xref: /openbmc/linux/net/ipv6/sit.c (revision 275876e2)
1 /*
2  *	IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
8  *
9  *	This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  *
14  *	Changes:
15  * Roger Venning <r.venning@telstra.com>:	6to4 support
16  * Nate Thompson <nate@thebog.net>:		6to4 support
17  * Fred Templin <fred.l.templin@boeing.com>:	isatap support
18  */
19 
20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
21 
22 #include <linux/module.h>
23 #include <linux/capability.h>
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/icmp.h>
33 #include <linux/slab.h>
34 #include <asm/uaccess.h>
35 #include <linux/init.h>
36 #include <linux/netfilter_ipv4.h>
37 #include <linux/if_ether.h>
38 
39 #include <net/sock.h>
40 #include <net/snmp.h>
41 
42 #include <net/ipv6.h>
43 #include <net/protocol.h>
44 #include <net/transp_v6.h>
45 #include <net/ip6_fib.h>
46 #include <net/ip6_route.h>
47 #include <net/ndisc.h>
48 #include <net/addrconf.h>
49 #include <net/ip.h>
50 #include <net/udp.h>
51 #include <net/icmp.h>
52 #include <net/ip_tunnels.h>
53 #include <net/inet_ecn.h>
54 #include <net/xfrm.h>
55 #include <net/dsfield.h>
56 #include <net/net_namespace.h>
57 #include <net/netns/generic.h>
58 
59 /*
60    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
61 
62    For comments look at net/ipv4/ip_gre.c --ANK
63  */
64 
65 #define HASH_SIZE  16
66 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
67 
68 static bool log_ecn_error = true;
69 module_param(log_ecn_error, bool, 0644);
70 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
71 
72 static int ipip6_tunnel_init(struct net_device *dev);
73 static void ipip6_tunnel_setup(struct net_device *dev);
74 static void ipip6_dev_free(struct net_device *dev);
75 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
76 		      __be32 *v4dst);
77 static struct rtnl_link_ops sit_link_ops __read_mostly;
78 
79 static int sit_net_id __read_mostly;
80 struct sit_net {
81 	struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
82 	struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
83 	struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
84 	struct ip_tunnel __rcu *tunnels_wc[1];
85 	struct ip_tunnel __rcu **tunnels[4];
86 
87 	struct net_device *fb_tunnel_dev;
88 };
89 
90 /*
91  * Must be invoked with rcu_read_lock
92  */
93 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
94 		struct net_device *dev, __be32 remote, __be32 local)
95 {
96 	unsigned int h0 = HASH(remote);
97 	unsigned int h1 = HASH(local);
98 	struct ip_tunnel *t;
99 	struct sit_net *sitn = net_generic(net, sit_net_id);
100 
101 	for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
102 		if (local == t->parms.iph.saddr &&
103 		    remote == t->parms.iph.daddr &&
104 		    (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
105 		    (t->dev->flags & IFF_UP))
106 			return t;
107 	}
108 	for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
109 		if (remote == t->parms.iph.daddr &&
110 		    (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
111 		    (t->dev->flags & IFF_UP))
112 			return t;
113 	}
114 	for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
115 		if (local == t->parms.iph.saddr &&
116 		    (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
117 		    (t->dev->flags & IFF_UP))
118 			return t;
119 	}
120 	t = rcu_dereference(sitn->tunnels_wc[0]);
121 	if ((t != NULL) && (t->dev->flags & IFF_UP))
122 		return t;
123 	return NULL;
124 }
125 
126 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
127 		struct ip_tunnel_parm *parms)
128 {
129 	__be32 remote = parms->iph.daddr;
130 	__be32 local = parms->iph.saddr;
131 	unsigned int h = 0;
132 	int prio = 0;
133 
134 	if (remote) {
135 		prio |= 2;
136 		h ^= HASH(remote);
137 	}
138 	if (local) {
139 		prio |= 1;
140 		h ^= HASH(local);
141 	}
142 	return &sitn->tunnels[prio][h];
143 }
144 
145 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
146 		struct ip_tunnel *t)
147 {
148 	return __ipip6_bucket(sitn, &t->parms);
149 }
150 
151 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
152 {
153 	struct ip_tunnel __rcu **tp;
154 	struct ip_tunnel *iter;
155 
156 	for (tp = ipip6_bucket(sitn, t);
157 	     (iter = rtnl_dereference(*tp)) != NULL;
158 	     tp = &iter->next) {
159 		if (t == iter) {
160 			rcu_assign_pointer(*tp, t->next);
161 			break;
162 		}
163 	}
164 }
165 
166 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
167 {
168 	struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
169 
170 	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
171 	rcu_assign_pointer(*tp, t);
172 }
173 
174 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
175 {
176 #ifdef CONFIG_IPV6_SIT_6RD
177 	struct ip_tunnel *t = netdev_priv(dev);
178 
179 	if (t->dev == sitn->fb_tunnel_dev) {
180 		ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
181 		t->ip6rd.relay_prefix = 0;
182 		t->ip6rd.prefixlen = 16;
183 		t->ip6rd.relay_prefixlen = 0;
184 	} else {
185 		struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
186 		memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
187 	}
188 #endif
189 }
190 
191 static int ipip6_tunnel_create(struct net_device *dev)
192 {
193 	struct ip_tunnel *t = netdev_priv(dev);
194 	struct net *net = dev_net(dev);
195 	struct sit_net *sitn = net_generic(net, sit_net_id);
196 	int err;
197 
198 	err = ipip6_tunnel_init(dev);
199 	if (err < 0)
200 		goto out;
201 	ipip6_tunnel_clone_6rd(dev, sitn);
202 
203 	if ((__force u16)t->parms.i_flags & SIT_ISATAP)
204 		dev->priv_flags |= IFF_ISATAP;
205 
206 	err = register_netdevice(dev);
207 	if (err < 0)
208 		goto out;
209 
210 	strcpy(t->parms.name, dev->name);
211 	dev->rtnl_link_ops = &sit_link_ops;
212 
213 	dev_hold(dev);
214 
215 	ipip6_tunnel_link(sitn, t);
216 	return 0;
217 
218 out:
219 	return err;
220 }
221 
222 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
223 		struct ip_tunnel_parm *parms, int create)
224 {
225 	__be32 remote = parms->iph.daddr;
226 	__be32 local = parms->iph.saddr;
227 	struct ip_tunnel *t, *nt;
228 	struct ip_tunnel __rcu **tp;
229 	struct net_device *dev;
230 	char name[IFNAMSIZ];
231 	struct sit_net *sitn = net_generic(net, sit_net_id);
232 
233 	for (tp = __ipip6_bucket(sitn, parms);
234 	    (t = rtnl_dereference(*tp)) != NULL;
235 	     tp = &t->next) {
236 		if (local == t->parms.iph.saddr &&
237 		    remote == t->parms.iph.daddr &&
238 		    parms->link == t->parms.link) {
239 			if (create)
240 				return NULL;
241 			else
242 				return t;
243 		}
244 	}
245 	if (!create)
246 		goto failed;
247 
248 	if (parms->name[0])
249 		strlcpy(name, parms->name, IFNAMSIZ);
250 	else
251 		strcpy(name, "sit%d");
252 
253 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
254 			   ipip6_tunnel_setup);
255 	if (dev == NULL)
256 		return NULL;
257 
258 	dev_net_set(dev, net);
259 
260 	nt = netdev_priv(dev);
261 
262 	nt->parms = *parms;
263 	if (ipip6_tunnel_create(dev) < 0)
264 		goto failed_free;
265 
266 	return nt;
267 
268 failed_free:
269 	ipip6_dev_free(dev);
270 failed:
271 	return NULL;
272 }
273 
274 #define for_each_prl_rcu(start)			\
275 	for (prl = rcu_dereference(start);	\
276 	     prl;				\
277 	     prl = rcu_dereference(prl->next))
278 
279 static struct ip_tunnel_prl_entry *
280 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
281 {
282 	struct ip_tunnel_prl_entry *prl;
283 
284 	for_each_prl_rcu(t->prl)
285 		if (prl->addr == addr)
286 			break;
287 	return prl;
288 
289 }
290 
291 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
292 				struct ip_tunnel_prl __user *a)
293 {
294 	struct ip_tunnel_prl kprl, *kp;
295 	struct ip_tunnel_prl_entry *prl;
296 	unsigned int cmax, c = 0, ca, len;
297 	int ret = 0;
298 
299 	if (copy_from_user(&kprl, a, sizeof(kprl)))
300 		return -EFAULT;
301 	cmax = kprl.datalen / sizeof(kprl);
302 	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
303 		cmax = 1;
304 
305 	/* For simple GET or for root users,
306 	 * we try harder to allocate.
307 	 */
308 	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
309 		kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
310 		NULL;
311 
312 	rcu_read_lock();
313 
314 	ca = t->prl_count < cmax ? t->prl_count : cmax;
315 
316 	if (!kp) {
317 		/* We don't try hard to allocate much memory for
318 		 * non-root users.
319 		 * For root users, retry allocating enough memory for
320 		 * the answer.
321 		 */
322 		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
323 		if (!kp) {
324 			ret = -ENOMEM;
325 			goto out;
326 		}
327 	}
328 
329 	c = 0;
330 	for_each_prl_rcu(t->prl) {
331 		if (c >= cmax)
332 			break;
333 		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
334 			continue;
335 		kp[c].addr = prl->addr;
336 		kp[c].flags = prl->flags;
337 		c++;
338 		if (kprl.addr != htonl(INADDR_ANY))
339 			break;
340 	}
341 out:
342 	rcu_read_unlock();
343 
344 	len = sizeof(*kp) * c;
345 	ret = 0;
346 	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
347 		ret = -EFAULT;
348 
349 	kfree(kp);
350 
351 	return ret;
352 }
353 
354 static int
355 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
356 {
357 	struct ip_tunnel_prl_entry *p;
358 	int err = 0;
359 
360 	if (a->addr == htonl(INADDR_ANY))
361 		return -EINVAL;
362 
363 	ASSERT_RTNL();
364 
365 	for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
366 		if (p->addr == a->addr) {
367 			if (chg) {
368 				p->flags = a->flags;
369 				goto out;
370 			}
371 			err = -EEXIST;
372 			goto out;
373 		}
374 	}
375 
376 	if (chg) {
377 		err = -ENXIO;
378 		goto out;
379 	}
380 
381 	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
382 	if (!p) {
383 		err = -ENOBUFS;
384 		goto out;
385 	}
386 
387 	p->next = t->prl;
388 	p->addr = a->addr;
389 	p->flags = a->flags;
390 	t->prl_count++;
391 	rcu_assign_pointer(t->prl, p);
392 out:
393 	return err;
394 }
395 
396 static void prl_list_destroy_rcu(struct rcu_head *head)
397 {
398 	struct ip_tunnel_prl_entry *p, *n;
399 
400 	p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
401 	do {
402 		n = rcu_dereference_protected(p->next, 1);
403 		kfree(p);
404 		p = n;
405 	} while (p);
406 }
407 
408 static int
409 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
410 {
411 	struct ip_tunnel_prl_entry *x;
412 	struct ip_tunnel_prl_entry __rcu **p;
413 	int err = 0;
414 
415 	ASSERT_RTNL();
416 
417 	if (a && a->addr != htonl(INADDR_ANY)) {
418 		for (p = &t->prl;
419 		     (x = rtnl_dereference(*p)) != NULL;
420 		     p = &x->next) {
421 			if (x->addr == a->addr) {
422 				*p = x->next;
423 				kfree_rcu(x, rcu_head);
424 				t->prl_count--;
425 				goto out;
426 			}
427 		}
428 		err = -ENXIO;
429 	} else {
430 		x = rtnl_dereference(t->prl);
431 		if (x) {
432 			t->prl_count = 0;
433 			call_rcu(&x->rcu_head, prl_list_destroy_rcu);
434 			t->prl = NULL;
435 		}
436 	}
437 out:
438 	return err;
439 }
440 
441 static int
442 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
443 {
444 	struct ip_tunnel_prl_entry *p;
445 	int ok = 1;
446 
447 	rcu_read_lock();
448 	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
449 	if (p) {
450 		if (p->flags & PRL_DEFAULT)
451 			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
452 		else
453 			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
454 	} else {
455 		const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
456 
457 		if (ipv6_addr_is_isatap(addr6) &&
458 		    (addr6->s6_addr32[3] == iph->saddr) &&
459 		    ipv6_chk_prefix(addr6, t->dev))
460 			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
461 		else
462 			ok = 0;
463 	}
464 	rcu_read_unlock();
465 	return ok;
466 }
467 
468 static void ipip6_tunnel_uninit(struct net_device *dev)
469 {
470 	struct ip_tunnel *tunnel = netdev_priv(dev);
471 	struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
472 
473 	if (dev == sitn->fb_tunnel_dev) {
474 		RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
475 	} else {
476 		ipip6_tunnel_unlink(sitn, tunnel);
477 		ipip6_tunnel_del_prl(tunnel, NULL);
478 	}
479 	ip_tunnel_dst_reset_all(tunnel);
480 	dev_put(dev);
481 }
482 
483 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
484  * if sufficient data bytes are available
485  */
486 static int ipip6_err_gen_icmpv6_unreach(struct sk_buff *skb)
487 {
488 	const struct iphdr *iph = (const struct iphdr *) skb->data;
489 	struct rt6_info *rt;
490 	struct sk_buff *skb2;
491 
492 	if (!pskb_may_pull(skb, iph->ihl * 4 + sizeof(struct ipv6hdr) + 8))
493 		return 1;
494 
495 	skb2 = skb_clone(skb, GFP_ATOMIC);
496 
497 	if (!skb2)
498 		return 1;
499 
500 	skb_dst_drop(skb2);
501 	skb_pull(skb2, iph->ihl * 4);
502 	skb_reset_network_header(skb2);
503 
504 	rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 0);
505 
506 	if (rt && rt->dst.dev)
507 		skb2->dev = rt->dst.dev;
508 
509 	icmpv6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
510 
511 	if (rt)
512 		ip6_rt_put(rt);
513 
514 	kfree_skb(skb2);
515 
516 	return 0;
517 }
518 
519 static int ipip6_err(struct sk_buff *skb, u32 info)
520 {
521 	const struct iphdr *iph = (const struct iphdr *)skb->data;
522 	const int type = icmp_hdr(skb)->type;
523 	const int code = icmp_hdr(skb)->code;
524 	struct ip_tunnel *t;
525 	int err;
526 
527 	switch (type) {
528 	default:
529 	case ICMP_PARAMETERPROB:
530 		return 0;
531 
532 	case ICMP_DEST_UNREACH:
533 		switch (code) {
534 		case ICMP_SR_FAILED:
535 			/* Impossible event. */
536 			return 0;
537 		default:
538 			/* All others are translated to HOST_UNREACH.
539 			   rfc2003 contains "deep thoughts" about NET_UNREACH,
540 			   I believe they are just ether pollution. --ANK
541 			 */
542 			break;
543 		}
544 		break;
545 	case ICMP_TIME_EXCEEDED:
546 		if (code != ICMP_EXC_TTL)
547 			return 0;
548 		break;
549 	case ICMP_REDIRECT:
550 		break;
551 	}
552 
553 	err = -ENOENT;
554 
555 	t = ipip6_tunnel_lookup(dev_net(skb->dev),
556 				skb->dev,
557 				iph->daddr,
558 				iph->saddr);
559 	if (t == NULL)
560 		goto out;
561 
562 	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
563 		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
564 				 t->parms.link, 0, IPPROTO_IPV6, 0);
565 		err = 0;
566 		goto out;
567 	}
568 	if (type == ICMP_REDIRECT) {
569 		ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
570 			      IPPROTO_IPV6, 0);
571 		err = 0;
572 		goto out;
573 	}
574 
575 	if (t->parms.iph.daddr == 0)
576 		goto out;
577 
578 	err = 0;
579 	if (!ipip6_err_gen_icmpv6_unreach(skb))
580 		goto out;
581 
582 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
583 		goto out;
584 
585 	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
586 		t->err_count++;
587 	else
588 		t->err_count = 1;
589 	t->err_time = jiffies;
590 out:
591 	return err;
592 }
593 
594 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
595 				  const struct in6_addr *v6addr)
596 {
597 	__be32 v4embed = 0;
598 	if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
599 		return true;
600 	return false;
601 }
602 
603 /* Checks if an address matches an address on the tunnel interface.
604  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
605  * Long story:
606  * This function is called after we considered the packet as spoofed
607  * in is_spoofed_6rd.
608  * We may have a router that is doing NAT for proto 41 packets
609  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
610  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
611  * function will return true, dropping the packet.
612  * But, we can still check if is spoofed against the IP
613  * addresses associated with the interface.
614  */
615 static bool only_dnatted(const struct ip_tunnel *tunnel,
616 	const struct in6_addr *v6dst)
617 {
618 	int prefix_len;
619 
620 #ifdef CONFIG_IPV6_SIT_6RD
621 	prefix_len = tunnel->ip6rd.prefixlen + 32
622 		- tunnel->ip6rd.relay_prefixlen;
623 #else
624 	prefix_len = 48;
625 #endif
626 	return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
627 }
628 
629 /* Returns true if a packet is spoofed */
630 static bool packet_is_spoofed(struct sk_buff *skb,
631 			      const struct iphdr *iph,
632 			      struct ip_tunnel *tunnel)
633 {
634 	const struct ipv6hdr *ipv6h;
635 
636 	if (tunnel->dev->priv_flags & IFF_ISATAP) {
637 		if (!isatap_chksrc(skb, iph, tunnel))
638 			return true;
639 
640 		return false;
641 	}
642 
643 	if (tunnel->dev->flags & IFF_POINTOPOINT)
644 		return false;
645 
646 	ipv6h = ipv6_hdr(skb);
647 
648 	if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
649 		net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
650 				     &iph->saddr, &ipv6h->saddr,
651 				     &iph->daddr, &ipv6h->daddr);
652 		return true;
653 	}
654 
655 	if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
656 		return false;
657 
658 	if (only_dnatted(tunnel, &ipv6h->daddr))
659 		return false;
660 
661 	net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
662 			     &iph->saddr, &ipv6h->saddr,
663 			     &iph->daddr, &ipv6h->daddr);
664 	return true;
665 }
666 
667 static int ipip6_rcv(struct sk_buff *skb)
668 {
669 	const struct iphdr *iph = ip_hdr(skb);
670 	struct ip_tunnel *tunnel;
671 	int err;
672 
673 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
674 				     iph->saddr, iph->daddr);
675 	if (tunnel != NULL) {
676 		struct pcpu_sw_netstats *tstats;
677 
678 		if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
679 		    tunnel->parms.iph.protocol != 0)
680 			goto out;
681 
682 		skb->mac_header = skb->network_header;
683 		skb_reset_network_header(skb);
684 		IPCB(skb)->flags = 0;
685 		skb->protocol = htons(ETH_P_IPV6);
686 
687 		if (packet_is_spoofed(skb, iph, tunnel)) {
688 			tunnel->dev->stats.rx_errors++;
689 			goto out;
690 		}
691 
692 		__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
693 
694 		err = IP_ECN_decapsulate(iph, skb);
695 		if (unlikely(err)) {
696 			if (log_ecn_error)
697 				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
698 						     &iph->saddr, iph->tos);
699 			if (err > 1) {
700 				++tunnel->dev->stats.rx_frame_errors;
701 				++tunnel->dev->stats.rx_errors;
702 				goto out;
703 			}
704 		}
705 
706 		tstats = this_cpu_ptr(tunnel->dev->tstats);
707 		u64_stats_update_begin(&tstats->syncp);
708 		tstats->rx_packets++;
709 		tstats->rx_bytes += skb->len;
710 		u64_stats_update_end(&tstats->syncp);
711 
712 		netif_rx(skb);
713 
714 		return 0;
715 	}
716 
717 	/* no tunnel matched,  let upstream know, ipsec may handle it */
718 	return 1;
719 out:
720 	kfree_skb(skb);
721 	return 0;
722 }
723 
724 static const struct tnl_ptk_info tpi = {
725 	/* no tunnel info required for ipip. */
726 	.proto = htons(ETH_P_IP),
727 };
728 
729 static int ipip_rcv(struct sk_buff *skb)
730 {
731 	const struct iphdr *iph;
732 	struct ip_tunnel *tunnel;
733 
734 	iph = ip_hdr(skb);
735 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
736 				     iph->saddr, iph->daddr);
737 	if (tunnel != NULL) {
738 		if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
739 		    tunnel->parms.iph.protocol != 0)
740 			goto drop;
741 
742 		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
743 			goto drop;
744 		if (iptunnel_pull_header(skb, 0, tpi.proto))
745 			goto drop;
746 		return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
747 	}
748 
749 	return 1;
750 
751 drop:
752 	kfree_skb(skb);
753 	return 0;
754 }
755 
756 /*
757  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
758  * stores the embedded IPv4 address in v4dst and returns true.
759  */
760 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
761 		      __be32 *v4dst)
762 {
763 #ifdef CONFIG_IPV6_SIT_6RD
764 	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
765 			      tunnel->ip6rd.prefixlen)) {
766 		unsigned int pbw0, pbi0;
767 		int pbi1;
768 		u32 d;
769 
770 		pbw0 = tunnel->ip6rd.prefixlen >> 5;
771 		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
772 
773 		d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
774 		    tunnel->ip6rd.relay_prefixlen;
775 
776 		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
777 		if (pbi1 > 0)
778 			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
779 			     (32 - pbi1);
780 
781 		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
782 		return true;
783 	}
784 #else
785 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
786 		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
787 		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
788 		return true;
789 	}
790 #endif
791 	return false;
792 }
793 
794 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
795 			     const struct in6_addr *v6dst)
796 {
797 	__be32 dst = 0;
798 	check_6rd(tunnel, v6dst, &dst);
799 	return dst;
800 }
801 
802 /*
803  *	This function assumes it is being called from dev_queue_xmit()
804  *	and that skb is filled properly by that function.
805  */
806 
807 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
808 				     struct net_device *dev)
809 {
810 	struct ip_tunnel *tunnel = netdev_priv(dev);
811 	const struct iphdr  *tiph = &tunnel->parms.iph;
812 	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
813 	u8     tos = tunnel->parms.iph.tos;
814 	__be16 df = tiph->frag_off;
815 	struct rtable *rt;     			/* Route to the other host */
816 	struct net_device *tdev;		/* Device to other host */
817 	unsigned int max_headroom;		/* The extra header space needed */
818 	__be32 dst = tiph->daddr;
819 	struct flowi4 fl4;
820 	int    mtu;
821 	const struct in6_addr *addr6;
822 	int addr_type;
823 	u8 ttl;
824 	int err;
825 
826 	if (skb->protocol != htons(ETH_P_IPV6))
827 		goto tx_error;
828 
829 	if (tos == 1)
830 		tos = ipv6_get_dsfield(iph6);
831 
832 	/* ISATAP (RFC4214) - must come before 6to4 */
833 	if (dev->priv_flags & IFF_ISATAP) {
834 		struct neighbour *neigh = NULL;
835 		bool do_tx_error = false;
836 
837 		if (skb_dst(skb))
838 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
839 
840 		if (neigh == NULL) {
841 			net_dbg_ratelimited("nexthop == NULL\n");
842 			goto tx_error;
843 		}
844 
845 		addr6 = (const struct in6_addr *)&neigh->primary_key;
846 		addr_type = ipv6_addr_type(addr6);
847 
848 		if ((addr_type & IPV6_ADDR_UNICAST) &&
849 		     ipv6_addr_is_isatap(addr6))
850 			dst = addr6->s6_addr32[3];
851 		else
852 			do_tx_error = true;
853 
854 		neigh_release(neigh);
855 		if (do_tx_error)
856 			goto tx_error;
857 	}
858 
859 	if (!dst)
860 		dst = try_6rd(tunnel, &iph6->daddr);
861 
862 	if (!dst) {
863 		struct neighbour *neigh = NULL;
864 		bool do_tx_error = false;
865 
866 		if (skb_dst(skb))
867 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
868 
869 		if (neigh == NULL) {
870 			net_dbg_ratelimited("nexthop == NULL\n");
871 			goto tx_error;
872 		}
873 
874 		addr6 = (const struct in6_addr *)&neigh->primary_key;
875 		addr_type = ipv6_addr_type(addr6);
876 
877 		if (addr_type == IPV6_ADDR_ANY) {
878 			addr6 = &ipv6_hdr(skb)->daddr;
879 			addr_type = ipv6_addr_type(addr6);
880 		}
881 
882 		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
883 			dst = addr6->s6_addr32[3];
884 		else
885 			do_tx_error = true;
886 
887 		neigh_release(neigh);
888 		if (do_tx_error)
889 			goto tx_error;
890 	}
891 
892 	rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
893 				   dst, tiph->saddr,
894 				   0, 0,
895 				   IPPROTO_IPV6, RT_TOS(tos),
896 				   tunnel->parms.link);
897 	if (IS_ERR(rt)) {
898 		dev->stats.tx_carrier_errors++;
899 		goto tx_error_icmp;
900 	}
901 	if (rt->rt_type != RTN_UNICAST) {
902 		ip_rt_put(rt);
903 		dev->stats.tx_carrier_errors++;
904 		goto tx_error_icmp;
905 	}
906 	tdev = rt->dst.dev;
907 
908 	if (tdev == dev) {
909 		ip_rt_put(rt);
910 		dev->stats.collisions++;
911 		goto tx_error;
912 	}
913 
914 	if (df) {
915 		mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
916 
917 		if (mtu < 68) {
918 			dev->stats.collisions++;
919 			ip_rt_put(rt);
920 			goto tx_error;
921 		}
922 
923 		if (mtu < IPV6_MIN_MTU) {
924 			mtu = IPV6_MIN_MTU;
925 			df = 0;
926 		}
927 
928 		if (tunnel->parms.iph.daddr && skb_dst(skb))
929 			skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
930 
931 		if (skb->len > mtu && !skb_is_gso(skb)) {
932 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
933 			ip_rt_put(rt);
934 			goto tx_error;
935 		}
936 	}
937 
938 	if (tunnel->err_count > 0) {
939 		if (time_before(jiffies,
940 				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
941 			tunnel->err_count--;
942 			dst_link_failure(skb);
943 		} else
944 			tunnel->err_count = 0;
945 	}
946 
947 	/*
948 	 * Okay, now see if we can stuff it in the buffer as-is.
949 	 */
950 	max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
951 
952 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
953 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
954 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
955 		if (!new_skb) {
956 			ip_rt_put(rt);
957 			dev->stats.tx_dropped++;
958 			kfree_skb(skb);
959 			return NETDEV_TX_OK;
960 		}
961 		if (skb->sk)
962 			skb_set_owner_w(new_skb, skb->sk);
963 		dev_kfree_skb(skb);
964 		skb = new_skb;
965 		iph6 = ipv6_hdr(skb);
966 	}
967 	ttl = tiph->ttl;
968 	if (ttl == 0)
969 		ttl = iph6->hop_limit;
970 	tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
971 
972 	skb = iptunnel_handle_offloads(skb, false, SKB_GSO_SIT);
973 	if (IS_ERR(skb)) {
974 		ip_rt_put(rt);
975 		goto out;
976 	}
977 
978 	err = iptunnel_xmit(skb->sk, rt, skb, fl4.saddr, fl4.daddr,
979 			    IPPROTO_IPV6, tos, ttl, df,
980 			    !net_eq(tunnel->net, dev_net(dev)));
981 	iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
982 	return NETDEV_TX_OK;
983 
984 tx_error_icmp:
985 	dst_link_failure(skb);
986 tx_error:
987 	kfree_skb(skb);
988 out:
989 	dev->stats.tx_errors++;
990 	return NETDEV_TX_OK;
991 }
992 
993 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
994 {
995 	struct ip_tunnel *tunnel = netdev_priv(dev);
996 	const struct iphdr  *tiph = &tunnel->parms.iph;
997 
998 	skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
999 	if (IS_ERR(skb))
1000 		goto out;
1001 
1002 	ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
1003 	return NETDEV_TX_OK;
1004 out:
1005 	dev->stats.tx_errors++;
1006 	return NETDEV_TX_OK;
1007 }
1008 
1009 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1010 				   struct net_device *dev)
1011 {
1012 	switch (skb->protocol) {
1013 	case htons(ETH_P_IP):
1014 		ipip_tunnel_xmit(skb, dev);
1015 		break;
1016 	case htons(ETH_P_IPV6):
1017 		ipip6_tunnel_xmit(skb, dev);
1018 		break;
1019 	default:
1020 		goto tx_err;
1021 	}
1022 
1023 	return NETDEV_TX_OK;
1024 
1025 tx_err:
1026 	dev->stats.tx_errors++;
1027 	kfree_skb(skb);
1028 	return NETDEV_TX_OK;
1029 
1030 }
1031 
1032 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1033 {
1034 	struct net_device *tdev = NULL;
1035 	struct ip_tunnel *tunnel;
1036 	const struct iphdr *iph;
1037 	struct flowi4 fl4;
1038 
1039 	tunnel = netdev_priv(dev);
1040 	iph = &tunnel->parms.iph;
1041 
1042 	if (iph->daddr) {
1043 		struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1044 							  NULL,
1045 							  iph->daddr, iph->saddr,
1046 							  0, 0,
1047 							  IPPROTO_IPV6,
1048 							  RT_TOS(iph->tos),
1049 							  tunnel->parms.link);
1050 
1051 		if (!IS_ERR(rt)) {
1052 			tdev = rt->dst.dev;
1053 			ip_rt_put(rt);
1054 		}
1055 		dev->flags |= IFF_POINTOPOINT;
1056 	}
1057 
1058 	if (!tdev && tunnel->parms.link)
1059 		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1060 
1061 	if (tdev) {
1062 		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1063 		dev->mtu = tdev->mtu - sizeof(struct iphdr);
1064 		if (dev->mtu < IPV6_MIN_MTU)
1065 			dev->mtu = IPV6_MIN_MTU;
1066 	}
1067 	dev->iflink = tunnel->parms.link;
1068 }
1069 
1070 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
1071 {
1072 	struct net *net = t->net;
1073 	struct sit_net *sitn = net_generic(net, sit_net_id);
1074 
1075 	ipip6_tunnel_unlink(sitn, t);
1076 	synchronize_net();
1077 	t->parms.iph.saddr = p->iph.saddr;
1078 	t->parms.iph.daddr = p->iph.daddr;
1079 	memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1080 	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1081 	ipip6_tunnel_link(sitn, t);
1082 	t->parms.iph.ttl = p->iph.ttl;
1083 	t->parms.iph.tos = p->iph.tos;
1084 	if (t->parms.link != p->link) {
1085 		t->parms.link = p->link;
1086 		ipip6_tunnel_bind_dev(t->dev);
1087 	}
1088 	ip_tunnel_dst_reset_all(t);
1089 	netdev_state_change(t->dev);
1090 }
1091 
1092 #ifdef CONFIG_IPV6_SIT_6RD
1093 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1094 				   struct ip_tunnel_6rd *ip6rd)
1095 {
1096 	struct in6_addr prefix;
1097 	__be32 relay_prefix;
1098 
1099 	if (ip6rd->relay_prefixlen > 32 ||
1100 	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1101 		return -EINVAL;
1102 
1103 	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1104 	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1105 		return -EINVAL;
1106 	if (ip6rd->relay_prefixlen)
1107 		relay_prefix = ip6rd->relay_prefix &
1108 			       htonl(0xffffffffUL <<
1109 				     (32 - ip6rd->relay_prefixlen));
1110 	else
1111 		relay_prefix = 0;
1112 	if (relay_prefix != ip6rd->relay_prefix)
1113 		return -EINVAL;
1114 
1115 	t->ip6rd.prefix = prefix;
1116 	t->ip6rd.relay_prefix = relay_prefix;
1117 	t->ip6rd.prefixlen = ip6rd->prefixlen;
1118 	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1119 	ip_tunnel_dst_reset_all(t);
1120 	netdev_state_change(t->dev);
1121 	return 0;
1122 }
1123 #endif
1124 
1125 static int
1126 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1127 {
1128 	int err = 0;
1129 	struct ip_tunnel_parm p;
1130 	struct ip_tunnel_prl prl;
1131 	struct ip_tunnel *t = netdev_priv(dev);
1132 	struct net *net = t->net;
1133 	struct sit_net *sitn = net_generic(net, sit_net_id);
1134 #ifdef CONFIG_IPV6_SIT_6RD
1135 	struct ip_tunnel_6rd ip6rd;
1136 #endif
1137 
1138 	switch (cmd) {
1139 	case SIOCGETTUNNEL:
1140 #ifdef CONFIG_IPV6_SIT_6RD
1141 	case SIOCGET6RD:
1142 #endif
1143 		if (dev == sitn->fb_tunnel_dev) {
1144 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1145 				err = -EFAULT;
1146 				break;
1147 			}
1148 			t = ipip6_tunnel_locate(net, &p, 0);
1149 			if (t == NULL)
1150 				t = netdev_priv(dev);
1151 		}
1152 
1153 		err = -EFAULT;
1154 		if (cmd == SIOCGETTUNNEL) {
1155 			memcpy(&p, &t->parms, sizeof(p));
1156 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1157 					 sizeof(p)))
1158 				goto done;
1159 #ifdef CONFIG_IPV6_SIT_6RD
1160 		} else {
1161 			ip6rd.prefix = t->ip6rd.prefix;
1162 			ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1163 			ip6rd.prefixlen = t->ip6rd.prefixlen;
1164 			ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1165 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1166 					 sizeof(ip6rd)))
1167 				goto done;
1168 #endif
1169 		}
1170 		err = 0;
1171 		break;
1172 
1173 	case SIOCADDTUNNEL:
1174 	case SIOCCHGTUNNEL:
1175 		err = -EPERM;
1176 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1177 			goto done;
1178 
1179 		err = -EFAULT;
1180 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1181 			goto done;
1182 
1183 		err = -EINVAL;
1184 		if (p.iph.protocol != IPPROTO_IPV6 &&
1185 		    p.iph.protocol != IPPROTO_IPIP &&
1186 		    p.iph.protocol != 0)
1187 			goto done;
1188 		if (p.iph.version != 4 ||
1189 		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1190 			goto done;
1191 		if (p.iph.ttl)
1192 			p.iph.frag_off |= htons(IP_DF);
1193 
1194 		t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1195 
1196 		if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1197 			if (t != NULL) {
1198 				if (t->dev != dev) {
1199 					err = -EEXIST;
1200 					break;
1201 				}
1202 			} else {
1203 				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1204 				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1205 					err = -EINVAL;
1206 					break;
1207 				}
1208 				t = netdev_priv(dev);
1209 			}
1210 
1211 			ipip6_tunnel_update(t, &p);
1212 		}
1213 
1214 		if (t) {
1215 			err = 0;
1216 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1217 				err = -EFAULT;
1218 		} else
1219 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1220 		break;
1221 
1222 	case SIOCDELTUNNEL:
1223 		err = -EPERM;
1224 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1225 			goto done;
1226 
1227 		if (dev == sitn->fb_tunnel_dev) {
1228 			err = -EFAULT;
1229 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1230 				goto done;
1231 			err = -ENOENT;
1232 			if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1233 				goto done;
1234 			err = -EPERM;
1235 			if (t == netdev_priv(sitn->fb_tunnel_dev))
1236 				goto done;
1237 			dev = t->dev;
1238 		}
1239 		unregister_netdevice(dev);
1240 		err = 0;
1241 		break;
1242 
1243 	case SIOCGETPRL:
1244 		err = -EINVAL;
1245 		if (dev == sitn->fb_tunnel_dev)
1246 			goto done;
1247 		err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1248 		break;
1249 
1250 	case SIOCADDPRL:
1251 	case SIOCDELPRL:
1252 	case SIOCCHGPRL:
1253 		err = -EPERM;
1254 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1255 			goto done;
1256 		err = -EINVAL;
1257 		if (dev == sitn->fb_tunnel_dev)
1258 			goto done;
1259 		err = -EFAULT;
1260 		if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1261 			goto done;
1262 
1263 		switch (cmd) {
1264 		case SIOCDELPRL:
1265 			err = ipip6_tunnel_del_prl(t, &prl);
1266 			break;
1267 		case SIOCADDPRL:
1268 		case SIOCCHGPRL:
1269 			err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1270 			break;
1271 		}
1272 		ip_tunnel_dst_reset_all(t);
1273 		netdev_state_change(dev);
1274 		break;
1275 
1276 #ifdef CONFIG_IPV6_SIT_6RD
1277 	case SIOCADD6RD:
1278 	case SIOCCHG6RD:
1279 	case SIOCDEL6RD:
1280 		err = -EPERM;
1281 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1282 			goto done;
1283 
1284 		err = -EFAULT;
1285 		if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1286 				   sizeof(ip6rd)))
1287 			goto done;
1288 
1289 		if (cmd != SIOCDEL6RD) {
1290 			err = ipip6_tunnel_update_6rd(t, &ip6rd);
1291 			if (err < 0)
1292 				goto done;
1293 		} else
1294 			ipip6_tunnel_clone_6rd(dev, sitn);
1295 
1296 		err = 0;
1297 		break;
1298 #endif
1299 
1300 	default:
1301 		err = -EINVAL;
1302 	}
1303 
1304 done:
1305 	return err;
1306 }
1307 
1308 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1309 {
1310 	if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1311 		return -EINVAL;
1312 	dev->mtu = new_mtu;
1313 	return 0;
1314 }
1315 
1316 static const struct net_device_ops ipip6_netdev_ops = {
1317 	.ndo_uninit	= ipip6_tunnel_uninit,
1318 	.ndo_start_xmit	= sit_tunnel_xmit,
1319 	.ndo_do_ioctl	= ipip6_tunnel_ioctl,
1320 	.ndo_change_mtu	= ipip6_tunnel_change_mtu,
1321 	.ndo_get_stats64 = ip_tunnel_get_stats64,
1322 };
1323 
1324 static void ipip6_dev_free(struct net_device *dev)
1325 {
1326 	struct ip_tunnel *tunnel = netdev_priv(dev);
1327 
1328 	free_percpu(tunnel->dst_cache);
1329 	free_percpu(dev->tstats);
1330 	free_netdev(dev);
1331 }
1332 
1333 #define SIT_FEATURES (NETIF_F_SG	   | \
1334 		      NETIF_F_FRAGLIST	   | \
1335 		      NETIF_F_HIGHDMA	   | \
1336 		      NETIF_F_GSO_SOFTWARE | \
1337 		      NETIF_F_HW_CSUM)
1338 
1339 static void ipip6_tunnel_setup(struct net_device *dev)
1340 {
1341 	dev->netdev_ops		= &ipip6_netdev_ops;
1342 	dev->destructor 	= ipip6_dev_free;
1343 
1344 	dev->type		= ARPHRD_SIT;
1345 	dev->hard_header_len 	= LL_MAX_HEADER + sizeof(struct iphdr);
1346 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr);
1347 	dev->flags		= IFF_NOARP;
1348 	dev->priv_flags	       &= ~IFF_XMIT_DST_RELEASE;
1349 	dev->iflink		= 0;
1350 	dev->addr_len		= 4;
1351 	dev->features		|= NETIF_F_LLTX;
1352 	dev->features		|= SIT_FEATURES;
1353 	dev->hw_features	|= SIT_FEATURES;
1354 }
1355 
1356 static int ipip6_tunnel_init(struct net_device *dev)
1357 {
1358 	struct ip_tunnel *tunnel = netdev_priv(dev);
1359 
1360 	tunnel->dev = dev;
1361 	tunnel->net = dev_net(dev);
1362 
1363 	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1364 	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1365 
1366 	ipip6_tunnel_bind_dev(dev);
1367 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1368 	if (!dev->tstats)
1369 		return -ENOMEM;
1370 
1371 	tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1372 	if (!tunnel->dst_cache) {
1373 		free_percpu(dev->tstats);
1374 		return -ENOMEM;
1375 	}
1376 
1377 	return 0;
1378 }
1379 
1380 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1381 {
1382 	struct ip_tunnel *tunnel = netdev_priv(dev);
1383 	struct iphdr *iph = &tunnel->parms.iph;
1384 	struct net *net = dev_net(dev);
1385 	struct sit_net *sitn = net_generic(net, sit_net_id);
1386 
1387 	tunnel->dev = dev;
1388 	tunnel->net = dev_net(dev);
1389 	strcpy(tunnel->parms.name, dev->name);
1390 
1391 	iph->version		= 4;
1392 	iph->protocol		= IPPROTO_IPV6;
1393 	iph->ihl		= 5;
1394 	iph->ttl		= 64;
1395 
1396 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1397 	if (!dev->tstats)
1398 		return -ENOMEM;
1399 
1400 	tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1401 	if (!tunnel->dst_cache) {
1402 		free_percpu(dev->tstats);
1403 		return -ENOMEM;
1404 	}
1405 
1406 	dev_hold(dev);
1407 	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1408 	return 0;
1409 }
1410 
1411 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1412 {
1413 	u8 proto;
1414 
1415 	if (!data || !data[IFLA_IPTUN_PROTO])
1416 		return 0;
1417 
1418 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1419 	if (proto != IPPROTO_IPV6 &&
1420 	    proto != IPPROTO_IPIP &&
1421 	    proto != 0)
1422 		return -EINVAL;
1423 
1424 	return 0;
1425 }
1426 
1427 static void ipip6_netlink_parms(struct nlattr *data[],
1428 				struct ip_tunnel_parm *parms)
1429 {
1430 	memset(parms, 0, sizeof(*parms));
1431 
1432 	parms->iph.version = 4;
1433 	parms->iph.protocol = IPPROTO_IPV6;
1434 	parms->iph.ihl = 5;
1435 	parms->iph.ttl = 64;
1436 
1437 	if (!data)
1438 		return;
1439 
1440 	if (data[IFLA_IPTUN_LINK])
1441 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1442 
1443 	if (data[IFLA_IPTUN_LOCAL])
1444 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1445 
1446 	if (data[IFLA_IPTUN_REMOTE])
1447 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1448 
1449 	if (data[IFLA_IPTUN_TTL]) {
1450 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1451 		if (parms->iph.ttl)
1452 			parms->iph.frag_off = htons(IP_DF);
1453 	}
1454 
1455 	if (data[IFLA_IPTUN_TOS])
1456 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1457 
1458 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1459 		parms->iph.frag_off = htons(IP_DF);
1460 
1461 	if (data[IFLA_IPTUN_FLAGS])
1462 		parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1463 
1464 	if (data[IFLA_IPTUN_PROTO])
1465 		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1466 
1467 }
1468 
1469 #ifdef CONFIG_IPV6_SIT_6RD
1470 /* This function returns true when 6RD attributes are present in the nl msg */
1471 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1472 				    struct ip_tunnel_6rd *ip6rd)
1473 {
1474 	bool ret = false;
1475 	memset(ip6rd, 0, sizeof(*ip6rd));
1476 
1477 	if (!data)
1478 		return ret;
1479 
1480 	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1481 		ret = true;
1482 		nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1483 			   sizeof(struct in6_addr));
1484 	}
1485 
1486 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1487 		ret = true;
1488 		ip6rd->relay_prefix =
1489 			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1490 	}
1491 
1492 	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1493 		ret = true;
1494 		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1495 	}
1496 
1497 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1498 		ret = true;
1499 		ip6rd->relay_prefixlen =
1500 			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1501 	}
1502 
1503 	return ret;
1504 }
1505 #endif
1506 
1507 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1508 			 struct nlattr *tb[], struct nlattr *data[])
1509 {
1510 	struct net *net = dev_net(dev);
1511 	struct ip_tunnel *nt;
1512 #ifdef CONFIG_IPV6_SIT_6RD
1513 	struct ip_tunnel_6rd ip6rd;
1514 #endif
1515 	int err;
1516 
1517 	nt = netdev_priv(dev);
1518 	ipip6_netlink_parms(data, &nt->parms);
1519 
1520 	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1521 		return -EEXIST;
1522 
1523 	err = ipip6_tunnel_create(dev);
1524 	if (err < 0)
1525 		return err;
1526 
1527 #ifdef CONFIG_IPV6_SIT_6RD
1528 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1529 		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1530 #endif
1531 
1532 	return err;
1533 }
1534 
1535 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1536 			  struct nlattr *data[])
1537 {
1538 	struct ip_tunnel *t = netdev_priv(dev);
1539 	struct ip_tunnel_parm p;
1540 	struct net *net = t->net;
1541 	struct sit_net *sitn = net_generic(net, sit_net_id);
1542 #ifdef CONFIG_IPV6_SIT_6RD
1543 	struct ip_tunnel_6rd ip6rd;
1544 #endif
1545 
1546 	if (dev == sitn->fb_tunnel_dev)
1547 		return -EINVAL;
1548 
1549 	ipip6_netlink_parms(data, &p);
1550 
1551 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1552 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1553 		return -EINVAL;
1554 
1555 	t = ipip6_tunnel_locate(net, &p, 0);
1556 
1557 	if (t) {
1558 		if (t->dev != dev)
1559 			return -EEXIST;
1560 	} else
1561 		t = netdev_priv(dev);
1562 
1563 	ipip6_tunnel_update(t, &p);
1564 
1565 #ifdef CONFIG_IPV6_SIT_6RD
1566 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1567 		return ipip6_tunnel_update_6rd(t, &ip6rd);
1568 #endif
1569 
1570 	return 0;
1571 }
1572 
1573 static size_t ipip6_get_size(const struct net_device *dev)
1574 {
1575 	return
1576 		/* IFLA_IPTUN_LINK */
1577 		nla_total_size(4) +
1578 		/* IFLA_IPTUN_LOCAL */
1579 		nla_total_size(4) +
1580 		/* IFLA_IPTUN_REMOTE */
1581 		nla_total_size(4) +
1582 		/* IFLA_IPTUN_TTL */
1583 		nla_total_size(1) +
1584 		/* IFLA_IPTUN_TOS */
1585 		nla_total_size(1) +
1586 		/* IFLA_IPTUN_PMTUDISC */
1587 		nla_total_size(1) +
1588 		/* IFLA_IPTUN_FLAGS */
1589 		nla_total_size(2) +
1590 		/* IFLA_IPTUN_PROTO */
1591 		nla_total_size(1) +
1592 #ifdef CONFIG_IPV6_SIT_6RD
1593 		/* IFLA_IPTUN_6RD_PREFIX */
1594 		nla_total_size(sizeof(struct in6_addr)) +
1595 		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1596 		nla_total_size(4) +
1597 		/* IFLA_IPTUN_6RD_PREFIXLEN */
1598 		nla_total_size(2) +
1599 		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1600 		nla_total_size(2) +
1601 #endif
1602 		0;
1603 }
1604 
1605 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1606 {
1607 	struct ip_tunnel *tunnel = netdev_priv(dev);
1608 	struct ip_tunnel_parm *parm = &tunnel->parms;
1609 
1610 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1611 	    nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1612 	    nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1613 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1614 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1615 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1616 		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1617 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1618 	    nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1619 		goto nla_put_failure;
1620 
1621 #ifdef CONFIG_IPV6_SIT_6RD
1622 	if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1623 		    &tunnel->ip6rd.prefix) ||
1624 	    nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1625 			 tunnel->ip6rd.relay_prefix) ||
1626 	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1627 			tunnel->ip6rd.prefixlen) ||
1628 	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1629 			tunnel->ip6rd.relay_prefixlen))
1630 		goto nla_put_failure;
1631 #endif
1632 
1633 	return 0;
1634 
1635 nla_put_failure:
1636 	return -EMSGSIZE;
1637 }
1638 
1639 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1640 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1641 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1642 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1643 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1644 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1645 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1646 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1647 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1648 #ifdef CONFIG_IPV6_SIT_6RD
1649 	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1650 	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1651 	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1652 	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1653 #endif
1654 };
1655 
1656 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1657 {
1658 	struct net *net = dev_net(dev);
1659 	struct sit_net *sitn = net_generic(net, sit_net_id);
1660 
1661 	if (dev != sitn->fb_tunnel_dev)
1662 		unregister_netdevice_queue(dev, head);
1663 }
1664 
1665 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1666 	.kind		= "sit",
1667 	.maxtype	= IFLA_IPTUN_MAX,
1668 	.policy		= ipip6_policy,
1669 	.priv_size	= sizeof(struct ip_tunnel),
1670 	.setup		= ipip6_tunnel_setup,
1671 	.validate	= ipip6_validate,
1672 	.newlink	= ipip6_newlink,
1673 	.changelink	= ipip6_changelink,
1674 	.get_size	= ipip6_get_size,
1675 	.fill_info	= ipip6_fill_info,
1676 	.dellink	= ipip6_dellink,
1677 };
1678 
1679 static struct xfrm_tunnel sit_handler __read_mostly = {
1680 	.handler	=	ipip6_rcv,
1681 	.err_handler	=	ipip6_err,
1682 	.priority	=	1,
1683 };
1684 
1685 static struct xfrm_tunnel ipip_handler __read_mostly = {
1686 	.handler	=	ipip_rcv,
1687 	.err_handler	=	ipip6_err,
1688 	.priority	=	2,
1689 };
1690 
1691 static void __net_exit sit_destroy_tunnels(struct net *net,
1692 					   struct list_head *head)
1693 {
1694 	struct sit_net *sitn = net_generic(net, sit_net_id);
1695 	struct net_device *dev, *aux;
1696 	int prio;
1697 
1698 	for_each_netdev_safe(net, dev, aux)
1699 		if (dev->rtnl_link_ops == &sit_link_ops)
1700 			unregister_netdevice_queue(dev, head);
1701 
1702 	for (prio = 1; prio < 4; prio++) {
1703 		int h;
1704 		for (h = 0; h < HASH_SIZE; h++) {
1705 			struct ip_tunnel *t;
1706 
1707 			t = rtnl_dereference(sitn->tunnels[prio][h]);
1708 			while (t != NULL) {
1709 				/* If dev is in the same netns, it has already
1710 				 * been added to the list by the previous loop.
1711 				 */
1712 				if (!net_eq(dev_net(t->dev), net))
1713 					unregister_netdevice_queue(t->dev,
1714 								   head);
1715 				t = rtnl_dereference(t->next);
1716 			}
1717 		}
1718 	}
1719 }
1720 
1721 static int __net_init sit_init_net(struct net *net)
1722 {
1723 	struct sit_net *sitn = net_generic(net, sit_net_id);
1724 	struct ip_tunnel *t;
1725 	int err;
1726 
1727 	sitn->tunnels[0] = sitn->tunnels_wc;
1728 	sitn->tunnels[1] = sitn->tunnels_l;
1729 	sitn->tunnels[2] = sitn->tunnels_r;
1730 	sitn->tunnels[3] = sitn->tunnels_r_l;
1731 
1732 	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1733 					   NET_NAME_UNKNOWN,
1734 					   ipip6_tunnel_setup);
1735 	if (!sitn->fb_tunnel_dev) {
1736 		err = -ENOMEM;
1737 		goto err_alloc_dev;
1738 	}
1739 	dev_net_set(sitn->fb_tunnel_dev, net);
1740 	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1741 	/* FB netdevice is special: we have one, and only one per netns.
1742 	 * Allowing to move it to another netns is clearly unsafe.
1743 	 */
1744 	sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1745 
1746 	err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1747 	if (err)
1748 		goto err_dev_free;
1749 
1750 	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1751 
1752 	if ((err = register_netdev(sitn->fb_tunnel_dev)))
1753 		goto err_reg_dev;
1754 
1755 	t = netdev_priv(sitn->fb_tunnel_dev);
1756 
1757 	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1758 	return 0;
1759 
1760 err_reg_dev:
1761 	dev_put(sitn->fb_tunnel_dev);
1762 err_dev_free:
1763 	ipip6_dev_free(sitn->fb_tunnel_dev);
1764 err_alloc_dev:
1765 	return err;
1766 }
1767 
1768 static void __net_exit sit_exit_net(struct net *net)
1769 {
1770 	LIST_HEAD(list);
1771 
1772 	rtnl_lock();
1773 	sit_destroy_tunnels(net, &list);
1774 	unregister_netdevice_many(&list);
1775 	rtnl_unlock();
1776 }
1777 
1778 static struct pernet_operations sit_net_ops = {
1779 	.init = sit_init_net,
1780 	.exit = sit_exit_net,
1781 	.id   = &sit_net_id,
1782 	.size = sizeof(struct sit_net),
1783 };
1784 
1785 static void __exit sit_cleanup(void)
1786 {
1787 	rtnl_link_unregister(&sit_link_ops);
1788 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1789 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1790 
1791 	unregister_pernet_device(&sit_net_ops);
1792 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1793 }
1794 
1795 static int __init sit_init(void)
1796 {
1797 	int err;
1798 
1799 	pr_info("IPv6 over IPv4 tunneling driver\n");
1800 
1801 	err = register_pernet_device(&sit_net_ops);
1802 	if (err < 0)
1803 		return err;
1804 	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1805 	if (err < 0) {
1806 		pr_info("%s: can't register ip6ip4\n", __func__);
1807 		goto xfrm_tunnel_failed;
1808 	}
1809 	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1810 	if (err < 0) {
1811 		pr_info("%s: can't register ip4ip4\n", __func__);
1812 		goto xfrm_tunnel4_failed;
1813 	}
1814 	err = rtnl_link_register(&sit_link_ops);
1815 	if (err < 0)
1816 		goto rtnl_link_failed;
1817 
1818 out:
1819 	return err;
1820 
1821 rtnl_link_failed:
1822 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1823 xfrm_tunnel4_failed:
1824 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1825 xfrm_tunnel_failed:
1826 	unregister_pernet_device(&sit_net_ops);
1827 	goto out;
1828 }
1829 
1830 module_init(sit_init);
1831 module_exit(sit_cleanup);
1832 MODULE_LICENSE("GPL");
1833 MODULE_ALIAS_RTNL_LINK("sit");
1834 MODULE_ALIAS_NETDEV("sit0");
1835