xref: /openbmc/linux/drivers/net/bareudp.c (revision 82df5b73)
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 pcpu_sw_netstats *stats;
58 	struct bareudp_dev *bareudp;
59 	unsigned short family;
60 	unsigned int len;
61 	__be16 proto;
62 	void *oiph;
63 	int err;
64 
65 	bareudp = rcu_dereference_sk_user_data(sk);
66 	if (!bareudp)
67 		goto drop;
68 
69 	if (skb->protocol ==  htons(ETH_P_IP))
70 		family = AF_INET;
71 	else
72 		family = AF_INET6;
73 
74 	if (bareudp->ethertype == htons(ETH_P_IP)) {
75 		struct iphdr *iphdr;
76 
77 		iphdr = (struct iphdr *)(skb->data + BAREUDP_BASE_HLEN);
78 		if (iphdr->version == 4) {
79 			proto = bareudp->ethertype;
80 		} else if (bareudp->multi_proto_mode && (iphdr->version == 6)) {
81 			proto = htons(ETH_P_IPV6);
82 		} else {
83 			bareudp->dev->stats.rx_dropped++;
84 			goto drop;
85 		}
86 	} else if (bareudp->ethertype == htons(ETH_P_MPLS_UC)) {
87 		struct iphdr *tunnel_hdr;
88 
89 		tunnel_hdr = (struct iphdr *)skb_network_header(skb);
90 		if (tunnel_hdr->version == 4) {
91 			if (!ipv4_is_multicast(tunnel_hdr->daddr)) {
92 				proto = bareudp->ethertype;
93 			} else if (bareudp->multi_proto_mode &&
94 				   ipv4_is_multicast(tunnel_hdr->daddr)) {
95 				proto = htons(ETH_P_MPLS_MC);
96 			} else {
97 				bareudp->dev->stats.rx_dropped++;
98 				goto drop;
99 			}
100 		} else {
101 			int addr_type;
102 			struct ipv6hdr *tunnel_hdr_v6;
103 
104 			tunnel_hdr_v6 = (struct ipv6hdr *)skb_network_header(skb);
105 			addr_type =
106 			ipv6_addr_type((struct in6_addr *)&tunnel_hdr_v6->daddr);
107 			if (!(addr_type & IPV6_ADDR_MULTICAST)) {
108 				proto = bareudp->ethertype;
109 			} else if (bareudp->multi_proto_mode &&
110 				   (addr_type & IPV6_ADDR_MULTICAST)) {
111 				proto = htons(ETH_P_MPLS_MC);
112 			} else {
113 				bareudp->dev->stats.rx_dropped++;
114 				goto drop;
115 			}
116 		}
117 	} else {
118 		proto = bareudp->ethertype;
119 	}
120 
121 	if (iptunnel_pull_header(skb, BAREUDP_BASE_HLEN,
122 				 proto,
123 				 !net_eq(bareudp->net,
124 				 dev_net(bareudp->dev)))) {
125 		bareudp->dev->stats.rx_dropped++;
126 		goto drop;
127 	}
128 
129 	tun_dst = udp_tun_rx_dst(skb, family, TUNNEL_KEY, 0, 0);
130 	if (!tun_dst) {
131 		bareudp->dev->stats.rx_dropped++;
132 		goto drop;
133 	}
134 	skb_dst_set(skb, &tun_dst->dst);
135 	skb->dev = bareudp->dev;
136 	oiph = skb_network_header(skb);
137 	skb_reset_network_header(skb);
138 
139 	if (!IS_ENABLED(CONFIG_IPV6) || family == AF_INET)
140 		err = IP_ECN_decapsulate(oiph, skb);
141 	else
142 		err = IP6_ECN_decapsulate(oiph, skb);
143 
144 	if (unlikely(err)) {
145 		if (log_ecn_error) {
146 			if  (!IS_ENABLED(CONFIG_IPV6) || family == AF_INET)
147 				net_info_ratelimited("non-ECT from %pI4 "
148 						     "with TOS=%#x\n",
149 						     &((struct iphdr *)oiph)->saddr,
150 						     ((struct iphdr *)oiph)->tos);
151 			else
152 				net_info_ratelimited("non-ECT from %pI6\n",
153 						     &((struct ipv6hdr *)oiph)->saddr);
154 		}
155 		if (err > 1) {
156 			++bareudp->dev->stats.rx_frame_errors;
157 			++bareudp->dev->stats.rx_errors;
158 			goto drop;
159 		}
160 	}
161 
162 	len = skb->len;
163 	err = gro_cells_receive(&bareudp->gro_cells, skb);
164 	if (likely(err == NET_RX_SUCCESS)) {
165 		stats = this_cpu_ptr(bareudp->dev->tstats);
166 		u64_stats_update_begin(&stats->syncp);
167 		stats->rx_packets++;
168 		stats->rx_bytes += len;
169 		u64_stats_update_end(&stats->syncp);
170 	}
171 	return 0;
172 drop:
173 	/* Consume bad packet */
174 	kfree_skb(skb);
175 
176 	return 0;
177 }
178 
179 static int bareudp_err_lookup(struct sock *sk, struct sk_buff *skb)
180 {
181 	return 0;
182 }
183 
184 static int bareudp_init(struct net_device *dev)
185 {
186 	struct bareudp_dev *bareudp = netdev_priv(dev);
187 	int err;
188 
189 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
190 	if (!dev->tstats)
191 		return -ENOMEM;
192 
193 	err = gro_cells_init(&bareudp->gro_cells, dev);
194 	if (err) {
195 		free_percpu(dev->tstats);
196 		return err;
197 	}
198 	return 0;
199 }
200 
201 static void bareudp_uninit(struct net_device *dev)
202 {
203 	struct bareudp_dev *bareudp = netdev_priv(dev);
204 
205 	gro_cells_destroy(&bareudp->gro_cells);
206 	free_percpu(dev->tstats);
207 }
208 
209 static struct socket *bareudp_create_sock(struct net *net, __be16 port)
210 {
211 	struct udp_port_cfg udp_conf;
212 	struct socket *sock;
213 	int err;
214 
215 	memset(&udp_conf, 0, sizeof(udp_conf));
216 #if IS_ENABLED(CONFIG_IPV6)
217 	udp_conf.family = AF_INET6;
218 #else
219 	udp_conf.family = AF_INET;
220 #endif
221 	udp_conf.local_udp_port = port;
222 	/* Open UDP socket */
223 	err = udp_sock_create(net, &udp_conf, &sock);
224 	if (err < 0)
225 		return ERR_PTR(err);
226 
227 	return sock;
228 }
229 
230 /* Create new listen socket if needed */
231 static int bareudp_socket_create(struct bareudp_dev *bareudp, __be16 port)
232 {
233 	struct udp_tunnel_sock_cfg tunnel_cfg;
234 	struct socket *sock;
235 
236 	sock = bareudp_create_sock(bareudp->net, port);
237 	if (IS_ERR(sock))
238 		return PTR_ERR(sock);
239 
240 	/* Mark socket as an encapsulation socket */
241 	memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
242 	tunnel_cfg.sk_user_data = bareudp;
243 	tunnel_cfg.encap_type = 1;
244 	tunnel_cfg.encap_rcv = bareudp_udp_encap_recv;
245 	tunnel_cfg.encap_err_lookup = bareudp_err_lookup;
246 	tunnel_cfg.encap_destroy = NULL;
247 	setup_udp_tunnel_sock(bareudp->net, sock, &tunnel_cfg);
248 
249 	/* As the setup_udp_tunnel_sock does not call udp_encap_enable if the
250 	 * socket type is v6 an explicit call to udp_encap_enable is needed.
251 	 */
252 	if (sock->sk->sk_family == AF_INET6)
253 		udp_encap_enable();
254 
255 	rcu_assign_pointer(bareudp->sock, sock);
256 	return 0;
257 }
258 
259 static int bareudp_open(struct net_device *dev)
260 {
261 	struct bareudp_dev *bareudp = netdev_priv(dev);
262 	int ret = 0;
263 
264 	ret =  bareudp_socket_create(bareudp, bareudp->port);
265 	return ret;
266 }
267 
268 static void bareudp_sock_release(struct bareudp_dev *bareudp)
269 {
270 	struct socket *sock;
271 
272 	sock = bareudp->sock;
273 	rcu_assign_pointer(bareudp->sock, NULL);
274 	synchronize_net();
275 	udp_tunnel_sock_release(sock);
276 }
277 
278 static int bareudp_stop(struct net_device *dev)
279 {
280 	struct bareudp_dev *bareudp = netdev_priv(dev);
281 
282 	bareudp_sock_release(bareudp);
283 	return 0;
284 }
285 
286 static int bareudp_xmit_skb(struct sk_buff *skb, struct net_device *dev,
287 			    struct bareudp_dev *bareudp,
288 			    const struct ip_tunnel_info *info)
289 {
290 	bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
291 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
292 	struct socket *sock = rcu_dereference(bareudp->sock);
293 	bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
294 	const struct ip_tunnel_key *key = &info->key;
295 	struct rtable *rt;
296 	__be16 sport, df;
297 	int min_headroom;
298 	__u8 tos, ttl;
299 	__be32 saddr;
300 	int err;
301 
302 	if (!sock)
303 		return -ESHUTDOWN;
304 
305 	rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr, info,
306 				    IPPROTO_UDP, use_cache);
307 
308 	if (IS_ERR(rt))
309 		return PTR_ERR(rt);
310 
311 	skb_tunnel_check_pmtu(skb, &rt->dst,
312 			      BAREUDP_IPV4_HLEN + info->options_len);
313 
314 	sport = udp_flow_src_port(bareudp->net, skb,
315 				  bareudp->sport_min, USHRT_MAX,
316 				  true);
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 	dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock, &saddr, info,
369 				    IPPROTO_UDP, use_cache);
370 	if (IS_ERR(dst))
371 		return PTR_ERR(dst);
372 
373 	skb_tunnel_check_pmtu(skb, dst, BAREUDP_IPV6_HLEN + info->options_len);
374 
375 	sport = udp_flow_src_port(bareudp->net, skb,
376 				  bareudp->sport_min, USHRT_MAX,
377 				  true);
378 	prio = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
379 	ttl = key->ttl;
380 
381 	skb_scrub_packet(skb, xnet);
382 
383 	err = -ENOSPC;
384 	if (!skb_pull(skb, skb_network_offset(skb)))
385 		goto free_dst;
386 
387 	min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len +
388 		BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr);
389 
390 	err = skb_cow_head(skb, min_headroom);
391 	if (unlikely(err))
392 		goto free_dst;
393 
394 	err = udp_tunnel_handle_offloads(skb, udp_sum);
395 	if (err)
396 		goto free_dst;
397 
398 	daddr = info->key.u.ipv6.dst;
399 	udp_tunnel6_xmit_skb(dst, sock->sk, skb, dev,
400 			     &saddr, &daddr, prio, ttl,
401 			     info->key.label, sport, bareudp->port,
402 			     !(info->key.tun_flags & TUNNEL_CSUM));
403 	return 0;
404 
405 free_dst:
406 	dst_release(dst);
407 	return err;
408 }
409 
410 static netdev_tx_t bareudp_xmit(struct sk_buff *skb, struct net_device *dev)
411 {
412 	struct bareudp_dev *bareudp = netdev_priv(dev);
413 	struct ip_tunnel_info *info = NULL;
414 	int err;
415 
416 	if (skb->protocol != bareudp->ethertype) {
417 		if (!bareudp->multi_proto_mode ||
418 		    (skb->protocol !=  htons(ETH_P_MPLS_MC) &&
419 		     skb->protocol !=  htons(ETH_P_IPV6))) {
420 			err = -EINVAL;
421 			goto tx_error;
422 		}
423 	}
424 
425 	info = skb_tunnel_info(skb);
426 	if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) {
427 		err = -EINVAL;
428 		goto tx_error;
429 	}
430 
431 	rcu_read_lock();
432 	if (IS_ENABLED(CONFIG_IPV6) && info->mode & IP_TUNNEL_INFO_IPV6)
433 		err = bareudp6_xmit_skb(skb, dev, bareudp, info);
434 	else
435 		err = bareudp_xmit_skb(skb, dev, bareudp, info);
436 
437 	rcu_read_unlock();
438 
439 	if (likely(!err))
440 		return NETDEV_TX_OK;
441 tx_error:
442 	dev_kfree_skb(skb);
443 
444 	if (err == -ELOOP)
445 		dev->stats.collisions++;
446 	else if (err == -ENETUNREACH)
447 		dev->stats.tx_carrier_errors++;
448 
449 	dev->stats.tx_errors++;
450 	return NETDEV_TX_OK;
451 }
452 
453 static int bareudp_fill_metadata_dst(struct net_device *dev,
454 				     struct sk_buff *skb)
455 {
456 	struct ip_tunnel_info *info = skb_tunnel_info(skb);
457 	struct bareudp_dev *bareudp = netdev_priv(dev);
458 	bool use_cache;
459 
460 	use_cache = ip_tunnel_dst_cache_usable(skb, info);
461 
462 	if (!IS_ENABLED(CONFIG_IPV6) || ip_tunnel_info_af(info) == AF_INET) {
463 		struct rtable *rt;
464 		__be32 saddr;
465 
466 		rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr,
467 					    info, IPPROTO_UDP, use_cache);
468 		if (IS_ERR(rt))
469 			return PTR_ERR(rt);
470 
471 		ip_rt_put(rt);
472 		info->key.u.ipv4.src = saddr;
473 	} else if (ip_tunnel_info_af(info) == AF_INET6) {
474 		struct dst_entry *dst;
475 		struct in6_addr saddr;
476 		struct socket *sock = rcu_dereference(bareudp->sock);
477 
478 		dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock,
479 					    &saddr, info, IPPROTO_UDP,
480 					    use_cache);
481 		if (IS_ERR(dst))
482 			return PTR_ERR(dst);
483 
484 		dst_release(dst);
485 		info->key.u.ipv6.src = saddr;
486 	} else {
487 		return -EINVAL;
488 	}
489 
490 	info->key.tp_src = udp_flow_src_port(bareudp->net, skb,
491 					     bareudp->sport_min,
492 			USHRT_MAX, true);
493 	info->key.tp_dst = bareudp->port;
494 	return 0;
495 }
496 
497 static const struct net_device_ops bareudp_netdev_ops = {
498 	.ndo_init               = bareudp_init,
499 	.ndo_uninit             = bareudp_uninit,
500 	.ndo_open               = bareudp_open,
501 	.ndo_stop               = bareudp_stop,
502 	.ndo_start_xmit         = bareudp_xmit,
503 	.ndo_get_stats64        = ip_tunnel_get_stats64,
504 	.ndo_fill_metadata_dst  = bareudp_fill_metadata_dst,
505 };
506 
507 static const struct nla_policy bareudp_policy[IFLA_BAREUDP_MAX + 1] = {
508 	[IFLA_BAREUDP_PORT]                = { .type = NLA_U16 },
509 	[IFLA_BAREUDP_ETHERTYPE]	   = { .type = NLA_U16 },
510 	[IFLA_BAREUDP_SRCPORT_MIN]         = { .type = NLA_U16 },
511 	[IFLA_BAREUDP_MULTIPROTO_MODE]     = { .type = NLA_FLAG },
512 };
513 
514 /* Info for udev, that this is a virtual tunnel endpoint */
515 static struct device_type bareudp_type = {
516 	.name = "bareudp",
517 };
518 
519 /* Initialize the device structure. */
520 static void bareudp_setup(struct net_device *dev)
521 {
522 	dev->netdev_ops = &bareudp_netdev_ops;
523 	dev->needs_free_netdev = true;
524 	SET_NETDEV_DEVTYPE(dev, &bareudp_type);
525 	dev->features    |= NETIF_F_SG | NETIF_F_HW_CSUM;
526 	dev->features    |= NETIF_F_RXCSUM;
527 	dev->features    |= NETIF_F_GSO_SOFTWARE;
528 	dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
529 	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
530 	dev->hard_header_len = 0;
531 	dev->addr_len = 0;
532 	dev->mtu = ETH_DATA_LEN;
533 	dev->min_mtu = IPV4_MIN_MTU;
534 	dev->max_mtu = IP_MAX_MTU - BAREUDP_BASE_HLEN;
535 	dev->type = ARPHRD_NONE;
536 	netif_keep_dst(dev);
537 	dev->priv_flags |= IFF_NO_QUEUE;
538 	dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
539 }
540 
541 static int bareudp_validate(struct nlattr *tb[], struct nlattr *data[],
542 			    struct netlink_ext_ack *extack)
543 {
544 	if (!data) {
545 		NL_SET_ERR_MSG(extack,
546 			       "Not enough attributes provided to perform the operation");
547 		return -EINVAL;
548 	}
549 	return 0;
550 }
551 
552 static int bareudp2info(struct nlattr *data[], struct bareudp_conf *conf,
553 			struct netlink_ext_ack *extack)
554 {
555 	if (!data[IFLA_BAREUDP_PORT]) {
556 		NL_SET_ERR_MSG(extack, "port not specified");
557 		return -EINVAL;
558 	}
559 	if (!data[IFLA_BAREUDP_ETHERTYPE]) {
560 		NL_SET_ERR_MSG(extack, "ethertype not specified");
561 		return -EINVAL;
562 	}
563 
564 	if (data[IFLA_BAREUDP_PORT])
565 		conf->port =  nla_get_u16(data[IFLA_BAREUDP_PORT]);
566 
567 	if (data[IFLA_BAREUDP_ETHERTYPE])
568 		conf->ethertype =  nla_get_u16(data[IFLA_BAREUDP_ETHERTYPE]);
569 
570 	if (data[IFLA_BAREUDP_SRCPORT_MIN])
571 		conf->sport_min =  nla_get_u16(data[IFLA_BAREUDP_SRCPORT_MIN]);
572 
573 	return 0;
574 }
575 
576 static struct bareudp_dev *bareudp_find_dev(struct bareudp_net *bn,
577 					    const struct bareudp_conf *conf)
578 {
579 	struct bareudp_dev *bareudp, *t = NULL;
580 
581 	list_for_each_entry(bareudp, &bn->bareudp_list, next) {
582 		if (conf->port == bareudp->port)
583 			t = bareudp;
584 	}
585 	return t;
586 }
587 
588 static int bareudp_configure(struct net *net, struct net_device *dev,
589 			     struct bareudp_conf *conf)
590 {
591 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
592 	struct bareudp_dev *t, *bareudp = netdev_priv(dev);
593 	int err;
594 
595 	bareudp->net = net;
596 	bareudp->dev = dev;
597 	t = bareudp_find_dev(bn, conf);
598 	if (t)
599 		return -EBUSY;
600 
601 	if (conf->multi_proto_mode &&
602 	    (conf->ethertype != htons(ETH_P_MPLS_UC) &&
603 	     conf->ethertype != htons(ETH_P_IP)))
604 		return -EINVAL;
605 
606 	bareudp->port = conf->port;
607 	bareudp->ethertype = conf->ethertype;
608 	bareudp->sport_min = conf->sport_min;
609 	bareudp->multi_proto_mode = conf->multi_proto_mode;
610 	err = register_netdevice(dev);
611 	if (err)
612 		return err;
613 
614 	list_add(&bareudp->next, &bn->bareudp_list);
615 	return 0;
616 }
617 
618 static int bareudp_link_config(struct net_device *dev,
619 			       struct nlattr *tb[])
620 {
621 	int err;
622 
623 	if (tb[IFLA_MTU]) {
624 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
625 		if (err)
626 			return err;
627 	}
628 	return 0;
629 }
630 
631 static int bareudp_newlink(struct net *net, struct net_device *dev,
632 			   struct nlattr *tb[], struct nlattr *data[],
633 			   struct netlink_ext_ack *extack)
634 {
635 	struct bareudp_conf conf;
636 	int err;
637 
638 	err = bareudp2info(data, &conf, extack);
639 	if (err)
640 		return err;
641 
642 	err = bareudp_configure(net, dev, &conf);
643 	if (err)
644 		return err;
645 
646 	err = bareudp_link_config(dev, tb);
647 	if (err)
648 		return err;
649 
650 	return 0;
651 }
652 
653 static void bareudp_dellink(struct net_device *dev, struct list_head *head)
654 {
655 	struct bareudp_dev *bareudp = netdev_priv(dev);
656 
657 	list_del(&bareudp->next);
658 	unregister_netdevice_queue(dev, head);
659 }
660 
661 static size_t bareudp_get_size(const struct net_device *dev)
662 {
663 	return  nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_PORT */
664 		nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_ETHERTYPE */
665 		nla_total_size(sizeof(__u16))  +  /* IFLA_BAREUDP_SRCPORT_MIN */
666 		nla_total_size(0)              +  /* IFLA_BAREUDP_MULTIPROTO_MODE */
667 		0;
668 }
669 
670 static int bareudp_fill_info(struct sk_buff *skb, const struct net_device *dev)
671 {
672 	struct bareudp_dev *bareudp = netdev_priv(dev);
673 
674 	if (nla_put_be16(skb, IFLA_BAREUDP_PORT, bareudp->port))
675 		goto nla_put_failure;
676 	if (nla_put_be16(skb, IFLA_BAREUDP_ETHERTYPE, bareudp->ethertype))
677 		goto nla_put_failure;
678 	if (nla_put_u16(skb, IFLA_BAREUDP_SRCPORT_MIN, bareudp->sport_min))
679 		goto nla_put_failure;
680 	if (bareudp->multi_proto_mode &&
681 	    nla_put_flag(skb, IFLA_BAREUDP_MULTIPROTO_MODE))
682 		goto nla_put_failure;
683 
684 	return 0;
685 
686 nla_put_failure:
687 	return -EMSGSIZE;
688 }
689 
690 static struct rtnl_link_ops bareudp_link_ops __read_mostly = {
691 	.kind           = "bareudp",
692 	.maxtype        = IFLA_BAREUDP_MAX,
693 	.policy         = bareudp_policy,
694 	.priv_size      = sizeof(struct bareudp_dev),
695 	.setup          = bareudp_setup,
696 	.validate       = bareudp_validate,
697 	.newlink        = bareudp_newlink,
698 	.dellink        = bareudp_dellink,
699 	.get_size       = bareudp_get_size,
700 	.fill_info      = bareudp_fill_info,
701 };
702 
703 struct net_device *bareudp_dev_create(struct net *net, const char *name,
704 				      u8 name_assign_type,
705 				      struct bareudp_conf *conf)
706 {
707 	struct nlattr *tb[IFLA_MAX + 1];
708 	struct net_device *dev;
709 	LIST_HEAD(list_kill);
710 	int err;
711 
712 	memset(tb, 0, sizeof(tb));
713 	dev = rtnl_create_link(net, name, name_assign_type,
714 			       &bareudp_link_ops, tb, NULL);
715 	if (IS_ERR(dev))
716 		return dev;
717 
718 	err = bareudp_configure(net, dev, conf);
719 	if (err) {
720 		free_netdev(dev);
721 		return ERR_PTR(err);
722 	}
723 	err = dev_set_mtu(dev, IP_MAX_MTU - BAREUDP_BASE_HLEN);
724 	if (err)
725 		goto err;
726 
727 	err = rtnl_configure_link(dev, NULL);
728 	if (err < 0)
729 		goto err;
730 
731 	return dev;
732 err:
733 	bareudp_dellink(dev, &list_kill);
734 	unregister_netdevice_many(&list_kill);
735 	return ERR_PTR(err);
736 }
737 EXPORT_SYMBOL_GPL(bareudp_dev_create);
738 
739 static __net_init int bareudp_init_net(struct net *net)
740 {
741 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
742 
743 	INIT_LIST_HEAD(&bn->bareudp_list);
744 	return 0;
745 }
746 
747 static void bareudp_destroy_tunnels(struct net *net, struct list_head *head)
748 {
749 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
750 	struct bareudp_dev *bareudp, *next;
751 
752 	list_for_each_entry_safe(bareudp, next, &bn->bareudp_list, next)
753 		unregister_netdevice_queue(bareudp->dev, head);
754 }
755 
756 static void __net_exit bareudp_exit_batch_net(struct list_head *net_list)
757 {
758 	struct net *net;
759 	LIST_HEAD(list);
760 
761 	rtnl_lock();
762 	list_for_each_entry(net, net_list, exit_list)
763 		bareudp_destroy_tunnels(net, &list);
764 
765 	/* unregister the devices gathered above */
766 	unregister_netdevice_many(&list);
767 	rtnl_unlock();
768 }
769 
770 static struct pernet_operations bareudp_net_ops = {
771 	.init = bareudp_init_net,
772 	.exit_batch = bareudp_exit_batch_net,
773 	.id   = &bareudp_net_id,
774 	.size = sizeof(struct bareudp_net),
775 };
776 
777 static int __init bareudp_init_module(void)
778 {
779 	int rc;
780 
781 	rc = register_pernet_subsys(&bareudp_net_ops);
782 	if (rc)
783 		goto out1;
784 
785 	rc = rtnl_link_register(&bareudp_link_ops);
786 	if (rc)
787 		goto out2;
788 
789 	return 0;
790 out2:
791 	unregister_pernet_subsys(&bareudp_net_ops);
792 out1:
793 	return rc;
794 }
795 late_initcall(bareudp_init_module);
796 
797 static void __exit bareudp_cleanup_module(void)
798 {
799 	rtnl_link_unregister(&bareudp_link_ops);
800 	unregister_pernet_subsys(&bareudp_net_ops);
801 }
802 module_exit(bareudp_cleanup_module);
803 
804 MODULE_ALIAS_RTNL_LINK("bareudp");
805 MODULE_LICENSE("GPL");
806 MODULE_AUTHOR("Martin Varghese <martin.varghese@nokia.com>");
807 MODULE_DESCRIPTION("Interface driver for UDP encapsulated traffic");
808