xref: /openbmc/linux/drivers/net/bareudp.c (revision 022dacdd)
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 (family == AF_INET)
140 		err = IP_ECN_decapsulate(oiph, skb);
141 #if IS_ENABLED(CONFIG_IPV6)
142 	else
143 		err = IP6_ECN_decapsulate(oiph, skb);
144 #endif
145 
146 	if (unlikely(err)) {
147 		if (log_ecn_error) {
148 			if  (family == AF_INET)
149 				net_info_ratelimited("non-ECT from %pI4 "
150 						     "with TOS=%#x\n",
151 						     &((struct iphdr *)oiph)->saddr,
152 						     ((struct iphdr *)oiph)->tos);
153 #if IS_ENABLED(CONFIG_IPV6)
154 			else
155 				net_info_ratelimited("non-ECT from %pI6\n",
156 						     &((struct ipv6hdr *)oiph)->saddr);
157 #endif
158 		}
159 		if (err > 1) {
160 			++bareudp->dev->stats.rx_frame_errors;
161 			++bareudp->dev->stats.rx_errors;
162 			goto drop;
163 		}
164 	}
165 
166 	len = skb->len;
167 	err = gro_cells_receive(&bareudp->gro_cells, skb);
168 	if (likely(err == NET_RX_SUCCESS)) {
169 		stats = this_cpu_ptr(bareudp->dev->tstats);
170 		u64_stats_update_begin(&stats->syncp);
171 		stats->rx_packets++;
172 		stats->rx_bytes += len;
173 		u64_stats_update_end(&stats->syncp);
174 	}
175 	return 0;
176 drop:
177 	/* Consume bad packet */
178 	kfree_skb(skb);
179 
180 	return 0;
181 }
182 
183 static int bareudp_err_lookup(struct sock *sk, struct sk_buff *skb)
184 {
185 	return 0;
186 }
187 
188 static int bareudp_init(struct net_device *dev)
189 {
190 	struct bareudp_dev *bareudp = netdev_priv(dev);
191 	int err;
192 
193 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
194 	if (!dev->tstats)
195 		return -ENOMEM;
196 
197 	err = gro_cells_init(&bareudp->gro_cells, dev);
198 	if (err) {
199 		free_percpu(dev->tstats);
200 		return err;
201 	}
202 	return 0;
203 }
204 
205 static void bareudp_uninit(struct net_device *dev)
206 {
207 	struct bareudp_dev *bareudp = netdev_priv(dev);
208 
209 	gro_cells_destroy(&bareudp->gro_cells);
210 	free_percpu(dev->tstats);
211 }
212 
213 static struct socket *bareudp_create_sock(struct net *net, __be16 port)
214 {
215 	struct udp_port_cfg udp_conf;
216 	struct socket *sock;
217 	int err;
218 
219 	memset(&udp_conf, 0, sizeof(udp_conf));
220 #if IS_ENABLED(CONFIG_IPV6)
221 	udp_conf.family = AF_INET6;
222 #else
223 	udp_conf.family = AF_INET;
224 #endif
225 	udp_conf.local_udp_port = port;
226 	/* Open UDP socket */
227 	err = udp_sock_create(net, &udp_conf, &sock);
228 	if (err < 0)
229 		return ERR_PTR(err);
230 
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 	/* As the setup_udp_tunnel_sock does not call udp_encap_enable if the
254 	 * socket type is v6 an explicit call to udp_encap_enable is needed.
255 	 */
256 	if (sock->sk->sk_family == AF_INET6)
257 		udp_encap_enable();
258 
259 	rcu_assign_pointer(bareudp->sock, sock);
260 	return 0;
261 }
262 
263 static int bareudp_open(struct net_device *dev)
264 {
265 	struct bareudp_dev *bareudp = netdev_priv(dev);
266 	int ret = 0;
267 
268 	ret =  bareudp_socket_create(bareudp, bareudp->port);
269 	return ret;
270 }
271 
272 static void bareudp_sock_release(struct bareudp_dev *bareudp)
273 {
274 	struct socket *sock;
275 
276 	sock = bareudp->sock;
277 	rcu_assign_pointer(bareudp->sock, NULL);
278 	synchronize_net();
279 	udp_tunnel_sock_release(sock);
280 }
281 
282 static int bareudp_stop(struct net_device *dev)
283 {
284 	struct bareudp_dev *bareudp = netdev_priv(dev);
285 
286 	bareudp_sock_release(bareudp);
287 	return 0;
288 }
289 
290 static int bareudp_xmit_skb(struct sk_buff *skb, struct net_device *dev,
291 			    struct bareudp_dev *bareudp,
292 			    const struct ip_tunnel_info *info)
293 {
294 	bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
295 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
296 	struct socket *sock = rcu_dereference(bareudp->sock);
297 	bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
298 	const struct ip_tunnel_key *key = &info->key;
299 	struct rtable *rt;
300 	__be16 sport, df;
301 	int min_headroom;
302 	__u8 tos, ttl;
303 	__be32 saddr;
304 	int err;
305 
306 	if (!sock)
307 		return -ESHUTDOWN;
308 
309 	rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr, info,
310 				    IPPROTO_UDP, use_cache);
311 
312 	if (IS_ERR(rt))
313 		return PTR_ERR(rt);
314 
315 	skb_tunnel_check_pmtu(skb, &rt->dst,
316 			      BAREUDP_IPV4_HLEN + info->options_len);
317 
318 	sport = udp_flow_src_port(bareudp->net, skb,
319 				  bareudp->sport_min, USHRT_MAX,
320 				  true);
321 	tos = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
322 	ttl = key->ttl;
323 	df = key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
324 	skb_scrub_packet(skb, xnet);
325 
326 	err = -ENOSPC;
327 	if (!skb_pull(skb, skb_network_offset(skb)))
328 		goto free_dst;
329 
330 	min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len +
331 		BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr);
332 
333 	err = skb_cow_head(skb, min_headroom);
334 	if (unlikely(err))
335 		goto free_dst;
336 
337 	err = udp_tunnel_handle_offloads(skb, udp_sum);
338 	if (err)
339 		goto free_dst;
340 
341 	skb_set_inner_protocol(skb, bareudp->ethertype);
342 	udp_tunnel_xmit_skb(rt, sock->sk, skb, saddr, info->key.u.ipv4.dst,
343 			    tos, ttl, df, sport, bareudp->port,
344 			    !net_eq(bareudp->net, dev_net(bareudp->dev)),
345 			    !(info->key.tun_flags & TUNNEL_CSUM));
346 	return 0;
347 
348 free_dst:
349 	dst_release(&rt->dst);
350 	return err;
351 }
352 
353 #if IS_ENABLED(CONFIG_IPV6)
354 static int bareudp6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
355 			     struct bareudp_dev *bareudp,
356 			     const struct ip_tunnel_info *info)
357 {
358 	bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
359 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
360 	struct socket *sock  = rcu_dereference(bareudp->sock);
361 	bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
362 	const struct ip_tunnel_key *key = &info->key;
363 	struct dst_entry *dst = NULL;
364 	struct in6_addr saddr, daddr;
365 	int min_headroom;
366 	__u8 prio, ttl;
367 	__be16 sport;
368 	int err;
369 
370 	if (!sock)
371 		return -ESHUTDOWN;
372 
373 	dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock, &saddr, info,
374 				    IPPROTO_UDP, use_cache);
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 
380 	sport = udp_flow_src_port(bareudp->net, skb,
381 				  bareudp->sport_min, USHRT_MAX,
382 				  true);
383 	prio = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
384 	ttl = key->ttl;
385 
386 	skb_scrub_packet(skb, xnet);
387 
388 	err = -ENOSPC;
389 	if (!skb_pull(skb, skb_network_offset(skb)))
390 		goto free_dst;
391 
392 	min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len +
393 		BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr);
394 
395 	err = skb_cow_head(skb, min_headroom);
396 	if (unlikely(err))
397 		goto free_dst;
398 
399 	err = udp_tunnel_handle_offloads(skb, udp_sum);
400 	if (err)
401 		goto free_dst;
402 
403 	daddr = info->key.u.ipv6.dst;
404 	udp_tunnel6_xmit_skb(dst, sock->sk, skb, dev,
405 			     &saddr, &daddr, prio, ttl,
406 			     info->key.label, sport, bareudp->port,
407 			     !(info->key.tun_flags & TUNNEL_CSUM));
408 	return 0;
409 
410 free_dst:
411 	dst_release(dst);
412 	return err;
413 }
414 #endif
415 
416 static netdev_tx_t bareudp_xmit(struct sk_buff *skb, struct net_device *dev)
417 {
418 	struct bareudp_dev *bareudp = netdev_priv(dev);
419 	struct ip_tunnel_info *info = NULL;
420 	int err;
421 
422 	if (skb->protocol != bareudp->ethertype) {
423 		if (!bareudp->multi_proto_mode ||
424 		    (skb->protocol !=  htons(ETH_P_MPLS_MC) &&
425 		     skb->protocol !=  htons(ETH_P_IPV6))) {
426 			err = -EINVAL;
427 			goto tx_error;
428 		}
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)
439 	if (info->mode & IP_TUNNEL_INFO_IPV6)
440 		err = bareudp6_xmit_skb(skb, dev, bareudp, info);
441 	else
442 #endif
443 		err = bareudp_xmit_skb(skb, dev, bareudp, info);
444 
445 	rcu_read_unlock();
446 
447 	if (likely(!err))
448 		return NETDEV_TX_OK;
449 tx_error:
450 	dev_kfree_skb(skb);
451 
452 	if (err == -ELOOP)
453 		dev->stats.collisions++;
454 	else if (err == -ENETUNREACH)
455 		dev->stats.tx_carrier_errors++;
456 
457 	dev->stats.tx_errors++;
458 	return NETDEV_TX_OK;
459 }
460 
461 static int bareudp_fill_metadata_dst(struct net_device *dev,
462 				     struct sk_buff *skb)
463 {
464 	struct ip_tunnel_info *info = skb_tunnel_info(skb);
465 	struct bareudp_dev *bareudp = netdev_priv(dev);
466 	bool use_cache;
467 
468 	use_cache = ip_tunnel_dst_cache_usable(skb, info);
469 
470 	if (ip_tunnel_info_af(info) == AF_INET) {
471 		struct rtable *rt;
472 		__be32 saddr;
473 
474 		rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr,
475 					    info, IPPROTO_UDP, use_cache);
476 		if (IS_ERR(rt))
477 			return PTR_ERR(rt);
478 
479 		ip_rt_put(rt);
480 		info->key.u.ipv4.src = saddr;
481 #if IS_ENABLED(CONFIG_IPV6)
482 	} else if (ip_tunnel_info_af(info) == AF_INET6) {
483 		struct dst_entry *dst;
484 		struct in6_addr saddr;
485 		struct socket *sock = rcu_dereference(bareudp->sock);
486 
487 		dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock,
488 					    &saddr, info, IPPROTO_UDP,
489 					    use_cache);
490 		if (IS_ERR(dst))
491 			return PTR_ERR(dst);
492 
493 		dst_release(dst);
494 		info->key.u.ipv6.src = saddr;
495 #endif
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        = ip_tunnel_get_stats64,
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 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_GSO_SOFTWARE;
538 	dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
539 	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
540 	dev->hard_header_len = 0;
541 	dev->addr_len = 0;
542 	dev->mtu = ETH_DATA_LEN;
543 	dev->min_mtu = IPV4_MIN_MTU;
544 	dev->max_mtu = IP_MAX_MTU - BAREUDP_BASE_HLEN;
545 	dev->type = ARPHRD_NONE;
546 	netif_keep_dst(dev);
547 	dev->priv_flags |= IFF_NO_QUEUE;
548 	dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
549 }
550 
551 static int bareudp_validate(struct nlattr *tb[], struct nlattr *data[],
552 			    struct netlink_ext_ack *extack)
553 {
554 	if (!data) {
555 		NL_SET_ERR_MSG(extack,
556 			       "Not enough attributes provided to perform the operation");
557 		return -EINVAL;
558 	}
559 	return 0;
560 }
561 
562 static int bareudp2info(struct nlattr *data[], struct bareudp_conf *conf,
563 			struct netlink_ext_ack *extack)
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 	return 0;
584 }
585 
586 static struct bareudp_dev *bareudp_find_dev(struct bareudp_net *bn,
587 					    const struct bareudp_conf *conf)
588 {
589 	struct bareudp_dev *bareudp, *t = NULL;
590 
591 	list_for_each_entry(bareudp, &bn->bareudp_list, next) {
592 		if (conf->port == bareudp->port)
593 			t = bareudp;
594 	}
595 	return t;
596 }
597 
598 static int bareudp_configure(struct net *net, struct net_device *dev,
599 			     struct bareudp_conf *conf)
600 {
601 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
602 	struct bareudp_dev *t, *bareudp = netdev_priv(dev);
603 	int err;
604 
605 	bareudp->net = net;
606 	bareudp->dev = dev;
607 	t = bareudp_find_dev(bn, conf);
608 	if (t)
609 		return -EBUSY;
610 
611 	if (conf->multi_proto_mode &&
612 	    (conf->ethertype != htons(ETH_P_MPLS_UC) &&
613 	     conf->ethertype != htons(ETH_P_IP)))
614 		return -EINVAL;
615 
616 	bareudp->port = conf->port;
617 	bareudp->ethertype = conf->ethertype;
618 	bareudp->sport_min = conf->sport_min;
619 	bareudp->multi_proto_mode = conf->multi_proto_mode;
620 	err = register_netdevice(dev);
621 	if (err)
622 		return err;
623 
624 	list_add(&bareudp->next, &bn->bareudp_list);
625 	return 0;
626 }
627 
628 static int bareudp_link_config(struct net_device *dev,
629 			       struct nlattr *tb[])
630 {
631 	int err;
632 
633 	if (tb[IFLA_MTU]) {
634 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
635 		if (err)
636 			return err;
637 	}
638 	return 0;
639 }
640 
641 static int bareudp_newlink(struct net *net, struct net_device *dev,
642 			   struct nlattr *tb[], struct nlattr *data[],
643 			   struct netlink_ext_ack *extack)
644 {
645 	struct bareudp_conf conf;
646 	int err;
647 
648 	err = bareudp2info(data, &conf, extack);
649 	if (err)
650 		return err;
651 
652 	err = bareudp_configure(net, dev, &conf);
653 	if (err)
654 		return err;
655 
656 	err = bareudp_link_config(dev, tb);
657 	if (err)
658 		return err;
659 
660 	return 0;
661 }
662 
663 static void bareudp_dellink(struct net_device *dev, struct list_head *head)
664 {
665 	struct bareudp_dev *bareudp = netdev_priv(dev);
666 
667 	list_del(&bareudp->next);
668 	unregister_netdevice_queue(dev, head);
669 }
670 
671 static size_t bareudp_get_size(const struct net_device *dev)
672 {
673 	return  nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_PORT */
674 		nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_ETHERTYPE */
675 		nla_total_size(sizeof(__u16))  +  /* IFLA_BAREUDP_SRCPORT_MIN */
676 		nla_total_size(0)              +  /* IFLA_BAREUDP_MULTIPROTO_MODE */
677 		0;
678 }
679 
680 static int bareudp_fill_info(struct sk_buff *skb, const struct net_device *dev)
681 {
682 	struct bareudp_dev *bareudp = netdev_priv(dev);
683 
684 	if (nla_put_be16(skb, IFLA_BAREUDP_PORT, bareudp->port))
685 		goto nla_put_failure;
686 	if (nla_put_be16(skb, IFLA_BAREUDP_ETHERTYPE, bareudp->ethertype))
687 		goto nla_put_failure;
688 	if (nla_put_u16(skb, IFLA_BAREUDP_SRCPORT_MIN, bareudp->sport_min))
689 		goto nla_put_failure;
690 	if (bareudp->multi_proto_mode &&
691 	    nla_put_flag(skb, IFLA_BAREUDP_MULTIPROTO_MODE))
692 		goto nla_put_failure;
693 
694 	return 0;
695 
696 nla_put_failure:
697 	return -EMSGSIZE;
698 }
699 
700 static struct rtnl_link_ops bareudp_link_ops __read_mostly = {
701 	.kind           = "bareudp",
702 	.maxtype        = IFLA_BAREUDP_MAX,
703 	.policy         = bareudp_policy,
704 	.priv_size      = sizeof(struct bareudp_dev),
705 	.setup          = bareudp_setup,
706 	.validate       = bareudp_validate,
707 	.newlink        = bareudp_newlink,
708 	.dellink        = bareudp_dellink,
709 	.get_size       = bareudp_get_size,
710 	.fill_info      = bareudp_fill_info,
711 };
712 
713 struct net_device *bareudp_dev_create(struct net *net, const char *name,
714 				      u8 name_assign_type,
715 				      struct bareudp_conf *conf)
716 {
717 	struct nlattr *tb[IFLA_MAX + 1];
718 	struct net_device *dev;
719 	LIST_HEAD(list_kill);
720 	int err;
721 
722 	memset(tb, 0, sizeof(tb));
723 	dev = rtnl_create_link(net, name, name_assign_type,
724 			       &bareudp_link_ops, tb, NULL);
725 	if (IS_ERR(dev))
726 		return dev;
727 
728 	err = bareudp_configure(net, dev, conf);
729 	if (err) {
730 		free_netdev(dev);
731 		return ERR_PTR(err);
732 	}
733 	err = dev_set_mtu(dev, IP_MAX_MTU - BAREUDP_BASE_HLEN);
734 	if (err)
735 		goto err;
736 
737 	err = rtnl_configure_link(dev, NULL);
738 	if (err < 0)
739 		goto err;
740 
741 	return dev;
742 err:
743 	bareudp_dellink(dev, &list_kill);
744 	unregister_netdevice_many(&list_kill);
745 	return ERR_PTR(err);
746 }
747 EXPORT_SYMBOL_GPL(bareudp_dev_create);
748 
749 static __net_init int bareudp_init_net(struct net *net)
750 {
751 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
752 
753 	INIT_LIST_HEAD(&bn->bareudp_list);
754 	return 0;
755 }
756 
757 static void bareudp_destroy_tunnels(struct net *net, struct list_head *head)
758 {
759 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
760 	struct bareudp_dev *bareudp, *next;
761 
762 	list_for_each_entry_safe(bareudp, next, &bn->bareudp_list, next)
763 		unregister_netdevice_queue(bareudp->dev, head);
764 }
765 
766 static void __net_exit bareudp_exit_batch_net(struct list_head *net_list)
767 {
768 	struct net *net;
769 	LIST_HEAD(list);
770 
771 	rtnl_lock();
772 	list_for_each_entry(net, net_list, exit_list)
773 		bareudp_destroy_tunnels(net, &list);
774 
775 	/* unregister the devices gathered above */
776 	unregister_netdevice_many(&list);
777 	rtnl_unlock();
778 }
779 
780 static struct pernet_operations bareudp_net_ops = {
781 	.init = bareudp_init_net,
782 	.exit_batch = bareudp_exit_batch_net,
783 	.id   = &bareudp_net_id,
784 	.size = sizeof(struct bareudp_net),
785 };
786 
787 static int __init bareudp_init_module(void)
788 {
789 	int rc;
790 
791 	rc = register_pernet_subsys(&bareudp_net_ops);
792 	if (rc)
793 		goto out1;
794 
795 	rc = rtnl_link_register(&bareudp_link_ops);
796 	if (rc)
797 		goto out2;
798 
799 	return 0;
800 out2:
801 	unregister_pernet_subsys(&bareudp_net_ops);
802 out1:
803 	return rc;
804 }
805 late_initcall(bareudp_init_module);
806 
807 static void __exit bareudp_cleanup_module(void)
808 {
809 	rtnl_link_unregister(&bareudp_link_ops);
810 	unregister_pernet_subsys(&bareudp_net_ops);
811 }
812 module_exit(bareudp_cleanup_module);
813 
814 MODULE_ALIAS_RTNL_LINK("bareudp");
815 MODULE_LICENSE("GPL");
816 MODULE_AUTHOR("Martin Varghese <martin.varghese@nokia.com>");
817 MODULE_DESCRIPTION("Interface driver for UDP encapsulated traffic");
818