xref: /linux/net/ipv6/sit.c (revision f884ab15afdc5514e88105c92a4e2e1e6539869a)
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 net *net = dev_net(dev);
470 	struct sit_net *sitn = net_generic(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, netdev_priv(dev));
476 		ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
477 	}
478 	dev_put(dev);
479 }
480 
481 
482 static int ipip6_err(struct sk_buff *skb, u32 info)
483 {
484 
485 /* All the routers (except for Linux) return only
486    8 bytes of packet payload. It means, that precise relaying of
487    ICMP in the real Internet is absolutely infeasible.
488  */
489 	const struct iphdr *iph = (const struct iphdr *)skb->data;
490 	const int type = icmp_hdr(skb)->type;
491 	const int code = icmp_hdr(skb)->code;
492 	struct ip_tunnel *t;
493 	int err;
494 
495 	switch (type) {
496 	default:
497 	case ICMP_PARAMETERPROB:
498 		return 0;
499 
500 	case ICMP_DEST_UNREACH:
501 		switch (code) {
502 		case ICMP_SR_FAILED:
503 		case ICMP_PORT_UNREACH:
504 			/* Impossible event. */
505 			return 0;
506 		default:
507 			/* All others are translated to HOST_UNREACH.
508 			   rfc2003 contains "deep thoughts" about NET_UNREACH,
509 			   I believe they are just ether pollution. --ANK
510 			 */
511 			break;
512 		}
513 		break;
514 	case ICMP_TIME_EXCEEDED:
515 		if (code != ICMP_EXC_TTL)
516 			return 0;
517 		break;
518 	case ICMP_REDIRECT:
519 		break;
520 	}
521 
522 	err = -ENOENT;
523 
524 	t = ipip6_tunnel_lookup(dev_net(skb->dev),
525 				skb->dev,
526 				iph->daddr,
527 				iph->saddr);
528 	if (t == NULL)
529 		goto out;
530 
531 	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
532 		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
533 				 t->dev->ifindex, 0, IPPROTO_IPV6, 0);
534 		err = 0;
535 		goto out;
536 	}
537 	if (type == ICMP_REDIRECT) {
538 		ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
539 			      IPPROTO_IPV6, 0);
540 		err = 0;
541 		goto out;
542 	}
543 
544 	if (t->parms.iph.daddr == 0)
545 		goto out;
546 
547 	err = 0;
548 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
549 		goto out;
550 
551 	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
552 		t->err_count++;
553 	else
554 		t->err_count = 1;
555 	t->err_time = jiffies;
556 out:
557 	return err;
558 }
559 
560 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
561 				  const struct in6_addr *v6addr)
562 {
563 	__be32 v4embed = 0;
564 	if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
565 		return true;
566 	return false;
567 }
568 
569 static int ipip6_rcv(struct sk_buff *skb)
570 {
571 	const struct iphdr *iph = ip_hdr(skb);
572 	struct ip_tunnel *tunnel;
573 	int err;
574 
575 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
576 				     iph->saddr, iph->daddr);
577 	if (tunnel != NULL) {
578 		struct pcpu_tstats *tstats;
579 
580 		secpath_reset(skb);
581 		skb->mac_header = skb->network_header;
582 		skb_reset_network_header(skb);
583 		IPCB(skb)->flags = 0;
584 		skb->protocol = htons(ETH_P_IPV6);
585 		skb->pkt_type = PACKET_HOST;
586 
587 		if (tunnel->dev->priv_flags & IFF_ISATAP) {
588 			if (!isatap_chksrc(skb, iph, tunnel)) {
589 				tunnel->dev->stats.rx_errors++;
590 				goto out;
591 			}
592 		} else {
593 			if (is_spoofed_6rd(tunnel, iph->saddr,
594 					   &ipv6_hdr(skb)->saddr) ||
595 			    is_spoofed_6rd(tunnel, iph->daddr,
596 					   &ipv6_hdr(skb)->daddr)) {
597 				tunnel->dev->stats.rx_errors++;
598 				goto out;
599 			}
600 		}
601 
602 		__skb_tunnel_rx(skb, tunnel->dev);
603 
604 		err = IP_ECN_decapsulate(iph, skb);
605 		if (unlikely(err)) {
606 			if (log_ecn_error)
607 				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
608 						     &iph->saddr, iph->tos);
609 			if (err > 1) {
610 				++tunnel->dev->stats.rx_frame_errors;
611 				++tunnel->dev->stats.rx_errors;
612 				goto out;
613 			}
614 		}
615 
616 		tstats = this_cpu_ptr(tunnel->dev->tstats);
617 		tstats->rx_packets++;
618 		tstats->rx_bytes += skb->len;
619 
620 		netif_rx(skb);
621 
622 		return 0;
623 	}
624 
625 	/* no tunnel matched,  let upstream know, ipsec may handle it */
626 	return 1;
627 out:
628 	kfree_skb(skb);
629 	return 0;
630 }
631 
632 /*
633  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
634  * stores the embedded IPv4 address in v4dst and returns true.
635  */
636 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
637 		      __be32 *v4dst)
638 {
639 #ifdef CONFIG_IPV6_SIT_6RD
640 	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
641 			      tunnel->ip6rd.prefixlen)) {
642 		unsigned int pbw0, pbi0;
643 		int pbi1;
644 		u32 d;
645 
646 		pbw0 = tunnel->ip6rd.prefixlen >> 5;
647 		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
648 
649 		d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
650 		    tunnel->ip6rd.relay_prefixlen;
651 
652 		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
653 		if (pbi1 > 0)
654 			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
655 			     (32 - pbi1);
656 
657 		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
658 		return true;
659 	}
660 #else
661 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
662 		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
663 		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
664 		return true;
665 	}
666 #endif
667 	return false;
668 }
669 
670 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
671 			     const struct in6_addr *v6dst)
672 {
673 	__be32 dst = 0;
674 	check_6rd(tunnel, v6dst, &dst);
675 	return dst;
676 }
677 
678 /*
679  *	This function assumes it is being called from dev_queue_xmit()
680  *	and that skb is filled properly by that function.
681  */
682 
683 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
684 				     struct net_device *dev)
685 {
686 	struct ip_tunnel *tunnel = netdev_priv(dev);
687 	const struct iphdr  *tiph = &tunnel->parms.iph;
688 	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
689 	u8     tos = tunnel->parms.iph.tos;
690 	__be16 df = tiph->frag_off;
691 	struct rtable *rt;     			/* Route to the other host */
692 	struct net_device *tdev;		/* Device to other host */
693 	struct iphdr  *iph;			/* Our new IP header */
694 	unsigned int max_headroom;		/* The extra header space needed */
695 	__be32 dst = tiph->daddr;
696 	struct flowi4 fl4;
697 	int    mtu;
698 	const struct in6_addr *addr6;
699 	int addr_type;
700 
701 	if (skb->protocol != htons(ETH_P_IPV6))
702 		goto tx_error;
703 
704 	if (tos == 1)
705 		tos = ipv6_get_dsfield(iph6);
706 
707 	/* ISATAP (RFC4214) - must come before 6to4 */
708 	if (dev->priv_flags & IFF_ISATAP) {
709 		struct neighbour *neigh = NULL;
710 		bool do_tx_error = false;
711 
712 		if (skb_dst(skb))
713 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
714 
715 		if (neigh == NULL) {
716 			net_dbg_ratelimited("sit: nexthop == NULL\n");
717 			goto tx_error;
718 		}
719 
720 		addr6 = (const struct in6_addr *)&neigh->primary_key;
721 		addr_type = ipv6_addr_type(addr6);
722 
723 		if ((addr_type & IPV6_ADDR_UNICAST) &&
724 		     ipv6_addr_is_isatap(addr6))
725 			dst = addr6->s6_addr32[3];
726 		else
727 			do_tx_error = true;
728 
729 		neigh_release(neigh);
730 		if (do_tx_error)
731 			goto tx_error;
732 	}
733 
734 	if (!dst)
735 		dst = try_6rd(tunnel, &iph6->daddr);
736 
737 	if (!dst) {
738 		struct neighbour *neigh = NULL;
739 		bool do_tx_error = false;
740 
741 		if (skb_dst(skb))
742 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
743 
744 		if (neigh == NULL) {
745 			net_dbg_ratelimited("sit: nexthop == NULL\n");
746 			goto tx_error;
747 		}
748 
749 		addr6 = (const struct in6_addr *)&neigh->primary_key;
750 		addr_type = ipv6_addr_type(addr6);
751 
752 		if (addr_type == IPV6_ADDR_ANY) {
753 			addr6 = &ipv6_hdr(skb)->daddr;
754 			addr_type = ipv6_addr_type(addr6);
755 		}
756 
757 		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
758 			dst = addr6->s6_addr32[3];
759 		else
760 			do_tx_error = true;
761 
762 		neigh_release(neigh);
763 		if (do_tx_error)
764 			goto tx_error;
765 	}
766 
767 	rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
768 				   dst, tiph->saddr,
769 				   0, 0,
770 				   IPPROTO_IPV6, RT_TOS(tos),
771 				   tunnel->parms.link);
772 	if (IS_ERR(rt)) {
773 		dev->stats.tx_carrier_errors++;
774 		goto tx_error_icmp;
775 	}
776 	if (rt->rt_type != RTN_UNICAST) {
777 		ip_rt_put(rt);
778 		dev->stats.tx_carrier_errors++;
779 		goto tx_error_icmp;
780 	}
781 	tdev = rt->dst.dev;
782 
783 	if (tdev == dev) {
784 		ip_rt_put(rt);
785 		dev->stats.collisions++;
786 		goto tx_error;
787 	}
788 
789 	if (df) {
790 		mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
791 
792 		if (mtu < 68) {
793 			dev->stats.collisions++;
794 			ip_rt_put(rt);
795 			goto tx_error;
796 		}
797 
798 		if (mtu < IPV6_MIN_MTU) {
799 			mtu = IPV6_MIN_MTU;
800 			df = 0;
801 		}
802 
803 		if (tunnel->parms.iph.daddr && skb_dst(skb))
804 			skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
805 
806 		if (skb->len > mtu) {
807 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
808 			ip_rt_put(rt);
809 			goto tx_error;
810 		}
811 	}
812 
813 	if (tunnel->err_count > 0) {
814 		if (time_before(jiffies,
815 				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
816 			tunnel->err_count--;
817 			dst_link_failure(skb);
818 		} else
819 			tunnel->err_count = 0;
820 	}
821 
822 	/*
823 	 * Okay, now see if we can stuff it in the buffer as-is.
824 	 */
825 	max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
826 
827 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
828 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
829 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
830 		if (!new_skb) {
831 			ip_rt_put(rt);
832 			dev->stats.tx_dropped++;
833 			dev_kfree_skb(skb);
834 			return NETDEV_TX_OK;
835 		}
836 		if (skb->sk)
837 			skb_set_owner_w(new_skb, skb->sk);
838 		dev_kfree_skb(skb);
839 		skb = new_skb;
840 		iph6 = ipv6_hdr(skb);
841 	}
842 
843 	skb->transport_header = skb->network_header;
844 	skb_push(skb, sizeof(struct iphdr));
845 	skb_reset_network_header(skb);
846 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
847 	IPCB(skb)->flags = 0;
848 	skb_dst_drop(skb);
849 	skb_dst_set(skb, &rt->dst);
850 
851 	/*
852 	 *	Push down and install the IPIP header.
853 	 */
854 
855 	iph 			=	ip_hdr(skb);
856 	iph->version		=	4;
857 	iph->ihl		=	sizeof(struct iphdr)>>2;
858 	iph->frag_off		=	df;
859 	iph->protocol		=	IPPROTO_IPV6;
860 	iph->tos		=	INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
861 	iph->daddr		=	fl4.daddr;
862 	iph->saddr		=	fl4.saddr;
863 
864 	if ((iph->ttl = tiph->ttl) == 0)
865 		iph->ttl	=	iph6->hop_limit;
866 
867 	skb->ip_summed = CHECKSUM_NONE;
868 	ip_select_ident(iph, skb_dst(skb), NULL);
869 	iptunnel_xmit(skb, dev);
870 	return NETDEV_TX_OK;
871 
872 tx_error_icmp:
873 	dst_link_failure(skb);
874 tx_error:
875 	dev->stats.tx_errors++;
876 	dev_kfree_skb(skb);
877 	return NETDEV_TX_OK;
878 }
879 
880 static void ipip6_tunnel_bind_dev(struct net_device *dev)
881 {
882 	struct net_device *tdev = NULL;
883 	struct ip_tunnel *tunnel;
884 	const struct iphdr *iph;
885 	struct flowi4 fl4;
886 
887 	tunnel = netdev_priv(dev);
888 	iph = &tunnel->parms.iph;
889 
890 	if (iph->daddr) {
891 		struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
892 							  iph->daddr, iph->saddr,
893 							  0, 0,
894 							  IPPROTO_IPV6,
895 							  RT_TOS(iph->tos),
896 							  tunnel->parms.link);
897 
898 		if (!IS_ERR(rt)) {
899 			tdev = rt->dst.dev;
900 			ip_rt_put(rt);
901 		}
902 		dev->flags |= IFF_POINTOPOINT;
903 	}
904 
905 	if (!tdev && tunnel->parms.link)
906 		tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
907 
908 	if (tdev) {
909 		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
910 		dev->mtu = tdev->mtu - sizeof(struct iphdr);
911 		if (dev->mtu < IPV6_MIN_MTU)
912 			dev->mtu = IPV6_MIN_MTU;
913 	}
914 	dev->iflink = tunnel->parms.link;
915 }
916 
917 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
918 {
919 	struct net *net = dev_net(t->dev);
920 	struct sit_net *sitn = net_generic(net, sit_net_id);
921 
922 	ipip6_tunnel_unlink(sitn, t);
923 	synchronize_net();
924 	t->parms.iph.saddr = p->iph.saddr;
925 	t->parms.iph.daddr = p->iph.daddr;
926 	memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
927 	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
928 	ipip6_tunnel_link(sitn, t);
929 	t->parms.iph.ttl = p->iph.ttl;
930 	t->parms.iph.tos = p->iph.tos;
931 	if (t->parms.link != p->link) {
932 		t->parms.link = p->link;
933 		ipip6_tunnel_bind_dev(t->dev);
934 	}
935 	netdev_state_change(t->dev);
936 }
937 
938 #ifdef CONFIG_IPV6_SIT_6RD
939 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
940 				   struct ip_tunnel_6rd *ip6rd)
941 {
942 	struct in6_addr prefix;
943 	__be32 relay_prefix;
944 
945 	if (ip6rd->relay_prefixlen > 32 ||
946 	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
947 		return -EINVAL;
948 
949 	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
950 	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
951 		return -EINVAL;
952 	if (ip6rd->relay_prefixlen)
953 		relay_prefix = ip6rd->relay_prefix &
954 			       htonl(0xffffffffUL <<
955 				     (32 - ip6rd->relay_prefixlen));
956 	else
957 		relay_prefix = 0;
958 	if (relay_prefix != ip6rd->relay_prefix)
959 		return -EINVAL;
960 
961 	t->ip6rd.prefix = prefix;
962 	t->ip6rd.relay_prefix = relay_prefix;
963 	t->ip6rd.prefixlen = ip6rd->prefixlen;
964 	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
965 	netdev_state_change(t->dev);
966 	return 0;
967 }
968 #endif
969 
970 static int
971 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
972 {
973 	int err = 0;
974 	struct ip_tunnel_parm p;
975 	struct ip_tunnel_prl prl;
976 	struct ip_tunnel *t;
977 	struct net *net = dev_net(dev);
978 	struct sit_net *sitn = net_generic(net, sit_net_id);
979 #ifdef CONFIG_IPV6_SIT_6RD
980 	struct ip_tunnel_6rd ip6rd;
981 #endif
982 
983 	switch (cmd) {
984 	case SIOCGETTUNNEL:
985 #ifdef CONFIG_IPV6_SIT_6RD
986 	case SIOCGET6RD:
987 #endif
988 		t = NULL;
989 		if (dev == sitn->fb_tunnel_dev) {
990 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
991 				err = -EFAULT;
992 				break;
993 			}
994 			t = ipip6_tunnel_locate(net, &p, 0);
995 		}
996 		if (t == NULL)
997 			t = netdev_priv(dev);
998 
999 		err = -EFAULT;
1000 		if (cmd == SIOCGETTUNNEL) {
1001 			memcpy(&p, &t->parms, sizeof(p));
1002 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1003 					 sizeof(p)))
1004 				goto done;
1005 #ifdef CONFIG_IPV6_SIT_6RD
1006 		} else {
1007 			ip6rd.prefix = t->ip6rd.prefix;
1008 			ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1009 			ip6rd.prefixlen = t->ip6rd.prefixlen;
1010 			ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1011 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1012 					 sizeof(ip6rd)))
1013 				goto done;
1014 #endif
1015 		}
1016 		err = 0;
1017 		break;
1018 
1019 	case SIOCADDTUNNEL:
1020 	case SIOCCHGTUNNEL:
1021 		err = -EPERM;
1022 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1023 			goto done;
1024 
1025 		err = -EFAULT;
1026 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1027 			goto done;
1028 
1029 		err = -EINVAL;
1030 		if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
1031 		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1032 			goto done;
1033 		if (p.iph.ttl)
1034 			p.iph.frag_off |= htons(IP_DF);
1035 
1036 		t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1037 
1038 		if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1039 			if (t != NULL) {
1040 				if (t->dev != dev) {
1041 					err = -EEXIST;
1042 					break;
1043 				}
1044 			} else {
1045 				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1046 				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1047 					err = -EINVAL;
1048 					break;
1049 				}
1050 				t = netdev_priv(dev);
1051 			}
1052 
1053 			ipip6_tunnel_update(t, &p);
1054 		}
1055 
1056 		if (t) {
1057 			err = 0;
1058 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1059 				err = -EFAULT;
1060 		} else
1061 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1062 		break;
1063 
1064 	case SIOCDELTUNNEL:
1065 		err = -EPERM;
1066 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1067 			goto done;
1068 
1069 		if (dev == sitn->fb_tunnel_dev) {
1070 			err = -EFAULT;
1071 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1072 				goto done;
1073 			err = -ENOENT;
1074 			if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1075 				goto done;
1076 			err = -EPERM;
1077 			if (t == netdev_priv(sitn->fb_tunnel_dev))
1078 				goto done;
1079 			dev = t->dev;
1080 		}
1081 		unregister_netdevice(dev);
1082 		err = 0;
1083 		break;
1084 
1085 	case SIOCGETPRL:
1086 		err = -EINVAL;
1087 		if (dev == sitn->fb_tunnel_dev)
1088 			goto done;
1089 		err = -ENOENT;
1090 		if (!(t = netdev_priv(dev)))
1091 			goto done;
1092 		err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1093 		break;
1094 
1095 	case SIOCADDPRL:
1096 	case SIOCDELPRL:
1097 	case SIOCCHGPRL:
1098 		err = -EPERM;
1099 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1100 			goto done;
1101 		err = -EINVAL;
1102 		if (dev == sitn->fb_tunnel_dev)
1103 			goto done;
1104 		err = -EFAULT;
1105 		if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1106 			goto done;
1107 		err = -ENOENT;
1108 		if (!(t = netdev_priv(dev)))
1109 			goto done;
1110 
1111 		switch (cmd) {
1112 		case SIOCDELPRL:
1113 			err = ipip6_tunnel_del_prl(t, &prl);
1114 			break;
1115 		case SIOCADDPRL:
1116 		case SIOCCHGPRL:
1117 			err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1118 			break;
1119 		}
1120 		netdev_state_change(dev);
1121 		break;
1122 
1123 #ifdef CONFIG_IPV6_SIT_6RD
1124 	case SIOCADD6RD:
1125 	case SIOCCHG6RD:
1126 	case SIOCDEL6RD:
1127 		err = -EPERM;
1128 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1129 			goto done;
1130 
1131 		err = -EFAULT;
1132 		if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1133 				   sizeof(ip6rd)))
1134 			goto done;
1135 
1136 		t = netdev_priv(dev);
1137 
1138 		if (cmd != SIOCDEL6RD) {
1139 			err = ipip6_tunnel_update_6rd(t, &ip6rd);
1140 			if (err < 0)
1141 				goto done;
1142 		} else
1143 			ipip6_tunnel_clone_6rd(dev, sitn);
1144 
1145 		err = 0;
1146 		break;
1147 #endif
1148 
1149 	default:
1150 		err = -EINVAL;
1151 	}
1152 
1153 done:
1154 	return err;
1155 }
1156 
1157 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1158 {
1159 	if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1160 		return -EINVAL;
1161 	dev->mtu = new_mtu;
1162 	return 0;
1163 }
1164 
1165 static const struct net_device_ops ipip6_netdev_ops = {
1166 	.ndo_uninit	= ipip6_tunnel_uninit,
1167 	.ndo_start_xmit	= ipip6_tunnel_xmit,
1168 	.ndo_do_ioctl	= ipip6_tunnel_ioctl,
1169 	.ndo_change_mtu	= ipip6_tunnel_change_mtu,
1170 	.ndo_get_stats64 = ip_tunnel_get_stats64,
1171 };
1172 
1173 static void ipip6_dev_free(struct net_device *dev)
1174 {
1175 	free_percpu(dev->tstats);
1176 	free_netdev(dev);
1177 }
1178 
1179 static void ipip6_tunnel_setup(struct net_device *dev)
1180 {
1181 	dev->netdev_ops		= &ipip6_netdev_ops;
1182 	dev->destructor 	= ipip6_dev_free;
1183 
1184 	dev->type		= ARPHRD_SIT;
1185 	dev->hard_header_len 	= LL_MAX_HEADER + sizeof(struct iphdr);
1186 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr);
1187 	dev->flags		= IFF_NOARP;
1188 	dev->priv_flags	       &= ~IFF_XMIT_DST_RELEASE;
1189 	dev->iflink		= 0;
1190 	dev->addr_len		= 4;
1191 	dev->features		|= NETIF_F_NETNS_LOCAL;
1192 	dev->features		|= NETIF_F_LLTX;
1193 }
1194 
1195 static int ipip6_tunnel_init(struct net_device *dev)
1196 {
1197 	struct ip_tunnel *tunnel = netdev_priv(dev);
1198 
1199 	tunnel->dev = dev;
1200 
1201 	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1202 	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1203 
1204 	ipip6_tunnel_bind_dev(dev);
1205 	dev->tstats = alloc_percpu(struct pcpu_tstats);
1206 	if (!dev->tstats)
1207 		return -ENOMEM;
1208 
1209 	return 0;
1210 }
1211 
1212 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1213 {
1214 	struct ip_tunnel *tunnel = netdev_priv(dev);
1215 	struct iphdr *iph = &tunnel->parms.iph;
1216 	struct net *net = dev_net(dev);
1217 	struct sit_net *sitn = net_generic(net, sit_net_id);
1218 
1219 	tunnel->dev = dev;
1220 	strcpy(tunnel->parms.name, dev->name);
1221 
1222 	iph->version		= 4;
1223 	iph->protocol		= IPPROTO_IPV6;
1224 	iph->ihl		= 5;
1225 	iph->ttl		= 64;
1226 
1227 	dev->tstats = alloc_percpu(struct pcpu_tstats);
1228 	if (!dev->tstats)
1229 		return -ENOMEM;
1230 	dev_hold(dev);
1231 	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1232 	return 0;
1233 }
1234 
1235 static void ipip6_netlink_parms(struct nlattr *data[],
1236 				struct ip_tunnel_parm *parms)
1237 {
1238 	memset(parms, 0, sizeof(*parms));
1239 
1240 	parms->iph.version = 4;
1241 	parms->iph.protocol = IPPROTO_IPV6;
1242 	parms->iph.ihl = 5;
1243 	parms->iph.ttl = 64;
1244 
1245 	if (!data)
1246 		return;
1247 
1248 	if (data[IFLA_IPTUN_LINK])
1249 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1250 
1251 	if (data[IFLA_IPTUN_LOCAL])
1252 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1253 
1254 	if (data[IFLA_IPTUN_REMOTE])
1255 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1256 
1257 	if (data[IFLA_IPTUN_TTL]) {
1258 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1259 		if (parms->iph.ttl)
1260 			parms->iph.frag_off = htons(IP_DF);
1261 	}
1262 
1263 	if (data[IFLA_IPTUN_TOS])
1264 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1265 
1266 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1267 		parms->iph.frag_off = htons(IP_DF);
1268 
1269 	if (data[IFLA_IPTUN_FLAGS])
1270 		parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1271 }
1272 
1273 #ifdef CONFIG_IPV6_SIT_6RD
1274 /* This function returns true when 6RD attributes are present in the nl msg */
1275 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1276 				    struct ip_tunnel_6rd *ip6rd)
1277 {
1278 	bool ret = false;
1279 	memset(ip6rd, 0, sizeof(*ip6rd));
1280 
1281 	if (!data)
1282 		return ret;
1283 
1284 	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1285 		ret = true;
1286 		nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1287 			   sizeof(struct in6_addr));
1288 	}
1289 
1290 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1291 		ret = true;
1292 		ip6rd->relay_prefix =
1293 			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1294 	}
1295 
1296 	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1297 		ret = true;
1298 		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1299 	}
1300 
1301 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1302 		ret = true;
1303 		ip6rd->relay_prefixlen =
1304 			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1305 	}
1306 
1307 	return ret;
1308 }
1309 #endif
1310 
1311 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1312 			 struct nlattr *tb[], struct nlattr *data[])
1313 {
1314 	struct net *net = dev_net(dev);
1315 	struct ip_tunnel *nt;
1316 #ifdef CONFIG_IPV6_SIT_6RD
1317 	struct ip_tunnel_6rd ip6rd;
1318 #endif
1319 	int err;
1320 
1321 	nt = netdev_priv(dev);
1322 	ipip6_netlink_parms(data, &nt->parms);
1323 
1324 	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1325 		return -EEXIST;
1326 
1327 	err = ipip6_tunnel_create(dev);
1328 	if (err < 0)
1329 		return err;
1330 
1331 #ifdef CONFIG_IPV6_SIT_6RD
1332 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1333 		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1334 #endif
1335 
1336 	return err;
1337 }
1338 
1339 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1340 			  struct nlattr *data[])
1341 {
1342 	struct ip_tunnel *t;
1343 	struct ip_tunnel_parm p;
1344 	struct net *net = dev_net(dev);
1345 	struct sit_net *sitn = net_generic(net, sit_net_id);
1346 #ifdef CONFIG_IPV6_SIT_6RD
1347 	struct ip_tunnel_6rd ip6rd;
1348 #endif
1349 
1350 	if (dev == sitn->fb_tunnel_dev)
1351 		return -EINVAL;
1352 
1353 	ipip6_netlink_parms(data, &p);
1354 
1355 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1356 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1357 		return -EINVAL;
1358 
1359 	t = ipip6_tunnel_locate(net, &p, 0);
1360 
1361 	if (t) {
1362 		if (t->dev != dev)
1363 			return -EEXIST;
1364 	} else
1365 		t = netdev_priv(dev);
1366 
1367 	ipip6_tunnel_update(t, &p);
1368 
1369 #ifdef CONFIG_IPV6_SIT_6RD
1370 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1371 		return ipip6_tunnel_update_6rd(t, &ip6rd);
1372 #endif
1373 
1374 	return 0;
1375 }
1376 
1377 static size_t ipip6_get_size(const struct net_device *dev)
1378 {
1379 	return
1380 		/* IFLA_IPTUN_LINK */
1381 		nla_total_size(4) +
1382 		/* IFLA_IPTUN_LOCAL */
1383 		nla_total_size(4) +
1384 		/* IFLA_IPTUN_REMOTE */
1385 		nla_total_size(4) +
1386 		/* IFLA_IPTUN_TTL */
1387 		nla_total_size(1) +
1388 		/* IFLA_IPTUN_TOS */
1389 		nla_total_size(1) +
1390 		/* IFLA_IPTUN_PMTUDISC */
1391 		nla_total_size(1) +
1392 		/* IFLA_IPTUN_FLAGS */
1393 		nla_total_size(2) +
1394 #ifdef CONFIG_IPV6_SIT_6RD
1395 		/* IFLA_IPTUN_6RD_PREFIX */
1396 		nla_total_size(sizeof(struct in6_addr)) +
1397 		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1398 		nla_total_size(4) +
1399 		/* IFLA_IPTUN_6RD_PREFIXLEN */
1400 		nla_total_size(2) +
1401 		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1402 		nla_total_size(2) +
1403 #endif
1404 		0;
1405 }
1406 
1407 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1408 {
1409 	struct ip_tunnel *tunnel = netdev_priv(dev);
1410 	struct ip_tunnel_parm *parm = &tunnel->parms;
1411 
1412 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1413 	    nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1414 	    nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1415 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1416 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1417 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1418 		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1419 	    nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1420 		goto nla_put_failure;
1421 
1422 #ifdef CONFIG_IPV6_SIT_6RD
1423 	if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1424 		    &tunnel->ip6rd.prefix) ||
1425 	    nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1426 			 tunnel->ip6rd.relay_prefix) ||
1427 	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1428 			tunnel->ip6rd.prefixlen) ||
1429 	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1430 			tunnel->ip6rd.relay_prefixlen))
1431 		goto nla_put_failure;
1432 #endif
1433 
1434 	return 0;
1435 
1436 nla_put_failure:
1437 	return -EMSGSIZE;
1438 }
1439 
1440 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1441 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1442 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1443 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1444 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1445 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1446 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1447 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1448 #ifdef CONFIG_IPV6_SIT_6RD
1449 	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1450 	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1451 	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1452 	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1453 #endif
1454 };
1455 
1456 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1457 	.kind		= "sit",
1458 	.maxtype	= IFLA_IPTUN_MAX,
1459 	.policy		= ipip6_policy,
1460 	.priv_size	= sizeof(struct ip_tunnel),
1461 	.setup		= ipip6_tunnel_setup,
1462 	.newlink	= ipip6_newlink,
1463 	.changelink	= ipip6_changelink,
1464 	.get_size	= ipip6_get_size,
1465 	.fill_info	= ipip6_fill_info,
1466 };
1467 
1468 static struct xfrm_tunnel sit_handler __read_mostly = {
1469 	.handler	=	ipip6_rcv,
1470 	.err_handler	=	ipip6_err,
1471 	.priority	=	1,
1472 };
1473 
1474 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1475 {
1476 	int prio;
1477 
1478 	for (prio = 1; prio < 4; prio++) {
1479 		int h;
1480 		for (h = 0; h < HASH_SIZE; h++) {
1481 			struct ip_tunnel *t;
1482 
1483 			t = rtnl_dereference(sitn->tunnels[prio][h]);
1484 			while (t != NULL) {
1485 				unregister_netdevice_queue(t->dev, head);
1486 				t = rtnl_dereference(t->next);
1487 			}
1488 		}
1489 	}
1490 }
1491 
1492 static int __net_init sit_init_net(struct net *net)
1493 {
1494 	struct sit_net *sitn = net_generic(net, sit_net_id);
1495 	struct ip_tunnel *t;
1496 	int err;
1497 
1498 	sitn->tunnels[0] = sitn->tunnels_wc;
1499 	sitn->tunnels[1] = sitn->tunnels_l;
1500 	sitn->tunnels[2] = sitn->tunnels_r;
1501 	sitn->tunnels[3] = sitn->tunnels_r_l;
1502 
1503 	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1504 					   ipip6_tunnel_setup);
1505 	if (!sitn->fb_tunnel_dev) {
1506 		err = -ENOMEM;
1507 		goto err_alloc_dev;
1508 	}
1509 	dev_net_set(sitn->fb_tunnel_dev, net);
1510 
1511 	err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1512 	if (err)
1513 		goto err_dev_free;
1514 
1515 	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1516 
1517 	if ((err = register_netdev(sitn->fb_tunnel_dev)))
1518 		goto err_reg_dev;
1519 
1520 	t = netdev_priv(sitn->fb_tunnel_dev);
1521 
1522 	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1523 	return 0;
1524 
1525 err_reg_dev:
1526 	dev_put(sitn->fb_tunnel_dev);
1527 err_dev_free:
1528 	ipip6_dev_free(sitn->fb_tunnel_dev);
1529 err_alloc_dev:
1530 	return err;
1531 }
1532 
1533 static void __net_exit sit_exit_net(struct net *net)
1534 {
1535 	struct sit_net *sitn = net_generic(net, sit_net_id);
1536 	LIST_HEAD(list);
1537 
1538 	rtnl_lock();
1539 	sit_destroy_tunnels(sitn, &list);
1540 	unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1541 	unregister_netdevice_many(&list);
1542 	rtnl_unlock();
1543 }
1544 
1545 static struct pernet_operations sit_net_ops = {
1546 	.init = sit_init_net,
1547 	.exit = sit_exit_net,
1548 	.id   = &sit_net_id,
1549 	.size = sizeof(struct sit_net),
1550 };
1551 
1552 static void __exit sit_cleanup(void)
1553 {
1554 	rtnl_link_unregister(&sit_link_ops);
1555 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1556 
1557 	unregister_pernet_device(&sit_net_ops);
1558 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1559 }
1560 
1561 static int __init sit_init(void)
1562 {
1563 	int err;
1564 
1565 	pr_info("IPv6 over IPv4 tunneling driver\n");
1566 
1567 	err = register_pernet_device(&sit_net_ops);
1568 	if (err < 0)
1569 		return err;
1570 	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1571 	if (err < 0) {
1572 		pr_info("%s: can't add protocol\n", __func__);
1573 		goto xfrm_tunnel_failed;
1574 	}
1575 	err = rtnl_link_register(&sit_link_ops);
1576 	if (err < 0)
1577 		goto rtnl_link_failed;
1578 
1579 out:
1580 	return err;
1581 
1582 rtnl_link_failed:
1583 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1584 xfrm_tunnel_failed:
1585 	unregister_pernet_device(&sit_net_ops);
1586 	goto out;
1587 }
1588 
1589 module_init(sit_init);
1590 module_exit(sit_cleanup);
1591 MODULE_LICENSE("GPL");
1592 MODULE_ALIAS_NETDEV("sit0");
1593