xref: /openbmc/linux/drivers/net/bareudp.c (revision ee7da21a)
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 		__u8 ipversion;
75 
76 		if (skb_copy_bits(skb, BAREUDP_BASE_HLEN, &ipversion,
77 				  sizeof(ipversion))) {
78 			bareudp->dev->stats.rx_dropped++;
79 			goto drop;
80 		}
81 		ipversion >>= 4;
82 
83 		if (ipversion == 4) {
84 			proto = htons(ETH_P_IP);
85 		} else if (ipversion == 6 && bareudp->multi_proto_mode) {
86 			proto = htons(ETH_P_IPV6);
87 		} else {
88 			bareudp->dev->stats.rx_dropped++;
89 			goto drop;
90 		}
91 	} else if (bareudp->ethertype == htons(ETH_P_MPLS_UC)) {
92 		struct iphdr *tunnel_hdr;
93 
94 		tunnel_hdr = (struct iphdr *)skb_network_header(skb);
95 		if (tunnel_hdr->version == 4) {
96 			if (!ipv4_is_multicast(tunnel_hdr->daddr)) {
97 				proto = bareudp->ethertype;
98 			} else if (bareudp->multi_proto_mode &&
99 				   ipv4_is_multicast(tunnel_hdr->daddr)) {
100 				proto = htons(ETH_P_MPLS_MC);
101 			} else {
102 				bareudp->dev->stats.rx_dropped++;
103 				goto drop;
104 			}
105 		} else {
106 			int addr_type;
107 			struct ipv6hdr *tunnel_hdr_v6;
108 
109 			tunnel_hdr_v6 = (struct ipv6hdr *)skb_network_header(skb);
110 			addr_type =
111 			ipv6_addr_type((struct in6_addr *)&tunnel_hdr_v6->daddr);
112 			if (!(addr_type & IPV6_ADDR_MULTICAST)) {
113 				proto = bareudp->ethertype;
114 			} else if (bareudp->multi_proto_mode &&
115 				   (addr_type & IPV6_ADDR_MULTICAST)) {
116 				proto = htons(ETH_P_MPLS_MC);
117 			} else {
118 				bareudp->dev->stats.rx_dropped++;
119 				goto drop;
120 			}
121 		}
122 	} else {
123 		proto = bareudp->ethertype;
124 	}
125 
126 	if (iptunnel_pull_header(skb, BAREUDP_BASE_HLEN,
127 				 proto,
128 				 !net_eq(bareudp->net,
129 				 dev_net(bareudp->dev)))) {
130 		bareudp->dev->stats.rx_dropped++;
131 		goto drop;
132 	}
133 	tun_dst = udp_tun_rx_dst(skb, family, TUNNEL_KEY, 0, 0);
134 	if (!tun_dst) {
135 		bareudp->dev->stats.rx_dropped++;
136 		goto drop;
137 	}
138 	skb_dst_set(skb, &tun_dst->dst);
139 	skb->dev = bareudp->dev;
140 	oiph = skb_network_header(skb);
141 	skb_reset_network_header(skb);
142 	skb_reset_mac_header(skb);
143 
144 	if (!IS_ENABLED(CONFIG_IPV6) || family == AF_INET)
145 		err = IP_ECN_decapsulate(oiph, skb);
146 	else
147 		err = IP6_ECN_decapsulate(oiph, skb);
148 
149 	if (unlikely(err)) {
150 		if (log_ecn_error) {
151 			if  (!IS_ENABLED(CONFIG_IPV6) || family == AF_INET)
152 				net_info_ratelimited("non-ECT from %pI4 "
153 						     "with TOS=%#x\n",
154 						     &((struct iphdr *)oiph)->saddr,
155 						     ((struct iphdr *)oiph)->tos);
156 			else
157 				net_info_ratelimited("non-ECT from %pI6\n",
158 						     &((struct ipv6hdr *)oiph)->saddr);
159 		}
160 		if (err > 1) {
161 			++bareudp->dev->stats.rx_frame_errors;
162 			++bareudp->dev->stats.rx_errors;
163 			goto drop;
164 		}
165 	}
166 
167 	len = skb->len;
168 	err = gro_cells_receive(&bareudp->gro_cells, skb);
169 	if (likely(err == NET_RX_SUCCESS))
170 		dev_sw_netstats_rx_add(bareudp->dev, len);
171 
172 	return 0;
173 drop:
174 	/* Consume bad packet */
175 	kfree_skb(skb);
176 
177 	return 0;
178 }
179 
180 static int bareudp_err_lookup(struct sock *sk, struct sk_buff *skb)
181 {
182 	return 0;
183 }
184 
185 static int bareudp_init(struct net_device *dev)
186 {
187 	struct bareudp_dev *bareudp = netdev_priv(dev);
188 	int err;
189 
190 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
191 	if (!dev->tstats)
192 		return -ENOMEM;
193 
194 	err = gro_cells_init(&bareudp->gro_cells, dev);
195 	if (err) {
196 		free_percpu(dev->tstats);
197 		return err;
198 	}
199 	return 0;
200 }
201 
202 static void bareudp_uninit(struct net_device *dev)
203 {
204 	struct bareudp_dev *bareudp = netdev_priv(dev);
205 
206 	gro_cells_destroy(&bareudp->gro_cells);
207 	free_percpu(dev->tstats);
208 }
209 
210 static struct socket *bareudp_create_sock(struct net *net, __be16 port)
211 {
212 	struct udp_port_cfg udp_conf;
213 	struct socket *sock;
214 	int err;
215 
216 	memset(&udp_conf, 0, sizeof(udp_conf));
217 #if IS_ENABLED(CONFIG_IPV6)
218 	udp_conf.family = AF_INET6;
219 #else
220 	udp_conf.family = AF_INET;
221 #endif
222 	udp_conf.local_udp_port = port;
223 	/* Open UDP socket */
224 	err = udp_sock_create(net, &udp_conf, &sock);
225 	if (err < 0)
226 		return ERR_PTR(err);
227 
228 	udp_allow_gso(sock->sk);
229 	return sock;
230 }
231 
232 /* Create new listen socket if needed */
233 static int bareudp_socket_create(struct bareudp_dev *bareudp, __be16 port)
234 {
235 	struct udp_tunnel_sock_cfg tunnel_cfg;
236 	struct socket *sock;
237 
238 	sock = bareudp_create_sock(bareudp->net, port);
239 	if (IS_ERR(sock))
240 		return PTR_ERR(sock);
241 
242 	/* Mark socket as an encapsulation socket */
243 	memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
244 	tunnel_cfg.sk_user_data = bareudp;
245 	tunnel_cfg.encap_type = 1;
246 	tunnel_cfg.encap_rcv = bareudp_udp_encap_recv;
247 	tunnel_cfg.encap_err_lookup = bareudp_err_lookup;
248 	tunnel_cfg.encap_destroy = NULL;
249 	setup_udp_tunnel_sock(bareudp->net, sock, &tunnel_cfg);
250 
251 	rcu_assign_pointer(bareudp->sock, sock);
252 	return 0;
253 }
254 
255 static int bareudp_open(struct net_device *dev)
256 {
257 	struct bareudp_dev *bareudp = netdev_priv(dev);
258 	int ret = 0;
259 
260 	ret =  bareudp_socket_create(bareudp, bareudp->port);
261 	return ret;
262 }
263 
264 static void bareudp_sock_release(struct bareudp_dev *bareudp)
265 {
266 	struct socket *sock;
267 
268 	sock = bareudp->sock;
269 	rcu_assign_pointer(bareudp->sock, NULL);
270 	synchronize_net();
271 	udp_tunnel_sock_release(sock);
272 }
273 
274 static int bareudp_stop(struct net_device *dev)
275 {
276 	struct bareudp_dev *bareudp = netdev_priv(dev);
277 
278 	bareudp_sock_release(bareudp);
279 	return 0;
280 }
281 
282 static int bareudp_xmit_skb(struct sk_buff *skb, struct net_device *dev,
283 			    struct bareudp_dev *bareudp,
284 			    const struct ip_tunnel_info *info)
285 {
286 	bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
287 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
288 	struct socket *sock = rcu_dereference(bareudp->sock);
289 	bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
290 	const struct ip_tunnel_key *key = &info->key;
291 	struct rtable *rt;
292 	__be16 sport, df;
293 	int min_headroom;
294 	__u8 tos, ttl;
295 	__be32 saddr;
296 	int err;
297 
298 	if (!sock)
299 		return -ESHUTDOWN;
300 
301 	rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr, info,
302 				    IPPROTO_UDP, use_cache);
303 
304 	if (IS_ERR(rt))
305 		return PTR_ERR(rt);
306 
307 	skb_tunnel_check_pmtu(skb, &rt->dst,
308 			      BAREUDP_IPV4_HLEN + info->options_len, false);
309 
310 	sport = udp_flow_src_port(bareudp->net, skb,
311 				  bareudp->sport_min, USHRT_MAX,
312 				  true);
313 	tos = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
314 	ttl = key->ttl;
315 	df = key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
316 	skb_scrub_packet(skb, xnet);
317 
318 	err = -ENOSPC;
319 	if (!skb_pull(skb, skb_network_offset(skb)))
320 		goto free_dst;
321 
322 	min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len +
323 		BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr);
324 
325 	err = skb_cow_head(skb, min_headroom);
326 	if (unlikely(err))
327 		goto free_dst;
328 
329 	err = udp_tunnel_handle_offloads(skb, udp_sum);
330 	if (err)
331 		goto free_dst;
332 
333 	skb_set_inner_protocol(skb, bareudp->ethertype);
334 	udp_tunnel_xmit_skb(rt, sock->sk, skb, saddr, info->key.u.ipv4.dst,
335 			    tos, ttl, df, sport, bareudp->port,
336 			    !net_eq(bareudp->net, dev_net(bareudp->dev)),
337 			    !(info->key.tun_flags & TUNNEL_CSUM));
338 	return 0;
339 
340 free_dst:
341 	dst_release(&rt->dst);
342 	return err;
343 }
344 
345 static int bareudp6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
346 			     struct bareudp_dev *bareudp,
347 			     const struct ip_tunnel_info *info)
348 {
349 	bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
350 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
351 	struct socket *sock  = rcu_dereference(bareudp->sock);
352 	bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
353 	const struct ip_tunnel_key *key = &info->key;
354 	struct dst_entry *dst = NULL;
355 	struct in6_addr saddr, daddr;
356 	int min_headroom;
357 	__u8 prio, ttl;
358 	__be16 sport;
359 	int err;
360 
361 	if (!sock)
362 		return -ESHUTDOWN;
363 
364 	dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock, &saddr, info,
365 				    IPPROTO_UDP, use_cache);
366 	if (IS_ERR(dst))
367 		return PTR_ERR(dst);
368 
369 	skb_tunnel_check_pmtu(skb, dst, BAREUDP_IPV6_HLEN + info->options_len,
370 			      false);
371 
372 	sport = udp_flow_src_port(bareudp->net, skb,
373 				  bareudp->sport_min, USHRT_MAX,
374 				  true);
375 	prio = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
376 	ttl = key->ttl;
377 
378 	skb_scrub_packet(skb, xnet);
379 
380 	err = -ENOSPC;
381 	if (!skb_pull(skb, skb_network_offset(skb)))
382 		goto free_dst;
383 
384 	min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len +
385 		BAREUDP_BASE_HLEN + info->options_len + sizeof(struct ipv6hdr);
386 
387 	err = skb_cow_head(skb, min_headroom);
388 	if (unlikely(err))
389 		goto free_dst;
390 
391 	err = udp_tunnel_handle_offloads(skb, udp_sum);
392 	if (err)
393 		goto free_dst;
394 
395 	daddr = info->key.u.ipv6.dst;
396 	udp_tunnel6_xmit_skb(dst, sock->sk, skb, dev,
397 			     &saddr, &daddr, prio, ttl,
398 			     info->key.label, sport, bareudp->port,
399 			     !(info->key.tun_flags & TUNNEL_CSUM));
400 	return 0;
401 
402 free_dst:
403 	dst_release(dst);
404 	return err;
405 }
406 
407 static bool bareudp_proto_valid(struct bareudp_dev *bareudp, __be16 proto)
408 {
409 	if (bareudp->ethertype == proto)
410 		return true;
411 
412 	if (!bareudp->multi_proto_mode)
413 		return false;
414 
415 	if (bareudp->ethertype == htons(ETH_P_MPLS_UC) &&
416 	    proto == htons(ETH_P_MPLS_MC))
417 		return true;
418 
419 	if (bareudp->ethertype == htons(ETH_P_IP) &&
420 	    proto == htons(ETH_P_IPV6))
421 		return true;
422 
423 	return false;
424 }
425 
426 static netdev_tx_t bareudp_xmit(struct sk_buff *skb, struct net_device *dev)
427 {
428 	struct bareudp_dev *bareudp = netdev_priv(dev);
429 	struct ip_tunnel_info *info = NULL;
430 	int err;
431 
432 	if (!bareudp_proto_valid(bareudp, skb->protocol)) {
433 		err = -EINVAL;
434 		goto tx_error;
435 	}
436 
437 	info = skb_tunnel_info(skb);
438 	if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) {
439 		err = -EINVAL;
440 		goto tx_error;
441 	}
442 
443 	rcu_read_lock();
444 	if (IS_ENABLED(CONFIG_IPV6) && info->mode & IP_TUNNEL_INFO_IPV6)
445 		err = bareudp6_xmit_skb(skb, dev, bareudp, info);
446 	else
447 		err = bareudp_xmit_skb(skb, dev, bareudp, info);
448 
449 	rcu_read_unlock();
450 
451 	if (likely(!err))
452 		return NETDEV_TX_OK;
453 tx_error:
454 	dev_kfree_skb(skb);
455 
456 	if (err == -ELOOP)
457 		dev->stats.collisions++;
458 	else if (err == -ENETUNREACH)
459 		dev->stats.tx_carrier_errors++;
460 
461 	dev->stats.tx_errors++;
462 	return NETDEV_TX_OK;
463 }
464 
465 static int bareudp_fill_metadata_dst(struct net_device *dev,
466 				     struct sk_buff *skb)
467 {
468 	struct ip_tunnel_info *info = skb_tunnel_info(skb);
469 	struct bareudp_dev *bareudp = netdev_priv(dev);
470 	bool use_cache;
471 
472 	use_cache = ip_tunnel_dst_cache_usable(skb, info);
473 
474 	if (!IS_ENABLED(CONFIG_IPV6) || ip_tunnel_info_af(info) == AF_INET) {
475 		struct rtable *rt;
476 		__be32 saddr;
477 
478 		rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr,
479 					    info, IPPROTO_UDP, use_cache);
480 		if (IS_ERR(rt))
481 			return PTR_ERR(rt);
482 
483 		ip_rt_put(rt);
484 		info->key.u.ipv4.src = saddr;
485 	} else if (ip_tunnel_info_af(info) == AF_INET6) {
486 		struct dst_entry *dst;
487 		struct in6_addr saddr;
488 		struct socket *sock = rcu_dereference(bareudp->sock);
489 
490 		dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock,
491 					    &saddr, info, IPPROTO_UDP,
492 					    use_cache);
493 		if (IS_ERR(dst))
494 			return PTR_ERR(dst);
495 
496 		dst_release(dst);
497 		info->key.u.ipv6.src = saddr;
498 	} else {
499 		return -EINVAL;
500 	}
501 
502 	info->key.tp_src = udp_flow_src_port(bareudp->net, skb,
503 					     bareudp->sport_min,
504 			USHRT_MAX, true);
505 	info->key.tp_dst = bareudp->port;
506 	return 0;
507 }
508 
509 static const struct net_device_ops bareudp_netdev_ops = {
510 	.ndo_init               = bareudp_init,
511 	.ndo_uninit             = bareudp_uninit,
512 	.ndo_open               = bareudp_open,
513 	.ndo_stop               = bareudp_stop,
514 	.ndo_start_xmit         = bareudp_xmit,
515 	.ndo_get_stats64        = dev_get_tstats64,
516 	.ndo_fill_metadata_dst  = bareudp_fill_metadata_dst,
517 };
518 
519 static const struct nla_policy bareudp_policy[IFLA_BAREUDP_MAX + 1] = {
520 	[IFLA_BAREUDP_PORT]                = { .type = NLA_U16 },
521 	[IFLA_BAREUDP_ETHERTYPE]	   = { .type = NLA_U16 },
522 	[IFLA_BAREUDP_SRCPORT_MIN]         = { .type = NLA_U16 },
523 	[IFLA_BAREUDP_MULTIPROTO_MODE]     = { .type = NLA_FLAG },
524 };
525 
526 /* Info for udev, that this is a virtual tunnel endpoint */
527 static const struct device_type bareudp_type = {
528 	.name = "bareudp",
529 };
530 
531 /* Initialize the device structure. */
532 static void bareudp_setup(struct net_device *dev)
533 {
534 	dev->netdev_ops = &bareudp_netdev_ops;
535 	dev->needs_free_netdev = true;
536 	SET_NETDEV_DEVTYPE(dev, &bareudp_type);
537 	dev->features    |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
538 	dev->features    |= NETIF_F_RXCSUM;
539 	dev->features    |= NETIF_F_LLTX;
540 	dev->features    |= NETIF_F_GSO_SOFTWARE;
541 	dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
542 	dev->hw_features |= NETIF_F_RXCSUM;
543 	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
544 	dev->hard_header_len = 0;
545 	dev->addr_len = 0;
546 	dev->mtu = ETH_DATA_LEN;
547 	dev->min_mtu = IPV4_MIN_MTU;
548 	dev->max_mtu = IP_MAX_MTU - BAREUDP_BASE_HLEN;
549 	dev->type = ARPHRD_NONE;
550 	netif_keep_dst(dev);
551 	dev->priv_flags |= IFF_NO_QUEUE;
552 	dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
553 }
554 
555 static int bareudp_validate(struct nlattr *tb[], struct nlattr *data[],
556 			    struct netlink_ext_ack *extack)
557 {
558 	if (!data) {
559 		NL_SET_ERR_MSG(extack,
560 			       "Not enough attributes provided to perform the operation");
561 		return -EINVAL;
562 	}
563 	return 0;
564 }
565 
566 static int bareudp2info(struct nlattr *data[], struct bareudp_conf *conf,
567 			struct netlink_ext_ack *extack)
568 {
569 	memset(conf, 0, sizeof(*conf));
570 
571 	if (!data[IFLA_BAREUDP_PORT]) {
572 		NL_SET_ERR_MSG(extack, "port not specified");
573 		return -EINVAL;
574 	}
575 	if (!data[IFLA_BAREUDP_ETHERTYPE]) {
576 		NL_SET_ERR_MSG(extack, "ethertype not specified");
577 		return -EINVAL;
578 	}
579 
580 	if (data[IFLA_BAREUDP_PORT])
581 		conf->port =  nla_get_u16(data[IFLA_BAREUDP_PORT]);
582 
583 	if (data[IFLA_BAREUDP_ETHERTYPE])
584 		conf->ethertype =  nla_get_u16(data[IFLA_BAREUDP_ETHERTYPE]);
585 
586 	if (data[IFLA_BAREUDP_SRCPORT_MIN])
587 		conf->sport_min =  nla_get_u16(data[IFLA_BAREUDP_SRCPORT_MIN]);
588 
589 	if (data[IFLA_BAREUDP_MULTIPROTO_MODE])
590 		conf->multi_proto_mode = true;
591 
592 	return 0;
593 }
594 
595 static struct bareudp_dev *bareudp_find_dev(struct bareudp_net *bn,
596 					    const struct bareudp_conf *conf)
597 {
598 	struct bareudp_dev *bareudp, *t = NULL;
599 
600 	list_for_each_entry(bareudp, &bn->bareudp_list, next) {
601 		if (conf->port == bareudp->port)
602 			t = bareudp;
603 	}
604 	return t;
605 }
606 
607 static int bareudp_configure(struct net *net, struct net_device *dev,
608 			     struct bareudp_conf *conf)
609 {
610 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
611 	struct bareudp_dev *t, *bareudp = netdev_priv(dev);
612 	int err;
613 
614 	bareudp->net = net;
615 	bareudp->dev = dev;
616 	t = bareudp_find_dev(bn, conf);
617 	if (t)
618 		return -EBUSY;
619 
620 	if (conf->multi_proto_mode &&
621 	    (conf->ethertype != htons(ETH_P_MPLS_UC) &&
622 	     conf->ethertype != htons(ETH_P_IP)))
623 		return -EINVAL;
624 
625 	bareudp->port = conf->port;
626 	bareudp->ethertype = conf->ethertype;
627 	bareudp->sport_min = conf->sport_min;
628 	bareudp->multi_proto_mode = conf->multi_proto_mode;
629 
630 	err = register_netdevice(dev);
631 	if (err)
632 		return err;
633 
634 	list_add(&bareudp->next, &bn->bareudp_list);
635 	return 0;
636 }
637 
638 static int bareudp_link_config(struct net_device *dev,
639 			       struct nlattr *tb[])
640 {
641 	int err;
642 
643 	if (tb[IFLA_MTU]) {
644 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
645 		if (err)
646 			return err;
647 	}
648 	return 0;
649 }
650 
651 static void bareudp_dellink(struct net_device *dev, struct list_head *head)
652 {
653 	struct bareudp_dev *bareudp = netdev_priv(dev);
654 
655 	list_del(&bareudp->next);
656 	unregister_netdevice_queue(dev, head);
657 }
658 
659 static int bareudp_newlink(struct net *net, struct net_device *dev,
660 			   struct nlattr *tb[], struct nlattr *data[],
661 			   struct netlink_ext_ack *extack)
662 {
663 	struct bareudp_conf conf;
664 	int err;
665 
666 	err = bareudp2info(data, &conf, extack);
667 	if (err)
668 		return err;
669 
670 	err = bareudp_configure(net, dev, &conf);
671 	if (err)
672 		return err;
673 
674 	err = bareudp_link_config(dev, tb);
675 	if (err)
676 		goto err_unconfig;
677 
678 	return 0;
679 
680 err_unconfig:
681 	bareudp_dellink(dev, NULL);
682 	return err;
683 }
684 
685 static size_t bareudp_get_size(const struct net_device *dev)
686 {
687 	return  nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_PORT */
688 		nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_ETHERTYPE */
689 		nla_total_size(sizeof(__u16))  +  /* IFLA_BAREUDP_SRCPORT_MIN */
690 		nla_total_size(0)              +  /* IFLA_BAREUDP_MULTIPROTO_MODE */
691 		0;
692 }
693 
694 static int bareudp_fill_info(struct sk_buff *skb, const struct net_device *dev)
695 {
696 	struct bareudp_dev *bareudp = netdev_priv(dev);
697 
698 	if (nla_put_be16(skb, IFLA_BAREUDP_PORT, bareudp->port))
699 		goto nla_put_failure;
700 	if (nla_put_be16(skb, IFLA_BAREUDP_ETHERTYPE, bareudp->ethertype))
701 		goto nla_put_failure;
702 	if (nla_put_u16(skb, IFLA_BAREUDP_SRCPORT_MIN, bareudp->sport_min))
703 		goto nla_put_failure;
704 	if (bareudp->multi_proto_mode &&
705 	    nla_put_flag(skb, IFLA_BAREUDP_MULTIPROTO_MODE))
706 		goto nla_put_failure;
707 
708 	return 0;
709 
710 nla_put_failure:
711 	return -EMSGSIZE;
712 }
713 
714 static struct rtnl_link_ops bareudp_link_ops __read_mostly = {
715 	.kind           = "bareudp",
716 	.maxtype        = IFLA_BAREUDP_MAX,
717 	.policy         = bareudp_policy,
718 	.priv_size      = sizeof(struct bareudp_dev),
719 	.setup          = bareudp_setup,
720 	.validate       = bareudp_validate,
721 	.newlink        = bareudp_newlink,
722 	.dellink        = bareudp_dellink,
723 	.get_size       = bareudp_get_size,
724 	.fill_info      = bareudp_fill_info,
725 };
726 
727 struct net_device *bareudp_dev_create(struct net *net, const char *name,
728 				      u8 name_assign_type,
729 				      struct bareudp_conf *conf)
730 {
731 	struct nlattr *tb[IFLA_MAX + 1];
732 	struct net_device *dev;
733 	int err;
734 
735 	memset(tb, 0, sizeof(tb));
736 	dev = rtnl_create_link(net, name, name_assign_type,
737 			       &bareudp_link_ops, tb, NULL);
738 	if (IS_ERR(dev))
739 		return dev;
740 
741 	err = bareudp_configure(net, dev, conf);
742 	if (err) {
743 		free_netdev(dev);
744 		return ERR_PTR(err);
745 	}
746 	err = dev_set_mtu(dev, IP_MAX_MTU - BAREUDP_BASE_HLEN);
747 	if (err)
748 		goto err;
749 
750 	err = rtnl_configure_link(dev, NULL);
751 	if (err < 0)
752 		goto err;
753 
754 	return dev;
755 err:
756 	bareudp_dellink(dev, NULL);
757 	return ERR_PTR(err);
758 }
759 EXPORT_SYMBOL_GPL(bareudp_dev_create);
760 
761 static __net_init int bareudp_init_net(struct net *net)
762 {
763 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
764 
765 	INIT_LIST_HEAD(&bn->bareudp_list);
766 	return 0;
767 }
768 
769 static void bareudp_destroy_tunnels(struct net *net, struct list_head *head)
770 {
771 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
772 	struct bareudp_dev *bareudp, *next;
773 
774 	list_for_each_entry_safe(bareudp, next, &bn->bareudp_list, next)
775 		unregister_netdevice_queue(bareudp->dev, head);
776 }
777 
778 static void __net_exit bareudp_exit_batch_net(struct list_head *net_list)
779 {
780 	struct net *net;
781 	LIST_HEAD(list);
782 
783 	rtnl_lock();
784 	list_for_each_entry(net, net_list, exit_list)
785 		bareudp_destroy_tunnels(net, &list);
786 
787 	/* unregister the devices gathered above */
788 	unregister_netdevice_many(&list);
789 	rtnl_unlock();
790 }
791 
792 static struct pernet_operations bareudp_net_ops = {
793 	.init = bareudp_init_net,
794 	.exit_batch = bareudp_exit_batch_net,
795 	.id   = &bareudp_net_id,
796 	.size = sizeof(struct bareudp_net),
797 };
798 
799 static int __init bareudp_init_module(void)
800 {
801 	int rc;
802 
803 	rc = register_pernet_subsys(&bareudp_net_ops);
804 	if (rc)
805 		goto out1;
806 
807 	rc = rtnl_link_register(&bareudp_link_ops);
808 	if (rc)
809 		goto out2;
810 
811 	return 0;
812 out2:
813 	unregister_pernet_subsys(&bareudp_net_ops);
814 out1:
815 	return rc;
816 }
817 late_initcall(bareudp_init_module);
818 
819 static void __exit bareudp_cleanup_module(void)
820 {
821 	rtnl_link_unregister(&bareudp_link_ops);
822 	unregister_pernet_subsys(&bareudp_net_ops);
823 }
824 module_exit(bareudp_cleanup_module);
825 
826 MODULE_ALIAS_RTNL_LINK("bareudp");
827 MODULE_LICENSE("GPL");
828 MODULE_AUTHOR("Martin Varghese <martin.varghese@nokia.com>");
829 MODULE_DESCRIPTION("Interface driver for UDP encapsulated traffic");
830