xref: /linux/net/ipv6/sit.c (revision aaa44952bbd1d4db14a4d676bf9595bb5db7e7b0)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
9  *
10  *	Changes:
11  * Roger Venning <r.venning@telstra.com>:	6to4 support
12  * Nate Thompson <nate@thebog.net>:		6to4 support
13  * Fred Templin <fred.l.templin@boeing.com>:	isatap support
14  */
15 
16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
17 
18 #include <linux/module.h>
19 #include <linux/capability.h>
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/sockios.h>
24 #include <linux/net.h>
25 #include <linux/in6.h>
26 #include <linux/netdevice.h>
27 #include <linux/if_arp.h>
28 #include <linux/icmp.h>
29 #include <linux/slab.h>
30 #include <linux/uaccess.h>
31 #include <linux/init.h>
32 #include <linux/netfilter_ipv4.h>
33 #include <linux/if_ether.h>
34 
35 #include <net/sock.h>
36 #include <net/snmp.h>
37 
38 #include <net/ipv6.h>
39 #include <net/protocol.h>
40 #include <net/transp_v6.h>
41 #include <net/ip6_fib.h>
42 #include <net/ip6_route.h>
43 #include <net/ndisc.h>
44 #include <net/addrconf.h>
45 #include <net/ip.h>
46 #include <net/udp.h>
47 #include <net/icmp.h>
48 #include <net/ip_tunnels.h>
49 #include <net/inet_ecn.h>
50 #include <net/xfrm.h>
51 #include <net/dsfield.h>
52 #include <net/net_namespace.h>
53 #include <net/netns/generic.h>
54 
55 /*
56    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
57 
58    For comments look at net/ipv4/ip_gre.c --ANK
59  */
60 
61 #define IP6_SIT_HASH_SIZE  16
62 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
63 
64 static bool log_ecn_error = true;
65 module_param(log_ecn_error, bool, 0644);
66 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
67 
68 static int ipip6_tunnel_init(struct net_device *dev);
69 static void ipip6_tunnel_setup(struct net_device *dev);
70 static void ipip6_dev_free(struct net_device *dev);
71 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
72 		      __be32 *v4dst);
73 static struct rtnl_link_ops sit_link_ops __read_mostly;
74 
75 static unsigned int sit_net_id __read_mostly;
76 struct sit_net {
77 	struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
78 	struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
79 	struct ip_tunnel __rcu *tunnels_l[IP6_SIT_HASH_SIZE];
80 	struct ip_tunnel __rcu *tunnels_wc[1];
81 	struct ip_tunnel __rcu **tunnels[4];
82 
83 	struct net_device *fb_tunnel_dev;
84 };
85 
86 static inline struct sit_net *dev_to_sit_net(struct net_device *dev)
87 {
88 	struct ip_tunnel *t = netdev_priv(dev);
89 
90 	return net_generic(t->net, sit_net_id);
91 }
92 
93 /*
94  * Must be invoked with rcu_read_lock
95  */
96 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
97 					     struct net_device *dev,
98 					     __be32 remote, __be32 local,
99 					     int sifindex)
100 {
101 	unsigned int h0 = HASH(remote);
102 	unsigned int h1 = HASH(local);
103 	struct ip_tunnel *t;
104 	struct sit_net *sitn = net_generic(net, sit_net_id);
105 	int ifindex = dev ? dev->ifindex : 0;
106 
107 	for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
108 		if (local == t->parms.iph.saddr &&
109 		    remote == t->parms.iph.daddr &&
110 		    (!dev || !t->parms.link || ifindex == t->parms.link ||
111 		     sifindex == t->parms.link) &&
112 		    (t->dev->flags & IFF_UP))
113 			return t;
114 	}
115 	for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
116 		if (remote == t->parms.iph.daddr &&
117 		    (!dev || !t->parms.link || ifindex == t->parms.link ||
118 		     sifindex == t->parms.link) &&
119 		    (t->dev->flags & IFF_UP))
120 			return t;
121 	}
122 	for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
123 		if (local == t->parms.iph.saddr &&
124 		    (!dev || !t->parms.link || ifindex == t->parms.link ||
125 		     sifindex == t->parms.link) &&
126 		    (t->dev->flags & IFF_UP))
127 			return t;
128 	}
129 	t = rcu_dereference(sitn->tunnels_wc[0]);
130 	if (t && (t->dev->flags & IFF_UP))
131 		return t;
132 	return NULL;
133 }
134 
135 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
136 		struct ip_tunnel_parm *parms)
137 {
138 	__be32 remote = parms->iph.daddr;
139 	__be32 local = parms->iph.saddr;
140 	unsigned int h = 0;
141 	int prio = 0;
142 
143 	if (remote) {
144 		prio |= 2;
145 		h ^= HASH(remote);
146 	}
147 	if (local) {
148 		prio |= 1;
149 		h ^= HASH(local);
150 	}
151 	return &sitn->tunnels[prio][h];
152 }
153 
154 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
155 		struct ip_tunnel *t)
156 {
157 	return __ipip6_bucket(sitn, &t->parms);
158 }
159 
160 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
161 {
162 	struct ip_tunnel __rcu **tp;
163 	struct ip_tunnel *iter;
164 
165 	for (tp = ipip6_bucket(sitn, t);
166 	     (iter = rtnl_dereference(*tp)) != NULL;
167 	     tp = &iter->next) {
168 		if (t == iter) {
169 			rcu_assign_pointer(*tp, t->next);
170 			break;
171 		}
172 	}
173 }
174 
175 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
176 {
177 	struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
178 
179 	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
180 	rcu_assign_pointer(*tp, t);
181 }
182 
183 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
184 {
185 #ifdef CONFIG_IPV6_SIT_6RD
186 	struct ip_tunnel *t = netdev_priv(dev);
187 
188 	if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
189 		ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
190 		t->ip6rd.relay_prefix = 0;
191 		t->ip6rd.prefixlen = 16;
192 		t->ip6rd.relay_prefixlen = 0;
193 	} else {
194 		struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
195 		memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
196 	}
197 #endif
198 }
199 
200 static int ipip6_tunnel_create(struct net_device *dev)
201 {
202 	struct ip_tunnel *t = netdev_priv(dev);
203 	struct net *net = dev_net(dev);
204 	struct sit_net *sitn = net_generic(net, sit_net_id);
205 	int err;
206 
207 	memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
208 	memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
209 
210 	if ((__force u16)t->parms.i_flags & SIT_ISATAP)
211 		dev->priv_flags |= IFF_ISATAP;
212 
213 	dev->rtnl_link_ops = &sit_link_ops;
214 
215 	err = register_netdevice(dev);
216 	if (err < 0)
217 		goto out;
218 
219 	ipip6_tunnel_clone_6rd(dev, sitn);
220 
221 	ipip6_tunnel_link(sitn, t);
222 	return 0;
223 
224 out:
225 	return err;
226 }
227 
228 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
229 		struct ip_tunnel_parm *parms, int create)
230 {
231 	__be32 remote = parms->iph.daddr;
232 	__be32 local = parms->iph.saddr;
233 	struct ip_tunnel *t, *nt;
234 	struct ip_tunnel __rcu **tp;
235 	struct net_device *dev;
236 	char name[IFNAMSIZ];
237 	struct sit_net *sitn = net_generic(net, sit_net_id);
238 
239 	for (tp = __ipip6_bucket(sitn, parms);
240 	    (t = rtnl_dereference(*tp)) != NULL;
241 	     tp = &t->next) {
242 		if (local == t->parms.iph.saddr &&
243 		    remote == t->parms.iph.daddr &&
244 		    parms->link == t->parms.link) {
245 			if (create)
246 				return NULL;
247 			else
248 				return t;
249 		}
250 	}
251 	if (!create)
252 		goto failed;
253 
254 	if (parms->name[0]) {
255 		if (!dev_valid_name(parms->name))
256 			goto failed;
257 		strlcpy(name, parms->name, IFNAMSIZ);
258 	} else {
259 		strcpy(name, "sit%d");
260 	}
261 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
262 			   ipip6_tunnel_setup);
263 	if (!dev)
264 		return NULL;
265 
266 	dev_net_set(dev, net);
267 
268 	nt = netdev_priv(dev);
269 
270 	nt->parms = *parms;
271 	if (ipip6_tunnel_create(dev) < 0)
272 		goto failed_free;
273 
274 	return nt;
275 
276 failed_free:
277 	free_netdev(dev);
278 failed:
279 	return NULL;
280 }
281 
282 #define for_each_prl_rcu(start)			\
283 	for (prl = rcu_dereference(start);	\
284 	     prl;				\
285 	     prl = rcu_dereference(prl->next))
286 
287 static struct ip_tunnel_prl_entry *
288 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
289 {
290 	struct ip_tunnel_prl_entry *prl;
291 
292 	for_each_prl_rcu(t->prl)
293 		if (prl->addr == addr)
294 			break;
295 	return prl;
296 
297 }
298 
299 static int ipip6_tunnel_get_prl(struct net_device *dev, struct ifreq *ifr)
300 {
301 	struct ip_tunnel_prl __user *a = ifr->ifr_ifru.ifru_data;
302 	struct ip_tunnel *t = netdev_priv(dev);
303 	struct ip_tunnel_prl kprl, *kp;
304 	struct ip_tunnel_prl_entry *prl;
305 	unsigned int cmax, c = 0, ca, len;
306 	int ret = 0;
307 
308 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
309 		return -EINVAL;
310 
311 	if (copy_from_user(&kprl, a, sizeof(kprl)))
312 		return -EFAULT;
313 	cmax = kprl.datalen / sizeof(kprl);
314 	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
315 		cmax = 1;
316 
317 	/* For simple GET or for root users,
318 	 * we try harder to allocate.
319 	 */
320 	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
321 		kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
322 		NULL;
323 
324 	rcu_read_lock();
325 
326 	ca = min(t->prl_count, cmax);
327 
328 	if (!kp) {
329 		/* We don't try hard to allocate much memory for
330 		 * non-root users.
331 		 * For root users, retry allocating enough memory for
332 		 * the answer.
333 		 */
334 		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
335 		if (!kp) {
336 			ret = -ENOMEM;
337 			goto out;
338 		}
339 	}
340 
341 	c = 0;
342 	for_each_prl_rcu(t->prl) {
343 		if (c >= cmax)
344 			break;
345 		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
346 			continue;
347 		kp[c].addr = prl->addr;
348 		kp[c].flags = prl->flags;
349 		c++;
350 		if (kprl.addr != htonl(INADDR_ANY))
351 			break;
352 	}
353 out:
354 	rcu_read_unlock();
355 
356 	len = sizeof(*kp) * c;
357 	ret = 0;
358 	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
359 		ret = -EFAULT;
360 
361 	kfree(kp);
362 
363 	return ret;
364 }
365 
366 static int
367 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
368 {
369 	struct ip_tunnel_prl_entry *p;
370 	int err = 0;
371 
372 	if (a->addr == htonl(INADDR_ANY))
373 		return -EINVAL;
374 
375 	ASSERT_RTNL();
376 
377 	for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
378 		if (p->addr == a->addr) {
379 			if (chg) {
380 				p->flags = a->flags;
381 				goto out;
382 			}
383 			err = -EEXIST;
384 			goto out;
385 		}
386 	}
387 
388 	if (chg) {
389 		err = -ENXIO;
390 		goto out;
391 	}
392 
393 	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
394 	if (!p) {
395 		err = -ENOBUFS;
396 		goto out;
397 	}
398 
399 	p->next = t->prl;
400 	p->addr = a->addr;
401 	p->flags = a->flags;
402 	t->prl_count++;
403 	rcu_assign_pointer(t->prl, p);
404 out:
405 	return err;
406 }
407 
408 static void prl_list_destroy_rcu(struct rcu_head *head)
409 {
410 	struct ip_tunnel_prl_entry *p, *n;
411 
412 	p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
413 	do {
414 		n = rcu_dereference_protected(p->next, 1);
415 		kfree(p);
416 		p = n;
417 	} while (p);
418 }
419 
420 static int
421 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
422 {
423 	struct ip_tunnel_prl_entry *x;
424 	struct ip_tunnel_prl_entry __rcu **p;
425 	int err = 0;
426 
427 	ASSERT_RTNL();
428 
429 	if (a && a->addr != htonl(INADDR_ANY)) {
430 		for (p = &t->prl;
431 		     (x = rtnl_dereference(*p)) != NULL;
432 		     p = &x->next) {
433 			if (x->addr == a->addr) {
434 				*p = x->next;
435 				kfree_rcu(x, rcu_head);
436 				t->prl_count--;
437 				goto out;
438 			}
439 		}
440 		err = -ENXIO;
441 	} else {
442 		x = rtnl_dereference(t->prl);
443 		if (x) {
444 			t->prl_count = 0;
445 			call_rcu(&x->rcu_head, prl_list_destroy_rcu);
446 			t->prl = NULL;
447 		}
448 	}
449 out:
450 	return err;
451 }
452 
453 static int ipip6_tunnel_prl_ctl(struct net_device *dev, struct ifreq *ifr,
454 		int cmd)
455 {
456 	struct ip_tunnel *t = netdev_priv(dev);
457 	struct ip_tunnel_prl prl;
458 	int err;
459 
460 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
461 		return -EPERM;
462 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
463 		return -EINVAL;
464 
465 	if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
466 		return -EFAULT;
467 
468 	switch (cmd) {
469 	case SIOCDELPRL:
470 		err = ipip6_tunnel_del_prl(t, &prl);
471 		break;
472 	case SIOCADDPRL:
473 	case SIOCCHGPRL:
474 		err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
475 		break;
476 	}
477 	dst_cache_reset(&t->dst_cache);
478 	netdev_state_change(dev);
479 	return err;
480 }
481 
482 static int
483 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
484 {
485 	struct ip_tunnel_prl_entry *p;
486 	int ok = 1;
487 
488 	rcu_read_lock();
489 	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
490 	if (p) {
491 		if (p->flags & PRL_DEFAULT)
492 			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
493 		else
494 			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
495 	} else {
496 		const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
497 
498 		if (ipv6_addr_is_isatap(addr6) &&
499 		    (addr6->s6_addr32[3] == iph->saddr) &&
500 		    ipv6_chk_prefix(addr6, t->dev))
501 			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
502 		else
503 			ok = 0;
504 	}
505 	rcu_read_unlock();
506 	return ok;
507 }
508 
509 static void ipip6_tunnel_uninit(struct net_device *dev)
510 {
511 	struct ip_tunnel *tunnel = netdev_priv(dev);
512 	struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
513 
514 	if (dev == sitn->fb_tunnel_dev) {
515 		RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
516 	} else {
517 		ipip6_tunnel_unlink(sitn, tunnel);
518 		ipip6_tunnel_del_prl(tunnel, NULL);
519 	}
520 	dst_cache_reset(&tunnel->dst_cache);
521 	dev_put(dev);
522 }
523 
524 static int ipip6_err(struct sk_buff *skb, u32 info)
525 {
526 	const struct iphdr *iph = (const struct iphdr *)skb->data;
527 	const int type = icmp_hdr(skb)->type;
528 	const int code = icmp_hdr(skb)->code;
529 	unsigned int data_len = 0;
530 	struct ip_tunnel *t;
531 	int sifindex;
532 	int err;
533 
534 	switch (type) {
535 	default:
536 	case ICMP_PARAMETERPROB:
537 		return 0;
538 
539 	case ICMP_DEST_UNREACH:
540 		switch (code) {
541 		case ICMP_SR_FAILED:
542 			/* Impossible event. */
543 			return 0;
544 		default:
545 			/* All others are translated to HOST_UNREACH.
546 			   rfc2003 contains "deep thoughts" about NET_UNREACH,
547 			   I believe they are just ether pollution. --ANK
548 			 */
549 			break;
550 		}
551 		break;
552 	case ICMP_TIME_EXCEEDED:
553 		if (code != ICMP_EXC_TTL)
554 			return 0;
555 		data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
556 		break;
557 	case ICMP_REDIRECT:
558 		break;
559 	}
560 
561 	err = -ENOENT;
562 
563 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
564 	t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
565 				iph->daddr, iph->saddr, sifindex);
566 	if (!t)
567 		goto out;
568 
569 	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
570 		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
571 				 t->parms.link, iph->protocol);
572 		err = 0;
573 		goto out;
574 	}
575 	if (type == ICMP_REDIRECT) {
576 		ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
577 			      iph->protocol);
578 		err = 0;
579 		goto out;
580 	}
581 
582 	err = 0;
583 	if (__in6_dev_get(skb->dev) &&
584 	    !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
585 		goto out;
586 
587 	if (t->parms.iph.daddr == 0)
588 		goto out;
589 
590 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
591 		goto out;
592 
593 	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
594 		t->err_count++;
595 	else
596 		t->err_count = 1;
597 	t->err_time = jiffies;
598 out:
599 	return err;
600 }
601 
602 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
603 				  const struct in6_addr *v6addr)
604 {
605 	__be32 v4embed = 0;
606 	if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
607 		return true;
608 	return false;
609 }
610 
611 /* Checks if an address matches an address on the tunnel interface.
612  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
613  * Long story:
614  * This function is called after we considered the packet as spoofed
615  * in is_spoofed_6rd.
616  * We may have a router that is doing NAT for proto 41 packets
617  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
618  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
619  * function will return true, dropping the packet.
620  * But, we can still check if is spoofed against the IP
621  * addresses associated with the interface.
622  */
623 static bool only_dnatted(const struct ip_tunnel *tunnel,
624 	const struct in6_addr *v6dst)
625 {
626 	int prefix_len;
627 
628 #ifdef CONFIG_IPV6_SIT_6RD
629 	prefix_len = tunnel->ip6rd.prefixlen + 32
630 		- tunnel->ip6rd.relay_prefixlen;
631 #else
632 	prefix_len = 48;
633 #endif
634 	return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
635 }
636 
637 /* Returns true if a packet is spoofed */
638 static bool packet_is_spoofed(struct sk_buff *skb,
639 			      const struct iphdr *iph,
640 			      struct ip_tunnel *tunnel)
641 {
642 	const struct ipv6hdr *ipv6h;
643 
644 	if (tunnel->dev->priv_flags & IFF_ISATAP) {
645 		if (!isatap_chksrc(skb, iph, tunnel))
646 			return true;
647 
648 		return false;
649 	}
650 
651 	if (tunnel->dev->flags & IFF_POINTOPOINT)
652 		return false;
653 
654 	ipv6h = ipv6_hdr(skb);
655 
656 	if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
657 		net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
658 				     &iph->saddr, &ipv6h->saddr,
659 				     &iph->daddr, &ipv6h->daddr);
660 		return true;
661 	}
662 
663 	if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
664 		return false;
665 
666 	if (only_dnatted(tunnel, &ipv6h->daddr))
667 		return false;
668 
669 	net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
670 			     &iph->saddr, &ipv6h->saddr,
671 			     &iph->daddr, &ipv6h->daddr);
672 	return true;
673 }
674 
675 static int ipip6_rcv(struct sk_buff *skb)
676 {
677 	const struct iphdr *iph = ip_hdr(skb);
678 	struct ip_tunnel *tunnel;
679 	int sifindex;
680 	int err;
681 
682 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
683 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
684 				     iph->saddr, iph->daddr, sifindex);
685 	if (tunnel) {
686 		struct pcpu_sw_netstats *tstats;
687 
688 		if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
689 		    tunnel->parms.iph.protocol != 0)
690 			goto out;
691 
692 		skb->mac_header = skb->network_header;
693 		skb_reset_network_header(skb);
694 		IPCB(skb)->flags = 0;
695 		skb->dev = tunnel->dev;
696 
697 		if (packet_is_spoofed(skb, iph, tunnel)) {
698 			tunnel->dev->stats.rx_errors++;
699 			goto out;
700 		}
701 
702 		if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
703 		    !net_eq(tunnel->net, dev_net(tunnel->dev))))
704 			goto out;
705 
706 		/* skb can be uncloned in iptunnel_pull_header, so
707 		 * old iph is no longer valid
708 		 */
709 		iph = (const struct iphdr *)skb_mac_header(skb);
710 		err = IP_ECN_decapsulate(iph, skb);
711 		if (unlikely(err)) {
712 			if (log_ecn_error)
713 				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
714 						     &iph->saddr, iph->tos);
715 			if (err > 1) {
716 				++tunnel->dev->stats.rx_frame_errors;
717 				++tunnel->dev->stats.rx_errors;
718 				goto out;
719 			}
720 		}
721 
722 		tstats = this_cpu_ptr(tunnel->dev->tstats);
723 		u64_stats_update_begin(&tstats->syncp);
724 		tstats->rx_packets++;
725 		tstats->rx_bytes += skb->len;
726 		u64_stats_update_end(&tstats->syncp);
727 
728 		netif_rx(skb);
729 
730 		return 0;
731 	}
732 
733 	/* no tunnel matched,  let upstream know, ipsec may handle it */
734 	return 1;
735 out:
736 	kfree_skb(skb);
737 	return 0;
738 }
739 
740 static const struct tnl_ptk_info ipip_tpi = {
741 	/* no tunnel info required for ipip. */
742 	.proto = htons(ETH_P_IP),
743 };
744 
745 #if IS_ENABLED(CONFIG_MPLS)
746 static const struct tnl_ptk_info mplsip_tpi = {
747 	/* no tunnel info required for mplsip. */
748 	.proto = htons(ETH_P_MPLS_UC),
749 };
750 #endif
751 
752 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
753 {
754 	const struct iphdr *iph;
755 	struct ip_tunnel *tunnel;
756 	int sifindex;
757 
758 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
759 
760 	iph = ip_hdr(skb);
761 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
762 				     iph->saddr, iph->daddr, sifindex);
763 	if (tunnel) {
764 		const struct tnl_ptk_info *tpi;
765 
766 		if (tunnel->parms.iph.protocol != ipproto &&
767 		    tunnel->parms.iph.protocol != 0)
768 			goto drop;
769 
770 		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
771 			goto drop;
772 #if IS_ENABLED(CONFIG_MPLS)
773 		if (ipproto == IPPROTO_MPLS)
774 			tpi = &mplsip_tpi;
775 		else
776 #endif
777 			tpi = &ipip_tpi;
778 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
779 			goto drop;
780 		return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
781 	}
782 
783 	return 1;
784 
785 drop:
786 	kfree_skb(skb);
787 	return 0;
788 }
789 
790 static int ipip_rcv(struct sk_buff *skb)
791 {
792 	return sit_tunnel_rcv(skb, IPPROTO_IPIP);
793 }
794 
795 #if IS_ENABLED(CONFIG_MPLS)
796 static int mplsip_rcv(struct sk_buff *skb)
797 {
798 	return sit_tunnel_rcv(skb, IPPROTO_MPLS);
799 }
800 #endif
801 
802 /*
803  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
804  * stores the embedded IPv4 address in v4dst and returns true.
805  */
806 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
807 		      __be32 *v4dst)
808 {
809 #ifdef CONFIG_IPV6_SIT_6RD
810 	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
811 			      tunnel->ip6rd.prefixlen)) {
812 		unsigned int pbw0, pbi0;
813 		int pbi1;
814 		u32 d;
815 
816 		pbw0 = tunnel->ip6rd.prefixlen >> 5;
817 		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
818 
819 		d = tunnel->ip6rd.relay_prefixlen < 32 ?
820 			(ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
821 		    tunnel->ip6rd.relay_prefixlen : 0;
822 
823 		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
824 		if (pbi1 > 0)
825 			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
826 			     (32 - pbi1);
827 
828 		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
829 		return true;
830 	}
831 #else
832 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
833 		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
834 		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
835 		return true;
836 	}
837 #endif
838 	return false;
839 }
840 
841 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
842 			     const struct in6_addr *v6dst)
843 {
844 	__be32 dst = 0;
845 	check_6rd(tunnel, v6dst, &dst);
846 	return dst;
847 }
848 
849 /*
850  *	This function assumes it is being called from dev_queue_xmit()
851  *	and that skb is filled properly by that function.
852  */
853 
854 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
855 				     struct net_device *dev)
856 {
857 	struct ip_tunnel *tunnel = netdev_priv(dev);
858 	const struct iphdr  *tiph = &tunnel->parms.iph;
859 	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
860 	u8     tos = tunnel->parms.iph.tos;
861 	__be16 df = tiph->frag_off;
862 	struct rtable *rt;		/* Route to the other host */
863 	struct net_device *tdev;	/* Device to other host */
864 	unsigned int max_headroom;	/* The extra header space needed */
865 	__be32 dst = tiph->daddr;
866 	struct flowi4 fl4;
867 	int    mtu;
868 	const struct in6_addr *addr6;
869 	int addr_type;
870 	u8 ttl;
871 	u8 protocol = IPPROTO_IPV6;
872 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
873 
874 	if (tos == 1)
875 		tos = ipv6_get_dsfield(iph6);
876 
877 	/* ISATAP (RFC4214) - must come before 6to4 */
878 	if (dev->priv_flags & IFF_ISATAP) {
879 		struct neighbour *neigh = NULL;
880 		bool do_tx_error = false;
881 
882 		if (skb_dst(skb))
883 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
884 
885 		if (!neigh) {
886 			net_dbg_ratelimited("nexthop == NULL\n");
887 			goto tx_error;
888 		}
889 
890 		addr6 = (const struct in6_addr *)&neigh->primary_key;
891 		addr_type = ipv6_addr_type(addr6);
892 
893 		if ((addr_type & IPV6_ADDR_UNICAST) &&
894 		     ipv6_addr_is_isatap(addr6))
895 			dst = addr6->s6_addr32[3];
896 		else
897 			do_tx_error = true;
898 
899 		neigh_release(neigh);
900 		if (do_tx_error)
901 			goto tx_error;
902 	}
903 
904 	if (!dst)
905 		dst = try_6rd(tunnel, &iph6->daddr);
906 
907 	if (!dst) {
908 		struct neighbour *neigh = NULL;
909 		bool do_tx_error = false;
910 
911 		if (skb_dst(skb))
912 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
913 
914 		if (!neigh) {
915 			net_dbg_ratelimited("nexthop == NULL\n");
916 			goto tx_error;
917 		}
918 
919 		addr6 = (const struct in6_addr *)&neigh->primary_key;
920 		addr_type = ipv6_addr_type(addr6);
921 
922 		if (addr_type == IPV6_ADDR_ANY) {
923 			addr6 = &ipv6_hdr(skb)->daddr;
924 			addr_type = ipv6_addr_type(addr6);
925 		}
926 
927 		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
928 			dst = addr6->s6_addr32[3];
929 		else
930 			do_tx_error = true;
931 
932 		neigh_release(neigh);
933 		if (do_tx_error)
934 			goto tx_error;
935 	}
936 
937 	flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
938 			   RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
939 			   0, dst, tiph->saddr, 0, 0,
940 			   sock_net_uid(tunnel->net, NULL));
941 
942 	rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
943 	if (!rt) {
944 		rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
945 		if (IS_ERR(rt)) {
946 			dev->stats.tx_carrier_errors++;
947 			goto tx_error_icmp;
948 		}
949 		dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
950 	}
951 
952 	if (rt->rt_type != RTN_UNICAST) {
953 		ip_rt_put(rt);
954 		dev->stats.tx_carrier_errors++;
955 		goto tx_error_icmp;
956 	}
957 	tdev = rt->dst.dev;
958 
959 	if (tdev == dev) {
960 		ip_rt_put(rt);
961 		dev->stats.collisions++;
962 		goto tx_error;
963 	}
964 
965 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
966 		ip_rt_put(rt);
967 		goto tx_error;
968 	}
969 
970 	if (df) {
971 		mtu = dst_mtu(&rt->dst) - t_hlen;
972 
973 		if (mtu < 68) {
974 			dev->stats.collisions++;
975 			ip_rt_put(rt);
976 			goto tx_error;
977 		}
978 
979 		if (mtu < IPV6_MIN_MTU) {
980 			mtu = IPV6_MIN_MTU;
981 			df = 0;
982 		}
983 
984 		if (tunnel->parms.iph.daddr)
985 			skb_dst_update_pmtu_no_confirm(skb, mtu);
986 
987 		if (skb->len > mtu && !skb_is_gso(skb)) {
988 			icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
989 			ip_rt_put(rt);
990 			goto tx_error;
991 		}
992 	}
993 
994 	if (tunnel->err_count > 0) {
995 		if (time_before(jiffies,
996 				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
997 			tunnel->err_count--;
998 			dst_link_failure(skb);
999 		} else
1000 			tunnel->err_count = 0;
1001 	}
1002 
1003 	/*
1004 	 * Okay, now see if we can stuff it in the buffer as-is.
1005 	 */
1006 	max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
1007 
1008 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1009 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1010 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
1011 		if (!new_skb) {
1012 			ip_rt_put(rt);
1013 			dev->stats.tx_dropped++;
1014 			kfree_skb(skb);
1015 			return NETDEV_TX_OK;
1016 		}
1017 		if (skb->sk)
1018 			skb_set_owner_w(new_skb, skb->sk);
1019 		dev_kfree_skb(skb);
1020 		skb = new_skb;
1021 		iph6 = ipv6_hdr(skb);
1022 	}
1023 	ttl = tiph->ttl;
1024 	if (ttl == 0)
1025 		ttl = iph6->hop_limit;
1026 	tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
1027 
1028 	if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
1029 		ip_rt_put(rt);
1030 		goto tx_error;
1031 	}
1032 
1033 	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1034 
1035 	iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
1036 		      df, !net_eq(tunnel->net, dev_net(dev)));
1037 	return NETDEV_TX_OK;
1038 
1039 tx_error_icmp:
1040 	dst_link_failure(skb);
1041 tx_error:
1042 	kfree_skb(skb);
1043 	dev->stats.tx_errors++;
1044 	return NETDEV_TX_OK;
1045 }
1046 
1047 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1048 				     struct net_device *dev, u8 ipproto)
1049 {
1050 	struct ip_tunnel *tunnel = netdev_priv(dev);
1051 	const struct iphdr  *tiph = &tunnel->parms.iph;
1052 
1053 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1054 		goto tx_error;
1055 
1056 	skb_set_inner_ipproto(skb, ipproto);
1057 
1058 	ip_tunnel_xmit(skb, dev, tiph, ipproto);
1059 	return NETDEV_TX_OK;
1060 tx_error:
1061 	kfree_skb(skb);
1062 	dev->stats.tx_errors++;
1063 	return NETDEV_TX_OK;
1064 }
1065 
1066 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1067 				   struct net_device *dev)
1068 {
1069 	if (!pskb_inet_may_pull(skb))
1070 		goto tx_err;
1071 
1072 	switch (skb->protocol) {
1073 	case htons(ETH_P_IP):
1074 		sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1075 		break;
1076 	case htons(ETH_P_IPV6):
1077 		ipip6_tunnel_xmit(skb, dev);
1078 		break;
1079 #if IS_ENABLED(CONFIG_MPLS)
1080 	case htons(ETH_P_MPLS_UC):
1081 		sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1082 		break;
1083 #endif
1084 	default:
1085 		goto tx_err;
1086 	}
1087 
1088 	return NETDEV_TX_OK;
1089 
1090 tx_err:
1091 	dev->stats.tx_errors++;
1092 	kfree_skb(skb);
1093 	return NETDEV_TX_OK;
1094 
1095 }
1096 
1097 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1098 {
1099 	struct net_device *tdev = NULL;
1100 	struct ip_tunnel *tunnel;
1101 	const struct iphdr *iph;
1102 	struct flowi4 fl4;
1103 
1104 	tunnel = netdev_priv(dev);
1105 	iph = &tunnel->parms.iph;
1106 
1107 	if (iph->daddr) {
1108 		struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1109 							  NULL,
1110 							  iph->daddr, iph->saddr,
1111 							  0, 0,
1112 							  IPPROTO_IPV6,
1113 							  RT_TOS(iph->tos),
1114 							  tunnel->parms.link);
1115 
1116 		if (!IS_ERR(rt)) {
1117 			tdev = rt->dst.dev;
1118 			ip_rt_put(rt);
1119 		}
1120 		dev->flags |= IFF_POINTOPOINT;
1121 	}
1122 
1123 	if (!tdev && tunnel->parms.link)
1124 		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1125 
1126 	if (tdev && !netif_is_l3_master(tdev)) {
1127 		int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1128 
1129 		dev->mtu = tdev->mtu - t_hlen;
1130 		if (dev->mtu < IPV6_MIN_MTU)
1131 			dev->mtu = IPV6_MIN_MTU;
1132 	}
1133 }
1134 
1135 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1136 				__u32 fwmark)
1137 {
1138 	struct net *net = t->net;
1139 	struct sit_net *sitn = net_generic(net, sit_net_id);
1140 
1141 	ipip6_tunnel_unlink(sitn, t);
1142 	synchronize_net();
1143 	t->parms.iph.saddr = p->iph.saddr;
1144 	t->parms.iph.daddr = p->iph.daddr;
1145 	memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1146 	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1147 	ipip6_tunnel_link(sitn, t);
1148 	t->parms.iph.ttl = p->iph.ttl;
1149 	t->parms.iph.tos = p->iph.tos;
1150 	t->parms.iph.frag_off = p->iph.frag_off;
1151 	if (t->parms.link != p->link || t->fwmark != fwmark) {
1152 		t->parms.link = p->link;
1153 		t->fwmark = fwmark;
1154 		ipip6_tunnel_bind_dev(t->dev);
1155 	}
1156 	dst_cache_reset(&t->dst_cache);
1157 	netdev_state_change(t->dev);
1158 }
1159 
1160 #ifdef CONFIG_IPV6_SIT_6RD
1161 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1162 				   struct ip_tunnel_6rd *ip6rd)
1163 {
1164 	struct in6_addr prefix;
1165 	__be32 relay_prefix;
1166 
1167 	if (ip6rd->relay_prefixlen > 32 ||
1168 	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1169 		return -EINVAL;
1170 
1171 	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1172 	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1173 		return -EINVAL;
1174 	if (ip6rd->relay_prefixlen)
1175 		relay_prefix = ip6rd->relay_prefix &
1176 			       htonl(0xffffffffUL <<
1177 				     (32 - ip6rd->relay_prefixlen));
1178 	else
1179 		relay_prefix = 0;
1180 	if (relay_prefix != ip6rd->relay_prefix)
1181 		return -EINVAL;
1182 
1183 	t->ip6rd.prefix = prefix;
1184 	t->ip6rd.relay_prefix = relay_prefix;
1185 	t->ip6rd.prefixlen = ip6rd->prefixlen;
1186 	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1187 	dst_cache_reset(&t->dst_cache);
1188 	netdev_state_change(t->dev);
1189 	return 0;
1190 }
1191 
1192 static int
1193 ipip6_tunnel_get6rd(struct net_device *dev, struct ifreq *ifr)
1194 {
1195 	struct ip_tunnel *t = netdev_priv(dev);
1196 	struct ip_tunnel_6rd ip6rd;
1197 	struct ip_tunnel_parm p;
1198 
1199 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1200 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1201 			return -EFAULT;
1202 		t = ipip6_tunnel_locate(t->net, &p, 0);
1203 	}
1204 	if (!t)
1205 		t = netdev_priv(dev);
1206 
1207 	ip6rd.prefix = t->ip6rd.prefix;
1208 	ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1209 	ip6rd.prefixlen = t->ip6rd.prefixlen;
1210 	ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1211 	if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd, sizeof(ip6rd)))
1212 		return -EFAULT;
1213 	return 0;
1214 }
1215 
1216 static int
1217 ipip6_tunnel_6rdctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1218 {
1219 	struct ip_tunnel *t = netdev_priv(dev);
1220 	struct ip_tunnel_6rd ip6rd;
1221 	int err;
1222 
1223 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1224 		return -EPERM;
1225 	if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data, sizeof(ip6rd)))
1226 		return -EFAULT;
1227 
1228 	if (cmd != SIOCDEL6RD) {
1229 		err = ipip6_tunnel_update_6rd(t, &ip6rd);
1230 		if (err < 0)
1231 			return err;
1232 	} else
1233 		ipip6_tunnel_clone_6rd(dev, dev_to_sit_net(dev));
1234 	return 0;
1235 }
1236 
1237 #endif /* CONFIG_IPV6_SIT_6RD */
1238 
1239 static bool ipip6_valid_ip_proto(u8 ipproto)
1240 {
1241 	return ipproto == IPPROTO_IPV6 ||
1242 		ipproto == IPPROTO_IPIP ||
1243 #if IS_ENABLED(CONFIG_MPLS)
1244 		ipproto == IPPROTO_MPLS ||
1245 #endif
1246 		ipproto == 0;
1247 }
1248 
1249 static int
1250 __ipip6_tunnel_ioctl_validate(struct net *net, struct ip_tunnel_parm *p)
1251 {
1252 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1253 		return -EPERM;
1254 
1255 	if (!ipip6_valid_ip_proto(p->iph.protocol))
1256 		return -EINVAL;
1257 	if (p->iph.version != 4 ||
1258 	    p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
1259 		return -EINVAL;
1260 
1261 	if (p->iph.ttl)
1262 		p->iph.frag_off |= htons(IP_DF);
1263 	return 0;
1264 }
1265 
1266 static int
1267 ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm *p)
1268 {
1269 	struct ip_tunnel *t = netdev_priv(dev);
1270 
1271 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
1272 		t = ipip6_tunnel_locate(t->net, p, 0);
1273 	if (!t)
1274 		t = netdev_priv(dev);
1275 	memcpy(p, &t->parms, sizeof(*p));
1276 	return 0;
1277 }
1278 
1279 static int
1280 ipip6_tunnel_add(struct net_device *dev, struct ip_tunnel_parm *p)
1281 {
1282 	struct ip_tunnel *t = netdev_priv(dev);
1283 	int err;
1284 
1285 	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1286 	if (err)
1287 		return err;
1288 
1289 	t = ipip6_tunnel_locate(t->net, p, 1);
1290 	if (!t)
1291 		return -ENOBUFS;
1292 	return 0;
1293 }
1294 
1295 static int
1296 ipip6_tunnel_change(struct net_device *dev, struct ip_tunnel_parm *p)
1297 {
1298 	struct ip_tunnel *t = netdev_priv(dev);
1299 	int err;
1300 
1301 	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1302 	if (err)
1303 		return err;
1304 
1305 	t = ipip6_tunnel_locate(t->net, p, 0);
1306 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1307 		if (!t)
1308 			return -ENOENT;
1309 	} else {
1310 		if (t) {
1311 			if (t->dev != dev)
1312 				return -EEXIST;
1313 		} else {
1314 			if (((dev->flags & IFF_POINTOPOINT) && !p->iph.daddr) ||
1315 			    (!(dev->flags & IFF_POINTOPOINT) && p->iph.daddr))
1316 				return -EINVAL;
1317 			t = netdev_priv(dev);
1318 		}
1319 
1320 		ipip6_tunnel_update(t, p, t->fwmark);
1321 	}
1322 
1323 	return 0;
1324 }
1325 
1326 static int
1327 ipip6_tunnel_del(struct net_device *dev, struct ip_tunnel_parm *p)
1328 {
1329 	struct ip_tunnel *t = netdev_priv(dev);
1330 
1331 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1332 		return -EPERM;
1333 
1334 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1335 		t = ipip6_tunnel_locate(t->net, p, 0);
1336 		if (!t)
1337 			return -ENOENT;
1338 		if (t == netdev_priv(dev_to_sit_net(dev)->fb_tunnel_dev))
1339 			return -EPERM;
1340 		dev = t->dev;
1341 	}
1342 	unregister_netdevice(dev);
1343 	return 0;
1344 }
1345 
1346 static int
1347 ipip6_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
1348 {
1349 	switch (cmd) {
1350 	case SIOCGETTUNNEL:
1351 		return ipip6_tunnel_get(dev, p);
1352 	case SIOCADDTUNNEL:
1353 		return ipip6_tunnel_add(dev, p);
1354 	case SIOCCHGTUNNEL:
1355 		return ipip6_tunnel_change(dev, p);
1356 	case SIOCDELTUNNEL:
1357 		return ipip6_tunnel_del(dev, p);
1358 	default:
1359 		return -EINVAL;
1360 	}
1361 }
1362 
1363 static int
1364 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1365 {
1366 	switch (cmd) {
1367 	case SIOCGETTUNNEL:
1368 	case SIOCADDTUNNEL:
1369 	case SIOCCHGTUNNEL:
1370 	case SIOCDELTUNNEL:
1371 		return ip_tunnel_ioctl(dev, ifr, cmd);
1372 	case SIOCGETPRL:
1373 		return ipip6_tunnel_get_prl(dev, ifr);
1374 	case SIOCADDPRL:
1375 	case SIOCDELPRL:
1376 	case SIOCCHGPRL:
1377 		return ipip6_tunnel_prl_ctl(dev, ifr, cmd);
1378 #ifdef CONFIG_IPV6_SIT_6RD
1379 	case SIOCGET6RD:
1380 		return ipip6_tunnel_get6rd(dev, ifr);
1381 	case SIOCADD6RD:
1382 	case SIOCCHG6RD:
1383 	case SIOCDEL6RD:
1384 		return ipip6_tunnel_6rdctl(dev, ifr, cmd);
1385 #endif
1386 	default:
1387 		return -EINVAL;
1388 	}
1389 }
1390 
1391 static const struct net_device_ops ipip6_netdev_ops = {
1392 	.ndo_init	= ipip6_tunnel_init,
1393 	.ndo_uninit	= ipip6_tunnel_uninit,
1394 	.ndo_start_xmit	= sit_tunnel_xmit,
1395 	.ndo_do_ioctl	= ipip6_tunnel_ioctl,
1396 	.ndo_get_stats64 = dev_get_tstats64,
1397 	.ndo_get_iflink = ip_tunnel_get_iflink,
1398 	.ndo_tunnel_ctl = ipip6_tunnel_ctl,
1399 };
1400 
1401 static void ipip6_dev_free(struct net_device *dev)
1402 {
1403 	struct ip_tunnel *tunnel = netdev_priv(dev);
1404 
1405 	dst_cache_destroy(&tunnel->dst_cache);
1406 	free_percpu(dev->tstats);
1407 }
1408 
1409 #define SIT_FEATURES (NETIF_F_SG	   | \
1410 		      NETIF_F_FRAGLIST	   | \
1411 		      NETIF_F_HIGHDMA	   | \
1412 		      NETIF_F_GSO_SOFTWARE | \
1413 		      NETIF_F_HW_CSUM)
1414 
1415 static void ipip6_tunnel_setup(struct net_device *dev)
1416 {
1417 	struct ip_tunnel *tunnel = netdev_priv(dev);
1418 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1419 
1420 	dev->netdev_ops		= &ipip6_netdev_ops;
1421 	dev->header_ops		= &ip_tunnel_header_ops;
1422 	dev->needs_free_netdev	= true;
1423 	dev->priv_destructor	= ipip6_dev_free;
1424 
1425 	dev->type		= ARPHRD_SIT;
1426 	dev->mtu		= ETH_DATA_LEN - t_hlen;
1427 	dev->min_mtu		= IPV6_MIN_MTU;
1428 	dev->max_mtu		= IP6_MAX_MTU - t_hlen;
1429 	dev->flags		= IFF_NOARP;
1430 	netif_keep_dst(dev);
1431 	dev->addr_len		= 4;
1432 	dev->features		|= NETIF_F_LLTX;
1433 	dev->features		|= SIT_FEATURES;
1434 	dev->hw_features	|= SIT_FEATURES;
1435 }
1436 
1437 static int ipip6_tunnel_init(struct net_device *dev)
1438 {
1439 	struct ip_tunnel *tunnel = netdev_priv(dev);
1440 	int err;
1441 
1442 	tunnel->dev = dev;
1443 	tunnel->net = dev_net(dev);
1444 	strcpy(tunnel->parms.name, dev->name);
1445 
1446 	ipip6_tunnel_bind_dev(dev);
1447 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1448 	if (!dev->tstats)
1449 		return -ENOMEM;
1450 
1451 	err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1452 	if (err) {
1453 		free_percpu(dev->tstats);
1454 		dev->tstats = NULL;
1455 		return err;
1456 	}
1457 	dev_hold(dev);
1458 	return 0;
1459 }
1460 
1461 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1462 {
1463 	struct ip_tunnel *tunnel = netdev_priv(dev);
1464 	struct iphdr *iph = &tunnel->parms.iph;
1465 	struct net *net = dev_net(dev);
1466 	struct sit_net *sitn = net_generic(net, sit_net_id);
1467 
1468 	iph->version		= 4;
1469 	iph->protocol		= IPPROTO_IPV6;
1470 	iph->ihl		= 5;
1471 	iph->ttl		= 64;
1472 
1473 	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1474 }
1475 
1476 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1477 			  struct netlink_ext_ack *extack)
1478 {
1479 	u8 proto;
1480 
1481 	if (!data || !data[IFLA_IPTUN_PROTO])
1482 		return 0;
1483 
1484 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1485 	if (!ipip6_valid_ip_proto(proto))
1486 		return -EINVAL;
1487 
1488 	return 0;
1489 }
1490 
1491 static void ipip6_netlink_parms(struct nlattr *data[],
1492 				struct ip_tunnel_parm *parms,
1493 				__u32 *fwmark)
1494 {
1495 	memset(parms, 0, sizeof(*parms));
1496 
1497 	parms->iph.version = 4;
1498 	parms->iph.protocol = IPPROTO_IPV6;
1499 	parms->iph.ihl = 5;
1500 	parms->iph.ttl = 64;
1501 
1502 	if (!data)
1503 		return;
1504 
1505 	if (data[IFLA_IPTUN_LINK])
1506 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1507 
1508 	if (data[IFLA_IPTUN_LOCAL])
1509 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1510 
1511 	if (data[IFLA_IPTUN_REMOTE])
1512 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1513 
1514 	if (data[IFLA_IPTUN_TTL]) {
1515 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1516 		if (parms->iph.ttl)
1517 			parms->iph.frag_off = htons(IP_DF);
1518 	}
1519 
1520 	if (data[IFLA_IPTUN_TOS])
1521 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1522 
1523 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1524 		parms->iph.frag_off = htons(IP_DF);
1525 
1526 	if (data[IFLA_IPTUN_FLAGS])
1527 		parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1528 
1529 	if (data[IFLA_IPTUN_PROTO])
1530 		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1531 
1532 	if (data[IFLA_IPTUN_FWMARK])
1533 		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1534 }
1535 
1536 /* This function returns true when ENCAP attributes are present in the nl msg */
1537 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1538 				      struct ip_tunnel_encap *ipencap)
1539 {
1540 	bool ret = false;
1541 
1542 	memset(ipencap, 0, sizeof(*ipencap));
1543 
1544 	if (!data)
1545 		return ret;
1546 
1547 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1548 		ret = true;
1549 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1550 	}
1551 
1552 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1553 		ret = true;
1554 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1555 	}
1556 
1557 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1558 		ret = true;
1559 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1560 	}
1561 
1562 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1563 		ret = true;
1564 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1565 	}
1566 
1567 	return ret;
1568 }
1569 
1570 #ifdef CONFIG_IPV6_SIT_6RD
1571 /* This function returns true when 6RD attributes are present in the nl msg */
1572 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1573 				    struct ip_tunnel_6rd *ip6rd)
1574 {
1575 	bool ret = false;
1576 	memset(ip6rd, 0, sizeof(*ip6rd));
1577 
1578 	if (!data)
1579 		return ret;
1580 
1581 	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1582 		ret = true;
1583 		ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1584 	}
1585 
1586 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1587 		ret = true;
1588 		ip6rd->relay_prefix =
1589 			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1590 	}
1591 
1592 	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1593 		ret = true;
1594 		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1595 	}
1596 
1597 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1598 		ret = true;
1599 		ip6rd->relay_prefixlen =
1600 			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1601 	}
1602 
1603 	return ret;
1604 }
1605 #endif
1606 
1607 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1608 			 struct nlattr *tb[], struct nlattr *data[],
1609 			 struct netlink_ext_ack *extack)
1610 {
1611 	struct net *net = dev_net(dev);
1612 	struct ip_tunnel *nt;
1613 	struct ip_tunnel_encap ipencap;
1614 #ifdef CONFIG_IPV6_SIT_6RD
1615 	struct ip_tunnel_6rd ip6rd;
1616 #endif
1617 	int err;
1618 
1619 	nt = netdev_priv(dev);
1620 
1621 	if (ipip6_netlink_encap_parms(data, &ipencap)) {
1622 		err = ip_tunnel_encap_setup(nt, &ipencap);
1623 		if (err < 0)
1624 			return err;
1625 	}
1626 
1627 	ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1628 
1629 	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1630 		return -EEXIST;
1631 
1632 	err = ipip6_tunnel_create(dev);
1633 	if (err < 0)
1634 		return err;
1635 
1636 	if (tb[IFLA_MTU]) {
1637 		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1638 
1639 		if (mtu >= IPV6_MIN_MTU &&
1640 		    mtu <= IP6_MAX_MTU - dev->hard_header_len)
1641 			dev->mtu = mtu;
1642 	}
1643 
1644 #ifdef CONFIG_IPV6_SIT_6RD
1645 	if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1646 		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1647 		if (err < 0)
1648 			unregister_netdevice_queue(dev, NULL);
1649 	}
1650 #endif
1651 
1652 	return err;
1653 }
1654 
1655 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1656 			    struct nlattr *data[],
1657 			    struct netlink_ext_ack *extack)
1658 {
1659 	struct ip_tunnel *t = netdev_priv(dev);
1660 	struct ip_tunnel_parm p;
1661 	struct ip_tunnel_encap ipencap;
1662 	struct net *net = t->net;
1663 	struct sit_net *sitn = net_generic(net, sit_net_id);
1664 #ifdef CONFIG_IPV6_SIT_6RD
1665 	struct ip_tunnel_6rd ip6rd;
1666 #endif
1667 	__u32 fwmark = t->fwmark;
1668 	int err;
1669 
1670 	if (dev == sitn->fb_tunnel_dev)
1671 		return -EINVAL;
1672 
1673 	if (ipip6_netlink_encap_parms(data, &ipencap)) {
1674 		err = ip_tunnel_encap_setup(t, &ipencap);
1675 		if (err < 0)
1676 			return err;
1677 	}
1678 
1679 	ipip6_netlink_parms(data, &p, &fwmark);
1680 
1681 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1682 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1683 		return -EINVAL;
1684 
1685 	t = ipip6_tunnel_locate(net, &p, 0);
1686 
1687 	if (t) {
1688 		if (t->dev != dev)
1689 			return -EEXIST;
1690 	} else
1691 		t = netdev_priv(dev);
1692 
1693 	ipip6_tunnel_update(t, &p, fwmark);
1694 
1695 #ifdef CONFIG_IPV6_SIT_6RD
1696 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1697 		return ipip6_tunnel_update_6rd(t, &ip6rd);
1698 #endif
1699 
1700 	return 0;
1701 }
1702 
1703 static size_t ipip6_get_size(const struct net_device *dev)
1704 {
1705 	return
1706 		/* IFLA_IPTUN_LINK */
1707 		nla_total_size(4) +
1708 		/* IFLA_IPTUN_LOCAL */
1709 		nla_total_size(4) +
1710 		/* IFLA_IPTUN_REMOTE */
1711 		nla_total_size(4) +
1712 		/* IFLA_IPTUN_TTL */
1713 		nla_total_size(1) +
1714 		/* IFLA_IPTUN_TOS */
1715 		nla_total_size(1) +
1716 		/* IFLA_IPTUN_PMTUDISC */
1717 		nla_total_size(1) +
1718 		/* IFLA_IPTUN_FLAGS */
1719 		nla_total_size(2) +
1720 		/* IFLA_IPTUN_PROTO */
1721 		nla_total_size(1) +
1722 #ifdef CONFIG_IPV6_SIT_6RD
1723 		/* IFLA_IPTUN_6RD_PREFIX */
1724 		nla_total_size(sizeof(struct in6_addr)) +
1725 		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1726 		nla_total_size(4) +
1727 		/* IFLA_IPTUN_6RD_PREFIXLEN */
1728 		nla_total_size(2) +
1729 		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1730 		nla_total_size(2) +
1731 #endif
1732 		/* IFLA_IPTUN_ENCAP_TYPE */
1733 		nla_total_size(2) +
1734 		/* IFLA_IPTUN_ENCAP_FLAGS */
1735 		nla_total_size(2) +
1736 		/* IFLA_IPTUN_ENCAP_SPORT */
1737 		nla_total_size(2) +
1738 		/* IFLA_IPTUN_ENCAP_DPORT */
1739 		nla_total_size(2) +
1740 		/* IFLA_IPTUN_FWMARK */
1741 		nla_total_size(4) +
1742 		0;
1743 }
1744 
1745 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1746 {
1747 	struct ip_tunnel *tunnel = netdev_priv(dev);
1748 	struct ip_tunnel_parm *parm = &tunnel->parms;
1749 
1750 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1751 	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1752 	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1753 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1754 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1755 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1756 		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1757 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1758 	    nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1759 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1760 		goto nla_put_failure;
1761 
1762 #ifdef CONFIG_IPV6_SIT_6RD
1763 	if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1764 			     &tunnel->ip6rd.prefix) ||
1765 	    nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1766 			    tunnel->ip6rd.relay_prefix) ||
1767 	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1768 			tunnel->ip6rd.prefixlen) ||
1769 	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1770 			tunnel->ip6rd.relay_prefixlen))
1771 		goto nla_put_failure;
1772 #endif
1773 
1774 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1775 			tunnel->encap.type) ||
1776 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1777 			tunnel->encap.sport) ||
1778 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1779 			tunnel->encap.dport) ||
1780 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1781 			tunnel->encap.flags))
1782 		goto nla_put_failure;
1783 
1784 	return 0;
1785 
1786 nla_put_failure:
1787 	return -EMSGSIZE;
1788 }
1789 
1790 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1791 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1792 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1793 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1794 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1795 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1796 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1797 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1798 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1799 #ifdef CONFIG_IPV6_SIT_6RD
1800 	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1801 	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1802 	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1803 	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1804 #endif
1805 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
1806 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
1807 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
1808 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
1809 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
1810 };
1811 
1812 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1813 {
1814 	struct net *net = dev_net(dev);
1815 	struct sit_net *sitn = net_generic(net, sit_net_id);
1816 
1817 	if (dev != sitn->fb_tunnel_dev)
1818 		unregister_netdevice_queue(dev, head);
1819 }
1820 
1821 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1822 	.kind		= "sit",
1823 	.maxtype	= IFLA_IPTUN_MAX,
1824 	.policy		= ipip6_policy,
1825 	.priv_size	= sizeof(struct ip_tunnel),
1826 	.setup		= ipip6_tunnel_setup,
1827 	.validate	= ipip6_validate,
1828 	.newlink	= ipip6_newlink,
1829 	.changelink	= ipip6_changelink,
1830 	.get_size	= ipip6_get_size,
1831 	.fill_info	= ipip6_fill_info,
1832 	.dellink	= ipip6_dellink,
1833 	.get_link_net	= ip_tunnel_get_link_net,
1834 };
1835 
1836 static struct xfrm_tunnel sit_handler __read_mostly = {
1837 	.handler	=	ipip6_rcv,
1838 	.err_handler	=	ipip6_err,
1839 	.priority	=	1,
1840 };
1841 
1842 static struct xfrm_tunnel ipip_handler __read_mostly = {
1843 	.handler	=	ipip_rcv,
1844 	.err_handler	=	ipip6_err,
1845 	.priority	=	2,
1846 };
1847 
1848 #if IS_ENABLED(CONFIG_MPLS)
1849 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1850 	.handler	=	mplsip_rcv,
1851 	.err_handler	=	ipip6_err,
1852 	.priority	=	2,
1853 };
1854 #endif
1855 
1856 static void __net_exit sit_destroy_tunnels(struct net *net,
1857 					   struct list_head *head)
1858 {
1859 	struct sit_net *sitn = net_generic(net, sit_net_id);
1860 	struct net_device *dev, *aux;
1861 	int prio;
1862 
1863 	for_each_netdev_safe(net, dev, aux)
1864 		if (dev->rtnl_link_ops == &sit_link_ops)
1865 			unregister_netdevice_queue(dev, head);
1866 
1867 	for (prio = 0; prio < 4; prio++) {
1868 		int h;
1869 		for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1870 			struct ip_tunnel *t;
1871 
1872 			t = rtnl_dereference(sitn->tunnels[prio][h]);
1873 			while (t) {
1874 				/* If dev is in the same netns, it has already
1875 				 * been added to the list by the previous loop.
1876 				 */
1877 				if (!net_eq(dev_net(t->dev), net))
1878 					unregister_netdevice_queue(t->dev,
1879 								   head);
1880 				t = rtnl_dereference(t->next);
1881 			}
1882 		}
1883 	}
1884 }
1885 
1886 static int __net_init sit_init_net(struct net *net)
1887 {
1888 	struct sit_net *sitn = net_generic(net, sit_net_id);
1889 	struct ip_tunnel *t;
1890 	int err;
1891 
1892 	sitn->tunnels[0] = sitn->tunnels_wc;
1893 	sitn->tunnels[1] = sitn->tunnels_l;
1894 	sitn->tunnels[2] = sitn->tunnels_r;
1895 	sitn->tunnels[3] = sitn->tunnels_r_l;
1896 
1897 	if (!net_has_fallback_tunnels(net))
1898 		return 0;
1899 
1900 	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1901 					   NET_NAME_UNKNOWN,
1902 					   ipip6_tunnel_setup);
1903 	if (!sitn->fb_tunnel_dev) {
1904 		err = -ENOMEM;
1905 		goto err_alloc_dev;
1906 	}
1907 	dev_net_set(sitn->fb_tunnel_dev, net);
1908 	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1909 	/* FB netdevice is special: we have one, and only one per netns.
1910 	 * Allowing to move it to another netns is clearly unsafe.
1911 	 */
1912 	sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1913 
1914 	err = register_netdev(sitn->fb_tunnel_dev);
1915 	if (err)
1916 		goto err_reg_dev;
1917 
1918 	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1919 	ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1920 
1921 	t = netdev_priv(sitn->fb_tunnel_dev);
1922 
1923 	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1924 	return 0;
1925 
1926 err_reg_dev:
1927 	ipip6_dev_free(sitn->fb_tunnel_dev);
1928 	free_netdev(sitn->fb_tunnel_dev);
1929 err_alloc_dev:
1930 	return err;
1931 }
1932 
1933 static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1934 {
1935 	LIST_HEAD(list);
1936 	struct net *net;
1937 
1938 	rtnl_lock();
1939 	list_for_each_entry(net, net_list, exit_list)
1940 		sit_destroy_tunnels(net, &list);
1941 
1942 	unregister_netdevice_many(&list);
1943 	rtnl_unlock();
1944 }
1945 
1946 static struct pernet_operations sit_net_ops = {
1947 	.init = sit_init_net,
1948 	.exit_batch = sit_exit_batch_net,
1949 	.id   = &sit_net_id,
1950 	.size = sizeof(struct sit_net),
1951 };
1952 
1953 static void __exit sit_cleanup(void)
1954 {
1955 	rtnl_link_unregister(&sit_link_ops);
1956 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1957 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1958 #if IS_ENABLED(CONFIG_MPLS)
1959 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1960 #endif
1961 
1962 	unregister_pernet_device(&sit_net_ops);
1963 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1964 }
1965 
1966 static int __init sit_init(void)
1967 {
1968 	int err;
1969 
1970 	pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1971 
1972 	err = register_pernet_device(&sit_net_ops);
1973 	if (err < 0)
1974 		return err;
1975 	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1976 	if (err < 0) {
1977 		pr_info("%s: can't register ip6ip4\n", __func__);
1978 		goto xfrm_tunnel_failed;
1979 	}
1980 	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1981 	if (err < 0) {
1982 		pr_info("%s: can't register ip4ip4\n", __func__);
1983 		goto xfrm_tunnel4_failed;
1984 	}
1985 #if IS_ENABLED(CONFIG_MPLS)
1986 	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1987 	if (err < 0) {
1988 		pr_info("%s: can't register mplsip\n", __func__);
1989 		goto xfrm_tunnel_mpls_failed;
1990 	}
1991 #endif
1992 	err = rtnl_link_register(&sit_link_ops);
1993 	if (err < 0)
1994 		goto rtnl_link_failed;
1995 
1996 out:
1997 	return err;
1998 
1999 rtnl_link_failed:
2000 #if IS_ENABLED(CONFIG_MPLS)
2001 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
2002 xfrm_tunnel_mpls_failed:
2003 #endif
2004 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
2005 xfrm_tunnel4_failed:
2006 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
2007 xfrm_tunnel_failed:
2008 	unregister_pernet_device(&sit_net_ops);
2009 	goto out;
2010 }
2011 
2012 module_init(sit_init);
2013 module_exit(sit_cleanup);
2014 MODULE_LICENSE("GPL");
2015 MODULE_ALIAS_RTNL_LINK("sit");
2016 MODULE_ALIAS_NETDEV("sit0");
2017