• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 	t = netdev_priv(dev);
261 
262 	dev->rtnl_link_ops = &ip6_link_ops;
263 	err = register_netdevice(dev);
264 	if (err < 0)
265 		goto out;
266 
267 	strcpy(t->parms.name, dev->name);
268 
269 	ip6_tnl_link(ip6n, t);
270 	return 0;
271 
272 out:
273 	return err;
274 }
275 
276 /**
277  * ip6_tnl_create - create a new tunnel
278  *   @net: network namespace
279  *   @p: tunnel parameters
280  *
281  * Description:
282  *   Create tunnel matching given parameters.
283  *
284  * Return:
285  *   created tunnel or error pointer
286  **/
287 
ip6_tnl_create(struct net * net,struct __ip6_tnl_parm * p)288 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
289 {
290 	struct net_device *dev;
291 	struct ip6_tnl *t;
292 	char name[IFNAMSIZ];
293 	int err = -E2BIG;
294 
295 	if (p->name[0]) {
296 		if (!dev_valid_name(p->name))
297 			goto failed;
298 		strlcpy(name, p->name, IFNAMSIZ);
299 	} else {
300 		sprintf(name, "ip6tnl%%d");
301 	}
302 	err = -ENOMEM;
303 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
304 			   ip6_tnl_dev_setup);
305 	if (!dev)
306 		goto failed;
307 
308 	dev_net_set(dev, net);
309 
310 	t = netdev_priv(dev);
311 	t->parms = *p;
312 	t->net = dev_net(dev);
313 	err = ip6_tnl_create2(dev);
314 	if (err < 0)
315 		goto failed_free;
316 
317 	return t;
318 
319 failed_free:
320 	free_netdev(dev);
321 failed:
322 	return ERR_PTR(err);
323 }
324 
325 /**
326  * ip6_tnl_locate - find or create tunnel matching given parameters
327  *   @net: network namespace
328  *   @p: tunnel parameters
329  *   @create: != 0 if allowed to create new tunnel if no match found
330  *
331  * Description:
332  *   ip6_tnl_locate() first tries to locate an existing tunnel
333  *   based on @parms. If this is unsuccessful, but @create is set a new
334  *   tunnel device is created and registered for use.
335  *
336  * Return:
337  *   matching tunnel or error pointer
338  **/
339 
ip6_tnl_locate(struct net * net,struct __ip6_tnl_parm * p,int create)340 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
341 		struct __ip6_tnl_parm *p, int create)
342 {
343 	const struct in6_addr *remote = &p->raddr;
344 	const struct in6_addr *local = &p->laddr;
345 	struct ip6_tnl __rcu **tp;
346 	struct ip6_tnl *t;
347 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
348 
349 	for (tp = ip6_tnl_bucket(ip6n, p);
350 	     (t = rtnl_dereference(*tp)) != NULL;
351 	     tp = &t->next) {
352 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
353 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
354 		    p->link == t->parms.link) {
355 			if (create)
356 				return ERR_PTR(-EEXIST);
357 
358 			return t;
359 		}
360 	}
361 	if (!create)
362 		return ERR_PTR(-ENODEV);
363 	return ip6_tnl_create(net, p);
364 }
365 
366 /**
367  * ip6_tnl_dev_uninit - tunnel device uninitializer
368  *   @dev: the device to be destroyed
369  *
370  * Description:
371  *   ip6_tnl_dev_uninit() removes tunnel from its list
372  **/
373 
374 static void
ip6_tnl_dev_uninit(struct net_device * dev)375 ip6_tnl_dev_uninit(struct net_device *dev)
376 {
377 	struct ip6_tnl *t = netdev_priv(dev);
378 	struct net *net = t->net;
379 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
380 
381 	if (dev == ip6n->fb_tnl_dev)
382 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
383 	else
384 		ip6_tnl_unlink(ip6n, t);
385 	dst_cache_reset(&t->dst_cache);
386 	dev_put(dev);
387 }
388 
389 /**
390  * ip6_tnl_parse_tlv_enc_lim - handle encapsulation limit option
391  *   @skb: received socket buffer
392  *   @raw: the ICMPv6 error message data
393  *
394  * Return:
395  *   0 if none was found,
396  *   else index to encapsulation limit
397  **/
398 
ip6_tnl_parse_tlv_enc_lim(struct sk_buff * skb,__u8 * raw)399 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
400 {
401 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
402 	unsigned int nhoff = raw - skb->data;
403 	unsigned int off = nhoff + sizeof(*ipv6h);
404 	u8 nexthdr = ipv6h->nexthdr;
405 
406 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
407 		struct ipv6_opt_hdr *hdr;
408 		u16 optlen;
409 
410 		if (!pskb_may_pull(skb, off + sizeof(*hdr)))
411 			break;
412 
413 		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
414 		if (nexthdr == NEXTHDR_FRAGMENT) {
415 			optlen = 8;
416 		} else if (nexthdr == NEXTHDR_AUTH) {
417 			optlen = ipv6_authlen(hdr);
418 		} else {
419 			optlen = ipv6_optlen(hdr);
420 		}
421 
422 		if (!pskb_may_pull(skb, off + optlen))
423 			break;
424 
425 		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
426 		if (nexthdr == NEXTHDR_FRAGMENT) {
427 			struct frag_hdr *frag_hdr = (struct frag_hdr *)hdr;
428 
429 			if (frag_hdr->frag_off)
430 				break;
431 		}
432 		if (nexthdr == NEXTHDR_DEST) {
433 			u16 i = 2;
434 
435 			while (1) {
436 				struct ipv6_tlv_tnl_enc_lim *tel;
437 
438 				/* No more room for encapsulation limit */
439 				if (i + sizeof(*tel) > optlen)
440 					break;
441 
442 				tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
443 				/* return index of option if found and valid */
444 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
445 				    tel->length == 1)
446 					return i + off - nhoff;
447 				/* else jump to next option */
448 				if (tel->type)
449 					i += tel->length + 2;
450 				else
451 					i++;
452 			}
453 		}
454 		nexthdr = hdr->nexthdr;
455 		off += optlen;
456 	}
457 	return 0;
458 }
459 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
460 
461 /* ip6_tnl_err() should handle errors in the tunnel according to the
462  * specifications in RFC 2473.
463  */
464 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)465 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
466 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
467 {
468 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
469 	struct net *net = dev_net(skb->dev);
470 	u8 rel_type = ICMPV6_DEST_UNREACH;
471 	u8 rel_code = ICMPV6_ADDR_UNREACH;
472 	__u32 rel_info = 0;
473 	struct ip6_tnl *t;
474 	int err = -ENOENT;
475 	int rel_msg = 0;
476 	u8 tproto;
477 	__u16 len;
478 
479 	/* If the packet doesn't contain the original IPv6 header we are
480 	   in trouble since we might need the source address for further
481 	   processing of the error. */
482 
483 	rcu_read_lock();
484 	t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->daddr, &ipv6h->saddr);
485 	if (!t)
486 		goto out;
487 
488 	tproto = READ_ONCE(t->parms.proto);
489 	if (tproto != ipproto && tproto != 0)
490 		goto out;
491 
492 	err = 0;
493 
494 	switch (*type) {
495 	case ICMPV6_DEST_UNREACH:
496 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
497 				    t->parms.name);
498 		rel_msg = 1;
499 		break;
500 	case ICMPV6_TIME_EXCEED:
501 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
502 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
503 					    t->parms.name);
504 			rel_msg = 1;
505 		}
506 		break;
507 	case ICMPV6_PARAMPROB: {
508 		struct ipv6_tlv_tnl_enc_lim *tel;
509 		__u32 teli;
510 
511 		teli = 0;
512 		if ((*code) == ICMPV6_HDR_FIELD)
513 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
514 
515 		if (teli && teli == *info - 2) {
516 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
517 			if (tel->encap_limit == 0) {
518 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
519 						    t->parms.name);
520 				rel_msg = 1;
521 			}
522 		} else {
523 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
524 					    t->parms.name);
525 		}
526 		break;
527 	}
528 	case ICMPV6_PKT_TOOBIG: {
529 		__u32 mtu;
530 
531 		ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
532 				sock_net_uid(net, NULL));
533 		mtu = *info - offset;
534 		if (mtu < IPV6_MIN_MTU)
535 			mtu = IPV6_MIN_MTU;
536 		len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
537 		if (len > mtu) {
538 			rel_type = ICMPV6_PKT_TOOBIG;
539 			rel_code = 0;
540 			rel_info = mtu;
541 			rel_msg = 1;
542 		}
543 		break;
544 	}
545 	case NDISC_REDIRECT:
546 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
547 			     sock_net_uid(net, NULL));
548 		break;
549 	}
550 
551 	*type = rel_type;
552 	*code = rel_code;
553 	*info = rel_info;
554 	*msg = rel_msg;
555 
556 out:
557 	rcu_read_unlock();
558 	return err;
559 }
560 
561 static int
ip4ip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)562 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
563 	   u8 type, u8 code, int offset, __be32 info)
564 {
565 	__u32 rel_info = ntohl(info);
566 	const struct iphdr *eiph;
567 	struct sk_buff *skb2;
568 	int err, rel_msg = 0;
569 	u8 rel_type = type;
570 	u8 rel_code = code;
571 	struct rtable *rt;
572 	struct flowi4 fl4;
573 
574 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
575 			  &rel_msg, &rel_info, offset);
576 	if (err < 0)
577 		return err;
578 
579 	if (rel_msg == 0)
580 		return 0;
581 
582 	switch (rel_type) {
583 	case ICMPV6_DEST_UNREACH:
584 		if (rel_code != ICMPV6_ADDR_UNREACH)
585 			return 0;
586 		rel_type = ICMP_DEST_UNREACH;
587 		rel_code = ICMP_HOST_UNREACH;
588 		break;
589 	case ICMPV6_PKT_TOOBIG:
590 		if (rel_code != 0)
591 			return 0;
592 		rel_type = ICMP_DEST_UNREACH;
593 		rel_code = ICMP_FRAG_NEEDED;
594 		break;
595 	default:
596 		return 0;
597 	}
598 
599 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
600 		return 0;
601 
602 	skb2 = skb_clone(skb, GFP_ATOMIC);
603 	if (!skb2)
604 		return 0;
605 
606 	skb_dst_drop(skb2);
607 
608 	skb_pull(skb2, offset);
609 	skb_reset_network_header(skb2);
610 	eiph = ip_hdr(skb2);
611 
612 	/* Try to guess incoming interface */
613 	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
614 				   0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
615 	if (IS_ERR(rt))
616 		goto out;
617 
618 	skb2->dev = rt->dst.dev;
619 	ip_rt_put(rt);
620 
621 	/* route "incoming" packet */
622 	if (rt->rt_flags & RTCF_LOCAL) {
623 		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
624 					   eiph->daddr, eiph->saddr, 0, 0,
625 					   IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
626 		if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) {
627 			if (!IS_ERR(rt))
628 				ip_rt_put(rt);
629 			goto out;
630 		}
631 		skb_dst_set(skb2, &rt->dst);
632 	} else {
633 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
634 				   skb2->dev) ||
635 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
636 			goto out;
637 	}
638 
639 	/* change mtu on this route */
640 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
641 		if (rel_info > dst_mtu(skb_dst(skb2)))
642 			goto out;
643 
644 		skb_dst_update_pmtu_no_confirm(skb2, rel_info);
645 	}
646 
647 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
648 
649 out:
650 	kfree_skb(skb2);
651 	return 0;
652 }
653 
654 static int
ip6ip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)655 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
656 	   u8 type, u8 code, int offset, __be32 info)
657 {
658 	__u32 rel_info = ntohl(info);
659 	int err, rel_msg = 0;
660 	u8 rel_type = type;
661 	u8 rel_code = code;
662 
663 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
664 			  &rel_msg, &rel_info, offset);
665 	if (err < 0)
666 		return err;
667 
668 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
669 		struct rt6_info *rt;
670 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
671 
672 		if (!skb2)
673 			return 0;
674 
675 		skb_dst_drop(skb2);
676 		skb_pull(skb2, offset);
677 		skb_reset_network_header(skb2);
678 
679 		/* Try to guess incoming interface */
680 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
681 				NULL, 0, skb2, 0);
682 
683 		if (rt && rt->dst.dev)
684 			skb2->dev = rt->dst.dev;
685 
686 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
687 
688 		ip6_rt_put(rt);
689 
690 		kfree_skb(skb2);
691 	}
692 
693 	return 0;
694 }
695 
696 static int
mplsip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)697 mplsip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
698 	    u8 type, u8 code, int offset, __be32 info)
699 {
700 	__u32 rel_info = ntohl(info);
701 	int err, rel_msg = 0;
702 	u8 rel_type = type;
703 	u8 rel_code = code;
704 
705 	err = ip6_tnl_err(skb, IPPROTO_MPLS, opt, &rel_type, &rel_code,
706 			  &rel_msg, &rel_info, offset);
707 	return err;
708 }
709 
ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)710 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
711 				       const struct ipv6hdr *ipv6h,
712 				       struct sk_buff *skb)
713 {
714 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
715 
716 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
717 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
718 
719 	return IP6_ECN_decapsulate(ipv6h, skb);
720 }
721 
ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)722 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
723 				       const struct ipv6hdr *ipv6h,
724 				       struct sk_buff *skb)
725 {
726 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
727 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
728 
729 	return IP6_ECN_decapsulate(ipv6h, skb);
730 }
731 
mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)732 static inline int mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
733 					       const struct ipv6hdr *ipv6h,
734 					       struct sk_buff *skb)
735 {
736 	/* ECN is not supported in AF_MPLS */
737 	return 0;
738 }
739 
ip6_tnl_get_cap(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)740 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
741 			     const struct in6_addr *laddr,
742 			     const struct in6_addr *raddr)
743 {
744 	struct __ip6_tnl_parm *p = &t->parms;
745 	int ltype = ipv6_addr_type(laddr);
746 	int rtype = ipv6_addr_type(raddr);
747 	__u32 flags = 0;
748 
749 	if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
750 		flags = IP6_TNL_F_CAP_PER_PACKET;
751 	} else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
752 		   rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
753 		   !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
754 		   (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
755 		if (ltype&IPV6_ADDR_UNICAST)
756 			flags |= IP6_TNL_F_CAP_XMIT;
757 		if (rtype&IPV6_ADDR_UNICAST)
758 			flags |= IP6_TNL_F_CAP_RCV;
759 	}
760 	return flags;
761 }
762 EXPORT_SYMBOL(ip6_tnl_get_cap);
763 
764 /* called with rcu_read_lock() */
ip6_tnl_rcv_ctl(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)765 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
766 				  const struct in6_addr *laddr,
767 				  const struct in6_addr *raddr)
768 {
769 	struct __ip6_tnl_parm *p = &t->parms;
770 	int ret = 0;
771 	struct net *net = t->net;
772 
773 	if ((p->flags & IP6_TNL_F_CAP_RCV) ||
774 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
775 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
776 		struct net_device *ldev = NULL;
777 
778 		if (p->link)
779 			ldev = dev_get_by_index_rcu(net, p->link);
780 
781 		if ((ipv6_addr_is_multicast(laddr) ||
782 		     likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
783 						    0, IFA_F_TENTATIVE))) &&
784 		    ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
785 		     likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
786 						     0, IFA_F_TENTATIVE))))
787 			ret = 1;
788 	}
789 	return ret;
790 }
791 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
792 
__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)793 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
794 			 const struct tnl_ptk_info *tpi,
795 			 struct metadata_dst *tun_dst,
796 			 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
797 						const struct ipv6hdr *ipv6h,
798 						struct sk_buff *skb),
799 			 bool log_ecn_err)
800 {
801 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
802 	int err;
803 
804 	if ((!(tpi->flags & TUNNEL_CSUM) &&
805 	     (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
806 	    ((tpi->flags & TUNNEL_CSUM) &&
807 	     !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
808 		tunnel->dev->stats.rx_crc_errors++;
809 		tunnel->dev->stats.rx_errors++;
810 		goto drop;
811 	}
812 
813 	if (tunnel->parms.i_flags & TUNNEL_SEQ) {
814 		if (!(tpi->flags & TUNNEL_SEQ) ||
815 		    (tunnel->i_seqno &&
816 		     (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
817 			tunnel->dev->stats.rx_fifo_errors++;
818 			tunnel->dev->stats.rx_errors++;
819 			goto drop;
820 		}
821 		tunnel->i_seqno = ntohl(tpi->seq) + 1;
822 	}
823 
824 	skb->protocol = tpi->proto;
825 
826 	/* Warning: All skb pointers will be invalidated! */
827 	if (tunnel->dev->type == ARPHRD_ETHER) {
828 		if (!pskb_may_pull(skb, ETH_HLEN)) {
829 			tunnel->dev->stats.rx_length_errors++;
830 			tunnel->dev->stats.rx_errors++;
831 			goto drop;
832 		}
833 
834 		ipv6h = ipv6_hdr(skb);
835 		skb->protocol = eth_type_trans(skb, tunnel->dev);
836 		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
837 	} else {
838 		skb->dev = tunnel->dev;
839 		skb_reset_mac_header(skb);
840 	}
841 
842 	skb_reset_network_header(skb);
843 	memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
844 
845 	__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
846 
847 	err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
848 	if (unlikely(err)) {
849 		if (log_ecn_err)
850 			net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
851 					     &ipv6h->saddr,
852 					     ipv6_get_dsfield(ipv6h));
853 		if (err > 1) {
854 			++tunnel->dev->stats.rx_frame_errors;
855 			++tunnel->dev->stats.rx_errors;
856 			goto drop;
857 		}
858 	}
859 
860 	dev_sw_netstats_rx_add(tunnel->dev, skb->len);
861 
862 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
863 
864 	if (tun_dst)
865 		skb_dst_set(skb, (struct dst_entry *)tun_dst);
866 
867 	gro_cells_receive(&tunnel->gro_cells, skb);
868 	return 0;
869 
870 drop:
871 	if (tun_dst)
872 		dst_release((struct dst_entry *)tun_dst);
873 	kfree_skb(skb);
874 	return 0;
875 }
876 
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)877 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
878 		const struct tnl_ptk_info *tpi,
879 		struct metadata_dst *tun_dst,
880 		bool log_ecn_err)
881 {
882 	int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
883 				    const struct ipv6hdr *ipv6h,
884 				    struct sk_buff *skb);
885 
886 	dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
887 	if (tpi->proto == htons(ETH_P_IP))
888 		dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
889 
890 	return __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
891 			     log_ecn_err);
892 }
893 EXPORT_SYMBOL(ip6_tnl_rcv);
894 
895 static const struct tnl_ptk_info tpi_v6 = {
896 	/* no tunnel info required for ipxip6. */
897 	.proto = htons(ETH_P_IPV6),
898 };
899 
900 static const struct tnl_ptk_info tpi_v4 = {
901 	/* no tunnel info required for ipxip6. */
902 	.proto = htons(ETH_P_IP),
903 };
904 
905 static const struct tnl_ptk_info tpi_mpls = {
906 	/* no tunnel info required for mplsip6. */
907 	.proto = htons(ETH_P_MPLS_UC),
908 };
909 
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))910 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
911 		      const struct tnl_ptk_info *tpi,
912 		      int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
913 						  const struct ipv6hdr *ipv6h,
914 						  struct sk_buff *skb))
915 {
916 	struct ip6_tnl *t;
917 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
918 	struct metadata_dst *tun_dst = NULL;
919 	int ret = -1;
920 
921 	rcu_read_lock();
922 	t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr);
923 
924 	if (t) {
925 		u8 tproto = READ_ONCE(t->parms.proto);
926 
927 		if (tproto != ipproto && tproto != 0)
928 			goto drop;
929 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
930 			goto drop;
931 		ipv6h = ipv6_hdr(skb);
932 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
933 			goto drop;
934 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
935 			goto drop;
936 		if (t->parms.collect_md) {
937 			tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
938 			if (!tun_dst)
939 				goto drop;
940 		}
941 		ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
942 				    log_ecn_error);
943 	}
944 
945 	rcu_read_unlock();
946 
947 	return ret;
948 
949 drop:
950 	rcu_read_unlock();
951 	kfree_skb(skb);
952 	return 0;
953 }
954 
ip4ip6_rcv(struct sk_buff * skb)955 static int ip4ip6_rcv(struct sk_buff *skb)
956 {
957 	return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
958 			  ip4ip6_dscp_ecn_decapsulate);
959 }
960 
ip6ip6_rcv(struct sk_buff * skb)961 static int ip6ip6_rcv(struct sk_buff *skb)
962 {
963 	return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
964 			  ip6ip6_dscp_ecn_decapsulate);
965 }
966 
mplsip6_rcv(struct sk_buff * skb)967 static int mplsip6_rcv(struct sk_buff *skb)
968 {
969 	return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls,
970 			  mplsip6_dscp_ecn_decapsulate);
971 }
972 
973 struct ipv6_tel_txoption {
974 	struct ipv6_txoptions ops;
975 	__u8 dst_opt[8];
976 };
977 
init_tel_txopt(struct ipv6_tel_txoption * opt,__u8 encap_limit)978 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
979 {
980 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
981 
982 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
983 	opt->dst_opt[3] = 1;
984 	opt->dst_opt[4] = encap_limit;
985 	opt->dst_opt[5] = IPV6_TLV_PADN;
986 	opt->dst_opt[6] = 1;
987 
988 	opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
989 	opt->ops.opt_nflen = 8;
990 }
991 
992 /**
993  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
994  *   @t: the outgoing tunnel device
995  *   @hdr: IPv6 header from the incoming packet
996  *
997  * Description:
998  *   Avoid trivial tunneling loop by checking that tunnel exit-point
999  *   doesn't match source of incoming packet.
1000  *
1001  * Return:
1002  *   1 if conflict,
1003  *   0 else
1004  **/
1005 
1006 static inline bool
ip6_tnl_addr_conflict(const struct ip6_tnl * t,const struct ipv6hdr * hdr)1007 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
1008 {
1009 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
1010 }
1011 
ip6_tnl_xmit_ctl(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)1012 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
1013 		     const struct in6_addr *laddr,
1014 		     const struct in6_addr *raddr)
1015 {
1016 	struct __ip6_tnl_parm *p = &t->parms;
1017 	int ret = 0;
1018 	struct net *net = t->net;
1019 
1020 	if (t->parms.collect_md)
1021 		return 1;
1022 
1023 	if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1024 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1025 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1026 		struct net_device *ldev = NULL;
1027 
1028 		rcu_read_lock();
1029 		if (p->link)
1030 			ldev = dev_get_by_index_rcu(net, p->link);
1031 
1032 		if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
1033 						      0, IFA_F_TENTATIVE)))
1034 			pr_warn_ratelimited("%s xmit: Local address not yet configured!\n",
1035 					    p->name);
1036 		else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1037 			 !ipv6_addr_is_multicast(raddr) &&
1038 			 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1039 							  true, 0, IFA_F_TENTATIVE)))
1040 			pr_warn_ratelimited("%s xmit: Routing loop! Remote address found on this node!\n",
1041 					    p->name);
1042 		else
1043 			ret = 1;
1044 		rcu_read_unlock();
1045 	}
1046 	return ret;
1047 }
1048 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1049 
1050 /**
1051  * ip6_tnl_xmit - encapsulate packet and send
1052  *   @skb: the outgoing socket buffer
1053  *   @dev: the outgoing tunnel device
1054  *   @dsfield: dscp code for outer header
1055  *   @fl6: flow of tunneled packet
1056  *   @encap_limit: encapsulation limit
1057  *   @pmtu: Path MTU is stored if packet is too big
1058  *   @proto: next header value
1059  *
1060  * Description:
1061  *   Build new header and do some sanity checks on the packet before sending
1062  *   it.
1063  *
1064  * Return:
1065  *   0 on success
1066  *   -1 fail
1067  *   %-EMSGSIZE message too big. return mtu in this case.
1068  **/
1069 
ip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev,__u8 dsfield,struct flowi6 * fl6,int encap_limit,__u32 * pmtu,__u8 proto)1070 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1071 		 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1072 		 __u8 proto)
1073 {
1074 	struct ip6_tnl *t = netdev_priv(dev);
1075 	struct net *net = t->net;
1076 	struct net_device_stats *stats = &t->dev->stats;
1077 	struct ipv6hdr *ipv6h;
1078 	struct ipv6_tel_txoption opt;
1079 	struct dst_entry *dst = NULL, *ndst = NULL;
1080 	struct net_device *tdev;
1081 	int mtu;
1082 	unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1083 	unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1084 	unsigned int max_headroom = psh_hlen;
1085 	bool use_cache = false;
1086 	u8 hop_limit;
1087 	int err = -1;
1088 
1089 	if (t->parms.collect_md) {
1090 		hop_limit = skb_tunnel_info(skb)->key.ttl;
1091 		goto route_lookup;
1092 	} else {
1093 		hop_limit = t->parms.hop_limit;
1094 	}
1095 
1096 	/* NBMA tunnel */
1097 	if (ipv6_addr_any(&t->parms.raddr)) {
1098 		if (skb->protocol == htons(ETH_P_IPV6)) {
1099 			struct in6_addr *addr6;
1100 			struct neighbour *neigh;
1101 			int addr_type;
1102 
1103 			if (!skb_dst(skb))
1104 				goto tx_err_link_failure;
1105 
1106 			neigh = dst_neigh_lookup(skb_dst(skb),
1107 						 &ipv6_hdr(skb)->daddr);
1108 			if (!neigh)
1109 				goto tx_err_link_failure;
1110 
1111 			addr6 = (struct in6_addr *)&neigh->primary_key;
1112 			addr_type = ipv6_addr_type(addr6);
1113 
1114 			if (addr_type == IPV6_ADDR_ANY)
1115 				addr6 = &ipv6_hdr(skb)->daddr;
1116 
1117 			memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1118 			neigh_release(neigh);
1119 		}
1120 	} else if (t->parms.proto != 0 && !(t->parms.flags &
1121 					    (IP6_TNL_F_USE_ORIG_TCLASS |
1122 					     IP6_TNL_F_USE_ORIG_FWMARK))) {
1123 		/* enable the cache only if neither the outer protocol nor the
1124 		 * routing decision depends on the current inner header value
1125 		 */
1126 		use_cache = true;
1127 	}
1128 
1129 	if (use_cache)
1130 		dst = dst_cache_get(&t->dst_cache);
1131 
1132 	if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1133 		goto tx_err_link_failure;
1134 
1135 	if (!dst) {
1136 route_lookup:
1137 		/* add dsfield to flowlabel for route lookup */
1138 		fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1139 
1140 		dst = ip6_route_output(net, NULL, fl6);
1141 
1142 		if (dst->error)
1143 			goto tx_err_link_failure;
1144 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1145 		if (IS_ERR(dst)) {
1146 			err = PTR_ERR(dst);
1147 			dst = NULL;
1148 			goto tx_err_link_failure;
1149 		}
1150 		if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1151 		    ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1152 				       &fl6->daddr, 0, &fl6->saddr))
1153 			goto tx_err_link_failure;
1154 		ndst = dst;
1155 	}
1156 
1157 	tdev = dst->dev;
1158 
1159 	if (tdev == dev) {
1160 		stats->collisions++;
1161 		net_warn_ratelimited("%s: Local routing loop detected!\n",
1162 				     t->parms.name);
1163 		goto tx_err_dst_release;
1164 	}
1165 	mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1166 	if (encap_limit >= 0) {
1167 		max_headroom += 8;
1168 		mtu -= 8;
1169 	}
1170 	mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1171 		       IPV6_MIN_MTU : IPV4_MIN_MTU);
1172 
1173 	skb_dst_update_pmtu_no_confirm(skb, mtu);
1174 	if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1175 		*pmtu = mtu;
1176 		err = -EMSGSIZE;
1177 		goto tx_err_dst_release;
1178 	}
1179 
1180 	if (t->err_count > 0) {
1181 		if (time_before(jiffies,
1182 				t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1183 			t->err_count--;
1184 
1185 			dst_link_failure(skb);
1186 		} else {
1187 			t->err_count = 0;
1188 		}
1189 	}
1190 
1191 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1192 
1193 	/*
1194 	 * Okay, now see if we can stuff it in the buffer as-is.
1195 	 */
1196 	max_headroom += LL_RESERVED_SPACE(tdev);
1197 
1198 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1199 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1200 		struct sk_buff *new_skb;
1201 
1202 		new_skb = skb_realloc_headroom(skb, max_headroom);
1203 		if (!new_skb)
1204 			goto tx_err_dst_release;
1205 
1206 		if (skb->sk)
1207 			skb_set_owner_w(new_skb, skb->sk);
1208 		consume_skb(skb);
1209 		skb = new_skb;
1210 	}
1211 
1212 	if (t->parms.collect_md) {
1213 		if (t->encap.type != TUNNEL_ENCAP_NONE)
1214 			goto tx_err_dst_release;
1215 	} else {
1216 		if (use_cache && ndst)
1217 			dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1218 	}
1219 	skb_dst_set(skb, dst);
1220 
1221 	if (hop_limit == 0) {
1222 		if (skb->protocol == htons(ETH_P_IP))
1223 			hop_limit = ip_hdr(skb)->ttl;
1224 		else if (skb->protocol == htons(ETH_P_IPV6))
1225 			hop_limit = ipv6_hdr(skb)->hop_limit;
1226 		else
1227 			hop_limit = ip6_dst_hoplimit(dst);
1228 	}
1229 
1230 	/* Calculate max headroom for all the headers and adjust
1231 	 * needed_headroom if necessary.
1232 	 */
1233 	max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1234 			+ dst->header_len + t->hlen;
1235 	if (max_headroom > READ_ONCE(dev->needed_headroom))
1236 		WRITE_ONCE(dev->needed_headroom, max_headroom);
1237 
1238 	err = ip6_tnl_encap(skb, t, &proto, fl6);
1239 	if (err)
1240 		return err;
1241 
1242 	if (encap_limit >= 0) {
1243 		init_tel_txopt(&opt, encap_limit);
1244 		ipv6_push_frag_opts(skb, &opt.ops, &proto);
1245 	}
1246 
1247 	skb_push(skb, sizeof(struct ipv6hdr));
1248 	skb_reset_network_header(skb);
1249 	ipv6h = ipv6_hdr(skb);
1250 	ip6_flow_hdr(ipv6h, dsfield,
1251 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1252 	ipv6h->hop_limit = hop_limit;
1253 	ipv6h->nexthdr = proto;
1254 	ipv6h->saddr = fl6->saddr;
1255 	ipv6h->daddr = fl6->daddr;
1256 	ip6tunnel_xmit(NULL, skb, dev);
1257 	return 0;
1258 tx_err_link_failure:
1259 	stats->tx_carrier_errors++;
1260 	dst_link_failure(skb);
1261 tx_err_dst_release:
1262 	dst_release(dst);
1263 	return err;
1264 }
1265 EXPORT_SYMBOL(ip6_tnl_xmit);
1266 
1267 static inline int
ipxip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev,u8 protocol)1268 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev,
1269 		u8 protocol)
1270 {
1271 	struct ip6_tnl *t = netdev_priv(dev);
1272 	struct ipv6hdr *ipv6h;
1273 	const struct iphdr  *iph;
1274 	int encap_limit = -1;
1275 	__u16 offset;
1276 	struct flowi6 fl6;
1277 	__u8 dsfield, orig_dsfield;
1278 	__u32 mtu;
1279 	u8 tproto;
1280 	int err;
1281 
1282 	tproto = READ_ONCE(t->parms.proto);
1283 	if (tproto != protocol && tproto != 0)
1284 		return -1;
1285 
1286 	if (t->parms.collect_md) {
1287 		struct ip_tunnel_info *tun_info;
1288 		const struct ip_tunnel_key *key;
1289 
1290 		tun_info = skb_tunnel_info(skb);
1291 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1292 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1293 			return -1;
1294 		key = &tun_info->key;
1295 		memset(&fl6, 0, sizeof(fl6));
1296 		fl6.flowi6_proto = protocol;
1297 		fl6.saddr = key->u.ipv6.src;
1298 		fl6.daddr = key->u.ipv6.dst;
1299 		fl6.flowlabel = key->label;
1300 		dsfield =  key->tos;
1301 		switch (protocol) {
1302 		case IPPROTO_IPIP:
1303 			iph = ip_hdr(skb);
1304 			orig_dsfield = ipv4_get_dsfield(iph);
1305 			break;
1306 		case IPPROTO_IPV6:
1307 			ipv6h = ipv6_hdr(skb);
1308 			orig_dsfield = ipv6_get_dsfield(ipv6h);
1309 			break;
1310 		default:
1311 			orig_dsfield = dsfield;
1312 			break;
1313 		}
1314 	} else {
1315 		if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1316 			encap_limit = t->parms.encap_limit;
1317 		if (protocol == IPPROTO_IPV6) {
1318 			offset = ip6_tnl_parse_tlv_enc_lim(skb,
1319 						skb_network_header(skb));
1320 			/* ip6_tnl_parse_tlv_enc_lim() might have
1321 			 * reallocated skb->head
1322 			 */
1323 			if (offset > 0) {
1324 				struct ipv6_tlv_tnl_enc_lim *tel;
1325 
1326 				tel = (void *)&skb_network_header(skb)[offset];
1327 				if (tel->encap_limit == 0) {
1328 					icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
1329 							ICMPV6_HDR_FIELD, offset + 2);
1330 					return -1;
1331 				}
1332 				encap_limit = tel->encap_limit - 1;
1333 			}
1334 		}
1335 
1336 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1337 		fl6.flowi6_proto = protocol;
1338 
1339 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1340 			fl6.flowi6_mark = skb->mark;
1341 		else
1342 			fl6.flowi6_mark = t->parms.fwmark;
1343 		switch (protocol) {
1344 		case IPPROTO_IPIP:
1345 			iph = ip_hdr(skb);
1346 			orig_dsfield = ipv4_get_dsfield(iph);
1347 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1348 				dsfield = orig_dsfield;
1349 			else
1350 				dsfield = ip6_tclass(t->parms.flowinfo);
1351 			break;
1352 		case IPPROTO_IPV6:
1353 			ipv6h = ipv6_hdr(skb);
1354 			orig_dsfield = ipv6_get_dsfield(ipv6h);
1355 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1356 				dsfield = orig_dsfield;
1357 			else
1358 				dsfield = ip6_tclass(t->parms.flowinfo);
1359 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1360 				fl6.flowlabel |= ip6_flowlabel(ipv6h);
1361 			break;
1362 		default:
1363 			orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo);
1364 			break;
1365 		}
1366 	}
1367 
1368 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1369 	dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield);
1370 
1371 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1372 		return -1;
1373 
1374 	skb_set_inner_ipproto(skb, protocol);
1375 
1376 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1377 			   protocol);
1378 	if (err != 0) {
1379 		/* XXX: send ICMP error even if DF is not set. */
1380 		if (err == -EMSGSIZE)
1381 			switch (protocol) {
1382 			case IPPROTO_IPIP:
1383 				icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1384 					      ICMP_FRAG_NEEDED, htonl(mtu));
1385 				break;
1386 			case IPPROTO_IPV6:
1387 				icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1388 				break;
1389 			default:
1390 				break;
1391 			}
1392 		return -1;
1393 	}
1394 
1395 	return 0;
1396 }
1397 
1398 static netdev_tx_t
ip6_tnl_start_xmit(struct sk_buff * skb,struct net_device * dev)1399 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1400 {
1401 	struct ip6_tnl *t = netdev_priv(dev);
1402 	struct net_device_stats *stats = &t->dev->stats;
1403 	u8 ipproto;
1404 	int ret;
1405 
1406 	if (!pskb_inet_may_pull(skb))
1407 		goto tx_err;
1408 
1409 	switch (skb->protocol) {
1410 	case htons(ETH_P_IP):
1411 		ipproto = IPPROTO_IPIP;
1412 		break;
1413 	case htons(ETH_P_IPV6):
1414 		if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb)))
1415 			goto tx_err;
1416 		ipproto = IPPROTO_IPV6;
1417 		break;
1418 	case htons(ETH_P_MPLS_UC):
1419 		ipproto = IPPROTO_MPLS;
1420 		break;
1421 	default:
1422 		goto tx_err;
1423 	}
1424 
1425 	ret = ipxip6_tnl_xmit(skb, dev, ipproto);
1426 	if (ret < 0)
1427 		goto tx_err;
1428 
1429 	return NETDEV_TX_OK;
1430 
1431 tx_err:
1432 	stats->tx_errors++;
1433 	stats->tx_dropped++;
1434 	kfree_skb(skb);
1435 	return NETDEV_TX_OK;
1436 }
1437 
ip6_tnl_link_config(struct ip6_tnl * t)1438 static void ip6_tnl_link_config(struct ip6_tnl *t)
1439 {
1440 	struct net_device *dev = t->dev;
1441 	struct net_device *tdev = NULL;
1442 	struct __ip6_tnl_parm *p = &t->parms;
1443 	struct flowi6 *fl6 = &t->fl.u.ip6;
1444 	int t_hlen;
1445 	int mtu;
1446 
1447 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1448 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1449 
1450 	/* Set up flowi template */
1451 	fl6->saddr = p->laddr;
1452 	fl6->daddr = p->raddr;
1453 	fl6->flowi6_oif = p->link;
1454 	fl6->flowlabel = 0;
1455 
1456 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1457 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1458 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1459 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1460 
1461 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1462 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1463 
1464 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1465 		dev->flags |= IFF_POINTOPOINT;
1466 	else
1467 		dev->flags &= ~IFF_POINTOPOINT;
1468 
1469 	t->tun_hlen = 0;
1470 	t->hlen = t->encap_hlen + t->tun_hlen;
1471 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1472 
1473 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1474 		int strict = (ipv6_addr_type(&p->raddr) &
1475 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1476 
1477 		struct rt6_info *rt = rt6_lookup(t->net,
1478 						 &p->raddr, &p->laddr,
1479 						 p->link, NULL, strict);
1480 		if (rt) {
1481 			tdev = rt->dst.dev;
1482 			ip6_rt_put(rt);
1483 		}
1484 
1485 		if (!tdev && p->link)
1486 			tdev = __dev_get_by_index(t->net, p->link);
1487 
1488 		if (tdev) {
1489 			dev->hard_header_len = tdev->hard_header_len + t_hlen;
1490 			mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU);
1491 
1492 			mtu = mtu - t_hlen;
1493 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1494 				mtu -= 8;
1495 
1496 			if (mtu < IPV6_MIN_MTU)
1497 				mtu = IPV6_MIN_MTU;
1498 			WRITE_ONCE(dev->mtu, mtu);
1499 		}
1500 	}
1501 }
1502 
1503 /**
1504  * ip6_tnl_change - update the tunnel parameters
1505  *   @t: tunnel to be changed
1506  *   @p: tunnel configuration parameters
1507  *
1508  * Description:
1509  *   ip6_tnl_change() updates the tunnel parameters
1510  **/
1511 
1512 static int
ip6_tnl_change(struct ip6_tnl * t,const struct __ip6_tnl_parm * p)1513 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1514 {
1515 	t->parms.laddr = p->laddr;
1516 	t->parms.raddr = p->raddr;
1517 	t->parms.flags = p->flags;
1518 	t->parms.hop_limit = p->hop_limit;
1519 	t->parms.encap_limit = p->encap_limit;
1520 	t->parms.flowinfo = p->flowinfo;
1521 	t->parms.link = p->link;
1522 	t->parms.proto = p->proto;
1523 	t->parms.fwmark = p->fwmark;
1524 	dst_cache_reset(&t->dst_cache);
1525 	ip6_tnl_link_config(t);
1526 	return 0;
1527 }
1528 
ip6_tnl_update(struct ip6_tnl * t,struct __ip6_tnl_parm * p)1529 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1530 {
1531 	struct net *net = t->net;
1532 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1533 	int err;
1534 
1535 	ip6_tnl_unlink(ip6n, t);
1536 	synchronize_net();
1537 	err = ip6_tnl_change(t, p);
1538 	ip6_tnl_link(ip6n, t);
1539 	netdev_state_change(t->dev);
1540 	return err;
1541 }
1542 
ip6_tnl0_update(struct ip6_tnl * t,struct __ip6_tnl_parm * p)1543 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1544 {
1545 	/* for default tnl0 device allow to change only the proto */
1546 	t->parms.proto = p->proto;
1547 	netdev_state_change(t->dev);
1548 	return 0;
1549 }
1550 
1551 static void
ip6_tnl_parm_from_user(struct __ip6_tnl_parm * p,const struct ip6_tnl_parm * u)1552 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1553 {
1554 	p->laddr = u->laddr;
1555 	p->raddr = u->raddr;
1556 	p->flags = u->flags;
1557 	p->hop_limit = u->hop_limit;
1558 	p->encap_limit = u->encap_limit;
1559 	p->flowinfo = u->flowinfo;
1560 	p->link = u->link;
1561 	p->proto = u->proto;
1562 	memcpy(p->name, u->name, sizeof(u->name));
1563 }
1564 
1565 static void
ip6_tnl_parm_to_user(struct ip6_tnl_parm * u,const struct __ip6_tnl_parm * p)1566 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1567 {
1568 	u->laddr = p->laddr;
1569 	u->raddr = p->raddr;
1570 	u->flags = p->flags;
1571 	u->hop_limit = p->hop_limit;
1572 	u->encap_limit = p->encap_limit;
1573 	u->flowinfo = p->flowinfo;
1574 	u->link = p->link;
1575 	u->proto = p->proto;
1576 	memcpy(u->name, p->name, sizeof(u->name));
1577 }
1578 
1579 /**
1580  * ip6_tnl_siocdevprivate - configure ipv6 tunnels from userspace
1581  *   @dev: virtual device associated with tunnel
1582  *   @ifr: unused
1583  *   @data: parameters passed from userspace
1584  *   @cmd: command to be performed
1585  *
1586  * Description:
1587  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1588  *   from userspace.
1589  *
1590  *   The possible commands are the following:
1591  *     %SIOCGETTUNNEL: get tunnel parameters for device
1592  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1593  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1594  *     %SIOCDELTUNNEL: delete tunnel
1595  *
1596  *   The fallback device "ip6tnl0", created during module
1597  *   initialization, can be used for creating other tunnel devices.
1598  *
1599  * Return:
1600  *   0 on success,
1601  *   %-EFAULT if unable to copy data to or from userspace,
1602  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1603  *   %-EINVAL if passed tunnel parameters are invalid,
1604  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1605  *   %-ENODEV if attempting to change or delete a nonexisting device
1606  **/
1607 
1608 static int
ip6_tnl_siocdevprivate(struct net_device * dev,struct ifreq * ifr,void __user * data,int cmd)1609 ip6_tnl_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1610 		       void __user *data, int cmd)
1611 {
1612 	int err = 0;
1613 	struct ip6_tnl_parm p;
1614 	struct __ip6_tnl_parm p1;
1615 	struct ip6_tnl *t = netdev_priv(dev);
1616 	struct net *net = t->net;
1617 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1618 
1619 	memset(&p1, 0, sizeof(p1));
1620 
1621 	switch (cmd) {
1622 	case SIOCGETTUNNEL:
1623 		if (dev == ip6n->fb_tnl_dev) {
1624 			if (copy_from_user(&p, data, sizeof(p))) {
1625 				err = -EFAULT;
1626 				break;
1627 			}
1628 			ip6_tnl_parm_from_user(&p1, &p);
1629 			t = ip6_tnl_locate(net, &p1, 0);
1630 			if (IS_ERR(t))
1631 				t = netdev_priv(dev);
1632 		} else {
1633 			memset(&p, 0, sizeof(p));
1634 		}
1635 		ip6_tnl_parm_to_user(&p, &t->parms);
1636 		if (copy_to_user(data, &p, sizeof(p)))
1637 			err = -EFAULT;
1638 		break;
1639 	case SIOCADDTUNNEL:
1640 	case SIOCCHGTUNNEL:
1641 		err = -EPERM;
1642 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1643 			break;
1644 		err = -EFAULT;
1645 		if (copy_from_user(&p, data, sizeof(p)))
1646 			break;
1647 		err = -EINVAL;
1648 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1649 		    p.proto != 0)
1650 			break;
1651 		ip6_tnl_parm_from_user(&p1, &p);
1652 		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1653 		if (cmd == SIOCCHGTUNNEL) {
1654 			if (!IS_ERR(t)) {
1655 				if (t->dev != dev) {
1656 					err = -EEXIST;
1657 					break;
1658 				}
1659 			} else
1660 				t = netdev_priv(dev);
1661 			if (dev == ip6n->fb_tnl_dev)
1662 				err = ip6_tnl0_update(t, &p1);
1663 			else
1664 				err = ip6_tnl_update(t, &p1);
1665 		}
1666 		if (!IS_ERR(t)) {
1667 			err = 0;
1668 			ip6_tnl_parm_to_user(&p, &t->parms);
1669 			if (copy_to_user(data, &p, sizeof(p)))
1670 				err = -EFAULT;
1671 
1672 		} else {
1673 			err = PTR_ERR(t);
1674 		}
1675 		break;
1676 	case SIOCDELTUNNEL:
1677 		err = -EPERM;
1678 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1679 			break;
1680 
1681 		if (dev == ip6n->fb_tnl_dev) {
1682 			err = -EFAULT;
1683 			if (copy_from_user(&p, data, sizeof(p)))
1684 				break;
1685 			err = -ENOENT;
1686 			ip6_tnl_parm_from_user(&p1, &p);
1687 			t = ip6_tnl_locate(net, &p1, 0);
1688 			if (IS_ERR(t))
1689 				break;
1690 			err = -EPERM;
1691 			if (t->dev == ip6n->fb_tnl_dev)
1692 				break;
1693 			dev = t->dev;
1694 		}
1695 		err = 0;
1696 		unregister_netdevice(dev);
1697 		break;
1698 	default:
1699 		err = -EINVAL;
1700 	}
1701 	return err;
1702 }
1703 
1704 /**
1705  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1706  *   @dev: virtual device associated with tunnel
1707  *   @new_mtu: the new mtu
1708  *
1709  * Return:
1710  *   0 on success,
1711  *   %-EINVAL if mtu too small
1712  **/
1713 
ip6_tnl_change_mtu(struct net_device * dev,int new_mtu)1714 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1715 {
1716 	struct ip6_tnl *tnl = netdev_priv(dev);
1717 
1718 	if (tnl->parms.proto == IPPROTO_IPV6) {
1719 		if (new_mtu < IPV6_MIN_MTU)
1720 			return -EINVAL;
1721 	} else {
1722 		if (new_mtu < ETH_MIN_MTU)
1723 			return -EINVAL;
1724 	}
1725 	if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1726 		if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1727 			return -EINVAL;
1728 	} else {
1729 		if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1730 			return -EINVAL;
1731 	}
1732 	dev->mtu = new_mtu;
1733 	return 0;
1734 }
1735 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1736 
ip6_tnl_get_iflink(const struct net_device * dev)1737 int ip6_tnl_get_iflink(const struct net_device *dev)
1738 {
1739 	struct ip6_tnl *t = netdev_priv(dev);
1740 
1741 	return t->parms.link;
1742 }
1743 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1744 
ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops * ops,unsigned int num)1745 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1746 			  unsigned int num)
1747 {
1748 	if (num >= MAX_IPTUN_ENCAP_OPS)
1749 		return -ERANGE;
1750 
1751 	return !cmpxchg((const struct ip6_tnl_encap_ops **)
1752 			&ip6tun_encaps[num],
1753 			NULL, ops) ? 0 : -1;
1754 }
1755 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1756 
ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops * ops,unsigned int num)1757 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1758 			  unsigned int num)
1759 {
1760 	int ret;
1761 
1762 	if (num >= MAX_IPTUN_ENCAP_OPS)
1763 		return -ERANGE;
1764 
1765 	ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1766 		       &ip6tun_encaps[num],
1767 		       ops, NULL) == ops) ? 0 : -1;
1768 
1769 	synchronize_net();
1770 
1771 	return ret;
1772 }
1773 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1774 
ip6_tnl_encap_setup(struct ip6_tnl * t,struct ip_tunnel_encap * ipencap)1775 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1776 			struct ip_tunnel_encap *ipencap)
1777 {
1778 	int hlen;
1779 
1780 	memset(&t->encap, 0, sizeof(t->encap));
1781 
1782 	hlen = ip6_encap_hlen(ipencap);
1783 	if (hlen < 0)
1784 		return hlen;
1785 
1786 	t->encap.type = ipencap->type;
1787 	t->encap.sport = ipencap->sport;
1788 	t->encap.dport = ipencap->dport;
1789 	t->encap.flags = ipencap->flags;
1790 
1791 	t->encap_hlen = hlen;
1792 	t->hlen = t->encap_hlen + t->tun_hlen;
1793 
1794 	return 0;
1795 }
1796 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1797 
1798 static const struct net_device_ops ip6_tnl_netdev_ops = {
1799 	.ndo_init	= ip6_tnl_dev_init,
1800 	.ndo_uninit	= ip6_tnl_dev_uninit,
1801 	.ndo_start_xmit = ip6_tnl_start_xmit,
1802 	.ndo_siocdevprivate = ip6_tnl_siocdevprivate,
1803 	.ndo_change_mtu = ip6_tnl_change_mtu,
1804 	.ndo_get_stats64 = dev_get_tstats64,
1805 	.ndo_get_iflink = ip6_tnl_get_iflink,
1806 };
1807 
1808 #define IPXIPX_FEATURES (NETIF_F_SG |		\
1809 			 NETIF_F_FRAGLIST |	\
1810 			 NETIF_F_HIGHDMA |	\
1811 			 NETIF_F_GSO_SOFTWARE |	\
1812 			 NETIF_F_HW_CSUM)
1813 
1814 /**
1815  * ip6_tnl_dev_setup - setup virtual tunnel device
1816  *   @dev: virtual device associated with tunnel
1817  *
1818  * Description:
1819  *   Initialize function pointers and device parameters
1820  **/
1821 
ip6_tnl_dev_setup(struct net_device * dev)1822 static void ip6_tnl_dev_setup(struct net_device *dev)
1823 {
1824 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1825 	dev->header_ops = &ip_tunnel_header_ops;
1826 	dev->needs_free_netdev = true;
1827 	dev->priv_destructor = ip6_dev_free;
1828 
1829 	dev->type = ARPHRD_TUNNEL6;
1830 	dev->flags |= IFF_NOARP;
1831 	dev->addr_len = sizeof(struct in6_addr);
1832 	dev->features |= NETIF_F_LLTX;
1833 	netif_keep_dst(dev);
1834 
1835 	dev->features		|= IPXIPX_FEATURES;
1836 	dev->hw_features	|= IPXIPX_FEATURES;
1837 
1838 	/* This perm addr will be used as interface identifier by IPv6 */
1839 	dev->addr_assign_type = NET_ADDR_RANDOM;
1840 	eth_random_addr(dev->perm_addr);
1841 }
1842 
1843 
1844 /**
1845  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1846  *   @dev: virtual device associated with tunnel
1847  **/
1848 
1849 static inline int
ip6_tnl_dev_init_gen(struct net_device * dev)1850 ip6_tnl_dev_init_gen(struct net_device *dev)
1851 {
1852 	struct ip6_tnl *t = netdev_priv(dev);
1853 	int ret;
1854 	int t_hlen;
1855 
1856 	t->dev = dev;
1857 	t->net = dev_net(dev);
1858 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1859 	if (!dev->tstats)
1860 		return -ENOMEM;
1861 
1862 	ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1863 	if (ret)
1864 		goto free_stats;
1865 
1866 	ret = gro_cells_init(&t->gro_cells, dev);
1867 	if (ret)
1868 		goto destroy_dst;
1869 
1870 	t->tun_hlen = 0;
1871 	t->hlen = t->encap_hlen + t->tun_hlen;
1872 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1873 
1874 	dev->type = ARPHRD_TUNNEL6;
1875 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1876 	dev->mtu = ETH_DATA_LEN - t_hlen;
1877 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1878 		dev->mtu -= 8;
1879 	dev->min_mtu = ETH_MIN_MTU;
1880 	dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1881 
1882 	dev_hold(dev);
1883 	return 0;
1884 
1885 destroy_dst:
1886 	dst_cache_destroy(&t->dst_cache);
1887 free_stats:
1888 	free_percpu(dev->tstats);
1889 	dev->tstats = NULL;
1890 
1891 	return ret;
1892 }
1893 
1894 /**
1895  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1896  *   @dev: virtual device associated with tunnel
1897  **/
1898 
ip6_tnl_dev_init(struct net_device * dev)1899 static int ip6_tnl_dev_init(struct net_device *dev)
1900 {
1901 	struct ip6_tnl *t = netdev_priv(dev);
1902 	int err = ip6_tnl_dev_init_gen(dev);
1903 
1904 	if (err)
1905 		return err;
1906 	ip6_tnl_link_config(t);
1907 	if (t->parms.collect_md)
1908 		netif_keep_dst(dev);
1909 	return 0;
1910 }
1911 
1912 /**
1913  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1914  *   @dev: fallback device
1915  *
1916  * Return: 0
1917  **/
1918 
ip6_fb_tnl_dev_init(struct net_device * dev)1919 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1920 {
1921 	struct ip6_tnl *t = netdev_priv(dev);
1922 	struct net *net = dev_net(dev);
1923 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1924 
1925 	t->parms.proto = IPPROTO_IPV6;
1926 
1927 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1928 	return 0;
1929 }
1930 
ip6_tnl_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1931 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1932 			    struct netlink_ext_ack *extack)
1933 {
1934 	u8 proto;
1935 
1936 	if (!data || !data[IFLA_IPTUN_PROTO])
1937 		return 0;
1938 
1939 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1940 	if (proto != IPPROTO_IPV6 &&
1941 	    proto != IPPROTO_IPIP &&
1942 	    proto != 0)
1943 		return -EINVAL;
1944 
1945 	return 0;
1946 }
1947 
ip6_tnl_netlink_parms(struct nlattr * data[],struct __ip6_tnl_parm * parms)1948 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1949 				  struct __ip6_tnl_parm *parms)
1950 {
1951 	memset(parms, 0, sizeof(*parms));
1952 
1953 	if (!data)
1954 		return;
1955 
1956 	if (data[IFLA_IPTUN_LINK])
1957 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1958 
1959 	if (data[IFLA_IPTUN_LOCAL])
1960 		parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1961 
1962 	if (data[IFLA_IPTUN_REMOTE])
1963 		parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1964 
1965 	if (data[IFLA_IPTUN_TTL])
1966 		parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1967 
1968 	if (data[IFLA_IPTUN_ENCAP_LIMIT])
1969 		parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1970 
1971 	if (data[IFLA_IPTUN_FLOWINFO])
1972 		parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1973 
1974 	if (data[IFLA_IPTUN_FLAGS])
1975 		parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1976 
1977 	if (data[IFLA_IPTUN_PROTO])
1978 		parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1979 
1980 	if (data[IFLA_IPTUN_COLLECT_METADATA])
1981 		parms->collect_md = true;
1982 
1983 	if (data[IFLA_IPTUN_FWMARK])
1984 		parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1985 }
1986 
ip6_tnl_netlink_encap_parms(struct nlattr * data[],struct ip_tunnel_encap * ipencap)1987 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1988 					struct ip_tunnel_encap *ipencap)
1989 {
1990 	bool ret = false;
1991 
1992 	memset(ipencap, 0, sizeof(*ipencap));
1993 
1994 	if (!data)
1995 		return ret;
1996 
1997 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1998 		ret = true;
1999 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
2000 	}
2001 
2002 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
2003 		ret = true;
2004 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
2005 	}
2006 
2007 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
2008 		ret = true;
2009 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
2010 	}
2011 
2012 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
2013 		ret = true;
2014 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
2015 	}
2016 
2017 	return ret;
2018 }
2019 
ip6_tnl_newlink(struct net * src_net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)2020 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
2021 			   struct nlattr *tb[], struct nlattr *data[],
2022 			   struct netlink_ext_ack *extack)
2023 {
2024 	struct net *net = dev_net(dev);
2025 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2026 	struct ip_tunnel_encap ipencap;
2027 	struct ip6_tnl *nt, *t;
2028 	int err;
2029 
2030 	nt = netdev_priv(dev);
2031 
2032 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2033 		err = ip6_tnl_encap_setup(nt, &ipencap);
2034 		if (err < 0)
2035 			return err;
2036 	}
2037 
2038 	ip6_tnl_netlink_parms(data, &nt->parms);
2039 
2040 	if (nt->parms.collect_md) {
2041 		if (rtnl_dereference(ip6n->collect_md_tun))
2042 			return -EEXIST;
2043 	} else {
2044 		t = ip6_tnl_locate(net, &nt->parms, 0);
2045 		if (!IS_ERR(t))
2046 			return -EEXIST;
2047 	}
2048 
2049 	err = ip6_tnl_create2(dev);
2050 	if (!err && tb[IFLA_MTU])
2051 		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2052 
2053 	return err;
2054 }
2055 
ip6_tnl_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)2056 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2057 			      struct nlattr *data[],
2058 			      struct netlink_ext_ack *extack)
2059 {
2060 	struct ip6_tnl *t = netdev_priv(dev);
2061 	struct __ip6_tnl_parm p;
2062 	struct net *net = t->net;
2063 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2064 	struct ip_tunnel_encap ipencap;
2065 
2066 	if (dev == ip6n->fb_tnl_dev)
2067 		return -EINVAL;
2068 
2069 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2070 		int err = ip6_tnl_encap_setup(t, &ipencap);
2071 
2072 		if (err < 0)
2073 			return err;
2074 	}
2075 	ip6_tnl_netlink_parms(data, &p);
2076 	if (p.collect_md)
2077 		return -EINVAL;
2078 
2079 	t = ip6_tnl_locate(net, &p, 0);
2080 	if (!IS_ERR(t)) {
2081 		if (t->dev != dev)
2082 			return -EEXIST;
2083 	} else
2084 		t = netdev_priv(dev);
2085 
2086 	return ip6_tnl_update(t, &p);
2087 }
2088 
ip6_tnl_dellink(struct net_device * dev,struct list_head * head)2089 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2090 {
2091 	struct net *net = dev_net(dev);
2092 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2093 
2094 	if (dev != ip6n->fb_tnl_dev)
2095 		unregister_netdevice_queue(dev, head);
2096 }
2097 
ip6_tnl_get_size(const struct net_device * dev)2098 static size_t ip6_tnl_get_size(const struct net_device *dev)
2099 {
2100 	return
2101 		/* IFLA_IPTUN_LINK */
2102 		nla_total_size(4) +
2103 		/* IFLA_IPTUN_LOCAL */
2104 		nla_total_size(sizeof(struct in6_addr)) +
2105 		/* IFLA_IPTUN_REMOTE */
2106 		nla_total_size(sizeof(struct in6_addr)) +
2107 		/* IFLA_IPTUN_TTL */
2108 		nla_total_size(1) +
2109 		/* IFLA_IPTUN_ENCAP_LIMIT */
2110 		nla_total_size(1) +
2111 		/* IFLA_IPTUN_FLOWINFO */
2112 		nla_total_size(4) +
2113 		/* IFLA_IPTUN_FLAGS */
2114 		nla_total_size(4) +
2115 		/* IFLA_IPTUN_PROTO */
2116 		nla_total_size(1) +
2117 		/* IFLA_IPTUN_ENCAP_TYPE */
2118 		nla_total_size(2) +
2119 		/* IFLA_IPTUN_ENCAP_FLAGS */
2120 		nla_total_size(2) +
2121 		/* IFLA_IPTUN_ENCAP_SPORT */
2122 		nla_total_size(2) +
2123 		/* IFLA_IPTUN_ENCAP_DPORT */
2124 		nla_total_size(2) +
2125 		/* IFLA_IPTUN_COLLECT_METADATA */
2126 		nla_total_size(0) +
2127 		/* IFLA_IPTUN_FWMARK */
2128 		nla_total_size(4) +
2129 		0;
2130 }
2131 
ip6_tnl_fill_info(struct sk_buff * skb,const struct net_device * dev)2132 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2133 {
2134 	struct ip6_tnl *tunnel = netdev_priv(dev);
2135 	struct __ip6_tnl_parm *parm = &tunnel->parms;
2136 
2137 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2138 	    nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2139 	    nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2140 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2141 	    nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2142 	    nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2143 	    nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2144 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2145 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2146 		goto nla_put_failure;
2147 
2148 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2149 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2150 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2151 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2152 		goto nla_put_failure;
2153 
2154 	if (parm->collect_md)
2155 		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2156 			goto nla_put_failure;
2157 
2158 	return 0;
2159 
2160 nla_put_failure:
2161 	return -EMSGSIZE;
2162 }
2163 
ip6_tnl_get_link_net(const struct net_device * dev)2164 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2165 {
2166 	struct ip6_tnl *tunnel = netdev_priv(dev);
2167 
2168 	return tunnel->net;
2169 }
2170 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2171 
2172 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2173 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
2174 	[IFLA_IPTUN_LOCAL]		= { .len = sizeof(struct in6_addr) },
2175 	[IFLA_IPTUN_REMOTE]		= { .len = sizeof(struct in6_addr) },
2176 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
2177 	[IFLA_IPTUN_ENCAP_LIMIT]	= { .type = NLA_U8 },
2178 	[IFLA_IPTUN_FLOWINFO]		= { .type = NLA_U32 },
2179 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U32 },
2180 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
2181 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
2182 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
2183 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
2184 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
2185 	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
2186 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
2187 };
2188 
2189 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2190 	.kind		= "ip6tnl",
2191 	.maxtype	= IFLA_IPTUN_MAX,
2192 	.policy		= ip6_tnl_policy,
2193 	.priv_size	= sizeof(struct ip6_tnl),
2194 	.setup		= ip6_tnl_dev_setup,
2195 	.validate	= ip6_tnl_validate,
2196 	.newlink	= ip6_tnl_newlink,
2197 	.changelink	= ip6_tnl_changelink,
2198 	.dellink	= ip6_tnl_dellink,
2199 	.get_size	= ip6_tnl_get_size,
2200 	.fill_info	= ip6_tnl_fill_info,
2201 	.get_link_net	= ip6_tnl_get_link_net,
2202 };
2203 
2204 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2205 	.handler	= ip4ip6_rcv,
2206 	.err_handler	= ip4ip6_err,
2207 	.priority	=	1,
2208 };
2209 
2210 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2211 	.handler	= ip6ip6_rcv,
2212 	.err_handler	= ip6ip6_err,
2213 	.priority	=	1,
2214 };
2215 
2216 static struct xfrm6_tunnel mplsip6_handler __read_mostly = {
2217 	.handler	= mplsip6_rcv,
2218 	.err_handler	= mplsip6_err,
2219 	.priority	=	1,
2220 };
2221 
ip6_tnl_destroy_tunnels(struct net * net,struct list_head * list)2222 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2223 {
2224 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2225 	struct net_device *dev, *aux;
2226 	int h;
2227 	struct ip6_tnl *t;
2228 
2229 	for_each_netdev_safe(net, dev, aux)
2230 		if (dev->rtnl_link_ops == &ip6_link_ops)
2231 			unregister_netdevice_queue(dev, list);
2232 
2233 	for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2234 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
2235 		while (t) {
2236 			/* If dev is in the same netns, it has already
2237 			 * been added to the list by the previous loop.
2238 			 */
2239 			if (!net_eq(dev_net(t->dev), net))
2240 				unregister_netdevice_queue(t->dev, list);
2241 			t = rtnl_dereference(t->next);
2242 		}
2243 	}
2244 
2245 	t = rtnl_dereference(ip6n->tnls_wc[0]);
2246 	while (t) {
2247 		/* If dev is in the same netns, it has already
2248 		 * been added to the list by the previous loop.
2249 		 */
2250 		if (!net_eq(dev_net(t->dev), net))
2251 			unregister_netdevice_queue(t->dev, list);
2252 		t = rtnl_dereference(t->next);
2253 	}
2254 }
2255 
ip6_tnl_init_net(struct net * net)2256 static int __net_init ip6_tnl_init_net(struct net *net)
2257 {
2258 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2259 	struct ip6_tnl *t = NULL;
2260 	int err;
2261 
2262 	ip6n->tnls[0] = ip6n->tnls_wc;
2263 	ip6n->tnls[1] = ip6n->tnls_r_l;
2264 
2265 	if (!net_has_fallback_tunnels(net))
2266 		return 0;
2267 	err = -ENOMEM;
2268 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2269 					NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2270 
2271 	if (!ip6n->fb_tnl_dev)
2272 		goto err_alloc_dev;
2273 	dev_net_set(ip6n->fb_tnl_dev, net);
2274 	ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2275 	/* FB netdevice is special: we have one, and only one per netns.
2276 	 * Allowing to move it to another netns is clearly unsafe.
2277 	 */
2278 	ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2279 
2280 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2281 	if (err < 0)
2282 		goto err_register;
2283 
2284 	err = register_netdev(ip6n->fb_tnl_dev);
2285 	if (err < 0)
2286 		goto err_register;
2287 
2288 	t = netdev_priv(ip6n->fb_tnl_dev);
2289 
2290 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2291 	return 0;
2292 
2293 err_register:
2294 	free_netdev(ip6n->fb_tnl_dev);
2295 err_alloc_dev:
2296 	return err;
2297 }
2298 
ip6_tnl_exit_batch_net(struct list_head * net_list)2299 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2300 {
2301 	struct net *net;
2302 	LIST_HEAD(list);
2303 
2304 	rtnl_lock();
2305 	list_for_each_entry(net, net_list, exit_list)
2306 		ip6_tnl_destroy_tunnels(net, &list);
2307 	unregister_netdevice_many(&list);
2308 	rtnl_unlock();
2309 }
2310 
2311 static struct pernet_operations ip6_tnl_net_ops = {
2312 	.init = ip6_tnl_init_net,
2313 	.exit_batch = ip6_tnl_exit_batch_net,
2314 	.id   = &ip6_tnl_net_id,
2315 	.size = sizeof(struct ip6_tnl_net),
2316 };
2317 
2318 /**
2319  * ip6_tunnel_init - register protocol and reserve needed resources
2320  *
2321  * Return: 0 on success
2322  **/
2323 
ip6_tunnel_init(void)2324 static int __init ip6_tunnel_init(void)
2325 {
2326 	int  err;
2327 
2328 	if (!ipv6_mod_enabled())
2329 		return -EOPNOTSUPP;
2330 
2331 	err = register_pernet_device(&ip6_tnl_net_ops);
2332 	if (err < 0)
2333 		goto out_pernet;
2334 
2335 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2336 	if (err < 0) {
2337 		pr_err("%s: can't register ip4ip6\n", __func__);
2338 		goto out_ip4ip6;
2339 	}
2340 
2341 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2342 	if (err < 0) {
2343 		pr_err("%s: can't register ip6ip6\n", __func__);
2344 		goto out_ip6ip6;
2345 	}
2346 
2347 	if (ip6_tnl_mpls_supported()) {
2348 		err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS);
2349 		if (err < 0) {
2350 			pr_err("%s: can't register mplsip6\n", __func__);
2351 			goto out_mplsip6;
2352 		}
2353 	}
2354 
2355 	err = rtnl_link_register(&ip6_link_ops);
2356 	if (err < 0)
2357 		goto rtnl_link_failed;
2358 
2359 	return 0;
2360 
2361 rtnl_link_failed:
2362 	if (ip6_tnl_mpls_supported())
2363 		xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS);
2364 out_mplsip6:
2365 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2366 out_ip6ip6:
2367 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2368 out_ip4ip6:
2369 	unregister_pernet_device(&ip6_tnl_net_ops);
2370 out_pernet:
2371 	return err;
2372 }
2373 
2374 /**
2375  * ip6_tunnel_cleanup - free resources and unregister protocol
2376  **/
2377 
ip6_tunnel_cleanup(void)2378 static void __exit ip6_tunnel_cleanup(void)
2379 {
2380 	rtnl_link_unregister(&ip6_link_ops);
2381 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2382 		pr_info("%s: can't deregister ip4ip6\n", __func__);
2383 
2384 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2385 		pr_info("%s: can't deregister ip6ip6\n", __func__);
2386 
2387 	if (ip6_tnl_mpls_supported() &&
2388 	    xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS))
2389 		pr_info("%s: can't deregister mplsip6\n", __func__);
2390 	unregister_pernet_device(&ip6_tnl_net_ops);
2391 }
2392 
2393 module_init(ip6_tunnel_init);
2394 module_exit(ip6_tunnel_cleanup);
2395