb
));
348 vlan_set_tci(skb
, 0);
359 skb
= __vlan_put_tag(skb
, vlan_tx_tag_get(skb
));
362 vlan_set_tci(skb
, 0);
375 /* Returns null if this device is not attached to a datapath. */
376 struct vport
*ovs_netdev_get_vport(struct net_device
*dev
)
378 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
379 defined HAVE_RHEL_OVS_HOOK
380 #if IFF_OVS_DATAPATH != 0
381 if (likely(dev
->priv_flags
& IFF_OVS_DATAPATH
))
383 if (likely(rcu_access_pointer(dev
->rx_handler
) == netdev_frame_hook
))
385 #ifdef HAVE_RHEL_OVS_HOOK
386 return (struct vport
*)rcu_dereference_rtnl(dev
->ax25_ptr
);
388 return (struct vport
*)rcu_dereference_rtnl(dev
->rx_handler_data
);
393 return (struct vport
*)rcu_dereference_rtnl(dev
->br_port
);
397 const struct vport_ops ovs_netdev_vport_ops
= {
398 .type
= OVS_VPORT_TYPE_NETDEV
,
399 .create
= netdev_create
,
400 .destroy
= netdev_destroy
,
401 .get_name
= ovs_netdev_get_name
,
405 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36) && \
406 !defined HAVE_RHEL_OVS_HOOK
408 * Enforces, mutual exclusion with the Linux bridge module, by declaring and
409 * exporting br_should_route_hook. Because the bridge module also exports the
410 * same symbol, the module loader will refuse to load both modules at the same
411 * time (e.g. "bridge: exports duplicate symbol br_should_route_hook (owned by
414 * Before Linux 2.6.36, Open vSwitch cannot safely coexist with the Linux
415 * bridge module, so openvswitch uses this macro in those versions. In
416 * Linux 2.6.36 and later, Open vSwitch can coexist with the bridge module.
418 * The use of "typeof" here avoids the need to track changes in the type of
419 * br_should_route_hook over various kernel versions.
421 typeof(br_should_route_hook
) br_should_route_hook
;
422 EXPORT_SYMBOL(br_should_route_hook
);