• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 
bareudp_udp_encap_recv(struct sock * sk,struct sk_buff * skb)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 
143 	if (!ipv6_mod_enabled() || family == AF_INET)
144 		err = IP_ECN_decapsulate(oiph, skb);
145 	else
146 		err = IP6_ECN_decapsulate(oiph, skb);
147 
148 	if (unlikely(err)) {
149 		if (log_ecn_error) {
150 			if  (!ipv6_mod_enabled() || family == AF_INET)
151 				net_info_ratelimited("non-ECT from %pI4 "
152 						     "with TOS=%#x\n",
153 						     &((struct iphdr *)oiph)->saddr,
154 						     ((struct iphdr *)oiph)->tos);
155 			else
156 				net_info_ratelimited("non-ECT from %pI6\n",
157 						     &((struct ipv6hdr *)oiph)->saddr);
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 		dev_sw_netstats_rx_add(bareudp->dev, len);
170 
171 	return 0;
172 drop:
173 	/* Consume bad packet */
174 	kfree_skb(skb);
175 
176 	return 0;
177 }
178 
bareudp_err_lookup(struct sock * sk,struct sk_buff * skb)179 static int bareudp_err_lookup(struct sock *sk, struct sk_buff *skb)
180 {
181 	return 0;
182 }
183 
bareudp_init(struct net_device * dev)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 
bareudp_uninit(struct net_device * dev)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 
bareudp_create_sock(struct net * net,__be16 port)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 
217 	if (ipv6_mod_enabled())
218 		udp_conf.family = AF_INET6;
219 	else
220 		udp_conf.family = AF_INET;
221 
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 	return sock;
229 }
230 
231 /* Create new listen socket if needed */
bareudp_socket_create(struct bareudp_dev * bareudp,__be16 port)232 static int bareudp_socket_create(struct bareudp_dev *bareudp, __be16 port)
233 {
234 	struct udp_tunnel_sock_cfg tunnel_cfg;
235 	struct socket *sock;
236 
237 	sock = bareudp_create_sock(bareudp->net, port);
238 	if (IS_ERR(sock))
239 		return PTR_ERR(sock);
240 
241 	/* Mark socket as an encapsulation socket */
242 	memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
243 	tunnel_cfg.sk_user_data = bareudp;
244 	tunnel_cfg.encap_type = 1;
245 	tunnel_cfg.encap_rcv = bareudp_udp_encap_recv;
246 	tunnel_cfg.encap_err_lookup = bareudp_err_lookup;
247 	tunnel_cfg.encap_destroy = NULL;
248 	setup_udp_tunnel_sock(bareudp->net, sock, &tunnel_cfg);
249 
250 	rcu_assign_pointer(bareudp->sock, sock);
251 	return 0;
252 }
253 
bareudp_open(struct net_device * dev)254 static int bareudp_open(struct net_device *dev)
255 {
256 	struct bareudp_dev *bareudp = netdev_priv(dev);
257 	int ret = 0;
258 
259 	ret =  bareudp_socket_create(bareudp, bareudp->port);
260 	return ret;
261 }
262 
bareudp_sock_release(struct bareudp_dev * bareudp)263 static void bareudp_sock_release(struct bareudp_dev *bareudp)
264 {
265 	struct socket *sock;
266 
267 	sock = bareudp->sock;
268 	rcu_assign_pointer(bareudp->sock, NULL);
269 	synchronize_net();
270 	udp_tunnel_sock_release(sock);
271 }
272 
bareudp_stop(struct net_device * dev)273 static int bareudp_stop(struct net_device *dev)
274 {
275 	struct bareudp_dev *bareudp = netdev_priv(dev);
276 
277 	bareudp_sock_release(bareudp);
278 	return 0;
279 }
280 
bareudp_xmit_skb(struct sk_buff * skb,struct net_device * dev,struct bareudp_dev * bareudp,const struct ip_tunnel_info * info)281 static int bareudp_xmit_skb(struct sk_buff *skb, struct net_device *dev,
282 			    struct bareudp_dev *bareudp,
283 			    const struct ip_tunnel_info *info)
284 {
285 	bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
286 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
287 	struct socket *sock = rcu_dereference(bareudp->sock);
288 	bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
289 	const struct ip_tunnel_key *key = &info->key;
290 	struct rtable *rt;
291 	__be16 sport, df;
292 	int min_headroom;
293 	__u8 tos, ttl;
294 	__be32 saddr;
295 	int err;
296 
297 	if (!sock)
298 		return -ESHUTDOWN;
299 
300 	rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr, info,
301 				    IPPROTO_UDP, use_cache);
302 
303 	if (IS_ERR(rt))
304 		return PTR_ERR(rt);
305 
306 	skb_tunnel_check_pmtu(skb, &rt->dst,
307 			      BAREUDP_IPV4_HLEN + info->options_len, false);
308 
309 	sport = udp_flow_src_port(bareudp->net, skb,
310 				  bareudp->sport_min, USHRT_MAX,
311 				  true);
312 	tos = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
313 	ttl = key->ttl;
314 	df = key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
315 	skb_scrub_packet(skb, xnet);
316 
317 	err = -ENOSPC;
318 	if (!skb_pull(skb, skb_network_offset(skb)))
319 		goto free_dst;
320 
321 	min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len +
322 		BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr);
323 
324 	err = skb_cow_head(skb, min_headroom);
325 	if (unlikely(err))
326 		goto free_dst;
327 
328 	err = udp_tunnel_handle_offloads(skb, udp_sum);
329 	if (err)
330 		goto free_dst;
331 
332 	skb_set_inner_protocol(skb, bareudp->ethertype);
333 	udp_tunnel_xmit_skb(rt, sock->sk, skb, saddr, info->key.u.ipv4.dst,
334 			    tos, ttl, df, sport, bareudp->port,
335 			    !net_eq(bareudp->net, dev_net(bareudp->dev)),
336 			    !(info->key.tun_flags & TUNNEL_CSUM));
337 	return 0;
338 
339 free_dst:
340 	dst_release(&rt->dst);
341 	return err;
342 }
343 
bareudp6_xmit_skb(struct sk_buff * skb,struct net_device * dev,struct bareudp_dev * bareudp,const struct ip_tunnel_info * info)344 static int bareudp6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
345 			     struct bareudp_dev *bareudp,
346 			     const struct ip_tunnel_info *info)
347 {
348 	bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
349 	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
350 	struct socket *sock  = rcu_dereference(bareudp->sock);
351 	bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
352 	const struct ip_tunnel_key *key = &info->key;
353 	struct dst_entry *dst = NULL;
354 	struct in6_addr saddr, daddr;
355 	int min_headroom;
356 	__u8 prio, ttl;
357 	__be16 sport;
358 	int err;
359 
360 	if (!sock)
361 		return -ESHUTDOWN;
362 
363 	dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock, &saddr, info,
364 				    IPPROTO_UDP, use_cache);
365 	if (IS_ERR(dst))
366 		return PTR_ERR(dst);
367 
368 	skb_tunnel_check_pmtu(skb, dst, BAREUDP_IPV6_HLEN + info->options_len,
369 			      false);
370 
371 	sport = udp_flow_src_port(bareudp->net, skb,
372 				  bareudp->sport_min, USHRT_MAX,
373 				  true);
374 	prio = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
375 	ttl = key->ttl;
376 
377 	skb_scrub_packet(skb, xnet);
378 
379 	err = -ENOSPC;
380 	if (!skb_pull(skb, skb_network_offset(skb)))
381 		goto free_dst;
382 
383 	min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len +
384 		BAREUDP_BASE_HLEN + info->options_len + sizeof(struct ipv6hdr);
385 
386 	err = skb_cow_head(skb, min_headroom);
387 	if (unlikely(err))
388 		goto free_dst;
389 
390 	err = udp_tunnel_handle_offloads(skb, udp_sum);
391 	if (err)
392 		goto free_dst;
393 
394 	daddr = info->key.u.ipv6.dst;
395 	udp_tunnel6_xmit_skb(dst, sock->sk, skb, dev,
396 			     &saddr, &daddr, prio, ttl,
397 			     info->key.label, sport, bareudp->port,
398 			     !(info->key.tun_flags & TUNNEL_CSUM));
399 	return 0;
400 
401 free_dst:
402 	dst_release(dst);
403 	return err;
404 }
405 
bareudp_proto_valid(struct bareudp_dev * bareudp,__be16 proto)406 static bool bareudp_proto_valid(struct bareudp_dev *bareudp, __be16 proto)
407 {
408 	if (bareudp->ethertype == proto)
409 		return true;
410 
411 	if (!bareudp->multi_proto_mode)
412 		return false;
413 
414 	if (bareudp->ethertype == htons(ETH_P_MPLS_UC) &&
415 	    proto == htons(ETH_P_MPLS_MC))
416 		return true;
417 
418 	if (bareudp->ethertype == htons(ETH_P_IP) &&
419 	    proto == htons(ETH_P_IPV6))
420 		return true;
421 
422 	return false;
423 }
424 
bareudp_xmit(struct sk_buff * skb,struct net_device * dev)425 static netdev_tx_t bareudp_xmit(struct sk_buff *skb, struct net_device *dev)
426 {
427 	struct bareudp_dev *bareudp = netdev_priv(dev);
428 	struct ip_tunnel_info *info = NULL;
429 	int err;
430 
431 	if (!bareudp_proto_valid(bareudp, skb->protocol)) {
432 		err = -EINVAL;
433 		goto tx_error;
434 	}
435 
436 	info = skb_tunnel_info(skb);
437 	if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) {
438 		err = -EINVAL;
439 		goto tx_error;
440 	}
441 
442 	rcu_read_lock();
443 	if (ipv6_mod_enabled() && info->mode & IP_TUNNEL_INFO_IPV6)
444 		err = bareudp6_xmit_skb(skb, dev, bareudp, info);
445 	else
446 		err = bareudp_xmit_skb(skb, dev, bareudp, info);
447 
448 	rcu_read_unlock();
449 
450 	if (likely(!err))
451 		return NETDEV_TX_OK;
452 tx_error:
453 	dev_kfree_skb(skb);
454 
455 	if (err == -ELOOP)
456 		dev->stats.collisions++;
457 	else if (err == -ENETUNREACH)
458 		dev->stats.tx_carrier_errors++;
459 
460 	dev->stats.tx_errors++;
461 	return NETDEV_TX_OK;
462 }
463 
bareudp_fill_metadata_dst(struct net_device * dev,struct sk_buff * skb)464 static int bareudp_fill_metadata_dst(struct net_device *dev,
465 				     struct sk_buff *skb)
466 {
467 	struct ip_tunnel_info *info = skb_tunnel_info(skb);
468 	struct bareudp_dev *bareudp = netdev_priv(dev);
469 	bool use_cache;
470 
471 	use_cache = ip_tunnel_dst_cache_usable(skb, info);
472 
473 	if (!ipv6_mod_enabled() || ip_tunnel_info_af(info) == AF_INET) {
474 		struct rtable *rt;
475 		__be32 saddr;
476 
477 		rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr,
478 					    info, IPPROTO_UDP, use_cache);
479 		if (IS_ERR(rt))
480 			return PTR_ERR(rt);
481 
482 		ip_rt_put(rt);
483 		info->key.u.ipv4.src = saddr;
484 	} else if (ip_tunnel_info_af(info) == AF_INET6) {
485 		struct dst_entry *dst;
486 		struct in6_addr saddr;
487 		struct socket *sock = rcu_dereference(bareudp->sock);
488 
489 		dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock,
490 					    &saddr, info, IPPROTO_UDP,
491 					    use_cache);
492 		if (IS_ERR(dst))
493 			return PTR_ERR(dst);
494 
495 		dst_release(dst);
496 		info->key.u.ipv6.src = saddr;
497 	} else {
498 		return -EINVAL;
499 	}
500 
501 	info->key.tp_src = udp_flow_src_port(bareudp->net, skb,
502 					     bareudp->sport_min,
503 			USHRT_MAX, true);
504 	info->key.tp_dst = bareudp->port;
505 	return 0;
506 }
507 
508 static const struct net_device_ops bareudp_netdev_ops = {
509 	.ndo_init               = bareudp_init,
510 	.ndo_uninit             = bareudp_uninit,
511 	.ndo_open               = bareudp_open,
512 	.ndo_stop               = bareudp_stop,
513 	.ndo_start_xmit         = bareudp_xmit,
514 	.ndo_get_stats64        = ip_tunnel_get_stats64,
515 	.ndo_fill_metadata_dst  = bareudp_fill_metadata_dst,
516 };
517 
518 static const struct nla_policy bareudp_policy[IFLA_BAREUDP_MAX + 1] = {
519 	[IFLA_BAREUDP_PORT]                = { .type = NLA_U16 },
520 	[IFLA_BAREUDP_ETHERTYPE]	   = { .type = NLA_U16 },
521 	[IFLA_BAREUDP_SRCPORT_MIN]         = { .type = NLA_U16 },
522 	[IFLA_BAREUDP_MULTIPROTO_MODE]     = { .type = NLA_FLAG },
523 };
524 
525 /* Info for udev, that this is a virtual tunnel endpoint */
526 static struct device_type bareudp_type = {
527 	.name = "bareudp",
528 };
529 
530 /* Initialize the device structure. */
bareudp_setup(struct net_device * dev)531 static void bareudp_setup(struct net_device *dev)
532 {
533 	dev->netdev_ops = &bareudp_netdev_ops;
534 	dev->needs_free_netdev = true;
535 	SET_NETDEV_DEVTYPE(dev, &bareudp_type);
536 	dev->features    |= NETIF_F_SG | NETIF_F_HW_CSUM;
537 	dev->features    |= NETIF_F_RXCSUM;
538 	dev->features    |= NETIF_F_LLTX;
539 	dev->features    |= NETIF_F_GSO_SOFTWARE;
540 	dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
541 	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
542 	dev->hard_header_len = 0;
543 	dev->addr_len = 0;
544 	dev->mtu = ETH_DATA_LEN;
545 	dev->min_mtu = IPV4_MIN_MTU;
546 	dev->max_mtu = IP_MAX_MTU - BAREUDP_BASE_HLEN;
547 	dev->type = ARPHRD_NONE;
548 	netif_keep_dst(dev);
549 	dev->priv_flags |= IFF_NO_QUEUE;
550 	dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
551 }
552 
bareudp_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)553 static int bareudp_validate(struct nlattr *tb[], struct nlattr *data[],
554 			    struct netlink_ext_ack *extack)
555 {
556 	if (!data) {
557 		NL_SET_ERR_MSG(extack,
558 			       "Not enough attributes provided to perform the operation");
559 		return -EINVAL;
560 	}
561 	return 0;
562 }
563 
bareudp2info(struct nlattr * data[],struct bareudp_conf * conf,struct netlink_ext_ack * extack)564 static int bareudp2info(struct nlattr *data[], struct bareudp_conf *conf,
565 			struct netlink_ext_ack *extack)
566 {
567 	memset(conf, 0, sizeof(*conf));
568 
569 	if (!data[IFLA_BAREUDP_PORT]) {
570 		NL_SET_ERR_MSG(extack, "port not specified");
571 		return -EINVAL;
572 	}
573 	if (!data[IFLA_BAREUDP_ETHERTYPE]) {
574 		NL_SET_ERR_MSG(extack, "ethertype not specified");
575 		return -EINVAL;
576 	}
577 
578 	if (data[IFLA_BAREUDP_PORT])
579 		conf->port =  nla_get_u16(data[IFLA_BAREUDP_PORT]);
580 
581 	if (data[IFLA_BAREUDP_ETHERTYPE])
582 		conf->ethertype =  nla_get_u16(data[IFLA_BAREUDP_ETHERTYPE]);
583 
584 	if (data[IFLA_BAREUDP_SRCPORT_MIN])
585 		conf->sport_min =  nla_get_u16(data[IFLA_BAREUDP_SRCPORT_MIN]);
586 
587 	if (data[IFLA_BAREUDP_MULTIPROTO_MODE])
588 		conf->multi_proto_mode = true;
589 
590 	return 0;
591 }
592 
bareudp_find_dev(struct bareudp_net * bn,const struct bareudp_conf * conf)593 static struct bareudp_dev *bareudp_find_dev(struct bareudp_net *bn,
594 					    const struct bareudp_conf *conf)
595 {
596 	struct bareudp_dev *bareudp, *t = NULL;
597 
598 	list_for_each_entry(bareudp, &bn->bareudp_list, next) {
599 		if (conf->port == bareudp->port)
600 			t = bareudp;
601 	}
602 	return t;
603 }
604 
bareudp_configure(struct net * net,struct net_device * dev,struct bareudp_conf * conf)605 static int bareudp_configure(struct net *net, struct net_device *dev,
606 			     struct bareudp_conf *conf)
607 {
608 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
609 	struct bareudp_dev *t, *bareudp = netdev_priv(dev);
610 	int err;
611 
612 	bareudp->net = net;
613 	bareudp->dev = dev;
614 	t = bareudp_find_dev(bn, conf);
615 	if (t)
616 		return -EBUSY;
617 
618 	if (conf->multi_proto_mode &&
619 	    (conf->ethertype != htons(ETH_P_MPLS_UC) &&
620 	     conf->ethertype != htons(ETH_P_IP)))
621 		return -EINVAL;
622 
623 	bareudp->port = conf->port;
624 	bareudp->ethertype = conf->ethertype;
625 	bareudp->sport_min = conf->sport_min;
626 	bareudp->multi_proto_mode = conf->multi_proto_mode;
627 
628 	err = register_netdevice(dev);
629 	if (err)
630 		return err;
631 
632 	list_add(&bareudp->next, &bn->bareudp_list);
633 	return 0;
634 }
635 
bareudp_link_config(struct net_device * dev,struct nlattr * tb[])636 static int bareudp_link_config(struct net_device *dev,
637 			       struct nlattr *tb[])
638 {
639 	int err;
640 
641 	if (tb[IFLA_MTU]) {
642 		err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
643 		if (err)
644 			return err;
645 	}
646 	return 0;
647 }
648 
bareudp_dellink(struct net_device * dev,struct list_head * head)649 static void bareudp_dellink(struct net_device *dev, struct list_head *head)
650 {
651 	struct bareudp_dev *bareudp = netdev_priv(dev);
652 
653 	list_del(&bareudp->next);
654 	unregister_netdevice_queue(dev, head);
655 }
656 
bareudp_newlink(struct net * net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)657 static int bareudp_newlink(struct net *net, struct net_device *dev,
658 			   struct nlattr *tb[], struct nlattr *data[],
659 			   struct netlink_ext_ack *extack)
660 {
661 	struct bareudp_conf conf;
662 	LIST_HEAD(list_kill);
663 	int err;
664 
665 	err = bareudp2info(data, &conf, extack);
666 	if (err)
667 		return err;
668 
669 	err = bareudp_configure(net, dev, &conf);
670 	if (err)
671 		return err;
672 
673 	err = bareudp_link_config(dev, tb);
674 	if (err)
675 		goto err_unconfig;
676 
677 	return 0;
678 
679 err_unconfig:
680 	bareudp_dellink(dev, &list_kill);
681 	unregister_netdevice_many(&list_kill);
682 	return err;
683 }
684 
bareudp_get_size(const struct net_device * dev)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 
bareudp_fill_info(struct sk_buff * skb,const struct net_device * dev)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 
bareudp_dev_create(struct net * net,const char * name,u8 name_assign_type,struct bareudp_conf * conf)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 	LIST_HEAD(list_kill);
734 	int err;
735 
736 	memset(tb, 0, sizeof(tb));
737 	dev = rtnl_create_link(net, name, name_assign_type,
738 			       &bareudp_link_ops, tb, NULL);
739 	if (IS_ERR(dev))
740 		return dev;
741 
742 	err = bareudp_configure(net, dev, conf);
743 	if (err) {
744 		free_netdev(dev);
745 		return ERR_PTR(err);
746 	}
747 	err = dev_set_mtu(dev, IP_MAX_MTU - BAREUDP_BASE_HLEN);
748 	if (err)
749 		goto err;
750 
751 	err = rtnl_configure_link(dev, NULL);
752 	if (err < 0)
753 		goto err;
754 
755 	return dev;
756 err:
757 	bareudp_dellink(dev, &list_kill);
758 	unregister_netdevice_many(&list_kill);
759 	return ERR_PTR(err);
760 }
761 EXPORT_SYMBOL_GPL(bareudp_dev_create);
762 
bareudp_init_net(struct net * net)763 static __net_init int bareudp_init_net(struct net *net)
764 {
765 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
766 
767 	INIT_LIST_HEAD(&bn->bareudp_list);
768 	return 0;
769 }
770 
bareudp_destroy_tunnels(struct net * net,struct list_head * head)771 static void bareudp_destroy_tunnels(struct net *net, struct list_head *head)
772 {
773 	struct bareudp_net *bn = net_generic(net, bareudp_net_id);
774 	struct bareudp_dev *bareudp, *next;
775 
776 	list_for_each_entry_safe(bareudp, next, &bn->bareudp_list, next)
777 		unregister_netdevice_queue(bareudp->dev, head);
778 }
779 
bareudp_exit_batch_net(struct list_head * net_list)780 static void __net_exit bareudp_exit_batch_net(struct list_head *net_list)
781 {
782 	struct net *net;
783 	LIST_HEAD(list);
784 
785 	rtnl_lock();
786 	list_for_each_entry(net, net_list, exit_list)
787 		bareudp_destroy_tunnels(net, &list);
788 
789 	/* unregister the devices gathered above */
790 	unregister_netdevice_many(&list);
791 	rtnl_unlock();
792 }
793 
794 static struct pernet_operations bareudp_net_ops = {
795 	.init = bareudp_init_net,
796 	.exit_batch = bareudp_exit_batch_net,
797 	.id   = &bareudp_net_id,
798 	.size = sizeof(struct bareudp_net),
799 };
800 
bareudp_init_module(void)801 static int __init bareudp_init_module(void)
802 {
803 	int rc;
804 
805 	rc = register_pernet_subsys(&bareudp_net_ops);
806 	if (rc)
807 		goto out1;
808 
809 	rc = rtnl_link_register(&bareudp_link_ops);
810 	if (rc)
811 		goto out2;
812 
813 	return 0;
814 out2:
815 	unregister_pernet_subsys(&bareudp_net_ops);
816 out1:
817 	return rc;
818 }
819 late_initcall(bareudp_init_module);
820 
bareudp_cleanup_module(void)821 static void __exit bareudp_cleanup_module(void)
822 {
823 	rtnl_link_unregister(&bareudp_link_ops);
824 	unregister_pernet_subsys(&bareudp_net_ops);
825 }
826 module_exit(bareudp_cleanup_module);
827 
828 MODULE_ALIAS_RTNL_LINK("bareudp");
829 MODULE_LICENSE("GPL");
830 MODULE_AUTHOR("Martin Varghese <martin.varghese@nokia.com>");
831 MODULE_DESCRIPTION("Interface driver for UDP encapsulated traffic");
832