xref: /openbmc/linux/net/ipv6/sit.c (revision b627b4ed) 
1 /*
2  *	IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
8  *
9  *	This program is free software; you can redistribute it and/or
10  *      modify it under the terms of the GNU General Public License
11  *      as published by the Free Software Foundation; either version
12  *      2 of the License, or (at your option) any later version.
13  *
14  *	Changes:
15  * Roger Venning <r.venning@telstra.com>:	6to4 support
16  * Nate Thompson <nate@thebog.net>:		6to4 support
17  * Fred Templin <fred.l.templin@boeing.com>:	isatap support
18  */
19 
20 #include <linux/module.h>
21 #include <linux/capability.h>
22 #include <linux/errno.h>
23 #include <linux/types.h>
24 #include <linux/socket.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/in6.h>
28 #include <linux/netdevice.h>
29 #include <linux/if_arp.h>
30 #include <linux/icmp.h>
31 #include <asm/uaccess.h>
32 #include <linux/init.h>
33 #include <linux/netfilter_ipv4.h>
34 #include <linux/if_ether.h>
35 
36 #include <net/sock.h>
37 #include <net/snmp.h>
38 
39 #include <net/ipv6.h>
40 #include <net/protocol.h>
41 #include <net/transp_v6.h>
42 #include <net/ip6_fib.h>
43 #include <net/ip6_route.h>
44 #include <net/ndisc.h>
45 #include <net/addrconf.h>
46 #include <net/ip.h>
47 #include <net/udp.h>
48 #include <net/icmp.h>
49 #include <net/ipip.h>
50 #include <net/inet_ecn.h>
51 #include <net/xfrm.h>
52 #include <net/dsfield.h>
53 #include <net/net_namespace.h>
54 #include <net/netns/generic.h>
55 
56 /*
57    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
58 
59    For comments look at net/ipv4/ip_gre.c --ANK
60  */
61 
62 #define HASH_SIZE  16
63 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
64 
65 static void ipip6_fb_tunnel_init(struct net_device *dev);
66 static void 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 
192 	nt->parms = *parms;
193 	ipip6_tunnel_init(dev);
194 
195 	if (parms->i_flags & SIT_ISATAP)
196 		dev->priv_flags |= IFF_ISATAP;
197 
198 	if (register_netdevice(dev) < 0)
199 		goto failed_free;
200 
201 	dev_hold(dev);
202 
203 	ipip6_tunnel_link(sitn, nt);
204 	return nt;
205 
206 failed_free:
207 	free_netdev(dev);
208 failed:
209 	return NULL;
210 }
211 
212 static struct ip_tunnel_prl_entry *
213 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
214 {
215 	struct ip_tunnel_prl_entry *p = (struct ip_tunnel_prl_entry *)NULL;
216 
217 	for (p = t->prl; p; p = p->next)
218 		if (p->addr == addr)
219 			break;
220 	return p;
221 
222 }
223 
224 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
225 				struct ip_tunnel_prl __user *a)
226 {
227 	struct ip_tunnel_prl kprl, *kp;
228 	struct ip_tunnel_prl_entry *prl;
229 	unsigned int cmax, c = 0, ca, len;
230 	int ret = 0;
231 
232 	if (copy_from_user(&kprl, a, sizeof(kprl)))
233 		return -EFAULT;
234 	cmax = kprl.datalen / sizeof(kprl);
235 	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
236 		cmax = 1;
237 
238 	/* For simple GET or for root users,
239 	 * we try harder to allocate.
240 	 */
241 	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
242 		kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
243 		NULL;
244 
245 	read_lock(&ipip6_lock);
246 
247 	ca = t->prl_count < cmax ? t->prl_count : cmax;
248 
249 	if (!kp) {
250 		/* We don't try hard to allocate much memory for
251 		 * non-root users.
252 		 * For root users, retry allocating enough memory for
253 		 * the answer.
254 		 */
255 		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
256 		if (!kp) {
257 			ret = -ENOMEM;
258 			goto out;
259 		}
260 	}
261 
262 	c = 0;
263 	for (prl = t->prl; prl; prl = prl->next) {
264 		if (c > cmax)
265 			break;
266 		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
267 			continue;
268 		kp[c].addr = prl->addr;
269 		kp[c].flags = prl->flags;
270 		c++;
271 		if (kprl.addr != htonl(INADDR_ANY))
272 			break;
273 	}
274 out:
275 	read_unlock(&ipip6_lock);
276 
277 	len = sizeof(*kp) * c;
278 	ret = 0;
279 	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
280 		ret = -EFAULT;
281 
282 	kfree(kp);
283 
284 	return ret;
285 }
286 
287 static int
288 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
289 {
290 	struct ip_tunnel_prl_entry *p;
291 	int err = 0;
292 
293 	if (a->addr == htonl(INADDR_ANY))
294 		return -EINVAL;
295 
296 	write_lock(&ipip6_lock);
297 
298 	for (p = t->prl; p; p = p->next) {
299 		if (p->addr == a->addr) {
300 			if (chg)
301 				goto update;
302 			err = -EEXIST;
303 			goto out;
304 		}
305 	}
306 
307 	if (chg) {
308 		err = -ENXIO;
309 		goto out;
310 	}
311 
312 	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
313 	if (!p) {
314 		err = -ENOBUFS;
315 		goto out;
316 	}
317 
318 	p->next = t->prl;
319 	t->prl = p;
320 	t->prl_count++;
321 update:
322 	p->addr = a->addr;
323 	p->flags = a->flags;
324 out:
325 	write_unlock(&ipip6_lock);
326 	return err;
327 }
328 
329 static int
330 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
331 {
332 	struct ip_tunnel_prl_entry *x, **p;
333 	int err = 0;
334 
335 	write_lock(&ipip6_lock);
336 
337 	if (a && a->addr != htonl(INADDR_ANY)) {
338 		for (p = &t->prl; *p; p = &(*p)->next) {
339 			if ((*p)->addr == a->addr) {
340 				x = *p;
341 				*p = x->next;
342 				kfree(x);
343 				t->prl_count--;
344 				goto out;
345 			}
346 		}
347 		err = -ENXIO;
348 	} else {
349 		while (t->prl) {
350 			x = t->prl;
351 			t->prl = t->prl->next;
352 			kfree(x);
353 			t->prl_count--;
354 		}
355 	}
356 out:
357 	write_unlock(&ipip6_lock);
358 	return 0;
359 }
360 
361 static int
362 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
363 {
364 	struct ip_tunnel_prl_entry *p;
365 	int ok = 1;
366 
367 	read_lock(&ipip6_lock);
368 	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
369 	if (p) {
370 		if (p->flags & PRL_DEFAULT)
371 			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
372 		else
373 			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
374 	} else {
375 		struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
376 		if (ipv6_addr_is_isatap(addr6) &&
377 		    (addr6->s6_addr32[3] == iph->saddr) &&
378 		    ipv6_chk_prefix(addr6, t->dev))
379 			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
380 		else
381 			ok = 0;
382 	}
383 	read_unlock(&ipip6_lock);
384 	return ok;
385 }
386 
387 static void ipip6_tunnel_uninit(struct net_device *dev)
388 {
389 	struct net *net = dev_net(dev);
390 	struct sit_net *sitn = net_generic(net, sit_net_id);
391 
392 	if (dev == sitn->fb_tunnel_dev) {
393 		write_lock_bh(&ipip6_lock);
394 		sitn->tunnels_wc[0] = NULL;
395 		write_unlock_bh(&ipip6_lock);
396 		dev_put(dev);
397 	} else {
398 		ipip6_tunnel_unlink(sitn, netdev_priv(dev));
399 		ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
400 		dev_put(dev);
401 	}
402 }
403 
404 
405 static int ipip6_err(struct sk_buff *skb, u32 info)
406 {
407 
408 /* All the routers (except for Linux) return only
409    8 bytes of packet payload. It means, that precise relaying of
410    ICMP in the real Internet is absolutely infeasible.
411  */
412 	struct iphdr *iph = (struct iphdr*)skb->data;
413 	const int type = icmp_hdr(skb)->type;
414 	const int code = icmp_hdr(skb)->code;
415 	struct ip_tunnel *t;
416 	int err;
417 
418 	switch (type) {
419 	default:
420 	case ICMP_PARAMETERPROB:
421 		return 0;
422 
423 	case ICMP_DEST_UNREACH:
424 		switch (code) {
425 		case ICMP_SR_FAILED:
426 		case ICMP_PORT_UNREACH:
427 			/* Impossible event. */
428 			return 0;
429 		case ICMP_FRAG_NEEDED:
430 			/* Soft state for pmtu is maintained by IP core. */
431 			return 0;
432 		default:
433 			/* All others are translated to HOST_UNREACH.
434 			   rfc2003 contains "deep thoughts" about NET_UNREACH,
435 			   I believe they are just ether pollution. --ANK
436 			 */
437 			break;
438 		}
439 		break;
440 	case ICMP_TIME_EXCEEDED:
441 		if (code != ICMP_EXC_TTL)
442 			return 0;
443 		break;
444 	}
445 
446 	err = -ENOENT;
447 
448 	read_lock(&ipip6_lock);
449 	t = ipip6_tunnel_lookup(dev_net(skb->dev), iph->daddr, iph->saddr);
450 	if (t == NULL || t->parms.iph.daddr == 0)
451 		goto out;
452 
453 	err = 0;
454 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
455 		goto out;
456 
457 	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
458 		t->err_count++;
459 	else
460 		t->err_count = 1;
461 	t->err_time = jiffies;
462 out:
463 	read_unlock(&ipip6_lock);
464 	return err;
465 }
466 
467 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
468 {
469 	if (INET_ECN_is_ce(iph->tos))
470 		IP6_ECN_set_ce(ipv6_hdr(skb));
471 }
472 
473 static int ipip6_rcv(struct sk_buff *skb)
474 {
475 	struct iphdr *iph;
476 	struct ip_tunnel *tunnel;
477 
478 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
479 		goto out;
480 
481 	iph = ip_hdr(skb);
482 
483 	read_lock(&ipip6_lock);
484 	if ((tunnel = ipip6_tunnel_lookup(dev_net(skb->dev),
485 					iph->saddr, iph->daddr)) != NULL) {
486 		secpath_reset(skb);
487 		skb->mac_header = skb->network_header;
488 		skb_reset_network_header(skb);
489 		IPCB(skb)->flags = 0;
490 		skb->protocol = htons(ETH_P_IPV6);
491 		skb->pkt_type = PACKET_HOST;
492 
493 		if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
494 		    !isatap_chksrc(skb, iph, tunnel)) {
495 			tunnel->dev->stats.rx_errors++;
496 			read_unlock(&ipip6_lock);
497 			kfree_skb(skb);
498 			return 0;
499 		}
500 		tunnel->dev->stats.rx_packets++;
501 		tunnel->dev->stats.rx_bytes += skb->len;
502 		skb->dev = tunnel->dev;
503 		dst_release(skb->dst);
504 		skb->dst = NULL;
505 		nf_reset(skb);
506 		ipip6_ecn_decapsulate(iph, skb);
507 		netif_rx(skb);
508 		read_unlock(&ipip6_lock);
509 		return 0;
510 	}
511 
512 	icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
513 	read_unlock(&ipip6_lock);
514 out:
515 	kfree_skb(skb);
516 	return 0;
517 }
518 
519 /* Returns the embedded IPv4 address if the IPv6 address
520    comes from 6to4 (RFC 3056) addr space */
521 
522 static inline __be32 try_6to4(struct in6_addr *v6dst)
523 {
524 	__be32 dst = 0;
525 
526 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
527 		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
528 		memcpy(&dst, &v6dst->s6_addr16[1], 4);
529 	}
530 	return dst;
531 }
532 
533 /*
534  *	This function assumes it is being called from dev_queue_xmit()
535  *	and that skb is filled properly by that function.
536  */
537 
538 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
539 {
540 	struct ip_tunnel *tunnel = netdev_priv(dev);
541 	struct net_device_stats *stats = &tunnel->dev->stats;
542 	struct iphdr  *tiph = &tunnel->parms.iph;
543 	struct ipv6hdr *iph6 = ipv6_hdr(skb);
544 	u8     tos = tunnel->parms.iph.tos;
545 	struct rtable *rt;     			/* Route to the other host */
546 	struct net_device *tdev;			/* Device to other host */
547 	struct iphdr  *iph;			/* Our new IP header */
548 	unsigned int max_headroom;		/* The extra header space needed */
549 	__be32 dst = tiph->daddr;
550 	int    mtu;
551 	struct in6_addr *addr6;
552 	int addr_type;
553 
554 	if (tunnel->recursion++) {
555 		stats->collisions++;
556 		goto tx_error;
557 	}
558 
559 	if (skb->protocol != htons(ETH_P_IPV6))
560 		goto tx_error;
561 
562 	/* ISATAP (RFC4214) - must come before 6to4 */
563 	if (dev->priv_flags & IFF_ISATAP) {
564 		struct neighbour *neigh = NULL;
565 
566 		if (skb->dst)
567 			neigh = skb->dst->neighbour;
568 
569 		if (neigh == NULL) {
570 			if (net_ratelimit())
571 				printk(KERN_DEBUG "sit: nexthop == NULL\n");
572 			goto tx_error;
573 		}
574 
575 		addr6 = (struct in6_addr*)&neigh->primary_key;
576 		addr_type = ipv6_addr_type(addr6);
577 
578 		if ((addr_type & IPV6_ADDR_UNICAST) &&
579 		     ipv6_addr_is_isatap(addr6))
580 			dst = addr6->s6_addr32[3];
581 		else
582 			goto tx_error;
583 	}
584 
585 	if (!dst)
586 		dst = try_6to4(&iph6->daddr);
587 
588 	if (!dst) {
589 		struct neighbour *neigh = NULL;
590 
591 		if (skb->dst)
592 			neigh = skb->dst->neighbour;
593 
594 		if (neigh == NULL) {
595 			if (net_ratelimit())
596 				printk(KERN_DEBUG "sit: nexthop == NULL\n");
597 			goto tx_error;
598 		}
599 
600 		addr6 = (struct in6_addr*)&neigh->primary_key;
601 		addr_type = ipv6_addr_type(addr6);
602 
603 		if (addr_type == IPV6_ADDR_ANY) {
604 			addr6 = &ipv6_hdr(skb)->daddr;
605 			addr_type = ipv6_addr_type(addr6);
606 		}
607 
608 		if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
609 			goto tx_error_icmp;
610 
611 		dst = addr6->s6_addr32[3];
612 	}
613 
614 	{
615 		struct flowi fl = { .nl_u = { .ip4_u =
616 					      { .daddr = dst,
617 						.saddr = tiph->saddr,
618 						.tos = RT_TOS(tos) } },
619 				    .oif = tunnel->parms.link,
620 				    .proto = IPPROTO_IPV6 };
621 		if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
622 			stats->tx_carrier_errors++;
623 			goto tx_error_icmp;
624 		}
625 	}
626 	if (rt->rt_type != RTN_UNICAST) {
627 		ip_rt_put(rt);
628 		stats->tx_carrier_errors++;
629 		goto tx_error_icmp;
630 	}
631 	tdev = rt->u.dst.dev;
632 
633 	if (tdev == dev) {
634 		ip_rt_put(rt);
635 		stats->collisions++;
636 		goto tx_error;
637 	}
638 
639 	if (tiph->frag_off)
640 		mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
641 	else
642 		mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
643 
644 	if (mtu < 68) {
645 		stats->collisions++;
646 		ip_rt_put(rt);
647 		goto tx_error;
648 	}
649 	if (mtu < IPV6_MIN_MTU)
650 		mtu = IPV6_MIN_MTU;
651 	if (tunnel->parms.iph.daddr && skb->dst)
652 		skb->dst->ops->update_pmtu(skb->dst, mtu);
653 
654 	if (skb->len > mtu) {
655 		icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
656 		ip_rt_put(rt);
657 		goto tx_error;
658 	}
659 
660 	if (tunnel->err_count > 0) {
661 		if (time_before(jiffies,
662 				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
663 			tunnel->err_count--;
664 			dst_link_failure(skb);
665 		} else
666 			tunnel->err_count = 0;
667 	}
668 
669 	/*
670 	 * Okay, now see if we can stuff it in the buffer as-is.
671 	 */
672 	max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
673 
674 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
675 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
676 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
677 		if (!new_skb) {
678 			ip_rt_put(rt);
679 			stats->tx_dropped++;
680 			dev_kfree_skb(skb);
681 			tunnel->recursion--;
682 			return 0;
683 		}
684 		if (skb->sk)
685 			skb_set_owner_w(new_skb, skb->sk);
686 		dev_kfree_skb(skb);
687 		skb = new_skb;
688 		iph6 = ipv6_hdr(skb);
689 	}
690 
691 	skb->transport_header = skb->network_header;
692 	skb_push(skb, sizeof(struct iphdr));
693 	skb_reset_network_header(skb);
694 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
695 	IPCB(skb)->flags = 0;
696 	dst_release(skb->dst);
697 	skb->dst = &rt->u.dst;
698 
699 	/*
700 	 *	Push down and install the IPIP header.
701 	 */
702 
703 	iph 			=	ip_hdr(skb);
704 	iph->version		=	4;
705 	iph->ihl		=	sizeof(struct iphdr)>>2;
706 	if (mtu > IPV6_MIN_MTU)
707 		iph->frag_off	=	htons(IP_DF);
708 	else
709 		iph->frag_off	=	0;
710 
711 	iph->protocol		=	IPPROTO_IPV6;
712 	iph->tos		=	INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
713 	iph->daddr		=	rt->rt_dst;
714 	iph->saddr		=	rt->rt_src;
715 
716 	if ((iph->ttl = tiph->ttl) == 0)
717 		iph->ttl	=	iph6->hop_limit;
718 
719 	nf_reset(skb);
720 
721 	IPTUNNEL_XMIT();
722 	tunnel->recursion--;
723 	return 0;
724 
725 tx_error_icmp:
726 	dst_link_failure(skb);
727 tx_error:
728 	stats->tx_errors++;
729 	dev_kfree_skb(skb);
730 	tunnel->recursion--;
731 	return 0;
732 }
733 
734 static void ipip6_tunnel_bind_dev(struct net_device *dev)
735 {
736 	struct net_device *tdev = NULL;
737 	struct ip_tunnel *tunnel;
738 	struct iphdr *iph;
739 
740 	tunnel = netdev_priv(dev);
741 	iph = &tunnel->parms.iph;
742 
743 	if (iph->daddr) {
744 		struct flowi fl = { .nl_u = { .ip4_u =
745 					      { .daddr = iph->daddr,
746 						.saddr = iph->saddr,
747 						.tos = RT_TOS(iph->tos) } },
748 				    .oif = tunnel->parms.link,
749 				    .proto = IPPROTO_IPV6 };
750 		struct rtable *rt;
751 		if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
752 			tdev = rt->u.dst.dev;
753 			ip_rt_put(rt);
754 		}
755 		dev->flags |= IFF_POINTOPOINT;
756 	}
757 
758 	if (!tdev && tunnel->parms.link)
759 		tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
760 
761 	if (tdev) {
762 		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
763 		dev->mtu = tdev->mtu - sizeof(struct iphdr);
764 		if (dev->mtu < IPV6_MIN_MTU)
765 			dev->mtu = IPV6_MIN_MTU;
766 	}
767 	dev->iflink = tunnel->parms.link;
768 }
769 
770 static int
771 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
772 {
773 	int err = 0;
774 	struct ip_tunnel_parm p;
775 	struct ip_tunnel_prl prl;
776 	struct ip_tunnel *t;
777 	struct net *net = dev_net(dev);
778 	struct sit_net *sitn = net_generic(net, sit_net_id);
779 
780 	switch (cmd) {
781 	case SIOCGETTUNNEL:
782 		t = NULL;
783 		if (dev == sitn->fb_tunnel_dev) {
784 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
785 				err = -EFAULT;
786 				break;
787 			}
788 			t = ipip6_tunnel_locate(net, &p, 0);
789 		}
790 		if (t == NULL)
791 			t = netdev_priv(dev);
792 		memcpy(&p, &t->parms, sizeof(p));
793 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
794 			err = -EFAULT;
795 		break;
796 
797 	case SIOCADDTUNNEL:
798 	case SIOCCHGTUNNEL:
799 		err = -EPERM;
800 		if (!capable(CAP_NET_ADMIN))
801 			goto done;
802 
803 		err = -EFAULT;
804 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
805 			goto done;
806 
807 		err = -EINVAL;
808 		if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
809 		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
810 			goto done;
811 		if (p.iph.ttl)
812 			p.iph.frag_off |= htons(IP_DF);
813 
814 		t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
815 
816 		if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
817 			if (t != NULL) {
818 				if (t->dev != dev) {
819 					err = -EEXIST;
820 					break;
821 				}
822 			} else {
823 				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
824 				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
825 					err = -EINVAL;
826 					break;
827 				}
828 				t = netdev_priv(dev);
829 				ipip6_tunnel_unlink(sitn, t);
830 				t->parms.iph.saddr = p.iph.saddr;
831 				t->parms.iph.daddr = p.iph.daddr;
832 				memcpy(dev->dev_addr, &p.iph.saddr, 4);
833 				memcpy(dev->broadcast, &p.iph.daddr, 4);
834 				ipip6_tunnel_link(sitn, t);
835 				netdev_state_change(dev);
836 			}
837 		}
838 
839 		if (t) {
840 			err = 0;
841 			if (cmd == SIOCCHGTUNNEL) {
842 				t->parms.iph.ttl = p.iph.ttl;
843 				t->parms.iph.tos = p.iph.tos;
844 				if (t->parms.link != p.link) {
845 					t->parms.link = p.link;
846 					ipip6_tunnel_bind_dev(dev);
847 					netdev_state_change(dev);
848 				}
849 			}
850 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
851 				err = -EFAULT;
852 		} else
853 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
854 		break;
855 
856 	case SIOCDELTUNNEL:
857 		err = -EPERM;
858 		if (!capable(CAP_NET_ADMIN))
859 			goto done;
860 
861 		if (dev == sitn->fb_tunnel_dev) {
862 			err = -EFAULT;
863 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
864 				goto done;
865 			err = -ENOENT;
866 			if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
867 				goto done;
868 			err = -EPERM;
869 			if (t == netdev_priv(sitn->fb_tunnel_dev))
870 				goto done;
871 			dev = t->dev;
872 		}
873 		unregister_netdevice(dev);
874 		err = 0;
875 		break;
876 
877 	case SIOCGETPRL:
878 		err = -EINVAL;
879 		if (dev == sitn->fb_tunnel_dev)
880 			goto done;
881 		err = -ENOENT;
882 		if (!(t = netdev_priv(dev)))
883 			goto done;
884 		err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
885 		break;
886 
887 	case SIOCADDPRL:
888 	case SIOCDELPRL:
889 	case SIOCCHGPRL:
890 		err = -EPERM;
891 		if (!capable(CAP_NET_ADMIN))
892 			goto done;
893 		err = -EINVAL;
894 		if (dev == sitn->fb_tunnel_dev)
895 			goto done;
896 		err = -EFAULT;
897 		if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
898 			goto done;
899 		err = -ENOENT;
900 		if (!(t = netdev_priv(dev)))
901 			goto done;
902 
903 		switch (cmd) {
904 		case SIOCDELPRL:
905 			err = ipip6_tunnel_del_prl(t, &prl);
906 			break;
907 		case SIOCADDPRL:
908 		case SIOCCHGPRL:
909 			err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
910 			break;
911 		}
912 		netdev_state_change(dev);
913 		break;
914 
915 	default:
916 		err = -EINVAL;
917 	}
918 
919 done:
920 	return err;
921 }
922 
923 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
924 {
925 	if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
926 		return -EINVAL;
927 	dev->mtu = new_mtu;
928 	return 0;
929 }
930 
931 static const struct net_device_ops ipip6_netdev_ops = {
932 	.ndo_uninit	= ipip6_tunnel_uninit,
933 	.ndo_start_xmit	= ipip6_tunnel_xmit,
934 	.ndo_do_ioctl	= ipip6_tunnel_ioctl,
935 	.ndo_change_mtu	= ipip6_tunnel_change_mtu,
936 };
937 
938 static void ipip6_tunnel_setup(struct net_device *dev)
939 {
940 	dev->netdev_ops		= &ipip6_netdev_ops;
941 	dev->destructor 	= free_netdev;
942 
943 	dev->type		= ARPHRD_SIT;
944 	dev->hard_header_len 	= LL_MAX_HEADER + sizeof(struct iphdr);
945 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr);
946 	dev->flags		= IFF_NOARP;
947 	dev->iflink		= 0;
948 	dev->addr_len		= 4;
949 	dev->features		|= NETIF_F_NETNS_LOCAL;
950 }
951 
952 static void ipip6_tunnel_init(struct net_device *dev)
953 {
954 	struct ip_tunnel *tunnel = netdev_priv(dev);
955 
956 	tunnel->dev = dev;
957 	strcpy(tunnel->parms.name, dev->name);
958 
959 	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
960 	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
961 
962 	ipip6_tunnel_bind_dev(dev);
963 }
964 
965 static void ipip6_fb_tunnel_init(struct net_device *dev)
966 {
967 	struct ip_tunnel *tunnel = netdev_priv(dev);
968 	struct iphdr *iph = &tunnel->parms.iph;
969 	struct net *net = dev_net(dev);
970 	struct sit_net *sitn = net_generic(net, sit_net_id);
971 
972 	tunnel->dev = dev;
973 	strcpy(tunnel->parms.name, dev->name);
974 
975 	iph->version		= 4;
976 	iph->protocol		= IPPROTO_IPV6;
977 	iph->ihl		= 5;
978 	iph->ttl		= 64;
979 
980 	dev_hold(dev);
981 	sitn->tunnels_wc[0]	= tunnel;
982 }
983 
984 static struct xfrm_tunnel sit_handler = {
985 	.handler	=	ipip6_rcv,
986 	.err_handler	=	ipip6_err,
987 	.priority	=	1,
988 };
989 
990 static void sit_destroy_tunnels(struct sit_net *sitn)
991 {
992 	int prio;
993 
994 	for (prio = 1; prio < 4; prio++) {
995 		int h;
996 		for (h = 0; h < HASH_SIZE; h++) {
997 			struct ip_tunnel *t;
998 			while ((t = sitn->tunnels[prio][h]) != NULL)
999 				unregister_netdevice(t->dev);
1000 		}
1001 	}
1002 }
1003 
1004 static int sit_init_net(struct net *net)
1005 {
1006 	int err;
1007 	struct sit_net *sitn;
1008 
1009 	err = -ENOMEM;
1010 	sitn = kzalloc(sizeof(struct sit_net), GFP_KERNEL);
1011 	if (sitn == NULL)
1012 		goto err_alloc;
1013 
1014 	err = net_assign_generic(net, sit_net_id, sitn);
1015 	if (err < 0)
1016 		goto err_assign;
1017 
1018 	sitn->tunnels[0] = sitn->tunnels_wc;
1019 	sitn->tunnels[1] = sitn->tunnels_l;
1020 	sitn->tunnels[2] = sitn->tunnels_r;
1021 	sitn->tunnels[3] = sitn->tunnels_r_l;
1022 
1023 	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1024 					   ipip6_tunnel_setup);
1025 	if (!sitn->fb_tunnel_dev) {
1026 		err = -ENOMEM;
1027 		goto err_alloc_dev;
1028 	}
1029 	dev_net_set(sitn->fb_tunnel_dev, net);
1030 
1031 	ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1032 
1033 	if ((err = register_netdev(sitn->fb_tunnel_dev)))
1034 		goto err_reg_dev;
1035 
1036 	return 0;
1037 
1038 err_reg_dev:
1039 	dev_put(sitn->fb_tunnel_dev);
1040 	free_netdev(sitn->fb_tunnel_dev);
1041 err_alloc_dev:
1042 	/* nothing */
1043 err_assign:
1044 	kfree(sitn);
1045 err_alloc:
1046 	return err;
1047 }
1048 
1049 static void sit_exit_net(struct net *net)
1050 {
1051 	struct sit_net *sitn;
1052 
1053 	sitn = net_generic(net, sit_net_id);
1054 	rtnl_lock();
1055 	sit_destroy_tunnels(sitn);
1056 	unregister_netdevice(sitn->fb_tunnel_dev);
1057 	rtnl_unlock();
1058 	kfree(sitn);
1059 }
1060 
1061 static struct pernet_operations sit_net_ops = {
1062 	.init = sit_init_net,
1063 	.exit = sit_exit_net,
1064 };
1065 
1066 static void __exit sit_cleanup(void)
1067 {
1068 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1069 
1070 	unregister_pernet_gen_device(sit_net_id, &sit_net_ops);
1071 }
1072 
1073 static int __init sit_init(void)
1074 {
1075 	int err;
1076 
1077 	printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1078 
1079 	if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
1080 		printk(KERN_INFO "sit init: Can't add protocol\n");
1081 		return -EAGAIN;
1082 	}
1083 
1084 	err = register_pernet_gen_device(&sit_net_id, &sit_net_ops);
1085 	if (err < 0)
1086 		xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1087 
1088 	return err;
1089 }
1090 
1091 module_init(sit_init);
1092 module_exit(sit_cleanup);
1093 MODULE_LICENSE("GPL");
1094 MODULE_ALIAS("sit0");
1095