xref: /openbmc/linux/net/ipv6/sit.c (revision e0f6d1a5)
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, 0, iph->protocol, 0);
538 		err = 0;
539 		goto out;
540 	}
541 	if (type == ICMP_REDIRECT) {
542 		ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
543 			      iph->protocol, 0);
544 		err = 0;
545 		goto out;
546 	}
547 
548 	err = 0;
549 	if (!ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
550 		goto out;
551 
552 	if (t->parms.iph.daddr == 0)
553 		goto out;
554 
555 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
556 		goto out;
557 
558 	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
559 		t->err_count++;
560 	else
561 		t->err_count = 1;
562 	t->err_time = jiffies;
563 out:
564 	return err;
565 }
566 
567 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
568 				  const struct in6_addr *v6addr)
569 {
570 	__be32 v4embed = 0;
571 	if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
572 		return true;
573 	return false;
574 }
575 
576 /* Checks if an address matches an address on the tunnel interface.
577  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
578  * Long story:
579  * This function is called after we considered the packet as spoofed
580  * in is_spoofed_6rd.
581  * We may have a router that is doing NAT for proto 41 packets
582  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
583  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
584  * function will return true, dropping the packet.
585  * But, we can still check if is spoofed against the IP
586  * addresses associated with the interface.
587  */
588 static bool only_dnatted(const struct ip_tunnel *tunnel,
589 	const struct in6_addr *v6dst)
590 {
591 	int prefix_len;
592 
593 #ifdef CONFIG_IPV6_SIT_6RD
594 	prefix_len = tunnel->ip6rd.prefixlen + 32
595 		- tunnel->ip6rd.relay_prefixlen;
596 #else
597 	prefix_len = 48;
598 #endif
599 	return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
600 }
601 
602 /* Returns true if a packet is spoofed */
603 static bool packet_is_spoofed(struct sk_buff *skb,
604 			      const struct iphdr *iph,
605 			      struct ip_tunnel *tunnel)
606 {
607 	const struct ipv6hdr *ipv6h;
608 
609 	if (tunnel->dev->priv_flags & IFF_ISATAP) {
610 		if (!isatap_chksrc(skb, iph, tunnel))
611 			return true;
612 
613 		return false;
614 	}
615 
616 	if (tunnel->dev->flags & IFF_POINTOPOINT)
617 		return false;
618 
619 	ipv6h = ipv6_hdr(skb);
620 
621 	if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
622 		net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
623 				     &iph->saddr, &ipv6h->saddr,
624 				     &iph->daddr, &ipv6h->daddr);
625 		return true;
626 	}
627 
628 	if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
629 		return false;
630 
631 	if (only_dnatted(tunnel, &ipv6h->daddr))
632 		return false;
633 
634 	net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
635 			     &iph->saddr, &ipv6h->saddr,
636 			     &iph->daddr, &ipv6h->daddr);
637 	return true;
638 }
639 
640 static int ipip6_rcv(struct sk_buff *skb)
641 {
642 	const struct iphdr *iph = ip_hdr(skb);
643 	struct ip_tunnel *tunnel;
644 	int sifindex;
645 	int err;
646 
647 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
648 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
649 				     iph->saddr, iph->daddr, sifindex);
650 	if (tunnel) {
651 		struct pcpu_sw_netstats *tstats;
652 
653 		if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
654 		    tunnel->parms.iph.protocol != 0)
655 			goto out;
656 
657 		skb->mac_header = skb->network_header;
658 		skb_reset_network_header(skb);
659 		IPCB(skb)->flags = 0;
660 		skb->dev = tunnel->dev;
661 
662 		if (packet_is_spoofed(skb, iph, tunnel)) {
663 			tunnel->dev->stats.rx_errors++;
664 			goto out;
665 		}
666 
667 		if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
668 		    !net_eq(tunnel->net, dev_net(tunnel->dev))))
669 			goto out;
670 
671 		err = IP_ECN_decapsulate(iph, skb);
672 		if (unlikely(err)) {
673 			if (log_ecn_error)
674 				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
675 						     &iph->saddr, iph->tos);
676 			if (err > 1) {
677 				++tunnel->dev->stats.rx_frame_errors;
678 				++tunnel->dev->stats.rx_errors;
679 				goto out;
680 			}
681 		}
682 
683 		tstats = this_cpu_ptr(tunnel->dev->tstats);
684 		u64_stats_update_begin(&tstats->syncp);
685 		tstats->rx_packets++;
686 		tstats->rx_bytes += skb->len;
687 		u64_stats_update_end(&tstats->syncp);
688 
689 		netif_rx(skb);
690 
691 		return 0;
692 	}
693 
694 	/* no tunnel matched,  let upstream know, ipsec may handle it */
695 	return 1;
696 out:
697 	kfree_skb(skb);
698 	return 0;
699 }
700 
701 static const struct tnl_ptk_info ipip_tpi = {
702 	/* no tunnel info required for ipip. */
703 	.proto = htons(ETH_P_IP),
704 };
705 
706 #if IS_ENABLED(CONFIG_MPLS)
707 static const struct tnl_ptk_info mplsip_tpi = {
708 	/* no tunnel info required for mplsip. */
709 	.proto = htons(ETH_P_MPLS_UC),
710 };
711 #endif
712 
713 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
714 {
715 	const struct iphdr *iph;
716 	struct ip_tunnel *tunnel;
717 	int sifindex;
718 
719 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
720 
721 	iph = ip_hdr(skb);
722 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
723 				     iph->saddr, iph->daddr, sifindex);
724 	if (tunnel) {
725 		const struct tnl_ptk_info *tpi;
726 
727 		if (tunnel->parms.iph.protocol != ipproto &&
728 		    tunnel->parms.iph.protocol != 0)
729 			goto drop;
730 
731 		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
732 			goto drop;
733 #if IS_ENABLED(CONFIG_MPLS)
734 		if (ipproto == IPPROTO_MPLS)
735 			tpi = &mplsip_tpi;
736 		else
737 #endif
738 			tpi = &ipip_tpi;
739 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
740 			goto drop;
741 		return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
742 	}
743 
744 	return 1;
745 
746 drop:
747 	kfree_skb(skb);
748 	return 0;
749 }
750 
751 static int ipip_rcv(struct sk_buff *skb)
752 {
753 	return sit_tunnel_rcv(skb, IPPROTO_IPIP);
754 }
755 
756 #if IS_ENABLED(CONFIG_MPLS)
757 static int mplsip_rcv(struct sk_buff *skb)
758 {
759 	return sit_tunnel_rcv(skb, IPPROTO_MPLS);
760 }
761 #endif
762 
763 /*
764  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
765  * stores the embedded IPv4 address in v4dst and returns true.
766  */
767 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
768 		      __be32 *v4dst)
769 {
770 #ifdef CONFIG_IPV6_SIT_6RD
771 	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
772 			      tunnel->ip6rd.prefixlen)) {
773 		unsigned int pbw0, pbi0;
774 		int pbi1;
775 		u32 d;
776 
777 		pbw0 = tunnel->ip6rd.prefixlen >> 5;
778 		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
779 
780 		d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
781 		    tunnel->ip6rd.relay_prefixlen;
782 
783 		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
784 		if (pbi1 > 0)
785 			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
786 			     (32 - pbi1);
787 
788 		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
789 		return true;
790 	}
791 #else
792 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
793 		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
794 		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
795 		return true;
796 	}
797 #endif
798 	return false;
799 }
800 
801 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
802 			     const struct in6_addr *v6dst)
803 {
804 	__be32 dst = 0;
805 	check_6rd(tunnel, v6dst, &dst);
806 	return dst;
807 }
808 
809 /*
810  *	This function assumes it is being called from dev_queue_xmit()
811  *	and that skb is filled properly by that function.
812  */
813 
814 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
815 				     struct net_device *dev)
816 {
817 	struct ip_tunnel *tunnel = netdev_priv(dev);
818 	const struct iphdr  *tiph = &tunnel->parms.iph;
819 	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
820 	u8     tos = tunnel->parms.iph.tos;
821 	__be16 df = tiph->frag_off;
822 	struct rtable *rt;		/* Route to the other host */
823 	struct net_device *tdev;	/* Device to other host */
824 	unsigned int max_headroom;	/* The extra header space needed */
825 	__be32 dst = tiph->daddr;
826 	struct flowi4 fl4;
827 	int    mtu;
828 	const struct in6_addr *addr6;
829 	int addr_type;
830 	u8 ttl;
831 	u8 protocol = IPPROTO_IPV6;
832 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
833 
834 	if (tos == 1)
835 		tos = ipv6_get_dsfield(iph6);
836 
837 	/* ISATAP (RFC4214) - must come before 6to4 */
838 	if (dev->priv_flags & IFF_ISATAP) {
839 		struct neighbour *neigh = NULL;
840 		bool do_tx_error = false;
841 
842 		if (skb_dst(skb))
843 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
844 
845 		if (!neigh) {
846 			net_dbg_ratelimited("nexthop == NULL\n");
847 			goto tx_error;
848 		}
849 
850 		addr6 = (const struct in6_addr *)&neigh->primary_key;
851 		addr_type = ipv6_addr_type(addr6);
852 
853 		if ((addr_type & IPV6_ADDR_UNICAST) &&
854 		     ipv6_addr_is_isatap(addr6))
855 			dst = addr6->s6_addr32[3];
856 		else
857 			do_tx_error = true;
858 
859 		neigh_release(neigh);
860 		if (do_tx_error)
861 			goto tx_error;
862 	}
863 
864 	if (!dst)
865 		dst = try_6rd(tunnel, &iph6->daddr);
866 
867 	if (!dst) {
868 		struct neighbour *neigh = NULL;
869 		bool do_tx_error = false;
870 
871 		if (skb_dst(skb))
872 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
873 
874 		if (!neigh) {
875 			net_dbg_ratelimited("nexthop == NULL\n");
876 			goto tx_error;
877 		}
878 
879 		addr6 = (const struct in6_addr *)&neigh->primary_key;
880 		addr_type = ipv6_addr_type(addr6);
881 
882 		if (addr_type == IPV6_ADDR_ANY) {
883 			addr6 = &ipv6_hdr(skb)->daddr;
884 			addr_type = ipv6_addr_type(addr6);
885 		}
886 
887 		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
888 			dst = addr6->s6_addr32[3];
889 		else
890 			do_tx_error = true;
891 
892 		neigh_release(neigh);
893 		if (do_tx_error)
894 			goto tx_error;
895 	}
896 
897 	flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
898 			   RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
899 			   0, dst, tiph->saddr, 0, 0,
900 			   sock_net_uid(tunnel->net, NULL));
901 	rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
902 
903 	if (IS_ERR(rt)) {
904 		dev->stats.tx_carrier_errors++;
905 		goto tx_error_icmp;
906 	}
907 	if (rt->rt_type != RTN_UNICAST) {
908 		ip_rt_put(rt);
909 		dev->stats.tx_carrier_errors++;
910 		goto tx_error_icmp;
911 	}
912 	tdev = rt->dst.dev;
913 
914 	if (tdev == dev) {
915 		ip_rt_put(rt);
916 		dev->stats.collisions++;
917 		goto tx_error;
918 	}
919 
920 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
921 		ip_rt_put(rt);
922 		goto tx_error;
923 	}
924 
925 	if (df) {
926 		mtu = dst_mtu(&rt->dst) - t_hlen;
927 
928 		if (mtu < 68) {
929 			dev->stats.collisions++;
930 			ip_rt_put(rt);
931 			goto tx_error;
932 		}
933 
934 		if (mtu < IPV6_MIN_MTU) {
935 			mtu = IPV6_MIN_MTU;
936 			df = 0;
937 		}
938 
939 		if (tunnel->parms.iph.daddr)
940 			skb_dst_update_pmtu(skb, mtu);
941 
942 		if (skb->len > mtu && !skb_is_gso(skb)) {
943 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
944 			ip_rt_put(rt);
945 			goto tx_error;
946 		}
947 	}
948 
949 	if (tunnel->err_count > 0) {
950 		if (time_before(jiffies,
951 				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
952 			tunnel->err_count--;
953 			dst_link_failure(skb);
954 		} else
955 			tunnel->err_count = 0;
956 	}
957 
958 	/*
959 	 * Okay, now see if we can stuff it in the buffer as-is.
960 	 */
961 	max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
962 
963 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
964 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
965 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
966 		if (!new_skb) {
967 			ip_rt_put(rt);
968 			dev->stats.tx_dropped++;
969 			kfree_skb(skb);
970 			return NETDEV_TX_OK;
971 		}
972 		if (skb->sk)
973 			skb_set_owner_w(new_skb, skb->sk);
974 		dev_kfree_skb(skb);
975 		skb = new_skb;
976 		iph6 = ipv6_hdr(skb);
977 	}
978 	ttl = tiph->ttl;
979 	if (ttl == 0)
980 		ttl = iph6->hop_limit;
981 	tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
982 
983 	if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
984 		ip_rt_put(rt);
985 		goto tx_error;
986 	}
987 
988 	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
989 
990 	iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
991 		      df, !net_eq(tunnel->net, dev_net(dev)));
992 	return NETDEV_TX_OK;
993 
994 tx_error_icmp:
995 	dst_link_failure(skb);
996 tx_error:
997 	kfree_skb(skb);
998 	dev->stats.tx_errors++;
999 	return NETDEV_TX_OK;
1000 }
1001 
1002 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1003 				     struct net_device *dev, u8 ipproto)
1004 {
1005 	struct ip_tunnel *tunnel = netdev_priv(dev);
1006 	const struct iphdr  *tiph = &tunnel->parms.iph;
1007 
1008 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1009 		goto tx_error;
1010 
1011 	skb_set_inner_ipproto(skb, ipproto);
1012 
1013 	ip_tunnel_xmit(skb, dev, tiph, ipproto);
1014 	return NETDEV_TX_OK;
1015 tx_error:
1016 	kfree_skb(skb);
1017 	dev->stats.tx_errors++;
1018 	return NETDEV_TX_OK;
1019 }
1020 
1021 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1022 				   struct net_device *dev)
1023 {
1024 	switch (skb->protocol) {
1025 	case htons(ETH_P_IP):
1026 		sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1027 		break;
1028 	case htons(ETH_P_IPV6):
1029 		ipip6_tunnel_xmit(skb, dev);
1030 		break;
1031 #if IS_ENABLED(CONFIG_MPLS)
1032 	case htons(ETH_P_MPLS_UC):
1033 		sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1034 		break;
1035 #endif
1036 	default:
1037 		goto tx_err;
1038 	}
1039 
1040 	return NETDEV_TX_OK;
1041 
1042 tx_err:
1043 	dev->stats.tx_errors++;
1044 	kfree_skb(skb);
1045 	return NETDEV_TX_OK;
1046 
1047 }
1048 
1049 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1050 {
1051 	struct net_device *tdev = NULL;
1052 	struct ip_tunnel *tunnel;
1053 	const struct iphdr *iph;
1054 	struct flowi4 fl4;
1055 
1056 	tunnel = netdev_priv(dev);
1057 	iph = &tunnel->parms.iph;
1058 
1059 	if (iph->daddr) {
1060 		struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1061 							  NULL,
1062 							  iph->daddr, iph->saddr,
1063 							  0, 0,
1064 							  IPPROTO_IPV6,
1065 							  RT_TOS(iph->tos),
1066 							  tunnel->parms.link);
1067 
1068 		if (!IS_ERR(rt)) {
1069 			tdev = rt->dst.dev;
1070 			ip_rt_put(rt);
1071 		}
1072 		dev->flags |= IFF_POINTOPOINT;
1073 	}
1074 
1075 	if (!tdev && tunnel->parms.link)
1076 		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1077 
1078 	if (tdev) {
1079 		int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1080 
1081 		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1082 		dev->mtu = tdev->mtu - t_hlen;
1083 		if (dev->mtu < IPV6_MIN_MTU)
1084 			dev->mtu = IPV6_MIN_MTU;
1085 	}
1086 }
1087 
1088 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1089 				__u32 fwmark)
1090 {
1091 	struct net *net = t->net;
1092 	struct sit_net *sitn = net_generic(net, sit_net_id);
1093 
1094 	ipip6_tunnel_unlink(sitn, t);
1095 	synchronize_net();
1096 	t->parms.iph.saddr = p->iph.saddr;
1097 	t->parms.iph.daddr = p->iph.daddr;
1098 	memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1099 	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1100 	ipip6_tunnel_link(sitn, t);
1101 	t->parms.iph.ttl = p->iph.ttl;
1102 	t->parms.iph.tos = p->iph.tos;
1103 	t->parms.iph.frag_off = p->iph.frag_off;
1104 	if (t->parms.link != p->link || t->fwmark != fwmark) {
1105 		t->parms.link = p->link;
1106 		t->fwmark = fwmark;
1107 		ipip6_tunnel_bind_dev(t->dev);
1108 	}
1109 	dst_cache_reset(&t->dst_cache);
1110 	netdev_state_change(t->dev);
1111 }
1112 
1113 #ifdef CONFIG_IPV6_SIT_6RD
1114 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1115 				   struct ip_tunnel_6rd *ip6rd)
1116 {
1117 	struct in6_addr prefix;
1118 	__be32 relay_prefix;
1119 
1120 	if (ip6rd->relay_prefixlen > 32 ||
1121 	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1122 		return -EINVAL;
1123 
1124 	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1125 	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1126 		return -EINVAL;
1127 	if (ip6rd->relay_prefixlen)
1128 		relay_prefix = ip6rd->relay_prefix &
1129 			       htonl(0xffffffffUL <<
1130 				     (32 - ip6rd->relay_prefixlen));
1131 	else
1132 		relay_prefix = 0;
1133 	if (relay_prefix != ip6rd->relay_prefix)
1134 		return -EINVAL;
1135 
1136 	t->ip6rd.prefix = prefix;
1137 	t->ip6rd.relay_prefix = relay_prefix;
1138 	t->ip6rd.prefixlen = ip6rd->prefixlen;
1139 	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1140 	dst_cache_reset(&t->dst_cache);
1141 	netdev_state_change(t->dev);
1142 	return 0;
1143 }
1144 #endif
1145 
1146 static bool ipip6_valid_ip_proto(u8 ipproto)
1147 {
1148 	return ipproto == IPPROTO_IPV6 ||
1149 		ipproto == IPPROTO_IPIP ||
1150 #if IS_ENABLED(CONFIG_MPLS)
1151 		ipproto == IPPROTO_MPLS ||
1152 #endif
1153 		ipproto == 0;
1154 }
1155 
1156 static int
1157 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1158 {
1159 	int err = 0;
1160 	struct ip_tunnel_parm p;
1161 	struct ip_tunnel_prl prl;
1162 	struct ip_tunnel *t = netdev_priv(dev);
1163 	struct net *net = t->net;
1164 	struct sit_net *sitn = net_generic(net, sit_net_id);
1165 #ifdef CONFIG_IPV6_SIT_6RD
1166 	struct ip_tunnel_6rd ip6rd;
1167 #endif
1168 
1169 	switch (cmd) {
1170 	case SIOCGETTUNNEL:
1171 #ifdef CONFIG_IPV6_SIT_6RD
1172 	case SIOCGET6RD:
1173 #endif
1174 		if (dev == sitn->fb_tunnel_dev) {
1175 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1176 				err = -EFAULT;
1177 				break;
1178 			}
1179 			t = ipip6_tunnel_locate(net, &p, 0);
1180 			if (!t)
1181 				t = netdev_priv(dev);
1182 		}
1183 
1184 		err = -EFAULT;
1185 		if (cmd == SIOCGETTUNNEL) {
1186 			memcpy(&p, &t->parms, sizeof(p));
1187 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1188 					 sizeof(p)))
1189 				goto done;
1190 #ifdef CONFIG_IPV6_SIT_6RD
1191 		} else {
1192 			ip6rd.prefix = t->ip6rd.prefix;
1193 			ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1194 			ip6rd.prefixlen = t->ip6rd.prefixlen;
1195 			ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1196 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1197 					 sizeof(ip6rd)))
1198 				goto done;
1199 #endif
1200 		}
1201 		err = 0;
1202 		break;
1203 
1204 	case SIOCADDTUNNEL:
1205 	case SIOCCHGTUNNEL:
1206 		err = -EPERM;
1207 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1208 			goto done;
1209 
1210 		err = -EFAULT;
1211 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1212 			goto done;
1213 
1214 		err = -EINVAL;
1215 		if (!ipip6_valid_ip_proto(p.iph.protocol))
1216 			goto done;
1217 		if (p.iph.version != 4 ||
1218 		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1219 			goto done;
1220 		if (p.iph.ttl)
1221 			p.iph.frag_off |= htons(IP_DF);
1222 
1223 		t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1224 
1225 		if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1226 			if (t) {
1227 				if (t->dev != dev) {
1228 					err = -EEXIST;
1229 					break;
1230 				}
1231 			} else {
1232 				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1233 				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1234 					err = -EINVAL;
1235 					break;
1236 				}
1237 				t = netdev_priv(dev);
1238 			}
1239 
1240 			ipip6_tunnel_update(t, &p, t->fwmark);
1241 		}
1242 
1243 		if (t) {
1244 			err = 0;
1245 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1246 				err = -EFAULT;
1247 		} else
1248 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1249 		break;
1250 
1251 	case SIOCDELTUNNEL:
1252 		err = -EPERM;
1253 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1254 			goto done;
1255 
1256 		if (dev == sitn->fb_tunnel_dev) {
1257 			err = -EFAULT;
1258 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1259 				goto done;
1260 			err = -ENOENT;
1261 			t = ipip6_tunnel_locate(net, &p, 0);
1262 			if (!t)
1263 				goto done;
1264 			err = -EPERM;
1265 			if (t == netdev_priv(sitn->fb_tunnel_dev))
1266 				goto done;
1267 			dev = t->dev;
1268 		}
1269 		unregister_netdevice(dev);
1270 		err = 0;
1271 		break;
1272 
1273 	case SIOCGETPRL:
1274 		err = -EINVAL;
1275 		if (dev == sitn->fb_tunnel_dev)
1276 			goto done;
1277 		err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1278 		break;
1279 
1280 	case SIOCADDPRL:
1281 	case SIOCDELPRL:
1282 	case SIOCCHGPRL:
1283 		err = -EPERM;
1284 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1285 			goto done;
1286 		err = -EINVAL;
1287 		if (dev == sitn->fb_tunnel_dev)
1288 			goto done;
1289 		err = -EFAULT;
1290 		if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1291 			goto done;
1292 
1293 		switch (cmd) {
1294 		case SIOCDELPRL:
1295 			err = ipip6_tunnel_del_prl(t, &prl);
1296 			break;
1297 		case SIOCADDPRL:
1298 		case SIOCCHGPRL:
1299 			err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1300 			break;
1301 		}
1302 		dst_cache_reset(&t->dst_cache);
1303 		netdev_state_change(dev);
1304 		break;
1305 
1306 #ifdef CONFIG_IPV6_SIT_6RD
1307 	case SIOCADD6RD:
1308 	case SIOCCHG6RD:
1309 	case SIOCDEL6RD:
1310 		err = -EPERM;
1311 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1312 			goto done;
1313 
1314 		err = -EFAULT;
1315 		if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1316 				   sizeof(ip6rd)))
1317 			goto done;
1318 
1319 		if (cmd != SIOCDEL6RD) {
1320 			err = ipip6_tunnel_update_6rd(t, &ip6rd);
1321 			if (err < 0)
1322 				goto done;
1323 		} else
1324 			ipip6_tunnel_clone_6rd(dev, sitn);
1325 
1326 		err = 0;
1327 		break;
1328 #endif
1329 
1330 	default:
1331 		err = -EINVAL;
1332 	}
1333 
1334 done:
1335 	return err;
1336 }
1337 
1338 static const struct net_device_ops ipip6_netdev_ops = {
1339 	.ndo_init	= ipip6_tunnel_init,
1340 	.ndo_uninit	= ipip6_tunnel_uninit,
1341 	.ndo_start_xmit	= sit_tunnel_xmit,
1342 	.ndo_do_ioctl	= ipip6_tunnel_ioctl,
1343 	.ndo_get_stats64 = ip_tunnel_get_stats64,
1344 	.ndo_get_iflink = ip_tunnel_get_iflink,
1345 };
1346 
1347 static void ipip6_dev_free(struct net_device *dev)
1348 {
1349 	struct ip_tunnel *tunnel = netdev_priv(dev);
1350 
1351 	dst_cache_destroy(&tunnel->dst_cache);
1352 	free_percpu(dev->tstats);
1353 }
1354 
1355 #define SIT_FEATURES (NETIF_F_SG	   | \
1356 		      NETIF_F_FRAGLIST	   | \
1357 		      NETIF_F_HIGHDMA	   | \
1358 		      NETIF_F_GSO_SOFTWARE | \
1359 		      NETIF_F_HW_CSUM)
1360 
1361 static void ipip6_tunnel_setup(struct net_device *dev)
1362 {
1363 	struct ip_tunnel *tunnel = netdev_priv(dev);
1364 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1365 
1366 	dev->netdev_ops		= &ipip6_netdev_ops;
1367 	dev->needs_free_netdev	= true;
1368 	dev->priv_destructor	= ipip6_dev_free;
1369 
1370 	dev->type		= ARPHRD_SIT;
1371 	dev->hard_header_len	= LL_MAX_HEADER + t_hlen;
1372 	dev->mtu		= ETH_DATA_LEN - t_hlen;
1373 	dev->min_mtu		= IPV6_MIN_MTU;
1374 	dev->max_mtu		= 0xFFF8 - t_hlen;
1375 	dev->flags		= IFF_NOARP;
1376 	netif_keep_dst(dev);
1377 	dev->addr_len		= 4;
1378 	dev->features		|= NETIF_F_LLTX;
1379 	dev->features		|= SIT_FEATURES;
1380 	dev->hw_features	|= SIT_FEATURES;
1381 }
1382 
1383 static int ipip6_tunnel_init(struct net_device *dev)
1384 {
1385 	struct ip_tunnel *tunnel = netdev_priv(dev);
1386 	int err;
1387 
1388 	tunnel->dev = dev;
1389 	tunnel->net = dev_net(dev);
1390 	strcpy(tunnel->parms.name, dev->name);
1391 
1392 	ipip6_tunnel_bind_dev(dev);
1393 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1394 	if (!dev->tstats)
1395 		return -ENOMEM;
1396 
1397 	err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1398 	if (err) {
1399 		free_percpu(dev->tstats);
1400 		dev->tstats = NULL;
1401 		return err;
1402 	}
1403 
1404 	return 0;
1405 }
1406 
1407 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1408 {
1409 	struct ip_tunnel *tunnel = netdev_priv(dev);
1410 	struct iphdr *iph = &tunnel->parms.iph;
1411 	struct net *net = dev_net(dev);
1412 	struct sit_net *sitn = net_generic(net, sit_net_id);
1413 
1414 	iph->version		= 4;
1415 	iph->protocol		= IPPROTO_IPV6;
1416 	iph->ihl		= 5;
1417 	iph->ttl		= 64;
1418 
1419 	dev_hold(dev);
1420 	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1421 }
1422 
1423 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1424 			  struct netlink_ext_ack *extack)
1425 {
1426 	u8 proto;
1427 
1428 	if (!data || !data[IFLA_IPTUN_PROTO])
1429 		return 0;
1430 
1431 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1432 	if (!ipip6_valid_ip_proto(proto))
1433 		return -EINVAL;
1434 
1435 	return 0;
1436 }
1437 
1438 static void ipip6_netlink_parms(struct nlattr *data[],
1439 				struct ip_tunnel_parm *parms,
1440 				__u32 *fwmark)
1441 {
1442 	memset(parms, 0, sizeof(*parms));
1443 
1444 	parms->iph.version = 4;
1445 	parms->iph.protocol = IPPROTO_IPV6;
1446 	parms->iph.ihl = 5;
1447 	parms->iph.ttl = 64;
1448 
1449 	if (!data)
1450 		return;
1451 
1452 	if (data[IFLA_IPTUN_LINK])
1453 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1454 
1455 	if (data[IFLA_IPTUN_LOCAL])
1456 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1457 
1458 	if (data[IFLA_IPTUN_REMOTE])
1459 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1460 
1461 	if (data[IFLA_IPTUN_TTL]) {
1462 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1463 		if (parms->iph.ttl)
1464 			parms->iph.frag_off = htons(IP_DF);
1465 	}
1466 
1467 	if (data[IFLA_IPTUN_TOS])
1468 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1469 
1470 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1471 		parms->iph.frag_off = htons(IP_DF);
1472 
1473 	if (data[IFLA_IPTUN_FLAGS])
1474 		parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1475 
1476 	if (data[IFLA_IPTUN_PROTO])
1477 		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1478 
1479 	if (data[IFLA_IPTUN_FWMARK])
1480 		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1481 }
1482 
1483 /* This function returns true when ENCAP attributes are present in the nl msg */
1484 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1485 				      struct ip_tunnel_encap *ipencap)
1486 {
1487 	bool ret = false;
1488 
1489 	memset(ipencap, 0, sizeof(*ipencap));
1490 
1491 	if (!data)
1492 		return ret;
1493 
1494 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1495 		ret = true;
1496 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1497 	}
1498 
1499 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1500 		ret = true;
1501 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1502 	}
1503 
1504 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1505 		ret = true;
1506 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1507 	}
1508 
1509 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1510 		ret = true;
1511 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1512 	}
1513 
1514 	return ret;
1515 }
1516 
1517 #ifdef CONFIG_IPV6_SIT_6RD
1518 /* This function returns true when 6RD attributes are present in the nl msg */
1519 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1520 				    struct ip_tunnel_6rd *ip6rd)
1521 {
1522 	bool ret = false;
1523 	memset(ip6rd, 0, sizeof(*ip6rd));
1524 
1525 	if (!data)
1526 		return ret;
1527 
1528 	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1529 		ret = true;
1530 		ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1531 	}
1532 
1533 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1534 		ret = true;
1535 		ip6rd->relay_prefix =
1536 			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1537 	}
1538 
1539 	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1540 		ret = true;
1541 		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1542 	}
1543 
1544 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1545 		ret = true;
1546 		ip6rd->relay_prefixlen =
1547 			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1548 	}
1549 
1550 	return ret;
1551 }
1552 #endif
1553 
1554 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1555 			 struct nlattr *tb[], struct nlattr *data[],
1556 			 struct netlink_ext_ack *extack)
1557 {
1558 	struct net *net = dev_net(dev);
1559 	struct ip_tunnel *nt;
1560 	struct ip_tunnel_encap ipencap;
1561 #ifdef CONFIG_IPV6_SIT_6RD
1562 	struct ip_tunnel_6rd ip6rd;
1563 #endif
1564 	int err;
1565 
1566 	nt = netdev_priv(dev);
1567 
1568 	if (ipip6_netlink_encap_parms(data, &ipencap)) {
1569 		err = ip_tunnel_encap_setup(nt, &ipencap);
1570 		if (err < 0)
1571 			return err;
1572 	}
1573 
1574 	ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1575 
1576 	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1577 		return -EEXIST;
1578 
1579 	err = ipip6_tunnel_create(dev);
1580 	if (err < 0)
1581 		return err;
1582 
1583 	if (tb[IFLA_MTU]) {
1584 		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1585 
1586 		if (mtu >= IPV6_MIN_MTU && mtu <= 0xFFF8 - dev->hard_header_len)
1587 			dev->mtu = mtu;
1588 	}
1589 
1590 #ifdef CONFIG_IPV6_SIT_6RD
1591 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1592 		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1593 #endif
1594 
1595 	return err;
1596 }
1597 
1598 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1599 			    struct nlattr *data[],
1600 			    struct netlink_ext_ack *extack)
1601 {
1602 	struct ip_tunnel *t = netdev_priv(dev);
1603 	struct ip_tunnel_parm p;
1604 	struct ip_tunnel_encap ipencap;
1605 	struct net *net = t->net;
1606 	struct sit_net *sitn = net_generic(net, sit_net_id);
1607 #ifdef CONFIG_IPV6_SIT_6RD
1608 	struct ip_tunnel_6rd ip6rd;
1609 #endif
1610 	__u32 fwmark = t->fwmark;
1611 	int err;
1612 
1613 	if (dev == sitn->fb_tunnel_dev)
1614 		return -EINVAL;
1615 
1616 	if (ipip6_netlink_encap_parms(data, &ipencap)) {
1617 		err = ip_tunnel_encap_setup(t, &ipencap);
1618 		if (err < 0)
1619 			return err;
1620 	}
1621 
1622 	ipip6_netlink_parms(data, &p, &fwmark);
1623 
1624 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1625 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1626 		return -EINVAL;
1627 
1628 	t = ipip6_tunnel_locate(net, &p, 0);
1629 
1630 	if (t) {
1631 		if (t->dev != dev)
1632 			return -EEXIST;
1633 	} else
1634 		t = netdev_priv(dev);
1635 
1636 	ipip6_tunnel_update(t, &p, fwmark);
1637 
1638 #ifdef CONFIG_IPV6_SIT_6RD
1639 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1640 		return ipip6_tunnel_update_6rd(t, &ip6rd);
1641 #endif
1642 
1643 	return 0;
1644 }
1645 
1646 static size_t ipip6_get_size(const struct net_device *dev)
1647 {
1648 	return
1649 		/* IFLA_IPTUN_LINK */
1650 		nla_total_size(4) +
1651 		/* IFLA_IPTUN_LOCAL */
1652 		nla_total_size(4) +
1653 		/* IFLA_IPTUN_REMOTE */
1654 		nla_total_size(4) +
1655 		/* IFLA_IPTUN_TTL */
1656 		nla_total_size(1) +
1657 		/* IFLA_IPTUN_TOS */
1658 		nla_total_size(1) +
1659 		/* IFLA_IPTUN_PMTUDISC */
1660 		nla_total_size(1) +
1661 		/* IFLA_IPTUN_FLAGS */
1662 		nla_total_size(2) +
1663 		/* IFLA_IPTUN_PROTO */
1664 		nla_total_size(1) +
1665 #ifdef CONFIG_IPV6_SIT_6RD
1666 		/* IFLA_IPTUN_6RD_PREFIX */
1667 		nla_total_size(sizeof(struct in6_addr)) +
1668 		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1669 		nla_total_size(4) +
1670 		/* IFLA_IPTUN_6RD_PREFIXLEN */
1671 		nla_total_size(2) +
1672 		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1673 		nla_total_size(2) +
1674 #endif
1675 		/* IFLA_IPTUN_ENCAP_TYPE */
1676 		nla_total_size(2) +
1677 		/* IFLA_IPTUN_ENCAP_FLAGS */
1678 		nla_total_size(2) +
1679 		/* IFLA_IPTUN_ENCAP_SPORT */
1680 		nla_total_size(2) +
1681 		/* IFLA_IPTUN_ENCAP_DPORT */
1682 		nla_total_size(2) +
1683 		/* IFLA_IPTUN_FWMARK */
1684 		nla_total_size(4) +
1685 		0;
1686 }
1687 
1688 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1689 {
1690 	struct ip_tunnel *tunnel = netdev_priv(dev);
1691 	struct ip_tunnel_parm *parm = &tunnel->parms;
1692 
1693 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1694 	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1695 	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1696 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1697 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1698 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1699 		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1700 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1701 	    nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1702 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1703 		goto nla_put_failure;
1704 
1705 #ifdef CONFIG_IPV6_SIT_6RD
1706 	if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1707 			     &tunnel->ip6rd.prefix) ||
1708 	    nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1709 			    tunnel->ip6rd.relay_prefix) ||
1710 	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1711 			tunnel->ip6rd.prefixlen) ||
1712 	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1713 			tunnel->ip6rd.relay_prefixlen))
1714 		goto nla_put_failure;
1715 #endif
1716 
1717 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1718 			tunnel->encap.type) ||
1719 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1720 			tunnel->encap.sport) ||
1721 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1722 			tunnel->encap.dport) ||
1723 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1724 			tunnel->encap.flags))
1725 		goto nla_put_failure;
1726 
1727 	return 0;
1728 
1729 nla_put_failure:
1730 	return -EMSGSIZE;
1731 }
1732 
1733 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1734 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1735 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1736 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1737 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1738 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1739 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1740 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1741 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1742 #ifdef CONFIG_IPV6_SIT_6RD
1743 	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1744 	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1745 	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1746 	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1747 #endif
1748 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
1749 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
1750 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
1751 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
1752 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
1753 };
1754 
1755 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1756 {
1757 	struct net *net = dev_net(dev);
1758 	struct sit_net *sitn = net_generic(net, sit_net_id);
1759 
1760 	if (dev != sitn->fb_tunnel_dev)
1761 		unregister_netdevice_queue(dev, head);
1762 }
1763 
1764 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1765 	.kind		= "sit",
1766 	.maxtype	= IFLA_IPTUN_MAX,
1767 	.policy		= ipip6_policy,
1768 	.priv_size	= sizeof(struct ip_tunnel),
1769 	.setup		= ipip6_tunnel_setup,
1770 	.validate	= ipip6_validate,
1771 	.newlink	= ipip6_newlink,
1772 	.changelink	= ipip6_changelink,
1773 	.get_size	= ipip6_get_size,
1774 	.fill_info	= ipip6_fill_info,
1775 	.dellink	= ipip6_dellink,
1776 	.get_link_net	= ip_tunnel_get_link_net,
1777 };
1778 
1779 static struct xfrm_tunnel sit_handler __read_mostly = {
1780 	.handler	=	ipip6_rcv,
1781 	.err_handler	=	ipip6_err,
1782 	.priority	=	1,
1783 };
1784 
1785 static struct xfrm_tunnel ipip_handler __read_mostly = {
1786 	.handler	=	ipip_rcv,
1787 	.err_handler	=	ipip6_err,
1788 	.priority	=	2,
1789 };
1790 
1791 #if IS_ENABLED(CONFIG_MPLS)
1792 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1793 	.handler	=	mplsip_rcv,
1794 	.err_handler	=	ipip6_err,
1795 	.priority	=	2,
1796 };
1797 #endif
1798 
1799 static void __net_exit sit_destroy_tunnels(struct net *net,
1800 					   struct list_head *head)
1801 {
1802 	struct sit_net *sitn = net_generic(net, sit_net_id);
1803 	struct net_device *dev, *aux;
1804 	int prio;
1805 
1806 	for_each_netdev_safe(net, dev, aux)
1807 		if (dev->rtnl_link_ops == &sit_link_ops)
1808 			unregister_netdevice_queue(dev, head);
1809 
1810 	for (prio = 1; prio < 4; prio++) {
1811 		int h;
1812 		for (h = 0; h < IP6_SIT_HASH_SIZE; h++) {
1813 			struct ip_tunnel *t;
1814 
1815 			t = rtnl_dereference(sitn->tunnels[prio][h]);
1816 			while (t) {
1817 				/* If dev is in the same netns, it has already
1818 				 * been added to the list by the previous loop.
1819 				 */
1820 				if (!net_eq(dev_net(t->dev), net))
1821 					unregister_netdevice_queue(t->dev,
1822 								   head);
1823 				t = rtnl_dereference(t->next);
1824 			}
1825 		}
1826 	}
1827 }
1828 
1829 static int __net_init sit_init_net(struct net *net)
1830 {
1831 	struct sit_net *sitn = net_generic(net, sit_net_id);
1832 	struct ip_tunnel *t;
1833 	int err;
1834 
1835 	sitn->tunnels[0] = sitn->tunnels_wc;
1836 	sitn->tunnels[1] = sitn->tunnels_l;
1837 	sitn->tunnels[2] = sitn->tunnels_r;
1838 	sitn->tunnels[3] = sitn->tunnels_r_l;
1839 
1840 	if (!net_has_fallback_tunnels(net))
1841 		return 0;
1842 
1843 	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1844 					   NET_NAME_UNKNOWN,
1845 					   ipip6_tunnel_setup);
1846 	if (!sitn->fb_tunnel_dev) {
1847 		err = -ENOMEM;
1848 		goto err_alloc_dev;
1849 	}
1850 	dev_net_set(sitn->fb_tunnel_dev, net);
1851 	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1852 	/* FB netdevice is special: we have one, and only one per netns.
1853 	 * Allowing to move it to another netns is clearly unsafe.
1854 	 */
1855 	sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1856 
1857 	err = register_netdev(sitn->fb_tunnel_dev);
1858 	if (err)
1859 		goto err_reg_dev;
1860 
1861 	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1862 	ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1863 
1864 	t = netdev_priv(sitn->fb_tunnel_dev);
1865 
1866 	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1867 	return 0;
1868 
1869 err_reg_dev:
1870 	ipip6_dev_free(sitn->fb_tunnel_dev);
1871 err_alloc_dev:
1872 	return err;
1873 }
1874 
1875 static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1876 {
1877 	LIST_HEAD(list);
1878 	struct net *net;
1879 
1880 	rtnl_lock();
1881 	list_for_each_entry(net, net_list, exit_list)
1882 		sit_destroy_tunnels(net, &list);
1883 
1884 	unregister_netdevice_many(&list);
1885 	rtnl_unlock();
1886 }
1887 
1888 static struct pernet_operations sit_net_ops = {
1889 	.init = sit_init_net,
1890 	.exit_batch = sit_exit_batch_net,
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_LICENSE("GPL");
1957 MODULE_ALIAS_RTNL_LINK("sit");
1958 MODULE_ALIAS_NETDEV("sit0");
1959