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/ip_tunnels.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
68static bool log_ecn_error = true;
69module_param(log_ecn_error, bool, 0644);
70MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
71
72static int ipip6_tunnel_init(struct net_device *dev);
73static void ipip6_tunnel_setup(struct net_device *dev);
74static void ipip6_dev_free(struct net_device *dev);
75static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
76		      __be32 *v4dst);
77static struct rtnl_link_ops sit_link_ops __read_mostly;
78
79static int sit_net_id __read_mostly;
80struct 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/*
91 * Must be invoked with rcu_read_lock
92 */
93static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
94		struct net_device *dev, __be32 remote, __be32 local)
95{
96	unsigned int h0 = HASH(remote);
97	unsigned int h1 = HASH(local);
98	struct ip_tunnel *t;
99	struct sit_net *sitn = net_generic(net, sit_net_id);
100
101	for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
102		if (local == t->parms.iph.saddr &&
103		    remote == t->parms.iph.daddr &&
104		    (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
105		    (t->dev->flags & IFF_UP))
106			return t;
107	}
108	for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
109		if (remote == t->parms.iph.daddr &&
110		    (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
111		    (t->dev->flags & IFF_UP))
112			return t;
113	}
114	for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
115		if (local == t->parms.iph.saddr &&
116		    (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
117		    (t->dev->flags & IFF_UP))
118			return t;
119	}
120	t = rcu_dereference(sitn->tunnels_wc[0]);
121	if (t && (t->dev->flags & IFF_UP))
122		return t;
123	return NULL;
124}
125
126static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
127		struct ip_tunnel_parm *parms)
128{
129	__be32 remote = parms->iph.daddr;
130	__be32 local = parms->iph.saddr;
131	unsigned int h = 0;
132	int prio = 0;
133
134	if (remote) {
135		prio |= 2;
136		h ^= HASH(remote);
137	}
138	if (local) {
139		prio |= 1;
140		h ^= HASH(local);
141	}
142	return &sitn->tunnels[prio][h];
143}
144
145static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
146		struct ip_tunnel *t)
147{
148	return __ipip6_bucket(sitn, &t->parms);
149}
150
151static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
152{
153	struct ip_tunnel __rcu **tp;
154	struct ip_tunnel *iter;
155
156	for (tp = ipip6_bucket(sitn, t);
157	     (iter = rtnl_dereference(*tp)) != NULL;
158	     tp = &iter->next) {
159		if (t == iter) {
160			rcu_assign_pointer(*tp, t->next);
161			break;
162		}
163	}
164}
165
166static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
167{
168	struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
169
170	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
171	rcu_assign_pointer(*tp, t);
172}
173
174static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
175{
176#ifdef CONFIG_IPV6_SIT_6RD
177	struct ip_tunnel *t = netdev_priv(dev);
178
179	if (t->dev == sitn->fb_tunnel_dev) {
180		ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
181		t->ip6rd.relay_prefix = 0;
182		t->ip6rd.prefixlen = 16;
183		t->ip6rd.relay_prefixlen = 0;
184	} else {
185		struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
186		memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
187	}
188#endif
189}
190
191static int ipip6_tunnel_create(struct net_device *dev)
192{
193	struct ip_tunnel *t = netdev_priv(dev);
194	struct net *net = dev_net(dev);
195	struct sit_net *sitn = net_generic(net, sit_net_id);
196	int err;
197
198	memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
199	memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
200
201	if ((__force u16)t->parms.i_flags & SIT_ISATAP)
202		dev->priv_flags |= IFF_ISATAP;
203
204	err = register_netdevice(dev);
205	if (err < 0)
206		goto out;
207
208	ipip6_tunnel_clone_6rd(dev, sitn);
209
210	dev->rtnl_link_ops = &sit_link_ops;
211
212	dev_hold(dev);
213
214	ipip6_tunnel_link(sitn, t);
215	return 0;
216
217out:
218	return err;
219}
220
221static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
222		struct ip_tunnel_parm *parms, int create)
223{
224	__be32 remote = parms->iph.daddr;
225	__be32 local = parms->iph.saddr;
226	struct ip_tunnel *t, *nt;
227	struct ip_tunnel __rcu **tp;
228	struct net_device *dev;
229	char name[IFNAMSIZ];
230	struct sit_net *sitn = net_generic(net, sit_net_id);
231
232	for (tp = __ipip6_bucket(sitn, parms);
233	    (t = rtnl_dereference(*tp)) != NULL;
234	     tp = &t->next) {
235		if (local == t->parms.iph.saddr &&
236		    remote == t->parms.iph.daddr &&
237		    parms->link == t->parms.link) {
238			if (create)
239				return NULL;
240			else
241				return t;
242		}
243	}
244	if (!create)
245		goto failed;
246
247	if (parms->name[0])
248		strlcpy(name, parms->name, IFNAMSIZ);
249	else
250		strcpy(name, "sit%d");
251
252	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
253			   ipip6_tunnel_setup);
254	if (!dev)
255		return NULL;
256
257	dev_net_set(dev, net);
258
259	nt = netdev_priv(dev);
260
261	nt->parms = *parms;
262	if (ipip6_tunnel_create(dev) < 0)
263		goto failed_free;
264
265	return nt;
266
267failed_free:
268	ipip6_dev_free(dev);
269failed:
270	return NULL;
271}
272
273#define for_each_prl_rcu(start)			\
274	for (prl = rcu_dereference(start);	\
275	     prl;				\
276	     prl = rcu_dereference(prl->next))
277
278static struct ip_tunnel_prl_entry *
279__ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
280{
281	struct ip_tunnel_prl_entry *prl;
282
283	for_each_prl_rcu(t->prl)
284		if (prl->addr == addr)
285			break;
286	return prl;
287
288}
289
290static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
291				struct ip_tunnel_prl __user *a)
292{
293	struct ip_tunnel_prl kprl, *kp;
294	struct ip_tunnel_prl_entry *prl;
295	unsigned int cmax, c = 0, ca, len;
296	int ret = 0;
297
298	if (copy_from_user(&kprl, a, sizeof(kprl)))
299		return -EFAULT;
300	cmax = kprl.datalen / sizeof(kprl);
301	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
302		cmax = 1;
303
304	/* For simple GET or for root users,
305	 * we try harder to allocate.
306	 */
307	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
308		kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
309		NULL;
310
311	rcu_read_lock();
312
313	ca = t->prl_count < cmax ? t->prl_count : cmax;
314
315	if (!kp) {
316		/* We don't try hard to allocate much memory for
317		 * non-root users.
318		 * For root users, retry allocating enough memory for
319		 * the answer.
320		 */
321		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
322		if (!kp) {
323			ret = -ENOMEM;
324			goto out;
325		}
326	}
327
328	c = 0;
329	for_each_prl_rcu(t->prl) {
330		if (c >= cmax)
331			break;
332		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
333			continue;
334		kp[c].addr = prl->addr;
335		kp[c].flags = prl->flags;
336		c++;
337		if (kprl.addr != htonl(INADDR_ANY))
338			break;
339	}
340out:
341	rcu_read_unlock();
342
343	len = sizeof(*kp) * c;
344	ret = 0;
345	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
346		ret = -EFAULT;
347
348	kfree(kp);
349
350	return ret;
351}
352
353static int
354ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
355{
356	struct ip_tunnel_prl_entry *p;
357	int err = 0;
358
359	if (a->addr == htonl(INADDR_ANY))
360		return -EINVAL;
361
362	ASSERT_RTNL();
363
364	for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
365		if (p->addr == a->addr) {
366			if (chg) {
367				p->flags = a->flags;
368				goto out;
369			}
370			err = -EEXIST;
371			goto out;
372		}
373	}
374
375	if (chg) {
376		err = -ENXIO;
377		goto out;
378	}
379
380	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
381	if (!p) {
382		err = -ENOBUFS;
383		goto out;
384	}
385
386	p->next = t->prl;
387	p->addr = a->addr;
388	p->flags = a->flags;
389	t->prl_count++;
390	rcu_assign_pointer(t->prl, p);
391out:
392	return err;
393}
394
395static void prl_list_destroy_rcu(struct rcu_head *head)
396{
397	struct ip_tunnel_prl_entry *p, *n;
398
399	p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
400	do {
401		n = rcu_dereference_protected(p->next, 1);
402		kfree(p);
403		p = n;
404	} while (p);
405}
406
407static int
408ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
409{
410	struct ip_tunnel_prl_entry *x;
411	struct ip_tunnel_prl_entry __rcu **p;
412	int err = 0;
413
414	ASSERT_RTNL();
415
416	if (a && a->addr != htonl(INADDR_ANY)) {
417		for (p = &t->prl;
418		     (x = rtnl_dereference(*p)) != NULL;
419		     p = &x->next) {
420			if (x->addr == a->addr) {
421				*p = x->next;
422				kfree_rcu(x, rcu_head);
423				t->prl_count--;
424				goto out;
425			}
426		}
427		err = -ENXIO;
428	} else {
429		x = rtnl_dereference(t->prl);
430		if (x) {
431			t->prl_count = 0;
432			call_rcu(&x->rcu_head, prl_list_destroy_rcu);
433			t->prl = NULL;
434		}
435	}
436out:
437	return err;
438}
439
440static int
441isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
442{
443	struct ip_tunnel_prl_entry *p;
444	int ok = 1;
445
446	rcu_read_lock();
447	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
448	if (p) {
449		if (p->flags & PRL_DEFAULT)
450			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
451		else
452			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
453	} else {
454		const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
455
456		if (ipv6_addr_is_isatap(addr6) &&
457		    (addr6->s6_addr32[3] == iph->saddr) &&
458		    ipv6_chk_prefix(addr6, t->dev))
459			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
460		else
461			ok = 0;
462	}
463	rcu_read_unlock();
464	return ok;
465}
466
467static void ipip6_tunnel_uninit(struct net_device *dev)
468{
469	struct ip_tunnel *tunnel = netdev_priv(dev);
470	struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
471
472	if (dev == sitn->fb_tunnel_dev) {
473		RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
474	} else {
475		ipip6_tunnel_unlink(sitn, tunnel);
476		ipip6_tunnel_del_prl(tunnel, NULL);
477	}
478	ip_tunnel_dst_reset_all(tunnel);
479	dev_put(dev);
480}
481
482/* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
483 * if sufficient data bytes are available
484 */
485static int ipip6_err_gen_icmpv6_unreach(struct sk_buff *skb)
486{
487	int ihl = ((const struct iphdr *)skb->data)->ihl*4;
488	struct rt6_info *rt;
489	struct sk_buff *skb2;
490
491	if (!pskb_may_pull(skb, ihl + sizeof(struct ipv6hdr) + 8))
492		return 1;
493
494	skb2 = skb_clone(skb, GFP_ATOMIC);
495
496	if (!skb2)
497		return 1;
498
499	skb_dst_drop(skb2);
500	skb_pull(skb2, ihl);
501	skb_reset_network_header(skb2);
502
503	rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 0);
504
505	if (rt && rt->dst.dev)
506		skb2->dev = rt->dst.dev;
507
508	icmpv6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
509
510	if (rt)
511		ip6_rt_put(rt);
512
513	kfree_skb(skb2);
514
515	return 0;
516}
517
518static int ipip6_err(struct sk_buff *skb, u32 info)
519{
520	const struct iphdr *iph = (const struct iphdr *)skb->data;
521	const int type = icmp_hdr(skb)->type;
522	const int code = icmp_hdr(skb)->code;
523	struct ip_tunnel *t;
524	int err;
525
526	switch (type) {
527	default:
528	case ICMP_PARAMETERPROB:
529		return 0;
530
531	case ICMP_DEST_UNREACH:
532		switch (code) {
533		case ICMP_SR_FAILED:
534			/* Impossible event. */
535			return 0;
536		default:
537			/* All others are translated to HOST_UNREACH.
538			   rfc2003 contains "deep thoughts" about NET_UNREACH,
539			   I believe they are just ether pollution. --ANK
540			 */
541			break;
542		}
543		break;
544	case ICMP_TIME_EXCEEDED:
545		if (code != ICMP_EXC_TTL)
546			return 0;
547		break;
548	case ICMP_REDIRECT:
549		break;
550	}
551
552	err = -ENOENT;
553
554	t = ipip6_tunnel_lookup(dev_net(skb->dev),
555				skb->dev,
556				iph->daddr,
557				iph->saddr);
558	if (!t)
559		goto out;
560
561	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
562		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
563				 t->parms.link, 0, IPPROTO_IPV6, 0);
564		err = 0;
565		goto out;
566	}
567	if (type == ICMP_REDIRECT) {
568		ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
569			      IPPROTO_IPV6, 0);
570		err = 0;
571		goto out;
572	}
573
574	if (t->parms.iph.daddr == 0)
575		goto out;
576
577	err = 0;
578	if (!ipip6_err_gen_icmpv6_unreach(skb))
579		goto out;
580
581	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
582		goto out;
583
584	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
585		t->err_count++;
586	else
587		t->err_count = 1;
588	t->err_time = jiffies;
589out:
590	return err;
591}
592
593static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
594				  const struct in6_addr *v6addr)
595{
596	__be32 v4embed = 0;
597	if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
598		return true;
599	return false;
600}
601
602/* Checks if an address matches an address on the tunnel interface.
603 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
604 * Long story:
605 * This function is called after we considered the packet as spoofed
606 * in is_spoofed_6rd.
607 * We may have a router that is doing NAT for proto 41 packets
608 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
609 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
610 * function will return true, dropping the packet.
611 * But, we can still check if is spoofed against the IP
612 * addresses associated with the interface.
613 */
614static bool only_dnatted(const struct ip_tunnel *tunnel,
615	const struct in6_addr *v6dst)
616{
617	int prefix_len;
618
619#ifdef CONFIG_IPV6_SIT_6RD
620	prefix_len = tunnel->ip6rd.prefixlen + 32
621		- tunnel->ip6rd.relay_prefixlen;
622#else
623	prefix_len = 48;
624#endif
625	return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
626}
627
628/* Returns true if a packet is spoofed */
629static bool packet_is_spoofed(struct sk_buff *skb,
630			      const struct iphdr *iph,
631			      struct ip_tunnel *tunnel)
632{
633	const struct ipv6hdr *ipv6h;
634
635	if (tunnel->dev->priv_flags & IFF_ISATAP) {
636		if (!isatap_chksrc(skb, iph, tunnel))
637			return true;
638
639		return false;
640	}
641
642	if (tunnel->dev->flags & IFF_POINTOPOINT)
643		return false;
644
645	ipv6h = ipv6_hdr(skb);
646
647	if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
648		net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
649				     &iph->saddr, &ipv6h->saddr,
650				     &iph->daddr, &ipv6h->daddr);
651		return true;
652	}
653
654	if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
655		return false;
656
657	if (only_dnatted(tunnel, &ipv6h->daddr))
658		return false;
659
660	net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
661			     &iph->saddr, &ipv6h->saddr,
662			     &iph->daddr, &ipv6h->daddr);
663	return true;
664}
665
666static int ipip6_rcv(struct sk_buff *skb)
667{
668	const struct iphdr *iph = ip_hdr(skb);
669	struct ip_tunnel *tunnel;
670	int err;
671
672	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
673				     iph->saddr, iph->daddr);
674	if (tunnel) {
675		struct pcpu_sw_netstats *tstats;
676
677		if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
678		    tunnel->parms.iph.protocol != 0)
679			goto out;
680
681		skb->mac_header = skb->network_header;
682		skb_reset_network_header(skb);
683		IPCB(skb)->flags = 0;
684		skb->protocol = htons(ETH_P_IPV6);
685
686		if (packet_is_spoofed(skb, iph, tunnel)) {
687			tunnel->dev->stats.rx_errors++;
688			goto out;
689		}
690
691		__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
692
693		err = IP_ECN_decapsulate(iph, skb);
694		if (unlikely(err)) {
695			if (log_ecn_error)
696				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
697						     &iph->saddr, iph->tos);
698			if (err > 1) {
699				++tunnel->dev->stats.rx_frame_errors;
700				++tunnel->dev->stats.rx_errors;
701				goto out;
702			}
703		}
704
705		tstats = this_cpu_ptr(tunnel->dev->tstats);
706		u64_stats_update_begin(&tstats->syncp);
707		tstats->rx_packets++;
708		tstats->rx_bytes += skb->len;
709		u64_stats_update_end(&tstats->syncp);
710
711		netif_rx(skb);
712
713		return 0;
714	}
715
716	/* no tunnel matched,  let upstream know, ipsec may handle it */
717	return 1;
718out:
719	kfree_skb(skb);
720	return 0;
721}
722
723static const struct tnl_ptk_info tpi = {
724	/* no tunnel info required for ipip. */
725	.proto = htons(ETH_P_IP),
726};
727
728static int ipip_rcv(struct sk_buff *skb)
729{
730	const struct iphdr *iph;
731	struct ip_tunnel *tunnel;
732
733	iph = ip_hdr(skb);
734	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
735				     iph->saddr, iph->daddr);
736	if (tunnel) {
737		if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
738		    tunnel->parms.iph.protocol != 0)
739			goto drop;
740
741		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
742			goto drop;
743		if (iptunnel_pull_header(skb, 0, tpi.proto))
744			goto drop;
745		return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
746	}
747
748	return 1;
749
750drop:
751	kfree_skb(skb);
752	return 0;
753}
754
755/*
756 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
757 * stores the embedded IPv4 address in v4dst and returns true.
758 */
759static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
760		      __be32 *v4dst)
761{
762#ifdef CONFIG_IPV6_SIT_6RD
763	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
764			      tunnel->ip6rd.prefixlen)) {
765		unsigned int pbw0, pbi0;
766		int pbi1;
767		u32 d;
768
769		pbw0 = tunnel->ip6rd.prefixlen >> 5;
770		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
771
772		d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
773		    tunnel->ip6rd.relay_prefixlen;
774
775		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
776		if (pbi1 > 0)
777			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
778			     (32 - pbi1);
779
780		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
781		return true;
782	}
783#else
784	if (v6dst->s6_addr16[0] == htons(0x2002)) {
785		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
786		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
787		return true;
788	}
789#endif
790	return false;
791}
792
793static inline __be32 try_6rd(struct ip_tunnel *tunnel,
794			     const struct in6_addr *v6dst)
795{
796	__be32 dst = 0;
797	check_6rd(tunnel, v6dst, &dst);
798	return dst;
799}
800
801/*
802 *	This function assumes it is being called from dev_queue_xmit()
803 *	and that skb is filled properly by that function.
804 */
805
806static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
807				     struct net_device *dev)
808{
809	struct ip_tunnel *tunnel = netdev_priv(dev);
810	const struct iphdr  *tiph = &tunnel->parms.iph;
811	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
812	u8     tos = tunnel->parms.iph.tos;
813	__be16 df = tiph->frag_off;
814	struct rtable *rt;		/* Route to the other host */
815	struct net_device *tdev;	/* Device to other host */
816	unsigned int max_headroom;	/* The extra header space needed */
817	__be32 dst = tiph->daddr;
818	struct flowi4 fl4;
819	int    mtu;
820	const struct in6_addr *addr6;
821	int addr_type;
822	u8 ttl;
823	int err;
824	u8 protocol = IPPROTO_IPV6;
825	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
826
827	if (skb->protocol != htons(ETH_P_IPV6))
828		goto tx_error;
829
830	if (tos == 1)
831		tos = ipv6_get_dsfield(iph6);
832
833	/* ISATAP (RFC4214) - must come before 6to4 */
834	if (dev->priv_flags & IFF_ISATAP) {
835		struct neighbour *neigh = NULL;
836		bool do_tx_error = false;
837
838		if (skb_dst(skb))
839			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
840
841		if (!neigh) {
842			net_dbg_ratelimited("nexthop == NULL\n");
843			goto tx_error;
844		}
845
846		addr6 = (const struct in6_addr *)&neigh->primary_key;
847		addr_type = ipv6_addr_type(addr6);
848
849		if ((addr_type & IPV6_ADDR_UNICAST) &&
850		     ipv6_addr_is_isatap(addr6))
851			dst = addr6->s6_addr32[3];
852		else
853			do_tx_error = true;
854
855		neigh_release(neigh);
856		if (do_tx_error)
857			goto tx_error;
858	}
859
860	if (!dst)
861		dst = try_6rd(tunnel, &iph6->daddr);
862
863	if (!dst) {
864		struct neighbour *neigh = NULL;
865		bool do_tx_error = false;
866
867		if (skb_dst(skb))
868			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
869
870		if (!neigh) {
871			net_dbg_ratelimited("nexthop == NULL\n");
872			goto tx_error;
873		}
874
875		addr6 = (const struct in6_addr *)&neigh->primary_key;
876		addr_type = ipv6_addr_type(addr6);
877
878		if (addr_type == IPV6_ADDR_ANY) {
879			addr6 = &ipv6_hdr(skb)->daddr;
880			addr_type = ipv6_addr_type(addr6);
881		}
882
883		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
884			dst = addr6->s6_addr32[3];
885		else
886			do_tx_error = true;
887
888		neigh_release(neigh);
889		if (do_tx_error)
890			goto tx_error;
891	}
892
893	rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
894				   dst, tiph->saddr,
895				   0, 0,
896				   IPPROTO_IPV6, RT_TOS(tos),
897				   tunnel->parms.link);
898	if (IS_ERR(rt)) {
899		dev->stats.tx_carrier_errors++;
900		goto tx_error_icmp;
901	}
902	if (rt->rt_type != RTN_UNICAST) {
903		ip_rt_put(rt);
904		dev->stats.tx_carrier_errors++;
905		goto tx_error_icmp;
906	}
907	tdev = rt->dst.dev;
908
909	if (tdev == dev) {
910		ip_rt_put(rt);
911		dev->stats.collisions++;
912		goto tx_error;
913	}
914
915	skb = iptunnel_handle_offloads(skb, false, SKB_GSO_SIT);
916	if (IS_ERR(skb)) {
917		ip_rt_put(rt);
918		goto out;
919	}
920
921	if (df) {
922		mtu = dst_mtu(&rt->dst) - t_hlen;
923
924		if (mtu < 68) {
925			dev->stats.collisions++;
926			ip_rt_put(rt);
927			goto tx_error;
928		}
929
930		if (mtu < IPV6_MIN_MTU) {
931			mtu = IPV6_MIN_MTU;
932			df = 0;
933		}
934
935		if (tunnel->parms.iph.daddr && skb_dst(skb))
936			skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
937
938		if (skb->len > mtu && !skb_is_gso(skb)) {
939			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
940			ip_rt_put(rt);
941			goto tx_error;
942		}
943	}
944
945	if (tunnel->err_count > 0) {
946		if (time_before(jiffies,
947				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
948			tunnel->err_count--;
949			dst_link_failure(skb);
950		} else
951			tunnel->err_count = 0;
952	}
953
954	/*
955	 * Okay, now see if we can stuff it in the buffer as-is.
956	 */
957	max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
958
959	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
960	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
961		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
962		if (!new_skb) {
963			ip_rt_put(rt);
964			dev->stats.tx_dropped++;
965			kfree_skb(skb);
966			return NETDEV_TX_OK;
967		}
968		if (skb->sk)
969			skb_set_owner_w(new_skb, skb->sk);
970		dev_kfree_skb(skb);
971		skb = new_skb;
972		iph6 = ipv6_hdr(skb);
973	}
974	ttl = tiph->ttl;
975	if (ttl == 0)
976		ttl = iph6->hop_limit;
977	tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
978
979	if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
980		ip_rt_put(rt);
981		goto tx_error;
982	}
983
984	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
985
986	err = iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr,
987			    protocol, tos, ttl, df,
988			    !net_eq(tunnel->net, dev_net(dev)));
989	iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
990	return NETDEV_TX_OK;
991
992tx_error_icmp:
993	dst_link_failure(skb);
994tx_error:
995	kfree_skb(skb);
996out:
997	dev->stats.tx_errors++;
998	return NETDEV_TX_OK;
999}
1000
1001static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
1002{
1003	struct ip_tunnel *tunnel = netdev_priv(dev);
1004	const struct iphdr  *tiph = &tunnel->parms.iph;
1005
1006	skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
1007	if (IS_ERR(skb))
1008		goto out;
1009
1010	skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1011
1012	ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
1013	return NETDEV_TX_OK;
1014out:
1015	dev->stats.tx_errors++;
1016	return NETDEV_TX_OK;
1017}
1018
1019static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1020				   struct net_device *dev)
1021{
1022	switch (skb->protocol) {
1023	case htons(ETH_P_IP):
1024		ipip_tunnel_xmit(skb, dev);
1025		break;
1026	case htons(ETH_P_IPV6):
1027		ipip6_tunnel_xmit(skb, dev);
1028		break;
1029	default:
1030		goto tx_err;
1031	}
1032
1033	return NETDEV_TX_OK;
1034
1035tx_err:
1036	dev->stats.tx_errors++;
1037	kfree_skb(skb);
1038	return NETDEV_TX_OK;
1039
1040}
1041
1042static void ipip6_tunnel_bind_dev(struct net_device *dev)
1043{
1044	struct net_device *tdev = NULL;
1045	struct ip_tunnel *tunnel;
1046	const struct iphdr *iph;
1047	struct flowi4 fl4;
1048
1049	tunnel = netdev_priv(dev);
1050	iph = &tunnel->parms.iph;
1051
1052	if (iph->daddr) {
1053		struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1054							  NULL,
1055							  iph->daddr, iph->saddr,
1056							  0, 0,
1057							  IPPROTO_IPV6,
1058							  RT_TOS(iph->tos),
1059							  tunnel->parms.link);
1060
1061		if (!IS_ERR(rt)) {
1062			tdev = rt->dst.dev;
1063			ip_rt_put(rt);
1064		}
1065		dev->flags |= IFF_POINTOPOINT;
1066	}
1067
1068	if (!tdev && tunnel->parms.link)
1069		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1070
1071	if (tdev) {
1072		int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1073
1074		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1075		dev->mtu = tdev->mtu - t_hlen;
1076		if (dev->mtu < IPV6_MIN_MTU)
1077			dev->mtu = IPV6_MIN_MTU;
1078	}
1079}
1080
1081static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
1082{
1083	struct net *net = t->net;
1084	struct sit_net *sitn = net_generic(net, sit_net_id);
1085
1086	ipip6_tunnel_unlink(sitn, t);
1087	synchronize_net();
1088	t->parms.iph.saddr = p->iph.saddr;
1089	t->parms.iph.daddr = p->iph.daddr;
1090	memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1091	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1092	ipip6_tunnel_link(sitn, t);
1093	t->parms.iph.ttl = p->iph.ttl;
1094	t->parms.iph.tos = p->iph.tos;
1095	if (t->parms.link != p->link) {
1096		t->parms.link = p->link;
1097		ipip6_tunnel_bind_dev(t->dev);
1098	}
1099	ip_tunnel_dst_reset_all(t);
1100	netdev_state_change(t->dev);
1101}
1102
1103#ifdef CONFIG_IPV6_SIT_6RD
1104static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1105				   struct ip_tunnel_6rd *ip6rd)
1106{
1107	struct in6_addr prefix;
1108	__be32 relay_prefix;
1109
1110	if (ip6rd->relay_prefixlen > 32 ||
1111	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1112		return -EINVAL;
1113
1114	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1115	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1116		return -EINVAL;
1117	if (ip6rd->relay_prefixlen)
1118		relay_prefix = ip6rd->relay_prefix &
1119			       htonl(0xffffffffUL <<
1120				     (32 - ip6rd->relay_prefixlen));
1121	else
1122		relay_prefix = 0;
1123	if (relay_prefix != ip6rd->relay_prefix)
1124		return -EINVAL;
1125
1126	t->ip6rd.prefix = prefix;
1127	t->ip6rd.relay_prefix = relay_prefix;
1128	t->ip6rd.prefixlen = ip6rd->prefixlen;
1129	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1130	ip_tunnel_dst_reset_all(t);
1131	netdev_state_change(t->dev);
1132	return 0;
1133}
1134#endif
1135
1136static int
1137ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1138{
1139	int err = 0;
1140	struct ip_tunnel_parm p;
1141	struct ip_tunnel_prl prl;
1142	struct ip_tunnel *t = netdev_priv(dev);
1143	struct net *net = t->net;
1144	struct sit_net *sitn = net_generic(net, sit_net_id);
1145#ifdef CONFIG_IPV6_SIT_6RD
1146	struct ip_tunnel_6rd ip6rd;
1147#endif
1148
1149	switch (cmd) {
1150	case SIOCGETTUNNEL:
1151#ifdef CONFIG_IPV6_SIT_6RD
1152	case SIOCGET6RD:
1153#endif
1154		if (dev == sitn->fb_tunnel_dev) {
1155			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1156				err = -EFAULT;
1157				break;
1158			}
1159			t = ipip6_tunnel_locate(net, &p, 0);
1160			if (!t)
1161				t = netdev_priv(dev);
1162		}
1163
1164		err = -EFAULT;
1165		if (cmd == SIOCGETTUNNEL) {
1166			memcpy(&p, &t->parms, sizeof(p));
1167			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1168					 sizeof(p)))
1169				goto done;
1170#ifdef CONFIG_IPV6_SIT_6RD
1171		} else {
1172			ip6rd.prefix = t->ip6rd.prefix;
1173			ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1174			ip6rd.prefixlen = t->ip6rd.prefixlen;
1175			ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1176			if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1177					 sizeof(ip6rd)))
1178				goto done;
1179#endif
1180		}
1181		err = 0;
1182		break;
1183
1184	case SIOCADDTUNNEL:
1185	case SIOCCHGTUNNEL:
1186		err = -EPERM;
1187		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1188			goto done;
1189
1190		err = -EFAULT;
1191		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1192			goto done;
1193
1194		err = -EINVAL;
1195		if (p.iph.protocol != IPPROTO_IPV6 &&
1196		    p.iph.protocol != IPPROTO_IPIP &&
1197		    p.iph.protocol != 0)
1198			goto done;
1199		if (p.iph.version != 4 ||
1200		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1201			goto done;
1202		if (p.iph.ttl)
1203			p.iph.frag_off |= htons(IP_DF);
1204
1205		t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1206
1207		if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1208			if (t) {
1209				if (t->dev != dev) {
1210					err = -EEXIST;
1211					break;
1212				}
1213			} else {
1214				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1215				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1216					err = -EINVAL;
1217					break;
1218				}
1219				t = netdev_priv(dev);
1220			}
1221
1222			ipip6_tunnel_update(t, &p);
1223		}
1224
1225		if (t) {
1226			err = 0;
1227			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1228				err = -EFAULT;
1229		} else
1230			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1231		break;
1232
1233	case SIOCDELTUNNEL:
1234		err = -EPERM;
1235		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1236			goto done;
1237
1238		if (dev == sitn->fb_tunnel_dev) {
1239			err = -EFAULT;
1240			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1241				goto done;
1242			err = -ENOENT;
1243			t = ipip6_tunnel_locate(net, &p, 0);
1244			if (!t)
1245				goto done;
1246			err = -EPERM;
1247			if (t == netdev_priv(sitn->fb_tunnel_dev))
1248				goto done;
1249			dev = t->dev;
1250		}
1251		unregister_netdevice(dev);
1252		err = 0;
1253		break;
1254
1255	case SIOCGETPRL:
1256		err = -EINVAL;
1257		if (dev == sitn->fb_tunnel_dev)
1258			goto done;
1259		err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1260		break;
1261
1262	case SIOCADDPRL:
1263	case SIOCDELPRL:
1264	case SIOCCHGPRL:
1265		err = -EPERM;
1266		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1267			goto done;
1268		err = -EINVAL;
1269		if (dev == sitn->fb_tunnel_dev)
1270			goto done;
1271		err = -EFAULT;
1272		if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1273			goto done;
1274
1275		switch (cmd) {
1276		case SIOCDELPRL:
1277			err = ipip6_tunnel_del_prl(t, &prl);
1278			break;
1279		case SIOCADDPRL:
1280		case SIOCCHGPRL:
1281			err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1282			break;
1283		}
1284		ip_tunnel_dst_reset_all(t);
1285		netdev_state_change(dev);
1286		break;
1287
1288#ifdef CONFIG_IPV6_SIT_6RD
1289	case SIOCADD6RD:
1290	case SIOCCHG6RD:
1291	case SIOCDEL6RD:
1292		err = -EPERM;
1293		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1294			goto done;
1295
1296		err = -EFAULT;
1297		if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1298				   sizeof(ip6rd)))
1299			goto done;
1300
1301		if (cmd != SIOCDEL6RD) {
1302			err = ipip6_tunnel_update_6rd(t, &ip6rd);
1303			if (err < 0)
1304				goto done;
1305		} else
1306			ipip6_tunnel_clone_6rd(dev, sitn);
1307
1308		err = 0;
1309		break;
1310#endif
1311
1312	default:
1313		err = -EINVAL;
1314	}
1315
1316done:
1317	return err;
1318}
1319
1320static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1321{
1322	struct ip_tunnel *tunnel = netdev_priv(dev);
1323	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1324
1325	if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - t_hlen)
1326		return -EINVAL;
1327	dev->mtu = new_mtu;
1328	return 0;
1329}
1330
1331static const struct net_device_ops ipip6_netdev_ops = {
1332	.ndo_init	= ipip6_tunnel_init,
1333	.ndo_uninit	= ipip6_tunnel_uninit,
1334	.ndo_start_xmit	= sit_tunnel_xmit,
1335	.ndo_do_ioctl	= ipip6_tunnel_ioctl,
1336	.ndo_change_mtu	= ipip6_tunnel_change_mtu,
1337	.ndo_get_stats64 = ip_tunnel_get_stats64,
1338	.ndo_get_iflink = ip_tunnel_get_iflink,
1339};
1340
1341static void ipip6_dev_free(struct net_device *dev)
1342{
1343	struct ip_tunnel *tunnel = netdev_priv(dev);
1344
1345	free_percpu(tunnel->dst_cache);
1346	free_percpu(dev->tstats);
1347	free_netdev(dev);
1348}
1349
1350#define SIT_FEATURES (NETIF_F_SG	   | \
1351		      NETIF_F_FRAGLIST	   | \
1352		      NETIF_F_HIGHDMA	   | \
1353		      NETIF_F_GSO_SOFTWARE | \
1354		      NETIF_F_HW_CSUM)
1355
1356static void ipip6_tunnel_setup(struct net_device *dev)
1357{
1358	struct ip_tunnel *tunnel = netdev_priv(dev);
1359	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1360
1361	dev->netdev_ops		= &ipip6_netdev_ops;
1362	dev->destructor		= ipip6_dev_free;
1363
1364	dev->type		= ARPHRD_SIT;
1365	dev->hard_header_len	= LL_MAX_HEADER + t_hlen;
1366	dev->mtu		= ETH_DATA_LEN - t_hlen;
1367	dev->flags		= IFF_NOARP;
1368	netif_keep_dst(dev);
1369	dev->addr_len		= 4;
1370	dev->features		|= NETIF_F_LLTX;
1371	dev->features		|= SIT_FEATURES;
1372	dev->hw_features	|= SIT_FEATURES;
1373}
1374
1375static int ipip6_tunnel_init(struct net_device *dev)
1376{
1377	struct ip_tunnel *tunnel = netdev_priv(dev);
1378
1379	tunnel->dev = dev;
1380	tunnel->net = dev_net(dev);
1381	strcpy(tunnel->parms.name, dev->name);
1382
1383	ipip6_tunnel_bind_dev(dev);
1384	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1385	if (!dev->tstats)
1386		return -ENOMEM;
1387
1388	tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1389	if (!tunnel->dst_cache) {
1390		free_percpu(dev->tstats);
1391		return -ENOMEM;
1392	}
1393
1394	return 0;
1395}
1396
1397static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1398{
1399	struct ip_tunnel *tunnel = netdev_priv(dev);
1400	struct iphdr *iph = &tunnel->parms.iph;
1401	struct net *net = dev_net(dev);
1402	struct sit_net *sitn = net_generic(net, sit_net_id);
1403
1404	iph->version		= 4;
1405	iph->protocol		= IPPROTO_IPV6;
1406	iph->ihl		= 5;
1407	iph->ttl		= 64;
1408
1409	dev_hold(dev);
1410	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1411}
1412
1413static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1414{
1415	u8 proto;
1416
1417	if (!data || !data[IFLA_IPTUN_PROTO])
1418		return 0;
1419
1420	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1421	if (proto != IPPROTO_IPV6 &&
1422	    proto != IPPROTO_IPIP &&
1423	    proto != 0)
1424		return -EINVAL;
1425
1426	return 0;
1427}
1428
1429static void ipip6_netlink_parms(struct nlattr *data[],
1430				struct ip_tunnel_parm *parms)
1431{
1432	memset(parms, 0, sizeof(*parms));
1433
1434	parms->iph.version = 4;
1435	parms->iph.protocol = IPPROTO_IPV6;
1436	parms->iph.ihl = 5;
1437	parms->iph.ttl = 64;
1438
1439	if (!data)
1440		return;
1441
1442	if (data[IFLA_IPTUN_LINK])
1443		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1444
1445	if (data[IFLA_IPTUN_LOCAL])
1446		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1447
1448	if (data[IFLA_IPTUN_REMOTE])
1449		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1450
1451	if (data[IFLA_IPTUN_TTL]) {
1452		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1453		if (parms->iph.ttl)
1454			parms->iph.frag_off = htons(IP_DF);
1455	}
1456
1457	if (data[IFLA_IPTUN_TOS])
1458		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1459
1460	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1461		parms->iph.frag_off = htons(IP_DF);
1462
1463	if (data[IFLA_IPTUN_FLAGS])
1464		parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1465
1466	if (data[IFLA_IPTUN_PROTO])
1467		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1468
1469}
1470
1471/* This function returns true when ENCAP attributes are present in the nl msg */
1472static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1473				      struct ip_tunnel_encap *ipencap)
1474{
1475	bool ret = false;
1476
1477	memset(ipencap, 0, sizeof(*ipencap));
1478
1479	if (!data)
1480		return ret;
1481
1482	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1483		ret = true;
1484		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1485	}
1486
1487	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1488		ret = true;
1489		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1490	}
1491
1492	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1493		ret = true;
1494		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1495	}
1496
1497	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1498		ret = true;
1499		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1500	}
1501
1502	return ret;
1503}
1504
1505#ifdef CONFIG_IPV6_SIT_6RD
1506/* This function returns true when 6RD attributes are present in the nl msg */
1507static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1508				    struct ip_tunnel_6rd *ip6rd)
1509{
1510	bool ret = false;
1511	memset(ip6rd, 0, sizeof(*ip6rd));
1512
1513	if (!data)
1514		return ret;
1515
1516	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1517		ret = true;
1518		ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1519	}
1520
1521	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1522		ret = true;
1523		ip6rd->relay_prefix =
1524			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1525	}
1526
1527	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1528		ret = true;
1529		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1530	}
1531
1532	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1533		ret = true;
1534		ip6rd->relay_prefixlen =
1535			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1536	}
1537
1538	return ret;
1539}
1540#endif
1541
1542static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1543			 struct nlattr *tb[], struct nlattr *data[])
1544{
1545	struct net *net = dev_net(dev);
1546	struct ip_tunnel *nt;
1547	struct ip_tunnel_encap ipencap;
1548#ifdef CONFIG_IPV6_SIT_6RD
1549	struct ip_tunnel_6rd ip6rd;
1550#endif
1551	int err;
1552
1553	nt = netdev_priv(dev);
1554
1555	if (ipip6_netlink_encap_parms(data, &ipencap)) {
1556		err = ip_tunnel_encap_setup(nt, &ipencap);
1557		if (err < 0)
1558			return err;
1559	}
1560
1561	ipip6_netlink_parms(data, &nt->parms);
1562
1563	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1564		return -EEXIST;
1565
1566	err = ipip6_tunnel_create(dev);
1567	if (err < 0)
1568		return err;
1569
1570#ifdef CONFIG_IPV6_SIT_6RD
1571	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1572		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1573#endif
1574
1575	return err;
1576}
1577
1578static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1579			  struct nlattr *data[])
1580{
1581	struct ip_tunnel *t = netdev_priv(dev);
1582	struct ip_tunnel_parm p;
1583	struct ip_tunnel_encap ipencap;
1584	struct net *net = t->net;
1585	struct sit_net *sitn = net_generic(net, sit_net_id);
1586#ifdef CONFIG_IPV6_SIT_6RD
1587	struct ip_tunnel_6rd ip6rd;
1588#endif
1589	int err;
1590
1591	if (dev == sitn->fb_tunnel_dev)
1592		return -EINVAL;
1593
1594	if (ipip6_netlink_encap_parms(data, &ipencap)) {
1595		err = ip_tunnel_encap_setup(t, &ipencap);
1596		if (err < 0)
1597			return err;
1598	}
1599
1600	ipip6_netlink_parms(data, &p);
1601
1602	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1603	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1604		return -EINVAL;
1605
1606	t = ipip6_tunnel_locate(net, &p, 0);
1607
1608	if (t) {
1609		if (t->dev != dev)
1610			return -EEXIST;
1611	} else
1612		t = netdev_priv(dev);
1613
1614	ipip6_tunnel_update(t, &p);
1615
1616#ifdef CONFIG_IPV6_SIT_6RD
1617	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1618		return ipip6_tunnel_update_6rd(t, &ip6rd);
1619#endif
1620
1621	return 0;
1622}
1623
1624static size_t ipip6_get_size(const struct net_device *dev)
1625{
1626	return
1627		/* IFLA_IPTUN_LINK */
1628		nla_total_size(4) +
1629		/* IFLA_IPTUN_LOCAL */
1630		nla_total_size(4) +
1631		/* IFLA_IPTUN_REMOTE */
1632		nla_total_size(4) +
1633		/* IFLA_IPTUN_TTL */
1634		nla_total_size(1) +
1635		/* IFLA_IPTUN_TOS */
1636		nla_total_size(1) +
1637		/* IFLA_IPTUN_PMTUDISC */
1638		nla_total_size(1) +
1639		/* IFLA_IPTUN_FLAGS */
1640		nla_total_size(2) +
1641		/* IFLA_IPTUN_PROTO */
1642		nla_total_size(1) +
1643#ifdef CONFIG_IPV6_SIT_6RD
1644		/* IFLA_IPTUN_6RD_PREFIX */
1645		nla_total_size(sizeof(struct in6_addr)) +
1646		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1647		nla_total_size(4) +
1648		/* IFLA_IPTUN_6RD_PREFIXLEN */
1649		nla_total_size(2) +
1650		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1651		nla_total_size(2) +
1652#endif
1653		/* IFLA_IPTUN_ENCAP_TYPE */
1654		nla_total_size(2) +
1655		/* IFLA_IPTUN_ENCAP_FLAGS */
1656		nla_total_size(2) +
1657		/* IFLA_IPTUN_ENCAP_SPORT */
1658		nla_total_size(2) +
1659		/* IFLA_IPTUN_ENCAP_DPORT */
1660		nla_total_size(2) +
1661		0;
1662}
1663
1664static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1665{
1666	struct ip_tunnel *tunnel = netdev_priv(dev);
1667	struct ip_tunnel_parm *parm = &tunnel->parms;
1668
1669	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1670	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1671	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1672	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1673	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1674	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1675		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1676	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1677	    nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1678		goto nla_put_failure;
1679
1680#ifdef CONFIG_IPV6_SIT_6RD
1681	if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1682			     &tunnel->ip6rd.prefix) ||
1683	    nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1684			    tunnel->ip6rd.relay_prefix) ||
1685	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1686			tunnel->ip6rd.prefixlen) ||
1687	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1688			tunnel->ip6rd.relay_prefixlen))
1689		goto nla_put_failure;
1690#endif
1691
1692	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1693			tunnel->encap.type) ||
1694	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1695			tunnel->encap.sport) ||
1696	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1697			tunnel->encap.dport) ||
1698	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1699			tunnel->encap.flags))
1700		goto nla_put_failure;
1701
1702	return 0;
1703
1704nla_put_failure:
1705	return -EMSGSIZE;
1706}
1707
1708static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1709	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1710	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1711	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1712	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1713	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1714	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1715	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1716	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1717#ifdef CONFIG_IPV6_SIT_6RD
1718	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1719	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1720	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1721	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1722#endif
1723	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
1724	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
1725	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
1726	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
1727};
1728
1729static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1730{
1731	struct net *net = dev_net(dev);
1732	struct sit_net *sitn = net_generic(net, sit_net_id);
1733
1734	if (dev != sitn->fb_tunnel_dev)
1735		unregister_netdevice_queue(dev, head);
1736}
1737
1738static struct rtnl_link_ops sit_link_ops __read_mostly = {
1739	.kind		= "sit",
1740	.maxtype	= IFLA_IPTUN_MAX,
1741	.policy		= ipip6_policy,
1742	.priv_size	= sizeof(struct ip_tunnel),
1743	.setup		= ipip6_tunnel_setup,
1744	.validate	= ipip6_validate,
1745	.newlink	= ipip6_newlink,
1746	.changelink	= ipip6_changelink,
1747	.get_size	= ipip6_get_size,
1748	.fill_info	= ipip6_fill_info,
1749	.dellink	= ipip6_dellink,
1750	.get_link_net	= ip_tunnel_get_link_net,
1751};
1752
1753static struct xfrm_tunnel sit_handler __read_mostly = {
1754	.handler	=	ipip6_rcv,
1755	.err_handler	=	ipip6_err,
1756	.priority	=	1,
1757};
1758
1759static struct xfrm_tunnel ipip_handler __read_mostly = {
1760	.handler	=	ipip_rcv,
1761	.err_handler	=	ipip6_err,
1762	.priority	=	2,
1763};
1764
1765static void __net_exit sit_destroy_tunnels(struct net *net,
1766					   struct list_head *head)
1767{
1768	struct sit_net *sitn = net_generic(net, sit_net_id);
1769	struct net_device *dev, *aux;
1770	int prio;
1771
1772	for_each_netdev_safe(net, dev, aux)
1773		if (dev->rtnl_link_ops == &sit_link_ops)
1774			unregister_netdevice_queue(dev, head);
1775
1776	for (prio = 1; prio < 4; prio++) {
1777		int h;
1778		for (h = 0; h < HASH_SIZE; h++) {
1779			struct ip_tunnel *t;
1780
1781			t = rtnl_dereference(sitn->tunnels[prio][h]);
1782			while (t) {
1783				/* If dev is in the same netns, it has already
1784				 * been added to the list by the previous loop.
1785				 */
1786				if (!net_eq(dev_net(t->dev), net))
1787					unregister_netdevice_queue(t->dev,
1788								   head);
1789				t = rtnl_dereference(t->next);
1790			}
1791		}
1792	}
1793}
1794
1795static int __net_init sit_init_net(struct net *net)
1796{
1797	struct sit_net *sitn = net_generic(net, sit_net_id);
1798	struct ip_tunnel *t;
1799	int err;
1800
1801	sitn->tunnels[0] = sitn->tunnels_wc;
1802	sitn->tunnels[1] = sitn->tunnels_l;
1803	sitn->tunnels[2] = sitn->tunnels_r;
1804	sitn->tunnels[3] = sitn->tunnels_r_l;
1805
1806	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1807					   NET_NAME_UNKNOWN,
1808					   ipip6_tunnel_setup);
1809	if (!sitn->fb_tunnel_dev) {
1810		err = -ENOMEM;
1811		goto err_alloc_dev;
1812	}
1813	dev_net_set(sitn->fb_tunnel_dev, net);
1814	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1815	/* FB netdevice is special: we have one, and only one per netns.
1816	 * Allowing to move it to another netns is clearly unsafe.
1817	 */
1818	sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1819
1820	err = register_netdev(sitn->fb_tunnel_dev);
1821	if (err)
1822		goto err_reg_dev;
1823
1824	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1825	ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1826
1827	t = netdev_priv(sitn->fb_tunnel_dev);
1828
1829	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1830	return 0;
1831
1832err_reg_dev:
1833	ipip6_dev_free(sitn->fb_tunnel_dev);
1834err_alloc_dev:
1835	return err;
1836}
1837
1838static void __net_exit sit_exit_net(struct net *net)
1839{
1840	LIST_HEAD(list);
1841
1842	rtnl_lock();
1843	sit_destroy_tunnels(net, &list);
1844	unregister_netdevice_many(&list);
1845	rtnl_unlock();
1846}
1847
1848static struct pernet_operations sit_net_ops = {
1849	.init = sit_init_net,
1850	.exit = sit_exit_net,
1851	.id   = &sit_net_id,
1852	.size = sizeof(struct sit_net),
1853};
1854
1855static void __exit sit_cleanup(void)
1856{
1857	rtnl_link_unregister(&sit_link_ops);
1858	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1859	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1860
1861	unregister_pernet_device(&sit_net_ops);
1862	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1863}
1864
1865static int __init sit_init(void)
1866{
1867	int err;
1868
1869	pr_info("IPv6 over IPv4 tunneling driver\n");
1870
1871	err = register_pernet_device(&sit_net_ops);
1872	if (err < 0)
1873		return err;
1874	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1875	if (err < 0) {
1876		pr_info("%s: can't register ip6ip4\n", __func__);
1877		goto xfrm_tunnel_failed;
1878	}
1879	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1880	if (err < 0) {
1881		pr_info("%s: can't register ip4ip4\n", __func__);
1882		goto xfrm_tunnel4_failed;
1883	}
1884	err = rtnl_link_register(&sit_link_ops);
1885	if (err < 0)
1886		goto rtnl_link_failed;
1887
1888out:
1889	return err;
1890
1891rtnl_link_failed:
1892	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1893xfrm_tunnel4_failed:
1894	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1895xfrm_tunnel_failed:
1896	unregister_pernet_device(&sit_net_ops);
1897	goto out;
1898}
1899
1900module_init(sit_init);
1901module_exit(sit_cleanup);
1902MODULE_LICENSE("GPL");
1903MODULE_ALIAS_RTNL_LINK("sit");
1904MODULE_ALIAS_NETDEV("sit0");
1905