xref: /openbmc/linux/net/ipv6/sit.c (revision a1e58bbd)
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  * Fred L. Templin <fltemplin@acm.org>:		isatap support
20  */
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 <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) (((__force u32)addr^((__force u32)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(__be32 remote, __be32 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_parm *parms)
104 {
105 	__be32 remote = parms->iph.daddr;
106 	__be32 local = 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 inline struct ip_tunnel **ipip6_bucket(struct ip_tunnel *t)
122 {
123 	return __ipip6_bucket(&t->parms);
124 }
125 
126 static void ipip6_tunnel_unlink(struct ip_tunnel *t)
127 {
128 	struct ip_tunnel **tp;
129 
130 	for (tp = ipip6_bucket(t); *tp; tp = &(*tp)->next) {
131 		if (t == *tp) {
132 			write_lock_bh(&ipip6_lock);
133 			*tp = t->next;
134 			write_unlock_bh(&ipip6_lock);
135 			break;
136 		}
137 	}
138 }
139 
140 static void ipip6_tunnel_link(struct ip_tunnel *t)
141 {
142 	struct ip_tunnel **tp = ipip6_bucket(t);
143 
144 	t->next = *tp;
145 	write_lock_bh(&ipip6_lock);
146 	*tp = t;
147 	write_unlock_bh(&ipip6_lock);
148 }
149 
150 static struct ip_tunnel * ipip6_tunnel_locate(struct ip_tunnel_parm *parms, int create)
151 {
152 	__be32 remote = parms->iph.daddr;
153 	__be32 local = parms->iph.saddr;
154 	struct ip_tunnel *t, **tp, *nt;
155 	struct net_device *dev;
156 	char name[IFNAMSIZ];
157 
158 	for (tp = __ipip6_bucket(parms); (t = *tp) != NULL; tp = &t->next) {
159 		if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
160 			return t;
161 	}
162 	if (!create)
163 		goto failed;
164 
165 	if (parms->name[0])
166 		strlcpy(name, parms->name, IFNAMSIZ);
167 	else
168 		sprintf(name, "sit%%d");
169 
170 	dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
171 	if (dev == NULL)
172 		return NULL;
173 
174 	if (strchr(name, '%')) {
175 		if (dev_alloc_name(dev, name) < 0)
176 			goto failed_free;
177 	}
178 
179 	nt = netdev_priv(dev);
180 	dev->init = ipip6_tunnel_init;
181 	nt->parms = *parms;
182 
183 	if (parms->i_flags & SIT_ISATAP)
184 		dev->priv_flags |= IFF_ISATAP;
185 
186 	if (register_netdevice(dev) < 0)
187 		goto failed_free;
188 
189 	dev_hold(dev);
190 
191 	ipip6_tunnel_link(nt);
192 	return nt;
193 
194 failed_free:
195 	free_netdev(dev);
196 failed:
197 	return NULL;
198 }
199 
200 static void ipip6_tunnel_uninit(struct net_device *dev)
201 {
202 	if (dev == ipip6_fb_tunnel_dev) {
203 		write_lock_bh(&ipip6_lock);
204 		tunnels_wc[0] = NULL;
205 		write_unlock_bh(&ipip6_lock);
206 		dev_put(dev);
207 	} else {
208 		ipip6_tunnel_unlink(netdev_priv(dev));
209 		dev_put(dev);
210 	}
211 }
212 
213 
214 static int ipip6_err(struct sk_buff *skb, u32 info)
215 {
216 #ifndef I_WISH_WORLD_WERE_PERFECT
217 
218 /* It is not :-( All the routers (except for Linux) return only
219    8 bytes of packet payload. It means, that precise relaying of
220    ICMP in the real Internet is absolutely infeasible.
221  */
222 	struct iphdr *iph = (struct iphdr*)skb->data;
223 	const int type = icmp_hdr(skb)->type;
224 	const int code = icmp_hdr(skb)->code;
225 	struct ip_tunnel *t;
226 	int err;
227 
228 	switch (type) {
229 	default:
230 	case ICMP_PARAMETERPROB:
231 		return 0;
232 
233 	case ICMP_DEST_UNREACH:
234 		switch (code) {
235 		case ICMP_SR_FAILED:
236 		case ICMP_PORT_UNREACH:
237 			/* Impossible event. */
238 			return 0;
239 		case ICMP_FRAG_NEEDED:
240 			/* Soft state for pmtu is maintained by IP core. */
241 			return 0;
242 		default:
243 			/* All others are translated to HOST_UNREACH.
244 			   rfc2003 contains "deep thoughts" about NET_UNREACH,
245 			   I believe they are just ether pollution. --ANK
246 			 */
247 			break;
248 		}
249 		break;
250 	case ICMP_TIME_EXCEEDED:
251 		if (code != ICMP_EXC_TTL)
252 			return 0;
253 		break;
254 	}
255 
256 	err = -ENOENT;
257 
258 	read_lock(&ipip6_lock);
259 	t = ipip6_tunnel_lookup(iph->daddr, iph->saddr);
260 	if (t == NULL || t->parms.iph.daddr == 0)
261 		goto out;
262 
263 	err = 0;
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 err;
275 #else
276 	struct iphdr *iph = (struct iphdr*)dp;
277 	int hlen = iph->ihl<<2;
278 	struct ipv6hdr *iph6;
279 	const int type = icmp_hdr(skb)->type;
280 	const int code = icmp_hdr(skb)->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 (icmp_hdr(skb)->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 = icmp_hdr(skb)->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 0;
336 	dst_release(skb2->dst);
337 	skb2->dst = NULL;
338 	skb_pull(skb2, skb->data - (u8*)iph6);
339 	skb_reset_network_header(skb2);
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 0;
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(ipv6_hdr(skb));
366 }
367 
368 /* ISATAP (RFC4214) - check source address */
369 static int
370 isatap_srcok(struct sk_buff *skb, struct iphdr *iph, struct net_device *dev)
371 {
372 	struct neighbour *neigh;
373 	struct dst_entry *dst;
374 	struct rt6_info *rt;
375 	struct flowi fl;
376 	struct in6_addr *addr6;
377 	struct in6_addr rtr;
378 	struct ipv6hdr *iph6;
379 	int ok = 0;
380 
381 	/* from onlink default router */
382 	ipv6_addr_set(&rtr,  htonl(0xFE800000), 0, 0, 0);
383 	ipv6_isatap_eui64(rtr.s6_addr + 8, iph->saddr);
384 	if ((rt = rt6_get_dflt_router(&rtr, dev))) {
385 		dst_release(&rt->u.dst);
386 		return 1;
387 	}
388 
389 	iph6 = ipv6_hdr(skb);
390 	memset(&fl, 0, sizeof(fl));
391 	fl.proto = iph6->nexthdr;
392 	ipv6_addr_copy(&fl.fl6_dst, &iph6->saddr);
393 	fl.oif = dev->ifindex;
394 	security_skb_classify_flow(skb, &fl);
395 
396 	dst = ip6_route_output(NULL, &fl);
397 	if (!dst->error && (dst->dev == dev) && (neigh = dst->neighbour)) {
398 
399 		addr6 = (struct in6_addr*)&neigh->primary_key;
400 
401 		/* from correct previous hop */
402 		if (ipv6_addr_is_isatap(addr6) &&
403 		    (addr6->s6_addr32[3] == iph->saddr))
404 			ok = 1;
405 	}
406 	dst_release(dst);
407 	return ok;
408 }
409 
410 static int ipip6_rcv(struct sk_buff *skb)
411 {
412 	struct iphdr *iph;
413 	struct ip_tunnel *tunnel;
414 
415 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
416 		goto out;
417 
418 	iph = ip_hdr(skb);
419 
420 	read_lock(&ipip6_lock);
421 	if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
422 		secpath_reset(skb);
423 		skb->mac_header = skb->network_header;
424 		skb_reset_network_header(skb);
425 		IPCB(skb)->flags = 0;
426 		skb->protocol = htons(ETH_P_IPV6);
427 		skb->pkt_type = PACKET_HOST;
428 
429 		if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
430 		    !isatap_srcok(skb, iph, tunnel->dev)) {
431 			tunnel->stat.rx_errors++;
432 			read_unlock(&ipip6_lock);
433 			kfree_skb(skb);
434 			return 0;
435 		}
436 		tunnel->stat.rx_packets++;
437 		tunnel->stat.rx_bytes += skb->len;
438 		skb->dev = tunnel->dev;
439 		dst_release(skb->dst);
440 		skb->dst = NULL;
441 		nf_reset(skb);
442 		ipip6_ecn_decapsulate(iph, skb);
443 		netif_rx(skb);
444 		read_unlock(&ipip6_lock);
445 		return 0;
446 	}
447 
448 	icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
449 	kfree_skb(skb);
450 	read_unlock(&ipip6_lock);
451 out:
452 	return 0;
453 }
454 
455 /* Returns the embedded IPv4 address if the IPv6 address
456    comes from 6to4 (RFC 3056) addr space */
457 
458 static inline __be32 try_6to4(struct in6_addr *v6dst)
459 {
460 	__be32 dst = 0;
461 
462 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
463 		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
464 		memcpy(&dst, &v6dst->s6_addr16[1], 4);
465 	}
466 	return dst;
467 }
468 
469 /*
470  *	This function assumes it is being called from dev_queue_xmit()
471  *	and that skb is filled properly by that function.
472  */
473 
474 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
475 {
476 	struct ip_tunnel *tunnel = netdev_priv(dev);
477 	struct net_device_stats *stats = &tunnel->stat;
478 	struct iphdr  *tiph = &tunnel->parms.iph;
479 	struct ipv6hdr *iph6 = ipv6_hdr(skb);
480 	u8     tos = tunnel->parms.iph.tos;
481 	struct rtable *rt;     			/* Route to the other host */
482 	struct net_device *tdev;			/* Device to other host */
483 	struct iphdr  *iph;			/* Our new IP header */
484 	unsigned int max_headroom;		/* The extra header space needed */
485 	__be32 dst = tiph->daddr;
486 	int    mtu;
487 	struct in6_addr *addr6;
488 	int addr_type;
489 
490 	if (tunnel->recursion++) {
491 		tunnel->stat.collisions++;
492 		goto tx_error;
493 	}
494 
495 	if (skb->protocol != htons(ETH_P_IPV6))
496 		goto tx_error;
497 
498 	/* ISATAP (RFC4214) - must come before 6to4 */
499 	if (dev->priv_flags & IFF_ISATAP) {
500 		struct neighbour *neigh = NULL;
501 
502 		if (skb->dst)
503 			neigh = skb->dst->neighbour;
504 
505 		if (neigh == NULL) {
506 			if (net_ratelimit())
507 				printk(KERN_DEBUG "sit: nexthop == NULL\n");
508 			goto tx_error;
509 		}
510 
511 		addr6 = (struct in6_addr*)&neigh->primary_key;
512 		addr_type = ipv6_addr_type(addr6);
513 
514 		if ((addr_type & IPV6_ADDR_UNICAST) &&
515 		     ipv6_addr_is_isatap(addr6))
516 			dst = addr6->s6_addr32[3];
517 		else
518 			goto tx_error;
519 	}
520 
521 	if (!dst)
522 		dst = try_6to4(&iph6->daddr);
523 
524 	if (!dst) {
525 		struct neighbour *neigh = NULL;
526 
527 		if (skb->dst)
528 			neigh = skb->dst->neighbour;
529 
530 		if (neigh == NULL) {
531 			if (net_ratelimit())
532 				printk(KERN_DEBUG "sit: nexthop == NULL\n");
533 			goto tx_error;
534 		}
535 
536 		addr6 = (struct in6_addr*)&neigh->primary_key;
537 		addr_type = ipv6_addr_type(addr6);
538 
539 		if (addr_type == IPV6_ADDR_ANY) {
540 			addr6 = &ipv6_hdr(skb)->daddr;
541 			addr_type = ipv6_addr_type(addr6);
542 		}
543 
544 		if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
545 			goto tx_error_icmp;
546 
547 		dst = addr6->s6_addr32[3];
548 	}
549 
550 	{
551 		struct flowi fl = { .nl_u = { .ip4_u =
552 					      { .daddr = dst,
553 						.saddr = tiph->saddr,
554 						.tos = RT_TOS(tos) } },
555 				    .oif = tunnel->parms.link,
556 				    .proto = IPPROTO_IPV6 };
557 		if (ip_route_output_key(&init_net, &rt, &fl)) {
558 			tunnel->stat.tx_carrier_errors++;
559 			goto tx_error_icmp;
560 		}
561 	}
562 	if (rt->rt_type != RTN_UNICAST) {
563 		ip_rt_put(rt);
564 		tunnel->stat.tx_carrier_errors++;
565 		goto tx_error_icmp;
566 	}
567 	tdev = rt->u.dst.dev;
568 
569 	if (tdev == dev) {
570 		ip_rt_put(rt);
571 		tunnel->stat.collisions++;
572 		goto tx_error;
573 	}
574 
575 	if (tiph->frag_off)
576 		mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
577 	else
578 		mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
579 
580 	if (mtu < 68) {
581 		tunnel->stat.collisions++;
582 		ip_rt_put(rt);
583 		goto tx_error;
584 	}
585 	if (mtu < IPV6_MIN_MTU)
586 		mtu = IPV6_MIN_MTU;
587 	if (tunnel->parms.iph.daddr && skb->dst)
588 		skb->dst->ops->update_pmtu(skb->dst, mtu);
589 
590 	if (skb->len > mtu) {
591 		icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
592 		ip_rt_put(rt);
593 		goto tx_error;
594 	}
595 
596 	if (tunnel->err_count > 0) {
597 		if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
598 			tunnel->err_count--;
599 			dst_link_failure(skb);
600 		} else
601 			tunnel->err_count = 0;
602 	}
603 
604 	/*
605 	 * Okay, now see if we can stuff it in the buffer as-is.
606 	 */
607 	max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
608 
609 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
610 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
611 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
612 		if (!new_skb) {
613 			ip_rt_put(rt);
614 			stats->tx_dropped++;
615 			dev_kfree_skb(skb);
616 			tunnel->recursion--;
617 			return 0;
618 		}
619 		if (skb->sk)
620 			skb_set_owner_w(new_skb, skb->sk);
621 		dev_kfree_skb(skb);
622 		skb = new_skb;
623 		iph6 = ipv6_hdr(skb);
624 	}
625 
626 	skb->transport_header = skb->network_header;
627 	skb_push(skb, sizeof(struct iphdr));
628 	skb_reset_network_header(skb);
629 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
630 	IPCB(skb)->flags = 0;
631 	dst_release(skb->dst);
632 	skb->dst = &rt->u.dst;
633 
634 	/*
635 	 *	Push down and install the IPIP header.
636 	 */
637 
638 	iph 			=	ip_hdr(skb);
639 	iph->version		=	4;
640 	iph->ihl		=	sizeof(struct iphdr)>>2;
641 	if (mtu > IPV6_MIN_MTU)
642 		iph->frag_off	=	htons(IP_DF);
643 	else
644 		iph->frag_off	=	0;
645 
646 	iph->protocol		=	IPPROTO_IPV6;
647 	iph->tos		=	INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
648 	iph->daddr		=	rt->rt_dst;
649 	iph->saddr		=	rt->rt_src;
650 
651 	if ((iph->ttl = tiph->ttl) == 0)
652 		iph->ttl	=	iph6->hop_limit;
653 
654 	nf_reset(skb);
655 
656 	IPTUNNEL_XMIT();
657 	tunnel->recursion--;
658 	return 0;
659 
660 tx_error_icmp:
661 	dst_link_failure(skb);
662 tx_error:
663 	stats->tx_errors++;
664 	dev_kfree_skb(skb);
665 	tunnel->recursion--;
666 	return 0;
667 }
668 
669 static void ipip6_tunnel_bind_dev(struct net_device *dev)
670 {
671 	struct net_device *tdev = NULL;
672 	struct ip_tunnel *tunnel;
673 	struct iphdr *iph;
674 
675 	tunnel = netdev_priv(dev);
676 	iph = &tunnel->parms.iph;
677 
678 	if (iph->daddr) {
679 		struct flowi fl = { .nl_u = { .ip4_u =
680 					      { .daddr = iph->daddr,
681 						.saddr = iph->saddr,
682 						.tos = RT_TOS(iph->tos) } },
683 				    .oif = tunnel->parms.link,
684 				    .proto = IPPROTO_IPV6 };
685 		struct rtable *rt;
686 		if (!ip_route_output_key(&init_net, &rt, &fl)) {
687 			tdev = rt->u.dst.dev;
688 			ip_rt_put(rt);
689 		}
690 		dev->flags |= IFF_POINTOPOINT;
691 	}
692 
693 	if (!tdev && tunnel->parms.link)
694 		tdev = __dev_get_by_index(&init_net, tunnel->parms.link);
695 
696 	if (tdev) {
697 		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
698 		dev->mtu = tdev->mtu - sizeof(struct iphdr);
699 		if (dev->mtu < IPV6_MIN_MTU)
700 			dev->mtu = IPV6_MIN_MTU;
701 	}
702 	dev->iflink = tunnel->parms.link;
703 }
704 
705 static int
706 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
707 {
708 	int err = 0;
709 	struct ip_tunnel_parm p;
710 	struct ip_tunnel *t;
711 
712 	switch (cmd) {
713 	case SIOCGETTUNNEL:
714 		t = NULL;
715 		if (dev == ipip6_fb_tunnel_dev) {
716 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
717 				err = -EFAULT;
718 				break;
719 			}
720 			t = ipip6_tunnel_locate(&p, 0);
721 		}
722 		if (t == NULL)
723 			t = netdev_priv(dev);
724 		memcpy(&p, &t->parms, sizeof(p));
725 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
726 			err = -EFAULT;
727 		break;
728 
729 	case SIOCADDTUNNEL:
730 	case SIOCCHGTUNNEL:
731 		err = -EPERM;
732 		if (!capable(CAP_NET_ADMIN))
733 			goto done;
734 
735 		err = -EFAULT;
736 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
737 			goto done;
738 
739 		err = -EINVAL;
740 		if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
741 		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
742 			goto done;
743 		if (p.iph.ttl)
744 			p.iph.frag_off |= htons(IP_DF);
745 
746 		t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
747 
748 		if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
749 			if (t != NULL) {
750 				if (t->dev != dev) {
751 					err = -EEXIST;
752 					break;
753 				}
754 			} else {
755 				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
756 				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
757 					err = -EINVAL;
758 					break;
759 				}
760 				t = netdev_priv(dev);
761 				ipip6_tunnel_unlink(t);
762 				t->parms.iph.saddr = p.iph.saddr;
763 				t->parms.iph.daddr = p.iph.daddr;
764 				memcpy(dev->dev_addr, &p.iph.saddr, 4);
765 				memcpy(dev->broadcast, &p.iph.daddr, 4);
766 				ipip6_tunnel_link(t);
767 				netdev_state_change(dev);
768 			}
769 		}
770 
771 		if (t) {
772 			err = 0;
773 			if (cmd == SIOCCHGTUNNEL) {
774 				t->parms.iph.ttl = p.iph.ttl;
775 				t->parms.iph.tos = p.iph.tos;
776 				if (t->parms.link != p.link) {
777 					t->parms.link = p.link;
778 					ipip6_tunnel_bind_dev(dev);
779 					netdev_state_change(dev);
780 				}
781 			}
782 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
783 				err = -EFAULT;
784 		} else
785 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
786 		break;
787 
788 	case SIOCDELTUNNEL:
789 		err = -EPERM;
790 		if (!capable(CAP_NET_ADMIN))
791 			goto done;
792 
793 		if (dev == ipip6_fb_tunnel_dev) {
794 			err = -EFAULT;
795 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
796 				goto done;
797 			err = -ENOENT;
798 			if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
799 				goto done;
800 			err = -EPERM;
801 			if (t == netdev_priv(ipip6_fb_tunnel_dev))
802 				goto done;
803 			dev = t->dev;
804 		}
805 		unregister_netdevice(dev);
806 		err = 0;
807 		break;
808 
809 	default:
810 		err = -EINVAL;
811 	}
812 
813 done:
814 	return err;
815 }
816 
817 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
818 {
819 	return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
820 }
821 
822 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
823 {
824 	if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
825 		return -EINVAL;
826 	dev->mtu = new_mtu;
827 	return 0;
828 }
829 
830 static void ipip6_tunnel_setup(struct net_device *dev)
831 {
832 	dev->uninit		= ipip6_tunnel_uninit;
833 	dev->destructor 	= free_netdev;
834 	dev->hard_start_xmit	= ipip6_tunnel_xmit;
835 	dev->get_stats		= ipip6_tunnel_get_stats;
836 	dev->do_ioctl		= ipip6_tunnel_ioctl;
837 	dev->change_mtu		= ipip6_tunnel_change_mtu;
838 
839 	dev->type		= ARPHRD_SIT;
840 	dev->hard_header_len 	= LL_MAX_HEADER + sizeof(struct iphdr);
841 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr);
842 	dev->flags		= IFF_NOARP;
843 	dev->iflink		= 0;
844 	dev->addr_len		= 4;
845 }
846 
847 static int ipip6_tunnel_init(struct net_device *dev)
848 {
849 	struct ip_tunnel *tunnel;
850 
851 	tunnel = netdev_priv(dev);
852 
853 	tunnel->dev = dev;
854 	strcpy(tunnel->parms.name, dev->name);
855 
856 	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
857 	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
858 
859 	ipip6_tunnel_bind_dev(dev);
860 
861 	return 0;
862 }
863 
864 static int __init ipip6_fb_tunnel_init(struct net_device *dev)
865 {
866 	struct ip_tunnel *tunnel = netdev_priv(dev);
867 	struct iphdr *iph = &tunnel->parms.iph;
868 
869 	tunnel->dev = dev;
870 	strcpy(tunnel->parms.name, dev->name);
871 
872 	iph->version		= 4;
873 	iph->protocol		= IPPROTO_IPV6;
874 	iph->ihl		= 5;
875 	iph->ttl		= 64;
876 
877 	dev_hold(dev);
878 	tunnels_wc[0]		= tunnel;
879 	return 0;
880 }
881 
882 static struct xfrm_tunnel sit_handler = {
883 	.handler	=	ipip6_rcv,
884 	.err_handler	=	ipip6_err,
885 	.priority	=	1,
886 };
887 
888 static void __exit sit_destroy_tunnels(void)
889 {
890 	int prio;
891 
892 	for (prio = 1; prio < 4; prio++) {
893 		int h;
894 		for (h = 0; h < HASH_SIZE; h++) {
895 			struct ip_tunnel *t;
896 			while ((t = tunnels[prio][h]) != NULL)
897 				unregister_netdevice(t->dev);
898 		}
899 	}
900 }
901 
902 static void __exit sit_cleanup(void)
903 {
904 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
905 
906 	rtnl_lock();
907 	sit_destroy_tunnels();
908 	unregister_netdevice(ipip6_fb_tunnel_dev);
909 	rtnl_unlock();
910 }
911 
912 static int __init sit_init(void)
913 {
914 	int err;
915 
916 	printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
917 
918 	if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
919 		printk(KERN_INFO "sit init: Can't add protocol\n");
920 		return -EAGAIN;
921 	}
922 
923 	ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
924 					   ipip6_tunnel_setup);
925 	if (!ipip6_fb_tunnel_dev) {
926 		err = -ENOMEM;
927 		goto err1;
928 	}
929 
930 	ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
931 
932 	if ((err =  register_netdev(ipip6_fb_tunnel_dev)))
933 		goto err2;
934 
935  out:
936 	return err;
937  err2:
938 	free_netdev(ipip6_fb_tunnel_dev);
939  err1:
940 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
941 	goto out;
942 }
943 
944 module_init(sit_init);
945 module_exit(sit_cleanup);
946 MODULE_LICENSE("GPL");
947 MODULE_ALIAS("sit0");
948