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 		strlcpy(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 	dev_put_track(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 next, 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 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
414 			if (frag_hdr->frag_off)
415 				break;
416 			optlen = 8;
417 		} else if (nexthdr == NEXTHDR_AUTH) {
418 			optlen = ipv6_authlen(hdr);
419 		} else {
420 			optlen = ipv6_optlen(hdr);
421 		}
422 		/* cache hdr->nexthdr, since pskb_may_pull() might
423 		 * invalidate hdr
424 		 */
425 		next = hdr->nexthdr;
426 		if (nexthdr == NEXTHDR_DEST) {
427 			u16 i = 2;
428 
429 			/* Remember : hdr is no longer valid at this point. */
430 			if (!pskb_may_pull(skb, off + optlen))
431 				break;
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 = next;
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 	struct pcpu_sw_netstats *tstats;
800 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
801 	int err;
802 
803 	if ((!(tpi->flags & TUNNEL_CSUM) &&
804 	     (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
805 	    ((tpi->flags & TUNNEL_CSUM) &&
806 	     !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
807 		tunnel->dev->stats.rx_crc_errors++;
808 		tunnel->dev->stats.rx_errors++;
809 		goto drop;
810 	}
811 
812 	if (tunnel->parms.i_flags & TUNNEL_SEQ) {
813 		if (!(tpi->flags & TUNNEL_SEQ) ||
814 		    (tunnel->i_seqno &&
815 		     (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
816 			tunnel->dev->stats.rx_fifo_errors++;
817 			tunnel->dev->stats.rx_errors++;
818 			goto drop;
819 		}
820 		tunnel->i_seqno = ntohl(tpi->seq) + 1;
821 	}
822 
823 	skb->protocol = tpi->proto;
824 
825 	/* Warning: All skb pointers will be invalidated! */
826 	if (tunnel->dev->type == ARPHRD_ETHER) {
827 		if (!pskb_may_pull(skb, ETH_HLEN)) {
828 			tunnel->dev->stats.rx_length_errors++;
829 			tunnel->dev->stats.rx_errors++;
830 			goto drop;
831 		}
832 
833 		ipv6h = ipv6_hdr(skb);
834 		skb->protocol = eth_type_trans(skb, tunnel->dev);
835 		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
836 	} else {
837 		skb->dev = tunnel->dev;
838 		skb_reset_mac_header(skb);
839 	}
840 
841 	skb_reset_network_header(skb);
842 	memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
843 
844 	__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
845 
846 	err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
847 	if (unlikely(err)) {
848 		if (log_ecn_err)
849 			net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
850 					     &ipv6h->saddr,
851 					     ipv6_get_dsfield(ipv6h));
852 		if (err > 1) {
853 			++tunnel->dev->stats.rx_frame_errors;
854 			++tunnel->dev->stats.rx_errors;
855 			goto drop;
856 		}
857 	}
858 
859 	tstats = this_cpu_ptr(tunnel->dev->tstats);
860 	u64_stats_update_begin(&tstats->syncp);
861 	tstats->rx_packets++;
862 	tstats->rx_bytes += skb->len;
863 	u64_stats_update_end(&tstats->syncp);
864 
865 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
866 
867 	if (tun_dst)
868 		skb_dst_set(skb, (struct dst_entry *)tun_dst);
869 
870 	gro_cells_receive(&tunnel->gro_cells, skb);
871 	return 0;
872 
873 drop:
874 	if (tun_dst)
875 		dst_release((struct dst_entry *)tun_dst);
876 	kfree_skb(skb);
877 	return 0;
878 }
879 
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)880 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
881 		const struct tnl_ptk_info *tpi,
882 		struct metadata_dst *tun_dst,
883 		bool log_ecn_err)
884 {
885 	int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
886 				    const struct ipv6hdr *ipv6h,
887 				    struct sk_buff *skb);
888 
889 	dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
890 	if (tpi->proto == htons(ETH_P_IP))
891 		dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
892 
893 	return __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
894 			     log_ecn_err);
895 }
896 EXPORT_SYMBOL(ip6_tnl_rcv);
897 
898 static const struct tnl_ptk_info tpi_v6 = {
899 	/* no tunnel info required for ipxip6. */
900 	.proto = htons(ETH_P_IPV6),
901 };
902 
903 static const struct tnl_ptk_info tpi_v4 = {
904 	/* no tunnel info required for ipxip6. */
905 	.proto = htons(ETH_P_IP),
906 };
907 
908 static const struct tnl_ptk_info tpi_mpls = {
909 	/* no tunnel info required for mplsip6. */
910 	.proto = htons(ETH_P_MPLS_UC),
911 };
912 
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))913 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
914 		      const struct tnl_ptk_info *tpi,
915 		      int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
916 						  const struct ipv6hdr *ipv6h,
917 						  struct sk_buff *skb))
918 {
919 	struct ip6_tnl *t;
920 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
921 	struct metadata_dst *tun_dst = NULL;
922 	int ret = -1;
923 
924 	rcu_read_lock();
925 	t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr);
926 
927 	if (t) {
928 		u8 tproto = READ_ONCE(t->parms.proto);
929 
930 		if (tproto != ipproto && tproto != 0)
931 			goto drop;
932 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
933 			goto drop;
934 		ipv6h = ipv6_hdr(skb);
935 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
936 			goto drop;
937 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
938 			goto drop;
939 		if (t->parms.collect_md) {
940 			tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
941 			if (!tun_dst)
942 				goto drop;
943 		}
944 		ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
945 				    log_ecn_error);
946 	}
947 
948 	rcu_read_unlock();
949 
950 	return ret;
951 
952 drop:
953 	rcu_read_unlock();
954 	kfree_skb(skb);
955 	return 0;
956 }
957 
ip4ip6_rcv(struct sk_buff * skb)958 static int ip4ip6_rcv(struct sk_buff *skb)
959 {
960 	return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
961 			  ip4ip6_dscp_ecn_decapsulate);
962 }
963 
ip6ip6_rcv(struct sk_buff * skb)964 static int ip6ip6_rcv(struct sk_buff *skb)
965 {
966 	return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
967 			  ip6ip6_dscp_ecn_decapsulate);
968 }
969 
mplsip6_rcv(struct sk_buff * skb)970 static int mplsip6_rcv(struct sk_buff *skb)
971 {
972 	return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls,
973 			  mplsip6_dscp_ecn_decapsulate);
974 }
975 
976 struct ipv6_tel_txoption {
977 	struct ipv6_txoptions ops;
978 	__u8 dst_opt[8];
979 };
980 
init_tel_txopt(struct ipv6_tel_txoption * opt,__u8 encap_limit)981 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
982 {
983 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
984 
985 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
986 	opt->dst_opt[3] = 1;
987 	opt->dst_opt[4] = encap_limit;
988 	opt->dst_opt[5] = IPV6_TLV_PADN;
989 	opt->dst_opt[6] = 1;
990 
991 	opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
992 	opt->ops.opt_nflen = 8;
993 }
994 
995 /**
996  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
997  *   @t: the outgoing tunnel device
998  *   @hdr: IPv6 header from the incoming packet
999  *
1000  * Description:
1001  *   Avoid trivial tunneling loop by checking that tunnel exit-point
1002  *   doesn't match source of incoming packet.
1003  *
1004  * Return:
1005  *   1 if conflict,
1006  *   0 else
1007  **/
1008 
1009 static inline bool
ip6_tnl_addr_conflict(const struct ip6_tnl * t,const struct ipv6hdr * hdr)1010 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
1011 {
1012 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
1013 }
1014 
ip6_tnl_xmit_ctl(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)1015 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
1016 		     const struct in6_addr *laddr,
1017 		     const struct in6_addr *raddr)
1018 {
1019 	struct __ip6_tnl_parm *p = &t->parms;
1020 	int ret = 0;
1021 	struct net *net = t->net;
1022 
1023 	if (t->parms.collect_md)
1024 		return 1;
1025 
1026 	if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1027 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1028 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1029 		struct net_device *ldev = NULL;
1030 
1031 		rcu_read_lock();
1032 		if (p->link)
1033 			ldev = dev_get_by_index_rcu(net, p->link);
1034 
1035 		if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
1036 						      0, IFA_F_TENTATIVE)))
1037 			pr_warn_ratelimited("%s xmit: Local address not yet configured!\n",
1038 					    p->name);
1039 		else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1040 			 !ipv6_addr_is_multicast(raddr) &&
1041 			 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1042 							  true, 0, IFA_F_TENTATIVE)))
1043 			pr_warn_ratelimited("%s xmit: Routing loop! Remote address found on this node!\n",
1044 					    p->name);
1045 		else
1046 			ret = 1;
1047 		rcu_read_unlock();
1048 	}
1049 	return ret;
1050 }
1051 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1052 
1053 /**
1054  * ip6_tnl_xmit - encapsulate packet and send
1055  *   @skb: the outgoing socket buffer
1056  *   @dev: the outgoing tunnel device
1057  *   @dsfield: dscp code for outer header
1058  *   @fl6: flow of tunneled packet
1059  *   @encap_limit: encapsulation limit
1060  *   @pmtu: Path MTU is stored if packet is too big
1061  *   @proto: next header value
1062  *
1063  * Description:
1064  *   Build new header and do some sanity checks on the packet before sending
1065  *   it.
1066  *
1067  * Return:
1068  *   0 on success
1069  *   -1 fail
1070  *   %-EMSGSIZE message too big. return mtu in this case.
1071  **/
1072 
ip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev,__u8 dsfield,struct flowi6 * fl6,int encap_limit,__u32 * pmtu,__u8 proto)1073 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1074 		 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1075 		 __u8 proto)
1076 {
1077 	struct ip6_tnl *t = netdev_priv(dev);
1078 	struct net *net = t->net;
1079 	struct net_device_stats *stats = &t->dev->stats;
1080 	struct ipv6hdr *ipv6h;
1081 	struct ipv6_tel_txoption opt;
1082 	struct dst_entry *dst = NULL, *ndst = NULL;
1083 	struct net_device *tdev;
1084 	int mtu;
1085 	unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1086 	unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1087 	unsigned int max_headroom = psh_hlen;
1088 	bool use_cache = false;
1089 	u8 hop_limit;
1090 	int err = -1;
1091 
1092 	if (t->parms.collect_md) {
1093 		hop_limit = skb_tunnel_info(skb)->key.ttl;
1094 		goto route_lookup;
1095 	} else {
1096 		hop_limit = t->parms.hop_limit;
1097 	}
1098 
1099 	/* NBMA tunnel */
1100 	if (ipv6_addr_any(&t->parms.raddr)) {
1101 		if (skb->protocol == htons(ETH_P_IPV6)) {
1102 			struct in6_addr *addr6;
1103 			struct neighbour *neigh;
1104 			int addr_type;
1105 
1106 			if (!skb_dst(skb))
1107 				goto tx_err_link_failure;
1108 
1109 			neigh = dst_neigh_lookup(skb_dst(skb),
1110 						 &ipv6_hdr(skb)->daddr);
1111 			if (!neigh)
1112 				goto tx_err_link_failure;
1113 
1114 			addr6 = (struct in6_addr *)&neigh->primary_key;
1115 			addr_type = ipv6_addr_type(addr6);
1116 
1117 			if (addr_type == IPV6_ADDR_ANY)
1118 				addr6 = &ipv6_hdr(skb)->daddr;
1119 
1120 			memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1121 			neigh_release(neigh);
1122 		} else if (skb->protocol == htons(ETH_P_IP)) {
1123 			const struct rtable *rt = skb_rtable(skb);
1124 
1125 			if (!rt)
1126 				goto tx_err_link_failure;
1127 
1128 			if (rt->rt_gw_family == AF_INET6)
1129 				memcpy(&fl6->daddr, &rt->rt_gw6, sizeof(fl6->daddr));
1130 		}
1131 	} else if (t->parms.proto != 0 && !(t->parms.flags &
1132 					    (IP6_TNL_F_USE_ORIG_TCLASS |
1133 					     IP6_TNL_F_USE_ORIG_FWMARK))) {
1134 		/* enable the cache only if neither the outer protocol nor the
1135 		 * routing decision depends on the current inner header value
1136 		 */
1137 		use_cache = true;
1138 	}
1139 
1140 	if (use_cache)
1141 		dst = dst_cache_get(&t->dst_cache);
1142 
1143 	if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1144 		goto tx_err_link_failure;
1145 
1146 	if (!dst) {
1147 route_lookup:
1148 		/* add dsfield to flowlabel for route lookup */
1149 		fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1150 
1151 		dst = ip6_route_output(net, NULL, fl6);
1152 
1153 		if (dst->error)
1154 			goto tx_err_link_failure;
1155 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1156 		if (IS_ERR(dst)) {
1157 			err = PTR_ERR(dst);
1158 			dst = NULL;
1159 			goto tx_err_link_failure;
1160 		}
1161 		if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1162 		    ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1163 				       &fl6->daddr, 0, &fl6->saddr))
1164 			goto tx_err_link_failure;
1165 		ndst = dst;
1166 	}
1167 
1168 	tdev = dst->dev;
1169 
1170 	if (tdev == dev) {
1171 		stats->collisions++;
1172 		net_warn_ratelimited("%s: Local routing loop detected!\n",
1173 				     t->parms.name);
1174 		goto tx_err_dst_release;
1175 	}
1176 	mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1177 	if (encap_limit >= 0) {
1178 		max_headroom += 8;
1179 		mtu -= 8;
1180 	}
1181 	mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1182 		       IPV6_MIN_MTU : IPV4_MIN_MTU);
1183 
1184 	skb_dst_update_pmtu_no_confirm(skb, mtu);
1185 	if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1186 		*pmtu = mtu;
1187 		err = -EMSGSIZE;
1188 		goto tx_err_dst_release;
1189 	}
1190 
1191 	if (t->err_count > 0) {
1192 		if (time_before(jiffies,
1193 				t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1194 			t->err_count--;
1195 
1196 			dst_link_failure(skb);
1197 		} else {
1198 			t->err_count = 0;
1199 		}
1200 	}
1201 
1202 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1203 
1204 	/*
1205 	 * Okay, now see if we can stuff it in the buffer as-is.
1206 	 */
1207 	max_headroom += LL_RESERVED_SPACE(tdev);
1208 
1209 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1210 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1211 		struct sk_buff *new_skb;
1212 
1213 		new_skb = skb_realloc_headroom(skb, max_headroom);
1214 		if (!new_skb)
1215 			goto tx_err_dst_release;
1216 
1217 		if (skb->sk)
1218 			skb_set_owner_w(new_skb, skb->sk);
1219 		consume_skb(skb);
1220 		skb = new_skb;
1221 	}
1222 
1223 	if (t->parms.collect_md) {
1224 		if (t->encap.type != TUNNEL_ENCAP_NONE)
1225 			goto tx_err_dst_release;
1226 	} else {
1227 		if (use_cache && ndst)
1228 			dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1229 	}
1230 	skb_dst_set(skb, dst);
1231 
1232 	if (hop_limit == 0) {
1233 		if (skb->protocol == htons(ETH_P_IP))
1234 			hop_limit = ip_hdr(skb)->ttl;
1235 		else if (skb->protocol == htons(ETH_P_IPV6))
1236 			hop_limit = ipv6_hdr(skb)->hop_limit;
1237 		else
1238 			hop_limit = ip6_dst_hoplimit(dst);
1239 	}
1240 
1241 	/* Calculate max headroom for all the headers and adjust
1242 	 * needed_headroom if necessary.
1243 	 */
1244 	max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1245 			+ dst->header_len + t->hlen;
1246 	if (max_headroom > dev->needed_headroom)
1247 		dev->needed_headroom = max_headroom;
1248 
1249 	err = ip6_tnl_encap(skb, t, &proto, fl6);
1250 	if (err)
1251 		return err;
1252 
1253 	if (encap_limit >= 0) {
1254 		init_tel_txopt(&opt, encap_limit);
1255 		ipv6_push_frag_opts(skb, &opt.ops, &proto);
1256 	}
1257 
1258 	skb_push(skb, sizeof(struct ipv6hdr));
1259 	skb_reset_network_header(skb);
1260 	ipv6h = ipv6_hdr(skb);
1261 	ip6_flow_hdr(ipv6h, dsfield,
1262 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1263 	ipv6h->hop_limit = hop_limit;
1264 	ipv6h->nexthdr = proto;
1265 	ipv6h->saddr = fl6->saddr;
1266 	ipv6h->daddr = fl6->daddr;
1267 	ip6tunnel_xmit(NULL, skb, dev);
1268 	return 0;
1269 tx_err_link_failure:
1270 	stats->tx_carrier_errors++;
1271 	dst_link_failure(skb);
1272 tx_err_dst_release:
1273 	dst_release(dst);
1274 	return err;
1275 }
1276 EXPORT_SYMBOL(ip6_tnl_xmit);
1277 
1278 static inline int
ipxip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev,u8 protocol)1279 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev,
1280 		u8 protocol)
1281 {
1282 	struct ip6_tnl *t = netdev_priv(dev);
1283 	struct ipv6hdr *ipv6h;
1284 	const struct iphdr  *iph;
1285 	int encap_limit = -1;
1286 	__u16 offset;
1287 	struct flowi6 fl6;
1288 	__u8 dsfield, orig_dsfield;
1289 	__u32 mtu;
1290 	u8 tproto;
1291 	int err;
1292 
1293 	tproto = READ_ONCE(t->parms.proto);
1294 	if (tproto != protocol && tproto != 0)
1295 		return -1;
1296 
1297 	if (t->parms.collect_md) {
1298 		struct ip_tunnel_info *tun_info;
1299 		const struct ip_tunnel_key *key;
1300 
1301 		tun_info = skb_tunnel_info(skb);
1302 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1303 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1304 			return -1;
1305 		key = &tun_info->key;
1306 		memset(&fl6, 0, sizeof(fl6));
1307 		fl6.flowi6_proto = protocol;
1308 		fl6.saddr = key->u.ipv6.src;
1309 		fl6.daddr = key->u.ipv6.dst;
1310 		fl6.flowlabel = key->label;
1311 		dsfield =  key->tos;
1312 		switch (protocol) {
1313 		case IPPROTO_IPIP:
1314 			iph = ip_hdr(skb);
1315 			orig_dsfield = ipv4_get_dsfield(iph);
1316 			break;
1317 		case IPPROTO_IPV6:
1318 			ipv6h = ipv6_hdr(skb);
1319 			orig_dsfield = ipv6_get_dsfield(ipv6h);
1320 			break;
1321 		default:
1322 			orig_dsfield = dsfield;
1323 			break;
1324 		}
1325 	} else {
1326 		if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1327 			encap_limit = t->parms.encap_limit;
1328 		if (protocol == IPPROTO_IPV6) {
1329 			offset = ip6_tnl_parse_tlv_enc_lim(skb,
1330 						skb_network_header(skb));
1331 			/* ip6_tnl_parse_tlv_enc_lim() might have
1332 			 * reallocated skb->head
1333 			 */
1334 			if (offset > 0) {
1335 				struct ipv6_tlv_tnl_enc_lim *tel;
1336 
1337 				tel = (void *)&skb_network_header(skb)[offset];
1338 				if (tel->encap_limit == 0) {
1339 					icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
1340 							ICMPV6_HDR_FIELD, offset + 2);
1341 					return -1;
1342 				}
1343 				encap_limit = tel->encap_limit - 1;
1344 			}
1345 		}
1346 
1347 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1348 		fl6.flowi6_proto = protocol;
1349 
1350 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1351 			fl6.flowi6_mark = skb->mark;
1352 		else
1353 			fl6.flowi6_mark = t->parms.fwmark;
1354 		switch (protocol) {
1355 		case IPPROTO_IPIP:
1356 			iph = ip_hdr(skb);
1357 			orig_dsfield = ipv4_get_dsfield(iph);
1358 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1359 				dsfield = orig_dsfield;
1360 			else
1361 				dsfield = ip6_tclass(t->parms.flowinfo);
1362 			break;
1363 		case IPPROTO_IPV6:
1364 			ipv6h = ipv6_hdr(skb);
1365 			orig_dsfield = ipv6_get_dsfield(ipv6h);
1366 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1367 				dsfield = orig_dsfield;
1368 			else
1369 				dsfield = ip6_tclass(t->parms.flowinfo);
1370 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1371 				fl6.flowlabel |= ip6_flowlabel(ipv6h);
1372 			break;
1373 		default:
1374 			orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo);
1375 			break;
1376 		}
1377 	}
1378 
1379 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1380 	dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield);
1381 
1382 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1383 		return -1;
1384 
1385 	skb_set_inner_ipproto(skb, protocol);
1386 
1387 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1388 			   protocol);
1389 	if (err != 0) {
1390 		/* XXX: send ICMP error even if DF is not set. */
1391 		if (err == -EMSGSIZE)
1392 			switch (protocol) {
1393 			case IPPROTO_IPIP:
1394 				icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1395 					      ICMP_FRAG_NEEDED, htonl(mtu));
1396 				break;
1397 			case IPPROTO_IPV6:
1398 				icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1399 				break;
1400 			default:
1401 				break;
1402 			}
1403 		return -1;
1404 	}
1405 
1406 	return 0;
1407 }
1408 
1409 static netdev_tx_t
ip6_tnl_start_xmit(struct sk_buff * skb,struct net_device * dev)1410 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1411 {
1412 	struct ip6_tnl *t = netdev_priv(dev);
1413 	struct net_device_stats *stats = &t->dev->stats;
1414 	u8 ipproto;
1415 	int ret;
1416 
1417 	if (!pskb_inet_may_pull(skb))
1418 		goto tx_err;
1419 
1420 	switch (skb->protocol) {
1421 	case htons(ETH_P_IP):
1422 		ipproto = IPPROTO_IPIP;
1423 		break;
1424 	case htons(ETH_P_IPV6):
1425 		if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb)))
1426 			goto tx_err;
1427 		ipproto = IPPROTO_IPV6;
1428 		break;
1429 	case htons(ETH_P_MPLS_UC):
1430 		ipproto = IPPROTO_MPLS;
1431 		break;
1432 	default:
1433 		goto tx_err;
1434 	}
1435 
1436 	ret = ipxip6_tnl_xmit(skb, dev, ipproto);
1437 	if (ret < 0)
1438 		goto tx_err;
1439 
1440 	return NETDEV_TX_OK;
1441 
1442 tx_err:
1443 	stats->tx_errors++;
1444 	stats->tx_dropped++;
1445 	kfree_skb(skb);
1446 	return NETDEV_TX_OK;
1447 }
1448 
ip6_tnl_link_config(struct ip6_tnl * t)1449 static void ip6_tnl_link_config(struct ip6_tnl *t)
1450 {
1451 	struct net_device *dev = t->dev;
1452 	struct net_device *tdev = NULL;
1453 	struct __ip6_tnl_parm *p = &t->parms;
1454 	struct flowi6 *fl6 = &t->fl.u.ip6;
1455 	unsigned int mtu;
1456 	int t_hlen;
1457 
1458 	__dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
1459 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1460 
1461 	/* Set up flowi template */
1462 	fl6->saddr = p->laddr;
1463 	fl6->daddr = p->raddr;
1464 	fl6->flowi6_oif = p->link;
1465 	fl6->flowlabel = 0;
1466 
1467 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1468 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1469 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1470 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1471 
1472 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1473 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1474 
1475 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1476 		dev->flags |= IFF_POINTOPOINT;
1477 	else
1478 		dev->flags &= ~IFF_POINTOPOINT;
1479 
1480 	t->tun_hlen = 0;
1481 	t->hlen = t->encap_hlen + t->tun_hlen;
1482 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1483 
1484 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1485 		int strict = (ipv6_addr_type(&p->raddr) &
1486 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1487 
1488 		struct rt6_info *rt = rt6_lookup(t->net,
1489 						 &p->raddr, &p->laddr,
1490 						 p->link, NULL, strict);
1491 		if (rt) {
1492 			tdev = rt->dst.dev;
1493 			ip6_rt_put(rt);
1494 		}
1495 
1496 		if (!tdev && p->link)
1497 			tdev = __dev_get_by_index(t->net, p->link);
1498 
1499 		if (tdev) {
1500 			dev->hard_header_len = tdev->hard_header_len + t_hlen;
1501 			mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU);
1502 
1503 			dev->mtu = mtu - t_hlen;
1504 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1505 				dev->mtu -= 8;
1506 
1507 			if (dev->mtu < IPV6_MIN_MTU)
1508 				dev->mtu = IPV6_MIN_MTU;
1509 		}
1510 	}
1511 }
1512 
1513 /**
1514  * ip6_tnl_change - update the tunnel parameters
1515  *   @t: tunnel to be changed
1516  *   @p: tunnel configuration parameters
1517  *
1518  * Description:
1519  *   ip6_tnl_change() updates the tunnel parameters
1520  **/
1521 
1522 static int
ip6_tnl_change(struct ip6_tnl * t,const struct __ip6_tnl_parm * p)1523 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1524 {
1525 	t->parms.laddr = p->laddr;
1526 	t->parms.raddr = p->raddr;
1527 	t->parms.flags = p->flags;
1528 	t->parms.hop_limit = p->hop_limit;
1529 	t->parms.encap_limit = p->encap_limit;
1530 	t->parms.flowinfo = p->flowinfo;
1531 	t->parms.link = p->link;
1532 	t->parms.proto = p->proto;
1533 	t->parms.fwmark = p->fwmark;
1534 	dst_cache_reset(&t->dst_cache);
1535 	ip6_tnl_link_config(t);
1536 	return 0;
1537 }
1538 
ip6_tnl_update(struct ip6_tnl * t,struct __ip6_tnl_parm * p)1539 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1540 {
1541 	struct net *net = t->net;
1542 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1543 	int err;
1544 
1545 	ip6_tnl_unlink(ip6n, t);
1546 	synchronize_net();
1547 	err = ip6_tnl_change(t, p);
1548 	ip6_tnl_link(ip6n, t);
1549 	netdev_state_change(t->dev);
1550 	return err;
1551 }
1552 
ip6_tnl0_update(struct ip6_tnl * t,struct __ip6_tnl_parm * p)1553 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1554 {
1555 	/* for default tnl0 device allow to change only the proto */
1556 	t->parms.proto = p->proto;
1557 	netdev_state_change(t->dev);
1558 	return 0;
1559 }
1560 
1561 static void
ip6_tnl_parm_from_user(struct __ip6_tnl_parm * p,const struct ip6_tnl_parm * u)1562 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1563 {
1564 	p->laddr = u->laddr;
1565 	p->raddr = u->raddr;
1566 	p->flags = u->flags;
1567 	p->hop_limit = u->hop_limit;
1568 	p->encap_limit = u->encap_limit;
1569 	p->flowinfo = u->flowinfo;
1570 	p->link = u->link;
1571 	p->proto = u->proto;
1572 	memcpy(p->name, u->name, sizeof(u->name));
1573 }
1574 
1575 static void
ip6_tnl_parm_to_user(struct ip6_tnl_parm * u,const struct __ip6_tnl_parm * p)1576 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1577 {
1578 	u->laddr = p->laddr;
1579 	u->raddr = p->raddr;
1580 	u->flags = p->flags;
1581 	u->hop_limit = p->hop_limit;
1582 	u->encap_limit = p->encap_limit;
1583 	u->flowinfo = p->flowinfo;
1584 	u->link = p->link;
1585 	u->proto = p->proto;
1586 	memcpy(u->name, p->name, sizeof(u->name));
1587 }
1588 
1589 /**
1590  * ip6_tnl_siocdevprivate - configure ipv6 tunnels from userspace
1591  *   @dev: virtual device associated with tunnel
1592  *   @ifr: unused
1593  *   @data: parameters passed from userspace
1594  *   @cmd: command to be performed
1595  *
1596  * Description:
1597  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1598  *   from userspace.
1599  *
1600  *   The possible commands are the following:
1601  *     %SIOCGETTUNNEL: get tunnel parameters for device
1602  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1603  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1604  *     %SIOCDELTUNNEL: delete tunnel
1605  *
1606  *   The fallback device "ip6tnl0", created during module
1607  *   initialization, can be used for creating other tunnel devices.
1608  *
1609  * Return:
1610  *   0 on success,
1611  *   %-EFAULT if unable to copy data to or from userspace,
1612  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1613  *   %-EINVAL if passed tunnel parameters are invalid,
1614  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1615  *   %-ENODEV if attempting to change or delete a nonexisting device
1616  **/
1617 
1618 static int
ip6_tnl_siocdevprivate(struct net_device * dev,struct ifreq * ifr,void __user * data,int cmd)1619 ip6_tnl_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1620 		       void __user *data, int cmd)
1621 {
1622 	int err = 0;
1623 	struct ip6_tnl_parm p;
1624 	struct __ip6_tnl_parm p1;
1625 	struct ip6_tnl *t = netdev_priv(dev);
1626 	struct net *net = t->net;
1627 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1628 
1629 	memset(&p1, 0, sizeof(p1));
1630 
1631 	switch (cmd) {
1632 	case SIOCGETTUNNEL:
1633 		if (dev == ip6n->fb_tnl_dev) {
1634 			if (copy_from_user(&p, data, sizeof(p))) {
1635 				err = -EFAULT;
1636 				break;
1637 			}
1638 			ip6_tnl_parm_from_user(&p1, &p);
1639 			t = ip6_tnl_locate(net, &p1, 0);
1640 			if (IS_ERR(t))
1641 				t = netdev_priv(dev);
1642 		} else {
1643 			memset(&p, 0, sizeof(p));
1644 		}
1645 		ip6_tnl_parm_to_user(&p, &t->parms);
1646 		if (copy_to_user(data, &p, sizeof(p)))
1647 			err = -EFAULT;
1648 		break;
1649 	case SIOCADDTUNNEL:
1650 	case SIOCCHGTUNNEL:
1651 		err = -EPERM;
1652 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1653 			break;
1654 		err = -EFAULT;
1655 		if (copy_from_user(&p, data, sizeof(p)))
1656 			break;
1657 		err = -EINVAL;
1658 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1659 		    p.proto != 0)
1660 			break;
1661 		ip6_tnl_parm_from_user(&p1, &p);
1662 		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1663 		if (cmd == SIOCCHGTUNNEL) {
1664 			if (!IS_ERR(t)) {
1665 				if (t->dev != dev) {
1666 					err = -EEXIST;
1667 					break;
1668 				}
1669 			} else
1670 				t = netdev_priv(dev);
1671 			if (dev == ip6n->fb_tnl_dev)
1672 				err = ip6_tnl0_update(t, &p1);
1673 			else
1674 				err = ip6_tnl_update(t, &p1);
1675 		}
1676 		if (!IS_ERR(t)) {
1677 			err = 0;
1678 			ip6_tnl_parm_to_user(&p, &t->parms);
1679 			if (copy_to_user(data, &p, sizeof(p)))
1680 				err = -EFAULT;
1681 
1682 		} else {
1683 			err = PTR_ERR(t);
1684 		}
1685 		break;
1686 	case SIOCDELTUNNEL:
1687 		err = -EPERM;
1688 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1689 			break;
1690 
1691 		if (dev == ip6n->fb_tnl_dev) {
1692 			err = -EFAULT;
1693 			if (copy_from_user(&p, data, sizeof(p)))
1694 				break;
1695 			err = -ENOENT;
1696 			ip6_tnl_parm_from_user(&p1, &p);
1697 			t = ip6_tnl_locate(net, &p1, 0);
1698 			if (IS_ERR(t))
1699 				break;
1700 			err = -EPERM;
1701 			if (t->dev == ip6n->fb_tnl_dev)
1702 				break;
1703 			dev = t->dev;
1704 		}
1705 		err = 0;
1706 		unregister_netdevice(dev);
1707 		break;
1708 	default:
1709 		err = -EINVAL;
1710 	}
1711 	return err;
1712 }
1713 
1714 /**
1715  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1716  *   @dev: virtual device associated with tunnel
1717  *   @new_mtu: the new mtu
1718  *
1719  * Return:
1720  *   0 on success,
1721  *   %-EINVAL if mtu too small
1722  **/
1723 
ip6_tnl_change_mtu(struct net_device * dev,int new_mtu)1724 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1725 {
1726 	struct ip6_tnl *tnl = netdev_priv(dev);
1727 
1728 	if (tnl->parms.proto == IPPROTO_IPV6) {
1729 		if (new_mtu < IPV6_MIN_MTU)
1730 			return -EINVAL;
1731 	} else {
1732 		if (new_mtu < ETH_MIN_MTU)
1733 			return -EINVAL;
1734 	}
1735 	if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1736 		if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1737 			return -EINVAL;
1738 	} else {
1739 		if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1740 			return -EINVAL;
1741 	}
1742 	dev->mtu = new_mtu;
1743 	return 0;
1744 }
1745 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1746 
ip6_tnl_get_iflink(const struct net_device * dev)1747 int ip6_tnl_get_iflink(const struct net_device *dev)
1748 {
1749 	struct ip6_tnl *t = netdev_priv(dev);
1750 
1751 	return t->parms.link;
1752 }
1753 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1754 
ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops * ops,unsigned int num)1755 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1756 			  unsigned int num)
1757 {
1758 	if (num >= MAX_IPTUN_ENCAP_OPS)
1759 		return -ERANGE;
1760 
1761 	return !cmpxchg((const struct ip6_tnl_encap_ops **)
1762 			&ip6tun_encaps[num],
1763 			NULL, ops) ? 0 : -1;
1764 }
1765 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1766 
ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops * ops,unsigned int num)1767 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1768 			  unsigned int num)
1769 {
1770 	int ret;
1771 
1772 	if (num >= MAX_IPTUN_ENCAP_OPS)
1773 		return -ERANGE;
1774 
1775 	ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1776 		       &ip6tun_encaps[num],
1777 		       ops, NULL) == ops) ? 0 : -1;
1778 
1779 	synchronize_net();
1780 
1781 	return ret;
1782 }
1783 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1784 
ip6_tnl_encap_setup(struct ip6_tnl * t,struct ip_tunnel_encap * ipencap)1785 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1786 			struct ip_tunnel_encap *ipencap)
1787 {
1788 	int hlen;
1789 
1790 	memset(&t->encap, 0, sizeof(t->encap));
1791 
1792 	hlen = ip6_encap_hlen(ipencap);
1793 	if (hlen < 0)
1794 		return hlen;
1795 
1796 	t->encap.type = ipencap->type;
1797 	t->encap.sport = ipencap->sport;
1798 	t->encap.dport = ipencap->dport;
1799 	t->encap.flags = ipencap->flags;
1800 
1801 	t->encap_hlen = hlen;
1802 	t->hlen = t->encap_hlen + t->tun_hlen;
1803 
1804 	return 0;
1805 }
1806 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1807 
1808 static const struct net_device_ops ip6_tnl_netdev_ops = {
1809 	.ndo_init	= ip6_tnl_dev_init,
1810 	.ndo_uninit	= ip6_tnl_dev_uninit,
1811 	.ndo_start_xmit = ip6_tnl_start_xmit,
1812 	.ndo_siocdevprivate = ip6_tnl_siocdevprivate,
1813 	.ndo_change_mtu = ip6_tnl_change_mtu,
1814 	.ndo_get_stats64 = dev_get_tstats64,
1815 	.ndo_get_iflink = ip6_tnl_get_iflink,
1816 };
1817 
1818 #define IPXIPX_FEATURES (NETIF_F_SG |		\
1819 			 NETIF_F_FRAGLIST |	\
1820 			 NETIF_F_HIGHDMA |	\
1821 			 NETIF_F_GSO_SOFTWARE |	\
1822 			 NETIF_F_HW_CSUM)
1823 
1824 /**
1825  * ip6_tnl_dev_setup - setup virtual tunnel device
1826  *   @dev: virtual device associated with tunnel
1827  *
1828  * Description:
1829  *   Initialize function pointers and device parameters
1830  **/
1831 
ip6_tnl_dev_setup(struct net_device * dev)1832 static void ip6_tnl_dev_setup(struct net_device *dev)
1833 {
1834 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1835 	dev->header_ops = &ip_tunnel_header_ops;
1836 	dev->needs_free_netdev = true;
1837 	dev->priv_destructor = ip6_dev_free;
1838 
1839 	dev->type = ARPHRD_TUNNEL6;
1840 	dev->flags |= IFF_NOARP;
1841 	dev->addr_len = sizeof(struct in6_addr);
1842 	dev->features |= NETIF_F_LLTX;
1843 	netif_keep_dst(dev);
1844 
1845 	dev->features		|= IPXIPX_FEATURES;
1846 	dev->hw_features	|= IPXIPX_FEATURES;
1847 
1848 	/* This perm addr will be used as interface identifier by IPv6 */
1849 	dev->addr_assign_type = NET_ADDR_RANDOM;
1850 	eth_random_addr(dev->perm_addr);
1851 }
1852 
1853 
1854 /**
1855  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1856  *   @dev: virtual device associated with tunnel
1857  **/
1858 
1859 static inline int
ip6_tnl_dev_init_gen(struct net_device * dev)1860 ip6_tnl_dev_init_gen(struct net_device *dev)
1861 {
1862 	struct ip6_tnl *t = netdev_priv(dev);
1863 	int ret;
1864 	int t_hlen;
1865 
1866 	t->dev = dev;
1867 	t->net = dev_net(dev);
1868 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1869 	if (!dev->tstats)
1870 		return -ENOMEM;
1871 
1872 	ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1873 	if (ret)
1874 		goto free_stats;
1875 
1876 	ret = gro_cells_init(&t->gro_cells, dev);
1877 	if (ret)
1878 		goto destroy_dst;
1879 
1880 	t->tun_hlen = 0;
1881 	t->hlen = t->encap_hlen + t->tun_hlen;
1882 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1883 
1884 	dev->type = ARPHRD_TUNNEL6;
1885 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1886 	dev->mtu = ETH_DATA_LEN - t_hlen;
1887 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1888 		dev->mtu -= 8;
1889 	dev->min_mtu = ETH_MIN_MTU;
1890 	dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1891 
1892 	dev_hold_track(dev, &t->dev_tracker, GFP_KERNEL);
1893 	return 0;
1894 
1895 destroy_dst:
1896 	dst_cache_destroy(&t->dst_cache);
1897 free_stats:
1898 	free_percpu(dev->tstats);
1899 	dev->tstats = NULL;
1900 
1901 	return ret;
1902 }
1903 
1904 /**
1905  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1906  *   @dev: virtual device associated with tunnel
1907  **/
1908 
ip6_tnl_dev_init(struct net_device * dev)1909 static int ip6_tnl_dev_init(struct net_device *dev)
1910 {
1911 	struct ip6_tnl *t = netdev_priv(dev);
1912 	int err = ip6_tnl_dev_init_gen(dev);
1913 
1914 	if (err)
1915 		return err;
1916 	ip6_tnl_link_config(t);
1917 	if (t->parms.collect_md)
1918 		netif_keep_dst(dev);
1919 	return 0;
1920 }
1921 
1922 /**
1923  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1924  *   @dev: fallback device
1925  *
1926  * Return: 0
1927  **/
1928 
ip6_fb_tnl_dev_init(struct net_device * dev)1929 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1930 {
1931 	struct ip6_tnl *t = netdev_priv(dev);
1932 	struct net *net = dev_net(dev);
1933 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1934 
1935 	t->parms.proto = IPPROTO_IPV6;
1936 
1937 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1938 	return 0;
1939 }
1940 
ip6_tnl_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1941 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1942 			    struct netlink_ext_ack *extack)
1943 {
1944 	u8 proto;
1945 
1946 	if (!data || !data[IFLA_IPTUN_PROTO])
1947 		return 0;
1948 
1949 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1950 	if (proto != IPPROTO_IPV6 &&
1951 	    proto != IPPROTO_IPIP &&
1952 	    proto != 0)
1953 		return -EINVAL;
1954 
1955 	return 0;
1956 }
1957 
ip6_tnl_netlink_parms(struct nlattr * data[],struct __ip6_tnl_parm * parms)1958 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1959 				  struct __ip6_tnl_parm *parms)
1960 {
1961 	memset(parms, 0, sizeof(*parms));
1962 
1963 	if (!data)
1964 		return;
1965 
1966 	if (data[IFLA_IPTUN_LINK])
1967 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1968 
1969 	if (data[IFLA_IPTUN_LOCAL])
1970 		parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1971 
1972 	if (data[IFLA_IPTUN_REMOTE])
1973 		parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1974 
1975 	if (data[IFLA_IPTUN_TTL])
1976 		parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1977 
1978 	if (data[IFLA_IPTUN_ENCAP_LIMIT])
1979 		parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1980 
1981 	if (data[IFLA_IPTUN_FLOWINFO])
1982 		parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1983 
1984 	if (data[IFLA_IPTUN_FLAGS])
1985 		parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1986 
1987 	if (data[IFLA_IPTUN_PROTO])
1988 		parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1989 
1990 	if (data[IFLA_IPTUN_COLLECT_METADATA])
1991 		parms->collect_md = true;
1992 
1993 	if (data[IFLA_IPTUN_FWMARK])
1994 		parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1995 }
1996 
ip6_tnl_netlink_encap_parms(struct nlattr * data[],struct ip_tunnel_encap * ipencap)1997 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1998 					struct ip_tunnel_encap *ipencap)
1999 {
2000 	bool ret = false;
2001 
2002 	memset(ipencap, 0, sizeof(*ipencap));
2003 
2004 	if (!data)
2005 		return ret;
2006 
2007 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
2008 		ret = true;
2009 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
2010 	}
2011 
2012 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
2013 		ret = true;
2014 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
2015 	}
2016 
2017 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
2018 		ret = true;
2019 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
2020 	}
2021 
2022 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
2023 		ret = true;
2024 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
2025 	}
2026 
2027 	return ret;
2028 }
2029 
ip6_tnl_newlink(struct net * src_net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)2030 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
2031 			   struct nlattr *tb[], struct nlattr *data[],
2032 			   struct netlink_ext_ack *extack)
2033 {
2034 	struct net *net = dev_net(dev);
2035 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2036 	struct ip_tunnel_encap ipencap;
2037 	struct ip6_tnl *nt, *t;
2038 	int err;
2039 
2040 	nt = netdev_priv(dev);
2041 
2042 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2043 		err = ip6_tnl_encap_setup(nt, &ipencap);
2044 		if (err < 0)
2045 			return err;
2046 	}
2047 
2048 	ip6_tnl_netlink_parms(data, &nt->parms);
2049 
2050 	if (nt->parms.collect_md) {
2051 		if (rtnl_dereference(ip6n->collect_md_tun))
2052 			return -EEXIST;
2053 	} else {
2054 		t = ip6_tnl_locate(net, &nt->parms, 0);
2055 		if (!IS_ERR(t))
2056 			return -EEXIST;
2057 	}
2058 
2059 	err = ip6_tnl_create2(dev);
2060 	if (!err && tb[IFLA_MTU])
2061 		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2062 
2063 	return err;
2064 }
2065 
ip6_tnl_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)2066 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2067 			      struct nlattr *data[],
2068 			      struct netlink_ext_ack *extack)
2069 {
2070 	struct ip6_tnl *t = netdev_priv(dev);
2071 	struct __ip6_tnl_parm p;
2072 	struct net *net = t->net;
2073 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2074 	struct ip_tunnel_encap ipencap;
2075 
2076 	if (dev == ip6n->fb_tnl_dev)
2077 		return -EINVAL;
2078 
2079 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2080 		int err = ip6_tnl_encap_setup(t, &ipencap);
2081 
2082 		if (err < 0)
2083 			return err;
2084 	}
2085 	ip6_tnl_netlink_parms(data, &p);
2086 	if (p.collect_md)
2087 		return -EINVAL;
2088 
2089 	t = ip6_tnl_locate(net, &p, 0);
2090 	if (!IS_ERR(t)) {
2091 		if (t->dev != dev)
2092 			return -EEXIST;
2093 	} else
2094 		t = netdev_priv(dev);
2095 
2096 	return ip6_tnl_update(t, &p);
2097 }
2098 
ip6_tnl_dellink(struct net_device * dev,struct list_head * head)2099 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2100 {
2101 	struct net *net = dev_net(dev);
2102 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2103 
2104 	if (dev != ip6n->fb_tnl_dev)
2105 		unregister_netdevice_queue(dev, head);
2106 }
2107 
ip6_tnl_get_size(const struct net_device * dev)2108 static size_t ip6_tnl_get_size(const struct net_device *dev)
2109 {
2110 	return
2111 		/* IFLA_IPTUN_LINK */
2112 		nla_total_size(4) +
2113 		/* IFLA_IPTUN_LOCAL */
2114 		nla_total_size(sizeof(struct in6_addr)) +
2115 		/* IFLA_IPTUN_REMOTE */
2116 		nla_total_size(sizeof(struct in6_addr)) +
2117 		/* IFLA_IPTUN_TTL */
2118 		nla_total_size(1) +
2119 		/* IFLA_IPTUN_ENCAP_LIMIT */
2120 		nla_total_size(1) +
2121 		/* IFLA_IPTUN_FLOWINFO */
2122 		nla_total_size(4) +
2123 		/* IFLA_IPTUN_FLAGS */
2124 		nla_total_size(4) +
2125 		/* IFLA_IPTUN_PROTO */
2126 		nla_total_size(1) +
2127 		/* IFLA_IPTUN_ENCAP_TYPE */
2128 		nla_total_size(2) +
2129 		/* IFLA_IPTUN_ENCAP_FLAGS */
2130 		nla_total_size(2) +
2131 		/* IFLA_IPTUN_ENCAP_SPORT */
2132 		nla_total_size(2) +
2133 		/* IFLA_IPTUN_ENCAP_DPORT */
2134 		nla_total_size(2) +
2135 		/* IFLA_IPTUN_COLLECT_METADATA */
2136 		nla_total_size(0) +
2137 		/* IFLA_IPTUN_FWMARK */
2138 		nla_total_size(4) +
2139 		0;
2140 }
2141 
ip6_tnl_fill_info(struct sk_buff * skb,const struct net_device * dev)2142 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2143 {
2144 	struct ip6_tnl *tunnel = netdev_priv(dev);
2145 	struct __ip6_tnl_parm *parm = &tunnel->parms;
2146 
2147 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2148 	    nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2149 	    nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2150 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2151 	    nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2152 	    nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2153 	    nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2154 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2155 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2156 		goto nla_put_failure;
2157 
2158 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2159 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2160 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2161 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2162 		goto nla_put_failure;
2163 
2164 	if (parm->collect_md)
2165 		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2166 			goto nla_put_failure;
2167 
2168 	return 0;
2169 
2170 nla_put_failure:
2171 	return -EMSGSIZE;
2172 }
2173 
ip6_tnl_get_link_net(const struct net_device * dev)2174 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2175 {
2176 	struct ip6_tnl *tunnel = netdev_priv(dev);
2177 
2178 	return tunnel->net;
2179 }
2180 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2181 
2182 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2183 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
2184 	[IFLA_IPTUN_LOCAL]		= { .len = sizeof(struct in6_addr) },
2185 	[IFLA_IPTUN_REMOTE]		= { .len = sizeof(struct in6_addr) },
2186 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
2187 	[IFLA_IPTUN_ENCAP_LIMIT]	= { .type = NLA_U8 },
2188 	[IFLA_IPTUN_FLOWINFO]		= { .type = NLA_U32 },
2189 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U32 },
2190 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
2191 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
2192 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
2193 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
2194 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
2195 	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
2196 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
2197 };
2198 
2199 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2200 	.kind		= "ip6tnl",
2201 	.maxtype	= IFLA_IPTUN_MAX,
2202 	.policy		= ip6_tnl_policy,
2203 	.priv_size	= sizeof(struct ip6_tnl),
2204 	.setup		= ip6_tnl_dev_setup,
2205 	.validate	= ip6_tnl_validate,
2206 	.newlink	= ip6_tnl_newlink,
2207 	.changelink	= ip6_tnl_changelink,
2208 	.dellink	= ip6_tnl_dellink,
2209 	.get_size	= ip6_tnl_get_size,
2210 	.fill_info	= ip6_tnl_fill_info,
2211 	.get_link_net	= ip6_tnl_get_link_net,
2212 };
2213 
2214 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2215 	.handler	= ip4ip6_rcv,
2216 	.err_handler	= ip4ip6_err,
2217 	.priority	=	1,
2218 };
2219 
2220 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2221 	.handler	= ip6ip6_rcv,
2222 	.err_handler	= ip6ip6_err,
2223 	.priority	=	1,
2224 };
2225 
2226 static struct xfrm6_tunnel mplsip6_handler __read_mostly = {
2227 	.handler	= mplsip6_rcv,
2228 	.err_handler	= mplsip6_err,
2229 	.priority	=	1,
2230 };
2231 
ip6_tnl_destroy_tunnels(struct net * net,struct list_head * list)2232 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2233 {
2234 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2235 	struct net_device *dev, *aux;
2236 	int h;
2237 	struct ip6_tnl *t;
2238 
2239 	for_each_netdev_safe(net, dev, aux)
2240 		if (dev->rtnl_link_ops == &ip6_link_ops)
2241 			unregister_netdevice_queue(dev, list);
2242 
2243 	for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2244 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
2245 		while (t) {
2246 			/* If dev is in the same netns, it has already
2247 			 * been added to the list by the previous loop.
2248 			 */
2249 			if (!net_eq(dev_net(t->dev), net))
2250 				unregister_netdevice_queue(t->dev, list);
2251 			t = rtnl_dereference(t->next);
2252 		}
2253 	}
2254 
2255 	t = rtnl_dereference(ip6n->tnls_wc[0]);
2256 	while (t) {
2257 		/* If dev is in the same netns, it has already
2258 		 * been added to the list by the previous loop.
2259 		 */
2260 		if (!net_eq(dev_net(t->dev), net))
2261 			unregister_netdevice_queue(t->dev, list);
2262 		t = rtnl_dereference(t->next);
2263 	}
2264 }
2265 
ip6_tnl_init_net(struct net * net)2266 static int __net_init ip6_tnl_init_net(struct net *net)
2267 {
2268 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2269 	struct ip6_tnl *t = NULL;
2270 	int err;
2271 
2272 	ip6n->tnls[0] = ip6n->tnls_wc;
2273 	ip6n->tnls[1] = ip6n->tnls_r_l;
2274 
2275 	if (!net_has_fallback_tunnels(net))
2276 		return 0;
2277 	err = -ENOMEM;
2278 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2279 					NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2280 
2281 	if (!ip6n->fb_tnl_dev)
2282 		goto err_alloc_dev;
2283 	dev_net_set(ip6n->fb_tnl_dev, net);
2284 	ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2285 	/* FB netdevice is special: we have one, and only one per netns.
2286 	 * Allowing to move it to another netns is clearly unsafe.
2287 	 */
2288 	ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2289 
2290 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2291 	if (err < 0)
2292 		goto err_register;
2293 
2294 	err = register_netdev(ip6n->fb_tnl_dev);
2295 	if (err < 0)
2296 		goto err_register;
2297 
2298 	t = netdev_priv(ip6n->fb_tnl_dev);
2299 
2300 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2301 	return 0;
2302 
2303 err_register:
2304 	free_netdev(ip6n->fb_tnl_dev);
2305 err_alloc_dev:
2306 	return err;
2307 }
2308 
ip6_tnl_exit_batch_net(struct list_head * net_list)2309 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2310 {
2311 	struct net *net;
2312 	LIST_HEAD(list);
2313 
2314 	rtnl_lock();
2315 	list_for_each_entry(net, net_list, exit_list)
2316 		ip6_tnl_destroy_tunnels(net, &list);
2317 	unregister_netdevice_many(&list);
2318 	rtnl_unlock();
2319 }
2320 
2321 static struct pernet_operations ip6_tnl_net_ops = {
2322 	.init = ip6_tnl_init_net,
2323 	.exit_batch = ip6_tnl_exit_batch_net,
2324 	.id   = &ip6_tnl_net_id,
2325 	.size = sizeof(struct ip6_tnl_net),
2326 };
2327 
2328 /**
2329  * ip6_tunnel_init - register protocol and reserve needed resources
2330  *
2331  * Return: 0 on success
2332  **/
2333 
ip6_tunnel_init(void)2334 static int __init ip6_tunnel_init(void)
2335 {
2336 	int  err;
2337 
2338 	if (!ipv6_mod_enabled())
2339 		return -EOPNOTSUPP;
2340 
2341 	err = register_pernet_device(&ip6_tnl_net_ops);
2342 	if (err < 0)
2343 		goto out_pernet;
2344 
2345 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2346 	if (err < 0) {
2347 		pr_err("%s: can't register ip4ip6\n", __func__);
2348 		goto out_ip4ip6;
2349 	}
2350 
2351 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2352 	if (err < 0) {
2353 		pr_err("%s: can't register ip6ip6\n", __func__);
2354 		goto out_ip6ip6;
2355 	}
2356 
2357 	if (ip6_tnl_mpls_supported()) {
2358 		err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS);
2359 		if (err < 0) {
2360 			pr_err("%s: can't register mplsip6\n", __func__);
2361 			goto out_mplsip6;
2362 		}
2363 	}
2364 
2365 	err = rtnl_link_register(&ip6_link_ops);
2366 	if (err < 0)
2367 		goto rtnl_link_failed;
2368 
2369 	return 0;
2370 
2371 rtnl_link_failed:
2372 	if (ip6_tnl_mpls_supported())
2373 		xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS);
2374 out_mplsip6:
2375 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2376 out_ip6ip6:
2377 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2378 out_ip4ip6:
2379 	unregister_pernet_device(&ip6_tnl_net_ops);
2380 out_pernet:
2381 	return err;
2382 }
2383 
2384 /**
2385  * ip6_tunnel_cleanup - free resources and unregister protocol
2386  **/
2387 
ip6_tunnel_cleanup(void)2388 static void __exit ip6_tunnel_cleanup(void)
2389 {
2390 	rtnl_link_unregister(&ip6_link_ops);
2391 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2392 		pr_info("%s: can't deregister ip4ip6\n", __func__);
2393 
2394 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2395 		pr_info("%s: can't deregister ip6ip6\n", __func__);
2396 
2397 	if (ip6_tnl_mpls_supported() &&
2398 	    xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS))
2399 		pr_info("%s: can't deregister mplsip6\n", __func__);
2400 	unregister_pernet_device(&ip6_tnl_net_ops);
2401 }
2402 
2403 module_init(ip6_tunnel_init);
2404 module_exit(ip6_tunnel_cleanup);
2405