xref: /openbmc/linux/net/ipv6/sit.c (revision 87c2ce3b)
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/config.h>
22 #include <linux/module.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 		memset(&(IPCB(skb)->opt), 0, sizeof(struct ip_options));
384 		IPCB(skb)->flags = 0;
385 		skb->protocol = htons(ETH_P_IPV6);
386 		skb->pkt_type = PACKET_HOST;
387 		tunnel->stat.rx_packets++;
388 		tunnel->stat.rx_bytes += skb->len;
389 		skb->dev = tunnel->dev;
390 		dst_release(skb->dst);
391 		skb->dst = NULL;
392 		nf_reset(skb);
393 		ipip6_ecn_decapsulate(iph, skb);
394 		netif_rx(skb);
395 		read_unlock(&ipip6_lock);
396 		return 0;
397 	}
398 
399 	icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PROT_UNREACH, 0);
400 	kfree_skb(skb);
401 	read_unlock(&ipip6_lock);
402 out:
403 	return 0;
404 }
405 
406 /* Returns the embedded IPv4 address if the IPv6 address
407    comes from 6to4 (RFC 3056) addr space */
408 
409 static inline u32 try_6to4(struct in6_addr *v6dst)
410 {
411 	u32 dst = 0;
412 
413 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
414 	        /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
415 		memcpy(&dst, &v6dst->s6_addr16[1], 4);
416 	}
417 	return dst;
418 }
419 
420 /*
421  *	This function assumes it is being called from dev_queue_xmit()
422  *	and that skb is filled properly by that function.
423  */
424 
425 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
426 {
427 	struct ip_tunnel *tunnel = netdev_priv(dev);
428 	struct net_device_stats *stats = &tunnel->stat;
429 	struct iphdr  *tiph = &tunnel->parms.iph;
430 	struct ipv6hdr *iph6 = skb->nh.ipv6h;
431 	u8     tos = tunnel->parms.iph.tos;
432 	struct rtable *rt;     			/* Route to the other host */
433 	struct net_device *tdev;			/* Device to other host */
434 	struct iphdr  *iph;			/* Our new IP header */
435 	int    max_headroom;			/* The extra header space needed */
436 	u32    dst = tiph->daddr;
437 	int    mtu;
438 	struct in6_addr *addr6;
439 	int addr_type;
440 
441 	if (tunnel->recursion++) {
442 		tunnel->stat.collisions++;
443 		goto tx_error;
444 	}
445 
446 	if (skb->protocol != htons(ETH_P_IPV6))
447 		goto tx_error;
448 
449 	if (!dst)
450 		dst = try_6to4(&iph6->daddr);
451 
452 	if (!dst) {
453 		struct neighbour *neigh = NULL;
454 
455 		if (skb->dst)
456 			neigh = skb->dst->neighbour;
457 
458 		if (neigh == NULL) {
459 			if (net_ratelimit())
460 				printk(KERN_DEBUG "sit: nexthop == NULL\n");
461 			goto tx_error;
462 		}
463 
464 		addr6 = (struct in6_addr*)&neigh->primary_key;
465 		addr_type = ipv6_addr_type(addr6);
466 
467 		if (addr_type == IPV6_ADDR_ANY) {
468 			addr6 = &skb->nh.ipv6h->daddr;
469 			addr_type = ipv6_addr_type(addr6);
470 		}
471 
472 		if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
473 			goto tx_error_icmp;
474 
475 		dst = addr6->s6_addr32[3];
476 	}
477 
478 	{
479 		struct flowi fl = { .nl_u = { .ip4_u =
480 					      { .daddr = dst,
481 						.saddr = tiph->saddr,
482 						.tos = RT_TOS(tos) } },
483 				    .oif = tunnel->parms.link,
484 				    .proto = IPPROTO_IPV6 };
485 		if (ip_route_output_key(&rt, &fl)) {
486 			tunnel->stat.tx_carrier_errors++;
487 			goto tx_error_icmp;
488 		}
489 	}
490 	if (rt->rt_type != RTN_UNICAST) {
491 		ip_rt_put(rt);
492 		tunnel->stat.tx_carrier_errors++;
493 		goto tx_error_icmp;
494 	}
495 	tdev = rt->u.dst.dev;
496 
497 	if (tdev == dev) {
498 		ip_rt_put(rt);
499 		tunnel->stat.collisions++;
500 		goto tx_error;
501 	}
502 
503 	if (tiph->frag_off)
504 		mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
505 	else
506 		mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
507 
508 	if (mtu < 68) {
509 		tunnel->stat.collisions++;
510 		ip_rt_put(rt);
511 		goto tx_error;
512 	}
513 	if (mtu < IPV6_MIN_MTU)
514 		mtu = IPV6_MIN_MTU;
515 	if (tunnel->parms.iph.daddr && skb->dst)
516 		skb->dst->ops->update_pmtu(skb->dst, mtu);
517 
518 	if (skb->len > mtu) {
519 		icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
520 		ip_rt_put(rt);
521 		goto tx_error;
522 	}
523 
524 	if (tunnel->err_count > 0) {
525 		if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
526 			tunnel->err_count--;
527 			dst_link_failure(skb);
528 		} else
529 			tunnel->err_count = 0;
530 	}
531 
532 	/*
533 	 * Okay, now see if we can stuff it in the buffer as-is.
534 	 */
535 	max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
536 
537 	if (skb_headroom(skb) < max_headroom || skb_cloned(skb) || skb_shared(skb)) {
538 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
539 		if (!new_skb) {
540 			ip_rt_put(rt);
541   			stats->tx_dropped++;
542 			dev_kfree_skb(skb);
543 			tunnel->recursion--;
544 			return 0;
545 		}
546 		if (skb->sk)
547 			skb_set_owner_w(new_skb, skb->sk);
548 		dev_kfree_skb(skb);
549 		skb = new_skb;
550 		iph6 = skb->nh.ipv6h;
551 	}
552 
553 	skb->h.raw = skb->nh.raw;
554 	skb->nh.raw = skb_push(skb, sizeof(struct iphdr));
555 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
556 	IPCB(skb)->flags = 0;
557 	dst_release(skb->dst);
558 	skb->dst = &rt->u.dst;
559 
560 	/*
561 	 *	Push down and install the IPIP header.
562 	 */
563 
564 	iph 			=	skb->nh.iph;
565 	iph->version		=	4;
566 	iph->ihl		=	sizeof(struct iphdr)>>2;
567 	if (mtu > IPV6_MIN_MTU)
568 		iph->frag_off	=	htons(IP_DF);
569 	else
570 		iph->frag_off	=	0;
571 
572 	iph->protocol		=	IPPROTO_IPV6;
573 	iph->tos		=	INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
574 	iph->daddr		=	rt->rt_dst;
575 	iph->saddr		=	rt->rt_src;
576 
577 	if ((iph->ttl = tiph->ttl) == 0)
578 		iph->ttl	=	iph6->hop_limit;
579 
580 	nf_reset(skb);
581 
582 	IPTUNNEL_XMIT();
583 	tunnel->recursion--;
584 	return 0;
585 
586 tx_error_icmp:
587 	dst_link_failure(skb);
588 tx_error:
589 	stats->tx_errors++;
590 	dev_kfree_skb(skb);
591 	tunnel->recursion--;
592 	return 0;
593 }
594 
595 static int
596 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
597 {
598 	int err = 0;
599 	struct ip_tunnel_parm p;
600 	struct ip_tunnel *t;
601 
602 	switch (cmd) {
603 	case SIOCGETTUNNEL:
604 		t = NULL;
605 		if (dev == ipip6_fb_tunnel_dev) {
606 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
607 				err = -EFAULT;
608 				break;
609 			}
610 			t = ipip6_tunnel_locate(&p, 0);
611 		}
612 		if (t == NULL)
613 			t = netdev_priv(dev);
614 		memcpy(&p, &t->parms, sizeof(p));
615 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
616 			err = -EFAULT;
617 		break;
618 
619 	case SIOCADDTUNNEL:
620 	case SIOCCHGTUNNEL:
621 		err = -EPERM;
622 		if (!capable(CAP_NET_ADMIN))
623 			goto done;
624 
625 		err = -EFAULT;
626 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
627 			goto done;
628 
629 		err = -EINVAL;
630 		if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
631 		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
632 			goto done;
633 		if (p.iph.ttl)
634 			p.iph.frag_off |= htons(IP_DF);
635 
636 		t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
637 
638 		if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
639 			if (t != NULL) {
640 				if (t->dev != dev) {
641 					err = -EEXIST;
642 					break;
643 				}
644 			} else {
645 				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
646 				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
647 					err = -EINVAL;
648 					break;
649 				}
650 				t = netdev_priv(dev);
651 				ipip6_tunnel_unlink(t);
652 				t->parms.iph.saddr = p.iph.saddr;
653 				t->parms.iph.daddr = p.iph.daddr;
654 				memcpy(dev->dev_addr, &p.iph.saddr, 4);
655 				memcpy(dev->broadcast, &p.iph.daddr, 4);
656 				ipip6_tunnel_link(t);
657 				netdev_state_change(dev);
658 			}
659 		}
660 
661 		if (t) {
662 			err = 0;
663 			if (cmd == SIOCCHGTUNNEL) {
664 				t->parms.iph.ttl = p.iph.ttl;
665 				t->parms.iph.tos = p.iph.tos;
666 			}
667 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
668 				err = -EFAULT;
669 		} else
670 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
671 		break;
672 
673 	case SIOCDELTUNNEL:
674 		err = -EPERM;
675 		if (!capable(CAP_NET_ADMIN))
676 			goto done;
677 
678 		if (dev == ipip6_fb_tunnel_dev) {
679 			err = -EFAULT;
680 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
681 				goto done;
682 			err = -ENOENT;
683 			if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
684 				goto done;
685 			err = -EPERM;
686 			if (t == netdev_priv(ipip6_fb_tunnel_dev))
687 				goto done;
688 			dev = t->dev;
689 		}
690 		err = unregister_netdevice(dev);
691 		break;
692 
693 	default:
694 		err = -EINVAL;
695 	}
696 
697 done:
698 	return err;
699 }
700 
701 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
702 {
703 	return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
704 }
705 
706 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
707 {
708 	if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
709 		return -EINVAL;
710 	dev->mtu = new_mtu;
711 	return 0;
712 }
713 
714 static void ipip6_tunnel_setup(struct net_device *dev)
715 {
716 	SET_MODULE_OWNER(dev);
717 	dev->uninit		= ipip6_tunnel_uninit;
718 	dev->destructor 	= free_netdev;
719 	dev->hard_start_xmit	= ipip6_tunnel_xmit;
720 	dev->get_stats		= ipip6_tunnel_get_stats;
721 	dev->do_ioctl		= ipip6_tunnel_ioctl;
722 	dev->change_mtu		= ipip6_tunnel_change_mtu;
723 
724 	dev->type		= ARPHRD_SIT;
725 	dev->hard_header_len 	= LL_MAX_HEADER + sizeof(struct iphdr);
726 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr);
727 	dev->flags		= IFF_NOARP;
728 	dev->iflink		= 0;
729 	dev->addr_len		= 4;
730 }
731 
732 static int ipip6_tunnel_init(struct net_device *dev)
733 {
734 	struct net_device *tdev = NULL;
735 	struct ip_tunnel *tunnel;
736 	struct iphdr *iph;
737 
738 	tunnel = netdev_priv(dev);
739 	iph = &tunnel->parms.iph;
740 
741 	tunnel->dev = dev;
742 	strcpy(tunnel->parms.name, dev->name);
743 
744 	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
745 	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
746 
747 	if (iph->daddr) {
748 		struct flowi fl = { .nl_u = { .ip4_u =
749 					      { .daddr = iph->daddr,
750 						.saddr = iph->saddr,
751 						.tos = RT_TOS(iph->tos) } },
752 				    .oif = tunnel->parms.link,
753 				    .proto = IPPROTO_IPV6 };
754 		struct rtable *rt;
755 		if (!ip_route_output_key(&rt, &fl)) {
756 			tdev = rt->u.dst.dev;
757 			ip_rt_put(rt);
758 		}
759 		dev->flags |= IFF_POINTOPOINT;
760 	}
761 
762 	if (!tdev && tunnel->parms.link)
763 		tdev = __dev_get_by_index(tunnel->parms.link);
764 
765 	if (tdev) {
766 		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
767 		dev->mtu = tdev->mtu - sizeof(struct iphdr);
768 		if (dev->mtu < IPV6_MIN_MTU)
769 			dev->mtu = IPV6_MIN_MTU;
770 	}
771 	dev->iflink = tunnel->parms.link;
772 
773 	return 0;
774 }
775 
776 static int __init ipip6_fb_tunnel_init(struct net_device *dev)
777 {
778 	struct ip_tunnel *tunnel = netdev_priv(dev);
779 	struct iphdr *iph = &tunnel->parms.iph;
780 
781 	tunnel->dev = dev;
782 	strcpy(tunnel->parms.name, dev->name);
783 
784 	iph->version		= 4;
785 	iph->protocol		= IPPROTO_IPV6;
786 	iph->ihl		= 5;
787 	iph->ttl		= 64;
788 
789 	dev_hold(dev);
790 	tunnels_wc[0]		= tunnel;
791 	return 0;
792 }
793 
794 static struct net_protocol sit_protocol = {
795 	.handler	=	ipip6_rcv,
796 	.err_handler	=	ipip6_err,
797 };
798 
799 static void __exit sit_destroy_tunnels(void)
800 {
801 	int prio;
802 
803 	for (prio = 1; prio < 4; prio++) {
804 		int h;
805 		for (h = 0; h < HASH_SIZE; h++) {
806 			struct ip_tunnel *t;
807 			while ((t = tunnels[prio][h]) != NULL)
808 				unregister_netdevice(t->dev);
809 		}
810 	}
811 }
812 
813 void __exit sit_cleanup(void)
814 {
815 	inet_del_protocol(&sit_protocol, IPPROTO_IPV6);
816 
817 	rtnl_lock();
818 	sit_destroy_tunnels();
819 	unregister_netdevice(ipip6_fb_tunnel_dev);
820 	rtnl_unlock();
821 }
822 
823 int __init sit_init(void)
824 {
825 	int err;
826 
827 	printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
828 
829 	if (inet_add_protocol(&sit_protocol, IPPROTO_IPV6) < 0) {
830 		printk(KERN_INFO "sit init: Can't add protocol\n");
831 		return -EAGAIN;
832 	}
833 
834 	ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
835 					   ipip6_tunnel_setup);
836 	if (!ipip6_fb_tunnel_dev) {
837 		err = -ENOMEM;
838 		goto err1;
839 	}
840 
841 	ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
842 
843 	if ((err =  register_netdev(ipip6_fb_tunnel_dev)))
844 		goto err2;
845 
846  out:
847 	return err;
848  err2:
849 	free_netdev(ipip6_fb_tunnel_dev);
850  err1:
851 	inet_del_protocol(&sit_protocol, IPPROTO_IPV6);
852 	goto out;
853 }
854