xref: /openbmc/linux/net/ipv6/ip6_tunnel.c (revision 0c874100)
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 <linux/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 #include <net/dst_metadata.h>
61 
62 MODULE_AUTHOR("Ville Nuorvala");
63 MODULE_DESCRIPTION("IPv6 tunneling device");
64 MODULE_LICENSE("GPL");
65 MODULE_ALIAS_RTNL_LINK("ip6tnl");
66 MODULE_ALIAS_NETDEV("ip6tnl0");
67 
68 #define IP6_TUNNEL_HASH_SIZE_SHIFT  5
69 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
70 
71 static bool log_ecn_error = true;
72 module_param(log_ecn_error, bool, 0644);
73 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
74 
75 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
76 {
77 	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
78 
79 	return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
80 }
81 
82 static int ip6_tnl_dev_init(struct net_device *dev);
83 static void ip6_tnl_dev_setup(struct net_device *dev);
84 static struct rtnl_link_ops ip6_link_ops __read_mostly;
85 
86 static unsigned int ip6_tnl_net_id __read_mostly;
87 struct ip6_tnl_net {
88 	/* the IPv6 tunnel fallback device */
89 	struct net_device *fb_tnl_dev;
90 	/* lists for storing tunnels in use */
91 	struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
92 	struct ip6_tnl __rcu *tnls_wc[1];
93 	struct ip6_tnl __rcu **tnls[2];
94 	struct ip6_tnl __rcu *collect_md_tun;
95 };
96 
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 /**
128  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
129  *   @remote: the address of the tunnel exit-point
130  *   @local: the address of the tunnel entry-point
131  *
132  * Return:
133  *   tunnel matching given end-points if found,
134  *   else fallback tunnel if its device is up,
135  *   else %NULL
136  **/
137 
138 #define for_each_ip6_tunnel_rcu(start) \
139 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
140 
141 static struct ip6_tnl *
142 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
143 {
144 	unsigned int hash = HASH(remote, local);
145 	struct ip6_tnl *t;
146 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
147 	struct in6_addr any;
148 
149 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
150 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
151 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
152 		    (t->dev->flags & IFF_UP))
153 			return t;
154 	}
155 
156 	memset(&any, 0, sizeof(any));
157 	hash = HASH(&any, local);
158 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
159 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
160 		    ipv6_addr_any(&t->parms.raddr) &&
161 		    (t->dev->flags & IFF_UP))
162 			return t;
163 	}
164 
165 	hash = HASH(remote, &any);
166 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
167 		if (ipv6_addr_equal(remote, &t->parms.raddr) &&
168 		    ipv6_addr_any(&t->parms.laddr) &&
169 		    (t->dev->flags & IFF_UP))
170 			return t;
171 	}
172 
173 	t = rcu_dereference(ip6n->collect_md_tun);
174 	if (t && t->dev->flags & IFF_UP)
175 		return t;
176 
177 	t = rcu_dereference(ip6n->tnls_wc[0]);
178 	if (t && (t->dev->flags & IFF_UP))
179 		return t;
180 
181 	return NULL;
182 }
183 
184 /**
185  * ip6_tnl_bucket - get head of list matching given tunnel parameters
186  *   @p: parameters containing tunnel end-points
187  *
188  * Description:
189  *   ip6_tnl_bucket() returns the head of the list matching the
190  *   &struct in6_addr entries laddr and raddr in @p.
191  *
192  * Return: head of IPv6 tunnel list
193  **/
194 
195 static struct ip6_tnl __rcu **
196 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
197 {
198 	const struct in6_addr *remote = &p->raddr;
199 	const struct in6_addr *local = &p->laddr;
200 	unsigned int h = 0;
201 	int prio = 0;
202 
203 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
204 		prio = 1;
205 		h = HASH(remote, local);
206 	}
207 	return &ip6n->tnls[prio][h];
208 }
209 
210 /**
211  * ip6_tnl_link - add tunnel to hash table
212  *   @t: tunnel to be added
213  **/
214 
215 static void
216 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
217 {
218 	struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
219 
220 	if (t->parms.collect_md)
221 		rcu_assign_pointer(ip6n->collect_md_tun, t);
222 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
223 	rcu_assign_pointer(*tp, t);
224 }
225 
226 /**
227  * ip6_tnl_unlink - remove tunnel from hash table
228  *   @t: tunnel to be removed
229  **/
230 
231 static void
232 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
233 {
234 	struct ip6_tnl __rcu **tp;
235 	struct ip6_tnl *iter;
236 
237 	if (t->parms.collect_md)
238 		rcu_assign_pointer(ip6n->collect_md_tun, NULL);
239 
240 	for (tp = ip6_tnl_bucket(ip6n, &t->parms);
241 	     (iter = rtnl_dereference(*tp)) != NULL;
242 	     tp = &iter->next) {
243 		if (t == iter) {
244 			rcu_assign_pointer(*tp, t->next);
245 			break;
246 		}
247 	}
248 }
249 
250 static void ip6_dev_free(struct net_device *dev)
251 {
252 	struct ip6_tnl *t = netdev_priv(dev);
253 
254 	gro_cells_destroy(&t->gro_cells);
255 	dst_cache_destroy(&t->dst_cache);
256 	free_percpu(dev->tstats);
257 }
258 
259 static int ip6_tnl_create2(struct net_device *dev)
260 {
261 	struct ip6_tnl *t = netdev_priv(dev);
262 	struct net *net = dev_net(dev);
263 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
264 	int err;
265 
266 	t = netdev_priv(dev);
267 
268 	dev->rtnl_link_ops = &ip6_link_ops;
269 	err = register_netdevice(dev);
270 	if (err < 0)
271 		goto out;
272 
273 	strcpy(t->parms.name, dev->name);
274 
275 	dev_hold(dev);
276 	ip6_tnl_link(ip6n, t);
277 	return 0;
278 
279 out:
280 	return err;
281 }
282 
283 /**
284  * ip6_tnl_create - create a new tunnel
285  *   @p: tunnel parameters
286  *   @pt: pointer to new tunnel
287  *
288  * Description:
289  *   Create tunnel matching given parameters.
290  *
291  * Return:
292  *   created tunnel or error pointer
293  **/
294 
295 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
296 {
297 	struct net_device *dev;
298 	struct ip6_tnl *t;
299 	char name[IFNAMSIZ];
300 	int err = -E2BIG;
301 
302 	if (p->name[0]) {
303 		if (!dev_valid_name(p->name))
304 			goto failed;
305 		strlcpy(name, p->name, IFNAMSIZ);
306 	} else {
307 		sprintf(name, "ip6tnl%%d");
308 	}
309 	err = -ENOMEM;
310 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
311 			   ip6_tnl_dev_setup);
312 	if (!dev)
313 		goto failed;
314 
315 	dev_net_set(dev, net);
316 
317 	t = netdev_priv(dev);
318 	t->parms = *p;
319 	t->net = dev_net(dev);
320 	err = ip6_tnl_create2(dev);
321 	if (err < 0)
322 		goto failed_free;
323 
324 	return t;
325 
326 failed_free:
327 	free_netdev(dev);
328 failed:
329 	return ERR_PTR(err);
330 }
331 
332 /**
333  * ip6_tnl_locate - find or create tunnel matching given parameters
334  *   @p: tunnel parameters
335  *   @create: != 0 if allowed to create new tunnel if no match found
336  *
337  * Description:
338  *   ip6_tnl_locate() first tries to locate an existing tunnel
339  *   based on @parms. If this is unsuccessful, but @create is set a new
340  *   tunnel device is created and registered for use.
341  *
342  * Return:
343  *   matching tunnel or error pointer
344  **/
345 
346 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
347 		struct __ip6_tnl_parm *p, int create)
348 {
349 	const struct in6_addr *remote = &p->raddr;
350 	const struct in6_addr *local = &p->laddr;
351 	struct ip6_tnl __rcu **tp;
352 	struct ip6_tnl *t;
353 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
354 
355 	for (tp = ip6_tnl_bucket(ip6n, p);
356 	     (t = rtnl_dereference(*tp)) != NULL;
357 	     tp = &t->next) {
358 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
359 		    ipv6_addr_equal(remote, &t->parms.raddr)) {
360 			if (create)
361 				return ERR_PTR(-EEXIST);
362 
363 			return t;
364 		}
365 	}
366 	if (!create)
367 		return ERR_PTR(-ENODEV);
368 	return ip6_tnl_create(net, p);
369 }
370 
371 /**
372  * ip6_tnl_dev_uninit - tunnel device uninitializer
373  *   @dev: the device to be destroyed
374  *
375  * Description:
376  *   ip6_tnl_dev_uninit() removes tunnel from its list
377  **/
378 
379 static void
380 ip6_tnl_dev_uninit(struct net_device *dev)
381 {
382 	struct ip6_tnl *t = netdev_priv(dev);
383 	struct net *net = t->net;
384 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
385 
386 	if (dev == ip6n->fb_tnl_dev)
387 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
388 	else
389 		ip6_tnl_unlink(ip6n, t);
390 	dst_cache_reset(&t->dst_cache);
391 	dev_put(dev);
392 }
393 
394 /**
395  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
396  *   @skb: received socket buffer
397  *
398  * Return:
399  *   0 if none was found,
400  *   else index to encapsulation limit
401  **/
402 
403 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
404 {
405 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
406 	unsigned int nhoff = raw - skb->data;
407 	unsigned int off = nhoff + sizeof(*ipv6h);
408 	u8 next, nexthdr = ipv6h->nexthdr;
409 
410 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
411 		struct ipv6_opt_hdr *hdr;
412 		u16 optlen;
413 
414 		if (!pskb_may_pull(skb, off + sizeof(*hdr)))
415 			break;
416 
417 		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
418 		if (nexthdr == NEXTHDR_FRAGMENT) {
419 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
420 			if (frag_hdr->frag_off)
421 				break;
422 			optlen = 8;
423 		} else if (nexthdr == NEXTHDR_AUTH) {
424 			optlen = (hdr->hdrlen + 2) << 2;
425 		} else {
426 			optlen = ipv6_optlen(hdr);
427 		}
428 		/* cache hdr->nexthdr, since pskb_may_pull() might
429 		 * invalidate hdr
430 		 */
431 		next = hdr->nexthdr;
432 		if (nexthdr == NEXTHDR_DEST) {
433 			u16 i = 2;
434 
435 			/* Remember : hdr is no longer valid at this point. */
436 			if (!pskb_may_pull(skb, off + optlen))
437 				break;
438 
439 			while (1) {
440 				struct ipv6_tlv_tnl_enc_lim *tel;
441 
442 				/* No more room for encapsulation limit */
443 				if (i + sizeof(*tel) > optlen)
444 					break;
445 
446 				tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
447 				/* return index of option if found and valid */
448 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
449 				    tel->length == 1)
450 					return i + off - nhoff;
451 				/* else jump to next option */
452 				if (tel->type)
453 					i += tel->length + 2;
454 				else
455 					i++;
456 			}
457 		}
458 		nexthdr = next;
459 		off += optlen;
460 	}
461 	return 0;
462 }
463 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
464 
465 /**
466  * ip6_tnl_err - tunnel error handler
467  *
468  * Description:
469  *   ip6_tnl_err() should handle errors in the tunnel according
470  *   to the specifications in RFC 2473.
471  **/
472 
473 static int
474 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
475 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
476 {
477 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
478 	struct net *net = dev_net(skb->dev);
479 	u8 rel_type = ICMPV6_DEST_UNREACH;
480 	u8 rel_code = ICMPV6_ADDR_UNREACH;
481 	__u32 rel_info = 0;
482 	struct ip6_tnl *t;
483 	int err = -ENOENT;
484 	int rel_msg = 0;
485 	u8 tproto;
486 	__u16 len;
487 
488 	/* If the packet doesn't contain the original IPv6 header we are
489 	   in trouble since we might need the source address for further
490 	   processing of the error. */
491 
492 	rcu_read_lock();
493 	t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
494 	if (!t)
495 		goto out;
496 
497 	tproto = READ_ONCE(t->parms.proto);
498 	if (tproto != ipproto && tproto != 0)
499 		goto out;
500 
501 	err = 0;
502 
503 	switch (*type) {
504 		struct ipv6_tlv_tnl_enc_lim *tel;
505 		__u32 mtu, teli;
506 	case ICMPV6_DEST_UNREACH:
507 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
508 				    t->parms.name);
509 		rel_msg = 1;
510 		break;
511 	case ICMPV6_TIME_EXCEED:
512 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
513 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
514 					    t->parms.name);
515 			rel_msg = 1;
516 		}
517 		break;
518 	case ICMPV6_PARAMPROB:
519 		teli = 0;
520 		if ((*code) == ICMPV6_HDR_FIELD)
521 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
522 
523 		if (teli && teli == *info - 2) {
524 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
525 			if (tel->encap_limit == 0) {
526 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
527 						    t->parms.name);
528 				rel_msg = 1;
529 			}
530 		} else {
531 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
532 					    t->parms.name);
533 		}
534 		break;
535 	case ICMPV6_PKT_TOOBIG:
536 		ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
537 				sock_net_uid(net, NULL));
538 		mtu = *info - offset;
539 		if (mtu < IPV6_MIN_MTU)
540 			mtu = IPV6_MIN_MTU;
541 		len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
542 		if (len > mtu) {
543 			rel_type = ICMPV6_PKT_TOOBIG;
544 			rel_code = 0;
545 			rel_info = mtu;
546 			rel_msg = 1;
547 		}
548 		break;
549 	case NDISC_REDIRECT:
550 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
551 			     sock_net_uid(net, NULL));
552 		break;
553 	}
554 
555 	*type = rel_type;
556 	*code = rel_code;
557 	*info = rel_info;
558 	*msg = rel_msg;
559 
560 out:
561 	rcu_read_unlock();
562 	return err;
563 }
564 
565 static int
566 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
567 	   u8 type, u8 code, int offset, __be32 info)
568 {
569 	__u32 rel_info = ntohl(info);
570 	const struct iphdr *eiph;
571 	struct sk_buff *skb2;
572 	int err, rel_msg = 0;
573 	u8 rel_type = type;
574 	u8 rel_code = code;
575 	struct rtable *rt;
576 	struct flowi4 fl4;
577 
578 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
579 			  &rel_msg, &rel_info, offset);
580 	if (err < 0)
581 		return err;
582 
583 	if (rel_msg == 0)
584 		return 0;
585 
586 	switch (rel_type) {
587 	case ICMPV6_DEST_UNREACH:
588 		if (rel_code != ICMPV6_ADDR_UNREACH)
589 			return 0;
590 		rel_type = ICMP_DEST_UNREACH;
591 		rel_code = ICMP_HOST_UNREACH;
592 		break;
593 	case ICMPV6_PKT_TOOBIG:
594 		if (rel_code != 0)
595 			return 0;
596 		rel_type = ICMP_DEST_UNREACH;
597 		rel_code = ICMP_FRAG_NEEDED;
598 		break;
599 	default:
600 		return 0;
601 	}
602 
603 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
604 		return 0;
605 
606 	skb2 = skb_clone(skb, GFP_ATOMIC);
607 	if (!skb2)
608 		return 0;
609 
610 	skb_dst_drop(skb2);
611 
612 	skb_pull(skb2, offset);
613 	skb_reset_network_header(skb2);
614 	eiph = ip_hdr(skb2);
615 
616 	/* Try to guess incoming interface */
617 	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
618 				   0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
619 	if (IS_ERR(rt))
620 		goto out;
621 
622 	skb2->dev = rt->dst.dev;
623 	ip_rt_put(rt);
624 
625 	/* route "incoming" packet */
626 	if (rt->rt_flags & RTCF_LOCAL) {
627 		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
628 					   eiph->daddr, eiph->saddr, 0, 0,
629 					   IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
630 		if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL) {
631 			if (!IS_ERR(rt))
632 				ip_rt_put(rt);
633 			goto out;
634 		}
635 		skb_dst_set(skb2, &rt->dst);
636 	} else {
637 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
638 				   skb2->dev) ||
639 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
640 			goto out;
641 	}
642 
643 	/* change mtu on this route */
644 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
645 		if (rel_info > dst_mtu(skb_dst(skb2)))
646 			goto out;
647 
648 		skb_dst_update_pmtu(skb2, rel_info);
649 	}
650 
651 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
652 
653 out:
654 	kfree_skb(skb2);
655 	return 0;
656 }
657 
658 static int
659 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
660 	   u8 type, u8 code, int offset, __be32 info)
661 {
662 	__u32 rel_info = ntohl(info);
663 	int err, rel_msg = 0;
664 	u8 rel_type = type;
665 	u8 rel_code = code;
666 
667 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
668 			  &rel_msg, &rel_info, offset);
669 	if (err < 0)
670 		return err;
671 
672 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
673 		struct rt6_info *rt;
674 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
675 
676 		if (!skb2)
677 			return 0;
678 
679 		skb_dst_drop(skb2);
680 		skb_pull(skb2, offset);
681 		skb_reset_network_header(skb2);
682 
683 		/* Try to guess incoming interface */
684 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
685 				NULL, 0, skb2, 0);
686 
687 		if (rt && rt->dst.dev)
688 			skb2->dev = rt->dst.dev;
689 
690 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
691 
692 		ip6_rt_put(rt);
693 
694 		kfree_skb(skb2);
695 	}
696 
697 	return 0;
698 }
699 
700 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
701 				       const struct ipv6hdr *ipv6h,
702 				       struct sk_buff *skb)
703 {
704 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
705 
706 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
707 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
708 
709 	return IP6_ECN_decapsulate(ipv6h, skb);
710 }
711 
712 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
713 				       const struct ipv6hdr *ipv6h,
714 				       struct sk_buff *skb)
715 {
716 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
717 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
718 
719 	return IP6_ECN_decapsulate(ipv6h, skb);
720 }
721 
722 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
723 			     const struct in6_addr *laddr,
724 			     const struct in6_addr *raddr)
725 {
726 	struct __ip6_tnl_parm *p = &t->parms;
727 	int ltype = ipv6_addr_type(laddr);
728 	int rtype = ipv6_addr_type(raddr);
729 	__u32 flags = 0;
730 
731 	if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
732 		flags = IP6_TNL_F_CAP_PER_PACKET;
733 	} else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
734 		   rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
735 		   !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
736 		   (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
737 		if (ltype&IPV6_ADDR_UNICAST)
738 			flags |= IP6_TNL_F_CAP_XMIT;
739 		if (rtype&IPV6_ADDR_UNICAST)
740 			flags |= IP6_TNL_F_CAP_RCV;
741 	}
742 	return flags;
743 }
744 EXPORT_SYMBOL(ip6_tnl_get_cap);
745 
746 /* called with rcu_read_lock() */
747 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
748 				  const struct in6_addr *laddr,
749 				  const struct in6_addr *raddr)
750 {
751 	struct __ip6_tnl_parm *p = &t->parms;
752 	int ret = 0;
753 	struct net *net = t->net;
754 
755 	if ((p->flags & IP6_TNL_F_CAP_RCV) ||
756 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
757 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
758 		struct net_device *ldev = NULL;
759 
760 		if (p->link)
761 			ldev = dev_get_by_index_rcu(net, p->link);
762 
763 		if ((ipv6_addr_is_multicast(laddr) ||
764 		     likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
765 						    0, IFA_F_TENTATIVE))) &&
766 		    ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
767 		     likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
768 						     0, IFA_F_TENTATIVE))))
769 			ret = 1;
770 	}
771 	return ret;
772 }
773 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
774 
775 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
776 			 const struct tnl_ptk_info *tpi,
777 			 struct metadata_dst *tun_dst,
778 			 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
779 						const struct ipv6hdr *ipv6h,
780 						struct sk_buff *skb),
781 			 bool log_ecn_err)
782 {
783 	struct pcpu_sw_netstats *tstats;
784 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
785 	int err;
786 
787 	if ((!(tpi->flags & TUNNEL_CSUM) &&
788 	     (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
789 	    ((tpi->flags & TUNNEL_CSUM) &&
790 	     !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
791 		tunnel->dev->stats.rx_crc_errors++;
792 		tunnel->dev->stats.rx_errors++;
793 		goto drop;
794 	}
795 
796 	if (tunnel->parms.i_flags & TUNNEL_SEQ) {
797 		if (!(tpi->flags & TUNNEL_SEQ) ||
798 		    (tunnel->i_seqno &&
799 		     (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
800 			tunnel->dev->stats.rx_fifo_errors++;
801 			tunnel->dev->stats.rx_errors++;
802 			goto drop;
803 		}
804 		tunnel->i_seqno = ntohl(tpi->seq) + 1;
805 	}
806 
807 	skb->protocol = tpi->proto;
808 
809 	/* Warning: All skb pointers will be invalidated! */
810 	if (tunnel->dev->type == ARPHRD_ETHER) {
811 		if (!pskb_may_pull(skb, ETH_HLEN)) {
812 			tunnel->dev->stats.rx_length_errors++;
813 			tunnel->dev->stats.rx_errors++;
814 			goto drop;
815 		}
816 
817 		ipv6h = ipv6_hdr(skb);
818 		skb->protocol = eth_type_trans(skb, tunnel->dev);
819 		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
820 	} else {
821 		skb->dev = tunnel->dev;
822 	}
823 
824 	skb_reset_network_header(skb);
825 	memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
826 
827 	__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
828 
829 	err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
830 	if (unlikely(err)) {
831 		if (log_ecn_err)
832 			net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
833 					     &ipv6h->saddr,
834 					     ipv6_get_dsfield(ipv6h));
835 		if (err > 1) {
836 			++tunnel->dev->stats.rx_frame_errors;
837 			++tunnel->dev->stats.rx_errors;
838 			goto drop;
839 		}
840 	}
841 
842 	tstats = this_cpu_ptr(tunnel->dev->tstats);
843 	u64_stats_update_begin(&tstats->syncp);
844 	tstats->rx_packets++;
845 	tstats->rx_bytes += skb->len;
846 	u64_stats_update_end(&tstats->syncp);
847 
848 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
849 
850 	if (tun_dst)
851 		skb_dst_set(skb, (struct dst_entry *)tun_dst);
852 
853 	gro_cells_receive(&tunnel->gro_cells, skb);
854 	return 0;
855 
856 drop:
857 	if (tun_dst)
858 		dst_release((struct dst_entry *)tun_dst);
859 	kfree_skb(skb);
860 	return 0;
861 }
862 
863 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
864 		const struct tnl_ptk_info *tpi,
865 		struct metadata_dst *tun_dst,
866 		bool log_ecn_err)
867 {
868 	return __ip6_tnl_rcv(t, skb, tpi, tun_dst, ip6ip6_dscp_ecn_decapsulate,
869 			     log_ecn_err);
870 }
871 EXPORT_SYMBOL(ip6_tnl_rcv);
872 
873 static const struct tnl_ptk_info tpi_v6 = {
874 	/* no tunnel info required for ipxip6. */
875 	.proto = htons(ETH_P_IPV6),
876 };
877 
878 static const struct tnl_ptk_info tpi_v4 = {
879 	/* no tunnel info required for ipxip6. */
880 	.proto = htons(ETH_P_IP),
881 };
882 
883 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
884 		      const struct tnl_ptk_info *tpi,
885 		      int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
886 						  const struct ipv6hdr *ipv6h,
887 						  struct sk_buff *skb))
888 {
889 	struct ip6_tnl *t;
890 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
891 	struct metadata_dst *tun_dst = NULL;
892 	int ret = -1;
893 
894 	rcu_read_lock();
895 	t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
896 
897 	if (t) {
898 		u8 tproto = READ_ONCE(t->parms.proto);
899 
900 		if (tproto != ipproto && tproto != 0)
901 			goto drop;
902 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
903 			goto drop;
904 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
905 			goto drop;
906 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
907 			goto drop;
908 		if (t->parms.collect_md) {
909 			tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
910 			if (!tun_dst)
911 				goto drop;
912 		}
913 		ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
914 				    log_ecn_error);
915 	}
916 
917 	rcu_read_unlock();
918 
919 	return ret;
920 
921 drop:
922 	rcu_read_unlock();
923 	kfree_skb(skb);
924 	return 0;
925 }
926 
927 static int ip4ip6_rcv(struct sk_buff *skb)
928 {
929 	return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
930 			  ip4ip6_dscp_ecn_decapsulate);
931 }
932 
933 static int ip6ip6_rcv(struct sk_buff *skb)
934 {
935 	return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
936 			  ip6ip6_dscp_ecn_decapsulate);
937 }
938 
939 struct ipv6_tel_txoption {
940 	struct ipv6_txoptions ops;
941 	__u8 dst_opt[8];
942 };
943 
944 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
945 {
946 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
947 
948 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
949 	opt->dst_opt[3] = 1;
950 	opt->dst_opt[4] = encap_limit;
951 	opt->dst_opt[5] = IPV6_TLV_PADN;
952 	opt->dst_opt[6] = 1;
953 
954 	opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
955 	opt->ops.opt_nflen = 8;
956 }
957 
958 /**
959  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
960  *   @t: the outgoing tunnel device
961  *   @hdr: IPv6 header from the incoming packet
962  *
963  * Description:
964  *   Avoid trivial tunneling loop by checking that tunnel exit-point
965  *   doesn't match source of incoming packet.
966  *
967  * Return:
968  *   1 if conflict,
969  *   0 else
970  **/
971 
972 static inline bool
973 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
974 {
975 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
976 }
977 
978 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
979 		     const struct in6_addr *laddr,
980 		     const struct in6_addr *raddr)
981 {
982 	struct __ip6_tnl_parm *p = &t->parms;
983 	int ret = 0;
984 	struct net *net = t->net;
985 
986 	if (t->parms.collect_md)
987 		return 1;
988 
989 	if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
990 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
991 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
992 		struct net_device *ldev = NULL;
993 
994 		rcu_read_lock();
995 		if (p->link)
996 			ldev = dev_get_by_index_rcu(net, p->link);
997 
998 		if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
999 						      0, IFA_F_TENTATIVE)))
1000 			pr_warn("%s xmit: Local address not yet configured!\n",
1001 				p->name);
1002 		else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1003 			 !ipv6_addr_is_multicast(raddr) &&
1004 			 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1005 							  true, 0, IFA_F_TENTATIVE)))
1006 			pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1007 				p->name);
1008 		else
1009 			ret = 1;
1010 		rcu_read_unlock();
1011 	}
1012 	return ret;
1013 }
1014 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1015 
1016 /**
1017  * ip6_tnl_xmit - encapsulate packet and send
1018  *   @skb: the outgoing socket buffer
1019  *   @dev: the outgoing tunnel device
1020  *   @dsfield: dscp code for outer header
1021  *   @fl6: flow of tunneled packet
1022  *   @encap_limit: encapsulation limit
1023  *   @pmtu: Path MTU is stored if packet is too big
1024  *   @proto: next header value
1025  *
1026  * Description:
1027  *   Build new header and do some sanity checks on the packet before sending
1028  *   it.
1029  *
1030  * Return:
1031  *   0 on success
1032  *   -1 fail
1033  *   %-EMSGSIZE message too big. return mtu in this case.
1034  **/
1035 
1036 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1037 		 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1038 		 __u8 proto)
1039 {
1040 	struct ip6_tnl *t = netdev_priv(dev);
1041 	struct net *net = t->net;
1042 	struct net_device_stats *stats = &t->dev->stats;
1043 	struct ipv6hdr *ipv6h;
1044 	struct ipv6_tel_txoption opt;
1045 	struct dst_entry *dst = NULL, *ndst = NULL;
1046 	struct net_device *tdev;
1047 	int mtu;
1048 	unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1049 	unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1050 	unsigned int max_headroom = psh_hlen;
1051 	bool use_cache = false;
1052 	u8 hop_limit;
1053 	int err = -1;
1054 
1055 	if (t->parms.collect_md) {
1056 		hop_limit = skb_tunnel_info(skb)->key.ttl;
1057 		goto route_lookup;
1058 	} else {
1059 		hop_limit = t->parms.hop_limit;
1060 	}
1061 
1062 	/* NBMA tunnel */
1063 	if (ipv6_addr_any(&t->parms.raddr)) {
1064 		if (skb->protocol == htons(ETH_P_IPV6)) {
1065 			struct in6_addr *addr6;
1066 			struct neighbour *neigh;
1067 			int addr_type;
1068 
1069 			if (!skb_dst(skb))
1070 				goto tx_err_link_failure;
1071 
1072 			neigh = dst_neigh_lookup(skb_dst(skb),
1073 						 &ipv6_hdr(skb)->daddr);
1074 			if (!neigh)
1075 				goto tx_err_link_failure;
1076 
1077 			addr6 = (struct in6_addr *)&neigh->primary_key;
1078 			addr_type = ipv6_addr_type(addr6);
1079 
1080 			if (addr_type == IPV6_ADDR_ANY)
1081 				addr6 = &ipv6_hdr(skb)->daddr;
1082 
1083 			memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1084 			neigh_release(neigh);
1085 		}
1086 	} else if (t->parms.proto != 0 && !(t->parms.flags &
1087 					    (IP6_TNL_F_USE_ORIG_TCLASS |
1088 					     IP6_TNL_F_USE_ORIG_FWMARK))) {
1089 		/* enable the cache only if neither the outer protocol nor the
1090 		 * routing decision depends on the current inner header value
1091 		 */
1092 		use_cache = true;
1093 	}
1094 
1095 	if (use_cache)
1096 		dst = dst_cache_get(&t->dst_cache);
1097 
1098 	if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1099 		goto tx_err_link_failure;
1100 
1101 	if (!dst) {
1102 route_lookup:
1103 		/* add dsfield to flowlabel for route lookup */
1104 		fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1105 
1106 		dst = ip6_route_output(net, NULL, fl6);
1107 
1108 		if (dst->error)
1109 			goto tx_err_link_failure;
1110 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1111 		if (IS_ERR(dst)) {
1112 			err = PTR_ERR(dst);
1113 			dst = NULL;
1114 			goto tx_err_link_failure;
1115 		}
1116 		if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1117 		    ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1118 				       &fl6->daddr, 0, &fl6->saddr))
1119 			goto tx_err_link_failure;
1120 		ndst = dst;
1121 	}
1122 
1123 	tdev = dst->dev;
1124 
1125 	if (tdev == dev) {
1126 		stats->collisions++;
1127 		net_warn_ratelimited("%s: Local routing loop detected!\n",
1128 				     t->parms.name);
1129 		goto tx_err_dst_release;
1130 	}
1131 	mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1132 	if (encap_limit >= 0) {
1133 		max_headroom += 8;
1134 		mtu -= 8;
1135 	}
1136 	mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1137 		       IPV6_MIN_MTU : IPV4_MIN_MTU);
1138 
1139 	skb_dst_update_pmtu(skb, mtu);
1140 	if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1141 		*pmtu = mtu;
1142 		err = -EMSGSIZE;
1143 		goto tx_err_dst_release;
1144 	}
1145 
1146 	if (t->err_count > 0) {
1147 		if (time_before(jiffies,
1148 				t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1149 			t->err_count--;
1150 
1151 			dst_link_failure(skb);
1152 		} else {
1153 			t->err_count = 0;
1154 		}
1155 	}
1156 
1157 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1158 
1159 	/*
1160 	 * Okay, now see if we can stuff it in the buffer as-is.
1161 	 */
1162 	max_headroom += LL_RESERVED_SPACE(tdev);
1163 
1164 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1165 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1166 		struct sk_buff *new_skb;
1167 
1168 		new_skb = skb_realloc_headroom(skb, max_headroom);
1169 		if (!new_skb)
1170 			goto tx_err_dst_release;
1171 
1172 		if (skb->sk)
1173 			skb_set_owner_w(new_skb, skb->sk);
1174 		consume_skb(skb);
1175 		skb = new_skb;
1176 	}
1177 
1178 	if (t->parms.collect_md) {
1179 		if (t->encap.type != TUNNEL_ENCAP_NONE)
1180 			goto tx_err_dst_release;
1181 	} else {
1182 		if (use_cache && ndst)
1183 			dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1184 	}
1185 	skb_dst_set(skb, dst);
1186 
1187 	if (hop_limit == 0) {
1188 		if (skb->protocol == htons(ETH_P_IP))
1189 			hop_limit = ip_hdr(skb)->ttl;
1190 		else if (skb->protocol == htons(ETH_P_IPV6))
1191 			hop_limit = ipv6_hdr(skb)->hop_limit;
1192 		else
1193 			hop_limit = ip6_dst_hoplimit(dst);
1194 	}
1195 
1196 	/* Calculate max headroom for all the headers and adjust
1197 	 * needed_headroom if necessary.
1198 	 */
1199 	max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1200 			+ dst->header_len + t->hlen;
1201 	if (max_headroom > dev->needed_headroom)
1202 		dev->needed_headroom = max_headroom;
1203 
1204 	err = ip6_tnl_encap(skb, t, &proto, fl6);
1205 	if (err)
1206 		return err;
1207 
1208 	if (encap_limit >= 0) {
1209 		init_tel_txopt(&opt, encap_limit);
1210 		ipv6_push_frag_opts(skb, &opt.ops, &proto);
1211 	}
1212 
1213 	skb_push(skb, sizeof(struct ipv6hdr));
1214 	skb_reset_network_header(skb);
1215 	ipv6h = ipv6_hdr(skb);
1216 	ip6_flow_hdr(ipv6h, dsfield,
1217 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1218 	ipv6h->hop_limit = hop_limit;
1219 	ipv6h->nexthdr = proto;
1220 	ipv6h->saddr = fl6->saddr;
1221 	ipv6h->daddr = fl6->daddr;
1222 	ip6tunnel_xmit(NULL, skb, dev);
1223 	return 0;
1224 tx_err_link_failure:
1225 	stats->tx_carrier_errors++;
1226 	dst_link_failure(skb);
1227 tx_err_dst_release:
1228 	dst_release(dst);
1229 	return err;
1230 }
1231 EXPORT_SYMBOL(ip6_tnl_xmit);
1232 
1233 static inline int
1234 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1235 {
1236 	struct ip6_tnl *t = netdev_priv(dev);
1237 	const struct iphdr  *iph;
1238 	int encap_limit = -1;
1239 	struct flowi6 fl6;
1240 	__u8 dsfield;
1241 	__u32 mtu;
1242 	u8 tproto;
1243 	int err;
1244 
1245 	/* ensure we can access the full inner ip header */
1246 	if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1247 		return -1;
1248 
1249 	iph = ip_hdr(skb);
1250 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1251 
1252 	tproto = READ_ONCE(t->parms.proto);
1253 	if (tproto != IPPROTO_IPIP && tproto != 0)
1254 		return -1;
1255 
1256 	if (t->parms.collect_md) {
1257 		struct ip_tunnel_info *tun_info;
1258 		const struct ip_tunnel_key *key;
1259 
1260 		tun_info = skb_tunnel_info(skb);
1261 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1262 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1263 			return -1;
1264 		key = &tun_info->key;
1265 		memset(&fl6, 0, sizeof(fl6));
1266 		fl6.flowi6_proto = IPPROTO_IPIP;
1267 		fl6.saddr = key->u.ipv6.src;
1268 		fl6.daddr = key->u.ipv6.dst;
1269 		fl6.flowlabel = key->label;
1270 		dsfield =  key->tos;
1271 	} else {
1272 		if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1273 			encap_limit = t->parms.encap_limit;
1274 
1275 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1276 		fl6.flowi6_proto = IPPROTO_IPIP;
1277 
1278 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1279 			dsfield = ipv4_get_dsfield(iph);
1280 		else
1281 			dsfield = ip6_tclass(t->parms.flowinfo);
1282 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1283 			fl6.flowi6_mark = skb->mark;
1284 		else
1285 			fl6.flowi6_mark = t->parms.fwmark;
1286 	}
1287 
1288 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1289 
1290 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1291 		return -1;
1292 
1293 	dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1294 
1295 	skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1296 
1297 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1298 			   IPPROTO_IPIP);
1299 	if (err != 0) {
1300 		/* XXX: send ICMP error even if DF is not set. */
1301 		if (err == -EMSGSIZE)
1302 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1303 				  htonl(mtu));
1304 		return -1;
1305 	}
1306 
1307 	return 0;
1308 }
1309 
1310 static inline int
1311 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1312 {
1313 	struct ip6_tnl *t = netdev_priv(dev);
1314 	struct ipv6hdr *ipv6h;
1315 	int encap_limit = -1;
1316 	__u16 offset;
1317 	struct flowi6 fl6;
1318 	__u8 dsfield;
1319 	__u32 mtu;
1320 	u8 tproto;
1321 	int err;
1322 
1323 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
1324 		return -1;
1325 
1326 	ipv6h = ipv6_hdr(skb);
1327 	tproto = READ_ONCE(t->parms.proto);
1328 	if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1329 	    ip6_tnl_addr_conflict(t, ipv6h))
1330 		return -1;
1331 
1332 	if (t->parms.collect_md) {
1333 		struct ip_tunnel_info *tun_info;
1334 		const struct ip_tunnel_key *key;
1335 
1336 		tun_info = skb_tunnel_info(skb);
1337 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1338 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1339 			return -1;
1340 		key = &tun_info->key;
1341 		memset(&fl6, 0, sizeof(fl6));
1342 		fl6.flowi6_proto = IPPROTO_IPV6;
1343 		fl6.saddr = key->u.ipv6.src;
1344 		fl6.daddr = key->u.ipv6.dst;
1345 		fl6.flowlabel = key->label;
1346 		dsfield = key->tos;
1347 	} else {
1348 		offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1349 		/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1350 		ipv6h = ipv6_hdr(skb);
1351 		if (offset > 0) {
1352 			struct ipv6_tlv_tnl_enc_lim *tel;
1353 
1354 			tel = (void *)&skb_network_header(skb)[offset];
1355 			if (tel->encap_limit == 0) {
1356 				icmpv6_send(skb, ICMPV6_PARAMPROB,
1357 					    ICMPV6_HDR_FIELD, offset + 2);
1358 				return -1;
1359 			}
1360 			encap_limit = tel->encap_limit - 1;
1361 		} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1362 			encap_limit = t->parms.encap_limit;
1363 		}
1364 
1365 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1366 		fl6.flowi6_proto = IPPROTO_IPV6;
1367 
1368 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1369 			dsfield = ipv6_get_dsfield(ipv6h);
1370 		else
1371 			dsfield = ip6_tclass(t->parms.flowinfo);
1372 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1373 			fl6.flowlabel |= ip6_flowlabel(ipv6h);
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 	}
1379 
1380 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1381 
1382 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1383 		return -1;
1384 
1385 	dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1386 
1387 	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1388 
1389 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1390 			   IPPROTO_IPV6);
1391 	if (err != 0) {
1392 		if (err == -EMSGSIZE)
1393 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1394 		return -1;
1395 	}
1396 
1397 	return 0;
1398 }
1399 
1400 static netdev_tx_t
1401 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1402 {
1403 	struct ip6_tnl *t = netdev_priv(dev);
1404 	struct net_device_stats *stats = &t->dev->stats;
1405 	int ret;
1406 
1407 	switch (skb->protocol) {
1408 	case htons(ETH_P_IP):
1409 		ret = ip4ip6_tnl_xmit(skb, dev);
1410 		break;
1411 	case htons(ETH_P_IPV6):
1412 		ret = ip6ip6_tnl_xmit(skb, dev);
1413 		break;
1414 	default:
1415 		goto tx_err;
1416 	}
1417 
1418 	if (ret < 0)
1419 		goto tx_err;
1420 
1421 	return NETDEV_TX_OK;
1422 
1423 tx_err:
1424 	stats->tx_errors++;
1425 	stats->tx_dropped++;
1426 	kfree_skb(skb);
1427 	return NETDEV_TX_OK;
1428 }
1429 
1430 static void ip6_tnl_link_config(struct ip6_tnl *t)
1431 {
1432 	struct net_device *dev = t->dev;
1433 	struct __ip6_tnl_parm *p = &t->parms;
1434 	struct flowi6 *fl6 = &t->fl.u.ip6;
1435 	int t_hlen;
1436 
1437 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1438 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1439 
1440 	/* Set up flowi template */
1441 	fl6->saddr = p->laddr;
1442 	fl6->daddr = p->raddr;
1443 	fl6->flowi6_oif = p->link;
1444 	fl6->flowlabel = 0;
1445 
1446 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1447 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1448 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1449 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1450 
1451 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1452 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1453 
1454 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1455 		dev->flags |= IFF_POINTOPOINT;
1456 	else
1457 		dev->flags &= ~IFF_POINTOPOINT;
1458 
1459 	t->tun_hlen = 0;
1460 	t->hlen = t->encap_hlen + t->tun_hlen;
1461 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1462 
1463 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1464 		int strict = (ipv6_addr_type(&p->raddr) &
1465 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1466 
1467 		struct rt6_info *rt = rt6_lookup(t->net,
1468 						 &p->raddr, &p->laddr,
1469 						 p->link, NULL, strict);
1470 
1471 		if (!rt)
1472 			return;
1473 
1474 		if (rt->dst.dev) {
1475 			dev->hard_header_len = rt->dst.dev->hard_header_len +
1476 				t_hlen;
1477 
1478 			dev->mtu = rt->dst.dev->mtu - t_hlen;
1479 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1480 				dev->mtu -= 8;
1481 
1482 			if (dev->mtu < IPV6_MIN_MTU)
1483 				dev->mtu = IPV6_MIN_MTU;
1484 		}
1485 		ip6_rt_put(rt);
1486 	}
1487 }
1488 
1489 /**
1490  * ip6_tnl_change - update the tunnel parameters
1491  *   @t: tunnel to be changed
1492  *   @p: tunnel configuration parameters
1493  *
1494  * Description:
1495  *   ip6_tnl_change() updates the tunnel parameters
1496  **/
1497 
1498 static int
1499 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1500 {
1501 	t->parms.laddr = p->laddr;
1502 	t->parms.raddr = p->raddr;
1503 	t->parms.flags = p->flags;
1504 	t->parms.hop_limit = p->hop_limit;
1505 	t->parms.encap_limit = p->encap_limit;
1506 	t->parms.flowinfo = p->flowinfo;
1507 	t->parms.link = p->link;
1508 	t->parms.proto = p->proto;
1509 	t->parms.fwmark = p->fwmark;
1510 	dst_cache_reset(&t->dst_cache);
1511 	ip6_tnl_link_config(t);
1512 	return 0;
1513 }
1514 
1515 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1516 {
1517 	struct net *net = t->net;
1518 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1519 	int err;
1520 
1521 	ip6_tnl_unlink(ip6n, t);
1522 	synchronize_net();
1523 	err = ip6_tnl_change(t, p);
1524 	ip6_tnl_link(ip6n, t);
1525 	netdev_state_change(t->dev);
1526 	return err;
1527 }
1528 
1529 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1530 {
1531 	/* for default tnl0 device allow to change only the proto */
1532 	t->parms.proto = p->proto;
1533 	netdev_state_change(t->dev);
1534 	return 0;
1535 }
1536 
1537 static void
1538 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1539 {
1540 	p->laddr = u->laddr;
1541 	p->raddr = u->raddr;
1542 	p->flags = u->flags;
1543 	p->hop_limit = u->hop_limit;
1544 	p->encap_limit = u->encap_limit;
1545 	p->flowinfo = u->flowinfo;
1546 	p->link = u->link;
1547 	p->proto = u->proto;
1548 	memcpy(p->name, u->name, sizeof(u->name));
1549 }
1550 
1551 static void
1552 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1553 {
1554 	u->laddr = p->laddr;
1555 	u->raddr = p->raddr;
1556 	u->flags = p->flags;
1557 	u->hop_limit = p->hop_limit;
1558 	u->encap_limit = p->encap_limit;
1559 	u->flowinfo = p->flowinfo;
1560 	u->link = p->link;
1561 	u->proto = p->proto;
1562 	memcpy(u->name, p->name, sizeof(u->name));
1563 }
1564 
1565 /**
1566  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1567  *   @dev: virtual device associated with tunnel
1568  *   @ifr: parameters passed from userspace
1569  *   @cmd: command to be performed
1570  *
1571  * Description:
1572  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1573  *   from userspace.
1574  *
1575  *   The possible commands are the following:
1576  *     %SIOCGETTUNNEL: get tunnel parameters for device
1577  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1578  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1579  *     %SIOCDELTUNNEL: delete tunnel
1580  *
1581  *   The fallback device "ip6tnl0", created during module
1582  *   initialization, can be used for creating other tunnel devices.
1583  *
1584  * Return:
1585  *   0 on success,
1586  *   %-EFAULT if unable to copy data to or from userspace,
1587  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1588  *   %-EINVAL if passed tunnel parameters are invalid,
1589  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1590  *   %-ENODEV if attempting to change or delete a nonexisting device
1591  **/
1592 
1593 static int
1594 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1595 {
1596 	int err = 0;
1597 	struct ip6_tnl_parm p;
1598 	struct __ip6_tnl_parm p1;
1599 	struct ip6_tnl *t = netdev_priv(dev);
1600 	struct net *net = t->net;
1601 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1602 
1603 	memset(&p1, 0, sizeof(p1));
1604 
1605 	switch (cmd) {
1606 	case SIOCGETTUNNEL:
1607 		if (dev == ip6n->fb_tnl_dev) {
1608 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1609 				err = -EFAULT;
1610 				break;
1611 			}
1612 			ip6_tnl_parm_from_user(&p1, &p);
1613 			t = ip6_tnl_locate(net, &p1, 0);
1614 			if (IS_ERR(t))
1615 				t = netdev_priv(dev);
1616 		} else {
1617 			memset(&p, 0, sizeof(p));
1618 		}
1619 		ip6_tnl_parm_to_user(&p, &t->parms);
1620 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1621 			err = -EFAULT;
1622 		}
1623 		break;
1624 	case SIOCADDTUNNEL:
1625 	case SIOCCHGTUNNEL:
1626 		err = -EPERM;
1627 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1628 			break;
1629 		err = -EFAULT;
1630 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1631 			break;
1632 		err = -EINVAL;
1633 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1634 		    p.proto != 0)
1635 			break;
1636 		ip6_tnl_parm_from_user(&p1, &p);
1637 		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1638 		if (cmd == SIOCCHGTUNNEL) {
1639 			if (!IS_ERR(t)) {
1640 				if (t->dev != dev) {
1641 					err = -EEXIST;
1642 					break;
1643 				}
1644 			} else
1645 				t = netdev_priv(dev);
1646 			if (dev == ip6n->fb_tnl_dev)
1647 				err = ip6_tnl0_update(t, &p1);
1648 			else
1649 				err = ip6_tnl_update(t, &p1);
1650 		}
1651 		if (!IS_ERR(t)) {
1652 			err = 0;
1653 			ip6_tnl_parm_to_user(&p, &t->parms);
1654 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1655 				err = -EFAULT;
1656 
1657 		} else {
1658 			err = PTR_ERR(t);
1659 		}
1660 		break;
1661 	case SIOCDELTUNNEL:
1662 		err = -EPERM;
1663 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1664 			break;
1665 
1666 		if (dev == ip6n->fb_tnl_dev) {
1667 			err = -EFAULT;
1668 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1669 				break;
1670 			err = -ENOENT;
1671 			ip6_tnl_parm_from_user(&p1, &p);
1672 			t = ip6_tnl_locate(net, &p1, 0);
1673 			if (IS_ERR(t))
1674 				break;
1675 			err = -EPERM;
1676 			if (t->dev == ip6n->fb_tnl_dev)
1677 				break;
1678 			dev = t->dev;
1679 		}
1680 		err = 0;
1681 		unregister_netdevice(dev);
1682 		break;
1683 	default:
1684 		err = -EINVAL;
1685 	}
1686 	return err;
1687 }
1688 
1689 /**
1690  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1691  *   @dev: virtual device associated with tunnel
1692  *   @new_mtu: the new mtu
1693  *
1694  * Return:
1695  *   0 on success,
1696  *   %-EINVAL if mtu too small
1697  **/
1698 
1699 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1700 {
1701 	struct ip6_tnl *tnl = netdev_priv(dev);
1702 
1703 	if (tnl->parms.proto == IPPROTO_IPV6) {
1704 		if (new_mtu < IPV6_MIN_MTU)
1705 			return -EINVAL;
1706 	} else {
1707 		if (new_mtu < ETH_MIN_MTU)
1708 			return -EINVAL;
1709 	}
1710 	if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1711 		if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1712 			return -EINVAL;
1713 	} else {
1714 		if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1715 			return -EINVAL;
1716 	}
1717 	dev->mtu = new_mtu;
1718 	return 0;
1719 }
1720 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1721 
1722 int ip6_tnl_get_iflink(const struct net_device *dev)
1723 {
1724 	struct ip6_tnl *t = netdev_priv(dev);
1725 
1726 	return t->parms.link;
1727 }
1728 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1729 
1730 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1731 			  unsigned int num)
1732 {
1733 	if (num >= MAX_IPTUN_ENCAP_OPS)
1734 		return -ERANGE;
1735 
1736 	return !cmpxchg((const struct ip6_tnl_encap_ops **)
1737 			&ip6tun_encaps[num],
1738 			NULL, ops) ? 0 : -1;
1739 }
1740 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1741 
1742 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1743 			  unsigned int num)
1744 {
1745 	int ret;
1746 
1747 	if (num >= MAX_IPTUN_ENCAP_OPS)
1748 		return -ERANGE;
1749 
1750 	ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1751 		       &ip6tun_encaps[num],
1752 		       ops, NULL) == ops) ? 0 : -1;
1753 
1754 	synchronize_net();
1755 
1756 	return ret;
1757 }
1758 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1759 
1760 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1761 			struct ip_tunnel_encap *ipencap)
1762 {
1763 	int hlen;
1764 
1765 	memset(&t->encap, 0, sizeof(t->encap));
1766 
1767 	hlen = ip6_encap_hlen(ipencap);
1768 	if (hlen < 0)
1769 		return hlen;
1770 
1771 	t->encap.type = ipencap->type;
1772 	t->encap.sport = ipencap->sport;
1773 	t->encap.dport = ipencap->dport;
1774 	t->encap.flags = ipencap->flags;
1775 
1776 	t->encap_hlen = hlen;
1777 	t->hlen = t->encap_hlen + t->tun_hlen;
1778 
1779 	return 0;
1780 }
1781 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1782 
1783 static const struct net_device_ops ip6_tnl_netdev_ops = {
1784 	.ndo_init	= ip6_tnl_dev_init,
1785 	.ndo_uninit	= ip6_tnl_dev_uninit,
1786 	.ndo_start_xmit = ip6_tnl_start_xmit,
1787 	.ndo_do_ioctl	= ip6_tnl_ioctl,
1788 	.ndo_change_mtu = ip6_tnl_change_mtu,
1789 	.ndo_get_stats	= ip6_get_stats,
1790 	.ndo_get_iflink = ip6_tnl_get_iflink,
1791 };
1792 
1793 #define IPXIPX_FEATURES (NETIF_F_SG |		\
1794 			 NETIF_F_FRAGLIST |	\
1795 			 NETIF_F_HIGHDMA |	\
1796 			 NETIF_F_GSO_SOFTWARE |	\
1797 			 NETIF_F_HW_CSUM)
1798 
1799 /**
1800  * ip6_tnl_dev_setup - setup virtual tunnel device
1801  *   @dev: virtual device associated with tunnel
1802  *
1803  * Description:
1804  *   Initialize function pointers and device parameters
1805  **/
1806 
1807 static void ip6_tnl_dev_setup(struct net_device *dev)
1808 {
1809 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1810 	dev->needs_free_netdev = true;
1811 	dev->priv_destructor = ip6_dev_free;
1812 
1813 	dev->type = ARPHRD_TUNNEL6;
1814 	dev->flags |= IFF_NOARP;
1815 	dev->addr_len = sizeof(struct in6_addr);
1816 	dev->features |= NETIF_F_LLTX;
1817 	netif_keep_dst(dev);
1818 
1819 	dev->features		|= IPXIPX_FEATURES;
1820 	dev->hw_features	|= IPXIPX_FEATURES;
1821 
1822 	/* This perm addr will be used as interface identifier by IPv6 */
1823 	dev->addr_assign_type = NET_ADDR_RANDOM;
1824 	eth_random_addr(dev->perm_addr);
1825 }
1826 
1827 
1828 /**
1829  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1830  *   @dev: virtual device associated with tunnel
1831  **/
1832 
1833 static inline int
1834 ip6_tnl_dev_init_gen(struct net_device *dev)
1835 {
1836 	struct ip6_tnl *t = netdev_priv(dev);
1837 	int ret;
1838 	int t_hlen;
1839 
1840 	t->dev = dev;
1841 	t->net = dev_net(dev);
1842 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1843 	if (!dev->tstats)
1844 		return -ENOMEM;
1845 
1846 	ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1847 	if (ret)
1848 		goto free_stats;
1849 
1850 	ret = gro_cells_init(&t->gro_cells, dev);
1851 	if (ret)
1852 		goto destroy_dst;
1853 
1854 	t->tun_hlen = 0;
1855 	t->hlen = t->encap_hlen + t->tun_hlen;
1856 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1857 
1858 	dev->type = ARPHRD_TUNNEL6;
1859 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1860 	dev->mtu = ETH_DATA_LEN - t_hlen;
1861 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1862 		dev->mtu -= 8;
1863 	dev->min_mtu = ETH_MIN_MTU;
1864 	dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1865 
1866 	return 0;
1867 
1868 destroy_dst:
1869 	dst_cache_destroy(&t->dst_cache);
1870 free_stats:
1871 	free_percpu(dev->tstats);
1872 	dev->tstats = NULL;
1873 
1874 	return ret;
1875 }
1876 
1877 /**
1878  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1879  *   @dev: virtual device associated with tunnel
1880  **/
1881 
1882 static int ip6_tnl_dev_init(struct net_device *dev)
1883 {
1884 	struct ip6_tnl *t = netdev_priv(dev);
1885 	int err = ip6_tnl_dev_init_gen(dev);
1886 
1887 	if (err)
1888 		return err;
1889 	ip6_tnl_link_config(t);
1890 	if (t->parms.collect_md) {
1891 		dev->features |= NETIF_F_NETNS_LOCAL;
1892 		netif_keep_dst(dev);
1893 	}
1894 	return 0;
1895 }
1896 
1897 /**
1898  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1899  *   @dev: fallback device
1900  *
1901  * Return: 0
1902  **/
1903 
1904 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1905 {
1906 	struct ip6_tnl *t = netdev_priv(dev);
1907 	struct net *net = dev_net(dev);
1908 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1909 
1910 	t->parms.proto = IPPROTO_IPV6;
1911 	dev_hold(dev);
1912 
1913 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1914 	return 0;
1915 }
1916 
1917 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1918 			    struct netlink_ext_ack *extack)
1919 {
1920 	u8 proto;
1921 
1922 	if (!data || !data[IFLA_IPTUN_PROTO])
1923 		return 0;
1924 
1925 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1926 	if (proto != IPPROTO_IPV6 &&
1927 	    proto != IPPROTO_IPIP &&
1928 	    proto != 0)
1929 		return -EINVAL;
1930 
1931 	return 0;
1932 }
1933 
1934 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1935 				  struct __ip6_tnl_parm *parms)
1936 {
1937 	memset(parms, 0, sizeof(*parms));
1938 
1939 	if (!data)
1940 		return;
1941 
1942 	if (data[IFLA_IPTUN_LINK])
1943 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1944 
1945 	if (data[IFLA_IPTUN_LOCAL])
1946 		parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1947 
1948 	if (data[IFLA_IPTUN_REMOTE])
1949 		parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1950 
1951 	if (data[IFLA_IPTUN_TTL])
1952 		parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1953 
1954 	if (data[IFLA_IPTUN_ENCAP_LIMIT])
1955 		parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1956 
1957 	if (data[IFLA_IPTUN_FLOWINFO])
1958 		parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1959 
1960 	if (data[IFLA_IPTUN_FLAGS])
1961 		parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1962 
1963 	if (data[IFLA_IPTUN_PROTO])
1964 		parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1965 
1966 	if (data[IFLA_IPTUN_COLLECT_METADATA])
1967 		parms->collect_md = true;
1968 
1969 	if (data[IFLA_IPTUN_FWMARK])
1970 		parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1971 }
1972 
1973 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1974 					struct ip_tunnel_encap *ipencap)
1975 {
1976 	bool ret = false;
1977 
1978 	memset(ipencap, 0, sizeof(*ipencap));
1979 
1980 	if (!data)
1981 		return ret;
1982 
1983 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1984 		ret = true;
1985 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1986 	}
1987 
1988 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1989 		ret = true;
1990 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1991 	}
1992 
1993 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1994 		ret = true;
1995 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1996 	}
1997 
1998 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1999 		ret = true;
2000 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
2001 	}
2002 
2003 	return ret;
2004 }
2005 
2006 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
2007 			   struct nlattr *tb[], struct nlattr *data[],
2008 			   struct netlink_ext_ack *extack)
2009 {
2010 	struct net *net = dev_net(dev);
2011 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2012 	struct ip_tunnel_encap ipencap;
2013 	struct ip6_tnl *nt, *t;
2014 	int err;
2015 
2016 	nt = netdev_priv(dev);
2017 
2018 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2019 		err = ip6_tnl_encap_setup(nt, &ipencap);
2020 		if (err < 0)
2021 			return err;
2022 	}
2023 
2024 	ip6_tnl_netlink_parms(data, &nt->parms);
2025 
2026 	if (nt->parms.collect_md) {
2027 		if (rtnl_dereference(ip6n->collect_md_tun))
2028 			return -EEXIST;
2029 	} else {
2030 		t = ip6_tnl_locate(net, &nt->parms, 0);
2031 		if (!IS_ERR(t))
2032 			return -EEXIST;
2033 	}
2034 
2035 	err = ip6_tnl_create2(dev);
2036 	if (!err && tb[IFLA_MTU])
2037 		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2038 
2039 	return err;
2040 }
2041 
2042 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2043 			      struct nlattr *data[],
2044 			      struct netlink_ext_ack *extack)
2045 {
2046 	struct ip6_tnl *t = netdev_priv(dev);
2047 	struct __ip6_tnl_parm p;
2048 	struct net *net = t->net;
2049 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2050 	struct ip_tunnel_encap ipencap;
2051 
2052 	if (dev == ip6n->fb_tnl_dev)
2053 		return -EINVAL;
2054 
2055 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2056 		int err = ip6_tnl_encap_setup(t, &ipencap);
2057 
2058 		if (err < 0)
2059 			return err;
2060 	}
2061 	ip6_tnl_netlink_parms(data, &p);
2062 	if (p.collect_md)
2063 		return -EINVAL;
2064 
2065 	t = ip6_tnl_locate(net, &p, 0);
2066 	if (!IS_ERR(t)) {
2067 		if (t->dev != dev)
2068 			return -EEXIST;
2069 	} else
2070 		t = netdev_priv(dev);
2071 
2072 	return ip6_tnl_update(t, &p);
2073 }
2074 
2075 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2076 {
2077 	struct net *net = dev_net(dev);
2078 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2079 
2080 	if (dev != ip6n->fb_tnl_dev)
2081 		unregister_netdevice_queue(dev, head);
2082 }
2083 
2084 static size_t ip6_tnl_get_size(const struct net_device *dev)
2085 {
2086 	return
2087 		/* IFLA_IPTUN_LINK */
2088 		nla_total_size(4) +
2089 		/* IFLA_IPTUN_LOCAL */
2090 		nla_total_size(sizeof(struct in6_addr)) +
2091 		/* IFLA_IPTUN_REMOTE */
2092 		nla_total_size(sizeof(struct in6_addr)) +
2093 		/* IFLA_IPTUN_TTL */
2094 		nla_total_size(1) +
2095 		/* IFLA_IPTUN_ENCAP_LIMIT */
2096 		nla_total_size(1) +
2097 		/* IFLA_IPTUN_FLOWINFO */
2098 		nla_total_size(4) +
2099 		/* IFLA_IPTUN_FLAGS */
2100 		nla_total_size(4) +
2101 		/* IFLA_IPTUN_PROTO */
2102 		nla_total_size(1) +
2103 		/* IFLA_IPTUN_ENCAP_TYPE */
2104 		nla_total_size(2) +
2105 		/* IFLA_IPTUN_ENCAP_FLAGS */
2106 		nla_total_size(2) +
2107 		/* IFLA_IPTUN_ENCAP_SPORT */
2108 		nla_total_size(2) +
2109 		/* IFLA_IPTUN_ENCAP_DPORT */
2110 		nla_total_size(2) +
2111 		/* IFLA_IPTUN_COLLECT_METADATA */
2112 		nla_total_size(0) +
2113 		/* IFLA_IPTUN_FWMARK */
2114 		nla_total_size(4) +
2115 		0;
2116 }
2117 
2118 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2119 {
2120 	struct ip6_tnl *tunnel = netdev_priv(dev);
2121 	struct __ip6_tnl_parm *parm = &tunnel->parms;
2122 
2123 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2124 	    nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2125 	    nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2126 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2127 	    nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2128 	    nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2129 	    nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2130 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2131 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2132 		goto nla_put_failure;
2133 
2134 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2135 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2136 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2137 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2138 		goto nla_put_failure;
2139 
2140 	if (parm->collect_md)
2141 		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2142 			goto nla_put_failure;
2143 
2144 	return 0;
2145 
2146 nla_put_failure:
2147 	return -EMSGSIZE;
2148 }
2149 
2150 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2151 {
2152 	struct ip6_tnl *tunnel = netdev_priv(dev);
2153 
2154 	return tunnel->net;
2155 }
2156 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2157 
2158 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2159 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
2160 	[IFLA_IPTUN_LOCAL]		= { .len = sizeof(struct in6_addr) },
2161 	[IFLA_IPTUN_REMOTE]		= { .len = sizeof(struct in6_addr) },
2162 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
2163 	[IFLA_IPTUN_ENCAP_LIMIT]	= { .type = NLA_U8 },
2164 	[IFLA_IPTUN_FLOWINFO]		= { .type = NLA_U32 },
2165 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U32 },
2166 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
2167 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
2168 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
2169 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
2170 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
2171 	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
2172 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
2173 };
2174 
2175 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2176 	.kind		= "ip6tnl",
2177 	.maxtype	= IFLA_IPTUN_MAX,
2178 	.policy		= ip6_tnl_policy,
2179 	.priv_size	= sizeof(struct ip6_tnl),
2180 	.setup		= ip6_tnl_dev_setup,
2181 	.validate	= ip6_tnl_validate,
2182 	.newlink	= ip6_tnl_newlink,
2183 	.changelink	= ip6_tnl_changelink,
2184 	.dellink	= ip6_tnl_dellink,
2185 	.get_size	= ip6_tnl_get_size,
2186 	.fill_info	= ip6_tnl_fill_info,
2187 	.get_link_net	= ip6_tnl_get_link_net,
2188 };
2189 
2190 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2191 	.handler	= ip4ip6_rcv,
2192 	.err_handler	= ip4ip6_err,
2193 	.priority	=	1,
2194 };
2195 
2196 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2197 	.handler	= ip6ip6_rcv,
2198 	.err_handler	= ip6ip6_err,
2199 	.priority	=	1,
2200 };
2201 
2202 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2203 {
2204 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2205 	struct net_device *dev, *aux;
2206 	int h;
2207 	struct ip6_tnl *t;
2208 
2209 	for_each_netdev_safe(net, dev, aux)
2210 		if (dev->rtnl_link_ops == &ip6_link_ops)
2211 			unregister_netdevice_queue(dev, list);
2212 
2213 	for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2214 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
2215 		while (t) {
2216 			/* If dev is in the same netns, it has already
2217 			 * been added to the list by the previous loop.
2218 			 */
2219 			if (!net_eq(dev_net(t->dev), net))
2220 				unregister_netdevice_queue(t->dev, list);
2221 			t = rtnl_dereference(t->next);
2222 		}
2223 	}
2224 }
2225 
2226 static int __net_init ip6_tnl_init_net(struct net *net)
2227 {
2228 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2229 	struct ip6_tnl *t = NULL;
2230 	int err;
2231 
2232 	ip6n->tnls[0] = ip6n->tnls_wc;
2233 	ip6n->tnls[1] = ip6n->tnls_r_l;
2234 
2235 	if (!net_has_fallback_tunnels(net))
2236 		return 0;
2237 	err = -ENOMEM;
2238 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2239 					NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2240 
2241 	if (!ip6n->fb_tnl_dev)
2242 		goto err_alloc_dev;
2243 	dev_net_set(ip6n->fb_tnl_dev, net);
2244 	ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2245 	/* FB netdevice is special: we have one, and only one per netns.
2246 	 * Allowing to move it to another netns is clearly unsafe.
2247 	 */
2248 	ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2249 
2250 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2251 	if (err < 0)
2252 		goto err_register;
2253 
2254 	err = register_netdev(ip6n->fb_tnl_dev);
2255 	if (err < 0)
2256 		goto err_register;
2257 
2258 	t = netdev_priv(ip6n->fb_tnl_dev);
2259 
2260 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2261 	return 0;
2262 
2263 err_register:
2264 	free_netdev(ip6n->fb_tnl_dev);
2265 err_alloc_dev:
2266 	return err;
2267 }
2268 
2269 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2270 {
2271 	struct net *net;
2272 	LIST_HEAD(list);
2273 
2274 	rtnl_lock();
2275 	list_for_each_entry(net, net_list, exit_list)
2276 		ip6_tnl_destroy_tunnels(net, &list);
2277 	unregister_netdevice_many(&list);
2278 	rtnl_unlock();
2279 }
2280 
2281 static struct pernet_operations ip6_tnl_net_ops = {
2282 	.init = ip6_tnl_init_net,
2283 	.exit_batch = ip6_tnl_exit_batch_net,
2284 	.id   = &ip6_tnl_net_id,
2285 	.size = sizeof(struct ip6_tnl_net),
2286 };
2287 
2288 /**
2289  * ip6_tunnel_init - register protocol and reserve needed resources
2290  *
2291  * Return: 0 on success
2292  **/
2293 
2294 static int __init ip6_tunnel_init(void)
2295 {
2296 	int  err;
2297 
2298 	if (!ipv6_mod_enabled())
2299 		return -EOPNOTSUPP;
2300 
2301 	err = register_pernet_device(&ip6_tnl_net_ops);
2302 	if (err < 0)
2303 		goto out_pernet;
2304 
2305 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2306 	if (err < 0) {
2307 		pr_err("%s: can't register ip4ip6\n", __func__);
2308 		goto out_ip4ip6;
2309 	}
2310 
2311 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2312 	if (err < 0) {
2313 		pr_err("%s: can't register ip6ip6\n", __func__);
2314 		goto out_ip6ip6;
2315 	}
2316 	err = rtnl_link_register(&ip6_link_ops);
2317 	if (err < 0)
2318 		goto rtnl_link_failed;
2319 
2320 	return 0;
2321 
2322 rtnl_link_failed:
2323 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2324 out_ip6ip6:
2325 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2326 out_ip4ip6:
2327 	unregister_pernet_device(&ip6_tnl_net_ops);
2328 out_pernet:
2329 	return err;
2330 }
2331 
2332 /**
2333  * ip6_tunnel_cleanup - free resources and unregister protocol
2334  **/
2335 
2336 static void __exit ip6_tunnel_cleanup(void)
2337 {
2338 	rtnl_link_unregister(&ip6_link_ops);
2339 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2340 		pr_info("%s: can't deregister ip4ip6\n", __func__);
2341 
2342 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2343 		pr_info("%s: can't deregister ip6ip6\n", __func__);
2344 
2345 	unregister_pernet_device(&ip6_tnl_net_ops);
2346 }
2347 
2348 module_init(ip6_tunnel_init);
2349 module_exit(ip6_tunnel_cleanup);
2350