xref: /openbmc/linux/net/ipv6/sit.c (revision 384740dc) 
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 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 int sit_net_id;
70 struct sit_net {
71 	struct ip_tunnel *tunnels_r_l[HASH_SIZE];
72 	struct ip_tunnel *tunnels_r[HASH_SIZE];
73 	struct ip_tunnel *tunnels_l[HASH_SIZE];
74 	struct ip_tunnel *tunnels_wc[1];
75 	struct ip_tunnel **tunnels[4];
76 
77 	struct net_device *fb_tunnel_dev;
78 };
79 
80 static DEFINE_RWLOCK(ipip6_lock);
81 
82 static struct ip_tunnel * ipip6_tunnel_lookup(struct net *net,
83 		__be32 remote, __be32 local)
84 {
85 	unsigned h0 = HASH(remote);
86 	unsigned h1 = HASH(local);
87 	struct ip_tunnel *t;
88 	struct sit_net *sitn = net_generic(net, sit_net_id);
89 
90 	for (t = sitn->tunnels_r_l[h0^h1]; t; t = t->next) {
91 		if (local == t->parms.iph.saddr &&
92 		    remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
93 			return t;
94 	}
95 	for (t = sitn->tunnels_r[h0]; t; t = t->next) {
96 		if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
97 			return t;
98 	}
99 	for (t = sitn->tunnels_l[h1]; t; t = t->next) {
100 		if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
101 			return t;
102 	}
103 	if ((t = sitn->tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
104 		return t;
105 	return NULL;
106 }
107 
108 static struct ip_tunnel **__ipip6_bucket(struct sit_net *sitn,
109 		struct ip_tunnel_parm *parms)
110 {
111 	__be32 remote = parms->iph.daddr;
112 	__be32 local = parms->iph.saddr;
113 	unsigned h = 0;
114 	int prio = 0;
115 
116 	if (remote) {
117 		prio |= 2;
118 		h ^= HASH(remote);
119 	}
120 	if (local) {
121 		prio |= 1;
122 		h ^= HASH(local);
123 	}
124 	return &sitn->tunnels[prio][h];
125 }
126 
127 static inline struct ip_tunnel **ipip6_bucket(struct sit_net *sitn,
128 		struct ip_tunnel *t)
129 {
130 	return __ipip6_bucket(sitn, &t->parms);
131 }
132 
133 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
134 {
135 	struct ip_tunnel **tp;
136 
137 	for (tp = ipip6_bucket(sitn, t); *tp; tp = &(*tp)->next) {
138 		if (t == *tp) {
139 			write_lock_bh(&ipip6_lock);
140 			*tp = t->next;
141 			write_unlock_bh(&ipip6_lock);
142 			break;
143 		}
144 	}
145 }
146 
147 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
148 {
149 	struct ip_tunnel **tp = ipip6_bucket(sitn, t);
150 
151 	t->next = *tp;
152 	write_lock_bh(&ipip6_lock);
153 	*tp = t;
154 	write_unlock_bh(&ipip6_lock);
155 }
156 
157 static struct ip_tunnel * ipip6_tunnel_locate(struct net *net,
158 		struct ip_tunnel_parm *parms, int create)
159 {
160 	__be32 remote = parms->iph.daddr;
161 	__be32 local = parms->iph.saddr;
162 	struct ip_tunnel *t, **tp, *nt;
163 	struct net_device *dev;
164 	char name[IFNAMSIZ];
165 	struct sit_net *sitn = net_generic(net, sit_net_id);
166 
167 	for (tp = __ipip6_bucket(sitn, parms); (t = *tp) != NULL; tp = &t->next) {
168 		if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
169 			return t;
170 	}
171 	if (!create)
172 		goto failed;
173 
174 	if (parms->name[0])
175 		strlcpy(name, parms->name, IFNAMSIZ);
176 	else
177 		sprintf(name, "sit%%d");
178 
179 	dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
180 	if (dev == NULL)
181 		return NULL;
182 
183 	dev_net_set(dev, net);
184 
185 	if (strchr(name, '%')) {
186 		if (dev_alloc_name(dev, name) < 0)
187 			goto failed_free;
188 	}
189 
190 	nt = netdev_priv(dev);
191 	dev->init = ipip6_tunnel_init;
192 	nt->parms = *parms;
193 
194 	if (parms->i_flags & SIT_ISATAP)
195 		dev->priv_flags |= IFF_ISATAP;
196 
197 	if (register_netdevice(dev) < 0)
198 		goto failed_free;
199 
200 	dev_hold(dev);
201 
202 	ipip6_tunnel_link(sitn, nt);
203 	return nt;
204 
205 failed_free:
206 	free_netdev(dev);
207 failed:
208 	return NULL;
209 }
210 
211 static struct ip_tunnel_prl_entry *
212 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
213 {
214 	struct ip_tunnel_prl_entry *p = (struct ip_tunnel_prl_entry *)NULL;
215 
216 	for (p = t->prl; p; p = p->next)
217 		if (p->addr == addr)
218 			break;
219 	return p;
220 
221 }
222 
223 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
224 				struct ip_tunnel_prl __user *a)
225 {
226 	struct ip_tunnel_prl kprl, *kp;
227 	struct ip_tunnel_prl_entry *prl;
228 	unsigned int cmax, c = 0, ca, len;
229 	int ret = 0;
230 
231 	if (copy_from_user(&kprl, a, sizeof(kprl)))
232 		return -EFAULT;
233 	cmax = kprl.datalen / sizeof(kprl);
234 	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
235 		cmax = 1;
236 
237 	/* For simple GET or for root users,
238 	 * we try harder to allocate.
239 	 */
240 	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
241 		kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
242 		NULL;
243 
244 	read_lock(&ipip6_lock);
245 
246 	ca = t->prl_count < cmax ? t->prl_count : cmax;
247 
248 	if (!kp) {
249 		/* We don't try hard to allocate much memory for
250 		 * non-root users.
251 		 * For root users, retry allocating enough memory for
252 		 * the answer.
253 		 */
254 		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
255 		if (!kp) {
256 			ret = -ENOMEM;
257 			goto out;
258 		}
259 	}
260 
261 	c = 0;
262 	for (prl = t->prl; prl; prl = prl->next) {
263 		if (c > cmax)
264 			break;
265 		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
266 			continue;
267 		kp[c].addr = prl->addr;
268 		kp[c].flags = prl->flags;
269 		c++;
270 		if (kprl.addr != htonl(INADDR_ANY))
271 			break;
272 	}
273 out:
274 	read_unlock(&ipip6_lock);
275 
276 	len = sizeof(*kp) * c;
277 	ret = 0;
278 	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
279 		ret = -EFAULT;
280 
281 	kfree(kp);
282 
283 	return ret;
284 }
285 
286 static int
287 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
288 {
289 	struct ip_tunnel_prl_entry *p;
290 	int err = 0;
291 
292 	if (a->addr == htonl(INADDR_ANY))
293 		return -EINVAL;
294 
295 	write_lock(&ipip6_lock);
296 
297 	for (p = t->prl; p; p = p->next) {
298 		if (p->addr == a->addr) {
299 			if (chg)
300 				goto update;
301 			err = -EEXIST;
302 			goto out;
303 		}
304 	}
305 
306 	if (chg) {
307 		err = -ENXIO;
308 		goto out;
309 	}
310 
311 	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
312 	if (!p) {
313 		err = -ENOBUFS;
314 		goto out;
315 	}
316 
317 	p->next = t->prl;
318 	t->prl = p;
319 	t->prl_count++;
320 update:
321 	p->addr = a->addr;
322 	p->flags = a->flags;
323 out:
324 	write_unlock(&ipip6_lock);
325 	return err;
326 }
327 
328 static int
329 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
330 {
331 	struct ip_tunnel_prl_entry *x, **p;
332 	int err = 0;
333 
334 	write_lock(&ipip6_lock);
335 
336 	if (a && a->addr != htonl(INADDR_ANY)) {
337 		for (p = &t->prl; *p; p = &(*p)->next) {
338 			if ((*p)->addr == a->addr) {
339 				x = *p;
340 				*p = x->next;
341 				kfree(x);
342 				t->prl_count--;
343 				goto out;
344 			}
345 		}
346 		err = -ENXIO;
347 	} else {
348 		while (t->prl) {
349 			x = t->prl;
350 			t->prl = t->prl->next;
351 			kfree(x);
352 			t->prl_count--;
353 		}
354 	}
355 out:
356 	write_unlock(&ipip6_lock);
357 	return 0;
358 }
359 
360 static int
361 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
362 {
363 	struct ip_tunnel_prl_entry *p;
364 	int ok = 1;
365 
366 	read_lock(&ipip6_lock);
367 	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
368 	if (p) {
369 		if (p->flags & PRL_DEFAULT)
370 			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
371 		else
372 			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
373 	} else {
374 		struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
375 		if (ipv6_addr_is_isatap(addr6) &&
376 		    (addr6->s6_addr32[3] == iph->saddr) &&
377 		    ipv6_chk_prefix(addr6, t->dev))
378 			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
379 		else
380 			ok = 0;
381 	}
382 	read_unlock(&ipip6_lock);
383 	return ok;
384 }
385 
386 static void ipip6_tunnel_uninit(struct net_device *dev)
387 {
388 	struct net *net = dev_net(dev);
389 	struct sit_net *sitn = net_generic(net, sit_net_id);
390 
391 	if (dev == sitn->fb_tunnel_dev) {
392 		write_lock_bh(&ipip6_lock);
393 		sitn->tunnels_wc[0] = NULL;
394 		write_unlock_bh(&ipip6_lock);
395 		dev_put(dev);
396 	} else {
397 		ipip6_tunnel_unlink(sitn, netdev_priv(dev));
398 		ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
399 		dev_put(dev);
400 	}
401 }
402 
403 
404 static int ipip6_err(struct sk_buff *skb, u32 info)
405 {
406 
407 /* All the routers (except for Linux) return only
408    8 bytes of packet payload. It means, that precise relaying of
409    ICMP in the real Internet is absolutely infeasible.
410  */
411 	struct iphdr *iph = (struct iphdr*)skb->data;
412 	const int type = icmp_hdr(skb)->type;
413 	const int code = icmp_hdr(skb)->code;
414 	struct ip_tunnel *t;
415 	int err;
416 
417 	switch (type) {
418 	default:
419 	case ICMP_PARAMETERPROB:
420 		return 0;
421 
422 	case ICMP_DEST_UNREACH:
423 		switch (code) {
424 		case ICMP_SR_FAILED:
425 		case ICMP_PORT_UNREACH:
426 			/* Impossible event. */
427 			return 0;
428 		case ICMP_FRAG_NEEDED:
429 			/* Soft state for pmtu is maintained by IP core. */
430 			return 0;
431 		default:
432 			/* All others are translated to HOST_UNREACH.
433 			   rfc2003 contains "deep thoughts" about NET_UNREACH,
434 			   I believe they are just ether pollution. --ANK
435 			 */
436 			break;
437 		}
438 		break;
439 	case ICMP_TIME_EXCEEDED:
440 		if (code != ICMP_EXC_TTL)
441 			return 0;
442 		break;
443 	}
444 
445 	err = -ENOENT;
446 
447 	read_lock(&ipip6_lock);
448 	t = ipip6_tunnel_lookup(dev_net(skb->dev), iph->daddr, iph->saddr);
449 	if (t == NULL || t->parms.iph.daddr == 0)
450 		goto out;
451 
452 	err = 0;
453 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
454 		goto out;
455 
456 	if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
457 		t->err_count++;
458 	else
459 		t->err_count = 1;
460 	t->err_time = jiffies;
461 out:
462 	read_unlock(&ipip6_lock);
463 	return err;
464 }
465 
466 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
467 {
468 	if (INET_ECN_is_ce(iph->tos))
469 		IP6_ECN_set_ce(ipv6_hdr(skb));
470 }
471 
472 static int ipip6_rcv(struct sk_buff *skb)
473 {
474 	struct iphdr *iph;
475 	struct ip_tunnel *tunnel;
476 
477 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
478 		goto out;
479 
480 	iph = ip_hdr(skb);
481 
482 	read_lock(&ipip6_lock);
483 	if ((tunnel = ipip6_tunnel_lookup(dev_net(skb->dev),
484 					iph->saddr, iph->daddr)) != NULL) {
485 		secpath_reset(skb);
486 		skb->mac_header = skb->network_header;
487 		skb_reset_network_header(skb);
488 		IPCB(skb)->flags = 0;
489 		skb->protocol = htons(ETH_P_IPV6);
490 		skb->pkt_type = PACKET_HOST;
491 
492 		if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
493 		    !isatap_chksrc(skb, iph, tunnel)) {
494 			tunnel->dev->stats.rx_errors++;
495 			read_unlock(&ipip6_lock);
496 			kfree_skb(skb);
497 			return 0;
498 		}
499 		tunnel->dev->stats.rx_packets++;
500 		tunnel->dev->stats.rx_bytes += skb->len;
501 		skb->dev = tunnel->dev;
502 		dst_release(skb->dst);
503 		skb->dst = NULL;
504 		nf_reset(skb);
505 		ipip6_ecn_decapsulate(iph, skb);
506 		netif_rx(skb);
507 		read_unlock(&ipip6_lock);
508 		return 0;
509 	}
510 
511 	icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
512 	read_unlock(&ipip6_lock);
513 out:
514 	kfree_skb(skb);
515 	return 0;
516 }
517 
518 /* Returns the embedded IPv4 address if the IPv6 address
519    comes from 6to4 (RFC 3056) addr space */
520 
521 static inline __be32 try_6to4(struct in6_addr *v6dst)
522 {
523 	__be32 dst = 0;
524 
525 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
526 		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
527 		memcpy(&dst, &v6dst->s6_addr16[1], 4);
528 	}
529 	return dst;
530 }
531 
532 /*
533  *	This function assumes it is being called from dev_queue_xmit()
534  *	and that skb is filled properly by that function.
535  */
536 
537 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
538 {
539 	struct ip_tunnel *tunnel = netdev_priv(dev);
540 	struct net_device_stats *stats = &tunnel->dev->stats;
541 	struct iphdr  *tiph = &tunnel->parms.iph;
542 	struct ipv6hdr *iph6 = ipv6_hdr(skb);
543 	u8     tos = tunnel->parms.iph.tos;
544 	struct rtable *rt;     			/* Route to the other host */
545 	struct net_device *tdev;			/* Device to other host */
546 	struct iphdr  *iph;			/* Our new IP header */
547 	unsigned int max_headroom;		/* The extra header space needed */
548 	__be32 dst = tiph->daddr;
549 	int    mtu;
550 	struct in6_addr *addr6;
551 	int addr_type;
552 
553 	if (tunnel->recursion++) {
554 		stats->collisions++;
555 		goto tx_error;
556 	}
557 
558 	if (skb->protocol != htons(ETH_P_IPV6))
559 		goto tx_error;
560 
561 	/* ISATAP (RFC4214) - must come before 6to4 */
562 	if (dev->priv_flags & IFF_ISATAP) {
563 		struct neighbour *neigh = NULL;
564 
565 		if (skb->dst)
566 			neigh = skb->dst->neighbour;
567 
568 		if (neigh == NULL) {
569 			if (net_ratelimit())
570 				printk(KERN_DEBUG "sit: nexthop == NULL\n");
571 			goto tx_error;
572 		}
573 
574 		addr6 = (struct in6_addr*)&neigh->primary_key;
575 		addr_type = ipv6_addr_type(addr6);
576 
577 		if ((addr_type & IPV6_ADDR_UNICAST) &&
578 		     ipv6_addr_is_isatap(addr6))
579 			dst = addr6->s6_addr32[3];
580 		else
581 			goto tx_error;
582 	}
583 
584 	if (!dst)
585 		dst = try_6to4(&iph6->daddr);
586 
587 	if (!dst) {
588 		struct neighbour *neigh = NULL;
589 
590 		if (skb->dst)
591 			neigh = skb->dst->neighbour;
592 
593 		if (neigh == NULL) {
594 			if (net_ratelimit())
595 				printk(KERN_DEBUG "sit: nexthop == NULL\n");
596 			goto tx_error;
597 		}
598 
599 		addr6 = (struct in6_addr*)&neigh->primary_key;
600 		addr_type = ipv6_addr_type(addr6);
601 
602 		if (addr_type == IPV6_ADDR_ANY) {
603 			addr6 = &ipv6_hdr(skb)->daddr;
604 			addr_type = ipv6_addr_type(addr6);
605 		}
606 
607 		if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
608 			goto tx_error_icmp;
609 
610 		dst = addr6->s6_addr32[3];
611 	}
612 
613 	{
614 		struct flowi fl = { .nl_u = { .ip4_u =
615 					      { .daddr = dst,
616 						.saddr = tiph->saddr,
617 						.tos = RT_TOS(tos) } },
618 				    .oif = tunnel->parms.link,
619 				    .proto = IPPROTO_IPV6 };
620 		if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
621 			stats->tx_carrier_errors++;
622 			goto tx_error_icmp;
623 		}
624 	}
625 	if (rt->rt_type != RTN_UNICAST) {
626 		ip_rt_put(rt);
627 		stats->tx_carrier_errors++;
628 		goto tx_error_icmp;
629 	}
630 	tdev = rt->u.dst.dev;
631 
632 	if (tdev == dev) {
633 		ip_rt_put(rt);
634 		stats->collisions++;
635 		goto tx_error;
636 	}
637 
638 	if (tiph->frag_off)
639 		mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
640 	else
641 		mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
642 
643 	if (mtu < 68) {
644 		stats->collisions++;
645 		ip_rt_put(rt);
646 		goto tx_error;
647 	}
648 	if (mtu < IPV6_MIN_MTU)
649 		mtu = IPV6_MIN_MTU;
650 	if (tunnel->parms.iph.daddr && skb->dst)
651 		skb->dst->ops->update_pmtu(skb->dst, mtu);
652 
653 	if (skb->len > mtu) {
654 		icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
655 		ip_rt_put(rt);
656 		goto tx_error;
657 	}
658 
659 	if (tunnel->err_count > 0) {
660 		if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
661 			tunnel->err_count--;
662 			dst_link_failure(skb);
663 		} else
664 			tunnel->err_count = 0;
665 	}
666 
667 	/*
668 	 * Okay, now see if we can stuff it in the buffer as-is.
669 	 */
670 	max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
671 
672 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
673 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
674 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
675 		if (!new_skb) {
676 			ip_rt_put(rt);
677 			stats->tx_dropped++;
678 			dev_kfree_skb(skb);
679 			tunnel->recursion--;
680 			return 0;
681 		}
682 		if (skb->sk)
683 			skb_set_owner_w(new_skb, skb->sk);
684 		dev_kfree_skb(skb);
685 		skb = new_skb;
686 		iph6 = ipv6_hdr(skb);
687 	}
688 
689 	skb->transport_header = skb->network_header;
690 	skb_push(skb, sizeof(struct iphdr));
691 	skb_reset_network_header(skb);
692 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
693 	IPCB(skb)->flags = 0;
694 	dst_release(skb->dst);
695 	skb->dst = &rt->u.dst;
696 
697 	/*
698 	 *	Push down and install the IPIP header.
699 	 */
700 
701 	iph 			=	ip_hdr(skb);
702 	iph->version		=	4;
703 	iph->ihl		=	sizeof(struct iphdr)>>2;
704 	if (mtu > IPV6_MIN_MTU)
705 		iph->frag_off	=	htons(IP_DF);
706 	else
707 		iph->frag_off	=	0;
708 
709 	iph->protocol		=	IPPROTO_IPV6;
710 	iph->tos		=	INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
711 	iph->daddr		=	rt->rt_dst;
712 	iph->saddr		=	rt->rt_src;
713 
714 	if ((iph->ttl = tiph->ttl) == 0)
715 		iph->ttl	=	iph6->hop_limit;
716 
717 	nf_reset(skb);
718 
719 	IPTUNNEL_XMIT();
720 	tunnel->recursion--;
721 	return 0;
722 
723 tx_error_icmp:
724 	dst_link_failure(skb);
725 tx_error:
726 	stats->tx_errors++;
727 	dev_kfree_skb(skb);
728 	tunnel->recursion--;
729 	return 0;
730 }
731 
732 static void ipip6_tunnel_bind_dev(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 	if (iph->daddr) {
742 		struct flowi fl = { .nl_u = { .ip4_u =
743 					      { .daddr = iph->daddr,
744 						.saddr = iph->saddr,
745 						.tos = RT_TOS(iph->tos) } },
746 				    .oif = tunnel->parms.link,
747 				    .proto = IPPROTO_IPV6 };
748 		struct rtable *rt;
749 		if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
750 			tdev = rt->u.dst.dev;
751 			ip_rt_put(rt);
752 		}
753 		dev->flags |= IFF_POINTOPOINT;
754 	}
755 
756 	if (!tdev && tunnel->parms.link)
757 		tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
758 
759 	if (tdev) {
760 		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
761 		dev->mtu = tdev->mtu - sizeof(struct iphdr);
762 		if (dev->mtu < IPV6_MIN_MTU)
763 			dev->mtu = IPV6_MIN_MTU;
764 	}
765 	dev->iflink = tunnel->parms.link;
766 }
767 
768 static int
769 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
770 {
771 	int err = 0;
772 	struct ip_tunnel_parm p;
773 	struct ip_tunnel_prl prl;
774 	struct ip_tunnel *t;
775 	struct net *net = dev_net(dev);
776 	struct sit_net *sitn = net_generic(net, sit_net_id);
777 
778 	switch (cmd) {
779 	case SIOCGETTUNNEL:
780 		t = NULL;
781 		if (dev == sitn->fb_tunnel_dev) {
782 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
783 				err = -EFAULT;
784 				break;
785 			}
786 			t = ipip6_tunnel_locate(net, &p, 0);
787 		}
788 		if (t == NULL)
789 			t = netdev_priv(dev);
790 		memcpy(&p, &t->parms, sizeof(p));
791 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
792 			err = -EFAULT;
793 		break;
794 
795 	case SIOCADDTUNNEL:
796 	case SIOCCHGTUNNEL:
797 		err = -EPERM;
798 		if (!capable(CAP_NET_ADMIN))
799 			goto done;
800 
801 		err = -EFAULT;
802 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
803 			goto done;
804 
805 		err = -EINVAL;
806 		if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
807 		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
808 			goto done;
809 		if (p.iph.ttl)
810 			p.iph.frag_off |= htons(IP_DF);
811 
812 		t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
813 
814 		if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
815 			if (t != NULL) {
816 				if (t->dev != dev) {
817 					err = -EEXIST;
818 					break;
819 				}
820 			} else {
821 				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
822 				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
823 					err = -EINVAL;
824 					break;
825 				}
826 				t = netdev_priv(dev);
827 				ipip6_tunnel_unlink(sitn, t);
828 				t->parms.iph.saddr = p.iph.saddr;
829 				t->parms.iph.daddr = p.iph.daddr;
830 				memcpy(dev->dev_addr, &p.iph.saddr, 4);
831 				memcpy(dev->broadcast, &p.iph.daddr, 4);
832 				ipip6_tunnel_link(sitn, t);
833 				netdev_state_change(dev);
834 			}
835 		}
836 
837 		if (t) {
838 			err = 0;
839 			if (cmd == SIOCCHGTUNNEL) {
840 				t->parms.iph.ttl = p.iph.ttl;
841 				t->parms.iph.tos = p.iph.tos;
842 				if (t->parms.link != p.link) {
843 					t->parms.link = p.link;
844 					ipip6_tunnel_bind_dev(dev);
845 					netdev_state_change(dev);
846 				}
847 			}
848 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
849 				err = -EFAULT;
850 		} else
851 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
852 		break;
853 
854 	case SIOCDELTUNNEL:
855 		err = -EPERM;
856 		if (!capable(CAP_NET_ADMIN))
857 			goto done;
858 
859 		if (dev == sitn->fb_tunnel_dev) {
860 			err = -EFAULT;
861 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
862 				goto done;
863 			err = -ENOENT;
864 			if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
865 				goto done;
866 			err = -EPERM;
867 			if (t == netdev_priv(sitn->fb_tunnel_dev))
868 				goto done;
869 			dev = t->dev;
870 		}
871 		unregister_netdevice(dev);
872 		err = 0;
873 		break;
874 
875 	case SIOCGETPRL:
876 		err = -EINVAL;
877 		if (dev == sitn->fb_tunnel_dev)
878 			goto done;
879 		err = -ENOENT;
880 		if (!(t = netdev_priv(dev)))
881 			goto done;
882 		err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
883 		break;
884 
885 	case SIOCADDPRL:
886 	case SIOCDELPRL:
887 	case SIOCCHGPRL:
888 		err = -EPERM;
889 		if (!capable(CAP_NET_ADMIN))
890 			goto done;
891 		err = -EINVAL;
892 		if (dev == sitn->fb_tunnel_dev)
893 			goto done;
894 		err = -EFAULT;
895 		if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
896 			goto done;
897 		err = -ENOENT;
898 		if (!(t = netdev_priv(dev)))
899 			goto done;
900 
901 		switch (cmd) {
902 		case SIOCDELPRL:
903 			err = ipip6_tunnel_del_prl(t, &prl);
904 			break;
905 		case SIOCADDPRL:
906 		case SIOCCHGPRL:
907 			err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
908 			break;
909 		}
910 		netdev_state_change(dev);
911 		break;
912 
913 	default:
914 		err = -EINVAL;
915 	}
916 
917 done:
918 	return err;
919 }
920 
921 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
922 {
923 	if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
924 		return -EINVAL;
925 	dev->mtu = new_mtu;
926 	return 0;
927 }
928 
929 static void ipip6_tunnel_setup(struct net_device *dev)
930 {
931 	dev->uninit		= ipip6_tunnel_uninit;
932 	dev->destructor 	= free_netdev;
933 	dev->hard_start_xmit	= ipip6_tunnel_xmit;
934 	dev->do_ioctl		= ipip6_tunnel_ioctl;
935 	dev->change_mtu		= ipip6_tunnel_change_mtu;
936 
937 	dev->type		= ARPHRD_SIT;
938 	dev->hard_header_len 	= LL_MAX_HEADER + sizeof(struct iphdr);
939 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr);
940 	dev->flags		= IFF_NOARP;
941 	dev->iflink		= 0;
942 	dev->addr_len		= 4;
943 	dev->features		|= NETIF_F_NETNS_LOCAL;
944 }
945 
946 static int ipip6_tunnel_init(struct net_device *dev)
947 {
948 	struct ip_tunnel *tunnel;
949 
950 	tunnel = netdev_priv(dev);
951 
952 	tunnel->dev = dev;
953 	strcpy(tunnel->parms.name, dev->name);
954 
955 	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
956 	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
957 
958 	ipip6_tunnel_bind_dev(dev);
959 
960 	return 0;
961 }
962 
963 static int ipip6_fb_tunnel_init(struct net_device *dev)
964 {
965 	struct ip_tunnel *tunnel = netdev_priv(dev);
966 	struct iphdr *iph = &tunnel->parms.iph;
967 	struct net *net = dev_net(dev);
968 	struct sit_net *sitn = net_generic(net, sit_net_id);
969 
970 	tunnel->dev = dev;
971 	strcpy(tunnel->parms.name, dev->name);
972 
973 	iph->version		= 4;
974 	iph->protocol		= IPPROTO_IPV6;
975 	iph->ihl		= 5;
976 	iph->ttl		= 64;
977 
978 	dev_hold(dev);
979 	sitn->tunnels_wc[0]	= tunnel;
980 	return 0;
981 }
982 
983 static struct xfrm_tunnel sit_handler = {
984 	.handler	=	ipip6_rcv,
985 	.err_handler	=	ipip6_err,
986 	.priority	=	1,
987 };
988 
989 static void sit_destroy_tunnels(struct sit_net *sitn)
990 {
991 	int prio;
992 
993 	for (prio = 1; prio < 4; prio++) {
994 		int h;
995 		for (h = 0; h < HASH_SIZE; h++) {
996 			struct ip_tunnel *t;
997 			while ((t = sitn->tunnels[prio][h]) != NULL)
998 				unregister_netdevice(t->dev);
999 		}
1000 	}
1001 }
1002 
1003 static int sit_init_net(struct net *net)
1004 {
1005 	int err;
1006 	struct sit_net *sitn;
1007 
1008 	err = -ENOMEM;
1009 	sitn = kzalloc(sizeof(struct sit_net), GFP_KERNEL);
1010 	if (sitn == NULL)
1011 		goto err_alloc;
1012 
1013 	err = net_assign_generic(net, sit_net_id, sitn);
1014 	if (err < 0)
1015 		goto err_assign;
1016 
1017 	sitn->tunnels[0] = sitn->tunnels_wc;
1018 	sitn->tunnels[1] = sitn->tunnels_l;
1019 	sitn->tunnels[2] = sitn->tunnels_r;
1020 	sitn->tunnels[3] = sitn->tunnels_r_l;
1021 
1022 	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1023 					   ipip6_tunnel_setup);
1024 	if (!sitn->fb_tunnel_dev) {
1025 		err = -ENOMEM;
1026 		goto err_alloc_dev;
1027 	}
1028 
1029 	sitn->fb_tunnel_dev->init = ipip6_fb_tunnel_init;
1030 	dev_net_set(sitn->fb_tunnel_dev, net);
1031 
1032 	if ((err = register_netdev(sitn->fb_tunnel_dev)))
1033 		goto err_reg_dev;
1034 
1035 	return 0;
1036 
1037 err_reg_dev:
1038 	free_netdev(sitn->fb_tunnel_dev);
1039 err_alloc_dev:
1040 	/* nothing */
1041 err_assign:
1042 	kfree(sitn);
1043 err_alloc:
1044 	return err;
1045 }
1046 
1047 static void sit_exit_net(struct net *net)
1048 {
1049 	struct sit_net *sitn;
1050 
1051 	sitn = net_generic(net, sit_net_id);
1052 	rtnl_lock();
1053 	sit_destroy_tunnels(sitn);
1054 	unregister_netdevice(sitn->fb_tunnel_dev);
1055 	rtnl_unlock();
1056 	kfree(sitn);
1057 }
1058 
1059 static struct pernet_operations sit_net_ops = {
1060 	.init = sit_init_net,
1061 	.exit = sit_exit_net,
1062 };
1063 
1064 static void __exit sit_cleanup(void)
1065 {
1066 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1067 
1068 	unregister_pernet_gen_device(sit_net_id, &sit_net_ops);
1069 }
1070 
1071 static int __init sit_init(void)
1072 {
1073 	int err;
1074 
1075 	printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1076 
1077 	if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
1078 		printk(KERN_INFO "sit init: Can't add protocol\n");
1079 		return -EAGAIN;
1080 	}
1081 
1082 	err = register_pernet_gen_device(&sit_net_id, &sit_net_ops);
1083 	if (err < 0)
1084 		xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1085 
1086 	return err;
1087 }
1088 
1089 module_init(sit_init);
1090 module_exit(sit_cleanup);
1091 MODULE_LICENSE("GPL");
1092 MODULE_ALIAS("sit0");
1093