xref: /openbmc/linux/net/ipv6/ip6_tunnel.c (revision baa7eb025ab14f3cba2e35c0a8648f9c9f01d24f)
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 #include <linux/module.h>
22 #include <linux/capability.h>
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/sockios.h>
26 #include <linux/icmp.h>
27 #include <linux/if.h>
28 #include <linux/in.h>
29 #include <linux/ip.h>
30 #include <linux/if_tunnel.h>
31 #include <linux/net.h>
32 #include <linux/in6.h>
33 #include <linux/netdevice.h>
34 #include <linux/if_arp.h>
35 #include <linux/icmpv6.h>
36 #include <linux/init.h>
37 #include <linux/route.h>
38 #include <linux/rtnetlink.h>
39 #include <linux/netfilter_ipv6.h>
40 #include <linux/slab.h>
41 
42 #include <asm/uaccess.h>
43 #include <asm/atomic.h>
44 
45 #include <net/icmp.h>
46 #include <net/ip.h>
47 #include <net/ipv6.h>
48 #include <net/ip6_route.h>
49 #include <net/addrconf.h>
50 #include <net/ip6_tunnel.h>
51 #include <net/xfrm.h>
52 #include <net/dsfield.h>
53 #include <net/inet_ecn.h>
54 #include <net/net_namespace.h>
55 #include <net/netns/generic.h>
56 
57 MODULE_AUTHOR("Ville Nuorvala");
58 MODULE_DESCRIPTION("IPv6 tunneling device");
59 MODULE_LICENSE("GPL");
60 
61 #ifdef IP6_TNL_DEBUG
62 #define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __func__)
63 #else
64 #define IP6_TNL_TRACE(x...) do {;} while(0)
65 #endif
66 
67 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
68 #define IPV6_TCLASS_SHIFT 20
69 
70 #define HASH_SIZE  32
71 
72 #define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
73 		     (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
74 		    (HASH_SIZE - 1))
75 
76 static int ip6_tnl_dev_init(struct net_device *dev);
77 static void ip6_tnl_dev_setup(struct net_device *dev);
78 
79 static int ip6_tnl_net_id __read_mostly;
80 struct ip6_tnl_net {
81 	/* the IPv6 tunnel fallback device */
82 	struct net_device *fb_tnl_dev;
83 	/* lists for storing tunnels in use */
84 	struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
85 	struct ip6_tnl __rcu *tnls_wc[1];
86 	struct ip6_tnl __rcu **tnls[2];
87 };
88 
89 /* often modified stats are per cpu, other are shared (netdev->stats) */
90 struct pcpu_tstats {
91 	unsigned long	rx_packets;
92 	unsigned long	rx_bytes;
93 	unsigned long	tx_packets;
94 	unsigned long	tx_bytes;
95 };
96 
97 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
98 {
99 	struct pcpu_tstats sum = { 0 };
100 	int i;
101 
102 	for_each_possible_cpu(i) {
103 		const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
104 
105 		sum.rx_packets += tstats->rx_packets;
106 		sum.rx_bytes   += tstats->rx_bytes;
107 		sum.tx_packets += tstats->tx_packets;
108 		sum.tx_bytes   += tstats->tx_bytes;
109 	}
110 	dev->stats.rx_packets = sum.rx_packets;
111 	dev->stats.rx_bytes   = sum.rx_bytes;
112 	dev->stats.tx_packets = sum.tx_packets;
113 	dev->stats.tx_bytes   = sum.tx_bytes;
114 	return &dev->stats;
115 }
116 
117 /*
118  * Locking : hash tables are protected by RCU and RTNL
119  */
120 
121 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
122 {
123 	struct dst_entry *dst = t->dst_cache;
124 
125 	if (dst && dst->obsolete &&
126 	    dst->ops->check(dst, t->dst_cookie) == NULL) {
127 		t->dst_cache = NULL;
128 		dst_release(dst);
129 		return NULL;
130 	}
131 
132 	return dst;
133 }
134 
135 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
136 {
137 	dst_release(t->dst_cache);
138 	t->dst_cache = NULL;
139 }
140 
141 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
142 {
143 	struct rt6_info *rt = (struct rt6_info *) dst;
144 	t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
145 	dst_release(t->dst_cache);
146 	t->dst_cache = dst;
147 }
148 
149 /**
150  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
151  *   @remote: the address of the tunnel exit-point
152  *   @local: the address of the tunnel entry-point
153  *
154  * Return:
155  *   tunnel matching given end-points if found,
156  *   else fallback tunnel if its device is up,
157  *   else %NULL
158  **/
159 
160 #define for_each_ip6_tunnel_rcu(start) \
161 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
162 
163 static struct ip6_tnl *
164 ip6_tnl_lookup(struct net *net, struct in6_addr *remote, struct in6_addr *local)
165 {
166 	unsigned int h0 = HASH(remote);
167 	unsigned int h1 = HASH(local);
168 	struct ip6_tnl *t;
169 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
170 
171 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[h0 ^ h1]) {
172 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
173 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
174 		    (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, struct ip6_tnl_parm *p)
197 {
198 	struct in6_addr *remote = &p->raddr;
199 	struct in6_addr *local = &p->laddr;
200 	unsigned h = 0;
201 	int prio = 0;
202 
203 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
204 		prio = 1;
205 		h = HASH(remote) ^ HASH(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 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
221 	rcu_assign_pointer(*tp, t);
222 }
223 
224 /**
225  * ip6_tnl_unlink - remove tunnel from hash table
226  *   @t: tunnel to be removed
227  **/
228 
229 static void
230 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
231 {
232 	struct ip6_tnl __rcu **tp;
233 	struct ip6_tnl *iter;
234 
235 	for (tp = ip6_tnl_bucket(ip6n, &t->parms);
236 	     (iter = rtnl_dereference(*tp)) != NULL;
237 	     tp = &iter->next) {
238 		if (t == iter) {
239 			rcu_assign_pointer(*tp, t->next);
240 			break;
241 		}
242 	}
243 }
244 
245 static void ip6_dev_free(struct net_device *dev)
246 {
247 	free_percpu(dev->tstats);
248 	free_netdev(dev);
249 }
250 
251 /**
252  * ip6_tnl_create() - create a new tunnel
253  *   @p: tunnel parameters
254  *   @pt: pointer to new tunnel
255  *
256  * Description:
257  *   Create tunnel matching given parameters.
258  *
259  * Return:
260  *   created tunnel or NULL
261  **/
262 
263 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p)
264 {
265 	struct net_device *dev;
266 	struct ip6_tnl *t;
267 	char name[IFNAMSIZ];
268 	int err;
269 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
270 
271 	if (p->name[0])
272 		strlcpy(name, p->name, IFNAMSIZ);
273 	else
274 		sprintf(name, "ip6tnl%%d");
275 
276 	dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
277 	if (dev == NULL)
278 		goto failed;
279 
280 	dev_net_set(dev, net);
281 
282 	if (strchr(name, '%')) {
283 		if (dev_alloc_name(dev, name) < 0)
284 			goto failed_free;
285 	}
286 
287 	t = netdev_priv(dev);
288 	t->parms = *p;
289 	err = ip6_tnl_dev_init(dev);
290 	if (err < 0)
291 		goto failed_free;
292 
293 	if ((err = register_netdevice(dev)) < 0)
294 		goto failed_free;
295 
296 	dev_hold(dev);
297 	ip6_tnl_link(ip6n, t);
298 	return t;
299 
300 failed_free:
301 	ip6_dev_free(dev);
302 failed:
303 	return NULL;
304 }
305 
306 /**
307  * ip6_tnl_locate - find or create tunnel matching given parameters
308  *   @p: tunnel parameters
309  *   @create: != 0 if allowed to create new tunnel if no match found
310  *
311  * Description:
312  *   ip6_tnl_locate() first tries to locate an existing tunnel
313  *   based on @parms. If this is unsuccessful, but @create is set a new
314  *   tunnel device is created and registered for use.
315  *
316  * Return:
317  *   matching tunnel or NULL
318  **/
319 
320 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
321 		struct ip6_tnl_parm *p, int create)
322 {
323 	struct in6_addr *remote = &p->raddr;
324 	struct in6_addr *local = &p->laddr;
325 	struct ip6_tnl __rcu **tp;
326 	struct ip6_tnl *t;
327 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
328 
329 	for (tp = ip6_tnl_bucket(ip6n, p);
330 	     (t = rtnl_dereference(*tp)) != NULL;
331 	     tp = &t->next) {
332 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
333 		    ipv6_addr_equal(remote, &t->parms.raddr))
334 			return t;
335 	}
336 	if (!create)
337 		return NULL;
338 	return ip6_tnl_create(net, p);
339 }
340 
341 /**
342  * ip6_tnl_dev_uninit - tunnel device uninitializer
343  *   @dev: the device to be destroyed
344  *
345  * Description:
346  *   ip6_tnl_dev_uninit() removes tunnel from its list
347  **/
348 
349 static void
350 ip6_tnl_dev_uninit(struct net_device *dev)
351 {
352 	struct ip6_tnl *t = netdev_priv(dev);
353 	struct net *net = dev_net(dev);
354 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
355 
356 	if (dev == ip6n->fb_tnl_dev)
357 		rcu_assign_pointer(ip6n->tnls_wc[0], NULL);
358 	else
359 		ip6_tnl_unlink(ip6n, t);
360 	ip6_tnl_dst_reset(t);
361 	dev_put(dev);
362 }
363 
364 /**
365  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
366  *   @skb: received socket buffer
367  *
368  * Return:
369  *   0 if none was found,
370  *   else index to encapsulation limit
371  **/
372 
373 static __u16
374 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
375 {
376 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
377 	__u8 nexthdr = ipv6h->nexthdr;
378 	__u16 off = sizeof (*ipv6h);
379 
380 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
381 		__u16 optlen = 0;
382 		struct ipv6_opt_hdr *hdr;
383 		if (raw + off + sizeof (*hdr) > skb->data &&
384 		    !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
385 			break;
386 
387 		hdr = (struct ipv6_opt_hdr *) (raw + off);
388 		if (nexthdr == NEXTHDR_FRAGMENT) {
389 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
390 			if (frag_hdr->frag_off)
391 				break;
392 			optlen = 8;
393 		} else if (nexthdr == NEXTHDR_AUTH) {
394 			optlen = (hdr->hdrlen + 2) << 2;
395 		} else {
396 			optlen = ipv6_optlen(hdr);
397 		}
398 		if (nexthdr == NEXTHDR_DEST) {
399 			__u16 i = off + 2;
400 			while (1) {
401 				struct ipv6_tlv_tnl_enc_lim *tel;
402 
403 				/* No more room for encapsulation limit */
404 				if (i + sizeof (*tel) > off + optlen)
405 					break;
406 
407 				tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
408 				/* return index of option if found and valid */
409 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
410 				    tel->length == 1)
411 					return i;
412 				/* else jump to next option */
413 				if (tel->type)
414 					i += tel->length + 2;
415 				else
416 					i++;
417 			}
418 		}
419 		nexthdr = hdr->nexthdr;
420 		off += optlen;
421 	}
422 	return 0;
423 }
424 
425 /**
426  * ip6_tnl_err - tunnel error handler
427  *
428  * Description:
429  *   ip6_tnl_err() should handle errors in the tunnel according
430  *   to the specifications in RFC 2473.
431  **/
432 
433 static int
434 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
435 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
436 {
437 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
438 	struct ip6_tnl *t;
439 	int rel_msg = 0;
440 	u8 rel_type = ICMPV6_DEST_UNREACH;
441 	u8 rel_code = ICMPV6_ADDR_UNREACH;
442 	__u32 rel_info = 0;
443 	__u16 len;
444 	int err = -ENOENT;
445 
446 	/* If the packet doesn't contain the original IPv6 header we are
447 	   in trouble since we might need the source address for further
448 	   processing of the error. */
449 
450 	rcu_read_lock();
451 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
452 					&ipv6h->saddr)) == NULL)
453 		goto out;
454 
455 	if (t->parms.proto != ipproto && t->parms.proto != 0)
456 		goto out;
457 
458 	err = 0;
459 
460 	switch (*type) {
461 		__u32 teli;
462 		struct ipv6_tlv_tnl_enc_lim *tel;
463 		__u32 mtu;
464 	case ICMPV6_DEST_UNREACH:
465 		if (net_ratelimit())
466 			printk(KERN_WARNING
467 			       "%s: Path to destination invalid "
468 			       "or inactive!\n", t->parms.name);
469 		rel_msg = 1;
470 		break;
471 	case ICMPV6_TIME_EXCEED:
472 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
473 			if (net_ratelimit())
474 				printk(KERN_WARNING
475 				       "%s: Too small hop limit or "
476 				       "routing loop in tunnel!\n",
477 				       t->parms.name);
478 			rel_msg = 1;
479 		}
480 		break;
481 	case ICMPV6_PARAMPROB:
482 		teli = 0;
483 		if ((*code) == ICMPV6_HDR_FIELD)
484 			teli = parse_tlv_tnl_enc_lim(skb, skb->data);
485 
486 		if (teli && teli == *info - 2) {
487 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
488 			if (tel->encap_limit == 0) {
489 				if (net_ratelimit())
490 					printk(KERN_WARNING
491 					       "%s: Too small encapsulation "
492 					       "limit or routing loop in "
493 					       "tunnel!\n", t->parms.name);
494 				rel_msg = 1;
495 			}
496 		} else if (net_ratelimit()) {
497 			printk(KERN_WARNING
498 			       "%s: Recipient unable to parse tunneled "
499 			       "packet!\n ", t->parms.name);
500 		}
501 		break;
502 	case ICMPV6_PKT_TOOBIG:
503 		mtu = *info - offset;
504 		if (mtu < IPV6_MIN_MTU)
505 			mtu = IPV6_MIN_MTU;
506 		t->dev->mtu = mtu;
507 
508 		if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
509 			rel_type = ICMPV6_PKT_TOOBIG;
510 			rel_code = 0;
511 			rel_info = mtu;
512 			rel_msg = 1;
513 		}
514 		break;
515 	}
516 
517 	*type = rel_type;
518 	*code = rel_code;
519 	*info = rel_info;
520 	*msg = rel_msg;
521 
522 out:
523 	rcu_read_unlock();
524 	return err;
525 }
526 
527 static int
528 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
529 	   u8 type, u8 code, int offset, __be32 info)
530 {
531 	int rel_msg = 0;
532 	u8 rel_type = type;
533 	u8 rel_code = code;
534 	__u32 rel_info = ntohl(info);
535 	int err;
536 	struct sk_buff *skb2;
537 	struct iphdr *eiph;
538 	struct flowi fl;
539 	struct rtable *rt;
540 
541 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
542 			  &rel_msg, &rel_info, offset);
543 	if (err < 0)
544 		return err;
545 
546 	if (rel_msg == 0)
547 		return 0;
548 
549 	switch (rel_type) {
550 	case ICMPV6_DEST_UNREACH:
551 		if (rel_code != ICMPV6_ADDR_UNREACH)
552 			return 0;
553 		rel_type = ICMP_DEST_UNREACH;
554 		rel_code = ICMP_HOST_UNREACH;
555 		break;
556 	case ICMPV6_PKT_TOOBIG:
557 		if (rel_code != 0)
558 			return 0;
559 		rel_type = ICMP_DEST_UNREACH;
560 		rel_code = ICMP_FRAG_NEEDED;
561 		break;
562 	default:
563 		return 0;
564 	}
565 
566 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
567 		return 0;
568 
569 	skb2 = skb_clone(skb, GFP_ATOMIC);
570 	if (!skb2)
571 		return 0;
572 
573 	skb_dst_drop(skb2);
574 
575 	skb_pull(skb2, offset);
576 	skb_reset_network_header(skb2);
577 	eiph = ip_hdr(skb2);
578 
579 	/* Try to guess incoming interface */
580 	memset(&fl, 0, sizeof(fl));
581 	fl.fl4_dst = eiph->saddr;
582 	fl.fl4_tos = RT_TOS(eiph->tos);
583 	fl.proto = IPPROTO_IPIP;
584 	if (ip_route_output_key(dev_net(skb->dev), &rt, &fl))
585 		goto out;
586 
587 	skb2->dev = rt->dst.dev;
588 
589 	/* route "incoming" packet */
590 	if (rt->rt_flags & RTCF_LOCAL) {
591 		ip_rt_put(rt);
592 		rt = NULL;
593 		fl.fl4_dst = eiph->daddr;
594 		fl.fl4_src = eiph->saddr;
595 		fl.fl4_tos = eiph->tos;
596 		if (ip_route_output_key(dev_net(skb->dev), &rt, &fl) ||
597 		    rt->dst.dev->type != ARPHRD_TUNNEL) {
598 			ip_rt_put(rt);
599 			goto out;
600 		}
601 		skb_dst_set(skb2, (struct dst_entry *)rt);
602 	} else {
603 		ip_rt_put(rt);
604 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
605 				   skb2->dev) ||
606 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
607 			goto out;
608 	}
609 
610 	/* change mtu on this route */
611 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
612 		if (rel_info > dst_mtu(skb_dst(skb2)))
613 			goto out;
614 
615 		skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info);
616 	}
617 
618 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
619 
620 out:
621 	kfree_skb(skb2);
622 	return 0;
623 }
624 
625 static int
626 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
627 	   u8 type, u8 code, int offset, __be32 info)
628 {
629 	int rel_msg = 0;
630 	u8 rel_type = type;
631 	u8 rel_code = code;
632 	__u32 rel_info = ntohl(info);
633 	int err;
634 
635 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
636 			  &rel_msg, &rel_info, offset);
637 	if (err < 0)
638 		return err;
639 
640 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
641 		struct rt6_info *rt;
642 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
643 
644 		if (!skb2)
645 			return 0;
646 
647 		skb_dst_drop(skb2);
648 		skb_pull(skb2, offset);
649 		skb_reset_network_header(skb2);
650 
651 		/* Try to guess incoming interface */
652 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
653 				NULL, 0, 0);
654 
655 		if (rt && rt->rt6i_dev)
656 			skb2->dev = rt->rt6i_dev;
657 
658 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
659 
660 		if (rt)
661 			dst_release(&rt->dst);
662 
663 		kfree_skb(skb2);
664 	}
665 
666 	return 0;
667 }
668 
669 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
670 					struct ipv6hdr *ipv6h,
671 					struct sk_buff *skb)
672 {
673 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
674 
675 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
676 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
677 
678 	if (INET_ECN_is_ce(dsfield))
679 		IP_ECN_set_ce(ip_hdr(skb));
680 }
681 
682 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
683 					struct ipv6hdr *ipv6h,
684 					struct sk_buff *skb)
685 {
686 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
687 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
688 
689 	if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
690 		IP6_ECN_set_ce(ipv6_hdr(skb));
691 }
692 
693 /* called with rcu_read_lock() */
694 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
695 {
696 	struct ip6_tnl_parm *p = &t->parms;
697 	int ret = 0;
698 	struct net *net = dev_net(t->dev);
699 
700 	if (p->flags & IP6_TNL_F_CAP_RCV) {
701 		struct net_device *ldev = NULL;
702 
703 		if (p->link)
704 			ldev = dev_get_by_index_rcu(net, p->link);
705 
706 		if ((ipv6_addr_is_multicast(&p->laddr) ||
707 		     likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) &&
708 		    likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0)))
709 			ret = 1;
710 
711 	}
712 	return ret;
713 }
714 
715 /**
716  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
717  *   @skb: received socket buffer
718  *   @protocol: ethernet protocol ID
719  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
720  *
721  * Return: 0
722  **/
723 
724 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
725 		       __u8 ipproto,
726 		       void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
727 						    struct ipv6hdr *ipv6h,
728 						    struct sk_buff *skb))
729 {
730 	struct ip6_tnl *t;
731 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
732 
733 	rcu_read_lock();
734 
735 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
736 					&ipv6h->daddr)) != NULL) {
737 		struct pcpu_tstats *tstats;
738 
739 		if (t->parms.proto != ipproto && t->parms.proto != 0) {
740 			rcu_read_unlock();
741 			goto discard;
742 		}
743 
744 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
745 			rcu_read_unlock();
746 			goto discard;
747 		}
748 
749 		if (!ip6_tnl_rcv_ctl(t)) {
750 			t->dev->stats.rx_dropped++;
751 			rcu_read_unlock();
752 			goto discard;
753 		}
754 		secpath_reset(skb);
755 		skb->mac_header = skb->network_header;
756 		skb_reset_network_header(skb);
757 		skb->protocol = htons(protocol);
758 		skb->pkt_type = PACKET_HOST;
759 		memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
760 
761 		tstats = this_cpu_ptr(t->dev->tstats);
762 		tstats->rx_packets++;
763 		tstats->rx_bytes += skb->len;
764 
765 		__skb_tunnel_rx(skb, t->dev);
766 
767 		dscp_ecn_decapsulate(t, ipv6h, skb);
768 
769 		netif_rx(skb);
770 
771 		rcu_read_unlock();
772 		return 0;
773 	}
774 	rcu_read_unlock();
775 	return 1;
776 
777 discard:
778 	kfree_skb(skb);
779 	return 0;
780 }
781 
782 static int ip4ip6_rcv(struct sk_buff *skb)
783 {
784 	return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
785 			   ip4ip6_dscp_ecn_decapsulate);
786 }
787 
788 static int ip6ip6_rcv(struct sk_buff *skb)
789 {
790 	return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
791 			   ip6ip6_dscp_ecn_decapsulate);
792 }
793 
794 struct ipv6_tel_txoption {
795 	struct ipv6_txoptions ops;
796 	__u8 dst_opt[8];
797 };
798 
799 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
800 {
801 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
802 
803 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
804 	opt->dst_opt[3] = 1;
805 	opt->dst_opt[4] = encap_limit;
806 	opt->dst_opt[5] = IPV6_TLV_PADN;
807 	opt->dst_opt[6] = 1;
808 
809 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
810 	opt->ops.opt_nflen = 8;
811 }
812 
813 /**
814  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
815  *   @t: the outgoing tunnel device
816  *   @hdr: IPv6 header from the incoming packet
817  *
818  * Description:
819  *   Avoid trivial tunneling loop by checking that tunnel exit-point
820  *   doesn't match source of incoming packet.
821  *
822  * Return:
823  *   1 if conflict,
824  *   0 else
825  **/
826 
827 static inline int
828 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
829 {
830 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
831 }
832 
833 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
834 {
835 	struct ip6_tnl_parm *p = &t->parms;
836 	int ret = 0;
837 	struct net *net = dev_net(t->dev);
838 
839 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
840 		struct net_device *ldev = NULL;
841 
842 		rcu_read_lock();
843 		if (p->link)
844 			ldev = dev_get_by_index_rcu(net, p->link);
845 
846 		if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
847 			printk(KERN_WARNING
848 			       "%s xmit: Local address not yet configured!\n",
849 			       p->name);
850 		else if (!ipv6_addr_is_multicast(&p->raddr) &&
851 			 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
852 			printk(KERN_WARNING
853 			       "%s xmit: Routing loop! "
854 			       "Remote address found on this node!\n",
855 			       p->name);
856 		else
857 			ret = 1;
858 		rcu_read_unlock();
859 	}
860 	return ret;
861 }
862 /**
863  * ip6_tnl_xmit2 - encapsulate packet and send
864  *   @skb: the outgoing socket buffer
865  *   @dev: the outgoing tunnel device
866  *   @dsfield: dscp code for outer header
867  *   @fl: flow of tunneled packet
868  *   @encap_limit: encapsulation limit
869  *   @pmtu: Path MTU is stored if packet is too big
870  *
871  * Description:
872  *   Build new header and do some sanity checks on the packet before sending
873  *   it.
874  *
875  * Return:
876  *   0 on success
877  *   -1 fail
878  *   %-EMSGSIZE message too big. return mtu in this case.
879  **/
880 
881 static int ip6_tnl_xmit2(struct sk_buff *skb,
882 			 struct net_device *dev,
883 			 __u8 dsfield,
884 			 struct flowi *fl,
885 			 int encap_limit,
886 			 __u32 *pmtu)
887 {
888 	struct net *net = dev_net(dev);
889 	struct ip6_tnl *t = netdev_priv(dev);
890 	struct net_device_stats *stats = &t->dev->stats;
891 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
892 	struct ipv6_tel_txoption opt;
893 	struct dst_entry *dst;
894 	struct net_device *tdev;
895 	int mtu;
896 	unsigned int max_headroom = sizeof(struct ipv6hdr);
897 	u8 proto;
898 	int err = -1;
899 	int pkt_len;
900 
901 	if ((dst = ip6_tnl_dst_check(t)) != NULL)
902 		dst_hold(dst);
903 	else {
904 		dst = ip6_route_output(net, NULL, fl);
905 
906 		if (dst->error || xfrm_lookup(net, &dst, fl, NULL, 0) < 0)
907 			goto tx_err_link_failure;
908 	}
909 
910 	tdev = dst->dev;
911 
912 	if (tdev == dev) {
913 		stats->collisions++;
914 		if (net_ratelimit())
915 			printk(KERN_WARNING
916 			       "%s: Local routing loop detected!\n",
917 			       t->parms.name);
918 		goto tx_err_dst_release;
919 	}
920 	mtu = dst_mtu(dst) - sizeof (*ipv6h);
921 	if (encap_limit >= 0) {
922 		max_headroom += 8;
923 		mtu -= 8;
924 	}
925 	if (mtu < IPV6_MIN_MTU)
926 		mtu = IPV6_MIN_MTU;
927 	if (skb_dst(skb))
928 		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
929 	if (skb->len > mtu) {
930 		*pmtu = mtu;
931 		err = -EMSGSIZE;
932 		goto tx_err_dst_release;
933 	}
934 
935 	/*
936 	 * Okay, now see if we can stuff it in the buffer as-is.
937 	 */
938 	max_headroom += LL_RESERVED_SPACE(tdev);
939 
940 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
941 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
942 		struct sk_buff *new_skb;
943 
944 		if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
945 			goto tx_err_dst_release;
946 
947 		if (skb->sk)
948 			skb_set_owner_w(new_skb, skb->sk);
949 		kfree_skb(skb);
950 		skb = new_skb;
951 	}
952 	skb_dst_drop(skb);
953 	skb_dst_set(skb, dst_clone(dst));
954 
955 	skb->transport_header = skb->network_header;
956 
957 	proto = fl->proto;
958 	if (encap_limit >= 0) {
959 		init_tel_txopt(&opt, encap_limit);
960 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
961 	}
962 	skb_push(skb, sizeof(struct ipv6hdr));
963 	skb_reset_network_header(skb);
964 	ipv6h = ipv6_hdr(skb);
965 	*(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
966 	dsfield = INET_ECN_encapsulate(0, dsfield);
967 	ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
968 	ipv6h->hop_limit = t->parms.hop_limit;
969 	ipv6h->nexthdr = proto;
970 	ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
971 	ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
972 	nf_reset(skb);
973 	pkt_len = skb->len;
974 	err = ip6_local_out(skb);
975 
976 	if (net_xmit_eval(err) == 0) {
977 		struct pcpu_tstats *tstats = this_cpu_ptr(t->dev->tstats);
978 
979 		tstats->tx_bytes += pkt_len;
980 		tstats->tx_packets++;
981 	} else {
982 		stats->tx_errors++;
983 		stats->tx_aborted_errors++;
984 	}
985 	ip6_tnl_dst_store(t, dst);
986 	return 0;
987 tx_err_link_failure:
988 	stats->tx_carrier_errors++;
989 	dst_link_failure(skb);
990 tx_err_dst_release:
991 	dst_release(dst);
992 	return err;
993 }
994 
995 static inline int
996 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
997 {
998 	struct ip6_tnl *t = netdev_priv(dev);
999 	struct iphdr  *iph = ip_hdr(skb);
1000 	int encap_limit = -1;
1001 	struct flowi fl;
1002 	__u8 dsfield;
1003 	__u32 mtu;
1004 	int err;
1005 
1006 	if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1007 	    !ip6_tnl_xmit_ctl(t))
1008 		return -1;
1009 
1010 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1011 		encap_limit = t->parms.encap_limit;
1012 
1013 	memcpy(&fl, &t->fl, sizeof (fl));
1014 	fl.proto = IPPROTO_IPIP;
1015 
1016 	dsfield = ipv4_get_dsfield(iph);
1017 
1018 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1019 		fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1020 					  & IPV6_TCLASS_MASK;
1021 
1022 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1023 	if (err != 0) {
1024 		/* XXX: send ICMP error even if DF is not set. */
1025 		if (err == -EMSGSIZE)
1026 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1027 				  htonl(mtu));
1028 		return -1;
1029 	}
1030 
1031 	return 0;
1032 }
1033 
1034 static inline int
1035 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1036 {
1037 	struct ip6_tnl *t = netdev_priv(dev);
1038 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1039 	int encap_limit = -1;
1040 	__u16 offset;
1041 	struct flowi fl;
1042 	__u8 dsfield;
1043 	__u32 mtu;
1044 	int err;
1045 
1046 	if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1047 	    !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1048 		return -1;
1049 
1050 	offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1051 	if (offset > 0) {
1052 		struct ipv6_tlv_tnl_enc_lim *tel;
1053 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1054 		if (tel->encap_limit == 0) {
1055 			icmpv6_send(skb, ICMPV6_PARAMPROB,
1056 				    ICMPV6_HDR_FIELD, offset + 2);
1057 			return -1;
1058 		}
1059 		encap_limit = tel->encap_limit - 1;
1060 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1061 		encap_limit = t->parms.encap_limit;
1062 
1063 	memcpy(&fl, &t->fl, sizeof (fl));
1064 	fl.proto = IPPROTO_IPV6;
1065 
1066 	dsfield = ipv6_get_dsfield(ipv6h);
1067 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1068 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1069 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1070 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1071 
1072 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1073 	if (err != 0) {
1074 		if (err == -EMSGSIZE)
1075 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1076 		return -1;
1077 	}
1078 
1079 	return 0;
1080 }
1081 
1082 static netdev_tx_t
1083 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1084 {
1085 	struct ip6_tnl *t = netdev_priv(dev);
1086 	struct net_device_stats *stats = &t->dev->stats;
1087 	int ret;
1088 
1089 	switch (skb->protocol) {
1090 	case htons(ETH_P_IP):
1091 		ret = ip4ip6_tnl_xmit(skb, dev);
1092 		break;
1093 	case htons(ETH_P_IPV6):
1094 		ret = ip6ip6_tnl_xmit(skb, dev);
1095 		break;
1096 	default:
1097 		goto tx_err;
1098 	}
1099 
1100 	if (ret < 0)
1101 		goto tx_err;
1102 
1103 	return NETDEV_TX_OK;
1104 
1105 tx_err:
1106 	stats->tx_errors++;
1107 	stats->tx_dropped++;
1108 	kfree_skb(skb);
1109 	return NETDEV_TX_OK;
1110 }
1111 
1112 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1113 {
1114 	struct ip6_tnl_parm *p = &t->parms;
1115 	int ltype = ipv6_addr_type(&p->laddr);
1116 	int rtype = ipv6_addr_type(&p->raddr);
1117 
1118 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1119 
1120 	if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1121 	    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1122 	    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1123 	    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1124 		if (ltype&IPV6_ADDR_UNICAST)
1125 			p->flags |= IP6_TNL_F_CAP_XMIT;
1126 		if (rtype&IPV6_ADDR_UNICAST)
1127 			p->flags |= IP6_TNL_F_CAP_RCV;
1128 	}
1129 }
1130 
1131 static void ip6_tnl_link_config(struct ip6_tnl *t)
1132 {
1133 	struct net_device *dev = t->dev;
1134 	struct ip6_tnl_parm *p = &t->parms;
1135 	struct flowi *fl = &t->fl;
1136 
1137 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1138 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1139 
1140 	/* Set up flowi template */
1141 	ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1142 	ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1143 	fl->oif = p->link;
1144 	fl->fl6_flowlabel = 0;
1145 
1146 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1147 		fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1148 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1149 		fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1150 
1151 	ip6_tnl_set_cap(t);
1152 
1153 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1154 		dev->flags |= IFF_POINTOPOINT;
1155 	else
1156 		dev->flags &= ~IFF_POINTOPOINT;
1157 
1158 	dev->iflink = p->link;
1159 
1160 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1161 		int strict = (ipv6_addr_type(&p->raddr) &
1162 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1163 
1164 		struct rt6_info *rt = rt6_lookup(dev_net(dev),
1165 						 &p->raddr, &p->laddr,
1166 						 p->link, strict);
1167 
1168 		if (rt == NULL)
1169 			return;
1170 
1171 		if (rt->rt6i_dev) {
1172 			dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1173 				sizeof (struct ipv6hdr);
1174 
1175 			dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1176 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1177 				dev->mtu-=8;
1178 
1179 			if (dev->mtu < IPV6_MIN_MTU)
1180 				dev->mtu = IPV6_MIN_MTU;
1181 		}
1182 		dst_release(&rt->dst);
1183 	}
1184 }
1185 
1186 /**
1187  * ip6_tnl_change - update the tunnel parameters
1188  *   @t: tunnel to be changed
1189  *   @p: tunnel configuration parameters
1190  *
1191  * Description:
1192  *   ip6_tnl_change() updates the tunnel parameters
1193  **/
1194 
1195 static int
1196 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1197 {
1198 	ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1199 	ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1200 	t->parms.flags = p->flags;
1201 	t->parms.hop_limit = p->hop_limit;
1202 	t->parms.encap_limit = p->encap_limit;
1203 	t->parms.flowinfo = p->flowinfo;
1204 	t->parms.link = p->link;
1205 	t->parms.proto = p->proto;
1206 	ip6_tnl_dst_reset(t);
1207 	ip6_tnl_link_config(t);
1208 	return 0;
1209 }
1210 
1211 /**
1212  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1213  *   @dev: virtual device associated with tunnel
1214  *   @ifr: parameters passed from userspace
1215  *   @cmd: command to be performed
1216  *
1217  * Description:
1218  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1219  *   from userspace.
1220  *
1221  *   The possible commands are the following:
1222  *     %SIOCGETTUNNEL: get tunnel parameters for device
1223  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1224  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1225  *     %SIOCDELTUNNEL: delete tunnel
1226  *
1227  *   The fallback device "ip6tnl0", created during module
1228  *   initialization, can be used for creating other tunnel devices.
1229  *
1230  * Return:
1231  *   0 on success,
1232  *   %-EFAULT if unable to copy data to or from userspace,
1233  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1234  *   %-EINVAL if passed tunnel parameters are invalid,
1235  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1236  *   %-ENODEV if attempting to change or delete a nonexisting device
1237  **/
1238 
1239 static int
1240 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1241 {
1242 	int err = 0;
1243 	struct ip6_tnl_parm p;
1244 	struct ip6_tnl *t = NULL;
1245 	struct net *net = dev_net(dev);
1246 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1247 
1248 	switch (cmd) {
1249 	case SIOCGETTUNNEL:
1250 		if (dev == ip6n->fb_tnl_dev) {
1251 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1252 				err = -EFAULT;
1253 				break;
1254 			}
1255 			t = ip6_tnl_locate(net, &p, 0);
1256 		}
1257 		if (t == NULL)
1258 			t = netdev_priv(dev);
1259 		memcpy(&p, &t->parms, sizeof (p));
1260 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1261 			err = -EFAULT;
1262 		}
1263 		break;
1264 	case SIOCADDTUNNEL:
1265 	case SIOCCHGTUNNEL:
1266 		err = -EPERM;
1267 		if (!capable(CAP_NET_ADMIN))
1268 			break;
1269 		err = -EFAULT;
1270 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1271 			break;
1272 		err = -EINVAL;
1273 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1274 		    p.proto != 0)
1275 			break;
1276 		t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
1277 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1278 			if (t != NULL) {
1279 				if (t->dev != dev) {
1280 					err = -EEXIST;
1281 					break;
1282 				}
1283 			} else
1284 				t = netdev_priv(dev);
1285 
1286 			ip6_tnl_unlink(ip6n, t);
1287 			synchronize_net();
1288 			err = ip6_tnl_change(t, &p);
1289 			ip6_tnl_link(ip6n, t);
1290 			netdev_state_change(dev);
1291 		}
1292 		if (t) {
1293 			err = 0;
1294 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1295 				err = -EFAULT;
1296 
1297 		} else
1298 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1299 		break;
1300 	case SIOCDELTUNNEL:
1301 		err = -EPERM;
1302 		if (!capable(CAP_NET_ADMIN))
1303 			break;
1304 
1305 		if (dev == ip6n->fb_tnl_dev) {
1306 			err = -EFAULT;
1307 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1308 				break;
1309 			err = -ENOENT;
1310 			if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
1311 				break;
1312 			err = -EPERM;
1313 			if (t->dev == ip6n->fb_tnl_dev)
1314 				break;
1315 			dev = t->dev;
1316 		}
1317 		err = 0;
1318 		unregister_netdevice(dev);
1319 		break;
1320 	default:
1321 		err = -EINVAL;
1322 	}
1323 	return err;
1324 }
1325 
1326 /**
1327  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1328  *   @dev: virtual device associated with tunnel
1329  *   @new_mtu: the new mtu
1330  *
1331  * Return:
1332  *   0 on success,
1333  *   %-EINVAL if mtu too small
1334  **/
1335 
1336 static int
1337 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1338 {
1339 	if (new_mtu < IPV6_MIN_MTU) {
1340 		return -EINVAL;
1341 	}
1342 	dev->mtu = new_mtu;
1343 	return 0;
1344 }
1345 
1346 
1347 static const struct net_device_ops ip6_tnl_netdev_ops = {
1348 	.ndo_uninit	= ip6_tnl_dev_uninit,
1349 	.ndo_start_xmit = ip6_tnl_xmit,
1350 	.ndo_do_ioctl	= ip6_tnl_ioctl,
1351 	.ndo_change_mtu = ip6_tnl_change_mtu,
1352 	.ndo_get_stats	= ip6_get_stats,
1353 };
1354 
1355 
1356 /**
1357  * ip6_tnl_dev_setup - setup virtual tunnel device
1358  *   @dev: virtual device associated with tunnel
1359  *
1360  * Description:
1361  *   Initialize function pointers and device parameters
1362  **/
1363 
1364 static void ip6_tnl_dev_setup(struct net_device *dev)
1365 {
1366 	struct ip6_tnl *t;
1367 
1368 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1369 	dev->destructor = ip6_dev_free;
1370 
1371 	dev->type = ARPHRD_TUNNEL6;
1372 	dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1373 	dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1374 	t = netdev_priv(dev);
1375 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1376 		dev->mtu-=8;
1377 	dev->flags |= IFF_NOARP;
1378 	dev->addr_len = sizeof(struct in6_addr);
1379 	dev->features |= NETIF_F_NETNS_LOCAL;
1380 	dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1381 }
1382 
1383 
1384 /**
1385  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1386  *   @dev: virtual device associated with tunnel
1387  **/
1388 
1389 static inline int
1390 ip6_tnl_dev_init_gen(struct net_device *dev)
1391 {
1392 	struct ip6_tnl *t = netdev_priv(dev);
1393 
1394 	t->dev = dev;
1395 	strcpy(t->parms.name, dev->name);
1396 	dev->tstats = alloc_percpu(struct pcpu_tstats);
1397 	if (!dev->tstats)
1398 		return -ENOMEM;
1399 	return 0;
1400 }
1401 
1402 /**
1403  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1404  *   @dev: virtual device associated with tunnel
1405  **/
1406 
1407 static int ip6_tnl_dev_init(struct net_device *dev)
1408 {
1409 	struct ip6_tnl *t = netdev_priv(dev);
1410 	int err = ip6_tnl_dev_init_gen(dev);
1411 
1412 	if (err)
1413 		return err;
1414 	ip6_tnl_link_config(t);
1415 	return 0;
1416 }
1417 
1418 /**
1419  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1420  *   @dev: fallback device
1421  *
1422  * Return: 0
1423  **/
1424 
1425 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1426 {
1427 	struct ip6_tnl *t = netdev_priv(dev);
1428 	struct net *net = dev_net(dev);
1429 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1430 	int err = ip6_tnl_dev_init_gen(dev);
1431 
1432 	if (err)
1433 		return err;
1434 
1435 	t->parms.proto = IPPROTO_IPV6;
1436 	dev_hold(dev);
1437 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1438 	return 0;
1439 }
1440 
1441 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1442 	.handler	= ip4ip6_rcv,
1443 	.err_handler	= ip4ip6_err,
1444 	.priority	=	1,
1445 };
1446 
1447 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1448 	.handler	= ip6ip6_rcv,
1449 	.err_handler	= ip6ip6_err,
1450 	.priority	=	1,
1451 };
1452 
1453 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1454 {
1455 	int h;
1456 	struct ip6_tnl *t;
1457 	LIST_HEAD(list);
1458 
1459 	for (h = 0; h < HASH_SIZE; h++) {
1460 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1461 		while (t != NULL) {
1462 			unregister_netdevice_queue(t->dev, &list);
1463 			t = rtnl_dereference(t->next);
1464 		}
1465 	}
1466 
1467 	t = rtnl_dereference(ip6n->tnls_wc[0]);
1468 	unregister_netdevice_queue(t->dev, &list);
1469 	unregister_netdevice_many(&list);
1470 }
1471 
1472 static int __net_init ip6_tnl_init_net(struct net *net)
1473 {
1474 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1475 	int err;
1476 
1477 	ip6n->tnls[0] = ip6n->tnls_wc;
1478 	ip6n->tnls[1] = ip6n->tnls_r_l;
1479 
1480 	err = -ENOMEM;
1481 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1482 				      ip6_tnl_dev_setup);
1483 
1484 	if (!ip6n->fb_tnl_dev)
1485 		goto err_alloc_dev;
1486 	dev_net_set(ip6n->fb_tnl_dev, net);
1487 
1488 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1489 	if (err < 0)
1490 		goto err_register;
1491 
1492 	err = register_netdev(ip6n->fb_tnl_dev);
1493 	if (err < 0)
1494 		goto err_register;
1495 	return 0;
1496 
1497 err_register:
1498 	ip6_dev_free(ip6n->fb_tnl_dev);
1499 err_alloc_dev:
1500 	return err;
1501 }
1502 
1503 static void __net_exit ip6_tnl_exit_net(struct net *net)
1504 {
1505 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1506 
1507 	rtnl_lock();
1508 	ip6_tnl_destroy_tunnels(ip6n);
1509 	rtnl_unlock();
1510 }
1511 
1512 static struct pernet_operations ip6_tnl_net_ops = {
1513 	.init = ip6_tnl_init_net,
1514 	.exit = ip6_tnl_exit_net,
1515 	.id   = &ip6_tnl_net_id,
1516 	.size = sizeof(struct ip6_tnl_net),
1517 };
1518 
1519 /**
1520  * ip6_tunnel_init - register protocol and reserve needed resources
1521  *
1522  * Return: 0 on success
1523  **/
1524 
1525 static int __init ip6_tunnel_init(void)
1526 {
1527 	int  err;
1528 
1529 	err = register_pernet_device(&ip6_tnl_net_ops);
1530 	if (err < 0)
1531 		goto out_pernet;
1532 
1533 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1534 	if (err < 0) {
1535 		printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1536 		goto out_ip4ip6;
1537 	}
1538 
1539 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1540 	if (err < 0) {
1541 		printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1542 		goto out_ip6ip6;
1543 	}
1544 
1545 	return 0;
1546 
1547 out_ip6ip6:
1548 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1549 out_ip4ip6:
1550 	unregister_pernet_device(&ip6_tnl_net_ops);
1551 out_pernet:
1552 	return err;
1553 }
1554 
1555 /**
1556  * ip6_tunnel_cleanup - free resources and unregister protocol
1557  **/
1558 
1559 static void __exit ip6_tunnel_cleanup(void)
1560 {
1561 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1562 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1563 
1564 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1565 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1566 
1567 	unregister_pernet_device(&ip6_tnl_net_ops);
1568 }
1569 
1570 module_init(ip6_tunnel_init);
1571 module_exit(ip6_tunnel_cleanup);
1572