xref: /openbmc/linux/net/ipv6/sit.c (revision 35267cea)
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 		skb_reset_mac_header(skb);
714 
715 		err = IP_ECN_decapsulate(iph, skb);
716 		if (unlikely(err)) {
717 			if (log_ecn_error)
718 				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
719 						     &iph->saddr, iph->tos);
720 			if (err > 1) {
721 				++tunnel->dev->stats.rx_frame_errors;
722 				++tunnel->dev->stats.rx_errors;
723 				goto out;
724 			}
725 		}
726 
727 		tstats = this_cpu_ptr(tunnel->dev->tstats);
728 		u64_stats_update_begin(&tstats->syncp);
729 		tstats->rx_packets++;
730 		tstats->rx_bytes += skb->len;
731 		u64_stats_update_end(&tstats->syncp);
732 
733 		netif_rx(skb);
734 
735 		return 0;
736 	}
737 
738 	/* no tunnel matched,  let upstream know, ipsec may handle it */
739 	return 1;
740 out:
741 	kfree_skb(skb);
742 	return 0;
743 }
744 
745 static const struct tnl_ptk_info ipip_tpi = {
746 	/* no tunnel info required for ipip. */
747 	.proto = htons(ETH_P_IP),
748 };
749 
750 #if IS_ENABLED(CONFIG_MPLS)
751 static const struct tnl_ptk_info mplsip_tpi = {
752 	/* no tunnel info required for mplsip. */
753 	.proto = htons(ETH_P_MPLS_UC),
754 };
755 #endif
756 
757 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
758 {
759 	const struct iphdr *iph;
760 	struct ip_tunnel *tunnel;
761 	int sifindex;
762 
763 	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
764 
765 	iph = ip_hdr(skb);
766 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
767 				     iph->saddr, iph->daddr, sifindex);
768 	if (tunnel) {
769 		const struct tnl_ptk_info *tpi;
770 
771 		if (tunnel->parms.iph.protocol != ipproto &&
772 		    tunnel->parms.iph.protocol != 0)
773 			goto drop;
774 
775 		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
776 			goto drop;
777 #if IS_ENABLED(CONFIG_MPLS)
778 		if (ipproto == IPPROTO_MPLS)
779 			tpi = &mplsip_tpi;
780 		else
781 #endif
782 			tpi = &ipip_tpi;
783 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
784 			goto drop;
785 		skb_reset_mac_header(skb);
786 
787 		return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
788 	}
789 
790 	return 1;
791 
792 drop:
793 	kfree_skb(skb);
794 	return 0;
795 }
796 
797 static int ipip_rcv(struct sk_buff *skb)
798 {
799 	return sit_tunnel_rcv(skb, IPPROTO_IPIP);
800 }
801 
802 #if IS_ENABLED(CONFIG_MPLS)
803 static int mplsip_rcv(struct sk_buff *skb)
804 {
805 	return sit_tunnel_rcv(skb, IPPROTO_MPLS);
806 }
807 #endif
808 
809 /*
810  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
811  * stores the embedded IPv4 address in v4dst and returns true.
812  */
813 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
814 		      __be32 *v4dst)
815 {
816 #ifdef CONFIG_IPV6_SIT_6RD
817 	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
818 			      tunnel->ip6rd.prefixlen)) {
819 		unsigned int pbw0, pbi0;
820 		int pbi1;
821 		u32 d;
822 
823 		pbw0 = tunnel->ip6rd.prefixlen >> 5;
824 		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
825 
826 		d = tunnel->ip6rd.relay_prefixlen < 32 ?
827 			(ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
828 		    tunnel->ip6rd.relay_prefixlen : 0;
829 
830 		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
831 		if (pbi1 > 0)
832 			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
833 			     (32 - pbi1);
834 
835 		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
836 		return true;
837 	}
838 #else
839 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
840 		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
841 		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
842 		return true;
843 	}
844 #endif
845 	return false;
846 }
847 
848 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
849 			     const struct in6_addr *v6dst)
850 {
851 	__be32 dst = 0;
852 	check_6rd(tunnel, v6dst, &dst);
853 	return dst;
854 }
855 
856 /*
857  *	This function assumes it is being called from dev_queue_xmit()
858  *	and that skb is filled properly by that function.
859  */
860 
861 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
862 				     struct net_device *dev)
863 {
864 	struct ip_tunnel *tunnel = netdev_priv(dev);
865 	const struct iphdr  *tiph = &tunnel->parms.iph;
866 	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
867 	u8     tos = tunnel->parms.iph.tos;
868 	__be16 df = tiph->frag_off;
869 	struct rtable *rt;		/* Route to the other host */
870 	struct net_device *tdev;	/* Device to other host */
871 	unsigned int max_headroom;	/* The extra header space needed */
872 	__be32 dst = tiph->daddr;
873 	struct flowi4 fl4;
874 	int    mtu;
875 	const struct in6_addr *addr6;
876 	int addr_type;
877 	u8 ttl;
878 	u8 protocol = IPPROTO_IPV6;
879 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
880 
881 	if (tos == 1)
882 		tos = ipv6_get_dsfield(iph6);
883 
884 	/* ISATAP (RFC4214) - must come before 6to4 */
885 	if (dev->priv_flags & IFF_ISATAP) {
886 		struct neighbour *neigh = NULL;
887 		bool do_tx_error = false;
888 
889 		if (skb_dst(skb))
890 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
891 
892 		if (!neigh) {
893 			net_dbg_ratelimited("nexthop == NULL\n");
894 			goto tx_error;
895 		}
896 
897 		addr6 = (const struct in6_addr *)&neigh->primary_key;
898 		addr_type = ipv6_addr_type(addr6);
899 
900 		if ((addr_type & IPV6_ADDR_UNICAST) &&
901 		     ipv6_addr_is_isatap(addr6))
902 			dst = addr6->s6_addr32[3];
903 		else
904 			do_tx_error = true;
905 
906 		neigh_release(neigh);
907 		if (do_tx_error)
908 			goto tx_error;
909 	}
910 
911 	if (!dst)
912 		dst = try_6rd(tunnel, &iph6->daddr);
913 
914 	if (!dst) {
915 		struct neighbour *neigh = NULL;
916 		bool do_tx_error = false;
917 
918 		if (skb_dst(skb))
919 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
920 
921 		if (!neigh) {
922 			net_dbg_ratelimited("nexthop == NULL\n");
923 			goto tx_error;
924 		}
925 
926 		addr6 = (const struct in6_addr *)&neigh->primary_key;
927 		addr_type = ipv6_addr_type(addr6);
928 
929 		if (addr_type == IPV6_ADDR_ANY) {
930 			addr6 = &ipv6_hdr(skb)->daddr;
931 			addr_type = ipv6_addr_type(addr6);
932 		}
933 
934 		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
935 			dst = addr6->s6_addr32[3];
936 		else
937 			do_tx_error = true;
938 
939 		neigh_release(neigh);
940 		if (do_tx_error)
941 			goto tx_error;
942 	}
943 
944 	flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
945 			   RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
946 			   0, dst, tiph->saddr, 0, 0,
947 			   sock_net_uid(tunnel->net, NULL));
948 
949 	rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
950 	if (!rt) {
951 		rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
952 		if (IS_ERR(rt)) {
953 			dev->stats.tx_carrier_errors++;
954 			goto tx_error_icmp;
955 		}
956 		dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
957 	}
958 
959 	if (rt->rt_type != RTN_UNICAST) {
960 		ip_rt_put(rt);
961 		dev->stats.tx_carrier_errors++;
962 		goto tx_error_icmp;
963 	}
964 	tdev = rt->dst.dev;
965 
966 	if (tdev == dev) {
967 		ip_rt_put(rt);
968 		dev->stats.collisions++;
969 		goto tx_error;
970 	}
971 
972 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
973 		ip_rt_put(rt);
974 		goto tx_error;
975 	}
976 
977 	if (df) {
978 		mtu = dst_mtu(&rt->dst) - t_hlen;
979 
980 		if (mtu < IPV4_MIN_MTU) {
981 			dev->stats.collisions++;
982 			ip_rt_put(rt);
983 			goto tx_error;
984 		}
985 
986 		if (mtu < IPV6_MIN_MTU) {
987 			mtu = IPV6_MIN_MTU;
988 			df = 0;
989 		}
990 
991 		if (tunnel->parms.iph.daddr)
992 			skb_dst_update_pmtu_no_confirm(skb, mtu);
993 
994 		if (skb->len > mtu && !skb_is_gso(skb)) {
995 			icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
996 			ip_rt_put(rt);
997 			goto tx_error;
998 		}
999 	}
1000 
1001 	if (tunnel->err_count > 0) {
1002 		if (time_before(jiffies,
1003 				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
1004 			tunnel->err_count--;
1005 			dst_link_failure(skb);
1006 		} else
1007 			tunnel->err_count = 0;
1008 	}
1009 
1010 	/*
1011 	 * Okay, now see if we can stuff it in the buffer as-is.
1012 	 */
1013 	max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
1014 
1015 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1016 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1017 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
1018 		if (!new_skb) {
1019 			ip_rt_put(rt);
1020 			dev->stats.tx_dropped++;
1021 			kfree_skb(skb);
1022 			return NETDEV_TX_OK;
1023 		}
1024 		if (skb->sk)
1025 			skb_set_owner_w(new_skb, skb->sk);
1026 		dev_kfree_skb(skb);
1027 		skb = new_skb;
1028 		iph6 = ipv6_hdr(skb);
1029 	}
1030 	ttl = tiph->ttl;
1031 	if (ttl == 0)
1032 		ttl = iph6->hop_limit;
1033 	tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
1034 
1035 	if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
1036 		ip_rt_put(rt);
1037 		goto tx_error;
1038 	}
1039 
1040 	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1041 
1042 	iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
1043 		      df, !net_eq(tunnel->net, dev_net(dev)));
1044 	return NETDEV_TX_OK;
1045 
1046 tx_error_icmp:
1047 	dst_link_failure(skb);
1048 tx_error:
1049 	kfree_skb(skb);
1050 	dev->stats.tx_errors++;
1051 	return NETDEV_TX_OK;
1052 }
1053 
1054 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1055 				     struct net_device *dev, u8 ipproto)
1056 {
1057 	struct ip_tunnel *tunnel = netdev_priv(dev);
1058 	const struct iphdr  *tiph = &tunnel->parms.iph;
1059 
1060 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1061 		goto tx_error;
1062 
1063 	skb_set_inner_ipproto(skb, ipproto);
1064 
1065 	ip_tunnel_xmit(skb, dev, tiph, ipproto);
1066 	return NETDEV_TX_OK;
1067 tx_error:
1068 	kfree_skb(skb);
1069 	dev->stats.tx_errors++;
1070 	return NETDEV_TX_OK;
1071 }
1072 
1073 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1074 				   struct net_device *dev)
1075 {
1076 	if (!pskb_inet_may_pull(skb))
1077 		goto tx_err;
1078 
1079 	switch (skb->protocol) {
1080 	case htons(ETH_P_IP):
1081 		sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1082 		break;
1083 	case htons(ETH_P_IPV6):
1084 		ipip6_tunnel_xmit(skb, dev);
1085 		break;
1086 #if IS_ENABLED(CONFIG_MPLS)
1087 	case htons(ETH_P_MPLS_UC):
1088 		sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1089 		break;
1090 #endif
1091 	default:
1092 		goto tx_err;
1093 	}
1094 
1095 	return NETDEV_TX_OK;
1096 
1097 tx_err:
1098 	dev->stats.tx_errors++;
1099 	kfree_skb(skb);
1100 	return NETDEV_TX_OK;
1101 
1102 }
1103 
1104 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1105 {
1106 	struct net_device *tdev = NULL;
1107 	struct ip_tunnel *tunnel;
1108 	const struct iphdr *iph;
1109 	struct flowi4 fl4;
1110 
1111 	tunnel = netdev_priv(dev);
1112 	iph = &tunnel->parms.iph;
1113 
1114 	if (iph->daddr) {
1115 		struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1116 							  NULL,
1117 							  iph->daddr, iph->saddr,
1118 							  0, 0,
1119 							  IPPROTO_IPV6,
1120 							  RT_TOS(iph->tos),
1121 							  tunnel->parms.link);
1122 
1123 		if (!IS_ERR(rt)) {
1124 			tdev = rt->dst.dev;
1125 			ip_rt_put(rt);
1126 		}
1127 		dev->flags |= IFF_POINTOPOINT;
1128 	}
1129 
1130 	if (!tdev && tunnel->parms.link)
1131 		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1132 
1133 	if (tdev && !netif_is_l3_master(tdev)) {
1134 		int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1135 
1136 		dev->mtu = tdev->mtu - t_hlen;
1137 		if (dev->mtu < IPV6_MIN_MTU)
1138 			dev->mtu = IPV6_MIN_MTU;
1139 	}
1140 }
1141 
1142 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1143 				__u32 fwmark)
1144 {
1145 	struct net *net = t->net;
1146 	struct sit_net *sitn = net_generic(net, sit_net_id);
1147 
1148 	ipip6_tunnel_unlink(sitn, t);
1149 	synchronize_net();
1150 	t->parms.iph.saddr = p->iph.saddr;
1151 	t->parms.iph.daddr = p->iph.daddr;
1152 	memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1153 	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1154 	ipip6_tunnel_link(sitn, t);
1155 	t->parms.iph.ttl = p->iph.ttl;
1156 	t->parms.iph.tos = p->iph.tos;
1157 	t->parms.iph.frag_off = p->iph.frag_off;
1158 	if (t->parms.link != p->link || t->fwmark != fwmark) {
1159 		t->parms.link = p->link;
1160 		t->fwmark = fwmark;
1161 		ipip6_tunnel_bind_dev(t->dev);
1162 	}
1163 	dst_cache_reset(&t->dst_cache);
1164 	netdev_state_change(t->dev);
1165 }
1166 
1167 #ifdef CONFIG_IPV6_SIT_6RD
1168 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1169 				   struct ip_tunnel_6rd *ip6rd)
1170 {
1171 	struct in6_addr prefix;
1172 	__be32 relay_prefix;
1173 
1174 	if (ip6rd->relay_prefixlen > 32 ||
1175 	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1176 		return -EINVAL;
1177 
1178 	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1179 	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1180 		return -EINVAL;
1181 	if (ip6rd->relay_prefixlen)
1182 		relay_prefix = ip6rd->relay_prefix &
1183 			       htonl(0xffffffffUL <<
1184 				     (32 - ip6rd->relay_prefixlen));
1185 	else
1186 		relay_prefix = 0;
1187 	if (relay_prefix != ip6rd->relay_prefix)
1188 		return -EINVAL;
1189 
1190 	t->ip6rd.prefix = prefix;
1191 	t->ip6rd.relay_prefix = relay_prefix;
1192 	t->ip6rd.prefixlen = ip6rd->prefixlen;
1193 	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1194 	dst_cache_reset(&t->dst_cache);
1195 	netdev_state_change(t->dev);
1196 	return 0;
1197 }
1198 
1199 static int
1200 ipip6_tunnel_get6rd(struct net_device *dev, struct ifreq *ifr)
1201 {
1202 	struct ip_tunnel *t = netdev_priv(dev);
1203 	struct ip_tunnel_6rd ip6rd;
1204 	struct ip_tunnel_parm p;
1205 
1206 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1207 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1208 			return -EFAULT;
1209 		t = ipip6_tunnel_locate(t->net, &p, 0);
1210 	}
1211 	if (!t)
1212 		t = netdev_priv(dev);
1213 
1214 	ip6rd.prefix = t->ip6rd.prefix;
1215 	ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1216 	ip6rd.prefixlen = t->ip6rd.prefixlen;
1217 	ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1218 	if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd, sizeof(ip6rd)))
1219 		return -EFAULT;
1220 	return 0;
1221 }
1222 
1223 static int
1224 ipip6_tunnel_6rdctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1225 {
1226 	struct ip_tunnel *t = netdev_priv(dev);
1227 	struct ip_tunnel_6rd ip6rd;
1228 	int err;
1229 
1230 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1231 		return -EPERM;
1232 	if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data, sizeof(ip6rd)))
1233 		return -EFAULT;
1234 
1235 	if (cmd != SIOCDEL6RD) {
1236 		err = ipip6_tunnel_update_6rd(t, &ip6rd);
1237 		if (err < 0)
1238 			return err;
1239 	} else
1240 		ipip6_tunnel_clone_6rd(dev, dev_to_sit_net(dev));
1241 	return 0;
1242 }
1243 
1244 #endif /* CONFIG_IPV6_SIT_6RD */
1245 
1246 static bool ipip6_valid_ip_proto(u8 ipproto)
1247 {
1248 	return ipproto == IPPROTO_IPV6 ||
1249 		ipproto == IPPROTO_IPIP ||
1250 #if IS_ENABLED(CONFIG_MPLS)
1251 		ipproto == IPPROTO_MPLS ||
1252 #endif
1253 		ipproto == 0;
1254 }
1255 
1256 static int
1257 __ipip6_tunnel_ioctl_validate(struct net *net, struct ip_tunnel_parm *p)
1258 {
1259 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1260 		return -EPERM;
1261 
1262 	if (!ipip6_valid_ip_proto(p->iph.protocol))
1263 		return -EINVAL;
1264 	if (p->iph.version != 4 ||
1265 	    p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
1266 		return -EINVAL;
1267 
1268 	if (p->iph.ttl)
1269 		p->iph.frag_off |= htons(IP_DF);
1270 	return 0;
1271 }
1272 
1273 static int
1274 ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm *p)
1275 {
1276 	struct ip_tunnel *t = netdev_priv(dev);
1277 
1278 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
1279 		t = ipip6_tunnel_locate(t->net, p, 0);
1280 	if (!t)
1281 		t = netdev_priv(dev);
1282 	memcpy(p, &t->parms, sizeof(*p));
1283 	return 0;
1284 }
1285 
1286 static int
1287 ipip6_tunnel_add(struct net_device *dev, struct ip_tunnel_parm *p)
1288 {
1289 	struct ip_tunnel *t = netdev_priv(dev);
1290 	int err;
1291 
1292 	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1293 	if (err)
1294 		return err;
1295 
1296 	t = ipip6_tunnel_locate(t->net, p, 1);
1297 	if (!t)
1298 		return -ENOBUFS;
1299 	return 0;
1300 }
1301 
1302 static int
1303 ipip6_tunnel_change(struct net_device *dev, struct ip_tunnel_parm *p)
1304 {
1305 	struct ip_tunnel *t = netdev_priv(dev);
1306 	int err;
1307 
1308 	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1309 	if (err)
1310 		return err;
1311 
1312 	t = ipip6_tunnel_locate(t->net, p, 0);
1313 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1314 		if (!t)
1315 			return -ENOENT;
1316 	} else {
1317 		if (t) {
1318 			if (t->dev != dev)
1319 				return -EEXIST;
1320 		} else {
1321 			if (((dev->flags & IFF_POINTOPOINT) && !p->iph.daddr) ||
1322 			    (!(dev->flags & IFF_POINTOPOINT) && p->iph.daddr))
1323 				return -EINVAL;
1324 			t = netdev_priv(dev);
1325 		}
1326 
1327 		ipip6_tunnel_update(t, p, t->fwmark);
1328 	}
1329 
1330 	return 0;
1331 }
1332 
1333 static int
1334 ipip6_tunnel_del(struct net_device *dev, struct ip_tunnel_parm *p)
1335 {
1336 	struct ip_tunnel *t = netdev_priv(dev);
1337 
1338 	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1339 		return -EPERM;
1340 
1341 	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1342 		t = ipip6_tunnel_locate(t->net, p, 0);
1343 		if (!t)
1344 			return -ENOENT;
1345 		if (t == netdev_priv(dev_to_sit_net(dev)->fb_tunnel_dev))
1346 			return -EPERM;
1347 		dev = t->dev;
1348 	}
1349 	unregister_netdevice(dev);
1350 	return 0;
1351 }
1352 
1353 static int
1354 ipip6_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
1355 {
1356 	switch (cmd) {
1357 	case SIOCGETTUNNEL:
1358 		return ipip6_tunnel_get(dev, p);
1359 	case SIOCADDTUNNEL:
1360 		return ipip6_tunnel_add(dev, p);
1361 	case SIOCCHGTUNNEL:
1362 		return ipip6_tunnel_change(dev, p);
1363 	case SIOCDELTUNNEL:
1364 		return ipip6_tunnel_del(dev, p);
1365 	default:
1366 		return -EINVAL;
1367 	}
1368 }
1369 
1370 static int
1371 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1372 {
1373 	switch (cmd) {
1374 	case SIOCGETTUNNEL:
1375 	case SIOCADDTUNNEL:
1376 	case SIOCCHGTUNNEL:
1377 	case SIOCDELTUNNEL:
1378 		return ip_tunnel_ioctl(dev, ifr, cmd);
1379 	case SIOCGETPRL:
1380 		return ipip6_tunnel_get_prl(dev, ifr);
1381 	case SIOCADDPRL:
1382 	case SIOCDELPRL:
1383 	case SIOCCHGPRL:
1384 		return ipip6_tunnel_prl_ctl(dev, ifr, cmd);
1385 #ifdef CONFIG_IPV6_SIT_6RD
1386 	case SIOCGET6RD:
1387 		return ipip6_tunnel_get6rd(dev, ifr);
1388 	case SIOCADD6RD:
1389 	case SIOCCHG6RD:
1390 	case SIOCDEL6RD:
1391 		return ipip6_tunnel_6rdctl(dev, ifr, cmd);
1392 #endif
1393 	default:
1394 		return -EINVAL;
1395 	}
1396 }
1397 
1398 static const struct net_device_ops ipip6_netdev_ops = {
1399 	.ndo_init	= ipip6_tunnel_init,
1400 	.ndo_uninit	= ipip6_tunnel_uninit,
1401 	.ndo_start_xmit	= sit_tunnel_xmit,
1402 	.ndo_do_ioctl	= ipip6_tunnel_ioctl,
1403 	.ndo_get_stats64 = dev_get_tstats64,
1404 	.ndo_get_iflink = ip_tunnel_get_iflink,
1405 	.ndo_tunnel_ctl = ipip6_tunnel_ctl,
1406 };
1407 
1408 static void ipip6_dev_free(struct net_device *dev)
1409 {
1410 	struct ip_tunnel *tunnel = netdev_priv(dev);
1411 
1412 	dst_cache_destroy(&tunnel->dst_cache);
1413 	free_percpu(dev->tstats);
1414 }
1415 
1416 #define SIT_FEATURES (NETIF_F_SG	   | \
1417 		      NETIF_F_FRAGLIST	   | \
1418 		      NETIF_F_HIGHDMA	   | \
1419 		      NETIF_F_GSO_SOFTWARE | \
1420 		      NETIF_F_HW_CSUM)
1421 
1422 static void ipip6_tunnel_setup(struct net_device *dev)
1423 {
1424 	struct ip_tunnel *tunnel = netdev_priv(dev);
1425 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1426 
1427 	dev->netdev_ops		= &ipip6_netdev_ops;
1428 	dev->header_ops		= &ip_tunnel_header_ops;
1429 	dev->needs_free_netdev	= true;
1430 	dev->priv_destructor	= ipip6_dev_free;
1431 
1432 	dev->type		= ARPHRD_SIT;
1433 	dev->mtu		= ETH_DATA_LEN - t_hlen;
1434 	dev->min_mtu		= IPV6_MIN_MTU;
1435 	dev->max_mtu		= IP6_MAX_MTU - t_hlen;
1436 	dev->flags		= IFF_NOARP;
1437 	netif_keep_dst(dev);
1438 	dev->addr_len		= 4;
1439 	dev->features		|= NETIF_F_LLTX;
1440 	dev->features		|= SIT_FEATURES;
1441 	dev->hw_features	|= SIT_FEATURES;
1442 }
1443 
1444 static int ipip6_tunnel_init(struct net_device *dev)
1445 {
1446 	struct ip_tunnel *tunnel = netdev_priv(dev);
1447 	int err;
1448 
1449 	tunnel->dev = dev;
1450 	tunnel->net = dev_net(dev);
1451 	strcpy(tunnel->parms.name, dev->name);
1452 
1453 	ipip6_tunnel_bind_dev(dev);
1454 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1455 	if (!dev->tstats)
1456 		return -ENOMEM;
1457 
1458 	err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1459 	if (err) {
1460 		free_percpu(dev->tstats);
1461 		dev->tstats = NULL;
1462 		return err;
1463 	}
1464 	dev_hold(dev);
1465 	return 0;
1466 }
1467 
1468 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1469 {
1470 	struct ip_tunnel *tunnel = netdev_priv(dev);
1471 	struct iphdr *iph = &tunnel->parms.iph;
1472 	struct net *net = dev_net(dev);
1473 	struct sit_net *sitn = net_generic(net, sit_net_id);
1474 
1475 	iph->version		= 4;
1476 	iph->protocol		= IPPROTO_IPV6;
1477 	iph->ihl		= 5;
1478 	iph->ttl		= 64;
1479 
1480 	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1481 }
1482 
1483 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1484 			  struct netlink_ext_ack *extack)
1485 {
1486 	u8 proto;
1487 
1488 	if (!data || !data[IFLA_IPTUN_PROTO])
1489 		return 0;
1490 
1491 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1492 	if (!ipip6_valid_ip_proto(proto))
1493 		return -EINVAL;
1494 
1495 	return 0;
1496 }
1497 
1498 static void ipip6_netlink_parms(struct nlattr *data[],
1499 				struct ip_tunnel_parm *parms,
1500 				__u32 *fwmark)
1501 {
1502 	memset(parms, 0, sizeof(*parms));
1503 
1504 	parms->iph.version = 4;
1505 	parms->iph.protocol = IPPROTO_IPV6;
1506 	parms->iph.ihl = 5;
1507 	parms->iph.ttl = 64;
1508 
1509 	if (!data)
1510 		return;
1511 
1512 	if (data[IFLA_IPTUN_LINK])
1513 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1514 
1515 	if (data[IFLA_IPTUN_LOCAL])
1516 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1517 
1518 	if (data[IFLA_IPTUN_REMOTE])
1519 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1520 
1521 	if (data[IFLA_IPTUN_TTL]) {
1522 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1523 		if (parms->iph.ttl)
1524 			parms->iph.frag_off = htons(IP_DF);
1525 	}
1526 
1527 	if (data[IFLA_IPTUN_TOS])
1528 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1529 
1530 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1531 		parms->iph.frag_off = htons(IP_DF);
1532 
1533 	if (data[IFLA_IPTUN_FLAGS])
1534 		parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1535 
1536 	if (data[IFLA_IPTUN_PROTO])
1537 		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1538 
1539 	if (data[IFLA_IPTUN_FWMARK])
1540 		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1541 }
1542 
1543 /* This function returns true when ENCAP attributes are present in the nl msg */
1544 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1545 				      struct ip_tunnel_encap *ipencap)
1546 {
1547 	bool ret = false;
1548 
1549 	memset(ipencap, 0, sizeof(*ipencap));
1550 
1551 	if (!data)
1552 		return ret;
1553 
1554 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1555 		ret = true;
1556 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1557 	}
1558 
1559 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1560 		ret = true;
1561 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1562 	}
1563 
1564 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1565 		ret = true;
1566 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1567 	}
1568 
1569 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1570 		ret = true;
1571 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1572 	}
1573 
1574 	return ret;
1575 }
1576 
1577 #ifdef CONFIG_IPV6_SIT_6RD
1578 /* This function returns true when 6RD attributes are present in the nl msg */
1579 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1580 				    struct ip_tunnel_6rd *ip6rd)
1581 {
1582 	bool ret = false;
1583 	memset(ip6rd, 0, sizeof(*ip6rd));
1584 
1585 	if (!data)
1586 		return ret;
1587 
1588 	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1589 		ret = true;
1590 		ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1591 	}
1592 
1593 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1594 		ret = true;
1595 		ip6rd->relay_prefix =
1596 			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1597 	}
1598 
1599 	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1600 		ret = true;
1601 		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1602 	}
1603 
1604 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1605 		ret = true;
1606 		ip6rd->relay_prefixlen =
1607 			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1608 	}
1609 
1610 	return ret;
1611 }
1612 #endif
1613 
1614 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1615 			 struct nlattr *tb[], struct nlattr *data[],
1616 			 struct netlink_ext_ack *extack)
1617 {
1618 	struct net *net = dev_net(dev);
1619 	struct ip_tunnel *nt;
1620 	struct ip_tunnel_encap ipencap;
1621 #ifdef CONFIG_IPV6_SIT_6RD
1622 	struct ip_tunnel_6rd ip6rd;
1623 #endif
1624 	int err;
1625 
1626 	nt = netdev_priv(dev);
1627 
1628 	if (ipip6_netlink_encap_parms(data, &ipencap)) {
1629 		err = ip_tunnel_encap_setup(nt, &ipencap);
1630 		if (err < 0)
1631 			return err;
1632 	}
1633 
1634 	ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1635 
1636 	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1637 		return -EEXIST;
1638 
1639 	err = ipip6_tunnel_create(dev);
1640 	if (err < 0)
1641 		return err;
1642 
1643 	if (tb[IFLA_MTU]) {
1644 		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1645 
1646 		if (mtu >= IPV6_MIN_MTU &&
1647 		    mtu <= IP6_MAX_MTU - dev->hard_header_len)
1648 			dev->mtu = mtu;
1649 	}
1650 
1651 #ifdef CONFIG_IPV6_SIT_6RD
1652 	if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1653 		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1654 		if (err < 0)
1655 			unregister_netdevice_queue(dev, NULL);
1656 	}
1657 #endif
1658 
1659 	return err;
1660 }
1661 
1662 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1663 			    struct nlattr *data[],
1664 			    struct netlink_ext_ack *extack)
1665 {
1666 	struct ip_tunnel *t = netdev_priv(dev);
1667 	struct ip_tunnel_parm p;
1668 	struct ip_tunnel_encap ipencap;
1669 	struct net *net = t->net;
1670 	struct sit_net *sitn = net_generic(net, sit_net_id);
1671 #ifdef CONFIG_IPV6_SIT_6RD
1672 	struct ip_tunnel_6rd ip6rd;
1673 #endif
1674 	__u32 fwmark = t->fwmark;
1675 	int err;
1676 
1677 	if (dev == sitn->fb_tunnel_dev)
1678 		return -EINVAL;
1679 
1680 	if (ipip6_netlink_encap_parms(data, &ipencap)) {
1681 		err = ip_tunnel_encap_setup(t, &ipencap);
1682 		if (err < 0)
1683 			return err;
1684 	}
1685 
1686 	ipip6_netlink_parms(data, &p, &fwmark);
1687 
1688 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1689 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1690 		return -EINVAL;
1691 
1692 	t = ipip6_tunnel_locate(net, &p, 0);
1693 
1694 	if (t) {
1695 		if (t->dev != dev)
1696 			return -EEXIST;
1697 	} else
1698 		t = netdev_priv(dev);
1699 
1700 	ipip6_tunnel_update(t, &p, fwmark);
1701 
1702 #ifdef CONFIG_IPV6_SIT_6RD
1703 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1704 		return ipip6_tunnel_update_6rd(t, &ip6rd);
1705 #endif
1706 
1707 	return 0;
1708 }
1709 
1710 static size_t ipip6_get_size(const struct net_device *dev)
1711 {
1712 	return
1713 		/* IFLA_IPTUN_LINK */
1714 		nla_total_size(4) +
1715 		/* IFLA_IPTUN_LOCAL */
1716 		nla_total_size(4) +
1717 		/* IFLA_IPTUN_REMOTE */
1718 		nla_total_size(4) +
1719 		/* IFLA_IPTUN_TTL */
1720 		nla_total_size(1) +
1721 		/* IFLA_IPTUN_TOS */
1722 		nla_total_size(1) +
1723 		/* IFLA_IPTUN_PMTUDISC */
1724 		nla_total_size(1) +
1725 		/* IFLA_IPTUN_FLAGS */
1726 		nla_total_size(2) +
1727 		/* IFLA_IPTUN_PROTO */
1728 		nla_total_size(1) +
1729 #ifdef CONFIG_IPV6_SIT_6RD
1730 		/* IFLA_IPTUN_6RD_PREFIX */
1731 		nla_total_size(sizeof(struct in6_addr)) +
1732 		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1733 		nla_total_size(4) +
1734 		/* IFLA_IPTUN_6RD_PREFIXLEN */
1735 		nla_total_size(2) +
1736 		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1737 		nla_total_size(2) +
1738 #endif
1739 		/* IFLA_IPTUN_ENCAP_TYPE */
1740 		nla_total_size(2) +
1741 		/* IFLA_IPTUN_ENCAP_FLAGS */
1742 		nla_total_size(2) +
1743 		/* IFLA_IPTUN_ENCAP_SPORT */
1744 		nla_total_size(2) +
1745 		/* IFLA_IPTUN_ENCAP_DPORT */
1746 		nla_total_size(2) +
1747 		/* IFLA_IPTUN_FWMARK */
1748 		nla_total_size(4) +
1749 		0;
1750 }
1751 
1752 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1753 {
1754 	struct ip_tunnel *tunnel = netdev_priv(dev);
1755 	struct ip_tunnel_parm *parm = &tunnel->parms;
1756 
1757 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1758 	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1759 	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1760 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1761 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1762 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1763 		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1764 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1765 	    nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1766 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1767 		goto nla_put_failure;
1768 
1769 #ifdef CONFIG_IPV6_SIT_6RD
1770 	if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1771 			     &tunnel->ip6rd.prefix) ||
1772 	    nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1773 			    tunnel->ip6rd.relay_prefix) ||
1774 	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1775 			tunnel->ip6rd.prefixlen) ||
1776 	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1777 			tunnel->ip6rd.relay_prefixlen))
1778 		goto nla_put_failure;
1779 #endif
1780 
1781 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1782 			tunnel->encap.type) ||
1783 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1784 			tunnel->encap.sport) ||
1785 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1786 			tunnel->encap.dport) ||
1787 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1788 			tunnel->encap.flags))
1789 		goto nla_put_failure;
1790 
1791 	return 0;
1792 
1793 nla_put_failure:
1794 	return -EMSGSIZE;
1795 }
1796 
1797 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1798 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1799 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1800 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1801 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1802 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1803 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1804 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1805 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1806 #ifdef CONFIG_IPV6_SIT_6RD
1807 	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1808 	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1809 	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1810 	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1811 #endif
1812 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
1813 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
1814 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
1815 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
1816 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
1817 };
1818 
1819 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1820 {
1821 	struct net *net = dev_net(dev);
1822 	struct sit_net *sitn = net_generic(net, sit_net_id);
1823 
1824 	if (dev != sitn->fb_tunnel_dev)
1825 		unregister_netdevice_queue(dev, head);
1826 }
1827 
1828 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1829 	.kind		= "sit",
1830 	.maxtype	= IFLA_IPTUN_MAX,
1831 	.policy		= ipip6_policy,
1832 	.priv_size	= sizeof(struct ip_tunnel),
1833 	.setup		= ipip6_tunnel_setup,
1834 	.validate	= ipip6_validate,
1835 	.newlink	= ipip6_newlink,
1836 	.changelink	= ipip6_changelink,
1837 	.get_size	= ipip6_get_size,
1838 	.fill_info	= ipip6_fill_info,
1839 	.dellink	= ipip6_dellink,
1840 	.get_link_net	= ip_tunnel_get_link_net,
1841 };
1842 
1843 static struct xfrm_tunnel sit_handler __read_mostly = {
1844 	.handler	=	ipip6_rcv,
1845 	.err_handler	=	ipip6_err,
1846 	.priority	=	1,
1847 };
1848 
1849 static struct xfrm_tunnel ipip_handler __read_mostly = {
1850 	.handler	=	ipip_rcv,
1851 	.err_handler	=	ipip6_err,
1852 	.priority	=	2,
1853 };
1854 
1855 #if IS_ENABLED(CONFIG_MPLS)
1856 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1857 	.handler	=	mplsip_rcv,
1858 	.err_handler	=	ipip6_err,
1859 	.priority	=	2,
1860 };
1861 #endif
1862 
1863 static void __net_exit sit_destroy_tunnels(struct net *net,
1864 					   struct list_head *head)
1865 {
1866 	struct sit_net *sitn = net_generic(net, sit_net_id);
1867 	struct net_device *dev, *aux;
1868 	int prio;
1869 
1870 	for_each_netdev_safe(net, dev, aux)
1871 		if (dev->rtnl_link_ops == &sit_link_ops)
1872 			unregister_netdevice_queue(dev, head);
1873 
1874 	for (prio = 0; prio < 4; prio++) {
1875 		int h;
1876 		for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1877 			struct ip_tunnel *t;
1878 
1879 			t = rtnl_dereference(sitn->tunnels[prio][h]);
1880 			while (t) {
1881 				/* If dev is in the same netns, it has already
1882 				 * been added to the list by the previous loop.
1883 				 */
1884 				if (!net_eq(dev_net(t->dev), net))
1885 					unregister_netdevice_queue(t->dev,
1886 								   head);
1887 				t = rtnl_dereference(t->next);
1888 			}
1889 		}
1890 	}
1891 }
1892 
1893 static int __net_init sit_init_net(struct net *net)
1894 {
1895 	struct sit_net *sitn = net_generic(net, sit_net_id);
1896 	struct ip_tunnel *t;
1897 	int err;
1898 
1899 	sitn->tunnels[0] = sitn->tunnels_wc;
1900 	sitn->tunnels[1] = sitn->tunnels_l;
1901 	sitn->tunnels[2] = sitn->tunnels_r;
1902 	sitn->tunnels[3] = sitn->tunnels_r_l;
1903 
1904 	if (!net_has_fallback_tunnels(net))
1905 		return 0;
1906 
1907 	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1908 					   NET_NAME_UNKNOWN,
1909 					   ipip6_tunnel_setup);
1910 	if (!sitn->fb_tunnel_dev) {
1911 		err = -ENOMEM;
1912 		goto err_alloc_dev;
1913 	}
1914 	dev_net_set(sitn->fb_tunnel_dev, net);
1915 	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1916 	/* FB netdevice is special: we have one, and only one per netns.
1917 	 * Allowing to move it to another netns is clearly unsafe.
1918 	 */
1919 	sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1920 
1921 	err = register_netdev(sitn->fb_tunnel_dev);
1922 	if (err)
1923 		goto err_reg_dev;
1924 
1925 	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1926 	ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1927 
1928 	t = netdev_priv(sitn->fb_tunnel_dev);
1929 
1930 	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1931 	return 0;
1932 
1933 err_reg_dev:
1934 	ipip6_dev_free(sitn->fb_tunnel_dev);
1935 	free_netdev(sitn->fb_tunnel_dev);
1936 err_alloc_dev:
1937 	return err;
1938 }
1939 
1940 static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1941 {
1942 	LIST_HEAD(list);
1943 	struct net *net;
1944 
1945 	rtnl_lock();
1946 	list_for_each_entry(net, net_list, exit_list)
1947 		sit_destroy_tunnels(net, &list);
1948 
1949 	unregister_netdevice_many(&list);
1950 	rtnl_unlock();
1951 }
1952 
1953 static struct pernet_operations sit_net_ops = {
1954 	.init = sit_init_net,
1955 	.exit_batch = sit_exit_batch_net,
1956 	.id   = &sit_net_id,
1957 	.size = sizeof(struct sit_net),
1958 };
1959 
1960 static void __exit sit_cleanup(void)
1961 {
1962 	rtnl_link_unregister(&sit_link_ops);
1963 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1964 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1965 #if IS_ENABLED(CONFIG_MPLS)
1966 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1967 #endif
1968 
1969 	unregister_pernet_device(&sit_net_ops);
1970 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1971 }
1972 
1973 static int __init sit_init(void)
1974 {
1975 	int err;
1976 
1977 	pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1978 
1979 	err = register_pernet_device(&sit_net_ops);
1980 	if (err < 0)
1981 		return err;
1982 	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1983 	if (err < 0) {
1984 		pr_info("%s: can't register ip6ip4\n", __func__);
1985 		goto xfrm_tunnel_failed;
1986 	}
1987 	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1988 	if (err < 0) {
1989 		pr_info("%s: can't register ip4ip4\n", __func__);
1990 		goto xfrm_tunnel4_failed;
1991 	}
1992 #if IS_ENABLED(CONFIG_MPLS)
1993 	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1994 	if (err < 0) {
1995 		pr_info("%s: can't register mplsip\n", __func__);
1996 		goto xfrm_tunnel_mpls_failed;
1997 	}
1998 #endif
1999 	err = rtnl_link_register(&sit_link_ops);
2000 	if (err < 0)
2001 		goto rtnl_link_failed;
2002 
2003 out:
2004 	return err;
2005 
2006 rtnl_link_failed:
2007 #if IS_ENABLED(CONFIG_MPLS)
2008 	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
2009 xfrm_tunnel_mpls_failed:
2010 #endif
2011 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
2012 xfrm_tunnel4_failed:
2013 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
2014 xfrm_tunnel_failed:
2015 	unregister_pernet_device(&sit_net_ops);
2016 	goto out;
2017 }
2018 
2019 module_init(sit_init);
2020 module_exit(sit_cleanup);
2021 MODULE_LICENSE("GPL");
2022 MODULE_ALIAS_RTNL_LINK("sit");
2023 MODULE_ALIAS_NETDEV("sit0");
2024