xref: /openbmc/linux/net/ipv6/sit.c (revision 3213486f)
1 /*
2  *	IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
8  *
9  *	This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  *
14  *	Changes:
15  * Roger Venning <r.venning@telstra.com>:	6to4 support
16  * Nate Thompson <nate@thebog.net>:		6to4 support
17  * Fred Templin <fred.l.templin@boeing.com>:	isatap support
18  */
19 
20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
21 
22 #include <linux/module.h>
23 #include <linux/capability.h>
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/icmp.h>
33 #include <linux/slab.h>
34 #include <linux/uaccess.h>
35 #include <linux/init.h>
36 #include <linux/netfilter_ipv4.h>
37 #include <linux/if_ether.h>
38 
39 #include <net/sock.h>
40 #include <net/snmp.h>
41 
42 #include <net/ipv6.h>
43 #include <net/protocol.h>
44 #include <net/transp_v6.h>
45 #include <net/ip6_fib.h>
46 #include <net/ip6_route.h>
47 #include <net/ndisc.h>
48 #include <net/addrconf.h>
49 #include <net/ip.h>
50 #include <net/udp.h>
51 #include <net/icmp.h>
52 #include <net/ip_tunnels.h>
53 #include <net/inet_ecn.h>
54 #include <net/xfrm.h>
55 #include <net/dsfield.h>
56 #include <net/net_namespace.h>
57 #include <net/netns/generic.h>
58 
59 /*
60    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
61 
62    For comments look at net/ipv4/ip_gre.c --ANK
63  */
64 
65 #define IP6_SIT_HASH_SIZE  16
66 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
67 
68 static bool log_ecn_error = true;
69 module_param(log_ecn_error, bool, 0644);
70 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
71 
72 static int ipip6_tunnel_init(struct net_device *dev);
73 static void ipip6_tunnel_setup(struct net_device *dev);
74 static void ipip6_dev_free(struct net_device *dev);
75 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
76 		      __be32 *v4dst);
77 static struct rtnl_link_ops sit_link_ops __read_mostly;
78 
79 static unsigned int sit_net_id __read_mostly;
80 struct sit_net {
81 	struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
82 	struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
83 	struct ip_tunnel __rcu *tunnels_l[IP6_SIT_HASH_SIZE];
84 	struct ip_tunnel __rcu *tunnels_wc[1];
85 	struct ip_tunnel __rcu **tunnels[4];
86 
87 	struct net_device *fb_tunnel_dev;
88 };
89 
90 /*
91  * Must be invoked with rcu_read_lock
92  */
93 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
94 					     struct net_device *dev,
95 					     __be32 remote, __be32 local,
96 					     int sifindex)
97 {
98 	unsigned int h0 = HASH(remote);
99 	unsigned int h1 = HASH(local);
100 	struct ip_tunnel *t;
101 	struct sit_net *sitn = net_generic(net, sit_net_id);
102 	int ifindex = dev ? dev->ifindex : 0;
103 
104 	for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
105 		if (local == t->parms.iph.saddr &&
106 		    remote == t->parms.iph.daddr &&
107 		    (!dev || !t->parms.link || ifindex == t->parms.link ||
108 		     sifindex == t->parms.link) &&
109 		    (t->dev->flags & IFF_UP))
110 			return t;
111 	}
112 	for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
113 		if (remote == t->parms.iph.daddr &&
114 		    (!dev || !t->parms.link || ifindex == t->parms.link ||
115 		     sifindex == t->parms.link) &&
116 		    (t->dev->flags & IFF_UP))
117 			return t;
118 	}
119 	for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
120 		if (local == t->parms.iph.saddr &&
121 		    (!dev || !t->parms.link || ifindex == t->parms.link ||
122 		     sifindex == t->parms.link) &&
123 		    (t->dev->flags & IFF_UP))
124 			return t;
125 	}
126 	t = rcu_dereference(sitn->tunnels_wc[0]);
127 	if (t && (t->dev->flags & IFF_UP))
128 		return t;
129 	return NULL;
130 }
131 
132 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
133 		struct ip_tunnel_parm *parms)
134 {
135 	__be32 remote = parms->iph.daddr;
136 	__be32 local = parms->iph.saddr;
137 	unsigned int h = 0;
138 	int prio = 0;
139 
140 	if (remote) {
141 		prio |= 2;
142 		h ^= HASH(remote);
143 	}
144 	if (local) {
145 		prio |= 1;
146 		h ^= HASH(local);
147 	}
148 	return &sitn->tunnels[prio][h];
149 }
150 
151 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
152 		struct ip_tunnel *t)
153 {
154 	return __ipip6_bucket(sitn, &t->parms);
155 }
156 
157 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
158 {
159 	struct ip_tunnel __rcu **tp;
160 	struct ip_tunnel *iter;
161 
162 	for (tp = ipip6_bucket(sitn, t);
163 	     (iter = rtnl_dereference(*tp)) != NULL;
164 	     tp = &iter->next) {
165 		if (t == iter) {
166 			rcu_assign_pointer(*tp, t->next);
167 			break;
168 		}
169 	}
170 }
171 
172 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
173 {
174 	struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
175 
176 	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
177 	rcu_assign_pointer(*tp, t);
178 }
179 
180 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
181 {
182 #ifdef CONFIG_IPV6_SIT_6RD
183 	struct ip_tunnel *t = netdev_priv(dev);
184 
185 	if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
186 		ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
187 		t->ip6rd.relay_prefix = 0;
188 		t->ip6rd.prefixlen = 16;
189 		t->ip6rd.relay_prefixlen = 0;
190 	} else {
191 		struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
192 		memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
193 	}
194 #endif
195 }
196 
197 static int ipip6_tunnel_create(struct net_device *dev)
198 {
199 	struct ip_tunnel *t = netdev_priv(dev);
200 	struct net *net = dev_net(dev);
201 	struct sit_net *sitn = net_generic(net, sit_net_id);
202 	int err;
203 
204 	memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
205 	memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
206 
207 	if ((__force u16)t->parms.i_flags & SIT_ISATAP)
208 		dev->priv_flags |= IFF_ISATAP;
209 
210 	dev->rtnl_link_ops = &sit_link_ops;
211 
212 	err = register_netdevice(dev);
213 	if (err < 0)
214 		goto out;
215 
216 	ipip6_tunnel_clone_6rd(dev, sitn);
217 
218 	dev_hold(dev);
219 
220 	ipip6_tunnel_link(sitn, t);
221 	return 0;
222 
223 out:
224 	return err;
225 }
226 
227 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
228 		struct ip_tunnel_parm *parms, int create)
229 {
230 	__be32 remote = parms->iph.daddr;
231 	__be32 local = parms->iph.saddr;
232 	struct ip_tunnel *t, *nt;
233 	struct ip_tunnel __rcu **tp;
234 	struct net_device *dev;
235 	char name[IFNAMSIZ];
236 	struct sit_net *sitn = net_generic(net, sit_net_id);
237 
238 	for (tp = __ipip6_bucket(sitn, parms);
239 	    (t = rtnl_dereference(*tp)) != NULL;
240 	     tp = &t->next) {
241 		if (local == t->parms.iph.saddr &&
242 		    remote == t->parms.iph.daddr &&
243 		    parms->link == t->parms.link) {
244 			if (create)
245 				return NULL;
246 			else
247 				return t;
248 		}
249 	}
250 	if (!create)
251 		goto failed;
252 
253 	if (parms->name[0]) {
254 		if (!dev_valid_name(parms->name))
255 			goto failed;
256 		strlcpy(name, parms->name, IFNAMSIZ);
257 	} else {
258 		strcpy(name, "sit%d");
259 	}
260 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
261 			   ipip6_tunnel_setup);
262 	if (!dev)
263 		return NULL;
264 
265 	dev_net_set(dev, net);
266 
267 	nt = netdev_priv(dev);
268 
269 	nt->parms = *parms;
270 	if (ipip6_tunnel_create(dev) < 0)
271 		goto failed_free;
272 
273 	return nt;
274 
275 failed_free:
276 	free_netdev(dev);
277 failed:
278 	return NULL;
279 }
280 
281 #define for_each_prl_rcu(start)			\
282 	for (prl = rcu_dereference(start);	\
283 	     prl;				\
284 	     prl = rcu_dereference(prl->next))
285 
286 static struct ip_tunnel_prl_entry *
287 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
288 {
289 	struct ip_tunnel_prl_entry *prl;
290 
291 	for_each_prl_rcu(t->prl)
292 		if (prl->addr == addr)
293 			break;
294 	return prl;
295 
296 }
297 
298 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
299 				struct ip_tunnel_prl __user *a)
300 {
301 	struct ip_tunnel_prl kprl, *kp;
302 	struct ip_tunnel_prl_entry *prl;
303 	unsigned int cmax, c = 0, ca, len;
304 	int ret = 0;
305 
306 	if (copy_from_user(&kprl, a, sizeof(kprl)))
307 		return -EFAULT;
308 	cmax = kprl.datalen / sizeof(kprl);
309 	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
310 		cmax = 1;
311 
312 	/* For simple GET or for root users,
313 	 * we try harder to allocate.
314 	 */
315 	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
316 		kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
317 		NULL;
318 
319 	rcu_read_lock();
320 
321 	ca = t->prl_count < cmax ? t->prl_count : cmax;
322 
323 	if (!kp) {
324 		/* We don't try hard to allocate much memory for
325 		 * non-root users.
326 		 * For root users, retry allocating enough memory for
327 		 * the answer.
328 		 */
329 		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
330 		if (!kp) {
331 			ret = -ENOMEM;
332 			goto out;
333 		}
334 	}
335 
336 	c = 0;
337 	for_each_prl_rcu(t->prl) {
338 		if (c >= cmax)
339 			break;
340 		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
341 			continue;
342 		kp[c].addr = prl->addr;
343 		kp[c].flags = prl->flags;
344 		c++;
345 		if (kprl.addr != htonl(INADDR_ANY))
346 			break;
347 	}
348 out:
349 	rcu_read_unlock();
350 
351 	len = sizeof(*kp) * c;
352 	ret = 0;
353 	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
354 		ret = -EFAULT;
355 
356 	kfree(kp);
357 
358 	return ret;
359 }
360 
361 static int
362 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
363 {
364 	struct ip_tunnel_prl_entry *p;
365 	int err = 0;
366 
367 	if (a->addr == htonl(INADDR_ANY))
368 		return -EINVAL;
369 
370 	ASSERT_RTNL();
371 
372 	for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
373 		if (p->addr == a->addr) {
374 			if (chg) {
375 				p->flags = a->flags;
376 				goto out;
377 			}
378 			err = -EEXIST;
379 			goto out;
380 		}
381 	}
382 
383 	if (chg) {
384 		err = -ENXIO;
385 		goto out;
386 	}
387 
388 	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
389 	if (!p) {
390 		err = -ENOBUFS;
391 		goto out;
392 	}
393 
394 	p->next = t->prl;
395 	p->addr = a->addr;
396 	p->flags = a->flags;
397 	t->prl_count++;
398 	rcu_assign_pointer(t->prl, p);
399 out:
400 	return err;
401 }
402 
403 static void prl_list_destroy_rcu(struct rcu_head *head)
404 {
405 	struct ip_tunnel_prl_entry *p, *n;
406 
407 	p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
408 	do {
409 		n = rcu_dereference_protected(p->next, 1);
410 		kfree(p);
411 		p = n;
412 	} while (p);
413 }
414 
415 static int
416 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
417 {
418 	struct ip_tunnel_prl_entry *x;
419 	struct ip_tunnel_prl_entry __rcu **p;
420 	int err = 0;
421 
422 	ASSERT_RTNL();
423 
424 	if (a && a->addr != htonl(INADDR_ANY)) {
425 		for (p = &t->prl;
426 		     (x = rtnl_dereference(*p)) != NULL;
427 		     p = &x->next) {
428 			if (x->addr == a->addr) {
429 				*p = x->next;
430 				kfree_rcu(x, rcu_head);
431 				t->prl_count--;
432 				goto out;
433 			}
434 		}
435 		err = -ENXIO;
436 	} else {
437 		x = rtnl_dereference(t->prl);
438 		if (x) {
439 			t->prl_count = 0;
440 			call_rcu(&x->rcu_head, prl_list_destroy_rcu);
441 			t->prl = NULL;
442 		}
443 	}
444 out:
445 	return err;
446 }
447 
448 static int
449 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
450 {
451 	struct ip_tunnel_prl_entry *p;
452 	int ok = 1;
453 
454 	rcu_read_lock();
455 	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
456 	if (p) {
457 		if (p->flags & PRL_DEFAULT)
458 			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
459 		else
460 			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
461 	} else {
462 		const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
463 
464 		if (ipv6_addr_is_isatap(addr6) &&
465 		    (addr6->s6_addr32[3] == iph->saddr) &&
466 		    ipv6_chk_prefix(addr6, t->dev))
467 			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
468 		else
469 			ok = 0;
470 	}
471 	rcu_read_unlock();
472 	return ok;
473 }
474 
475 static void ipip6_tunnel_uninit(struct net_device *dev)
476 {
477 	struct ip_tunnel *tunnel = netdev_priv(dev);
478 	struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
479 
480 	if (dev == sitn->fb_tunnel_dev) {
481 		RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
482 	} else {
483 		ipip6_tunnel_unlink(sitn, tunnel);
484 		ipip6_tunnel_del_prl(tunnel, NULL);
485 	}
486 	dst_cache_reset(&tunnel->dst_cache);
487 	dev_put(dev);
488 }
489 
490 static int ipip6_err(struct sk_buff *skb, u32 info)
491 {
492 	const struct iphdr *iph = (const struct iphdr *)skb->data;
493 	const int type = icmp_hdr(skb)->type;
494 	const int code = icmp_hdr(skb)->code;
495 	unsigned int data_len = 0;
496 	struct ip_tunnel *t;
497 	int sifindex;
498 	int err;
499 
500 	switch (type) {
501 	default:
502 	case ICMP_PARAMETERPROB:
503 		return 0;
504 
505 	case ICMP_DEST_UNREACH:
506 		switch (code) {
507 		case ICMP_SR_FAILED:
508 			/* Impossible event. */
509 			return 0;
510 		default:
511 			/* All others are translated to HOST_UNREACH.
512 			   rfc2003 contains "deep thoughts" about NET_UNREACH,
513 			   I believe they are just ether pollution. --ANK
514 			 */
515 			break;
516 		}
517 		break;
518 	case ICMP_TIME_EXCEEDED:
519 		if (code != ICMP_EXC_TTL)
520 			return 0;
521 		data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
522 		break;
523 	case ICMP_REDIRECT:
524 		break;
525 	}
526 
527 	err = -ENOENT;
528 
529 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
530 	t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
531 				iph->daddr, iph->saddr, sifindex);
532 	if (!t)
533 		goto out;
534 
535 	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
536 		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
537 				 t->parms.link, iph->protocol);
538 		err = 0;
539 		goto out;
540 	}
541 	if (type == ICMP_REDIRECT) {
542 		ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
543 			      iph->protocol);
544 		err = 0;
545 		goto out;
546 	}
547 
548 	err = 0;
549 	if (__in6_dev_get(skb->dev) &&
550 	    !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
551 		goto out;
552 
553 	if (t->parms.iph.daddr == 0)
554 		goto out;
555 
556 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
557 		goto out;
558 
559 	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
560 		t->err_count++;
561 	else
562 		t->err_count = 1;
563 	t->err_time = jiffies;
564 out:
565 	return err;
566 }
567 
568 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
569 				  const struct in6_addr *v6addr)
570 {
571 	__be32 v4embed = 0;
572 	if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
573 		return true;
574 	return false;
575 }
576 
577 /* Checks if an address matches an address on the tunnel interface.
578  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
579  * Long story:
580  * This function is called after we considered the packet as spoofed
581  * in is_spoofed_6rd.
582  * We may have a router that is doing NAT for proto 41 packets
583  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
584  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
585  * function will return true, dropping the packet.
586  * But, we can still check if is spoofed against the IP
587  * addresses associated with the interface.
588  */
589 static bool only_dnatted(const struct ip_tunnel *tunnel,
590 	const struct in6_addr *v6dst)
591 {
592 	int prefix_len;
593 
594 #ifdef CONFIG_IPV6_SIT_6RD
595 	prefix_len = tunnel->ip6rd.prefixlen + 32
596 		- tunnel->ip6rd.relay_prefixlen;
597 #else
598 	prefix_len = 48;
599 #endif
600 	return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
601 }
602 
603 /* Returns true if a packet is spoofed */
604 static bool packet_is_spoofed(struct sk_buff *skb,
605 			      const struct iphdr *iph,
606 			      struct ip_tunnel *tunnel)
607 {
608 	const struct ipv6hdr *ipv6h;
609 
610 	if (tunnel->dev->priv_flags & IFF_ISATAP) {
611 		if (!isatap_chksrc(skb, iph, tunnel))
612 			return true;
613 
614 		return false;
615 	}
616 
617 	if (tunnel->dev->flags & IFF_POINTOPOINT)
618 		return false;
619 
620 	ipv6h = ipv6_hdr(skb);
621 
622 	if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
623 		net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
624 				     &iph->saddr, &ipv6h->saddr,
625 				     &iph->daddr, &ipv6h->daddr);
626 		return true;
627 	}
628 
629 	if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
630 		return false;
631 
632 	if (only_dnatted(tunnel, &ipv6h->daddr))
633 		return false;
634 
635 	net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
636 			     &iph->saddr, &ipv6h->saddr,
637 			     &iph->daddr, &ipv6h->daddr);
638 	return true;
639 }
640 
641 static int ipip6_rcv(struct sk_buff *skb)
642 {
643 	const struct iphdr *iph = ip_hdr(skb);
644 	struct ip_tunnel *tunnel;
645 	int sifindex;
646 	int err;
647 
648 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
649 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
650 				     iph->saddr, iph->daddr, sifindex);
651 	if (tunnel) {
652 		struct pcpu_sw_netstats *tstats;
653 
654 		if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
655 		    tunnel->parms.iph.protocol != 0)
656 			goto out;
657 
658 		skb->mac_header = skb->network_header;
659 		skb_reset_network_header(skb);
660 		IPCB(skb)->flags = 0;
661 		skb->dev = tunnel->dev;
662 
663 		if (packet_is_spoofed(skb, iph, tunnel)) {
664 			tunnel->dev->stats.rx_errors++;
665 			goto out;
666 		}
667 
668 		if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
669 		    !net_eq(tunnel->net, dev_net(tunnel->dev))))
670 			goto out;
671 
672 		/* skb can be uncloned in iptunnel_pull_header, so
673 		 * old iph is no longer valid
674 		 */
675 		iph = (const struct iphdr *)skb_mac_header(skb);
676 		err = IP_ECN_decapsulate(iph, skb);
677 		if (unlikely(err)) {
678 			if (log_ecn_error)
679 				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
680 						     &iph->saddr, iph->tos);
681 			if (err > 1) {
682 				++tunnel->dev->stats.rx_frame_errors;
683 				++tunnel->dev->stats.rx_errors;
684 				goto out;
685 			}
686 		}
687 
688 		tstats = this_cpu_ptr(tunnel->dev->tstats);
689 		u64_stats_update_begin(&tstats->syncp);
690 		tstats->rx_packets++;
691 		tstats->rx_bytes += skb->len;
692 		u64_stats_update_end(&tstats->syncp);
693 
694 		netif_rx(skb);
695 
696 		return 0;
697 	}
698 
699 	/* no tunnel matched,  let upstream know, ipsec may handle it */
700 	return 1;
701 out:
702 	kfree_skb(skb);
703 	return 0;
704 }
705 
706 static const struct tnl_ptk_info ipip_tpi = {
707 	/* no tunnel info required for ipip. */
708 	.proto = htons(ETH_P_IP),
709 };
710 
711 #if IS_ENABLED(CONFIG_MPLS)
712 static const struct tnl_ptk_info mplsip_tpi = {
713 	/* no tunnel info required for mplsip. */
714 	.proto = htons(ETH_P_MPLS_UC),
715 };
716 #endif
717 
718 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
719 {
720 	const struct iphdr *iph;
721 	struct ip_tunnel *tunnel;
722 	int sifindex;
723 
724 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
725 
726 	iph = ip_hdr(skb);
727 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
728 				     iph->saddr, iph->daddr, sifindex);
729 	if (tunnel) {
730 		const struct tnl_ptk_info *tpi;
731 
732 		if (tunnel->parms.iph.protocol != ipproto &&
733 		    tunnel->parms.iph.protocol != 0)
734 			goto drop;
735 
736 		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
737 			goto drop;
738 #if IS_ENABLED(CONFIG_MPLS)
739 		if (ipproto == IPPROTO_MPLS)
740 			tpi = &mplsip_tpi;
741 		else
742 #endif
743 			tpi = &ipip_tpi;
744 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
745 			goto drop;
746 		return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
747 	}
748 
749 	return 1;
750 
751 drop:
752 	kfree_skb(skb);
753 	return 0;
754 }
755 
756 static int ipip_rcv(struct sk_buff *skb)
757 {
758 	return sit_tunnel_rcv(skb, IPPROTO_IPIP);
759 }
760 
761 #if IS_ENABLED(CONFIG_MPLS)
762 static int mplsip_rcv(struct sk_buff *skb)
763 {
764 	return sit_tunnel_rcv(skb, IPPROTO_MPLS);
765 }
766 #endif
767 
768 /*
769  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
770  * stores the embedded IPv4 address in v4dst and returns true.
771  */
772 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
773 		      __be32 *v4dst)
774 {
775 #ifdef CONFIG_IPV6_SIT_6RD
776 	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
777 			      tunnel->ip6rd.prefixlen)) {
778 		unsigned int pbw0, pbi0;
779 		int pbi1;
780 		u32 d;
781 
782 		pbw0 = tunnel->ip6rd.prefixlen >> 5;
783 		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
784 
785 		d = tunnel->ip6rd.relay_prefixlen < 32 ?
786 			(ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
787 		    tunnel->ip6rd.relay_prefixlen : 0;
788 
789 		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
790 		if (pbi1 > 0)
791 			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
792 			     (32 - pbi1);
793 
794 		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
795 		return true;
796 	}
797 #else
798 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
799 		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
800 		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
801 		return true;
802 	}
803 #endif
804 	return false;
805 }
806 
807 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
808 			     const struct in6_addr *v6dst)
809 {
810 	__be32 dst = 0;
811 	check_6rd(tunnel, v6dst, &dst);
812 	return dst;
813 }
814 
815 /*
816  *	This function assumes it is being called from dev_queue_xmit()
817  *	and that skb is filled properly by that function.
818  */
819 
820 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
821 				     struct net_device *dev)
822 {
823 	struct ip_tunnel *tunnel = netdev_priv(dev);
824 	const struct iphdr  *tiph = &tunnel->parms.iph;
825 	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
826 	u8     tos = tunnel->parms.iph.tos;
827 	__be16 df = tiph->frag_off;
828 	struct rtable *rt;		/* Route to the other host */
829 	struct net_device *tdev;	/* Device to other host */
830 	unsigned int max_headroom;	/* The extra header space needed */
831 	__be32 dst = tiph->daddr;
832 	struct flowi4 fl4;
833 	int    mtu;
834 	const struct in6_addr *addr6;
835 	int addr_type;
836 	u8 ttl;
837 	u8 protocol = IPPROTO_IPV6;
838 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
839 
840 	if (tos == 1)
841 		tos = ipv6_get_dsfield(iph6);
842 
843 	/* ISATAP (RFC4214) - must come before 6to4 */
844 	if (dev->priv_flags & IFF_ISATAP) {
845 		struct neighbour *neigh = NULL;
846 		bool do_tx_error = false;
847 
848 		if (skb_dst(skb))
849 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
850 
851 		if (!neigh) {
852 			net_dbg_ratelimited("nexthop == NULL\n");
853 			goto tx_error;
854 		}
855 
856 		addr6 = (const struct in6_addr *)&neigh->primary_key;
857 		addr_type = ipv6_addr_type(addr6);
858 
859 		if ((addr_type & IPV6_ADDR_UNICAST) &&
860 		     ipv6_addr_is_isatap(addr6))
861 			dst = addr6->s6_addr32[3];
862 		else
863 			do_tx_error = true;
864 
865 		neigh_release(neigh);
866 		if (do_tx_error)
867 			goto tx_error;
868 	}
869 
870 	if (!dst)
871 		dst = try_6rd(tunnel, &iph6->daddr);
872 
873 	if (!dst) {
874 		struct neighbour *neigh = NULL;
875 		bool do_tx_error = false;
876 
877 		if (skb_dst(skb))
878 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
879 
880 		if (!neigh) {
881 			net_dbg_ratelimited("nexthop == NULL\n");
882 			goto tx_error;
883 		}
884 
885 		addr6 = (const struct in6_addr *)&neigh->primary_key;
886 		addr_type = ipv6_addr_type(addr6);
887 
888 		if (addr_type == IPV6_ADDR_ANY) {
889 			addr6 = &ipv6_hdr(skb)->daddr;
890 			addr_type = ipv6_addr_type(addr6);
891 		}
892 
893 		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
894 			dst = addr6->s6_addr32[3];
895 		else
896 			do_tx_error = true;
897 
898 		neigh_release(neigh);
899 		if (do_tx_error)
900 			goto tx_error;
901 	}
902 
903 	flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
904 			   RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
905 			   0, dst, tiph->saddr, 0, 0,
906 			   sock_net_uid(tunnel->net, NULL));
907 	rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
908 
909 	if (IS_ERR(rt)) {
910 		dev->stats.tx_carrier_errors++;
911 		goto tx_error_icmp;
912 	}
913 	if (rt->rt_type != RTN_UNICAST) {
914 		ip_rt_put(rt);
915 		dev->stats.tx_carrier_errors++;
916 		goto tx_error_icmp;
917 	}
918 	tdev = rt->dst.dev;
919 
920 	if (tdev == dev) {
921 		ip_rt_put(rt);
922 		dev->stats.collisions++;
923 		goto tx_error;
924 	}
925 
926 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
927 		ip_rt_put(rt);
928 		goto tx_error;
929 	}
930 
931 	if (df) {
932 		mtu = dst_mtu(&rt->dst) - t_hlen;
933 
934 		if (mtu < 68) {
935 			dev->stats.collisions++;
936 			ip_rt_put(rt);
937 			goto tx_error;
938 		}
939 
940 		if (mtu < IPV6_MIN_MTU) {
941 			mtu = IPV6_MIN_MTU;
942 			df = 0;
943 		}
944 
945 		if (tunnel->parms.iph.daddr)
946 			skb_dst_update_pmtu(skb, mtu);
947 
948 		if (skb->len > mtu && !skb_is_gso(skb)) {
949 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
950 			ip_rt_put(rt);
951 			goto tx_error;
952 		}
953 	}
954 
955 	if (tunnel->err_count > 0) {
956 		if (time_before(jiffies,
957 				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
958 			tunnel->err_count--;
959 			dst_link_failure(skb);
960 		} else
961 			tunnel->err_count = 0;
962 	}
963 
964 	/*
965 	 * Okay, now see if we can stuff it in the buffer as-is.
966 	 */
967 	max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
968 
969 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
970 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
971 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
972 		if (!new_skb) {
973 			ip_rt_put(rt);
974 			dev->stats.tx_dropped++;
975 			kfree_skb(skb);
976 			return NETDEV_TX_OK;
977 		}
978 		if (skb->sk)
979 			skb_set_owner_w(new_skb, skb->sk);
980 		dev_kfree_skb(skb);
981 		skb = new_skb;
982 		iph6 = ipv6_hdr(skb);
983 	}
984 	ttl = tiph->ttl;
985 	if (ttl == 0)
986 		ttl = iph6->hop_limit;
987 	tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
988 
989 	if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
990 		ip_rt_put(rt);
991 		goto tx_error;
992 	}
993 
994 	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
995 
996 	iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
997 		      df, !net_eq(tunnel->net, dev_net(dev)));
998 	return NETDEV_TX_OK;
999 
1000 tx_error_icmp:
1001 	dst_link_failure(skb);
1002 tx_error:
1003 	kfree_skb(skb);
1004 	dev->stats.tx_errors++;
1005 	return NETDEV_TX_OK;
1006 }
1007 
1008 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1009 				     struct net_device *dev, u8 ipproto)
1010 {
1011 	struct ip_tunnel *tunnel = netdev_priv(dev);
1012 	const struct iphdr  *tiph = &tunnel->parms.iph;
1013 
1014 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1015 		goto tx_error;
1016 
1017 	skb_set_inner_ipproto(skb, ipproto);
1018 
1019 	ip_tunnel_xmit(skb, dev, tiph, ipproto);
1020 	return NETDEV_TX_OK;
1021 tx_error:
1022 	kfree_skb(skb);
1023 	dev->stats.tx_errors++;
1024 	return NETDEV_TX_OK;
1025 }
1026 
1027 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1028 				   struct net_device *dev)
1029 {
1030 	if (!pskb_inet_may_pull(skb))
1031 		goto tx_err;
1032 
1033 	switch (skb->protocol) {
1034 	case htons(ETH_P_IP):
1035 		sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1036 		break;
1037 	case htons(ETH_P_IPV6):
1038 		ipip6_tunnel_xmit(skb, dev);
1039 		break;
1040 #if IS_ENABLED(CONFIG_MPLS)
1041 	case htons(ETH_P_MPLS_UC):
1042 		sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1043 		break;
1044 #endif
1045 	default:
1046 		goto tx_err;
1047 	}
1048 
1049 	return NETDEV_TX_OK;
1050 
1051 tx_err:
1052 	dev->stats.tx_errors++;
1053 	kfree_skb(skb);
1054 	return NETDEV_TX_OK;
1055 
1056 }
1057 
1058 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1059 {
1060 	struct net_device *tdev = NULL;
1061 	struct ip_tunnel *tunnel;
1062 	const struct iphdr *iph;
1063 	struct flowi4 fl4;
1064 
1065 	tunnel = netdev_priv(dev);
1066 	iph = &tunnel->parms.iph;
1067 
1068 	if (iph->daddr) {
1069 		struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1070 							  NULL,
1071 							  iph->daddr, iph->saddr,
1072 							  0, 0,
1073 							  IPPROTO_IPV6,
1074 							  RT_TOS(iph->tos),
1075 							  tunnel->parms.link);
1076 
1077 		if (!IS_ERR(rt)) {
1078 			tdev = rt->dst.dev;
1079 			ip_rt_put(rt);
1080 		}
1081 		dev->flags |= IFF_POINTOPOINT;
1082 	}
1083 
1084 	if (!tdev && tunnel->parms.link)
1085 		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1086 
1087 	if (tdev) {
1088 		int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1089 
1090 		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1091 		dev->mtu = tdev->mtu - t_hlen;
1092 		if (dev->mtu < IPV6_MIN_MTU)
1093 			dev->mtu = IPV6_MIN_MTU;
1094 	}
1095 }
1096 
1097 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1098 				__u32 fwmark)
1099 {
1100 	struct net *net = t->net;
1101 	struct sit_net *sitn = net_generic(net, sit_net_id);
1102 
1103 	ipip6_tunnel_unlink(sitn, t);
1104 	synchronize_net();
1105 	t->parms.iph.saddr = p->iph.saddr;
1106 	t->parms.iph.daddr = p->iph.daddr;
1107 	memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1108 	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1109 	ipip6_tunnel_link(sitn, t);
1110 	t->parms.iph.ttl = p->iph.ttl;
1111 	t->parms.iph.tos = p->iph.tos;
1112 	t->parms.iph.frag_off = p->iph.frag_off;
1113 	if (t->parms.link != p->link || t->fwmark != fwmark) {
1114 		t->parms.link = p->link;
1115 		t->fwmark = fwmark;
1116 		ipip6_tunnel_bind_dev(t->dev);
1117 	}
1118 	dst_cache_reset(&t->dst_cache);
1119 	netdev_state_change(t->dev);
1120 }
1121 
1122 #ifdef CONFIG_IPV6_SIT_6RD
1123 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1124 				   struct ip_tunnel_6rd *ip6rd)
1125 {
1126 	struct in6_addr prefix;
1127 	__be32 relay_prefix;
1128 
1129 	if (ip6rd->relay_prefixlen > 32 ||
1130 	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1131 		return -EINVAL;
1132 
1133 	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1134 	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1135 		return -EINVAL;
1136 	if (ip6rd->relay_prefixlen)
1137 		relay_prefix = ip6rd->relay_prefix &
1138 			       htonl(0xffffffffUL <<
1139 				     (32 - ip6rd->relay_prefixlen));
1140 	else
1141 		relay_prefix = 0;
1142 	if (relay_prefix != ip6rd->relay_prefix)
1143 		return -EINVAL;
1144 
1145 	t->ip6rd.prefix = prefix;
1146 	t->ip6rd.relay_prefix = relay_prefix;
1147 	t->ip6rd.prefixlen = ip6rd->prefixlen;
1148 	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1149 	dst_cache_reset(&t->dst_cache);
1150 	netdev_state_change(t->dev);
1151 	return 0;
1152 }
1153 #endif
1154 
1155 static bool ipip6_valid_ip_proto(u8 ipproto)
1156 {
1157 	return ipproto == IPPROTO_IPV6 ||
1158 		ipproto == IPPROTO_IPIP ||
1159 #if IS_ENABLED(CONFIG_MPLS)
1160 		ipproto == IPPROTO_MPLS ||
1161 #endif
1162 		ipproto == 0;
1163 }
1164 
1165 static int
1166 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1167 {
1168 	int err = 0;
1169 	struct ip_tunnel_parm p;
1170 	struct ip_tunnel_prl prl;
1171 	struct ip_tunnel *t = netdev_priv(dev);
1172 	struct net *net = t->net;
1173 	struct sit_net *sitn = net_generic(net, sit_net_id);
1174 #ifdef CONFIG_IPV6_SIT_6RD
1175 	struct ip_tunnel_6rd ip6rd;
1176 #endif
1177 
1178 	switch (cmd) {
1179 	case SIOCGETTUNNEL:
1180 #ifdef CONFIG_IPV6_SIT_6RD
1181 	case SIOCGET6RD:
1182 #endif
1183 		if (dev == sitn->fb_tunnel_dev) {
1184 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1185 				err = -EFAULT;
1186 				break;
1187 			}
1188 			t = ipip6_tunnel_locate(net, &p, 0);
1189 			if (!t)
1190 				t = netdev_priv(dev);
1191 		}
1192 
1193 		err = -EFAULT;
1194 		if (cmd == SIOCGETTUNNEL) {
1195 			memcpy(&p, &t->parms, sizeof(p));
1196 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1197 					 sizeof(p)))
1198 				goto done;
1199 #ifdef CONFIG_IPV6_SIT_6RD
1200 		} else {
1201 			ip6rd.prefix = t->ip6rd.prefix;
1202 			ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1203 			ip6rd.prefixlen = t->ip6rd.prefixlen;
1204 			ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1205 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1206 					 sizeof(ip6rd)))
1207 				goto done;
1208 #endif
1209 		}
1210 		err = 0;
1211 		break;
1212 
1213 	case SIOCADDTUNNEL:
1214 	case SIOCCHGTUNNEL:
1215 		err = -EPERM;
1216 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1217 			goto done;
1218 
1219 		err = -EFAULT;
1220 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1221 			goto done;
1222 
1223 		err = -EINVAL;
1224 		if (!ipip6_valid_ip_proto(p.iph.protocol))
1225 			goto done;
1226 		if (p.iph.version != 4 ||
1227 		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1228 			goto done;
1229 		if (p.iph.ttl)
1230 			p.iph.frag_off |= htons(IP_DF);
1231 
1232 		t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1233 
1234 		if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1235 			if (t) {
1236 				if (t->dev != dev) {
1237 					err = -EEXIST;
1238 					break;
1239 				}
1240 			} else {
1241 				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1242 				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1243 					err = -EINVAL;
1244 					break;
1245 				}
1246 				t = netdev_priv(dev);
1247 			}
1248 
1249 			ipip6_tunnel_update(t, &p, t->fwmark);
1250 		}
1251 
1252 		if (t) {
1253 			err = 0;
1254 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1255 				err = -EFAULT;
1256 		} else
1257 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1258 		break;
1259 
1260 	case SIOCDELTUNNEL:
1261 		err = -EPERM;
1262 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1263 			goto done;
1264 
1265 		if (dev == sitn->fb_tunnel_dev) {
1266 			err = -EFAULT;
1267 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1268 				goto done;
1269 			err = -ENOENT;
1270 			t = ipip6_tunnel_locate(net, &p, 0);
1271 			if (!t)
1272 				goto done;
1273 			err = -EPERM;
1274 			if (t == netdev_priv(sitn->fb_tunnel_dev))
1275 				goto done;
1276 			dev = t->dev;
1277 		}
1278 		unregister_netdevice(dev);
1279 		err = 0;
1280 		break;
1281 
1282 	case SIOCGETPRL:
1283 		err = -EINVAL;
1284 		if (dev == sitn->fb_tunnel_dev)
1285 			goto done;
1286 		err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1287 		break;
1288 
1289 	case SIOCADDPRL:
1290 	case SIOCDELPRL:
1291 	case SIOCCHGPRL:
1292 		err = -EPERM;
1293 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1294 			goto done;
1295 		err = -EINVAL;
1296 		if (dev == sitn->fb_tunnel_dev)
1297 			goto done;
1298 		err = -EFAULT;
1299 		if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1300 			goto done;
1301 
1302 		switch (cmd) {
1303 		case SIOCDELPRL:
1304 			err = ipip6_tunnel_del_prl(t, &prl);
1305 			break;
1306 		case SIOCADDPRL:
1307 		case SIOCCHGPRL:
1308 			err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1309 			break;
1310 		}
1311 		dst_cache_reset(&t->dst_cache);
1312 		netdev_state_change(dev);
1313 		break;
1314 
1315 #ifdef CONFIG_IPV6_SIT_6RD
1316 	case SIOCADD6RD:
1317 	case SIOCCHG6RD:
1318 	case SIOCDEL6RD:
1319 		err = -EPERM;
1320 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1321 			goto done;
1322 
1323 		err = -EFAULT;
1324 		if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1325 				   sizeof(ip6rd)))
1326 			goto done;
1327 
1328 		if (cmd != SIOCDEL6RD) {
1329 			err = ipip6_tunnel_update_6rd(t, &ip6rd);
1330 			if (err < 0)
1331 				goto done;
1332 		} else
1333 			ipip6_tunnel_clone_6rd(dev, sitn);
1334 
1335 		err = 0;
1336 		break;
1337 #endif
1338 
1339 	default:
1340 		err = -EINVAL;
1341 	}
1342 
1343 done:
1344 	return err;
1345 }
1346 
1347 static const struct net_device_ops ipip6_netdev_ops = {
1348 	.ndo_init	= ipip6_tunnel_init,
1349 	.ndo_uninit	= ipip6_tunnel_uninit,
1350 	.ndo_start_xmit	= sit_tunnel_xmit,
1351 	.ndo_do_ioctl	= ipip6_tunnel_ioctl,
1352 	.ndo_get_stats64 = ip_tunnel_get_stats64,
1353 	.ndo_get_iflink = ip_tunnel_get_iflink,
1354 };
1355 
1356 static void ipip6_dev_free(struct net_device *dev)
1357 {
1358 	struct ip_tunnel *tunnel = netdev_priv(dev);
1359 
1360 	dst_cache_destroy(&tunnel->dst_cache);
1361 	free_percpu(dev->tstats);
1362 }
1363 
1364 #define SIT_FEATURES (NETIF_F_SG	   | \
1365 		      NETIF_F_FRAGLIST	   | \
1366 		      NETIF_F_HIGHDMA	   | \
1367 		      NETIF_F_GSO_SOFTWARE | \
1368 		      NETIF_F_HW_CSUM)
1369 
1370 static void ipip6_tunnel_setup(struct net_device *dev)
1371 {
1372 	struct ip_tunnel *tunnel = netdev_priv(dev);
1373 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1374 
1375 	dev->netdev_ops		= &ipip6_netdev_ops;
1376 	dev->needs_free_netdev	= true;
1377 	dev->priv_destructor	= ipip6_dev_free;
1378 
1379 	dev->type		= ARPHRD_SIT;
1380 	dev->hard_header_len	= LL_MAX_HEADER + t_hlen;
1381 	dev->mtu		= ETH_DATA_LEN - t_hlen;
1382 	dev->min_mtu		= IPV6_MIN_MTU;
1383 	dev->max_mtu		= IP6_MAX_MTU - t_hlen;
1384 	dev->flags		= IFF_NOARP;
1385 	netif_keep_dst(dev);
1386 	dev->addr_len		= 4;
1387 	dev->features		|= NETIF_F_LLTX;
1388 	dev->features		|= SIT_FEATURES;
1389 	dev->hw_features	|= SIT_FEATURES;
1390 }
1391 
1392 static int ipip6_tunnel_init(struct net_device *dev)
1393 {
1394 	struct ip_tunnel *tunnel = netdev_priv(dev);
1395 	int err;
1396 
1397 	tunnel->dev = dev;
1398 	tunnel->net = dev_net(dev);
1399 	strcpy(tunnel->parms.name, dev->name);
1400 
1401 	ipip6_tunnel_bind_dev(dev);
1402 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1403 	if (!dev->tstats)
1404 		return -ENOMEM;
1405 
1406 	err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1407 	if (err) {
1408 		free_percpu(dev->tstats);
1409 		dev->tstats = NULL;
1410 		return err;
1411 	}
1412 
1413 	return 0;
1414 }
1415 
1416 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1417 {
1418 	struct ip_tunnel *tunnel = netdev_priv(dev);
1419 	struct iphdr *iph = &tunnel->parms.iph;
1420 	struct net *net = dev_net(dev);
1421 	struct sit_net *sitn = net_generic(net, sit_net_id);
1422 
1423 	iph->version		= 4;
1424 	iph->protocol		= IPPROTO_IPV6;
1425 	iph->ihl		= 5;
1426 	iph->ttl		= 64;
1427 
1428 	dev_hold(dev);
1429 	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1430 }
1431 
1432 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1433 			  struct netlink_ext_ack *extack)
1434 {
1435 	u8 proto;
1436 
1437 	if (!data || !data[IFLA_IPTUN_PROTO])
1438 		return 0;
1439 
1440 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1441 	if (!ipip6_valid_ip_proto(proto))
1442 		return -EINVAL;
1443 
1444 	return 0;
1445 }
1446 
1447 static void ipip6_netlink_parms(struct nlattr *data[],
1448 				struct ip_tunnel_parm *parms,
1449 				__u32 *fwmark)
1450 {
1451 	memset(parms, 0, sizeof(*parms));
1452 
1453 	parms->iph.version = 4;
1454 	parms->iph.protocol = IPPROTO_IPV6;
1455 	parms->iph.ihl = 5;
1456 	parms->iph.ttl = 64;
1457 
1458 	if (!data)
1459 		return;
1460 
1461 	if (data[IFLA_IPTUN_LINK])
1462 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1463 
1464 	if (data[IFLA_IPTUN_LOCAL])
1465 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1466 
1467 	if (data[IFLA_IPTUN_REMOTE])
1468 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1469 
1470 	if (data[IFLA_IPTUN_TTL]) {
1471 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1472 		if (parms->iph.ttl)
1473 			parms->iph.frag_off = htons(IP_DF);
1474 	}
1475 
1476 	if (data[IFLA_IPTUN_TOS])
1477 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1478 
1479 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1480 		parms->iph.frag_off = htons(IP_DF);
1481 
1482 	if (data[IFLA_IPTUN_FLAGS])
1483 		parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1484 
1485 	if (data[IFLA_IPTUN_PROTO])
1486 		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1487 
1488 	if (data[IFLA_IPTUN_FWMARK])
1489 		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1490 }
1491 
1492 /* This function returns true when ENCAP attributes are present in the nl msg */
1493 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1494 				      struct ip_tunnel_encap *ipencap)
1495 {
1496 	bool ret = false;
1497 
1498 	memset(ipencap, 0, sizeof(*ipencap));
1499 
1500 	if (!data)
1501 		return ret;
1502 
1503 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1504 		ret = true;
1505 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1506 	}
1507 
1508 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1509 		ret = true;
1510 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1511 	}
1512 
1513 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1514 		ret = true;
1515 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1516 	}
1517 
1518 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1519 		ret = true;
1520 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1521 	}
1522 
1523 	return ret;
1524 }
1525 
1526 #ifdef CONFIG_IPV6_SIT_6RD
1527 /* This function returns true when 6RD attributes are present in the nl msg */
1528 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1529 				    struct ip_tunnel_6rd *ip6rd)
1530 {
1531 	bool ret = false;
1532 	memset(ip6rd, 0, sizeof(*ip6rd));
1533 
1534 	if (!data)
1535 		return ret;
1536 
1537 	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1538 		ret = true;
1539 		ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1540 	}
1541 
1542 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1543 		ret = true;
1544 		ip6rd->relay_prefix =
1545 			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1546 	}
1547 
1548 	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1549 		ret = true;
1550 		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1551 	}
1552 
1553 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1554 		ret = true;
1555 		ip6rd->relay_prefixlen =
1556 			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1557 	}
1558 
1559 	return ret;
1560 }
1561 #endif
1562 
1563 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1564 			 struct nlattr *tb[], struct nlattr *data[],
1565 			 struct netlink_ext_ack *extack)
1566 {
1567 	struct net *net = dev_net(dev);
1568 	struct ip_tunnel *nt;
1569 	struct ip_tunnel_encap ipencap;
1570 #ifdef CONFIG_IPV6_SIT_6RD
1571 	struct ip_tunnel_6rd ip6rd;
1572 #endif
1573 	int err;
1574 
1575 	nt = netdev_priv(dev);
1576 
1577 	if (ipip6_netlink_encap_parms(data, &ipencap)) {
1578 		err = ip_tunnel_encap_setup(nt, &ipencap);
1579 		if (err < 0)
1580 			return err;
1581 	}
1582 
1583 	ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1584 
1585 	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1586 		return -EEXIST;
1587 
1588 	err = ipip6_tunnel_create(dev);
1589 	if (err < 0)
1590 		return err;
1591 
1592 	if (tb[IFLA_MTU]) {
1593 		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1594 
1595 		if (mtu >= IPV6_MIN_MTU &&
1596 		    mtu <= IP6_MAX_MTU - dev->hard_header_len)
1597 			dev->mtu = mtu;
1598 	}
1599 
1600 #ifdef CONFIG_IPV6_SIT_6RD
1601 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1602 		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1603 #endif
1604 
1605 	return err;
1606 }
1607 
1608 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1609 			    struct nlattr *data[],
1610 			    struct netlink_ext_ack *extack)
1611 {
1612 	struct ip_tunnel *t = netdev_priv(dev);
1613 	struct ip_tunnel_parm p;
1614 	struct ip_tunnel_encap ipencap;
1615 	struct net *net = t->net;
1616 	struct sit_net *sitn = net_generic(net, sit_net_id);
1617 #ifdef CONFIG_IPV6_SIT_6RD
1618 	struct ip_tunnel_6rd ip6rd;
1619 #endif
1620 	__u32 fwmark = t->fwmark;
1621 	int err;
1622 
1623 	if (dev == sitn->fb_tunnel_dev)
1624 		return -EINVAL;
1625 
1626 	if (ipip6_netlink_encap_parms(data, &ipencap)) {
1627 		err = ip_tunnel_encap_setup(t, &ipencap);
1628 		if (err < 0)
1629 			return err;
1630 	}
1631 
1632 	ipip6_netlink_parms(data, &p, &fwmark);
1633 
1634 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1635 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1636 		return -EINVAL;
1637 
1638 	t = ipip6_tunnel_locate(net, &p, 0);
1639 
1640 	if (t) {
1641 		if (t->dev != dev)
1642 			return -EEXIST;
1643 	} else
1644 		t = netdev_priv(dev);
1645 
1646 	ipip6_tunnel_update(t, &p, fwmark);
1647 
1648 #ifdef CONFIG_IPV6_SIT_6RD
1649 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1650 		return ipip6_tunnel_update_6rd(t, &ip6rd);
1651 #endif
1652 
1653 	return 0;
1654 }
1655 
1656 static size_t ipip6_get_size(const struct net_device *dev)
1657 {
1658 	return
1659 		/* IFLA_IPTUN_LINK */
1660 		nla_total_size(4) +
1661 		/* IFLA_IPTUN_LOCAL */
1662 		nla_total_size(4) +
1663 		/* IFLA_IPTUN_REMOTE */
1664 		nla_total_size(4) +
1665 		/* IFLA_IPTUN_TTL */
1666 		nla_total_size(1) +
1667 		/* IFLA_IPTUN_TOS */
1668 		nla_total_size(1) +
1669 		/* IFLA_IPTUN_PMTUDISC */
1670 		nla_total_size(1) +
1671 		/* IFLA_IPTUN_FLAGS */
1672 		nla_total_size(2) +
1673 		/* IFLA_IPTUN_PROTO */
1674 		nla_total_size(1) +
1675 #ifdef CONFIG_IPV6_SIT_6RD
1676 		/* IFLA_IPTUN_6RD_PREFIX */
1677 		nla_total_size(sizeof(struct in6_addr)) +
1678 		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1679 		nla_total_size(4) +
1680 		/* IFLA_IPTUN_6RD_PREFIXLEN */
1681 		nla_total_size(2) +
1682 		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1683 		nla_total_size(2) +
1684 #endif
1685 		/* IFLA_IPTUN_ENCAP_TYPE */
1686 		nla_total_size(2) +
1687 		/* IFLA_IPTUN_ENCAP_FLAGS */
1688 		nla_total_size(2) +
1689 		/* IFLA_IPTUN_ENCAP_SPORT */
1690 		nla_total_size(2) +
1691 		/* IFLA_IPTUN_ENCAP_DPORT */
1692 		nla_total_size(2) +
1693 		/* IFLA_IPTUN_FWMARK */
1694 		nla_total_size(4) +
1695 		0;
1696 }
1697 
1698 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1699 {
1700 	struct ip_tunnel *tunnel = netdev_priv(dev);
1701 	struct ip_tunnel_parm *parm = &tunnel->parms;
1702 
1703 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1704 	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1705 	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1706 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1707 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1708 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1709 		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1710 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1711 	    nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1712 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1713 		goto nla_put_failure;
1714 
1715 #ifdef CONFIG_IPV6_SIT_6RD
1716 	if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1717 			     &tunnel->ip6rd.prefix) ||
1718 	    nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1719 			    tunnel->ip6rd.relay_prefix) ||
1720 	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1721 			tunnel->ip6rd.prefixlen) ||
1722 	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1723 			tunnel->ip6rd.relay_prefixlen))
1724 		goto nla_put_failure;
1725 #endif
1726 
1727 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1728 			tunnel->encap.type) ||
1729 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1730 			tunnel->encap.sport) ||
1731 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1732 			tunnel->encap.dport) ||
1733 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1734 			tunnel->encap.flags))
1735 		goto nla_put_failure;
1736 
1737 	return 0;
1738 
1739 nla_put_failure:
1740 	return -EMSGSIZE;
1741 }
1742 
1743 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1744 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1745 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1746 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1747 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1748 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1749 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1750 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1751 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1752 #ifdef CONFIG_IPV6_SIT_6RD
1753 	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1754 	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1755 	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1756 	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1757 #endif
1758 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
1759 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
1760 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
1761 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
1762 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
1763 };
1764 
1765 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1766 {
1767 	struct net *net = dev_net(dev);
1768 	struct sit_net *sitn = net_generic(net, sit_net_id);
1769 
1770 	if (dev != sitn->fb_tunnel_dev)
1771 		unregister_netdevice_queue(dev, head);
1772 }
1773 
1774 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1775 	.kind		= "sit",
1776 	.maxtype	= IFLA_IPTUN_MAX,
1777 	.policy		= ipip6_policy,
1778 	.priv_size	= sizeof(struct ip_tunnel),
1779 	.setup		= ipip6_tunnel_setup,
1780 	.validate	= ipip6_validate,
1781 	.newlink	= ipip6_newlink,
1782 	.changelink	= ipip6_changelink,
1783 	.get_size	= ipip6_get_size,
1784 	.fill_info	= ipip6_fill_info,
1785 	.dellink	= ipip6_dellink,
1786 	.get_link_net	= ip_tunnel_get_link_net,
1787 };
1788 
1789 static struct xfrm_tunnel sit_handler __read_mostly = {
1790 	.handler	=	ipip6_rcv,
1791 	.err_handler	=	ipip6_err,
1792 	.priority	=	1,
1793 };
1794 
1795 static struct xfrm_tunnel ipip_handler __read_mostly = {
1796 	.handler	=	ipip_rcv,
1797 	.err_handler	=	ipip6_err,
1798 	.priority	=	2,
1799 };
1800 
1801 #if IS_ENABLED(CONFIG_MPLS)
1802 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1803 	.handler	=	mplsip_rcv,
1804 	.err_handler	=	ipip6_err,
1805 	.priority	=	2,
1806 };
1807 #endif
1808 
1809 static void __net_exit sit_destroy_tunnels(struct net *net,
1810 					   struct list_head *head)
1811 {
1812 	struct sit_net *sitn = net_generic(net, sit_net_id);
1813 	struct net_device *dev, *aux;
1814 	int prio;
1815 
1816 	for_each_netdev_safe(net, dev, aux)
1817 		if (dev->rtnl_link_ops == &sit_link_ops)
1818 			unregister_netdevice_queue(dev, head);
1819 
1820 	for (prio = 1; prio < 4; prio++) {
1821 		int h;
1822 		for (h = 0; h < IP6_SIT_HASH_SIZE; h++) {
1823 			struct ip_tunnel *t;
1824 
1825 			t = rtnl_dereference(sitn->tunnels[prio][h]);
1826 			while (t) {
1827 				/* If dev is in the same netns, it has already
1828 				 * been added to the list by the previous loop.
1829 				 */
1830 				if (!net_eq(dev_net(t->dev), net))
1831 					unregister_netdevice_queue(t->dev,
1832 								   head);
1833 				t = rtnl_dereference(t->next);
1834 			}
1835 		}
1836 	}
1837 }
1838 
1839 static int __net_init sit_init_net(struct net *net)
1840 {
1841 	struct sit_net *sitn = net_generic(net, sit_net_id);
1842 	struct ip_tunnel *t;
1843 	int err;
1844 
1845 	sitn->tunnels[0] = sitn->tunnels_wc;
1846 	sitn->tunnels[1] = sitn->tunnels_l;
1847 	sitn->tunnels[2] = sitn->tunnels_r;
1848 	sitn->tunnels[3] = sitn->tunnels_r_l;
1849 
1850 	if (!net_has_fallback_tunnels(net))
1851 		return 0;
1852 
1853 	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1854 					   NET_NAME_UNKNOWN,
1855 					   ipip6_tunnel_setup);
1856 	if (!sitn->fb_tunnel_dev) {
1857 		err = -ENOMEM;
1858 		goto err_alloc_dev;
1859 	}
1860 	dev_net_set(sitn->fb_tunnel_dev, net);
1861 	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1862 	/* FB netdevice is special: we have one, and only one per netns.
1863 	 * Allowing to move it to another netns is clearly unsafe.
1864 	 */
1865 	sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1866 
1867 	err = register_netdev(sitn->fb_tunnel_dev);
1868 	if (err)
1869 		goto err_reg_dev;
1870 
1871 	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1872 	ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1873 
1874 	t = netdev_priv(sitn->fb_tunnel_dev);
1875 
1876 	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1877 	return 0;
1878 
1879 err_reg_dev:
1880 	ipip6_dev_free(sitn->fb_tunnel_dev);
1881 	free_netdev(sitn->fb_tunnel_dev);
1882 err_alloc_dev:
1883 	return err;
1884 }
1885 
1886 static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1887 {
1888 	LIST_HEAD(list);
1889 	struct net *net;
1890 
1891 	rtnl_lock();
1892 	list_for_each_entry(net, net_list, exit_list)
1893 		sit_destroy_tunnels(net, &list);
1894 
1895 	unregister_netdevice_many(&list);
1896 	rtnl_unlock();
1897 }
1898 
1899 static struct pernet_operations sit_net_ops = {
1900 	.init = sit_init_net,
1901 	.exit_batch = sit_exit_batch_net,
1902 	.id   = &sit_net_id,
1903 	.size = sizeof(struct sit_net),
1904 };
1905 
1906 static void __exit sit_cleanup(void)
1907 {
1908 	rtnl_link_unregister(&sit_link_ops);
1909 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1910 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1911 #if IS_ENABLED(CONFIG_MPLS)
1912 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1913 #endif
1914 
1915 	unregister_pernet_device(&sit_net_ops);
1916 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1917 }
1918 
1919 static int __init sit_init(void)
1920 {
1921 	int err;
1922 
1923 	pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1924 
1925 	err = register_pernet_device(&sit_net_ops);
1926 	if (err < 0)
1927 		return err;
1928 	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1929 	if (err < 0) {
1930 		pr_info("%s: can't register ip6ip4\n", __func__);
1931 		goto xfrm_tunnel_failed;
1932 	}
1933 	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1934 	if (err < 0) {
1935 		pr_info("%s: can't register ip4ip4\n", __func__);
1936 		goto xfrm_tunnel4_failed;
1937 	}
1938 #if IS_ENABLED(CONFIG_MPLS)
1939 	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1940 	if (err < 0) {
1941 		pr_info("%s: can't register mplsip\n", __func__);
1942 		goto xfrm_tunnel_mpls_failed;
1943 	}
1944 #endif
1945 	err = rtnl_link_register(&sit_link_ops);
1946 	if (err < 0)
1947 		goto rtnl_link_failed;
1948 
1949 out:
1950 	return err;
1951 
1952 rtnl_link_failed:
1953 #if IS_ENABLED(CONFIG_MPLS)
1954 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1955 xfrm_tunnel_mpls_failed:
1956 #endif
1957 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1958 xfrm_tunnel4_failed:
1959 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1960 xfrm_tunnel_failed:
1961 	unregister_pernet_device(&sit_net_ops);
1962 	goto out;
1963 }
1964 
1965 module_init(sit_init);
1966 module_exit(sit_cleanup);
1967 MODULE_LICENSE("GPL");
1968 MODULE_ALIAS_RTNL_LINK("sit");
1969 MODULE_ALIAS_NETDEV("sit0");
1970