• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IPv6 tunneling device
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Ville Nuorvala		<vnuorval@tcs.hut.fi>
8  *	Yasuyuki Kozakai	<kozakai@linux-ipv6.org>
9  *
10  *      Based on:
11  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
12  *
13  *      RFC 2473
14  */
15 
16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
17 
18 #include <linux/module.h>
19 #include <linux/capability.h>
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/sockios.h>
23 #include <linux/icmp.h>
24 #include <linux/if.h>
25 #include <linux/in.h>
26 #include <linux/ip.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmpv6.h>
32 #include <linux/init.h>
33 #include <linux/route.h>
34 #include <linux/rtnetlink.h>
35 #include <linux/netfilter_ipv6.h>
36 #include <linux/slab.h>
37 #include <linux/hash.h>
38 #include <linux/etherdevice.h>
39 
40 #include <linux/uaccess.h>
41 #include <linux/atomic.h>
42 
43 #include <net/icmp.h>
44 #include <net/ip.h>
45 #include <net/ip_tunnels.h>
46 #include <net/ipv6.h>
47 #include <net/ip6_route.h>
48 #include <net/addrconf.h>
49 #include <net/ip6_tunnel.h>
50 #include <net/xfrm.h>
51 #include <net/dsfield.h>
52 #include <net/inet_ecn.h>
53 #include <net/net_namespace.h>
54 #include <net/netns/generic.h>
55 #include <net/dst_metadata.h>
56 
57 MODULE_AUTHOR("Ville Nuorvala");
58 MODULE_DESCRIPTION("IPv6 tunneling device");
59 MODULE_LICENSE("GPL");
60 MODULE_ALIAS_RTNL_LINK("ip6tnl");
61 MODULE_ALIAS_NETDEV("ip6tnl0");
62 
63 #define IP6_TUNNEL_HASH_SIZE_SHIFT  5
64 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
65 
66 static bool log_ecn_error = true;
67 module_param(log_ecn_error, bool, 0644);
68 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
69 
HASH(const struct in6_addr * addr1,const struct in6_addr * addr2)70 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
71 {
72 	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
73 
74 	return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
75 }
76 
77 static int ip6_tnl_dev_init(struct net_device *dev);
78 static void ip6_tnl_dev_setup(struct net_device *dev);
79 static struct rtnl_link_ops ip6_link_ops __read_mostly;
80 
81 static unsigned int ip6_tnl_net_id __read_mostly;
82 struct ip6_tnl_net {
83 	/* the IPv6 tunnel fallback device */
84 	struct net_device *fb_tnl_dev;
85 	/* lists for storing tunnels in use */
86 	struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
87 	struct ip6_tnl __rcu *tnls_wc[1];
88 	struct ip6_tnl __rcu **tnls[2];
89 	struct ip6_tnl __rcu *collect_md_tun;
90 };
91 
ip6_tnl_mpls_supported(void)92 static inline int ip6_tnl_mpls_supported(void)
93 {
94 	return IS_ENABLED(CONFIG_MPLS);
95 }
96 
ip6_get_stats(struct net_device * dev)97 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
98 {
99 	struct pcpu_sw_netstats tmp, sum = { 0 };
100 	int i;
101 
102 	for_each_possible_cpu(i) {
103 		unsigned int start;
104 		const struct pcpu_sw_netstats *tstats =
105 						   per_cpu_ptr(dev->tstats, i);
106 
107 		do {
108 			start = u64_stats_fetch_begin_irq(&tstats->syncp);
109 			tmp.rx_packets = tstats->rx_packets;
110 			tmp.rx_bytes = tstats->rx_bytes;
111 			tmp.tx_packets = tstats->tx_packets;
112 			tmp.tx_bytes =  tstats->tx_bytes;
113 		} while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
114 
115 		sum.rx_packets += tmp.rx_packets;
116 		sum.rx_bytes   += tmp.rx_bytes;
117 		sum.tx_packets += tmp.tx_packets;
118 		sum.tx_bytes   += tmp.tx_bytes;
119 	}
120 	dev->stats.rx_packets = sum.rx_packets;
121 	dev->stats.rx_bytes   = sum.rx_bytes;
122 	dev->stats.tx_packets = sum.tx_packets;
123 	dev->stats.tx_bytes   = sum.tx_bytes;
124 	return &dev->stats;
125 }
126 
127 #define for_each_ip6_tunnel_rcu(start) \
128 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
129 
130 /**
131  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
132  *   @net: network namespace
133  *   @link: ifindex of underlying interface
134  *   @remote: the address of the tunnel exit-point
135  *   @local: the address of the tunnel entry-point
136  *
137  * Return:
138  *   tunnel matching given end-points if found,
139  *   else fallback tunnel if its device is up,
140  *   else %NULL
141  **/
142 
143 static struct ip6_tnl *
ip6_tnl_lookup(struct net * net,int link,const struct in6_addr * remote,const struct in6_addr * local)144 ip6_tnl_lookup(struct net *net, int link,
145 	       const struct in6_addr *remote, const struct in6_addr *local)
146 {
147 	unsigned int hash = HASH(remote, local);
148 	struct ip6_tnl *t, *cand = NULL;
149 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
150 	struct in6_addr any;
151 
152 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
153 		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
154 		    !ipv6_addr_equal(remote, &t->parms.raddr) ||
155 		    !(t->dev->flags & IFF_UP))
156 			continue;
157 
158 		if (link == t->parms.link)
159 			return t;
160 		else
161 			cand = t;
162 	}
163 
164 	memset(&any, 0, sizeof(any));
165 	hash = HASH(&any, local);
166 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
167 		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
168 		    !ipv6_addr_any(&t->parms.raddr) ||
169 		    !(t->dev->flags & IFF_UP))
170 			continue;
171 
172 		if (link == t->parms.link)
173 			return t;
174 		else if (!cand)
175 			cand = t;
176 	}
177 
178 	hash = HASH(remote, &any);
179 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
180 		if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
181 		    !ipv6_addr_any(&t->parms.laddr) ||
182 		    !(t->dev->flags & IFF_UP))
183 			continue;
184 
185 		if (link == t->parms.link)
186 			return t;
187 		else if (!cand)
188 			cand = t;
189 	}
190 
191 	if (cand)
192 		return cand;
193 
194 	t = rcu_dereference(ip6n->collect_md_tun);
195 	if (t && t->dev->flags & IFF_UP)
196 		return t;
197 
198 	t = rcu_dereference(ip6n->tnls_wc[0]);
199 	if (t && (t->dev->flags & IFF_UP))
200 		return t;
201 
202 	return NULL;
203 }
204 
205 /**
206  * ip6_tnl_bucket - get head of list matching given tunnel parameters
207  *   @p: parameters containing tunnel end-points
208  *
209  * Description:
210  *   ip6_tnl_bucket() returns the head of the list matching the
211  *   &struct in6_addr entries laddr and raddr in @p.
212  *
213  * Return: head of IPv6 tunnel list
214  **/
215 
216 static struct ip6_tnl __rcu **
ip6_tnl_bucket(struct ip6_tnl_net * ip6n,const struct __ip6_tnl_parm * p)217 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
218 {
219 	const struct in6_addr *remote = &p->raddr;
220 	const struct in6_addr *local = &p->laddr;
221 	unsigned int h = 0;
222 	int prio = 0;
223 
224 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
225 		prio = 1;
226 		h = HASH(remote, local);
227 	}
228 	return &ip6n->tnls[prio][h];
229 }
230 
231 /**
232  * ip6_tnl_link - add tunnel to hash table
233  *   @t: tunnel to be added
234  **/
235 
236 static void
ip6_tnl_link(struct ip6_tnl_net * ip6n,struct ip6_tnl * t)237 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
238 {
239 	struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
240 
241 	if (t->parms.collect_md)
242 		rcu_assign_pointer(ip6n->collect_md_tun, t);
243 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
244 	rcu_assign_pointer(*tp, t);
245 }
246 
247 /**
248  * ip6_tnl_unlink - remove tunnel from hash table
249  *   @t: tunnel to be removed
250  **/
251 
252 static void
ip6_tnl_unlink(struct ip6_tnl_net * ip6n,struct ip6_tnl * t)253 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
254 {
255 	struct ip6_tnl __rcu **tp;
256 	struct ip6_tnl *iter;
257 
258 	if (t->parms.collect_md)
259 		rcu_assign_pointer(ip6n->collect_md_tun, NULL);
260 
261 	for (tp = ip6_tnl_bucket(ip6n, &t->parms);
262 	     (iter = rtnl_dereference(*tp)) != NULL;
263 	     tp = &iter->next) {
264 		if (t == iter) {
265 			rcu_assign_pointer(*tp, t->next);
266 			break;
267 		}
268 	}
269 }
270 
ip6_dev_free(struct net_device * dev)271 static void ip6_dev_free(struct net_device *dev)
272 {
273 	struct ip6_tnl *t = netdev_priv(dev);
274 
275 	gro_cells_destroy(&t->gro_cells);
276 	dst_cache_destroy(&t->dst_cache);
277 	free_percpu(dev->tstats);
278 }
279 
ip6_tnl_create2(struct net_device * dev)280 static int ip6_tnl_create2(struct net_device *dev)
281 {
282 	struct ip6_tnl *t = netdev_priv(dev);
283 	struct net *net = dev_net(dev);
284 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
285 	int err;
286 
287 	t = netdev_priv(dev);
288 
289 	dev->rtnl_link_ops = &ip6_link_ops;
290 	err = register_netdevice(dev);
291 	if (err < 0)
292 		goto out;
293 
294 	strcpy(t->parms.name, dev->name);
295 
296 	ip6_tnl_link(ip6n, t);
297 	return 0;
298 
299 out:
300 	return err;
301 }
302 
303 /**
304  * ip6_tnl_create - create a new tunnel
305  *   @net: network namespace
306  *   @p: tunnel parameters
307  *
308  * Description:
309  *   Create tunnel matching given parameters.
310  *
311  * Return:
312  *   created tunnel or error pointer
313  **/
314 
ip6_tnl_create(struct net * net,struct __ip6_tnl_parm * p)315 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
316 {
317 	struct net_device *dev;
318 	struct ip6_tnl *t;
319 	char name[IFNAMSIZ];
320 	int err = -E2BIG;
321 
322 	if (p->name[0]) {
323 		if (!dev_valid_name(p->name))
324 			goto failed;
325 		strlcpy(name, p->name, IFNAMSIZ);
326 	} else {
327 		sprintf(name, "ip6tnl%%d");
328 	}
329 	err = -ENOMEM;
330 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
331 			   ip6_tnl_dev_setup);
332 	if (!dev)
333 		goto failed;
334 
335 	dev_net_set(dev, net);
336 
337 	t = netdev_priv(dev);
338 	t->parms = *p;
339 	t->net = dev_net(dev);
340 	err = ip6_tnl_create2(dev);
341 	if (err < 0)
342 		goto failed_free;
343 
344 	return t;
345 
346 failed_free:
347 	free_netdev(dev);
348 failed:
349 	return ERR_PTR(err);
350 }
351 
352 /**
353  * ip6_tnl_locate - find or create tunnel matching given parameters
354  *   @net: network namespace
355  *   @p: tunnel parameters
356  *   @create: != 0 if allowed to create new tunnel if no match found
357  *
358  * Description:
359  *   ip6_tnl_locate() first tries to locate an existing tunnel
360  *   based on @parms. If this is unsuccessful, but @create is set a new
361  *   tunnel device is created and registered for use.
362  *
363  * Return:
364  *   matching tunnel or error pointer
365  **/
366 
ip6_tnl_locate(struct net * net,struct __ip6_tnl_parm * p,int create)367 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
368 		struct __ip6_tnl_parm *p, int create)
369 {
370 	const struct in6_addr *remote = &p->raddr;
371 	const struct in6_addr *local = &p->laddr;
372 	struct ip6_tnl __rcu **tp;
373 	struct ip6_tnl *t;
374 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
375 
376 	for (tp = ip6_tnl_bucket(ip6n, p);
377 	     (t = rtnl_dereference(*tp)) != NULL;
378 	     tp = &t->next) {
379 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
380 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
381 		    p->link == t->parms.link) {
382 			if (create)
383 				return ERR_PTR(-EEXIST);
384 
385 			return t;
386 		}
387 	}
388 	if (!create)
389 		return ERR_PTR(-ENODEV);
390 	return ip6_tnl_create(net, p);
391 }
392 
393 /**
394  * ip6_tnl_dev_uninit - tunnel device uninitializer
395  *   @dev: the device to be destroyed
396  *
397  * Description:
398  *   ip6_tnl_dev_uninit() removes tunnel from its list
399  **/
400 
401 static void
ip6_tnl_dev_uninit(struct net_device * dev)402 ip6_tnl_dev_uninit(struct net_device *dev)
403 {
404 	struct ip6_tnl *t = netdev_priv(dev);
405 	struct net *net = t->net;
406 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
407 
408 	if (dev == ip6n->fb_tnl_dev)
409 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
410 	else
411 		ip6_tnl_unlink(ip6n, t);
412 	dst_cache_reset(&t->dst_cache);
413 	dev_put(dev);
414 }
415 
416 /**
417  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
418  *   @skb: received socket buffer
419  *
420  * Return:
421  *   0 if none was found,
422  *   else index to encapsulation limit
423  **/
424 
ip6_tnl_parse_tlv_enc_lim(struct sk_buff * skb,__u8 * raw)425 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
426 {
427 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
428 	unsigned int nhoff = raw - skb->data;
429 	unsigned int off = nhoff + sizeof(*ipv6h);
430 	u8 next, nexthdr = ipv6h->nexthdr;
431 
432 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
433 		struct ipv6_opt_hdr *hdr;
434 		u16 optlen;
435 
436 		if (!pskb_may_pull(skb, off + sizeof(*hdr)))
437 			break;
438 
439 		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
440 		if (nexthdr == NEXTHDR_FRAGMENT) {
441 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
442 			if (frag_hdr->frag_off)
443 				break;
444 			optlen = 8;
445 		} else if (nexthdr == NEXTHDR_AUTH) {
446 			optlen = ipv6_authlen(hdr);
447 		} else {
448 			optlen = ipv6_optlen(hdr);
449 		}
450 		/* cache hdr->nexthdr, since pskb_may_pull() might
451 		 * invalidate hdr
452 		 */
453 		next = hdr->nexthdr;
454 		if (nexthdr == NEXTHDR_DEST) {
455 			u16 i = 2;
456 
457 			/* Remember : hdr is no longer valid at this point. */
458 			if (!pskb_may_pull(skb, off + optlen))
459 				break;
460 
461 			while (1) {
462 				struct ipv6_tlv_tnl_enc_lim *tel;
463 
464 				/* No more room for encapsulation limit */
465 				if (i + sizeof(*tel) > optlen)
466 					break;
467 
468 				tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
469 				/* return index of option if found and valid */
470 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
471 				    tel->length == 1)
472 					return i + off - nhoff;
473 				/* else jump to next option */
474 				if (tel->type)
475 					i += tel->length + 2;
476 				else
477 					i++;
478 			}
479 		}
480 		nexthdr = next;
481 		off += optlen;
482 	}
483 	return 0;
484 }
485 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
486 
487 /**
488  * ip6_tnl_err - tunnel error handler
489  *
490  * Description:
491  *   ip6_tnl_err() should handle errors in the tunnel according
492  *   to the specifications in RFC 2473.
493  **/
494 
495 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)496 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
497 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
498 {
499 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
500 	struct net *net = dev_net(skb->dev);
501 	u8 rel_type = ICMPV6_DEST_UNREACH;
502 	u8 rel_code = ICMPV6_ADDR_UNREACH;
503 	__u32 rel_info = 0;
504 	struct ip6_tnl *t;
505 	int err = -ENOENT;
506 	int rel_msg = 0;
507 	u8 tproto;
508 	__u16 len;
509 
510 	/* If the packet doesn't contain the original IPv6 header we are
511 	   in trouble since we might need the source address for further
512 	   processing of the error. */
513 
514 	rcu_read_lock();
515 	t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->daddr, &ipv6h->saddr);
516 	if (!t)
517 		goto out;
518 
519 	tproto = READ_ONCE(t->parms.proto);
520 	if (tproto != ipproto && tproto != 0)
521 		goto out;
522 
523 	err = 0;
524 
525 	switch (*type) {
526 	case ICMPV6_DEST_UNREACH:
527 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
528 				    t->parms.name);
529 		rel_msg = 1;
530 		break;
531 	case ICMPV6_TIME_EXCEED:
532 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
533 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
534 					    t->parms.name);
535 			rel_msg = 1;
536 		}
537 		break;
538 	case ICMPV6_PARAMPROB: {
539 		struct ipv6_tlv_tnl_enc_lim *tel;
540 		__u32 teli;
541 
542 		teli = 0;
543 		if ((*code) == ICMPV6_HDR_FIELD)
544 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
545 
546 		if (teli && teli == *info - 2) {
547 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
548 			if (tel->encap_limit == 0) {
549 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
550 						    t->parms.name);
551 				rel_msg = 1;
552 			}
553 		} else {
554 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
555 					    t->parms.name);
556 		}
557 		break;
558 	}
559 	case ICMPV6_PKT_TOOBIG: {
560 		__u32 mtu;
561 
562 		ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
563 				sock_net_uid(net, NULL));
564 		mtu = *info - offset;
565 		if (mtu < IPV6_MIN_MTU)
566 			mtu = IPV6_MIN_MTU;
567 		len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
568 		if (len > mtu) {
569 			rel_type = ICMPV6_PKT_TOOBIG;
570 			rel_code = 0;
571 			rel_info = mtu;
572 			rel_msg = 1;
573 		}
574 		break;
575 	}
576 	case NDISC_REDIRECT:
577 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
578 			     sock_net_uid(net, NULL));
579 		break;
580 	}
581 
582 	*type = rel_type;
583 	*code = rel_code;
584 	*info = rel_info;
585 	*msg = rel_msg;
586 
587 out:
588 	rcu_read_unlock();
589 	return err;
590 }
591 
592 static int
ip4ip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)593 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
594 	   u8 type, u8 code, int offset, __be32 info)
595 {
596 	__u32 rel_info = ntohl(info);
597 	const struct iphdr *eiph;
598 	struct sk_buff *skb2;
599 	int err, rel_msg = 0;
600 	u8 rel_type = type;
601 	u8 rel_code = code;
602 	struct rtable *rt;
603 	struct flowi4 fl4;
604 
605 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
606 			  &rel_msg, &rel_info, offset);
607 	if (err < 0)
608 		return err;
609 
610 	if (rel_msg == 0)
611 		return 0;
612 
613 	switch (rel_type) {
614 	case ICMPV6_DEST_UNREACH:
615 		if (rel_code != ICMPV6_ADDR_UNREACH)
616 			return 0;
617 		rel_type = ICMP_DEST_UNREACH;
618 		rel_code = ICMP_HOST_UNREACH;
619 		break;
620 	case ICMPV6_PKT_TOOBIG:
621 		if (rel_code != 0)
622 			return 0;
623 		rel_type = ICMP_DEST_UNREACH;
624 		rel_code = ICMP_FRAG_NEEDED;
625 		break;
626 	default:
627 		return 0;
628 	}
629 
630 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
631 		return 0;
632 
633 	skb2 = skb_clone(skb, GFP_ATOMIC);
634 	if (!skb2)
635 		return 0;
636 
637 	skb_dst_drop(skb2);
638 
639 	skb_pull(skb2, offset);
640 	skb_reset_network_header(skb2);
641 	eiph = ip_hdr(skb2);
642 
643 	/* Try to guess incoming interface */
644 	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
645 				   0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
646 	if (IS_ERR(rt))
647 		goto out;
648 
649 	skb2->dev = rt->dst.dev;
650 	ip_rt_put(rt);
651 
652 	/* route "incoming" packet */
653 	if (rt->rt_flags & RTCF_LOCAL) {
654 		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
655 					   eiph->daddr, eiph->saddr, 0, 0,
656 					   IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
657 		if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) {
658 			if (!IS_ERR(rt))
659 				ip_rt_put(rt);
660 			goto out;
661 		}
662 		skb_dst_set(skb2, &rt->dst);
663 	} else {
664 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
665 				   skb2->dev) ||
666 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
667 			goto out;
668 	}
669 
670 	/* change mtu on this route */
671 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
672 		if (rel_info > dst_mtu(skb_dst(skb2)))
673 			goto out;
674 
675 		skb_dst_update_pmtu_no_confirm(skb2, rel_info);
676 	}
677 
678 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
679 
680 out:
681 	kfree_skb(skb2);
682 	return 0;
683 }
684 
685 static int
ip6ip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)686 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
687 	   u8 type, u8 code, int offset, __be32 info)
688 {
689 	__u32 rel_info = ntohl(info);
690 	int err, rel_msg = 0;
691 	u8 rel_type = type;
692 	u8 rel_code = code;
693 
694 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
695 			  &rel_msg, &rel_info, offset);
696 	if (err < 0)
697 		return err;
698 
699 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
700 		struct rt6_info *rt;
701 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
702 
703 		if (!skb2)
704 			return 0;
705 
706 		skb_dst_drop(skb2);
707 		skb_pull(skb2, offset);
708 		skb_reset_network_header(skb2);
709 
710 		/* Try to guess incoming interface */
711 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
712 				NULL, 0, skb2, 0);
713 
714 		if (rt && rt->dst.dev)
715 			skb2->dev = rt->dst.dev;
716 
717 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
718 
719 		ip6_rt_put(rt);
720 
721 		kfree_skb(skb2);
722 	}
723 
724 	return 0;
725 }
726 
727 static int
mplsip6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)728 mplsip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
729 	    u8 type, u8 code, int offset, __be32 info)
730 {
731 	__u32 rel_info = ntohl(info);
732 	int err, rel_msg = 0;
733 	u8 rel_type = type;
734 	u8 rel_code = code;
735 
736 	err = ip6_tnl_err(skb, IPPROTO_MPLS, opt, &rel_type, &rel_code,
737 			  &rel_msg, &rel_info, offset);
738 	return err;
739 }
740 
ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)741 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
742 				       const struct ipv6hdr *ipv6h,
743 				       struct sk_buff *skb)
744 {
745 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
746 
747 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
748 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
749 
750 	return IP6_ECN_decapsulate(ipv6h, skb);
751 }
752 
ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)753 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
754 				       const struct ipv6hdr *ipv6h,
755 				       struct sk_buff *skb)
756 {
757 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
758 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
759 
760 	return IP6_ECN_decapsulate(ipv6h, skb);
761 }
762 
mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl * t,const struct ipv6hdr * ipv6h,struct sk_buff * skb)763 static inline int mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
764 					       const struct ipv6hdr *ipv6h,
765 					       struct sk_buff *skb)
766 {
767 	/* ECN is not supported in AF_MPLS */
768 	return 0;
769 }
770 
ip6_tnl_get_cap(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)771 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
772 			     const struct in6_addr *laddr,
773 			     const struct in6_addr *raddr)
774 {
775 	struct __ip6_tnl_parm *p = &t->parms;
776 	int ltype = ipv6_addr_type(laddr);
777 	int rtype = ipv6_addr_type(raddr);
778 	__u32 flags = 0;
779 
780 	if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
781 		flags = IP6_TNL_F_CAP_PER_PACKET;
782 	} else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
783 		   rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
784 		   !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
785 		   (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
786 		if (ltype&IPV6_ADDR_UNICAST)
787 			flags |= IP6_TNL_F_CAP_XMIT;
788 		if (rtype&IPV6_ADDR_UNICAST)
789 			flags |= IP6_TNL_F_CAP_RCV;
790 	}
791 	return flags;
792 }
793 EXPORT_SYMBOL(ip6_tnl_get_cap);
794 
795 /* called with rcu_read_lock() */
ip6_tnl_rcv_ctl(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)796 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
797 				  const struct in6_addr *laddr,
798 				  const struct in6_addr *raddr)
799 {
800 	struct __ip6_tnl_parm *p = &t->parms;
801 	int ret = 0;
802 	struct net *net = t->net;
803 
804 	if ((p->flags & IP6_TNL_F_CAP_RCV) ||
805 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
806 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
807 		struct net_device *ldev = NULL;
808 
809 		if (p->link)
810 			ldev = dev_get_by_index_rcu(net, p->link);
811 
812 		if ((ipv6_addr_is_multicast(laddr) ||
813 		     likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
814 						    0, IFA_F_TENTATIVE))) &&
815 		    ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
816 		     likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
817 						     0, IFA_F_TENTATIVE))))
818 			ret = 1;
819 	}
820 	return ret;
821 }
822 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
823 
__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)824 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
825 			 const struct tnl_ptk_info *tpi,
826 			 struct metadata_dst *tun_dst,
827 			 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
828 						const struct ipv6hdr *ipv6h,
829 						struct sk_buff *skb),
830 			 bool log_ecn_err)
831 {
832 	struct pcpu_sw_netstats *tstats;
833 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
834 	int err;
835 
836 	if ((!(tpi->flags & TUNNEL_CSUM) &&
837 	     (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
838 	    ((tpi->flags & TUNNEL_CSUM) &&
839 	     !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
840 		tunnel->dev->stats.rx_crc_errors++;
841 		tunnel->dev->stats.rx_errors++;
842 		goto drop;
843 	}
844 
845 	if (tunnel->parms.i_flags & TUNNEL_SEQ) {
846 		if (!(tpi->flags & TUNNEL_SEQ) ||
847 		    (tunnel->i_seqno &&
848 		     (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
849 			tunnel->dev->stats.rx_fifo_errors++;
850 			tunnel->dev->stats.rx_errors++;
851 			goto drop;
852 		}
853 		tunnel->i_seqno = ntohl(tpi->seq) + 1;
854 	}
855 
856 	skb->protocol = tpi->proto;
857 
858 	/* Warning: All skb pointers will be invalidated! */
859 	if (tunnel->dev->type == ARPHRD_ETHER) {
860 		if (!pskb_may_pull(skb, ETH_HLEN)) {
861 			tunnel->dev->stats.rx_length_errors++;
862 			tunnel->dev->stats.rx_errors++;
863 			goto drop;
864 		}
865 
866 		ipv6h = ipv6_hdr(skb);
867 		skb->protocol = eth_type_trans(skb, tunnel->dev);
868 		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
869 	} else {
870 		skb->dev = tunnel->dev;
871 	}
872 
873 	skb_reset_network_header(skb);
874 	memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
875 
876 	__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
877 
878 	err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
879 	if (unlikely(err)) {
880 		if (log_ecn_err)
881 			net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
882 					     &ipv6h->saddr,
883 					     ipv6_get_dsfield(ipv6h));
884 		if (err > 1) {
885 			++tunnel->dev->stats.rx_frame_errors;
886 			++tunnel->dev->stats.rx_errors;
887 			goto drop;
888 		}
889 	}
890 
891 	tstats = this_cpu_ptr(tunnel->dev->tstats);
892 	u64_stats_update_begin(&tstats->syncp);
893 	tstats->rx_packets++;
894 	tstats->rx_bytes += skb->len;
895 	u64_stats_update_end(&tstats->syncp);
896 
897 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
898 
899 	if (tun_dst)
900 		skb_dst_set(skb, (struct dst_entry *)tun_dst);
901 
902 	gro_cells_receive(&tunnel->gro_cells, skb);
903 	return 0;
904 
905 drop:
906 	if (tun_dst)
907 		dst_release((struct dst_entry *)tun_dst);
908 	kfree_skb(skb);
909 	return 0;
910 }
911 
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)912 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
913 		const struct tnl_ptk_info *tpi,
914 		struct metadata_dst *tun_dst,
915 		bool log_ecn_err)
916 {
917 	int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
918 				    const struct ipv6hdr *ipv6h,
919 				    struct sk_buff *skb);
920 
921 	dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
922 	if (tpi->proto == htons(ETH_P_IP))
923 		dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
924 
925 	return __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
926 			     log_ecn_err);
927 }
928 EXPORT_SYMBOL(ip6_tnl_rcv);
929 
930 static const struct tnl_ptk_info tpi_v6 = {
931 	/* no tunnel info required for ipxip6. */
932 	.proto = htons(ETH_P_IPV6),
933 };
934 
935 static const struct tnl_ptk_info tpi_v4 = {
936 	/* no tunnel info required for ipxip6. */
937 	.proto = htons(ETH_P_IP),
938 };
939 
940 static const struct tnl_ptk_info tpi_mpls = {
941 	/* no tunnel info required for mplsip6. */
942 	.proto = htons(ETH_P_MPLS_UC),
943 };
944 
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))945 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
946 		      const struct tnl_ptk_info *tpi,
947 		      int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
948 						  const struct ipv6hdr *ipv6h,
949 						  struct sk_buff *skb))
950 {
951 	struct ip6_tnl *t;
952 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
953 	struct metadata_dst *tun_dst = NULL;
954 	int ret = -1;
955 
956 	rcu_read_lock();
957 	t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr);
958 
959 	if (t) {
960 		u8 tproto = READ_ONCE(t->parms.proto);
961 
962 		if (tproto != ipproto && tproto != 0)
963 			goto drop;
964 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
965 			goto drop;
966 		ipv6h = ipv6_hdr(skb);
967 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
968 			goto drop;
969 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
970 			goto drop;
971 		if (t->parms.collect_md) {
972 			tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
973 			if (!tun_dst)
974 				goto drop;
975 		}
976 		ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
977 				    log_ecn_error);
978 	}
979 
980 	rcu_read_unlock();
981 
982 	return ret;
983 
984 drop:
985 	rcu_read_unlock();
986 	kfree_skb(skb);
987 	return 0;
988 }
989 
ip4ip6_rcv(struct sk_buff * skb)990 static int ip4ip6_rcv(struct sk_buff *skb)
991 {
992 	return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
993 			  ip4ip6_dscp_ecn_decapsulate);
994 }
995 
ip6ip6_rcv(struct sk_buff * skb)996 static int ip6ip6_rcv(struct sk_buff *skb)
997 {
998 	return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
999 			  ip6ip6_dscp_ecn_decapsulate);
1000 }
1001 
mplsip6_rcv(struct sk_buff * skb)1002 static int mplsip6_rcv(struct sk_buff *skb)
1003 {
1004 	return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls,
1005 			  mplsip6_dscp_ecn_decapsulate);
1006 }
1007 
1008 struct ipv6_tel_txoption {
1009 	struct ipv6_txoptions ops;
1010 	__u8 dst_opt[8];
1011 };
1012 
init_tel_txopt(struct ipv6_tel_txoption * opt,__u8 encap_limit)1013 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
1014 {
1015 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
1016 
1017 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
1018 	opt->dst_opt[3] = 1;
1019 	opt->dst_opt[4] = encap_limit;
1020 	opt->dst_opt[5] = IPV6_TLV_PADN;
1021 	opt->dst_opt[6] = 1;
1022 
1023 	opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
1024 	opt->ops.opt_nflen = 8;
1025 }
1026 
1027 /**
1028  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
1029  *   @t: the outgoing tunnel device
1030  *   @hdr: IPv6 header from the incoming packet
1031  *
1032  * Description:
1033  *   Avoid trivial tunneling loop by checking that tunnel exit-point
1034  *   doesn't match source of incoming packet.
1035  *
1036  * Return:
1037  *   1 if conflict,
1038  *   0 else
1039  **/
1040 
1041 static inline bool
ip6_tnl_addr_conflict(const struct ip6_tnl * t,const struct ipv6hdr * hdr)1042 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
1043 {
1044 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
1045 }
1046 
ip6_tnl_xmit_ctl(struct ip6_tnl * t,const struct in6_addr * laddr,const struct in6_addr * raddr)1047 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
1048 		     const struct in6_addr *laddr,
1049 		     const struct in6_addr *raddr)
1050 {
1051 	struct __ip6_tnl_parm *p = &t->parms;
1052 	int ret = 0;
1053 	struct net *net = t->net;
1054 
1055 	if (t->parms.collect_md)
1056 		return 1;
1057 
1058 	if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1059 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1060 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1061 		struct net_device *ldev = NULL;
1062 
1063 		rcu_read_lock();
1064 		if (p->link)
1065 			ldev = dev_get_by_index_rcu(net, p->link);
1066 
1067 		if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
1068 						      0, IFA_F_TENTATIVE)))
1069 			pr_warn_ratelimited("%s xmit: Local address not yet configured!\n",
1070 					    p->name);
1071 		else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1072 			 !ipv6_addr_is_multicast(raddr) &&
1073 			 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1074 							  true, 0, IFA_F_TENTATIVE)))
1075 			pr_warn_ratelimited("%s xmit: Routing loop! Remote address found on this node!\n",
1076 					    p->name);
1077 		else
1078 			ret = 1;
1079 		rcu_read_unlock();
1080 	}
1081 	return ret;
1082 }
1083 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1084 
1085 /**
1086  * ip6_tnl_xmit - encapsulate packet and send
1087  *   @skb: the outgoing socket buffer
1088  *   @dev: the outgoing tunnel device
1089  *   @dsfield: dscp code for outer header
1090  *   @fl6: flow of tunneled packet
1091  *   @encap_limit: encapsulation limit
1092  *   @pmtu: Path MTU is stored if packet is too big
1093  *   @proto: next header value
1094  *
1095  * Description:
1096  *   Build new header and do some sanity checks on the packet before sending
1097  *   it.
1098  *
1099  * Return:
1100  *   0 on success
1101  *   -1 fail
1102  *   %-EMSGSIZE message too big. return mtu in this case.
1103  **/
1104 
ip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev,__u8 dsfield,struct flowi6 * fl6,int encap_limit,__u32 * pmtu,__u8 proto)1105 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1106 		 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1107 		 __u8 proto)
1108 {
1109 	struct ip6_tnl *t = netdev_priv(dev);
1110 	struct net *net = t->net;
1111 	struct net_device_stats *stats = &t->dev->stats;
1112 	struct ipv6hdr *ipv6h;
1113 	struct ipv6_tel_txoption opt;
1114 	struct dst_entry *dst = NULL, *ndst = NULL;
1115 	struct net_device *tdev;
1116 	int mtu;
1117 	unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1118 	unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1119 	unsigned int max_headroom = psh_hlen;
1120 	bool use_cache = false;
1121 	u8 hop_limit;
1122 	int err = -1;
1123 
1124 	if (t->parms.collect_md) {
1125 		hop_limit = skb_tunnel_info(skb)->key.ttl;
1126 		goto route_lookup;
1127 	} else {
1128 		hop_limit = t->parms.hop_limit;
1129 	}
1130 
1131 	/* NBMA tunnel */
1132 	if (ipv6_addr_any(&t->parms.raddr)) {
1133 		if (skb->protocol == htons(ETH_P_IPV6)) {
1134 			struct in6_addr *addr6;
1135 			struct neighbour *neigh;
1136 			int addr_type;
1137 
1138 			if (!skb_dst(skb))
1139 				goto tx_err_link_failure;
1140 
1141 			neigh = dst_neigh_lookup(skb_dst(skb),
1142 						 &ipv6_hdr(skb)->daddr);
1143 			if (!neigh)
1144 				goto tx_err_link_failure;
1145 
1146 			addr6 = (struct in6_addr *)&neigh->primary_key;
1147 			addr_type = ipv6_addr_type(addr6);
1148 
1149 			if (addr_type == IPV6_ADDR_ANY)
1150 				addr6 = &ipv6_hdr(skb)->daddr;
1151 
1152 			memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1153 			neigh_release(neigh);
1154 		}
1155 	} else if (t->parms.proto != 0 && !(t->parms.flags &
1156 					    (IP6_TNL_F_USE_ORIG_TCLASS |
1157 					     IP6_TNL_F_USE_ORIG_FWMARK))) {
1158 		/* enable the cache only if neither the outer protocol nor the
1159 		 * routing decision depends on the current inner header value
1160 		 */
1161 		use_cache = true;
1162 	}
1163 
1164 	if (use_cache)
1165 		dst = dst_cache_get(&t->dst_cache);
1166 
1167 	if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1168 		goto tx_err_link_failure;
1169 
1170 	if (!dst) {
1171 route_lookup:
1172 		/* add dsfield to flowlabel for route lookup */
1173 		fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1174 
1175 		dst = ip6_route_output(net, NULL, fl6);
1176 
1177 		if (dst->error)
1178 			goto tx_err_link_failure;
1179 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1180 		if (IS_ERR(dst)) {
1181 			err = PTR_ERR(dst);
1182 			dst = NULL;
1183 			goto tx_err_link_failure;
1184 		}
1185 		if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1186 		    ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1187 				       &fl6->daddr, 0, &fl6->saddr))
1188 			goto tx_err_link_failure;
1189 		ndst = dst;
1190 	}
1191 
1192 	tdev = dst->dev;
1193 
1194 	if (tdev == dev) {
1195 		stats->collisions++;
1196 		net_warn_ratelimited("%s: Local routing loop detected!\n",
1197 				     t->parms.name);
1198 		goto tx_err_dst_release;
1199 	}
1200 	mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1201 	if (encap_limit >= 0) {
1202 		max_headroom += 8;
1203 		mtu -= 8;
1204 	}
1205 	mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1206 		       IPV6_MIN_MTU : IPV4_MIN_MTU);
1207 
1208 	skb_dst_update_pmtu_no_confirm(skb, mtu);
1209 	if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1210 		*pmtu = mtu;
1211 		err = -EMSGSIZE;
1212 		goto tx_err_dst_release;
1213 	}
1214 
1215 	if (t->err_count > 0) {
1216 		if (time_before(jiffies,
1217 				t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1218 			t->err_count--;
1219 
1220 			dst_link_failure(skb);
1221 		} else {
1222 			t->err_count = 0;
1223 		}
1224 	}
1225 
1226 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1227 
1228 	/*
1229 	 * Okay, now see if we can stuff it in the buffer as-is.
1230 	 */
1231 	max_headroom += LL_RESERVED_SPACE(tdev);
1232 
1233 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1234 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1235 		struct sk_buff *new_skb;
1236 
1237 		new_skb = skb_realloc_headroom(skb, max_headroom);
1238 		if (!new_skb)
1239 			goto tx_err_dst_release;
1240 
1241 		if (skb->sk)
1242 			skb_set_owner_w(new_skb, skb->sk);
1243 		consume_skb(skb);
1244 		skb = new_skb;
1245 	}
1246 
1247 	if (t->parms.collect_md) {
1248 		if (t->encap.type != TUNNEL_ENCAP_NONE)
1249 			goto tx_err_dst_release;
1250 	} else {
1251 		if (use_cache && ndst)
1252 			dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1253 	}
1254 	skb_dst_set(skb, dst);
1255 
1256 	if (hop_limit == 0) {
1257 		if (skb->protocol == htons(ETH_P_IP))
1258 			hop_limit = ip_hdr(skb)->ttl;
1259 		else if (skb->protocol == htons(ETH_P_IPV6))
1260 			hop_limit = ipv6_hdr(skb)->hop_limit;
1261 		else
1262 			hop_limit = ip6_dst_hoplimit(dst);
1263 	}
1264 
1265 	/* Calculate max headroom for all the headers and adjust
1266 	 * needed_headroom if necessary.
1267 	 */
1268 	max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1269 			+ dst->header_len + t->hlen;
1270 	if (max_headroom > READ_ONCE(dev->needed_headroom))
1271 		WRITE_ONCE(dev->needed_headroom, max_headroom);
1272 
1273 	err = ip6_tnl_encap(skb, t, &proto, fl6);
1274 	if (err)
1275 		return err;
1276 
1277 	if (encap_limit >= 0) {
1278 		init_tel_txopt(&opt, encap_limit);
1279 		ipv6_push_frag_opts(skb, &opt.ops, &proto);
1280 	}
1281 
1282 	skb_push(skb, sizeof(struct ipv6hdr));
1283 	skb_reset_network_header(skb);
1284 	ipv6h = ipv6_hdr(skb);
1285 	ip6_flow_hdr(ipv6h, dsfield,
1286 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1287 	ipv6h->hop_limit = hop_limit;
1288 	ipv6h->nexthdr = proto;
1289 	ipv6h->saddr = fl6->saddr;
1290 	ipv6h->daddr = fl6->daddr;
1291 	ip6tunnel_xmit(NULL, skb, dev);
1292 	return 0;
1293 tx_err_link_failure:
1294 	stats->tx_carrier_errors++;
1295 	dst_link_failure(skb);
1296 tx_err_dst_release:
1297 	dst_release(dst);
1298 	return err;
1299 }
1300 EXPORT_SYMBOL(ip6_tnl_xmit);
1301 
1302 static inline int
ipxip6_tnl_xmit(struct sk_buff * skb,struct net_device * dev,u8 protocol)1303 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev,
1304 		u8 protocol)
1305 {
1306 	struct ip6_tnl *t = netdev_priv(dev);
1307 	struct ipv6hdr *ipv6h;
1308 	const struct iphdr  *iph;
1309 	int encap_limit = -1;
1310 	__u16 offset;
1311 	struct flowi6 fl6;
1312 	__u8 dsfield, orig_dsfield;
1313 	__u32 mtu;
1314 	u8 tproto;
1315 	int err;
1316 
1317 	tproto = READ_ONCE(t->parms.proto);
1318 	if (tproto != protocol && tproto != 0)
1319 		return -1;
1320 
1321 	if (t->parms.collect_md) {
1322 		struct ip_tunnel_info *tun_info;
1323 		const struct ip_tunnel_key *key;
1324 
1325 		tun_info = skb_tunnel_info(skb);
1326 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1327 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1328 			return -1;
1329 		key = &tun_info->key;
1330 		memset(&fl6, 0, sizeof(fl6));
1331 		fl6.flowi6_proto = protocol;
1332 		fl6.saddr = key->u.ipv6.src;
1333 		fl6.daddr = key->u.ipv6.dst;
1334 		fl6.flowlabel = key->label;
1335 		dsfield =  key->tos;
1336 		switch (protocol) {
1337 		case IPPROTO_IPIP:
1338 			iph = ip_hdr(skb);
1339 			orig_dsfield = ipv4_get_dsfield(iph);
1340 			break;
1341 		case IPPROTO_IPV6:
1342 			ipv6h = ipv6_hdr(skb);
1343 			orig_dsfield = ipv6_get_dsfield(ipv6h);
1344 			break;
1345 		default:
1346 			orig_dsfield = dsfield;
1347 			break;
1348 		}
1349 	} else {
1350 		if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1351 			encap_limit = t->parms.encap_limit;
1352 		if (protocol == IPPROTO_IPV6) {
1353 			offset = ip6_tnl_parse_tlv_enc_lim(skb,
1354 						skb_network_header(skb));
1355 			/* ip6_tnl_parse_tlv_enc_lim() might have
1356 			 * reallocated skb->head
1357 			 */
1358 			if (offset > 0) {
1359 				struct ipv6_tlv_tnl_enc_lim *tel;
1360 
1361 				tel = (void *)&skb_network_header(skb)[offset];
1362 				if (tel->encap_limit == 0) {
1363 					icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
1364 							ICMPV6_HDR_FIELD, offset + 2);
1365 					return -1;
1366 				}
1367 				encap_limit = tel->encap_limit - 1;
1368 			}
1369 		}
1370 
1371 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1372 		fl6.flowi6_proto = protocol;
1373 
1374 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1375 			fl6.flowi6_mark = skb->mark;
1376 		else
1377 			fl6.flowi6_mark = t->parms.fwmark;
1378 		switch (protocol) {
1379 		case IPPROTO_IPIP:
1380 			iph = ip_hdr(skb);
1381 			orig_dsfield = ipv4_get_dsfield(iph);
1382 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1383 				dsfield = orig_dsfield;
1384 			else
1385 				dsfield = ip6_tclass(t->parms.flowinfo);
1386 			break;
1387 		case IPPROTO_IPV6:
1388 			ipv6h = ipv6_hdr(skb);
1389 			orig_dsfield = ipv6_get_dsfield(ipv6h);
1390 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1391 				dsfield = orig_dsfield;
1392 			else
1393 				dsfield = ip6_tclass(t->parms.flowinfo);
1394 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1395 				fl6.flowlabel |= ip6_flowlabel(ipv6h);
1396 			break;
1397 		default:
1398 			orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo);
1399 			break;
1400 		}
1401 	}
1402 
1403 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1404 	dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield);
1405 
1406 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1407 		return -1;
1408 
1409 	skb_set_inner_ipproto(skb, protocol);
1410 
1411 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1412 			   protocol);
1413 	if (err != 0) {
1414 		/* XXX: send ICMP error even if DF is not set. */
1415 		if (err == -EMSGSIZE)
1416 			switch (protocol) {
1417 			case IPPROTO_IPIP:
1418 				icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1419 					      ICMP_FRAG_NEEDED, htonl(mtu));
1420 				break;
1421 			case IPPROTO_IPV6:
1422 				icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1423 				break;
1424 			default:
1425 				break;
1426 			}
1427 		return -1;
1428 	}
1429 
1430 	return 0;
1431 }
1432 
1433 static netdev_tx_t
ip6_tnl_start_xmit(struct sk_buff * skb,struct net_device * dev)1434 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1435 {
1436 	struct ip6_tnl *t = netdev_priv(dev);
1437 	struct net_device_stats *stats = &t->dev->stats;
1438 	u8 ipproto;
1439 	int ret;
1440 
1441 	if (!pskb_inet_may_pull(skb))
1442 		goto tx_err;
1443 
1444 	switch (skb->protocol) {
1445 	case htons(ETH_P_IP):
1446 		ipproto = IPPROTO_IPIP;
1447 		break;
1448 	case htons(ETH_P_IPV6):
1449 		if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb)))
1450 			goto tx_err;
1451 		ipproto = IPPROTO_IPV6;
1452 		break;
1453 	case htons(ETH_P_MPLS_UC):
1454 		ipproto = IPPROTO_MPLS;
1455 		break;
1456 	default:
1457 		goto tx_err;
1458 	}
1459 
1460 	ret = ipxip6_tnl_xmit(skb, dev, ipproto);
1461 	if (ret < 0)
1462 		goto tx_err;
1463 
1464 	return NETDEV_TX_OK;
1465 
1466 tx_err:
1467 	stats->tx_errors++;
1468 	stats->tx_dropped++;
1469 	kfree_skb(skb);
1470 	return NETDEV_TX_OK;
1471 }
1472 
ip6_tnl_link_config(struct ip6_tnl * t)1473 static void ip6_tnl_link_config(struct ip6_tnl *t)
1474 {
1475 	struct net_device *dev = t->dev;
1476 	struct net_device *tdev = NULL;
1477 	struct __ip6_tnl_parm *p = &t->parms;
1478 	struct flowi6 *fl6 = &t->fl.u.ip6;
1479 	int t_hlen;
1480 	int mtu;
1481 
1482 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1483 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1484 
1485 	/* Set up flowi template */
1486 	fl6->saddr = p->laddr;
1487 	fl6->daddr = p->raddr;
1488 	fl6->flowi6_oif = p->link;
1489 	fl6->flowlabel = 0;
1490 
1491 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1492 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1493 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1494 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1495 
1496 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1497 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1498 
1499 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1500 		dev->flags |= IFF_POINTOPOINT;
1501 	else
1502 		dev->flags &= ~IFF_POINTOPOINT;
1503 
1504 	t->tun_hlen = 0;
1505 	t->hlen = t->encap_hlen + t->tun_hlen;
1506 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1507 
1508 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1509 		int strict = (ipv6_addr_type(&p->raddr) &
1510 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1511 
1512 		struct rt6_info *rt = rt6_lookup(t->net,
1513 						 &p->raddr, &p->laddr,
1514 						 p->link, NULL, strict);
1515 		if (rt) {
1516 			tdev = rt->dst.dev;
1517 			ip6_rt_put(rt);
1518 		}
1519 
1520 		if (!tdev && p->link)
1521 			tdev = __dev_get_by_index(t->net, p->link);
1522 
1523 		if (tdev) {
1524 			dev->hard_header_len = tdev->hard_header_len + t_hlen;
1525 			mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU);
1526 
1527 			mtu = mtu - t_hlen;
1528 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1529 				mtu -= 8;
1530 
1531 			if (mtu < IPV6_MIN_MTU)
1532 				mtu = IPV6_MIN_MTU;
1533 			WRITE_ONCE(dev->mtu, mtu);
1534 		}
1535 	}
1536 }
1537 
1538 /**
1539  * ip6_tnl_change - update the tunnel parameters
1540  *   @t: tunnel to be changed
1541  *   @p: tunnel configuration parameters
1542  *
1543  * Description:
1544  *   ip6_tnl_change() updates the tunnel parameters
1545  **/
1546 
1547 static int
ip6_tnl_change(struct ip6_tnl * t,const struct __ip6_tnl_parm * p)1548 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1549 {
1550 	t->parms.laddr = p->laddr;
1551 	t->parms.raddr = p->raddr;
1552 	t->parms.flags = p->flags;
1553 	t->parms.hop_limit = p->hop_limit;
1554 	t->parms.encap_limit = p->encap_limit;
1555 	t->parms.flowinfo = p->flowinfo;
1556 	t->parms.link = p->link;
1557 	t->parms.proto = p->proto;
1558 	t->parms.fwmark = p->fwmark;
1559 	dst_cache_reset(&t->dst_cache);
1560 	ip6_tnl_link_config(t);
1561 	return 0;
1562 }
1563 
ip6_tnl_update(struct ip6_tnl * t,struct __ip6_tnl_parm * p)1564 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1565 {
1566 	struct net *net = t->net;
1567 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1568 	int err;
1569 
1570 	ip6_tnl_unlink(ip6n, t);
1571 	synchronize_net();
1572 	err = ip6_tnl_change(t, p);
1573 	ip6_tnl_link(ip6n, t);
1574 	netdev_state_change(t->dev);
1575 	return err;
1576 }
1577 
ip6_tnl0_update(struct ip6_tnl * t,struct __ip6_tnl_parm * p)1578 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1579 {
1580 	/* for default tnl0 device allow to change only the proto */
1581 	t->parms.proto = p->proto;
1582 	netdev_state_change(t->dev);
1583 	return 0;
1584 }
1585 
1586 static void
ip6_tnl_parm_from_user(struct __ip6_tnl_parm * p,const struct ip6_tnl_parm * u)1587 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1588 {
1589 	p->laddr = u->laddr;
1590 	p->raddr = u->raddr;
1591 	p->flags = u->flags;
1592 	p->hop_limit = u->hop_limit;
1593 	p->encap_limit = u->encap_limit;
1594 	p->flowinfo = u->flowinfo;
1595 	p->link = u->link;
1596 	p->proto = u->proto;
1597 	memcpy(p->name, u->name, sizeof(u->name));
1598 }
1599 
1600 static void
ip6_tnl_parm_to_user(struct ip6_tnl_parm * u,const struct __ip6_tnl_parm * p)1601 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1602 {
1603 	u->laddr = p->laddr;
1604 	u->raddr = p->raddr;
1605 	u->flags = p->flags;
1606 	u->hop_limit = p->hop_limit;
1607 	u->encap_limit = p->encap_limit;
1608 	u->flowinfo = p->flowinfo;
1609 	u->link = p->link;
1610 	u->proto = p->proto;
1611 	memcpy(u->name, p->name, sizeof(u->name));
1612 }
1613 
1614 /**
1615  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1616  *   @dev: virtual device associated with tunnel
1617  *   @ifr: parameters passed from userspace
1618  *   @cmd: command to be performed
1619  *
1620  * Description:
1621  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1622  *   from userspace.
1623  *
1624  *   The possible commands are the following:
1625  *     %SIOCGETTUNNEL: get tunnel parameters for device
1626  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1627  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1628  *     %SIOCDELTUNNEL: delete tunnel
1629  *
1630  *   The fallback device "ip6tnl0", created during module
1631  *   initialization, can be used for creating other tunnel devices.
1632  *
1633  * Return:
1634  *   0 on success,
1635  *   %-EFAULT if unable to copy data to or from userspace,
1636  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1637  *   %-EINVAL if passed tunnel parameters are invalid,
1638  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1639  *   %-ENODEV if attempting to change or delete a nonexisting device
1640  **/
1641 
1642 static int
ip6_tnl_ioctl(struct net_device * dev,struct ifreq * ifr,int cmd)1643 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1644 {
1645 	int err = 0;
1646 	struct ip6_tnl_parm p;
1647 	struct __ip6_tnl_parm p1;
1648 	struct ip6_tnl *t = netdev_priv(dev);
1649 	struct net *net = t->net;
1650 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1651 
1652 	memset(&p1, 0, sizeof(p1));
1653 
1654 	switch (cmd) {
1655 	case SIOCGETTUNNEL:
1656 		if (dev == ip6n->fb_tnl_dev) {
1657 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1658 				err = -EFAULT;
1659 				break;
1660 			}
1661 			ip6_tnl_parm_from_user(&p1, &p);
1662 			t = ip6_tnl_locate(net, &p1, 0);
1663 			if (IS_ERR(t))
1664 				t = netdev_priv(dev);
1665 		} else {
1666 			memset(&p, 0, sizeof(p));
1667 		}
1668 		ip6_tnl_parm_to_user(&p, &t->parms);
1669 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1670 			err = -EFAULT;
1671 		}
1672 		break;
1673 	case SIOCADDTUNNEL:
1674 	case SIOCCHGTUNNEL:
1675 		err = -EPERM;
1676 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1677 			break;
1678 		err = -EFAULT;
1679 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1680 			break;
1681 		err = -EINVAL;
1682 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1683 		    p.proto != 0)
1684 			break;
1685 		ip6_tnl_parm_from_user(&p1, &p);
1686 		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1687 		if (cmd == SIOCCHGTUNNEL) {
1688 			if (!IS_ERR(t)) {
1689 				if (t->dev != dev) {
1690 					err = -EEXIST;
1691 					break;
1692 				}
1693 			} else
1694 				t = netdev_priv(dev);
1695 			if (dev == ip6n->fb_tnl_dev)
1696 				err = ip6_tnl0_update(t, &p1);
1697 			else
1698 				err = ip6_tnl_update(t, &p1);
1699 		}
1700 		if (!IS_ERR(t)) {
1701 			err = 0;
1702 			ip6_tnl_parm_to_user(&p, &t->parms);
1703 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1704 				err = -EFAULT;
1705 
1706 		} else {
1707 			err = PTR_ERR(t);
1708 		}
1709 		break;
1710 	case SIOCDELTUNNEL:
1711 		err = -EPERM;
1712 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1713 			break;
1714 
1715 		if (dev == ip6n->fb_tnl_dev) {
1716 			err = -EFAULT;
1717 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1718 				break;
1719 			err = -ENOENT;
1720 			ip6_tnl_parm_from_user(&p1, &p);
1721 			t = ip6_tnl_locate(net, &p1, 0);
1722 			if (IS_ERR(t))
1723 				break;
1724 			err = -EPERM;
1725 			if (t->dev == ip6n->fb_tnl_dev)
1726 				break;
1727 			dev = t->dev;
1728 		}
1729 		err = 0;
1730 		unregister_netdevice(dev);
1731 		break;
1732 	default:
1733 		err = -EINVAL;
1734 	}
1735 	return err;
1736 }
1737 
1738 /**
1739  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1740  *   @dev: virtual device associated with tunnel
1741  *   @new_mtu: the new mtu
1742  *
1743  * Return:
1744  *   0 on success,
1745  *   %-EINVAL if mtu too small
1746  **/
1747 
ip6_tnl_change_mtu(struct net_device * dev,int new_mtu)1748 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1749 {
1750 	struct ip6_tnl *tnl = netdev_priv(dev);
1751 
1752 	if (tnl->parms.proto == IPPROTO_IPV6) {
1753 		if (new_mtu < IPV6_MIN_MTU)
1754 			return -EINVAL;
1755 	} else {
1756 		if (new_mtu < ETH_MIN_MTU)
1757 			return -EINVAL;
1758 	}
1759 	if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1760 		if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1761 			return -EINVAL;
1762 	} else {
1763 		if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1764 			return -EINVAL;
1765 	}
1766 	dev->mtu = new_mtu;
1767 	return 0;
1768 }
1769 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1770 
ip6_tnl_get_iflink(const struct net_device * dev)1771 int ip6_tnl_get_iflink(const struct net_device *dev)
1772 {
1773 	struct ip6_tnl *t = netdev_priv(dev);
1774 
1775 	return t->parms.link;
1776 }
1777 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1778 
ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops * ops,unsigned int num)1779 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1780 			  unsigned int num)
1781 {
1782 	if (num >= MAX_IPTUN_ENCAP_OPS)
1783 		return -ERANGE;
1784 
1785 	return !cmpxchg((const struct ip6_tnl_encap_ops **)
1786 			&ip6tun_encaps[num],
1787 			NULL, ops) ? 0 : -1;
1788 }
1789 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1790 
ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops * ops,unsigned int num)1791 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1792 			  unsigned int num)
1793 {
1794 	int ret;
1795 
1796 	if (num >= MAX_IPTUN_ENCAP_OPS)
1797 		return -ERANGE;
1798 
1799 	ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1800 		       &ip6tun_encaps[num],
1801 		       ops, NULL) == ops) ? 0 : -1;
1802 
1803 	synchronize_net();
1804 
1805 	return ret;
1806 }
1807 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1808 
ip6_tnl_encap_setup(struct ip6_tnl * t,struct ip_tunnel_encap * ipencap)1809 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1810 			struct ip_tunnel_encap *ipencap)
1811 {
1812 	int hlen;
1813 
1814 	memset(&t->encap, 0, sizeof(t->encap));
1815 
1816 	hlen = ip6_encap_hlen(ipencap);
1817 	if (hlen < 0)
1818 		return hlen;
1819 
1820 	t->encap.type = ipencap->type;
1821 	t->encap.sport = ipencap->sport;
1822 	t->encap.dport = ipencap->dport;
1823 	t->encap.flags = ipencap->flags;
1824 
1825 	t->encap_hlen = hlen;
1826 	t->hlen = t->encap_hlen + t->tun_hlen;
1827 
1828 	return 0;
1829 }
1830 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1831 
1832 static const struct net_device_ops ip6_tnl_netdev_ops = {
1833 	.ndo_init	= ip6_tnl_dev_init,
1834 	.ndo_uninit	= ip6_tnl_dev_uninit,
1835 	.ndo_start_xmit = ip6_tnl_start_xmit,
1836 	.ndo_do_ioctl	= ip6_tnl_ioctl,
1837 	.ndo_change_mtu = ip6_tnl_change_mtu,
1838 	.ndo_get_stats	= ip6_get_stats,
1839 	.ndo_get_iflink = ip6_tnl_get_iflink,
1840 };
1841 
1842 #define IPXIPX_FEATURES (NETIF_F_SG |		\
1843 			 NETIF_F_FRAGLIST |	\
1844 			 NETIF_F_HIGHDMA |	\
1845 			 NETIF_F_GSO_SOFTWARE |	\
1846 			 NETIF_F_HW_CSUM)
1847 
1848 /**
1849  * ip6_tnl_dev_setup - setup virtual tunnel device
1850  *   @dev: virtual device associated with tunnel
1851  *
1852  * Description:
1853  *   Initialize function pointers and device parameters
1854  **/
1855 
ip6_tnl_dev_setup(struct net_device * dev)1856 static void ip6_tnl_dev_setup(struct net_device *dev)
1857 {
1858 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1859 	dev->header_ops = &ip_tunnel_header_ops;
1860 	dev->needs_free_netdev = true;
1861 	dev->priv_destructor = ip6_dev_free;
1862 
1863 	dev->type = ARPHRD_TUNNEL6;
1864 	dev->flags |= IFF_NOARP;
1865 	dev->addr_len = sizeof(struct in6_addr);
1866 	dev->features |= NETIF_F_LLTX;
1867 	netif_keep_dst(dev);
1868 
1869 	dev->features		|= IPXIPX_FEATURES;
1870 	dev->hw_features	|= IPXIPX_FEATURES;
1871 
1872 	/* This perm addr will be used as interface identifier by IPv6 */
1873 	dev->addr_assign_type = NET_ADDR_RANDOM;
1874 	eth_random_addr(dev->perm_addr);
1875 }
1876 
1877 
1878 /**
1879  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1880  *   @dev: virtual device associated with tunnel
1881  **/
1882 
1883 static inline int
ip6_tnl_dev_init_gen(struct net_device * dev)1884 ip6_tnl_dev_init_gen(struct net_device *dev)
1885 {
1886 	struct ip6_tnl *t = netdev_priv(dev);
1887 	int ret;
1888 	int t_hlen;
1889 
1890 	t->dev = dev;
1891 	t->net = dev_net(dev);
1892 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1893 	if (!dev->tstats)
1894 		return -ENOMEM;
1895 
1896 	ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1897 	if (ret)
1898 		goto free_stats;
1899 
1900 	ret = gro_cells_init(&t->gro_cells, dev);
1901 	if (ret)
1902 		goto destroy_dst;
1903 
1904 	t->tun_hlen = 0;
1905 	t->hlen = t->encap_hlen + t->tun_hlen;
1906 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1907 
1908 	dev->type = ARPHRD_TUNNEL6;
1909 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1910 	dev->mtu = ETH_DATA_LEN - t_hlen;
1911 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1912 		dev->mtu -= 8;
1913 	dev->min_mtu = ETH_MIN_MTU;
1914 	dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1915 
1916 	dev_hold(dev);
1917 	return 0;
1918 
1919 destroy_dst:
1920 	dst_cache_destroy(&t->dst_cache);
1921 free_stats:
1922 	free_percpu(dev->tstats);
1923 	dev->tstats = NULL;
1924 
1925 	return ret;
1926 }
1927 
1928 /**
1929  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1930  *   @dev: virtual device associated with tunnel
1931  **/
1932 
ip6_tnl_dev_init(struct net_device * dev)1933 static int ip6_tnl_dev_init(struct net_device *dev)
1934 {
1935 	struct ip6_tnl *t = netdev_priv(dev);
1936 	int err = ip6_tnl_dev_init_gen(dev);
1937 
1938 	if (err)
1939 		return err;
1940 	ip6_tnl_link_config(t);
1941 	if (t->parms.collect_md)
1942 		netif_keep_dst(dev);
1943 	return 0;
1944 }
1945 
1946 /**
1947  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1948  *   @dev: fallback device
1949  *
1950  * Return: 0
1951  **/
1952 
ip6_fb_tnl_dev_init(struct net_device * dev)1953 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1954 {
1955 	struct ip6_tnl *t = netdev_priv(dev);
1956 	struct net *net = dev_net(dev);
1957 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1958 
1959 	t->parms.proto = IPPROTO_IPV6;
1960 
1961 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1962 	return 0;
1963 }
1964 
ip6_tnl_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1965 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1966 			    struct netlink_ext_ack *extack)
1967 {
1968 	u8 proto;
1969 
1970 	if (!data || !data[IFLA_IPTUN_PROTO])
1971 		return 0;
1972 
1973 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1974 	if (proto != IPPROTO_IPV6 &&
1975 	    proto != IPPROTO_IPIP &&
1976 	    proto != 0)
1977 		return -EINVAL;
1978 
1979 	return 0;
1980 }
1981 
ip6_tnl_netlink_parms(struct nlattr * data[],struct __ip6_tnl_parm * parms)1982 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1983 				  struct __ip6_tnl_parm *parms)
1984 {
1985 	memset(parms, 0, sizeof(*parms));
1986 
1987 	if (!data)
1988 		return;
1989 
1990 	if (data[IFLA_IPTUN_LINK])
1991 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1992 
1993 	if (data[IFLA_IPTUN_LOCAL])
1994 		parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1995 
1996 	if (data[IFLA_IPTUN_REMOTE])
1997 		parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1998 
1999 	if (data[IFLA_IPTUN_TTL])
2000 		parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
2001 
2002 	if (data[IFLA_IPTUN_ENCAP_LIMIT])
2003 		parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
2004 
2005 	if (data[IFLA_IPTUN_FLOWINFO])
2006 		parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
2007 
2008 	if (data[IFLA_IPTUN_FLAGS])
2009 		parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
2010 
2011 	if (data[IFLA_IPTUN_PROTO])
2012 		parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
2013 
2014 	if (data[IFLA_IPTUN_COLLECT_METADATA])
2015 		parms->collect_md = true;
2016 
2017 	if (data[IFLA_IPTUN_FWMARK])
2018 		parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
2019 }
2020 
ip6_tnl_netlink_encap_parms(struct nlattr * data[],struct ip_tunnel_encap * ipencap)2021 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
2022 					struct ip_tunnel_encap *ipencap)
2023 {
2024 	bool ret = false;
2025 
2026 	memset(ipencap, 0, sizeof(*ipencap));
2027 
2028 	if (!data)
2029 		return ret;
2030 
2031 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
2032 		ret = true;
2033 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
2034 	}
2035 
2036 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
2037 		ret = true;
2038 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
2039 	}
2040 
2041 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
2042 		ret = true;
2043 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
2044 	}
2045 
2046 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
2047 		ret = true;
2048 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
2049 	}
2050 
2051 	return ret;
2052 }
2053 
ip6_tnl_newlink(struct net * src_net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)2054 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
2055 			   struct nlattr *tb[], struct nlattr *data[],
2056 			   struct netlink_ext_ack *extack)
2057 {
2058 	struct net *net = dev_net(dev);
2059 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2060 	struct ip_tunnel_encap ipencap;
2061 	struct ip6_tnl *nt, *t;
2062 	int err;
2063 
2064 	nt = netdev_priv(dev);
2065 
2066 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2067 		err = ip6_tnl_encap_setup(nt, &ipencap);
2068 		if (err < 0)
2069 			return err;
2070 	}
2071 
2072 	ip6_tnl_netlink_parms(data, &nt->parms);
2073 
2074 	if (nt->parms.collect_md) {
2075 		if (rtnl_dereference(ip6n->collect_md_tun))
2076 			return -EEXIST;
2077 	} else {
2078 		t = ip6_tnl_locate(net, &nt->parms, 0);
2079 		if (!IS_ERR(t))
2080 			return -EEXIST;
2081 	}
2082 
2083 	err = ip6_tnl_create2(dev);
2084 	if (!err && tb[IFLA_MTU])
2085 		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2086 
2087 	return err;
2088 }
2089 
ip6_tnl_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)2090 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2091 			      struct nlattr *data[],
2092 			      struct netlink_ext_ack *extack)
2093 {
2094 	struct ip6_tnl *t = netdev_priv(dev);
2095 	struct __ip6_tnl_parm p;
2096 	struct net *net = t->net;
2097 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2098 	struct ip_tunnel_encap ipencap;
2099 
2100 	if (dev == ip6n->fb_tnl_dev)
2101 		return -EINVAL;
2102 
2103 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2104 		int err = ip6_tnl_encap_setup(t, &ipencap);
2105 
2106 		if (err < 0)
2107 			return err;
2108 	}
2109 	ip6_tnl_netlink_parms(data, &p);
2110 	if (p.collect_md)
2111 		return -EINVAL;
2112 
2113 	t = ip6_tnl_locate(net, &p, 0);
2114 	if (!IS_ERR(t)) {
2115 		if (t->dev != dev)
2116 			return -EEXIST;
2117 	} else
2118 		t = netdev_priv(dev);
2119 
2120 	return ip6_tnl_update(t, &p);
2121 }
2122 
ip6_tnl_dellink(struct net_device * dev,struct list_head * head)2123 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2124 {
2125 	struct net *net = dev_net(dev);
2126 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2127 
2128 	if (dev != ip6n->fb_tnl_dev)
2129 		unregister_netdevice_queue(dev, head);
2130 }
2131 
ip6_tnl_get_size(const struct net_device * dev)2132 static size_t ip6_tnl_get_size(const struct net_device *dev)
2133 {
2134 	return
2135 		/* IFLA_IPTUN_LINK */
2136 		nla_total_size(4) +
2137 		/* IFLA_IPTUN_LOCAL */
2138 		nla_total_size(sizeof(struct in6_addr)) +
2139 		/* IFLA_IPTUN_REMOTE */
2140 		nla_total_size(sizeof(struct in6_addr)) +
2141 		/* IFLA_IPTUN_TTL */
2142 		nla_total_size(1) +
2143 		/* IFLA_IPTUN_ENCAP_LIMIT */
2144 		nla_total_size(1) +
2145 		/* IFLA_IPTUN_FLOWINFO */
2146 		nla_total_size(4) +
2147 		/* IFLA_IPTUN_FLAGS */
2148 		nla_total_size(4) +
2149 		/* IFLA_IPTUN_PROTO */
2150 		nla_total_size(1) +
2151 		/* IFLA_IPTUN_ENCAP_TYPE */
2152 		nla_total_size(2) +
2153 		/* IFLA_IPTUN_ENCAP_FLAGS */
2154 		nla_total_size(2) +
2155 		/* IFLA_IPTUN_ENCAP_SPORT */
2156 		nla_total_size(2) +
2157 		/* IFLA_IPTUN_ENCAP_DPORT */
2158 		nla_total_size(2) +
2159 		/* IFLA_IPTUN_COLLECT_METADATA */
2160 		nla_total_size(0) +
2161 		/* IFLA_IPTUN_FWMARK */
2162 		nla_total_size(4) +
2163 		0;
2164 }
2165 
ip6_tnl_fill_info(struct sk_buff * skb,const struct net_device * dev)2166 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2167 {
2168 	struct ip6_tnl *tunnel = netdev_priv(dev);
2169 	struct __ip6_tnl_parm *parm = &tunnel->parms;
2170 
2171 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2172 	    nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2173 	    nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2174 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2175 	    nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2176 	    nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2177 	    nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2178 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2179 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2180 		goto nla_put_failure;
2181 
2182 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2183 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2184 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2185 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2186 		goto nla_put_failure;
2187 
2188 	if (parm->collect_md)
2189 		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2190 			goto nla_put_failure;
2191 
2192 	return 0;
2193 
2194 nla_put_failure:
2195 	return -EMSGSIZE;
2196 }
2197 
ip6_tnl_get_link_net(const struct net_device * dev)2198 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2199 {
2200 	struct ip6_tnl *tunnel = netdev_priv(dev);
2201 
2202 	return tunnel->net;
2203 }
2204 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2205 
2206 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2207 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
2208 	[IFLA_IPTUN_LOCAL]		= { .len = sizeof(struct in6_addr) },
2209 	[IFLA_IPTUN_REMOTE]		= { .len = sizeof(struct in6_addr) },
2210 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
2211 	[IFLA_IPTUN_ENCAP_LIMIT]	= { .type = NLA_U8 },
2212 	[IFLA_IPTUN_FLOWINFO]		= { .type = NLA_U32 },
2213 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U32 },
2214 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
2215 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
2216 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
2217 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
2218 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
2219 	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
2220 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
2221 };
2222 
2223 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2224 	.kind		= "ip6tnl",
2225 	.maxtype	= IFLA_IPTUN_MAX,
2226 	.policy		= ip6_tnl_policy,
2227 	.priv_size	= sizeof(struct ip6_tnl),
2228 	.setup		= ip6_tnl_dev_setup,
2229 	.validate	= ip6_tnl_validate,
2230 	.newlink	= ip6_tnl_newlink,
2231 	.changelink	= ip6_tnl_changelink,
2232 	.dellink	= ip6_tnl_dellink,
2233 	.get_size	= ip6_tnl_get_size,
2234 	.fill_info	= ip6_tnl_fill_info,
2235 	.get_link_net	= ip6_tnl_get_link_net,
2236 };
2237 
2238 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2239 	.handler	= ip4ip6_rcv,
2240 	.err_handler	= ip4ip6_err,
2241 	.priority	=	1,
2242 };
2243 
2244 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2245 	.handler	= ip6ip6_rcv,
2246 	.err_handler	= ip6ip6_err,
2247 	.priority	=	1,
2248 };
2249 
2250 static struct xfrm6_tunnel mplsip6_handler __read_mostly = {
2251 	.handler	= mplsip6_rcv,
2252 	.err_handler	= mplsip6_err,
2253 	.priority	=	1,
2254 };
2255 
ip6_tnl_destroy_tunnels(struct net * net,struct list_head * list)2256 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2257 {
2258 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2259 	struct net_device *dev, *aux;
2260 	int h;
2261 	struct ip6_tnl *t;
2262 
2263 	for_each_netdev_safe(net, dev, aux)
2264 		if (dev->rtnl_link_ops == &ip6_link_ops)
2265 			unregister_netdevice_queue(dev, list);
2266 
2267 	for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2268 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
2269 		while (t) {
2270 			/* If dev is in the same netns, it has already
2271 			 * been added to the list by the previous loop.
2272 			 */
2273 			if (!net_eq(dev_net(t->dev), net))
2274 				unregister_netdevice_queue(t->dev, list);
2275 			t = rtnl_dereference(t->next);
2276 		}
2277 	}
2278 
2279 	t = rtnl_dereference(ip6n->tnls_wc[0]);
2280 	while (t) {
2281 		/* If dev is in the same netns, it has already
2282 		 * been added to the list by the previous loop.
2283 		 */
2284 		if (!net_eq(dev_net(t->dev), net))
2285 			unregister_netdevice_queue(t->dev, list);
2286 		t = rtnl_dereference(t->next);
2287 	}
2288 }
2289 
ip6_tnl_init_net(struct net * net)2290 static int __net_init ip6_tnl_init_net(struct net *net)
2291 {
2292 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2293 	struct ip6_tnl *t = NULL;
2294 	int err;
2295 
2296 	ip6n->tnls[0] = ip6n->tnls_wc;
2297 	ip6n->tnls[1] = ip6n->tnls_r_l;
2298 
2299 	if (!net_has_fallback_tunnels(net))
2300 		return 0;
2301 	err = -ENOMEM;
2302 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2303 					NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2304 
2305 	if (!ip6n->fb_tnl_dev)
2306 		goto err_alloc_dev;
2307 	dev_net_set(ip6n->fb_tnl_dev, net);
2308 	ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2309 	/* FB netdevice is special: we have one, and only one per netns.
2310 	 * Allowing to move it to another netns is clearly unsafe.
2311 	 */
2312 	ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2313 
2314 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2315 	if (err < 0)
2316 		goto err_register;
2317 
2318 	err = register_netdev(ip6n->fb_tnl_dev);
2319 	if (err < 0)
2320 		goto err_register;
2321 
2322 	t = netdev_priv(ip6n->fb_tnl_dev);
2323 
2324 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2325 	return 0;
2326 
2327 err_register:
2328 	free_netdev(ip6n->fb_tnl_dev);
2329 err_alloc_dev:
2330 	return err;
2331 }
2332 
ip6_tnl_exit_batch_net(struct list_head * net_list)2333 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2334 {
2335 	struct net *net;
2336 	LIST_HEAD(list);
2337 
2338 	rtnl_lock();
2339 	list_for_each_entry(net, net_list, exit_list)
2340 		ip6_tnl_destroy_tunnels(net, &list);
2341 	unregister_netdevice_many(&list);
2342 	rtnl_unlock();
2343 }
2344 
2345 static struct pernet_operations ip6_tnl_net_ops = {
2346 	.init = ip6_tnl_init_net,
2347 	.exit_batch = ip6_tnl_exit_batch_net,
2348 	.id   = &ip6_tnl_net_id,
2349 	.size = sizeof(struct ip6_tnl_net),
2350 };
2351 
2352 /**
2353  * ip6_tunnel_init - register protocol and reserve needed resources
2354  *
2355  * Return: 0 on success
2356  **/
2357 
ip6_tunnel_init(void)2358 static int __init ip6_tunnel_init(void)
2359 {
2360 	int  err;
2361 
2362 	if (!ipv6_mod_enabled())
2363 		return -EOPNOTSUPP;
2364 
2365 	err = register_pernet_device(&ip6_tnl_net_ops);
2366 	if (err < 0)
2367 		goto out_pernet;
2368 
2369 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2370 	if (err < 0) {
2371 		pr_err("%s: can't register ip4ip6\n", __func__);
2372 		goto out_ip4ip6;
2373 	}
2374 
2375 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2376 	if (err < 0) {
2377 		pr_err("%s: can't register ip6ip6\n", __func__);
2378 		goto out_ip6ip6;
2379 	}
2380 
2381 	if (ip6_tnl_mpls_supported()) {
2382 		err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS);
2383 		if (err < 0) {
2384 			pr_err("%s: can't register mplsip6\n", __func__);
2385 			goto out_mplsip6;
2386 		}
2387 	}
2388 
2389 	err = rtnl_link_register(&ip6_link_ops);
2390 	if (err < 0)
2391 		goto rtnl_link_failed;
2392 
2393 	return 0;
2394 
2395 rtnl_link_failed:
2396 	if (ip6_tnl_mpls_supported())
2397 		xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS);
2398 out_mplsip6:
2399 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2400 out_ip6ip6:
2401 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2402 out_ip4ip6:
2403 	unregister_pernet_device(&ip6_tnl_net_ops);
2404 out_pernet:
2405 	return err;
2406 }
2407 
2408 /**
2409  * ip6_tunnel_cleanup - free resources and unregister protocol
2410  **/
2411 
ip6_tunnel_cleanup(void)2412 static void __exit ip6_tunnel_cleanup(void)
2413 {
2414 	rtnl_link_unregister(&ip6_link_ops);
2415 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2416 		pr_info("%s: can't deregister ip4ip6\n", __func__);
2417 
2418 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2419 		pr_info("%s: can't deregister ip6ip6\n", __func__);
2420 
2421 	if (ip6_tnl_mpls_supported() &&
2422 	    xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS))
2423 		pr_info("%s: can't deregister mplsip6\n", __func__);
2424 	unregister_pernet_device(&ip6_tnl_net_ops);
2425 }
2426 
2427 module_init(ip6_tunnel_init);
2428 module_exit(ip6_tunnel_cleanup);
2429