xref: /openbmc/linux/drivers/net/bareudp.c (revision 27ab1c1c)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Bareudp: UDP  tunnel encasulation for different Payload types like
3  * MPLS, NSH, IP, etc.
4  * Copyright (c) 2019 Nokia, Inc.
5  * Authors:  Martin Varghese, <martin.varghese@nokia.com>
6  */
7 
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9 
10 #include <linux/kernel.h>
11 #include <linux/module.h>
12 #include <linux/etherdevice.h>
13 #include <linux/hash.h>
14 #include <net/dst_metadata.h>
15 #include <net/gro_cells.h>
16 #include <net/rtnetlink.h>
17 #include <net/protocol.h>
18 #include <net/ip6_tunnel.h>
19 #include <net/ip_tunnels.h>
20 #include <net/udp_tunnel.h>
21 #include <net/bareudp.h>
22 
23 #define BAREUDP_BASE_HLEN sizeof(struct udphdr)
24 #define BAREUDP_IPV4_HLEN (sizeof(struct iphdr) + \
25 			   sizeof(struct udphdr))
26 #define BAREUDP_IPV6_HLEN (sizeof(struct ipv6hdr) + \
27 			   sizeof(struct udphdr))
28 
29 static bool log_ecn_error = true;
30 module_param(log_ecn_error, bool, 0644);
31 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
32 
33 /* per-network namespace private data for this module */
34 
35 static unsigned int bareudp_net_id;
36 
37 struct bareudp_net {
38 	struct list_head        bareudp_list;
39 };
40 
41 /* Pseudo network device */
42 struct bareudp_dev {
43 	struct net         *net;        /* netns for packet i/o */
44 	struct net_device  *dev;        /* netdev for bareudp tunnel */
45 	__be16		   ethertype;
46 	__be16             port;
47 	u16	           sport_min;
48 	bool               multi_proto_mode;
49 	struct socket      __rcu *sock;
50 	struct list_head   next;        /* bareudp node  on namespace list */
51 	struct gro_cells   gro_cells;
52 };
53 
54 static int bareudp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
55 {
56 	struct metadata_dst *tun_dst = NULL;
57 	struct bareudp_dev *bareudp;
58 	unsigned short family;
59 	unsigned int len;
60 	__be16 proto;
61 	void *oiph;
62 	int err;
63 
64 	bareudp = rcu_dereference_sk_user_data(sk);
65 	if (!bareudp)
66 		goto drop;
67 
68 	if (skb->protocol ==  htons(ETH_P_IP))
69 		family = AF_INET;
70 	else
71 		family = AF_INET6;
72 
73 	if (bareudp->ethertype == htons(ETH_P_IP)) {
74 		struct iphdr *iphdr;
75 
76 		iphdr = (struct iphdr *)(skb->data + BAREUDP_BASE_HLEN);
77 		if (iphdr->version == 4) {
78 			proto = bareudp->ethertype;
79 		} else if (bareudp->multi_proto_mode && (iphdr->version == 6)) {
80 			proto = htons(ETH_P_IPV6);
81 		} else {
82 			bareudp->dev->stats.rx_dropped++;
83 			goto drop;
84 		}
85 	} else if (bareudp->ethertype == htons(ETH_P_MPLS_UC)) {
86 		struct iphdr *tunnel_hdr;
87 
88 		tunnel_hdr = (struct iphdr *)skb_network_header(skb);
89 		if (tunnel_hdr->version == 4) {
90 			if (!ipv4_is_multicast(tunnel_hdr->daddr)) {
91 				proto = bareudp->ethertype;
92 			} else if (bareudp->multi_proto_mode &&
93 				   ipv4_is_multicast(tunnel_hdr->daddr)) {
94 				proto = htons(ETH_P_MPLS_MC);
95 			} else {
96 				bareudp->dev->stats.rx_dropped++;
97 				goto drop;
98 			}
99 		} else {
100 			int addr_type;
101 			struct ipv6hdr *tunnel_hdr_v6;
102 
103 			tunnel_hdr_v6 = (struct ipv6hdr *)skb_network_header(skb);
104 			addr_type =
105 			ipv6_addr_type((struct in6_addr *)&tunnel_hdr_v6->daddr);
106 			if (!(addr_type & IPV6_ADDR_MULTICAST)) {
107 				proto = bareudp->ethertype;
108 			} else if (bareudp->multi_proto_mode &&
109 				   (addr_type & IPV6_ADDR_MULTICAST)) {
110 				proto = htons(ETH_P_MPLS_MC);
111 			} else {
112 				bareudp->dev->stats.rx_dropped++;
113 				goto drop;
114 			}
115 		}
116 	} else {
117 		proto = bareudp->ethertype;
118 	}
119 
120 	if (iptunnel_pull_header(skb, BAREUDP_BASE_HLEN,
121 				 proto,
122 				 !net_eq(bareudp->net,
123 				 dev_net(bareudp->dev)))) {
124 		bareudp->dev->stats.rx_dropped++;
125 		goto drop;
126 	}
127 	tun_dst = udp_tun_rx_dst(skb, family, TUNNEL_KEY, 0, 0);
128 	if (!tun_dst) {
129 		bareudp->dev->stats.rx_dropped++;
130 		goto drop;
131 	}
132 	skb_dst_set(skb, &tun_dst->dst);
133 	skb->dev = bareudp->dev;
134 	oiph = skb_network_header(skb);
135 	skb_reset_network_header(skb);
136 
137 	if (!IS_ENABLED(CONFIG_IPV6) || family == AF_INET)
138 		err = IP_ECN_decapsulate(oiph, skb);
139 	else
140 		err = IP6_ECN_decapsulate(oiph, skb);
141 
142 	if (unlikely(err)) {
143 		if (log_ecn_error) {
144 			if  (!IS_ENABLED(CONFIG_IPV6) || family == AF_INET)
145 				net_info_ratelimited("non-ECT from %pI4 "
146 						     "with TOS=%#x\n",
147 						     &((struct iphdr *)oiph)->saddr,
148 						     ((struct iphdr *)oiph)->tos);
149 			else
150 				net_info_ratelimited("non-ECT from %pI6\n",
151 						     &((struct ipv6hdr *)oiph)->saddr);
152 		}
153 		if (err > 1) {
154 			++bareudp->dev->stats.rx_frame_errors;
155 			++bareudp->dev->stats.rx_errors;
156 			goto drop;
157 		}
158 	}
159 
160 	len = skb->len;
161 	err = gro_cells_receive(&bareudp->gro_cells, skb);
162 	if (likely(err == NET_RX_SUCCESS))
163 		dev_sw_netstats_rx_add(bareudp->dev, len);
164 
165 	return 0;
166 drop:
167 	/* Consume bad packet */
168 	kfree_skb(skb);
169 
170 	return 0;
171 }
172 
173 static int bareudp_err_lookup(struct sock *sk, struct sk_buff *skb)
174 {
175 	return 0;
176 }
177 
178 static int bareudp_init(struct net_device *dev)
179 {
180 	struct bareudp_dev *bareudp = netdev_priv(dev);
181 	int err;
182 
183 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
184 	if (!dev->tstats)
185 		return -ENOMEM;
186 
187 	err = gro_cells_init(&bareudp->gro_cells, dev);
188 	if (err) {
189 		free_percpu(dev->tstats);
190 		return err;
191 	}
192 	return 0;
193 }
194 
195 static void bareudp_uninit(struct net_device *dev)
196 {
197 	struct bareudp_dev *bareudp = netdev_priv(dev);
198 
199 	gro_cells_destroy(&bareudp->gro_cells);
200 	free_percpu(dev->tstats);
201 }
202 
203 static struct socket *bareudp_create_sock(struct net *net, __be16 port)
204 {
205 	struct udp_port_cfg udp_conf;
206 	struct socket *sock;
207 	int err;
208 
209 	memset(&udp_conf, 0, sizeof(udp_conf));
210 #if IS_ENABLED(CONFIG_IPV6)
211 	udp_conf.family = AF_INET6;
212 #else
213 	udp_conf.family = AF_INET;
214 #endif
215 	udp_conf.local_udp_port = port;
216 	/* Open UDP socket */
217 	err = udp_sock_create(net, &udp_conf, &sock);
218 	if (err < 0)
219 		return ERR_PTR(err);
220 
221 	return sock;
222 }
223 
224 /* Create new listen socket if needed */
225 static int bareudp_socket_create(struct bareudp_dev *bareudp, __be16 port)
226 {
227 	struct udp_tunnel_sock_cfg tunnel_cfg;
228 	struct socket *sock;
229 
230 	sock = bareudp_create_sock(bareudp->net, port);
231 	if (IS_ERR(sock))
232 		return PTR_ERR(sock);
233 
234 	/* Mark socket as an encapsulation socket */
235 	memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
236 	tunnel_cfg.sk_user_data = bareudp;
237 	tunnel_cfg.encap_type = 1;
238 	tunnel_cfg.encap_rcv = bareudp_udp_encap_recv;
239 	tunnel_cfg.encap_err_lookup = bareudp_err_lookup;
240 	tunnel_cfg.encap_destroy = NULL;
241 	setup_udp_tunnel_sock(bareudp->net, sock, &tunnel_cfg);
242 
243 	/* As the setup_udp_tunnel_sock does not call udp_encap_enable if the
244 	 * socket type is v6 an explicit call to udp_encap_enable is needed.
245 	 */
246 	if (sock->sk->sk_family == AF_INET6)
247 		udp_encap_enable();
248 
249 	rcu_assign_pointer(bareudp->sock, sock);
250 	return 0;
251 }
252 
253 static int bareudp_open(struct net_device *dev)
254 {
255 	struct bareudp_dev *bareudp = netdev_priv(dev);
256 	int ret = 0;
257 
258 	ret =  bareudp_socket_create(bareudp, bareudp->port);
259 	return ret;
260 }
261 
262 static void bareudp_sock_release(struct bareudp_dev *bareudp)
263 {
264 	struct socket *sock;
265 
266 	sock = bareudp->sock;
267 	rcu_assign_pointer(bareudp->sock, NULL);
268 	synchronize_net();
269 	udp_tunnel_sock_release(sock);
270 }
271 
272 static int bareudp_stop(struct net_device *dev)
273 {
274 	struct bareudp_dev *bareudp = netdev_priv(dev);
275 
276 	bareudp_sock_release(bareudp);
277 	return 0;
278 }
279 
280 static int bareudp_xmit_skb(struct sk_buff *skb, struct net_device *dev,
281 			    struct bareudp_dev *bareudp,
282 			    const struct ip_tunnel_info *info)
283 {
284 	bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
285 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
286 	struct socket *sock = rcu_dereference(bareudp->sock);
287 	bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
288 	const struct ip_tunnel_key *key = &info->key;
289 	struct rtable *rt;
290 	__be16 sport, df;
291 	int min_headroom;
292 	__u8 tos, ttl;
293 	__be32 saddr;
294 	int err;
295 
296 	if (!sock)
297 		return -ESHUTDOWN;
298 
299 	rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr, info,
300 				    IPPROTO_UDP, use_cache);
301 
302 	if (IS_ERR(rt))
303 		return PTR_ERR(rt);
304 
305 	skb_tunnel_check_pmtu(skb, &rt->dst,
306 			      BAREUDP_IPV4_HLEN + info->options_len, false);
307 
308 	sport = udp_flow_src_port(bareudp->net, skb,
309 				  bareudp->sport_min, USHRT_MAX,
310 				  true);
311 	tos = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
312 	ttl = key->ttl;
313 	df = key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
314 	skb_scrub_packet(skb, xnet);
315 
316 	err = -ENOSPC;
317 	if (!skb_pull(skb, skb_network_offset(skb)))
318 		goto free_dst;
319 
320 	min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len +
321 		BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr);
322 
323 	err = skb_cow_head(skb, min_headroom);
324 	if (unlikely(err))
325 		goto free_dst;
326 
327 	err = udp_tunnel_handle_offloads(skb, udp_sum);
328 	if (err)
329 		goto free_dst;
330 
331 	skb_set_inner_protocol(skb, bareudp->ethertype);
332 	udp_tunnel_xmit_skb(rt, sock->sk, skb, saddr, info->key.u.ipv4.dst,
333 			    tos, ttl, df, sport, bareudp->port,
334 			    !net_eq(bareudp->net, dev_net(bareudp->dev)),
335 			    !(info->key.tun_flags & TUNNEL_CSUM));
336 	return 0;
337 
338 free_dst:
339 	dst_release(&rt->dst);
340 	return err;
341 }
342 
343 static int bareudp6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
344 			     struct bareudp_dev *bareudp,
345 			     const struct ip_tunnel_info *info)
346 {
347 	bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
348 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
349 	struct socket *sock  = rcu_dereference(bareudp->sock);
350 	bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
351 	const struct ip_tunnel_key *key = &info->key;
352 	struct dst_entry *dst = NULL;
353 	struct in6_addr saddr, daddr;
354 	int min_headroom;
355 	__u8 prio, ttl;
356 	__be16 sport;
357 	int err;
358 
359 	if (!sock)
360 		return -ESHUTDOWN;
361 
362 	dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock, &saddr, info,
363 				    IPPROTO_UDP, use_cache);
364 	if (IS_ERR(dst))
365 		return PTR_ERR(dst);
366 
367 	skb_tunnel_check_pmtu(skb, dst, BAREUDP_IPV6_HLEN + info->options_len,
368 			      false);
369 
370 	sport = udp_flow_src_port(bareudp->net, skb,
371 				  bareudp->sport_min, USHRT_MAX,
372 				  true);
373 	prio = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
374 	ttl = key->ttl;
375 
376 	skb_scrub_packet(skb, xnet);
377 
378 	err = -ENOSPC;
379 	if (!skb_pull(skb, skb_network_offset(skb)))
380 		goto free_dst;
381 
382 	min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len +
383 		BAREUDP_BASE_HLEN + info->options_len + sizeof(struct ipv6hdr);
384 
385 	err = skb_cow_head(skb, min_headroom);
386 	if (unlikely(err))
387 		goto free_dst;
388 
389 	err = udp_tunnel_handle_offloads(skb, udp_sum);
390 	if (err)
391 		goto free_dst;
392 
393 	daddr = info->key.u.ipv6.dst;
394 	udp_tunnel6_xmit_skb(dst, sock->sk, skb, dev,
395 			     &saddr, &daddr, prio, ttl,
396 			     info->key.label, sport, bareudp->port,
397 			     !(info->key.tun_flags & TUNNEL_CSUM));
398 	return 0;
399 
400 free_dst:
401 	dst_release(dst);
402 	return err;
403 }
404 
405 static bool bareudp_proto_valid(struct bareudp_dev *bareudp, __be16 proto)
406 {
407 	if (bareudp->ethertype == proto)
408 		return true;
409 
410 	if (!bareudp->multi_proto_mode)
411 		return false;
412 
413 	if (bareudp->ethertype == htons(ETH_P_MPLS_UC) &&
414 	    proto == htons(ETH_P_MPLS_MC))
415 		return true;
416 
417 	if (bareudp->ethertype == htons(ETH_P_IP) &&
418 	    proto == htons(ETH_P_IPV6))
419 		return true;
420 
421 	return false;
422 }
423 
424 static netdev_tx_t bareudp_xmit(struct sk_buff *skb, struct net_device *dev)
425 {
426 	struct bareudp_dev *bareudp = netdev_priv(dev);
427 	struct ip_tunnel_info *info = NULL;
428 	int err;
429 
430 	if (!bareudp_proto_valid(bareudp, skb->protocol)) {
431 		err = -EINVAL;
432 		goto tx_error;
433 	}
434 
435 	info = skb_tunnel_info(skb);
436 	if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) {
437 		err = -EINVAL;
438 		goto tx_error;
439 	}
440 
441 	rcu_read_lock();
442 	if (IS_ENABLED(CONFIG_IPV6) && info->mode & IP_TUNNEL_INFO_IPV6)
443 		err = bareudp6_xmit_skb(skb, dev, bareudp, info);
444 	else
445 		err = bareudp_xmit_skb(skb, dev, bareudp, info);
446 
447 	rcu_read_unlock();
448 
449 	if (likely(!err))
450 		return NETDEV_TX_OK;
451 tx_error:
452 	dev_kfree_skb(skb);
453 
454 	if (err == -ELOOP)
455 		dev->stats.collisions++;
456 	else if (err == -ENETUNREACH)
457 		dev->stats.tx_carrier_errors++;
458 
459 	dev->stats.tx_errors++;
460 	return NETDEV_TX_OK;
461 }
462 
463 static int bareudp_fill_metadata_dst(struct net_device *dev,
464 				     struct sk_buff *skb)
465 {
466 	struct ip_tunnel_info *info = skb_tunnel_info(skb);
467 	struct bareudp_dev *bareudp = netdev_priv(dev);
468 	bool use_cache;
469 
470 	use_cache = ip_tunnel_dst_cache_usable(skb, info);
471 
472 	if (!IS_ENABLED(CONFIG_IPV6) || ip_tunnel_info_af(info) == AF_INET) {
473 		struct rtable *rt;
474 		__be32 saddr;
475 
476 		rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr,
477 					    info, IPPROTO_UDP, use_cache);
478 		if (IS_ERR(rt))
479 			return PTR_ERR(rt);
480 
481 		ip_rt_put(rt);
482 		info->key.u.ipv4.src = saddr;
483 	} else if (ip_tunnel_info_af(info) == AF_INET6) {
484 		struct dst_entry *dst;
485 		struct in6_addr saddr;
486 		struct socket *sock = rcu_dereference(bareudp->sock);
487 
488 		dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock,
489 					    &saddr, info, IPPROTO_UDP,
490 					    use_cache);
491 		if (IS_ERR(dst))
492 			return PTR_ERR(dst);
493 
494 		dst_release(dst);
495 		info->key.u.ipv6.src = saddr;
496 	} else {
497 		return -EINVAL;
498 	}
499 
500 	info->key.tp_src = udp_flow_src_port(bareudp->net, skb,
501 					     bareudp->sport_min,
502 			USHRT_MAX, true);
503 	info->key.tp_dst = bareudp->port;
504 	return 0;
505 }
506 
507 static const struct net_device_ops bareudp_netdev_ops = {
508 	.ndo_init               = bareudp_init,
509 	.ndo_uninit             = bareudp_uninit,
510 	.ndo_open               = bareudp_open,
511 	.ndo_stop               = bareudp_stop,
512 	.ndo_start_xmit         = bareudp_xmit,
513 	.ndo_get_stats64        = dev_get_tstats64,
514 	.ndo_fill_metadata_dst  = bareudp_fill_metadata_dst,
515 };
516 
517 static const struct nla_policy bareudp_policy[IFLA_BAREUDP_MAX + 1] = {
518 	[IFLA_BAREUDP_PORT]                = { .type = NLA_U16 },
519 	[IFLA_BAREUDP_ETHERTYPE]	   = { .type = NLA_U16 },
520 	[IFLA_BAREUDP_SRCPORT_MIN]         = { .type = NLA_U16 },
521 	[IFLA_BAREUDP_MULTIPROTO_MODE]     = { .type = NLA_FLAG },
522 };
523 
524 /* Info for udev, that this is a virtual tunnel endpoint */
525 static const struct device_type bareudp_type = {
526 	.name = "bareudp",
527 };
528 
529 /* Initialize the device structure. */
530 static void bareudp_setup(struct net_device *dev)
531 {
532 	dev->netdev_ops = &bareudp_netdev_ops;
533 	dev->needs_free_netdev = true;
534 	SET_NETDEV_DEVTYPE(dev, &bareudp_type);
535 	dev->features    |= NETIF_F_SG | NETIF_F_HW_CSUM;
536 	dev->features    |= NETIF_F_RXCSUM;
537 	dev->features    |= NETIF_F_LLTX;
538 	dev->features    |= NETIF_F_GSO_SOFTWARE;
539 	dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
540 	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
541 	dev->hard_header_len = 0;
542 	dev->addr_len = 0;
543 	dev->mtu = ETH_DATA_LEN;
544 	dev->min_mtu = IPV4_MIN_MTU;
545 	dev->max_mtu = IP_MAX_MTU - BAREUDP_BASE_HLEN;
546 	dev->type = ARPHRD_NONE;
547 	netif_keep_dst(dev);
548 	dev->priv_flags |= IFF_NO_QUEUE;
549 	dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
550 }
551 
552 static int bareudp_validate(struct nlattr *tb[], struct nlattr *data[],
553 			    struct netlink_ext_ack *extack)
554 {
555 	if (!data) {
556 		NL_SET_ERR_MSG(extack,
557 			       "Not enough attributes provided to perform the operation");
558 		return -EINVAL;
559 	}
560 	return 0;
561 }
562 
563 static int bareudp2info(struct nlattr *data[], struct bareudp_conf *conf,
564 			struct netlink_ext_ack *extack)
565 {
566 	memset(conf, 0, sizeof(*conf));
567 
568 	if (!data[IFLA_BAREUDP_PORT]) {
569 		NL_SET_ERR_MSG(extack, "port not specified");
570 		return -EINVAL;
571 	}
572 	if (!data[IFLA_BAREUDP_ETHERTYPE]) {
573 		NL_SET_ERR_MSG(extack, "ethertype not specified");
574 		return -EINVAL;
575 	}
576 
577 	if (data[IFLA_BAREUDP_PORT])
578 		conf->port =  nla_get_u16(data[IFLA_BAREUDP_PORT]);
579 
580 	if (data[IFLA_BAREUDP_ETHERTYPE])
581 		conf->ethertype =  nla_get_u16(data[IFLA_BAREUDP_ETHERTYPE]);
582 
583 	if (data[IFLA_BAREUDP_SRCPORT_MIN])
584 		conf->sport_min =  nla_get_u16(data[IFLA_BAREUDP_SRCPORT_MIN]);
585 
586 	if (data[IFLA_BAREUDP_MULTIPROTO_MODE])
587 		conf->multi_proto_mode = true;
588 
589 	return 0;
590 }
591 
592 static struct bareudp_dev *bareudp_find_dev(struct bareudp_net *bn,
593 					    const struct bareudp_conf *conf)
594 {
595 	struct bareudp_dev *bareudp, *t = NULL;
596 
597 	list_for_each_entry(bareudp, &bn->bareudp_list, next) {
598 		if (conf->port == bareudp->port)
599 			t = bareudp;
600 	}
601 	return t;
602 }
603 
604 static int bareudp_configure(struct net *net, struct net_device *dev,
605 			     struct bareudp_conf *conf)
606 {
607 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
608 	struct bareudp_dev *t, *bareudp = netdev_priv(dev);
609 	int err;
610 
611 	bareudp->net = net;
612 	bareudp->dev = dev;
613 	t = bareudp_find_dev(bn, conf);
614 	if (t)
615 		return -EBUSY;
616 
617 	if (conf->multi_proto_mode &&
618 	    (conf->ethertype != htons(ETH_P_MPLS_UC) &&
619 	     conf->ethertype != htons(ETH_P_IP)))
620 		return -EINVAL;
621 
622 	bareudp->port = conf->port;
623 	bareudp->ethertype = conf->ethertype;
624 	bareudp->sport_min = conf->sport_min;
625 	bareudp->multi_proto_mode = conf->multi_proto_mode;
626 
627 	err = register_netdevice(dev);
628 	if (err)
629 		return err;
630 
631 	list_add(&bareudp->next, &bn->bareudp_list);
632 	return 0;
633 }
634 
635 static int bareudp_link_config(struct net_device *dev,
636 			       struct nlattr *tb[])
637 {
638 	int err;
639 
640 	if (tb[IFLA_MTU]) {
641 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
642 		if (err)
643 			return err;
644 	}
645 	return 0;
646 }
647 
648 static void bareudp_dellink(struct net_device *dev, struct list_head *head)
649 {
650 	struct bareudp_dev *bareudp = netdev_priv(dev);
651 
652 	list_del(&bareudp->next);
653 	unregister_netdevice_queue(dev, head);
654 }
655 
656 static int bareudp_newlink(struct net *net, struct net_device *dev,
657 			   struct nlattr *tb[], struct nlattr *data[],
658 			   struct netlink_ext_ack *extack)
659 {
660 	struct bareudp_conf conf;
661 	LIST_HEAD(list_kill);
662 	int err;
663 
664 	err = bareudp2info(data, &conf, extack);
665 	if (err)
666 		return err;
667 
668 	err = bareudp_configure(net, dev, &conf);
669 	if (err)
670 		return err;
671 
672 	err = bareudp_link_config(dev, tb);
673 	if (err)
674 		goto err_unconfig;
675 
676 	return 0;
677 
678 err_unconfig:
679 	bareudp_dellink(dev, &list_kill);
680 	unregister_netdevice_many(&list_kill);
681 	return err;
682 }
683 
684 static size_t bareudp_get_size(const struct net_device *dev)
685 {
686 	return  nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_PORT */
687 		nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_ETHERTYPE */
688 		nla_total_size(sizeof(__u16))  +  /* IFLA_BAREUDP_SRCPORT_MIN */
689 		nla_total_size(0)              +  /* IFLA_BAREUDP_MULTIPROTO_MODE */
690 		0;
691 }
692 
693 static int bareudp_fill_info(struct sk_buff *skb, const struct net_device *dev)
694 {
695 	struct bareudp_dev *bareudp = netdev_priv(dev);
696 
697 	if (nla_put_be16(skb, IFLA_BAREUDP_PORT, bareudp->port))
698 		goto nla_put_failure;
699 	if (nla_put_be16(skb, IFLA_BAREUDP_ETHERTYPE, bareudp->ethertype))
700 		goto nla_put_failure;
701 	if (nla_put_u16(skb, IFLA_BAREUDP_SRCPORT_MIN, bareudp->sport_min))
702 		goto nla_put_failure;
703 	if (bareudp->multi_proto_mode &&
704 	    nla_put_flag(skb, IFLA_BAREUDP_MULTIPROTO_MODE))
705 		goto nla_put_failure;
706 
707 	return 0;
708 
709 nla_put_failure:
710 	return -EMSGSIZE;
711 }
712 
713 static struct rtnl_link_ops bareudp_link_ops __read_mostly = {
714 	.kind           = "bareudp",
715 	.maxtype        = IFLA_BAREUDP_MAX,
716 	.policy         = bareudp_policy,
717 	.priv_size      = sizeof(struct bareudp_dev),
718 	.setup          = bareudp_setup,
719 	.validate       = bareudp_validate,
720 	.newlink        = bareudp_newlink,
721 	.dellink        = bareudp_dellink,
722 	.get_size       = bareudp_get_size,
723 	.fill_info      = bareudp_fill_info,
724 };
725 
726 struct net_device *bareudp_dev_create(struct net *net, const char *name,
727 				      u8 name_assign_type,
728 				      struct bareudp_conf *conf)
729 {
730 	struct nlattr *tb[IFLA_MAX + 1];
731 	struct net_device *dev;
732 	LIST_HEAD(list_kill);
733 	int err;
734 
735 	memset(tb, 0, sizeof(tb));
736 	dev = rtnl_create_link(net, name, name_assign_type,
737 			       &bareudp_link_ops, tb, NULL);
738 	if (IS_ERR(dev))
739 		return dev;
740 
741 	err = bareudp_configure(net, dev, conf);
742 	if (err) {
743 		free_netdev(dev);
744 		return ERR_PTR(err);
745 	}
746 	err = dev_set_mtu(dev, IP_MAX_MTU - BAREUDP_BASE_HLEN);
747 	if (err)
748 		goto err;
749 
750 	err = rtnl_configure_link(dev, NULL);
751 	if (err < 0)
752 		goto err;
753 
754 	return dev;
755 err:
756 	bareudp_dellink(dev, &list_kill);
757 	unregister_netdevice_many(&list_kill);
758 	return ERR_PTR(err);
759 }
760 EXPORT_SYMBOL_GPL(bareudp_dev_create);
761 
762 static __net_init int bareudp_init_net(struct net *net)
763 {
764 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
765 
766 	INIT_LIST_HEAD(&bn->bareudp_list);
767 	return 0;
768 }
769 
770 static void bareudp_destroy_tunnels(struct net *net, struct list_head *head)
771 {
772 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
773 	struct bareudp_dev *bareudp, *next;
774 
775 	list_for_each_entry_safe(bareudp, next, &bn->bareudp_list, next)
776 		unregister_netdevice_queue(bareudp->dev, head);
777 }
778 
779 static void __net_exit bareudp_exit_batch_net(struct list_head *net_list)
780 {
781 	struct net *net;
782 	LIST_HEAD(list);
783 
784 	rtnl_lock();
785 	list_for_each_entry(net, net_list, exit_list)
786 		bareudp_destroy_tunnels(net, &list);
787 
788 	/* unregister the devices gathered above */
789 	unregister_netdevice_many(&list);
790 	rtnl_unlock();
791 }
792 
793 static struct pernet_operations bareudp_net_ops = {
794 	.init = bareudp_init_net,
795 	.exit_batch = bareudp_exit_batch_net,
796 	.id   = &bareudp_net_id,
797 	.size = sizeof(struct bareudp_net),
798 };
799 
800 static int __init bareudp_init_module(void)
801 {
802 	int rc;
803 
804 	rc = register_pernet_subsys(&bareudp_net_ops);
805 	if (rc)
806 		goto out1;
807 
808 	rc = rtnl_link_register(&bareudp_link_ops);
809 	if (rc)
810 		goto out2;
811 
812 	return 0;
813 out2:
814 	unregister_pernet_subsys(&bareudp_net_ops);
815 out1:
816 	return rc;
817 }
818 late_initcall(bareudp_init_module);
819 
820 static void __exit bareudp_cleanup_module(void)
821 {
822 	rtnl_link_unregister(&bareudp_link_ops);
823 	unregister_pernet_subsys(&bareudp_net_ops);
824 }
825 module_exit(bareudp_cleanup_module);
826 
827 MODULE_ALIAS_RTNL_LINK("bareudp");
828 MODULE_LICENSE("GPL");
829 MODULE_AUTHOR("Martin Varghese <martin.varghese@nokia.com>");
830 MODULE_DESCRIPTION("Interface driver for UDP encapsulated traffic");
831