xref: /openbmc/linux/net/ipv6/sit.c (revision bc000245)
1 /*
2  *	IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
8  *
9  *	This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  *
14  *	Changes:
15  * Roger Venning <r.venning@telstra.com>:	6to4 support
16  * Nate Thompson <nate@thebog.net>:		6to4 support
17  * Fred Templin <fred.l.templin@boeing.com>:	isatap support
18  */
19 
20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
21 
22 #include <linux/module.h>
23 #include <linux/capability.h>
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/icmp.h>
33 #include <linux/slab.h>
34 #include <asm/uaccess.h>
35 #include <linux/init.h>
36 #include <linux/netfilter_ipv4.h>
37 #include <linux/if_ether.h>
38 
39 #include <net/sock.h>
40 #include <net/snmp.h>
41 
42 #include <net/ipv6.h>
43 #include <net/protocol.h>
44 #include <net/transp_v6.h>
45 #include <net/ip6_fib.h>
46 #include <net/ip6_route.h>
47 #include <net/ndisc.h>
48 #include <net/addrconf.h>
49 #include <net/ip.h>
50 #include <net/udp.h>
51 #include <net/icmp.h>
52 #include <net/ip_tunnels.h>
53 #include <net/inet_ecn.h>
54 #include <net/xfrm.h>
55 #include <net/dsfield.h>
56 #include <net/net_namespace.h>
57 #include <net/netns/generic.h>
58 
59 /*
60    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
61 
62    For comments look at net/ipv4/ip_gre.c --ANK
63  */
64 
65 #define HASH_SIZE  16
66 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
67 
68 static bool log_ecn_error = true;
69 module_param(log_ecn_error, bool, 0644);
70 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
71 
72 static int ipip6_tunnel_init(struct net_device *dev);
73 static void ipip6_tunnel_setup(struct net_device *dev);
74 static void ipip6_dev_free(struct net_device *dev);
75 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
76 		      __be32 *v4dst);
77 static struct rtnl_link_ops sit_link_ops __read_mostly;
78 
79 static int sit_net_id __read_mostly;
80 struct sit_net {
81 	struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
82 	struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
83 	struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
84 	struct ip_tunnel __rcu *tunnels_wc[1];
85 	struct ip_tunnel __rcu **tunnels[4];
86 
87 	struct net_device *fb_tunnel_dev;
88 };
89 
90 /*
91  * Must be invoked with rcu_read_lock
92  */
93 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
94 		struct net_device *dev, __be32 remote, __be32 local)
95 {
96 	unsigned int h0 = HASH(remote);
97 	unsigned int h1 = HASH(local);
98 	struct ip_tunnel *t;
99 	struct sit_net *sitn = net_generic(net, sit_net_id);
100 
101 	for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
102 		if (local == t->parms.iph.saddr &&
103 		    remote == t->parms.iph.daddr &&
104 		    (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
105 		    (t->dev->flags & IFF_UP))
106 			return t;
107 	}
108 	for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
109 		if (remote == t->parms.iph.daddr &&
110 		    (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
111 		    (t->dev->flags & IFF_UP))
112 			return t;
113 	}
114 	for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
115 		if (local == t->parms.iph.saddr &&
116 		    (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
117 		    (t->dev->flags & IFF_UP))
118 			return t;
119 	}
120 	t = rcu_dereference(sitn->tunnels_wc[0]);
121 	if ((t != NULL) && (t->dev->flags & IFF_UP))
122 		return t;
123 	return NULL;
124 }
125 
126 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
127 		struct ip_tunnel_parm *parms)
128 {
129 	__be32 remote = parms->iph.daddr;
130 	__be32 local = parms->iph.saddr;
131 	unsigned int h = 0;
132 	int prio = 0;
133 
134 	if (remote) {
135 		prio |= 2;
136 		h ^= HASH(remote);
137 	}
138 	if (local) {
139 		prio |= 1;
140 		h ^= HASH(local);
141 	}
142 	return &sitn->tunnels[prio][h];
143 }
144 
145 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
146 		struct ip_tunnel *t)
147 {
148 	return __ipip6_bucket(sitn, &t->parms);
149 }
150 
151 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
152 {
153 	struct ip_tunnel __rcu **tp;
154 	struct ip_tunnel *iter;
155 
156 	for (tp = ipip6_bucket(sitn, t);
157 	     (iter = rtnl_dereference(*tp)) != NULL;
158 	     tp = &iter->next) {
159 		if (t == iter) {
160 			rcu_assign_pointer(*tp, t->next);
161 			break;
162 		}
163 	}
164 }
165 
166 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
167 {
168 	struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
169 
170 	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
171 	rcu_assign_pointer(*tp, t);
172 }
173 
174 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
175 {
176 #ifdef CONFIG_IPV6_SIT_6RD
177 	struct ip_tunnel *t = netdev_priv(dev);
178 
179 	if (t->dev == sitn->fb_tunnel_dev) {
180 		ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
181 		t->ip6rd.relay_prefix = 0;
182 		t->ip6rd.prefixlen = 16;
183 		t->ip6rd.relay_prefixlen = 0;
184 	} else {
185 		struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
186 		memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
187 	}
188 #endif
189 }
190 
191 static int ipip6_tunnel_create(struct net_device *dev)
192 {
193 	struct ip_tunnel *t = netdev_priv(dev);
194 	struct net *net = dev_net(dev);
195 	struct sit_net *sitn = net_generic(net, sit_net_id);
196 	int err;
197 
198 	err = ipip6_tunnel_init(dev);
199 	if (err < 0)
200 		goto out;
201 	ipip6_tunnel_clone_6rd(dev, sitn);
202 
203 	if ((__force u16)t->parms.i_flags & SIT_ISATAP)
204 		dev->priv_flags |= IFF_ISATAP;
205 
206 	err = register_netdevice(dev);
207 	if (err < 0)
208 		goto out;
209 
210 	strcpy(t->parms.name, dev->name);
211 	dev->rtnl_link_ops = &sit_link_ops;
212 
213 	dev_hold(dev);
214 
215 	ipip6_tunnel_link(sitn, t);
216 	return 0;
217 
218 out:
219 	return err;
220 }
221 
222 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
223 		struct ip_tunnel_parm *parms, int create)
224 {
225 	__be32 remote = parms->iph.daddr;
226 	__be32 local = parms->iph.saddr;
227 	struct ip_tunnel *t, *nt;
228 	struct ip_tunnel __rcu **tp;
229 	struct net_device *dev;
230 	char name[IFNAMSIZ];
231 	struct sit_net *sitn = net_generic(net, sit_net_id);
232 
233 	for (tp = __ipip6_bucket(sitn, parms);
234 	    (t = rtnl_dereference(*tp)) != NULL;
235 	     tp = &t->next) {
236 		if (local == t->parms.iph.saddr &&
237 		    remote == t->parms.iph.daddr &&
238 		    parms->link == t->parms.link) {
239 			if (create)
240 				return NULL;
241 			else
242 				return t;
243 		}
244 	}
245 	if (!create)
246 		goto failed;
247 
248 	if (parms->name[0])
249 		strlcpy(name, parms->name, IFNAMSIZ);
250 	else
251 		strcpy(name, "sit%d");
252 
253 	dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
254 	if (dev == NULL)
255 		return NULL;
256 
257 	dev_net_set(dev, net);
258 
259 	nt = netdev_priv(dev);
260 
261 	nt->parms = *parms;
262 	if (ipip6_tunnel_create(dev) < 0)
263 		goto failed_free;
264 
265 	return nt;
266 
267 failed_free:
268 	ipip6_dev_free(dev);
269 failed:
270 	return NULL;
271 }
272 
273 #define for_each_prl_rcu(start)			\
274 	for (prl = rcu_dereference(start);	\
275 	     prl;				\
276 	     prl = rcu_dereference(prl->next))
277 
278 static struct ip_tunnel_prl_entry *
279 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
280 {
281 	struct ip_tunnel_prl_entry *prl;
282 
283 	for_each_prl_rcu(t->prl)
284 		if (prl->addr == addr)
285 			break;
286 	return prl;
287 
288 }
289 
290 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
291 				struct ip_tunnel_prl __user *a)
292 {
293 	struct ip_tunnel_prl kprl, *kp;
294 	struct ip_tunnel_prl_entry *prl;
295 	unsigned int cmax, c = 0, ca, len;
296 	int ret = 0;
297 
298 	if (copy_from_user(&kprl, a, sizeof(kprl)))
299 		return -EFAULT;
300 	cmax = kprl.datalen / sizeof(kprl);
301 	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
302 		cmax = 1;
303 
304 	/* For simple GET or for root users,
305 	 * we try harder to allocate.
306 	 */
307 	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
308 		kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
309 		NULL;
310 
311 	rcu_read_lock();
312 
313 	ca = t->prl_count < cmax ? t->prl_count : cmax;
314 
315 	if (!kp) {
316 		/* We don't try hard to allocate much memory for
317 		 * non-root users.
318 		 * For root users, retry allocating enough memory for
319 		 * the answer.
320 		 */
321 		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
322 		if (!kp) {
323 			ret = -ENOMEM;
324 			goto out;
325 		}
326 	}
327 
328 	c = 0;
329 	for_each_prl_rcu(t->prl) {
330 		if (c >= cmax)
331 			break;
332 		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
333 			continue;
334 		kp[c].addr = prl->addr;
335 		kp[c].flags = prl->flags;
336 		c++;
337 		if (kprl.addr != htonl(INADDR_ANY))
338 			break;
339 	}
340 out:
341 	rcu_read_unlock();
342 
343 	len = sizeof(*kp) * c;
344 	ret = 0;
345 	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
346 		ret = -EFAULT;
347 
348 	kfree(kp);
349 
350 	return ret;
351 }
352 
353 static int
354 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
355 {
356 	struct ip_tunnel_prl_entry *p;
357 	int err = 0;
358 
359 	if (a->addr == htonl(INADDR_ANY))
360 		return -EINVAL;
361 
362 	ASSERT_RTNL();
363 
364 	for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
365 		if (p->addr == a->addr) {
366 			if (chg) {
367 				p->flags = a->flags;
368 				goto out;
369 			}
370 			err = -EEXIST;
371 			goto out;
372 		}
373 	}
374 
375 	if (chg) {
376 		err = -ENXIO;
377 		goto out;
378 	}
379 
380 	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
381 	if (!p) {
382 		err = -ENOBUFS;
383 		goto out;
384 	}
385 
386 	p->next = t->prl;
387 	p->addr = a->addr;
388 	p->flags = a->flags;
389 	t->prl_count++;
390 	rcu_assign_pointer(t->prl, p);
391 out:
392 	return err;
393 }
394 
395 static void prl_list_destroy_rcu(struct rcu_head *head)
396 {
397 	struct ip_tunnel_prl_entry *p, *n;
398 
399 	p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
400 	do {
401 		n = rcu_dereference_protected(p->next, 1);
402 		kfree(p);
403 		p = n;
404 	} while (p);
405 }
406 
407 static int
408 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
409 {
410 	struct ip_tunnel_prl_entry *x;
411 	struct ip_tunnel_prl_entry __rcu **p;
412 	int err = 0;
413 
414 	ASSERT_RTNL();
415 
416 	if (a && a->addr != htonl(INADDR_ANY)) {
417 		for (p = &t->prl;
418 		     (x = rtnl_dereference(*p)) != NULL;
419 		     p = &x->next) {
420 			if (x->addr == a->addr) {
421 				*p = x->next;
422 				kfree_rcu(x, rcu_head);
423 				t->prl_count--;
424 				goto out;
425 			}
426 		}
427 		err = -ENXIO;
428 	} else {
429 		x = rtnl_dereference(t->prl);
430 		if (x) {
431 			t->prl_count = 0;
432 			call_rcu(&x->rcu_head, prl_list_destroy_rcu);
433 			t->prl = NULL;
434 		}
435 	}
436 out:
437 	return err;
438 }
439 
440 static int
441 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
442 {
443 	struct ip_tunnel_prl_entry *p;
444 	int ok = 1;
445 
446 	rcu_read_lock();
447 	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
448 	if (p) {
449 		if (p->flags & PRL_DEFAULT)
450 			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
451 		else
452 			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
453 	} else {
454 		const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
455 
456 		if (ipv6_addr_is_isatap(addr6) &&
457 		    (addr6->s6_addr32[3] == iph->saddr) &&
458 		    ipv6_chk_prefix(addr6, t->dev))
459 			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
460 		else
461 			ok = 0;
462 	}
463 	rcu_read_unlock();
464 	return ok;
465 }
466 
467 static void ipip6_tunnel_uninit(struct net_device *dev)
468 {
469 	struct ip_tunnel *tunnel = netdev_priv(dev);
470 	struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
471 
472 	if (dev == sitn->fb_tunnel_dev) {
473 		RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
474 	} else {
475 		ipip6_tunnel_unlink(sitn, tunnel);
476 		ipip6_tunnel_del_prl(tunnel, NULL);
477 	}
478 	dev_put(dev);
479 }
480 
481 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
482  * if sufficient data bytes are available
483  */
484 static int ipip6_err_gen_icmpv6_unreach(struct sk_buff *skb)
485 {
486 	const struct iphdr *iph = (const struct iphdr *) skb->data;
487 	struct rt6_info *rt;
488 	struct sk_buff *skb2;
489 
490 	if (!pskb_may_pull(skb, iph->ihl * 4 + sizeof(struct ipv6hdr) + 8))
491 		return 1;
492 
493 	skb2 = skb_clone(skb, GFP_ATOMIC);
494 
495 	if (!skb2)
496 		return 1;
497 
498 	skb_dst_drop(skb2);
499 	skb_pull(skb2, iph->ihl * 4);
500 	skb_reset_network_header(skb2);
501 
502 	rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 0);
503 
504 	if (rt && rt->dst.dev)
505 		skb2->dev = rt->dst.dev;
506 
507 	icmpv6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
508 
509 	if (rt)
510 		ip6_rt_put(rt);
511 
512 	kfree_skb(skb2);
513 
514 	return 0;
515 }
516 
517 static int ipip6_err(struct sk_buff *skb, u32 info)
518 {
519 	const struct iphdr *iph = (const struct iphdr *)skb->data;
520 	const int type = icmp_hdr(skb)->type;
521 	const int code = icmp_hdr(skb)->code;
522 	struct ip_tunnel *t;
523 	int err;
524 
525 	switch (type) {
526 	default:
527 	case ICMP_PARAMETERPROB:
528 		return 0;
529 
530 	case ICMP_DEST_UNREACH:
531 		switch (code) {
532 		case ICMP_SR_FAILED:
533 			/* Impossible event. */
534 			return 0;
535 		default:
536 			/* All others are translated to HOST_UNREACH.
537 			   rfc2003 contains "deep thoughts" about NET_UNREACH,
538 			   I believe they are just ether pollution. --ANK
539 			 */
540 			break;
541 		}
542 		break;
543 	case ICMP_TIME_EXCEEDED:
544 		if (code != ICMP_EXC_TTL)
545 			return 0;
546 		break;
547 	case ICMP_REDIRECT:
548 		break;
549 	}
550 
551 	err = -ENOENT;
552 
553 	t = ipip6_tunnel_lookup(dev_net(skb->dev),
554 				skb->dev,
555 				iph->daddr,
556 				iph->saddr);
557 	if (t == NULL)
558 		goto out;
559 
560 	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
561 		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
562 				 t->dev->ifindex, 0, IPPROTO_IPV6, 0);
563 		err = 0;
564 		goto out;
565 	}
566 	if (type == ICMP_REDIRECT) {
567 		ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
568 			      IPPROTO_IPV6, 0);
569 		err = 0;
570 		goto out;
571 	}
572 
573 	if (t->parms.iph.daddr == 0)
574 		goto out;
575 
576 	err = 0;
577 	if (!ipip6_err_gen_icmpv6_unreach(skb))
578 		goto out;
579 
580 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
581 		goto out;
582 
583 	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
584 		t->err_count++;
585 	else
586 		t->err_count = 1;
587 	t->err_time = jiffies;
588 out:
589 	return err;
590 }
591 
592 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
593 				  const struct in6_addr *v6addr)
594 {
595 	__be32 v4embed = 0;
596 	if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
597 		return true;
598 	return false;
599 }
600 
601 /* Checks if an address matches an address on the tunnel interface.
602  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
603  * Long story:
604  * This function is called after we considered the packet as spoofed
605  * in is_spoofed_6rd.
606  * We may have a router that is doing NAT for proto 41 packets
607  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
608  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
609  * function will return true, dropping the packet.
610  * But, we can still check if is spoofed against the IP
611  * addresses associated with the interface.
612  */
613 static bool only_dnatted(const struct ip_tunnel *tunnel,
614 	const struct in6_addr *v6dst)
615 {
616 	int prefix_len;
617 
618 #ifdef CONFIG_IPV6_SIT_6RD
619 	prefix_len = tunnel->ip6rd.prefixlen + 32
620 		- tunnel->ip6rd.relay_prefixlen;
621 #else
622 	prefix_len = 48;
623 #endif
624 	return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
625 }
626 
627 /* Returns true if a packet is spoofed */
628 static bool packet_is_spoofed(struct sk_buff *skb,
629 			      const struct iphdr *iph,
630 			      struct ip_tunnel *tunnel)
631 {
632 	const struct ipv6hdr *ipv6h;
633 
634 	if (tunnel->dev->priv_flags & IFF_ISATAP) {
635 		if (!isatap_chksrc(skb, iph, tunnel))
636 			return true;
637 
638 		return false;
639 	}
640 
641 	if (tunnel->dev->flags & IFF_POINTOPOINT)
642 		return false;
643 
644 	ipv6h = ipv6_hdr(skb);
645 
646 	if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
647 		net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
648 				     &iph->saddr, &ipv6h->saddr,
649 				     &iph->daddr, &ipv6h->daddr);
650 		return true;
651 	}
652 
653 	if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
654 		return false;
655 
656 	if (only_dnatted(tunnel, &ipv6h->daddr))
657 		return false;
658 
659 	net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
660 			     &iph->saddr, &ipv6h->saddr,
661 			     &iph->daddr, &ipv6h->daddr);
662 	return true;
663 }
664 
665 static int ipip6_rcv(struct sk_buff *skb)
666 {
667 	const struct iphdr *iph = ip_hdr(skb);
668 	struct ip_tunnel *tunnel;
669 	int err;
670 
671 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
672 				     iph->saddr, iph->daddr);
673 	if (tunnel != NULL) {
674 		struct pcpu_tstats *tstats;
675 
676 		if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
677 		    tunnel->parms.iph.protocol != 0)
678 			goto out;
679 
680 		skb->mac_header = skb->network_header;
681 		skb_reset_network_header(skb);
682 		IPCB(skb)->flags = 0;
683 		skb->protocol = htons(ETH_P_IPV6);
684 
685 		if (packet_is_spoofed(skb, iph, tunnel)) {
686 			tunnel->dev->stats.rx_errors++;
687 			goto out;
688 		}
689 
690 		__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
691 
692 		err = IP_ECN_decapsulate(iph, skb);
693 		if (unlikely(err)) {
694 			if (log_ecn_error)
695 				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
696 						     &iph->saddr, iph->tos);
697 			if (err > 1) {
698 				++tunnel->dev->stats.rx_frame_errors;
699 				++tunnel->dev->stats.rx_errors;
700 				goto out;
701 			}
702 		}
703 
704 		tstats = this_cpu_ptr(tunnel->dev->tstats);
705 		tstats->rx_packets++;
706 		tstats->rx_bytes += skb->len;
707 
708 		netif_rx(skb);
709 
710 		return 0;
711 	}
712 
713 	/* no tunnel matched,  let upstream know, ipsec may handle it */
714 	return 1;
715 out:
716 	kfree_skb(skb);
717 	return 0;
718 }
719 
720 static const struct tnl_ptk_info tpi = {
721 	/* no tunnel info required for ipip. */
722 	.proto = htons(ETH_P_IP),
723 };
724 
725 static int ipip_rcv(struct sk_buff *skb)
726 {
727 	const struct iphdr *iph;
728 	struct ip_tunnel *tunnel;
729 
730 	iph = ip_hdr(skb);
731 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
732 				     iph->saddr, iph->daddr);
733 	if (tunnel != NULL) {
734 		if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
735 		    tunnel->parms.iph.protocol != 0)
736 			goto drop;
737 
738 		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
739 			goto drop;
740 		if (iptunnel_pull_header(skb, 0, tpi.proto))
741 			goto drop;
742 		return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
743 	}
744 
745 	return 1;
746 
747 drop:
748 	kfree_skb(skb);
749 	return 0;
750 }
751 
752 /*
753  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
754  * stores the embedded IPv4 address in v4dst and returns true.
755  */
756 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
757 		      __be32 *v4dst)
758 {
759 #ifdef CONFIG_IPV6_SIT_6RD
760 	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
761 			      tunnel->ip6rd.prefixlen)) {
762 		unsigned int pbw0, pbi0;
763 		int pbi1;
764 		u32 d;
765 
766 		pbw0 = tunnel->ip6rd.prefixlen >> 5;
767 		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
768 
769 		d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
770 		    tunnel->ip6rd.relay_prefixlen;
771 
772 		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
773 		if (pbi1 > 0)
774 			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
775 			     (32 - pbi1);
776 
777 		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
778 		return true;
779 	}
780 #else
781 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
782 		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
783 		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
784 		return true;
785 	}
786 #endif
787 	return false;
788 }
789 
790 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
791 			     const struct in6_addr *v6dst)
792 {
793 	__be32 dst = 0;
794 	check_6rd(tunnel, v6dst, &dst);
795 	return dst;
796 }
797 
798 /*
799  *	This function assumes it is being called from dev_queue_xmit()
800  *	and that skb is filled properly by that function.
801  */
802 
803 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
804 				     struct net_device *dev)
805 {
806 	struct ip_tunnel *tunnel = netdev_priv(dev);
807 	const struct iphdr  *tiph = &tunnel->parms.iph;
808 	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
809 	u8     tos = tunnel->parms.iph.tos;
810 	__be16 df = tiph->frag_off;
811 	struct rtable *rt;     			/* Route to the other host */
812 	struct net_device *tdev;		/* Device to other host */
813 	unsigned int max_headroom;		/* The extra header space needed */
814 	__be32 dst = tiph->daddr;
815 	struct flowi4 fl4;
816 	int    mtu;
817 	const struct in6_addr *addr6;
818 	int addr_type;
819 	u8 ttl;
820 	int err;
821 
822 	if (skb->protocol != htons(ETH_P_IPV6))
823 		goto tx_error;
824 
825 	if (tos == 1)
826 		tos = ipv6_get_dsfield(iph6);
827 
828 	/* ISATAP (RFC4214) - must come before 6to4 */
829 	if (dev->priv_flags & IFF_ISATAP) {
830 		struct neighbour *neigh = NULL;
831 		bool do_tx_error = false;
832 
833 		if (skb_dst(skb))
834 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
835 
836 		if (neigh == NULL) {
837 			net_dbg_ratelimited("nexthop == NULL\n");
838 			goto tx_error;
839 		}
840 
841 		addr6 = (const struct in6_addr *)&neigh->primary_key;
842 		addr_type = ipv6_addr_type(addr6);
843 
844 		if ((addr_type & IPV6_ADDR_UNICAST) &&
845 		     ipv6_addr_is_isatap(addr6))
846 			dst = addr6->s6_addr32[3];
847 		else
848 			do_tx_error = true;
849 
850 		neigh_release(neigh);
851 		if (do_tx_error)
852 			goto tx_error;
853 	}
854 
855 	if (!dst)
856 		dst = try_6rd(tunnel, &iph6->daddr);
857 
858 	if (!dst) {
859 		struct neighbour *neigh = NULL;
860 		bool do_tx_error = false;
861 
862 		if (skb_dst(skb))
863 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
864 
865 		if (neigh == NULL) {
866 			net_dbg_ratelimited("nexthop == NULL\n");
867 			goto tx_error;
868 		}
869 
870 		addr6 = (const struct in6_addr *)&neigh->primary_key;
871 		addr_type = ipv6_addr_type(addr6);
872 
873 		if (addr_type == IPV6_ADDR_ANY) {
874 			addr6 = &ipv6_hdr(skb)->daddr;
875 			addr_type = ipv6_addr_type(addr6);
876 		}
877 
878 		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
879 			dst = addr6->s6_addr32[3];
880 		else
881 			do_tx_error = true;
882 
883 		neigh_release(neigh);
884 		if (do_tx_error)
885 			goto tx_error;
886 	}
887 
888 	rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
889 				   dst, tiph->saddr,
890 				   0, 0,
891 				   IPPROTO_IPV6, RT_TOS(tos),
892 				   tunnel->parms.link);
893 	if (IS_ERR(rt)) {
894 		dev->stats.tx_carrier_errors++;
895 		goto tx_error_icmp;
896 	}
897 	if (rt->rt_type != RTN_UNICAST) {
898 		ip_rt_put(rt);
899 		dev->stats.tx_carrier_errors++;
900 		goto tx_error_icmp;
901 	}
902 	tdev = rt->dst.dev;
903 
904 	if (tdev == dev) {
905 		ip_rt_put(rt);
906 		dev->stats.collisions++;
907 		goto tx_error;
908 	}
909 
910 	if (df) {
911 		mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
912 
913 		if (mtu < 68) {
914 			dev->stats.collisions++;
915 			ip_rt_put(rt);
916 			goto tx_error;
917 		}
918 
919 		if (mtu < IPV6_MIN_MTU) {
920 			mtu = IPV6_MIN_MTU;
921 			df = 0;
922 		}
923 
924 		if (tunnel->parms.iph.daddr && skb_dst(skb))
925 			skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
926 
927 		if (skb->len > mtu) {
928 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
929 			ip_rt_put(rt);
930 			goto tx_error;
931 		}
932 	}
933 
934 	if (tunnel->err_count > 0) {
935 		if (time_before(jiffies,
936 				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
937 			tunnel->err_count--;
938 			dst_link_failure(skb);
939 		} else
940 			tunnel->err_count = 0;
941 	}
942 
943 	/*
944 	 * Okay, now see if we can stuff it in the buffer as-is.
945 	 */
946 	max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
947 
948 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
949 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
950 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
951 		if (!new_skb) {
952 			ip_rt_put(rt);
953 			dev->stats.tx_dropped++;
954 			kfree_skb(skb);
955 			return NETDEV_TX_OK;
956 		}
957 		if (skb->sk)
958 			skb_set_owner_w(new_skb, skb->sk);
959 		dev_kfree_skb(skb);
960 		skb = new_skb;
961 		iph6 = ipv6_hdr(skb);
962 	}
963 	ttl = tiph->ttl;
964 	if (ttl == 0)
965 		ttl = iph6->hop_limit;
966 	tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
967 
968 	skb = iptunnel_handle_offloads(skb, false, SKB_GSO_SIT);
969 	if (IS_ERR(skb))
970 		goto out;
971 
972 	err = iptunnel_xmit(rt, skb, fl4.saddr, fl4.daddr, IPPROTO_IPV6, tos,
973 			    ttl, df, !net_eq(tunnel->net, dev_net(dev)));
974 	iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
975 	return NETDEV_TX_OK;
976 
977 tx_error_icmp:
978 	dst_link_failure(skb);
979 tx_error:
980 	kfree_skb(skb);
981 out:
982 	dev->stats.tx_errors++;
983 	return NETDEV_TX_OK;
984 }
985 
986 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
987 {
988 	struct ip_tunnel *tunnel = netdev_priv(dev);
989 	const struct iphdr  *tiph = &tunnel->parms.iph;
990 
991 	skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
992 	if (IS_ERR(skb))
993 		goto out;
994 
995 	ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
996 	return NETDEV_TX_OK;
997 out:
998 	dev->stats.tx_errors++;
999 	return NETDEV_TX_OK;
1000 }
1001 
1002 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1003 				   struct net_device *dev)
1004 {
1005 	switch (skb->protocol) {
1006 	case htons(ETH_P_IP):
1007 		ipip_tunnel_xmit(skb, dev);
1008 		break;
1009 	case htons(ETH_P_IPV6):
1010 		ipip6_tunnel_xmit(skb, dev);
1011 		break;
1012 	default:
1013 		goto tx_err;
1014 	}
1015 
1016 	return NETDEV_TX_OK;
1017 
1018 tx_err:
1019 	dev->stats.tx_errors++;
1020 	kfree_skb(skb);
1021 	return NETDEV_TX_OK;
1022 
1023 }
1024 
1025 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1026 {
1027 	struct net_device *tdev = NULL;
1028 	struct ip_tunnel *tunnel;
1029 	const struct iphdr *iph;
1030 	struct flowi4 fl4;
1031 
1032 	tunnel = netdev_priv(dev);
1033 	iph = &tunnel->parms.iph;
1034 
1035 	if (iph->daddr) {
1036 		struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1037 							  NULL,
1038 							  iph->daddr, iph->saddr,
1039 							  0, 0,
1040 							  IPPROTO_IPV6,
1041 							  RT_TOS(iph->tos),
1042 							  tunnel->parms.link);
1043 
1044 		if (!IS_ERR(rt)) {
1045 			tdev = rt->dst.dev;
1046 			ip_rt_put(rt);
1047 		}
1048 		dev->flags |= IFF_POINTOPOINT;
1049 	}
1050 
1051 	if (!tdev && tunnel->parms.link)
1052 		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1053 
1054 	if (tdev) {
1055 		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1056 		dev->mtu = tdev->mtu - sizeof(struct iphdr);
1057 		if (dev->mtu < IPV6_MIN_MTU)
1058 			dev->mtu = IPV6_MIN_MTU;
1059 	}
1060 	dev->iflink = tunnel->parms.link;
1061 }
1062 
1063 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
1064 {
1065 	struct net *net = t->net;
1066 	struct sit_net *sitn = net_generic(net, sit_net_id);
1067 
1068 	ipip6_tunnel_unlink(sitn, t);
1069 	synchronize_net();
1070 	t->parms.iph.saddr = p->iph.saddr;
1071 	t->parms.iph.daddr = p->iph.daddr;
1072 	memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1073 	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1074 	ipip6_tunnel_link(sitn, t);
1075 	t->parms.iph.ttl = p->iph.ttl;
1076 	t->parms.iph.tos = p->iph.tos;
1077 	if (t->parms.link != p->link) {
1078 		t->parms.link = p->link;
1079 		ipip6_tunnel_bind_dev(t->dev);
1080 	}
1081 	netdev_state_change(t->dev);
1082 }
1083 
1084 #ifdef CONFIG_IPV6_SIT_6RD
1085 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1086 				   struct ip_tunnel_6rd *ip6rd)
1087 {
1088 	struct in6_addr prefix;
1089 	__be32 relay_prefix;
1090 
1091 	if (ip6rd->relay_prefixlen > 32 ||
1092 	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1093 		return -EINVAL;
1094 
1095 	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1096 	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1097 		return -EINVAL;
1098 	if (ip6rd->relay_prefixlen)
1099 		relay_prefix = ip6rd->relay_prefix &
1100 			       htonl(0xffffffffUL <<
1101 				     (32 - ip6rd->relay_prefixlen));
1102 	else
1103 		relay_prefix = 0;
1104 	if (relay_prefix != ip6rd->relay_prefix)
1105 		return -EINVAL;
1106 
1107 	t->ip6rd.prefix = prefix;
1108 	t->ip6rd.relay_prefix = relay_prefix;
1109 	t->ip6rd.prefixlen = ip6rd->prefixlen;
1110 	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1111 	netdev_state_change(t->dev);
1112 	return 0;
1113 }
1114 #endif
1115 
1116 static int
1117 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1118 {
1119 	int err = 0;
1120 	struct ip_tunnel_parm p;
1121 	struct ip_tunnel_prl prl;
1122 	struct ip_tunnel *t;
1123 	struct net *net = dev_net(dev);
1124 	struct sit_net *sitn = net_generic(net, sit_net_id);
1125 #ifdef CONFIG_IPV6_SIT_6RD
1126 	struct ip_tunnel_6rd ip6rd;
1127 #endif
1128 
1129 	switch (cmd) {
1130 	case SIOCGETTUNNEL:
1131 #ifdef CONFIG_IPV6_SIT_6RD
1132 	case SIOCGET6RD:
1133 #endif
1134 		t = NULL;
1135 		if (dev == sitn->fb_tunnel_dev) {
1136 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1137 				err = -EFAULT;
1138 				break;
1139 			}
1140 			t = ipip6_tunnel_locate(net, &p, 0);
1141 		}
1142 		if (t == NULL)
1143 			t = netdev_priv(dev);
1144 
1145 		err = -EFAULT;
1146 		if (cmd == SIOCGETTUNNEL) {
1147 			memcpy(&p, &t->parms, sizeof(p));
1148 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1149 					 sizeof(p)))
1150 				goto done;
1151 #ifdef CONFIG_IPV6_SIT_6RD
1152 		} else {
1153 			ip6rd.prefix = t->ip6rd.prefix;
1154 			ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1155 			ip6rd.prefixlen = t->ip6rd.prefixlen;
1156 			ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1157 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1158 					 sizeof(ip6rd)))
1159 				goto done;
1160 #endif
1161 		}
1162 		err = 0;
1163 		break;
1164 
1165 	case SIOCADDTUNNEL:
1166 	case SIOCCHGTUNNEL:
1167 		err = -EPERM;
1168 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1169 			goto done;
1170 
1171 		err = -EFAULT;
1172 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1173 			goto done;
1174 
1175 		err = -EINVAL;
1176 		if (p.iph.protocol != IPPROTO_IPV6 &&
1177 		    p.iph.protocol != IPPROTO_IPIP &&
1178 		    p.iph.protocol != 0)
1179 			goto done;
1180 		if (p.iph.version != 4 ||
1181 		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1182 			goto done;
1183 		if (p.iph.ttl)
1184 			p.iph.frag_off |= htons(IP_DF);
1185 
1186 		t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1187 
1188 		if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1189 			if (t != NULL) {
1190 				if (t->dev != dev) {
1191 					err = -EEXIST;
1192 					break;
1193 				}
1194 			} else {
1195 				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1196 				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1197 					err = -EINVAL;
1198 					break;
1199 				}
1200 				t = netdev_priv(dev);
1201 			}
1202 
1203 			ipip6_tunnel_update(t, &p);
1204 		}
1205 
1206 		if (t) {
1207 			err = 0;
1208 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1209 				err = -EFAULT;
1210 		} else
1211 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1212 		break;
1213 
1214 	case SIOCDELTUNNEL:
1215 		err = -EPERM;
1216 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1217 			goto done;
1218 
1219 		if (dev == sitn->fb_tunnel_dev) {
1220 			err = -EFAULT;
1221 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1222 				goto done;
1223 			err = -ENOENT;
1224 			if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1225 				goto done;
1226 			err = -EPERM;
1227 			if (t == netdev_priv(sitn->fb_tunnel_dev))
1228 				goto done;
1229 			dev = t->dev;
1230 		}
1231 		unregister_netdevice(dev);
1232 		err = 0;
1233 		break;
1234 
1235 	case SIOCGETPRL:
1236 		err = -EINVAL;
1237 		if (dev == sitn->fb_tunnel_dev)
1238 			goto done;
1239 		err = -ENOENT;
1240 		if (!(t = netdev_priv(dev)))
1241 			goto done;
1242 		err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1243 		break;
1244 
1245 	case SIOCADDPRL:
1246 	case SIOCDELPRL:
1247 	case SIOCCHGPRL:
1248 		err = -EPERM;
1249 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1250 			goto done;
1251 		err = -EINVAL;
1252 		if (dev == sitn->fb_tunnel_dev)
1253 			goto done;
1254 		err = -EFAULT;
1255 		if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1256 			goto done;
1257 		err = -ENOENT;
1258 		if (!(t = netdev_priv(dev)))
1259 			goto done;
1260 
1261 		switch (cmd) {
1262 		case SIOCDELPRL:
1263 			err = ipip6_tunnel_del_prl(t, &prl);
1264 			break;
1265 		case SIOCADDPRL:
1266 		case SIOCCHGPRL:
1267 			err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1268 			break;
1269 		}
1270 		netdev_state_change(dev);
1271 		break;
1272 
1273 #ifdef CONFIG_IPV6_SIT_6RD
1274 	case SIOCADD6RD:
1275 	case SIOCCHG6RD:
1276 	case SIOCDEL6RD:
1277 		err = -EPERM;
1278 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1279 			goto done;
1280 
1281 		err = -EFAULT;
1282 		if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1283 				   sizeof(ip6rd)))
1284 			goto done;
1285 
1286 		t = netdev_priv(dev);
1287 
1288 		if (cmd != SIOCDEL6RD) {
1289 			err = ipip6_tunnel_update_6rd(t, &ip6rd);
1290 			if (err < 0)
1291 				goto done;
1292 		} else
1293 			ipip6_tunnel_clone_6rd(dev, sitn);
1294 
1295 		err = 0;
1296 		break;
1297 #endif
1298 
1299 	default:
1300 		err = -EINVAL;
1301 	}
1302 
1303 done:
1304 	return err;
1305 }
1306 
1307 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1308 {
1309 	if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1310 		return -EINVAL;
1311 	dev->mtu = new_mtu;
1312 	return 0;
1313 }
1314 
1315 static const struct net_device_ops ipip6_netdev_ops = {
1316 	.ndo_uninit	= ipip6_tunnel_uninit,
1317 	.ndo_start_xmit	= sit_tunnel_xmit,
1318 	.ndo_do_ioctl	= ipip6_tunnel_ioctl,
1319 	.ndo_change_mtu	= ipip6_tunnel_change_mtu,
1320 	.ndo_get_stats64 = ip_tunnel_get_stats64,
1321 };
1322 
1323 static void ipip6_dev_free(struct net_device *dev)
1324 {
1325 	free_percpu(dev->tstats);
1326 	free_netdev(dev);
1327 }
1328 
1329 #define SIT_FEATURES (NETIF_F_SG	   | \
1330 		      NETIF_F_FRAGLIST	   | \
1331 		      NETIF_F_HIGHDMA	   | \
1332 		      NETIF_F_GSO_SOFTWARE | \
1333 		      NETIF_F_HW_CSUM)
1334 
1335 static void ipip6_tunnel_setup(struct net_device *dev)
1336 {
1337 	dev->netdev_ops		= &ipip6_netdev_ops;
1338 	dev->destructor 	= ipip6_dev_free;
1339 
1340 	dev->type		= ARPHRD_SIT;
1341 	dev->hard_header_len 	= LL_MAX_HEADER + sizeof(struct iphdr);
1342 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr);
1343 	dev->flags		= IFF_NOARP;
1344 	dev->priv_flags	       &= ~IFF_XMIT_DST_RELEASE;
1345 	dev->iflink		= 0;
1346 	dev->addr_len		= 4;
1347 	dev->features		|= NETIF_F_LLTX;
1348 	dev->features		|= SIT_FEATURES;
1349 	dev->hw_features	|= SIT_FEATURES;
1350 }
1351 
1352 static int ipip6_tunnel_init(struct net_device *dev)
1353 {
1354 	struct ip_tunnel *tunnel = netdev_priv(dev);
1355 	int i;
1356 
1357 	tunnel->dev = dev;
1358 	tunnel->net = dev_net(dev);
1359 
1360 	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1361 	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1362 
1363 	ipip6_tunnel_bind_dev(dev);
1364 	dev->tstats = alloc_percpu(struct pcpu_tstats);
1365 	if (!dev->tstats)
1366 		return -ENOMEM;
1367 
1368 	for_each_possible_cpu(i) {
1369 		struct pcpu_tstats *ipip6_tunnel_stats;
1370 		ipip6_tunnel_stats = per_cpu_ptr(dev->tstats, i);
1371 		u64_stats_init(&ipip6_tunnel_stats->syncp);
1372 	}
1373 
1374 	return 0;
1375 }
1376 
1377 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1378 {
1379 	struct ip_tunnel *tunnel = netdev_priv(dev);
1380 	struct iphdr *iph = &tunnel->parms.iph;
1381 	struct net *net = dev_net(dev);
1382 	struct sit_net *sitn = net_generic(net, sit_net_id);
1383 	int i;
1384 
1385 	tunnel->dev = dev;
1386 	tunnel->net = dev_net(dev);
1387 	strcpy(tunnel->parms.name, dev->name);
1388 
1389 	iph->version		= 4;
1390 	iph->protocol		= IPPROTO_IPV6;
1391 	iph->ihl		= 5;
1392 	iph->ttl		= 64;
1393 
1394 	dev->tstats = alloc_percpu(struct pcpu_tstats);
1395 	if (!dev->tstats)
1396 		return -ENOMEM;
1397 
1398 	for_each_possible_cpu(i) {
1399 		struct pcpu_tstats *ipip6_fb_stats;
1400 		ipip6_fb_stats = per_cpu_ptr(dev->tstats, i);
1401 		u64_stats_init(&ipip6_fb_stats->syncp);
1402 	}
1403 
1404 	dev_hold(dev);
1405 	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1406 	return 0;
1407 }
1408 
1409 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1410 {
1411 	u8 proto;
1412 
1413 	if (!data || !data[IFLA_IPTUN_PROTO])
1414 		return 0;
1415 
1416 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1417 	if (proto != IPPROTO_IPV6 &&
1418 	    proto != IPPROTO_IPIP &&
1419 	    proto != 0)
1420 		return -EINVAL;
1421 
1422 	return 0;
1423 }
1424 
1425 static void ipip6_netlink_parms(struct nlattr *data[],
1426 				struct ip_tunnel_parm *parms)
1427 {
1428 	memset(parms, 0, sizeof(*parms));
1429 
1430 	parms->iph.version = 4;
1431 	parms->iph.protocol = IPPROTO_IPV6;
1432 	parms->iph.ihl = 5;
1433 	parms->iph.ttl = 64;
1434 
1435 	if (!data)
1436 		return;
1437 
1438 	if (data[IFLA_IPTUN_LINK])
1439 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1440 
1441 	if (data[IFLA_IPTUN_LOCAL])
1442 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1443 
1444 	if (data[IFLA_IPTUN_REMOTE])
1445 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1446 
1447 	if (data[IFLA_IPTUN_TTL]) {
1448 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1449 		if (parms->iph.ttl)
1450 			parms->iph.frag_off = htons(IP_DF);
1451 	}
1452 
1453 	if (data[IFLA_IPTUN_TOS])
1454 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1455 
1456 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1457 		parms->iph.frag_off = htons(IP_DF);
1458 
1459 	if (data[IFLA_IPTUN_FLAGS])
1460 		parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1461 
1462 	if (data[IFLA_IPTUN_PROTO])
1463 		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1464 
1465 }
1466 
1467 #ifdef CONFIG_IPV6_SIT_6RD
1468 /* This function returns true when 6RD attributes are present in the nl msg */
1469 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1470 				    struct ip_tunnel_6rd *ip6rd)
1471 {
1472 	bool ret = false;
1473 	memset(ip6rd, 0, sizeof(*ip6rd));
1474 
1475 	if (!data)
1476 		return ret;
1477 
1478 	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1479 		ret = true;
1480 		nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1481 			   sizeof(struct in6_addr));
1482 	}
1483 
1484 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1485 		ret = true;
1486 		ip6rd->relay_prefix =
1487 			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1488 	}
1489 
1490 	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1491 		ret = true;
1492 		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1493 	}
1494 
1495 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1496 		ret = true;
1497 		ip6rd->relay_prefixlen =
1498 			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1499 	}
1500 
1501 	return ret;
1502 }
1503 #endif
1504 
1505 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1506 			 struct nlattr *tb[], struct nlattr *data[])
1507 {
1508 	struct net *net = dev_net(dev);
1509 	struct ip_tunnel *nt;
1510 #ifdef CONFIG_IPV6_SIT_6RD
1511 	struct ip_tunnel_6rd ip6rd;
1512 #endif
1513 	int err;
1514 
1515 	nt = netdev_priv(dev);
1516 	ipip6_netlink_parms(data, &nt->parms);
1517 
1518 	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1519 		return -EEXIST;
1520 
1521 	err = ipip6_tunnel_create(dev);
1522 	if (err < 0)
1523 		return err;
1524 
1525 #ifdef CONFIG_IPV6_SIT_6RD
1526 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1527 		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1528 #endif
1529 
1530 	return err;
1531 }
1532 
1533 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1534 			  struct nlattr *data[])
1535 {
1536 	struct ip_tunnel *t = netdev_priv(dev);
1537 	struct ip_tunnel_parm p;
1538 	struct net *net = t->net;
1539 	struct sit_net *sitn = net_generic(net, sit_net_id);
1540 #ifdef CONFIG_IPV6_SIT_6RD
1541 	struct ip_tunnel_6rd ip6rd;
1542 #endif
1543 
1544 	if (dev == sitn->fb_tunnel_dev)
1545 		return -EINVAL;
1546 
1547 	ipip6_netlink_parms(data, &p);
1548 
1549 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1550 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1551 		return -EINVAL;
1552 
1553 	t = ipip6_tunnel_locate(net, &p, 0);
1554 
1555 	if (t) {
1556 		if (t->dev != dev)
1557 			return -EEXIST;
1558 	} else
1559 		t = netdev_priv(dev);
1560 
1561 	ipip6_tunnel_update(t, &p);
1562 
1563 #ifdef CONFIG_IPV6_SIT_6RD
1564 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1565 		return ipip6_tunnel_update_6rd(t, &ip6rd);
1566 #endif
1567 
1568 	return 0;
1569 }
1570 
1571 static size_t ipip6_get_size(const struct net_device *dev)
1572 {
1573 	return
1574 		/* IFLA_IPTUN_LINK */
1575 		nla_total_size(4) +
1576 		/* IFLA_IPTUN_LOCAL */
1577 		nla_total_size(4) +
1578 		/* IFLA_IPTUN_REMOTE */
1579 		nla_total_size(4) +
1580 		/* IFLA_IPTUN_TTL */
1581 		nla_total_size(1) +
1582 		/* IFLA_IPTUN_TOS */
1583 		nla_total_size(1) +
1584 		/* IFLA_IPTUN_PMTUDISC */
1585 		nla_total_size(1) +
1586 		/* IFLA_IPTUN_FLAGS */
1587 		nla_total_size(2) +
1588 		/* IFLA_IPTUN_PROTO */
1589 		nla_total_size(1) +
1590 #ifdef CONFIG_IPV6_SIT_6RD
1591 		/* IFLA_IPTUN_6RD_PREFIX */
1592 		nla_total_size(sizeof(struct in6_addr)) +
1593 		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1594 		nla_total_size(4) +
1595 		/* IFLA_IPTUN_6RD_PREFIXLEN */
1596 		nla_total_size(2) +
1597 		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1598 		nla_total_size(2) +
1599 #endif
1600 		0;
1601 }
1602 
1603 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1604 {
1605 	struct ip_tunnel *tunnel = netdev_priv(dev);
1606 	struct ip_tunnel_parm *parm = &tunnel->parms;
1607 
1608 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1609 	    nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1610 	    nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1611 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1612 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1613 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1614 		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1615 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1616 	    nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1617 		goto nla_put_failure;
1618 
1619 #ifdef CONFIG_IPV6_SIT_6RD
1620 	if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1621 		    &tunnel->ip6rd.prefix) ||
1622 	    nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1623 			 tunnel->ip6rd.relay_prefix) ||
1624 	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1625 			tunnel->ip6rd.prefixlen) ||
1626 	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1627 			tunnel->ip6rd.relay_prefixlen))
1628 		goto nla_put_failure;
1629 #endif
1630 
1631 	return 0;
1632 
1633 nla_put_failure:
1634 	return -EMSGSIZE;
1635 }
1636 
1637 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1638 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1639 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1640 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1641 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1642 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1643 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1644 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1645 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1646 #ifdef CONFIG_IPV6_SIT_6RD
1647 	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1648 	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1649 	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1650 	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1651 #endif
1652 };
1653 
1654 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1655 {
1656 	struct net *net = dev_net(dev);
1657 	struct sit_net *sitn = net_generic(net, sit_net_id);
1658 
1659 	if (dev != sitn->fb_tunnel_dev)
1660 		unregister_netdevice_queue(dev, head);
1661 }
1662 
1663 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1664 	.kind		= "sit",
1665 	.maxtype	= IFLA_IPTUN_MAX,
1666 	.policy		= ipip6_policy,
1667 	.priv_size	= sizeof(struct ip_tunnel),
1668 	.setup		= ipip6_tunnel_setup,
1669 	.validate	= ipip6_validate,
1670 	.newlink	= ipip6_newlink,
1671 	.changelink	= ipip6_changelink,
1672 	.get_size	= ipip6_get_size,
1673 	.fill_info	= ipip6_fill_info,
1674 	.dellink	= ipip6_dellink,
1675 };
1676 
1677 static struct xfrm_tunnel sit_handler __read_mostly = {
1678 	.handler	=	ipip6_rcv,
1679 	.err_handler	=	ipip6_err,
1680 	.priority	=	1,
1681 };
1682 
1683 static struct xfrm_tunnel ipip_handler __read_mostly = {
1684 	.handler	=	ipip_rcv,
1685 	.err_handler	=	ipip6_err,
1686 	.priority	=	2,
1687 };
1688 
1689 static void __net_exit sit_destroy_tunnels(struct net *net,
1690 					   struct list_head *head)
1691 {
1692 	struct sit_net *sitn = net_generic(net, sit_net_id);
1693 	struct net_device *dev, *aux;
1694 	int prio;
1695 
1696 	for_each_netdev_safe(net, dev, aux)
1697 		if (dev->rtnl_link_ops == &sit_link_ops)
1698 			unregister_netdevice_queue(dev, head);
1699 
1700 	for (prio = 1; prio < 4; prio++) {
1701 		int h;
1702 		for (h = 0; h < HASH_SIZE; h++) {
1703 			struct ip_tunnel *t;
1704 
1705 			t = rtnl_dereference(sitn->tunnels[prio][h]);
1706 			while (t != NULL) {
1707 				/* If dev is in the same netns, it has already
1708 				 * been added to the list by the previous loop.
1709 				 */
1710 				if (!net_eq(dev_net(t->dev), net))
1711 					unregister_netdevice_queue(t->dev,
1712 								   head);
1713 				t = rtnl_dereference(t->next);
1714 			}
1715 		}
1716 	}
1717 }
1718 
1719 static int __net_init sit_init_net(struct net *net)
1720 {
1721 	struct sit_net *sitn = net_generic(net, sit_net_id);
1722 	struct ip_tunnel *t;
1723 	int err;
1724 
1725 	sitn->tunnels[0] = sitn->tunnels_wc;
1726 	sitn->tunnels[1] = sitn->tunnels_l;
1727 	sitn->tunnels[2] = sitn->tunnels_r;
1728 	sitn->tunnels[3] = sitn->tunnels_r_l;
1729 
1730 	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1731 					   ipip6_tunnel_setup);
1732 	if (!sitn->fb_tunnel_dev) {
1733 		err = -ENOMEM;
1734 		goto err_alloc_dev;
1735 	}
1736 	dev_net_set(sitn->fb_tunnel_dev, net);
1737 	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1738 	/* FB netdevice is special: we have one, and only one per netns.
1739 	 * Allowing to move it to another netns is clearly unsafe.
1740 	 */
1741 	sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1742 
1743 	err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1744 	if (err)
1745 		goto err_dev_free;
1746 
1747 	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1748 
1749 	if ((err = register_netdev(sitn->fb_tunnel_dev)))
1750 		goto err_reg_dev;
1751 
1752 	t = netdev_priv(sitn->fb_tunnel_dev);
1753 
1754 	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1755 	return 0;
1756 
1757 err_reg_dev:
1758 	dev_put(sitn->fb_tunnel_dev);
1759 err_dev_free:
1760 	ipip6_dev_free(sitn->fb_tunnel_dev);
1761 err_alloc_dev:
1762 	return err;
1763 }
1764 
1765 static void __net_exit sit_exit_net(struct net *net)
1766 {
1767 	LIST_HEAD(list);
1768 
1769 	rtnl_lock();
1770 	sit_destroy_tunnels(net, &list);
1771 	unregister_netdevice_many(&list);
1772 	rtnl_unlock();
1773 }
1774 
1775 static struct pernet_operations sit_net_ops = {
1776 	.init = sit_init_net,
1777 	.exit = sit_exit_net,
1778 	.id   = &sit_net_id,
1779 	.size = sizeof(struct sit_net),
1780 };
1781 
1782 static void __exit sit_cleanup(void)
1783 {
1784 	rtnl_link_unregister(&sit_link_ops);
1785 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1786 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1787 
1788 	unregister_pernet_device(&sit_net_ops);
1789 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1790 }
1791 
1792 static int __init sit_init(void)
1793 {
1794 	int err;
1795 
1796 	pr_info("IPv6 over IPv4 tunneling driver\n");
1797 
1798 	err = register_pernet_device(&sit_net_ops);
1799 	if (err < 0)
1800 		return err;
1801 	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1802 	if (err < 0) {
1803 		pr_info("%s: can't register ip6ip4\n", __func__);
1804 		goto xfrm_tunnel_failed;
1805 	}
1806 	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1807 	if (err < 0) {
1808 		pr_info("%s: can't register ip4ip4\n", __func__);
1809 		goto xfrm_tunnel4_failed;
1810 	}
1811 	err = rtnl_link_register(&sit_link_ops);
1812 	if (err < 0)
1813 		goto rtnl_link_failed;
1814 
1815 out:
1816 	return err;
1817 
1818 rtnl_link_failed:
1819 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1820 xfrm_tunnel4_failed:
1821 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1822 xfrm_tunnel_failed:
1823 	unregister_pernet_device(&sit_net_ops);
1824 	goto out;
1825 }
1826 
1827 module_init(sit_init);
1828 module_exit(sit_cleanup);
1829 MODULE_LICENSE("GPL");
1830 MODULE_ALIAS_NETDEV("sit0");
1831