xref: /openbmc/linux/drivers/net/bareudp.c (revision b868a02e)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Bareudp: UDP  tunnel encasulation for different Payload types like
3  * MPLS, NSH, IP, etc.
4  * Copyright (c) 2019 Nokia, Inc.
5  * Authors:  Martin Varghese, <martin.varghese@nokia.com>
6  */
7 
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9 
10 #include <linux/kernel.h>
11 #include <linux/module.h>
12 #include <linux/etherdevice.h>
13 #include <linux/hash.h>
14 #include <net/dst_metadata.h>
15 #include <net/gro_cells.h>
16 #include <net/rtnetlink.h>
17 #include <net/protocol.h>
18 #include <net/ip6_tunnel.h>
19 #include <net/ip_tunnels.h>
20 #include <net/udp_tunnel.h>
21 #include <net/bareudp.h>
22 
23 #define BAREUDP_BASE_HLEN sizeof(struct udphdr)
24 #define BAREUDP_IPV4_HLEN (sizeof(struct iphdr) + \
25 			   sizeof(struct udphdr))
26 #define BAREUDP_IPV6_HLEN (sizeof(struct ipv6hdr) + \
27 			   sizeof(struct udphdr))
28 
29 static bool log_ecn_error = true;
30 module_param(log_ecn_error, bool, 0644);
31 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
32 
33 /* per-network namespace private data for this module */
34 
35 static unsigned int bareudp_net_id;
36 
37 struct bareudp_net {
38 	struct list_head        bareudp_list;
39 };
40 
41 struct bareudp_conf {
42 	__be16 ethertype;
43 	__be16 port;
44 	u16 sport_min;
45 	bool multi_proto_mode;
46 };
47 
48 /* Pseudo network device */
49 struct bareudp_dev {
50 	struct net         *net;        /* netns for packet i/o */
51 	struct net_device  *dev;        /* netdev for bareudp tunnel */
52 	__be16		   ethertype;
53 	__be16             port;
54 	u16	           sport_min;
55 	bool               multi_proto_mode;
56 	struct socket      __rcu *sock;
57 	struct list_head   next;        /* bareudp node  on namespace list */
58 	struct gro_cells   gro_cells;
59 };
60 
61 static int bareudp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
62 {
63 	struct metadata_dst *tun_dst = NULL;
64 	struct bareudp_dev *bareudp;
65 	unsigned short family;
66 	unsigned int len;
67 	__be16 proto;
68 	void *oiph;
69 	int err;
70 
71 	bareudp = rcu_dereference_sk_user_data(sk);
72 	if (!bareudp)
73 		goto drop;
74 
75 	if (skb->protocol ==  htons(ETH_P_IP))
76 		family = AF_INET;
77 	else
78 		family = AF_INET6;
79 
80 	if (bareudp->ethertype == htons(ETH_P_IP)) {
81 		__u8 ipversion;
82 
83 		if (skb_copy_bits(skb, BAREUDP_BASE_HLEN, &ipversion,
84 				  sizeof(ipversion))) {
85 			bareudp->dev->stats.rx_dropped++;
86 			goto drop;
87 		}
88 		ipversion >>= 4;
89 
90 		if (ipversion == 4) {
91 			proto = htons(ETH_P_IP);
92 		} else if (ipversion == 6 && bareudp->multi_proto_mode) {
93 			proto = htons(ETH_P_IPV6);
94 		} else {
95 			bareudp->dev->stats.rx_dropped++;
96 			goto drop;
97 		}
98 	} else if (bareudp->ethertype == htons(ETH_P_MPLS_UC)) {
99 		struct iphdr *tunnel_hdr;
100 
101 		tunnel_hdr = (struct iphdr *)skb_network_header(skb);
102 		if (tunnel_hdr->version == 4) {
103 			if (!ipv4_is_multicast(tunnel_hdr->daddr)) {
104 				proto = bareudp->ethertype;
105 			} else if (bareudp->multi_proto_mode &&
106 				   ipv4_is_multicast(tunnel_hdr->daddr)) {
107 				proto = htons(ETH_P_MPLS_MC);
108 			} else {
109 				bareudp->dev->stats.rx_dropped++;
110 				goto drop;
111 			}
112 		} else {
113 			int addr_type;
114 			struct ipv6hdr *tunnel_hdr_v6;
115 
116 			tunnel_hdr_v6 = (struct ipv6hdr *)skb_network_header(skb);
117 			addr_type =
118 			ipv6_addr_type((struct in6_addr *)&tunnel_hdr_v6->daddr);
119 			if (!(addr_type & IPV6_ADDR_MULTICAST)) {
120 				proto = bareudp->ethertype;
121 			} else if (bareudp->multi_proto_mode &&
122 				   (addr_type & IPV6_ADDR_MULTICAST)) {
123 				proto = htons(ETH_P_MPLS_MC);
124 			} else {
125 				bareudp->dev->stats.rx_dropped++;
126 				goto drop;
127 			}
128 		}
129 	} else {
130 		proto = bareudp->ethertype;
131 	}
132 
133 	if (iptunnel_pull_header(skb, BAREUDP_BASE_HLEN,
134 				 proto,
135 				 !net_eq(bareudp->net,
136 				 dev_net(bareudp->dev)))) {
137 		bareudp->dev->stats.rx_dropped++;
138 		goto drop;
139 	}
140 	tun_dst = udp_tun_rx_dst(skb, family, TUNNEL_KEY, 0, 0);
141 	if (!tun_dst) {
142 		bareudp->dev->stats.rx_dropped++;
143 		goto drop;
144 	}
145 	skb_dst_set(skb, &tun_dst->dst);
146 	skb->dev = bareudp->dev;
147 	oiph = skb_network_header(skb);
148 	skb_reset_network_header(skb);
149 	skb_reset_mac_header(skb);
150 
151 	if (!ipv6_mod_enabled() || family == AF_INET)
152 		err = IP_ECN_decapsulate(oiph, skb);
153 	else
154 		err = IP6_ECN_decapsulate(oiph, skb);
155 
156 	if (unlikely(err)) {
157 		if (log_ecn_error) {
158 			if  (!ipv6_mod_enabled() || family == AF_INET)
159 				net_info_ratelimited("non-ECT from %pI4 "
160 						     "with TOS=%#x\n",
161 						     &((struct iphdr *)oiph)->saddr,
162 						     ((struct iphdr *)oiph)->tos);
163 			else
164 				net_info_ratelimited("non-ECT from %pI6\n",
165 						     &((struct ipv6hdr *)oiph)->saddr);
166 		}
167 		if (err > 1) {
168 			++bareudp->dev->stats.rx_frame_errors;
169 			++bareudp->dev->stats.rx_errors;
170 			goto drop;
171 		}
172 	}
173 
174 	len = skb->len;
175 	err = gro_cells_receive(&bareudp->gro_cells, skb);
176 	if (likely(err == NET_RX_SUCCESS))
177 		dev_sw_netstats_rx_add(bareudp->dev, len);
178 
179 	return 0;
180 drop:
181 	/* Consume bad packet */
182 	kfree_skb(skb);
183 
184 	return 0;
185 }
186 
187 static int bareudp_err_lookup(struct sock *sk, struct sk_buff *skb)
188 {
189 	return 0;
190 }
191 
192 static int bareudp_init(struct net_device *dev)
193 {
194 	struct bareudp_dev *bareudp = netdev_priv(dev);
195 	int err;
196 
197 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
198 	if (!dev->tstats)
199 		return -ENOMEM;
200 
201 	err = gro_cells_init(&bareudp->gro_cells, dev);
202 	if (err) {
203 		free_percpu(dev->tstats);
204 		return err;
205 	}
206 	return 0;
207 }
208 
209 static void bareudp_uninit(struct net_device *dev)
210 {
211 	struct bareudp_dev *bareudp = netdev_priv(dev);
212 
213 	gro_cells_destroy(&bareudp->gro_cells);
214 	free_percpu(dev->tstats);
215 }
216 
217 static struct socket *bareudp_create_sock(struct net *net, __be16 port)
218 {
219 	struct udp_port_cfg udp_conf;
220 	struct socket *sock;
221 	int err;
222 
223 	memset(&udp_conf, 0, sizeof(udp_conf));
224 
225 	if (ipv6_mod_enabled())
226 		udp_conf.family = AF_INET6;
227 	else
228 		udp_conf.family = AF_INET;
229 
230 	udp_conf.local_udp_port = port;
231 	/* Open UDP socket */
232 	err = udp_sock_create(net, &udp_conf, &sock);
233 	if (err < 0)
234 		return ERR_PTR(err);
235 
236 	udp_allow_gso(sock->sk);
237 	return sock;
238 }
239 
240 /* Create new listen socket if needed */
241 static int bareudp_socket_create(struct bareudp_dev *bareudp, __be16 port)
242 {
243 	struct udp_tunnel_sock_cfg tunnel_cfg;
244 	struct socket *sock;
245 
246 	sock = bareudp_create_sock(bareudp->net, port);
247 	if (IS_ERR(sock))
248 		return PTR_ERR(sock);
249 
250 	/* Mark socket as an encapsulation socket */
251 	memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
252 	tunnel_cfg.sk_user_data = bareudp;
253 	tunnel_cfg.encap_type = 1;
254 	tunnel_cfg.encap_rcv = bareudp_udp_encap_recv;
255 	tunnel_cfg.encap_err_lookup = bareudp_err_lookup;
256 	tunnel_cfg.encap_destroy = NULL;
257 	setup_udp_tunnel_sock(bareudp->net, sock, &tunnel_cfg);
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, false);
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 static int bareudp6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
354 			     struct bareudp_dev *bareudp,
355 			     const struct ip_tunnel_info *info)
356 {
357 	bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
358 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
359 	struct socket *sock  = rcu_dereference(bareudp->sock);
360 	bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
361 	const struct ip_tunnel_key *key = &info->key;
362 	struct dst_entry *dst = NULL;
363 	struct in6_addr saddr, daddr;
364 	int min_headroom;
365 	__u8 prio, ttl;
366 	__be16 sport;
367 	int err;
368 
369 	if (!sock)
370 		return -ESHUTDOWN;
371 
372 	dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock, &saddr, info,
373 				    IPPROTO_UDP, use_cache);
374 	if (IS_ERR(dst))
375 		return PTR_ERR(dst);
376 
377 	skb_tunnel_check_pmtu(skb, dst, BAREUDP_IPV6_HLEN + info->options_len,
378 			      false);
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 ipv6hdr);
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 
415 static bool bareudp_proto_valid(struct bareudp_dev *bareudp, __be16 proto)
416 {
417 	if (bareudp->ethertype == proto)
418 		return true;
419 
420 	if (!bareudp->multi_proto_mode)
421 		return false;
422 
423 	if (bareudp->ethertype == htons(ETH_P_MPLS_UC) &&
424 	    proto == htons(ETH_P_MPLS_MC))
425 		return true;
426 
427 	if (bareudp->ethertype == htons(ETH_P_IP) &&
428 	    proto == htons(ETH_P_IPV6))
429 		return true;
430 
431 	return false;
432 }
433 
434 static netdev_tx_t bareudp_xmit(struct sk_buff *skb, struct net_device *dev)
435 {
436 	struct bareudp_dev *bareudp = netdev_priv(dev);
437 	struct ip_tunnel_info *info = NULL;
438 	int err;
439 
440 	if (!bareudp_proto_valid(bareudp, skb->protocol)) {
441 		err = -EINVAL;
442 		goto tx_error;
443 	}
444 
445 	info = skb_tunnel_info(skb);
446 	if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) {
447 		err = -EINVAL;
448 		goto tx_error;
449 	}
450 
451 	rcu_read_lock();
452 	if (ipv6_mod_enabled() && info->mode & IP_TUNNEL_INFO_IPV6)
453 		err = bareudp6_xmit_skb(skb, dev, bareudp, info);
454 	else
455 		err = bareudp_xmit_skb(skb, dev, bareudp, info);
456 
457 	rcu_read_unlock();
458 
459 	if (likely(!err))
460 		return NETDEV_TX_OK;
461 tx_error:
462 	dev_kfree_skb(skb);
463 
464 	if (err == -ELOOP)
465 		dev->stats.collisions++;
466 	else if (err == -ENETUNREACH)
467 		dev->stats.tx_carrier_errors++;
468 
469 	dev->stats.tx_errors++;
470 	return NETDEV_TX_OK;
471 }
472 
473 static int bareudp_fill_metadata_dst(struct net_device *dev,
474 				     struct sk_buff *skb)
475 {
476 	struct ip_tunnel_info *info = skb_tunnel_info(skb);
477 	struct bareudp_dev *bareudp = netdev_priv(dev);
478 	bool use_cache;
479 
480 	use_cache = ip_tunnel_dst_cache_usable(skb, info);
481 
482 	if (!ipv6_mod_enabled() || ip_tunnel_info_af(info) == AF_INET) {
483 		struct rtable *rt;
484 		__be32 saddr;
485 
486 		rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr,
487 					    info, IPPROTO_UDP, use_cache);
488 		if (IS_ERR(rt))
489 			return PTR_ERR(rt);
490 
491 		ip_rt_put(rt);
492 		info->key.u.ipv4.src = saddr;
493 	} else if (ip_tunnel_info_af(info) == AF_INET6) {
494 		struct dst_entry *dst;
495 		struct in6_addr saddr;
496 		struct socket *sock = rcu_dereference(bareudp->sock);
497 
498 		dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock,
499 					    &saddr, info, IPPROTO_UDP,
500 					    use_cache);
501 		if (IS_ERR(dst))
502 			return PTR_ERR(dst);
503 
504 		dst_release(dst);
505 		info->key.u.ipv6.src = saddr;
506 	} else {
507 		return -EINVAL;
508 	}
509 
510 	info->key.tp_src = udp_flow_src_port(bareudp->net, skb,
511 					     bareudp->sport_min,
512 			USHRT_MAX, true);
513 	info->key.tp_dst = bareudp->port;
514 	return 0;
515 }
516 
517 static const struct net_device_ops bareudp_netdev_ops = {
518 	.ndo_init               = bareudp_init,
519 	.ndo_uninit             = bareudp_uninit,
520 	.ndo_open               = bareudp_open,
521 	.ndo_stop               = bareudp_stop,
522 	.ndo_start_xmit         = bareudp_xmit,
523 	.ndo_get_stats64        = dev_get_tstats64,
524 	.ndo_fill_metadata_dst  = bareudp_fill_metadata_dst,
525 };
526 
527 static const struct nla_policy bareudp_policy[IFLA_BAREUDP_MAX + 1] = {
528 	[IFLA_BAREUDP_PORT]                = { .type = NLA_U16 },
529 	[IFLA_BAREUDP_ETHERTYPE]	   = { .type = NLA_U16 },
530 	[IFLA_BAREUDP_SRCPORT_MIN]         = { .type = NLA_U16 },
531 	[IFLA_BAREUDP_MULTIPROTO_MODE]     = { .type = NLA_FLAG },
532 };
533 
534 /* Info for udev, that this is a virtual tunnel endpoint */
535 static const struct device_type bareudp_type = {
536 	.name = "bareudp",
537 };
538 
539 /* Initialize the device structure. */
540 static void bareudp_setup(struct net_device *dev)
541 {
542 	dev->netdev_ops = &bareudp_netdev_ops;
543 	dev->needs_free_netdev = true;
544 	SET_NETDEV_DEVTYPE(dev, &bareudp_type);
545 	dev->features    |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
546 	dev->features    |= NETIF_F_RXCSUM;
547 	dev->features    |= NETIF_F_LLTX;
548 	dev->features    |= NETIF_F_GSO_SOFTWARE;
549 	dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
550 	dev->hw_features |= NETIF_F_RXCSUM;
551 	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
552 	dev->hard_header_len = 0;
553 	dev->addr_len = 0;
554 	dev->mtu = ETH_DATA_LEN;
555 	dev->min_mtu = IPV4_MIN_MTU;
556 	dev->max_mtu = IP_MAX_MTU - BAREUDP_BASE_HLEN;
557 	dev->type = ARPHRD_NONE;
558 	netif_keep_dst(dev);
559 	dev->priv_flags |= IFF_NO_QUEUE;
560 	dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
561 }
562 
563 static int bareudp_validate(struct nlattr *tb[], struct nlattr *data[],
564 			    struct netlink_ext_ack *extack)
565 {
566 	if (!data) {
567 		NL_SET_ERR_MSG(extack,
568 			       "Not enough attributes provided to perform the operation");
569 		return -EINVAL;
570 	}
571 	return 0;
572 }
573 
574 static int bareudp2info(struct nlattr *data[], struct bareudp_conf *conf,
575 			struct netlink_ext_ack *extack)
576 {
577 	memset(conf, 0, sizeof(*conf));
578 
579 	if (!data[IFLA_BAREUDP_PORT]) {
580 		NL_SET_ERR_MSG(extack, "port not specified");
581 		return -EINVAL;
582 	}
583 	if (!data[IFLA_BAREUDP_ETHERTYPE]) {
584 		NL_SET_ERR_MSG(extack, "ethertype not specified");
585 		return -EINVAL;
586 	}
587 
588 	conf->port = nla_get_u16(data[IFLA_BAREUDP_PORT]);
589 	conf->ethertype = nla_get_u16(data[IFLA_BAREUDP_ETHERTYPE]);
590 
591 	if (data[IFLA_BAREUDP_SRCPORT_MIN])
592 		conf->sport_min =  nla_get_u16(data[IFLA_BAREUDP_SRCPORT_MIN]);
593 
594 	if (data[IFLA_BAREUDP_MULTIPROTO_MODE])
595 		conf->multi_proto_mode = true;
596 
597 	return 0;
598 }
599 
600 static struct bareudp_dev *bareudp_find_dev(struct bareudp_net *bn,
601 					    const struct bareudp_conf *conf)
602 {
603 	struct bareudp_dev *bareudp, *t = NULL;
604 
605 	list_for_each_entry(bareudp, &bn->bareudp_list, next) {
606 		if (conf->port == bareudp->port)
607 			t = bareudp;
608 	}
609 	return t;
610 }
611 
612 static int bareudp_configure(struct net *net, struct net_device *dev,
613 			     struct bareudp_conf *conf,
614 			     struct netlink_ext_ack *extack)
615 {
616 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
617 	struct bareudp_dev *t, *bareudp = netdev_priv(dev);
618 	int err;
619 
620 	bareudp->net = net;
621 	bareudp->dev = dev;
622 	t = bareudp_find_dev(bn, conf);
623 	if (t) {
624 		NL_SET_ERR_MSG(extack, "Another bareudp device using the same port already exists");
625 		return -EBUSY;
626 	}
627 
628 	if (conf->multi_proto_mode &&
629 	    (conf->ethertype != htons(ETH_P_MPLS_UC) &&
630 	     conf->ethertype != htons(ETH_P_IP))) {
631 		NL_SET_ERR_MSG(extack, "Cannot set multiproto mode for this ethertype (only IPv4 and unicast MPLS are supported)");
632 		return -EINVAL;
633 	}
634 
635 	bareudp->port = conf->port;
636 	bareudp->ethertype = conf->ethertype;
637 	bareudp->sport_min = conf->sport_min;
638 	bareudp->multi_proto_mode = conf->multi_proto_mode;
639 
640 	err = register_netdevice(dev);
641 	if (err)
642 		return err;
643 
644 	list_add(&bareudp->next, &bn->bareudp_list);
645 	return 0;
646 }
647 
648 static int bareudp_link_config(struct net_device *dev,
649 			       struct nlattr *tb[])
650 {
651 	int err;
652 
653 	if (tb[IFLA_MTU]) {
654 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
655 		if (err)
656 			return err;
657 	}
658 	return 0;
659 }
660 
661 static void bareudp_dellink(struct net_device *dev, struct list_head *head)
662 {
663 	struct bareudp_dev *bareudp = netdev_priv(dev);
664 
665 	list_del(&bareudp->next);
666 	unregister_netdevice_queue(dev, head);
667 }
668 
669 static int bareudp_newlink(struct net *net, struct net_device *dev,
670 			   struct nlattr *tb[], struct nlattr *data[],
671 			   struct netlink_ext_ack *extack)
672 {
673 	struct bareudp_conf conf;
674 	int err;
675 
676 	err = bareudp2info(data, &conf, extack);
677 	if (err)
678 		return err;
679 
680 	err = bareudp_configure(net, dev, &conf, extack);
681 	if (err)
682 		return err;
683 
684 	err = bareudp_link_config(dev, tb);
685 	if (err)
686 		goto err_unconfig;
687 
688 	return 0;
689 
690 err_unconfig:
691 	bareudp_dellink(dev, NULL);
692 	return err;
693 }
694 
695 static size_t bareudp_get_size(const struct net_device *dev)
696 {
697 	return  nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_PORT */
698 		nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_ETHERTYPE */
699 		nla_total_size(sizeof(__u16))  +  /* IFLA_BAREUDP_SRCPORT_MIN */
700 		nla_total_size(0)              +  /* IFLA_BAREUDP_MULTIPROTO_MODE */
701 		0;
702 }
703 
704 static int bareudp_fill_info(struct sk_buff *skb, const struct net_device *dev)
705 {
706 	struct bareudp_dev *bareudp = netdev_priv(dev);
707 
708 	if (nla_put_be16(skb, IFLA_BAREUDP_PORT, bareudp->port))
709 		goto nla_put_failure;
710 	if (nla_put_be16(skb, IFLA_BAREUDP_ETHERTYPE, bareudp->ethertype))
711 		goto nla_put_failure;
712 	if (nla_put_u16(skb, IFLA_BAREUDP_SRCPORT_MIN, bareudp->sport_min))
713 		goto nla_put_failure;
714 	if (bareudp->multi_proto_mode &&
715 	    nla_put_flag(skb, IFLA_BAREUDP_MULTIPROTO_MODE))
716 		goto nla_put_failure;
717 
718 	return 0;
719 
720 nla_put_failure:
721 	return -EMSGSIZE;
722 }
723 
724 static struct rtnl_link_ops bareudp_link_ops __read_mostly = {
725 	.kind           = "bareudp",
726 	.maxtype        = IFLA_BAREUDP_MAX,
727 	.policy         = bareudp_policy,
728 	.priv_size      = sizeof(struct bareudp_dev),
729 	.setup          = bareudp_setup,
730 	.validate       = bareudp_validate,
731 	.newlink        = bareudp_newlink,
732 	.dellink        = bareudp_dellink,
733 	.get_size       = bareudp_get_size,
734 	.fill_info      = bareudp_fill_info,
735 };
736 
737 static __net_init int bareudp_init_net(struct net *net)
738 {
739 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
740 
741 	INIT_LIST_HEAD(&bn->bareudp_list);
742 	return 0;
743 }
744 
745 static void bareudp_destroy_tunnels(struct net *net, struct list_head *head)
746 {
747 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
748 	struct bareudp_dev *bareudp, *next;
749 
750 	list_for_each_entry_safe(bareudp, next, &bn->bareudp_list, next)
751 		unregister_netdevice_queue(bareudp->dev, head);
752 }
753 
754 static void __net_exit bareudp_exit_batch_net(struct list_head *net_list)
755 {
756 	struct net *net;
757 	LIST_HEAD(list);
758 
759 	rtnl_lock();
760 	list_for_each_entry(net, net_list, exit_list)
761 		bareudp_destroy_tunnels(net, &list);
762 
763 	/* unregister the devices gathered above */
764 	unregister_netdevice_many(&list);
765 	rtnl_unlock();
766 }
767 
768 static struct pernet_operations bareudp_net_ops = {
769 	.init = bareudp_init_net,
770 	.exit_batch = bareudp_exit_batch_net,
771 	.id   = &bareudp_net_id,
772 	.size = sizeof(struct bareudp_net),
773 };
774 
775 static int __init bareudp_init_module(void)
776 {
777 	int rc;
778 
779 	rc = register_pernet_subsys(&bareudp_net_ops);
780 	if (rc)
781 		goto out1;
782 
783 	rc = rtnl_link_register(&bareudp_link_ops);
784 	if (rc)
785 		goto out2;
786 
787 	return 0;
788 out2:
789 	unregister_pernet_subsys(&bareudp_net_ops);
790 out1:
791 	return rc;
792 }
793 late_initcall(bareudp_init_module);
794 
795 static void __exit bareudp_cleanup_module(void)
796 {
797 	rtnl_link_unregister(&bareudp_link_ops);
798 	unregister_pernet_subsys(&bareudp_net_ops);
799 }
800 module_exit(bareudp_cleanup_module);
801 
802 MODULE_ALIAS_RTNL_LINK("bareudp");
803 MODULE_LICENSE("GPL");
804 MODULE_AUTHOR("Martin Varghese <martin.varghese@nokia.com>");
805 MODULE_DESCRIPTION("Interface driver for UDP encapsulated traffic");
806