xref: /linux/net/ipv6/ip6_tunnel.c (revision a5210135489ae7bc1ef1cb4a8157361dd7b468cd)
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/netdev_lock.h>
56 #include <net/dst_metadata.h>
57 #include <net/inet_dscp.h>
58 
59 MODULE_AUTHOR("Ville Nuorvala");
60 MODULE_DESCRIPTION("IPv6 tunneling device");
61 MODULE_LICENSE("GPL");
62 MODULE_ALIAS_RTNL_LINK("ip6tnl");
63 MODULE_ALIAS_NETDEV("ip6tnl0");
64 
65 #define IP6_TUNNEL_MAX_DEST_TLVS    8
66 
67 #define IP6_TUNNEL_HASH_SIZE_SHIFT  5
68 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
69 
70 static bool log_ecn_error = true;
71 module_param(log_ecn_error, bool, 0644);
72 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
73 
HASH(const struct in6_addr * addr1,const struct in6_addr * addr2)74 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
75 {
76 	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
77 
78 	return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
79 }
80 
81 static int ip6_tnl_dev_init(struct net_device *dev);
82 static void ip6_tnl_dev_setup(struct net_device *dev);
83 static struct rtnl_link_ops ip6_link_ops __read_mostly;
84 
85 static unsigned int ip6_tnl_net_id __read_mostly;
86 struct ip6_tnl_net {
87 	/* the IPv6 tunnel fallback device */
88 	struct net_device *fb_tnl_dev;
89 	/* lists for storing tunnels in use */
90 	struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
91 	struct ip6_tnl __rcu *tnls_wc[1];
92 	struct ip6_tnl __rcu **tnls[2];
93 	struct ip6_tnl __rcu *collect_md_tun;
94 };
95 
ip6_tnl_mpls_supported(void)96 static inline int ip6_tnl_mpls_supported(void)
97 {
98 	return IS_ENABLED(CONFIG_MPLS);
99 }
100 
101 /**
102  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
103  *   @net: network namespace
104  *   @link: ifindex of underlying interface
105  *   @remote: the address of the tunnel exit-point
106  *   @local: the address of the tunnel entry-point
107  *
108  * Return:
109  *   tunnel matching given end-points if found,
110  *   else fallback tunnel if its device is up,
111  *   else %NULL
112  **/
113 
114 static struct ip6_tnl *
ip6_tnl_lookup(struct net * net,int link,const struct in6_addr * remote,const struct in6_addr * local)115 ip6_tnl_lookup(struct net *net, int link,
116 	       const struct in6_addr *remote, const struct in6_addr *local)
117 {
118 	unsigned int hash = HASH(remote, local);
119 	struct ip6_tnl *t, *cand = NULL;
120 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
121 	struct in6_addr any;
122 
123 	for_each_ip_tunnel_rcu(t, ip6n->tnls_r_l[hash]) {
124 		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
125 		    !ipv6_addr_equal(remote, &t->parms.raddr) ||
126 		    !(t->dev->flags & IFF_UP))
127 			continue;
128 
129 		if (link == t->parms.link)
130 			return t;
131 		else
132 			cand = t;
133 	}
134 
135 	memset(&any, 0, sizeof(any));
136 	hash = HASH(&any, local);
137 	for_each_ip_tunnel_rcu(t, ip6n->tnls_r_l[hash]) {
138 		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
139 		    !ipv6_addr_any(&t->parms.raddr) ||
140 		    !(t->dev->flags & IFF_UP))
141 			continue;
142 
143 		if (link == t->parms.link)
144 			return t;
145 		else if (!cand)
146 			cand = t;
147 	}
148 
149 	hash = HASH(remote, &any);
150 	for_each_ip_tunnel_rcu(t, ip6n->tnls_r_l[hash]) {
151 		if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
152 		    !ipv6_addr_any(&t->parms.laddr) ||
153 		    !(t->dev->flags & IFF_UP))
154 			continue;
155 
156 		if (link == t->parms.link)
157 			return t;
158 		else if (!cand)
159 			cand = t;
160 	}
161 
162 	if (cand)
163 		return cand;
164 
165 	t = rcu_dereference(ip6n->collect_md_tun);
166 	if (t && t->dev->flags & IFF_UP)
167 		return t;
168 
169 	t = rcu_dereference(ip6n->tnls_wc[0]);
170 	if (t && (t->dev->flags & IFF_UP))
171 		return t;
172 
173 	return NULL;
174 }
175 
176 /**
177  * ip6_tnl_bucket - get head of list matching given tunnel parameters
178  *   @ip6n: the private data for ip6_vti in the netns
179  *   @p: parameters containing tunnel end-points
180  *
181  * Description:
182  *   ip6_tnl_bucket() returns the head of the list matching the
183  *   &struct in6_addr entries laddr and raddr in @p.
184  *
185  * Return: head of IPv6 tunnel list
186  **/
187 
188 static struct ip6_tnl __rcu **
ip6_tnl_bucket(struct ip6_tnl_net * ip6n,const struct __ip6_tnl_parm * p)189 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
190 {
191 	const struct in6_addr *remote = &p->raddr;
192 	const struct in6_addr *local = &p->laddr;
193 	unsigned int h = 0;
194 	int prio = 0;
195 
196 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
197 		prio = 1;
198 		h = HASH(remote, local);
199 	}
200 	return &ip6n->tnls[prio][h];
201 }
202 
203 /**
204  * ip6_tnl_link - add tunnel to hash table
205  *   @ip6n: the private data for ip6_vti in the netns
206  *   @t: tunnel to be added
207  **/
208 
209 static void
ip6_tnl_link(struct ip6_tnl_net * ip6n,struct ip6_tnl * t)210 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
211 {
212 	struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
213 
214 	if (t->parms.collect_md)
215 		rcu_assign_pointer(ip6n->collect_md_tun, t);
216 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
217 	rcu_assign_pointer(*tp, t);
218 }
219 
220 /**
221  * ip6_tnl_unlink - remove tunnel from hash table
222  *   @ip6n: the private data for ip6_vti in the netns
223  *   @t: tunnel to be removed
224  **/
225 
226 static void
ip6_tnl_unlink(struct ip6_tnl_net * ip6n,struct ip6_tnl * t)227 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
228 {
229 	struct ip6_tnl __rcu **tp;
230 	struct ip6_tnl *iter;
231 
232 	if (t->parms.collect_md)
233 		rcu_assign_pointer(ip6n->collect_md_tun, NULL);
234 
235 	for (tp = ip6_tnl_bucket(ip6n, &t->parms);
236 	     (iter = rtnl_dereference(*tp)) != NULL;
237 	     tp = &iter->next) {
238 		if (t == iter) {
239 			rcu_assign_pointer(*tp, t->next);
240 			break;
241 		}
242 	}
243 }
244 
ip6_dev_free(struct net_device * dev)245 static void ip6_dev_free(struct net_device *dev)
246 {
247 	struct ip6_tnl *t = netdev_priv(dev);
248 
249 	gro_cells_destroy(&t->gro_cells);
250 	dst_cache_destroy(&t->dst_cache);
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 ip6_tnl_net *ip6n = net_generic(t->net, ip6_tnl_net_id);
257 	int err;
258 
259 	dev->rtnl_link_ops = &ip6_link_ops;
260 	err = register_netdevice(dev);
261 	if (err < 0)
262 		goto out;
263 
264 	strcpy(t->parms.name, dev->name);
265 
266 	ip6_tnl_link(ip6n, t);
267 	return 0;
268 
269 out:
270 	return err;
271 }
272 
273 /**
274  * ip6_tnl_create - create a new tunnel
275  *   @net: network namespace
276  *   @p: tunnel parameters
277  *
278  * Description:
279  *   Create tunnel matching given parameters.
280  *
281  * Return:
282  *   created tunnel or error pointer
283  **/
284 
ip6_tnl_create(struct net * net,struct __ip6_tnl_parm * p)285 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
286 {
287 	struct net_device *dev;
288 	struct ip6_tnl *t;
289 	char name[IFNAMSIZ];
290 	int err = -E2BIG;
291 
292 	if (p->name[0]) {
293 		if (!dev_valid_name(p->name))
294 			goto failed;
295 		strscpy(name, p->name, IFNAMSIZ);
296 	} else {
297 		sprintf(name, "ip6tnl%%d");
298 	}
299 	err = -ENOMEM;
300 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
301 			   ip6_tnl_dev_setup);
302 	if (!dev)
303 		goto failed;
304 
305 	dev_net_set(dev, net);
306 
307 	t = netdev_priv(dev);
308 	t->parms = *p;
309 	t->net = dev_net(dev);
310 	err = ip6_tnl_create2(dev);
311 	if (err < 0)
312 		goto failed_free;
313 
314 	return t;
315 
316 failed_free:
317 	free_netdev(dev);
318 failed:
319 	return ERR_PTR(err);
320 }
321 
322 /**
323  * ip6_tnl_locate - find or create tunnel matching given parameters
324  *   @net: network namespace
325  *   @p: tunnel parameters
326  *   @create: != 0 if allowed to create new tunnel if no match found
327  *
328  * Description:
329  *   ip6_tnl_locate() first tries to locate an existing tunnel
330  *   based on @parms. If this is unsuccessful, but @create is set a new
331  *   tunnel device is created and registered for use.
332  *
333  * Return:
334  *   matching tunnel or error pointer
335  **/
336 
ip6_tnl_locate(struct net * net,struct __ip6_tnl_parm * p,int create)337 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
338 		struct __ip6_tnl_parm *p, int create)
339 {
340 	const struct in6_addr *remote = &p->raddr;
341 	const struct in6_addr *local = &p->laddr;
342 	struct ip6_tnl __rcu **tp;
343 	struct ip6_tnl *t;
344 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
345 
346 	for (tp = ip6_tnl_bucket(ip6n, p);
347 	     (t = rtnl_dereference(*tp)) != NULL;
348 	     tp = &t->next) {
349 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
350 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
351 		    p->link == t->parms.link) {
352 			if (create)
353 				return ERR_PTR(-EEXIST);
354 
355 			return t;
356 		}
357 	}
358 	if (!create)
359 		return ERR_PTR(-ENODEV);
360 	return ip6_tnl_create(net, p);
361 }
362 
363 /**
364  * ip6_tnl_dev_uninit - tunnel device uninitializer
365  *   @dev: the device to be destroyed
366  *
367  * Description:
368  *   ip6_tnl_dev_uninit() removes tunnel from its list
369  **/
370 
371 static void
ip6_tnl_dev_uninit(struct net_device * dev)372 ip6_tnl_dev_uninit(struct net_device *dev)
373 {
374 	struct ip6_tnl *t = netdev_priv(dev);
375 	struct net *net = t->net;
376 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
377 
378 	if (dev == ip6n->fb_tnl_dev)
379 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
380 	else
381 		ip6_tnl_unlink(ip6n, t);
382 	dst_cache_reset(&t->dst_cache);
383 	netdev_put(dev, &t->dev_tracker);
384 }
385 
386 /**
387  * ip6_tnl_parse_tlv_enc_lim - handle encapsulation limit option
388  *   @skb: received socket buffer
389  *   @raw: the ICMPv6 error message data
390  *
391  * Return:
392  *   0 if none was found,
393  *   else index to encapsulation limit
394  **/
395 
ip6_tnl_parse_tlv_enc_lim(struct sk_buff * skb,__u8 * raw)396 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
397 {
398 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
399 	unsigned int nhoff = raw - skb->data;
400 	unsigned int off = nhoff + sizeof(*ipv6h);
401 	u8 nexthdr = ipv6h->nexthdr;
402 
403 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
404 		struct ipv6_opt_hdr *hdr;
405 		u16 optlen;
406 
407 		if (!pskb_may_pull(skb, off + sizeof(*hdr)))
408 			break;
409 
410 		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
411 		if (nexthdr == NEXTHDR_FRAGMENT) {
412 			optlen = 8;
413 		} else if (nexthdr == NEXTHDR_AUTH) {
414 			optlen = ipv6_authlen(hdr);
415 		} else {
416 			optlen = ipv6_optlen(hdr);
417 		}
418 
419 		if (!pskb_may_pull(skb, off + optlen))
420 			break;
421 
422 		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
423 		if (nexthdr == NEXTHDR_FRAGMENT) {
424 			struct frag_hdr *frag_hdr = (struct frag_hdr *)hdr;
425 
426 			if (frag_hdr->frag_off)
427 				break;
428 		}
429 		if (nexthdr == NEXTHDR_DEST) {
430 			int tlv_cnt = 0;
431 			u16 i = 2;
432 
433 			while (1) {
434 				struct ipv6_tlv_tnl_enc_lim *tel;
435 
436 				if (unlikely(tlv_cnt++ >= IP6_TUNNEL_MAX_DEST_TLVS))
437 					break;
438 
439 				/* No more room for encapsulation limit */
440 				if (i + sizeof(*tel) > optlen)
441 					break;
442 
443 				tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
444 				/* return index of option if found and valid */
445 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
446 				    tel->length == 1)
447 					return i + off - nhoff;
448 				/* else jump to next option */
449 				if (tel->type)
450 					i += tel->length + 2;
451 				else
452 					i++;
453 			}
454 		}
455 		nexthdr = hdr->nexthdr;
456 		off += optlen;
457 	}
458 	return 0;
459 }
460 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
461 
462 /* ip6_tnl_err() should handle errors in the tunnel according to the
463  * specifications in RFC 2473.
464  */
465 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)466 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
467 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
468 {
469 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
470 	struct net *net = dev_net(skb->dev);
471 	u8 rel_type = ICMPV6_DEST_UNREACH;
472 	u8 rel_code = ICMPV6_ADDR_UNREACH;
473 	__u32 rel_info = 0;
474 	struct ip6_tnl *t;
475 	int err = -ENOENT;
476 	int rel_msg = 0;
477 	u8 tproto;
478 	__u16 len;
479 
480 	/* If the packet doesn't contain the original IPv6 header we are
481 	   in trouble since we might need the source address for further
482 	   processing of the error. */
483 
484 	rcu_read_lock();
485 	t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->daddr, &ipv6h->saddr);
486 	if (!t)
487 		goto out;
488 
489 	tproto = READ_ONCE(t->parms.proto);
490 	if (tproto != ipproto && tproto != 0)
491 		goto out;
492 
493 	err = 0;
494 
495 	switch (*type) {
496 	case ICMPV6_DEST_UNREACH:
497 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
498 				    t->parms.name);
499 		rel_msg = 1;
500 		break;
501 	case ICMPV6_TIME_EXCEED:
502 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
503 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
504 					    t->parms.name);
505 			rel_msg = 1;
506 		}
507 		break;
508 	case ICMPV6_PARAMPROB: {
509 		struct ipv6_tlv_tnl_enc_lim *tel;
510 		__u32 teli;
511 
512 		teli = 0;
513 		if ((*code) == ICMPV6_HDR_FIELD)
514 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
515 
516 		if (teli && teli == *info - 2) {
517 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
518 			if (tel->encap_limit == 0) {
519 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
520 						    t->parms.name);
521 				rel_msg = 1;
522 			}
523 		} else {
524 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
525 					    t->parms.name);
526 		}
527 		break;
528 	}
529 	case ICMPV6_PKT_TOOBIG: {
530 		__u32 mtu;
531 
532 		ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
533 				sock_net_uid(net, NULL));
534 		mtu = *info - offset;
535 		if (mtu < IPV6_MIN_MTU)
536 			mtu = IPV6_MIN_MTU;
537 		len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
538 		if (len > mtu) {
539 			rel_type = ICMPV6_PKT_TOOBIG;
540 			rel_code = 0;
541 			rel_info = mtu;
542 			rel_msg = 1;
543 		}
544 		break;
545 	}
546 	case NDISC_REDIRECT:
547 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
548 			     sock_net_uid(net, NULL));
549 		break;
550 	}
551 
552 	*type = rel_type;
553 	*code = rel_code;
554 	*info = rel_info;
555 	*msg = rel_msg;
556 
557 out:
558 	rcu_read_unlock();
559 	return err;
560 }
561 
562 static int
ip4ip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)563 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
564 	   u8 type, u8 code, int offset, __be32 info)
565 {
566 	__u32 rel_info = ntohl(info);
567 	const struct iphdr *eiph;
568 	struct sk_buff *skb2;
569 	int err, rel_msg = 0;
570 	u8 rel_type = type;
571 	u8 rel_code = code;
572 	struct rtable *rt;
573 	struct flowi4 fl4;
574 
575 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
576 			  &rel_msg, &rel_info, offset);
577 	if (err < 0)
578 		return err;
579 
580 	if (rel_msg == 0)
581 		return 0;
582 
583 	switch (rel_type) {
584 	case ICMPV6_DEST_UNREACH:
585 		if (rel_code != ICMPV6_ADDR_UNREACH)
586 			return 0;
587 		rel_type = ICMP_DEST_UNREACH;
588 		rel_code = ICMP_HOST_UNREACH;
589 		break;
590 	case ICMPV6_PKT_TOOBIG:
591 		if (rel_code != 0)
592 			return 0;
593 		rel_type = ICMP_DEST_UNREACH;
594 		rel_code = ICMP_FRAG_NEEDED;
595 		break;
596 	default:
597 		return 0;
598 	}
599 
600 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
601 		return 0;
602 
603 	skb2 = skb_clone(skb, GFP_ATOMIC);
604 	if (!skb2)
605 		return 0;
606 
607 	/* Remove debris left by IPv6 stack. */
608 	memset(IPCB(skb2), 0, sizeof(*IPCB(skb2)));
609 
610 	skb_dst_drop(skb2);
611 
612 	skb_pull(skb2, offset);
613 	skb_reset_network_header(skb2);
614 	eiph = ip_hdr(skb2);
615 	if (eiph->version != 4 || eiph->ihl < 5)
616 		goto out;
617 
618 	/* Try to guess incoming interface */
619 	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
620 				   0, 0, 0, IPPROTO_IPIP,
621 				   eiph->tos & INET_DSCP_MASK, 0);
622 	if (IS_ERR(rt))
623 		goto out;
624 
625 	skb2->dev = rt->dst.dev;
626 	ip_rt_put(rt);
627 
628 	/* route "incoming" packet */
629 	if (rt->rt_flags & RTCF_LOCAL) {
630 		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
631 					   eiph->daddr, eiph->saddr, 0, 0,
632 					   IPPROTO_IPIP,
633 					   eiph->tos & INET_DSCP_MASK, 0);
634 		if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) {
635 			if (!IS_ERR(rt))
636 				ip_rt_put(rt);
637 			goto out;
638 		}
639 		skb_dst_set(skb2, &rt->dst);
640 	} else {
641 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr,
642 				   ip4h_dscp(eiph), skb2->dev) ||
643 		    skb_dst_dev(skb2)->type != ARPHRD_TUNNEL6)
644 			goto out;
645 	}
646 
647 	/* change mtu on this route */
648 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
649 		if (rel_info > dst6_mtu(skb_dst(skb2)))
650 			goto out;
651 
652 		skb_dst_update_pmtu_no_confirm(skb2, rel_info);
653 	}
654 
655 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
656 
657 out:
658 	kfree_skb(skb2);
659 	return 0;
660 }
661 
662 static int
ip6ip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)663 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
664 	   u8 type, u8 code, int offset, __be32 info)
665 {
666 	__u32 rel_info = ntohl(info);
667 	int err, rel_msg = 0;
668 	u8 rel_type = type;
669 	u8 rel_code = code;
670 
671 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
672 			  &rel_msg, &rel_info, offset);
673 	if (err < 0)
674 		return err;
675 
676 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
677 		struct rt6_info *rt;
678 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
679 
680 		if (!skb2)
681 			return 0;
682 
683 		skb_dst_drop(skb2);
684 		skb_pull(skb2, offset);
685 		skb_reset_network_header(skb2);
686 
687 		/* Try to guess incoming interface */
688 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
689 				NULL, 0, skb2, 0);
690 
691 		if (rt && rt->dst.dev)
692 			skb2->dev = rt->dst.dev;
693 
694 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
695 
696 		ip6_rt_put(rt);
697 
698 		kfree_skb(skb2);
699 	}
700 
701 	return 0;
702 }
703 
704 static int
mplsip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)705 mplsip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
706 	    u8 type, u8 code, int offset, __be32 info)
707 {
708 	__u32 rel_info = ntohl(info);
709 	int err, rel_msg = 0;
710 	u8 rel_type = type;
711 	u8 rel_code = code;
712 
713 	err = ip6_tnl_err(skb, IPPROTO_MPLS, opt, &rel_type, &rel_code,
714 			  &rel_msg, &rel_info, offset);
715 	return err;
716 }
717 
ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)718 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
719 				       const struct ipv6hdr *ipv6h,
720 				       struct sk_buff *skb)
721 {
722 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
723 
724 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
725 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
726 
727 	return IP6_ECN_decapsulate(ipv6h, skb);
728 }
729 
ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)730 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
731 				       const struct ipv6hdr *ipv6h,
732 				       struct sk_buff *skb)
733 {
734 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
735 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
736 
737 	return IP6_ECN_decapsulate(ipv6h, skb);
738 }
739 
mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)740 static inline int mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
741 					       const struct ipv6hdr *ipv6h,
742 					       struct sk_buff *skb)
743 {
744 	/* ECN is not supported in AF_MPLS */
745 	return 0;
746 }
747 
ip6_tnl_get_cap(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)748 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
749 			     const struct in6_addr *laddr,
750 			     const struct in6_addr *raddr)
751 {
752 	struct __ip6_tnl_parm *p = &t->parms;
753 	int ltype = ipv6_addr_type(laddr);
754 	int rtype = ipv6_addr_type(raddr);
755 	__u32 flags = 0;
756 
757 	if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
758 		flags = IP6_TNL_F_CAP_PER_PACKET;
759 	} else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
760 		   rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
761 		   !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
762 		   (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
763 		if (ltype&IPV6_ADDR_UNICAST)
764 			flags |= IP6_TNL_F_CAP_XMIT;
765 		if (rtype&IPV6_ADDR_UNICAST)
766 			flags |= IP6_TNL_F_CAP_RCV;
767 	}
768 	return flags;
769 }
770 EXPORT_SYMBOL(ip6_tnl_get_cap);
771 
772 /* called with rcu_read_lock() */
ip6_tnl_rcv_ctl(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)773 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
774 				  const struct in6_addr *laddr,
775 				  const struct in6_addr *raddr)
776 {
777 	struct __ip6_tnl_parm *p = &t->parms;
778 	int ret = 0;
779 	struct net *net = t->net;
780 
781 	if ((p->flags & IP6_TNL_F_CAP_RCV) ||
782 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
783 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
784 		struct net_device *ldev = NULL;
785 
786 		if (p->link)
787 			ldev = dev_get_by_index_rcu(net, p->link);
788 
789 		if ((ipv6_addr_is_multicast(laddr) ||
790 		     likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
791 						    0, IFA_F_TENTATIVE))) &&
792 		    ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
793 		     likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
794 						     0, IFA_F_TENTATIVE))))
795 			ret = 1;
796 	}
797 	return ret;
798 }
799 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
800 
__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)801 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
802 			 const struct tnl_ptk_info *tpi,
803 			 struct metadata_dst *tun_dst,
804 			 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
805 						const struct ipv6hdr *ipv6h,
806 						struct sk_buff *skb),
807 			 bool log_ecn_err)
808 {
809 	const struct ipv6hdr *ipv6h;
810 	int nh, err;
811 
812 	if (test_bit(IP_TUNNEL_CSUM_BIT, tunnel->parms.i_flags) !=
813 	    test_bit(IP_TUNNEL_CSUM_BIT, tpi->flags)) {
814 		DEV_STATS_INC(tunnel->dev, rx_crc_errors);
815 		DEV_STATS_INC(tunnel->dev, rx_errors);
816 		goto drop;
817 	}
818 
819 	if (test_bit(IP_TUNNEL_SEQ_BIT, tunnel->parms.i_flags)) {
820 		if (!test_bit(IP_TUNNEL_SEQ_BIT, tpi->flags) ||
821 		    (tunnel->i_seqno &&
822 		     (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
823 			DEV_STATS_INC(tunnel->dev, rx_fifo_errors);
824 			DEV_STATS_INC(tunnel->dev, rx_errors);
825 			goto drop;
826 		}
827 		tunnel->i_seqno = ntohl(tpi->seq) + 1;
828 	}
829 
830 	skb->protocol = tpi->proto;
831 
832 	/* Warning: All skb pointers will be invalidated! */
833 	if (tunnel->dev->type == ARPHRD_ETHER) {
834 		if (!pskb_may_pull(skb, ETH_HLEN)) {
835 			DEV_STATS_INC(tunnel->dev, rx_length_errors);
836 			DEV_STATS_INC(tunnel->dev, rx_errors);
837 			goto drop;
838 		}
839 
840 		skb->protocol = eth_type_trans(skb, tunnel->dev);
841 		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
842 	} else {
843 		skb->dev = tunnel->dev;
844 		skb_reset_mac_header(skb);
845 	}
846 
847 	/* Save offset of outer header relative to skb->head,
848 	 * because we are going to reset the network header to the inner header
849 	 * and might change skb->head.
850 	 */
851 	nh = skb_network_header(skb) - skb->head;
852 
853 	skb_reset_network_header(skb);
854 
855 	if (skb_vlan_inet_prepare(skb, true)) {
856 		DEV_STATS_INC(tunnel->dev, rx_length_errors);
857 		DEV_STATS_INC(tunnel->dev, rx_errors);
858 		goto drop;
859 	}
860 
861 	/* Get the outer header. */
862 	ipv6h = (struct ipv6hdr *)(skb->head + nh);
863 
864 	memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
865 
866 	__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
867 
868 	err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
869 	if (unlikely(err)) {
870 		if (log_ecn_err)
871 			net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
872 					     &ipv6h->saddr,
873 					     ipv6_get_dsfield(ipv6h));
874 		if (err > 1) {
875 			DEV_STATS_INC(tunnel->dev, rx_frame_errors);
876 			DEV_STATS_INC(tunnel->dev, rx_errors);
877 			goto drop;
878 		}
879 	}
880 
881 	dev_sw_netstats_rx_add(tunnel->dev, skb->len);
882 
883 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
884 
885 	if (tun_dst)
886 		skb_dst_set(skb, (struct dst_entry *)tun_dst);
887 
888 	gro_cells_receive(&tunnel->gro_cells, skb);
889 	return 0;
890 
891 drop:
892 	if (tun_dst)
893 		dst_release((struct dst_entry *)tun_dst);
894 	kfree_skb(skb);
895 	return 0;
896 }
897 
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)898 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
899 		const struct tnl_ptk_info *tpi,
900 		struct metadata_dst *tun_dst,
901 		bool log_ecn_err)
902 {
903 	int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
904 				    const struct ipv6hdr *ipv6h,
905 				    struct sk_buff *skb);
906 
907 	dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
908 	if (tpi->proto == htons(ETH_P_IP))
909 		dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
910 
911 	return __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
912 			     log_ecn_err);
913 }
914 EXPORT_SYMBOL(ip6_tnl_rcv);
915 
916 static const struct tnl_ptk_info tpi_v6 = {
917 	/* no tunnel info required for ipxip6. */
918 	.proto = htons(ETH_P_IPV6),
919 };
920 
921 static const struct tnl_ptk_info tpi_v4 = {
922 	/* no tunnel info required for ipxip6. */
923 	.proto = htons(ETH_P_IP),
924 };
925 
926 static const struct tnl_ptk_info tpi_mpls = {
927 	/* no tunnel info required for mplsip6. */
928 	.proto = htons(ETH_P_MPLS_UC),
929 };
930 
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))931 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
932 		      const struct tnl_ptk_info *tpi,
933 		      int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
934 						  const struct ipv6hdr *ipv6h,
935 						  struct sk_buff *skb))
936 {
937 	struct ip6_tnl *t;
938 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
939 	struct metadata_dst *tun_dst = NULL;
940 	int ret = -1;
941 
942 	rcu_read_lock();
943 	t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr);
944 
945 	if (t) {
946 		u8 tproto = READ_ONCE(t->parms.proto);
947 
948 		if (tproto != ipproto && tproto != 0)
949 			goto drop;
950 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
951 			goto drop;
952 		ipv6h = ipv6_hdr(skb);
953 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
954 			goto drop;
955 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
956 			goto drop;
957 		if (t->parms.collect_md) {
958 			IP_TUNNEL_DECLARE_FLAGS(flags) = { };
959 
960 			tun_dst = ipv6_tun_rx_dst(skb, flags, 0, 0);
961 			if (!tun_dst)
962 				goto drop;
963 		}
964 		ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
965 				    log_ecn_error);
966 	}
967 
968 	rcu_read_unlock();
969 
970 	return ret;
971 
972 drop:
973 	rcu_read_unlock();
974 	kfree_skb(skb);
975 	return 0;
976 }
977 
ip4ip6_rcv(struct sk_buff * skb)978 static int ip4ip6_rcv(struct sk_buff *skb)
979 {
980 	return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
981 			  ip4ip6_dscp_ecn_decapsulate);
982 }
983 
ip6ip6_rcv(struct sk_buff * skb)984 static int ip6ip6_rcv(struct sk_buff *skb)
985 {
986 	return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
987 			  ip6ip6_dscp_ecn_decapsulate);
988 }
989 
mplsip6_rcv(struct sk_buff * skb)990 static int mplsip6_rcv(struct sk_buff *skb)
991 {
992 	return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls,
993 			  mplsip6_dscp_ecn_decapsulate);
994 }
995 
996 struct ipv6_tel_txoption {
997 	struct ipv6_txoptions ops;
998 	__u8 dst_opt[8];
999 };
1000 
init_tel_txopt(struct ipv6_tel_txoption * opt,__u8 encap_limit)1001 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
1002 {
1003 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
1004 
1005 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
1006 	opt->dst_opt[3] = 1;
1007 	opt->dst_opt[4] = encap_limit;
1008 	opt->dst_opt[5] = IPV6_TLV_PADN;
1009 	opt->dst_opt[6] = 1;
1010 
1011 	opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
1012 	opt->ops.opt_nflen = 8;
1013 }
1014 
1015 /**
1016  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
1017  *   @t: the outgoing tunnel device
1018  *   @hdr: IPv6 header from the incoming packet
1019  *
1020  * Description:
1021  *   Avoid trivial tunneling loop by checking that tunnel exit-point
1022  *   doesn't match source of incoming packet.
1023  *
1024  * Return:
1025  *   1 if conflict,
1026  *   0 else
1027  **/
1028 
1029 static inline bool
ip6_tnl_addr_conflict(const struct ip6_tnl * t,const struct ipv6hdr * hdr)1030 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
1031 {
1032 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
1033 }
1034 
ip6_tnl_xmit_ctl(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)1035 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
1036 		     const struct in6_addr *laddr,
1037 		     const struct in6_addr *raddr)
1038 {
1039 	struct __ip6_tnl_parm *p = &t->parms;
1040 	int ret = 0;
1041 	struct net *net = t->net;
1042 
1043 	if (t->parms.collect_md)
1044 		return 1;
1045 
1046 	if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1047 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1048 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1049 		struct net_device *ldev = NULL;
1050 
1051 		rcu_read_lock();
1052 		if (p->link)
1053 			ldev = dev_get_by_index_rcu(net, p->link);
1054 
1055 		if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
1056 						      0, IFA_F_TENTATIVE)))
1057 			pr_warn_ratelimited("%s xmit: Local address not yet configured!\n",
1058 					    p->name);
1059 		else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1060 			 !ipv6_addr_is_multicast(raddr) &&
1061 			 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1062 							  true, 0, IFA_F_TENTATIVE)))
1063 			pr_warn_ratelimited("%s xmit: Routing loop! Remote address found on this node!\n",
1064 					    p->name);
1065 		else
1066 			ret = 1;
1067 		rcu_read_unlock();
1068 	}
1069 	return ret;
1070 }
1071 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1072 
1073 /**
1074  * ip6_tnl_xmit - encapsulate packet and send
1075  *   @skb: the outgoing socket buffer
1076  *   @dev: the outgoing tunnel device
1077  *   @dsfield: dscp code for outer header
1078  *   @fl6: flow of tunneled packet
1079  *   @encap_limit: encapsulation limit
1080  *   @pmtu: Path MTU is stored if packet is too big
1081  *   @proto: next header value
1082  *
1083  * Description:
1084  *   Build new header and do some sanity checks on the packet before sending
1085  *   it.
1086  *
1087  * Return:
1088  *   0 on success
1089  *   -1 fail
1090  *   %-EMSGSIZE message too big. return mtu in this case.
1091  **/
1092 
ip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev,__u8 dsfield,struct flowi6 * fl6,int encap_limit,__u32 * pmtu,__u8 proto)1093 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1094 		 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1095 		 __u8 proto)
1096 {
1097 	struct ip6_tnl *t = netdev_priv(dev);
1098 	struct net *net = t->net;
1099 	struct ipv6hdr *ipv6h;
1100 	struct ipv6_tel_txoption opt;
1101 	struct dst_entry *dst = NULL, *ndst = NULL;
1102 	struct net_device *tdev;
1103 	int mtu;
1104 	unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1105 	unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1106 	unsigned int max_headroom = psh_hlen;
1107 	__be16 payload_protocol;
1108 	bool use_cache = false;
1109 	u8 hop_limit;
1110 	int err = -1;
1111 
1112 	payload_protocol = skb_protocol(skb, true);
1113 
1114 	if (t->parms.collect_md) {
1115 		hop_limit = skb_tunnel_info(skb)->key.ttl;
1116 		goto route_lookup;
1117 	} else {
1118 		hop_limit = t->parms.hop_limit;
1119 	}
1120 
1121 	/* NBMA tunnel */
1122 	if (ipv6_addr_any(&t->parms.raddr)) {
1123 		if (payload_protocol == htons(ETH_P_IPV6)) {
1124 			struct in6_addr *addr6;
1125 			struct neighbour *neigh;
1126 			int addr_type;
1127 
1128 			if (!skb_dst(skb))
1129 				goto tx_err_link_failure;
1130 
1131 			neigh = dst_neigh_lookup(skb_dst(skb),
1132 						 &ipv6_hdr(skb)->daddr);
1133 			if (!neigh)
1134 				goto tx_err_link_failure;
1135 
1136 			addr6 = (struct in6_addr *)&neigh->primary_key;
1137 			addr_type = ipv6_addr_type(addr6);
1138 
1139 			if (addr_type == IPV6_ADDR_ANY)
1140 				addr6 = &ipv6_hdr(skb)->daddr;
1141 
1142 			memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1143 			neigh_release(neigh);
1144 		} else if (payload_protocol == htons(ETH_P_IP)) {
1145 			const struct rtable *rt = skb_rtable(skb);
1146 
1147 			if (!rt)
1148 				goto tx_err_link_failure;
1149 
1150 			if (rt->rt_gw_family == AF_INET6)
1151 				memcpy(&fl6->daddr, &rt->rt_gw6, sizeof(fl6->daddr));
1152 		}
1153 	} else if (t->parms.proto != 0 && !(t->parms.flags &
1154 					    (IP6_TNL_F_USE_ORIG_TCLASS |
1155 					     IP6_TNL_F_USE_ORIG_FWMARK))) {
1156 		/* enable the cache only if neither the outer protocol nor the
1157 		 * routing decision depends on the current inner header value
1158 		 */
1159 		use_cache = true;
1160 	}
1161 
1162 	if (use_cache)
1163 		dst = dst_cache_get(&t->dst_cache);
1164 
1165 	if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1166 		goto tx_err_link_failure;
1167 
1168 	if (!dst) {
1169 route_lookup:
1170 		/* add dsfield to flowlabel for route lookup */
1171 		fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1172 
1173 		dst = ip6_route_output(net, NULL, fl6);
1174 
1175 		if (dst->error)
1176 			goto tx_err_link_failure;
1177 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1178 		if (IS_ERR(dst)) {
1179 			err = PTR_ERR(dst);
1180 			dst = NULL;
1181 			goto tx_err_link_failure;
1182 		}
1183 		if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1184 		    ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1185 				       &fl6->daddr, 0, &fl6->saddr))
1186 			goto tx_err_link_failure;
1187 		ndst = dst;
1188 	}
1189 
1190 	tdev = dst_dev(dst);
1191 
1192 	if (tdev == dev) {
1193 		DEV_STATS_INC(dev, collisions);
1194 		net_warn_ratelimited("%s: Local routing loop detected!\n",
1195 				     t->parms.name);
1196 		goto tx_err_dst_release;
1197 	}
1198 	mtu = dst6_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1199 	if (encap_limit >= 0) {
1200 		max_headroom += 8;
1201 		mtu -= 8;
1202 	}
1203 	mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1204 		       IPV6_MIN_MTU : IPV4_MIN_MTU);
1205 
1206 	skb_dst_update_pmtu_no_confirm(skb, mtu);
1207 	if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1208 		*pmtu = mtu;
1209 		err = -EMSGSIZE;
1210 		goto tx_err_dst_release;
1211 	}
1212 
1213 	if (t->err_count > 0) {
1214 		if (time_before(jiffies,
1215 				t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1216 			t->err_count--;
1217 
1218 			dst_link_failure(skb);
1219 		} else {
1220 			t->err_count = 0;
1221 		}
1222 	}
1223 
1224 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1225 
1226 	/*
1227 	 * Okay, now see if we can stuff it in the buffer as-is.
1228 	 */
1229 	max_headroom += LL_RESERVED_SPACE(tdev);
1230 
1231 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1232 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1233 		struct sk_buff *new_skb;
1234 
1235 		new_skb = skb_realloc_headroom(skb, max_headroom);
1236 		if (!new_skb)
1237 			goto tx_err_dst_release;
1238 
1239 		if (skb->sk)
1240 			skb_set_owner_w(new_skb, skb->sk);
1241 		consume_skb(skb);
1242 		skb = new_skb;
1243 	}
1244 
1245 	if (t->parms.collect_md) {
1246 		if (t->encap.type != TUNNEL_ENCAP_NONE)
1247 			goto tx_err_dst_release;
1248 	} else {
1249 		if (use_cache && ndst)
1250 			dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1251 	}
1252 	skb_dst_set(skb, dst);
1253 
1254 	if (hop_limit == 0) {
1255 		if (payload_protocol == htons(ETH_P_IP))
1256 			hop_limit = ip_hdr(skb)->ttl;
1257 		else if (payload_protocol == htons(ETH_P_IPV6))
1258 			hop_limit = ipv6_hdr(skb)->hop_limit;
1259 		else
1260 			hop_limit = ip6_dst_hoplimit(dst);
1261 	}
1262 
1263 	/* Calculate max headroom for all the headers and adjust
1264 	 * needed_headroom if necessary.
1265 	 */
1266 	max_headroom = LL_RESERVED_SPACE(tdev) + sizeof(struct ipv6hdr)
1267 			+ dst->header_len + t->hlen;
1268 	ip_tunnel_adj_headroom(dev, max_headroom);
1269 
1270 	err = ip6_tnl_encap(skb, t, &proto, fl6);
1271 	if (err)
1272 		return err;
1273 
1274 	if (encap_limit >= 0) {
1275 		init_tel_txopt(&opt, encap_limit);
1276 		proto = ipv6_push_frag_opts(skb, &opt.ops, proto);
1277 	}
1278 
1279 	skb_push(skb, sizeof(struct ipv6hdr));
1280 	skb_reset_network_header(skb);
1281 	ipv6h = ipv6_hdr(skb);
1282 	ip6_flow_hdr(ipv6h, dsfield,
1283 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1284 	ipv6h->hop_limit = hop_limit;
1285 	ipv6h->nexthdr = proto;
1286 	ipv6h->saddr = fl6->saddr;
1287 	ipv6h->daddr = fl6->daddr;
1288 	ip6tunnel_xmit(NULL, skb, dev, 0);
1289 	return 0;
1290 tx_err_link_failure:
1291 	DEV_STATS_INC(dev, tx_carrier_errors);
1292 	dst_link_failure(skb);
1293 tx_err_dst_release:
1294 	dst_release(dst);
1295 	return err;
1296 }
1297 EXPORT_SYMBOL(ip6_tnl_xmit);
1298 
1299 static inline int
ipxip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev,u8 protocol)1300 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev,
1301 		u8 protocol)
1302 {
1303 	struct ip6_tnl *t = netdev_priv(dev);
1304 	struct ipv6hdr *ipv6h;
1305 	const struct iphdr  *iph;
1306 	int encap_limit = -1;
1307 	__u16 offset;
1308 	struct flowi6 fl6;
1309 	__u8 dsfield, orig_dsfield;
1310 	__u32 mtu;
1311 	u8 tproto;
1312 	int err;
1313 
1314 	tproto = READ_ONCE(t->parms.proto);
1315 	if (tproto != protocol && tproto != 0)
1316 		return -1;
1317 
1318 	if (t->parms.collect_md) {
1319 		struct ip_tunnel_info *tun_info;
1320 		const struct ip_tunnel_key *key;
1321 
1322 		tun_info = skb_tunnel_info(skb);
1323 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1324 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1325 			return -1;
1326 		key = &tun_info->key;
1327 		memset(&fl6, 0, sizeof(fl6));
1328 		fl6.flowi6_proto = protocol;
1329 		fl6.saddr = key->u.ipv6.src;
1330 		fl6.daddr = key->u.ipv6.dst;
1331 		fl6.flowlabel = key->label;
1332 		dsfield =  key->tos;
1333 		switch (protocol) {
1334 		case IPPROTO_IPIP:
1335 			iph = ip_hdr(skb);
1336 			orig_dsfield = ipv4_get_dsfield(iph);
1337 			break;
1338 		case IPPROTO_IPV6:
1339 			ipv6h = ipv6_hdr(skb);
1340 			orig_dsfield = ipv6_get_dsfield(ipv6h);
1341 			break;
1342 		default:
1343 			orig_dsfield = dsfield;
1344 			break;
1345 		}
1346 	} else {
1347 		if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1348 			encap_limit = t->parms.encap_limit;
1349 		if (protocol == IPPROTO_IPV6) {
1350 			offset = ip6_tnl_parse_tlv_enc_lim(skb,
1351 						skb_network_header(skb));
1352 			/* ip6_tnl_parse_tlv_enc_lim() might have
1353 			 * reallocated skb->head
1354 			 */
1355 			if (offset > 0) {
1356 				struct ipv6_tlv_tnl_enc_lim *tel;
1357 
1358 				tel = (void *)&skb_network_header(skb)[offset];
1359 				if (tel->encap_limit == 0) {
1360 					icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
1361 							ICMPV6_HDR_FIELD, offset + 2);
1362 					return -1;
1363 				}
1364 				encap_limit = tel->encap_limit - 1;
1365 			}
1366 		}
1367 
1368 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1369 		fl6.flowi6_proto = protocol;
1370 
1371 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1372 			fl6.flowi6_mark = skb->mark;
1373 		else
1374 			fl6.flowi6_mark = t->parms.fwmark;
1375 		switch (protocol) {
1376 		case IPPROTO_IPIP:
1377 			iph = ip_hdr(skb);
1378 			orig_dsfield = ipv4_get_dsfield(iph);
1379 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1380 				dsfield = orig_dsfield;
1381 			else
1382 				dsfield = ip6_tclass(t->parms.flowinfo);
1383 			break;
1384 		case IPPROTO_IPV6:
1385 			ipv6h = ipv6_hdr(skb);
1386 			orig_dsfield = ipv6_get_dsfield(ipv6h);
1387 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1388 				dsfield = orig_dsfield;
1389 			else
1390 				dsfield = ip6_tclass(t->parms.flowinfo);
1391 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1392 				fl6.flowlabel |= ip6_flowlabel(ipv6h);
1393 			break;
1394 		default:
1395 			orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo);
1396 			break;
1397 		}
1398 	}
1399 
1400 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1401 	dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield);
1402 
1403 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1404 		return -1;
1405 
1406 	skb_set_inner_ipproto(skb, protocol);
1407 
1408 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1409 			   protocol);
1410 	if (err != 0) {
1411 		/* XXX: send ICMP error even if DF is not set. */
1412 		if (err == -EMSGSIZE)
1413 			switch (protocol) {
1414 			case IPPROTO_IPIP:
1415 				icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1416 					      ICMP_FRAG_NEEDED, htonl(mtu));
1417 				break;
1418 			case IPPROTO_IPV6:
1419 				icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1420 				break;
1421 			default:
1422 				break;
1423 			}
1424 		return -1;
1425 	}
1426 
1427 	return 0;
1428 }
1429 
1430 static netdev_tx_t
ip6_tnl_start_xmit(struct sk_buff * skb,struct net_device * dev)1431 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1432 {
1433 	struct ip6_tnl *t = netdev_priv(dev);
1434 	u8 ipproto;
1435 	int ret;
1436 
1437 	if (!pskb_inet_may_pull(skb))
1438 		goto tx_err;
1439 
1440 	switch (skb->protocol) {
1441 	case htons(ETH_P_IP):
1442 		ipproto = IPPROTO_IPIP;
1443 		break;
1444 	case htons(ETH_P_IPV6):
1445 		if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb)))
1446 			goto tx_err;
1447 		ipproto = IPPROTO_IPV6;
1448 		break;
1449 	case htons(ETH_P_MPLS_UC):
1450 		ipproto = IPPROTO_MPLS;
1451 		break;
1452 	default:
1453 		goto tx_err;
1454 	}
1455 
1456 	ret = ipxip6_tnl_xmit(skb, dev, ipproto);
1457 	if (ret < 0)
1458 		goto tx_err;
1459 
1460 	return NETDEV_TX_OK;
1461 
1462 tx_err:
1463 	DEV_STATS_INC(dev, tx_errors);
1464 	DEV_STATS_INC(dev, tx_dropped);
1465 	kfree_skb(skb);
1466 	return NETDEV_TX_OK;
1467 }
1468 
ip6_tnl_link_config(struct ip6_tnl * t)1469 static void ip6_tnl_link_config(struct ip6_tnl *t)
1470 {
1471 	struct net_device *dev = t->dev;
1472 	struct net_device *tdev = NULL;
1473 	struct __ip6_tnl_parm *p = &t->parms;
1474 	struct flowi6 *fl6 = &t->fl.u.ip6;
1475 	int t_hlen;
1476 	int mtu;
1477 
1478 	__dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
1479 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1480 
1481 	/* Set up flowi template */
1482 	fl6->saddr = p->laddr;
1483 	fl6->daddr = p->raddr;
1484 	fl6->flowi6_oif = p->link;
1485 	fl6->flowlabel = 0;
1486 
1487 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1488 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1489 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1490 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1491 
1492 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1493 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1494 
1495 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1496 		dev->flags |= IFF_POINTOPOINT;
1497 	else
1498 		dev->flags &= ~IFF_POINTOPOINT;
1499 
1500 	t->tun_hlen = 0;
1501 	t->hlen = t->encap_hlen + t->tun_hlen;
1502 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1503 
1504 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1505 		int strict = (ipv6_addr_type(&p->raddr) &
1506 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1507 
1508 		struct rt6_info *rt = rt6_lookup(t->net,
1509 						 &p->raddr, &p->laddr,
1510 						 p->link, NULL, strict);
1511 		if (rt) {
1512 			tdev = rt->dst.dev;
1513 			ip6_rt_put(rt);
1514 		}
1515 
1516 		if (!tdev && p->link)
1517 			tdev = __dev_get_by_index(t->net, p->link);
1518 
1519 		if (tdev) {
1520 			dev->needed_headroom = tdev->hard_header_len +
1521 				tdev->needed_headroom + t_hlen;
1522 			mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU);
1523 
1524 			mtu = mtu - t_hlen;
1525 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1526 				mtu -= 8;
1527 
1528 			if (mtu < IPV6_MIN_MTU)
1529 				mtu = IPV6_MIN_MTU;
1530 			WRITE_ONCE(dev->mtu, mtu);
1531 		}
1532 	}
1533 }
1534 
1535 /**
1536  * ip6_tnl_change - update the tunnel parameters
1537  *   @t: tunnel to be changed
1538  *   @p: tunnel configuration parameters
1539  *
1540  * Description:
1541  *   ip6_tnl_change() updates the tunnel parameters
1542  **/
1543 
1544 static void
ip6_tnl_change(struct ip6_tnl * t,const struct __ip6_tnl_parm * p)1545 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1546 {
1547 	t->parms.laddr = p->laddr;
1548 	t->parms.raddr = p->raddr;
1549 	t->parms.flags = p->flags;
1550 	t->parms.hop_limit = p->hop_limit;
1551 	t->parms.encap_limit = p->encap_limit;
1552 	t->parms.flowinfo = p->flowinfo;
1553 	t->parms.link = p->link;
1554 	t->parms.proto = p->proto;
1555 	t->parms.fwmark = p->fwmark;
1556 	dst_cache_reset(&t->dst_cache);
1557 	ip6_tnl_link_config(t);
1558 }
1559 
ip6_tnl_update(struct ip6_tnl * t,struct __ip6_tnl_parm * p)1560 static void ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1561 {
1562 	struct net *net = t->net;
1563 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1564 
1565 	ip6_tnl_unlink(ip6n, t);
1566 	synchronize_net();
1567 	ip6_tnl_change(t, p);
1568 	ip6_tnl_link(ip6n, t);
1569 	netdev_state_change(t->dev);
1570 }
1571 
ip6_tnl0_update(struct ip6_tnl * t,struct __ip6_tnl_parm * p,bool strict)1572 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p,
1573 			   bool strict)
1574 {
1575 	/* For the default ip6tnl0 device, allow changing only the protocol
1576 	 * (the IP6_TNL_F_CAP_PER_PACKET flag is set on ip6tnl0, and all other
1577 	 * parameters are 0).
1578 	 */
1579 	if (strict &&
1580 	    (!ipv6_addr_any(&p->laddr) || !ipv6_addr_any(&p->raddr) ||
1581 	     p->flags != t->parms.flags || p->hop_limit || p->encap_limit ||
1582 	     p->flowinfo || p->link || p->fwmark || p->collect_md))
1583 		return -EINVAL;
1584 
1585 	t->parms.proto = p->proto;
1586 	netdev_state_change(t->dev);
1587 	return 0;
1588 }
1589 
1590 static void
ip6_tnl_parm_from_user(struct __ip6_tnl_parm * p,const struct ip6_tnl_parm * u)1591 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1592 {
1593 	p->laddr = u->laddr;
1594 	p->raddr = u->raddr;
1595 	p->flags = u->flags;
1596 	p->hop_limit = u->hop_limit;
1597 	p->encap_limit = u->encap_limit;
1598 	p->flowinfo = u->flowinfo;
1599 	p->link = u->link;
1600 	p->proto = u->proto;
1601 	memcpy(p->name, u->name, sizeof(u->name));
1602 }
1603 
1604 static void
ip6_tnl_parm_to_user(struct ip6_tnl_parm * u,const struct __ip6_tnl_parm * p)1605 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1606 {
1607 	u->laddr = p->laddr;
1608 	u->raddr = p->raddr;
1609 	u->flags = p->flags;
1610 	u->hop_limit = p->hop_limit;
1611 	u->encap_limit = p->encap_limit;
1612 	u->flowinfo = p->flowinfo;
1613 	u->link = p->link;
1614 	u->proto = p->proto;
1615 	memcpy(u->name, p->name, sizeof(u->name));
1616 }
1617 
1618 /**
1619  * ip6_tnl_siocdevprivate - configure ipv6 tunnels from userspace
1620  *   @dev: virtual device associated with tunnel
1621  *   @ifr: unused
1622  *   @data: parameters passed from userspace
1623  *   @cmd: command to be performed
1624  *
1625  * Description:
1626  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1627  *   from userspace.
1628  *
1629  *   The possible commands are the following:
1630  *     %SIOCGETTUNNEL: get tunnel parameters for device
1631  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1632  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1633  *     %SIOCDELTUNNEL: delete tunnel
1634  *
1635  *   The fallback device "ip6tnl0", created during module
1636  *   initialization, can be used for creating other tunnel devices.
1637  *
1638  * Return:
1639  *   0 on success,
1640  *   %-EFAULT if unable to copy data to or from userspace,
1641  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1642  *   %-EINVAL if passed tunnel parameters are invalid,
1643  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1644  *   %-ENODEV if attempting to change or delete a nonexisting device
1645  **/
1646 
1647 static int
ip6_tnl_siocdevprivate(struct net_device * dev,struct ifreq * ifr,void __user * data,int cmd)1648 ip6_tnl_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1649 		       void __user *data, int cmd)
1650 {
1651 	int err = 0;
1652 	struct ip6_tnl_parm p;
1653 	struct __ip6_tnl_parm p1;
1654 	struct ip6_tnl *t = netdev_priv(dev);
1655 	struct net *net = t->net;
1656 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1657 
1658 	memset(&p1, 0, sizeof(p1));
1659 
1660 	switch (cmd) {
1661 	case SIOCGETTUNNEL:
1662 		if (dev == ip6n->fb_tnl_dev) {
1663 			if (copy_from_user(&p, data, sizeof(p))) {
1664 				err = -EFAULT;
1665 				break;
1666 			}
1667 			ip6_tnl_parm_from_user(&p1, &p);
1668 			t = ip6_tnl_locate(net, &p1, 0);
1669 			if (IS_ERR(t))
1670 				t = netdev_priv(dev);
1671 		} else {
1672 			memset(&p, 0, sizeof(p));
1673 		}
1674 		ip6_tnl_parm_to_user(&p, &t->parms);
1675 		if (copy_to_user(data, &p, sizeof(p)))
1676 			err = -EFAULT;
1677 		break;
1678 	case SIOCADDTUNNEL:
1679 	case SIOCCHGTUNNEL:
1680 		err = -EPERM;
1681 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1682 			break;
1683 		err = -EFAULT;
1684 		if (copy_from_user(&p, data, sizeof(p)))
1685 			break;
1686 		err = -EINVAL;
1687 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1688 		    p.proto != 0)
1689 			break;
1690 		ip6_tnl_parm_from_user(&p1, &p);
1691 		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1692 		if (cmd == SIOCCHGTUNNEL) {
1693 			if (!IS_ERR(t)) {
1694 				if (t->dev != dev) {
1695 					err = -EEXIST;
1696 					break;
1697 				}
1698 			} else
1699 				t = netdev_priv(dev);
1700 			if (dev == ip6n->fb_tnl_dev)
1701 				ip6_tnl0_update(t, &p1, false);
1702 			else
1703 				ip6_tnl_update(t, &p1);
1704 		}
1705 		if (!IS_ERR(t)) {
1706 			err = 0;
1707 			ip6_tnl_parm_to_user(&p, &t->parms);
1708 			if (copy_to_user(data, &p, sizeof(p)))
1709 				err = -EFAULT;
1710 
1711 		} else {
1712 			err = PTR_ERR(t);
1713 		}
1714 		break;
1715 	case SIOCDELTUNNEL:
1716 		err = -EPERM;
1717 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1718 			break;
1719 
1720 		if (dev == ip6n->fb_tnl_dev) {
1721 			err = -EFAULT;
1722 			if (copy_from_user(&p, data, sizeof(p)))
1723 				break;
1724 			err = -ENOENT;
1725 			ip6_tnl_parm_from_user(&p1, &p);
1726 			t = ip6_tnl_locate(net, &p1, 0);
1727 			if (IS_ERR(t))
1728 				break;
1729 			err = -EPERM;
1730 			if (t->dev == ip6n->fb_tnl_dev)
1731 				break;
1732 			dev = t->dev;
1733 		}
1734 		err = 0;
1735 		unregister_netdevice(dev);
1736 		break;
1737 	default:
1738 		err = -EINVAL;
1739 	}
1740 	return err;
1741 }
1742 
1743 /**
1744  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1745  *   @dev: virtual device associated with tunnel
1746  *   @new_mtu: the new mtu
1747  *
1748  * Return:
1749  *   0 on success,
1750  *   %-EINVAL if mtu too small
1751  **/
1752 
ip6_tnl_change_mtu(struct net_device * dev,int new_mtu)1753 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1754 {
1755 	struct ip6_tnl *tnl = netdev_priv(dev);
1756 	int t_hlen;
1757 
1758 	t_hlen = tnl->hlen + sizeof(struct ipv6hdr);
1759 	if (tnl->parms.proto == IPPROTO_IPV6) {
1760 		if (new_mtu < IPV6_MIN_MTU)
1761 			return -EINVAL;
1762 	} else {
1763 		if (new_mtu < ETH_MIN_MTU)
1764 			return -EINVAL;
1765 	}
1766 	if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1767 		if (new_mtu > IP6_MAX_MTU - dev->hard_header_len - t_hlen)
1768 			return -EINVAL;
1769 	} else {
1770 		if (new_mtu > IP_MAX_MTU - dev->hard_header_len - t_hlen)
1771 			return -EINVAL;
1772 	}
1773 	WRITE_ONCE(dev->mtu, new_mtu);
1774 	return 0;
1775 }
1776 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1777 
ip6_tnl_get_iflink(const struct net_device * dev)1778 int ip6_tnl_get_iflink(const struct net_device *dev)
1779 {
1780 	struct ip6_tnl *t = netdev_priv(dev);
1781 
1782 	return READ_ONCE(t->parms.link);
1783 }
1784 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1785 
ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops * ops,unsigned int num)1786 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1787 			  unsigned int num)
1788 {
1789 	if (num >= MAX_IPTUN_ENCAP_OPS)
1790 		return -ERANGE;
1791 
1792 	return !cmpxchg((const struct ip6_tnl_encap_ops **)
1793 			&ip6tun_encaps[num],
1794 			NULL, ops) ? 0 : -1;
1795 }
1796 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1797 
ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops * ops,unsigned int num)1798 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1799 			  unsigned int num)
1800 {
1801 	int ret;
1802 
1803 	if (num >= MAX_IPTUN_ENCAP_OPS)
1804 		return -ERANGE;
1805 
1806 	ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1807 		       &ip6tun_encaps[num],
1808 		       ops, NULL) == ops) ? 0 : -1;
1809 
1810 	synchronize_net();
1811 
1812 	return ret;
1813 }
1814 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1815 
ip6_tnl_encap_setup(struct ip6_tnl * t,struct ip_tunnel_encap * ipencap)1816 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1817 			struct ip_tunnel_encap *ipencap)
1818 {
1819 	int hlen;
1820 
1821 	memset(&t->encap, 0, sizeof(t->encap));
1822 
1823 	hlen = ip6_encap_hlen(ipencap);
1824 	if (hlen < 0)
1825 		return hlen;
1826 
1827 	t->encap.type = ipencap->type;
1828 	t->encap.sport = ipencap->sport;
1829 	t->encap.dport = ipencap->dport;
1830 	t->encap.flags = ipencap->flags;
1831 
1832 	t->encap_hlen = hlen;
1833 	t->hlen = t->encap_hlen + t->tun_hlen;
1834 
1835 	return 0;
1836 }
1837 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1838 
ip6_tnl_fill_forward_path(struct net_device_path_ctx * ctx,struct net_device_path * path)1839 static int ip6_tnl_fill_forward_path(struct net_device_path_ctx *ctx,
1840 				     struct net_device_path *path)
1841 {
1842 	struct ip6_tnl *t = netdev_priv(ctx->dev);
1843 	struct flowi6 fl6 = {
1844 		.daddr = t->parms.raddr,
1845 	};
1846 	struct dst_entry *dst;
1847 	int err;
1848 
1849 	dst = ip6_route_output(dev_net(ctx->dev), NULL, &fl6);
1850 	if (!dst->error) {
1851 		path->type = DEV_PATH_TUN;
1852 		path->tun.src_v6 = t->parms.laddr;
1853 		path->tun.dst_v6 = t->parms.raddr;
1854 		path->tun.l3_proto = IPPROTO_IPV6;
1855 		path->dev = ctx->dev;
1856 		ctx->dev = dst->dev;
1857 	}
1858 
1859 	err = dst->error;
1860 	dst_release(dst);
1861 
1862 	return err;
1863 }
1864 
1865 static const struct net_device_ops ip6_tnl_netdev_ops = {
1866 	.ndo_init	= ip6_tnl_dev_init,
1867 	.ndo_uninit	= ip6_tnl_dev_uninit,
1868 	.ndo_start_xmit = ip6_tnl_start_xmit,
1869 	.ndo_siocdevprivate = ip6_tnl_siocdevprivate,
1870 	.ndo_change_mtu = ip6_tnl_change_mtu,
1871 	.ndo_get_stats64 = dev_get_tstats64,
1872 	.ndo_get_iflink = ip6_tnl_get_iflink,
1873 	.ndo_fill_forward_path = ip6_tnl_fill_forward_path,
1874 };
1875 
1876 #define IPXIPX_FEATURES (NETIF_F_SG |		\
1877 			 NETIF_F_FRAGLIST |	\
1878 			 NETIF_F_HIGHDMA |	\
1879 			 NETIF_F_GSO_SOFTWARE |	\
1880 			 NETIF_F_HW_CSUM)
1881 
1882 /**
1883  * ip6_tnl_dev_setup - setup virtual tunnel device
1884  *   @dev: virtual device associated with tunnel
1885  *
1886  * Description:
1887  *   Initialize function pointers and device parameters
1888  **/
1889 
ip6_tnl_dev_setup(struct net_device * dev)1890 static void ip6_tnl_dev_setup(struct net_device *dev)
1891 {
1892 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1893 	dev->header_ops = &ip_tunnel_header_ops;
1894 	dev->needs_free_netdev = true;
1895 	dev->priv_destructor = ip6_dev_free;
1896 
1897 	dev->type = ARPHRD_TUNNEL6;
1898 	dev->flags |= IFF_NOARP;
1899 	dev->addr_len = sizeof(struct in6_addr);
1900 	dev->lltx = true;
1901 	dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
1902 	netif_keep_dst(dev);
1903 
1904 	dev->features		|= IPXIPX_FEATURES;
1905 	dev->hw_features	|= IPXIPX_FEATURES;
1906 
1907 	/* This perm addr will be used as interface identifier by IPv6 */
1908 	dev->addr_assign_type = NET_ADDR_RANDOM;
1909 	eth_random_addr(dev->perm_addr);
1910 }
1911 
1912 
1913 /**
1914  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1915  *   @dev: virtual device associated with tunnel
1916  **/
1917 
1918 static inline int
ip6_tnl_dev_init_gen(struct net_device * dev)1919 ip6_tnl_dev_init_gen(struct net_device *dev)
1920 {
1921 	struct ip6_tnl *t = netdev_priv(dev);
1922 	int ret;
1923 	int t_hlen;
1924 
1925 	t->dev = dev;
1926 
1927 	ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1928 	if (ret)
1929 		return ret;
1930 
1931 	ret = gro_cells_init(&t->gro_cells, dev);
1932 	if (ret)
1933 		goto destroy_dst;
1934 
1935 	t->tun_hlen = 0;
1936 	t->hlen = t->encap_hlen + t->tun_hlen;
1937 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1938 
1939 	dev->type = ARPHRD_TUNNEL6;
1940 	dev->mtu = ETH_DATA_LEN - t_hlen;
1941 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1942 		dev->mtu -= 8;
1943 	dev->min_mtu = ETH_MIN_MTU;
1944 	dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len - t_hlen;
1945 
1946 	netdev_hold(dev, &t->dev_tracker, GFP_KERNEL);
1947 	netdev_lockdep_set_classes(dev);
1948 	return 0;
1949 
1950 destroy_dst:
1951 	dst_cache_destroy(&t->dst_cache);
1952 
1953 	return ret;
1954 }
1955 
1956 /**
1957  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1958  *   @dev: virtual device associated with tunnel
1959  **/
1960 
ip6_tnl_dev_init(struct net_device * dev)1961 static int ip6_tnl_dev_init(struct net_device *dev)
1962 {
1963 	struct ip6_tnl *t = netdev_priv(dev);
1964 	int err = ip6_tnl_dev_init_gen(dev);
1965 
1966 	if (err)
1967 		return err;
1968 	ip6_tnl_link_config(t);
1969 	if (t->parms.collect_md)
1970 		netif_keep_dst(dev);
1971 	return 0;
1972 }
1973 
1974 /**
1975  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1976  *   @dev: fallback device
1977  *
1978  * Return: 0
1979  **/
1980 
ip6_fb_tnl_dev_init(struct net_device * dev)1981 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1982 {
1983 	struct ip6_tnl *t = netdev_priv(dev);
1984 	struct net *net = dev_net(dev);
1985 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1986 
1987 	t->net = net;
1988 	t->parms.proto = IPPROTO_IPV6;
1989 
1990 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1991 	return 0;
1992 }
1993 
ip6_tnl_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1994 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1995 			    struct netlink_ext_ack *extack)
1996 {
1997 	u8 proto;
1998 
1999 	if (!data || !data[IFLA_IPTUN_PROTO])
2000 		return 0;
2001 
2002 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
2003 	if (proto != IPPROTO_IPV6 &&
2004 	    proto != IPPROTO_IPIP &&
2005 	    proto != 0)
2006 		return -EINVAL;
2007 
2008 	return 0;
2009 }
2010 
ip6_tnl_netlink_parms(struct nlattr * data[],struct __ip6_tnl_parm * parms)2011 static void ip6_tnl_netlink_parms(struct nlattr *data[],
2012 				  struct __ip6_tnl_parm *parms)
2013 {
2014 	memset(parms, 0, sizeof(*parms));
2015 
2016 	if (!data)
2017 		return;
2018 
2019 	if (data[IFLA_IPTUN_LINK])
2020 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
2021 
2022 	if (data[IFLA_IPTUN_LOCAL])
2023 		parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
2024 
2025 	if (data[IFLA_IPTUN_REMOTE])
2026 		parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
2027 
2028 	if (data[IFLA_IPTUN_TTL])
2029 		parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
2030 
2031 	if (data[IFLA_IPTUN_ENCAP_LIMIT])
2032 		parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
2033 
2034 	if (data[IFLA_IPTUN_FLOWINFO])
2035 		parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
2036 
2037 	if (data[IFLA_IPTUN_FLAGS])
2038 		parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
2039 
2040 	if (data[IFLA_IPTUN_PROTO])
2041 		parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
2042 
2043 	if (data[IFLA_IPTUN_COLLECT_METADATA])
2044 		parms->collect_md = true;
2045 
2046 	if (data[IFLA_IPTUN_FWMARK])
2047 		parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
2048 }
2049 
ip6_tnl_newlink(struct net_device * dev,struct rtnl_newlink_params * params,struct netlink_ext_ack * extack)2050 static int ip6_tnl_newlink(struct net_device *dev,
2051 			   struct rtnl_newlink_params *params,
2052 			   struct netlink_ext_ack *extack)
2053 {
2054 	struct nlattr **data = params->data;
2055 	struct nlattr **tb = params->tb;
2056 	struct ip_tunnel_encap ipencap;
2057 	struct ip6_tnl_net *ip6n;
2058 	struct ip6_tnl *nt, *t;
2059 	struct net *net;
2060 	int err;
2061 
2062 	net = params->link_net ? : dev_net(dev);
2063 	ip6n = net_generic(net, ip6_tnl_net_id);
2064 	nt = netdev_priv(dev);
2065 	nt->net = net;
2066 
2067 	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
2068 		err = ip6_tnl_encap_setup(nt, &ipencap);
2069 		if (err < 0)
2070 			return err;
2071 	}
2072 
2073 	ip6_tnl_netlink_parms(data, &nt->parms);
2074 
2075 	if (nt->parms.collect_md) {
2076 		if (rtnl_dereference(ip6n->collect_md_tun))
2077 			return -EEXIST;
2078 	} else {
2079 		t = ip6_tnl_locate(net, &nt->parms, 0);
2080 		if (!IS_ERR(t))
2081 			return -EEXIST;
2082 	}
2083 
2084 	err = ip6_tnl_create2(dev);
2085 	if (!err && tb[IFLA_MTU])
2086 		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2087 
2088 	return err;
2089 }
2090 
ip6_tnl_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)2091 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2092 			      struct nlattr *data[],
2093 			      struct netlink_ext_ack *extack)
2094 {
2095 	struct ip6_tnl *t = netdev_priv(dev);
2096 	struct __ip6_tnl_parm p;
2097 	struct net *net = t->net;
2098 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2099 	struct ip_tunnel_encap ipencap;
2100 
2101 	if (dev == ip6n->fb_tnl_dev) {
2102 		if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
2103 			/* iproute2 always sets TUNNEL_ENCAP_FLAG_CSUM6, so
2104 			 * let's ignore this flag.
2105 			 */
2106 			ipencap.flags &= ~TUNNEL_ENCAP_FLAG_CSUM6;
2107 			if (memchr_inv(&ipencap, 0, sizeof(ipencap))) {
2108 				NL_SET_ERR_MSG(extack,
2109 					       "Only protocol can be changed for fallback tunnel, not encap params");
2110 				return -EINVAL;
2111 			}
2112 		}
2113 
2114 		ip6_tnl_netlink_parms(data, &p);
2115 		if (ip6_tnl0_update(t, &p, true) < 0) {
2116 			NL_SET_ERR_MSG(extack,
2117 				       "Only protocol can be changed for fallback tunnel");
2118 			return -EINVAL;
2119 		}
2120 
2121 		return 0;
2122 	}
2123 
2124 	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
2125 		int err = ip6_tnl_encap_setup(t, &ipencap);
2126 
2127 		if (err < 0)
2128 			return err;
2129 	}
2130 	ip6_tnl_netlink_parms(data, &p);
2131 	if (p.collect_md)
2132 		return -EINVAL;
2133 
2134 	t = ip6_tnl_locate(net, &p, 0);
2135 	if (!IS_ERR(t)) {
2136 		if (t->dev != dev)
2137 			return -EEXIST;
2138 	} else
2139 		t = netdev_priv(dev);
2140 
2141 	ip6_tnl_update(t, &p);
2142 	return 0;
2143 }
2144 
ip6_tnl_dellink(struct net_device * dev,struct list_head * head)2145 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2146 {
2147 	struct net *net = dev_net(dev);
2148 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2149 
2150 	if (dev != ip6n->fb_tnl_dev)
2151 		unregister_netdevice_queue(dev, head);
2152 }
2153 
ip6_tnl_get_size(const struct net_device * dev)2154 static size_t ip6_tnl_get_size(const struct net_device *dev)
2155 {
2156 	return
2157 		/* IFLA_IPTUN_LINK */
2158 		nla_total_size(4) +
2159 		/* IFLA_IPTUN_LOCAL */
2160 		nla_total_size(sizeof(struct in6_addr)) +
2161 		/* IFLA_IPTUN_REMOTE */
2162 		nla_total_size(sizeof(struct in6_addr)) +
2163 		/* IFLA_IPTUN_TTL */
2164 		nla_total_size(1) +
2165 		/* IFLA_IPTUN_ENCAP_LIMIT */
2166 		nla_total_size(1) +
2167 		/* IFLA_IPTUN_FLOWINFO */
2168 		nla_total_size(4) +
2169 		/* IFLA_IPTUN_FLAGS */
2170 		nla_total_size(4) +
2171 		/* IFLA_IPTUN_PROTO */
2172 		nla_total_size(1) +
2173 		/* IFLA_IPTUN_ENCAP_TYPE */
2174 		nla_total_size(2) +
2175 		/* IFLA_IPTUN_ENCAP_FLAGS */
2176 		nla_total_size(2) +
2177 		/* IFLA_IPTUN_ENCAP_SPORT */
2178 		nla_total_size(2) +
2179 		/* IFLA_IPTUN_ENCAP_DPORT */
2180 		nla_total_size(2) +
2181 		/* IFLA_IPTUN_COLLECT_METADATA */
2182 		nla_total_size(0) +
2183 		/* IFLA_IPTUN_FWMARK */
2184 		nla_total_size(4) +
2185 		0;
2186 }
2187 
ip6_tnl_fill_info(struct sk_buff * skb,const struct net_device * dev)2188 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2189 {
2190 	struct ip6_tnl *tunnel = netdev_priv(dev);
2191 	struct __ip6_tnl_parm *parm = &tunnel->parms;
2192 
2193 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2194 	    nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2195 	    nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2196 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2197 	    nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2198 	    nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2199 	    nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2200 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2201 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2202 		goto nla_put_failure;
2203 
2204 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2205 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2206 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2207 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2208 		goto nla_put_failure;
2209 
2210 	if (parm->collect_md)
2211 		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2212 			goto nla_put_failure;
2213 
2214 	return 0;
2215 
2216 nla_put_failure:
2217 	return -EMSGSIZE;
2218 }
2219 
ip6_tnl_get_link_net(const struct net_device * dev)2220 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2221 {
2222 	struct ip6_tnl *tunnel = netdev_priv(dev);
2223 
2224 	return READ_ONCE(tunnel->net);
2225 }
2226 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2227 
2228 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2229 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
2230 	[IFLA_IPTUN_LOCAL]		= { .len = sizeof(struct in6_addr) },
2231 	[IFLA_IPTUN_REMOTE]		= { .len = sizeof(struct in6_addr) },
2232 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
2233 	[IFLA_IPTUN_ENCAP_LIMIT]	= { .type = NLA_U8 },
2234 	[IFLA_IPTUN_FLOWINFO]		= { .type = NLA_U32 },
2235 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U32 },
2236 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
2237 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
2238 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
2239 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
2240 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
2241 	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
2242 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
2243 };
2244 
2245 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2246 	.kind		= "ip6tnl",
2247 	.maxtype	= IFLA_IPTUN_MAX,
2248 	.policy		= ip6_tnl_policy,
2249 	.priv_size	= sizeof(struct ip6_tnl),
2250 	.setup		= ip6_tnl_dev_setup,
2251 	.validate	= ip6_tnl_validate,
2252 	.newlink	= ip6_tnl_newlink,
2253 	.changelink	= ip6_tnl_changelink,
2254 	.dellink	= ip6_tnl_dellink,
2255 	.get_size	= ip6_tnl_get_size,
2256 	.fill_info	= ip6_tnl_fill_info,
2257 	.get_link_net	= ip6_tnl_get_link_net,
2258 };
2259 
2260 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2261 	.handler	= ip4ip6_rcv,
2262 	.err_handler	= ip4ip6_err,
2263 	.priority	=	1,
2264 };
2265 
2266 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2267 	.handler	= ip6ip6_rcv,
2268 	.err_handler	= ip6ip6_err,
2269 	.priority	=	1,
2270 };
2271 
2272 static struct xfrm6_tunnel mplsip6_handler __read_mostly = {
2273 	.handler	= mplsip6_rcv,
2274 	.err_handler	= mplsip6_err,
2275 	.priority	=	1,
2276 };
2277 
ip6_tnl_exit_rtnl_net(struct net * net,struct list_head * list)2278 static void __net_exit ip6_tnl_exit_rtnl_net(struct net *net, struct list_head *list)
2279 {
2280 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2281 	struct net_device *dev, *aux;
2282 	int h;
2283 	struct ip6_tnl *t;
2284 
2285 	for_each_netdev_safe(net, dev, aux)
2286 		if (dev->rtnl_link_ops == &ip6_link_ops)
2287 			unregister_netdevice_queue(dev, list);
2288 
2289 	for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2290 		t = rtnl_net_dereference(net, ip6n->tnls_r_l[h]);
2291 		while (t) {
2292 			/* If dev is in the same netns, it has already
2293 			 * been added to the list by the previous loop.
2294 			 */
2295 			if (!net_eq(dev_net(t->dev), net))
2296 				unregister_netdevice_queue(t->dev, list);
2297 
2298 			t = rtnl_net_dereference(net, t->next);
2299 		}
2300 	}
2301 
2302 	t = rtnl_net_dereference(net, ip6n->tnls_wc[0]);
2303 	while (t) {
2304 		/* If dev is in the same netns, it has already
2305 		 * been added to the list by the previous loop.
2306 		 */
2307 		if (!net_eq(dev_net(t->dev), net))
2308 			unregister_netdevice_queue(t->dev, list);
2309 
2310 		t = rtnl_net_dereference(net, t->next);
2311 	}
2312 }
2313 
ip6_tnl_init_net(struct net * net)2314 static int __net_init ip6_tnl_init_net(struct net *net)
2315 {
2316 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2317 	struct ip6_tnl *t = NULL;
2318 	int err;
2319 
2320 	ip6n->tnls[0] = ip6n->tnls_wc;
2321 	ip6n->tnls[1] = ip6n->tnls_r_l;
2322 
2323 	if (!net_has_fallback_tunnels(net))
2324 		return 0;
2325 	err = -ENOMEM;
2326 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2327 					NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2328 
2329 	if (!ip6n->fb_tnl_dev)
2330 		goto err_alloc_dev;
2331 	dev_net_set(ip6n->fb_tnl_dev, net);
2332 	ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2333 	/* FB netdevice is special: we have one, and only one per netns.
2334 	 * Allowing to move it to another netns is clearly unsafe.
2335 	 */
2336 	ip6n->fb_tnl_dev->netns_immutable = true;
2337 
2338 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2339 	if (err < 0)
2340 		goto err_register;
2341 
2342 	err = register_netdev(ip6n->fb_tnl_dev);
2343 	if (err < 0)
2344 		goto err_register;
2345 
2346 	t = netdev_priv(ip6n->fb_tnl_dev);
2347 
2348 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2349 	return 0;
2350 
2351 err_register:
2352 	free_netdev(ip6n->fb_tnl_dev);
2353 err_alloc_dev:
2354 	return err;
2355 }
2356 
2357 static struct pernet_operations ip6_tnl_net_ops = {
2358 	.init = ip6_tnl_init_net,
2359 	.exit_rtnl = ip6_tnl_exit_rtnl_net,
2360 	.id   = &ip6_tnl_net_id,
2361 	.size = sizeof(struct ip6_tnl_net),
2362 };
2363 
2364 /**
2365  * ip6_tunnel_init - register protocol and reserve needed resources
2366  *
2367  * Return: 0 on success
2368  **/
2369 
ip6_tunnel_init(void)2370 static int __init ip6_tunnel_init(void)
2371 {
2372 	int  err;
2373 
2374 	if (!ipv6_mod_enabled())
2375 		return -EOPNOTSUPP;
2376 
2377 	err = register_pernet_device(&ip6_tnl_net_ops);
2378 	if (err < 0)
2379 		goto out_pernet;
2380 
2381 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2382 	if (err < 0) {
2383 		pr_err("%s: can't register ip4ip6\n", __func__);
2384 		goto out_ip4ip6;
2385 	}
2386 
2387 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2388 	if (err < 0) {
2389 		pr_err("%s: can't register ip6ip6\n", __func__);
2390 		goto out_ip6ip6;
2391 	}
2392 
2393 	if (ip6_tnl_mpls_supported()) {
2394 		err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS);
2395 		if (err < 0) {
2396 			pr_err("%s: can't register mplsip6\n", __func__);
2397 			goto out_mplsip6;
2398 		}
2399 	}
2400 
2401 	err = rtnl_link_register(&ip6_link_ops);
2402 	if (err < 0)
2403 		goto rtnl_link_failed;
2404 
2405 	return 0;
2406 
2407 rtnl_link_failed:
2408 	if (ip6_tnl_mpls_supported())
2409 		xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS);
2410 out_mplsip6:
2411 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2412 out_ip6ip6:
2413 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2414 out_ip4ip6:
2415 	unregister_pernet_device(&ip6_tnl_net_ops);
2416 out_pernet:
2417 	return err;
2418 }
2419 
2420 /**
2421  * ip6_tunnel_cleanup - free resources and unregister protocol
2422  **/
2423 
ip6_tunnel_cleanup(void)2424 static void __exit ip6_tunnel_cleanup(void)
2425 {
2426 	rtnl_link_unregister(&ip6_link_ops);
2427 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2428 		pr_info("%s: can't deregister ip4ip6\n", __func__);
2429 
2430 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2431 		pr_info("%s: can't deregister ip6ip6\n", __func__);
2432 
2433 	if (ip6_tnl_mpls_supported() &&
2434 	    xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS))
2435 		pr_info("%s: can't deregister mplsip6\n", __func__);
2436 	unregister_pernet_device(&ip6_tnl_net_ops);
2437 }
2438 
2439 module_init(ip6_tunnel_init);
2440 module_exit(ip6_tunnel_cleanup);
2441