xref: /linux/drivers/net/bareudp.c (revision 4b660dbd9ee2059850fd30e0df420ca7a38a1856)
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 struct bareudp_conf {
42 	__be16 ethertype;
43 	__be16 port;
44 	u16 sport_min;
45 	bool multi_proto_mode;
46 };
47 
48 /* Pseudo network device */
49 struct bareudp_dev {
50 	struct net         *net;        /* netns for packet i/o */
51 	struct net_device  *dev;        /* netdev for bareudp tunnel */
52 	__be16		   ethertype;
53 	__be16             port;
54 	u16	           sport_min;
55 	bool               multi_proto_mode;
56 	struct socket      __rcu *sock;
57 	struct list_head   next;        /* bareudp node  on namespace list */
58 	struct gro_cells   gro_cells;
59 };
60 
61 static int bareudp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
62 {
63 	struct metadata_dst *tun_dst = NULL;
64 	struct bareudp_dev *bareudp;
65 	unsigned short family;
66 	unsigned int len;
67 	__be16 proto;
68 	void *oiph;
69 	int err;
70 
71 	bareudp = rcu_dereference_sk_user_data(sk);
72 	if (!bareudp)
73 		goto drop;
74 
75 	if (skb->protocol ==  htons(ETH_P_IP))
76 		family = AF_INET;
77 	else
78 		family = AF_INET6;
79 
80 	if (bareudp->ethertype == htons(ETH_P_IP)) {
81 		__u8 ipversion;
82 
83 		if (skb_copy_bits(skb, BAREUDP_BASE_HLEN, &ipversion,
84 				  sizeof(ipversion))) {
85 			bareudp->dev->stats.rx_dropped++;
86 			goto drop;
87 		}
88 		ipversion >>= 4;
89 
90 		if (ipversion == 4) {
91 			proto = htons(ETH_P_IP);
92 		} else if (ipversion == 6 && bareudp->multi_proto_mode) {
93 			proto = htons(ETH_P_IPV6);
94 		} else {
95 			bareudp->dev->stats.rx_dropped++;
96 			goto drop;
97 		}
98 	} else if (bareudp->ethertype == htons(ETH_P_MPLS_UC)) {
99 		struct iphdr *tunnel_hdr;
100 
101 		tunnel_hdr = (struct iphdr *)skb_network_header(skb);
102 		if (tunnel_hdr->version == 4) {
103 			if (!ipv4_is_multicast(tunnel_hdr->daddr)) {
104 				proto = bareudp->ethertype;
105 			} else if (bareudp->multi_proto_mode &&
106 				   ipv4_is_multicast(tunnel_hdr->daddr)) {
107 				proto = htons(ETH_P_MPLS_MC);
108 			} else {
109 				bareudp->dev->stats.rx_dropped++;
110 				goto drop;
111 			}
112 		} else {
113 			int addr_type;
114 			struct ipv6hdr *tunnel_hdr_v6;
115 
116 			tunnel_hdr_v6 = (struct ipv6hdr *)skb_network_header(skb);
117 			addr_type =
118 			ipv6_addr_type((struct in6_addr *)&tunnel_hdr_v6->daddr);
119 			if (!(addr_type & IPV6_ADDR_MULTICAST)) {
120 				proto = bareudp->ethertype;
121 			} else if (bareudp->multi_proto_mode &&
122 				   (addr_type & IPV6_ADDR_MULTICAST)) {
123 				proto = htons(ETH_P_MPLS_MC);
124 			} else {
125 				bareudp->dev->stats.rx_dropped++;
126 				goto drop;
127 			}
128 		}
129 	} else {
130 		proto = bareudp->ethertype;
131 	}
132 
133 	if (iptunnel_pull_header(skb, BAREUDP_BASE_HLEN,
134 				 proto,
135 				 !net_eq(bareudp->net,
136 				 dev_net(bareudp->dev)))) {
137 		bareudp->dev->stats.rx_dropped++;
138 		goto drop;
139 	}
140 	tun_dst = udp_tun_rx_dst(skb, family, TUNNEL_KEY, 0, 0);
141 	if (!tun_dst) {
142 		bareudp->dev->stats.rx_dropped++;
143 		goto drop;
144 	}
145 	skb_dst_set(skb, &tun_dst->dst);
146 	skb->dev = bareudp->dev;
147 	oiph = skb_network_header(skb);
148 	skb_reset_network_header(skb);
149 	skb_reset_mac_header(skb);
150 
151 	if (!ipv6_mod_enabled() || family == AF_INET)
152 		err = IP_ECN_decapsulate(oiph, skb);
153 	else
154 		err = IP6_ECN_decapsulate(oiph, skb);
155 
156 	if (unlikely(err)) {
157 		if (log_ecn_error) {
158 			if  (!ipv6_mod_enabled() || family == AF_INET)
159 				net_info_ratelimited("non-ECT from %pI4 "
160 						     "with TOS=%#x\n",
161 						     &((struct iphdr *)oiph)->saddr,
162 						     ((struct iphdr *)oiph)->tos);
163 			else
164 				net_info_ratelimited("non-ECT from %pI6\n",
165 						     &((struct ipv6hdr *)oiph)->saddr);
166 		}
167 		if (err > 1) {
168 			++bareudp->dev->stats.rx_frame_errors;
169 			++bareudp->dev->stats.rx_errors;
170 			goto drop;
171 		}
172 	}
173 
174 	len = skb->len;
175 	err = gro_cells_receive(&bareudp->gro_cells, skb);
176 	if (likely(err == NET_RX_SUCCESS))
177 		dev_sw_netstats_rx_add(bareudp->dev, len);
178 
179 	return 0;
180 drop:
181 	/* Consume bad packet */
182 	kfree_skb(skb);
183 
184 	return 0;
185 }
186 
187 static int bareudp_err_lookup(struct sock *sk, struct sk_buff *skb)
188 {
189 	return 0;
190 }
191 
192 static int bareudp_init(struct net_device *dev)
193 {
194 	struct bareudp_dev *bareudp = netdev_priv(dev);
195 	int err;
196 
197 	err = gro_cells_init(&bareudp->gro_cells, dev);
198 	if (err)
199 		return err;
200 
201 	return 0;
202 }
203 
204 static void bareudp_uninit(struct net_device *dev)
205 {
206 	struct bareudp_dev *bareudp = netdev_priv(dev);
207 
208 	gro_cells_destroy(&bareudp->gro_cells);
209 }
210 
211 static struct socket *bareudp_create_sock(struct net *net, __be16 port)
212 {
213 	struct udp_port_cfg udp_conf;
214 	struct socket *sock;
215 	int err;
216 
217 	memset(&udp_conf, 0, sizeof(udp_conf));
218 
219 	if (ipv6_mod_enabled())
220 		udp_conf.family = AF_INET6;
221 	else
222 		udp_conf.family = AF_INET;
223 
224 	udp_conf.local_udp_port = port;
225 	/* Open UDP socket */
226 	err = udp_sock_create(net, &udp_conf, &sock);
227 	if (err < 0)
228 		return ERR_PTR(err);
229 
230 	udp_allow_gso(sock->sk);
231 	return sock;
232 }
233 
234 /* Create new listen socket if needed */
235 static int bareudp_socket_create(struct bareudp_dev *bareudp, __be16 port)
236 {
237 	struct udp_tunnel_sock_cfg tunnel_cfg;
238 	struct socket *sock;
239 
240 	sock = bareudp_create_sock(bareudp->net, port);
241 	if (IS_ERR(sock))
242 		return PTR_ERR(sock);
243 
244 	/* Mark socket as an encapsulation socket */
245 	memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
246 	tunnel_cfg.sk_user_data = bareudp;
247 	tunnel_cfg.encap_type = 1;
248 	tunnel_cfg.encap_rcv = bareudp_udp_encap_recv;
249 	tunnel_cfg.encap_err_lookup = bareudp_err_lookup;
250 	tunnel_cfg.encap_destroy = NULL;
251 	setup_udp_tunnel_sock(bareudp->net, sock, &tunnel_cfg);
252 
253 	rcu_assign_pointer(bareudp->sock, sock);
254 	return 0;
255 }
256 
257 static int bareudp_open(struct net_device *dev)
258 {
259 	struct bareudp_dev *bareudp = netdev_priv(dev);
260 	int ret = 0;
261 
262 	ret =  bareudp_socket_create(bareudp, bareudp->port);
263 	return ret;
264 }
265 
266 static void bareudp_sock_release(struct bareudp_dev *bareudp)
267 {
268 	struct socket *sock;
269 
270 	sock = bareudp->sock;
271 	rcu_assign_pointer(bareudp->sock, NULL);
272 	synchronize_net();
273 	udp_tunnel_sock_release(sock);
274 }
275 
276 static int bareudp_stop(struct net_device *dev)
277 {
278 	struct bareudp_dev *bareudp = netdev_priv(dev);
279 
280 	bareudp_sock_release(bareudp);
281 	return 0;
282 }
283 
284 static int bareudp_xmit_skb(struct sk_buff *skb, struct net_device *dev,
285 			    struct bareudp_dev *bareudp,
286 			    const struct ip_tunnel_info *info)
287 {
288 	bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
289 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
290 	struct socket *sock = rcu_dereference(bareudp->sock);
291 	bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
292 	const struct ip_tunnel_key *key = &info->key;
293 	struct rtable *rt;
294 	__be16 sport, df;
295 	int min_headroom;
296 	__u8 tos, ttl;
297 	__be32 saddr;
298 	int err;
299 
300 	if (!sock)
301 		return -ESHUTDOWN;
302 
303 	sport = udp_flow_src_port(bareudp->net, skb,
304 				  bareudp->sport_min, USHRT_MAX,
305 				  true);
306 	rt = udp_tunnel_dst_lookup(skb, dev, bareudp->net, 0, &saddr, &info->key,
307 				   sport, bareudp->port, key->tos,
308 				   use_cache ?
309 				   (struct dst_cache *)&info->dst_cache : NULL);
310 
311 	if (IS_ERR(rt))
312 		return PTR_ERR(rt);
313 
314 	skb_tunnel_check_pmtu(skb, &rt->dst,
315 			      BAREUDP_IPV4_HLEN + info->options_len, false);
316 
317 	tos = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
318 	ttl = key->ttl;
319 	df = key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
320 	skb_scrub_packet(skb, xnet);
321 
322 	err = -ENOSPC;
323 	if (!skb_pull(skb, skb_network_offset(skb)))
324 		goto free_dst;
325 
326 	min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len +
327 		BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr);
328 
329 	err = skb_cow_head(skb, min_headroom);
330 	if (unlikely(err))
331 		goto free_dst;
332 
333 	err = udp_tunnel_handle_offloads(skb, udp_sum);
334 	if (err)
335 		goto free_dst;
336 
337 	skb_set_inner_protocol(skb, bareudp->ethertype);
338 	udp_tunnel_xmit_skb(rt, sock->sk, skb, saddr, info->key.u.ipv4.dst,
339 			    tos, ttl, df, sport, bareudp->port,
340 			    !net_eq(bareudp->net, dev_net(bareudp->dev)),
341 			    !(info->key.tun_flags & TUNNEL_CSUM));
342 	return 0;
343 
344 free_dst:
345 	dst_release(&rt->dst);
346 	return err;
347 }
348 
349 static int bareudp6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
350 			     struct bareudp_dev *bareudp,
351 			     const struct ip_tunnel_info *info)
352 {
353 	bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
354 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
355 	struct socket *sock  = rcu_dereference(bareudp->sock);
356 	bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
357 	const struct ip_tunnel_key *key = &info->key;
358 	struct dst_entry *dst = NULL;
359 	struct in6_addr saddr, daddr;
360 	int min_headroom;
361 	__u8 prio, ttl;
362 	__be16 sport;
363 	int err;
364 
365 	if (!sock)
366 		return -ESHUTDOWN;
367 
368 	sport = udp_flow_src_port(bareudp->net, skb,
369 				  bareudp->sport_min, USHRT_MAX,
370 				  true);
371 	dst = udp_tunnel6_dst_lookup(skb, dev, bareudp->net, sock, 0, &saddr,
372 				     key, sport, bareudp->port, key->tos,
373 				     use_cache ?
374 				     (struct dst_cache *) &info->dst_cache : NULL);
375 	if (IS_ERR(dst))
376 		return PTR_ERR(dst);
377 
378 	skb_tunnel_check_pmtu(skb, dst, BAREUDP_IPV6_HLEN + info->options_len,
379 			      false);
380 
381 	prio = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
382 	ttl = key->ttl;
383 
384 	skb_scrub_packet(skb, xnet);
385 
386 	err = -ENOSPC;
387 	if (!skb_pull(skb, skb_network_offset(skb)))
388 		goto free_dst;
389 
390 	min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len +
391 		BAREUDP_BASE_HLEN + info->options_len + sizeof(struct ipv6hdr);
392 
393 	err = skb_cow_head(skb, min_headroom);
394 	if (unlikely(err))
395 		goto free_dst;
396 
397 	err = udp_tunnel_handle_offloads(skb, udp_sum);
398 	if (err)
399 		goto free_dst;
400 
401 	daddr = info->key.u.ipv6.dst;
402 	udp_tunnel6_xmit_skb(dst, sock->sk, skb, dev,
403 			     &saddr, &daddr, prio, ttl,
404 			     info->key.label, sport, bareudp->port,
405 			     !(info->key.tun_flags & TUNNEL_CSUM));
406 	return 0;
407 
408 free_dst:
409 	dst_release(dst);
410 	return err;
411 }
412 
413 static bool bareudp_proto_valid(struct bareudp_dev *bareudp, __be16 proto)
414 {
415 	if (bareudp->ethertype == proto)
416 		return true;
417 
418 	if (!bareudp->multi_proto_mode)
419 		return false;
420 
421 	if (bareudp->ethertype == htons(ETH_P_MPLS_UC) &&
422 	    proto == htons(ETH_P_MPLS_MC))
423 		return true;
424 
425 	if (bareudp->ethertype == htons(ETH_P_IP) &&
426 	    proto == htons(ETH_P_IPV6))
427 		return true;
428 
429 	return false;
430 }
431 
432 static netdev_tx_t bareudp_xmit(struct sk_buff *skb, struct net_device *dev)
433 {
434 	struct bareudp_dev *bareudp = netdev_priv(dev);
435 	struct ip_tunnel_info *info = NULL;
436 	int err;
437 
438 	if (!bareudp_proto_valid(bareudp, skb->protocol)) {
439 		err = -EINVAL;
440 		goto tx_error;
441 	}
442 
443 	info = skb_tunnel_info(skb);
444 	if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) {
445 		err = -EINVAL;
446 		goto tx_error;
447 	}
448 
449 	rcu_read_lock();
450 	if (ipv6_mod_enabled() && info->mode & IP_TUNNEL_INFO_IPV6)
451 		err = bareudp6_xmit_skb(skb, dev, bareudp, info);
452 	else
453 		err = bareudp_xmit_skb(skb, dev, bareudp, info);
454 
455 	rcu_read_unlock();
456 
457 	if (likely(!err))
458 		return NETDEV_TX_OK;
459 tx_error:
460 	dev_kfree_skb(skb);
461 
462 	if (err == -ELOOP)
463 		dev->stats.collisions++;
464 	else if (err == -ENETUNREACH)
465 		dev->stats.tx_carrier_errors++;
466 
467 	dev->stats.tx_errors++;
468 	return NETDEV_TX_OK;
469 }
470 
471 static int bareudp_fill_metadata_dst(struct net_device *dev,
472 				     struct sk_buff *skb)
473 {
474 	struct ip_tunnel_info *info = skb_tunnel_info(skb);
475 	struct bareudp_dev *bareudp = netdev_priv(dev);
476 	bool use_cache;
477 	__be16 sport;
478 
479 	use_cache = ip_tunnel_dst_cache_usable(skb, info);
480 	sport = udp_flow_src_port(bareudp->net, skb,
481 				  bareudp->sport_min, USHRT_MAX,
482 				  true);
483 
484 	if (!ipv6_mod_enabled() || ip_tunnel_info_af(info) == AF_INET) {
485 		struct rtable *rt;
486 		__be32 saddr;
487 
488 		rt = udp_tunnel_dst_lookup(skb, dev, bareudp->net, 0, &saddr,
489 					   &info->key, sport, bareudp->port,
490 					   info->key.tos,
491 					   use_cache ? &info->dst_cache : NULL);
492 		if (IS_ERR(rt))
493 			return PTR_ERR(rt);
494 
495 		ip_rt_put(rt);
496 		info->key.u.ipv4.src = saddr;
497 	} else if (ip_tunnel_info_af(info) == AF_INET6) {
498 		struct dst_entry *dst;
499 		struct in6_addr saddr;
500 		struct socket *sock = rcu_dereference(bareudp->sock);
501 
502 		dst = udp_tunnel6_dst_lookup(skb, dev, bareudp->net, sock,
503 					     0, &saddr, &info->key,
504 					     sport, bareudp->port, info->key.tos,
505 					     use_cache ? &info->dst_cache : NULL);
506 		if (IS_ERR(dst))
507 			return PTR_ERR(dst);
508 
509 		dst_release(dst);
510 		info->key.u.ipv6.src = saddr;
511 	} else {
512 		return -EINVAL;
513 	}
514 
515 	info->key.tp_src = sport;
516 	info->key.tp_dst = bareudp->port;
517 	return 0;
518 }
519 
520 static const struct net_device_ops bareudp_netdev_ops = {
521 	.ndo_init               = bareudp_init,
522 	.ndo_uninit             = bareudp_uninit,
523 	.ndo_open               = bareudp_open,
524 	.ndo_stop               = bareudp_stop,
525 	.ndo_start_xmit         = bareudp_xmit,
526 	.ndo_fill_metadata_dst  = bareudp_fill_metadata_dst,
527 };
528 
529 static const struct nla_policy bareudp_policy[IFLA_BAREUDP_MAX + 1] = {
530 	[IFLA_BAREUDP_PORT]                = { .type = NLA_U16 },
531 	[IFLA_BAREUDP_ETHERTYPE]	   = { .type = NLA_U16 },
532 	[IFLA_BAREUDP_SRCPORT_MIN]         = { .type = NLA_U16 },
533 	[IFLA_BAREUDP_MULTIPROTO_MODE]     = { .type = NLA_FLAG },
534 };
535 
536 /* Info for udev, that this is a virtual tunnel endpoint */
537 static const struct device_type bareudp_type = {
538 	.name = "bareudp",
539 };
540 
541 /* Initialize the device structure. */
542 static void bareudp_setup(struct net_device *dev)
543 {
544 	dev->netdev_ops = &bareudp_netdev_ops;
545 	dev->needs_free_netdev = true;
546 	SET_NETDEV_DEVTYPE(dev, &bareudp_type);
547 	dev->features    |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
548 	dev->features    |= NETIF_F_RXCSUM;
549 	dev->features    |= NETIF_F_LLTX;
550 	dev->features    |= NETIF_F_GSO_SOFTWARE;
551 	dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
552 	dev->hw_features |= NETIF_F_RXCSUM;
553 	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
554 	dev->hard_header_len = 0;
555 	dev->addr_len = 0;
556 	dev->mtu = ETH_DATA_LEN;
557 	dev->min_mtu = IPV4_MIN_MTU;
558 	dev->max_mtu = IP_MAX_MTU - BAREUDP_BASE_HLEN;
559 	dev->type = ARPHRD_NONE;
560 	netif_keep_dst(dev);
561 	dev->priv_flags |= IFF_NO_QUEUE;
562 	dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
563 	dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
564 }
565 
566 static int bareudp_validate(struct nlattr *tb[], struct nlattr *data[],
567 			    struct netlink_ext_ack *extack)
568 {
569 	if (!data) {
570 		NL_SET_ERR_MSG(extack,
571 			       "Not enough attributes provided to perform the operation");
572 		return -EINVAL;
573 	}
574 	return 0;
575 }
576 
577 static int bareudp2info(struct nlattr *data[], struct bareudp_conf *conf,
578 			struct netlink_ext_ack *extack)
579 {
580 	memset(conf, 0, sizeof(*conf));
581 
582 	if (!data[IFLA_BAREUDP_PORT]) {
583 		NL_SET_ERR_MSG(extack, "port not specified");
584 		return -EINVAL;
585 	}
586 	if (!data[IFLA_BAREUDP_ETHERTYPE]) {
587 		NL_SET_ERR_MSG(extack, "ethertype not specified");
588 		return -EINVAL;
589 	}
590 
591 	conf->port = nla_get_u16(data[IFLA_BAREUDP_PORT]);
592 	conf->ethertype = nla_get_u16(data[IFLA_BAREUDP_ETHERTYPE]);
593 
594 	if (data[IFLA_BAREUDP_SRCPORT_MIN])
595 		conf->sport_min =  nla_get_u16(data[IFLA_BAREUDP_SRCPORT_MIN]);
596 
597 	if (data[IFLA_BAREUDP_MULTIPROTO_MODE])
598 		conf->multi_proto_mode = true;
599 
600 	return 0;
601 }
602 
603 static struct bareudp_dev *bareudp_find_dev(struct bareudp_net *bn,
604 					    const struct bareudp_conf *conf)
605 {
606 	struct bareudp_dev *bareudp, *t = NULL;
607 
608 	list_for_each_entry(bareudp, &bn->bareudp_list, next) {
609 		if (conf->port == bareudp->port)
610 			t = bareudp;
611 	}
612 	return t;
613 }
614 
615 static int bareudp_configure(struct net *net, struct net_device *dev,
616 			     struct bareudp_conf *conf,
617 			     struct netlink_ext_ack *extack)
618 {
619 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
620 	struct bareudp_dev *t, *bareudp = netdev_priv(dev);
621 	int err;
622 
623 	bareudp->net = net;
624 	bareudp->dev = dev;
625 	t = bareudp_find_dev(bn, conf);
626 	if (t) {
627 		NL_SET_ERR_MSG(extack, "Another bareudp device using the same port already exists");
628 		return -EBUSY;
629 	}
630 
631 	if (conf->multi_proto_mode &&
632 	    (conf->ethertype != htons(ETH_P_MPLS_UC) &&
633 	     conf->ethertype != htons(ETH_P_IP))) {
634 		NL_SET_ERR_MSG(extack, "Cannot set multiproto mode for this ethertype (only IPv4 and unicast MPLS are supported)");
635 		return -EINVAL;
636 	}
637 
638 	bareudp->port = conf->port;
639 	bareudp->ethertype = conf->ethertype;
640 	bareudp->sport_min = conf->sport_min;
641 	bareudp->multi_proto_mode = conf->multi_proto_mode;
642 
643 	err = register_netdevice(dev);
644 	if (err)
645 		return err;
646 
647 	list_add(&bareudp->next, &bn->bareudp_list);
648 	return 0;
649 }
650 
651 static int bareudp_link_config(struct net_device *dev,
652 			       struct nlattr *tb[])
653 {
654 	int err;
655 
656 	if (tb[IFLA_MTU]) {
657 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
658 		if (err)
659 			return err;
660 	}
661 	return 0;
662 }
663 
664 static void bareudp_dellink(struct net_device *dev, struct list_head *head)
665 {
666 	struct bareudp_dev *bareudp = netdev_priv(dev);
667 
668 	list_del(&bareudp->next);
669 	unregister_netdevice_queue(dev, head);
670 }
671 
672 static int bareudp_newlink(struct net *net, struct net_device *dev,
673 			   struct nlattr *tb[], struct nlattr *data[],
674 			   struct netlink_ext_ack *extack)
675 {
676 	struct bareudp_conf conf;
677 	int err;
678 
679 	err = bareudp2info(data, &conf, extack);
680 	if (err)
681 		return err;
682 
683 	err = bareudp_configure(net, dev, &conf, extack);
684 	if (err)
685 		return err;
686 
687 	err = bareudp_link_config(dev, tb);
688 	if (err)
689 		goto err_unconfig;
690 
691 	return 0;
692 
693 err_unconfig:
694 	bareudp_dellink(dev, NULL);
695 	return err;
696 }
697 
698 static size_t bareudp_get_size(const struct net_device *dev)
699 {
700 	return  nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_PORT */
701 		nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_ETHERTYPE */
702 		nla_total_size(sizeof(__u16))  +  /* IFLA_BAREUDP_SRCPORT_MIN */
703 		nla_total_size(0)              +  /* IFLA_BAREUDP_MULTIPROTO_MODE */
704 		0;
705 }
706 
707 static int bareudp_fill_info(struct sk_buff *skb, const struct net_device *dev)
708 {
709 	struct bareudp_dev *bareudp = netdev_priv(dev);
710 
711 	if (nla_put_be16(skb, IFLA_BAREUDP_PORT, bareudp->port))
712 		goto nla_put_failure;
713 	if (nla_put_be16(skb, IFLA_BAREUDP_ETHERTYPE, bareudp->ethertype))
714 		goto nla_put_failure;
715 	if (nla_put_u16(skb, IFLA_BAREUDP_SRCPORT_MIN, bareudp->sport_min))
716 		goto nla_put_failure;
717 	if (bareudp->multi_proto_mode &&
718 	    nla_put_flag(skb, IFLA_BAREUDP_MULTIPROTO_MODE))
719 		goto nla_put_failure;
720 
721 	return 0;
722 
723 nla_put_failure:
724 	return -EMSGSIZE;
725 }
726 
727 static struct rtnl_link_ops bareudp_link_ops __read_mostly = {
728 	.kind           = "bareudp",
729 	.maxtype        = IFLA_BAREUDP_MAX,
730 	.policy         = bareudp_policy,
731 	.priv_size      = sizeof(struct bareudp_dev),
732 	.setup          = bareudp_setup,
733 	.validate       = bareudp_validate,
734 	.newlink        = bareudp_newlink,
735 	.dellink        = bareudp_dellink,
736 	.get_size       = bareudp_get_size,
737 	.fill_info      = bareudp_fill_info,
738 };
739 
740 static __net_init int bareudp_init_net(struct net *net)
741 {
742 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
743 
744 	INIT_LIST_HEAD(&bn->bareudp_list);
745 	return 0;
746 }
747 
748 static void bareudp_destroy_tunnels(struct net *net, struct list_head *head)
749 {
750 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
751 	struct bareudp_dev *bareudp, *next;
752 
753 	list_for_each_entry_safe(bareudp, next, &bn->bareudp_list, next)
754 		unregister_netdevice_queue(bareudp->dev, head);
755 }
756 
757 static void __net_exit bareudp_exit_batch_rtnl(struct list_head *net_list,
758 					       struct list_head *dev_kill_list)
759 {
760 	struct net *net;
761 
762 	list_for_each_entry(net, net_list, exit_list)
763 		bareudp_destroy_tunnels(net, dev_kill_list);
764 }
765 
766 static struct pernet_operations bareudp_net_ops = {
767 	.init = bareudp_init_net,
768 	.exit_batch_rtnl = bareudp_exit_batch_rtnl,
769 	.id   = &bareudp_net_id,
770 	.size = sizeof(struct bareudp_net),
771 };
772 
773 static int __init bareudp_init_module(void)
774 {
775 	int rc;
776 
777 	rc = register_pernet_subsys(&bareudp_net_ops);
778 	if (rc)
779 		goto out1;
780 
781 	rc = rtnl_link_register(&bareudp_link_ops);
782 	if (rc)
783 		goto out2;
784 
785 	return 0;
786 out2:
787 	unregister_pernet_subsys(&bareudp_net_ops);
788 out1:
789 	return rc;
790 }
791 late_initcall(bareudp_init_module);
792 
793 static void __exit bareudp_cleanup_module(void)
794 {
795 	rtnl_link_unregister(&bareudp_link_ops);
796 	unregister_pernet_subsys(&bareudp_net_ops);
797 }
798 module_exit(bareudp_cleanup_module);
799 
800 MODULE_ALIAS_RTNL_LINK("bareudp");
801 MODULE_LICENSE("GPL");
802 MODULE_AUTHOR("Martin Varghese <martin.varghese@nokia.com>");
803 MODULE_DESCRIPTION("Interface driver for UDP encapsulated traffic");
804