xref: /openbmc/linux/net/ipv6/sit.c (revision e15a5365)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
9  *
10  *	Changes:
11  * Roger Venning <r.venning@telstra.com>:	6to4 support
12  * Nate Thompson <nate@thebog.net>:		6to4 support
13  * Fred Templin <fred.l.templin@boeing.com>:	isatap support
14  */
15 
16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
17 
18 #include <linux/module.h>
19 #include <linux/capability.h>
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/sockios.h>
24 #include <linux/net.h>
25 #include <linux/in6.h>
26 #include <linux/netdevice.h>
27 #include <linux/if_arp.h>
28 #include <linux/icmp.h>
29 #include <linux/slab.h>
30 #include <linux/uaccess.h>
31 #include <linux/init.h>
32 #include <linux/netfilter_ipv4.h>
33 #include <linux/if_ether.h>
34 
35 #include <net/sock.h>
36 #include <net/snmp.h>
37 
38 #include <net/ipv6.h>
39 #include <net/protocol.h>
40 #include <net/transp_v6.h>
41 #include <net/ip6_fib.h>
42 #include <net/ip6_route.h>
43 #include <net/ndisc.h>
44 #include <net/addrconf.h>
45 #include <net/ip.h>
46 #include <net/udp.h>
47 #include <net/icmp.h>
48 #include <net/ip_tunnels.h>
49 #include <net/inet_ecn.h>
50 #include <net/xfrm.h>
51 #include <net/dsfield.h>
52 #include <net/net_namespace.h>
53 #include <net/netns/generic.h>
54 
55 /*
56    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
57 
58    For comments look at net/ipv4/ip_gre.c --ANK
59  */
60 
61 #define IP6_SIT_HASH_SIZE  16
62 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
63 
64 static bool log_ecn_error = true;
65 module_param(log_ecn_error, bool, 0644);
66 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
67 
68 static int ipip6_tunnel_init(struct net_device *dev);
69 static void ipip6_tunnel_setup(struct net_device *dev);
70 static void ipip6_dev_free(struct net_device *dev);
71 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
72 		      __be32 *v4dst);
73 static struct rtnl_link_ops sit_link_ops __read_mostly;
74 
75 static unsigned int sit_net_id __read_mostly;
76 struct sit_net {
77 	struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
78 	struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
79 	struct ip_tunnel __rcu *tunnels_l[IP6_SIT_HASH_SIZE];
80 	struct ip_tunnel __rcu *tunnels_wc[1];
81 	struct ip_tunnel __rcu **tunnels[4];
82 
83 	struct net_device *fb_tunnel_dev;
84 };
85 
86 static inline struct sit_net *dev_to_sit_net(struct net_device *dev)
87 {
88 	struct ip_tunnel *t = netdev_priv(dev);
89 
90 	return net_generic(t->net, sit_net_id);
91 }
92 
93 /*
94  * Must be invoked with rcu_read_lock
95  */
96 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
97 					     struct net_device *dev,
98 					     __be32 remote, __be32 local,
99 					     int sifindex)
100 {
101 	unsigned int h0 = HASH(remote);
102 	unsigned int h1 = HASH(local);
103 	struct ip_tunnel *t;
104 	struct sit_net *sitn = net_generic(net, sit_net_id);
105 	int ifindex = dev ? dev->ifindex : 0;
106 
107 	for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
108 		if (local == t->parms.iph.saddr &&
109 		    remote == t->parms.iph.daddr &&
110 		    (!dev || !t->parms.link || ifindex == t->parms.link ||
111 		     sifindex == t->parms.link) &&
112 		    (t->dev->flags & IFF_UP))
113 			return t;
114 	}
115 	for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
116 		if (remote == t->parms.iph.daddr &&
117 		    (!dev || !t->parms.link || ifindex == t->parms.link ||
118 		     sifindex == t->parms.link) &&
119 		    (t->dev->flags & IFF_UP))
120 			return t;
121 	}
122 	for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
123 		if (local == t->parms.iph.saddr &&
124 		    (!dev || !t->parms.link || ifindex == t->parms.link ||
125 		     sifindex == t->parms.link) &&
126 		    (t->dev->flags & IFF_UP))
127 			return t;
128 	}
129 	t = rcu_dereference(sitn->tunnels_wc[0]);
130 	if (t && (t->dev->flags & IFF_UP))
131 		return t;
132 	return NULL;
133 }
134 
135 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
136 		struct ip_tunnel_parm *parms)
137 {
138 	__be32 remote = parms->iph.daddr;
139 	__be32 local = parms->iph.saddr;
140 	unsigned int h = 0;
141 	int prio = 0;
142 
143 	if (remote) {
144 		prio |= 2;
145 		h ^= HASH(remote);
146 	}
147 	if (local) {
148 		prio |= 1;
149 		h ^= HASH(local);
150 	}
151 	return &sitn->tunnels[prio][h];
152 }
153 
154 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
155 		struct ip_tunnel *t)
156 {
157 	return __ipip6_bucket(sitn, &t->parms);
158 }
159 
160 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
161 {
162 	struct ip_tunnel __rcu **tp;
163 	struct ip_tunnel *iter;
164 
165 	for (tp = ipip6_bucket(sitn, t);
166 	     (iter = rtnl_dereference(*tp)) != NULL;
167 	     tp = &iter->next) {
168 		if (t == iter) {
169 			rcu_assign_pointer(*tp, t->next);
170 			break;
171 		}
172 	}
173 }
174 
175 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
176 {
177 	struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
178 
179 	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
180 	rcu_assign_pointer(*tp, t);
181 }
182 
183 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
184 {
185 #ifdef CONFIG_IPV6_SIT_6RD
186 	struct ip_tunnel *t = netdev_priv(dev);
187 
188 	if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
189 		ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
190 		t->ip6rd.relay_prefix = 0;
191 		t->ip6rd.prefixlen = 16;
192 		t->ip6rd.relay_prefixlen = 0;
193 	} else {
194 		struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
195 		memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
196 	}
197 #endif
198 }
199 
200 static int ipip6_tunnel_create(struct net_device *dev)
201 {
202 	struct ip_tunnel *t = netdev_priv(dev);
203 	struct net *net = dev_net(dev);
204 	struct sit_net *sitn = net_generic(net, sit_net_id);
205 	int err;
206 
207 	memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
208 	memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
209 
210 	if ((__force u16)t->parms.i_flags & SIT_ISATAP)
211 		dev->priv_flags |= IFF_ISATAP;
212 
213 	dev->rtnl_link_ops = &sit_link_ops;
214 
215 	err = register_netdevice(dev);
216 	if (err < 0)
217 		goto out;
218 
219 	ipip6_tunnel_clone_6rd(dev, sitn);
220 
221 	dev_hold(dev);
222 
223 	ipip6_tunnel_link(sitn, t);
224 	return 0;
225 
226 out:
227 	return err;
228 }
229 
230 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
231 		struct ip_tunnel_parm *parms, int create)
232 {
233 	__be32 remote = parms->iph.daddr;
234 	__be32 local = parms->iph.saddr;
235 	struct ip_tunnel *t, *nt;
236 	struct ip_tunnel __rcu **tp;
237 	struct net_device *dev;
238 	char name[IFNAMSIZ];
239 	struct sit_net *sitn = net_generic(net, sit_net_id);
240 
241 	for (tp = __ipip6_bucket(sitn, parms);
242 	    (t = rtnl_dereference(*tp)) != NULL;
243 	     tp = &t->next) {
244 		if (local == t->parms.iph.saddr &&
245 		    remote == t->parms.iph.daddr &&
246 		    parms->link == t->parms.link) {
247 			if (create)
248 				return NULL;
249 			else
250 				return t;
251 		}
252 	}
253 	if (!create)
254 		goto failed;
255 
256 	if (parms->name[0]) {
257 		if (!dev_valid_name(parms->name))
258 			goto failed;
259 		strlcpy(name, parms->name, IFNAMSIZ);
260 	} else {
261 		strcpy(name, "sit%d");
262 	}
263 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
264 			   ipip6_tunnel_setup);
265 	if (!dev)
266 		return NULL;
267 
268 	dev_net_set(dev, net);
269 
270 	nt = netdev_priv(dev);
271 
272 	nt->parms = *parms;
273 	if (ipip6_tunnel_create(dev) < 0)
274 		goto failed_free;
275 
276 	return nt;
277 
278 failed_free:
279 	free_netdev(dev);
280 failed:
281 	return NULL;
282 }
283 
284 #define for_each_prl_rcu(start)			\
285 	for (prl = rcu_dereference(start);	\
286 	     prl;				\
287 	     prl = rcu_dereference(prl->next))
288 
289 static struct ip_tunnel_prl_entry *
290 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
291 {
292 	struct ip_tunnel_prl_entry *prl;
293 
294 	for_each_prl_rcu(t->prl)
295 		if (prl->addr == addr)
296 			break;
297 	return prl;
298 
299 }
300 
301 static int ipip6_tunnel_get_prl(struct net_device *dev, struct ifreq *ifr)
302 {
303 	struct ip_tunnel_prl __user *a = ifr->ifr_ifru.ifru_data;
304 	struct ip_tunnel *t = netdev_priv(dev);
305 	struct ip_tunnel_prl kprl, *kp;
306 	struct ip_tunnel_prl_entry *prl;
307 	unsigned int cmax, c = 0, ca, len;
308 	int ret = 0;
309 
310 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
311 		return -EINVAL;
312 
313 	if (copy_from_user(&kprl, a, sizeof(kprl)))
314 		return -EFAULT;
315 	cmax = kprl.datalen / sizeof(kprl);
316 	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
317 		cmax = 1;
318 
319 	/* For simple GET or for root users,
320 	 * we try harder to allocate.
321 	 */
322 	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
323 		kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
324 		NULL;
325 
326 	rcu_read_lock();
327 
328 	ca = t->prl_count < cmax ? 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);
337 		if (!kp) {
338 			ret = -ENOMEM;
339 			goto out;
340 		}
341 	}
342 
343 	c = 0;
344 	for_each_prl_rcu(t->prl) {
345 		if (c >= cmax)
346 			break;
347 		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
348 			continue;
349 		kp[c].addr = prl->addr;
350 		kp[c].flags = prl->flags;
351 		c++;
352 		if (kprl.addr != htonl(INADDR_ANY))
353 			break;
354 	}
355 out:
356 	rcu_read_unlock();
357 
358 	len = sizeof(*kp) * c;
359 	ret = 0;
360 	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
361 		ret = -EFAULT;
362 
363 	kfree(kp);
364 
365 	return ret;
366 }
367 
368 static int
369 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
370 {
371 	struct ip_tunnel_prl_entry *p;
372 	int err = 0;
373 
374 	if (a->addr == htonl(INADDR_ANY))
375 		return -EINVAL;
376 
377 	ASSERT_RTNL();
378 
379 	for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
380 		if (p->addr == a->addr) {
381 			if (chg) {
382 				p->flags = a->flags;
383 				goto out;
384 			}
385 			err = -EEXIST;
386 			goto out;
387 		}
388 	}
389 
390 	if (chg) {
391 		err = -ENXIO;
392 		goto out;
393 	}
394 
395 	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
396 	if (!p) {
397 		err = -ENOBUFS;
398 		goto out;
399 	}
400 
401 	p->next = t->prl;
402 	p->addr = a->addr;
403 	p->flags = a->flags;
404 	t->prl_count++;
405 	rcu_assign_pointer(t->prl, p);
406 out:
407 	return err;
408 }
409 
410 static void prl_list_destroy_rcu(struct rcu_head *head)
411 {
412 	struct ip_tunnel_prl_entry *p, *n;
413 
414 	p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
415 	do {
416 		n = rcu_dereference_protected(p->next, 1);
417 		kfree(p);
418 		p = n;
419 	} while (p);
420 }
421 
422 static int
423 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
424 {
425 	struct ip_tunnel_prl_entry *x;
426 	struct ip_tunnel_prl_entry __rcu **p;
427 	int err = 0;
428 
429 	ASSERT_RTNL();
430 
431 	if (a && a->addr != htonl(INADDR_ANY)) {
432 		for (p = &t->prl;
433 		     (x = rtnl_dereference(*p)) != NULL;
434 		     p = &x->next) {
435 			if (x->addr == a->addr) {
436 				*p = x->next;
437 				kfree_rcu(x, rcu_head);
438 				t->prl_count--;
439 				goto out;
440 			}
441 		}
442 		err = -ENXIO;
443 	} else {
444 		x = rtnl_dereference(t->prl);
445 		if (x) {
446 			t->prl_count = 0;
447 			call_rcu(&x->rcu_head, prl_list_destroy_rcu);
448 			t->prl = NULL;
449 		}
450 	}
451 out:
452 	return err;
453 }
454 
455 static int ipip6_tunnel_prl_ctl(struct net_device *dev, struct ifreq *ifr,
456 		int cmd)
457 {
458 	struct ip_tunnel *t = netdev_priv(dev);
459 	struct ip_tunnel_prl prl;
460 	int err;
461 
462 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
463 		return -EPERM;
464 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
465 		return -EINVAL;
466 
467 	if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
468 		return -EFAULT;
469 
470 	switch (cmd) {
471 	case SIOCDELPRL:
472 		err = ipip6_tunnel_del_prl(t, &prl);
473 		break;
474 	case SIOCADDPRL:
475 	case SIOCCHGPRL:
476 		err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
477 		break;
478 	}
479 	dst_cache_reset(&t->dst_cache);
480 	netdev_state_change(dev);
481 	return err;
482 }
483 
484 static int
485 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
486 {
487 	struct ip_tunnel_prl_entry *p;
488 	int ok = 1;
489 
490 	rcu_read_lock();
491 	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
492 	if (p) {
493 		if (p->flags & PRL_DEFAULT)
494 			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
495 		else
496 			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
497 	} else {
498 		const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
499 
500 		if (ipv6_addr_is_isatap(addr6) &&
501 		    (addr6->s6_addr32[3] == iph->saddr) &&
502 		    ipv6_chk_prefix(addr6, t->dev))
503 			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
504 		else
505 			ok = 0;
506 	}
507 	rcu_read_unlock();
508 	return ok;
509 }
510 
511 static void ipip6_tunnel_uninit(struct net_device *dev)
512 {
513 	struct ip_tunnel *tunnel = netdev_priv(dev);
514 	struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
515 
516 	if (dev == sitn->fb_tunnel_dev) {
517 		RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
518 	} else {
519 		ipip6_tunnel_unlink(sitn, tunnel);
520 		ipip6_tunnel_del_prl(tunnel, NULL);
521 	}
522 	dst_cache_reset(&tunnel->dst_cache);
523 	dev_put(dev);
524 }
525 
526 static int ipip6_err(struct sk_buff *skb, u32 info)
527 {
528 	const struct iphdr *iph = (const struct iphdr *)skb->data;
529 	const int type = icmp_hdr(skb)->type;
530 	const int code = icmp_hdr(skb)->code;
531 	unsigned int data_len = 0;
532 	struct ip_tunnel *t;
533 	int sifindex;
534 	int err;
535 
536 	switch (type) {
537 	default:
538 	case ICMP_PARAMETERPROB:
539 		return 0;
540 
541 	case ICMP_DEST_UNREACH:
542 		switch (code) {
543 		case ICMP_SR_FAILED:
544 			/* Impossible event. */
545 			return 0;
546 		default:
547 			/* All others are translated to HOST_UNREACH.
548 			   rfc2003 contains "deep thoughts" about NET_UNREACH,
549 			   I believe they are just ether pollution. --ANK
550 			 */
551 			break;
552 		}
553 		break;
554 	case ICMP_TIME_EXCEEDED:
555 		if (code != ICMP_EXC_TTL)
556 			return 0;
557 		data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
558 		break;
559 	case ICMP_REDIRECT:
560 		break;
561 	}
562 
563 	err = -ENOENT;
564 
565 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
566 	t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
567 				iph->daddr, iph->saddr, sifindex);
568 	if (!t)
569 		goto out;
570 
571 	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
572 		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
573 				 t->parms.link, iph->protocol);
574 		err = 0;
575 		goto out;
576 	}
577 	if (type == ICMP_REDIRECT) {
578 		ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
579 			      iph->protocol);
580 		err = 0;
581 		goto out;
582 	}
583 
584 	err = 0;
585 	if (__in6_dev_get(skb->dev) &&
586 	    !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
587 		goto out;
588 
589 	if (t->parms.iph.daddr == 0)
590 		goto out;
591 
592 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
593 		goto out;
594 
595 	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
596 		t->err_count++;
597 	else
598 		t->err_count = 1;
599 	t->err_time = jiffies;
600 out:
601 	return err;
602 }
603 
604 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
605 				  const struct in6_addr *v6addr)
606 {
607 	__be32 v4embed = 0;
608 	if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
609 		return true;
610 	return false;
611 }
612 
613 /* Checks if an address matches an address on the tunnel interface.
614  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
615  * Long story:
616  * This function is called after we considered the packet as spoofed
617  * in is_spoofed_6rd.
618  * We may have a router that is doing NAT for proto 41 packets
619  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
620  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
621  * function will return true, dropping the packet.
622  * But, we can still check if is spoofed against the IP
623  * addresses associated with the interface.
624  */
625 static bool only_dnatted(const struct ip_tunnel *tunnel,
626 	const struct in6_addr *v6dst)
627 {
628 	int prefix_len;
629 
630 #ifdef CONFIG_IPV6_SIT_6RD
631 	prefix_len = tunnel->ip6rd.prefixlen + 32
632 		- tunnel->ip6rd.relay_prefixlen;
633 #else
634 	prefix_len = 48;
635 #endif
636 	return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
637 }
638 
639 /* Returns true if a packet is spoofed */
640 static bool packet_is_spoofed(struct sk_buff *skb,
641 			      const struct iphdr *iph,
642 			      struct ip_tunnel *tunnel)
643 {
644 	const struct ipv6hdr *ipv6h;
645 
646 	if (tunnel->dev->priv_flags & IFF_ISATAP) {
647 		if (!isatap_chksrc(skb, iph, tunnel))
648 			return true;
649 
650 		return false;
651 	}
652 
653 	if (tunnel->dev->flags & IFF_POINTOPOINT)
654 		return false;
655 
656 	ipv6h = ipv6_hdr(skb);
657 
658 	if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
659 		net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
660 				     &iph->saddr, &ipv6h->saddr,
661 				     &iph->daddr, &ipv6h->daddr);
662 		return true;
663 	}
664 
665 	if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
666 		return false;
667 
668 	if (only_dnatted(tunnel, &ipv6h->daddr))
669 		return false;
670 
671 	net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
672 			     &iph->saddr, &ipv6h->saddr,
673 			     &iph->daddr, &ipv6h->daddr);
674 	return true;
675 }
676 
677 static int ipip6_rcv(struct sk_buff *skb)
678 {
679 	const struct iphdr *iph = ip_hdr(skb);
680 	struct ip_tunnel *tunnel;
681 	int sifindex;
682 	int err;
683 
684 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
685 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
686 				     iph->saddr, iph->daddr, sifindex);
687 	if (tunnel) {
688 		struct pcpu_sw_netstats *tstats;
689 
690 		if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
691 		    tunnel->parms.iph.protocol != 0)
692 			goto out;
693 
694 		skb->mac_header = skb->network_header;
695 		skb_reset_network_header(skb);
696 		IPCB(skb)->flags = 0;
697 		skb->dev = tunnel->dev;
698 
699 		if (packet_is_spoofed(skb, iph, tunnel)) {
700 			tunnel->dev->stats.rx_errors++;
701 			goto out;
702 		}
703 
704 		if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
705 		    !net_eq(tunnel->net, dev_net(tunnel->dev))))
706 			goto out;
707 
708 		/* skb can be uncloned in iptunnel_pull_header, so
709 		 * old iph is no longer valid
710 		 */
711 		iph = (const struct iphdr *)skb_mac_header(skb);
712 		err = IP_ECN_decapsulate(iph, skb);
713 		if (unlikely(err)) {
714 			if (log_ecn_error)
715 				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
716 						     &iph->saddr, iph->tos);
717 			if (err > 1) {
718 				++tunnel->dev->stats.rx_frame_errors;
719 				++tunnel->dev->stats.rx_errors;
720 				goto out;
721 			}
722 		}
723 
724 		tstats = this_cpu_ptr(tunnel->dev->tstats);
725 		u64_stats_update_begin(&tstats->syncp);
726 		tstats->rx_packets++;
727 		tstats->rx_bytes += skb->len;
728 		u64_stats_update_end(&tstats->syncp);
729 
730 		netif_rx(skb);
731 
732 		return 0;
733 	}
734 
735 	/* no tunnel matched,  let upstream know, ipsec may handle it */
736 	return 1;
737 out:
738 	kfree_skb(skb);
739 	return 0;
740 }
741 
742 static const struct tnl_ptk_info ipip_tpi = {
743 	/* no tunnel info required for ipip. */
744 	.proto = htons(ETH_P_IP),
745 };
746 
747 #if IS_ENABLED(CONFIG_MPLS)
748 static const struct tnl_ptk_info mplsip_tpi = {
749 	/* no tunnel info required for mplsip. */
750 	.proto = htons(ETH_P_MPLS_UC),
751 };
752 #endif
753 
754 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
755 {
756 	const struct iphdr *iph;
757 	struct ip_tunnel *tunnel;
758 	int sifindex;
759 
760 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
761 
762 	iph = ip_hdr(skb);
763 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
764 				     iph->saddr, iph->daddr, sifindex);
765 	if (tunnel) {
766 		const struct tnl_ptk_info *tpi;
767 
768 		if (tunnel->parms.iph.protocol != ipproto &&
769 		    tunnel->parms.iph.protocol != 0)
770 			goto drop;
771 
772 		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
773 			goto drop;
774 #if IS_ENABLED(CONFIG_MPLS)
775 		if (ipproto == IPPROTO_MPLS)
776 			tpi = &mplsip_tpi;
777 		else
778 #endif
779 			tpi = &ipip_tpi;
780 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
781 			goto drop;
782 		return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
783 	}
784 
785 	return 1;
786 
787 drop:
788 	kfree_skb(skb);
789 	return 0;
790 }
791 
792 static int ipip_rcv(struct sk_buff *skb)
793 {
794 	return sit_tunnel_rcv(skb, IPPROTO_IPIP);
795 }
796 
797 #if IS_ENABLED(CONFIG_MPLS)
798 static int mplsip_rcv(struct sk_buff *skb)
799 {
800 	return sit_tunnel_rcv(skb, IPPROTO_MPLS);
801 }
802 #endif
803 
804 /*
805  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
806  * stores the embedded IPv4 address in v4dst and returns true.
807  */
808 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
809 		      __be32 *v4dst)
810 {
811 #ifdef CONFIG_IPV6_SIT_6RD
812 	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
813 			      tunnel->ip6rd.prefixlen)) {
814 		unsigned int pbw0, pbi0;
815 		int pbi1;
816 		u32 d;
817 
818 		pbw0 = tunnel->ip6rd.prefixlen >> 5;
819 		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
820 
821 		d = tunnel->ip6rd.relay_prefixlen < 32 ?
822 			(ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
823 		    tunnel->ip6rd.relay_prefixlen : 0;
824 
825 		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
826 		if (pbi1 > 0)
827 			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
828 			     (32 - pbi1);
829 
830 		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
831 		return true;
832 	}
833 #else
834 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
835 		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
836 		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
837 		return true;
838 	}
839 #endif
840 	return false;
841 }
842 
843 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
844 			     const struct in6_addr *v6dst)
845 {
846 	__be32 dst = 0;
847 	check_6rd(tunnel, v6dst, &dst);
848 	return dst;
849 }
850 
851 /*
852  *	This function assumes it is being called from dev_queue_xmit()
853  *	and that skb is filled properly by that function.
854  */
855 
856 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
857 				     struct net_device *dev)
858 {
859 	struct ip_tunnel *tunnel = netdev_priv(dev);
860 	const struct iphdr  *tiph = &tunnel->parms.iph;
861 	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
862 	u8     tos = tunnel->parms.iph.tos;
863 	__be16 df = tiph->frag_off;
864 	struct rtable *rt;		/* Route to the other host */
865 	struct net_device *tdev;	/* Device to other host */
866 	unsigned int max_headroom;	/* The extra header space needed */
867 	__be32 dst = tiph->daddr;
868 	struct flowi4 fl4;
869 	int    mtu;
870 	const struct in6_addr *addr6;
871 	int addr_type;
872 	u8 ttl;
873 	u8 protocol = IPPROTO_IPV6;
874 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
875 
876 	if (tos == 1)
877 		tos = ipv6_get_dsfield(iph6);
878 
879 	/* ISATAP (RFC4214) - must come before 6to4 */
880 	if (dev->priv_flags & IFF_ISATAP) {
881 		struct neighbour *neigh = NULL;
882 		bool do_tx_error = false;
883 
884 		if (skb_dst(skb))
885 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
886 
887 		if (!neigh) {
888 			net_dbg_ratelimited("nexthop == NULL\n");
889 			goto tx_error;
890 		}
891 
892 		addr6 = (const struct in6_addr *)&neigh->primary_key;
893 		addr_type = ipv6_addr_type(addr6);
894 
895 		if ((addr_type & IPV6_ADDR_UNICAST) &&
896 		     ipv6_addr_is_isatap(addr6))
897 			dst = addr6->s6_addr32[3];
898 		else
899 			do_tx_error = true;
900 
901 		neigh_release(neigh);
902 		if (do_tx_error)
903 			goto tx_error;
904 	}
905 
906 	if (!dst)
907 		dst = try_6rd(tunnel, &iph6->daddr);
908 
909 	if (!dst) {
910 		struct neighbour *neigh = NULL;
911 		bool do_tx_error = false;
912 
913 		if (skb_dst(skb))
914 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
915 
916 		if (!neigh) {
917 			net_dbg_ratelimited("nexthop == NULL\n");
918 			goto tx_error;
919 		}
920 
921 		addr6 = (const struct in6_addr *)&neigh->primary_key;
922 		addr_type = ipv6_addr_type(addr6);
923 
924 		if (addr_type == IPV6_ADDR_ANY) {
925 			addr6 = &ipv6_hdr(skb)->daddr;
926 			addr_type = ipv6_addr_type(addr6);
927 		}
928 
929 		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
930 			dst = addr6->s6_addr32[3];
931 		else
932 			do_tx_error = true;
933 
934 		neigh_release(neigh);
935 		if (do_tx_error)
936 			goto tx_error;
937 	}
938 
939 	flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
940 			   RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
941 			   0, dst, tiph->saddr, 0, 0,
942 			   sock_net_uid(tunnel->net, NULL));
943 
944 	rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
945 	if (!rt) {
946 		rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
947 		if (IS_ERR(rt)) {
948 			dev->stats.tx_carrier_errors++;
949 			goto tx_error_icmp;
950 		}
951 		dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
952 	}
953 
954 	if (rt->rt_type != RTN_UNICAST) {
955 		ip_rt_put(rt);
956 		dev->stats.tx_carrier_errors++;
957 		goto tx_error_icmp;
958 	}
959 	tdev = rt->dst.dev;
960 
961 	if (tdev == dev) {
962 		ip_rt_put(rt);
963 		dev->stats.collisions++;
964 		goto tx_error;
965 	}
966 
967 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
968 		ip_rt_put(rt);
969 		goto tx_error;
970 	}
971 
972 	if (df) {
973 		mtu = dst_mtu(&rt->dst) - t_hlen;
974 
975 		if (mtu < 68) {
976 			dev->stats.collisions++;
977 			ip_rt_put(rt);
978 			goto tx_error;
979 		}
980 
981 		if (mtu < IPV6_MIN_MTU) {
982 			mtu = IPV6_MIN_MTU;
983 			df = 0;
984 		}
985 
986 		if (tunnel->parms.iph.daddr)
987 			skb_dst_update_pmtu_no_confirm(skb, mtu);
988 
989 		if (skb->len > mtu && !skb_is_gso(skb)) {
990 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
991 			ip_rt_put(rt);
992 			goto tx_error;
993 		}
994 	}
995 
996 	if (tunnel->err_count > 0) {
997 		if (time_before(jiffies,
998 				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
999 			tunnel->err_count--;
1000 			dst_link_failure(skb);
1001 		} else
1002 			tunnel->err_count = 0;
1003 	}
1004 
1005 	/*
1006 	 * Okay, now see if we can stuff it in the buffer as-is.
1007 	 */
1008 	max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
1009 
1010 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1011 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1012 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
1013 		if (!new_skb) {
1014 			ip_rt_put(rt);
1015 			dev->stats.tx_dropped++;
1016 			kfree_skb(skb);
1017 			return NETDEV_TX_OK;
1018 		}
1019 		if (skb->sk)
1020 			skb_set_owner_w(new_skb, skb->sk);
1021 		dev_kfree_skb(skb);
1022 		skb = new_skb;
1023 		iph6 = ipv6_hdr(skb);
1024 	}
1025 	ttl = tiph->ttl;
1026 	if (ttl == 0)
1027 		ttl = iph6->hop_limit;
1028 	tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
1029 
1030 	if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
1031 		ip_rt_put(rt);
1032 		goto tx_error;
1033 	}
1034 
1035 	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1036 
1037 	iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
1038 		      df, !net_eq(tunnel->net, dev_net(dev)));
1039 	return NETDEV_TX_OK;
1040 
1041 tx_error_icmp:
1042 	dst_link_failure(skb);
1043 tx_error:
1044 	kfree_skb(skb);
1045 	dev->stats.tx_errors++;
1046 	return NETDEV_TX_OK;
1047 }
1048 
1049 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1050 				     struct net_device *dev, u8 ipproto)
1051 {
1052 	struct ip_tunnel *tunnel = netdev_priv(dev);
1053 	const struct iphdr  *tiph = &tunnel->parms.iph;
1054 
1055 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1056 		goto tx_error;
1057 
1058 	skb_set_inner_ipproto(skb, ipproto);
1059 
1060 	ip_tunnel_xmit(skb, dev, tiph, ipproto);
1061 	return NETDEV_TX_OK;
1062 tx_error:
1063 	kfree_skb(skb);
1064 	dev->stats.tx_errors++;
1065 	return NETDEV_TX_OK;
1066 }
1067 
1068 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1069 				   struct net_device *dev)
1070 {
1071 	if (!pskb_inet_may_pull(skb))
1072 		goto tx_err;
1073 
1074 	switch (skb->protocol) {
1075 	case htons(ETH_P_IP):
1076 		sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1077 		break;
1078 	case htons(ETH_P_IPV6):
1079 		ipip6_tunnel_xmit(skb, dev);
1080 		break;
1081 #if IS_ENABLED(CONFIG_MPLS)
1082 	case htons(ETH_P_MPLS_UC):
1083 		sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1084 		break;
1085 #endif
1086 	default:
1087 		goto tx_err;
1088 	}
1089 
1090 	return NETDEV_TX_OK;
1091 
1092 tx_err:
1093 	dev->stats.tx_errors++;
1094 	kfree_skb(skb);
1095 	return NETDEV_TX_OK;
1096 
1097 }
1098 
1099 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1100 {
1101 	struct net_device *tdev = NULL;
1102 	struct ip_tunnel *tunnel;
1103 	const struct iphdr *iph;
1104 	struct flowi4 fl4;
1105 
1106 	tunnel = netdev_priv(dev);
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 							  RT_TOS(iph->tos),
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 t_hlen = tunnel->hlen + sizeof(struct iphdr);
1130 
1131 		dev->mtu = tdev->mtu - t_hlen;
1132 		if (dev->mtu < IPV6_MIN_MTU)
1133 			dev->mtu = IPV6_MIN_MTU;
1134 	}
1135 }
1136 
1137 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1138 				__u32 fwmark)
1139 {
1140 	struct net *net = t->net;
1141 	struct sit_net *sitn = net_generic(net, sit_net_id);
1142 
1143 	ipip6_tunnel_unlink(sitn, t);
1144 	synchronize_net();
1145 	t->parms.iph.saddr = p->iph.saddr;
1146 	t->parms.iph.daddr = p->iph.daddr;
1147 	memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1148 	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1149 	ipip6_tunnel_link(sitn, t);
1150 	t->parms.iph.ttl = p->iph.ttl;
1151 	t->parms.iph.tos = p->iph.tos;
1152 	t->parms.iph.frag_off = p->iph.frag_off;
1153 	if (t->parms.link != p->link || t->fwmark != fwmark) {
1154 		t->parms.link = p->link;
1155 		t->fwmark = fwmark;
1156 		ipip6_tunnel_bind_dev(t->dev);
1157 	}
1158 	dst_cache_reset(&t->dst_cache);
1159 	netdev_state_change(t->dev);
1160 }
1161 
1162 #ifdef CONFIG_IPV6_SIT_6RD
1163 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1164 				   struct ip_tunnel_6rd *ip6rd)
1165 {
1166 	struct in6_addr prefix;
1167 	__be32 relay_prefix;
1168 
1169 	if (ip6rd->relay_prefixlen > 32 ||
1170 	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1171 		return -EINVAL;
1172 
1173 	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1174 	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1175 		return -EINVAL;
1176 	if (ip6rd->relay_prefixlen)
1177 		relay_prefix = ip6rd->relay_prefix &
1178 			       htonl(0xffffffffUL <<
1179 				     (32 - ip6rd->relay_prefixlen));
1180 	else
1181 		relay_prefix = 0;
1182 	if (relay_prefix != ip6rd->relay_prefix)
1183 		return -EINVAL;
1184 
1185 	t->ip6rd.prefix = prefix;
1186 	t->ip6rd.relay_prefix = relay_prefix;
1187 	t->ip6rd.prefixlen = ip6rd->prefixlen;
1188 	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1189 	dst_cache_reset(&t->dst_cache);
1190 	netdev_state_change(t->dev);
1191 	return 0;
1192 }
1193 
1194 static int
1195 ipip6_tunnel_get6rd(struct net_device *dev, struct ifreq *ifr)
1196 {
1197 	struct ip_tunnel *t = netdev_priv(dev);
1198 	struct ip_tunnel_6rd ip6rd;
1199 	struct ip_tunnel_parm p;
1200 
1201 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1202 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1203 			return -EFAULT;
1204 		t = ipip6_tunnel_locate(t->net, &p, 0);
1205 	}
1206 	if (!t)
1207 		t = netdev_priv(dev);
1208 
1209 	ip6rd.prefix = t->ip6rd.prefix;
1210 	ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1211 	ip6rd.prefixlen = t->ip6rd.prefixlen;
1212 	ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1213 	if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd, sizeof(ip6rd)))
1214 		return -EFAULT;
1215 	return 0;
1216 }
1217 
1218 static int
1219 ipip6_tunnel_6rdctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1220 {
1221 	struct ip_tunnel *t = netdev_priv(dev);
1222 	struct ip_tunnel_6rd ip6rd;
1223 	int err;
1224 
1225 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1226 		return -EPERM;
1227 	if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data, sizeof(ip6rd)))
1228 		return -EFAULT;
1229 
1230 	if (cmd != SIOCDEL6RD) {
1231 		err = ipip6_tunnel_update_6rd(t, &ip6rd);
1232 		if (err < 0)
1233 			return err;
1234 	} else
1235 		ipip6_tunnel_clone_6rd(dev, dev_to_sit_net(dev));
1236 	return 0;
1237 }
1238 
1239 #endif /* CONFIG_IPV6_SIT_6RD */
1240 
1241 static bool ipip6_valid_ip_proto(u8 ipproto)
1242 {
1243 	return ipproto == IPPROTO_IPV6 ||
1244 		ipproto == IPPROTO_IPIP ||
1245 #if IS_ENABLED(CONFIG_MPLS)
1246 		ipproto == IPPROTO_MPLS ||
1247 #endif
1248 		ipproto == 0;
1249 }
1250 
1251 static int
1252 __ipip6_tunnel_ioctl_validate(struct net *net, struct ip_tunnel_parm *p)
1253 {
1254 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1255 		return -EPERM;
1256 
1257 	if (!ipip6_valid_ip_proto(p->iph.protocol))
1258 		return -EINVAL;
1259 	if (p->iph.version != 4 ||
1260 	    p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
1261 		return -EINVAL;
1262 
1263 	if (p->iph.ttl)
1264 		p->iph.frag_off |= htons(IP_DF);
1265 	return 0;
1266 }
1267 
1268 static int
1269 ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm *p)
1270 {
1271 	struct ip_tunnel *t = netdev_priv(dev);
1272 
1273 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
1274 		t = ipip6_tunnel_locate(t->net, p, 0);
1275 	if (!t)
1276 		t = netdev_priv(dev);
1277 	memcpy(p, &t->parms, sizeof(*p));
1278 	return 0;
1279 }
1280 
1281 static int
1282 ipip6_tunnel_add(struct net_device *dev, struct ip_tunnel_parm *p)
1283 {
1284 	struct ip_tunnel *t = netdev_priv(dev);
1285 	int err;
1286 
1287 	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1288 	if (err)
1289 		return err;
1290 
1291 	t = ipip6_tunnel_locate(t->net, p, 1);
1292 	if (!t)
1293 		return -ENOBUFS;
1294 	return 0;
1295 }
1296 
1297 static int
1298 ipip6_tunnel_change(struct net_device *dev, struct ip_tunnel_parm *p)
1299 {
1300 	struct ip_tunnel *t = netdev_priv(dev);
1301 	int err;
1302 
1303 	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1304 	if (err)
1305 		return err;
1306 
1307 	t = ipip6_tunnel_locate(t->net, p, 0);
1308 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1309 		if (!t)
1310 			return -ENOENT;
1311 	} else {
1312 		if (t) {
1313 			if (t->dev != dev)
1314 				return -EEXIST;
1315 		} else {
1316 			if (((dev->flags & IFF_POINTOPOINT) && !p->iph.daddr) ||
1317 			    (!(dev->flags & IFF_POINTOPOINT) && p->iph.daddr))
1318 				return -EINVAL;
1319 			t = netdev_priv(dev);
1320 		}
1321 
1322 		ipip6_tunnel_update(t, p, t->fwmark);
1323 	}
1324 
1325 	return 0;
1326 }
1327 
1328 static int
1329 ipip6_tunnel_del(struct net_device *dev, struct ip_tunnel_parm *p)
1330 {
1331 	struct ip_tunnel *t = netdev_priv(dev);
1332 
1333 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1334 		return -EPERM;
1335 
1336 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1337 		t = ipip6_tunnel_locate(t->net, p, 0);
1338 		if (!t)
1339 			return -ENOENT;
1340 		if (t == netdev_priv(dev_to_sit_net(dev)->fb_tunnel_dev))
1341 			return -EPERM;
1342 		dev = t->dev;
1343 	}
1344 	unregister_netdevice(dev);
1345 	return 0;
1346 }
1347 
1348 static int
1349 ipip6_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
1350 {
1351 	switch (cmd) {
1352 	case SIOCGETTUNNEL:
1353 		return ipip6_tunnel_get(dev, p);
1354 	case SIOCADDTUNNEL:
1355 		return ipip6_tunnel_add(dev, p);
1356 	case SIOCCHGTUNNEL:
1357 		return ipip6_tunnel_change(dev, p);
1358 	case SIOCDELTUNNEL:
1359 		return ipip6_tunnel_del(dev, p);
1360 	default:
1361 		return -EINVAL;
1362 	}
1363 }
1364 
1365 static int
1366 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1367 {
1368 	switch (cmd) {
1369 	case SIOCGETTUNNEL:
1370 	case SIOCADDTUNNEL:
1371 	case SIOCCHGTUNNEL:
1372 	case SIOCDELTUNNEL:
1373 		return ip_tunnel_ioctl(dev, ifr, cmd);
1374 	case SIOCGETPRL:
1375 		return ipip6_tunnel_get_prl(dev, ifr);
1376 	case SIOCADDPRL:
1377 	case SIOCDELPRL:
1378 	case SIOCCHGPRL:
1379 		return ipip6_tunnel_prl_ctl(dev, ifr, cmd);
1380 #ifdef CONFIG_IPV6_SIT_6RD
1381 	case SIOCGET6RD:
1382 		return ipip6_tunnel_get6rd(dev, ifr);
1383 	case SIOCADD6RD:
1384 	case SIOCCHG6RD:
1385 	case SIOCDEL6RD:
1386 		return ipip6_tunnel_6rdctl(dev, ifr, cmd);
1387 #endif
1388 	default:
1389 		return -EINVAL;
1390 	}
1391 }
1392 
1393 static const struct net_device_ops ipip6_netdev_ops = {
1394 	.ndo_init	= ipip6_tunnel_init,
1395 	.ndo_uninit	= ipip6_tunnel_uninit,
1396 	.ndo_start_xmit	= sit_tunnel_xmit,
1397 	.ndo_do_ioctl	= ipip6_tunnel_ioctl,
1398 	.ndo_get_stats64 = ip_tunnel_get_stats64,
1399 	.ndo_get_iflink = ip_tunnel_get_iflink,
1400 	.ndo_tunnel_ctl = ipip6_tunnel_ctl,
1401 };
1402 
1403 static void ipip6_dev_free(struct net_device *dev)
1404 {
1405 	struct ip_tunnel *tunnel = netdev_priv(dev);
1406 
1407 	dst_cache_destroy(&tunnel->dst_cache);
1408 	free_percpu(dev->tstats);
1409 }
1410 
1411 #define SIT_FEATURES (NETIF_F_SG	   | \
1412 		      NETIF_F_FRAGLIST	   | \
1413 		      NETIF_F_HIGHDMA	   | \
1414 		      NETIF_F_GSO_SOFTWARE | \
1415 		      NETIF_F_HW_CSUM)
1416 
1417 static void ipip6_tunnel_setup(struct net_device *dev)
1418 {
1419 	struct ip_tunnel *tunnel = netdev_priv(dev);
1420 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1421 
1422 	dev->netdev_ops		= &ipip6_netdev_ops;
1423 	dev->header_ops		= &ip_tunnel_header_ops;
1424 	dev->needs_free_netdev	= true;
1425 	dev->priv_destructor	= ipip6_dev_free;
1426 
1427 	dev->type		= ARPHRD_SIT;
1428 	dev->mtu		= ETH_DATA_LEN - t_hlen;
1429 	dev->min_mtu		= IPV6_MIN_MTU;
1430 	dev->max_mtu		= IP6_MAX_MTU - t_hlen;
1431 	dev->flags		= IFF_NOARP;
1432 	netif_keep_dst(dev);
1433 	dev->addr_len		= 4;
1434 	dev->features		|= NETIF_F_LLTX;
1435 	dev->features		|= SIT_FEATURES;
1436 	dev->hw_features	|= SIT_FEATURES;
1437 }
1438 
1439 static int ipip6_tunnel_init(struct net_device *dev)
1440 {
1441 	struct ip_tunnel *tunnel = netdev_priv(dev);
1442 	int err;
1443 
1444 	tunnel->dev = dev;
1445 	tunnel->net = dev_net(dev);
1446 	strcpy(tunnel->parms.name, dev->name);
1447 
1448 	ipip6_tunnel_bind_dev(dev);
1449 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1450 	if (!dev->tstats)
1451 		return -ENOMEM;
1452 
1453 	err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1454 	if (err) {
1455 		free_percpu(dev->tstats);
1456 		dev->tstats = NULL;
1457 		return err;
1458 	}
1459 
1460 	return 0;
1461 }
1462 
1463 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1464 {
1465 	struct ip_tunnel *tunnel = netdev_priv(dev);
1466 	struct iphdr *iph = &tunnel->parms.iph;
1467 	struct net *net = dev_net(dev);
1468 	struct sit_net *sitn = net_generic(net, sit_net_id);
1469 
1470 	iph->version		= 4;
1471 	iph->protocol		= IPPROTO_IPV6;
1472 	iph->ihl		= 5;
1473 	iph->ttl		= 64;
1474 
1475 	dev_hold(dev);
1476 	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1477 }
1478 
1479 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1480 			  struct netlink_ext_ack *extack)
1481 {
1482 	u8 proto;
1483 
1484 	if (!data || !data[IFLA_IPTUN_PROTO])
1485 		return 0;
1486 
1487 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1488 	if (!ipip6_valid_ip_proto(proto))
1489 		return -EINVAL;
1490 
1491 	return 0;
1492 }
1493 
1494 static void ipip6_netlink_parms(struct nlattr *data[],
1495 				struct ip_tunnel_parm *parms,
1496 				__u32 *fwmark)
1497 {
1498 	memset(parms, 0, sizeof(*parms));
1499 
1500 	parms->iph.version = 4;
1501 	parms->iph.protocol = IPPROTO_IPV6;
1502 	parms->iph.ihl = 5;
1503 	parms->iph.ttl = 64;
1504 
1505 	if (!data)
1506 		return;
1507 
1508 	if (data[IFLA_IPTUN_LINK])
1509 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1510 
1511 	if (data[IFLA_IPTUN_LOCAL])
1512 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1513 
1514 	if (data[IFLA_IPTUN_REMOTE])
1515 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1516 
1517 	if (data[IFLA_IPTUN_TTL]) {
1518 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1519 		if (parms->iph.ttl)
1520 			parms->iph.frag_off = htons(IP_DF);
1521 	}
1522 
1523 	if (data[IFLA_IPTUN_TOS])
1524 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1525 
1526 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1527 		parms->iph.frag_off = htons(IP_DF);
1528 
1529 	if (data[IFLA_IPTUN_FLAGS])
1530 		parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1531 
1532 	if (data[IFLA_IPTUN_PROTO])
1533 		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1534 
1535 	if (data[IFLA_IPTUN_FWMARK])
1536 		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1537 }
1538 
1539 /* This function returns true when ENCAP attributes are present in the nl msg */
1540 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1541 				      struct ip_tunnel_encap *ipencap)
1542 {
1543 	bool ret = false;
1544 
1545 	memset(ipencap, 0, sizeof(*ipencap));
1546 
1547 	if (!data)
1548 		return ret;
1549 
1550 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1551 		ret = true;
1552 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1553 	}
1554 
1555 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1556 		ret = true;
1557 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1558 	}
1559 
1560 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1561 		ret = true;
1562 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1563 	}
1564 
1565 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1566 		ret = true;
1567 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1568 	}
1569 
1570 	return ret;
1571 }
1572 
1573 #ifdef CONFIG_IPV6_SIT_6RD
1574 /* This function returns true when 6RD attributes are present in the nl msg */
1575 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1576 				    struct ip_tunnel_6rd *ip6rd)
1577 {
1578 	bool ret = false;
1579 	memset(ip6rd, 0, sizeof(*ip6rd));
1580 
1581 	if (!data)
1582 		return ret;
1583 
1584 	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1585 		ret = true;
1586 		ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1587 	}
1588 
1589 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1590 		ret = true;
1591 		ip6rd->relay_prefix =
1592 			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1593 	}
1594 
1595 	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1596 		ret = true;
1597 		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1598 	}
1599 
1600 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1601 		ret = true;
1602 		ip6rd->relay_prefixlen =
1603 			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1604 	}
1605 
1606 	return ret;
1607 }
1608 #endif
1609 
1610 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1611 			 struct nlattr *tb[], struct nlattr *data[],
1612 			 struct netlink_ext_ack *extack)
1613 {
1614 	struct net *net = dev_net(dev);
1615 	struct ip_tunnel *nt;
1616 	struct ip_tunnel_encap ipencap;
1617 #ifdef CONFIG_IPV6_SIT_6RD
1618 	struct ip_tunnel_6rd ip6rd;
1619 #endif
1620 	int err;
1621 
1622 	nt = netdev_priv(dev);
1623 
1624 	if (ipip6_netlink_encap_parms(data, &ipencap)) {
1625 		err = ip_tunnel_encap_setup(nt, &ipencap);
1626 		if (err < 0)
1627 			return err;
1628 	}
1629 
1630 	ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1631 
1632 	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1633 		return -EEXIST;
1634 
1635 	err = ipip6_tunnel_create(dev);
1636 	if (err < 0)
1637 		return err;
1638 
1639 	if (tb[IFLA_MTU]) {
1640 		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1641 
1642 		if (mtu >= IPV6_MIN_MTU &&
1643 		    mtu <= IP6_MAX_MTU - dev->hard_header_len)
1644 			dev->mtu = mtu;
1645 	}
1646 
1647 #ifdef CONFIG_IPV6_SIT_6RD
1648 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1649 		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1650 #endif
1651 
1652 	return err;
1653 }
1654 
1655 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1656 			    struct nlattr *data[],
1657 			    struct netlink_ext_ack *extack)
1658 {
1659 	struct ip_tunnel *t = netdev_priv(dev);
1660 	struct ip_tunnel_parm p;
1661 	struct ip_tunnel_encap ipencap;
1662 	struct net *net = t->net;
1663 	struct sit_net *sitn = net_generic(net, sit_net_id);
1664 #ifdef CONFIG_IPV6_SIT_6RD
1665 	struct ip_tunnel_6rd ip6rd;
1666 #endif
1667 	__u32 fwmark = t->fwmark;
1668 	int err;
1669 
1670 	if (dev == sitn->fb_tunnel_dev)
1671 		return -EINVAL;
1672 
1673 	if (ipip6_netlink_encap_parms(data, &ipencap)) {
1674 		err = ip_tunnel_encap_setup(t, &ipencap);
1675 		if (err < 0)
1676 			return err;
1677 	}
1678 
1679 	ipip6_netlink_parms(data, &p, &fwmark);
1680 
1681 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1682 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1683 		return -EINVAL;
1684 
1685 	t = ipip6_tunnel_locate(net, &p, 0);
1686 
1687 	if (t) {
1688 		if (t->dev != dev)
1689 			return -EEXIST;
1690 	} else
1691 		t = netdev_priv(dev);
1692 
1693 	ipip6_tunnel_update(t, &p, fwmark);
1694 
1695 #ifdef CONFIG_IPV6_SIT_6RD
1696 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1697 		return ipip6_tunnel_update_6rd(t, &ip6rd);
1698 #endif
1699 
1700 	return 0;
1701 }
1702 
1703 static size_t ipip6_get_size(const struct net_device *dev)
1704 {
1705 	return
1706 		/* IFLA_IPTUN_LINK */
1707 		nla_total_size(4) +
1708 		/* IFLA_IPTUN_LOCAL */
1709 		nla_total_size(4) +
1710 		/* IFLA_IPTUN_REMOTE */
1711 		nla_total_size(4) +
1712 		/* IFLA_IPTUN_TTL */
1713 		nla_total_size(1) +
1714 		/* IFLA_IPTUN_TOS */
1715 		nla_total_size(1) +
1716 		/* IFLA_IPTUN_PMTUDISC */
1717 		nla_total_size(1) +
1718 		/* IFLA_IPTUN_FLAGS */
1719 		nla_total_size(2) +
1720 		/* IFLA_IPTUN_PROTO */
1721 		nla_total_size(1) +
1722 #ifdef CONFIG_IPV6_SIT_6RD
1723 		/* IFLA_IPTUN_6RD_PREFIX */
1724 		nla_total_size(sizeof(struct in6_addr)) +
1725 		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1726 		nla_total_size(4) +
1727 		/* IFLA_IPTUN_6RD_PREFIXLEN */
1728 		nla_total_size(2) +
1729 		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1730 		nla_total_size(2) +
1731 #endif
1732 		/* IFLA_IPTUN_ENCAP_TYPE */
1733 		nla_total_size(2) +
1734 		/* IFLA_IPTUN_ENCAP_FLAGS */
1735 		nla_total_size(2) +
1736 		/* IFLA_IPTUN_ENCAP_SPORT */
1737 		nla_total_size(2) +
1738 		/* IFLA_IPTUN_ENCAP_DPORT */
1739 		nla_total_size(2) +
1740 		/* IFLA_IPTUN_FWMARK */
1741 		nla_total_size(4) +
1742 		0;
1743 }
1744 
1745 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1746 {
1747 	struct ip_tunnel *tunnel = netdev_priv(dev);
1748 	struct ip_tunnel_parm *parm = &tunnel->parms;
1749 
1750 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1751 	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1752 	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1753 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1754 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1755 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1756 		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1757 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1758 	    nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1759 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1760 		goto nla_put_failure;
1761 
1762 #ifdef CONFIG_IPV6_SIT_6RD
1763 	if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1764 			     &tunnel->ip6rd.prefix) ||
1765 	    nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1766 			    tunnel->ip6rd.relay_prefix) ||
1767 	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1768 			tunnel->ip6rd.prefixlen) ||
1769 	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1770 			tunnel->ip6rd.relay_prefixlen))
1771 		goto nla_put_failure;
1772 #endif
1773 
1774 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1775 			tunnel->encap.type) ||
1776 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1777 			tunnel->encap.sport) ||
1778 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1779 			tunnel->encap.dport) ||
1780 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1781 			tunnel->encap.flags))
1782 		goto nla_put_failure;
1783 
1784 	return 0;
1785 
1786 nla_put_failure:
1787 	return -EMSGSIZE;
1788 }
1789 
1790 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1791 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1792 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1793 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1794 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1795 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1796 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1797 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1798 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1799 #ifdef CONFIG_IPV6_SIT_6RD
1800 	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1801 	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1802 	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1803 	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1804 #endif
1805 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
1806 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
1807 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
1808 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
1809 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
1810 };
1811 
1812 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1813 {
1814 	struct net *net = dev_net(dev);
1815 	struct sit_net *sitn = net_generic(net, sit_net_id);
1816 
1817 	if (dev != sitn->fb_tunnel_dev)
1818 		unregister_netdevice_queue(dev, head);
1819 }
1820 
1821 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1822 	.kind		= "sit",
1823 	.maxtype	= IFLA_IPTUN_MAX,
1824 	.policy		= ipip6_policy,
1825 	.priv_size	= sizeof(struct ip_tunnel),
1826 	.setup		= ipip6_tunnel_setup,
1827 	.validate	= ipip6_validate,
1828 	.newlink	= ipip6_newlink,
1829 	.changelink	= ipip6_changelink,
1830 	.get_size	= ipip6_get_size,
1831 	.fill_info	= ipip6_fill_info,
1832 	.dellink	= ipip6_dellink,
1833 	.get_link_net	= ip_tunnel_get_link_net,
1834 };
1835 
1836 static struct xfrm_tunnel sit_handler __read_mostly = {
1837 	.handler	=	ipip6_rcv,
1838 	.err_handler	=	ipip6_err,
1839 	.priority	=	1,
1840 };
1841 
1842 static struct xfrm_tunnel ipip_handler __read_mostly = {
1843 	.handler	=	ipip_rcv,
1844 	.err_handler	=	ipip6_err,
1845 	.priority	=	2,
1846 };
1847 
1848 #if IS_ENABLED(CONFIG_MPLS)
1849 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1850 	.handler	=	mplsip_rcv,
1851 	.err_handler	=	ipip6_err,
1852 	.priority	=	2,
1853 };
1854 #endif
1855 
1856 static void __net_exit sit_destroy_tunnels(struct net *net,
1857 					   struct list_head *head)
1858 {
1859 	struct sit_net *sitn = net_generic(net, sit_net_id);
1860 	struct net_device *dev, *aux;
1861 	int prio;
1862 
1863 	for_each_netdev_safe(net, dev, aux)
1864 		if (dev->rtnl_link_ops == &sit_link_ops)
1865 			unregister_netdevice_queue(dev, head);
1866 
1867 	for (prio = 1; prio < 4; prio++) {
1868 		int h;
1869 		for (h = 0; h < IP6_SIT_HASH_SIZE; h++) {
1870 			struct ip_tunnel *t;
1871 
1872 			t = rtnl_dereference(sitn->tunnels[prio][h]);
1873 			while (t) {
1874 				/* If dev is in the same netns, it has already
1875 				 * been added to the list by the previous loop.
1876 				 */
1877 				if (!net_eq(dev_net(t->dev), net))
1878 					unregister_netdevice_queue(t->dev,
1879 								   head);
1880 				t = rtnl_dereference(t->next);
1881 			}
1882 		}
1883 	}
1884 }
1885 
1886 static int __net_init sit_init_net(struct net *net)
1887 {
1888 	struct sit_net *sitn = net_generic(net, sit_net_id);
1889 	struct ip_tunnel *t;
1890 	int err;
1891 
1892 	sitn->tunnels[0] = sitn->tunnels_wc;
1893 	sitn->tunnels[1] = sitn->tunnels_l;
1894 	sitn->tunnels[2] = sitn->tunnels_r;
1895 	sitn->tunnels[3] = sitn->tunnels_r_l;
1896 
1897 	if (!net_has_fallback_tunnels(net))
1898 		return 0;
1899 
1900 	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1901 					   NET_NAME_UNKNOWN,
1902 					   ipip6_tunnel_setup);
1903 	if (!sitn->fb_tunnel_dev) {
1904 		err = -ENOMEM;
1905 		goto err_alloc_dev;
1906 	}
1907 	dev_net_set(sitn->fb_tunnel_dev, net);
1908 	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1909 	/* FB netdevice is special: we have one, and only one per netns.
1910 	 * Allowing to move it to another netns is clearly unsafe.
1911 	 */
1912 	sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1913 
1914 	err = register_netdev(sitn->fb_tunnel_dev);
1915 	if (err)
1916 		goto err_reg_dev;
1917 
1918 	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1919 	ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1920 
1921 	t = netdev_priv(sitn->fb_tunnel_dev);
1922 
1923 	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1924 	return 0;
1925 
1926 err_reg_dev:
1927 	ipip6_dev_free(sitn->fb_tunnel_dev);
1928 	free_netdev(sitn->fb_tunnel_dev);
1929 err_alloc_dev:
1930 	return err;
1931 }
1932 
1933 static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1934 {
1935 	LIST_HEAD(list);
1936 	struct net *net;
1937 
1938 	rtnl_lock();
1939 	list_for_each_entry(net, net_list, exit_list)
1940 		sit_destroy_tunnels(net, &list);
1941 
1942 	unregister_netdevice_many(&list);
1943 	rtnl_unlock();
1944 }
1945 
1946 static struct pernet_operations sit_net_ops = {
1947 	.init = sit_init_net,
1948 	.exit_batch = sit_exit_batch_net,
1949 	.id   = &sit_net_id,
1950 	.size = sizeof(struct sit_net),
1951 };
1952 
1953 static void __exit sit_cleanup(void)
1954 {
1955 	rtnl_link_unregister(&sit_link_ops);
1956 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1957 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1958 #if IS_ENABLED(CONFIG_MPLS)
1959 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1960 #endif
1961 
1962 	unregister_pernet_device(&sit_net_ops);
1963 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1964 }
1965 
1966 static int __init sit_init(void)
1967 {
1968 	int err;
1969 
1970 	pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1971 
1972 	err = register_pernet_device(&sit_net_ops);
1973 	if (err < 0)
1974 		return err;
1975 	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1976 	if (err < 0) {
1977 		pr_info("%s: can't register ip6ip4\n", __func__);
1978 		goto xfrm_tunnel_failed;
1979 	}
1980 	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1981 	if (err < 0) {
1982 		pr_info("%s: can't register ip4ip4\n", __func__);
1983 		goto xfrm_tunnel4_failed;
1984 	}
1985 #if IS_ENABLED(CONFIG_MPLS)
1986 	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1987 	if (err < 0) {
1988 		pr_info("%s: can't register mplsip\n", __func__);
1989 		goto xfrm_tunnel_mpls_failed;
1990 	}
1991 #endif
1992 	err = rtnl_link_register(&sit_link_ops);
1993 	if (err < 0)
1994 		goto rtnl_link_failed;
1995 
1996 out:
1997 	return err;
1998 
1999 rtnl_link_failed:
2000 #if IS_ENABLED(CONFIG_MPLS)
2001 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
2002 xfrm_tunnel_mpls_failed:
2003 #endif
2004 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
2005 xfrm_tunnel4_failed:
2006 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
2007 xfrm_tunnel_failed:
2008 	unregister_pernet_device(&sit_net_ops);
2009 	goto out;
2010 }
2011 
2012 module_init(sit_init);
2013 module_exit(sit_cleanup);
2014 MODULE_LICENSE("GPL");
2015 MODULE_ALIAS_RTNL_LINK("sit");
2016 MODULE_ALIAS_NETDEV("sit0");
2017