xref: /linux/net/ipv6/sit.c (revision 1cc3462159babb69c84c39cb1b4e262aef3ea325)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
9  *
10  *	Changes:
11  * Roger Venning <r.venning@telstra.com>:	6to4 support
12  * Nate Thompson <nate@thebog.net>:		6to4 support
13  * Fred Templin <fred.l.templin@boeing.com>:	isatap support
14  */
15 
16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
17 
18 #include <linux/module.h>
19 #include <linux/capability.h>
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/sockios.h>
24 #include <linux/net.h>
25 #include <linux/in6.h>
26 #include <linux/netdevice.h>
27 #include <linux/if_arp.h>
28 #include <linux/icmp.h>
29 #include <linux/slab.h>
30 #include <linux/uaccess.h>
31 #include <linux/init.h>
32 #include <linux/netfilter_ipv4.h>
33 #include <linux/if_ether.h>
34 
35 #include <net/sock.h>
36 #include <net/snmp.h>
37 
38 #include <net/ipv6.h>
39 #include <net/protocol.h>
40 #include <net/transp_v6.h>
41 #include <net/ip6_fib.h>
42 #include <net/ip6_route.h>
43 #include <net/ndisc.h>
44 #include <net/addrconf.h>
45 #include <net/ip.h>
46 #include <net/udp.h>
47 #include <net/icmp.h>
48 #include <net/ip_tunnels.h>
49 #include <net/inet_ecn.h>
50 #include <net/xfrm.h>
51 #include <net/dsfield.h>
52 #include <net/net_namespace.h>
53 #include <net/netns/generic.h>
54 #include <net/inet_dscp.h>
55 
56 /*
57    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
58 
59    For comments look at net/ipv4/ip_gre.c --ANK
60  */
61 
62 #define IP6_SIT_HASH_SIZE  16
63 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
64 
65 static bool log_ecn_error = true;
66 module_param(log_ecn_error, bool, 0644);
67 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
68 
69 static int ipip6_tunnel_init(struct net_device *dev);
70 static void ipip6_tunnel_setup(struct net_device *dev);
71 static void ipip6_dev_free(struct net_device *dev);
72 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
73 		      __be32 *v4dst);
74 static struct rtnl_link_ops sit_link_ops __read_mostly;
75 
76 static unsigned int sit_net_id __read_mostly;
77 struct sit_net {
78 	struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
79 	struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
80 	struct ip_tunnel __rcu *tunnels_l[IP6_SIT_HASH_SIZE];
81 	struct ip_tunnel __rcu *tunnels_wc[1];
82 	struct ip_tunnel __rcu **tunnels[4];
83 
84 	struct net_device *fb_tunnel_dev;
85 };
86 
87 static inline struct sit_net *dev_to_sit_net(struct net_device *dev)
88 {
89 	struct ip_tunnel *t = netdev_priv(dev);
90 
91 	return net_generic(t->net, sit_net_id);
92 }
93 
94 /*
95  * Must be invoked with rcu_read_lock
96  */
97 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
98 					     struct net_device *dev,
99 					     __be32 remote, __be32 local,
100 					     int sifindex)
101 {
102 	unsigned int h0 = HASH(remote);
103 	unsigned int h1 = HASH(local);
104 	struct ip_tunnel *t;
105 	struct sit_net *sitn = net_generic(net, sit_net_id);
106 	int ifindex = dev ? dev->ifindex : 0;
107 
108 	for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
109 		if (local == t->parms.iph.saddr &&
110 		    remote == t->parms.iph.daddr &&
111 		    (!dev || !t->parms.link || ifindex == t->parms.link ||
112 		     sifindex == t->parms.link) &&
113 		    (t->dev->flags & IFF_UP))
114 			return t;
115 	}
116 	for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
117 		if (remote == t->parms.iph.daddr &&
118 		    (!dev || !t->parms.link || ifindex == t->parms.link ||
119 		     sifindex == t->parms.link) &&
120 		    (t->dev->flags & IFF_UP))
121 			return t;
122 	}
123 	for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
124 		if (local == t->parms.iph.saddr &&
125 		    (!dev || !t->parms.link || ifindex == t->parms.link ||
126 		     sifindex == t->parms.link) &&
127 		    (t->dev->flags & IFF_UP))
128 			return t;
129 	}
130 	t = rcu_dereference(sitn->tunnels_wc[0]);
131 	if (t && (t->dev->flags & IFF_UP))
132 		return t;
133 	return NULL;
134 }
135 
136 static struct ip_tunnel __rcu **
137 __ipip6_bucket(struct sit_net *sitn, struct ip_tunnel_parm_kern *parms)
138 {
139 	__be32 remote = parms->iph.daddr;
140 	__be32 local = parms->iph.saddr;
141 	unsigned int h = 0;
142 	int prio = 0;
143 
144 	if (remote) {
145 		prio |= 2;
146 		h ^= HASH(remote);
147 	}
148 	if (local) {
149 		prio |= 1;
150 		h ^= HASH(local);
151 	}
152 	return &sitn->tunnels[prio][h];
153 }
154 
155 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
156 		struct ip_tunnel *t)
157 {
158 	return __ipip6_bucket(sitn, &t->parms);
159 }
160 
161 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
162 {
163 	struct ip_tunnel __rcu **tp;
164 	struct ip_tunnel *iter;
165 
166 	for (tp = ipip6_bucket(sitn, t);
167 	     (iter = rtnl_dereference(*tp)) != NULL;
168 	     tp = &iter->next) {
169 		if (t == iter) {
170 			rcu_assign_pointer(*tp, t->next);
171 			break;
172 		}
173 	}
174 }
175 
176 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
177 {
178 	struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
179 
180 	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
181 	rcu_assign_pointer(*tp, t);
182 }
183 
184 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
185 {
186 #ifdef CONFIG_IPV6_SIT_6RD
187 	struct ip_tunnel *t = netdev_priv(dev);
188 
189 	if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
190 		ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
191 		t->ip6rd.relay_prefix = 0;
192 		t->ip6rd.prefixlen = 16;
193 		t->ip6rd.relay_prefixlen = 0;
194 	} else {
195 		struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
196 		memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
197 	}
198 #endif
199 }
200 
201 static int ipip6_tunnel_create(struct net_device *dev)
202 {
203 	struct ip_tunnel *t = netdev_priv(dev);
204 	struct sit_net *sitn = net_generic(t->net, sit_net_id);
205 	int err;
206 
207 	__dev_addr_set(dev, &t->parms.iph.saddr, 4);
208 	memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
209 
210 	if (test_bit(IP_TUNNEL_SIT_ISATAP_BIT, t->parms.i_flags))
211 		dev->priv_flags |= IFF_ISATAP;
212 
213 	dev->rtnl_link_ops = &sit_link_ops;
214 
215 	err = register_netdevice(dev);
216 	if (err < 0)
217 		goto out;
218 
219 	ipip6_tunnel_clone_6rd(dev, sitn);
220 
221 	ipip6_tunnel_link(sitn, t);
222 	return 0;
223 
224 out:
225 	return err;
226 }
227 
228 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
229 					     struct ip_tunnel_parm_kern *parms,
230 					     int create)
231 {
232 	__be32 remote = parms->iph.daddr;
233 	__be32 local = parms->iph.saddr;
234 	struct ip_tunnel *t, *nt;
235 	struct ip_tunnel __rcu **tp;
236 	struct net_device *dev;
237 	char name[IFNAMSIZ];
238 	struct sit_net *sitn = net_generic(net, sit_net_id);
239 
240 	for (tp = __ipip6_bucket(sitn, parms);
241 	    (t = rtnl_dereference(*tp)) != NULL;
242 	     tp = &t->next) {
243 		if (local == t->parms.iph.saddr &&
244 		    remote == t->parms.iph.daddr &&
245 		    parms->link == t->parms.link) {
246 			if (create)
247 				return NULL;
248 			else
249 				return t;
250 		}
251 	}
252 	if (!create)
253 		goto failed;
254 
255 	if (parms->name[0]) {
256 		if (!dev_valid_name(parms->name))
257 			goto failed;
258 		strscpy(name, parms->name, IFNAMSIZ);
259 	} else {
260 		strcpy(name, "sit%d");
261 	}
262 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
263 			   ipip6_tunnel_setup);
264 	if (!dev)
265 		return NULL;
266 
267 	dev_net_set(dev, net);
268 
269 	nt = netdev_priv(dev);
270 
271 	nt->net = net;
272 	nt->parms = *parms;
273 	if (ipip6_tunnel_create(dev) < 0)
274 		goto failed_free;
275 
276 	if (!parms->name[0])
277 		strcpy(parms->name, dev->name);
278 
279 	return nt;
280 
281 failed_free:
282 	free_netdev(dev);
283 failed:
284 	return NULL;
285 }
286 
287 #define for_each_prl_rcu(start)			\
288 	for (prl = rcu_dereference(start);	\
289 	     prl;				\
290 	     prl = rcu_dereference(prl->next))
291 
292 static struct ip_tunnel_prl_entry *
293 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
294 {
295 	struct ip_tunnel_prl_entry *prl;
296 
297 	for_each_prl_rcu(t->prl)
298 		if (prl->addr == addr)
299 			break;
300 	return prl;
301 
302 }
303 
304 static int ipip6_tunnel_get_prl(struct net_device *dev, struct ip_tunnel_prl __user *a)
305 {
306 	struct ip_tunnel *t = netdev_priv(dev);
307 	struct ip_tunnel_prl kprl, *kp;
308 	struct ip_tunnel_prl_entry *prl;
309 	unsigned int cmax, c = 0, ca, len;
310 	int ret = 0;
311 
312 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
313 		return -EINVAL;
314 
315 	if (copy_from_user(&kprl, a, sizeof(kprl)))
316 		return -EFAULT;
317 	cmax = kprl.datalen / sizeof(kprl);
318 	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
319 		cmax = 1;
320 
321 	/* For simple GET or for root users,
322 	 * we try harder to allocate.
323 	 */
324 	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
325 		kcalloc(cmax, sizeof(*kp), GFP_KERNEL_ACCOUNT | __GFP_NOWARN) :
326 		NULL;
327 
328 	ca = min(t->prl_count, cmax);
329 
330 	if (!kp) {
331 		/* We don't try hard to allocate much memory for
332 		 * non-root users.
333 		 * For root users, retry allocating enough memory for
334 		 * the answer.
335 		 */
336 		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC | __GFP_ACCOUNT |
337 					      __GFP_NOWARN);
338 		if (!kp) {
339 			ret = -ENOMEM;
340 			goto out;
341 		}
342 	}
343 
344 	rcu_read_lock();
345 	for_each_prl_rcu(t->prl) {
346 		if (c >= cmax)
347 			break;
348 		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
349 			continue;
350 		kp[c].addr = prl->addr;
351 		kp[c].flags = prl->flags;
352 		c++;
353 		if (kprl.addr != htonl(INADDR_ANY))
354 			break;
355 	}
356 
357 	rcu_read_unlock();
358 
359 	len = sizeof(*kp) * c;
360 	ret = 0;
361 	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
362 		ret = -EFAULT;
363 
364 	kfree(kp);
365 out:
366 	return ret;
367 }
368 
369 static int
370 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
371 {
372 	struct ip_tunnel_prl_entry *p;
373 	int err = 0;
374 
375 	if (a->addr == htonl(INADDR_ANY))
376 		return -EINVAL;
377 
378 	ASSERT_RTNL();
379 
380 	for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
381 		if (p->addr == a->addr) {
382 			if (chg) {
383 				p->flags = a->flags;
384 				goto out;
385 			}
386 			err = -EEXIST;
387 			goto out;
388 		}
389 	}
390 
391 	if (chg) {
392 		err = -ENXIO;
393 		goto out;
394 	}
395 
396 	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
397 	if (!p) {
398 		err = -ENOBUFS;
399 		goto out;
400 	}
401 
402 	p->next = t->prl;
403 	p->addr = a->addr;
404 	p->flags = a->flags;
405 	t->prl_count++;
406 	rcu_assign_pointer(t->prl, p);
407 out:
408 	return err;
409 }
410 
411 static void prl_list_destroy_rcu(struct rcu_head *head)
412 {
413 	struct ip_tunnel_prl_entry *p, *n;
414 
415 	p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
416 	do {
417 		n = rcu_dereference_protected(p->next, 1);
418 		kfree(p);
419 		p = n;
420 	} while (p);
421 }
422 
423 static int
424 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
425 {
426 	struct ip_tunnel_prl_entry *x;
427 	struct ip_tunnel_prl_entry __rcu **p;
428 	int err = 0;
429 
430 	ASSERT_RTNL();
431 
432 	if (a && a->addr != htonl(INADDR_ANY)) {
433 		for (p = &t->prl;
434 		     (x = rtnl_dereference(*p)) != NULL;
435 		     p = &x->next) {
436 			if (x->addr == a->addr) {
437 				*p = x->next;
438 				kfree_rcu(x, rcu_head);
439 				t->prl_count--;
440 				goto out;
441 			}
442 		}
443 		err = -ENXIO;
444 	} else {
445 		x = rtnl_dereference(t->prl);
446 		if (x) {
447 			t->prl_count = 0;
448 			call_rcu(&x->rcu_head, prl_list_destroy_rcu);
449 			t->prl = NULL;
450 		}
451 	}
452 out:
453 	return err;
454 }
455 
456 static int ipip6_tunnel_prl_ctl(struct net_device *dev,
457 				struct ip_tunnel_prl __user *data, int cmd)
458 {
459 	struct ip_tunnel *t = netdev_priv(dev);
460 	struct ip_tunnel_prl prl;
461 	int err;
462 
463 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
464 		return -EPERM;
465 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
466 		return -EINVAL;
467 
468 	if (copy_from_user(&prl, data, sizeof(prl)))
469 		return -EFAULT;
470 
471 	switch (cmd) {
472 	case SIOCDELPRL:
473 		err = ipip6_tunnel_del_prl(t, &prl);
474 		break;
475 	case SIOCADDPRL:
476 	case SIOCCHGPRL:
477 		err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
478 		break;
479 	}
480 	dst_cache_reset(&t->dst_cache);
481 	netdev_state_change(dev);
482 	return err;
483 }
484 
485 static int
486 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
487 {
488 	struct ip_tunnel_prl_entry *p;
489 	int ok = 1;
490 
491 	rcu_read_lock();
492 	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
493 	if (p) {
494 		if (p->flags & PRL_DEFAULT)
495 			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
496 		else
497 			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
498 	} else {
499 		const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
500 
501 		if (ipv6_addr_is_isatap(addr6) &&
502 		    (addr6->s6_addr32[3] == iph->saddr) &&
503 		    ipv6_chk_prefix(addr6, t->dev))
504 			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
505 		else
506 			ok = 0;
507 	}
508 	rcu_read_unlock();
509 	return ok;
510 }
511 
512 static void ipip6_tunnel_uninit(struct net_device *dev)
513 {
514 	struct ip_tunnel *tunnel = netdev_priv(dev);
515 	struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
516 
517 	if (dev == sitn->fb_tunnel_dev) {
518 		RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
519 	} else {
520 		ipip6_tunnel_unlink(sitn, tunnel);
521 		ipip6_tunnel_del_prl(tunnel, NULL);
522 	}
523 	dst_cache_reset(&tunnel->dst_cache);
524 	netdev_put(dev, &tunnel->dev_tracker);
525 }
526 
527 static int ipip6_err(struct sk_buff *skb, u32 info)
528 {
529 	const struct iphdr *iph = (const struct iphdr *)skb->data;
530 	const int type = icmp_hdr(skb)->type;
531 	const int code = icmp_hdr(skb)->code;
532 	unsigned int data_len = 0;
533 	struct ip_tunnel *t;
534 	int sifindex;
535 	int err;
536 
537 	switch (type) {
538 	default:
539 	case ICMP_PARAMETERPROB:
540 		return 0;
541 
542 	case ICMP_DEST_UNREACH:
543 		switch (code) {
544 		case ICMP_SR_FAILED:
545 			/* Impossible event. */
546 			return 0;
547 		default:
548 			/* All others are translated to HOST_UNREACH.
549 			   rfc2003 contains "deep thoughts" about NET_UNREACH,
550 			   I believe they are just ether pollution. --ANK
551 			 */
552 			break;
553 		}
554 		break;
555 	case ICMP_TIME_EXCEEDED:
556 		if (code != ICMP_EXC_TTL)
557 			return 0;
558 		data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
559 		break;
560 	case ICMP_REDIRECT:
561 		break;
562 	}
563 
564 	err = -ENOENT;
565 
566 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
567 	t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
568 				iph->daddr, iph->saddr, sifindex);
569 	if (!t)
570 		goto out;
571 
572 	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
573 		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
574 				 t->parms.link, iph->protocol);
575 		err = 0;
576 		goto out;
577 	}
578 	if (type == ICMP_REDIRECT) {
579 		ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
580 			      iph->protocol);
581 		err = 0;
582 		goto out;
583 	}
584 
585 	err = 0;
586 	if (__in6_dev_get(skb->dev) &&
587 	    !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
588 		goto out;
589 
590 	if (t->parms.iph.daddr == 0)
591 		goto out;
592 
593 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
594 		goto out;
595 
596 	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
597 		t->err_count++;
598 	else
599 		t->err_count = 1;
600 	t->err_time = jiffies;
601 out:
602 	return err;
603 }
604 
605 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
606 				  const struct in6_addr *v6addr)
607 {
608 	__be32 v4embed = 0;
609 	if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
610 		return true;
611 	return false;
612 }
613 
614 /* Checks if an address matches an address on the tunnel interface.
615  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
616  * Long story:
617  * This function is called after we considered the packet as spoofed
618  * in is_spoofed_6rd.
619  * We may have a router that is doing NAT for proto 41 packets
620  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
621  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
622  * function will return true, dropping the packet.
623  * But, we can still check if is spoofed against the IP
624  * addresses associated with the interface.
625  */
626 static bool only_dnatted(const struct ip_tunnel *tunnel,
627 	const struct in6_addr *v6dst)
628 {
629 	int prefix_len;
630 
631 #ifdef CONFIG_IPV6_SIT_6RD
632 	prefix_len = tunnel->ip6rd.prefixlen + 32
633 		- tunnel->ip6rd.relay_prefixlen;
634 #else
635 	prefix_len = 48;
636 #endif
637 	return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
638 }
639 
640 /* Returns true if a packet is spoofed */
641 static bool packet_is_spoofed(struct sk_buff *skb,
642 			      const struct iphdr *iph,
643 			      struct ip_tunnel *tunnel)
644 {
645 	const struct ipv6hdr *ipv6h;
646 
647 	if (tunnel->dev->priv_flags & IFF_ISATAP) {
648 		if (!isatap_chksrc(skb, iph, tunnel))
649 			return true;
650 
651 		return false;
652 	}
653 
654 	if (tunnel->dev->flags & IFF_POINTOPOINT)
655 		return false;
656 
657 	ipv6h = ipv6_hdr(skb);
658 
659 	if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
660 		net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
661 				     &iph->saddr, &ipv6h->saddr,
662 				     &iph->daddr, &ipv6h->daddr);
663 		return true;
664 	}
665 
666 	if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
667 		return false;
668 
669 	if (only_dnatted(tunnel, &ipv6h->daddr))
670 		return false;
671 
672 	net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
673 			     &iph->saddr, &ipv6h->saddr,
674 			     &iph->daddr, &ipv6h->daddr);
675 	return true;
676 }
677 
678 static int ipip6_rcv(struct sk_buff *skb)
679 {
680 	const struct iphdr *iph = ip_hdr(skb);
681 	struct ip_tunnel *tunnel;
682 	int sifindex;
683 	int err;
684 
685 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
686 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
687 				     iph->saddr, iph->daddr, sifindex);
688 	if (tunnel) {
689 		if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
690 		    tunnel->parms.iph.protocol != 0)
691 			goto out;
692 
693 		skb->mac_header = skb->network_header;
694 		skb_reset_network_header(skb);
695 		IPCB(skb)->flags = 0;
696 		skb->dev = tunnel->dev;
697 
698 		if (packet_is_spoofed(skb, iph, tunnel)) {
699 			DEV_STATS_INC(tunnel->dev, rx_errors);
700 			goto out;
701 		}
702 
703 		if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
704 		    !net_eq(tunnel->net, dev_net(tunnel->dev))))
705 			goto out;
706 
707 		/* skb can be uncloned in iptunnel_pull_header, so
708 		 * old iph is no longer valid
709 		 */
710 		iph = (const struct iphdr *)skb_mac_header(skb);
711 		skb_reset_mac_header(skb);
712 
713 		err = IP_ECN_decapsulate(iph, skb);
714 		if (unlikely(err)) {
715 			if (log_ecn_error)
716 				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
717 						     &iph->saddr, iph->tos);
718 			if (err > 1) {
719 				DEV_STATS_INC(tunnel->dev, rx_frame_errors);
720 				DEV_STATS_INC(tunnel->dev, rx_errors);
721 				goto out;
722 			}
723 		}
724 
725 		dev_sw_netstats_rx_add(tunnel->dev, skb->len);
726 
727 		netif_rx(skb);
728 
729 		return 0;
730 	}
731 
732 	/* no tunnel matched,  let upstream know, ipsec may handle it */
733 	return 1;
734 out:
735 	kfree_skb(skb);
736 	return 0;
737 }
738 
739 static const struct tnl_ptk_info ipip_tpi = {
740 	/* no tunnel info required for ipip. */
741 	.proto = htons(ETH_P_IP),
742 };
743 
744 #if IS_ENABLED(CONFIG_MPLS)
745 static const struct tnl_ptk_info mplsip_tpi = {
746 	/* no tunnel info required for mplsip. */
747 	.proto = htons(ETH_P_MPLS_UC),
748 };
749 #endif
750 
751 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
752 {
753 	const struct iphdr *iph;
754 	struct ip_tunnel *tunnel;
755 	int sifindex;
756 
757 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
758 
759 	iph = ip_hdr(skb);
760 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
761 				     iph->saddr, iph->daddr, sifindex);
762 	if (tunnel) {
763 		const struct tnl_ptk_info *tpi;
764 
765 		if (tunnel->parms.iph.protocol != ipproto &&
766 		    tunnel->parms.iph.protocol != 0)
767 			goto drop;
768 
769 		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
770 			goto drop;
771 #if IS_ENABLED(CONFIG_MPLS)
772 		if (ipproto == IPPROTO_MPLS)
773 			tpi = &mplsip_tpi;
774 		else
775 #endif
776 			tpi = &ipip_tpi;
777 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
778 			goto drop;
779 		skb_reset_mac_header(skb);
780 
781 		return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
782 	}
783 
784 	return 1;
785 
786 drop:
787 	kfree_skb(skb);
788 	return 0;
789 }
790 
791 static int ipip_rcv(struct sk_buff *skb)
792 {
793 	return sit_tunnel_rcv(skb, IPPROTO_IPIP);
794 }
795 
796 #if IS_ENABLED(CONFIG_MPLS)
797 static int mplsip_rcv(struct sk_buff *skb)
798 {
799 	return sit_tunnel_rcv(skb, IPPROTO_MPLS);
800 }
801 #endif
802 
803 /*
804  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
805  * stores the embedded IPv4 address in v4dst and returns true.
806  */
807 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
808 		      __be32 *v4dst)
809 {
810 #ifdef CONFIG_IPV6_SIT_6RD
811 	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
812 			      tunnel->ip6rd.prefixlen)) {
813 		unsigned int pbw0, pbi0;
814 		int pbi1;
815 		u32 d;
816 
817 		pbw0 = tunnel->ip6rd.prefixlen >> 5;
818 		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
819 
820 		d = tunnel->ip6rd.relay_prefixlen < 32 ?
821 			(ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
822 		    tunnel->ip6rd.relay_prefixlen : 0;
823 
824 		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
825 		if (pbi1 > 0)
826 			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
827 			     (32 - pbi1);
828 
829 		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
830 		return true;
831 	}
832 #else
833 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
834 		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
835 		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
836 		return true;
837 	}
838 #endif
839 	return false;
840 }
841 
842 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
843 			     const struct in6_addr *v6dst)
844 {
845 	__be32 dst = 0;
846 	check_6rd(tunnel, v6dst, &dst);
847 	return dst;
848 }
849 
850 /*
851  *	This function assumes it is being called from dev_queue_xmit()
852  *	and that skb is filled properly by that function.
853  */
854 
855 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
856 				     struct net_device *dev)
857 {
858 	struct ip_tunnel *tunnel = netdev_priv(dev);
859 	const struct iphdr  *tiph = &tunnel->parms.iph;
860 	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
861 	u8     tos = tunnel->parms.iph.tos;
862 	__be16 df = tiph->frag_off;
863 	struct rtable *rt;		/* Route to the other host */
864 	struct net_device *tdev;	/* Device to other host */
865 	unsigned int max_headroom;	/* The extra header space needed */
866 	__be32 dst = tiph->daddr;
867 	struct flowi4 fl4;
868 	int    mtu;
869 	const struct in6_addr *addr6;
870 	int addr_type;
871 	u8 ttl;
872 	u8 protocol = IPPROTO_IPV6;
873 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
874 
875 	if (tos == 1)
876 		tos = ipv6_get_dsfield(iph6);
877 
878 	/* ISATAP (RFC4214) - must come before 6to4 */
879 	if (dev->priv_flags & IFF_ISATAP) {
880 		struct neighbour *neigh = NULL;
881 		bool do_tx_error = false;
882 
883 		if (skb_dst(skb))
884 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
885 
886 		if (!neigh) {
887 			net_dbg_ratelimited("nexthop == NULL\n");
888 			goto tx_error;
889 		}
890 
891 		addr6 = (const struct in6_addr *)&neigh->primary_key;
892 		addr_type = ipv6_addr_type(addr6);
893 
894 		if ((addr_type & IPV6_ADDR_UNICAST) &&
895 		     ipv6_addr_is_isatap(addr6))
896 			dst = addr6->s6_addr32[3];
897 		else
898 			do_tx_error = true;
899 
900 		neigh_release(neigh);
901 		if (do_tx_error)
902 			goto tx_error;
903 	}
904 
905 	if (!dst)
906 		dst = try_6rd(tunnel, &iph6->daddr);
907 
908 	if (!dst) {
909 		struct neighbour *neigh = NULL;
910 		bool do_tx_error = false;
911 
912 		if (skb_dst(skb))
913 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
914 
915 		if (!neigh) {
916 			net_dbg_ratelimited("nexthop == NULL\n");
917 			goto tx_error;
918 		}
919 
920 		addr6 = (const struct in6_addr *)&neigh->primary_key;
921 		addr_type = ipv6_addr_type(addr6);
922 
923 		if (addr_type == IPV6_ADDR_ANY) {
924 			addr6 = &ipv6_hdr(skb)->daddr;
925 			addr_type = ipv6_addr_type(addr6);
926 		}
927 
928 		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
929 			dst = addr6->s6_addr32[3];
930 		else
931 			do_tx_error = true;
932 
933 		neigh_release(neigh);
934 		if (do_tx_error)
935 			goto tx_error;
936 	}
937 
938 	flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
939 			   tos & INET_DSCP_MASK, RT_SCOPE_UNIVERSE,
940 			   IPPROTO_IPV6, 0, dst, tiph->saddr, 0, 0,
941 			   sock_net_uid(tunnel->net, NULL));
942 
943 	rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
944 	if (!rt) {
945 		rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
946 		if (IS_ERR(rt)) {
947 			DEV_STATS_INC(dev, tx_carrier_errors);
948 			goto tx_error_icmp;
949 		}
950 		dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
951 	}
952 
953 	if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
954 		ip_rt_put(rt);
955 		DEV_STATS_INC(dev, tx_carrier_errors);
956 		goto tx_error_icmp;
957 	}
958 	tdev = rt->dst.dev;
959 
960 	if (tdev == dev) {
961 		ip_rt_put(rt);
962 		DEV_STATS_INC(dev, collisions);
963 		goto tx_error;
964 	}
965 
966 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
967 		ip_rt_put(rt);
968 		goto tx_error;
969 	}
970 
971 	if (df) {
972 		mtu = dst_mtu(&rt->dst) - t_hlen;
973 
974 		if (mtu < IPV4_MIN_MTU) {
975 			DEV_STATS_INC(dev, collisions);
976 			ip_rt_put(rt);
977 			goto tx_error;
978 		}
979 
980 		if (mtu < IPV6_MIN_MTU) {
981 			mtu = IPV6_MIN_MTU;
982 			df = 0;
983 		}
984 
985 		if (tunnel->parms.iph.daddr)
986 			skb_dst_update_pmtu_no_confirm(skb, mtu);
987 
988 		if (skb->len > mtu && !skb_is_gso(skb)) {
989 			icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
990 			ip_rt_put(rt);
991 			goto tx_error;
992 		}
993 	}
994 
995 	if (tunnel->err_count > 0) {
996 		if (time_before(jiffies,
997 				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
998 			tunnel->err_count--;
999 			dst_link_failure(skb);
1000 		} else
1001 			tunnel->err_count = 0;
1002 	}
1003 
1004 	/*
1005 	 * Okay, now see if we can stuff it in the buffer as-is.
1006 	 */
1007 	max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
1008 
1009 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1010 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1011 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
1012 		if (!new_skb) {
1013 			ip_rt_put(rt);
1014 			DEV_STATS_INC(dev, tx_dropped);
1015 			kfree_skb(skb);
1016 			return NETDEV_TX_OK;
1017 		}
1018 		if (skb->sk)
1019 			skb_set_owner_w(new_skb, skb->sk);
1020 		dev_kfree_skb(skb);
1021 		skb = new_skb;
1022 		iph6 = ipv6_hdr(skb);
1023 	}
1024 	ttl = tiph->ttl;
1025 	if (ttl == 0)
1026 		ttl = iph6->hop_limit;
1027 	tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
1028 
1029 	if (ip_tunnel_encap(skb, &tunnel->encap, &protocol, &fl4) < 0) {
1030 		ip_rt_put(rt);
1031 		goto tx_error;
1032 	}
1033 
1034 	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1035 
1036 	iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
1037 		      df, !net_eq(tunnel->net, dev_net(dev)));
1038 	return NETDEV_TX_OK;
1039 
1040 tx_error_icmp:
1041 	dst_link_failure(skb);
1042 tx_error:
1043 	kfree_skb(skb);
1044 	DEV_STATS_INC(dev, tx_errors);
1045 	return NETDEV_TX_OK;
1046 }
1047 
1048 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1049 				     struct net_device *dev, u8 ipproto)
1050 {
1051 	struct ip_tunnel *tunnel = netdev_priv(dev);
1052 	const struct iphdr  *tiph = &tunnel->parms.iph;
1053 
1054 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1055 		goto tx_error;
1056 
1057 	skb_set_inner_ipproto(skb, ipproto);
1058 
1059 	ip_tunnel_xmit(skb, dev, tiph, ipproto);
1060 	return NETDEV_TX_OK;
1061 tx_error:
1062 	kfree_skb(skb);
1063 	DEV_STATS_INC(dev, tx_errors);
1064 	return NETDEV_TX_OK;
1065 }
1066 
1067 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1068 				   struct net_device *dev)
1069 {
1070 	if (!pskb_inet_may_pull(skb))
1071 		goto tx_err;
1072 
1073 	switch (skb->protocol) {
1074 	case htons(ETH_P_IP):
1075 		sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1076 		break;
1077 	case htons(ETH_P_IPV6):
1078 		ipip6_tunnel_xmit(skb, dev);
1079 		break;
1080 #if IS_ENABLED(CONFIG_MPLS)
1081 	case htons(ETH_P_MPLS_UC):
1082 		sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1083 		break;
1084 #endif
1085 	default:
1086 		goto tx_err;
1087 	}
1088 
1089 	return NETDEV_TX_OK;
1090 
1091 tx_err:
1092 	DEV_STATS_INC(dev, tx_errors);
1093 	kfree_skb(skb);
1094 	return NETDEV_TX_OK;
1095 
1096 }
1097 
1098 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1099 {
1100 	struct ip_tunnel *tunnel = netdev_priv(dev);
1101 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1102 	struct net_device *tdev = NULL;
1103 	int hlen = LL_MAX_HEADER;
1104 	const struct iphdr *iph;
1105 	struct flowi4 fl4;
1106 
1107 	iph = &tunnel->parms.iph;
1108 
1109 	if (iph->daddr) {
1110 		struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1111 							  NULL,
1112 							  iph->daddr, iph->saddr,
1113 							  0, 0,
1114 							  IPPROTO_IPV6,
1115 							  iph->tos & INET_DSCP_MASK,
1116 							  tunnel->parms.link);
1117 
1118 		if (!IS_ERR(rt)) {
1119 			tdev = rt->dst.dev;
1120 			ip_rt_put(rt);
1121 		}
1122 		dev->flags |= IFF_POINTOPOINT;
1123 	}
1124 
1125 	if (!tdev && tunnel->parms.link)
1126 		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1127 
1128 	if (tdev && !netif_is_l3_master(tdev)) {
1129 		int mtu;
1130 
1131 		mtu = tdev->mtu - t_hlen;
1132 		if (mtu < IPV6_MIN_MTU)
1133 			mtu = IPV6_MIN_MTU;
1134 		WRITE_ONCE(dev->mtu, mtu);
1135 		hlen = tdev->hard_header_len + tdev->needed_headroom;
1136 	}
1137 	dev->needed_headroom = t_hlen + hlen;
1138 }
1139 
1140 static void ipip6_tunnel_update(struct ip_tunnel *t,
1141 				struct ip_tunnel_parm_kern *p,
1142 				__u32 fwmark)
1143 {
1144 	struct net *net = t->net;
1145 	struct sit_net *sitn = net_generic(net, sit_net_id);
1146 
1147 	ipip6_tunnel_unlink(sitn, t);
1148 	synchronize_net();
1149 	t->parms.iph.saddr = p->iph.saddr;
1150 	t->parms.iph.daddr = p->iph.daddr;
1151 	__dev_addr_set(t->dev, &p->iph.saddr, 4);
1152 	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1153 	ipip6_tunnel_link(sitn, t);
1154 	t->parms.iph.ttl = p->iph.ttl;
1155 	t->parms.iph.tos = p->iph.tos;
1156 	t->parms.iph.frag_off = p->iph.frag_off;
1157 	if (t->parms.link != p->link || t->fwmark != fwmark) {
1158 		t->parms.link = p->link;
1159 		t->fwmark = fwmark;
1160 		ipip6_tunnel_bind_dev(t->dev);
1161 	}
1162 	dst_cache_reset(&t->dst_cache);
1163 	netdev_state_change(t->dev);
1164 }
1165 
1166 #ifdef CONFIG_IPV6_SIT_6RD
1167 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1168 				   struct ip_tunnel_6rd *ip6rd)
1169 {
1170 	struct in6_addr prefix;
1171 	__be32 relay_prefix;
1172 
1173 	if (ip6rd->relay_prefixlen > 32 ||
1174 	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1175 		return -EINVAL;
1176 
1177 	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1178 	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1179 		return -EINVAL;
1180 	if (ip6rd->relay_prefixlen)
1181 		relay_prefix = ip6rd->relay_prefix &
1182 			       htonl(0xffffffffUL <<
1183 				     (32 - ip6rd->relay_prefixlen));
1184 	else
1185 		relay_prefix = 0;
1186 	if (relay_prefix != ip6rd->relay_prefix)
1187 		return -EINVAL;
1188 
1189 	t->ip6rd.prefix = prefix;
1190 	t->ip6rd.relay_prefix = relay_prefix;
1191 	t->ip6rd.prefixlen = ip6rd->prefixlen;
1192 	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1193 	dst_cache_reset(&t->dst_cache);
1194 	netdev_state_change(t->dev);
1195 	return 0;
1196 }
1197 
1198 static int
1199 ipip6_tunnel_get6rd(struct net_device *dev, struct ip_tunnel_parm __user *data)
1200 {
1201 	struct ip_tunnel *t = netdev_priv(dev);
1202 	struct ip_tunnel_parm_kern p;
1203 	struct ip_tunnel_6rd ip6rd;
1204 
1205 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1206 		if (!ip_tunnel_parm_from_user(&p, data))
1207 			return -EFAULT;
1208 		t = ipip6_tunnel_locate(t->net, &p, 0);
1209 	}
1210 	if (!t)
1211 		t = netdev_priv(dev);
1212 
1213 	ip6rd.prefix = t->ip6rd.prefix;
1214 	ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1215 	ip6rd.prefixlen = t->ip6rd.prefixlen;
1216 	ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1217 	if (copy_to_user(data, &ip6rd, sizeof(ip6rd)))
1218 		return -EFAULT;
1219 	return 0;
1220 }
1221 
1222 static int
1223 ipip6_tunnel_6rdctl(struct net_device *dev, struct ip_tunnel_6rd __user *data,
1224 		    int cmd)
1225 {
1226 	struct ip_tunnel *t = netdev_priv(dev);
1227 	struct ip_tunnel_6rd ip6rd;
1228 	int err;
1229 
1230 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1231 		return -EPERM;
1232 	if (copy_from_user(&ip6rd, data, sizeof(ip6rd)))
1233 		return -EFAULT;
1234 
1235 	if (cmd != SIOCDEL6RD) {
1236 		err = ipip6_tunnel_update_6rd(t, &ip6rd);
1237 		if (err < 0)
1238 			return err;
1239 	} else
1240 		ipip6_tunnel_clone_6rd(dev, dev_to_sit_net(dev));
1241 	return 0;
1242 }
1243 
1244 #endif /* CONFIG_IPV6_SIT_6RD */
1245 
1246 static bool ipip6_valid_ip_proto(u8 ipproto)
1247 {
1248 	return ipproto == IPPROTO_IPV6 ||
1249 		ipproto == IPPROTO_IPIP ||
1250 #if IS_ENABLED(CONFIG_MPLS)
1251 		ipproto == IPPROTO_MPLS ||
1252 #endif
1253 		ipproto == 0;
1254 }
1255 
1256 static int
1257 __ipip6_tunnel_ioctl_validate(struct net *net, struct ip_tunnel_parm_kern *p)
1258 {
1259 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1260 		return -EPERM;
1261 
1262 	if (!ipip6_valid_ip_proto(p->iph.protocol))
1263 		return -EINVAL;
1264 	if (p->iph.version != 4 ||
1265 	    p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
1266 		return -EINVAL;
1267 
1268 	if (p->iph.ttl)
1269 		p->iph.frag_off |= htons(IP_DF);
1270 	return 0;
1271 }
1272 
1273 static int
1274 ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1275 {
1276 	struct ip_tunnel *t = netdev_priv(dev);
1277 
1278 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
1279 		t = ipip6_tunnel_locate(t->net, p, 0);
1280 	if (!t)
1281 		t = netdev_priv(dev);
1282 	memcpy(p, &t->parms, sizeof(*p));
1283 	return 0;
1284 }
1285 
1286 static int
1287 ipip6_tunnel_add(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1288 {
1289 	struct ip_tunnel *t = netdev_priv(dev);
1290 	int err;
1291 
1292 	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1293 	if (err)
1294 		return err;
1295 
1296 	t = ipip6_tunnel_locate(t->net, p, 1);
1297 	if (!t)
1298 		return -ENOBUFS;
1299 	return 0;
1300 }
1301 
1302 static int
1303 ipip6_tunnel_change(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1304 {
1305 	struct ip_tunnel *t = netdev_priv(dev);
1306 	int err;
1307 
1308 	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1309 	if (err)
1310 		return err;
1311 
1312 	t = ipip6_tunnel_locate(t->net, p, 0);
1313 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1314 		if (!t)
1315 			return -ENOENT;
1316 	} else {
1317 		if (t) {
1318 			if (t->dev != dev)
1319 				return -EEXIST;
1320 		} else {
1321 			if (((dev->flags & IFF_POINTOPOINT) && !p->iph.daddr) ||
1322 			    (!(dev->flags & IFF_POINTOPOINT) && p->iph.daddr))
1323 				return -EINVAL;
1324 			t = netdev_priv(dev);
1325 		}
1326 
1327 		ipip6_tunnel_update(t, p, t->fwmark);
1328 	}
1329 
1330 	return 0;
1331 }
1332 
1333 static int
1334 ipip6_tunnel_del(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1335 {
1336 	struct ip_tunnel *t = netdev_priv(dev);
1337 
1338 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1339 		return -EPERM;
1340 
1341 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1342 		t = ipip6_tunnel_locate(t->net, p, 0);
1343 		if (!t)
1344 			return -ENOENT;
1345 		if (t == netdev_priv(dev_to_sit_net(dev)->fb_tunnel_dev))
1346 			return -EPERM;
1347 		dev = t->dev;
1348 	}
1349 	unregister_netdevice(dev);
1350 	return 0;
1351 }
1352 
1353 static int
1354 ipip6_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm_kern *p,
1355 		 int cmd)
1356 {
1357 	switch (cmd) {
1358 	case SIOCGETTUNNEL:
1359 		return ipip6_tunnel_get(dev, p);
1360 	case SIOCADDTUNNEL:
1361 		return ipip6_tunnel_add(dev, p);
1362 	case SIOCCHGTUNNEL:
1363 		return ipip6_tunnel_change(dev, p);
1364 	case SIOCDELTUNNEL:
1365 		return ipip6_tunnel_del(dev, p);
1366 	default:
1367 		return -EINVAL;
1368 	}
1369 }
1370 
1371 static int
1372 ipip6_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1373 			    void __user *data, int cmd)
1374 {
1375 	switch (cmd) {
1376 	case SIOCGETTUNNEL:
1377 	case SIOCADDTUNNEL:
1378 	case SIOCCHGTUNNEL:
1379 	case SIOCDELTUNNEL:
1380 		return ip_tunnel_siocdevprivate(dev, ifr, data, cmd);
1381 	case SIOCGETPRL:
1382 		return ipip6_tunnel_get_prl(dev, data);
1383 	case SIOCADDPRL:
1384 	case SIOCDELPRL:
1385 	case SIOCCHGPRL:
1386 		return ipip6_tunnel_prl_ctl(dev, data, cmd);
1387 #ifdef CONFIG_IPV6_SIT_6RD
1388 	case SIOCGET6RD:
1389 		return ipip6_tunnel_get6rd(dev, data);
1390 	case SIOCADD6RD:
1391 	case SIOCCHG6RD:
1392 	case SIOCDEL6RD:
1393 		return ipip6_tunnel_6rdctl(dev, data, cmd);
1394 #endif
1395 	default:
1396 		return -EINVAL;
1397 	}
1398 }
1399 
1400 static const struct net_device_ops ipip6_netdev_ops = {
1401 	.ndo_init	= ipip6_tunnel_init,
1402 	.ndo_uninit	= ipip6_tunnel_uninit,
1403 	.ndo_start_xmit	= sit_tunnel_xmit,
1404 	.ndo_siocdevprivate = ipip6_tunnel_siocdevprivate,
1405 	.ndo_get_iflink = ip_tunnel_get_iflink,
1406 	.ndo_tunnel_ctl = ipip6_tunnel_ctl,
1407 };
1408 
1409 static void ipip6_dev_free(struct net_device *dev)
1410 {
1411 	struct ip_tunnel *tunnel = netdev_priv(dev);
1412 
1413 	dst_cache_destroy(&tunnel->dst_cache);
1414 }
1415 
1416 #define SIT_FEATURES (NETIF_F_SG	   | \
1417 		      NETIF_F_FRAGLIST	   | \
1418 		      NETIF_F_HIGHDMA	   | \
1419 		      NETIF_F_GSO_SOFTWARE | \
1420 		      NETIF_F_HW_CSUM)
1421 
1422 static void ipip6_tunnel_setup(struct net_device *dev)
1423 {
1424 	struct ip_tunnel *tunnel = netdev_priv(dev);
1425 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1426 
1427 	dev->netdev_ops		= &ipip6_netdev_ops;
1428 	dev->header_ops		= &ip_tunnel_header_ops;
1429 	dev->needs_free_netdev	= true;
1430 	dev->priv_destructor	= ipip6_dev_free;
1431 
1432 	dev->type		= ARPHRD_SIT;
1433 	dev->mtu		= ETH_DATA_LEN - t_hlen;
1434 	dev->min_mtu		= IPV6_MIN_MTU;
1435 	dev->max_mtu		= IP6_MAX_MTU - t_hlen;
1436 	dev->flags		= IFF_NOARP;
1437 	netif_keep_dst(dev);
1438 	dev->addr_len		= 4;
1439 	dev->lltx		= true;
1440 	dev->features		|= SIT_FEATURES;
1441 	dev->hw_features	|= SIT_FEATURES;
1442 	dev->pcpu_stat_type	= NETDEV_PCPU_STAT_TSTATS;
1443 
1444 }
1445 
1446 static int ipip6_tunnel_init(struct net_device *dev)
1447 {
1448 	struct ip_tunnel *tunnel = netdev_priv(dev);
1449 	int err;
1450 
1451 	tunnel->dev = dev;
1452 	strcpy(tunnel->parms.name, dev->name);
1453 
1454 	ipip6_tunnel_bind_dev(dev);
1455 
1456 	err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1457 	if (err)
1458 		return err;
1459 
1460 	netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1461 	netdev_lockdep_set_classes(dev);
1462 	return 0;
1463 }
1464 
1465 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1466 {
1467 	struct ip_tunnel *tunnel = netdev_priv(dev);
1468 	struct iphdr *iph = &tunnel->parms.iph;
1469 	struct net *net = dev_net(dev);
1470 	struct sit_net *sitn = net_generic(net, sit_net_id);
1471 
1472 	iph->version		= 4;
1473 	iph->protocol		= IPPROTO_IPV6;
1474 	iph->ihl		= 5;
1475 	iph->ttl		= 64;
1476 
1477 	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1478 }
1479 
1480 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1481 			  struct netlink_ext_ack *extack)
1482 {
1483 	u8 proto;
1484 
1485 	if (!data || !data[IFLA_IPTUN_PROTO])
1486 		return 0;
1487 
1488 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1489 	if (!ipip6_valid_ip_proto(proto))
1490 		return -EINVAL;
1491 
1492 	return 0;
1493 }
1494 
1495 static void ipip6_netlink_parms(struct nlattr *data[],
1496 				struct ip_tunnel_parm_kern *parms,
1497 				__u32 *fwmark)
1498 {
1499 	memset(parms, 0, sizeof(*parms));
1500 
1501 	parms->iph.version = 4;
1502 	parms->iph.protocol = IPPROTO_IPV6;
1503 	parms->iph.ihl = 5;
1504 	parms->iph.ttl = 64;
1505 
1506 	if (!data)
1507 		return;
1508 
1509 	ip_tunnel_netlink_parms(data, parms);
1510 
1511 	if (data[IFLA_IPTUN_FWMARK])
1512 		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1513 }
1514 
1515 #ifdef CONFIG_IPV6_SIT_6RD
1516 /* This function returns true when 6RD attributes are present in the nl msg */
1517 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1518 				    struct ip_tunnel_6rd *ip6rd)
1519 {
1520 	bool ret = false;
1521 	memset(ip6rd, 0, sizeof(*ip6rd));
1522 
1523 	if (!data)
1524 		return ret;
1525 
1526 	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1527 		ret = true;
1528 		ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1529 	}
1530 
1531 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1532 		ret = true;
1533 		ip6rd->relay_prefix =
1534 			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1535 	}
1536 
1537 	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1538 		ret = true;
1539 		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1540 	}
1541 
1542 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1543 		ret = true;
1544 		ip6rd->relay_prefixlen =
1545 			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1546 	}
1547 
1548 	return ret;
1549 }
1550 #endif
1551 
1552 static int ipip6_newlink(struct net_device *dev,
1553 			 struct rtnl_newlink_params *params,
1554 			 struct netlink_ext_ack *extack)
1555 {
1556 	struct nlattr **data = params->data;
1557 	struct nlattr **tb = params->tb;
1558 	struct ip_tunnel *nt;
1559 	struct ip_tunnel_encap ipencap;
1560 #ifdef CONFIG_IPV6_SIT_6RD
1561 	struct ip_tunnel_6rd ip6rd;
1562 #endif
1563 	struct net *net;
1564 	int err;
1565 
1566 	net = params->link_net ? : dev_net(dev);
1567 	nt = netdev_priv(dev);
1568 	nt->net = net;
1569 
1570 	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
1571 		err = ip_tunnel_encap_setup(nt, &ipencap);
1572 		if (err < 0)
1573 			return err;
1574 	}
1575 
1576 	ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1577 
1578 	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1579 		return -EEXIST;
1580 
1581 	err = ipip6_tunnel_create(dev);
1582 	if (err < 0)
1583 		return err;
1584 
1585 	if (tb[IFLA_MTU]) {
1586 		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1587 
1588 		if (mtu >= IPV6_MIN_MTU &&
1589 		    mtu <= IP6_MAX_MTU - dev->hard_header_len)
1590 			dev->mtu = mtu;
1591 	}
1592 
1593 #ifdef CONFIG_IPV6_SIT_6RD
1594 	if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1595 		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1596 		if (err < 0)
1597 			unregister_netdevice_queue(dev, NULL);
1598 	}
1599 #endif
1600 
1601 	return err;
1602 }
1603 
1604 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1605 			    struct nlattr *data[],
1606 			    struct netlink_ext_ack *extack)
1607 {
1608 	struct ip_tunnel *t = netdev_priv(dev);
1609 	struct ip_tunnel_encap ipencap;
1610 	struct ip_tunnel_parm_kern p;
1611 	struct net *net = t->net;
1612 	struct sit_net *sitn = net_generic(net, sit_net_id);
1613 #ifdef CONFIG_IPV6_SIT_6RD
1614 	struct ip_tunnel_6rd ip6rd;
1615 #endif
1616 	__u32 fwmark = t->fwmark;
1617 	int err;
1618 
1619 	if (dev == sitn->fb_tunnel_dev)
1620 		return -EINVAL;
1621 
1622 	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
1623 		err = ip_tunnel_encap_setup(t, &ipencap);
1624 		if (err < 0)
1625 			return err;
1626 	}
1627 
1628 	ipip6_netlink_parms(data, &p, &fwmark);
1629 
1630 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1631 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1632 		return -EINVAL;
1633 
1634 	t = ipip6_tunnel_locate(net, &p, 0);
1635 
1636 	if (t) {
1637 		if (t->dev != dev)
1638 			return -EEXIST;
1639 	} else
1640 		t = netdev_priv(dev);
1641 
1642 	ipip6_tunnel_update(t, &p, fwmark);
1643 
1644 #ifdef CONFIG_IPV6_SIT_6RD
1645 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1646 		return ipip6_tunnel_update_6rd(t, &ip6rd);
1647 #endif
1648 
1649 	return 0;
1650 }
1651 
1652 static size_t ipip6_get_size(const struct net_device *dev)
1653 {
1654 	return
1655 		/* IFLA_IPTUN_LINK */
1656 		nla_total_size(4) +
1657 		/* IFLA_IPTUN_LOCAL */
1658 		nla_total_size(4) +
1659 		/* IFLA_IPTUN_REMOTE */
1660 		nla_total_size(4) +
1661 		/* IFLA_IPTUN_TTL */
1662 		nla_total_size(1) +
1663 		/* IFLA_IPTUN_TOS */
1664 		nla_total_size(1) +
1665 		/* IFLA_IPTUN_PMTUDISC */
1666 		nla_total_size(1) +
1667 		/* IFLA_IPTUN_FLAGS */
1668 		nla_total_size(2) +
1669 		/* IFLA_IPTUN_PROTO */
1670 		nla_total_size(1) +
1671 #ifdef CONFIG_IPV6_SIT_6RD
1672 		/* IFLA_IPTUN_6RD_PREFIX */
1673 		nla_total_size(sizeof(struct in6_addr)) +
1674 		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1675 		nla_total_size(4) +
1676 		/* IFLA_IPTUN_6RD_PREFIXLEN */
1677 		nla_total_size(2) +
1678 		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1679 		nla_total_size(2) +
1680 #endif
1681 		/* IFLA_IPTUN_ENCAP_TYPE */
1682 		nla_total_size(2) +
1683 		/* IFLA_IPTUN_ENCAP_FLAGS */
1684 		nla_total_size(2) +
1685 		/* IFLA_IPTUN_ENCAP_SPORT */
1686 		nla_total_size(2) +
1687 		/* IFLA_IPTUN_ENCAP_DPORT */
1688 		nla_total_size(2) +
1689 		/* IFLA_IPTUN_FWMARK */
1690 		nla_total_size(4) +
1691 		0;
1692 }
1693 
1694 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1695 {
1696 	struct ip_tunnel *tunnel = netdev_priv(dev);
1697 	struct ip_tunnel_parm_kern *parm = &tunnel->parms;
1698 
1699 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1700 	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1701 	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1702 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1703 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1704 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1705 		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1706 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1707 	    nla_put_be16(skb, IFLA_IPTUN_FLAGS,
1708 			 ip_tunnel_flags_to_be16(parm->i_flags)) ||
1709 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1710 		goto nla_put_failure;
1711 
1712 #ifdef CONFIG_IPV6_SIT_6RD
1713 	if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1714 			     &tunnel->ip6rd.prefix) ||
1715 	    nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1716 			    tunnel->ip6rd.relay_prefix) ||
1717 	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1718 			tunnel->ip6rd.prefixlen) ||
1719 	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1720 			tunnel->ip6rd.relay_prefixlen))
1721 		goto nla_put_failure;
1722 #endif
1723 
1724 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1725 			tunnel->encap.type) ||
1726 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1727 			tunnel->encap.sport) ||
1728 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1729 			tunnel->encap.dport) ||
1730 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1731 			tunnel->encap.flags))
1732 		goto nla_put_failure;
1733 
1734 	return 0;
1735 
1736 nla_put_failure:
1737 	return -EMSGSIZE;
1738 }
1739 
1740 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1741 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1742 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1743 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1744 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1745 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1746 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1747 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1748 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1749 #ifdef CONFIG_IPV6_SIT_6RD
1750 	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1751 	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1752 	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1753 	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1754 #endif
1755 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
1756 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
1757 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
1758 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
1759 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
1760 };
1761 
1762 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1763 {
1764 	struct net *net = dev_net(dev);
1765 	struct sit_net *sitn = net_generic(net, sit_net_id);
1766 
1767 	if (dev != sitn->fb_tunnel_dev)
1768 		unregister_netdevice_queue(dev, head);
1769 }
1770 
1771 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1772 	.kind		= "sit",
1773 	.maxtype	= IFLA_IPTUN_MAX,
1774 	.policy		= ipip6_policy,
1775 	.priv_size	= sizeof(struct ip_tunnel),
1776 	.setup		= ipip6_tunnel_setup,
1777 	.validate	= ipip6_validate,
1778 	.newlink	= ipip6_newlink,
1779 	.changelink	= ipip6_changelink,
1780 	.get_size	= ipip6_get_size,
1781 	.fill_info	= ipip6_fill_info,
1782 	.dellink	= ipip6_dellink,
1783 	.get_link_net	= ip_tunnel_get_link_net,
1784 };
1785 
1786 static struct xfrm_tunnel sit_handler __read_mostly = {
1787 	.handler	=	ipip6_rcv,
1788 	.err_handler	=	ipip6_err,
1789 	.priority	=	1,
1790 };
1791 
1792 static struct xfrm_tunnel ipip_handler __read_mostly = {
1793 	.handler	=	ipip_rcv,
1794 	.err_handler	=	ipip6_err,
1795 	.priority	=	2,
1796 };
1797 
1798 #if IS_ENABLED(CONFIG_MPLS)
1799 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1800 	.handler	=	mplsip_rcv,
1801 	.err_handler	=	ipip6_err,
1802 	.priority	=	2,
1803 };
1804 #endif
1805 
1806 static void __net_exit sit_destroy_tunnels(struct net *net,
1807 					   struct list_head *head)
1808 {
1809 	struct sit_net *sitn = net_generic(net, sit_net_id);
1810 	struct net_device *dev, *aux;
1811 	int prio;
1812 
1813 	for_each_netdev_safe(net, dev, aux)
1814 		if (dev->rtnl_link_ops == &sit_link_ops)
1815 			unregister_netdevice_queue(dev, head);
1816 
1817 	for (prio = 0; prio < 4; prio++) {
1818 		int h;
1819 		for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1820 			struct ip_tunnel *t;
1821 
1822 			t = rtnl_dereference(sitn->tunnels[prio][h]);
1823 			while (t) {
1824 				/* If dev is in the same netns, it has already
1825 				 * been added to the list by the previous loop.
1826 				 */
1827 				if (!net_eq(dev_net(t->dev), net))
1828 					unregister_netdevice_queue(t->dev,
1829 								   head);
1830 				t = rtnl_dereference(t->next);
1831 			}
1832 		}
1833 	}
1834 }
1835 
1836 static int __net_init sit_init_net(struct net *net)
1837 {
1838 	struct sit_net *sitn = net_generic(net, sit_net_id);
1839 	struct ip_tunnel *t;
1840 	int err;
1841 
1842 	sitn->tunnels[0] = sitn->tunnels_wc;
1843 	sitn->tunnels[1] = sitn->tunnels_l;
1844 	sitn->tunnels[2] = sitn->tunnels_r;
1845 	sitn->tunnels[3] = sitn->tunnels_r_l;
1846 
1847 	if (!net_has_fallback_tunnels(net))
1848 		return 0;
1849 
1850 	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1851 					   NET_NAME_UNKNOWN,
1852 					   ipip6_tunnel_setup);
1853 	if (!sitn->fb_tunnel_dev) {
1854 		err = -ENOMEM;
1855 		goto err_alloc_dev;
1856 	}
1857 	dev_net_set(sitn->fb_tunnel_dev, net);
1858 	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1859 	/* FB netdevice is special: we have one, and only one per netns.
1860 	 * Allowing to move it to another netns is clearly unsafe.
1861 	 */
1862 	sitn->fb_tunnel_dev->netns_immutable = true;
1863 
1864 	t = netdev_priv(sitn->fb_tunnel_dev);
1865 	t->net = net;
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 	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1875 	return 0;
1876 
1877 err_reg_dev:
1878 	free_netdev(sitn->fb_tunnel_dev);
1879 err_alloc_dev:
1880 	return err;
1881 }
1882 
1883 static void __net_exit sit_exit_batch_rtnl(struct list_head *net_list,
1884 					   struct list_head *dev_to_kill)
1885 {
1886 	struct net *net;
1887 
1888 	ASSERT_RTNL();
1889 	list_for_each_entry(net, net_list, exit_list)
1890 		sit_destroy_tunnels(net, dev_to_kill);
1891 }
1892 
1893 static struct pernet_operations sit_net_ops = {
1894 	.init = sit_init_net,
1895 	.exit_batch_rtnl = sit_exit_batch_rtnl,
1896 	.id   = &sit_net_id,
1897 	.size = sizeof(struct sit_net),
1898 };
1899 
1900 static void __exit sit_cleanup(void)
1901 {
1902 	rtnl_link_unregister(&sit_link_ops);
1903 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1904 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1905 #if IS_ENABLED(CONFIG_MPLS)
1906 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1907 #endif
1908 
1909 	unregister_pernet_device(&sit_net_ops);
1910 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1911 }
1912 
1913 static int __init sit_init(void)
1914 {
1915 	int err;
1916 
1917 	pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1918 
1919 	err = register_pernet_device(&sit_net_ops);
1920 	if (err < 0)
1921 		return err;
1922 	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1923 	if (err < 0) {
1924 		pr_info("%s: can't register ip6ip4\n", __func__);
1925 		goto xfrm_tunnel_failed;
1926 	}
1927 	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1928 	if (err < 0) {
1929 		pr_info("%s: can't register ip4ip4\n", __func__);
1930 		goto xfrm_tunnel4_failed;
1931 	}
1932 #if IS_ENABLED(CONFIG_MPLS)
1933 	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1934 	if (err < 0) {
1935 		pr_info("%s: can't register mplsip\n", __func__);
1936 		goto xfrm_tunnel_mpls_failed;
1937 	}
1938 #endif
1939 	err = rtnl_link_register(&sit_link_ops);
1940 	if (err < 0)
1941 		goto rtnl_link_failed;
1942 
1943 out:
1944 	return err;
1945 
1946 rtnl_link_failed:
1947 #if IS_ENABLED(CONFIG_MPLS)
1948 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1949 xfrm_tunnel_mpls_failed:
1950 #endif
1951 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1952 xfrm_tunnel4_failed:
1953 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1954 xfrm_tunnel_failed:
1955 	unregister_pernet_device(&sit_net_ops);
1956 	goto out;
1957 }
1958 
1959 module_init(sit_init);
1960 module_exit(sit_cleanup);
1961 MODULE_DESCRIPTION("IPv6-in-IPv4 tunnel SIT driver");
1962 MODULE_LICENSE("GPL");
1963 MODULE_ALIAS_RTNL_LINK("sit");
1964 MODULE_ALIAS_NETDEV("sit0");
1965