xref: /linux/net/openvswitch/vport-netdev.c (revision 54c9ee39923c0ffa27b99a1b7321e0d07408c97b) 
1 /*
2  * Copyright (c) 2007-2012 Nicira, Inc.
3  *
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.
7  *
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.
12  *
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
16  * 02110-1301, USA
17  */
18 
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
20 
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>
29 
30 #include <net/udp.h>
31 #include <net/ip_tunnels.h>
32 #include <net/rtnetlink.h>
33 #include <net/vxlan.h>
34 
35 #include "datapath.h"
36 #include "vport.h"
37 #include "vport-internal_dev.h"
38 #include "vport-netdev.h"
39 
40 static struct vport_ops ovs_netdev_vport_ops;
41 
42 /* Must be called with rcu_read_lock. */
43 static void netdev_port_receive(struct vport *vport, struct sk_buff *skb)
44 {
45 	if (unlikely(!vport))
46 		goto error;
47 
48 	if (unlikely(skb_warn_if_lro(skb)))
49 		goto error;
50 
51 	/* Make our own copy of the packet.  Otherwise we will mangle the
52 	 * packet for anyone who came before us (e.g. tcpdump via AF_PACKET).
53 	 */
54 	skb = skb_share_check(skb, GFP_ATOMIC);
55 	if (unlikely(!skb))
56 		return;
57 
58 	skb_push(skb, ETH_HLEN);
59 	ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);
60 
61 	ovs_vport_receive(vport, skb, NULL);
62 	return;
63 
64 error:
65 	kfree_skb(skb);
66 }
67 
68 /* Called with rcu_read_lock and bottom-halves disabled. */
69 static rx_handler_result_t netdev_frame_hook(struct sk_buff **pskb)
70 {
71 	struct sk_buff *skb = *pskb;
72 	struct vport *vport;
73 
74 	if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
75 		return RX_HANDLER_PASS;
76 
77 	vport = ovs_netdev_get_vport(skb->dev);
78 
79 	netdev_port_receive(vport, skb);
80 
81 	return RX_HANDLER_CONSUMED;
82 }
83 
84 static struct net_device *get_dpdev(const struct datapath *dp)
85 {
86 	struct vport *local;
87 
88 	local = ovs_vport_ovsl(dp, OVSP_LOCAL);
89 	BUG_ON(!local);
90 	return local->dev;
91 }
92 
93 static struct vport *netdev_link(struct vport *vport, const char *name)
94 {
95 	int err;
96 
97 	vport->dev = dev_get_by_name(ovs_dp_get_net(vport->dp), name);
98 	if (!vport->dev) {
99 		err = -ENODEV;
100 		goto error_free_vport;
101 	}
102 
103 	if (vport->dev->flags & IFF_LOOPBACK ||
104 	    vport->dev->type != ARPHRD_ETHER ||
105 	    ovs_is_internal_dev(vport->dev)) {
106 		err = -EINVAL;
107 		goto error_put;
108 	}
109 
110 	rtnl_lock();
111 	err = netdev_master_upper_dev_link(vport->dev,
112 					   get_dpdev(vport->dp));
113 	if (err)
114 		goto error_unlock;
115 
116 	err = netdev_rx_handler_register(vport->dev, netdev_frame_hook,
117 					 vport);
118 	if (err)
119 		goto error_master_upper_dev_unlink;
120 
121 	dev_disable_lro(vport->dev);
122 	dev_set_promiscuity(vport->dev, 1);
123 	vport->dev->priv_flags |= IFF_OVS_DATAPATH;
124 	rtnl_unlock();
125 
126 	return vport;
127 
128 error_master_upper_dev_unlink:
129 	netdev_upper_dev_unlink(vport->dev, get_dpdev(vport->dp));
130 error_unlock:
131 	rtnl_unlock();
132 error_put:
133 	dev_put(vport->dev);
134 error_free_vport:
135 	ovs_vport_free(vport);
136 	return ERR_PTR(err);
137 }
138 
139 static struct vport *netdev_create(const struct vport_parms *parms)
140 {
141 	struct vport *vport;
142 
143 	vport = ovs_vport_alloc(0, &ovs_netdev_vport_ops, parms);
144 	if (IS_ERR(vport))
145 		return vport;
146 
147 	return netdev_link(vport, parms->name);
148 }
149 
150 static void free_port_rcu(struct rcu_head *rcu)
151 {
152 	struct vport *vport = container_of(rcu, struct vport, rcu);
153 
154 	if (vport->dev)
155 		dev_put(vport->dev);
156 	ovs_vport_free(vport);
157 }
158 
159 void ovs_netdev_detach_dev(struct vport *vport)
160 {
161 	ASSERT_RTNL();
162 	vport->dev->priv_flags &= ~IFF_OVS_DATAPATH;
163 	netdev_rx_handler_unregister(vport->dev);
164 	netdev_upper_dev_unlink(vport->dev,
165 				netdev_master_upper_dev_get(vport->dev));
166 	dev_set_promiscuity(vport->dev, -1);
167 }
168 
169 static void netdev_destroy(struct vport *vport)
170 {
171 	rtnl_lock();
172 	if (vport->dev->priv_flags & IFF_OVS_DATAPATH)
173 		ovs_netdev_detach_dev(vport);
174 	rtnl_unlock();
175 
176 	call_rcu(&vport->rcu, free_port_rcu);
177 }
178 
179 static unsigned int packet_length(const struct sk_buff *skb)
180 {
181 	unsigned int length = skb->len - ETH_HLEN;
182 
183 	if (skb->protocol == htons(ETH_P_8021Q))
184 		length -= VLAN_HLEN;
185 
186 	return length;
187 }
188 
189 static int netdev_send(struct vport *vport, struct sk_buff *skb)
190 {
191 	int mtu = vport->dev->mtu;
192 	int len;
193 
194 	if (unlikely(packet_length(skb) > mtu && !skb_is_gso(skb))) {
195 		net_warn_ratelimited("%s: dropped over-mtu packet: %d > %d\n",
196 				     vport->dev->name,
197 				     packet_length(skb), mtu);
198 		goto drop;
199 	}
200 
201 	skb->dev = vport->dev;
202 	len = skb->len;
203 	dev_queue_xmit(skb);
204 
205 	return len;
206 
207 drop:
208 	kfree_skb(skb);
209 	return 0;
210 }
211 
212 /* Returns null if this device is not attached to a datapath. */
213 struct vport *ovs_netdev_get_vport(struct net_device *dev)
214 {
215 	if (likely(dev->priv_flags & IFF_OVS_DATAPATH))
216 		return (struct vport *)
217 			rcu_dereference_rtnl(dev->rx_handler_data);
218 	else
219 		return NULL;
220 }
221 
222 static struct vport_ops ovs_netdev_vport_ops = {
223 	.type		= OVS_VPORT_TYPE_NETDEV,
224 	.create		= netdev_create,
225 	.destroy	= netdev_destroy,
226 	.send		= netdev_send,
227 };
228 
229 /* Compat code for old userspace. */
230 #if IS_ENABLED(CONFIG_VXLAN)
231 static struct vport_ops ovs_vxlan_netdev_vport_ops;
232 
233 static int vxlan_get_options(const struct vport *vport, struct sk_buff *skb)
234 {
235 	struct vxlan_dev *vxlan = netdev_priv(vport->dev);
236 	__be16 dst_port = vxlan->cfg.dst_port;
237 
238 	if (nla_put_u16(skb, OVS_TUNNEL_ATTR_DST_PORT, ntohs(dst_port)))
239 		return -EMSGSIZE;
240 
241 	if (vxlan->flags & VXLAN_F_GBP) {
242 		struct nlattr *exts;
243 
244 		exts = nla_nest_start(skb, OVS_TUNNEL_ATTR_EXTENSION);
245 		if (!exts)
246 			return -EMSGSIZE;
247 
248 		if (vxlan->flags & VXLAN_F_GBP &&
249 		    nla_put_flag(skb, OVS_VXLAN_EXT_GBP))
250 			return -EMSGSIZE;
251 
252 		nla_nest_end(skb, exts);
253 	}
254 
255 	return 0;
256 }
257 
258 static const struct nla_policy exts_policy[OVS_VXLAN_EXT_MAX + 1] = {
259 	[OVS_VXLAN_EXT_GBP]	= { .type = NLA_FLAG, },
260 };
261 
262 static int vxlan_configure_exts(struct vport *vport, struct nlattr *attr,
263 				struct vxlan_config *conf)
264 {
265 	struct nlattr *exts[OVS_VXLAN_EXT_MAX + 1];
266 	int err;
267 
268 	if (nla_len(attr) < sizeof(struct nlattr))
269 		return -EINVAL;
270 
271 	err = nla_parse_nested(exts, OVS_VXLAN_EXT_MAX, attr, exts_policy);
272 	if (err < 0)
273 		return err;
274 
275 	if (exts[OVS_VXLAN_EXT_GBP])
276 		conf->flags |= VXLAN_F_GBP;
277 
278 	return 0;
279 }
280 
281 static struct vport *vxlan_tnl_create(const struct vport_parms *parms)
282 {
283 	struct net *net = ovs_dp_get_net(parms->dp);
284 	struct nlattr *options = parms->options;
285 	struct net_device *dev;
286 	struct vport *vport;
287 	struct nlattr *a;
288 	int err;
289 	struct vxlan_config conf = {
290 		.no_share = true,
291 		.flags = VXLAN_F_FLOW_BASED | VXLAN_F_COLLECT_METADATA,
292 	};
293 
294 	if (!options) {
295 		err = -EINVAL;
296 		goto error;
297 	}
298 
299 	a = nla_find_nested(options, OVS_TUNNEL_ATTR_DST_PORT);
300 	if (a && nla_len(a) == sizeof(u16)) {
301 		conf.dst_port = htons(nla_get_u16(a));
302 	} else {
303 		/* Require destination port from userspace. */
304 		err = -EINVAL;
305 		goto error;
306 	}
307 
308 	vport = ovs_vport_alloc(0, &ovs_vxlan_netdev_vport_ops, parms);
309 	if (IS_ERR(vport))
310 		return vport;
311 
312 	a = nla_find_nested(options, OVS_TUNNEL_ATTR_EXTENSION);
313 	if (a) {
314 		err = vxlan_configure_exts(vport, a, &conf);
315 		if (err) {
316 			ovs_vport_free(vport);
317 			goto error;
318 		}
319 	}
320 
321 	rtnl_lock();
322 	dev = vxlan_dev_create(net, parms->name, NET_NAME_USER, &conf);
323 	if (IS_ERR(dev)) {
324 		rtnl_unlock();
325 		ovs_vport_free(vport);
326 		return ERR_CAST(dev);
327 	}
328 
329 	dev_change_flags(dev, dev->flags | IFF_UP);
330 	rtnl_unlock();
331 	return vport;
332 error:
333 	return ERR_PTR(err);
334 }
335 
336 static struct vport *vxlan_create(const struct vport_parms *parms)
337 {
338 	struct vport *vport;
339 
340 	vport = vxlan_tnl_create(parms);
341 	if (IS_ERR(vport))
342 		return vport;
343 
344 	return netdev_link(vport, parms->name);
345 }
346 
347 static void vxlan_destroy(struct vport *vport)
348 {
349 	rtnl_lock();
350 	if (vport->dev->priv_flags & IFF_OVS_DATAPATH)
351 		ovs_netdev_detach_dev(vport);
352 
353 	/* Early release so we can unregister the device */
354 	dev_put(vport->dev);
355 	rtnl_delete_link(vport->dev);
356 	vport->dev = NULL;
357 	rtnl_unlock();
358 
359 	call_rcu(&vport->rcu, free_port_rcu);
360 }
361 
362 static int vxlan_get_egress_tun_info(struct vport *vport, struct sk_buff *skb,
363 				     struct ip_tunnel_info *egress_tun_info)
364 {
365 	struct vxlan_dev *vxlan = netdev_priv(vport->dev);
366 	struct net *net = ovs_dp_get_net(vport->dp);
367 	__be16 dst_port = vxlan_dev_dst_port(vxlan);
368 	__be16 src_port;
369 	int port_min;
370 	int port_max;
371 
372 	inet_get_local_port_range(net, &port_min, &port_max);
373 	src_port = udp_flow_src_port(net, skb, 0, 0, true);
374 
375 	return ovs_tunnel_get_egress_info(egress_tun_info, net,
376 					  OVS_CB(skb)->egress_tun_info,
377 					  IPPROTO_UDP, skb->mark,
378 					  src_port, dst_port);
379 }
380 
381 static struct vport_ops ovs_vxlan_netdev_vport_ops = {
382 	.type		= OVS_VPORT_TYPE_VXLAN,
383 	.create		= vxlan_create,
384 	.destroy	= vxlan_destroy,
385 	.get_options	= vxlan_get_options,
386 	.send		= netdev_send,
387 	.get_egress_tun_info	= vxlan_get_egress_tun_info,
388 };
389 
390 static int vxlan_compat_init(void)
391 {
392 	return ovs_vport_ops_register(&ovs_vxlan_netdev_vport_ops);
393 }
394 
395 static void vxlan_compat_exit(void)
396 {
397 	ovs_vport_ops_unregister(&ovs_vxlan_netdev_vport_ops);
398 }
399 #else
400 static int vxlan_compat_init(void)
401 {
402 	return 0;
403 }
404 
405 static void vxlan_compat_exit(void)
406 {
407 }
408 #endif
409 
410 int __init ovs_netdev_init(void)
411 {
412 	int err;
413 
414 	err = ovs_vport_ops_register(&ovs_netdev_vport_ops);
415 	if (err)
416 		return err;
417 	err = vxlan_compat_init();
418 	if (err)
419 		vxlan_compat_exit();
420 	return err;
421 }
422 
423 void ovs_netdev_exit(void)
424 {
425 	ovs_vport_ops_unregister(&ovs_netdev_vport_ops);
426 	vxlan_compat_exit();
427 }
428