xref: /linux/net/ipv6/sit.c (revision bfb921b2a9d5d1123d1d10b196a39db629ddef87)
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 **
136 __ipip6_bucket(struct sit_net *sitn, struct ip_tunnel_parm_kern *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 	__dev_addr_set(dev, &t->parms.iph.saddr, 4);
208 	memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
209 
210 	if (test_bit(IP_TUNNEL_SIT_ISATAP_BIT, t->parms.i_flags))
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_kern *parms,
230 					     int create)
231 {
232 	__be32 remote = parms->iph.daddr;
233 	__be32 local = parms->iph.saddr;
234 	struct ip_tunnel *t, *nt;
235 	struct ip_tunnel __rcu **tp;
236 	struct net_device *dev;
237 	char name[IFNAMSIZ];
238 	struct sit_net *sitn = net_generic(net, sit_net_id);
239 
240 	for (tp = __ipip6_bucket(sitn, parms);
241 	    (t = rtnl_dereference(*tp)) != NULL;
242 	     tp = &t->next) {
243 		if (local == t->parms.iph.saddr &&
244 		    remote == t->parms.iph.daddr &&
245 		    parms->link == t->parms.link) {
246 			if (create)
247 				return NULL;
248 			else
249 				return t;
250 		}
251 	}
252 	if (!create)
253 		goto failed;
254 
255 	if (parms->name[0]) {
256 		if (!dev_valid_name(parms->name))
257 			goto failed;
258 		strscpy(name, parms->name, IFNAMSIZ);
259 	} else {
260 		strcpy(name, "sit%d");
261 	}
262 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
263 			   ipip6_tunnel_setup);
264 	if (!dev)
265 		return NULL;
266 
267 	dev_net_set(dev, net);
268 
269 	nt = netdev_priv(dev);
270 
271 	nt->parms = *parms;
272 	if (ipip6_tunnel_create(dev) < 0)
273 		goto failed_free;
274 
275 	if (!parms->name[0])
276 		strcpy(parms->name, dev->name);
277 
278 	return nt;
279 
280 failed_free:
281 	free_netdev(dev);
282 failed:
283 	return NULL;
284 }
285 
286 #define for_each_prl_rcu(start)			\
287 	for (prl = rcu_dereference(start);	\
288 	     prl;				\
289 	     prl = rcu_dereference(prl->next))
290 
291 static struct ip_tunnel_prl_entry *
292 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
293 {
294 	struct ip_tunnel_prl_entry *prl;
295 
296 	for_each_prl_rcu(t->prl)
297 		if (prl->addr == addr)
298 			break;
299 	return prl;
300 
301 }
302 
303 static int ipip6_tunnel_get_prl(struct net_device *dev, struct ip_tunnel_prl __user *a)
304 {
305 	struct ip_tunnel *t = netdev_priv(dev);
306 	struct ip_tunnel_prl kprl, *kp;
307 	struct ip_tunnel_prl_entry *prl;
308 	unsigned int cmax, c = 0, ca, len;
309 	int ret = 0;
310 
311 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
312 		return -EINVAL;
313 
314 	if (copy_from_user(&kprl, a, sizeof(kprl)))
315 		return -EFAULT;
316 	cmax = kprl.datalen / sizeof(kprl);
317 	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
318 		cmax = 1;
319 
320 	/* For simple GET or for root users,
321 	 * we try harder to allocate.
322 	 */
323 	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
324 		kcalloc(cmax, sizeof(*kp), GFP_KERNEL_ACCOUNT | __GFP_NOWARN) :
325 		NULL;
326 
327 	ca = min(t->prl_count, cmax);
328 
329 	if (!kp) {
330 		/* We don't try hard to allocate much memory for
331 		 * non-root users.
332 		 * For root users, retry allocating enough memory for
333 		 * the answer.
334 		 */
335 		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC | __GFP_ACCOUNT |
336 					      __GFP_NOWARN);
337 		if (!kp) {
338 			ret = -ENOMEM;
339 			goto out;
340 		}
341 	}
342 
343 	rcu_read_lock();
344 	for_each_prl_rcu(t->prl) {
345 		if (c >= cmax)
346 			break;
347 		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
348 			continue;
349 		kp[c].addr = prl->addr;
350 		kp[c].flags = prl->flags;
351 		c++;
352 		if (kprl.addr != htonl(INADDR_ANY))
353 			break;
354 	}
355 
356 	rcu_read_unlock();
357 
358 	len = sizeof(*kp) * c;
359 	ret = 0;
360 	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
361 		ret = -EFAULT;
362 
363 	kfree(kp);
364 out:
365 	return ret;
366 }
367 
368 static int
369 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
370 {
371 	struct ip_tunnel_prl_entry *p;
372 	int err = 0;
373 
374 	if (a->addr == htonl(INADDR_ANY))
375 		return -EINVAL;
376 
377 	ASSERT_RTNL();
378 
379 	for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
380 		if (p->addr == a->addr) {
381 			if (chg) {
382 				p->flags = a->flags;
383 				goto out;
384 			}
385 			err = -EEXIST;
386 			goto out;
387 		}
388 	}
389 
390 	if (chg) {
391 		err = -ENXIO;
392 		goto out;
393 	}
394 
395 	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
396 	if (!p) {
397 		err = -ENOBUFS;
398 		goto out;
399 	}
400 
401 	p->next = t->prl;
402 	p->addr = a->addr;
403 	p->flags = a->flags;
404 	t->prl_count++;
405 	rcu_assign_pointer(t->prl, p);
406 out:
407 	return err;
408 }
409 
410 static void prl_list_destroy_rcu(struct rcu_head *head)
411 {
412 	struct ip_tunnel_prl_entry *p, *n;
413 
414 	p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
415 	do {
416 		n = rcu_dereference_protected(p->next, 1);
417 		kfree(p);
418 		p = n;
419 	} while (p);
420 }
421 
422 static int
423 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
424 {
425 	struct ip_tunnel_prl_entry *x;
426 	struct ip_tunnel_prl_entry __rcu **p;
427 	int err = 0;
428 
429 	ASSERT_RTNL();
430 
431 	if (a && a->addr != htonl(INADDR_ANY)) {
432 		for (p = &t->prl;
433 		     (x = rtnl_dereference(*p)) != NULL;
434 		     p = &x->next) {
435 			if (x->addr == a->addr) {
436 				*p = x->next;
437 				kfree_rcu(x, rcu_head);
438 				t->prl_count--;
439 				goto out;
440 			}
441 		}
442 		err = -ENXIO;
443 	} else {
444 		x = rtnl_dereference(t->prl);
445 		if (x) {
446 			t->prl_count = 0;
447 			call_rcu(&x->rcu_head, prl_list_destroy_rcu);
448 			t->prl = NULL;
449 		}
450 	}
451 out:
452 	return err;
453 }
454 
455 static int ipip6_tunnel_prl_ctl(struct net_device *dev,
456 				struct ip_tunnel_prl __user *data, int cmd)
457 {
458 	struct ip_tunnel *t = netdev_priv(dev);
459 	struct ip_tunnel_prl prl;
460 	int err;
461 
462 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
463 		return -EPERM;
464 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
465 		return -EINVAL;
466 
467 	if (copy_from_user(&prl, data, sizeof(prl)))
468 		return -EFAULT;
469 
470 	switch (cmd) {
471 	case SIOCDELPRL:
472 		err = ipip6_tunnel_del_prl(t, &prl);
473 		break;
474 	case SIOCADDPRL:
475 	case SIOCCHGPRL:
476 		err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
477 		break;
478 	}
479 	dst_cache_reset(&t->dst_cache);
480 	netdev_state_change(dev);
481 	return err;
482 }
483 
484 static int
485 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
486 {
487 	struct ip_tunnel_prl_entry *p;
488 	int ok = 1;
489 
490 	rcu_read_lock();
491 	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
492 	if (p) {
493 		if (p->flags & PRL_DEFAULT)
494 			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
495 		else
496 			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
497 	} else {
498 		const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
499 
500 		if (ipv6_addr_is_isatap(addr6) &&
501 		    (addr6->s6_addr32[3] == iph->saddr) &&
502 		    ipv6_chk_prefix(addr6, t->dev))
503 			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
504 		else
505 			ok = 0;
506 	}
507 	rcu_read_unlock();
508 	return ok;
509 }
510 
511 static void ipip6_tunnel_uninit(struct net_device *dev)
512 {
513 	struct ip_tunnel *tunnel = netdev_priv(dev);
514 	struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
515 
516 	if (dev == sitn->fb_tunnel_dev) {
517 		RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
518 	} else {
519 		ipip6_tunnel_unlink(sitn, tunnel);
520 		ipip6_tunnel_del_prl(tunnel, NULL);
521 	}
522 	dst_cache_reset(&tunnel->dst_cache);
523 	netdev_put(dev, &tunnel->dev_tracker);
524 }
525 
526 static int ipip6_err(struct sk_buff *skb, u32 info)
527 {
528 	const struct iphdr *iph = (const struct iphdr *)skb->data;
529 	const int type = icmp_hdr(skb)->type;
530 	const int code = icmp_hdr(skb)->code;
531 	unsigned int data_len = 0;
532 	struct ip_tunnel *t;
533 	int sifindex;
534 	int err;
535 
536 	switch (type) {
537 	default:
538 	case ICMP_PARAMETERPROB:
539 		return 0;
540 
541 	case ICMP_DEST_UNREACH:
542 		switch (code) {
543 		case ICMP_SR_FAILED:
544 			/* Impossible event. */
545 			return 0;
546 		default:
547 			/* All others are translated to HOST_UNREACH.
548 			   rfc2003 contains "deep thoughts" about NET_UNREACH,
549 			   I believe they are just ether pollution. --ANK
550 			 */
551 			break;
552 		}
553 		break;
554 	case ICMP_TIME_EXCEEDED:
555 		if (code != ICMP_EXC_TTL)
556 			return 0;
557 		data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
558 		break;
559 	case ICMP_REDIRECT:
560 		break;
561 	}
562 
563 	err = -ENOENT;
564 
565 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
566 	t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
567 				iph->daddr, iph->saddr, sifindex);
568 	if (!t)
569 		goto out;
570 
571 	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
572 		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
573 				 t->parms.link, iph->protocol);
574 		err = 0;
575 		goto out;
576 	}
577 	if (type == ICMP_REDIRECT) {
578 		ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
579 			      iph->protocol);
580 		err = 0;
581 		goto out;
582 	}
583 
584 	err = 0;
585 	if (__in6_dev_get(skb->dev) &&
586 	    !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
587 		goto out;
588 
589 	if (t->parms.iph.daddr == 0)
590 		goto out;
591 
592 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
593 		goto out;
594 
595 	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
596 		t->err_count++;
597 	else
598 		t->err_count = 1;
599 	t->err_time = jiffies;
600 out:
601 	return err;
602 }
603 
604 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
605 				  const struct in6_addr *v6addr)
606 {
607 	__be32 v4embed = 0;
608 	if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
609 		return true;
610 	return false;
611 }
612 
613 /* Checks if an address matches an address on the tunnel interface.
614  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
615  * Long story:
616  * This function is called after we considered the packet as spoofed
617  * in is_spoofed_6rd.
618  * We may have a router that is doing NAT for proto 41 packets
619  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
620  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
621  * function will return true, dropping the packet.
622  * But, we can still check if is spoofed against the IP
623  * addresses associated with the interface.
624  */
625 static bool only_dnatted(const struct ip_tunnel *tunnel,
626 	const struct in6_addr *v6dst)
627 {
628 	int prefix_len;
629 
630 #ifdef CONFIG_IPV6_SIT_6RD
631 	prefix_len = tunnel->ip6rd.prefixlen + 32
632 		- tunnel->ip6rd.relay_prefixlen;
633 #else
634 	prefix_len = 48;
635 #endif
636 	return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
637 }
638 
639 /* Returns true if a packet is spoofed */
640 static bool packet_is_spoofed(struct sk_buff *skb,
641 			      const struct iphdr *iph,
642 			      struct ip_tunnel *tunnel)
643 {
644 	const struct ipv6hdr *ipv6h;
645 
646 	if (tunnel->dev->priv_flags & IFF_ISATAP) {
647 		if (!isatap_chksrc(skb, iph, tunnel))
648 			return true;
649 
650 		return false;
651 	}
652 
653 	if (tunnel->dev->flags & IFF_POINTOPOINT)
654 		return false;
655 
656 	ipv6h = ipv6_hdr(skb);
657 
658 	if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
659 		net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
660 				     &iph->saddr, &ipv6h->saddr,
661 				     &iph->daddr, &ipv6h->daddr);
662 		return true;
663 	}
664 
665 	if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
666 		return false;
667 
668 	if (only_dnatted(tunnel, &ipv6h->daddr))
669 		return false;
670 
671 	net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
672 			     &iph->saddr, &ipv6h->saddr,
673 			     &iph->daddr, &ipv6h->daddr);
674 	return true;
675 }
676 
677 static int ipip6_rcv(struct sk_buff *skb)
678 {
679 	const struct iphdr *iph = ip_hdr(skb);
680 	struct ip_tunnel *tunnel;
681 	int sifindex;
682 	int err;
683 
684 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
685 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
686 				     iph->saddr, iph->daddr, sifindex);
687 	if (tunnel) {
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 			DEV_STATS_INC(tunnel->dev, 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 		skb_reset_mac_header(skb);
711 
712 		err = IP_ECN_decapsulate(iph, skb);
713 		if (unlikely(err)) {
714 			if (log_ecn_error)
715 				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
716 						     &iph->saddr, iph->tos);
717 			if (err > 1) {
718 				DEV_STATS_INC(tunnel->dev, rx_frame_errors);
719 				DEV_STATS_INC(tunnel->dev, rx_errors);
720 				goto out;
721 			}
722 		}
723 
724 		dev_sw_netstats_rx_add(tunnel->dev, skb->len);
725 
726 		netif_rx(skb);
727 
728 		return 0;
729 	}
730 
731 	/* no tunnel matched,  let upstream know, ipsec may handle it */
732 	return 1;
733 out:
734 	kfree_skb(skb);
735 	return 0;
736 }
737 
738 static const struct tnl_ptk_info ipip_tpi = {
739 	/* no tunnel info required for ipip. */
740 	.proto = htons(ETH_P_IP),
741 };
742 
743 #if IS_ENABLED(CONFIG_MPLS)
744 static const struct tnl_ptk_info mplsip_tpi = {
745 	/* no tunnel info required for mplsip. */
746 	.proto = htons(ETH_P_MPLS_UC),
747 };
748 #endif
749 
750 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
751 {
752 	const struct iphdr *iph;
753 	struct ip_tunnel *tunnel;
754 	int sifindex;
755 
756 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
757 
758 	iph = ip_hdr(skb);
759 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
760 				     iph->saddr, iph->daddr, sifindex);
761 	if (tunnel) {
762 		const struct tnl_ptk_info *tpi;
763 
764 		if (tunnel->parms.iph.protocol != ipproto &&
765 		    tunnel->parms.iph.protocol != 0)
766 			goto drop;
767 
768 		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
769 			goto drop;
770 #if IS_ENABLED(CONFIG_MPLS)
771 		if (ipproto == IPPROTO_MPLS)
772 			tpi = &mplsip_tpi;
773 		else
774 #endif
775 			tpi = &ipip_tpi;
776 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
777 			goto drop;
778 		skb_reset_mac_header(skb);
779 
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_INC(dev, 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 && rt->rt_type != RTN_LOCAL) {
953 		ip_rt_put(rt);
954 		DEV_STATS_INC(dev, 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_INC(dev, 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 < IPV4_MIN_MTU) {
974 			DEV_STATS_INC(dev, 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_INC(dev, 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->encap, &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_INC(dev, 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_INC(dev, 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_INC(dev, 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 ip_tunnel *tunnel = netdev_priv(dev);
1100 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1101 	struct net_device *tdev = NULL;
1102 	int hlen = LL_MAX_HEADER;
1103 	const struct iphdr *iph;
1104 	struct flowi4 fl4;
1105 
1106 	iph = &tunnel->parms.iph;
1107 
1108 	if (iph->daddr) {
1109 		struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1110 							  NULL,
1111 							  iph->daddr, iph->saddr,
1112 							  0, 0,
1113 							  IPPROTO_IPV6,
1114 							  RT_TOS(iph->tos),
1115 							  tunnel->parms.link);
1116 
1117 		if (!IS_ERR(rt)) {
1118 			tdev = rt->dst.dev;
1119 			ip_rt_put(rt);
1120 		}
1121 		dev->flags |= IFF_POINTOPOINT;
1122 	}
1123 
1124 	if (!tdev && tunnel->parms.link)
1125 		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1126 
1127 	if (tdev && !netif_is_l3_master(tdev)) {
1128 		int mtu;
1129 
1130 		mtu = tdev->mtu - t_hlen;
1131 		if (mtu < IPV6_MIN_MTU)
1132 			mtu = IPV6_MIN_MTU;
1133 		WRITE_ONCE(dev->mtu, mtu);
1134 		hlen = tdev->hard_header_len + tdev->needed_headroom;
1135 	}
1136 	dev->needed_headroom = t_hlen + hlen;
1137 }
1138 
1139 static void ipip6_tunnel_update(struct ip_tunnel *t,
1140 				struct ip_tunnel_parm_kern *p,
1141 				__u32 fwmark)
1142 {
1143 	struct net *net = t->net;
1144 	struct sit_net *sitn = net_generic(net, sit_net_id);
1145 
1146 	ipip6_tunnel_unlink(sitn, t);
1147 	synchronize_net();
1148 	t->parms.iph.saddr = p->iph.saddr;
1149 	t->parms.iph.daddr = p->iph.daddr;
1150 	__dev_addr_set(t->dev, &p->iph.saddr, 4);
1151 	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1152 	ipip6_tunnel_link(sitn, t);
1153 	t->parms.iph.ttl = p->iph.ttl;
1154 	t->parms.iph.tos = p->iph.tos;
1155 	t->parms.iph.frag_off = p->iph.frag_off;
1156 	if (t->parms.link != p->link || t->fwmark != fwmark) {
1157 		t->parms.link = p->link;
1158 		t->fwmark = fwmark;
1159 		ipip6_tunnel_bind_dev(t->dev);
1160 	}
1161 	dst_cache_reset(&t->dst_cache);
1162 	netdev_state_change(t->dev);
1163 }
1164 
1165 #ifdef CONFIG_IPV6_SIT_6RD
1166 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1167 				   struct ip_tunnel_6rd *ip6rd)
1168 {
1169 	struct in6_addr prefix;
1170 	__be32 relay_prefix;
1171 
1172 	if (ip6rd->relay_prefixlen > 32 ||
1173 	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1174 		return -EINVAL;
1175 
1176 	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1177 	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1178 		return -EINVAL;
1179 	if (ip6rd->relay_prefixlen)
1180 		relay_prefix = ip6rd->relay_prefix &
1181 			       htonl(0xffffffffUL <<
1182 				     (32 - ip6rd->relay_prefixlen));
1183 	else
1184 		relay_prefix = 0;
1185 	if (relay_prefix != ip6rd->relay_prefix)
1186 		return -EINVAL;
1187 
1188 	t->ip6rd.prefix = prefix;
1189 	t->ip6rd.relay_prefix = relay_prefix;
1190 	t->ip6rd.prefixlen = ip6rd->prefixlen;
1191 	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1192 	dst_cache_reset(&t->dst_cache);
1193 	netdev_state_change(t->dev);
1194 	return 0;
1195 }
1196 
1197 static int
1198 ipip6_tunnel_get6rd(struct net_device *dev, struct ip_tunnel_parm __user *data)
1199 {
1200 	struct ip_tunnel *t = netdev_priv(dev);
1201 	struct ip_tunnel_parm_kern p;
1202 	struct ip_tunnel_6rd ip6rd;
1203 
1204 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1205 		if (!ip_tunnel_parm_from_user(&p, data))
1206 			return -EFAULT;
1207 		t = ipip6_tunnel_locate(t->net, &p, 0);
1208 	}
1209 	if (!t)
1210 		t = netdev_priv(dev);
1211 
1212 	ip6rd.prefix = t->ip6rd.prefix;
1213 	ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1214 	ip6rd.prefixlen = t->ip6rd.prefixlen;
1215 	ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1216 	if (copy_to_user(data, &ip6rd, sizeof(ip6rd)))
1217 		return -EFAULT;
1218 	return 0;
1219 }
1220 
1221 static int
1222 ipip6_tunnel_6rdctl(struct net_device *dev, struct ip_tunnel_6rd __user *data,
1223 		    int cmd)
1224 {
1225 	struct ip_tunnel *t = netdev_priv(dev);
1226 	struct ip_tunnel_6rd ip6rd;
1227 	int err;
1228 
1229 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1230 		return -EPERM;
1231 	if (copy_from_user(&ip6rd, data, sizeof(ip6rd)))
1232 		return -EFAULT;
1233 
1234 	if (cmd != SIOCDEL6RD) {
1235 		err = ipip6_tunnel_update_6rd(t, &ip6rd);
1236 		if (err < 0)
1237 			return err;
1238 	} else
1239 		ipip6_tunnel_clone_6rd(dev, dev_to_sit_net(dev));
1240 	return 0;
1241 }
1242 
1243 #endif /* CONFIG_IPV6_SIT_6RD */
1244 
1245 static bool ipip6_valid_ip_proto(u8 ipproto)
1246 {
1247 	return ipproto == IPPROTO_IPV6 ||
1248 		ipproto == IPPROTO_IPIP ||
1249 #if IS_ENABLED(CONFIG_MPLS)
1250 		ipproto == IPPROTO_MPLS ||
1251 #endif
1252 		ipproto == 0;
1253 }
1254 
1255 static int
1256 __ipip6_tunnel_ioctl_validate(struct net *net, struct ip_tunnel_parm_kern *p)
1257 {
1258 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1259 		return -EPERM;
1260 
1261 	if (!ipip6_valid_ip_proto(p->iph.protocol))
1262 		return -EINVAL;
1263 	if (p->iph.version != 4 ||
1264 	    p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
1265 		return -EINVAL;
1266 
1267 	if (p->iph.ttl)
1268 		p->iph.frag_off |= htons(IP_DF);
1269 	return 0;
1270 }
1271 
1272 static int
1273 ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1274 {
1275 	struct ip_tunnel *t = netdev_priv(dev);
1276 
1277 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
1278 		t = ipip6_tunnel_locate(t->net, p, 0);
1279 	if (!t)
1280 		t = netdev_priv(dev);
1281 	memcpy(p, &t->parms, sizeof(*p));
1282 	return 0;
1283 }
1284 
1285 static int
1286 ipip6_tunnel_add(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1287 {
1288 	struct ip_tunnel *t = netdev_priv(dev);
1289 	int err;
1290 
1291 	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1292 	if (err)
1293 		return err;
1294 
1295 	t = ipip6_tunnel_locate(t->net, p, 1);
1296 	if (!t)
1297 		return -ENOBUFS;
1298 	return 0;
1299 }
1300 
1301 static int
1302 ipip6_tunnel_change(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1303 {
1304 	struct ip_tunnel *t = netdev_priv(dev);
1305 	int err;
1306 
1307 	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1308 	if (err)
1309 		return err;
1310 
1311 	t = ipip6_tunnel_locate(t->net, p, 0);
1312 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1313 		if (!t)
1314 			return -ENOENT;
1315 	} else {
1316 		if (t) {
1317 			if (t->dev != dev)
1318 				return -EEXIST;
1319 		} else {
1320 			if (((dev->flags & IFF_POINTOPOINT) && !p->iph.daddr) ||
1321 			    (!(dev->flags & IFF_POINTOPOINT) && p->iph.daddr))
1322 				return -EINVAL;
1323 			t = netdev_priv(dev);
1324 		}
1325 
1326 		ipip6_tunnel_update(t, p, t->fwmark);
1327 	}
1328 
1329 	return 0;
1330 }
1331 
1332 static int
1333 ipip6_tunnel_del(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1334 {
1335 	struct ip_tunnel *t = netdev_priv(dev);
1336 
1337 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1338 		return -EPERM;
1339 
1340 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1341 		t = ipip6_tunnel_locate(t->net, p, 0);
1342 		if (!t)
1343 			return -ENOENT;
1344 		if (t == netdev_priv(dev_to_sit_net(dev)->fb_tunnel_dev))
1345 			return -EPERM;
1346 		dev = t->dev;
1347 	}
1348 	unregister_netdevice(dev);
1349 	return 0;
1350 }
1351 
1352 static int
1353 ipip6_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm_kern *p,
1354 		 int cmd)
1355 {
1356 	switch (cmd) {
1357 	case SIOCGETTUNNEL:
1358 		return ipip6_tunnel_get(dev, p);
1359 	case SIOCADDTUNNEL:
1360 		return ipip6_tunnel_add(dev, p);
1361 	case SIOCCHGTUNNEL:
1362 		return ipip6_tunnel_change(dev, p);
1363 	case SIOCDELTUNNEL:
1364 		return ipip6_tunnel_del(dev, p);
1365 	default:
1366 		return -EINVAL;
1367 	}
1368 }
1369 
1370 static int
1371 ipip6_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1372 			    void __user *data, int cmd)
1373 {
1374 	switch (cmd) {
1375 	case SIOCGETTUNNEL:
1376 	case SIOCADDTUNNEL:
1377 	case SIOCCHGTUNNEL:
1378 	case SIOCDELTUNNEL:
1379 		return ip_tunnel_siocdevprivate(dev, ifr, data, cmd);
1380 	case SIOCGETPRL:
1381 		return ipip6_tunnel_get_prl(dev, data);
1382 	case SIOCADDPRL:
1383 	case SIOCDELPRL:
1384 	case SIOCCHGPRL:
1385 		return ipip6_tunnel_prl_ctl(dev, data, cmd);
1386 #ifdef CONFIG_IPV6_SIT_6RD
1387 	case SIOCGET6RD:
1388 		return ipip6_tunnel_get6rd(dev, data);
1389 	case SIOCADD6RD:
1390 	case SIOCCHG6RD:
1391 	case SIOCDEL6RD:
1392 		return ipip6_tunnel_6rdctl(dev, data, cmd);
1393 #endif
1394 	default:
1395 		return -EINVAL;
1396 	}
1397 }
1398 
1399 static const struct net_device_ops ipip6_netdev_ops = {
1400 	.ndo_init	= ipip6_tunnel_init,
1401 	.ndo_uninit	= ipip6_tunnel_uninit,
1402 	.ndo_start_xmit	= sit_tunnel_xmit,
1403 	.ndo_siocdevprivate = ipip6_tunnel_siocdevprivate,
1404 	.ndo_get_iflink = ip_tunnel_get_iflink,
1405 	.ndo_tunnel_ctl = ipip6_tunnel_ctl,
1406 };
1407 
1408 static void ipip6_dev_free(struct net_device *dev)
1409 {
1410 	struct ip_tunnel *tunnel = netdev_priv(dev);
1411 
1412 	dst_cache_destroy(&tunnel->dst_cache);
1413 }
1414 
1415 #define SIT_FEATURES (NETIF_F_SG	   | \
1416 		      NETIF_F_FRAGLIST	   | \
1417 		      NETIF_F_HIGHDMA	   | \
1418 		      NETIF_F_GSO_SOFTWARE | \
1419 		      NETIF_F_HW_CSUM)
1420 
1421 static void ipip6_tunnel_setup(struct net_device *dev)
1422 {
1423 	struct ip_tunnel *tunnel = netdev_priv(dev);
1424 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1425 
1426 	dev->netdev_ops		= &ipip6_netdev_ops;
1427 	dev->header_ops		= &ip_tunnel_header_ops;
1428 	dev->needs_free_netdev	= true;
1429 	dev->priv_destructor	= ipip6_dev_free;
1430 
1431 	dev->type		= ARPHRD_SIT;
1432 	dev->mtu		= ETH_DATA_LEN - t_hlen;
1433 	dev->min_mtu		= IPV6_MIN_MTU;
1434 	dev->max_mtu		= IP6_MAX_MTU - t_hlen;
1435 	dev->flags		= IFF_NOARP;
1436 	netif_keep_dst(dev);
1437 	dev->addr_len		= 4;
1438 	dev->features		|= NETIF_F_LLTX;
1439 	dev->features		|= SIT_FEATURES;
1440 	dev->hw_features	|= SIT_FEATURES;
1441 	dev->pcpu_stat_type	= NETDEV_PCPU_STAT_TSTATS;
1442 
1443 }
1444 
1445 static int ipip6_tunnel_init(struct net_device *dev)
1446 {
1447 	struct ip_tunnel *tunnel = netdev_priv(dev);
1448 	int err;
1449 
1450 	tunnel->dev = dev;
1451 	tunnel->net = dev_net(dev);
1452 	strcpy(tunnel->parms.name, dev->name);
1453 
1454 	ipip6_tunnel_bind_dev(dev);
1455 
1456 	err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1457 	if (err)
1458 		return err;
1459 
1460 	netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1461 	netdev_lockdep_set_classes(dev);
1462 	return 0;
1463 }
1464 
1465 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1466 {
1467 	struct ip_tunnel *tunnel = netdev_priv(dev);
1468 	struct iphdr *iph = &tunnel->parms.iph;
1469 	struct net *net = dev_net(dev);
1470 	struct sit_net *sitn = net_generic(net, sit_net_id);
1471 
1472 	iph->version		= 4;
1473 	iph->protocol		= IPPROTO_IPV6;
1474 	iph->ihl		= 5;
1475 	iph->ttl		= 64;
1476 
1477 	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1478 }
1479 
1480 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1481 			  struct netlink_ext_ack *extack)
1482 {
1483 	u8 proto;
1484 
1485 	if (!data || !data[IFLA_IPTUN_PROTO])
1486 		return 0;
1487 
1488 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1489 	if (!ipip6_valid_ip_proto(proto))
1490 		return -EINVAL;
1491 
1492 	return 0;
1493 }
1494 
1495 static void ipip6_netlink_parms(struct nlattr *data[],
1496 				struct ip_tunnel_parm_kern *parms,
1497 				__u32 *fwmark)
1498 {
1499 	memset(parms, 0, sizeof(*parms));
1500 
1501 	parms->iph.version = 4;
1502 	parms->iph.protocol = IPPROTO_IPV6;
1503 	parms->iph.ihl = 5;
1504 	parms->iph.ttl = 64;
1505 
1506 	if (!data)
1507 		return;
1508 
1509 	ip_tunnel_netlink_parms(data, parms);
1510 
1511 	if (data[IFLA_IPTUN_FWMARK])
1512 		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1513 }
1514 
1515 #ifdef CONFIG_IPV6_SIT_6RD
1516 /* This function returns true when 6RD attributes are present in the nl msg */
1517 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1518 				    struct ip_tunnel_6rd *ip6rd)
1519 {
1520 	bool ret = false;
1521 	memset(ip6rd, 0, sizeof(*ip6rd));
1522 
1523 	if (!data)
1524 		return ret;
1525 
1526 	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1527 		ret = true;
1528 		ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1529 	}
1530 
1531 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1532 		ret = true;
1533 		ip6rd->relay_prefix =
1534 			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1535 	}
1536 
1537 	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1538 		ret = true;
1539 		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1540 	}
1541 
1542 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1543 		ret = true;
1544 		ip6rd->relay_prefixlen =
1545 			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1546 	}
1547 
1548 	return ret;
1549 }
1550 #endif
1551 
1552 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1553 			 struct nlattr *tb[], struct nlattr *data[],
1554 			 struct netlink_ext_ack *extack)
1555 {
1556 	struct net *net = dev_net(dev);
1557 	struct ip_tunnel *nt;
1558 	struct ip_tunnel_encap ipencap;
1559 #ifdef CONFIG_IPV6_SIT_6RD
1560 	struct ip_tunnel_6rd ip6rd;
1561 #endif
1562 	int err;
1563 
1564 	nt = netdev_priv(dev);
1565 
1566 	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
1567 		err = ip_tunnel_encap_setup(nt, &ipencap);
1568 		if (err < 0)
1569 			return err;
1570 	}
1571 
1572 	ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1573 
1574 	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1575 		return -EEXIST;
1576 
1577 	err = ipip6_tunnel_create(dev);
1578 	if (err < 0)
1579 		return err;
1580 
1581 	if (tb[IFLA_MTU]) {
1582 		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1583 
1584 		if (mtu >= IPV6_MIN_MTU &&
1585 		    mtu <= IP6_MAX_MTU - dev->hard_header_len)
1586 			dev->mtu = mtu;
1587 	}
1588 
1589 #ifdef CONFIG_IPV6_SIT_6RD
1590 	if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1591 		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1592 		if (err < 0)
1593 			unregister_netdevice_queue(dev, NULL);
1594 	}
1595 #endif
1596 
1597 	return err;
1598 }
1599 
1600 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1601 			    struct nlattr *data[],
1602 			    struct netlink_ext_ack *extack)
1603 {
1604 	struct ip_tunnel *t = netdev_priv(dev);
1605 	struct ip_tunnel_encap ipencap;
1606 	struct ip_tunnel_parm_kern p;
1607 	struct net *net = t->net;
1608 	struct sit_net *sitn = net_generic(net, sit_net_id);
1609 #ifdef CONFIG_IPV6_SIT_6RD
1610 	struct ip_tunnel_6rd ip6rd;
1611 #endif
1612 	__u32 fwmark = t->fwmark;
1613 	int err;
1614 
1615 	if (dev == sitn->fb_tunnel_dev)
1616 		return -EINVAL;
1617 
1618 	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
1619 		err = ip_tunnel_encap_setup(t, &ipencap);
1620 		if (err < 0)
1621 			return err;
1622 	}
1623 
1624 	ipip6_netlink_parms(data, &p, &fwmark);
1625 
1626 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1627 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1628 		return -EINVAL;
1629 
1630 	t = ipip6_tunnel_locate(net, &p, 0);
1631 
1632 	if (t) {
1633 		if (t->dev != dev)
1634 			return -EEXIST;
1635 	} else
1636 		t = netdev_priv(dev);
1637 
1638 	ipip6_tunnel_update(t, &p, fwmark);
1639 
1640 #ifdef CONFIG_IPV6_SIT_6RD
1641 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1642 		return ipip6_tunnel_update_6rd(t, &ip6rd);
1643 #endif
1644 
1645 	return 0;
1646 }
1647 
1648 static size_t ipip6_get_size(const struct net_device *dev)
1649 {
1650 	return
1651 		/* IFLA_IPTUN_LINK */
1652 		nla_total_size(4) +
1653 		/* IFLA_IPTUN_LOCAL */
1654 		nla_total_size(4) +
1655 		/* IFLA_IPTUN_REMOTE */
1656 		nla_total_size(4) +
1657 		/* IFLA_IPTUN_TTL */
1658 		nla_total_size(1) +
1659 		/* IFLA_IPTUN_TOS */
1660 		nla_total_size(1) +
1661 		/* IFLA_IPTUN_PMTUDISC */
1662 		nla_total_size(1) +
1663 		/* IFLA_IPTUN_FLAGS */
1664 		nla_total_size(2) +
1665 		/* IFLA_IPTUN_PROTO */
1666 		nla_total_size(1) +
1667 #ifdef CONFIG_IPV6_SIT_6RD
1668 		/* IFLA_IPTUN_6RD_PREFIX */
1669 		nla_total_size(sizeof(struct in6_addr)) +
1670 		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1671 		nla_total_size(4) +
1672 		/* IFLA_IPTUN_6RD_PREFIXLEN */
1673 		nla_total_size(2) +
1674 		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1675 		nla_total_size(2) +
1676 #endif
1677 		/* IFLA_IPTUN_ENCAP_TYPE */
1678 		nla_total_size(2) +
1679 		/* IFLA_IPTUN_ENCAP_FLAGS */
1680 		nla_total_size(2) +
1681 		/* IFLA_IPTUN_ENCAP_SPORT */
1682 		nla_total_size(2) +
1683 		/* IFLA_IPTUN_ENCAP_DPORT */
1684 		nla_total_size(2) +
1685 		/* IFLA_IPTUN_FWMARK */
1686 		nla_total_size(4) +
1687 		0;
1688 }
1689 
1690 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1691 {
1692 	struct ip_tunnel *tunnel = netdev_priv(dev);
1693 	struct ip_tunnel_parm_kern *parm = &tunnel->parms;
1694 
1695 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1696 	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1697 	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1698 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1699 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1700 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1701 		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1702 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1703 	    nla_put_be16(skb, IFLA_IPTUN_FLAGS,
1704 			 ip_tunnel_flags_to_be16(parm->i_flags)) ||
1705 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1706 		goto nla_put_failure;
1707 
1708 #ifdef CONFIG_IPV6_SIT_6RD
1709 	if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1710 			     &tunnel->ip6rd.prefix) ||
1711 	    nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1712 			    tunnel->ip6rd.relay_prefix) ||
1713 	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1714 			tunnel->ip6rd.prefixlen) ||
1715 	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1716 			tunnel->ip6rd.relay_prefixlen))
1717 		goto nla_put_failure;
1718 #endif
1719 
1720 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1721 			tunnel->encap.type) ||
1722 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1723 			tunnel->encap.sport) ||
1724 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1725 			tunnel->encap.dport) ||
1726 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1727 			tunnel->encap.flags))
1728 		goto nla_put_failure;
1729 
1730 	return 0;
1731 
1732 nla_put_failure:
1733 	return -EMSGSIZE;
1734 }
1735 
1736 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1737 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1738 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1739 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1740 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1741 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1742 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1743 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1744 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1745 #ifdef CONFIG_IPV6_SIT_6RD
1746 	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1747 	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1748 	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1749 	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1750 #endif
1751 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
1752 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
1753 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
1754 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
1755 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
1756 };
1757 
1758 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1759 {
1760 	struct net *net = dev_net(dev);
1761 	struct sit_net *sitn = net_generic(net, sit_net_id);
1762 
1763 	if (dev != sitn->fb_tunnel_dev)
1764 		unregister_netdevice_queue(dev, head);
1765 }
1766 
1767 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1768 	.kind		= "sit",
1769 	.maxtype	= IFLA_IPTUN_MAX,
1770 	.policy		= ipip6_policy,
1771 	.priv_size	= sizeof(struct ip_tunnel),
1772 	.setup		= ipip6_tunnel_setup,
1773 	.validate	= ipip6_validate,
1774 	.newlink	= ipip6_newlink,
1775 	.changelink	= ipip6_changelink,
1776 	.get_size	= ipip6_get_size,
1777 	.fill_info	= ipip6_fill_info,
1778 	.dellink	= ipip6_dellink,
1779 	.get_link_net	= ip_tunnel_get_link_net,
1780 };
1781 
1782 static struct xfrm_tunnel sit_handler __read_mostly = {
1783 	.handler	=	ipip6_rcv,
1784 	.err_handler	=	ipip6_err,
1785 	.priority	=	1,
1786 };
1787 
1788 static struct xfrm_tunnel ipip_handler __read_mostly = {
1789 	.handler	=	ipip_rcv,
1790 	.err_handler	=	ipip6_err,
1791 	.priority	=	2,
1792 };
1793 
1794 #if IS_ENABLED(CONFIG_MPLS)
1795 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1796 	.handler	=	mplsip_rcv,
1797 	.err_handler	=	ipip6_err,
1798 	.priority	=	2,
1799 };
1800 #endif
1801 
1802 static void __net_exit sit_destroy_tunnels(struct net *net,
1803 					   struct list_head *head)
1804 {
1805 	struct sit_net *sitn = net_generic(net, sit_net_id);
1806 	struct net_device *dev, *aux;
1807 	int prio;
1808 
1809 	for_each_netdev_safe(net, dev, aux)
1810 		if (dev->rtnl_link_ops == &sit_link_ops)
1811 			unregister_netdevice_queue(dev, head);
1812 
1813 	for (prio = 0; prio < 4; prio++) {
1814 		int h;
1815 		for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1816 			struct ip_tunnel *t;
1817 
1818 			t = rtnl_dereference(sitn->tunnels[prio][h]);
1819 			while (t) {
1820 				/* If dev is in the same netns, it has already
1821 				 * been added to the list by the previous loop.
1822 				 */
1823 				if (!net_eq(dev_net(t->dev), net))
1824 					unregister_netdevice_queue(t->dev,
1825 								   head);
1826 				t = rtnl_dereference(t->next);
1827 			}
1828 		}
1829 	}
1830 }
1831 
1832 static int __net_init sit_init_net(struct net *net)
1833 {
1834 	struct sit_net *sitn = net_generic(net, sit_net_id);
1835 	struct ip_tunnel *t;
1836 	int err;
1837 
1838 	sitn->tunnels[0] = sitn->tunnels_wc;
1839 	sitn->tunnels[1] = sitn->tunnels_l;
1840 	sitn->tunnels[2] = sitn->tunnels_r;
1841 	sitn->tunnels[3] = sitn->tunnels_r_l;
1842 
1843 	if (!net_has_fallback_tunnels(net))
1844 		return 0;
1845 
1846 	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1847 					   NET_NAME_UNKNOWN,
1848 					   ipip6_tunnel_setup);
1849 	if (!sitn->fb_tunnel_dev) {
1850 		err = -ENOMEM;
1851 		goto err_alloc_dev;
1852 	}
1853 	dev_net_set(sitn->fb_tunnel_dev, net);
1854 	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1855 	/* FB netdevice is special: we have one, and only one per netns.
1856 	 * Allowing to move it to another netns is clearly unsafe.
1857 	 */
1858 	sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1859 
1860 	err = register_netdev(sitn->fb_tunnel_dev);
1861 	if (err)
1862 		goto err_reg_dev;
1863 
1864 	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1865 	ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1866 
1867 	t = netdev_priv(sitn->fb_tunnel_dev);
1868 
1869 	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1870 	return 0;
1871 
1872 err_reg_dev:
1873 	free_netdev(sitn->fb_tunnel_dev);
1874 err_alloc_dev:
1875 	return err;
1876 }
1877 
1878 static void __net_exit sit_exit_batch_rtnl(struct list_head *net_list,
1879 					   struct list_head *dev_to_kill)
1880 {
1881 	struct net *net;
1882 
1883 	ASSERT_RTNL();
1884 	list_for_each_entry(net, net_list, exit_list)
1885 		sit_destroy_tunnels(net, dev_to_kill);
1886 }
1887 
1888 static struct pernet_operations sit_net_ops = {
1889 	.init = sit_init_net,
1890 	.exit_batch_rtnl = sit_exit_batch_rtnl,
1891 	.id   = &sit_net_id,
1892 	.size = sizeof(struct sit_net),
1893 };
1894 
1895 static void __exit sit_cleanup(void)
1896 {
1897 	rtnl_link_unregister(&sit_link_ops);
1898 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1899 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1900 #if IS_ENABLED(CONFIG_MPLS)
1901 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1902 #endif
1903 
1904 	unregister_pernet_device(&sit_net_ops);
1905 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1906 }
1907 
1908 static int __init sit_init(void)
1909 {
1910 	int err;
1911 
1912 	pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1913 
1914 	err = register_pernet_device(&sit_net_ops);
1915 	if (err < 0)
1916 		return err;
1917 	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1918 	if (err < 0) {
1919 		pr_info("%s: can't register ip6ip4\n", __func__);
1920 		goto xfrm_tunnel_failed;
1921 	}
1922 	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1923 	if (err < 0) {
1924 		pr_info("%s: can't register ip4ip4\n", __func__);
1925 		goto xfrm_tunnel4_failed;
1926 	}
1927 #if IS_ENABLED(CONFIG_MPLS)
1928 	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1929 	if (err < 0) {
1930 		pr_info("%s: can't register mplsip\n", __func__);
1931 		goto xfrm_tunnel_mpls_failed;
1932 	}
1933 #endif
1934 	err = rtnl_link_register(&sit_link_ops);
1935 	if (err < 0)
1936 		goto rtnl_link_failed;
1937 
1938 out:
1939 	return err;
1940 
1941 rtnl_link_failed:
1942 #if IS_ENABLED(CONFIG_MPLS)
1943 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1944 xfrm_tunnel_mpls_failed:
1945 #endif
1946 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1947 xfrm_tunnel4_failed:
1948 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1949 xfrm_tunnel_failed:
1950 	unregister_pernet_device(&sit_net_ops);
1951 	goto out;
1952 }
1953 
1954 module_init(sit_init);
1955 module_exit(sit_cleanup);
1956 MODULE_DESCRIPTION("IPv6-in-IPv4 tunnel SIT driver");
1957 MODULE_LICENSE("GPL");
1958 MODULE_ALIAS_RTNL_LINK("sit");
1959 MODULE_ALIAS_NETDEV("sit0");
1960