1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IPv6 tunneling device
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Ville Nuorvala		<vnuorval@tcs.hut.fi>
8  *	Yasuyuki Kozakai	<kozakai@linux-ipv6.org>
9  *
10  *      Based on:
11  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
12  *
13  *      RFC 2473
14  */
15 
16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
17 
18 #include <linux/module.h>
19 #include <linux/capability.h>
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/sockios.h>
23 #include <linux/icmp.h>
24 #include <linux/if.h>
25 #include <linux/in.h>
26 #include <linux/ip.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmpv6.h>
32 #include <linux/init.h>
33 #include <linux/route.h>
34 #include <linux/rtnetlink.h>
35 #include <linux/netfilter_ipv6.h>
36 #include <linux/slab.h>
37 #include <linux/hash.h>
38 #include <linux/etherdevice.h>
39 
40 #include <linux/uaccess.h>
41 #include <linux/atomic.h>
42 
43 #include <net/icmp.h>
44 #include <net/ip.h>
45 #include <net/ip_tunnels.h>
46 #include <net/ipv6.h>
47 #include <net/ip6_route.h>
48 #include <net/addrconf.h>
49 #include <net/ip6_tunnel.h>
50 #include <net/xfrm.h>
51 #include <net/dsfield.h>
52 #include <net/inet_ecn.h>
53 #include <net/net_namespace.h>
54 #include <net/netns/generic.h>
55 #include <net/dst_metadata.h>
56 
57 MODULE_AUTHOR("Ville Nuorvala");
58 MODULE_DESCRIPTION("IPv6 tunneling device");
59 MODULE_LICENSE("GPL");
60 MODULE_ALIAS_RTNL_LINK("ip6tnl");
61 MODULE_ALIAS_NETDEV("ip6tnl0");
62 
63 #define IP6_TUNNEL_HASH_SIZE_SHIFT  5
64 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
65 
66 static bool log_ecn_error = true;
67 module_param(log_ecn_error, bool, 0644);
68 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
69 
HASH(const struct in6_addr * addr1,const struct in6_addr * addr2)70 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
71 {
72 	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
73 
74 	return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
75 }
76 
77 static int ip6_tnl_dev_init(struct net_device *dev);
78 static void ip6_tnl_dev_setup(struct net_device *dev);
79 static struct rtnl_link_ops ip6_link_ops __read_mostly;
80 
81 static unsigned int ip6_tnl_net_id __read_mostly;
82 struct ip6_tnl_net {
83 	/* the IPv6 tunnel fallback device */
84 	struct net_device *fb_tnl_dev;
85 	/* lists for storing tunnels in use */
86 	struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
87 	struct ip6_tnl __rcu *tnls_wc[1];
88 	struct ip6_tnl __rcu **tnls[2];
89 	struct ip6_tnl __rcu *collect_md_tun;
90 };
91 
ip6_tnl_mpls_supported(void)92 static inline int ip6_tnl_mpls_supported(void)
93 {
94 	return IS_ENABLED(CONFIG_MPLS);
95 }
96 
97 #define for_each_ip6_tunnel_rcu(start) \
98 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
99 
100 /**
101  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
102  *   @net: network namespace
103  *   @link: ifindex of underlying interface
104  *   @remote: the address of the tunnel exit-point
105  *   @local: the address of the tunnel entry-point
106  *
107  * Return:
108  *   tunnel matching given end-points if found,
109  *   else fallback tunnel if its device is up,
110  *   else %NULL
111  **/
112 
113 static struct ip6_tnl *
ip6_tnl_lookup(struct net * net,int link,const struct in6_addr * remote,const struct in6_addr * local)114 ip6_tnl_lookup(struct net *net, int link,
115 	       const struct in6_addr *remote, const struct in6_addr *local)
116 {
117 	unsigned int hash = HASH(remote, local);
118 	struct ip6_tnl *t, *cand = NULL;
119 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
120 	struct in6_addr any;
121 
122 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
123 		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
124 		    !ipv6_addr_equal(remote, &t->parms.raddr) ||
125 		    !(t->dev->flags & IFF_UP))
126 			continue;
127 
128 		if (link == t->parms.link)
129 			return t;
130 		else
131 			cand = t;
132 	}
133 
134 	memset(&any, 0, sizeof(any));
135 	hash = HASH(&any, local);
136 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
137 		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
138 		    !ipv6_addr_any(&t->parms.raddr) ||
139 		    !(t->dev->flags & IFF_UP))
140 			continue;
141 
142 		if (link == t->parms.link)
143 			return t;
144 		else if (!cand)
145 			cand = t;
146 	}
147 
148 	hash = HASH(remote, &any);
149 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
150 		if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
151 		    !ipv6_addr_any(&t->parms.laddr) ||
152 		    !(t->dev->flags & IFF_UP))
153 			continue;
154 
155 		if (link == t->parms.link)
156 			return t;
157 		else if (!cand)
158 			cand = t;
159 	}
160 
161 	if (cand)
162 		return cand;
163 
164 	t = rcu_dereference(ip6n->collect_md_tun);
165 	if (t && t->dev->flags & IFF_UP)
166 		return t;
167 
168 	t = rcu_dereference(ip6n->tnls_wc[0]);
169 	if (t && (t->dev->flags & IFF_UP))
170 		return t;
171 
172 	return NULL;
173 }
174 
175 /**
176  * ip6_tnl_bucket - get head of list matching given tunnel parameters
177  *   @ip6n: the private data for ip6_vti in the netns
178  *   @p: parameters containing tunnel end-points
179  *
180  * Description:
181  *   ip6_tnl_bucket() returns the head of the list matching the
182  *   &struct in6_addr entries laddr and raddr in @p.
183  *
184  * Return: head of IPv6 tunnel list
185  **/
186 
187 static struct ip6_tnl __rcu **
ip6_tnl_bucket(struct ip6_tnl_net * ip6n,const struct __ip6_tnl_parm * p)188 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
189 {
190 	const struct in6_addr *remote = &p->raddr;
191 	const struct in6_addr *local = &p->laddr;
192 	unsigned int h = 0;
193 	int prio = 0;
194 
195 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
196 		prio = 1;
197 		h = HASH(remote, local);
198 	}
199 	return &ip6n->tnls[prio][h];
200 }
201 
202 /**
203  * ip6_tnl_link - add tunnel to hash table
204  *   @ip6n: the private data for ip6_vti in the netns
205  *   @t: tunnel to be added
206  **/
207 
208 static void
ip6_tnl_link(struct ip6_tnl_net * ip6n,struct ip6_tnl * t)209 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
210 {
211 	struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
212 
213 	if (t->parms.collect_md)
214 		rcu_assign_pointer(ip6n->collect_md_tun, t);
215 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
216 	rcu_assign_pointer(*tp, t);
217 }
218 
219 /**
220  * ip6_tnl_unlink - remove tunnel from hash table
221  *   @ip6n: the private data for ip6_vti in the netns
222  *   @t: tunnel to be removed
223  **/
224 
225 static void
ip6_tnl_unlink(struct ip6_tnl_net * ip6n,struct ip6_tnl * t)226 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
227 {
228 	struct ip6_tnl __rcu **tp;
229 	struct ip6_tnl *iter;
230 
231 	if (t->parms.collect_md)
232 		rcu_assign_pointer(ip6n->collect_md_tun, NULL);
233 
234 	for (tp = ip6_tnl_bucket(ip6n, &t->parms);
235 	     (iter = rtnl_dereference(*tp)) != NULL;
236 	     tp = &iter->next) {
237 		if (t == iter) {
238 			rcu_assign_pointer(*tp, t->next);
239 			break;
240 		}
241 	}
242 }
243 
ip6_dev_free(struct net_device * dev)244 static void ip6_dev_free(struct net_device *dev)
245 {
246 	struct ip6_tnl *t = netdev_priv(dev);
247 
248 	gro_cells_destroy(&t->gro_cells);
249 	dst_cache_destroy(&t->dst_cache);
250 	free_percpu(dev->tstats);
251 }
252 
ip6_tnl_create2(struct net_device * dev)253 static int ip6_tnl_create2(struct net_device *dev)
254 {
255 	struct ip6_tnl *t = netdev_priv(dev);
256 	struct net *net = dev_net(dev);
257 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
258 	int err;
259 
260 	dev->rtnl_link_ops = &ip6_link_ops;
261 	err = register_netdevice(dev);
262 	if (err < 0)
263 		goto out;
264 
265 	strcpy(t->parms.name, dev->name);
266 
267 	ip6_tnl_link(ip6n, t);
268 	return 0;
269 
270 out:
271 	return err;
272 }
273 
274 /**
275  * ip6_tnl_create - create a new tunnel
276  *   @net: network namespace
277  *   @p: tunnel parameters
278  *
279  * Description:
280  *   Create tunnel matching given parameters.
281  *
282  * Return:
283  *   created tunnel or error pointer
284  **/
285 
ip6_tnl_create(struct net * net,struct __ip6_tnl_parm * p)286 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
287 {
288 	struct net_device *dev;
289 	struct ip6_tnl *t;
290 	char name[IFNAMSIZ];
291 	int err = -E2BIG;
292 
293 	if (p->name[0]) {
294 		if (!dev_valid_name(p->name))
295 			goto failed;
296 		strscpy(name, p->name, IFNAMSIZ);
297 	} else {
298 		sprintf(name, "ip6tnl%%d");
299 	}
300 	err = -ENOMEM;
301 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
302 			   ip6_tnl_dev_setup);
303 	if (!dev)
304 		goto failed;
305 
306 	dev_net_set(dev, net);
307 
308 	t = netdev_priv(dev);
309 	t->parms = *p;
310 	t->net = dev_net(dev);
311 	err = ip6_tnl_create2(dev);
312 	if (err < 0)
313 		goto failed_free;
314 
315 	return t;
316 
317 failed_free:
318 	free_netdev(dev);
319 failed:
320 	return ERR_PTR(err);
321 }
322 
323 /**
324  * ip6_tnl_locate - find or create tunnel matching given parameters
325  *   @net: network namespace
326  *   @p: tunnel parameters
327  *   @create: != 0 if allowed to create new tunnel if no match found
328  *
329  * Description:
330  *   ip6_tnl_locate() first tries to locate an existing tunnel
331  *   based on @parms. If this is unsuccessful, but @create is set a new
332  *   tunnel device is created and registered for use.
333  *
334  * Return:
335  *   matching tunnel or error pointer
336  **/
337 
ip6_tnl_locate(struct net * net,struct __ip6_tnl_parm * p,int create)338 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
339 		struct __ip6_tnl_parm *p, int create)
340 {
341 	const struct in6_addr *remote = &p->raddr;
342 	const struct in6_addr *local = &p->laddr;
343 	struct ip6_tnl __rcu **tp;
344 	struct ip6_tnl *t;
345 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
346 
347 	for (tp = ip6_tnl_bucket(ip6n, p);
348 	     (t = rtnl_dereference(*tp)) != NULL;
349 	     tp = &t->next) {
350 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
351 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
352 		    p->link == t->parms.link) {
353 			if (create)
354 				return ERR_PTR(-EEXIST);
355 
356 			return t;
357 		}
358 	}
359 	if (!create)
360 		return ERR_PTR(-ENODEV);
361 	return ip6_tnl_create(net, p);
362 }
363 
364 /**
365  * ip6_tnl_dev_uninit - tunnel device uninitializer
366  *   @dev: the device to be destroyed
367  *
368  * Description:
369  *   ip6_tnl_dev_uninit() removes tunnel from its list
370  **/
371 
372 static void
ip6_tnl_dev_uninit(struct net_device * dev)373 ip6_tnl_dev_uninit(struct net_device *dev)
374 {
375 	struct ip6_tnl *t = netdev_priv(dev);
376 	struct net *net = t->net;
377 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
378 
379 	if (dev == ip6n->fb_tnl_dev)
380 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
381 	else
382 		ip6_tnl_unlink(ip6n, t);
383 	dst_cache_reset(&t->dst_cache);
384 	netdev_put(dev, &t->dev_tracker);
385 }
386 
387 /**
388  * ip6_tnl_parse_tlv_enc_lim - handle encapsulation limit option
389  *   @skb: received socket buffer
390  *   @raw: the ICMPv6 error message data
391  *
392  * Return:
393  *   0 if none was found,
394  *   else index to encapsulation limit
395  **/
396 
ip6_tnl_parse_tlv_enc_lim(struct sk_buff * skb,__u8 * raw)397 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
398 {
399 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
400 	unsigned int nhoff = raw - skb->data;
401 	unsigned int off = nhoff + sizeof(*ipv6h);
402 	u8 nexthdr = ipv6h->nexthdr;
403 
404 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
405 		struct ipv6_opt_hdr *hdr;
406 		u16 optlen;
407 
408 		if (!pskb_may_pull(skb, off + sizeof(*hdr)))
409 			break;
410 
411 		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
412 		if (nexthdr == NEXTHDR_FRAGMENT) {
413 			optlen = 8;
414 		} else if (nexthdr == NEXTHDR_AUTH) {
415 			optlen = ipv6_authlen(hdr);
416 		} else {
417 			optlen = ipv6_optlen(hdr);
418 		}
419 
420 		if (!pskb_may_pull(skb, off + optlen))
421 			break;
422 
423 		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
424 		if (nexthdr == NEXTHDR_FRAGMENT) {
425 			struct frag_hdr *frag_hdr = (struct frag_hdr *)hdr;
426 
427 			if (frag_hdr->frag_off)
428 				break;
429 		}
430 		if (nexthdr == NEXTHDR_DEST) {
431 			u16 i = 2;
432 
433 			while (1) {
434 				struct ipv6_tlv_tnl_enc_lim *tel;
435 
436 				/* No more room for encapsulation limit */
437 				if (i + sizeof(*tel) > optlen)
438 					break;
439 
440 				tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
441 				/* return index of option if found and valid */
442 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
443 				    tel->length == 1)
444 					return i + off - nhoff;
445 				/* else jump to next option */
446 				if (tel->type)
447 					i += tel->length + 2;
448 				else
449 					i++;
450 			}
451 		}
452 		nexthdr = hdr->nexthdr;
453 		off += optlen;
454 	}
455 	return 0;
456 }
457 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
458 
459 /* ip6_tnl_err() should handle errors in the tunnel according to the
460  * specifications in RFC 2473.
461  */
462 static int
ip6_tnl_err(struct sk_buff * skb,__u8 ipproto,struct inet6_skb_parm * opt,u8 * type,u8 * code,int * msg,__u32 * info,int offset)463 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
464 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
465 {
466 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
467 	struct net *net = dev_net(skb->dev);
468 	u8 rel_type = ICMPV6_DEST_UNREACH;
469 	u8 rel_code = ICMPV6_ADDR_UNREACH;
470 	__u32 rel_info = 0;
471 	struct ip6_tnl *t;
472 	int err = -ENOENT;
473 	int rel_msg = 0;
474 	u8 tproto;
475 	__u16 len;
476 
477 	/* If the packet doesn't contain the original IPv6 header we are
478 	   in trouble since we might need the source address for further
479 	   processing of the error. */
480 
481 	rcu_read_lock();
482 	t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->daddr, &ipv6h->saddr);
483 	if (!t)
484 		goto out;
485 
486 	tproto = READ_ONCE(t->parms.proto);
487 	if (tproto != ipproto && tproto != 0)
488 		goto out;
489 
490 	err = 0;
491 
492 	switch (*type) {
493 	case ICMPV6_DEST_UNREACH:
494 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
495 				    t->parms.name);
496 		rel_msg = 1;
497 		break;
498 	case ICMPV6_TIME_EXCEED:
499 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
500 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
501 					    t->parms.name);
502 			rel_msg = 1;
503 		}
504 		break;
505 	case ICMPV6_PARAMPROB: {
506 		struct ipv6_tlv_tnl_enc_lim *tel;
507 		__u32 teli;
508 
509 		teli = 0;
510 		if ((*code) == ICMPV6_HDR_FIELD)
511 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
512 
513 		if (teli && teli == *info - 2) {
514 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
515 			if (tel->encap_limit == 0) {
516 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
517 						    t->parms.name);
518 				rel_msg = 1;
519 			}
520 		} else {
521 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
522 					    t->parms.name);
523 		}
524 		break;
525 	}
526 	case ICMPV6_PKT_TOOBIG: {
527 		__u32 mtu;
528 
529 		ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
530 				sock_net_uid(net, NULL));
531 		mtu = *info - offset;
532 		if (mtu < IPV6_MIN_MTU)
533 			mtu = IPV6_MIN_MTU;
534 		len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
535 		if (len > mtu) {
536 			rel_type = ICMPV6_PKT_TOOBIG;
537 			rel_code = 0;
538 			rel_info = mtu;
539 			rel_msg = 1;
540 		}
541 		break;
542 	}
543 	case NDISC_REDIRECT:
544 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
545 			     sock_net_uid(net, NULL));
546 		break;
547 	}
548 
549 	*type = rel_type;
550 	*code = rel_code;
551 	*info = rel_info;
552 	*msg = rel_msg;
553 
554 out:
555 	rcu_read_unlock();
556 	return err;
557 }
558 
559 static int
ip4ip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)560 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
561 	   u8 type, u8 code, int offset, __be32 info)
562 {
563 	__u32 rel_info = ntohl(info);
564 	const struct iphdr *eiph;
565 	struct sk_buff *skb2;
566 	int err, rel_msg = 0;
567 	u8 rel_type = type;
568 	u8 rel_code = code;
569 	struct rtable *rt;
570 	struct flowi4 fl4;
571 
572 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
573 			  &rel_msg, &rel_info, offset);
574 	if (err < 0)
575 		return err;
576 
577 	if (rel_msg == 0)
578 		return 0;
579 
580 	switch (rel_type) {
581 	case ICMPV6_DEST_UNREACH:
582 		if (rel_code != ICMPV6_ADDR_UNREACH)
583 			return 0;
584 		rel_type = ICMP_DEST_UNREACH;
585 		rel_code = ICMP_HOST_UNREACH;
586 		break;
587 	case ICMPV6_PKT_TOOBIG:
588 		if (rel_code != 0)
589 			return 0;
590 		rel_type = ICMP_DEST_UNREACH;
591 		rel_code = ICMP_FRAG_NEEDED;
592 		break;
593 	default:
594 		return 0;
595 	}
596 
597 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
598 		return 0;
599 
600 	skb2 = skb_clone(skb, GFP_ATOMIC);
601 	if (!skb2)
602 		return 0;
603 
604 	skb_dst_drop(skb2);
605 
606 	skb_pull(skb2, offset);
607 	skb_reset_network_header(skb2);
608 	eiph = ip_hdr(skb2);
609 
610 	/* Try to guess incoming interface */
611 	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
612 				   0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
613 	if (IS_ERR(rt))
614 		goto out;
615 
616 	skb2->dev = rt->dst.dev;
617 	ip_rt_put(rt);
618 
619 	/* route "incoming" packet */
620 	if (rt->rt_flags & RTCF_LOCAL) {
621 		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
622 					   eiph->daddr, eiph->saddr, 0, 0,
623 					   IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
624 		if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) {
625 			if (!IS_ERR(rt))
626 				ip_rt_put(rt);
627 			goto out;
628 		}
629 		skb_dst_set(skb2, &rt->dst);
630 	} else {
631 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
632 				   skb2->dev) ||
633 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
634 			goto out;
635 	}
636 
637 	/* change mtu on this route */
638 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
639 		if (rel_info > dst_mtu(skb_dst(skb2)))
640 			goto out;
641 
642 		skb_dst_update_pmtu_no_confirm(skb2, rel_info);
643 	}
644 
645 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
646 
647 out:
648 	kfree_skb(skb2);
649 	return 0;
650 }
651 
652 static int
ip6ip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)653 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
654 	   u8 type, u8 code, int offset, __be32 info)
655 {
656 	__u32 rel_info = ntohl(info);
657 	int err, rel_msg = 0;
658 	u8 rel_type = type;
659 	u8 rel_code = code;
660 
661 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
662 			  &rel_msg, &rel_info, offset);
663 	if (err < 0)
664 		return err;
665 
666 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
667 		struct rt6_info *rt;
668 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
669 
670 		if (!skb2)
671 			return 0;
672 
673 		skb_dst_drop(skb2);
674 		skb_pull(skb2, offset);
675 		skb_reset_network_header(skb2);
676 
677 		/* Try to guess incoming interface */
678 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
679 				NULL, 0, skb2, 0);
680 
681 		if (rt && rt->dst.dev)
682 			skb2->dev = rt->dst.dev;
683 
684 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
685 
686 		ip6_rt_put(rt);
687 
688 		kfree_skb(skb2);
689 	}
690 
691 	return 0;
692 }
693 
694 static int
mplsip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)695 mplsip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
696 	    u8 type, u8 code, int offset, __be32 info)
697 {
698 	__u32 rel_info = ntohl(info);
699 	int err, rel_msg = 0;
700 	u8 rel_type = type;
701 	u8 rel_code = code;
702 
703 	err = ip6_tnl_err(skb, IPPROTO_MPLS, opt, &rel_type, &rel_code,
704 			  &rel_msg, &rel_info, offset);
705 	return err;
706 }
707 
ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)708 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
709 				       const struct ipv6hdr *ipv6h,
710 				       struct sk_buff *skb)
711 {
712 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
713 
714 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
715 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
716 
717 	return IP6_ECN_decapsulate(ipv6h, skb);
718 }
719 
ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)720 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
721 				       const struct ipv6hdr *ipv6h,
722 				       struct sk_buff *skb)
723 {
724 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
725 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
726 
727 	return IP6_ECN_decapsulate(ipv6h, skb);
728 }
729 
mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)730 static inline int mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
731 					       const struct ipv6hdr *ipv6h,
732 					       struct sk_buff *skb)
733 {
734 	/* ECN is not supported in AF_MPLS */
735 	return 0;
736 }
737 
ip6_tnl_get_cap(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)738 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
739 			     const struct in6_addr *laddr,
740 			     const struct in6_addr *raddr)
741 {
742 	struct __ip6_tnl_parm *p = &t->parms;
743 	int ltype = ipv6_addr_type(laddr);
744 	int rtype = ipv6_addr_type(raddr);
745 	__u32 flags = 0;
746 
747 	if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
748 		flags = IP6_TNL_F_CAP_PER_PACKET;
749 	} else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
750 		   rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
751 		   !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
752 		   (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
753 		if (ltype&IPV6_ADDR_UNICAST)
754 			flags |= IP6_TNL_F_CAP_XMIT;
755 		if (rtype&IPV6_ADDR_UNICAST)
756 			flags |= IP6_TNL_F_CAP_RCV;
757 	}
758 	return flags;
759 }
760 EXPORT_SYMBOL(ip6_tnl_get_cap);
761 
762 /* called with rcu_read_lock() */
ip6_tnl_rcv_ctl(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)763 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
764 				  const struct in6_addr *laddr,
765 				  const struct in6_addr *raddr)
766 {
767 	struct __ip6_tnl_parm *p = &t->parms;
768 	int ret = 0;
769 	struct net *net = t->net;
770 
771 	if ((p->flags & IP6_TNL_F_CAP_RCV) ||
772 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
773 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
774 		struct net_device *ldev = NULL;
775 
776 		if (p->link)
777 			ldev = dev_get_by_index_rcu(net, p->link);
778 
779 		if ((ipv6_addr_is_multicast(laddr) ||
780 		     likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
781 						    0, IFA_F_TENTATIVE))) &&
782 		    ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
783 		     likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
784 						     0, IFA_F_TENTATIVE))))
785 			ret = 1;
786 	}
787 	return ret;
788 }
789 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
790 
__ip6_tnl_rcv(struct ip6_tnl * tunnel,struct sk_buff * skb,const struct tnl_ptk_info * tpi,struct metadata_dst * tun_dst,int (* dscp_ecn_decapsulate)(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb),bool log_ecn_err)791 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
792 			 const struct tnl_ptk_info *tpi,
793 			 struct metadata_dst *tun_dst,
794 			 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
795 						const struct ipv6hdr *ipv6h,
796 						struct sk_buff *skb),
797 			 bool log_ecn_err)
798 {
799 	const struct ipv6hdr *ipv6h;
800 	int nh, err;
801 
802 	if ((!(tpi->flags & TUNNEL_CSUM) &&
803 	     (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
804 	    ((tpi->flags & TUNNEL_CSUM) &&
805 	     !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
806 		DEV_STATS_INC(tunnel->dev, rx_crc_errors);
807 		DEV_STATS_INC(tunnel->dev, rx_errors);
808 		goto drop;
809 	}
810 
811 	if (tunnel->parms.i_flags & TUNNEL_SEQ) {
812 		if (!(tpi->flags & TUNNEL_SEQ) ||
813 		    (tunnel->i_seqno &&
814 		     (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
815 			DEV_STATS_INC(tunnel->dev, rx_fifo_errors);
816 			DEV_STATS_INC(tunnel->dev, rx_errors);
817 			goto drop;
818 		}
819 		tunnel->i_seqno = ntohl(tpi->seq) + 1;
820 	}
821 
822 	skb->protocol = tpi->proto;
823 
824 	/* Warning: All skb pointers will be invalidated! */
825 	if (tunnel->dev->type == ARPHRD_ETHER) {
826 		if (!pskb_may_pull(skb, ETH_HLEN)) {
827 			DEV_STATS_INC(tunnel->dev, rx_length_errors);
828 			DEV_STATS_INC(tunnel->dev, rx_errors);
829 			goto drop;
830 		}
831 
832 		skb->protocol = eth_type_trans(skb, tunnel->dev);
833 		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
834 	} else {
835 		skb->dev = tunnel->dev;
836 		skb_reset_mac_header(skb);
837 	}
838 
839 	/* Save offset of outer header relative to skb->head,
840 	 * because we are going to reset the network header to the inner header
841 	 * and might change skb->head.
842 	 */
843 	nh = skb_network_header(skb) - skb->head;
844 
845 	skb_reset_network_header(skb);
846 
847 	if (!pskb_inet_may_pull(skb)) {
848 		DEV_STATS_INC(tunnel->dev, rx_length_errors);
849 		DEV_STATS_INC(tunnel->dev, rx_errors);
850 		goto drop;
851 	}
852 
853 	/* Get the outer header. */
854 	ipv6h = (struct ipv6hdr *)(skb->head + nh);
855 
856 	memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
857 
858 	__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
859 
860 	err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
861 	if (unlikely(err)) {
862 		if (log_ecn_err)
863 			net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
864 					     &ipv6h->saddr,
865 					     ipv6_get_dsfield(ipv6h));
866 		if (err > 1) {
867 			DEV_STATS_INC(tunnel->dev, rx_frame_errors);
868 			DEV_STATS_INC(tunnel->dev, rx_errors);
869 			goto drop;
870 		}
871 	}
872 
873 	dev_sw_netstats_rx_add(tunnel->dev, skb->len);
874 
875 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
876 
877 	if (tun_dst)
878 		skb_dst_set(skb, (struct dst_entry *)tun_dst);
879 
880 	gro_cells_receive(&tunnel->gro_cells, skb);
881 	return 0;
882 
883 drop:
884 	if (tun_dst)
885 		dst_release((struct dst_entry *)tun_dst);
886 	kfree_skb(skb);
887 	return 0;
888 }
889 
ip6_tnl_rcv(struct ip6_tnl * t,struct sk_buff * skb,const struct tnl_ptk_info * tpi,struct metadata_dst * tun_dst,bool log_ecn_err)890 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
891 		const struct tnl_ptk_info *tpi,
892 		struct metadata_dst *tun_dst,
893 		bool log_ecn_err)
894 {
895 	int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
896 				    const struct ipv6hdr *ipv6h,
897 				    struct sk_buff *skb);
898 
899 	dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
900 	if (tpi->proto == htons(ETH_P_IP))
901 		dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
902 
903 	return __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
904 			     log_ecn_err);
905 }
906 EXPORT_SYMBOL(ip6_tnl_rcv);
907 
908 static const struct tnl_ptk_info tpi_v6 = {
909 	/* no tunnel info required for ipxip6. */
910 	.proto = htons(ETH_P_IPV6),
911 };
912 
913 static const struct tnl_ptk_info tpi_v4 = {
914 	/* no tunnel info required for ipxip6. */
915 	.proto = htons(ETH_P_IP),
916 };
917 
918 static const struct tnl_ptk_info tpi_mpls = {
919 	/* no tunnel info required for mplsip6. */
920 	.proto = htons(ETH_P_MPLS_UC),
921 };
922 
ipxip6_rcv(struct sk_buff * skb,u8 ipproto,const struct tnl_ptk_info * tpi,int (* dscp_ecn_decapsulate)(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb))923 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
924 		      const struct tnl_ptk_info *tpi,
925 		      int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
926 						  const struct ipv6hdr *ipv6h,
927 						  struct sk_buff *skb))
928 {
929 	struct ip6_tnl *t;
930 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
931 	struct metadata_dst *tun_dst = NULL;
932 	int ret = -1;
933 
934 	rcu_read_lock();
935 	t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr);
936 
937 	if (t) {
938 		u8 tproto = READ_ONCE(t->parms.proto);
939 
940 		if (tproto != ipproto && tproto != 0)
941 			goto drop;
942 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
943 			goto drop;
944 		ipv6h = ipv6_hdr(skb);
945 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
946 			goto drop;
947 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
948 			goto drop;
949 		if (t->parms.collect_md) {
950 			tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
951 			if (!tun_dst)
952 				goto drop;
953 		}
954 		ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
955 				    log_ecn_error);
956 	}
957 
958 	rcu_read_unlock();
959 
960 	return ret;
961 
962 drop:
963 	rcu_read_unlock();
964 	kfree_skb(skb);
965 	return 0;
966 }
967 
ip4ip6_rcv(struct sk_buff * skb)968 static int ip4ip6_rcv(struct sk_buff *skb)
969 {
970 	return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
971 			  ip4ip6_dscp_ecn_decapsulate);
972 }
973 
ip6ip6_rcv(struct sk_buff * skb)974 static int ip6ip6_rcv(struct sk_buff *skb)
975 {
976 	return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
977 			  ip6ip6_dscp_ecn_decapsulate);
978 }
979 
mplsip6_rcv(struct sk_buff * skb)980 static int mplsip6_rcv(struct sk_buff *skb)
981 {
982 	return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls,
983 			  mplsip6_dscp_ecn_decapsulate);
984 }
985 
986 struct ipv6_tel_txoption {
987 	struct ipv6_txoptions ops;
988 	__u8 dst_opt[8];
989 };
990 
init_tel_txopt(struct ipv6_tel_txoption * opt,__u8 encap_limit)991 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
992 {
993 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
994 
995 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
996 	opt->dst_opt[3] = 1;
997 	opt->dst_opt[4] = encap_limit;
998 	opt->dst_opt[5] = IPV6_TLV_PADN;
999 	opt->dst_opt[6] = 1;
1000 
1001 	opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
1002 	opt->ops.opt_nflen = 8;
1003 }
1004 
1005 /**
1006  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
1007  *   @t: the outgoing tunnel device
1008  *   @hdr: IPv6 header from the incoming packet
1009  *
1010  * Description:
1011  *   Avoid trivial tunneling loop by checking that tunnel exit-point
1012  *   doesn't match source of incoming packet.
1013  *
1014  * Return:
1015  *   1 if conflict,
1016  *   0 else
1017  **/
1018 
1019 static inline bool
ip6_tnl_addr_conflict(const struct ip6_tnl * t,const struct ipv6hdr * hdr)1020 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
1021 {
1022 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
1023 }
1024 
ip6_tnl_xmit_ctl(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)1025 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
1026 		     const struct in6_addr *laddr,
1027 		     const struct in6_addr *raddr)
1028 {
1029 	struct __ip6_tnl_parm *p = &t->parms;
1030 	int ret = 0;
1031 	struct net *net = t->net;
1032 
1033 	if (t->parms.collect_md)
1034 		return 1;
1035 
1036 	if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1037 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1038 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1039 		struct net_device *ldev = NULL;
1040 
1041 		rcu_read_lock();
1042 		if (p->link)
1043 			ldev = dev_get_by_index_rcu(net, p->link);
1044 
1045 		if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
1046 						      0, IFA_F_TENTATIVE)))
1047 			pr_warn_ratelimited("%s xmit: Local address not yet configured!\n",
1048 					    p->name);
1049 		else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1050 			 !ipv6_addr_is_multicast(raddr) &&
1051 			 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1052 							  true, 0, IFA_F_TENTATIVE)))
1053 			pr_warn_ratelimited("%s xmit: Routing loop! Remote address found on this node!\n",
1054 					    p->name);
1055 		else
1056 			ret = 1;
1057 		rcu_read_unlock();
1058 	}
1059 	return ret;
1060 }
1061 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1062 
1063 /**
1064  * ip6_tnl_xmit - encapsulate packet and send
1065  *   @skb: the outgoing socket buffer
1066  *   @dev: the outgoing tunnel device
1067  *   @dsfield: dscp code for outer header
1068  *   @fl6: flow of tunneled packet
1069  *   @encap_limit: encapsulation limit
1070  *   @pmtu: Path MTU is stored if packet is too big
1071  *   @proto: next header value
1072  *
1073  * Description:
1074  *   Build new header and do some sanity checks on the packet before sending
1075  *   it.
1076  *
1077  * Return:
1078  *   0 on success
1079  *   -1 fail
1080  *   %-EMSGSIZE message too big. return mtu in this case.
1081  **/
1082 
ip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev,__u8 dsfield,struct flowi6 * fl6,int encap_limit,__u32 * pmtu,__u8 proto)1083 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1084 		 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1085 		 __u8 proto)
1086 {
1087 	struct ip6_tnl *t = netdev_priv(dev);
1088 	struct net *net = t->net;
1089 	struct ipv6hdr *ipv6h;
1090 	struct ipv6_tel_txoption opt;
1091 	struct dst_entry *dst = NULL, *ndst = NULL;
1092 	struct net_device *tdev;
1093 	int mtu;
1094 	unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1095 	unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1096 	unsigned int max_headroom = psh_hlen;
1097 	__be16 payload_protocol;
1098 	bool use_cache = false;
1099 	u8 hop_limit;
1100 	int err = -1;
1101 
1102 	payload_protocol = skb_protocol(skb, true);
1103 
1104 	if (t->parms.collect_md) {
1105 		hop_limit = skb_tunnel_info(skb)->key.ttl;
1106 		goto route_lookup;
1107 	} else {
1108 		hop_limit = t->parms.hop_limit;
1109 	}
1110 
1111 	/* NBMA tunnel */
1112 	if (ipv6_addr_any(&t->parms.raddr)) {
1113 		if (payload_protocol == htons(ETH_P_IPV6)) {
1114 			struct in6_addr *addr6;
1115 			struct neighbour *neigh;
1116 			int addr_type;
1117 
1118 			if (!skb_dst(skb))
1119 				goto tx_err_link_failure;
1120 
1121 			neigh = dst_neigh_lookup(skb_dst(skb),
1122 						 &ipv6_hdr(skb)->daddr);
1123 			if (!neigh)
1124 				goto tx_err_link_failure;
1125 
1126 			addr6 = (struct in6_addr *)&neigh->primary_key;
1127 			addr_type = ipv6_addr_type(addr6);
1128 
1129 			if (addr_type == IPV6_ADDR_ANY)
1130 				addr6 = &ipv6_hdr(skb)->daddr;
1131 
1132 			memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1133 			neigh_release(neigh);
1134 		} else if (payload_protocol == htons(ETH_P_IP)) {
1135 			const struct rtable *rt = skb_rtable(skb);
1136 
1137 			if (!rt)
1138 				goto tx_err_link_failure;
1139 
1140 			if (rt->rt_gw_family == AF_INET6)
1141 				memcpy(&fl6->daddr, &rt->rt_gw6, sizeof(fl6->daddr));
1142 		}
1143 	} else if (t->parms.proto != 0 && !(t->parms.flags &
1144 					    (IP6_TNL_F_USE_ORIG_TCLASS |
1145 					     IP6_TNL_F_USE_ORIG_FWMARK))) {
1146 		/* enable the cache only if neither the outer protocol nor the
1147 		 * routing decision depends on the current inner header value
1148 		 */
1149 		use_cache = true;
1150 	}
1151 
1152 	if (use_cache)
1153 		dst = dst_cache_get(&t->dst_cache);
1154 
1155 	if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1156 		goto tx_err_link_failure;
1157 
1158 	if (!dst) {
1159 route_lookup:
1160 		/* add dsfield to flowlabel for route lookup */
1161 		fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1162 
1163 		dst = ip6_route_output(net, NULL, fl6);
1164 
1165 		if (dst->error)
1166 			goto tx_err_link_failure;
1167 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1168 		if (IS_ERR(dst)) {
1169 			err = PTR_ERR(dst);
1170 			dst = NULL;
1171 			goto tx_err_link_failure;
1172 		}
1173 		if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1174 		    ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1175 				       &fl6->daddr, 0, &fl6->saddr))
1176 			goto tx_err_link_failure;
1177 		ndst = dst;
1178 	}
1179 
1180 	tdev = dst->dev;
1181 
1182 	if (tdev == dev) {
1183 		DEV_STATS_INC(dev, collisions);
1184 		net_warn_ratelimited("%s: Local routing loop detected!\n",
1185 				     t->parms.name);
1186 		goto tx_err_dst_release;
1187 	}
1188 	mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1189 	if (encap_limit >= 0) {
1190 		max_headroom += 8;
1191 		mtu -= 8;
1192 	}
1193 	mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1194 		       IPV6_MIN_MTU : IPV4_MIN_MTU);
1195 
1196 	skb_dst_update_pmtu_no_confirm(skb, mtu);
1197 	if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1198 		*pmtu = mtu;
1199 		err = -EMSGSIZE;
1200 		goto tx_err_dst_release;
1201 	}
1202 
1203 	if (t->err_count > 0) {
1204 		if (time_before(jiffies,
1205 				t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1206 			t->err_count--;
1207 
1208 			dst_link_failure(skb);
1209 		} else {
1210 			t->err_count = 0;
1211 		}
1212 	}
1213 
1214 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1215 
1216 	/*
1217 	 * Okay, now see if we can stuff it in the buffer as-is.
1218 	 */
1219 	max_headroom += LL_RESERVED_SPACE(tdev);
1220 
1221 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1222 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1223 		struct sk_buff *new_skb;
1224 
1225 		new_skb = skb_realloc_headroom(skb, max_headroom);
1226 		if (!new_skb)
1227 			goto tx_err_dst_release;
1228 
1229 		if (skb->sk)
1230 			skb_set_owner_w(new_skb, skb->sk);
1231 		consume_skb(skb);
1232 		skb = new_skb;
1233 	}
1234 
1235 	if (t->parms.collect_md) {
1236 		if (t->encap.type != TUNNEL_ENCAP_NONE)
1237 			goto tx_err_dst_release;
1238 	} else {
1239 		if (use_cache && ndst)
1240 			dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1241 	}
1242 	skb_dst_set(skb, dst);
1243 
1244 	if (hop_limit == 0) {
1245 		if (payload_protocol == htons(ETH_P_IP))
1246 			hop_limit = ip_hdr(skb)->ttl;
1247 		else if (payload_protocol == htons(ETH_P_IPV6))
1248 			hop_limit = ipv6_hdr(skb)->hop_limit;
1249 		else
1250 			hop_limit = ip6_dst_hoplimit(dst);
1251 	}
1252 
1253 	/* Calculate max headroom for all the headers and adjust
1254 	 * needed_headroom if necessary.
1255 	 */
1256 	max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1257 			+ dst->header_len + t->hlen;
1258 	if (max_headroom > READ_ONCE(dev->needed_headroom))
1259 		WRITE_ONCE(dev->needed_headroom, max_headroom);
1260 
1261 	err = ip6_tnl_encap(skb, t, &proto, fl6);
1262 	if (err)
1263 		return err;
1264 
1265 	if (encap_limit >= 0) {
1266 		init_tel_txopt(&opt, encap_limit);
1267 		ipv6_push_frag_opts(skb, &opt.ops, &proto);
1268 	}
1269 
1270 	skb_push(skb, sizeof(struct ipv6hdr));
1271 	skb_reset_network_header(skb);
1272 	ipv6h = ipv6_hdr(skb);
1273 	ip6_flow_hdr(ipv6h, dsfield,
1274 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1275 	ipv6h->hop_limit = hop_limit;
1276 	ipv6h->nexthdr = proto;
1277 	ipv6h->saddr = fl6->saddr;
1278 	ipv6h->daddr = fl6->daddr;
1279 	ip6tunnel_xmit(NULL, skb, dev);
1280 	return 0;
1281 tx_err_link_failure:
1282 	DEV_STATS_INC(dev, tx_carrier_errors);
1283 	dst_link_failure(skb);
1284 tx_err_dst_release:
1285 	dst_release(dst);
1286 	return err;
1287 }
1288 EXPORT_SYMBOL(ip6_tnl_xmit);
1289 
1290 static inline int
ipxip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev,u8 protocol)1291 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev,
1292 		u8 protocol)
1293 {
1294 	struct ip6_tnl *t = netdev_priv(dev);
1295 	struct ipv6hdr *ipv6h;
1296 	const struct iphdr  *iph;
1297 	int encap_limit = -1;
1298 	__u16 offset;
1299 	struct flowi6 fl6;
1300 	__u8 dsfield, orig_dsfield;
1301 	__u32 mtu;
1302 	u8 tproto;
1303 	int err;
1304 
1305 	tproto = READ_ONCE(t->parms.proto);
1306 	if (tproto != protocol && tproto != 0)
1307 		return -1;
1308 
1309 	if (t->parms.collect_md) {
1310 		struct ip_tunnel_info *tun_info;
1311 		const struct ip_tunnel_key *key;
1312 
1313 		tun_info = skb_tunnel_info(skb);
1314 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1315 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1316 			return -1;
1317 		key = &tun_info->key;
1318 		memset(&fl6, 0, sizeof(fl6));
1319 		fl6.flowi6_proto = protocol;
1320 		fl6.saddr = key->u.ipv6.src;
1321 		fl6.daddr = key->u.ipv6.dst;
1322 		fl6.flowlabel = key->label;
1323 		dsfield =  key->tos;
1324 		switch (protocol) {
1325 		case IPPROTO_IPIP:
1326 			iph = ip_hdr(skb);
1327 			orig_dsfield = ipv4_get_dsfield(iph);
1328 			break;
1329 		case IPPROTO_IPV6:
1330 			ipv6h = ipv6_hdr(skb);
1331 			orig_dsfield = ipv6_get_dsfield(ipv6h);
1332 			break;
1333 		default:
1334 			orig_dsfield = dsfield;
1335 			break;
1336 		}
1337 	} else {
1338 		if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1339 			encap_limit = t->parms.encap_limit;
1340 		if (protocol == IPPROTO_IPV6) {
1341 			offset = ip6_tnl_parse_tlv_enc_lim(skb,
1342 						skb_network_header(skb));
1343 			/* ip6_tnl_parse_tlv_enc_lim() might have
1344 			 * reallocated skb->head
1345 			 */
1346 			if (offset > 0) {
1347 				struct ipv6_tlv_tnl_enc_lim *tel;
1348 
1349 				tel = (void *)&skb_network_header(skb)[offset];
1350 				if (tel->encap_limit == 0) {
1351 					icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
1352 							ICMPV6_HDR_FIELD, offset + 2);
1353 					return -1;
1354 				}
1355 				encap_limit = tel->encap_limit - 1;
1356 			}
1357 		}
1358 
1359 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1360 		fl6.flowi6_proto = protocol;
1361 
1362 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1363 			fl6.flowi6_mark = skb->mark;
1364 		else
1365 			fl6.flowi6_mark = t->parms.fwmark;
1366 		switch (protocol) {
1367 		case IPPROTO_IPIP:
1368 			iph = ip_hdr(skb);
1369 			orig_dsfield = ipv4_get_dsfield(iph);
1370 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1371 				dsfield = orig_dsfield;
1372 			else
1373 				dsfield = ip6_tclass(t->parms.flowinfo);
1374 			break;
1375 		case IPPROTO_IPV6:
1376 			ipv6h = ipv6_hdr(skb);
1377 			orig_dsfield = ipv6_get_dsfield(ipv6h);
1378 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1379 				dsfield = orig_dsfield;
1380 			else
1381 				dsfield = ip6_tclass(t->parms.flowinfo);
1382 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1383 				fl6.flowlabel |= ip6_flowlabel(ipv6h);
1384 			break;
1385 		default:
1386 			orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo);
1387 			break;
1388 		}
1389 	}
1390 
1391 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1392 	dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield);
1393 
1394 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1395 		return -1;
1396 
1397 	skb_set_inner_ipproto(skb, protocol);
1398 
1399 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1400 			   protocol);
1401 	if (err != 0) {
1402 		/* XXX: send ICMP error even if DF is not set. */
1403 		if (err == -EMSGSIZE)
1404 			switch (protocol) {
1405 			case IPPROTO_IPIP:
1406 				icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1407 					      ICMP_FRAG_NEEDED, htonl(mtu));
1408 				break;
1409 			case IPPROTO_IPV6:
1410 				icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1411 				break;
1412 			default:
1413 				break;
1414 			}
1415 		return -1;
1416 	}
1417 
1418 	return 0;
1419 }
1420 
1421 static netdev_tx_t
ip6_tnl_start_xmit(struct sk_buff * skb,struct net_device * dev)1422 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1423 {
1424 	struct ip6_tnl *t = netdev_priv(dev);
1425 	u8 ipproto;
1426 	int ret;
1427 
1428 	if (!pskb_inet_may_pull(skb))
1429 		goto tx_err;
1430 
1431 	switch (skb->protocol) {
1432 	case htons(ETH_P_IP):
1433 		ipproto = IPPROTO_IPIP;
1434 		break;
1435 	case htons(ETH_P_IPV6):
1436 		if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb)))
1437 			goto tx_err;
1438 		ipproto = IPPROTO_IPV6;
1439 		break;
1440 	case htons(ETH_P_MPLS_UC):
1441 		ipproto = IPPROTO_MPLS;
1442 		break;
1443 	default:
1444 		goto tx_err;
1445 	}
1446 
1447 	ret = ipxip6_tnl_xmit(skb, dev, ipproto);
1448 	if (ret < 0)
1449 		goto tx_err;
1450 
1451 	return NETDEV_TX_OK;
1452 
1453 tx_err:
1454 	DEV_STATS_INC(dev, tx_errors);
1455 	DEV_STATS_INC(dev, tx_dropped);
1456 	kfree_skb(skb);
1457 	return NETDEV_TX_OK;
1458 }
1459 
ip6_tnl_link_config(struct ip6_tnl * t)1460 static void ip6_tnl_link_config(struct ip6_tnl *t)
1461 {
1462 	struct net_device *dev = t->dev;
1463 	struct net_device *tdev = NULL;
1464 	struct __ip6_tnl_parm *p = &t->parms;
1465 	struct flowi6 *fl6 = &t->fl.u.ip6;
1466 	int t_hlen;
1467 	int mtu;
1468 
1469 	__dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
1470 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1471 
1472 	/* Set up flowi template */
1473 	fl6->saddr = p->laddr;
1474 	fl6->daddr = p->raddr;
1475 	fl6->flowi6_oif = p->link;
1476 	fl6->flowlabel = 0;
1477 
1478 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1479 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1480 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1481 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1482 
1483 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1484 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1485 
1486 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1487 		dev->flags |= IFF_POINTOPOINT;
1488 	else
1489 		dev->flags &= ~IFF_POINTOPOINT;
1490 
1491 	t->tun_hlen = 0;
1492 	t->hlen = t->encap_hlen + t->tun_hlen;
1493 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1494 
1495 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1496 		int strict = (ipv6_addr_type(&p->raddr) &
1497 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1498 
1499 		struct rt6_info *rt = rt6_lookup(t->net,
1500 						 &p->raddr, &p->laddr,
1501 						 p->link, NULL, strict);
1502 		if (rt) {
1503 			tdev = rt->dst.dev;
1504 			ip6_rt_put(rt);
1505 		}
1506 
1507 		if (!tdev && p->link)
1508 			tdev = __dev_get_by_index(t->net, p->link);
1509 
1510 		if (tdev) {
1511 			dev->hard_header_len = tdev->hard_header_len + t_hlen;
1512 			mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU);
1513 
1514 			mtu = mtu - t_hlen;
1515 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1516 				mtu -= 8;
1517 
1518 			if (mtu < IPV6_MIN_MTU)
1519 				mtu = IPV6_MIN_MTU;
1520 			WRITE_ONCE(dev->mtu, mtu);
1521 		}
1522 	}
1523 }
1524 
1525 /**
1526  * ip6_tnl_change - update the tunnel parameters
1527  *   @t: tunnel to be changed
1528  *   @p: tunnel configuration parameters
1529  *
1530  * Description:
1531  *   ip6_tnl_change() updates the tunnel parameters
1532  **/
1533 
1534 static void
ip6_tnl_change(struct ip6_tnl * t,const struct __ip6_tnl_parm * p)1535 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1536 {
1537 	t->parms.laddr = p->laddr;
1538 	t->parms.raddr = p->raddr;
1539 	t->parms.flags = p->flags;
1540 	t->parms.hop_limit = p->hop_limit;
1541 	t->parms.encap_limit = p->encap_limit;
1542 	t->parms.flowinfo = p->flowinfo;
1543 	t->parms.link = p->link;
1544 	t->parms.proto = p->proto;
1545 	t->parms.fwmark = p->fwmark;
1546 	dst_cache_reset(&t->dst_cache);
1547 	ip6_tnl_link_config(t);
1548 }
1549 
ip6_tnl_update(struct ip6_tnl * t,struct __ip6_tnl_parm * p)1550 static void ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1551 {
1552 	struct net *net = t->net;
1553 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1554 
1555 	ip6_tnl_unlink(ip6n, t);
1556 	synchronize_net();
1557 	ip6_tnl_change(t, p);
1558 	ip6_tnl_link(ip6n, t);
1559 	netdev_state_change(t->dev);
1560 }
1561 
ip6_tnl0_update(struct ip6_tnl * t,struct __ip6_tnl_parm * p)1562 static void ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1563 {
1564 	/* for default tnl0 device allow to change only the proto */
1565 	t->parms.proto = p->proto;
1566 	netdev_state_change(t->dev);
1567 }
1568 
1569 static void
ip6_tnl_parm_from_user(struct __ip6_tnl_parm * p,const struct ip6_tnl_parm * u)1570 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1571 {
1572 	p->laddr = u->laddr;
1573 	p->raddr = u->raddr;
1574 	p->flags = u->flags;
1575 	p->hop_limit = u->hop_limit;
1576 	p->encap_limit = u->encap_limit;
1577 	p->flowinfo = u->flowinfo;
1578 	p->link = u->link;
1579 	p->proto = u->proto;
1580 	memcpy(p->name, u->name, sizeof(u->name));
1581 }
1582 
1583 static void
ip6_tnl_parm_to_user(struct ip6_tnl_parm * u,const struct __ip6_tnl_parm * p)1584 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1585 {
1586 	u->laddr = p->laddr;
1587 	u->raddr = p->raddr;
1588 	u->flags = p->flags;
1589 	u->hop_limit = p->hop_limit;
1590 	u->encap_limit = p->encap_limit;
1591 	u->flowinfo = p->flowinfo;
1592 	u->link = p->link;
1593 	u->proto = p->proto;
1594 	memcpy(u->name, p->name, sizeof(u->name));
1595 }
1596 
1597 /**
1598  * ip6_tnl_siocdevprivate - configure ipv6 tunnels from userspace
1599  *   @dev: virtual device associated with tunnel
1600  *   @ifr: unused
1601  *   @data: parameters passed from userspace
1602  *   @cmd: command to be performed
1603  *
1604  * Description:
1605  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1606  *   from userspace.
1607  *
1608  *   The possible commands are the following:
1609  *     %SIOCGETTUNNEL: get tunnel parameters for device
1610  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1611  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1612  *     %SIOCDELTUNNEL: delete tunnel
1613  *
1614  *   The fallback device "ip6tnl0", created during module
1615  *   initialization, can be used for creating other tunnel devices.
1616  *
1617  * Return:
1618  *   0 on success,
1619  *   %-EFAULT if unable to copy data to or from userspace,
1620  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1621  *   %-EINVAL if passed tunnel parameters are invalid,
1622  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1623  *   %-ENODEV if attempting to change or delete a nonexisting device
1624  **/
1625 
1626 static int
ip6_tnl_siocdevprivate(struct net_device * dev,struct ifreq * ifr,void __user * data,int cmd)1627 ip6_tnl_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1628 		       void __user *data, int cmd)
1629 {
1630 	int err = 0;
1631 	struct ip6_tnl_parm p;
1632 	struct __ip6_tnl_parm p1;
1633 	struct ip6_tnl *t = netdev_priv(dev);
1634 	struct net *net = t->net;
1635 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1636 
1637 	memset(&p1, 0, sizeof(p1));
1638 
1639 	switch (cmd) {
1640 	case SIOCGETTUNNEL:
1641 		if (dev == ip6n->fb_tnl_dev) {
1642 			if (copy_from_user(&p, data, sizeof(p))) {
1643 				err = -EFAULT;
1644 				break;
1645 			}
1646 			ip6_tnl_parm_from_user(&p1, &p);
1647 			t = ip6_tnl_locate(net, &p1, 0);
1648 			if (IS_ERR(t))
1649 				t = netdev_priv(dev);
1650 		} else {
1651 			memset(&p, 0, sizeof(p));
1652 		}
1653 		ip6_tnl_parm_to_user(&p, &t->parms);
1654 		if (copy_to_user(data, &p, sizeof(p)))
1655 			err = -EFAULT;
1656 		break;
1657 	case SIOCADDTUNNEL:
1658 	case SIOCCHGTUNNEL:
1659 		err = -EPERM;
1660 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1661 			break;
1662 		err = -EFAULT;
1663 		if (copy_from_user(&p, data, sizeof(p)))
1664 			break;
1665 		err = -EINVAL;
1666 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1667 		    p.proto != 0)
1668 			break;
1669 		ip6_tnl_parm_from_user(&p1, &p);
1670 		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1671 		if (cmd == SIOCCHGTUNNEL) {
1672 			if (!IS_ERR(t)) {
1673 				if (t->dev != dev) {
1674 					err = -EEXIST;
1675 					break;
1676 				}
1677 			} else
1678 				t = netdev_priv(dev);
1679 			if (dev == ip6n->fb_tnl_dev)
1680 				ip6_tnl0_update(t, &p1);
1681 			else
1682 				ip6_tnl_update(t, &p1);
1683 		}
1684 		if (!IS_ERR(t)) {
1685 			err = 0;
1686 			ip6_tnl_parm_to_user(&p, &t->parms);
1687 			if (copy_to_user(data, &p, sizeof(p)))
1688 				err = -EFAULT;
1689 
1690 		} else {
1691 			err = PTR_ERR(t);
1692 		}
1693 		break;
1694 	case SIOCDELTUNNEL:
1695 		err = -EPERM;
1696 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1697 			break;
1698 
1699 		if (dev == ip6n->fb_tnl_dev) {
1700 			err = -EFAULT;
1701 			if (copy_from_user(&p, data, sizeof(p)))
1702 				break;
1703 			err = -ENOENT;
1704 			ip6_tnl_parm_from_user(&p1, &p);
1705 			t = ip6_tnl_locate(net, &p1, 0);
1706 			if (IS_ERR(t))
1707 				break;
1708 			err = -EPERM;
1709 			if (t->dev == ip6n->fb_tnl_dev)
1710 				break;
1711 			dev = t->dev;
1712 		}
1713 		err = 0;
1714 		unregister_netdevice(dev);
1715 		break;
1716 	default:
1717 		err = -EINVAL;
1718 	}
1719 	return err;
1720 }
1721 
1722 /**
1723  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1724  *   @dev: virtual device associated with tunnel
1725  *   @new_mtu: the new mtu
1726  *
1727  * Return:
1728  *   0 on success,
1729  *   %-EINVAL if mtu too small
1730  **/
1731 
ip6_tnl_change_mtu(struct net_device * dev,int new_mtu)1732 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1733 {
1734 	struct ip6_tnl *tnl = netdev_priv(dev);
1735 
1736 	if (tnl->parms.proto == IPPROTO_IPV6) {
1737 		if (new_mtu < IPV6_MIN_MTU)
1738 			return -EINVAL;
1739 	} else {
1740 		if (new_mtu < ETH_MIN_MTU)
1741 			return -EINVAL;
1742 	}
1743 	if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1744 		if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1745 			return -EINVAL;
1746 	} else {
1747 		if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1748 			return -EINVAL;
1749 	}
1750 	dev->mtu = new_mtu;
1751 	return 0;
1752 }
1753 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1754 
ip6_tnl_get_iflink(const struct net_device * dev)1755 int ip6_tnl_get_iflink(const struct net_device *dev)
1756 {
1757 	struct ip6_tnl *t = netdev_priv(dev);
1758 
1759 	return t->parms.link;
1760 }
1761 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1762 
ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops * ops,unsigned int num)1763 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1764 			  unsigned int num)
1765 {
1766 	if (num >= MAX_IPTUN_ENCAP_OPS)
1767 		return -ERANGE;
1768 
1769 	return !cmpxchg((const struct ip6_tnl_encap_ops **)
1770 			&ip6tun_encaps[num],
1771 			NULL, ops) ? 0 : -1;
1772 }
1773 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1774 
ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops * ops,unsigned int num)1775 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1776 			  unsigned int num)
1777 {
1778 	int ret;
1779 
1780 	if (num >= MAX_IPTUN_ENCAP_OPS)
1781 		return -ERANGE;
1782 
1783 	ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1784 		       &ip6tun_encaps[num],
1785 		       ops, NULL) == ops) ? 0 : -1;
1786 
1787 	synchronize_net();
1788 
1789 	return ret;
1790 }
1791 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1792 
ip6_tnl_encap_setup(struct ip6_tnl * t,struct ip_tunnel_encap * ipencap)1793 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1794 			struct ip_tunnel_encap *ipencap)
1795 {
1796 	int hlen;
1797 
1798 	memset(&t->encap, 0, sizeof(t->encap));
1799 
1800 	hlen = ip6_encap_hlen(ipencap);
1801 	if (hlen < 0)
1802 		return hlen;
1803 
1804 	t->encap.type = ipencap->type;
1805 	t->encap.sport = ipencap->sport;
1806 	t->encap.dport = ipencap->dport;
1807 	t->encap.flags = ipencap->flags;
1808 
1809 	t->encap_hlen = hlen;
1810 	t->hlen = t->encap_hlen + t->tun_hlen;
1811 
1812 	return 0;
1813 }
1814 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1815 
1816 static const struct net_device_ops ip6_tnl_netdev_ops = {
1817 	.ndo_init	= ip6_tnl_dev_init,
1818 	.ndo_uninit	= ip6_tnl_dev_uninit,
1819 	.ndo_start_xmit = ip6_tnl_start_xmit,
1820 	.ndo_siocdevprivate = ip6_tnl_siocdevprivate,
1821 	.ndo_change_mtu = ip6_tnl_change_mtu,
1822 	.ndo_get_stats64 = dev_get_tstats64,
1823 	.ndo_get_iflink = ip6_tnl_get_iflink,
1824 };
1825 
1826 #define IPXIPX_FEATURES (NETIF_F_SG |		\
1827 			 NETIF_F_FRAGLIST |	\
1828 			 NETIF_F_HIGHDMA |	\
1829 			 NETIF_F_GSO_SOFTWARE |	\
1830 			 NETIF_F_HW_CSUM)
1831 
1832 /**
1833  * ip6_tnl_dev_setup - setup virtual tunnel device
1834  *   @dev: virtual device associated with tunnel
1835  *
1836  * Description:
1837  *   Initialize function pointers and device parameters
1838  **/
1839 
ip6_tnl_dev_setup(struct net_device * dev)1840 static void ip6_tnl_dev_setup(struct net_device *dev)
1841 {
1842 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1843 	dev->header_ops = &ip_tunnel_header_ops;
1844 	dev->needs_free_netdev = true;
1845 	dev->priv_destructor = ip6_dev_free;
1846 
1847 	dev->type = ARPHRD_TUNNEL6;
1848 	dev->flags |= IFF_NOARP;
1849 	dev->addr_len = sizeof(struct in6_addr);
1850 	dev->features |= NETIF_F_LLTX;
1851 	netif_keep_dst(dev);
1852 
1853 	dev->features		|= IPXIPX_FEATURES;
1854 	dev->hw_features	|= IPXIPX_FEATURES;
1855 
1856 	/* This perm addr will be used as interface identifier by IPv6 */
1857 	dev->addr_assign_type = NET_ADDR_RANDOM;
1858 	eth_random_addr(dev->perm_addr);
1859 }
1860 
1861 
1862 /**
1863  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1864  *   @dev: virtual device associated with tunnel
1865  **/
1866 
1867 static inline int
ip6_tnl_dev_init_gen(struct net_device * dev)1868 ip6_tnl_dev_init_gen(struct net_device *dev)
1869 {
1870 	struct ip6_tnl *t = netdev_priv(dev);
1871 	int ret;
1872 	int t_hlen;
1873 
1874 	t->dev = dev;
1875 	t->net = dev_net(dev);
1876 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1877 	if (!dev->tstats)
1878 		return -ENOMEM;
1879 
1880 	ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1881 	if (ret)
1882 		goto free_stats;
1883 
1884 	ret = gro_cells_init(&t->gro_cells, dev);
1885 	if (ret)
1886 		goto destroy_dst;
1887 
1888 	t->tun_hlen = 0;
1889 	t->hlen = t->encap_hlen + t->tun_hlen;
1890 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1891 
1892 	dev->type = ARPHRD_TUNNEL6;
1893 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1894 	dev->mtu = ETH_DATA_LEN - t_hlen;
1895 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1896 		dev->mtu -= 8;
1897 	dev->min_mtu = ETH_MIN_MTU;
1898 	dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1899 
1900 	netdev_hold(dev, &t->dev_tracker, GFP_KERNEL);
1901 	return 0;
1902 
1903 destroy_dst:
1904 	dst_cache_destroy(&t->dst_cache);
1905 free_stats:
1906 	free_percpu(dev->tstats);
1907 	dev->tstats = NULL;
1908 
1909 	return ret;
1910 }
1911 
1912 /**
1913  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1914  *   @dev: virtual device associated with tunnel
1915  **/
1916 
ip6_tnl_dev_init(struct net_device * dev)1917 static int ip6_tnl_dev_init(struct net_device *dev)
1918 {
1919 	struct ip6_tnl *t = netdev_priv(dev);
1920 	int err = ip6_tnl_dev_init_gen(dev);
1921 
1922 	if (err)
1923 		return err;
1924 	ip6_tnl_link_config(t);
1925 	if (t->parms.collect_md)
1926 		netif_keep_dst(dev);
1927 	return 0;
1928 }
1929 
1930 /**
1931  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1932  *   @dev: fallback device
1933  *
1934  * Return: 0
1935  **/
1936 
ip6_fb_tnl_dev_init(struct net_device * dev)1937 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1938 {
1939 	struct ip6_tnl *t = netdev_priv(dev);
1940 	struct net *net = dev_net(dev);
1941 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1942 
1943 	t->parms.proto = IPPROTO_IPV6;
1944 
1945 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1946 	return 0;
1947 }
1948 
ip6_tnl_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1949 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1950 			    struct netlink_ext_ack *extack)
1951 {
1952 	u8 proto;
1953 
1954 	if (!data || !data[IFLA_IPTUN_PROTO])
1955 		return 0;
1956 
1957 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1958 	if (proto != IPPROTO_IPV6 &&
1959 	    proto != IPPROTO_IPIP &&
1960 	    proto != 0)
1961 		return -EINVAL;
1962 
1963 	return 0;
1964 }
1965 
ip6_tnl_netlink_parms(struct nlattr * data[],struct __ip6_tnl_parm * parms)1966 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1967 				  struct __ip6_tnl_parm *parms)
1968 {
1969 	memset(parms, 0, sizeof(*parms));
1970 
1971 	if (!data)
1972 		return;
1973 
1974 	if (data[IFLA_IPTUN_LINK])
1975 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1976 
1977 	if (data[IFLA_IPTUN_LOCAL])
1978 		parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1979 
1980 	if (data[IFLA_IPTUN_REMOTE])
1981 		parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1982 
1983 	if (data[IFLA_IPTUN_TTL])
1984 		parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1985 
1986 	if (data[IFLA_IPTUN_ENCAP_LIMIT])
1987 		parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1988 
1989 	if (data[IFLA_IPTUN_FLOWINFO])
1990 		parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1991 
1992 	if (data[IFLA_IPTUN_FLAGS])
1993 		parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1994 
1995 	if (data[IFLA_IPTUN_PROTO])
1996 		parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1997 
1998 	if (data[IFLA_IPTUN_COLLECT_METADATA])
1999 		parms->collect_md = true;
2000 
2001 	if (data[IFLA_IPTUN_FWMARK])
2002 		parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
2003 }
2004 
ip6_tnl_newlink(struct net * src_net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)2005 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
2006 			   struct nlattr *tb[], struct nlattr *data[],
2007 			   struct netlink_ext_ack *extack)
2008 {
2009 	struct net *net = dev_net(dev);
2010 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2011 	struct ip_tunnel_encap ipencap;
2012 	struct ip6_tnl *nt, *t;
2013 	int err;
2014 
2015 	nt = netdev_priv(dev);
2016 
2017 	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
2018 		err = ip6_tnl_encap_setup(nt, &ipencap);
2019 		if (err < 0)
2020 			return err;
2021 	}
2022 
2023 	ip6_tnl_netlink_parms(data, &nt->parms);
2024 
2025 	if (nt->parms.collect_md) {
2026 		if (rtnl_dereference(ip6n->collect_md_tun))
2027 			return -EEXIST;
2028 	} else {
2029 		t = ip6_tnl_locate(net, &nt->parms, 0);
2030 		if (!IS_ERR(t))
2031 			return -EEXIST;
2032 	}
2033 
2034 	err = ip6_tnl_create2(dev);
2035 	if (!err && tb[IFLA_MTU])
2036 		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2037 
2038 	return err;
2039 }
2040 
ip6_tnl_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)2041 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2042 			      struct nlattr *data[],
2043 			      struct netlink_ext_ack *extack)
2044 {
2045 	struct ip6_tnl *t = netdev_priv(dev);
2046 	struct __ip6_tnl_parm p;
2047 	struct net *net = t->net;
2048 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2049 	struct ip_tunnel_encap ipencap;
2050 
2051 	if (dev == ip6n->fb_tnl_dev)
2052 		return -EINVAL;
2053 
2054 	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
2055 		int err = ip6_tnl_encap_setup(t, &ipencap);
2056 
2057 		if (err < 0)
2058 			return err;
2059 	}
2060 	ip6_tnl_netlink_parms(data, &p);
2061 	if (p.collect_md)
2062 		return -EINVAL;
2063 
2064 	t = ip6_tnl_locate(net, &p, 0);
2065 	if (!IS_ERR(t)) {
2066 		if (t->dev != dev)
2067 			return -EEXIST;
2068 	} else
2069 		t = netdev_priv(dev);
2070 
2071 	ip6_tnl_update(t, &p);
2072 	return 0;
2073 }
2074 
ip6_tnl_dellink(struct net_device * dev,struct list_head * head)2075 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2076 {
2077 	struct net *net = dev_net(dev);
2078 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2079 
2080 	if (dev != ip6n->fb_tnl_dev)
2081 		unregister_netdevice_queue(dev, head);
2082 }
2083 
ip6_tnl_get_size(const struct net_device * dev)2084 static size_t ip6_tnl_get_size(const struct net_device *dev)
2085 {
2086 	return
2087 		/* IFLA_IPTUN_LINK */
2088 		nla_total_size(4) +
2089 		/* IFLA_IPTUN_LOCAL */
2090 		nla_total_size(sizeof(struct in6_addr)) +
2091 		/* IFLA_IPTUN_REMOTE */
2092 		nla_total_size(sizeof(struct in6_addr)) +
2093 		/* IFLA_IPTUN_TTL */
2094 		nla_total_size(1) +
2095 		/* IFLA_IPTUN_ENCAP_LIMIT */
2096 		nla_total_size(1) +
2097 		/* IFLA_IPTUN_FLOWINFO */
2098 		nla_total_size(4) +
2099 		/* IFLA_IPTUN_FLAGS */
2100 		nla_total_size(4) +
2101 		/* IFLA_IPTUN_PROTO */
2102 		nla_total_size(1) +
2103 		/* IFLA_IPTUN_ENCAP_TYPE */
2104 		nla_total_size(2) +
2105 		/* IFLA_IPTUN_ENCAP_FLAGS */
2106 		nla_total_size(2) +
2107 		/* IFLA_IPTUN_ENCAP_SPORT */
2108 		nla_total_size(2) +
2109 		/* IFLA_IPTUN_ENCAP_DPORT */
2110 		nla_total_size(2) +
2111 		/* IFLA_IPTUN_COLLECT_METADATA */
2112 		nla_total_size(0) +
2113 		/* IFLA_IPTUN_FWMARK */
2114 		nla_total_size(4) +
2115 		0;
2116 }
2117 
ip6_tnl_fill_info(struct sk_buff * skb,const struct net_device * dev)2118 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2119 {
2120 	struct ip6_tnl *tunnel = netdev_priv(dev);
2121 	struct __ip6_tnl_parm *parm = &tunnel->parms;
2122 
2123 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2124 	    nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2125 	    nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2126 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2127 	    nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2128 	    nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2129 	    nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2130 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2131 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2132 		goto nla_put_failure;
2133 
2134 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2135 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2136 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2137 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2138 		goto nla_put_failure;
2139 
2140 	if (parm->collect_md)
2141 		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2142 			goto nla_put_failure;
2143 
2144 	return 0;
2145 
2146 nla_put_failure:
2147 	return -EMSGSIZE;
2148 }
2149 
ip6_tnl_get_link_net(const struct net_device * dev)2150 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2151 {
2152 	struct ip6_tnl *tunnel = netdev_priv(dev);
2153 
2154 	return tunnel->net;
2155 }
2156 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2157 
2158 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2159 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
2160 	[IFLA_IPTUN_LOCAL]		= { .len = sizeof(struct in6_addr) },
2161 	[IFLA_IPTUN_REMOTE]		= { .len = sizeof(struct in6_addr) },
2162 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
2163 	[IFLA_IPTUN_ENCAP_LIMIT]	= { .type = NLA_U8 },
2164 	[IFLA_IPTUN_FLOWINFO]		= { .type = NLA_U32 },
2165 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U32 },
2166 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
2167 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
2168 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
2169 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
2170 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
2171 	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
2172 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
2173 };
2174 
2175 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2176 	.kind		= "ip6tnl",
2177 	.maxtype	= IFLA_IPTUN_MAX,
2178 	.policy		= ip6_tnl_policy,
2179 	.priv_size	= sizeof(struct ip6_tnl),
2180 	.setup		= ip6_tnl_dev_setup,
2181 	.validate	= ip6_tnl_validate,
2182 	.newlink	= ip6_tnl_newlink,
2183 	.changelink	= ip6_tnl_changelink,
2184 	.dellink	= ip6_tnl_dellink,
2185 	.get_size	= ip6_tnl_get_size,
2186 	.fill_info	= ip6_tnl_fill_info,
2187 	.get_link_net	= ip6_tnl_get_link_net,
2188 };
2189 
2190 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2191 	.handler	= ip4ip6_rcv,
2192 	.err_handler	= ip4ip6_err,
2193 	.priority	=	1,
2194 };
2195 
2196 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2197 	.handler	= ip6ip6_rcv,
2198 	.err_handler	= ip6ip6_err,
2199 	.priority	=	1,
2200 };
2201 
2202 static struct xfrm6_tunnel mplsip6_handler __read_mostly = {
2203 	.handler	= mplsip6_rcv,
2204 	.err_handler	= mplsip6_err,
2205 	.priority	=	1,
2206 };
2207 
ip6_tnl_destroy_tunnels(struct net * net,struct list_head * list)2208 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2209 {
2210 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2211 	struct net_device *dev, *aux;
2212 	int h;
2213 	struct ip6_tnl *t;
2214 
2215 	for_each_netdev_safe(net, dev, aux)
2216 		if (dev->rtnl_link_ops == &ip6_link_ops)
2217 			unregister_netdevice_queue(dev, list);
2218 
2219 	for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2220 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
2221 		while (t) {
2222 			/* If dev is in the same netns, it has already
2223 			 * been added to the list by the previous loop.
2224 			 */
2225 			if (!net_eq(dev_net(t->dev), net))
2226 				unregister_netdevice_queue(t->dev, list);
2227 			t = rtnl_dereference(t->next);
2228 		}
2229 	}
2230 
2231 	t = rtnl_dereference(ip6n->tnls_wc[0]);
2232 	while (t) {
2233 		/* If dev is in the same netns, it has already
2234 		 * been added to the list by the previous loop.
2235 		 */
2236 		if (!net_eq(dev_net(t->dev), net))
2237 			unregister_netdevice_queue(t->dev, list);
2238 		t = rtnl_dereference(t->next);
2239 	}
2240 }
2241 
ip6_tnl_init_net(struct net * net)2242 static int __net_init ip6_tnl_init_net(struct net *net)
2243 {
2244 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2245 	struct ip6_tnl *t = NULL;
2246 	int err;
2247 
2248 	ip6n->tnls[0] = ip6n->tnls_wc;
2249 	ip6n->tnls[1] = ip6n->tnls_r_l;
2250 
2251 	if (!net_has_fallback_tunnels(net))
2252 		return 0;
2253 	err = -ENOMEM;
2254 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2255 					NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2256 
2257 	if (!ip6n->fb_tnl_dev)
2258 		goto err_alloc_dev;
2259 	dev_net_set(ip6n->fb_tnl_dev, net);
2260 	ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2261 	/* FB netdevice is special: we have one, and only one per netns.
2262 	 * Allowing to move it to another netns is clearly unsafe.
2263 	 */
2264 	ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2265 
2266 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2267 	if (err < 0)
2268 		goto err_register;
2269 
2270 	err = register_netdev(ip6n->fb_tnl_dev);
2271 	if (err < 0)
2272 		goto err_register;
2273 
2274 	t = netdev_priv(ip6n->fb_tnl_dev);
2275 
2276 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2277 	return 0;
2278 
2279 err_register:
2280 	free_netdev(ip6n->fb_tnl_dev);
2281 err_alloc_dev:
2282 	return err;
2283 }
2284 
ip6_tnl_exit_batch_net(struct list_head * net_list)2285 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2286 {
2287 	struct net *net;
2288 	LIST_HEAD(list);
2289 
2290 	rtnl_lock();
2291 	list_for_each_entry(net, net_list, exit_list)
2292 		ip6_tnl_destroy_tunnels(net, &list);
2293 	unregister_netdevice_many(&list);
2294 	rtnl_unlock();
2295 }
2296 
2297 static struct pernet_operations ip6_tnl_net_ops = {
2298 	.init = ip6_tnl_init_net,
2299 	.exit_batch = ip6_tnl_exit_batch_net,
2300 	.id   = &ip6_tnl_net_id,
2301 	.size = sizeof(struct ip6_tnl_net),
2302 };
2303 
2304 /**
2305  * ip6_tunnel_init - register protocol and reserve needed resources
2306  *
2307  * Return: 0 on success
2308  **/
2309 
ip6_tunnel_init(void)2310 static int __init ip6_tunnel_init(void)
2311 {
2312 	int  err;
2313 
2314 	if (!ipv6_mod_enabled())
2315 		return -EOPNOTSUPP;
2316 
2317 	err = register_pernet_device(&ip6_tnl_net_ops);
2318 	if (err < 0)
2319 		goto out_pernet;
2320 
2321 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2322 	if (err < 0) {
2323 		pr_err("%s: can't register ip4ip6\n", __func__);
2324 		goto out_ip4ip6;
2325 	}
2326 
2327 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2328 	if (err < 0) {
2329 		pr_err("%s: can't register ip6ip6\n", __func__);
2330 		goto out_ip6ip6;
2331 	}
2332 
2333 	if (ip6_tnl_mpls_supported()) {
2334 		err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS);
2335 		if (err < 0) {
2336 			pr_err("%s: can't register mplsip6\n", __func__);
2337 			goto out_mplsip6;
2338 		}
2339 	}
2340 
2341 	err = rtnl_link_register(&ip6_link_ops);
2342 	if (err < 0)
2343 		goto rtnl_link_failed;
2344 
2345 	return 0;
2346 
2347 rtnl_link_failed:
2348 	if (ip6_tnl_mpls_supported())
2349 		xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS);
2350 out_mplsip6:
2351 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2352 out_ip6ip6:
2353 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2354 out_ip4ip6:
2355 	unregister_pernet_device(&ip6_tnl_net_ops);
2356 out_pernet:
2357 	return err;
2358 }
2359 
2360 /**
2361  * ip6_tunnel_cleanup - free resources and unregister protocol
2362  **/
2363 
ip6_tunnel_cleanup(void)2364 static void __exit ip6_tunnel_cleanup(void)
2365 {
2366 	rtnl_link_unregister(&ip6_link_ops);
2367 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2368 		pr_info("%s: can't deregister ip4ip6\n", __func__);
2369 
2370 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2371 		pr_info("%s: can't deregister ip6ip6\n", __func__);
2372 
2373 	if (ip6_tnl_mpls_supported() &&
2374 	    xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS))
2375 		pr_info("%s: can't deregister mplsip6\n", __func__);
2376 	unregister_pernet_device(&ip6_tnl_net_ops);
2377 }
2378 
2379 module_init(ip6_tunnel_init);
2380 module_exit(ip6_tunnel_cleanup);
2381