xref: /openbmc/linux/net/ipv6/sit.c (revision f7af616c)
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 	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 *parms, int create)
230 {
231 	__be32 remote = parms->iph.daddr;
232 	__be32 local = parms->iph.saddr;
233 	struct ip_tunnel *t, *nt;
234 	struct ip_tunnel __rcu **tp;
235 	struct net_device *dev;
236 	char name[IFNAMSIZ];
237 	struct sit_net *sitn = net_generic(net, sit_net_id);
238 
239 	for (tp = __ipip6_bucket(sitn, parms);
240 	    (t = rtnl_dereference(*tp)) != NULL;
241 	     tp = &t->next) {
242 		if (local == t->parms.iph.saddr &&
243 		    remote == t->parms.iph.daddr &&
244 		    parms->link == t->parms.link) {
245 			if (create)
246 				return NULL;
247 			else
248 				return t;
249 		}
250 	}
251 	if (!create)
252 		goto failed;
253 
254 	if (parms->name[0]) {
255 		if (!dev_valid_name(parms->name))
256 			goto failed;
257 		strlcpy(name, parms->name, IFNAMSIZ);
258 	} else {
259 		strcpy(name, "sit%d");
260 	}
261 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
262 			   ipip6_tunnel_setup);
263 	if (!dev)
264 		return NULL;
265 
266 	dev_net_set(dev, net);
267 
268 	nt = netdev_priv(dev);
269 
270 	nt->parms = *parms;
271 	if (ipip6_tunnel_create(dev) < 0)
272 		goto failed_free;
273 
274 	if (!parms->name[0])
275 		strcpy(parms->name, dev->name);
276 
277 	return nt;
278 
279 failed_free:
280 	free_netdev(dev);
281 failed:
282 	return NULL;
283 }
284 
285 #define for_each_prl_rcu(start)			\
286 	for (prl = rcu_dereference(start);	\
287 	     prl;				\
288 	     prl = rcu_dereference(prl->next))
289 
290 static struct ip_tunnel_prl_entry *
291 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
292 {
293 	struct ip_tunnel_prl_entry *prl;
294 
295 	for_each_prl_rcu(t->prl)
296 		if (prl->addr == addr)
297 			break;
298 	return prl;
299 
300 }
301 
302 static int ipip6_tunnel_get_prl(struct net_device *dev, struct ifreq *ifr)
303 {
304 	struct ip_tunnel_prl __user *a = ifr->ifr_ifru.ifru_data;
305 	struct ip_tunnel *t = netdev_priv(dev);
306 	struct ip_tunnel_prl kprl, *kp;
307 	struct ip_tunnel_prl_entry *prl;
308 	unsigned int cmax, c = 0, ca, len;
309 	int ret = 0;
310 
311 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
312 		return -EINVAL;
313 
314 	if (copy_from_user(&kprl, a, sizeof(kprl)))
315 		return -EFAULT;
316 	cmax = kprl.datalen / sizeof(kprl);
317 	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
318 		cmax = 1;
319 
320 	/* For simple GET or for root users,
321 	 * we try harder to allocate.
322 	 */
323 	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
324 		kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
325 		NULL;
326 
327 	rcu_read_lock();
328 
329 	ca = min(t->prl_count, cmax);
330 
331 	if (!kp) {
332 		/* We don't try hard to allocate much memory for
333 		 * non-root users.
334 		 * For root users, retry allocating enough memory for
335 		 * the answer.
336 		 */
337 		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
338 		if (!kp) {
339 			ret = -ENOMEM;
340 			goto out;
341 		}
342 	}
343 
344 	c = 0;
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 out:
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 
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, struct ifreq *ifr,
457 		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, ifr->ifr_ifru.ifru_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 	dev_put(dev);
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 		struct pcpu_sw_netstats *tstats;
690 
691 		if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
692 		    tunnel->parms.iph.protocol != 0)
693 			goto out;
694 
695 		skb->mac_header = skb->network_header;
696 		skb_reset_network_header(skb);
697 		IPCB(skb)->flags = 0;
698 		skb->dev = tunnel->dev;
699 
700 		if (packet_is_spoofed(skb, iph, tunnel)) {
701 			tunnel->dev->stats.rx_errors++;
702 			goto out;
703 		}
704 
705 		if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
706 		    !net_eq(tunnel->net, dev_net(tunnel->dev))))
707 			goto out;
708 
709 		/* skb can be uncloned in iptunnel_pull_header, so
710 		 * old iph is no longer valid
711 		 */
712 		iph = (const struct iphdr *)skb_mac_header(skb);
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 				++tunnel->dev->stats.rx_frame_errors;
720 				++tunnel->dev->stats.rx_errors;
721 				goto out;
722 			}
723 		}
724 
725 		tstats = this_cpu_ptr(tunnel->dev->tstats);
726 		u64_stats_update_begin(&tstats->syncp);
727 		tstats->rx_packets++;
728 		tstats->rx_bytes += skb->len;
729 		u64_stats_update_end(&tstats->syncp);
730 
731 		netif_rx(skb);
732 
733 		return 0;
734 	}
735 
736 	/* no tunnel matched,  let upstream know, ipsec may handle it */
737 	return 1;
738 out:
739 	kfree_skb(skb);
740 	return 0;
741 }
742 
743 static const struct tnl_ptk_info ipip_tpi = {
744 	/* no tunnel info required for ipip. */
745 	.proto = htons(ETH_P_IP),
746 };
747 
748 #if IS_ENABLED(CONFIG_MPLS)
749 static const struct tnl_ptk_info mplsip_tpi = {
750 	/* no tunnel info required for mplsip. */
751 	.proto = htons(ETH_P_MPLS_UC),
752 };
753 #endif
754 
755 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
756 {
757 	const struct iphdr *iph;
758 	struct ip_tunnel *tunnel;
759 	int sifindex;
760 
761 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
762 
763 	iph = ip_hdr(skb);
764 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
765 				     iph->saddr, iph->daddr, sifindex);
766 	if (tunnel) {
767 		const struct tnl_ptk_info *tpi;
768 
769 		if (tunnel->parms.iph.protocol != ipproto &&
770 		    tunnel->parms.iph.protocol != 0)
771 			goto drop;
772 
773 		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
774 			goto drop;
775 #if IS_ENABLED(CONFIG_MPLS)
776 		if (ipproto == IPPROTO_MPLS)
777 			tpi = &mplsip_tpi;
778 		else
779 #endif
780 			tpi = &ipip_tpi;
781 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
782 			goto drop;
783 		return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
784 	}
785 
786 	return 1;
787 
788 drop:
789 	kfree_skb(skb);
790 	return 0;
791 }
792 
793 static int ipip_rcv(struct sk_buff *skb)
794 {
795 	return sit_tunnel_rcv(skb, IPPROTO_IPIP);
796 }
797 
798 #if IS_ENABLED(CONFIG_MPLS)
799 static int mplsip_rcv(struct sk_buff *skb)
800 {
801 	return sit_tunnel_rcv(skb, IPPROTO_MPLS);
802 }
803 #endif
804 
805 /*
806  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
807  * stores the embedded IPv4 address in v4dst and returns true.
808  */
809 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
810 		      __be32 *v4dst)
811 {
812 #ifdef CONFIG_IPV6_SIT_6RD
813 	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
814 			      tunnel->ip6rd.prefixlen)) {
815 		unsigned int pbw0, pbi0;
816 		int pbi1;
817 		u32 d;
818 
819 		pbw0 = tunnel->ip6rd.prefixlen >> 5;
820 		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
821 
822 		d = tunnel->ip6rd.relay_prefixlen < 32 ?
823 			(ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
824 		    tunnel->ip6rd.relay_prefixlen : 0;
825 
826 		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
827 		if (pbi1 > 0)
828 			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
829 			     (32 - pbi1);
830 
831 		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
832 		return true;
833 	}
834 #else
835 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
836 		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
837 		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
838 		return true;
839 	}
840 #endif
841 	return false;
842 }
843 
844 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
845 			     const struct in6_addr *v6dst)
846 {
847 	__be32 dst = 0;
848 	check_6rd(tunnel, v6dst, &dst);
849 	return dst;
850 }
851 
852 /*
853  *	This function assumes it is being called from dev_queue_xmit()
854  *	and that skb is filled properly by that function.
855  */
856 
857 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
858 				     struct net_device *dev)
859 {
860 	struct ip_tunnel *tunnel = netdev_priv(dev);
861 	const struct iphdr  *tiph = &tunnel->parms.iph;
862 	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
863 	u8     tos = tunnel->parms.iph.tos;
864 	__be16 df = tiph->frag_off;
865 	struct rtable *rt;		/* Route to the other host */
866 	struct net_device *tdev;	/* Device to other host */
867 	unsigned int max_headroom;	/* The extra header space needed */
868 	__be32 dst = tiph->daddr;
869 	struct flowi4 fl4;
870 	int    mtu;
871 	const struct in6_addr *addr6;
872 	int addr_type;
873 	u8 ttl;
874 	u8 protocol = IPPROTO_IPV6;
875 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
876 
877 	if (tos == 1)
878 		tos = ipv6_get_dsfield(iph6);
879 
880 	/* ISATAP (RFC4214) - must come before 6to4 */
881 	if (dev->priv_flags & IFF_ISATAP) {
882 		struct neighbour *neigh = NULL;
883 		bool do_tx_error = false;
884 
885 		if (skb_dst(skb))
886 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
887 
888 		if (!neigh) {
889 			net_dbg_ratelimited("nexthop == NULL\n");
890 			goto tx_error;
891 		}
892 
893 		addr6 = (const struct in6_addr *)&neigh->primary_key;
894 		addr_type = ipv6_addr_type(addr6);
895 
896 		if ((addr_type & IPV6_ADDR_UNICAST) &&
897 		     ipv6_addr_is_isatap(addr6))
898 			dst = addr6->s6_addr32[3];
899 		else
900 			do_tx_error = true;
901 
902 		neigh_release(neigh);
903 		if (do_tx_error)
904 			goto tx_error;
905 	}
906 
907 	if (!dst)
908 		dst = try_6rd(tunnel, &iph6->daddr);
909 
910 	if (!dst) {
911 		struct neighbour *neigh = NULL;
912 		bool do_tx_error = false;
913 
914 		if (skb_dst(skb))
915 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
916 
917 		if (!neigh) {
918 			net_dbg_ratelimited("nexthop == NULL\n");
919 			goto tx_error;
920 		}
921 
922 		addr6 = (const struct in6_addr *)&neigh->primary_key;
923 		addr_type = ipv6_addr_type(addr6);
924 
925 		if (addr_type == IPV6_ADDR_ANY) {
926 			addr6 = &ipv6_hdr(skb)->daddr;
927 			addr_type = ipv6_addr_type(addr6);
928 		}
929 
930 		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
931 			dst = addr6->s6_addr32[3];
932 		else
933 			do_tx_error = true;
934 
935 		neigh_release(neigh);
936 		if (do_tx_error)
937 			goto tx_error;
938 	}
939 
940 	flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
941 			   RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
942 			   0, dst, tiph->saddr, 0, 0,
943 			   sock_net_uid(tunnel->net, NULL));
944 
945 	rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
946 	if (!rt) {
947 		rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
948 		if (IS_ERR(rt)) {
949 			dev->stats.tx_carrier_errors++;
950 			goto tx_error_icmp;
951 		}
952 		dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
953 	}
954 
955 	if (rt->rt_type != RTN_UNICAST) {
956 		ip_rt_put(rt);
957 		dev->stats.tx_carrier_errors++;
958 		goto tx_error_icmp;
959 	}
960 	tdev = rt->dst.dev;
961 
962 	if (tdev == dev) {
963 		ip_rt_put(rt);
964 		dev->stats.collisions++;
965 		goto tx_error;
966 	}
967 
968 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
969 		ip_rt_put(rt);
970 		goto tx_error;
971 	}
972 
973 	if (df) {
974 		mtu = dst_mtu(&rt->dst) - t_hlen;
975 
976 		if (mtu < IPV4_MIN_MTU) {
977 			dev->stats.collisions++;
978 			ip_rt_put(rt);
979 			goto tx_error;
980 		}
981 
982 		if (mtu < IPV6_MIN_MTU) {
983 			mtu = IPV6_MIN_MTU;
984 			df = 0;
985 		}
986 
987 		if (tunnel->parms.iph.daddr)
988 			skb_dst_update_pmtu_no_confirm(skb, mtu);
989 
990 		if (skb->len > mtu && !skb_is_gso(skb)) {
991 			icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
992 			ip_rt_put(rt);
993 			goto tx_error;
994 		}
995 	}
996 
997 	if (tunnel->err_count > 0) {
998 		if (time_before(jiffies,
999 				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
1000 			tunnel->err_count--;
1001 			dst_link_failure(skb);
1002 		} else
1003 			tunnel->err_count = 0;
1004 	}
1005 
1006 	/*
1007 	 * Okay, now see if we can stuff it in the buffer as-is.
1008 	 */
1009 	max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
1010 
1011 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1012 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1013 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
1014 		if (!new_skb) {
1015 			ip_rt_put(rt);
1016 			dev->stats.tx_dropped++;
1017 			kfree_skb(skb);
1018 			return NETDEV_TX_OK;
1019 		}
1020 		if (skb->sk)
1021 			skb_set_owner_w(new_skb, skb->sk);
1022 		dev_kfree_skb(skb);
1023 		skb = new_skb;
1024 		iph6 = ipv6_hdr(skb);
1025 	}
1026 	ttl = tiph->ttl;
1027 	if (ttl == 0)
1028 		ttl = iph6->hop_limit;
1029 	tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
1030 
1031 	if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
1032 		ip_rt_put(rt);
1033 		goto tx_error;
1034 	}
1035 
1036 	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1037 
1038 	iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
1039 		      df, !net_eq(tunnel->net, dev_net(dev)));
1040 	return NETDEV_TX_OK;
1041 
1042 tx_error_icmp:
1043 	dst_link_failure(skb);
1044 tx_error:
1045 	kfree_skb(skb);
1046 	dev->stats.tx_errors++;
1047 	return NETDEV_TX_OK;
1048 }
1049 
1050 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1051 				     struct net_device *dev, u8 ipproto)
1052 {
1053 	struct ip_tunnel *tunnel = netdev_priv(dev);
1054 	const struct iphdr  *tiph = &tunnel->parms.iph;
1055 
1056 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1057 		goto tx_error;
1058 
1059 	skb_set_inner_ipproto(skb, ipproto);
1060 
1061 	ip_tunnel_xmit(skb, dev, tiph, ipproto);
1062 	return NETDEV_TX_OK;
1063 tx_error:
1064 	kfree_skb(skb);
1065 	dev->stats.tx_errors++;
1066 	return NETDEV_TX_OK;
1067 }
1068 
1069 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1070 				   struct net_device *dev)
1071 {
1072 	if (!pskb_inet_may_pull(skb))
1073 		goto tx_err;
1074 
1075 	switch (skb->protocol) {
1076 	case htons(ETH_P_IP):
1077 		sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1078 		break;
1079 	case htons(ETH_P_IPV6):
1080 		ipip6_tunnel_xmit(skb, dev);
1081 		break;
1082 #if IS_ENABLED(CONFIG_MPLS)
1083 	case htons(ETH_P_MPLS_UC):
1084 		sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1085 		break;
1086 #endif
1087 	default:
1088 		goto tx_err;
1089 	}
1090 
1091 	return NETDEV_TX_OK;
1092 
1093 tx_err:
1094 	dev->stats.tx_errors++;
1095 	kfree_skb(skb);
1096 	return NETDEV_TX_OK;
1097 
1098 }
1099 
1100 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1101 {
1102 	struct net_device *tdev = NULL;
1103 	struct ip_tunnel *tunnel;
1104 	const struct iphdr *iph;
1105 	struct flowi4 fl4;
1106 
1107 	tunnel = netdev_priv(dev);
1108 	iph = &tunnel->parms.iph;
1109 
1110 	if (iph->daddr) {
1111 		struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1112 							  NULL,
1113 							  iph->daddr, iph->saddr,
1114 							  0, 0,
1115 							  IPPROTO_IPV6,
1116 							  RT_TOS(iph->tos),
1117 							  tunnel->parms.link);
1118 
1119 		if (!IS_ERR(rt)) {
1120 			tdev = rt->dst.dev;
1121 			ip_rt_put(rt);
1122 		}
1123 		dev->flags |= IFF_POINTOPOINT;
1124 	}
1125 
1126 	if (!tdev && tunnel->parms.link)
1127 		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1128 
1129 	if (tdev && !netif_is_l3_master(tdev)) {
1130 		int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1131 
1132 		dev->mtu = tdev->mtu - t_hlen;
1133 		if (dev->mtu < IPV6_MIN_MTU)
1134 			dev->mtu = IPV6_MIN_MTU;
1135 	}
1136 }
1137 
1138 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1139 				__u32 fwmark)
1140 {
1141 	struct net *net = t->net;
1142 	struct sit_net *sitn = net_generic(net, sit_net_id);
1143 
1144 	ipip6_tunnel_unlink(sitn, t);
1145 	synchronize_net();
1146 	t->parms.iph.saddr = p->iph.saddr;
1147 	t->parms.iph.daddr = p->iph.daddr;
1148 	memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1149 	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1150 	ipip6_tunnel_link(sitn, t);
1151 	t->parms.iph.ttl = p->iph.ttl;
1152 	t->parms.iph.tos = p->iph.tos;
1153 	t->parms.iph.frag_off = p->iph.frag_off;
1154 	if (t->parms.link != p->link || t->fwmark != fwmark) {
1155 		t->parms.link = p->link;
1156 		t->fwmark = fwmark;
1157 		ipip6_tunnel_bind_dev(t->dev);
1158 	}
1159 	dst_cache_reset(&t->dst_cache);
1160 	netdev_state_change(t->dev);
1161 }
1162 
1163 #ifdef CONFIG_IPV6_SIT_6RD
1164 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1165 				   struct ip_tunnel_6rd *ip6rd)
1166 {
1167 	struct in6_addr prefix;
1168 	__be32 relay_prefix;
1169 
1170 	if (ip6rd->relay_prefixlen > 32 ||
1171 	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1172 		return -EINVAL;
1173 
1174 	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1175 	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1176 		return -EINVAL;
1177 	if (ip6rd->relay_prefixlen)
1178 		relay_prefix = ip6rd->relay_prefix &
1179 			       htonl(0xffffffffUL <<
1180 				     (32 - ip6rd->relay_prefixlen));
1181 	else
1182 		relay_prefix = 0;
1183 	if (relay_prefix != ip6rd->relay_prefix)
1184 		return -EINVAL;
1185 
1186 	t->ip6rd.prefix = prefix;
1187 	t->ip6rd.relay_prefix = relay_prefix;
1188 	t->ip6rd.prefixlen = ip6rd->prefixlen;
1189 	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1190 	dst_cache_reset(&t->dst_cache);
1191 	netdev_state_change(t->dev);
1192 	return 0;
1193 }
1194 
1195 static int
1196 ipip6_tunnel_get6rd(struct net_device *dev, struct ifreq *ifr)
1197 {
1198 	struct ip_tunnel *t = netdev_priv(dev);
1199 	struct ip_tunnel_6rd ip6rd;
1200 	struct ip_tunnel_parm p;
1201 
1202 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1203 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1204 			return -EFAULT;
1205 		t = ipip6_tunnel_locate(t->net, &p, 0);
1206 	}
1207 	if (!t)
1208 		t = netdev_priv(dev);
1209 
1210 	ip6rd.prefix = t->ip6rd.prefix;
1211 	ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1212 	ip6rd.prefixlen = t->ip6rd.prefixlen;
1213 	ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1214 	if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd, sizeof(ip6rd)))
1215 		return -EFAULT;
1216 	return 0;
1217 }
1218 
1219 static int
1220 ipip6_tunnel_6rdctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1221 {
1222 	struct ip_tunnel *t = netdev_priv(dev);
1223 	struct ip_tunnel_6rd ip6rd;
1224 	int err;
1225 
1226 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1227 		return -EPERM;
1228 	if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data, sizeof(ip6rd)))
1229 		return -EFAULT;
1230 
1231 	if (cmd != SIOCDEL6RD) {
1232 		err = ipip6_tunnel_update_6rd(t, &ip6rd);
1233 		if (err < 0)
1234 			return err;
1235 	} else
1236 		ipip6_tunnel_clone_6rd(dev, dev_to_sit_net(dev));
1237 	return 0;
1238 }
1239 
1240 #endif /* CONFIG_IPV6_SIT_6RD */
1241 
1242 static bool ipip6_valid_ip_proto(u8 ipproto)
1243 {
1244 	return ipproto == IPPROTO_IPV6 ||
1245 		ipproto == IPPROTO_IPIP ||
1246 #if IS_ENABLED(CONFIG_MPLS)
1247 		ipproto == IPPROTO_MPLS ||
1248 #endif
1249 		ipproto == 0;
1250 }
1251 
1252 static int
1253 __ipip6_tunnel_ioctl_validate(struct net *net, struct ip_tunnel_parm *p)
1254 {
1255 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1256 		return -EPERM;
1257 
1258 	if (!ipip6_valid_ip_proto(p->iph.protocol))
1259 		return -EINVAL;
1260 	if (p->iph.version != 4 ||
1261 	    p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
1262 		return -EINVAL;
1263 
1264 	if (p->iph.ttl)
1265 		p->iph.frag_off |= htons(IP_DF);
1266 	return 0;
1267 }
1268 
1269 static int
1270 ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm *p)
1271 {
1272 	struct ip_tunnel *t = netdev_priv(dev);
1273 
1274 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
1275 		t = ipip6_tunnel_locate(t->net, p, 0);
1276 	if (!t)
1277 		t = netdev_priv(dev);
1278 	memcpy(p, &t->parms, sizeof(*p));
1279 	return 0;
1280 }
1281 
1282 static int
1283 ipip6_tunnel_add(struct net_device *dev, struct ip_tunnel_parm *p)
1284 {
1285 	struct ip_tunnel *t = netdev_priv(dev);
1286 	int err;
1287 
1288 	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1289 	if (err)
1290 		return err;
1291 
1292 	t = ipip6_tunnel_locate(t->net, p, 1);
1293 	if (!t)
1294 		return -ENOBUFS;
1295 	return 0;
1296 }
1297 
1298 static int
1299 ipip6_tunnel_change(struct net_device *dev, struct ip_tunnel_parm *p)
1300 {
1301 	struct ip_tunnel *t = netdev_priv(dev);
1302 	int err;
1303 
1304 	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1305 	if (err)
1306 		return err;
1307 
1308 	t = ipip6_tunnel_locate(t->net, p, 0);
1309 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1310 		if (!t)
1311 			return -ENOENT;
1312 	} else {
1313 		if (t) {
1314 			if (t->dev != dev)
1315 				return -EEXIST;
1316 		} else {
1317 			if (((dev->flags & IFF_POINTOPOINT) && !p->iph.daddr) ||
1318 			    (!(dev->flags & IFF_POINTOPOINT) && p->iph.daddr))
1319 				return -EINVAL;
1320 			t = netdev_priv(dev);
1321 		}
1322 
1323 		ipip6_tunnel_update(t, p, t->fwmark);
1324 	}
1325 
1326 	return 0;
1327 }
1328 
1329 static int
1330 ipip6_tunnel_del(struct net_device *dev, struct ip_tunnel_parm *p)
1331 {
1332 	struct ip_tunnel *t = netdev_priv(dev);
1333 
1334 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1335 		return -EPERM;
1336 
1337 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1338 		t = ipip6_tunnel_locate(t->net, p, 0);
1339 		if (!t)
1340 			return -ENOENT;
1341 		if (t == netdev_priv(dev_to_sit_net(dev)->fb_tunnel_dev))
1342 			return -EPERM;
1343 		dev = t->dev;
1344 	}
1345 	unregister_netdevice(dev);
1346 	return 0;
1347 }
1348 
1349 static int
1350 ipip6_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
1351 {
1352 	switch (cmd) {
1353 	case SIOCGETTUNNEL:
1354 		return ipip6_tunnel_get(dev, p);
1355 	case SIOCADDTUNNEL:
1356 		return ipip6_tunnel_add(dev, p);
1357 	case SIOCCHGTUNNEL:
1358 		return ipip6_tunnel_change(dev, p);
1359 	case SIOCDELTUNNEL:
1360 		return ipip6_tunnel_del(dev, p);
1361 	default:
1362 		return -EINVAL;
1363 	}
1364 }
1365 
1366 static int
1367 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1368 {
1369 	switch (cmd) {
1370 	case SIOCGETTUNNEL:
1371 	case SIOCADDTUNNEL:
1372 	case SIOCCHGTUNNEL:
1373 	case SIOCDELTUNNEL:
1374 		return ip_tunnel_ioctl(dev, ifr, cmd);
1375 	case SIOCGETPRL:
1376 		return ipip6_tunnel_get_prl(dev, ifr);
1377 	case SIOCADDPRL:
1378 	case SIOCDELPRL:
1379 	case SIOCCHGPRL:
1380 		return ipip6_tunnel_prl_ctl(dev, ifr, cmd);
1381 #ifdef CONFIG_IPV6_SIT_6RD
1382 	case SIOCGET6RD:
1383 		return ipip6_tunnel_get6rd(dev, ifr);
1384 	case SIOCADD6RD:
1385 	case SIOCCHG6RD:
1386 	case SIOCDEL6RD:
1387 		return ipip6_tunnel_6rdctl(dev, ifr, cmd);
1388 #endif
1389 	default:
1390 		return -EINVAL;
1391 	}
1392 }
1393 
1394 static const struct net_device_ops ipip6_netdev_ops = {
1395 	.ndo_init	= ipip6_tunnel_init,
1396 	.ndo_uninit	= ipip6_tunnel_uninit,
1397 	.ndo_start_xmit	= sit_tunnel_xmit,
1398 	.ndo_do_ioctl	= ipip6_tunnel_ioctl,
1399 	.ndo_get_stats64 = dev_get_tstats64,
1400 	.ndo_get_iflink = ip_tunnel_get_iflink,
1401 	.ndo_tunnel_ctl = ipip6_tunnel_ctl,
1402 };
1403 
1404 static void ipip6_dev_free(struct net_device *dev)
1405 {
1406 	struct ip_tunnel *tunnel = netdev_priv(dev);
1407 
1408 	dst_cache_destroy(&tunnel->dst_cache);
1409 	free_percpu(dev->tstats);
1410 }
1411 
1412 #define SIT_FEATURES (NETIF_F_SG	   | \
1413 		      NETIF_F_FRAGLIST	   | \
1414 		      NETIF_F_HIGHDMA	   | \
1415 		      NETIF_F_GSO_SOFTWARE | \
1416 		      NETIF_F_HW_CSUM)
1417 
1418 static void ipip6_tunnel_setup(struct net_device *dev)
1419 {
1420 	struct ip_tunnel *tunnel = netdev_priv(dev);
1421 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1422 
1423 	dev->netdev_ops		= &ipip6_netdev_ops;
1424 	dev->header_ops		= &ip_tunnel_header_ops;
1425 	dev->needs_free_netdev	= true;
1426 	dev->priv_destructor	= ipip6_dev_free;
1427 
1428 	dev->type		= ARPHRD_SIT;
1429 	dev->mtu		= ETH_DATA_LEN - t_hlen;
1430 	dev->min_mtu		= IPV6_MIN_MTU;
1431 	dev->max_mtu		= IP6_MAX_MTU - t_hlen;
1432 	dev->flags		= IFF_NOARP;
1433 	netif_keep_dst(dev);
1434 	dev->addr_len		= 4;
1435 	dev->features		|= NETIF_F_LLTX;
1436 	dev->features		|= SIT_FEATURES;
1437 	dev->hw_features	|= SIT_FEATURES;
1438 }
1439 
1440 static int ipip6_tunnel_init(struct net_device *dev)
1441 {
1442 	struct ip_tunnel *tunnel = netdev_priv(dev);
1443 	int err;
1444 
1445 	tunnel->dev = dev;
1446 	tunnel->net = dev_net(dev);
1447 	strcpy(tunnel->parms.name, dev->name);
1448 
1449 	ipip6_tunnel_bind_dev(dev);
1450 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1451 	if (!dev->tstats)
1452 		return -ENOMEM;
1453 
1454 	err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1455 	if (err) {
1456 		free_percpu(dev->tstats);
1457 		dev->tstats = NULL;
1458 		return err;
1459 	}
1460 	dev_hold(dev);
1461 	return 0;
1462 }
1463 
1464 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1465 {
1466 	struct ip_tunnel *tunnel = netdev_priv(dev);
1467 	struct iphdr *iph = &tunnel->parms.iph;
1468 	struct net *net = dev_net(dev);
1469 	struct sit_net *sitn = net_generic(net, sit_net_id);
1470 
1471 	iph->version		= 4;
1472 	iph->protocol		= IPPROTO_IPV6;
1473 	iph->ihl		= 5;
1474 	iph->ttl		= 64;
1475 
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 		if (err < 0)
1651 			unregister_netdevice_queue(dev, NULL);
1652 	}
1653 #endif
1654 
1655 	return err;
1656 }
1657 
1658 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1659 			    struct nlattr *data[],
1660 			    struct netlink_ext_ack *extack)
1661 {
1662 	struct ip_tunnel *t = netdev_priv(dev);
1663 	struct ip_tunnel_parm p;
1664 	struct ip_tunnel_encap ipencap;
1665 	struct net *net = t->net;
1666 	struct sit_net *sitn = net_generic(net, sit_net_id);
1667 #ifdef CONFIG_IPV6_SIT_6RD
1668 	struct ip_tunnel_6rd ip6rd;
1669 #endif
1670 	__u32 fwmark = t->fwmark;
1671 	int err;
1672 
1673 	if (dev == sitn->fb_tunnel_dev)
1674 		return -EINVAL;
1675 
1676 	if (ipip6_netlink_encap_parms(data, &ipencap)) {
1677 		err = ip_tunnel_encap_setup(t, &ipencap);
1678 		if (err < 0)
1679 			return err;
1680 	}
1681 
1682 	ipip6_netlink_parms(data, &p, &fwmark);
1683 
1684 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1685 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1686 		return -EINVAL;
1687 
1688 	t = ipip6_tunnel_locate(net, &p, 0);
1689 
1690 	if (t) {
1691 		if (t->dev != dev)
1692 			return -EEXIST;
1693 	} else
1694 		t = netdev_priv(dev);
1695 
1696 	ipip6_tunnel_update(t, &p, fwmark);
1697 
1698 #ifdef CONFIG_IPV6_SIT_6RD
1699 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1700 		return ipip6_tunnel_update_6rd(t, &ip6rd);
1701 #endif
1702 
1703 	return 0;
1704 }
1705 
1706 static size_t ipip6_get_size(const struct net_device *dev)
1707 {
1708 	return
1709 		/* IFLA_IPTUN_LINK */
1710 		nla_total_size(4) +
1711 		/* IFLA_IPTUN_LOCAL */
1712 		nla_total_size(4) +
1713 		/* IFLA_IPTUN_REMOTE */
1714 		nla_total_size(4) +
1715 		/* IFLA_IPTUN_TTL */
1716 		nla_total_size(1) +
1717 		/* IFLA_IPTUN_TOS */
1718 		nla_total_size(1) +
1719 		/* IFLA_IPTUN_PMTUDISC */
1720 		nla_total_size(1) +
1721 		/* IFLA_IPTUN_FLAGS */
1722 		nla_total_size(2) +
1723 		/* IFLA_IPTUN_PROTO */
1724 		nla_total_size(1) +
1725 #ifdef CONFIG_IPV6_SIT_6RD
1726 		/* IFLA_IPTUN_6RD_PREFIX */
1727 		nla_total_size(sizeof(struct in6_addr)) +
1728 		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1729 		nla_total_size(4) +
1730 		/* IFLA_IPTUN_6RD_PREFIXLEN */
1731 		nla_total_size(2) +
1732 		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1733 		nla_total_size(2) +
1734 #endif
1735 		/* IFLA_IPTUN_ENCAP_TYPE */
1736 		nla_total_size(2) +
1737 		/* IFLA_IPTUN_ENCAP_FLAGS */
1738 		nla_total_size(2) +
1739 		/* IFLA_IPTUN_ENCAP_SPORT */
1740 		nla_total_size(2) +
1741 		/* IFLA_IPTUN_ENCAP_DPORT */
1742 		nla_total_size(2) +
1743 		/* IFLA_IPTUN_FWMARK */
1744 		nla_total_size(4) +
1745 		0;
1746 }
1747 
1748 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1749 {
1750 	struct ip_tunnel *tunnel = netdev_priv(dev);
1751 	struct ip_tunnel_parm *parm = &tunnel->parms;
1752 
1753 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1754 	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1755 	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1756 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1757 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1758 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1759 		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1760 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1761 	    nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1762 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1763 		goto nla_put_failure;
1764 
1765 #ifdef CONFIG_IPV6_SIT_6RD
1766 	if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1767 			     &tunnel->ip6rd.prefix) ||
1768 	    nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1769 			    tunnel->ip6rd.relay_prefix) ||
1770 	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1771 			tunnel->ip6rd.prefixlen) ||
1772 	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1773 			tunnel->ip6rd.relay_prefixlen))
1774 		goto nla_put_failure;
1775 #endif
1776 
1777 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1778 			tunnel->encap.type) ||
1779 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1780 			tunnel->encap.sport) ||
1781 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1782 			tunnel->encap.dport) ||
1783 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1784 			tunnel->encap.flags))
1785 		goto nla_put_failure;
1786 
1787 	return 0;
1788 
1789 nla_put_failure:
1790 	return -EMSGSIZE;
1791 }
1792 
1793 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1794 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1795 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1796 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1797 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1798 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1799 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1800 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1801 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1802 #ifdef CONFIG_IPV6_SIT_6RD
1803 	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1804 	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1805 	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1806 	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1807 #endif
1808 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
1809 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
1810 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
1811 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
1812 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
1813 };
1814 
1815 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1816 {
1817 	struct net *net = dev_net(dev);
1818 	struct sit_net *sitn = net_generic(net, sit_net_id);
1819 
1820 	if (dev != sitn->fb_tunnel_dev)
1821 		unregister_netdevice_queue(dev, head);
1822 }
1823 
1824 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1825 	.kind		= "sit",
1826 	.maxtype	= IFLA_IPTUN_MAX,
1827 	.policy		= ipip6_policy,
1828 	.priv_size	= sizeof(struct ip_tunnel),
1829 	.setup		= ipip6_tunnel_setup,
1830 	.validate	= ipip6_validate,
1831 	.newlink	= ipip6_newlink,
1832 	.changelink	= ipip6_changelink,
1833 	.get_size	= ipip6_get_size,
1834 	.fill_info	= ipip6_fill_info,
1835 	.dellink	= ipip6_dellink,
1836 	.get_link_net	= ip_tunnel_get_link_net,
1837 };
1838 
1839 static struct xfrm_tunnel sit_handler __read_mostly = {
1840 	.handler	=	ipip6_rcv,
1841 	.err_handler	=	ipip6_err,
1842 	.priority	=	1,
1843 };
1844 
1845 static struct xfrm_tunnel ipip_handler __read_mostly = {
1846 	.handler	=	ipip_rcv,
1847 	.err_handler	=	ipip6_err,
1848 	.priority	=	2,
1849 };
1850 
1851 #if IS_ENABLED(CONFIG_MPLS)
1852 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1853 	.handler	=	mplsip_rcv,
1854 	.err_handler	=	ipip6_err,
1855 	.priority	=	2,
1856 };
1857 #endif
1858 
1859 static void __net_exit sit_destroy_tunnels(struct net *net,
1860 					   struct list_head *head)
1861 {
1862 	struct sit_net *sitn = net_generic(net, sit_net_id);
1863 	struct net_device *dev, *aux;
1864 	int prio;
1865 
1866 	for_each_netdev_safe(net, dev, aux)
1867 		if (dev->rtnl_link_ops == &sit_link_ops)
1868 			unregister_netdevice_queue(dev, head);
1869 
1870 	for (prio = 0; prio < 4; prio++) {
1871 		int h;
1872 		for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1873 			struct ip_tunnel *t;
1874 
1875 			t = rtnl_dereference(sitn->tunnels[prio][h]);
1876 			while (t) {
1877 				/* If dev is in the same netns, it has already
1878 				 * been added to the list by the previous loop.
1879 				 */
1880 				if (!net_eq(dev_net(t->dev), net))
1881 					unregister_netdevice_queue(t->dev,
1882 								   head);
1883 				t = rtnl_dereference(t->next);
1884 			}
1885 		}
1886 	}
1887 }
1888 
1889 static int __net_init sit_init_net(struct net *net)
1890 {
1891 	struct sit_net *sitn = net_generic(net, sit_net_id);
1892 	struct ip_tunnel *t;
1893 	int err;
1894 
1895 	sitn->tunnels[0] = sitn->tunnels_wc;
1896 	sitn->tunnels[1] = sitn->tunnels_l;
1897 	sitn->tunnels[2] = sitn->tunnels_r;
1898 	sitn->tunnels[3] = sitn->tunnels_r_l;
1899 
1900 	if (!net_has_fallback_tunnels(net))
1901 		return 0;
1902 
1903 	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1904 					   NET_NAME_UNKNOWN,
1905 					   ipip6_tunnel_setup);
1906 	if (!sitn->fb_tunnel_dev) {
1907 		err = -ENOMEM;
1908 		goto err_alloc_dev;
1909 	}
1910 	dev_net_set(sitn->fb_tunnel_dev, net);
1911 	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1912 	/* FB netdevice is special: we have one, and only one per netns.
1913 	 * Allowing to move it to another netns is clearly unsafe.
1914 	 */
1915 	sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1916 
1917 	err = register_netdev(sitn->fb_tunnel_dev);
1918 	if (err)
1919 		goto err_reg_dev;
1920 
1921 	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1922 	ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1923 
1924 	t = netdev_priv(sitn->fb_tunnel_dev);
1925 
1926 	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1927 	return 0;
1928 
1929 err_reg_dev:
1930 	ipip6_dev_free(sitn->fb_tunnel_dev);
1931 	free_netdev(sitn->fb_tunnel_dev);
1932 err_alloc_dev:
1933 	return err;
1934 }
1935 
1936 static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1937 {
1938 	LIST_HEAD(list);
1939 	struct net *net;
1940 
1941 	rtnl_lock();
1942 	list_for_each_entry(net, net_list, exit_list)
1943 		sit_destroy_tunnels(net, &list);
1944 
1945 	unregister_netdevice_many(&list);
1946 	rtnl_unlock();
1947 }
1948 
1949 static struct pernet_operations sit_net_ops = {
1950 	.init = sit_init_net,
1951 	.exit_batch = sit_exit_batch_net,
1952 	.id   = &sit_net_id,
1953 	.size = sizeof(struct sit_net),
1954 };
1955 
1956 static void __exit sit_cleanup(void)
1957 {
1958 	rtnl_link_unregister(&sit_link_ops);
1959 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1960 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1961 #if IS_ENABLED(CONFIG_MPLS)
1962 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1963 #endif
1964 
1965 	unregister_pernet_device(&sit_net_ops);
1966 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1967 }
1968 
1969 static int __init sit_init(void)
1970 {
1971 	int err;
1972 
1973 	pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1974 
1975 	err = register_pernet_device(&sit_net_ops);
1976 	if (err < 0)
1977 		return err;
1978 	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1979 	if (err < 0) {
1980 		pr_info("%s: can't register ip6ip4\n", __func__);
1981 		goto xfrm_tunnel_failed;
1982 	}
1983 	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1984 	if (err < 0) {
1985 		pr_info("%s: can't register ip4ip4\n", __func__);
1986 		goto xfrm_tunnel4_failed;
1987 	}
1988 #if IS_ENABLED(CONFIG_MPLS)
1989 	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1990 	if (err < 0) {
1991 		pr_info("%s: can't register mplsip\n", __func__);
1992 		goto xfrm_tunnel_mpls_failed;
1993 	}
1994 #endif
1995 	err = rtnl_link_register(&sit_link_ops);
1996 	if (err < 0)
1997 		goto rtnl_link_failed;
1998 
1999 out:
2000 	return err;
2001 
2002 rtnl_link_failed:
2003 #if IS_ENABLED(CONFIG_MPLS)
2004 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
2005 xfrm_tunnel_mpls_failed:
2006 #endif
2007 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
2008 xfrm_tunnel4_failed:
2009 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
2010 xfrm_tunnel_failed:
2011 	unregister_pernet_device(&sit_net_ops);
2012 	goto out;
2013 }
2014 
2015 module_init(sit_init);
2016 module_exit(sit_cleanup);
2017 MODULE_LICENSE("GPL");
2018 MODULE_ALIAS_RTNL_LINK("sit");
2019 MODULE_ALIAS_NETDEV("sit0");
2020