xref: /linux/net/ipv6/sit.c (revision 6e8331ac6973435b1e7604c30f2ad394035b46e1)
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  *	$Id: sit.c,v 1.53 2001/09/25 05:09:53 davem Exp $
10  *
11  *	This program is free software; you can redistribute it and/or
12  *      modify it under the terms of the GNU General Public License
13  *      as published by the Free Software Foundation; either version
14  *      2 of the License, or (at your option) any later version.
15  *
16  *	Changes:
17  * Roger Venning <r.venning@telstra.com>:	6to4 support
18  * Nate Thompson <nate@thebog.net>:		6to4 support
19  */
20 
21 #include <linux/module.h>
22 #include <linux/capability.h>
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/socket.h>
26 #include <linux/sockios.h>
27 #include <linux/sched.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 <asm/uaccess.h>
34 #include <linux/init.h>
35 #include <linux/netfilter_ipv4.h>
36 #include <linux/if_ether.h>
37 
38 #include <net/sock.h>
39 #include <net/snmp.h>
40 
41 #include <net/ipv6.h>
42 #include <net/protocol.h>
43 #include <net/transp_v6.h>
44 #include <net/ip6_fib.h>
45 #include <net/ip6_route.h>
46 #include <net/ndisc.h>
47 #include <net/addrconf.h>
48 #include <net/ip.h>
49 #include <net/udp.h>
50 #include <net/icmp.h>
51 #include <net/ipip.h>
52 #include <net/inet_ecn.h>
53 #include <net/xfrm.h>
54 #include <net/dsfield.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) ((addr^(addr>>4))&0xF)
64 
65 static int ipip6_fb_tunnel_init(struct net_device *dev);
66 static int ipip6_tunnel_init(struct net_device *dev);
67 static void ipip6_tunnel_setup(struct net_device *dev);
68 
69 static struct net_device *ipip6_fb_tunnel_dev;
70 
71 static struct ip_tunnel *tunnels_r_l[HASH_SIZE];
72 static struct ip_tunnel *tunnels_r[HASH_SIZE];
73 static struct ip_tunnel *tunnels_l[HASH_SIZE];
74 static struct ip_tunnel *tunnels_wc[1];
75 static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l };
76 
77 static DEFINE_RWLOCK(ipip6_lock);
78 
79 static struct ip_tunnel * ipip6_tunnel_lookup(u32 remote, u32 local)
80 {
81 	unsigned h0 = HASH(remote);
82 	unsigned h1 = HASH(local);
83 	struct ip_tunnel *t;
84 
85 	for (t = tunnels_r_l[h0^h1]; t; t = t->next) {
86 		if (local == t->parms.iph.saddr &&
87 		    remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
88 			return t;
89 	}
90 	for (t = tunnels_r[h0]; t; t = t->next) {
91 		if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
92 			return t;
93 	}
94 	for (t = tunnels_l[h1]; t; t = t->next) {
95 		if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
96 			return t;
97 	}
98 	if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
99 		return t;
100 	return NULL;
101 }
102 
103 static struct ip_tunnel ** ipip6_bucket(struct ip_tunnel *t)
104 {
105 	u32 remote = t->parms.iph.daddr;
106 	u32 local = t->parms.iph.saddr;
107 	unsigned h = 0;
108 	int prio = 0;
109 
110 	if (remote) {
111 		prio |= 2;
112 		h ^= HASH(remote);
113 	}
114 	if (local) {
115 		prio |= 1;
116 		h ^= HASH(local);
117 	}
118 	return &tunnels[prio][h];
119 }
120 
121 static void ipip6_tunnel_unlink(struct ip_tunnel *t)
122 {
123 	struct ip_tunnel **tp;
124 
125 	for (tp = ipip6_bucket(t); *tp; tp = &(*tp)->next) {
126 		if (t == *tp) {
127 			write_lock_bh(&ipip6_lock);
128 			*tp = t->next;
129 			write_unlock_bh(&ipip6_lock);
130 			break;
131 		}
132 	}
133 }
134 
135 static void ipip6_tunnel_link(struct ip_tunnel *t)
136 {
137 	struct ip_tunnel **tp = ipip6_bucket(t);
138 
139 	t->next = *tp;
140 	write_lock_bh(&ipip6_lock);
141 	*tp = t;
142 	write_unlock_bh(&ipip6_lock);
143 }
144 
145 static struct ip_tunnel * ipip6_tunnel_locate(struct ip_tunnel_parm *parms, int create)
146 {
147 	u32 remote = parms->iph.daddr;
148 	u32 local = parms->iph.saddr;
149 	struct ip_tunnel *t, **tp, *nt;
150 	struct net_device *dev;
151 	unsigned h = 0;
152 	int prio = 0;
153 	char name[IFNAMSIZ];
154 
155 	if (remote) {
156 		prio |= 2;
157 		h ^= HASH(remote);
158 	}
159 	if (local) {
160 		prio |= 1;
161 		h ^= HASH(local);
162 	}
163 	for (tp = &tunnels[prio][h]; (t = *tp) != NULL; tp = &t->next) {
164 		if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
165 			return t;
166 	}
167 	if (!create)
168 		goto failed;
169 
170 	if (parms->name[0])
171 		strlcpy(name, parms->name, IFNAMSIZ);
172 	else {
173 		int i;
174 		for (i=1; i<100; i++) {
175 			sprintf(name, "sit%d", i);
176 			if (__dev_get_by_name(name) == NULL)
177 				break;
178 		}
179 		if (i==100)
180 			goto failed;
181 	}
182 
183 	dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
184 	if (dev == NULL)
185 		return NULL;
186 
187 	nt = netdev_priv(dev);
188 	dev->init = ipip6_tunnel_init;
189 	nt->parms = *parms;
190 
191 	if (register_netdevice(dev) < 0) {
192 		free_netdev(dev);
193 		goto failed;
194 	}
195 
196 	dev_hold(dev);
197 
198 	ipip6_tunnel_link(nt);
199 	return nt;
200 
201 failed:
202 	return NULL;
203 }
204 
205 static void ipip6_tunnel_uninit(struct net_device *dev)
206 {
207 	if (dev == ipip6_fb_tunnel_dev) {
208 		write_lock_bh(&ipip6_lock);
209 		tunnels_wc[0] = NULL;
210 		write_unlock_bh(&ipip6_lock);
211 		dev_put(dev);
212 	} else {
213 		ipip6_tunnel_unlink(netdev_priv(dev));
214 		dev_put(dev);
215 	}
216 }
217 
218 
219 static void ipip6_err(struct sk_buff *skb, u32 info)
220 {
221 #ifndef I_WISH_WORLD_WERE_PERFECT
222 
223 /* It is not :-( All the routers (except for Linux) return only
224    8 bytes of packet payload. It means, that precise relaying of
225    ICMP in the real Internet is absolutely infeasible.
226  */
227 	struct iphdr *iph = (struct iphdr*)skb->data;
228 	int type = skb->h.icmph->type;
229 	int code = skb->h.icmph->code;
230 	struct ip_tunnel *t;
231 
232 	switch (type) {
233 	default:
234 	case ICMP_PARAMETERPROB:
235 		return;
236 
237 	case ICMP_DEST_UNREACH:
238 		switch (code) {
239 		case ICMP_SR_FAILED:
240 		case ICMP_PORT_UNREACH:
241 			/* Impossible event. */
242 			return;
243 		case ICMP_FRAG_NEEDED:
244 			/* Soft state for pmtu is maintained by IP core. */
245 			return;
246 		default:
247 			/* All others are translated to HOST_UNREACH.
248 			   rfc2003 contains "deep thoughts" about NET_UNREACH,
249 			   I believe they are just ether pollution. --ANK
250 			 */
251 			break;
252 		}
253 		break;
254 	case ICMP_TIME_EXCEEDED:
255 		if (code != ICMP_EXC_TTL)
256 			return;
257 		break;
258 	}
259 
260 	read_lock(&ipip6_lock);
261 	t = ipip6_tunnel_lookup(iph->daddr, iph->saddr);
262 	if (t == NULL || t->parms.iph.daddr == 0)
263 		goto out;
264 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
265 		goto out;
266 
267 	if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
268 		t->err_count++;
269 	else
270 		t->err_count = 1;
271 	t->err_time = jiffies;
272 out:
273 	read_unlock(&ipip6_lock);
274 	return;
275 #else
276 	struct iphdr *iph = (struct iphdr*)dp;
277 	int hlen = iph->ihl<<2;
278 	struct ipv6hdr *iph6;
279 	int type = skb->h.icmph->type;
280 	int code = skb->h.icmph->code;
281 	int rel_type = 0;
282 	int rel_code = 0;
283 	int rel_info = 0;
284 	struct sk_buff *skb2;
285 	struct rt6_info *rt6i;
286 
287 	if (len < hlen + sizeof(struct ipv6hdr))
288 		return;
289 	iph6 = (struct ipv6hdr*)(dp + hlen);
290 
291 	switch (type) {
292 	default:
293 		return;
294 	case ICMP_PARAMETERPROB:
295 		if (skb->h.icmph->un.gateway < hlen)
296 			return;
297 
298 		/* So... This guy found something strange INSIDE encapsulated
299 		   packet. Well, he is fool, but what can we do ?
300 		 */
301 		rel_type = ICMPV6_PARAMPROB;
302 		rel_info = skb->h.icmph->un.gateway - hlen;
303 		break;
304 
305 	case ICMP_DEST_UNREACH:
306 		switch (code) {
307 		case ICMP_SR_FAILED:
308 		case ICMP_PORT_UNREACH:
309 			/* Impossible event. */
310 			return;
311 		case ICMP_FRAG_NEEDED:
312 			/* Too complicated case ... */
313 			return;
314 		default:
315 			/* All others are translated to HOST_UNREACH.
316 			   rfc2003 contains "deep thoughts" about NET_UNREACH,
317 			   I believe, it is just ether pollution. --ANK
318 			 */
319 			rel_type = ICMPV6_DEST_UNREACH;
320 			rel_code = ICMPV6_ADDR_UNREACH;
321 			break;
322 		}
323 		break;
324 	case ICMP_TIME_EXCEEDED:
325 		if (code != ICMP_EXC_TTL)
326 			return;
327 		rel_type = ICMPV6_TIME_EXCEED;
328 		rel_code = ICMPV6_EXC_HOPLIMIT;
329 		break;
330 	}
331 
332 	/* Prepare fake skb to feed it to icmpv6_send */
333 	skb2 = skb_clone(skb, GFP_ATOMIC);
334 	if (skb2 == NULL)
335 		return;
336 	dst_release(skb2->dst);
337 	skb2->dst = NULL;
338 	skb_pull(skb2, skb->data - (u8*)iph6);
339 	skb2->nh.raw = skb2->data;
340 
341 	/* Try to guess incoming interface */
342 	rt6i = rt6_lookup(&iph6->saddr, NULL, NULL, 0);
343 	if (rt6i && rt6i->rt6i_dev) {
344 		skb2->dev = rt6i->rt6i_dev;
345 
346 		rt6i = rt6_lookup(&iph6->daddr, &iph6->saddr, NULL, 0);
347 
348 		if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
349 			struct ip_tunnel *t = netdev_priv(rt6i->rt6i_dev);
350 			if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
351 				rel_type = ICMPV6_DEST_UNREACH;
352 				rel_code = ICMPV6_ADDR_UNREACH;
353 			}
354 			icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
355 		}
356 	}
357 	kfree_skb(skb2);
358 	return;
359 #endif
360 }
361 
362 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
363 {
364 	if (INET_ECN_is_ce(iph->tos))
365 		IP6_ECN_set_ce(skb->nh.ipv6h);
366 }
367 
368 static int ipip6_rcv(struct sk_buff *skb)
369 {
370 	struct iphdr *iph;
371 	struct ip_tunnel *tunnel;
372 
373 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
374 		goto out;
375 
376 	iph = skb->nh.iph;
377 
378 	read_lock(&ipip6_lock);
379 	if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
380 		secpath_reset(skb);
381 		skb->mac.raw = skb->nh.raw;
382 		skb->nh.raw = skb->data;
383 		IPCB(skb)->flags = 0;
384 		skb->protocol = htons(ETH_P_IPV6);
385 		skb->pkt_type = PACKET_HOST;
386 		tunnel->stat.rx_packets++;
387 		tunnel->stat.rx_bytes += skb->len;
388 		skb->dev = tunnel->dev;
389 		dst_release(skb->dst);
390 		skb->dst = NULL;
391 		nf_reset(skb);
392 		ipip6_ecn_decapsulate(iph, skb);
393 		netif_rx(skb);
394 		read_unlock(&ipip6_lock);
395 		return 0;
396 	}
397 
398 	icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
399 	kfree_skb(skb);
400 	read_unlock(&ipip6_lock);
401 out:
402 	return 0;
403 }
404 
405 /* Returns the embedded IPv4 address if the IPv6 address
406    comes from 6to4 (RFC 3056) addr space */
407 
408 static inline u32 try_6to4(struct in6_addr *v6dst)
409 {
410 	u32 dst = 0;
411 
412 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
413 	        /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
414 		memcpy(&dst, &v6dst->s6_addr16[1], 4);
415 	}
416 	return dst;
417 }
418 
419 /*
420  *	This function assumes it is being called from dev_queue_xmit()
421  *	and that skb is filled properly by that function.
422  */
423 
424 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
425 {
426 	struct ip_tunnel *tunnel = netdev_priv(dev);
427 	struct net_device_stats *stats = &tunnel->stat;
428 	struct iphdr  *tiph = &tunnel->parms.iph;
429 	struct ipv6hdr *iph6 = skb->nh.ipv6h;
430 	u8     tos = tunnel->parms.iph.tos;
431 	struct rtable *rt;     			/* Route to the other host */
432 	struct net_device *tdev;			/* Device to other host */
433 	struct iphdr  *iph;			/* Our new IP header */
434 	int    max_headroom;			/* The extra header space needed */
435 	u32    dst = tiph->daddr;
436 	int    mtu;
437 	struct in6_addr *addr6;
438 	int addr_type;
439 
440 	if (tunnel->recursion++) {
441 		tunnel->stat.collisions++;
442 		goto tx_error;
443 	}
444 
445 	if (skb->protocol != htons(ETH_P_IPV6))
446 		goto tx_error;
447 
448 	if (!dst)
449 		dst = try_6to4(&iph6->daddr);
450 
451 	if (!dst) {
452 		struct neighbour *neigh = NULL;
453 
454 		if (skb->dst)
455 			neigh = skb->dst->neighbour;
456 
457 		if (neigh == NULL) {
458 			if (net_ratelimit())
459 				printk(KERN_DEBUG "sit: nexthop == NULL\n");
460 			goto tx_error;
461 		}
462 
463 		addr6 = (struct in6_addr*)&neigh->primary_key;
464 		addr_type = ipv6_addr_type(addr6);
465 
466 		if (addr_type == IPV6_ADDR_ANY) {
467 			addr6 = &skb->nh.ipv6h->daddr;
468 			addr_type = ipv6_addr_type(addr6);
469 		}
470 
471 		if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
472 			goto tx_error_icmp;
473 
474 		dst = addr6->s6_addr32[3];
475 	}
476 
477 	{
478 		struct flowi fl = { .nl_u = { .ip4_u =
479 					      { .daddr = dst,
480 						.saddr = tiph->saddr,
481 						.tos = RT_TOS(tos) } },
482 				    .oif = tunnel->parms.link,
483 				    .proto = IPPROTO_IPV6 };
484 		if (ip_route_output_key(&rt, &fl)) {
485 			tunnel->stat.tx_carrier_errors++;
486 			goto tx_error_icmp;
487 		}
488 	}
489 	if (rt->rt_type != RTN_UNICAST) {
490 		ip_rt_put(rt);
491 		tunnel->stat.tx_carrier_errors++;
492 		goto tx_error_icmp;
493 	}
494 	tdev = rt->u.dst.dev;
495 
496 	if (tdev == dev) {
497 		ip_rt_put(rt);
498 		tunnel->stat.collisions++;
499 		goto tx_error;
500 	}
501 
502 	if (tiph->frag_off)
503 		mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
504 	else
505 		mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
506 
507 	if (mtu < 68) {
508 		tunnel->stat.collisions++;
509 		ip_rt_put(rt);
510 		goto tx_error;
511 	}
512 	if (mtu < IPV6_MIN_MTU)
513 		mtu = IPV6_MIN_MTU;
514 	if (tunnel->parms.iph.daddr && skb->dst)
515 		skb->dst->ops->update_pmtu(skb->dst, mtu);
516 
517 	if (skb->len > mtu) {
518 		icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
519 		ip_rt_put(rt);
520 		goto tx_error;
521 	}
522 
523 	if (tunnel->err_count > 0) {
524 		if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
525 			tunnel->err_count--;
526 			dst_link_failure(skb);
527 		} else
528 			tunnel->err_count = 0;
529 	}
530 
531 	/*
532 	 * Okay, now see if we can stuff it in the buffer as-is.
533 	 */
534 	max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
535 
536 	if (skb_headroom(skb) < max_headroom || skb_cloned(skb) || skb_shared(skb)) {
537 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
538 		if (!new_skb) {
539 			ip_rt_put(rt);
540   			stats->tx_dropped++;
541 			dev_kfree_skb(skb);
542 			tunnel->recursion--;
543 			return 0;
544 		}
545 		if (skb->sk)
546 			skb_set_owner_w(new_skb, skb->sk);
547 		dev_kfree_skb(skb);
548 		skb = new_skb;
549 		iph6 = skb->nh.ipv6h;
550 	}
551 
552 	skb->h.raw = skb->nh.raw;
553 	skb->nh.raw = skb_push(skb, sizeof(struct iphdr));
554 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
555 	IPCB(skb)->flags = 0;
556 	dst_release(skb->dst);
557 	skb->dst = &rt->u.dst;
558 
559 	/*
560 	 *	Push down and install the IPIP header.
561 	 */
562 
563 	iph 			=	skb->nh.iph;
564 	iph->version		=	4;
565 	iph->ihl		=	sizeof(struct iphdr)>>2;
566 	if (mtu > IPV6_MIN_MTU)
567 		iph->frag_off	=	htons(IP_DF);
568 	else
569 		iph->frag_off	=	0;
570 
571 	iph->protocol		=	IPPROTO_IPV6;
572 	iph->tos		=	INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
573 	iph->daddr		=	rt->rt_dst;
574 	iph->saddr		=	rt->rt_src;
575 
576 	if ((iph->ttl = tiph->ttl) == 0)
577 		iph->ttl	=	iph6->hop_limit;
578 
579 	nf_reset(skb);
580 
581 	IPTUNNEL_XMIT();
582 	tunnel->recursion--;
583 	return 0;
584 
585 tx_error_icmp:
586 	dst_link_failure(skb);
587 tx_error:
588 	stats->tx_errors++;
589 	dev_kfree_skb(skb);
590 	tunnel->recursion--;
591 	return 0;
592 }
593 
594 static int
595 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
596 {
597 	int err = 0;
598 	struct ip_tunnel_parm p;
599 	struct ip_tunnel *t;
600 
601 	switch (cmd) {
602 	case SIOCGETTUNNEL:
603 		t = NULL;
604 		if (dev == ipip6_fb_tunnel_dev) {
605 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
606 				err = -EFAULT;
607 				break;
608 			}
609 			t = ipip6_tunnel_locate(&p, 0);
610 		}
611 		if (t == NULL)
612 			t = netdev_priv(dev);
613 		memcpy(&p, &t->parms, sizeof(p));
614 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
615 			err = -EFAULT;
616 		break;
617 
618 	case SIOCADDTUNNEL:
619 	case SIOCCHGTUNNEL:
620 		err = -EPERM;
621 		if (!capable(CAP_NET_ADMIN))
622 			goto done;
623 
624 		err = -EFAULT;
625 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
626 			goto done;
627 
628 		err = -EINVAL;
629 		if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
630 		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
631 			goto done;
632 		if (p.iph.ttl)
633 			p.iph.frag_off |= htons(IP_DF);
634 
635 		t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
636 
637 		if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
638 			if (t != NULL) {
639 				if (t->dev != dev) {
640 					err = -EEXIST;
641 					break;
642 				}
643 			} else {
644 				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
645 				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
646 					err = -EINVAL;
647 					break;
648 				}
649 				t = netdev_priv(dev);
650 				ipip6_tunnel_unlink(t);
651 				t->parms.iph.saddr = p.iph.saddr;
652 				t->parms.iph.daddr = p.iph.daddr;
653 				memcpy(dev->dev_addr, &p.iph.saddr, 4);
654 				memcpy(dev->broadcast, &p.iph.daddr, 4);
655 				ipip6_tunnel_link(t);
656 				netdev_state_change(dev);
657 			}
658 		}
659 
660 		if (t) {
661 			err = 0;
662 			if (cmd == SIOCCHGTUNNEL) {
663 				t->parms.iph.ttl = p.iph.ttl;
664 				t->parms.iph.tos = p.iph.tos;
665 			}
666 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
667 				err = -EFAULT;
668 		} else
669 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
670 		break;
671 
672 	case SIOCDELTUNNEL:
673 		err = -EPERM;
674 		if (!capable(CAP_NET_ADMIN))
675 			goto done;
676 
677 		if (dev == ipip6_fb_tunnel_dev) {
678 			err = -EFAULT;
679 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
680 				goto done;
681 			err = -ENOENT;
682 			if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
683 				goto done;
684 			err = -EPERM;
685 			if (t == netdev_priv(ipip6_fb_tunnel_dev))
686 				goto done;
687 			dev = t->dev;
688 		}
689 		err = unregister_netdevice(dev);
690 		break;
691 
692 	default:
693 		err = -EINVAL;
694 	}
695 
696 done:
697 	return err;
698 }
699 
700 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
701 {
702 	return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
703 }
704 
705 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
706 {
707 	if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
708 		return -EINVAL;
709 	dev->mtu = new_mtu;
710 	return 0;
711 }
712 
713 static void ipip6_tunnel_setup(struct net_device *dev)
714 {
715 	SET_MODULE_OWNER(dev);
716 	dev->uninit		= ipip6_tunnel_uninit;
717 	dev->destructor 	= free_netdev;
718 	dev->hard_start_xmit	= ipip6_tunnel_xmit;
719 	dev->get_stats		= ipip6_tunnel_get_stats;
720 	dev->do_ioctl		= ipip6_tunnel_ioctl;
721 	dev->change_mtu		= ipip6_tunnel_change_mtu;
722 
723 	dev->type		= ARPHRD_SIT;
724 	dev->hard_header_len 	= LL_MAX_HEADER + sizeof(struct iphdr);
725 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr);
726 	dev->flags		= IFF_NOARP;
727 	dev->iflink		= 0;
728 	dev->addr_len		= 4;
729 }
730 
731 static int ipip6_tunnel_init(struct net_device *dev)
732 {
733 	struct net_device *tdev = NULL;
734 	struct ip_tunnel *tunnel;
735 	struct iphdr *iph;
736 
737 	tunnel = netdev_priv(dev);
738 	iph = &tunnel->parms.iph;
739 
740 	tunnel->dev = dev;
741 	strcpy(tunnel->parms.name, dev->name);
742 
743 	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
744 	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
745 
746 	if (iph->daddr) {
747 		struct flowi fl = { .nl_u = { .ip4_u =
748 					      { .daddr = iph->daddr,
749 						.saddr = iph->saddr,
750 						.tos = RT_TOS(iph->tos) } },
751 				    .oif = tunnel->parms.link,
752 				    .proto = IPPROTO_IPV6 };
753 		struct rtable *rt;
754 		if (!ip_route_output_key(&rt, &fl)) {
755 			tdev = rt->u.dst.dev;
756 			ip_rt_put(rt);
757 		}
758 		dev->flags |= IFF_POINTOPOINT;
759 	}
760 
761 	if (!tdev && tunnel->parms.link)
762 		tdev = __dev_get_by_index(tunnel->parms.link);
763 
764 	if (tdev) {
765 		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
766 		dev->mtu = tdev->mtu - sizeof(struct iphdr);
767 		if (dev->mtu < IPV6_MIN_MTU)
768 			dev->mtu = IPV6_MIN_MTU;
769 	}
770 	dev->iflink = tunnel->parms.link;
771 
772 	return 0;
773 }
774 
775 static int __init ipip6_fb_tunnel_init(struct net_device *dev)
776 {
777 	struct ip_tunnel *tunnel = netdev_priv(dev);
778 	struct iphdr *iph = &tunnel->parms.iph;
779 
780 	tunnel->dev = dev;
781 	strcpy(tunnel->parms.name, dev->name);
782 
783 	iph->version		= 4;
784 	iph->protocol		= IPPROTO_IPV6;
785 	iph->ihl		= 5;
786 	iph->ttl		= 64;
787 
788 	dev_hold(dev);
789 	tunnels_wc[0]		= tunnel;
790 	return 0;
791 }
792 
793 static struct net_protocol sit_protocol = {
794 	.handler	=	ipip6_rcv,
795 	.err_handler	=	ipip6_err,
796 };
797 
798 static void __exit sit_destroy_tunnels(void)
799 {
800 	int prio;
801 
802 	for (prio = 1; prio < 4; prio++) {
803 		int h;
804 		for (h = 0; h < HASH_SIZE; h++) {
805 			struct ip_tunnel *t;
806 			while ((t = tunnels[prio][h]) != NULL)
807 				unregister_netdevice(t->dev);
808 		}
809 	}
810 }
811 
812 void __exit sit_cleanup(void)
813 {
814 	inet_del_protocol(&sit_protocol, IPPROTO_IPV6);
815 
816 	rtnl_lock();
817 	sit_destroy_tunnels();
818 	unregister_netdevice(ipip6_fb_tunnel_dev);
819 	rtnl_unlock();
820 }
821 
822 int __init sit_init(void)
823 {
824 	int err;
825 
826 	printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
827 
828 	if (inet_add_protocol(&sit_protocol, IPPROTO_IPV6) < 0) {
829 		printk(KERN_INFO "sit init: Can't add protocol\n");
830 		return -EAGAIN;
831 	}
832 
833 	ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
834 					   ipip6_tunnel_setup);
835 	if (!ipip6_fb_tunnel_dev) {
836 		err = -ENOMEM;
837 		goto err1;
838 	}
839 
840 	ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
841 
842 	if ((err =  register_netdev(ipip6_fb_tunnel_dev)))
843 		goto err2;
844 
845  out:
846 	return err;
847  err2:
848 	free_netdev(ipip6_fb_tunnel_dev);
849  err1:
850 	inet_del_protocol(&sit_protocol, IPPROTO_IPV6);
851 	goto out;
852 }
853