xref: /linux/net/ipv6/sit.c (revision a115bc070b1fc57ab23f3972401425927b5b465c)
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 #include <linux/module.h>
21 #include <linux/capability.h>
22 #include <linux/errno.h>
23 #include <linux/types.h>
24 #include <linux/socket.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/in6.h>
28 #include <linux/netdevice.h>
29 #include <linux/if_arp.h>
30 #include <linux/icmp.h>
31 #include <asm/uaccess.h>
32 #include <linux/init.h>
33 #include <linux/netfilter_ipv4.h>
34 #include <linux/if_ether.h>
35 
36 #include <net/sock.h>
37 #include <net/snmp.h>
38 
39 #include <net/ipv6.h>
40 #include <net/protocol.h>
41 #include <net/transp_v6.h>
42 #include <net/ip6_fib.h>
43 #include <net/ip6_route.h>
44 #include <net/ndisc.h>
45 #include <net/addrconf.h>
46 #include <net/ip.h>
47 #include <net/udp.h>
48 #include <net/icmp.h>
49 #include <net/ipip.h>
50 #include <net/inet_ecn.h>
51 #include <net/xfrm.h>
52 #include <net/dsfield.h>
53 #include <net/net_namespace.h>
54 #include <net/netns/generic.h>
55 
56 /*
57    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
58 
59    For comments look at net/ipv4/ip_gre.c --ANK
60  */
61 
62 #define HASH_SIZE  16
63 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
64 
65 static void ipip6_tunnel_init(struct net_device *dev);
66 static void ipip6_tunnel_setup(struct net_device *dev);
67 
68 static int sit_net_id __read_mostly;
69 struct sit_net {
70 	struct ip_tunnel *tunnels_r_l[HASH_SIZE];
71 	struct ip_tunnel *tunnels_r[HASH_SIZE];
72 	struct ip_tunnel *tunnels_l[HASH_SIZE];
73 	struct ip_tunnel *tunnels_wc[1];
74 	struct ip_tunnel **tunnels[4];
75 
76 	struct net_device *fb_tunnel_dev;
77 };
78 
79 /*
80  * Locking : hash tables are protected by RCU and a spinlock
81  */
82 static DEFINE_SPINLOCK(ipip6_lock);
83 
84 #define for_each_ip_tunnel_rcu(start) \
85 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
86 
87 /*
88  * Must be invoked with rcu_read_lock
89  */
90 static struct ip_tunnel * ipip6_tunnel_lookup(struct net *net,
91 		struct net_device *dev, __be32 remote, __be32 local)
92 {
93 	unsigned h0 = HASH(remote);
94 	unsigned h1 = HASH(local);
95 	struct ip_tunnel *t;
96 	struct sit_net *sitn = net_generic(net, sit_net_id);
97 
98 	for_each_ip_tunnel_rcu(sitn->tunnels_r_l[h0 ^ h1]) {
99 		if (local == t->parms.iph.saddr &&
100 		    remote == t->parms.iph.daddr &&
101 		    (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
102 		    (t->dev->flags & IFF_UP))
103 			return t;
104 	}
105 	for_each_ip_tunnel_rcu(sitn->tunnels_r[h0]) {
106 		if (remote == t->parms.iph.daddr &&
107 		    (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
108 		    (t->dev->flags & IFF_UP))
109 			return t;
110 	}
111 	for_each_ip_tunnel_rcu(sitn->tunnels_l[h1]) {
112 		if (local == t->parms.iph.saddr &&
113 		    (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
114 		    (t->dev->flags & IFF_UP))
115 			return t;
116 	}
117 	t = rcu_dereference(sitn->tunnels_wc[0]);
118 	if ((t != NULL) && (t->dev->flags & IFF_UP))
119 		return t;
120 	return NULL;
121 }
122 
123 static struct ip_tunnel **__ipip6_bucket(struct sit_net *sitn,
124 		struct ip_tunnel_parm *parms)
125 {
126 	__be32 remote = parms->iph.daddr;
127 	__be32 local = parms->iph.saddr;
128 	unsigned h = 0;
129 	int prio = 0;
130 
131 	if (remote) {
132 		prio |= 2;
133 		h ^= HASH(remote);
134 	}
135 	if (local) {
136 		prio |= 1;
137 		h ^= HASH(local);
138 	}
139 	return &sitn->tunnels[prio][h];
140 }
141 
142 static inline struct ip_tunnel **ipip6_bucket(struct sit_net *sitn,
143 		struct ip_tunnel *t)
144 {
145 	return __ipip6_bucket(sitn, &t->parms);
146 }
147 
148 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
149 {
150 	struct ip_tunnel **tp;
151 
152 	for (tp = ipip6_bucket(sitn, t); *tp; tp = &(*tp)->next) {
153 		if (t == *tp) {
154 			spin_lock_bh(&ipip6_lock);
155 			*tp = t->next;
156 			spin_unlock_bh(&ipip6_lock);
157 			break;
158 		}
159 	}
160 }
161 
162 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
163 {
164 	struct ip_tunnel **tp = ipip6_bucket(sitn, t);
165 
166 	spin_lock_bh(&ipip6_lock);
167 	t->next = *tp;
168 	rcu_assign_pointer(*tp, t);
169 	spin_unlock_bh(&ipip6_lock);
170 }
171 
172 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
173 {
174 #ifdef CONFIG_IPV6_SIT_6RD
175 	struct ip_tunnel *t = netdev_priv(dev);
176 
177 	if (t->dev == sitn->fb_tunnel_dev) {
178 		ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
179 		t->ip6rd.relay_prefix = 0;
180 		t->ip6rd.prefixlen = 16;
181 		t->ip6rd.relay_prefixlen = 0;
182 	} else {
183 		struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
184 		memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
185 	}
186 #endif
187 }
188 
189 static struct ip_tunnel * ipip6_tunnel_locate(struct net *net,
190 		struct ip_tunnel_parm *parms, int create)
191 {
192 	__be32 remote = parms->iph.daddr;
193 	__be32 local = parms->iph.saddr;
194 	struct ip_tunnel *t, **tp, *nt;
195 	struct net_device *dev;
196 	char name[IFNAMSIZ];
197 	struct sit_net *sitn = net_generic(net, sit_net_id);
198 
199 	for (tp = __ipip6_bucket(sitn, parms); (t = *tp) != NULL; tp = &t->next) {
200 		if (local == t->parms.iph.saddr &&
201 		    remote == t->parms.iph.daddr &&
202 		    parms->link == t->parms.link) {
203 			if (create)
204 				return NULL;
205 			else
206 				return t;
207 		}
208 	}
209 	if (!create)
210 		goto failed;
211 
212 	if (parms->name[0])
213 		strlcpy(name, parms->name, IFNAMSIZ);
214 	else
215 		sprintf(name, "sit%%d");
216 
217 	dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
218 	if (dev == NULL)
219 		return NULL;
220 
221 	dev_net_set(dev, net);
222 
223 	if (strchr(name, '%')) {
224 		if (dev_alloc_name(dev, name) < 0)
225 			goto failed_free;
226 	}
227 
228 	nt = netdev_priv(dev);
229 
230 	nt->parms = *parms;
231 	ipip6_tunnel_init(dev);
232 	ipip6_tunnel_clone_6rd(dev, sitn);
233 
234 	if (parms->i_flags & SIT_ISATAP)
235 		dev->priv_flags |= IFF_ISATAP;
236 
237 	if (register_netdevice(dev) < 0)
238 		goto failed_free;
239 
240 	dev_hold(dev);
241 
242 	ipip6_tunnel_link(sitn, nt);
243 	return nt;
244 
245 failed_free:
246 	free_netdev(dev);
247 failed:
248 	return NULL;
249 }
250 
251 static DEFINE_SPINLOCK(ipip6_prl_lock);
252 
253 #define for_each_prl_rcu(start)			\
254 	for (prl = rcu_dereference(start);	\
255 	     prl;				\
256 	     prl = rcu_dereference(prl->next))
257 
258 static struct ip_tunnel_prl_entry *
259 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
260 {
261 	struct ip_tunnel_prl_entry *prl;
262 
263 	for_each_prl_rcu(t->prl)
264 		if (prl->addr == addr)
265 			break;
266 	return prl;
267 
268 }
269 
270 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
271 				struct ip_tunnel_prl __user *a)
272 {
273 	struct ip_tunnel_prl kprl, *kp;
274 	struct ip_tunnel_prl_entry *prl;
275 	unsigned int cmax, c = 0, ca, len;
276 	int ret = 0;
277 
278 	if (copy_from_user(&kprl, a, sizeof(kprl)))
279 		return -EFAULT;
280 	cmax = kprl.datalen / sizeof(kprl);
281 	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
282 		cmax = 1;
283 
284 	/* For simple GET or for root users,
285 	 * we try harder to allocate.
286 	 */
287 	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
288 		kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
289 		NULL;
290 
291 	rcu_read_lock();
292 
293 	ca = t->prl_count < cmax ? t->prl_count : cmax;
294 
295 	if (!kp) {
296 		/* We don't try hard to allocate much memory for
297 		 * non-root users.
298 		 * For root users, retry allocating enough memory for
299 		 * the answer.
300 		 */
301 		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
302 		if (!kp) {
303 			ret = -ENOMEM;
304 			goto out;
305 		}
306 	}
307 
308 	c = 0;
309 	for_each_prl_rcu(t->prl) {
310 		if (c >= cmax)
311 			break;
312 		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
313 			continue;
314 		kp[c].addr = prl->addr;
315 		kp[c].flags = prl->flags;
316 		c++;
317 		if (kprl.addr != htonl(INADDR_ANY))
318 			break;
319 	}
320 out:
321 	rcu_read_unlock();
322 
323 	len = sizeof(*kp) * c;
324 	ret = 0;
325 	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
326 		ret = -EFAULT;
327 
328 	kfree(kp);
329 
330 	return ret;
331 }
332 
333 static int
334 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
335 {
336 	struct ip_tunnel_prl_entry *p;
337 	int err = 0;
338 
339 	if (a->addr == htonl(INADDR_ANY))
340 		return -EINVAL;
341 
342 	spin_lock(&ipip6_prl_lock);
343 
344 	for (p = t->prl; p; p = p->next) {
345 		if (p->addr == a->addr) {
346 			if (chg) {
347 				p->flags = a->flags;
348 				goto out;
349 			}
350 			err = -EEXIST;
351 			goto out;
352 		}
353 	}
354 
355 	if (chg) {
356 		err = -ENXIO;
357 		goto out;
358 	}
359 
360 	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
361 	if (!p) {
362 		err = -ENOBUFS;
363 		goto out;
364 	}
365 
366 	p->next = t->prl;
367 	p->addr = a->addr;
368 	p->flags = a->flags;
369 	t->prl_count++;
370 	rcu_assign_pointer(t->prl, p);
371 out:
372 	spin_unlock(&ipip6_prl_lock);
373 	return err;
374 }
375 
376 static void prl_entry_destroy_rcu(struct rcu_head *head)
377 {
378 	kfree(container_of(head, struct ip_tunnel_prl_entry, rcu_head));
379 }
380 
381 static void prl_list_destroy_rcu(struct rcu_head *head)
382 {
383 	struct ip_tunnel_prl_entry *p, *n;
384 
385 	p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
386 	do {
387 		n = p->next;
388 		kfree(p);
389 		p = n;
390 	} while (p);
391 }
392 
393 static int
394 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
395 {
396 	struct ip_tunnel_prl_entry *x, **p;
397 	int err = 0;
398 
399 	spin_lock(&ipip6_prl_lock);
400 
401 	if (a && a->addr != htonl(INADDR_ANY)) {
402 		for (p = &t->prl; *p; p = &(*p)->next) {
403 			if ((*p)->addr == a->addr) {
404 				x = *p;
405 				*p = x->next;
406 				call_rcu(&x->rcu_head, prl_entry_destroy_rcu);
407 				t->prl_count--;
408 				goto out;
409 			}
410 		}
411 		err = -ENXIO;
412 	} else {
413 		if (t->prl) {
414 			t->prl_count = 0;
415 			x = t->prl;
416 			call_rcu(&x->rcu_head, prl_list_destroy_rcu);
417 			t->prl = NULL;
418 		}
419 	}
420 out:
421 	spin_unlock(&ipip6_prl_lock);
422 	return err;
423 }
424 
425 static int
426 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
427 {
428 	struct ip_tunnel_prl_entry *p;
429 	int ok = 1;
430 
431 	rcu_read_lock();
432 	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
433 	if (p) {
434 		if (p->flags & PRL_DEFAULT)
435 			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
436 		else
437 			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
438 	} else {
439 		struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
440 		if (ipv6_addr_is_isatap(addr6) &&
441 		    (addr6->s6_addr32[3] == iph->saddr) &&
442 		    ipv6_chk_prefix(addr6, t->dev))
443 			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
444 		else
445 			ok = 0;
446 	}
447 	rcu_read_unlock();
448 	return ok;
449 }
450 
451 static void ipip6_tunnel_uninit(struct net_device *dev)
452 {
453 	struct net *net = dev_net(dev);
454 	struct sit_net *sitn = net_generic(net, sit_net_id);
455 
456 	if (dev == sitn->fb_tunnel_dev) {
457 		spin_lock_bh(&ipip6_lock);
458 		sitn->tunnels_wc[0] = NULL;
459 		spin_unlock_bh(&ipip6_lock);
460 		dev_put(dev);
461 	} else {
462 		ipip6_tunnel_unlink(sitn, netdev_priv(dev));
463 		ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
464 		dev_put(dev);
465 	}
466 }
467 
468 
469 static int ipip6_err(struct sk_buff *skb, u32 info)
470 {
471 
472 /* All the routers (except for Linux) return only
473    8 bytes of packet payload. It means, that precise relaying of
474    ICMP in the real Internet is absolutely infeasible.
475  */
476 	struct iphdr *iph = (struct iphdr*)skb->data;
477 	const int type = icmp_hdr(skb)->type;
478 	const int code = icmp_hdr(skb)->code;
479 	struct ip_tunnel *t;
480 	int err;
481 
482 	switch (type) {
483 	default:
484 	case ICMP_PARAMETERPROB:
485 		return 0;
486 
487 	case ICMP_DEST_UNREACH:
488 		switch (code) {
489 		case ICMP_SR_FAILED:
490 		case ICMP_PORT_UNREACH:
491 			/* Impossible event. */
492 			return 0;
493 		case ICMP_FRAG_NEEDED:
494 			/* Soft state for pmtu is maintained by IP core. */
495 			return 0;
496 		default:
497 			/* All others are translated to HOST_UNREACH.
498 			   rfc2003 contains "deep thoughts" about NET_UNREACH,
499 			   I believe they are just ether pollution. --ANK
500 			 */
501 			break;
502 		}
503 		break;
504 	case ICMP_TIME_EXCEEDED:
505 		if (code != ICMP_EXC_TTL)
506 			return 0;
507 		break;
508 	}
509 
510 	err = -ENOENT;
511 
512 	rcu_read_lock();
513 	t = ipip6_tunnel_lookup(dev_net(skb->dev),
514 				skb->dev,
515 				iph->daddr,
516 				iph->saddr);
517 	if (t == NULL || t->parms.iph.daddr == 0)
518 		goto out;
519 
520 	err = 0;
521 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
522 		goto out;
523 
524 	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
525 		t->err_count++;
526 	else
527 		t->err_count = 1;
528 	t->err_time = jiffies;
529 out:
530 	rcu_read_unlock();
531 	return err;
532 }
533 
534 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
535 {
536 	if (INET_ECN_is_ce(iph->tos))
537 		IP6_ECN_set_ce(ipv6_hdr(skb));
538 }
539 
540 static int ipip6_rcv(struct sk_buff *skb)
541 {
542 	struct iphdr *iph;
543 	struct ip_tunnel *tunnel;
544 
545 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
546 		goto out;
547 
548 	iph = ip_hdr(skb);
549 
550 	rcu_read_lock();
551 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
552 				     iph->saddr, iph->daddr);
553 	if (tunnel != NULL) {
554 		secpath_reset(skb);
555 		skb->mac_header = skb->network_header;
556 		skb_reset_network_header(skb);
557 		IPCB(skb)->flags = 0;
558 		skb->protocol = htons(ETH_P_IPV6);
559 		skb->pkt_type = PACKET_HOST;
560 
561 		if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
562 		    !isatap_chksrc(skb, iph, tunnel)) {
563 			tunnel->dev->stats.rx_errors++;
564 			rcu_read_unlock();
565 			kfree_skb(skb);
566 			return 0;
567 		}
568 		tunnel->dev->stats.rx_packets++;
569 		tunnel->dev->stats.rx_bytes += skb->len;
570 		skb->dev = tunnel->dev;
571 		skb_dst_drop(skb);
572 		nf_reset(skb);
573 		ipip6_ecn_decapsulate(iph, skb);
574 		netif_rx(skb);
575 		rcu_read_unlock();
576 		return 0;
577 	}
578 
579 	icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
580 	rcu_read_unlock();
581 out:
582 	kfree_skb(skb);
583 	return 0;
584 }
585 
586 /*
587  * Returns the embedded IPv4 address if the IPv6 address
588  * comes from 6rd / 6to4 (RFC 3056) addr space.
589  */
590 static inline
591 __be32 try_6rd(struct in6_addr *v6dst, struct ip_tunnel *tunnel)
592 {
593 	__be32 dst = 0;
594 
595 #ifdef CONFIG_IPV6_SIT_6RD
596 	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
597 			      tunnel->ip6rd.prefixlen)) {
598 		unsigned pbw0, pbi0;
599 		int pbi1;
600 		u32 d;
601 
602 		pbw0 = tunnel->ip6rd.prefixlen >> 5;
603 		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
604 
605 		d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
606 		    tunnel->ip6rd.relay_prefixlen;
607 
608 		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
609 		if (pbi1 > 0)
610 			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
611 			     (32 - pbi1);
612 
613 		dst = tunnel->ip6rd.relay_prefix | htonl(d);
614 	}
615 #else
616 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
617 		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
618 		memcpy(&dst, &v6dst->s6_addr16[1], 4);
619 	}
620 #endif
621 	return dst;
622 }
623 
624 /*
625  *	This function assumes it is being called from dev_queue_xmit()
626  *	and that skb is filled properly by that function.
627  */
628 
629 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
630 				     struct net_device *dev)
631 {
632 	struct ip_tunnel *tunnel = netdev_priv(dev);
633 	struct net_device_stats *stats = &dev->stats;
634 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
635 	struct iphdr  *tiph = &tunnel->parms.iph;
636 	struct ipv6hdr *iph6 = ipv6_hdr(skb);
637 	u8     tos = tunnel->parms.iph.tos;
638 	__be16 df = tiph->frag_off;
639 	struct rtable *rt;     			/* Route to the other host */
640 	struct net_device *tdev;			/* Device to other host */
641 	struct iphdr  *iph;			/* Our new IP header */
642 	unsigned int max_headroom;		/* The extra header space needed */
643 	__be32 dst = tiph->daddr;
644 	int    mtu;
645 	struct in6_addr *addr6;
646 	int addr_type;
647 
648 	if (skb->protocol != htons(ETH_P_IPV6))
649 		goto tx_error;
650 
651 	/* ISATAP (RFC4214) - must come before 6to4 */
652 	if (dev->priv_flags & IFF_ISATAP) {
653 		struct neighbour *neigh = NULL;
654 
655 		if (skb_dst(skb))
656 			neigh = skb_dst(skb)->neighbour;
657 
658 		if (neigh == NULL) {
659 			if (net_ratelimit())
660 				printk(KERN_DEBUG "sit: nexthop == NULL\n");
661 			goto tx_error;
662 		}
663 
664 		addr6 = (struct in6_addr*)&neigh->primary_key;
665 		addr_type = ipv6_addr_type(addr6);
666 
667 		if ((addr_type & IPV6_ADDR_UNICAST) &&
668 		     ipv6_addr_is_isatap(addr6))
669 			dst = addr6->s6_addr32[3];
670 		else
671 			goto tx_error;
672 	}
673 
674 	if (!dst)
675 		dst = try_6rd(&iph6->daddr, tunnel);
676 
677 	if (!dst) {
678 		struct neighbour *neigh = NULL;
679 
680 		if (skb_dst(skb))
681 			neigh = skb_dst(skb)->neighbour;
682 
683 		if (neigh == NULL) {
684 			if (net_ratelimit())
685 				printk(KERN_DEBUG "sit: nexthop == NULL\n");
686 			goto tx_error;
687 		}
688 
689 		addr6 = (struct in6_addr*)&neigh->primary_key;
690 		addr_type = ipv6_addr_type(addr6);
691 
692 		if (addr_type == IPV6_ADDR_ANY) {
693 			addr6 = &ipv6_hdr(skb)->daddr;
694 			addr_type = ipv6_addr_type(addr6);
695 		}
696 
697 		if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
698 			goto tx_error_icmp;
699 
700 		dst = addr6->s6_addr32[3];
701 	}
702 
703 	{
704 		struct flowi fl = { .nl_u = { .ip4_u =
705 					      { .daddr = dst,
706 						.saddr = tiph->saddr,
707 						.tos = RT_TOS(tos) } },
708 				    .oif = tunnel->parms.link,
709 				    .proto = IPPROTO_IPV6 };
710 		if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
711 			stats->tx_carrier_errors++;
712 			goto tx_error_icmp;
713 		}
714 	}
715 	if (rt->rt_type != RTN_UNICAST) {
716 		ip_rt_put(rt);
717 		stats->tx_carrier_errors++;
718 		goto tx_error_icmp;
719 	}
720 	tdev = rt->u.dst.dev;
721 
722 	if (tdev == dev) {
723 		ip_rt_put(rt);
724 		stats->collisions++;
725 		goto tx_error;
726 	}
727 
728 	if (df) {
729 		mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
730 
731 		if (mtu < 68) {
732 			stats->collisions++;
733 			ip_rt_put(rt);
734 			goto tx_error;
735 		}
736 
737 		if (mtu < IPV6_MIN_MTU) {
738 			mtu = IPV6_MIN_MTU;
739 			df = 0;
740 		}
741 
742 		if (tunnel->parms.iph.daddr && skb_dst(skb))
743 			skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
744 
745 		if (skb->len > mtu) {
746 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
747 			ip_rt_put(rt);
748 			goto tx_error;
749 		}
750 	}
751 
752 	if (tunnel->err_count > 0) {
753 		if (time_before(jiffies,
754 				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
755 			tunnel->err_count--;
756 			dst_link_failure(skb);
757 		} else
758 			tunnel->err_count = 0;
759 	}
760 
761 	/*
762 	 * Okay, now see if we can stuff it in the buffer as-is.
763 	 */
764 	max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
765 
766 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
767 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
768 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
769 		if (!new_skb) {
770 			ip_rt_put(rt);
771 			txq->tx_dropped++;
772 			dev_kfree_skb(skb);
773 			return NETDEV_TX_OK;
774 		}
775 		if (skb->sk)
776 			skb_set_owner_w(new_skb, skb->sk);
777 		dev_kfree_skb(skb);
778 		skb = new_skb;
779 		iph6 = ipv6_hdr(skb);
780 	}
781 
782 	skb->transport_header = skb->network_header;
783 	skb_push(skb, sizeof(struct iphdr));
784 	skb_reset_network_header(skb);
785 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
786 	IPCB(skb)->flags = 0;
787 	skb_dst_drop(skb);
788 	skb_dst_set(skb, &rt->u.dst);
789 
790 	/*
791 	 *	Push down and install the IPIP header.
792 	 */
793 
794 	iph 			=	ip_hdr(skb);
795 	iph->version		=	4;
796 	iph->ihl		=	sizeof(struct iphdr)>>2;
797 	iph->frag_off		=	df;
798 	iph->protocol		=	IPPROTO_IPV6;
799 	iph->tos		=	INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
800 	iph->daddr		=	rt->rt_dst;
801 	iph->saddr		=	rt->rt_src;
802 
803 	if ((iph->ttl = tiph->ttl) == 0)
804 		iph->ttl	=	iph6->hop_limit;
805 
806 	nf_reset(skb);
807 
808 	IPTUNNEL_XMIT();
809 	return NETDEV_TX_OK;
810 
811 tx_error_icmp:
812 	dst_link_failure(skb);
813 tx_error:
814 	stats->tx_errors++;
815 	dev_kfree_skb(skb);
816 	return NETDEV_TX_OK;
817 }
818 
819 static void ipip6_tunnel_bind_dev(struct net_device *dev)
820 {
821 	struct net_device *tdev = NULL;
822 	struct ip_tunnel *tunnel;
823 	struct iphdr *iph;
824 
825 	tunnel = netdev_priv(dev);
826 	iph = &tunnel->parms.iph;
827 
828 	if (iph->daddr) {
829 		struct flowi fl = { .nl_u = { .ip4_u =
830 					      { .daddr = iph->daddr,
831 						.saddr = iph->saddr,
832 						.tos = RT_TOS(iph->tos) } },
833 				    .oif = tunnel->parms.link,
834 				    .proto = IPPROTO_IPV6 };
835 		struct rtable *rt;
836 		if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
837 			tdev = rt->u.dst.dev;
838 			ip_rt_put(rt);
839 		}
840 		dev->flags |= IFF_POINTOPOINT;
841 	}
842 
843 	if (!tdev && tunnel->parms.link)
844 		tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
845 
846 	if (tdev) {
847 		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
848 		dev->mtu = tdev->mtu - sizeof(struct iphdr);
849 		if (dev->mtu < IPV6_MIN_MTU)
850 			dev->mtu = IPV6_MIN_MTU;
851 	}
852 	dev->iflink = tunnel->parms.link;
853 }
854 
855 static int
856 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
857 {
858 	int err = 0;
859 	struct ip_tunnel_parm p;
860 	struct ip_tunnel_prl prl;
861 	struct ip_tunnel *t;
862 	struct net *net = dev_net(dev);
863 	struct sit_net *sitn = net_generic(net, sit_net_id);
864 #ifdef CONFIG_IPV6_SIT_6RD
865 	struct ip_tunnel_6rd ip6rd;
866 #endif
867 
868 	switch (cmd) {
869 	case SIOCGETTUNNEL:
870 #ifdef CONFIG_IPV6_SIT_6RD
871 	case SIOCGET6RD:
872 #endif
873 		t = NULL;
874 		if (dev == sitn->fb_tunnel_dev) {
875 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
876 				err = -EFAULT;
877 				break;
878 			}
879 			t = ipip6_tunnel_locate(net, &p, 0);
880 		}
881 		if (t == NULL)
882 			t = netdev_priv(dev);
883 
884 		err = -EFAULT;
885 		if (cmd == SIOCGETTUNNEL) {
886 			memcpy(&p, &t->parms, sizeof(p));
887 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
888 					 sizeof(p)))
889 				goto done;
890 #ifdef CONFIG_IPV6_SIT_6RD
891 		} else {
892 			ipv6_addr_copy(&ip6rd.prefix, &t->ip6rd.prefix);
893 			ip6rd.relay_prefix = t->ip6rd.relay_prefix;
894 			ip6rd.prefixlen = t->ip6rd.prefixlen;
895 			ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
896 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
897 					 sizeof(ip6rd)))
898 				goto done;
899 #endif
900 		}
901 		err = 0;
902 		break;
903 
904 	case SIOCADDTUNNEL:
905 	case SIOCCHGTUNNEL:
906 		err = -EPERM;
907 		if (!capable(CAP_NET_ADMIN))
908 			goto done;
909 
910 		err = -EFAULT;
911 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
912 			goto done;
913 
914 		err = -EINVAL;
915 		if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
916 		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
917 			goto done;
918 		if (p.iph.ttl)
919 			p.iph.frag_off |= htons(IP_DF);
920 
921 		t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
922 
923 		if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
924 			if (t != NULL) {
925 				if (t->dev != dev) {
926 					err = -EEXIST;
927 					break;
928 				}
929 			} else {
930 				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
931 				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
932 					err = -EINVAL;
933 					break;
934 				}
935 				t = netdev_priv(dev);
936 				ipip6_tunnel_unlink(sitn, t);
937 				t->parms.iph.saddr = p.iph.saddr;
938 				t->parms.iph.daddr = p.iph.daddr;
939 				memcpy(dev->dev_addr, &p.iph.saddr, 4);
940 				memcpy(dev->broadcast, &p.iph.daddr, 4);
941 				ipip6_tunnel_link(sitn, t);
942 				netdev_state_change(dev);
943 			}
944 		}
945 
946 		if (t) {
947 			err = 0;
948 			if (cmd == SIOCCHGTUNNEL) {
949 				t->parms.iph.ttl = p.iph.ttl;
950 				t->parms.iph.tos = p.iph.tos;
951 				if (t->parms.link != p.link) {
952 					t->parms.link = p.link;
953 					ipip6_tunnel_bind_dev(dev);
954 					netdev_state_change(dev);
955 				}
956 			}
957 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
958 				err = -EFAULT;
959 		} else
960 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
961 		break;
962 
963 	case SIOCDELTUNNEL:
964 		err = -EPERM;
965 		if (!capable(CAP_NET_ADMIN))
966 			goto done;
967 
968 		if (dev == sitn->fb_tunnel_dev) {
969 			err = -EFAULT;
970 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
971 				goto done;
972 			err = -ENOENT;
973 			if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
974 				goto done;
975 			err = -EPERM;
976 			if (t == netdev_priv(sitn->fb_tunnel_dev))
977 				goto done;
978 			dev = t->dev;
979 		}
980 		unregister_netdevice(dev);
981 		err = 0;
982 		break;
983 
984 	case SIOCGETPRL:
985 		err = -EINVAL;
986 		if (dev == sitn->fb_tunnel_dev)
987 			goto done;
988 		err = -ENOENT;
989 		if (!(t = netdev_priv(dev)))
990 			goto done;
991 		err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
992 		break;
993 
994 	case SIOCADDPRL:
995 	case SIOCDELPRL:
996 	case SIOCCHGPRL:
997 		err = -EPERM;
998 		if (!capable(CAP_NET_ADMIN))
999 			goto done;
1000 		err = -EINVAL;
1001 		if (dev == sitn->fb_tunnel_dev)
1002 			goto done;
1003 		err = -EFAULT;
1004 		if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1005 			goto done;
1006 		err = -ENOENT;
1007 		if (!(t = netdev_priv(dev)))
1008 			goto done;
1009 
1010 		switch (cmd) {
1011 		case SIOCDELPRL:
1012 			err = ipip6_tunnel_del_prl(t, &prl);
1013 			break;
1014 		case SIOCADDPRL:
1015 		case SIOCCHGPRL:
1016 			err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1017 			break;
1018 		}
1019 		netdev_state_change(dev);
1020 		break;
1021 
1022 #ifdef CONFIG_IPV6_SIT_6RD
1023 	case SIOCADD6RD:
1024 	case SIOCCHG6RD:
1025 	case SIOCDEL6RD:
1026 		err = -EPERM;
1027 		if (!capable(CAP_NET_ADMIN))
1028 			goto done;
1029 
1030 		err = -EFAULT;
1031 		if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1032 				   sizeof(ip6rd)))
1033 			goto done;
1034 
1035 		t = netdev_priv(dev);
1036 
1037 		if (cmd != SIOCDEL6RD) {
1038 			struct in6_addr prefix;
1039 			__be32 relay_prefix;
1040 
1041 			err = -EINVAL;
1042 			if (ip6rd.relay_prefixlen > 32 ||
1043 			    ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1044 				goto done;
1045 
1046 			ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1047 					 ip6rd.prefixlen);
1048 			if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1049 				goto done;
1050 			if (ip6rd.relay_prefixlen)
1051 				relay_prefix = ip6rd.relay_prefix &
1052 					       htonl(0xffffffffUL <<
1053 						     (32 - ip6rd.relay_prefixlen));
1054 			else
1055 				relay_prefix = 0;
1056 			if (relay_prefix != ip6rd.relay_prefix)
1057 				goto done;
1058 
1059 			ipv6_addr_copy(&t->ip6rd.prefix, &prefix);
1060 			t->ip6rd.relay_prefix = relay_prefix;
1061 			t->ip6rd.prefixlen = ip6rd.prefixlen;
1062 			t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1063 		} else
1064 			ipip6_tunnel_clone_6rd(dev, sitn);
1065 
1066 		err = 0;
1067 		break;
1068 #endif
1069 
1070 	default:
1071 		err = -EINVAL;
1072 	}
1073 
1074 done:
1075 	return err;
1076 }
1077 
1078 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1079 {
1080 	if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1081 		return -EINVAL;
1082 	dev->mtu = new_mtu;
1083 	return 0;
1084 }
1085 
1086 static const struct net_device_ops ipip6_netdev_ops = {
1087 	.ndo_uninit	= ipip6_tunnel_uninit,
1088 	.ndo_start_xmit	= ipip6_tunnel_xmit,
1089 	.ndo_do_ioctl	= ipip6_tunnel_ioctl,
1090 	.ndo_change_mtu	= ipip6_tunnel_change_mtu,
1091 };
1092 
1093 static void ipip6_tunnel_setup(struct net_device *dev)
1094 {
1095 	dev->netdev_ops		= &ipip6_netdev_ops;
1096 	dev->destructor 	= free_netdev;
1097 
1098 	dev->type		= ARPHRD_SIT;
1099 	dev->hard_header_len 	= LL_MAX_HEADER + sizeof(struct iphdr);
1100 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr);
1101 	dev->flags		= IFF_NOARP;
1102 	dev->priv_flags	       &= ~IFF_XMIT_DST_RELEASE;
1103 	dev->iflink		= 0;
1104 	dev->addr_len		= 4;
1105 	dev->features		|= NETIF_F_NETNS_LOCAL;
1106 }
1107 
1108 static void ipip6_tunnel_init(struct net_device *dev)
1109 {
1110 	struct ip_tunnel *tunnel = netdev_priv(dev);
1111 
1112 	tunnel->dev = dev;
1113 	strcpy(tunnel->parms.name, dev->name);
1114 
1115 	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1116 	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1117 
1118 	ipip6_tunnel_bind_dev(dev);
1119 }
1120 
1121 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1122 {
1123 	struct ip_tunnel *tunnel = netdev_priv(dev);
1124 	struct iphdr *iph = &tunnel->parms.iph;
1125 	struct net *net = dev_net(dev);
1126 	struct sit_net *sitn = net_generic(net, sit_net_id);
1127 
1128 	tunnel->dev = dev;
1129 	strcpy(tunnel->parms.name, dev->name);
1130 
1131 	iph->version		= 4;
1132 	iph->protocol		= IPPROTO_IPV6;
1133 	iph->ihl		= 5;
1134 	iph->ttl		= 64;
1135 
1136 	dev_hold(dev);
1137 	sitn->tunnels_wc[0]	= tunnel;
1138 }
1139 
1140 static struct xfrm_tunnel sit_handler = {
1141 	.handler	=	ipip6_rcv,
1142 	.err_handler	=	ipip6_err,
1143 	.priority	=	1,
1144 };
1145 
1146 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1147 {
1148 	int prio;
1149 
1150 	for (prio = 1; prio < 4; prio++) {
1151 		int h;
1152 		for (h = 0; h < HASH_SIZE; h++) {
1153 			struct ip_tunnel *t = sitn->tunnels[prio][h];
1154 
1155 			while (t != NULL) {
1156 				unregister_netdevice_queue(t->dev, head);
1157 				t = t->next;
1158 			}
1159 		}
1160 	}
1161 }
1162 
1163 static int __net_init sit_init_net(struct net *net)
1164 {
1165 	struct sit_net *sitn = net_generic(net, sit_net_id);
1166 	int err;
1167 
1168 	sitn->tunnels[0] = sitn->tunnels_wc;
1169 	sitn->tunnels[1] = sitn->tunnels_l;
1170 	sitn->tunnels[2] = sitn->tunnels_r;
1171 	sitn->tunnels[3] = sitn->tunnels_r_l;
1172 
1173 	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1174 					   ipip6_tunnel_setup);
1175 	if (!sitn->fb_tunnel_dev) {
1176 		err = -ENOMEM;
1177 		goto err_alloc_dev;
1178 	}
1179 	dev_net_set(sitn->fb_tunnel_dev, net);
1180 
1181 	ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1182 	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1183 
1184 	if ((err = register_netdev(sitn->fb_tunnel_dev)))
1185 		goto err_reg_dev;
1186 
1187 	return 0;
1188 
1189 err_reg_dev:
1190 	dev_put(sitn->fb_tunnel_dev);
1191 	free_netdev(sitn->fb_tunnel_dev);
1192 err_alloc_dev:
1193 	return err;
1194 }
1195 
1196 static void __net_exit sit_exit_net(struct net *net)
1197 {
1198 	struct sit_net *sitn = net_generic(net, sit_net_id);
1199 	LIST_HEAD(list);
1200 
1201 	rtnl_lock();
1202 	sit_destroy_tunnels(sitn, &list);
1203 	unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1204 	unregister_netdevice_many(&list);
1205 	rtnl_unlock();
1206 }
1207 
1208 static struct pernet_operations sit_net_ops = {
1209 	.init = sit_init_net,
1210 	.exit = sit_exit_net,
1211 	.id   = &sit_net_id,
1212 	.size = sizeof(struct sit_net),
1213 };
1214 
1215 static void __exit sit_cleanup(void)
1216 {
1217 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1218 
1219 	unregister_pernet_device(&sit_net_ops);
1220 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1221 }
1222 
1223 static int __init sit_init(void)
1224 {
1225 	int err;
1226 
1227 	printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1228 
1229 	err = register_pernet_device(&sit_net_ops);
1230 	if (err < 0)
1231 		return err;
1232 	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1233 	if (err < 0) {
1234 		unregister_pernet_device(&sit_net_ops);
1235 		printk(KERN_INFO "sit init: Can't add protocol\n");
1236 	}
1237 	return err;
1238 }
1239 
1240 module_init(sit_init);
1241 module_exit(sit_cleanup);
1242 MODULE_LICENSE("GPL");
1243 MODULE_ALIAS("sit0");
1244