xref: /openbmc/linux/drivers/net/bareudp.c (revision aa0dc6a7)
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 	skb_reset_mac_header(skb);
137 
138 	if (!IS_ENABLED(CONFIG_IPV6) || family == AF_INET)
139 		err = IP_ECN_decapsulate(oiph, skb);
140 	else
141 		err = IP6_ECN_decapsulate(oiph, skb);
142 
143 	if (unlikely(err)) {
144 		if (log_ecn_error) {
145 			if  (!IS_ENABLED(CONFIG_IPV6) || family == AF_INET)
146 				net_info_ratelimited("non-ECT from %pI4 "
147 						     "with TOS=%#x\n",
148 						     &((struct iphdr *)oiph)->saddr,
149 						     ((struct iphdr *)oiph)->tos);
150 			else
151 				net_info_ratelimited("non-ECT from %pI6\n",
152 						     &((struct ipv6hdr *)oiph)->saddr);
153 		}
154 		if (err > 1) {
155 			++bareudp->dev->stats.rx_frame_errors;
156 			++bareudp->dev->stats.rx_errors;
157 			goto drop;
158 		}
159 	}
160 
161 	len = skb->len;
162 	err = gro_cells_receive(&bareudp->gro_cells, skb);
163 	if (likely(err == NET_RX_SUCCESS))
164 		dev_sw_netstats_rx_add(bareudp->dev, len);
165 
166 	return 0;
167 drop:
168 	/* Consume bad packet */
169 	kfree_skb(skb);
170 
171 	return 0;
172 }
173 
174 static int bareudp_err_lookup(struct sock *sk, struct sk_buff *skb)
175 {
176 	return 0;
177 }
178 
179 static int bareudp_init(struct net_device *dev)
180 {
181 	struct bareudp_dev *bareudp = netdev_priv(dev);
182 	int err;
183 
184 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
185 	if (!dev->tstats)
186 		return -ENOMEM;
187 
188 	err = gro_cells_init(&bareudp->gro_cells, dev);
189 	if (err) {
190 		free_percpu(dev->tstats);
191 		return err;
192 	}
193 	return 0;
194 }
195 
196 static void bareudp_uninit(struct net_device *dev)
197 {
198 	struct bareudp_dev *bareudp = netdev_priv(dev);
199 
200 	gro_cells_destroy(&bareudp->gro_cells);
201 	free_percpu(dev->tstats);
202 }
203 
204 static struct socket *bareudp_create_sock(struct net *net, __be16 port)
205 {
206 	struct udp_port_cfg udp_conf;
207 	struct socket *sock;
208 	int err;
209 
210 	memset(&udp_conf, 0, sizeof(udp_conf));
211 #if IS_ENABLED(CONFIG_IPV6)
212 	udp_conf.family = AF_INET6;
213 #else
214 	udp_conf.family = AF_INET;
215 #endif
216 	udp_conf.local_udp_port = port;
217 	/* Open UDP socket */
218 	err = udp_sock_create(net, &udp_conf, &sock);
219 	if (err < 0)
220 		return ERR_PTR(err);
221 
222 	udp_allow_gso(sock->sk);
223 	return sock;
224 }
225 
226 /* Create new listen socket if needed */
227 static int bareudp_socket_create(struct bareudp_dev *bareudp, __be16 port)
228 {
229 	struct udp_tunnel_sock_cfg tunnel_cfg;
230 	struct socket *sock;
231 
232 	sock = bareudp_create_sock(bareudp->net, port);
233 	if (IS_ERR(sock))
234 		return PTR_ERR(sock);
235 
236 	/* Mark socket as an encapsulation socket */
237 	memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
238 	tunnel_cfg.sk_user_data = bareudp;
239 	tunnel_cfg.encap_type = 1;
240 	tunnel_cfg.encap_rcv = bareudp_udp_encap_recv;
241 	tunnel_cfg.encap_err_lookup = bareudp_err_lookup;
242 	tunnel_cfg.encap_destroy = NULL;
243 	setup_udp_tunnel_sock(bareudp->net, sock, &tunnel_cfg);
244 
245 	rcu_assign_pointer(bareudp->sock, sock);
246 	return 0;
247 }
248 
249 static int bareudp_open(struct net_device *dev)
250 {
251 	struct bareudp_dev *bareudp = netdev_priv(dev);
252 	int ret = 0;
253 
254 	ret =  bareudp_socket_create(bareudp, bareudp->port);
255 	return ret;
256 }
257 
258 static void bareudp_sock_release(struct bareudp_dev *bareudp)
259 {
260 	struct socket *sock;
261 
262 	sock = bareudp->sock;
263 	rcu_assign_pointer(bareudp->sock, NULL);
264 	synchronize_net();
265 	udp_tunnel_sock_release(sock);
266 }
267 
268 static int bareudp_stop(struct net_device *dev)
269 {
270 	struct bareudp_dev *bareudp = netdev_priv(dev);
271 
272 	bareudp_sock_release(bareudp);
273 	return 0;
274 }
275 
276 static int bareudp_xmit_skb(struct sk_buff *skb, struct net_device *dev,
277 			    struct bareudp_dev *bareudp,
278 			    const struct ip_tunnel_info *info)
279 {
280 	bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
281 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
282 	struct socket *sock = rcu_dereference(bareudp->sock);
283 	bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
284 	const struct ip_tunnel_key *key = &info->key;
285 	struct rtable *rt;
286 	__be16 sport, df;
287 	int min_headroom;
288 	__u8 tos, ttl;
289 	__be32 saddr;
290 	int err;
291 
292 	if (!sock)
293 		return -ESHUTDOWN;
294 
295 	rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr, info,
296 				    IPPROTO_UDP, use_cache);
297 
298 	if (IS_ERR(rt))
299 		return PTR_ERR(rt);
300 
301 	skb_tunnel_check_pmtu(skb, &rt->dst,
302 			      BAREUDP_IPV4_HLEN + info->options_len, false);
303 
304 	sport = udp_flow_src_port(bareudp->net, skb,
305 				  bareudp->sport_min, USHRT_MAX,
306 				  true);
307 	tos = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
308 	ttl = key->ttl;
309 	df = key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
310 	skb_scrub_packet(skb, xnet);
311 
312 	err = -ENOSPC;
313 	if (!skb_pull(skb, skb_network_offset(skb)))
314 		goto free_dst;
315 
316 	min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len +
317 		BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr);
318 
319 	err = skb_cow_head(skb, min_headroom);
320 	if (unlikely(err))
321 		goto free_dst;
322 
323 	err = udp_tunnel_handle_offloads(skb, udp_sum);
324 	if (err)
325 		goto free_dst;
326 
327 	skb_set_inner_protocol(skb, bareudp->ethertype);
328 	udp_tunnel_xmit_skb(rt, sock->sk, skb, saddr, info->key.u.ipv4.dst,
329 			    tos, ttl, df, sport, bareudp->port,
330 			    !net_eq(bareudp->net, dev_net(bareudp->dev)),
331 			    !(info->key.tun_flags & TUNNEL_CSUM));
332 	return 0;
333 
334 free_dst:
335 	dst_release(&rt->dst);
336 	return err;
337 }
338 
339 static int bareudp6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
340 			     struct bareudp_dev *bareudp,
341 			     const struct ip_tunnel_info *info)
342 {
343 	bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
344 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
345 	struct socket *sock  = rcu_dereference(bareudp->sock);
346 	bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
347 	const struct ip_tunnel_key *key = &info->key;
348 	struct dst_entry *dst = NULL;
349 	struct in6_addr saddr, daddr;
350 	int min_headroom;
351 	__u8 prio, ttl;
352 	__be16 sport;
353 	int err;
354 
355 	if (!sock)
356 		return -ESHUTDOWN;
357 
358 	dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock, &saddr, info,
359 				    IPPROTO_UDP, use_cache);
360 	if (IS_ERR(dst))
361 		return PTR_ERR(dst);
362 
363 	skb_tunnel_check_pmtu(skb, dst, BAREUDP_IPV6_HLEN + info->options_len,
364 			      false);
365 
366 	sport = udp_flow_src_port(bareudp->net, skb,
367 				  bareudp->sport_min, USHRT_MAX,
368 				  true);
369 	prio = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
370 	ttl = key->ttl;
371 
372 	skb_scrub_packet(skb, xnet);
373 
374 	err = -ENOSPC;
375 	if (!skb_pull(skb, skb_network_offset(skb)))
376 		goto free_dst;
377 
378 	min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len +
379 		BAREUDP_BASE_HLEN + info->options_len + sizeof(struct ipv6hdr);
380 
381 	err = skb_cow_head(skb, min_headroom);
382 	if (unlikely(err))
383 		goto free_dst;
384 
385 	err = udp_tunnel_handle_offloads(skb, udp_sum);
386 	if (err)
387 		goto free_dst;
388 
389 	daddr = info->key.u.ipv6.dst;
390 	udp_tunnel6_xmit_skb(dst, sock->sk, skb, dev,
391 			     &saddr, &daddr, prio, ttl,
392 			     info->key.label, sport, bareudp->port,
393 			     !(info->key.tun_flags & TUNNEL_CSUM));
394 	return 0;
395 
396 free_dst:
397 	dst_release(dst);
398 	return err;
399 }
400 
401 static bool bareudp_proto_valid(struct bareudp_dev *bareudp, __be16 proto)
402 {
403 	if (bareudp->ethertype == proto)
404 		return true;
405 
406 	if (!bareudp->multi_proto_mode)
407 		return false;
408 
409 	if (bareudp->ethertype == htons(ETH_P_MPLS_UC) &&
410 	    proto == htons(ETH_P_MPLS_MC))
411 		return true;
412 
413 	if (bareudp->ethertype == htons(ETH_P_IP) &&
414 	    proto == htons(ETH_P_IPV6))
415 		return true;
416 
417 	return false;
418 }
419 
420 static netdev_tx_t bareudp_xmit(struct sk_buff *skb, struct net_device *dev)
421 {
422 	struct bareudp_dev *bareudp = netdev_priv(dev);
423 	struct ip_tunnel_info *info = NULL;
424 	int err;
425 
426 	if (!bareudp_proto_valid(bareudp, skb->protocol)) {
427 		err = -EINVAL;
428 		goto tx_error;
429 	}
430 
431 	info = skb_tunnel_info(skb);
432 	if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) {
433 		err = -EINVAL;
434 		goto tx_error;
435 	}
436 
437 	rcu_read_lock();
438 	if (IS_ENABLED(CONFIG_IPV6) && info->mode & IP_TUNNEL_INFO_IPV6)
439 		err = bareudp6_xmit_skb(skb, dev, bareudp, info);
440 	else
441 		err = bareudp_xmit_skb(skb, dev, bareudp, info);
442 
443 	rcu_read_unlock();
444 
445 	if (likely(!err))
446 		return NETDEV_TX_OK;
447 tx_error:
448 	dev_kfree_skb(skb);
449 
450 	if (err == -ELOOP)
451 		dev->stats.collisions++;
452 	else if (err == -ENETUNREACH)
453 		dev->stats.tx_carrier_errors++;
454 
455 	dev->stats.tx_errors++;
456 	return NETDEV_TX_OK;
457 }
458 
459 static int bareudp_fill_metadata_dst(struct net_device *dev,
460 				     struct sk_buff *skb)
461 {
462 	struct ip_tunnel_info *info = skb_tunnel_info(skb);
463 	struct bareudp_dev *bareudp = netdev_priv(dev);
464 	bool use_cache;
465 
466 	use_cache = ip_tunnel_dst_cache_usable(skb, info);
467 
468 	if (!IS_ENABLED(CONFIG_IPV6) || ip_tunnel_info_af(info) == AF_INET) {
469 		struct rtable *rt;
470 		__be32 saddr;
471 
472 		rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr,
473 					    info, IPPROTO_UDP, use_cache);
474 		if (IS_ERR(rt))
475 			return PTR_ERR(rt);
476 
477 		ip_rt_put(rt);
478 		info->key.u.ipv4.src = saddr;
479 	} else if (ip_tunnel_info_af(info) == AF_INET6) {
480 		struct dst_entry *dst;
481 		struct in6_addr saddr;
482 		struct socket *sock = rcu_dereference(bareudp->sock);
483 
484 		dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock,
485 					    &saddr, info, IPPROTO_UDP,
486 					    use_cache);
487 		if (IS_ERR(dst))
488 			return PTR_ERR(dst);
489 
490 		dst_release(dst);
491 		info->key.u.ipv6.src = saddr;
492 	} else {
493 		return -EINVAL;
494 	}
495 
496 	info->key.tp_src = udp_flow_src_port(bareudp->net, skb,
497 					     bareudp->sport_min,
498 			USHRT_MAX, true);
499 	info->key.tp_dst = bareudp->port;
500 	return 0;
501 }
502 
503 static const struct net_device_ops bareudp_netdev_ops = {
504 	.ndo_init               = bareudp_init,
505 	.ndo_uninit             = bareudp_uninit,
506 	.ndo_open               = bareudp_open,
507 	.ndo_stop               = bareudp_stop,
508 	.ndo_start_xmit         = bareudp_xmit,
509 	.ndo_get_stats64        = dev_get_tstats64,
510 	.ndo_fill_metadata_dst  = bareudp_fill_metadata_dst,
511 };
512 
513 static const struct nla_policy bareudp_policy[IFLA_BAREUDP_MAX + 1] = {
514 	[IFLA_BAREUDP_PORT]                = { .type = NLA_U16 },
515 	[IFLA_BAREUDP_ETHERTYPE]	   = { .type = NLA_U16 },
516 	[IFLA_BAREUDP_SRCPORT_MIN]         = { .type = NLA_U16 },
517 	[IFLA_BAREUDP_MULTIPROTO_MODE]     = { .type = NLA_FLAG },
518 };
519 
520 /* Info for udev, that this is a virtual tunnel endpoint */
521 static const struct device_type bareudp_type = {
522 	.name = "bareudp",
523 };
524 
525 /* Initialize the device structure. */
526 static void bareudp_setup(struct net_device *dev)
527 {
528 	dev->netdev_ops = &bareudp_netdev_ops;
529 	dev->needs_free_netdev = true;
530 	SET_NETDEV_DEVTYPE(dev, &bareudp_type);
531 	dev->features    |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
532 	dev->features    |= NETIF_F_RXCSUM;
533 	dev->features    |= NETIF_F_LLTX;
534 	dev->features    |= NETIF_F_GSO_SOFTWARE;
535 	dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
536 	dev->hw_features |= NETIF_F_RXCSUM;
537 	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
538 	dev->hard_header_len = 0;
539 	dev->addr_len = 0;
540 	dev->mtu = ETH_DATA_LEN;
541 	dev->min_mtu = IPV4_MIN_MTU;
542 	dev->max_mtu = IP_MAX_MTU - BAREUDP_BASE_HLEN;
543 	dev->type = ARPHRD_NONE;
544 	netif_keep_dst(dev);
545 	dev->priv_flags |= IFF_NO_QUEUE;
546 	dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
547 }
548 
549 static int bareudp_validate(struct nlattr *tb[], struct nlattr *data[],
550 			    struct netlink_ext_ack *extack)
551 {
552 	if (!data) {
553 		NL_SET_ERR_MSG(extack,
554 			       "Not enough attributes provided to perform the operation");
555 		return -EINVAL;
556 	}
557 	return 0;
558 }
559 
560 static int bareudp2info(struct nlattr *data[], struct bareudp_conf *conf,
561 			struct netlink_ext_ack *extack)
562 {
563 	memset(conf, 0, sizeof(*conf));
564 
565 	if (!data[IFLA_BAREUDP_PORT]) {
566 		NL_SET_ERR_MSG(extack, "port not specified");
567 		return -EINVAL;
568 	}
569 	if (!data[IFLA_BAREUDP_ETHERTYPE]) {
570 		NL_SET_ERR_MSG(extack, "ethertype not specified");
571 		return -EINVAL;
572 	}
573 
574 	if (data[IFLA_BAREUDP_PORT])
575 		conf->port =  nla_get_u16(data[IFLA_BAREUDP_PORT]);
576 
577 	if (data[IFLA_BAREUDP_ETHERTYPE])
578 		conf->ethertype =  nla_get_u16(data[IFLA_BAREUDP_ETHERTYPE]);
579 
580 	if (data[IFLA_BAREUDP_SRCPORT_MIN])
581 		conf->sport_min =  nla_get_u16(data[IFLA_BAREUDP_SRCPORT_MIN]);
582 
583 	if (data[IFLA_BAREUDP_MULTIPROTO_MODE])
584 		conf->multi_proto_mode = true;
585 
586 	return 0;
587 }
588 
589 static struct bareudp_dev *bareudp_find_dev(struct bareudp_net *bn,
590 					    const struct bareudp_conf *conf)
591 {
592 	struct bareudp_dev *bareudp, *t = NULL;
593 
594 	list_for_each_entry(bareudp, &bn->bareudp_list, next) {
595 		if (conf->port == bareudp->port)
596 			t = bareudp;
597 	}
598 	return t;
599 }
600 
601 static int bareudp_configure(struct net *net, struct net_device *dev,
602 			     struct bareudp_conf *conf)
603 {
604 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
605 	struct bareudp_dev *t, *bareudp = netdev_priv(dev);
606 	int err;
607 
608 	bareudp->net = net;
609 	bareudp->dev = dev;
610 	t = bareudp_find_dev(bn, conf);
611 	if (t)
612 		return -EBUSY;
613 
614 	if (conf->multi_proto_mode &&
615 	    (conf->ethertype != htons(ETH_P_MPLS_UC) &&
616 	     conf->ethertype != htons(ETH_P_IP)))
617 		return -EINVAL;
618 
619 	bareudp->port = conf->port;
620 	bareudp->ethertype = conf->ethertype;
621 	bareudp->sport_min = conf->sport_min;
622 	bareudp->multi_proto_mode = conf->multi_proto_mode;
623 
624 	err = register_netdevice(dev);
625 	if (err)
626 		return err;
627 
628 	list_add(&bareudp->next, &bn->bareudp_list);
629 	return 0;
630 }
631 
632 static int bareudp_link_config(struct net_device *dev,
633 			       struct nlattr *tb[])
634 {
635 	int err;
636 
637 	if (tb[IFLA_MTU]) {
638 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
639 		if (err)
640 			return err;
641 	}
642 	return 0;
643 }
644 
645 static void bareudp_dellink(struct net_device *dev, struct list_head *head)
646 {
647 	struct bareudp_dev *bareudp = netdev_priv(dev);
648 
649 	list_del(&bareudp->next);
650 	unregister_netdevice_queue(dev, head);
651 }
652 
653 static int bareudp_newlink(struct net *net, struct net_device *dev,
654 			   struct nlattr *tb[], struct nlattr *data[],
655 			   struct netlink_ext_ack *extack)
656 {
657 	struct bareudp_conf conf;
658 	int err;
659 
660 	err = bareudp2info(data, &conf, extack);
661 	if (err)
662 		return err;
663 
664 	err = bareudp_configure(net, dev, &conf);
665 	if (err)
666 		return err;
667 
668 	err = bareudp_link_config(dev, tb);
669 	if (err)
670 		goto err_unconfig;
671 
672 	return 0;
673 
674 err_unconfig:
675 	bareudp_dellink(dev, NULL);
676 	return err;
677 }
678 
679 static size_t bareudp_get_size(const struct net_device *dev)
680 {
681 	return  nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_PORT */
682 		nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_ETHERTYPE */
683 		nla_total_size(sizeof(__u16))  +  /* IFLA_BAREUDP_SRCPORT_MIN */
684 		nla_total_size(0)              +  /* IFLA_BAREUDP_MULTIPROTO_MODE */
685 		0;
686 }
687 
688 static int bareudp_fill_info(struct sk_buff *skb, const struct net_device *dev)
689 {
690 	struct bareudp_dev *bareudp = netdev_priv(dev);
691 
692 	if (nla_put_be16(skb, IFLA_BAREUDP_PORT, bareudp->port))
693 		goto nla_put_failure;
694 	if (nla_put_be16(skb, IFLA_BAREUDP_ETHERTYPE, bareudp->ethertype))
695 		goto nla_put_failure;
696 	if (nla_put_u16(skb, IFLA_BAREUDP_SRCPORT_MIN, bareudp->sport_min))
697 		goto nla_put_failure;
698 	if (bareudp->multi_proto_mode &&
699 	    nla_put_flag(skb, IFLA_BAREUDP_MULTIPROTO_MODE))
700 		goto nla_put_failure;
701 
702 	return 0;
703 
704 nla_put_failure:
705 	return -EMSGSIZE;
706 }
707 
708 static struct rtnl_link_ops bareudp_link_ops __read_mostly = {
709 	.kind           = "bareudp",
710 	.maxtype        = IFLA_BAREUDP_MAX,
711 	.policy         = bareudp_policy,
712 	.priv_size      = sizeof(struct bareudp_dev),
713 	.setup          = bareudp_setup,
714 	.validate       = bareudp_validate,
715 	.newlink        = bareudp_newlink,
716 	.dellink        = bareudp_dellink,
717 	.get_size       = bareudp_get_size,
718 	.fill_info      = bareudp_fill_info,
719 };
720 
721 struct net_device *bareudp_dev_create(struct net *net, const char *name,
722 				      u8 name_assign_type,
723 				      struct bareudp_conf *conf)
724 {
725 	struct nlattr *tb[IFLA_MAX + 1];
726 	struct net_device *dev;
727 	int err;
728 
729 	memset(tb, 0, sizeof(tb));
730 	dev = rtnl_create_link(net, name, name_assign_type,
731 			       &bareudp_link_ops, tb, NULL);
732 	if (IS_ERR(dev))
733 		return dev;
734 
735 	err = bareudp_configure(net, dev, conf);
736 	if (err) {
737 		free_netdev(dev);
738 		return ERR_PTR(err);
739 	}
740 	err = dev_set_mtu(dev, IP_MAX_MTU - BAREUDP_BASE_HLEN);
741 	if (err)
742 		goto err;
743 
744 	err = rtnl_configure_link(dev, NULL);
745 	if (err < 0)
746 		goto err;
747 
748 	return dev;
749 err:
750 	bareudp_dellink(dev, NULL);
751 	return ERR_PTR(err);
752 }
753 EXPORT_SYMBOL_GPL(bareudp_dev_create);
754 
755 static __net_init int bareudp_init_net(struct net *net)
756 {
757 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
758 
759 	INIT_LIST_HEAD(&bn->bareudp_list);
760 	return 0;
761 }
762 
763 static void bareudp_destroy_tunnels(struct net *net, struct list_head *head)
764 {
765 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
766 	struct bareudp_dev *bareudp, *next;
767 
768 	list_for_each_entry_safe(bareudp, next, &bn->bareudp_list, next)
769 		unregister_netdevice_queue(bareudp->dev, head);
770 }
771 
772 static void __net_exit bareudp_exit_batch_net(struct list_head *net_list)
773 {
774 	struct net *net;
775 	LIST_HEAD(list);
776 
777 	rtnl_lock();
778 	list_for_each_entry(net, net_list, exit_list)
779 		bareudp_destroy_tunnels(net, &list);
780 
781 	/* unregister the devices gathered above */
782 	unregister_netdevice_many(&list);
783 	rtnl_unlock();
784 }
785 
786 static struct pernet_operations bareudp_net_ops = {
787 	.init = bareudp_init_net,
788 	.exit_batch = bareudp_exit_batch_net,
789 	.id   = &bareudp_net_id,
790 	.size = sizeof(struct bareudp_net),
791 };
792 
793 static int __init bareudp_init_module(void)
794 {
795 	int rc;
796 
797 	rc = register_pernet_subsys(&bareudp_net_ops);
798 	if (rc)
799 		goto out1;
800 
801 	rc = rtnl_link_register(&bareudp_link_ops);
802 	if (rc)
803 		goto out2;
804 
805 	return 0;
806 out2:
807 	unregister_pernet_subsys(&bareudp_net_ops);
808 out1:
809 	return rc;
810 }
811 late_initcall(bareudp_init_module);
812 
813 static void __exit bareudp_cleanup_module(void)
814 {
815 	rtnl_link_unregister(&bareudp_link_ops);
816 	unregister_pernet_subsys(&bareudp_net_ops);
817 }
818 module_exit(bareudp_cleanup_module);
819 
820 MODULE_ALIAS_RTNL_LINK("bareudp");
821 MODULE_LICENSE("GPL");
822 MODULE_AUTHOR("Martin Varghese <martin.varghese@nokia.com>");
823 MODULE_DESCRIPTION("Interface driver for UDP encapsulated traffic");
824