xref: /openbmc/linux/net/ipv6/sit.c (revision 840ef8b7cc584a23c4f9d05352f4dbaf8e56e5ab)
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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
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 <linux/slab.h>
34 #include <asm/uaccess.h>
35 #include <linux/init.h>
36 #include <linux/netfilter_ipv4.h>
37 #include <linux/if_ether.h>
38 
39 #include <net/sock.h>
40 #include <net/snmp.h>
41 
42 #include <net/ipv6.h>
43 #include <net/protocol.h>
44 #include <net/transp_v6.h>
45 #include <net/ip6_fib.h>
46 #include <net/ip6_route.h>
47 #include <net/ndisc.h>
48 #include <net/addrconf.h>
49 #include <net/ip.h>
50 #include <net/udp.h>
51 #include <net/icmp.h>
52 #include <net/ipip.h>
53 #include <net/inet_ecn.h>
54 #include <net/xfrm.h>
55 #include <net/dsfield.h>
56 #include <net/net_namespace.h>
57 #include <net/netns/generic.h>
58 
59 /*
60    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
61 
62    For comments look at net/ipv4/ip_gre.c --ANK
63  */
64 
65 #define HASH_SIZE  16
66 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
67 
68 static bool log_ecn_error = true;
69 module_param(log_ecn_error, bool, 0644);
70 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
71 
72 static int ipip6_tunnel_init(struct net_device *dev);
73 static void ipip6_tunnel_setup(struct net_device *dev);
74 static void ipip6_dev_free(struct net_device *dev);
75 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
76 		      __be32 *v4dst);
77 static struct rtnl_link_ops sit_link_ops __read_mostly;
78 
79 static int sit_net_id __read_mostly;
80 struct sit_net {
81 	struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
82 	struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
83 	struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
84 	struct ip_tunnel __rcu *tunnels_wc[1];
85 	struct ip_tunnel __rcu **tunnels[4];
86 
87 	struct net_device *fb_tunnel_dev;
88 };
89 
90 static struct rtnl_link_stats64 *ipip6_get_stats64(struct net_device *dev,
91 						   struct rtnl_link_stats64 *tot)
92 {
93 	int i;
94 
95 	for_each_possible_cpu(i) {
96 		const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
97 		u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
98 		unsigned int start;
99 
100 		do {
101 			start = u64_stats_fetch_begin_bh(&tstats->syncp);
102 			rx_packets = tstats->rx_packets;
103 			tx_packets = tstats->tx_packets;
104 			rx_bytes = tstats->rx_bytes;
105 			tx_bytes = tstats->tx_bytes;
106 		} while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
107 
108 		tot->rx_packets += rx_packets;
109 		tot->tx_packets += tx_packets;
110 		tot->rx_bytes   += rx_bytes;
111 		tot->tx_bytes   += tx_bytes;
112 	}
113 
114 	tot->rx_errors = dev->stats.rx_errors;
115 	tot->rx_frame_errors = dev->stats.rx_frame_errors;
116 	tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
117 	tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
118 	tot->tx_dropped = dev->stats.tx_dropped;
119 	tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
120 	tot->tx_errors = dev->stats.tx_errors;
121 
122 	return tot;
123 }
124 
125 /*
126  * Must be invoked with rcu_read_lock
127  */
128 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
129 		struct net_device *dev, __be32 remote, __be32 local)
130 {
131 	unsigned int h0 = HASH(remote);
132 	unsigned int h1 = HASH(local);
133 	struct ip_tunnel *t;
134 	struct sit_net *sitn = net_generic(net, sit_net_id);
135 
136 	for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
137 		if (local == t->parms.iph.saddr &&
138 		    remote == t->parms.iph.daddr &&
139 		    (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
140 		    (t->dev->flags & IFF_UP))
141 			return t;
142 	}
143 	for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
144 		if (remote == t->parms.iph.daddr &&
145 		    (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
146 		    (t->dev->flags & IFF_UP))
147 			return t;
148 	}
149 	for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
150 		if (local == t->parms.iph.saddr &&
151 		    (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
152 		    (t->dev->flags & IFF_UP))
153 			return t;
154 	}
155 	t = rcu_dereference(sitn->tunnels_wc[0]);
156 	if ((t != NULL) && (t->dev->flags & IFF_UP))
157 		return t;
158 	return NULL;
159 }
160 
161 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
162 		struct ip_tunnel_parm *parms)
163 {
164 	__be32 remote = parms->iph.daddr;
165 	__be32 local = parms->iph.saddr;
166 	unsigned int h = 0;
167 	int prio = 0;
168 
169 	if (remote) {
170 		prio |= 2;
171 		h ^= HASH(remote);
172 	}
173 	if (local) {
174 		prio |= 1;
175 		h ^= HASH(local);
176 	}
177 	return &sitn->tunnels[prio][h];
178 }
179 
180 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
181 		struct ip_tunnel *t)
182 {
183 	return __ipip6_bucket(sitn, &t->parms);
184 }
185 
186 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
187 {
188 	struct ip_tunnel __rcu **tp;
189 	struct ip_tunnel *iter;
190 
191 	for (tp = ipip6_bucket(sitn, t);
192 	     (iter = rtnl_dereference(*tp)) != NULL;
193 	     tp = &iter->next) {
194 		if (t == iter) {
195 			rcu_assign_pointer(*tp, t->next);
196 			break;
197 		}
198 	}
199 }
200 
201 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
202 {
203 	struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
204 
205 	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
206 	rcu_assign_pointer(*tp, t);
207 }
208 
209 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
210 {
211 #ifdef CONFIG_IPV6_SIT_6RD
212 	struct ip_tunnel *t = netdev_priv(dev);
213 
214 	if (t->dev == sitn->fb_tunnel_dev) {
215 		ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
216 		t->ip6rd.relay_prefix = 0;
217 		t->ip6rd.prefixlen = 16;
218 		t->ip6rd.relay_prefixlen = 0;
219 	} else {
220 		struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
221 		memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
222 	}
223 #endif
224 }
225 
226 static int ipip6_tunnel_create(struct net_device *dev)
227 {
228 	struct ip_tunnel *t = netdev_priv(dev);
229 	struct net *net = dev_net(dev);
230 	struct sit_net *sitn = net_generic(net, sit_net_id);
231 	int err;
232 
233 	err = ipip6_tunnel_init(dev);
234 	if (err < 0)
235 		goto out;
236 	ipip6_tunnel_clone_6rd(dev, sitn);
237 
238 	if ((__force u16)t->parms.i_flags & SIT_ISATAP)
239 		dev->priv_flags |= IFF_ISATAP;
240 
241 	err = register_netdevice(dev);
242 	if (err < 0)
243 		goto out;
244 
245 	strcpy(t->parms.name, dev->name);
246 	dev->rtnl_link_ops = &sit_link_ops;
247 
248 	dev_hold(dev);
249 
250 	ipip6_tunnel_link(sitn, t);
251 	return 0;
252 
253 out:
254 	return err;
255 }
256 
257 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
258 		struct ip_tunnel_parm *parms, int create)
259 {
260 	__be32 remote = parms->iph.daddr;
261 	__be32 local = parms->iph.saddr;
262 	struct ip_tunnel *t, *nt;
263 	struct ip_tunnel __rcu **tp;
264 	struct net_device *dev;
265 	char name[IFNAMSIZ];
266 	struct sit_net *sitn = net_generic(net, sit_net_id);
267 
268 	for (tp = __ipip6_bucket(sitn, parms);
269 	    (t = rtnl_dereference(*tp)) != NULL;
270 	     tp = &t->next) {
271 		if (local == t->parms.iph.saddr &&
272 		    remote == t->parms.iph.daddr &&
273 		    parms->link == t->parms.link) {
274 			if (create)
275 				return NULL;
276 			else
277 				return t;
278 		}
279 	}
280 	if (!create)
281 		goto failed;
282 
283 	if (parms->name[0])
284 		strlcpy(name, parms->name, IFNAMSIZ);
285 	else
286 		strcpy(name, "sit%d");
287 
288 	dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
289 	if (dev == NULL)
290 		return NULL;
291 
292 	dev_net_set(dev, net);
293 
294 	nt = netdev_priv(dev);
295 
296 	nt->parms = *parms;
297 	if (ipip6_tunnel_create(dev) < 0)
298 		goto failed_free;
299 
300 	return nt;
301 
302 failed_free:
303 	ipip6_dev_free(dev);
304 failed:
305 	return NULL;
306 }
307 
308 #define for_each_prl_rcu(start)			\
309 	for (prl = rcu_dereference(start);	\
310 	     prl;				\
311 	     prl = rcu_dereference(prl->next))
312 
313 static struct ip_tunnel_prl_entry *
314 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
315 {
316 	struct ip_tunnel_prl_entry *prl;
317 
318 	for_each_prl_rcu(t->prl)
319 		if (prl->addr == addr)
320 			break;
321 	return prl;
322 
323 }
324 
325 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
326 				struct ip_tunnel_prl __user *a)
327 {
328 	struct ip_tunnel_prl kprl, *kp;
329 	struct ip_tunnel_prl_entry *prl;
330 	unsigned int cmax, c = 0, ca, len;
331 	int ret = 0;
332 
333 	if (copy_from_user(&kprl, a, sizeof(kprl)))
334 		return -EFAULT;
335 	cmax = kprl.datalen / sizeof(kprl);
336 	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
337 		cmax = 1;
338 
339 	/* For simple GET or for root users,
340 	 * we try harder to allocate.
341 	 */
342 	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
343 		kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
344 		NULL;
345 
346 	rcu_read_lock();
347 
348 	ca = t->prl_count < cmax ? t->prl_count : cmax;
349 
350 	if (!kp) {
351 		/* We don't try hard to allocate much memory for
352 		 * non-root users.
353 		 * For root users, retry allocating enough memory for
354 		 * the answer.
355 		 */
356 		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
357 		if (!kp) {
358 			ret = -ENOMEM;
359 			goto out;
360 		}
361 	}
362 
363 	c = 0;
364 	for_each_prl_rcu(t->prl) {
365 		if (c >= cmax)
366 			break;
367 		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
368 			continue;
369 		kp[c].addr = prl->addr;
370 		kp[c].flags = prl->flags;
371 		c++;
372 		if (kprl.addr != htonl(INADDR_ANY))
373 			break;
374 	}
375 out:
376 	rcu_read_unlock();
377 
378 	len = sizeof(*kp) * c;
379 	ret = 0;
380 	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
381 		ret = -EFAULT;
382 
383 	kfree(kp);
384 
385 	return ret;
386 }
387 
388 static int
389 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
390 {
391 	struct ip_tunnel_prl_entry *p;
392 	int err = 0;
393 
394 	if (a->addr == htonl(INADDR_ANY))
395 		return -EINVAL;
396 
397 	ASSERT_RTNL();
398 
399 	for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
400 		if (p->addr == a->addr) {
401 			if (chg) {
402 				p->flags = a->flags;
403 				goto out;
404 			}
405 			err = -EEXIST;
406 			goto out;
407 		}
408 	}
409 
410 	if (chg) {
411 		err = -ENXIO;
412 		goto out;
413 	}
414 
415 	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
416 	if (!p) {
417 		err = -ENOBUFS;
418 		goto out;
419 	}
420 
421 	p->next = t->prl;
422 	p->addr = a->addr;
423 	p->flags = a->flags;
424 	t->prl_count++;
425 	rcu_assign_pointer(t->prl, p);
426 out:
427 	return err;
428 }
429 
430 static void prl_list_destroy_rcu(struct rcu_head *head)
431 {
432 	struct ip_tunnel_prl_entry *p, *n;
433 
434 	p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
435 	do {
436 		n = rcu_dereference_protected(p->next, 1);
437 		kfree(p);
438 		p = n;
439 	} while (p);
440 }
441 
442 static int
443 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
444 {
445 	struct ip_tunnel_prl_entry *x;
446 	struct ip_tunnel_prl_entry __rcu **p;
447 	int err = 0;
448 
449 	ASSERT_RTNL();
450 
451 	if (a && a->addr != htonl(INADDR_ANY)) {
452 		for (p = &t->prl;
453 		     (x = rtnl_dereference(*p)) != NULL;
454 		     p = &x->next) {
455 			if (x->addr == a->addr) {
456 				*p = x->next;
457 				kfree_rcu(x, rcu_head);
458 				t->prl_count--;
459 				goto out;
460 			}
461 		}
462 		err = -ENXIO;
463 	} else {
464 		x = rtnl_dereference(t->prl);
465 		if (x) {
466 			t->prl_count = 0;
467 			call_rcu(&x->rcu_head, prl_list_destroy_rcu);
468 			t->prl = NULL;
469 		}
470 	}
471 out:
472 	return err;
473 }
474 
475 static int
476 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
477 {
478 	struct ip_tunnel_prl_entry *p;
479 	int ok = 1;
480 
481 	rcu_read_lock();
482 	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
483 	if (p) {
484 		if (p->flags & PRL_DEFAULT)
485 			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
486 		else
487 			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
488 	} else {
489 		const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
490 
491 		if (ipv6_addr_is_isatap(addr6) &&
492 		    (addr6->s6_addr32[3] == iph->saddr) &&
493 		    ipv6_chk_prefix(addr6, t->dev))
494 			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
495 		else
496 			ok = 0;
497 	}
498 	rcu_read_unlock();
499 	return ok;
500 }
501 
502 static void ipip6_tunnel_uninit(struct net_device *dev)
503 {
504 	struct net *net = dev_net(dev);
505 	struct sit_net *sitn = net_generic(net, sit_net_id);
506 
507 	if (dev == sitn->fb_tunnel_dev) {
508 		RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
509 	} else {
510 		ipip6_tunnel_unlink(sitn, netdev_priv(dev));
511 		ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
512 	}
513 	dev_put(dev);
514 }
515 
516 
517 static int ipip6_err(struct sk_buff *skb, u32 info)
518 {
519 
520 /* All the routers (except for Linux) return only
521    8 bytes of packet payload. It means, that precise relaying of
522    ICMP in the real Internet is absolutely infeasible.
523  */
524 	const struct iphdr *iph = (const struct iphdr *)skb->data;
525 	const int type = icmp_hdr(skb)->type;
526 	const int code = icmp_hdr(skb)->code;
527 	struct ip_tunnel *t;
528 	int err;
529 
530 	switch (type) {
531 	default:
532 	case ICMP_PARAMETERPROB:
533 		return 0;
534 
535 	case ICMP_DEST_UNREACH:
536 		switch (code) {
537 		case ICMP_SR_FAILED:
538 		case ICMP_PORT_UNREACH:
539 			/* Impossible event. */
540 			return 0;
541 		default:
542 			/* All others are translated to HOST_UNREACH.
543 			   rfc2003 contains "deep thoughts" about NET_UNREACH,
544 			   I believe they are just ether pollution. --ANK
545 			 */
546 			break;
547 		}
548 		break;
549 	case ICMP_TIME_EXCEEDED:
550 		if (code != ICMP_EXC_TTL)
551 			return 0;
552 		break;
553 	case ICMP_REDIRECT:
554 		break;
555 	}
556 
557 	err = -ENOENT;
558 
559 	t = ipip6_tunnel_lookup(dev_net(skb->dev),
560 				skb->dev,
561 				iph->daddr,
562 				iph->saddr);
563 	if (t == NULL)
564 		goto out;
565 
566 	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
567 		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
568 				 t->dev->ifindex, 0, IPPROTO_IPV6, 0);
569 		err = 0;
570 		goto out;
571 	}
572 	if (type == ICMP_REDIRECT) {
573 		ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
574 			      IPPROTO_IPV6, 0);
575 		err = 0;
576 		goto out;
577 	}
578 
579 	if (t->parms.iph.daddr == 0)
580 		goto out;
581 
582 	err = 0;
583 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
584 		goto out;
585 
586 	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
587 		t->err_count++;
588 	else
589 		t->err_count = 1;
590 	t->err_time = jiffies;
591 out:
592 	return err;
593 }
594 
595 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
596 				  const struct in6_addr *v6addr)
597 {
598 	__be32 v4embed = 0;
599 	if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
600 		return true;
601 	return false;
602 }
603 
604 static int ipip6_rcv(struct sk_buff *skb)
605 {
606 	const struct iphdr *iph = ip_hdr(skb);
607 	struct ip_tunnel *tunnel;
608 	int err;
609 
610 	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
611 				     iph->saddr, iph->daddr);
612 	if (tunnel != NULL) {
613 		struct pcpu_tstats *tstats;
614 
615 		secpath_reset(skb);
616 		skb->mac_header = skb->network_header;
617 		skb_reset_network_header(skb);
618 		IPCB(skb)->flags = 0;
619 		skb->protocol = htons(ETH_P_IPV6);
620 		skb->pkt_type = PACKET_HOST;
621 
622 		if (tunnel->dev->priv_flags & IFF_ISATAP) {
623 			if (!isatap_chksrc(skb, iph, tunnel)) {
624 				tunnel->dev->stats.rx_errors++;
625 				goto out;
626 			}
627 		} else {
628 			if (is_spoofed_6rd(tunnel, iph->saddr,
629 					   &ipv6_hdr(skb)->saddr) ||
630 			    is_spoofed_6rd(tunnel, iph->daddr,
631 					   &ipv6_hdr(skb)->daddr)) {
632 				tunnel->dev->stats.rx_errors++;
633 				goto out;
634 			}
635 		}
636 
637 		__skb_tunnel_rx(skb, tunnel->dev);
638 
639 		err = IP_ECN_decapsulate(iph, skb);
640 		if (unlikely(err)) {
641 			if (log_ecn_error)
642 				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
643 						     &iph->saddr, iph->tos);
644 			if (err > 1) {
645 				++tunnel->dev->stats.rx_frame_errors;
646 				++tunnel->dev->stats.rx_errors;
647 				goto out;
648 			}
649 		}
650 
651 		tstats = this_cpu_ptr(tunnel->dev->tstats);
652 		tstats->rx_packets++;
653 		tstats->rx_bytes += skb->len;
654 
655 		netif_rx(skb);
656 
657 		return 0;
658 	}
659 
660 	/* no tunnel matched,  let upstream know, ipsec may handle it */
661 	return 1;
662 out:
663 	kfree_skb(skb);
664 	return 0;
665 }
666 
667 /*
668  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
669  * stores the embedded IPv4 address in v4dst and returns true.
670  */
671 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
672 		      __be32 *v4dst)
673 {
674 #ifdef CONFIG_IPV6_SIT_6RD
675 	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
676 			      tunnel->ip6rd.prefixlen)) {
677 		unsigned int pbw0, pbi0;
678 		int pbi1;
679 		u32 d;
680 
681 		pbw0 = tunnel->ip6rd.prefixlen >> 5;
682 		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
683 
684 		d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
685 		    tunnel->ip6rd.relay_prefixlen;
686 
687 		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
688 		if (pbi1 > 0)
689 			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
690 			     (32 - pbi1);
691 
692 		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
693 		return true;
694 	}
695 #else
696 	if (v6dst->s6_addr16[0] == htons(0x2002)) {
697 		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
698 		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
699 		return true;
700 	}
701 #endif
702 	return false;
703 }
704 
705 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
706 			     const struct in6_addr *v6dst)
707 {
708 	__be32 dst = 0;
709 	check_6rd(tunnel, v6dst, &dst);
710 	return dst;
711 }
712 
713 /*
714  *	This function assumes it is being called from dev_queue_xmit()
715  *	and that skb is filled properly by that function.
716  */
717 
718 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
719 				     struct net_device *dev)
720 {
721 	struct ip_tunnel *tunnel = netdev_priv(dev);
722 	const struct iphdr  *tiph = &tunnel->parms.iph;
723 	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
724 	u8     tos = tunnel->parms.iph.tos;
725 	__be16 df = tiph->frag_off;
726 	struct rtable *rt;     			/* Route to the other host */
727 	struct net_device *tdev;		/* Device to other host */
728 	struct iphdr  *iph;			/* Our new IP header */
729 	unsigned int max_headroom;		/* The extra header space needed */
730 	__be32 dst = tiph->daddr;
731 	struct flowi4 fl4;
732 	int    mtu;
733 	const struct in6_addr *addr6;
734 	int addr_type;
735 
736 	if (skb->protocol != htons(ETH_P_IPV6))
737 		goto tx_error;
738 
739 	if (tos == 1)
740 		tos = ipv6_get_dsfield(iph6);
741 
742 	/* ISATAP (RFC4214) - must come before 6to4 */
743 	if (dev->priv_flags & IFF_ISATAP) {
744 		struct neighbour *neigh = NULL;
745 		bool do_tx_error = false;
746 
747 		if (skb_dst(skb))
748 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
749 
750 		if (neigh == NULL) {
751 			net_dbg_ratelimited("sit: nexthop == NULL\n");
752 			goto tx_error;
753 		}
754 
755 		addr6 = (const struct in6_addr *)&neigh->primary_key;
756 		addr_type = ipv6_addr_type(addr6);
757 
758 		if ((addr_type & IPV6_ADDR_UNICAST) &&
759 		     ipv6_addr_is_isatap(addr6))
760 			dst = addr6->s6_addr32[3];
761 		else
762 			do_tx_error = true;
763 
764 		neigh_release(neigh);
765 		if (do_tx_error)
766 			goto tx_error;
767 	}
768 
769 	if (!dst)
770 		dst = try_6rd(tunnel, &iph6->daddr);
771 
772 	if (!dst) {
773 		struct neighbour *neigh = NULL;
774 		bool do_tx_error = false;
775 
776 		if (skb_dst(skb))
777 			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
778 
779 		if (neigh == NULL) {
780 			net_dbg_ratelimited("sit: nexthop == NULL\n");
781 			goto tx_error;
782 		}
783 
784 		addr6 = (const struct in6_addr *)&neigh->primary_key;
785 		addr_type = ipv6_addr_type(addr6);
786 
787 		if (addr_type == IPV6_ADDR_ANY) {
788 			addr6 = &ipv6_hdr(skb)->daddr;
789 			addr_type = ipv6_addr_type(addr6);
790 		}
791 
792 		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
793 			dst = addr6->s6_addr32[3];
794 		else
795 			do_tx_error = true;
796 
797 		neigh_release(neigh);
798 		if (do_tx_error)
799 			goto tx_error;
800 	}
801 
802 	rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
803 				   dst, tiph->saddr,
804 				   0, 0,
805 				   IPPROTO_IPV6, RT_TOS(tos),
806 				   tunnel->parms.link);
807 	if (IS_ERR(rt)) {
808 		dev->stats.tx_carrier_errors++;
809 		goto tx_error_icmp;
810 	}
811 	if (rt->rt_type != RTN_UNICAST) {
812 		ip_rt_put(rt);
813 		dev->stats.tx_carrier_errors++;
814 		goto tx_error_icmp;
815 	}
816 	tdev = rt->dst.dev;
817 
818 	if (tdev == dev) {
819 		ip_rt_put(rt);
820 		dev->stats.collisions++;
821 		goto tx_error;
822 	}
823 
824 	if (df) {
825 		mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
826 
827 		if (mtu < 68) {
828 			dev->stats.collisions++;
829 			ip_rt_put(rt);
830 			goto tx_error;
831 		}
832 
833 		if (mtu < IPV6_MIN_MTU) {
834 			mtu = IPV6_MIN_MTU;
835 			df = 0;
836 		}
837 
838 		if (tunnel->parms.iph.daddr && skb_dst(skb))
839 			skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
840 
841 		if (skb->len > mtu) {
842 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
843 			ip_rt_put(rt);
844 			goto tx_error;
845 		}
846 	}
847 
848 	if (tunnel->err_count > 0) {
849 		if (time_before(jiffies,
850 				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
851 			tunnel->err_count--;
852 			dst_link_failure(skb);
853 		} else
854 			tunnel->err_count = 0;
855 	}
856 
857 	/*
858 	 * Okay, now see if we can stuff it in the buffer as-is.
859 	 */
860 	max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
861 
862 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
863 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
864 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
865 		if (!new_skb) {
866 			ip_rt_put(rt);
867 			dev->stats.tx_dropped++;
868 			dev_kfree_skb(skb);
869 			return NETDEV_TX_OK;
870 		}
871 		if (skb->sk)
872 			skb_set_owner_w(new_skb, skb->sk);
873 		dev_kfree_skb(skb);
874 		skb = new_skb;
875 		iph6 = ipv6_hdr(skb);
876 	}
877 
878 	skb->transport_header = skb->network_header;
879 	skb_push(skb, sizeof(struct iphdr));
880 	skb_reset_network_header(skb);
881 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
882 	IPCB(skb)->flags = 0;
883 	skb_dst_drop(skb);
884 	skb_dst_set(skb, &rt->dst);
885 
886 	/*
887 	 *	Push down and install the IPIP header.
888 	 */
889 
890 	iph 			=	ip_hdr(skb);
891 	iph->version		=	4;
892 	iph->ihl		=	sizeof(struct iphdr)>>2;
893 	iph->frag_off		=	df;
894 	iph->protocol		=	IPPROTO_IPV6;
895 	iph->tos		=	INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
896 	iph->daddr		=	fl4.daddr;
897 	iph->saddr		=	fl4.saddr;
898 
899 	if ((iph->ttl = tiph->ttl) == 0)
900 		iph->ttl	=	iph6->hop_limit;
901 
902 	iptunnel_xmit(skb, dev);
903 	return NETDEV_TX_OK;
904 
905 tx_error_icmp:
906 	dst_link_failure(skb);
907 tx_error:
908 	dev->stats.tx_errors++;
909 	dev_kfree_skb(skb);
910 	return NETDEV_TX_OK;
911 }
912 
913 static void ipip6_tunnel_bind_dev(struct net_device *dev)
914 {
915 	struct net_device *tdev = NULL;
916 	struct ip_tunnel *tunnel;
917 	const struct iphdr *iph;
918 	struct flowi4 fl4;
919 
920 	tunnel = netdev_priv(dev);
921 	iph = &tunnel->parms.iph;
922 
923 	if (iph->daddr) {
924 		struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
925 							  iph->daddr, iph->saddr,
926 							  0, 0,
927 							  IPPROTO_IPV6,
928 							  RT_TOS(iph->tos),
929 							  tunnel->parms.link);
930 
931 		if (!IS_ERR(rt)) {
932 			tdev = rt->dst.dev;
933 			ip_rt_put(rt);
934 		}
935 		dev->flags |= IFF_POINTOPOINT;
936 	}
937 
938 	if (!tdev && tunnel->parms.link)
939 		tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
940 
941 	if (tdev) {
942 		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
943 		dev->mtu = tdev->mtu - sizeof(struct iphdr);
944 		if (dev->mtu < IPV6_MIN_MTU)
945 			dev->mtu = IPV6_MIN_MTU;
946 	}
947 	dev->iflink = tunnel->parms.link;
948 }
949 
950 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
951 {
952 	struct net *net = dev_net(t->dev);
953 	struct sit_net *sitn = net_generic(net, sit_net_id);
954 
955 	ipip6_tunnel_unlink(sitn, t);
956 	synchronize_net();
957 	t->parms.iph.saddr = p->iph.saddr;
958 	t->parms.iph.daddr = p->iph.daddr;
959 	memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
960 	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
961 	ipip6_tunnel_link(sitn, t);
962 	t->parms.iph.ttl = p->iph.ttl;
963 	t->parms.iph.tos = p->iph.tos;
964 	if (t->parms.link != p->link) {
965 		t->parms.link = p->link;
966 		ipip6_tunnel_bind_dev(t->dev);
967 	}
968 	netdev_state_change(t->dev);
969 }
970 
971 #ifdef CONFIG_IPV6_SIT_6RD
972 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
973 				   struct ip_tunnel_6rd *ip6rd)
974 {
975 	struct in6_addr prefix;
976 	__be32 relay_prefix;
977 
978 	if (ip6rd->relay_prefixlen > 32 ||
979 	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
980 		return -EINVAL;
981 
982 	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
983 	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
984 		return -EINVAL;
985 	if (ip6rd->relay_prefixlen)
986 		relay_prefix = ip6rd->relay_prefix &
987 			       htonl(0xffffffffUL <<
988 				     (32 - ip6rd->relay_prefixlen));
989 	else
990 		relay_prefix = 0;
991 	if (relay_prefix != ip6rd->relay_prefix)
992 		return -EINVAL;
993 
994 	t->ip6rd.prefix = prefix;
995 	t->ip6rd.relay_prefix = relay_prefix;
996 	t->ip6rd.prefixlen = ip6rd->prefixlen;
997 	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
998 	netdev_state_change(t->dev);
999 	return 0;
1000 }
1001 #endif
1002 
1003 static int
1004 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1005 {
1006 	int err = 0;
1007 	struct ip_tunnel_parm p;
1008 	struct ip_tunnel_prl prl;
1009 	struct ip_tunnel *t;
1010 	struct net *net = dev_net(dev);
1011 	struct sit_net *sitn = net_generic(net, sit_net_id);
1012 #ifdef CONFIG_IPV6_SIT_6RD
1013 	struct ip_tunnel_6rd ip6rd;
1014 #endif
1015 
1016 	switch (cmd) {
1017 	case SIOCGETTUNNEL:
1018 #ifdef CONFIG_IPV6_SIT_6RD
1019 	case SIOCGET6RD:
1020 #endif
1021 		t = NULL;
1022 		if (dev == sitn->fb_tunnel_dev) {
1023 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1024 				err = -EFAULT;
1025 				break;
1026 			}
1027 			t = ipip6_tunnel_locate(net, &p, 0);
1028 		}
1029 		if (t == NULL)
1030 			t = netdev_priv(dev);
1031 
1032 		err = -EFAULT;
1033 		if (cmd == SIOCGETTUNNEL) {
1034 			memcpy(&p, &t->parms, sizeof(p));
1035 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1036 					 sizeof(p)))
1037 				goto done;
1038 #ifdef CONFIG_IPV6_SIT_6RD
1039 		} else {
1040 			ip6rd.prefix = t->ip6rd.prefix;
1041 			ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1042 			ip6rd.prefixlen = t->ip6rd.prefixlen;
1043 			ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1044 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1045 					 sizeof(ip6rd)))
1046 				goto done;
1047 #endif
1048 		}
1049 		err = 0;
1050 		break;
1051 
1052 	case SIOCADDTUNNEL:
1053 	case SIOCCHGTUNNEL:
1054 		err = -EPERM;
1055 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1056 			goto done;
1057 
1058 		err = -EFAULT;
1059 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1060 			goto done;
1061 
1062 		err = -EINVAL;
1063 		if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
1064 		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1065 			goto done;
1066 		if (p.iph.ttl)
1067 			p.iph.frag_off |= htons(IP_DF);
1068 
1069 		t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1070 
1071 		if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1072 			if (t != NULL) {
1073 				if (t->dev != dev) {
1074 					err = -EEXIST;
1075 					break;
1076 				}
1077 			} else {
1078 				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1079 				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1080 					err = -EINVAL;
1081 					break;
1082 				}
1083 				t = netdev_priv(dev);
1084 			}
1085 
1086 			ipip6_tunnel_update(t, &p);
1087 		}
1088 
1089 		if (t) {
1090 			err = 0;
1091 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1092 				err = -EFAULT;
1093 		} else
1094 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1095 		break;
1096 
1097 	case SIOCDELTUNNEL:
1098 		err = -EPERM;
1099 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1100 			goto done;
1101 
1102 		if (dev == sitn->fb_tunnel_dev) {
1103 			err = -EFAULT;
1104 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1105 				goto done;
1106 			err = -ENOENT;
1107 			if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1108 				goto done;
1109 			err = -EPERM;
1110 			if (t == netdev_priv(sitn->fb_tunnel_dev))
1111 				goto done;
1112 			dev = t->dev;
1113 		}
1114 		unregister_netdevice(dev);
1115 		err = 0;
1116 		break;
1117 
1118 	case SIOCGETPRL:
1119 		err = -EINVAL;
1120 		if (dev == sitn->fb_tunnel_dev)
1121 			goto done;
1122 		err = -ENOENT;
1123 		if (!(t = netdev_priv(dev)))
1124 			goto done;
1125 		err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1126 		break;
1127 
1128 	case SIOCADDPRL:
1129 	case SIOCDELPRL:
1130 	case SIOCCHGPRL:
1131 		err = -EPERM;
1132 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1133 			goto done;
1134 		err = -EINVAL;
1135 		if (dev == sitn->fb_tunnel_dev)
1136 			goto done;
1137 		err = -EFAULT;
1138 		if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1139 			goto done;
1140 		err = -ENOENT;
1141 		if (!(t = netdev_priv(dev)))
1142 			goto done;
1143 
1144 		switch (cmd) {
1145 		case SIOCDELPRL:
1146 			err = ipip6_tunnel_del_prl(t, &prl);
1147 			break;
1148 		case SIOCADDPRL:
1149 		case SIOCCHGPRL:
1150 			err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1151 			break;
1152 		}
1153 		netdev_state_change(dev);
1154 		break;
1155 
1156 #ifdef CONFIG_IPV6_SIT_6RD
1157 	case SIOCADD6RD:
1158 	case SIOCCHG6RD:
1159 	case SIOCDEL6RD:
1160 		err = -EPERM;
1161 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1162 			goto done;
1163 
1164 		err = -EFAULT;
1165 		if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1166 				   sizeof(ip6rd)))
1167 			goto done;
1168 
1169 		t = netdev_priv(dev);
1170 
1171 		if (cmd != SIOCDEL6RD) {
1172 			err = ipip6_tunnel_update_6rd(t, &ip6rd);
1173 			if (err < 0)
1174 				goto done;
1175 		} else
1176 			ipip6_tunnel_clone_6rd(dev, sitn);
1177 
1178 		err = 0;
1179 		break;
1180 #endif
1181 
1182 	default:
1183 		err = -EINVAL;
1184 	}
1185 
1186 done:
1187 	return err;
1188 }
1189 
1190 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1191 {
1192 	if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1193 		return -EINVAL;
1194 	dev->mtu = new_mtu;
1195 	return 0;
1196 }
1197 
1198 static const struct net_device_ops ipip6_netdev_ops = {
1199 	.ndo_uninit	= ipip6_tunnel_uninit,
1200 	.ndo_start_xmit	= ipip6_tunnel_xmit,
1201 	.ndo_do_ioctl	= ipip6_tunnel_ioctl,
1202 	.ndo_change_mtu	= ipip6_tunnel_change_mtu,
1203 	.ndo_get_stats64= ipip6_get_stats64,
1204 };
1205 
1206 static void ipip6_dev_free(struct net_device *dev)
1207 {
1208 	free_percpu(dev->tstats);
1209 	free_netdev(dev);
1210 }
1211 
1212 static void ipip6_tunnel_setup(struct net_device *dev)
1213 {
1214 	dev->netdev_ops		= &ipip6_netdev_ops;
1215 	dev->destructor 	= ipip6_dev_free;
1216 
1217 	dev->type		= ARPHRD_SIT;
1218 	dev->hard_header_len 	= LL_MAX_HEADER + sizeof(struct iphdr);
1219 	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr);
1220 	dev->flags		= IFF_NOARP;
1221 	dev->priv_flags	       &= ~IFF_XMIT_DST_RELEASE;
1222 	dev->iflink		= 0;
1223 	dev->addr_len		= 4;
1224 	dev->features		|= NETIF_F_NETNS_LOCAL;
1225 	dev->features		|= NETIF_F_LLTX;
1226 }
1227 
1228 static int ipip6_tunnel_init(struct net_device *dev)
1229 {
1230 	struct ip_tunnel *tunnel = netdev_priv(dev);
1231 
1232 	tunnel->dev = dev;
1233 
1234 	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1235 	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1236 
1237 	ipip6_tunnel_bind_dev(dev);
1238 	dev->tstats = alloc_percpu(struct pcpu_tstats);
1239 	if (!dev->tstats)
1240 		return -ENOMEM;
1241 
1242 	return 0;
1243 }
1244 
1245 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1246 {
1247 	struct ip_tunnel *tunnel = netdev_priv(dev);
1248 	struct iphdr *iph = &tunnel->parms.iph;
1249 	struct net *net = dev_net(dev);
1250 	struct sit_net *sitn = net_generic(net, sit_net_id);
1251 
1252 	tunnel->dev = dev;
1253 	strcpy(tunnel->parms.name, dev->name);
1254 
1255 	iph->version		= 4;
1256 	iph->protocol		= IPPROTO_IPV6;
1257 	iph->ihl		= 5;
1258 	iph->ttl		= 64;
1259 
1260 	dev->tstats = alloc_percpu(struct pcpu_tstats);
1261 	if (!dev->tstats)
1262 		return -ENOMEM;
1263 	dev_hold(dev);
1264 	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1265 	return 0;
1266 }
1267 
1268 static void ipip6_netlink_parms(struct nlattr *data[],
1269 				struct ip_tunnel_parm *parms)
1270 {
1271 	memset(parms, 0, sizeof(*parms));
1272 
1273 	parms->iph.version = 4;
1274 	parms->iph.protocol = IPPROTO_IPV6;
1275 	parms->iph.ihl = 5;
1276 	parms->iph.ttl = 64;
1277 
1278 	if (!data)
1279 		return;
1280 
1281 	if (data[IFLA_IPTUN_LINK])
1282 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1283 
1284 	if (data[IFLA_IPTUN_LOCAL])
1285 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1286 
1287 	if (data[IFLA_IPTUN_REMOTE])
1288 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1289 
1290 	if (data[IFLA_IPTUN_TTL]) {
1291 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1292 		if (parms->iph.ttl)
1293 			parms->iph.frag_off = htons(IP_DF);
1294 	}
1295 
1296 	if (data[IFLA_IPTUN_TOS])
1297 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1298 
1299 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1300 		parms->iph.frag_off = htons(IP_DF);
1301 
1302 	if (data[IFLA_IPTUN_FLAGS])
1303 		parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1304 }
1305 
1306 #ifdef CONFIG_IPV6_SIT_6RD
1307 /* This function returns true when 6RD attributes are present in the nl msg */
1308 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1309 				    struct ip_tunnel_6rd *ip6rd)
1310 {
1311 	bool ret = false;
1312 	memset(ip6rd, 0, sizeof(*ip6rd));
1313 
1314 	if (!data)
1315 		return ret;
1316 
1317 	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1318 		ret = true;
1319 		nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1320 			   sizeof(struct in6_addr));
1321 	}
1322 
1323 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1324 		ret = true;
1325 		ip6rd->relay_prefix =
1326 			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1327 	}
1328 
1329 	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1330 		ret = true;
1331 		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1332 	}
1333 
1334 	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1335 		ret = true;
1336 		ip6rd->relay_prefixlen =
1337 			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1338 	}
1339 
1340 	return ret;
1341 }
1342 #endif
1343 
1344 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1345 			 struct nlattr *tb[], struct nlattr *data[])
1346 {
1347 	struct net *net = dev_net(dev);
1348 	struct ip_tunnel *nt;
1349 #ifdef CONFIG_IPV6_SIT_6RD
1350 	struct ip_tunnel_6rd ip6rd;
1351 #endif
1352 	int err;
1353 
1354 	nt = netdev_priv(dev);
1355 	ipip6_netlink_parms(data, &nt->parms);
1356 
1357 	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1358 		return -EEXIST;
1359 
1360 	err = ipip6_tunnel_create(dev);
1361 	if (err < 0)
1362 		return err;
1363 
1364 #ifdef CONFIG_IPV6_SIT_6RD
1365 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1366 		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1367 #endif
1368 
1369 	return err;
1370 }
1371 
1372 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1373 			  struct nlattr *data[])
1374 {
1375 	struct ip_tunnel *t;
1376 	struct ip_tunnel_parm p;
1377 	struct net *net = dev_net(dev);
1378 	struct sit_net *sitn = net_generic(net, sit_net_id);
1379 #ifdef CONFIG_IPV6_SIT_6RD
1380 	struct ip_tunnel_6rd ip6rd;
1381 #endif
1382 
1383 	if (dev == sitn->fb_tunnel_dev)
1384 		return -EINVAL;
1385 
1386 	ipip6_netlink_parms(data, &p);
1387 
1388 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1389 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1390 		return -EINVAL;
1391 
1392 	t = ipip6_tunnel_locate(net, &p, 0);
1393 
1394 	if (t) {
1395 		if (t->dev != dev)
1396 			return -EEXIST;
1397 	} else
1398 		t = netdev_priv(dev);
1399 
1400 	ipip6_tunnel_update(t, &p);
1401 
1402 #ifdef CONFIG_IPV6_SIT_6RD
1403 	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1404 		return ipip6_tunnel_update_6rd(t, &ip6rd);
1405 #endif
1406 
1407 	return 0;
1408 }
1409 
1410 static size_t ipip6_get_size(const struct net_device *dev)
1411 {
1412 	return
1413 		/* IFLA_IPTUN_LINK */
1414 		nla_total_size(4) +
1415 		/* IFLA_IPTUN_LOCAL */
1416 		nla_total_size(4) +
1417 		/* IFLA_IPTUN_REMOTE */
1418 		nla_total_size(4) +
1419 		/* IFLA_IPTUN_TTL */
1420 		nla_total_size(1) +
1421 		/* IFLA_IPTUN_TOS */
1422 		nla_total_size(1) +
1423 		/* IFLA_IPTUN_PMTUDISC */
1424 		nla_total_size(1) +
1425 		/* IFLA_IPTUN_FLAGS */
1426 		nla_total_size(2) +
1427 #ifdef CONFIG_IPV6_SIT_6RD
1428 		/* IFLA_IPTUN_6RD_PREFIX */
1429 		nla_total_size(sizeof(struct in6_addr)) +
1430 		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1431 		nla_total_size(4) +
1432 		/* IFLA_IPTUN_6RD_PREFIXLEN */
1433 		nla_total_size(2) +
1434 		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1435 		nla_total_size(2) +
1436 #endif
1437 		0;
1438 }
1439 
1440 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1441 {
1442 	struct ip_tunnel *tunnel = netdev_priv(dev);
1443 	struct ip_tunnel_parm *parm = &tunnel->parms;
1444 
1445 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1446 	    nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1447 	    nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1448 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1449 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1450 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1451 		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1452 	    nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1453 		goto nla_put_failure;
1454 
1455 #ifdef CONFIG_IPV6_SIT_6RD
1456 	if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1457 		    &tunnel->ip6rd.prefix) ||
1458 	    nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1459 			 tunnel->ip6rd.relay_prefix) ||
1460 	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1461 			tunnel->ip6rd.prefixlen) ||
1462 	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1463 			tunnel->ip6rd.relay_prefixlen))
1464 		goto nla_put_failure;
1465 #endif
1466 
1467 	return 0;
1468 
1469 nla_put_failure:
1470 	return -EMSGSIZE;
1471 }
1472 
1473 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1474 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1475 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1476 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1477 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1478 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1479 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1480 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1481 #ifdef CONFIG_IPV6_SIT_6RD
1482 	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1483 	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1484 	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1485 	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1486 #endif
1487 };
1488 
1489 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1490 	.kind		= "sit",
1491 	.maxtype	= IFLA_IPTUN_MAX,
1492 	.policy		= ipip6_policy,
1493 	.priv_size	= sizeof(struct ip_tunnel),
1494 	.setup		= ipip6_tunnel_setup,
1495 	.newlink	= ipip6_newlink,
1496 	.changelink	= ipip6_changelink,
1497 	.get_size	= ipip6_get_size,
1498 	.fill_info	= ipip6_fill_info,
1499 };
1500 
1501 static struct xfrm_tunnel sit_handler __read_mostly = {
1502 	.handler	=	ipip6_rcv,
1503 	.err_handler	=	ipip6_err,
1504 	.priority	=	1,
1505 };
1506 
1507 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1508 {
1509 	int prio;
1510 
1511 	for (prio = 1; prio < 4; prio++) {
1512 		int h;
1513 		for (h = 0; h < HASH_SIZE; h++) {
1514 			struct ip_tunnel *t;
1515 
1516 			t = rtnl_dereference(sitn->tunnels[prio][h]);
1517 			while (t != NULL) {
1518 				unregister_netdevice_queue(t->dev, head);
1519 				t = rtnl_dereference(t->next);
1520 			}
1521 		}
1522 	}
1523 }
1524 
1525 static int __net_init sit_init_net(struct net *net)
1526 {
1527 	struct sit_net *sitn = net_generic(net, sit_net_id);
1528 	struct ip_tunnel *t;
1529 	int err;
1530 
1531 	sitn->tunnels[0] = sitn->tunnels_wc;
1532 	sitn->tunnels[1] = sitn->tunnels_l;
1533 	sitn->tunnels[2] = sitn->tunnels_r;
1534 	sitn->tunnels[3] = sitn->tunnels_r_l;
1535 
1536 	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1537 					   ipip6_tunnel_setup);
1538 	if (!sitn->fb_tunnel_dev) {
1539 		err = -ENOMEM;
1540 		goto err_alloc_dev;
1541 	}
1542 	dev_net_set(sitn->fb_tunnel_dev, net);
1543 
1544 	err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1545 	if (err)
1546 		goto err_dev_free;
1547 
1548 	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1549 
1550 	if ((err = register_netdev(sitn->fb_tunnel_dev)))
1551 		goto err_reg_dev;
1552 
1553 	t = netdev_priv(sitn->fb_tunnel_dev);
1554 
1555 	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1556 	return 0;
1557 
1558 err_reg_dev:
1559 	dev_put(sitn->fb_tunnel_dev);
1560 err_dev_free:
1561 	ipip6_dev_free(sitn->fb_tunnel_dev);
1562 err_alloc_dev:
1563 	return err;
1564 }
1565 
1566 static void __net_exit sit_exit_net(struct net *net)
1567 {
1568 	struct sit_net *sitn = net_generic(net, sit_net_id);
1569 	LIST_HEAD(list);
1570 
1571 	rtnl_lock();
1572 	sit_destroy_tunnels(sitn, &list);
1573 	unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1574 	unregister_netdevice_many(&list);
1575 	rtnl_unlock();
1576 }
1577 
1578 static struct pernet_operations sit_net_ops = {
1579 	.init = sit_init_net,
1580 	.exit = sit_exit_net,
1581 	.id   = &sit_net_id,
1582 	.size = sizeof(struct sit_net),
1583 };
1584 
1585 static void __exit sit_cleanup(void)
1586 {
1587 	rtnl_link_unregister(&sit_link_ops);
1588 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1589 
1590 	unregister_pernet_device(&sit_net_ops);
1591 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1592 }
1593 
1594 static int __init sit_init(void)
1595 {
1596 	int err;
1597 
1598 	pr_info("IPv6 over IPv4 tunneling driver\n");
1599 
1600 	err = register_pernet_device(&sit_net_ops);
1601 	if (err < 0)
1602 		return err;
1603 	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1604 	if (err < 0) {
1605 		pr_info("%s: can't add protocol\n", __func__);
1606 		goto xfrm_tunnel_failed;
1607 	}
1608 	err = rtnl_link_register(&sit_link_ops);
1609 	if (err < 0)
1610 		goto rtnl_link_failed;
1611 
1612 out:
1613 	return err;
1614 
1615 rtnl_link_failed:
1616 	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1617 xfrm_tunnel_failed:
1618 	unregister_pernet_device(&sit_net_ops);
1619 	goto out;
1620 }
1621 
1622 module_init(sit_init);
1623 module_exit(sit_cleanup);
1624 MODULE_LICENSE("GPL");
1625 MODULE_ALIAS_NETDEV("sit0");
1626