• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  *	IPv6 tunneling device
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Ville Nuorvala		<vnuorval@tcs.hut.fi>
7  *	Yasuyuki Kozakai	<kozakai@linux-ipv6.org>
8  *
9  *      Based on:
10  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
11  *
12  *      RFC 2473
13  *
14  *	This program is free software; you can redistribute it and/or
15  *      modify it under the terms of the GNU General Public License
16  *      as published by the Free Software Foundation; either version
17  *      2 of the License, or (at your option) any later version.
18  *
19  */
20 
21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22 
23 #include <linux/module.h>
24 #include <linux/capability.h>
25 #include <linux/errno.h>
26 #include <linux/types.h>
27 #include <linux/sockios.h>
28 #include <linux/icmp.h>
29 #include <linux/if.h>
30 #include <linux/in.h>
31 #include <linux/ip.h>
32 #include <linux/net.h>
33 #include <linux/in6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/icmpv6.h>
37 #include <linux/init.h>
38 #include <linux/route.h>
39 #include <linux/rtnetlink.h>
40 #include <linux/netfilter_ipv6.h>
41 #include <linux/slab.h>
42 #include <linux/hash.h>
43 #include <linux/etherdevice.h>
44 
45 #include <asm/uaccess.h>
46 #include <linux/atomic.h>
47 
48 #include <net/icmp.h>
49 #include <net/ip.h>
50 #include <net/ip_tunnels.h>
51 #include <net/ipv6.h>
52 #include <net/ip6_route.h>
53 #include <net/addrconf.h>
54 #include <net/ip6_tunnel.h>
55 #include <net/xfrm.h>
56 #include <net/dsfield.h>
57 #include <net/inet_ecn.h>
58 #include <net/net_namespace.h>
59 #include <net/netns/generic.h>
60 
61 MODULE_AUTHOR("Ville Nuorvala");
62 MODULE_DESCRIPTION("IPv6 tunneling device");
63 MODULE_LICENSE("GPL");
64 MODULE_ALIAS_RTNL_LINK("ip6tnl");
65 MODULE_ALIAS_NETDEV("ip6tnl0");
66 
67 #ifdef IP6_TNL_DEBUG
68 #define IP6_TNL_TRACE(x...) pr_debug("%s:" x "\n", __func__)
69 #else
70 #define IP6_TNL_TRACE(x...) do {;} while(0)
71 #endif
72 
73 #define HASH_SIZE_SHIFT  5
74 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
75 
76 static bool log_ecn_error = true;
77 module_param(log_ecn_error, bool, 0644);
78 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
79 
HASH(const struct in6_addr * addr1,const struct in6_addr * addr2)80 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
81 {
82 	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
83 
84 	return hash_32(hash, HASH_SIZE_SHIFT);
85 }
86 
87 static int ip6_tnl_dev_init(struct net_device *dev);
88 static void ip6_tnl_dev_setup(struct net_device *dev);
89 static struct rtnl_link_ops ip6_link_ops __read_mostly;
90 
91 static int ip6_tnl_net_id __read_mostly;
92 struct ip6_tnl_net {
93 	/* the IPv6 tunnel fallback device */
94 	struct net_device *fb_tnl_dev;
95 	/* lists for storing tunnels in use */
96 	struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
97 	struct ip6_tnl __rcu *tnls_wc[1];
98 	struct ip6_tnl __rcu **tnls[2];
99 };
100 
ip6_get_stats(struct net_device * dev)101 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
102 {
103 	struct pcpu_sw_netstats tmp, sum = { 0 };
104 	int i;
105 
106 	for_each_possible_cpu(i) {
107 		unsigned int start;
108 		const struct pcpu_sw_netstats *tstats =
109 						   per_cpu_ptr(dev->tstats, i);
110 
111 		do {
112 			start = u64_stats_fetch_begin_irq(&tstats->syncp);
113 			tmp.rx_packets = tstats->rx_packets;
114 			tmp.rx_bytes = tstats->rx_bytes;
115 			tmp.tx_packets = tstats->tx_packets;
116 			tmp.tx_bytes =  tstats->tx_bytes;
117 		} while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
118 
119 		sum.rx_packets += tmp.rx_packets;
120 		sum.rx_bytes   += tmp.rx_bytes;
121 		sum.tx_packets += tmp.tx_packets;
122 		sum.tx_bytes   += tmp.tx_bytes;
123 	}
124 	dev->stats.rx_packets = sum.rx_packets;
125 	dev->stats.rx_bytes   = sum.rx_bytes;
126 	dev->stats.tx_packets = sum.tx_packets;
127 	dev->stats.tx_bytes   = sum.tx_bytes;
128 	return &dev->stats;
129 }
130 
131 /*
132  * Locking : hash tables are protected by RCU and RTNL
133  */
134 
ip6_tnl_dst_check(struct ip6_tnl * t)135 struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
136 {
137 	struct dst_entry *dst = t->dst_cache;
138 
139 	if (dst && dst->obsolete &&
140 	    dst->ops->check(dst, t->dst_cookie) == NULL) {
141 		t->dst_cache = NULL;
142 		dst_release(dst);
143 		return NULL;
144 	}
145 
146 	return dst;
147 }
148 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check);
149 
ip6_tnl_dst_reset(struct ip6_tnl * t)150 void ip6_tnl_dst_reset(struct ip6_tnl *t)
151 {
152 	dst_release(t->dst_cache);
153 	t->dst_cache = NULL;
154 }
155 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset);
156 
ip6_tnl_dst_store(struct ip6_tnl * t,struct dst_entry * dst)157 void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
158 {
159 	struct rt6_info *rt = (struct rt6_info *) dst;
160 	t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
161 	dst_release(t->dst_cache);
162 	t->dst_cache = dst;
163 }
164 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store);
165 
166 /**
167  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
168  *   @remote: the address of the tunnel exit-point
169  *   @local: the address of the tunnel entry-point
170  *
171  * Return:
172  *   tunnel matching given end-points if found,
173  *   else fallback tunnel if its device is up,
174  *   else %NULL
175  **/
176 
177 #define for_each_ip6_tunnel_rcu(start) \
178 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
179 
180 static struct ip6_tnl *
ip6_tnl_lookup(struct net * net,const struct in6_addr * remote,const struct in6_addr * local)181 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
182 {
183 	unsigned int hash = HASH(remote, local);
184 	struct ip6_tnl *t;
185 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
186 
187 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
188 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
189 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
190 		    (t->dev->flags & IFF_UP))
191 			return t;
192 	}
193 	t = rcu_dereference(ip6n->tnls_wc[0]);
194 	if (t && (t->dev->flags & IFF_UP))
195 		return t;
196 
197 	return NULL;
198 }
199 
200 /**
201  * ip6_tnl_bucket - get head of list matching given tunnel parameters
202  *   @p: parameters containing tunnel end-points
203  *
204  * Description:
205  *   ip6_tnl_bucket() returns the head of the list matching the
206  *   &struct in6_addr entries laddr and raddr in @p.
207  *
208  * Return: head of IPv6 tunnel list
209  **/
210 
211 static struct ip6_tnl __rcu **
ip6_tnl_bucket(struct ip6_tnl_net * ip6n,const struct __ip6_tnl_parm * p)212 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
213 {
214 	const struct in6_addr *remote = &p->raddr;
215 	const struct in6_addr *local = &p->laddr;
216 	unsigned int h = 0;
217 	int prio = 0;
218 
219 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
220 		prio = 1;
221 		h = HASH(remote, local);
222 	}
223 	return &ip6n->tnls[prio][h];
224 }
225 
226 /**
227  * ip6_tnl_link - add tunnel to hash table
228  *   @t: tunnel to be added
229  **/
230 
231 static void
ip6_tnl_link(struct ip6_tnl_net * ip6n,struct ip6_tnl * t)232 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
233 {
234 	struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
235 
236 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
237 	rcu_assign_pointer(*tp, t);
238 }
239 
240 /**
241  * ip6_tnl_unlink - remove tunnel from hash table
242  *   @t: tunnel to be removed
243  **/
244 
245 static void
ip6_tnl_unlink(struct ip6_tnl_net * ip6n,struct ip6_tnl * t)246 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
247 {
248 	struct ip6_tnl __rcu **tp;
249 	struct ip6_tnl *iter;
250 
251 	for (tp = ip6_tnl_bucket(ip6n, &t->parms);
252 	     (iter = rtnl_dereference(*tp)) != NULL;
253 	     tp = &iter->next) {
254 		if (t == iter) {
255 			rcu_assign_pointer(*tp, t->next);
256 			break;
257 		}
258 	}
259 }
260 
ip6_dev_free(struct net_device * dev)261 static void ip6_dev_free(struct net_device *dev)
262 {
263 	free_percpu(dev->tstats);
264 	free_netdev(dev);
265 }
266 
ip6_tnl_create2(struct net_device * dev)267 static int ip6_tnl_create2(struct net_device *dev)
268 {
269 	struct ip6_tnl *t = netdev_priv(dev);
270 	struct net *net = dev_net(dev);
271 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
272 	int err;
273 
274 	t = netdev_priv(dev);
275 
276 	dev->rtnl_link_ops = &ip6_link_ops;
277 	err = register_netdevice(dev);
278 	if (err < 0)
279 		goto out;
280 
281 	strcpy(t->parms.name, dev->name);
282 
283 	dev_hold(dev);
284 	ip6_tnl_link(ip6n, t);
285 	return 0;
286 
287 out:
288 	return err;
289 }
290 
291 /**
292  * ip6_tnl_create - create a new tunnel
293  *   @p: tunnel parameters
294  *   @pt: pointer to new tunnel
295  *
296  * Description:
297  *   Create tunnel matching given parameters.
298  *
299  * Return:
300  *   created tunnel or NULL
301  **/
302 
ip6_tnl_create(struct net * net,struct __ip6_tnl_parm * p)303 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
304 {
305 	struct net_device *dev;
306 	struct ip6_tnl *t;
307 	char name[IFNAMSIZ];
308 	int err;
309 
310 	if (p->name[0])
311 		strlcpy(name, p->name, IFNAMSIZ);
312 	else
313 		sprintf(name, "ip6tnl%%d");
314 
315 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
316 			   ip6_tnl_dev_setup);
317 	if (dev == NULL)
318 		goto failed;
319 
320 	dev_net_set(dev, net);
321 
322 	t = netdev_priv(dev);
323 	t->parms = *p;
324 	t->net = dev_net(dev);
325 	err = ip6_tnl_create2(dev);
326 	if (err < 0)
327 		goto failed_free;
328 
329 	return t;
330 
331 failed_free:
332 	ip6_dev_free(dev);
333 failed:
334 	return NULL;
335 }
336 
337 /**
338  * ip6_tnl_locate - find or create tunnel matching given parameters
339  *   @p: tunnel parameters
340  *   @create: != 0 if allowed to create new tunnel if no match found
341  *
342  * Description:
343  *   ip6_tnl_locate() first tries to locate an existing tunnel
344  *   based on @parms. If this is unsuccessful, but @create is set a new
345  *   tunnel device is created and registered for use.
346  *
347  * Return:
348  *   matching tunnel or NULL
349  **/
350 
ip6_tnl_locate(struct net * net,struct __ip6_tnl_parm * p,int create)351 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
352 		struct __ip6_tnl_parm *p, int create)
353 {
354 	const struct in6_addr *remote = &p->raddr;
355 	const struct in6_addr *local = &p->laddr;
356 	struct ip6_tnl __rcu **tp;
357 	struct ip6_tnl *t;
358 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
359 
360 	for (tp = ip6_tnl_bucket(ip6n, p);
361 	     (t = rtnl_dereference(*tp)) != NULL;
362 	     tp = &t->next) {
363 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
364 		    ipv6_addr_equal(remote, &t->parms.raddr)) {
365 			if (create)
366 				return NULL;
367 
368 			return t;
369 		}
370 	}
371 	if (!create)
372 		return NULL;
373 	return ip6_tnl_create(net, p);
374 }
375 
376 /**
377  * ip6_tnl_dev_uninit - tunnel device uninitializer
378  *   @dev: the device to be destroyed
379  *
380  * Description:
381  *   ip6_tnl_dev_uninit() removes tunnel from its list
382  **/
383 
384 static void
ip6_tnl_dev_uninit(struct net_device * dev)385 ip6_tnl_dev_uninit(struct net_device *dev)
386 {
387 	struct ip6_tnl *t = netdev_priv(dev);
388 	struct net *net = t->net;
389 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
390 
391 	if (dev == ip6n->fb_tnl_dev)
392 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
393 	else
394 		ip6_tnl_unlink(ip6n, t);
395 	ip6_tnl_dst_reset(t);
396 	dev_put(dev);
397 }
398 
399 /**
400  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
401  *   @skb: received socket buffer
402  *
403  * Return:
404  *   0 if none was found,
405  *   else index to encapsulation limit
406  **/
407 
ip6_tnl_parse_tlv_enc_lim(struct sk_buff * skb,__u8 * raw)408 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
409 {
410 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
411 	unsigned int nhoff = raw - skb->data;
412 	unsigned int off = nhoff + sizeof(*ipv6h);
413 	u8 next, nexthdr = ipv6h->nexthdr;
414 
415 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
416 		struct ipv6_opt_hdr *hdr;
417 		u16 optlen;
418 
419 		if (!pskb_may_pull(skb, off + sizeof(*hdr)))
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 			if (frag_hdr->frag_off)
426 				break;
427 			optlen = 8;
428 		} else if (nexthdr == NEXTHDR_AUTH) {
429 			optlen = (hdr->hdrlen + 2) << 2;
430 		} else {
431 			optlen = ipv6_optlen(hdr);
432 		}
433 		/* cache hdr->nexthdr, since pskb_may_pull() might
434 		 * invalidate hdr
435 		 */
436 		next = hdr->nexthdr;
437 		if (nexthdr == NEXTHDR_DEST) {
438 			u16 i = 2;
439 
440 			/* Remember : hdr is no longer valid at this point. */
441 			if (!pskb_may_pull(skb, off + optlen))
442 				break;
443 
444 			while (1) {
445 				struct ipv6_tlv_tnl_enc_lim *tel;
446 
447 				/* No more room for encapsulation limit */
448 				if (i + sizeof(*tel) > optlen)
449 					break;
450 
451 				tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
452 				/* return index of option if found and valid */
453 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
454 				    tel->length == 1)
455 					return i + off - nhoff;
456 				/* else jump to next option */
457 				if (tel->type)
458 					i += tel->length + 2;
459 				else
460 					i++;
461 			}
462 		}
463 		nexthdr = next;
464 		off += optlen;
465 	}
466 	return 0;
467 }
468 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
469 
470 /**
471  * ip6_tnl_err - tunnel error handler
472  *
473  * Description:
474  *   ip6_tnl_err() should handle errors in the tunnel according
475  *   to the specifications in RFC 2473.
476  **/
477 
478 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)479 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
480 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
481 {
482 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
483 	struct ip6_tnl *t;
484 	int rel_msg = 0;
485 	u8 rel_type = ICMPV6_DEST_UNREACH;
486 	u8 rel_code = ICMPV6_ADDR_UNREACH;
487 	__u32 rel_info = 0;
488 	__u16 len;
489 	int err = -ENOENT;
490 
491 	/* If the packet doesn't contain the original IPv6 header we are
492 	   in trouble since we might need the source address for further
493 	   processing of the error. */
494 
495 	rcu_read_lock();
496 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
497 					&ipv6h->saddr)) == NULL)
498 		goto out;
499 
500 	if (t->parms.proto != ipproto && t->parms.proto != 0)
501 		goto out;
502 
503 	err = 0;
504 
505 	switch (*type) {
506 		__u32 teli;
507 		struct ipv6_tlv_tnl_enc_lim *tel;
508 		__u32 mtu;
509 	case ICMPV6_DEST_UNREACH:
510 		net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
511 				     t->parms.name);
512 		rel_msg = 1;
513 		break;
514 	case ICMPV6_TIME_EXCEED:
515 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
516 			net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
517 					     t->parms.name);
518 			rel_msg = 1;
519 		}
520 		break;
521 	case ICMPV6_PARAMPROB:
522 		teli = 0;
523 		if ((*code) == ICMPV6_HDR_FIELD)
524 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
525 
526 		if (teli && teli == *info - 2) {
527 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
528 			if (tel->encap_limit == 0) {
529 				net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
530 						     t->parms.name);
531 				rel_msg = 1;
532 			}
533 		} else {
534 			net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
535 					     t->parms.name);
536 		}
537 		break;
538 	case ICMPV6_PKT_TOOBIG:
539 		mtu = *info - offset;
540 		if (mtu < IPV6_MIN_MTU)
541 			mtu = IPV6_MIN_MTU;
542 		t->dev->mtu = mtu;
543 
544 		if ((len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
545 			rel_type = ICMPV6_PKT_TOOBIG;
546 			rel_code = 0;
547 			rel_info = mtu;
548 			rel_msg = 1;
549 		}
550 		break;
551 	}
552 
553 	*type = rel_type;
554 	*code = rel_code;
555 	*info = rel_info;
556 	*msg = rel_msg;
557 
558 out:
559 	rcu_read_unlock();
560 	return err;
561 }
562 
563 static int
ip4ip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)564 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
565 	   u8 type, u8 code, int offset, __be32 info)
566 {
567 	int rel_msg = 0;
568 	u8 rel_type = type;
569 	u8 rel_code = code;
570 	__u32 rel_info = ntohl(info);
571 	int err;
572 	struct sk_buff *skb2;
573 	const struct iphdr *eiph;
574 	struct rtable *rt;
575 	struct flowi4 fl4;
576 
577 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
578 			  &rel_msg, &rel_info, offset);
579 	if (err < 0)
580 		return err;
581 
582 	if (rel_msg == 0)
583 		return 0;
584 
585 	switch (rel_type) {
586 	case ICMPV6_DEST_UNREACH:
587 		if (rel_code != ICMPV6_ADDR_UNREACH)
588 			return 0;
589 		rel_type = ICMP_DEST_UNREACH;
590 		rel_code = ICMP_HOST_UNREACH;
591 		break;
592 	case ICMPV6_PKT_TOOBIG:
593 		if (rel_code != 0)
594 			return 0;
595 		rel_type = ICMP_DEST_UNREACH;
596 		rel_code = ICMP_FRAG_NEEDED;
597 		break;
598 	case NDISC_REDIRECT:
599 		rel_type = ICMP_REDIRECT;
600 		rel_code = ICMP_REDIR_HOST;
601 	default:
602 		return 0;
603 	}
604 
605 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
606 		return 0;
607 
608 	skb2 = skb_clone(skb, GFP_ATOMIC);
609 	if (!skb2)
610 		return 0;
611 
612 	skb_dst_drop(skb2);
613 
614 	skb_pull(skb2, offset);
615 	skb_reset_network_header(skb2);
616 	eiph = ip_hdr(skb2);
617 
618 	/* Try to guess incoming interface */
619 	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
620 				   eiph->saddr, 0,
621 				   0, 0,
622 				   IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
623 	if (IS_ERR(rt))
624 		goto out;
625 
626 	skb2->dev = rt->dst.dev;
627 
628 	/* route "incoming" packet */
629 	if (rt->rt_flags & RTCF_LOCAL) {
630 		ip_rt_put(rt);
631 		rt = NULL;
632 		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
633 					   eiph->daddr, eiph->saddr,
634 					   0, 0,
635 					   IPPROTO_IPIP,
636 					   RT_TOS(eiph->tos), 0);
637 		if (IS_ERR(rt) ||
638 		    rt->dst.dev->type != ARPHRD_TUNNEL) {
639 			if (!IS_ERR(rt))
640 				ip_rt_put(rt);
641 			goto out;
642 		}
643 		skb_dst_set(skb2, &rt->dst);
644 	} else {
645 		ip_rt_put(rt);
646 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
647 				   skb2->dev) ||
648 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
649 			goto out;
650 	}
651 
652 	/* change mtu on this route */
653 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
654 		if (rel_info > dst_mtu(skb_dst(skb2)))
655 			goto out;
656 
657 		skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
658 	}
659 	if (rel_type == ICMP_REDIRECT)
660 		skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
661 
662 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
663 
664 out:
665 	kfree_skb(skb2);
666 	return 0;
667 }
668 
669 static int
ip6ip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)670 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
671 	   u8 type, u8 code, int offset, __be32 info)
672 {
673 	int rel_msg = 0;
674 	u8 rel_type = type;
675 	u8 rel_code = code;
676 	__u32 rel_info = ntohl(info);
677 	int err;
678 
679 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
680 			  &rel_msg, &rel_info, offset);
681 	if (err < 0)
682 		return err;
683 
684 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
685 		struct rt6_info *rt;
686 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
687 
688 		if (!skb2)
689 			return 0;
690 
691 		skb_dst_drop(skb2);
692 		skb_pull(skb2, offset);
693 		skb_reset_network_header(skb2);
694 
695 		/* Try to guess incoming interface */
696 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
697 				NULL, 0, 0);
698 
699 		if (rt && rt->dst.dev)
700 			skb2->dev = rt->dst.dev;
701 
702 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
703 
704 		ip6_rt_put(rt);
705 
706 		kfree_skb(skb2);
707 	}
708 
709 	return 0;
710 }
711 
ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)712 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
713 				       const struct ipv6hdr *ipv6h,
714 				       struct sk_buff *skb)
715 {
716 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
717 
718 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
719 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
720 
721 	return IP6_ECN_decapsulate(ipv6h, skb);
722 }
723 
ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)724 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
725 				       const struct ipv6hdr *ipv6h,
726 				       struct sk_buff *skb)
727 {
728 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
729 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
730 
731 	return IP6_ECN_decapsulate(ipv6h, skb);
732 }
733 
ip6_tnl_get_cap(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)734 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
735 			     const struct in6_addr *laddr,
736 			     const struct in6_addr *raddr)
737 {
738 	struct __ip6_tnl_parm *p = &t->parms;
739 	int ltype = ipv6_addr_type(laddr);
740 	int rtype = ipv6_addr_type(raddr);
741 	__u32 flags = 0;
742 
743 	if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
744 		flags = IP6_TNL_F_CAP_PER_PACKET;
745 	} else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
746 		   rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
747 		   !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
748 		   (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
749 		if (ltype&IPV6_ADDR_UNICAST)
750 			flags |= IP6_TNL_F_CAP_XMIT;
751 		if (rtype&IPV6_ADDR_UNICAST)
752 			flags |= IP6_TNL_F_CAP_RCV;
753 	}
754 	return flags;
755 }
756 EXPORT_SYMBOL(ip6_tnl_get_cap);
757 
758 /* called with rcu_read_lock() */
ip6_tnl_rcv_ctl(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)759 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
760 				  const struct in6_addr *laddr,
761 				  const struct in6_addr *raddr)
762 {
763 	struct __ip6_tnl_parm *p = &t->parms;
764 	int ret = 0;
765 	struct net *net = t->net;
766 
767 	if ((p->flags & IP6_TNL_F_CAP_RCV) ||
768 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
769 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
770 		struct net_device *ldev = NULL;
771 
772 		if (p->link)
773 			ldev = dev_get_by_index_rcu(net, p->link);
774 
775 		if ((ipv6_addr_is_multicast(laddr) ||
776 		     likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
777 		    likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
778 			ret = 1;
779 	}
780 	return ret;
781 }
782 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
783 
784 /**
785  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
786  *   @skb: received socket buffer
787  *   @protocol: ethernet protocol ID
788  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
789  *
790  * Return: 0
791  **/
792 
ip6_tnl_rcv(struct sk_buff * skb,__u16 protocol,__u8 ipproto,int (* dscp_ecn_decapsulate)(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb))793 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
794 		       __u8 ipproto,
795 		       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
796 						   const struct ipv6hdr *ipv6h,
797 						   struct sk_buff *skb))
798 {
799 	struct ip6_tnl *t;
800 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
801 	int err;
802 
803 	rcu_read_lock();
804 
805 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
806 					&ipv6h->daddr)) != NULL) {
807 		struct pcpu_sw_netstats *tstats;
808 
809 		if (t->parms.proto != ipproto && t->parms.proto != 0) {
810 			rcu_read_unlock();
811 			goto discard;
812 		}
813 
814 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
815 			rcu_read_unlock();
816 			goto discard;
817 		}
818 
819 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
820 			t->dev->stats.rx_dropped++;
821 			rcu_read_unlock();
822 			goto discard;
823 		}
824 		skb->mac_header = skb->network_header;
825 		skb_reset_network_header(skb);
826 		skb->protocol = htons(protocol);
827 		memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
828 
829 		__skb_tunnel_rx(skb, t->dev, t->net);
830 
831 		err = dscp_ecn_decapsulate(t, ipv6h, skb);
832 		if (unlikely(err)) {
833 			if (log_ecn_error)
834 				net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
835 						     &ipv6h->saddr,
836 						     ipv6_get_dsfield(ipv6h));
837 			if (err > 1) {
838 				++t->dev->stats.rx_frame_errors;
839 				++t->dev->stats.rx_errors;
840 				rcu_read_unlock();
841 				goto discard;
842 			}
843 		}
844 
845 		tstats = this_cpu_ptr(t->dev->tstats);
846 		u64_stats_update_begin(&tstats->syncp);
847 		tstats->rx_packets++;
848 		tstats->rx_bytes += skb->len;
849 		u64_stats_update_end(&tstats->syncp);
850 
851 		netif_rx(skb);
852 
853 		rcu_read_unlock();
854 		return 0;
855 	}
856 	rcu_read_unlock();
857 	return 1;
858 
859 discard:
860 	kfree_skb(skb);
861 	return 0;
862 }
863 
ip4ip6_rcv(struct sk_buff * skb)864 static int ip4ip6_rcv(struct sk_buff *skb)
865 {
866 	return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
867 			   ip4ip6_dscp_ecn_decapsulate);
868 }
869 
ip6ip6_rcv(struct sk_buff * skb)870 static int ip6ip6_rcv(struct sk_buff *skb)
871 {
872 	return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
873 			   ip6ip6_dscp_ecn_decapsulate);
874 }
875 
876 struct ipv6_tel_txoption {
877 	struct ipv6_txoptions ops;
878 	__u8 dst_opt[8];
879 };
880 
init_tel_txopt(struct ipv6_tel_txoption * opt,__u8 encap_limit)881 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
882 {
883 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
884 
885 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
886 	opt->dst_opt[3] = 1;
887 	opt->dst_opt[4] = encap_limit;
888 	opt->dst_opt[5] = IPV6_TLV_PADN;
889 	opt->dst_opt[6] = 1;
890 
891 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
892 	opt->ops.opt_nflen = 8;
893 }
894 
895 /**
896  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
897  *   @t: the outgoing tunnel device
898  *   @hdr: IPv6 header from the incoming packet
899  *
900  * Description:
901  *   Avoid trivial tunneling loop by checking that tunnel exit-point
902  *   doesn't match source of incoming packet.
903  *
904  * Return:
905  *   1 if conflict,
906  *   0 else
907  **/
908 
909 static inline bool
ip6_tnl_addr_conflict(const struct ip6_tnl * t,const struct ipv6hdr * hdr)910 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
911 {
912 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
913 }
914 
ip6_tnl_xmit_ctl(struct ip6_tnl * t)915 int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
916 {
917 	struct __ip6_tnl_parm *p = &t->parms;
918 	int ret = 0;
919 	struct net *net = t->net;
920 
921 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
922 		struct net_device *ldev = NULL;
923 
924 		rcu_read_lock();
925 		if (p->link)
926 			ldev = dev_get_by_index_rcu(net, p->link);
927 
928 		if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
929 			pr_warn("%s xmit: Local address not yet configured!\n",
930 				p->name);
931 		else if (!ipv6_addr_is_multicast(&p->raddr) &&
932 			 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
933 			pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
934 				p->name);
935 		else
936 			ret = 1;
937 		rcu_read_unlock();
938 	}
939 	return ret;
940 }
941 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
942 
943 /**
944  * ip6_tnl_xmit2 - encapsulate packet and send
945  *   @skb: the outgoing socket buffer
946  *   @dev: the outgoing tunnel device
947  *   @dsfield: dscp code for outer header
948  *   @fl: flow of tunneled packet
949  *   @encap_limit: encapsulation limit
950  *   @pmtu: Path MTU is stored if packet is too big
951  *
952  * Description:
953  *   Build new header and do some sanity checks on the packet before sending
954  *   it.
955  *
956  * Return:
957  *   0 on success
958  *   -1 fail
959  *   %-EMSGSIZE message too big. return mtu in this case.
960  **/
961 
ip6_tnl_xmit2(struct sk_buff * skb,struct net_device * dev,__u8 dsfield,struct flowi6 * fl6,int encap_limit,__u32 * pmtu)962 static int ip6_tnl_xmit2(struct sk_buff *skb,
963 			 struct net_device *dev,
964 			 __u8 dsfield,
965 			 struct flowi6 *fl6,
966 			 int encap_limit,
967 			 __u32 *pmtu)
968 {
969 	struct ip6_tnl *t = netdev_priv(dev);
970 	struct net *net = t->net;
971 	struct net_device_stats *stats = &t->dev->stats;
972 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
973 	struct ipv6_tel_txoption opt;
974 	struct dst_entry *dst = NULL, *ndst = NULL;
975 	struct net_device *tdev;
976 	int mtu;
977 	unsigned int max_headroom = sizeof(struct ipv6hdr);
978 	u8 proto;
979 	int err = -1;
980 
981 	if (!fl6->flowi6_mark)
982 		dst = ip6_tnl_dst_check(t);
983 	if (!dst) {
984 		ndst = ip6_route_output(net, NULL, fl6);
985 
986 		if (ndst->error)
987 			goto tx_err_link_failure;
988 		ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
989 		if (IS_ERR(ndst)) {
990 			err = PTR_ERR(ndst);
991 			ndst = NULL;
992 			goto tx_err_link_failure;
993 		}
994 		dst = ndst;
995 	}
996 
997 	tdev = dst->dev;
998 
999 	if (tdev == dev) {
1000 		stats->collisions++;
1001 		net_warn_ratelimited("%s: Local routing loop detected!\n",
1002 				     t->parms.name);
1003 		goto tx_err_dst_release;
1004 	}
1005 	mtu = dst_mtu(dst) - sizeof(*ipv6h);
1006 	if (encap_limit >= 0) {
1007 		max_headroom += 8;
1008 		mtu -= 8;
1009 	}
1010 	if (mtu < IPV6_MIN_MTU)
1011 		mtu = IPV6_MIN_MTU;
1012 	if (skb_dst(skb))
1013 		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1014 	if (skb->len > mtu) {
1015 		*pmtu = mtu;
1016 		err = -EMSGSIZE;
1017 		goto tx_err_dst_release;
1018 	}
1019 
1020 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1021 
1022 	/*
1023 	 * Okay, now see if we can stuff it in the buffer as-is.
1024 	 */
1025 	max_headroom += LL_RESERVED_SPACE(tdev);
1026 
1027 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1028 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1029 		struct sk_buff *new_skb;
1030 
1031 		if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
1032 			goto tx_err_dst_release;
1033 
1034 		if (skb->sk)
1035 			skb_set_owner_w(new_skb, skb->sk);
1036 		consume_skb(skb);
1037 		skb = new_skb;
1038 	}
1039 	if (fl6->flowi6_mark) {
1040 		skb_dst_set(skb, dst);
1041 		ndst = NULL;
1042 	} else {
1043 		skb_dst_set_noref(skb, dst);
1044 	}
1045 	skb->transport_header = skb->network_header;
1046 
1047 	proto = fl6->flowi6_proto;
1048 	if (encap_limit >= 0) {
1049 		init_tel_txopt(&opt, encap_limit);
1050 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1051 	}
1052 
1053 	if (likely(!skb->encapsulation)) {
1054 		skb_reset_inner_headers(skb);
1055 		skb->encapsulation = 1;
1056 	}
1057 
1058 	skb_push(skb, sizeof(struct ipv6hdr));
1059 	skb_reset_network_header(skb);
1060 	ipv6h = ipv6_hdr(skb);
1061 	ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1062 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, false));
1063 	ipv6h->hop_limit = t->parms.hop_limit;
1064 	ipv6h->nexthdr = proto;
1065 	ipv6h->saddr = fl6->saddr;
1066 	ipv6h->daddr = fl6->daddr;
1067 	ip6tunnel_xmit(skb, dev);
1068 	if (ndst)
1069 		ip6_tnl_dst_store(t, ndst);
1070 	return 0;
1071 tx_err_link_failure:
1072 	stats->tx_carrier_errors++;
1073 	dst_link_failure(skb);
1074 tx_err_dst_release:
1075 	dst_release(ndst);
1076 	return err;
1077 }
1078 
1079 static inline int
ip4ip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev)1080 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1081 {
1082 	struct ip6_tnl *t = netdev_priv(dev);
1083 	const struct iphdr  *iph = ip_hdr(skb);
1084 	int encap_limit = -1;
1085 	struct flowi6 fl6;
1086 	__u8 dsfield;
1087 	__u32 mtu;
1088 	int err;
1089 
1090 	if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1091 	    !ip6_tnl_xmit_ctl(t))
1092 		return -1;
1093 
1094 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1095 		encap_limit = t->parms.encap_limit;
1096 
1097 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1098 	fl6.flowi6_proto = IPPROTO_IPIP;
1099 
1100 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1101 
1102 	dsfield = ipv4_get_dsfield(iph);
1103 
1104 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1105 		fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1106 					  & IPV6_TCLASS_MASK;
1107 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1108 		fl6.flowi6_mark = skb->mark;
1109 
1110 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1111 	if (err != 0) {
1112 		/* XXX: send ICMP error even if DF is not set. */
1113 		if (err == -EMSGSIZE)
1114 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1115 				  htonl(mtu));
1116 		return -1;
1117 	}
1118 
1119 	return 0;
1120 }
1121 
1122 static inline int
ip6ip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev)1123 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1124 {
1125 	struct ip6_tnl *t = netdev_priv(dev);
1126 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1127 	int encap_limit = -1;
1128 	__u16 offset;
1129 	struct flowi6 fl6;
1130 	__u8 dsfield;
1131 	__u32 mtu;
1132 	int err;
1133 
1134 	if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1135 	    !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1136 		return -1;
1137 
1138 	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1139 	if (offset > 0) {
1140 		struct ipv6_tlv_tnl_enc_lim *tel;
1141 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1142 		if (tel->encap_limit == 0) {
1143 			icmpv6_send(skb, ICMPV6_PARAMPROB,
1144 				    ICMPV6_HDR_FIELD, offset + 2);
1145 			return -1;
1146 		}
1147 		encap_limit = tel->encap_limit - 1;
1148 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1149 		encap_limit = t->parms.encap_limit;
1150 
1151 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1152 	fl6.flowi6_proto = IPPROTO_IPV6;
1153 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1154 
1155 	dsfield = ipv6_get_dsfield(ipv6h);
1156 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1157 		fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1158 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1159 		fl6.flowlabel |= ip6_flowlabel(ipv6h);
1160 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1161 		fl6.flowi6_mark = skb->mark;
1162 
1163 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1164 	if (err != 0) {
1165 		if (err == -EMSGSIZE)
1166 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1167 		return -1;
1168 	}
1169 
1170 	return 0;
1171 }
1172 
1173 static netdev_tx_t
ip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev)1174 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1175 {
1176 	struct ip6_tnl *t = netdev_priv(dev);
1177 	struct net_device_stats *stats = &t->dev->stats;
1178 	int ret;
1179 
1180 	switch (skb->protocol) {
1181 	case htons(ETH_P_IP):
1182 		ret = ip4ip6_tnl_xmit(skb, dev);
1183 		break;
1184 	case htons(ETH_P_IPV6):
1185 		ret = ip6ip6_tnl_xmit(skb, dev);
1186 		break;
1187 	default:
1188 		goto tx_err;
1189 	}
1190 
1191 	if (ret < 0)
1192 		goto tx_err;
1193 
1194 	return NETDEV_TX_OK;
1195 
1196 tx_err:
1197 	stats->tx_errors++;
1198 	stats->tx_dropped++;
1199 	kfree_skb(skb);
1200 	return NETDEV_TX_OK;
1201 }
1202 
ip6_tnl_link_config(struct ip6_tnl * t)1203 static void ip6_tnl_link_config(struct ip6_tnl *t)
1204 {
1205 	struct net_device *dev = t->dev;
1206 	struct __ip6_tnl_parm *p = &t->parms;
1207 	struct flowi6 *fl6 = &t->fl.u.ip6;
1208 
1209 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1210 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1211 
1212 	/* Set up flowi template */
1213 	fl6->saddr = p->laddr;
1214 	fl6->daddr = p->raddr;
1215 	fl6->flowi6_oif = p->link;
1216 	fl6->flowlabel = 0;
1217 
1218 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1219 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1220 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1221 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1222 
1223 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1224 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1225 
1226 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1227 		dev->flags |= IFF_POINTOPOINT;
1228 	else
1229 		dev->flags &= ~IFF_POINTOPOINT;
1230 
1231 	dev->iflink = p->link;
1232 
1233 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1234 		int strict = (ipv6_addr_type(&p->raddr) &
1235 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1236 
1237 		struct rt6_info *rt = rt6_lookup(t->net,
1238 						 &p->raddr, &p->laddr,
1239 						 p->link, strict);
1240 
1241 		if (rt == NULL)
1242 			return;
1243 
1244 		if (rt->dst.dev) {
1245 			dev->hard_header_len = rt->dst.dev->hard_header_len +
1246 				sizeof(struct ipv6hdr);
1247 
1248 			dev->mtu = rt->dst.dev->mtu - sizeof(struct ipv6hdr);
1249 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1250 				dev->mtu -= 8;
1251 
1252 			if (dev->mtu < IPV6_MIN_MTU)
1253 				dev->mtu = IPV6_MIN_MTU;
1254 		}
1255 		ip6_rt_put(rt);
1256 	}
1257 }
1258 
1259 /**
1260  * ip6_tnl_change - update the tunnel parameters
1261  *   @t: tunnel to be changed
1262  *   @p: tunnel configuration parameters
1263  *
1264  * Description:
1265  *   ip6_tnl_change() updates the tunnel parameters
1266  **/
1267 
1268 static int
ip6_tnl_change(struct ip6_tnl * t,const struct __ip6_tnl_parm * p)1269 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1270 {
1271 	t->parms.laddr = p->laddr;
1272 	t->parms.raddr = p->raddr;
1273 	t->parms.flags = p->flags;
1274 	t->parms.hop_limit = p->hop_limit;
1275 	t->parms.encap_limit = p->encap_limit;
1276 	t->parms.flowinfo = p->flowinfo;
1277 	t->parms.link = p->link;
1278 	t->parms.proto = p->proto;
1279 	ip6_tnl_dst_reset(t);
1280 	ip6_tnl_link_config(t);
1281 	return 0;
1282 }
1283 
ip6_tnl_update(struct ip6_tnl * t,struct __ip6_tnl_parm * p)1284 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1285 {
1286 	struct net *net = t->net;
1287 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1288 	int err;
1289 
1290 	ip6_tnl_unlink(ip6n, t);
1291 	synchronize_net();
1292 	err = ip6_tnl_change(t, p);
1293 	ip6_tnl_link(ip6n, t);
1294 	netdev_state_change(t->dev);
1295 	return err;
1296 }
1297 
1298 static void
ip6_tnl_parm_from_user(struct __ip6_tnl_parm * p,const struct ip6_tnl_parm * u)1299 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1300 {
1301 	p->laddr = u->laddr;
1302 	p->raddr = u->raddr;
1303 	p->flags = u->flags;
1304 	p->hop_limit = u->hop_limit;
1305 	p->encap_limit = u->encap_limit;
1306 	p->flowinfo = u->flowinfo;
1307 	p->link = u->link;
1308 	p->proto = u->proto;
1309 	memcpy(p->name, u->name, sizeof(u->name));
1310 }
1311 
1312 static void
ip6_tnl_parm_to_user(struct ip6_tnl_parm * u,const struct __ip6_tnl_parm * p)1313 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1314 {
1315 	u->laddr = p->laddr;
1316 	u->raddr = p->raddr;
1317 	u->flags = p->flags;
1318 	u->hop_limit = p->hop_limit;
1319 	u->encap_limit = p->encap_limit;
1320 	u->flowinfo = p->flowinfo;
1321 	u->link = p->link;
1322 	u->proto = p->proto;
1323 	memcpy(u->name, p->name, sizeof(u->name));
1324 }
1325 
1326 /**
1327  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1328  *   @dev: virtual device associated with tunnel
1329  *   @ifr: parameters passed from userspace
1330  *   @cmd: command to be performed
1331  *
1332  * Description:
1333  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1334  *   from userspace.
1335  *
1336  *   The possible commands are the following:
1337  *     %SIOCGETTUNNEL: get tunnel parameters for device
1338  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1339  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1340  *     %SIOCDELTUNNEL: delete tunnel
1341  *
1342  *   The fallback device "ip6tnl0", created during module
1343  *   initialization, can be used for creating other tunnel devices.
1344  *
1345  * Return:
1346  *   0 on success,
1347  *   %-EFAULT if unable to copy data to or from userspace,
1348  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1349  *   %-EINVAL if passed tunnel parameters are invalid,
1350  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1351  *   %-ENODEV if attempting to change or delete a nonexisting device
1352  **/
1353 
1354 static int
ip6_tnl_ioctl(struct net_device * dev,struct ifreq * ifr,int cmd)1355 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1356 {
1357 	int err = 0;
1358 	struct ip6_tnl_parm p;
1359 	struct __ip6_tnl_parm p1;
1360 	struct ip6_tnl *t = netdev_priv(dev);
1361 	struct net *net = t->net;
1362 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1363 
1364 	switch (cmd) {
1365 	case SIOCGETTUNNEL:
1366 		if (dev == ip6n->fb_tnl_dev) {
1367 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1368 				err = -EFAULT;
1369 				break;
1370 			}
1371 			ip6_tnl_parm_from_user(&p1, &p);
1372 			t = ip6_tnl_locate(net, &p1, 0);
1373 			if (t == NULL)
1374 				t = netdev_priv(dev);
1375 		} else {
1376 			memset(&p, 0, sizeof(p));
1377 		}
1378 		ip6_tnl_parm_to_user(&p, &t->parms);
1379 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1380 			err = -EFAULT;
1381 		}
1382 		break;
1383 	case SIOCADDTUNNEL:
1384 	case SIOCCHGTUNNEL:
1385 		err = -EPERM;
1386 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1387 			break;
1388 		err = -EFAULT;
1389 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1390 			break;
1391 		err = -EINVAL;
1392 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1393 		    p.proto != 0)
1394 			break;
1395 		ip6_tnl_parm_from_user(&p1, &p);
1396 		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1397 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1398 			if (t != NULL) {
1399 				if (t->dev != dev) {
1400 					err = -EEXIST;
1401 					break;
1402 				}
1403 			} else
1404 				t = netdev_priv(dev);
1405 
1406 			err = ip6_tnl_update(t, &p1);
1407 		}
1408 		if (t) {
1409 			err = 0;
1410 			ip6_tnl_parm_to_user(&p, &t->parms);
1411 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1412 				err = -EFAULT;
1413 
1414 		} else
1415 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1416 		break;
1417 	case SIOCDELTUNNEL:
1418 		err = -EPERM;
1419 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1420 			break;
1421 
1422 		if (dev == ip6n->fb_tnl_dev) {
1423 			err = -EFAULT;
1424 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1425 				break;
1426 			err = -ENOENT;
1427 			ip6_tnl_parm_from_user(&p1, &p);
1428 			t = ip6_tnl_locate(net, &p1, 0);
1429 			if (t == NULL)
1430 				break;
1431 			err = -EPERM;
1432 			if (t->dev == ip6n->fb_tnl_dev)
1433 				break;
1434 			dev = t->dev;
1435 		}
1436 		err = 0;
1437 		unregister_netdevice(dev);
1438 		break;
1439 	default:
1440 		err = -EINVAL;
1441 	}
1442 	return err;
1443 }
1444 
1445 /**
1446  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1447  *   @dev: virtual device associated with tunnel
1448  *   @new_mtu: the new mtu
1449  *
1450  * Return:
1451  *   0 on success,
1452  *   %-EINVAL if mtu too small
1453  **/
1454 
1455 static int
ip6_tnl_change_mtu(struct net_device * dev,int new_mtu)1456 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1457 {
1458 	struct ip6_tnl *tnl = netdev_priv(dev);
1459 
1460 	if (tnl->parms.proto == IPPROTO_IPIP) {
1461 		if (new_mtu < 68)
1462 			return -EINVAL;
1463 	} else {
1464 		if (new_mtu < IPV6_MIN_MTU)
1465 			return -EINVAL;
1466 	}
1467 	if (new_mtu > 0xFFF8 - dev->hard_header_len)
1468 		return -EINVAL;
1469 	dev->mtu = new_mtu;
1470 	return 0;
1471 }
1472 
1473 
1474 static const struct net_device_ops ip6_tnl_netdev_ops = {
1475 	.ndo_init	= ip6_tnl_dev_init,
1476 	.ndo_uninit	= ip6_tnl_dev_uninit,
1477 	.ndo_start_xmit = ip6_tnl_xmit,
1478 	.ndo_do_ioctl	= ip6_tnl_ioctl,
1479 	.ndo_change_mtu = ip6_tnl_change_mtu,
1480 	.ndo_get_stats	= ip6_get_stats,
1481 };
1482 
1483 
1484 /**
1485  * ip6_tnl_dev_setup - setup virtual tunnel device
1486  *   @dev: virtual device associated with tunnel
1487  *
1488  * Description:
1489  *   Initialize function pointers and device parameters
1490  **/
1491 
ip6_tnl_dev_setup(struct net_device * dev)1492 static void ip6_tnl_dev_setup(struct net_device *dev)
1493 {
1494 	struct ip6_tnl *t;
1495 
1496 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1497 	dev->destructor = ip6_dev_free;
1498 
1499 	dev->type = ARPHRD_TUNNEL6;
1500 	dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
1501 	dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr);
1502 	t = netdev_priv(dev);
1503 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1504 		dev->mtu -= 8;
1505 	dev->flags |= IFF_NOARP;
1506 	dev->addr_len = sizeof(struct in6_addr);
1507 	netif_keep_dst(dev);
1508 	/* This perm addr will be used as interface identifier by IPv6 */
1509 	dev->addr_assign_type = NET_ADDR_RANDOM;
1510 	eth_random_addr(dev->perm_addr);
1511 }
1512 
1513 
1514 /**
1515  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1516  *   @dev: virtual device associated with tunnel
1517  **/
1518 
1519 static inline int
ip6_tnl_dev_init_gen(struct net_device * dev)1520 ip6_tnl_dev_init_gen(struct net_device *dev)
1521 {
1522 	struct ip6_tnl *t = netdev_priv(dev);
1523 
1524 	t->dev = dev;
1525 	t->net = dev_net(dev);
1526 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1527 	if (!dev->tstats)
1528 		return -ENOMEM;
1529 	return 0;
1530 }
1531 
1532 /**
1533  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1534  *   @dev: virtual device associated with tunnel
1535  **/
1536 
ip6_tnl_dev_init(struct net_device * dev)1537 static int ip6_tnl_dev_init(struct net_device *dev)
1538 {
1539 	struct ip6_tnl *t = netdev_priv(dev);
1540 	int err = ip6_tnl_dev_init_gen(dev);
1541 
1542 	if (err)
1543 		return err;
1544 	ip6_tnl_link_config(t);
1545 	return 0;
1546 }
1547 
1548 /**
1549  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1550  *   @dev: fallback device
1551  *
1552  * Return: 0
1553  **/
1554 
ip6_fb_tnl_dev_init(struct net_device * dev)1555 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1556 {
1557 	struct ip6_tnl *t = netdev_priv(dev);
1558 	struct net *net = dev_net(dev);
1559 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1560 
1561 	t->parms.proto = IPPROTO_IPV6;
1562 	dev_hold(dev);
1563 
1564 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1565 	return 0;
1566 }
1567 
ip6_tnl_validate(struct nlattr * tb[],struct nlattr * data[])1568 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1569 {
1570 	u8 proto;
1571 
1572 	if (!data || !data[IFLA_IPTUN_PROTO])
1573 		return 0;
1574 
1575 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1576 	if (proto != IPPROTO_IPV6 &&
1577 	    proto != IPPROTO_IPIP &&
1578 	    proto != 0)
1579 		return -EINVAL;
1580 
1581 	return 0;
1582 }
1583 
ip6_tnl_netlink_parms(struct nlattr * data[],struct __ip6_tnl_parm * parms)1584 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1585 				  struct __ip6_tnl_parm *parms)
1586 {
1587 	memset(parms, 0, sizeof(*parms));
1588 
1589 	if (!data)
1590 		return;
1591 
1592 	if (data[IFLA_IPTUN_LINK])
1593 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1594 
1595 	if (data[IFLA_IPTUN_LOCAL])
1596 		nla_memcpy(&parms->laddr, data[IFLA_IPTUN_LOCAL],
1597 			   sizeof(struct in6_addr));
1598 
1599 	if (data[IFLA_IPTUN_REMOTE])
1600 		nla_memcpy(&parms->raddr, data[IFLA_IPTUN_REMOTE],
1601 			   sizeof(struct in6_addr));
1602 
1603 	if (data[IFLA_IPTUN_TTL])
1604 		parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1605 
1606 	if (data[IFLA_IPTUN_ENCAP_LIMIT])
1607 		parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1608 
1609 	if (data[IFLA_IPTUN_FLOWINFO])
1610 		parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1611 
1612 	if (data[IFLA_IPTUN_FLAGS])
1613 		parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1614 
1615 	if (data[IFLA_IPTUN_PROTO])
1616 		parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1617 }
1618 
ip6_tnl_newlink(struct net * src_net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[])1619 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1620 			   struct nlattr *tb[], struct nlattr *data[])
1621 {
1622 	struct net *net = dev_net(dev);
1623 	struct ip6_tnl *nt;
1624 
1625 	nt = netdev_priv(dev);
1626 	ip6_tnl_netlink_parms(data, &nt->parms);
1627 
1628 	if (ip6_tnl_locate(net, &nt->parms, 0))
1629 		return -EEXIST;
1630 
1631 	return ip6_tnl_create2(dev);
1632 }
1633 
ip6_tnl_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[])1634 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1635 			      struct nlattr *data[])
1636 {
1637 	struct ip6_tnl *t = netdev_priv(dev);
1638 	struct __ip6_tnl_parm p;
1639 	struct net *net = t->net;
1640 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1641 
1642 	if (dev == ip6n->fb_tnl_dev)
1643 		return -EINVAL;
1644 
1645 	ip6_tnl_netlink_parms(data, &p);
1646 
1647 	t = ip6_tnl_locate(net, &p, 0);
1648 
1649 	if (t) {
1650 		if (t->dev != dev)
1651 			return -EEXIST;
1652 	} else
1653 		t = netdev_priv(dev);
1654 
1655 	return ip6_tnl_update(t, &p);
1656 }
1657 
ip6_tnl_dellink(struct net_device * dev,struct list_head * head)1658 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1659 {
1660 	struct net *net = dev_net(dev);
1661 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1662 
1663 	if (dev != ip6n->fb_tnl_dev)
1664 		unregister_netdevice_queue(dev, head);
1665 }
1666 
ip6_tnl_get_size(const struct net_device * dev)1667 static size_t ip6_tnl_get_size(const struct net_device *dev)
1668 {
1669 	return
1670 		/* IFLA_IPTUN_LINK */
1671 		nla_total_size(4) +
1672 		/* IFLA_IPTUN_LOCAL */
1673 		nla_total_size(sizeof(struct in6_addr)) +
1674 		/* IFLA_IPTUN_REMOTE */
1675 		nla_total_size(sizeof(struct in6_addr)) +
1676 		/* IFLA_IPTUN_TTL */
1677 		nla_total_size(1) +
1678 		/* IFLA_IPTUN_ENCAP_LIMIT */
1679 		nla_total_size(1) +
1680 		/* IFLA_IPTUN_FLOWINFO */
1681 		nla_total_size(4) +
1682 		/* IFLA_IPTUN_FLAGS */
1683 		nla_total_size(4) +
1684 		/* IFLA_IPTUN_PROTO */
1685 		nla_total_size(1) +
1686 		0;
1687 }
1688 
ip6_tnl_fill_info(struct sk_buff * skb,const struct net_device * dev)1689 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1690 {
1691 	struct ip6_tnl *tunnel = netdev_priv(dev);
1692 	struct __ip6_tnl_parm *parm = &tunnel->parms;
1693 
1694 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1695 	    nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr),
1696 		    &parm->laddr) ||
1697 	    nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr),
1698 		    &parm->raddr) ||
1699 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1700 	    nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1701 	    nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1702 	    nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1703 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1704 		goto nla_put_failure;
1705 	return 0;
1706 
1707 nla_put_failure:
1708 	return -EMSGSIZE;
1709 }
1710 
1711 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1712 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1713 	[IFLA_IPTUN_LOCAL]		= { .len = sizeof(struct in6_addr) },
1714 	[IFLA_IPTUN_REMOTE]		= { .len = sizeof(struct in6_addr) },
1715 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1716 	[IFLA_IPTUN_ENCAP_LIMIT]	= { .type = NLA_U8 },
1717 	[IFLA_IPTUN_FLOWINFO]		= { .type = NLA_U32 },
1718 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U32 },
1719 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1720 };
1721 
1722 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1723 	.kind		= "ip6tnl",
1724 	.maxtype	= IFLA_IPTUN_MAX,
1725 	.policy		= ip6_tnl_policy,
1726 	.priv_size	= sizeof(struct ip6_tnl),
1727 	.setup		= ip6_tnl_dev_setup,
1728 	.validate	= ip6_tnl_validate,
1729 	.newlink	= ip6_tnl_newlink,
1730 	.changelink	= ip6_tnl_changelink,
1731 	.dellink	= ip6_tnl_dellink,
1732 	.get_size	= ip6_tnl_get_size,
1733 	.fill_info	= ip6_tnl_fill_info,
1734 };
1735 
1736 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1737 	.handler	= ip4ip6_rcv,
1738 	.err_handler	= ip4ip6_err,
1739 	.priority	=	1,
1740 };
1741 
1742 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1743 	.handler	= ip6ip6_rcv,
1744 	.err_handler	= ip6ip6_err,
1745 	.priority	=	1,
1746 };
1747 
ip6_tnl_destroy_tunnels(struct net * net)1748 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
1749 {
1750 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1751 	struct net_device *dev, *aux;
1752 	int h;
1753 	struct ip6_tnl *t;
1754 	LIST_HEAD(list);
1755 
1756 	for_each_netdev_safe(net, dev, aux)
1757 		if (dev->rtnl_link_ops == &ip6_link_ops)
1758 			unregister_netdevice_queue(dev, &list);
1759 
1760 	for (h = 0; h < HASH_SIZE; h++) {
1761 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1762 		while (t != NULL) {
1763 			/* If dev is in the same netns, it has already
1764 			 * been added to the list by the previous loop.
1765 			 */
1766 			if (!net_eq(dev_net(t->dev), net))
1767 				unregister_netdevice_queue(t->dev, &list);
1768 			t = rtnl_dereference(t->next);
1769 		}
1770 	}
1771 
1772 	unregister_netdevice_many(&list);
1773 }
1774 
ip6_tnl_init_net(struct net * net)1775 static int __net_init ip6_tnl_init_net(struct net *net)
1776 {
1777 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1778 	struct ip6_tnl *t = NULL;
1779 	int err;
1780 
1781 	ip6n->tnls[0] = ip6n->tnls_wc;
1782 	ip6n->tnls[1] = ip6n->tnls_r_l;
1783 
1784 	err = -ENOMEM;
1785 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1786 					NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
1787 
1788 	if (!ip6n->fb_tnl_dev)
1789 		goto err_alloc_dev;
1790 	dev_net_set(ip6n->fb_tnl_dev, net);
1791 	ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1792 	/* FB netdevice is special: we have one, and only one per netns.
1793 	 * Allowing to move it to another netns is clearly unsafe.
1794 	 */
1795 	ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1796 
1797 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1798 	if (err < 0)
1799 		goto err_register;
1800 
1801 	err = register_netdev(ip6n->fb_tnl_dev);
1802 	if (err < 0)
1803 		goto err_register;
1804 
1805 	t = netdev_priv(ip6n->fb_tnl_dev);
1806 
1807 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1808 	return 0;
1809 
1810 err_register:
1811 	ip6_dev_free(ip6n->fb_tnl_dev);
1812 err_alloc_dev:
1813 	return err;
1814 }
1815 
ip6_tnl_exit_net(struct net * net)1816 static void __net_exit ip6_tnl_exit_net(struct net *net)
1817 {
1818 	rtnl_lock();
1819 	ip6_tnl_destroy_tunnels(net);
1820 	rtnl_unlock();
1821 }
1822 
1823 static struct pernet_operations ip6_tnl_net_ops = {
1824 	.init = ip6_tnl_init_net,
1825 	.exit = ip6_tnl_exit_net,
1826 	.id   = &ip6_tnl_net_id,
1827 	.size = sizeof(struct ip6_tnl_net),
1828 };
1829 
1830 /**
1831  * ip6_tunnel_init - register protocol and reserve needed resources
1832  *
1833  * Return: 0 on success
1834  **/
1835 
ip6_tunnel_init(void)1836 static int __init ip6_tunnel_init(void)
1837 {
1838 	int  err;
1839 
1840 	err = register_pernet_device(&ip6_tnl_net_ops);
1841 	if (err < 0)
1842 		goto out_pernet;
1843 
1844 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1845 	if (err < 0) {
1846 		pr_err("%s: can't register ip4ip6\n", __func__);
1847 		goto out_ip4ip6;
1848 	}
1849 
1850 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1851 	if (err < 0) {
1852 		pr_err("%s: can't register ip6ip6\n", __func__);
1853 		goto out_ip6ip6;
1854 	}
1855 	err = rtnl_link_register(&ip6_link_ops);
1856 	if (err < 0)
1857 		goto rtnl_link_failed;
1858 
1859 	return 0;
1860 
1861 rtnl_link_failed:
1862 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1863 out_ip6ip6:
1864 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1865 out_ip4ip6:
1866 	unregister_pernet_device(&ip6_tnl_net_ops);
1867 out_pernet:
1868 	return err;
1869 }
1870 
1871 /**
1872  * ip6_tunnel_cleanup - free resources and unregister protocol
1873  **/
1874 
ip6_tunnel_cleanup(void)1875 static void __exit ip6_tunnel_cleanup(void)
1876 {
1877 	rtnl_link_unregister(&ip6_link_ops);
1878 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1879 		pr_info("%s: can't deregister ip4ip6\n", __func__);
1880 
1881 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1882 		pr_info("%s: can't deregister ip6ip6\n", __func__);
1883 
1884 	unregister_pernet_device(&ip6_tnl_net_ops);
1885 }
1886 
1887 module_init(ip6_tunnel_init);
1888 module_exit(ip6_tunnel_cleanup);
1889